Reviews Of Accelerator Science And Technology - Volume 5: Applications Of Superconducting Technology To Accelerators : Applications of Superconducting Technology to Accelerators 9789814449953, 9789814449946

Citation preview

Reviews of Accelerator Science and Technology Volume 5 • 2012

RAST Vol.5-tp.indd 1

4/1/13 1:31 PM

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

This page intentionally left blank

1203001˙book

Reviews of Accelerator Science and Technology Volume 5 • 2012

Applications of Superconducting Technology to Accelerators

Editors

Alexander W. Chao SLAC National Accelerator Laboratory, USA

Weiren Chou Fermi National Accelerator Laboratory, USA

World Scientific NEW JERSEY

RAST Vol.5-tp.indd 2

•

LONDON

•

SINGAPORE

•

BEIJING

•

SHANGHAI

•

HONG KONG

•

TA I P E I

•

CHENNAI

4/1/13 1:31 PM

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

1203001˙book

Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.

REVIEWS OF ACCELERATOR SCIENCE AND TECHNOLOGY Volume 5: Applications of Superconducting Technology to Accelerators Copyright © 2012 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN 978-981-4449-94-6

Printed in Singapore by Mainland Press.

January 10, 2013

16:17

WSPC/253-RAST : SPI-J100

content˙book

Contents

Editorial Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overview of Superconductivity and Challenges in Applications Rene Fl¨ ukiger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

1

Superconducting Materials and Conductors: Fabrication and Limiting Parameters Luca Bottura and Arno Godeke . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Superconducting Magnets for Particle Accelerators Lucio Rossi and Luca Bottura . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Superconducting Magnets for Particle Detectors and Fusion Devices Akira Yamamoto and Thomas Taylor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 Superconducting Radio-Frequency Fundamentals for Particle Accelerators Alex Gurevich . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 Superconducting Radio-Frequency Systems for High-β Particle Accelerators Sergey Belomestnykh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators Michael Kelly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 Cryogenic Technology for Superconducting Accelerators Kenji Hosoyama . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Superconductivity in Medicine Jose R. Alonso and Timothy A. Antaya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 Industrialization of Superconducting RF Accelerator Technology Michael Peiniger, Michael Pekeler and Hanspeter Vogel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265 Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications Charles E. Reece and Gianluigi Ciovati . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 Educating and Training Accelerator Scientists and Technologists for Tomorrow William Barletta, Swapan Chattopadhyay and Andrei Seryi . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 Pursuit of Accelerator Projects at KEK in Japan Yoshitaka Kimura and Nobukazu Toge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

v

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

This page intentionally left blank

1203001˙book

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

1203001˙book

Reviews of Accelerator Science and Technology Vol. 5 (2012) vii–viii c World Scientific Publishing Company
DOI: 10.1142/S1793626812030014

Editorial Preface

Over the past several decades major advances in accelerators have resulted from breakthroughs in accelerator science and accelerator technology. After the introduction of a new accelerator physics concept or the implementation of a new technology, a leap in accelerator performance followed. A well-known representation of these advances is the Livingston chart, which shows an exponential growth of accelerator performance over the last seven or eight decades. One of the breakthrough accelerator technologies that supported this exponential growth was superconducting technology. Recognizing this major technological advance, we dedicate Volume 5 of Reviews of Accelerator Science and Technology to superconducting technology and its applications. Two major applications are superconducting magnets (SC magnets) and superconducting radio-frequency (SRF) cavities. SC magnets provide much higher magnetic field than room-temperature magnets, thus allowing accelerators to reach higher energies with comparable size as well as much reduced power consumption. SRF technology allows field energy storage for continuous wave applications and energy recovery, in addition to the advantage of tremendous power savings and better particle beam quality. In this volume, we describe both technologies and their applications. We also include discussion of the associated R&D in superconducting materials and future prospects for these technologies. This volume contains thirteen articles, all written by leading scientists in their respective fields. The first two articles are overviews of superconducting technology and applications. Rene Flukiger, a well-known veteran in the field of applied superconductivity, gives a historical and comprehensive review of superconductivity, superconductors and challenges in various applications including energy, medicine and communications. The article by Bottura and Godeke traces the evolution of superconducting materials for accelerators from the late 1960s through the 2020s covering NbTi, Nb3 Sn and high temperature superconductors (HTS). The next two articles discuss SC magnets: Rossi and Bottura on SC magnets for particle accelerators; Yamamoto and Taylor on SC magnets for particle detectors and fusion devices. They are followed by three articles addressing SRF systems. Gurevich gives an overview of SRF system fundamentals, and Belomestnykh and Kelly discuss specific features of two different SRF systems — one for relativistic and another for nonrelativistic accelerators. In the article by Hosoyama the description of superconducting technology includes a discussion of the supporting infrastructure of cryogenic and other related systems. A major application of superconductivity is in medicine. The article by Alonso and Antaya describes compact superconducting cyclotrons and applications in beam therapy and isotope production, as well as the application of SC magnets in isocentric gantries and diagnostic imaging. The following article by Peiniger, Pekeler and Vogel is a discussion about one of the key issues involved in the development of SC technology, namely, the industrialization of SRF accelerator technology. The article by Reece and Giovati includes a survey of SRF technology R&D and highlights the potential for future applications. It has been a tradition of the RAST journal that in each volume we feature one or two articles that concern the accelerator community worldwide but are not necessarily related to the theme. In this volume, we present two such articles. One is an article by Barletta, Chattopadhyay and Seryi, which discusses the education and training of accelerator physicists and engineers for tomorrow, a task of critical importance for our community. The article by Kimura and Toge is a review of the history of the KEK laboratory in Japan, including interesting stories about its two founders — Shigeki Suwa and Tetsuji Nishikawa.

vii

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

viii

Editorial Preface

1203001˙book

The RAST journal has continued to receive strong support from our accelerator community. Feedback from our readers is most encouraging. We are grateful for the generous support from the board of advisors, authors, and referees. Their selfless dedication and contributions are what make this journal possible and demonstrate the vitality of our field and our community. Alexander W. Chao SLAC National Accelerator Laboratory, USA [email protected] Weiren Chou Fermi National Accelerator Laboratory, USA [email protected]

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Reviews of Accelerator Science and Technology Vol. 5 (2012) 1–23 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300010

Overview of Superconductivity and Challenges in Applications Rene Fl¨ ukiger CERN, TE–MSC, Geneva 1211, Switzerland [email protected] Considerable progress has been achieved during the last few decades in the various fields of applied superconductivity, while the related low temperature technology has reached a high level. Magnetic resonance imaging (MRI) and nuclear magnetic resonance (NMR) are so far the most successful applications, with tens of thousands of units worldwide, but high potential can also be recognized in the energy sector, with high energy cables, transformers, motors, generators for wind turbines, fault current limiters and devices for magnetic energy storage. A large number of magnet and cable prototypes have been constructed, showing in all cases high reliability. Large projects involving the construction of magnets, solenoids as well as dipoles and quadrupoles are described in the present book. A very large project, the LHC, is currently in operation, demonstrating that superconductivity is a reliable technology, even in a device of unprecedented high complexity. A project of similar complexity is ITER, a fusion device that is presently under construction. This article starts with a brief historical introduction to superconductivity as a phenomenon, and some fundamental properties necessary for the understanding of the technical behavior of superconductors are described. The introduction of superconductivity in the industrial cycle faces many challenges, first for the properties of the base elements, e.g. the wires, tapes and thin films, then for the various applied devices, where a number of new difficulties had to be resolved. A variety of industrial applications in energy, medicine and communications are briefly presented, showing how superconductivity is now entering the market. Keywords: Superconductivity; Meissner effect; high temperature superconductors; magnetic resonance imaging; nuclear magnetic resonance; thermal stability of superconducting wires; magnetic energy storage; Large Hadron Collider; ITER; effect of high energy irradiation on superconducting properties.

1. Overview of Superconducting Properties

was destroyed by applying rather small currents or magnetic fields. This observation limited the interest in applied superconductivity for a few decades, but this changed dramatically after the discovery of type II superconductors, which are the basis of all present developments. At present, an increasing variety of devices are based on the effects of superconductivity. Several important industrial achievements could not have been made without the effect of vanishing electrical resistivity. This holds for magnets, where the possibility of persistent currents in NbTi and Nb3 Sn wires has revolutionized the field of spectroscopic analysis. The two main examples are represented by magnets for magnetic resonance imaging (MRI) for medical analysis and high field nuclear magnetic resonance (NMR) for industrial research in pharmacy, and molecular biology. Worldwide, the total number of magnets for resonance analysis has reached several tens of thousands and is still increasing.

1.1. Introduction Since the discovery of superconductivity by Kamerlingh Onnes in 1911, this phenomenon has exerted an unbroken fascination, and the search for new materials with higher transition temperatures goes on. Today, it is widely accepted that the discovery of new materials with considerably higher transition temperatures or improved fabricability with respect to the presently known ones is still possible, leading to unexpected applications. The ability to carry large currents in the superconducting state under operating conditions, even in the presence of high magnetic fields, is a stringent requirement for an industrial wire. Soon after the discovery of the sudden transition of the electrical resistivity of Hg to zero, however, Onnes observed that superconductivity in the materials known at that time (defined today as type I superconductors) 1

December 24, 2012

2

14:49

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

Superconductivity represents the only viable alternative for large applications in accelerators and energy: the LHC collider reaches fields of up to 8.8 T and is already in operation. At present, the upgrade with fields reaching 15 T in the quadrupole magnets is under study, while ITER, the experimental fusion magnet designed for 12 T, is already under construction. In all these cases, the electrical resistance in Cu cables would have led to prohibitively high levels of heating. A large number of applications in the energy sector requiring only low or intermediate magnetic fields have already been successfully tested on an industrial scale, e.g. fault current limiters, high current cables, transformers and generators. Considerable market penetration could be expected within the next few years. The list of superconducting applications would not be complete without mentioning the transportation issues: the construction of a levitated high speed train between Tokyo and Osaka has been decided and will be operational within a few decades. Finally, first studies undertaken regarding wind generators show that superconductivity would constitute a real alternative at power levels exceeding 10 MW. The history of superconductivity is marked by the sudden discovery of new material classes with new or improved properties. On the way from the discovery of a new material class to its application, considerable experimental and metallurgical difficulties had to be overcome, depending on the individual nature of the superconductor. The search for solutions to these high technology requirements has also led to substantial progress in various research fields: new methods had to be developed for fabricating complex wires of kilometer lengths based on Nb3 Sn, a very brittle superconductor, and to use them as a basis for high field magnets. Another example is given by the coated conductor tapes based on the superconductor YBa2 Cu3 O7 or R.E.Ba2 Cu3 O7 , called Y-123, in the following: the need for industrial lengths of high quality tapes with high critical currents has not only led to intensive development of complex thin film deposition techniques, but also to progress of analysis methods in the nanometric range. In the first part of this article, a brief introduction to superconductivity as a phenomenon is given. It would lead too far to give a detailed description of all aspects of superconductivity: the present

discussion will thus be limited to those fundamental properties necessary for understanding the technical behavior of superconductors. More detailed information about fundamental properties can be obtained from the series of books [2–6] which were used for preparing the present overview. The second part will treat selected material problems and their challenges, while the last part will be devoted to the main applications, in view of present and future developments. The challenges in a variety of industrial devices are discussed. 1.2. Historical Since 1911, a series of steps have led to the present understanding of superconductivity. A very complete overview can be found in the book 100 Years of Superconductivity [1]. The following steps mark the progress in superconductivity over the last decades: 1933: Expulsion of the magnetic field by Meissner and Ochsenfeld: perfect diamagnetism. 1934: London two-fluid model; prediction of penetration depth. 1937: Observation of a new behavior in superconductors in the presence of magnetic fields (later called type II), by Shubnikov. 1950: Phenomenological theory by Landau and Ginzburg, based on Landau’s theory of second order transitions. 1957: Theoretical prediction of a triangular flux line lattice by Abrikosov. 1957: Microscopic theory of superconductivity by Bardeen, Cooper and Schrieffer (BCS theory). 1954: Discovery of Nb3 Sn. 1962: First commercial superconducting NbTi wires and first superconducting magnet. 1962: Discovery of the Josephson effect. uller and G. 1987: High Tc superconductors, by A. M¨ Bednorz (Tc up to 130 K). 2001: Discovery of MgB2 (Tc = 39 K). 2008: First superconducting Fe-based pnictides, with Tc values up to 56 K. 1.3. Meissner effect and penetration length In 1933, Meissner and Ochsenfeld found that the magnetic flux of a superconducting sample cooled below the critical temperature, Tc , is completely

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

expelled from the interior of the sample: the magnetic induction B inside the material was found to be zero. This behavior characterizes the superconductors of type I, which were the first ones to be discovered. With a few exceptions (Nb, Ta, V, etc.), metallic elements are all type I superconductors. After finding a different behavior in new compounds, and based on theoretical arguments, Abrikosov in 1957 proposed classifying them as type II superconductors. The expulsion of the magnetic field (see definition in footnote a) from the interior is due to the critical currents at the vicinity of the surface, which cause a magnetic flux being opposed to the applied external field: B = µ0 (H +M ), with µ0 = 4π×10−7 . Thus, the Meissner effect with B = 0 corresponds to M = −H , i.e. the sample becomes a perfect diamagnet. For a superconducting wire with a given diameter d, a critical current Ic can be defined. For simple superconductors this current, which generates the critical field at the surface, is given by Silsbee’s rule: Ic = d/µ0 Bc . There is a critical magnetic flux Bc , the thermodynamical critical field, above which the surface supercurrents can no longer exclude flux from the inside. The variation of Bc with temperature for type I superconductors is well described by the empirical formula
 2  T . (1) Bc = B0 1 − Tc This parabolic behavior is strictly observed in type I superconductors (Fig. 1), but represents also a good approximation for type II superconductors. The first description of superconductivity has been given by the two-fluid model of Gorter and Casimir, who assumed a superconductor to be containing electrons of two different types: normal electrons and superelectrons. The normal electrons have the same properties as in normal metals, while the superelectrons present the new, unusual properties. For their explanation of the Meissner effect, F. and H. London proposed replacing in superconductors Ohm’s law J = σE by Newton’s law of motion ∂J = (1/Λ)E , ∂t a In

(2)

3

Fig. 1. Thermodynamical critical fields Bc vs. T for the type I superconductors Hg, Sn, In and Tl.

where Λ = me /ns e2 is the London parameter and ns the density of superconducting electrons. Using Maxwell’s equation, and taking into account that in the bulk of the superconductor, J = 0 and B = 0, they found that ∇(ΛJ s ) = −B,

(3)

the second London equation. Introducing Amp`ere’s rule and after several transformations, it is ∇2 B = B/λ2L .

(4)

This result implies that the external magnetic field (and current) can penetrate only a thin surface layer of a superconductor, the London penetration depth, defined as λL = (me /µ0 ns e2 )1/2 . The solution to this equation is represented by Ba (x) = B(0)e−x/λ and reflects the screening effect: near a plane surface, the magnetic flux and the supercurrent density both decay exponentially with depth x (Fig. 2). The treatment of the Meissner effect leading to a penetration depth λL became the basis of the electrodynamics of superconductors. The situation in a wire is drawn in Fig. 3. The value of λL varies between 40 and >100 nm, as illustrated for several superconductors in Table 1. The shielding of the magnetic field and current due to the Meissner effect in type II superconductors is the basis of magnetic levitation. Today, known applications of magnetic levitation are ultrarapid

vacuum, the quantities H (magnetic field) and B (magnetic induction) differ only by a multiplicative constant: B = µ0 H. Inside a superconductor, B = µµ0 H, where µ is the permeability. Informally, though, and formally for some recent textbooks, the term “magnetic field” is used to describe B as well as or in place of H. In the present overview, the symbol B will be used, but where necessary it will be distinguished between applied magnetic field strength and magnetic flux.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

4

the last 20 years were the main argument for the already-mentioned construction of the first levitating train between Tokyo and Osaka. There is another useful application of the Meissner effect: it is still the ultimate practical test when one is determining whether a newly discovered material is really superconducting or not. 1.4. Heat capacity of superconductors The free enthalpy of a superconductor with the volume V in the presence of an external field strength H and the magnetic flux B = µ0 H can be formulated: dG = −S dT + V dp − V · M · dB. Fig. 2. Exponential decrease of the magnetic flux B with distance x into a bulk superconductor, with the characteristic decay length λL (London penetration depth).

(5)

The difference between the free enthalpies Gn (T, B) for the normal state and Gs (T, B) for the superconducting state can be calculated taking into account the Meissner effect with M = −B/µ0 : Gs (T, B) = Gs (T, 0) + B 2 .

(6)

Since at the thermodynamical critical field Bc there is Gs (T, Bc ) = Gn (T, Bc ) and Gn (T, Bc ) = Gn (T, 0), it follows that Gn (T, 0) − Gs (T, 0) = Bc2 ,

Fig. 3. The current in a type I superconductor flows only in a surface layer. For currents I ≥ Ic = Bc ·2πR/µ0 , the thickness of this layer is zero and the wire becomes normal.

Table 1. Some values of the penetration depth at 0 K. Tc (K)

λL (nm)

Al

1.1

50

Nb

9.2

40

NbTi

9.5

60

Nb3 Sn

18.3

80

YBa2 Cu3 O7

92

400

centrifuges, while prototypes of levitating trains and flywheels have been tested successfully. Particularly noteworthy is the 18-km-long test track for levitating trains in Japan: a series of successful tests in

(7)

and Gn > Gs , the difference reflecting the stability of the superconducting state. Using S = −( ∂G ∂T )B , one can calculate the difference between the entropies: Sn − Ss = −V.

(8)

c < 0, there is Sn − Ss > 0; thus, Since ∂B ∂T the superconducting state has always a higher order. Since Bc (T ) = 0 for T → Tc ; it follows that Sn −Ss = 0 at T = Tc ; thus, in the absence of a latent heat, the superconducting transition is a phase transition of the second order. This is in contrast to the case where the normal state at T = const is caused by a magnetic field B > Bc ; in this case, the thermal energy (Sn −Ss )T has to be added, and the transition is of the first order. These considerations can be used to calculate the change of specific heat at the superconducting transition. With C = −T (∂ 2 G)/∂T 2 )p,B , the difference at T = Tc can be written as

Cs − Cn = (∂/∂T )2 .

(9)

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

5

(a)

Fig. 4. Discontinuity at T = Tc of the specific heat Cv of a superconductor.

This is the Rutgers formula, which implies that there must be a discontinuous jump at T = Tc in the specific heat as a function of temperature. This jump of Cv , reflecting the second order transition, is schematically represented in Fig. 4.

(b) Fig. 5. (a) Magnetization as a function of the applied field Ba = Ha /µ0 for a type II superconductor. A “Meissner” behavior is only observed for Ba < Bc1 , (b) Partial penetration of magnetic flux as a function of Ba .

1.5. The mixed state The difference between type I and II superconductors can be illustrated on the basis of Fig. 5, which shows the variation of magnetization as a function of the applied field Ba = Ha /µ0 . In type II superconductors, the flux is completely expelled up to the applied field Bc1 (the lower magnetic field), as in type I superconductors. However, with an increasing applied magnetic field, the flux partially penetrates until Bc2 , the upper critical magnetic field, is reached. The correlation between Bc , Bc2 and Bc is illustrated by the equality of the two hatched surfaces in the regions Bc1 < Ba < Bc and Bc < Ba < Bc2 . Figure 6 shows schematically the variation of the critical fields Bc1 and Bc2 as a function of temperature. The upper critical field Bc2 can reach values >100 T in HTS materials (see Table 2, where some type II superconductors are listed). On the basis of the Ginzburg–Landau theory, Abrikosov found in 1957 that superconductors of type I and type II can be characterized on the basis of the energy at the normal/superconducting boundary. This boundary is characterized by two energies: the energy for the expulsion of the magnetic field, ∆EB , and the energy decrease due to the condensation of the Cooper pairs,

Fig. 6. Variation of the critical field flux Bc1 and Bc2 as a function of temperature for type I and type II superconductors.

∆EC . For the total energy, it follows that ∆Etot = ∆EC − ∆EB = (ξ − λ)A · Bc2 /2µ0 ,

(10)

where A is the boundary surface while the volume for the energies corresponding to ∆EC and ∆EB is simplified by ξ · A and λ · A, respectively. For ξ > λ, ∆Etot > 0: when building up the boundary, the energy loss due to condensation is

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

6

Table 2. Upper critical fields and coherence lengths for several type II superconductors. The asterisk indicates a wide range of Bc2 and ξ0 values in HTS and FeAs superconductors. Compound NbTi

Tc (K)

Bc2 (T )

ξ0 (nm)

9.5

14

6

Nb3 Sn

18.3

28

4

Nb3 Al

19

33

4

MgB2 (bulk)

39

39

5

PbMo6 S8

15

60

2.2

Rb3 C60

29.3

50

∼2

La1−x Srx CuO4

38

65∗

∼1.5∗

YBa2 Cu3 O7 (123)

92

120∗

∼1∗

FeAsO1−x Fx (1111)

55

>50∗

∼2.5∗

(Ba0.6 K0.4 )Fe2 As2 (122)

38

70–135∗

∼2∗

larger than the expulsion energy. The most favorable energy configuration is thus the expulsion of the magnetic field flux, which is the condition for a type I superconductor. For ξ < λ, ∆Etot < 0: the energy associated with the boundary between superconducting and normal regions is negative; and this will favor the formation of as many boundaries between normal and superconducting regions as possible. This is the behavior of type II superconductors. Since the magnetic flux in the normal region is not Bc but generally B, the boundary energy is ∆Etot = ∆EC − ∆EB = ξ · Bc2 < λ · B 2 .

(11)

For ∆Etot < 0, there is ξ · Bc2 < λ · B 2 . With the definition κ = λ/ξ, the so-called Ginzburg–Landau parameter, one gets B 2 > Bc2 /κ. It follows that superconductors of type I and type II can clearly

Fig. 7.

be defined: κ> κ
Fp . This motion will stop once F = Fp is reached at all points: this physical regime is called the critical state. The variation of the field inside the sample is directly linked to the current Jc , which can pass through the sample without vortex motion: Fp = Jc B =

dH B. dx

(13)

For a superconducting slab of thickness d with an applied longitudinal field B, its penetration above Bc1 will reach a depth ∆ = dH dx . Figure 12 represents the situation for increasing values of a longitudinal applied magnetic field B up to the values B ∗ for full field penetration, followed by a decrease of B. The applied field at which the magnetic flux is fully penetrated in the rod is B ∗ = 12 µ0 Jc d. If the

Fig. 12. Schematic critical state flux profile: field penetration of a longitudinal field B in a superconducting rod. B first increases (1–3), then decreases (4–6). B ∗ is the field for full field penetration.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

applied field is again decreased to zero, a considerable quantity of flux remains trapped in the sample. During the whole cycle, there is dissipation of energy, necessary for the movement of the vortices. These losses due to hysteresis limit the potential efficiency of type II superconductors in ac applications. From a hysteresis cycle, one can deduce the critical current density, Jc (B). The difference ∆M (B) between the magnetization measured in increasing and decreasing fields can be calculated from Fig. 11. Applying the Bean model, one gets Jc [Am−2 ] = 2

∆M . d

(14)

Although this formula does not account for uniformity differences in real conductors, it is often used for the determination of Jc , as indicated in Fig. 11, as an alternative to the measurements by transport. However, particular care has to be given to the definition of the boundary conditions which may affect the result of the inductive Jc measurement. The critical state Bean model can also be used for describing the field penetration in a cylindrical wire where a current I is flowing, as shown in Fig. 13. For the following consideration, jc is assumed to be constant and x  R. For a current I  Ic , the flux penetrates up to the depth x1 , given by I = jc π(2Rx1 − x21 ), and x1 = I/2πRjc . For a current 2I, the penetration depth is x2 , which can be determined

9

to x2 = 2x1 . The full penetration xn = R will be reached for the current In = πR2 jc . If the current is decreased from 2I to I, the penetrated flux must again be expelled from the wire. Following the Bean model, this happens by inverting the direction of the current in the outermost region x4 of the wire. The thickness x4 can be found from x4 = x2 − x4 : x4 =

I x2 − . 2 4πRjc

(15)

1.7. Stabilization criteria for superconducting wires A superconducting wire in a magnet must be able to recover its equilibrium even after a perturbation which leads locally to zones where the temperature can exceed Tc . A simple presentation of the effects of a thermal perturbation on a wire of radius R in a bath of temperature Tb is shown in Fig. 14 and will be discussed in the following. Many reasons can lead to a perturbation causing a local enhancement of temperature in a superconducting wire. In the following it will be simply supposed that the perturbation has already taken place, regardless of the causes. After linearization, the temperature gradient in the perturbed zone is T  = dT/dx = 2/a(Tp − Tb ). The equilibrium conditions leading to a normal zone can be determined by the balance of the power terms

Fig. 13. Penetrating magnetic flux in a cylindrical wire under a current I. The regions through which the current flows are calculated with the Bean model. (a) Current I  Ic ; (b) increased current, I = 2Ic ; (c) decreasing current from 2I to I. (After P. Komarek [5].)

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

10

principles:

Fig. 14. Cryogenic stabilization of a technical superconducting wire.

• The difference (Tc − Tb ) in Fig. 14 should be as large as possible. • The radius of the wire R should be as small as possible. More generally, it is thermally more favorable to subdivide the wire in order to get a multifilamentary structure. • Heat transfer: the filaments should be enclosed in a highly conductive matrix, e.g. Cu or Al, to ensure a high thermal transmission to the cooling medium.

in the system:

1.8. Relation between Jc , Bc2 and Tc

(1) Joule heat in the normal zone. With j = I/πR2 and ρn , the normal resistivity at Tc , there is

The ability to carry large currents in the superconducting state even in the presence of high magnetic fields is the most important requirement for an industrial superconductor. The superconducting state in a wire carrying a current I is defined by the three quantities Jc , the critical current density, Bc2 , the upper critical magnetic field (the magnetic flux) and Tc , the critical temperature. These three quantities are correlated, as schematically shown in Fig. 15. If only one of these three critical quantities is exceeded, the wire becomes normal: only working points below the envelope are possible. For Bc2 (T ), the experimental results can be approximated by Bc2 (T ) = Bc2 (0)[1 − (T /Tc)2 ]. The critical current density depends on the density of pinning centers, which in turn is field-dependent, as described by E. Kramer in 1973 [8]. In order to understand the effect of pinning, one can consider an ideal flux line lattice and a random distribution of defects as pinning centers. For very small pinning forces, i.e.

PJoule = Rn I 2 = ρn aj 2 πR2 .

(16)

(2) Transmitted power from the normal zone to the neighboring superconductor, by thermal conductivity. With the thermal conductivity λ, Pthermal = 2πR2 λ

dT = 2πR2 λT  . dx

(17)

(3) Power transmitted to the cryogenic bath over a line dx of the normal zone, introducing the heat transition coefficient h: Pbath = 2πah[(Tc − Tb ) + aT  ].

(18)

The stability condition can thus be formulated as PJoule = Pthermal + Pbath , ρn J 2 = 2λT 

(19)

1 hT ∗ Tc − Tb + a + 2h a 2R R

≡ f (a).

(20)

The minimum of this equation can be found to be fmin = 2T (λh/R)1/2 + 2h

Tc − Tb . R

(21)

The stability condition ρn J 2 < fmin is now αst =

ρ2n < 1, fmin

(22)

where αst is the Stekly parameter [7]. After some improvements, the Stekly criterion is currently used for determining the stability of magnets. In order to improve the conditions of stability, various parameters in the equation of αst can be optimized, which leads to some general

Fig. 15. Correlation between critical current density Jc , upper critical magnetic induction Bc2 and transition temperature Tc in industrial superconducting wires.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

at low magnetic fields, the lattice will respond elastically: the regular flux line lattice will essentially subsist, and the pinning effect by defects or inclusions will still be small. The total pinning force Fp is smaller than would be possible taking into account all built-in defects. With an increasing field, the positional long-range order of the flux line lattice is lost. More and more flux lines will be trapped by defects, thus leading to an enhancement of Fp . With further increase of the magnetic field, the number of flux lines will increase. The global pinning force per volume, Fp , frequently exhibits scaling behavior with a reduced field b = B/Bc2 and κ: with magnetic field will in general have the following behavior: n Fp = BJc = S · Bc2 /κm bp (1 − b)q ,

(23)

where the constant S is a geometrical function of the microstructure [8]. For normal particle pinning, this approximation is better than a few percent for b > 0.4. As an example, the variation of Fp vs. B for a multifilamentary wire as measured by J. Ekin [9] is shown in Fig. 16. It shows that this representation is still valid after applying uniaxial tensile stress on the wire: Fpmax does not change for various strain values ε. This simple formulation fails at low flux densities and at phase lines where there is a change in the basic pinning mechanism due, for instance, to dimensional crossover or to the onset of thermally activated processes.

Fig. 16. Pinning force as a function of the applied magnetic field for a binary multifilamentary Nb3 Sn wire at 4.2 K for various strain values ε. The upper critical field Bc2 ∗ as determined from Kramer extrapolation is 20.8 T and the fitting parameters are p = 0.56 and q = 2.0. (From Ref. 9.)

11

2. Challenges for Superconducting Materials and Conductors 2.1. Time between discovery and application The list of envisageable applications involving superconductivity is very long, and an impressive number of prototype devices for a large variety of purposes have already been constructed and successfully tested. When one is studying the impact of superconductivity in the near future, a question arises about the R&D and the necessary time for reaching the step of industrialization. It is interesting to compare superconductivity to other high technology areas, e.g. semiconductors or fiber optics. In these areas, the time gap between ideas and preliminary experimental work on prototypes and market penetration and commercial success was of the order of 3–4 decades. For example, the initial development of transistors at AT&T Bell Laboratories started in 1947, but the first integrated circuit was built in 1958. More than 30 additional years were necessary to reach the present highly sophisticated industrial level. A second case is fiber optics communication: the concept of channeling light was produced in 1858 by J. Tyndall, while communication by light was realized in 1880 by Alexander Graham Bell. The first optical fibers were only developed in 1971 by Corning. In the case of superconductivity, it took more than 50 years between its discovery in 1911 and the first NbTi magnet. Nb3 Sn was discovered in 1954 by B.T. Matthias, and was rapidly followed by small prototype magnets. However, the first industrial Nb3 Sn magnet producing a field of 15 T was not available before the 1980s. In a similar way, the HTS materials were discovered in 1987 and the market penetration is starting only now, 25 years later, after a strong worldwide effort expended on a multitude of prototypes. These prototypes were developed for very different purposes, the most important ones being e.g. high current cables, fault current limiters, transformers, energy storage, generators and microelectronics. It is very important for the future developments that all prototypes involving superconductors were successful: it can be said that superconducting systems have demonstrated their reliability under industrial conditions.

December 24, 2012

14:49

12

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

Fig. 17. Highest Tc values in various superconducting material classes discovered since 1911. (The pnictides, discovered after 2008, were added to the original graph established by DOE.)

2.2. Superconducting materials with higher Tc values The superconducting material classes found since 1911 are shown in Fig. 17. The highest Tc value reached so far was produced by applying 30 GPa to the compound HgBaCaCuO, leading to an increase from 131 to 164 K [10]. The question arises whether new materials with even higher Tc values can still be found. However, it must be said that, up to the present day, all superconductors shown in Fig. 17 were found unexpectedly. No valid principle is known to predict new superconducting materials with higher Tc values. 2.2.1. The search for new superconductors In spite of the failure of the known strategies for the search for new superconductors with higher Tc , there is no reason to exclude the possibility that one of the known strategies, possibly after slight changes or in combination with other ideas, may lead to success in the future. The search for new superconductors being an important challenge, it may be of interest to briefly mention the strategies which have been applied so far. In the 1970s, B.T. Matthias started his search with a purely empirical principle: the number of electrons per atom in a given lattice. This guideline, which was not supported by any theory, led to interesting results, but was limited to the class

of binary superconductors based on transition metals, with a maximum Tc of 23 K for Nb3 Ge [11]. A new era started in 1987 with the discovery of HTS oxides by M¨ uller and Bednorz [12]. This discovery had the merit of stimulating the synthetization of a large variety of compounds where superconductivity was not expected. In the meantime, many new classes of superconductors were discovered, the latest one being the family of Fe-based pnictides (see Fig. 17). Now, the search was extended to crystal structures with four elements (quaternaries) and even to those with five elements and more. It is easily seen that the number of possible new combinations for a purely empirical search exceeds 107 . This unrealistic number can be reduced only if new restricting criteria are found. However, after the failure of the BCS theory for most compounds found after 1987, no new model is presently able to describe satisfactorily all aspects of HTS superconductivity. We are still far away from finding superconductors by design. The field to be investigated is extremely wide, as illustrated by the complexity of the various theoretical problems to be solved: • Electronic structure calculations by density functional theory; • Large scale phonon calculations in the nonlinear, anharmonic limit; • Formulation of “very strong” electron–phonon coupling (beyond Eliashberg theory);

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

• Determination of quantitative pairing mechanisms for high temperature superconductors. In spite of all these difficulties, the search for new superconductors goes on, as illustrated by the recent progress achieved in organic superconductors, where the highest Tc value for the alkali-metal-doped hydrocarbon with the formula 1,2:8,9-dibenzopentacene is now 33 K [13].

2.3. Superconducting wires and tapes for applications At the present time, there is still no ideal superconducting material with the potential for all temperatures up to Tc values in the presence of high magnetic fields. The choice of industrial superconductors is quite limited and does not even include the compounds with the highest Tc values shown in Fig. 17. Indeed, mainly due to metallurgical difficulties in preparing high quality wires and tapes based on the compounds HgBaCaCuO and TlBaCaCuO, the choice of industrial superconductors is limited to compounds with Tc values below 110 K (see Table 3). Within these limits, the choice of the superconducting material for industrial applications will essentially depend on the operation conditions, e.g. operation temperature, operation magnetic field, critical current density, ac losses, mechanical and thermal stability, sensitivity to high energy irradiation and, last but not least, production costs. One cannot establish general rules, since each application requires a different solution. Particular aspects of superconducting wires are briefly treated here.

Table 3. Transition temperature (Tc ), upper critical fields (Bc2 ), anisotropy (γ) and coherence lengths (ξ) of various superconducting high field compounds for industrial applications. Exception: FeAs-122, which is still not industrial. Material

NbTi Nb3 Sn+Ti

Tc (K)

Bc2 (0) (B)

γ

ξ (nm)

9.6

14

1

6

18.3

28

1

4

ξab (nm)

ξc (nm)

MgB2

39

40

∼5

5

∼2

YBCO

93

>100(//)

≤14

1.5

0.8

Bi-2212

92

>100(//)

Bi-2223

110

>100(//)

>100

38

>70(//)

>70

FeAs-122

13

The last material in Table 3, FeAs-122, has not reached the industrial level, but deserves a particular mention. It belongs to the large family of Fe pnictides, discovered in 2008 by Hosono et al. [14]. This material class exhibits lower Tc values than HTS compounds (up to 38 K), But has very high Bc2 values too, exceeding 70 T. Pnictides crystallizing in the 122 structure have a considerably reduced anisotropy of Bc2 , reflecting a smaller anisotropy of coherence length ξ. In contrast to HTS compounds, which require highly textured tapes, pnictides have the potential to be prepared as round wires. At present, the highest reported critical current density of the first round AsFe-122 wires is 2 × 104 A/cm2 at 10 T and 4.2 K [15]. Considering the high value of Bc2 , the decrease of Jc vs. B is very slow, showing that this material not only has potential for the fabrication of round wires, but may be promising for very high fields. 2.4. Wires for high field magnets A real challenge in the long term development of applied superconductivity consists in the production of very high fields in various magnet types, e.g. solenoids, dipoles and quadrupoles. For the present LHC collider at CERN, the dipoles were constructed for reaching 8.3 T at 1.8 K, but for the next generation, the high luminosity LHC upgrade, magnetic fields above 15 T are envisaged. A possible future development could be the muon collider with fields exceeding 30 T. The presently known superconducting materials are listed in Table 3, together with the transition temperatures and the upper critical fields. However, these data are not sufficient for deciding about the applicability of a given material for a very high field application. A choice can only be made taking into account the fabricability of wires or tapes on an industrial scale. In addition, it must be considered that in contrast to cubic Nb3 Sn and NbTi, the achievable magnetic field in the anisotropic compounds MgB2 and the HTS depends on the irreversibility field Birr , which is considerably lower than Bc2 , in particular for fields applied perpendicular to the axis. 2.4.1. Field ranges and challenges

2–3

1–1.5

The superconducting material used for a wide range of industrial applications is ordinarily taken

December 24, 2012

14:49

14

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

depending on the magnetic field range and the operation temperature. Some examples are given in the following, but this list is certainly not complete. The definition of the various field ranges is somewhat arbitrary, but is introduced here for better clarity. Low and middle field ranges: • Superconducting fault current limiters (SFCLs) and high current HTS cables usually operate at 77 K, at the self-field. Only the 13,000 A cables for the LHC upgrade (LINK project) [17] based on MgB2 wires are designed for operation at 10 K. • MRI magnets operating at 4.2 K consist of NbTi wires. Their magnetic field for medical applications presently reaches 3 T, but it can follow the future market requirements up to considerably higher magnetic fields. • The operation range for HTS motors or wind generators at 77 K and fields up to 3 T is envisaged, but still constitutes a challenge. • Due to the considerably lower costs of MgB2 wires with respect to HTS-coated conductors, such a conductor may be envisaged for motors and wind generators up to 25 K at fields up to 3 T and more. • Due to the low relaxation rate of MgB2 [18], this conductor wire is presently also used for MRI magnets at 20 K, the present operation field being of the order of 1 T. The challenge consists here in a substantial field increase, maybe to 3 T, in the next few years. • NbTi is used up to fields around 9 T at 4.2 K and is the only ductile high field superconductor. It can be produced in large quantities and is the most economical one among all known superconductors. It is used for the background field of all laboratory and NMR magnets producing fields above 10 T. • NbTi is also used for the poloidal field coils of ITER. High field range: • All laboratory magnets as well as the NMR magnets consist of Nb3 Sn magnets in a background field produced by external, concentric NbTi magnets. At 1.8 K, these magnets can produce magnetic fields close to 11 T. • The highest field achieved in the Nb3 Sn solenoid was realized by an industrial NMR magnet, which produced 23.5 T at 1.8 K [19]. However, it can be seen from the behavior of Bc2 vs. T in Fig. 18 that

Fig. 18. 2212 and axis. For identical.

Irreversibility fields for wires of Nb3 Sn, MgB2 , Bitapes of YBCO for applied fields B parallel to the the cubic Nb3 Sn, Bc2 (T ) and Birr (T ) are almost (From Ref. 16.)

this field is very close to the high field limit of Nb3 Sn. • The central coil as well as the Tokamak coils for ITER are based on Nb3 Sn wires and are designed for fields of the order of 12 T at 4.2 K. • The magnetic field in the quadrupoles for the future high luminosity LHC upgrade collider is designed for 15 T. The strong requirement for the noncopper Jc value, which should exceed 1,500 A/mm2 at 15 T and 4.2 K, has recently been fulfilled [20]. A substantial improvement of this value in Nb3 Sn wires would constitute a real challenge. Very high field range (above 23.5 T ): • There is a need for considerably higher fields than the 23.5 T achieved by the Nb3 Sn magnet. This goal can be achieved using HTS superconductors: recently, a prototype laboratory magnet based on Y-123-coated conductors has reached a magnetic field of 33.8 T at 4.2 K in a background field of 31 T [21]. Magnetic fields well above 20 T have also been produced in prototype magnets based on Bi-2212 and Bi-2223. • Very high fields would also be needed for other hypothetical future projects, such as the muon collider or the LHC Tripler. It follows from Fig. 18 that the known HTS conductors cannot produce high fields at 77 K. Thus,

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

high field magnets based on both Bi-based compounds must be operated at 4.2 K, while high fields at 30 K seem possible for YBCO. Several manufacturing companies are presently working on the project to build industrial HTS high field magnets. It is only a matter of time before the first industrial HTS magnet yielding, say, 30 T is produced. The main difficulties to be resolved for HTS magnets are related to thermal and mechanical stability. In the following, the challenges to various magnet types are mentioned.

2.5. Wires for magnets operating in the persistent mode An NMR magnet consisting of alloyed Nb3 Sn wires has reached the record field value of 23.5 T [19], thus corresponding to a resonance frequency of 1 GHz. However, so far it has not been possible to get satisfactory operation in the persistent mode with HTS conductors, which is inherently due to the particular situation at the HTS grain boundaries; the published value is of the order of 10−9 Ω, with the insulating outermost layer leading to a high contact resistance. The lowest contact resistance, as low as 10−12 Ω at the joints of an NMR magnet, is required for performing a high quality NMR operation. Such low contact resistances have so far only been obtained for NbTi or Nb3 Sn wires. It should be mentioned that joints between MgB2 wires, but also between MgB2 and NbTi, have been operated in the persistent mode too. In order to use HTS conductors for NMR applications at higher magnetic fields, the challenge consists in finding new ways for a considerable reduction of the contact resistance. Another possibility consists in replacing the persistent mode by a driven mode. However, for NMR this would require the development of a power supply with unprecedented stability characteristics, which also constitutes a real challenge. For spectrometers beyond 1 GHz, another possibility exists: long term NMR measurements. Indeed, waiting a sufficiently long time after coil excitation (of the order of ∼1000 h), the field drift rate approaches the field decay rate of the persistent current (10−8 h−1 ). This technique is envisaged in a 1 GHz currently under work in Japan, based on Bi2223 tapes.

15

2.6. Wires for accelerator magnets Engineering physics and construction requirements for accelerator magnets require Rutherford cables made by round wires. The situation is simplest below 9 T, where NbTi wires constitute a technically and economically satisfactory solution. For higher fields, e.g. 15 T as envisaged for the quadrupoles in the high luminosity LHC upgrade, the only available round wire material is Nb3 Sn, which has fulfilled the condition of noncopper Jc values of 1500 A/mm2 at 15 T [20]. Another hypothetical candidate material for round wires is Bi-2212. At present, the overall critical current density of Bi-2212 round wires is still lower than required, but may be improved in the light of recent progress at NHMFL [21]. Due to the difficulties in bending tape conductors around the normal to the flat surface, the fabrication of dipoles and quadrupoles based on tape conductors would constitute a real challenge. 2.7. Wires for fusion magnets The ITER fusion project comprises an assembly of various coil types, NbTi poloidal field coils and two types of Nb3 Sn coils: the D-shaped Tokamak coils and the central magnet. The magnetic field of both coils is of the order of 12 T at 4.2 K, and Nb3 Sn was chosen as the only industrial alternative. For future fusion devices requiring higher fields, e.g. DEMO, the possibility of using HTS-coated conductors is presently being discussed. The challenge will consist in finding a practical cable configuration. 2.8. The critical current density Regardless of the differences between low Tc (or LTS) superconductors and high Tc (or HTS) superconductors from the point of view of crystallographic, metallurgical or physical properties, the main challenge for a given wire or tape is to maximize its critical current at the operation temperature and field. This involves the following two conditions: • The pinning strength should be enhanced to its highest possible value; • The irreversibility field should be raised as high as possible (this holds for all superconductors with the exception of the cubic NbTi and Nb3 Sn, where Bc2 is very close to Birr ).

December 24, 2012

14:49

16

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

It is well known that the value of Bc2 in Nb3 Sn wires can be enhanced by up to 3 T by Ta or Ti additions. It was found that Ta substitutes the Nb sites, while Ti goes on the Sn sites [23]. In MgB2 wires, the values of Bc2 and Birr can even be enhanced by 5 T by the addition of SiC, resulting in a substitution of carbon on the B sites [24]. For the considerations about pinning, the notion of coherence length and penetration depth remains valid in all studied compounds. It turns out that the transport current behavior in a general compound is mainly controlled by two factors: the size of the coherence length, and the anisotropy of the physical properties. These two factors have to be taken into account when optimizing wire or tape conductors.

2.8.1. Pinning strength in LTS compounds In LTS compounds, the pinning strength is strongly connected with the conditions at the grain boundaries. This was first demonstrated on Nb3 Sn by Schauer and Schelb [25], who showed that the pinning force increases inversely to the grain size. This result has in the following been extended to other compounds, as Nb3 Sn, and even to MgB2 . In both cases, reaction at higher temperatures leads to larger grains and thus to a reduced pinning force. It follows that the maximum Jc value in these systems is always obtained after the lowest temperature where a reaction of the A15 phase is possible. At the grain boundaries, the periodicity of the lattice is broken, which causes the formation of dislocations and a local shift of the boundary atoms from the equilibrium positions. These defects create the condition for a greater number of vortices in the boundary region, thus explaining why the smallest grains lead to the highest pinning force, which is synonymous with enhanced Jc values. The precise width of this “damaged” region cannot be measured by diffraction measurements, due to the absence of periodicity. The thickness of the damaged zone at each grain boundary must of course be considerably smaller than the coherence length, but is not well known. From a comparison between LTS and HTS compounds, it follows that this thickness changes from compound to compound and that it also depends on the orientation of the grain surface relative to the crystal orientation.

So far, no way has been found to influence the thickness of this damaged zone at the grain boundaries: the challenge, at least for LTS compounds, for MgB2 and probably for FeAs-122, consists in reducing the grain size as much as possible. 2.8.2. Pinning strength in HTS compounds In contrast to LTS compounds, the behavior at the grain boundaries is not the dominant effect for the pinning behavior in HTS compounds. Here, the vortex dynamics inside the grains is decisive, in its static and dynamic configurations. The pinning behavior in HTS compounds can be influenced by including nanoparticles. As an example, we mention here the case of BaZrO3 nanoparticle addition to Y-123 by J. L. MacManus-Driscoll et al. [26]. The effect of these additives is to enhance the pinning force, an additional beneficial effect being the reduction of anisotropy. Several manufacturers have accomplished strong enhancements on Jc of R.E.-123–coated conductor tapes using nanoparticles. As an example, the enhancement of Jc , in the tape surface in the course of the last few years by SuperPower, is illustrated in Fig. 19 [27] with the applied field parallel to the tape surface (Jc is here defined as the critical current density of a 10-mm-wide tape, and has the unit A/cm-w). This figure also shows that the enhancement of Jc is partly due to an increased thickness of the superconductor layer, provided that the orientation remains constant during the growth process. It is expected that both nanoparticle additives and improved deposition techniques will lead to further enhancements of Jc . The highest reported values for

Fig. 19. Enhancement of critical current density at 77 K, zero field in R.E.-123–coated conductor tapes with various nanosize additives. (From Ref. 27.)

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

a 4-mm-wide tape are close to Ic = 320 A, corresponding to 830 A/cm-w. However, it must be mentioned that this value has so far only been measured on samples several meters long. The Jc values of standard coated conductor tapes with lengths of several hundreds of meters are somewhat lower, of the order of 100–200 A for a 4-mm-width. These results illustrate that there is still a challenge to reach higher values. A reduction of anisotropy has also been obtained by the addition of R.E.2 O3 nanoparticles [28]. A convincing example for the considerable lowering of anisotropy achieved by American Superconductors is shown in Fig. 20 [29]. With further progress it is expected that, in future, R.E.-123–coated conductor tapes Jc may behave “isotropically.” These examples show that the development of improved coated conductor tapes still constitutes a valuable challenge. For more information, see article 2 of the present book. 2.8.3. High energy irradiation of superconductors The energy level in the high luminosity LHC as well as ITER involves high energy irradiation that will have an impact on the material properties of the superconducting magnets. The situation is, however, very different for the two cases, and the effects on the superconducting material (Nb3 Sn) as well as on the Cu stabilizer are expected to be different.

17

is well documented. The overwhelming amount of neutron irradiation data on LTS superconductors (in the bulk as well as in the filamentary state) has been performed prior to 1987 and has recently been reviewed [30]. The effect of neutron irradiation on Jc of Nb3 Sn wires is shown by the bronze route technique, while the present wires are fabricated using either the internal Sn diffusion or the powder-in-tube (PIT) technique. It is expected that these wire types may behave differently, based on recent specific heat measurements, represented in Fig. 22 [33]. These measurements reveal a difference of the average Sn content between PIT and bronze route wires of the order of ∼1 at. % Sn, thus suggesting atomic ordering effects. This question is presently being studied at CERN, in collaboration with the Atominstitut in Vienna. In the last few years, the effect of high energy irradiation has been concentrated on HTS-coated conductor tapes [34] (Fig. 23) and in MgB2 wires [35]. For all measured materials, e.g. Nb3 Sn, MgB2 and R.E.-123, the variation of Jc /Jco vs. fluence φt shows an increase at higher magnetic fields, which is attributed to the enhancement of the normal state electrical resistivity with neutron irradiation. (b) The high energy irradiation spectrum of the LHC The irradiation spectrum at the high luminosity LHC upgrade is shown in Fig. 21 [31]. The maximum

(a) High energy neutron irradiation The radiation in ITER consists of 1 MeV neutrons and by photons. The effect of photons at this energy level on the superconducting properties of Nb3 Sn is assumed to be negligible, while the effect of neutrons

Fig. 20. Anisotropy of Jc in HTS-coated conductor tapes at 77 K/1 T. (From Ref. 29.)

Fig. 21. Variation of the ratio Jc /Jco between irradiated and unirradiated Nb3 Sn bronze route wires. The maximum Jc /Jco is markedly higher for binary Nb3 Sn wires than for Ti-alloyed wires. (From Ref. 31.)

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

18

Fig. 22. Tc distribution in the filaments of Nb3 Sn wires prepared by the bronze route (GAP #24), internal Sn diffusion (OST) and PIT (SMI), and in Nb3 Al wires prepared at NIMS. (From Ref. 33.)

Fig. 24. Energy spectrum of neutrons in the LHC, showing a main peak at ∼1 MeV and a shoulder at around 100 MeV [30, 36].

Fig. 25. Energy spectrum of protons in the LHC, with a flat maximum at 200 MeV [36].

Fig. 23. Variation of Jc /Jco for an industrial HTS-coated conductor tape of AMSC, showing an increase of Jc at higher applied magnetic fields [34].

Table 4. Composition of the radiation spectrum reaching the quadrupole Q2a of the LHC [30, 36]. High energy source

Photons Neutrons Electrons Positrons Pions (+/−) Protons

Composition (%)

Energy maximum

88 7 4 2.5 0.45 0.15

1 MeV 1 MeV (100 MeV) 10 MeV 10 MeV 1 GeV 200 MeV

of Jc /Jco for the binary Nb3 Sn wires is observed at much higher fluences than for Ti-alloyed wires, which was confirmed by independent neutron irradiation measurements [32]. The measurements in Fig. 21 were performed on Nb3 Sn wires and were produced by collider calculated by F. Cerutti at CERN [36] is very complex. More details are contained in a recent publication [30] treating the problems of superconductivity in the quadrupoles of the LHC. The composition of the various high energy sources is listed in Table 4, showing that the main contribution is due to the photons. The highest fluence will be reached at the inner winding of the quadrupole Q2a, at a distance of 33 m from the LHC collision point. The neutron spectrum is centered at 1 MeV (similar to the situation in ITER), but with a shoulder at 100 MeV (Fig. 24). The expected maximum radiation fluence was calculated to 2 × 1021 n/cm2 per

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

year, assuming an operation of 200 days a year [30]. The effect of these fluences on Tc and Jc of Nb3 Sn wires can be estimated using the neutron irradiation data reported on binary and ternary alloyed superconducting wires at neutron energies between 1 and 14 MeV [37, 31, 38, 32]. It is particularly interesting to analyze the effect of the other high energy sources in the LHC upgrade. The various energy spectra have all been calculated [36] and they show in all cases flat maxima. The energies corresponding to these maxima are listed in Table 4; they reach from 1 MeV for photons to 10 MeV for electrons and finally to 1 GeV for pions. The impact of photon and electron irradiation on the superconducting properties of Nb3 Sn wires is in general assumed to be small and will be neglected here. The proton spectrum is of particular interest and is shown in Fig. 25, as an example for charged high energy particles. From Table 4, it is seen that the ratio between the charged high energy particles (proton and pions) and neutrons in the LHC is around 8%. From the results of proton irradiation at 3 MeV [39, 40] and neutron irradiation at 1 MeV [31], a direct comparison of the respective damage can be performed. It is found that the maximum of Jcun /Jcirr for proton irradiation of binary Nb3 Sn is obtained at a fluence of 7 × 1020 p/m2 [39, 40], which is around 30 times lower than the corresponding value for neutron irradiation, 2 × 1022 n/m2 [31]. This shows that charged protons (but in general charged particles) have a considerably higher potential of damage than neutrons. The published proton irradiation data were obtained on Nb3 Sn samples which were thin enough to avoid a Bragg peak inside the sample. However, a realistic damage estimation for charged high energy particles has to be taken into account for the enhanced local damage in the region of the Bragg peak. This condition has to be studied, since in LHC quadrupoles almost all regions where the Bragg peaks occur will be situated in the magnet windings. From these remarks, it follows that the total effect produced by the 8% charged particles on the superconducting properties of Nb3 Sn wires will be considerably higher than that of the 92% neutrons. After this brief discussion, it follows that we do not yet have all elements for estimating the behavior of the critical current densities in Nb3 Sn quadrupoles of the high luminescence LHC upgrade

19

collider during operation. In conclusion, it can be said that the effect of high energy irradiation on the Nb3 Sn quadrupoles is somewhat higher than expected, but that the operational safety margin for the quadrupoles is still sufficient. A series of experiments are presently underway or meeting the challenge of a precise calculation of the irradiation effects in the LHC upgrade. 3. Challenges in Superconducting Applications 3.1. Applications in energy In the preceding sections, the application of superconductors for resonance purposes (MRI and NMR) as well as for accelerators and fusion devices has been briefly discussed. In the following, several emerging energy applications will be mentioned. 3.1.1. Superconducting magnetic energy storage Reliable power delivery will require effective electric storage solutions. Presently, renewable energy sources, such as wind and solar energy, are important sources of power, and it can be foreseen that their importance will increase in future. However, these energy sources have inherently a variable and somewhat uncertain output. The variability of these sources has led to challenges regarding the reliability of the electrical grid. Superconducting magnetic energy storage (SMES) is expected to offer improved performance and efficiency compared with other utility devices. For example, high capacity superconducting cables, with their inherent high efficiencies, will enable a new generation of transmission and distribution electrical grids that can meet the steady increase in demand for electrical power. The performance and efficiency of SMES (98%–99%) are higher than for fuel cells. The energy is stored in the magnetic field of a dc current that flows in a superconducting wire or tape. The consequence of the dc current is low energy losses. A high reliability is expected, the major part of SMES being static. The advantages of SMES are: • Higher power quality, reliability across the transmission and distribution networks;

December 24, 2012

14:49

20

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

• Flexibility and fast response in energy storage and delivery; • Continuous electricity supply by storing excess capacity from intermittent renewable energy generation sources. 3.1.2. Power cables Superconducting cables carry considerably higher current than Cu wires having the same cross section. This is of particular interest in crowded urban areas, where they can replace already existing Cu cables, thus avoiding expensive underground constructions. This procedure, called “retrofit,” leads to strongly enhanced capacity. The last few years saw the implantation of several superconducting cables, mainly using first generation Bi-2223 tapes, with lengths varying by up to 600 m. At present, a cable 1000 m long is under construction in Essen (Germany). First attempts have been made to replace short lengths of Bi-2223 cables by cables based on second generation Y-123–coated conductors (Long Island Port Authority, LIPA, New York, 215 MVA, 138 kV). New cables based on Y-123–coated conductors are planned in Korea (1000 MVA, 154 kV) and in Japan (275 kV@3 kA). However, in spite of these promising projects, the adoption of HTS by the electrical utility industry may possibly be restricted to a small number of cables for very special purposes. In the USA, severe reductions have recently been made in funding for superconducting cables. A high obstacle to introducing superconducting cables is the fact that the cable market is a very conservative one, the average lifetime of a conventional high power Cu cable ranging between 40 and 60 years. In addition, it must be mentioned that, today, overhead transmission and distribution is less expensive than underground transmission. Under these conditions, a reduction in the production costs of HTS-coated conductor tapes in the next few years appears to be essential. In spite of the present situation, it is worthwhile to mention the advantages of superconducting cables in cases where the retrofit solution is envisageable: • Enhancement of transmitted current through existing rights of way; • Enhancement of current over conventional lines by a factor between 5 and 8;

• Transmission of higher currents at lower voltages; • Reduced stray fields. 3.1.3. Fault current limiters Superconducting fault current limiters (SFCLs) limit the current in a crucial branch of the circuit so that no component in the system becomes overloaded. SFCLs work by inserting impedance in a conductor when there is an excess of current on utility distribution and transmission networks. They are invisible to the system, having nearly zero resistance to the steady-state current. At the occurrence of an excess of electricity (otherwise known as a fault current), the SFCL enters into action dissipating it, thus protecting the transmission equipment on the line. It is obvious that the SFCL has the potential to contribute in reducing wide-area blackouts, thus making the modern grid more efficient by protecting high voltage power lines and the electrical grid equipment from damage. The advantages over conventional FCLs are: • Enhanced safety, reliability and efficiency of power delivery systems; reduced blackouts; • Unlike conventional FCLs, SFCLs work in a reproducible way, i.e. there are no circuit breakers to be replaced after a limiting operation; • SFCLs have a reduced recovery time after the occurrence of a disruption; • Possibility of building a fault-current-limiting transformer. 3.1.4. Motors and generators The possibility of carrying higher currents than by Cu cables leads to smaller and more efficient motors and generators. The successful operation of motors up to 36 MW by several manufacturers has demonstrated their reliability. The strong reduction in the volume and total mass of superconducting motors compared to conventional motors leads to enhanced performance. There is a wide field of applications for motors, including magnetically levitated trains, ship propulsion systems, electric cars and even aeroplanes. The advantage of mass reduction appears clearly when one is considering wind generators with power levels of 10 MW and more, where the total mass would be reduced from 800 tons to 400 tons, with strong consequences for the mechanical

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

stability of the whole, >100-m-high system. Superconducting rotating machines have the following advantages: • The higher efficiency will lead to energy conservation; • Reduction by ∼50% in total mass and volume leads to higher flexibility; • Enablement of new transport technologies; • Increased environmental protection through decreased emissions (no CO2 emissions).

3.2. Applications in medicine As mentioned before, there is already a large market for MRI magnets at fields of up to 3 T operating at 1.8 K. The first MRI systems with the MgB2 magnets at higher operation temperatures, around 20 K, have already been built, following the principle of the “open sky magnet.” Here the patient is not enclosed in a narrow tube, but is located in an open space which even allows a direct medical treatment during the analysis. It can be projected that future developments will also include MRI magnets operating at fields well above 3 T. Since for MRI the requirement for the contact resistance is considerably less stringent than for NMR, this will open the choice to HTS materials. Finally, the use of noninvasive diagnosis of human organs like the heart and brain by means of magnetic source imaging and magnetocardiology systems is envisaged.

3.3. Applications in communications Superconductivity has also opened the door to thin film technologies: it has allowed the construction of digital circuits based on rapid single flux quantum technology as well as rf and microwave filters for mobile-phone base stations. Superconductors are used to build Josephson junctions, which are the building blocks of SQUIDs (superconducting quantum interference devices), the most sensitive magnetometers known at present. As a major result, the SI volt unit has been calibrated by a series of Josephson junctions allowing a very precise voltageto-frequency conversion, combined with the cesium133 time reference. Finally, SQUIDs are also used in scanning SQUID microscopes and in medical applications, e.g. magnetoencephalography.

21

3.4. Energy storage Persistent current makes possible two different storage devices: superconducting magnet energy storage (SMES) and flywheel electrical energy storage (FEES). • SMES stores energy in the magnetic field of a coil. The energy is stored because the current persists without substantial power input power, i.e. negligible energy is dissipated in the coil — there is no resistance and no loss. • FEES stores energy in a rotating mass in which rotation persists because magnetic bearings are used. Unlike conventional magnetic bearings, which are unstable without energy-consuming power electronics that provide feedback, superconducting bearings provide stability as well as levitation. 4. Conclusions Storage also provides a necessary (but not sufficient) condition for the widespread use of wind and solar generation. It is often said that the fluctuating, unpredictable availability of wind will limit its use to roughly 10%–20% of generation. Economical energy storage would permit one to considerably increase this number. It is noteworthy that superconductivity presents possibilities for both electrical and electronic applications. There is no doubt that superconducting materials will have a significant role and will contribute to the progress of selected industrial and scientific applications. The main problem of superconducting applications is of course connected with the cooling costs, which may in many cases influence the choice in favor of conventional systems. Major benefits are expected in various sectors, including energy, environment and healthcare. The greatest opportunity for HTS to reduce the threats to the environment, safety and health is the role it may play in contributing to technology for storing shaft power or electrical power, thereby reducing the use of fossil and nuclear fuels. References [1] H. Rogalla and P. H. Kes, 100 Years of Superconductivity (Taylor and Francis, New York, 2012). [2] J. R. Waldram, Superconductivity of Metals and Cuprates (IOP, Bristol, 1996).

December 24, 2012

14:49

22

WSPC/253-RAST : SPI-J100

1230001

R. Fl¨ ukiger

[3] A. C. Rose-Innes and E. H. Rhoderick, Introduction to Superconductivity, 2nd edn. (Pergamon, 1994). [4] C. P. Poole, Jr., H. A. Farach and R. J. Creswick, Superconductivity (Academic, 1995). [5] P. Komarek, Hochstromanwendung der Supraleitung (Teubner Studienb¨ ucher, Angewandte Physik, 1995). [6] T. P. Orlando and K. A. Devlin, Foundations of Applied Superconductivity (Addison-Wesley, 1991). [7] Z. J. J. Stekly, J. L. Zar and L. Hoppie, Design of Superconducting Magnet Systems, AFAPL-TR66-126, Ohio, USA (1967). [8] E. J. Kramer, J. Appl. Phys. 44, 1360 (1973). [9] J. W. Ekin, Cryogenics 20, 611 (1980). [10] L. Gao, Y. Y. Xue, F. Chen, Q. Xiong, R. L. Meng D. Ramirez and C. W. Chu, Phys. Rev. B 50, 4260 (1994). [11] J. R. Gavaler, M. A. Janocko and C. K. Jones, Proc. LT13, Vol. 3 (Plenum, New York, 1972), pp. 558– 562. [12] K. A. M¨ uller, M. Takashige and J. G. Bednorz, Phys. Rev. Lett. 58, 1143 (1987). [13] M. Xue, T. Cao, D. Wang, Y. Wu, H. Yang, X. Dong, J. He, F. Li and G. F. Chen, Scientific Reports (Nature.com), 2, article No. 389, Apr. 2012 (doi:10.1038/srep00389). [14] Y. Kamihara, T. Watanabe, M. Hirano and H. Hosono, J. Am. Chem. Soc. 130, 3296 (2008). [15] D. Weiss, C. Tarantini, J. Jang, F. Kametani, A. A. Polyanskii, D. C. Larbalestier and E. E. Hellstrom, Nat. Mater. 11, 682 (2012). [16] D. Larbalestier, see Ref. 1, pp. 627–648. [17] LINK project, CERN, directed by A. Ballarino. [18] C. Senatore, P. Lezza and R. Fl¨ ukiger, Adv. Cryo. Engrg. 52, 654 (2006). [19] Bruker BioSpin press release, June 1, 2009, Dr. T. Thiel, Director, Marketing Communications. [20] T. Boutboul, L. Oberli, A. den Ouden, D. Pedrini, B. Seeber and G. Volpini, IEEE Appl. Supercond. 19, 2564 (2009). [21] H. W. Weijers, U. P. Trociewitz, W. D. Markiewicz, J. Jiang, D. Myers, E. E. Hellstrom, A. Xu, J. Jaroszynski, P. Noyes, Y. Viouchkov and D. C. Larbalestier, IEEE Trans. Appl. Supercond. 20, 576 (2010).

[22] F. Kametani, T. Shen, J. Jiang, C. Scheuerlein, A. Malagoli, M. di Michiel, Y. Huang, H. Miao, J. A. Parrell, E. E. Hellstrom and D. C. Larbalestier, Supercond. Sci. Technol. 24, 075009 (2011). [23] R. Fl¨ ukiger, D. Uglietti, C. Senatore and F. Buta, Cryogenics 48, 293 (2008). This article summarizes the effects of Ta and Ti additives on Bc2 of Nb3 Sn. [24] S. Soltanian, X. L. Wang, I. Kusevic, E. Babic, A. H. Li, H. K. Liu, E. W. Collings and S. X. Dou, Physica C 361, 84 (2001). [25] W. Schauer and W. Schelb, IEEE Trans. Magn. MAG-17, 374 (1981). [26] J. L. MacManus-Driscoll et al., Nat. Mater. 3, 439 (2004). [27] V. Selvamanickam, DOE HTS Peer Review (2010). [28] T. Holesinger et al., in BES Report on Basic Research Needs for Superconductivity (Mar. 6, 2006). [29] M. Rupich, J. McCall and C. Thieme, DOE Peer Review (2010). [30] R. Fl¨ ukiger, The irradiation effects in low Tc superconductors, CERN Internal Report (2008). [31] H. W. Weber, Adv. Cryo. Engrg. 32, 853 (1986). [32] F. Weiss, R. Fl¨ ukiger, W. Maurer, P. A. Hahn and M. W. Guinan, IEEE Trans. Magn. MAG-23, 976 (1987). [33] C. Senatore, V. Ab¨ acherli, M. Cantoni and R. Fl¨ ukiger, Supercond. Sci. Technol. 20, S217 (2007). [34] M. Eisterer, R. Prokopec, R. K. Maix, H. Fillunger, T. Baumgartner and H. W. Weber, presented at the RESMM’12 Workshop at Fermilab (Feb. 2012). [35] M. Putti, R. Vaglio and J. M. Rowell, Supercond. Sci. Technol. 21, 043001 (2008). [36] F. Cerutti, presented at the WAMSDO Workshop at CERN (Nov. 24, 2011). [37] A. R. Sweedler, D. G. Schweizer and G. W. Webb, Phys. Rev. Lett. 33, 168 (1974). [38] M. W. Guinan, R. A. van Konynenburg and J. B. Mitchell, Internal Report UCID-20048. Lawrence Livermore Lab., USA (1984). [39] H. J. Bode and K. Wohlleben, Phys. Lett. A 24, 25 (1967). [40] I. V. Voronova, N. N. Mihailov, G. V. Sotnikov and V. J. Zaikin, J. Nucl. Mater. 72, 129 (1978).

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230001

Overview of Superconductivity and Challenges in Applications

Ren´ e Louis Fl¨ ukiger received his Ph.D. from the University of Geneva in 1972. He was an invited scientist at the Francis Bitter National Magnet Laboratory, MIT, Cambridge, USA from 1977 to 1979, and then served as department leader at the Inst. Tech. Physics, Nuclear Research Center, Karlsruhe, Germany till 1990. In 1990, he became Full Professor for Experimental Physics and Director of the Solid State Physics Department at the University of Geneva. Professor Fl¨ ukiger won the IEEE Award (International Electronic Engineering Association) for “lifetime developments in the system Nb3 Sn” in 2005, and the ICMC Cryogenic Materials Award for “lifetime’s achievement in advancing the knowledge of cryogenic materials” in 2011. He was a Guest Professor at the Institute of Electric Engineering, Beijing, China in 2008, and has been a CERN Associate with the Magnet group since 2008. His recent research subjects include fabrication and study of prototype MgB2 superconductors and study of irradiation problems in the LHC at CERN. Professor Fl¨ ukiger is married with two children.

23

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

This page intentionally left blank

1203001˙book

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Reviews of Accelerator Science and Technology Vol. 5 (2012) 25–50 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300022

Superconducting Materials and Conductors: Fabrication and Limiting Parameters Luca Bottura CERN, CH-1211 Geneve 23, Switzerland [email protected] Arno Godeke Ernest Orlando Lawrence Berkeley National Laboratory, 1 Cyclotron Rd., Berkeley, CA 94720, USA [email protected] Superconductivity is the technology that enabled the construction of the most recent generation of high-energy particle accelerators, the largest scientific instruments ever built. In this review we trace the evolution of superconducting materials for particle accelerator magnets, from the first steps in the late 1960s, through the rise and glory of Nb– Ti in the 1970s, till the 2010s, and the promises of Nb3 Sn for the 2020s. We conclude with a perspective on the opportunities for high-temperature superconductors (HTSs). Many such reviews have been written in the past, as witnessed by the long list of references provided. In this review we put particular emphasis on the practical aspects of wire and tape manufacturing, cabling, engineering performance, and potential for use in accelerator magnets, while leaving in the background matters such as the physics of superconductivity and fundamental material issues. Keywords: Superconductors; LTS materials; HTS materials; superconducting cables; superconducting accelerator magnets.

1. A Perspective on Applied Superconductivity for HEP Magnets

used to produce, for example, the background field in a particle detector. The initial applications were plagued by flux-jump instabilities and required much ingenuity to achieve practical results. Swift progress in the technology, and an undeniable pioneering spirit, made it possible to achieve remarkable results. The most relevant examples are the large-size bubble chambers at Argonne National Laboratory [3–5], BNL [6], and the Big European Bubble Chamber at CERN [7], which, with a stored energy of 800 MJ, remained for some 35 years the largest-size single magnet ever built. Various materials were considered in that early age. Cold-drawn Nb wires were the initial step toward large current-carrying capability [8, 9]. Mo3 Re and Nb3 Sn were the first high-field superconductors [10, 11], but their use did not find widespread application due to the difficulty inherent with the material brittleness, and large filaments. On the other hand, the earlier recognition of the effect of cold work and precipitates in Nb wires led to the development of Nb–Zr, the first commercially available

High-energy physics (HEP) has been a strong driver in the development of technical superconductors. The potential of superconductors for HEP experiments became clear already in the early 1960s, when the groups at Bell Labs, Westinghouse, Atomics International, Lincoln Labs, and Oak Ridge Laboratories reached fields well above 2 T, the typical range of classical electromagnets. By 1961, fields of 6 T were achieved in solenoids using a cold-drawn Nb–Zr alloy by Hulm from Westinghouse and Richard Hake and Ted Berlincourt from Atomics International, and record values of 7 T were reported by John Eugene Kunzler, at Bell Labs [1], using a wire made of Nb tubes filled with crushed powder of Nb and Sn, and heat-treated at 1000◦C to chemically react the precursors into Nb3 Sn [2]. The superconducting wires developed at that time had critical current densities in the range of a few hundred to a thousand A/mm2 , and could be employed to wind magnets of medium-to-large size 25

December 24, 2012

14:49

26

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

superconductor, and later Nb–Ti, which has become the material with the largest application basis, and undoubted success [12]. Indeed, Nb–Ti has been the workhorse for the vast majority of HEP applications. Multifilamentary Nb–Ti wires became widespread in the late 1960s and 1970s, while scientists understood the need for twisting and cabling [13, 14]. Different choices of cable configurations fueled the dispute between the Energy Doubler/Saver [15], known later as the Tevatron [16], and Isabelle [17], later christened the Colliding Beam Accelerator (CBA) and terminated by DOE in favor of the Superconducting Super Collider [18]. The SSC prototypes [18], the proton ring at the Hadron Electron Ring Accelerator (HERA) [19], the UNK prototypes [20], and the Relativistic Heavy Ion Collider (RHIC) [21] were all built using Rutherfordtype cables made of multifilamentary Nb–Ti wires. The latest achievement is the Large Hadron Collider (LHC) [22], which will soon be pushing the operating field to the 8 T range, the practical boundary of Nb–Ti for this class of applications. The companion paper [23] contains an abridged history of the research and achievements in the field of superconducting accelerator magnets. Similarly, Nb–Ti cables have found widespread application in the magnets providing the bending field for HEP detectors. The main advantage with respect to resistive electromagnets is the possibility of generating the desired magnetic field configuration and strength with a very thin winding, i.e. transparent to the particles to be detected and tracked. The first detector solenoids were built in the 1970s at the CERN ISR [24], PEP-4 at SLAC [25], and CLEO1 at Cornell University [24]. A major technological advance was the Al-stabilized Nb–Ti conductor that was developed in the late 1970s at CEN-Saclay, for the CELLO detector solenoid at DESY [26]. The Alstabilized conductor had the advantage of improved stability to external perturbations (see later), which is a very interesting feature when one is dealing with magnets of large size that are subjected to continuous radiation loads. This conductor, in a number of configurations, has found applications in all large HEP magnets built during the last 40 years, from CDF at the Tevatron [27], through ALEPH [28] and DELPHI [29] at LEP, to the most recent examples of ATLAS [30] and CMS [31] at the LHC. This last magnet, with a stored energy of 2.6 GJ, holds today

the record for the single magnet with the largest stored energy. At this point in time Nb–Ti has reached industrial maturity and is operating close to its intrinsic limitations (refer also to the later discussion). It is clear that it will be difficult, if not impossible, to surpass the latest achievements without resorting to new materials. To exemplify this situation, we report in Fig. 1 the evolution in time of the critical current density Jc of industrial Nb–Ti wires in the past 35 years. Throughout this article we define Jc as the ratio of the critical current Ic to the non-stabilizer area in the wire Anon-stabilizer : Jc =

Ic . Anon-stabilizer

(1)

As heritage from LTS materials, the non-stabilizer cross section is commonly referred to as non-Cu, which is applicable to Nb–Ti and Nb3 Sn, but no longer appropriate for HTS materials. We observe in Fig. 1 that the non-Cu critical current density at 5 T and 4.2 K was dramatically increased during the transition from the Tevatron production to the SSC R&D, in the 1980s (see also later). This gain was beneficial to industrial applications, foremost being MRI and NMR magnets, and was later exploited in the construction of the LHC. We see that the level achieved in the late 1980s has improved by as little as 10% over the past 20 years, which suggests that the full industrial potential of this material has been reached. In the same plot we report the evolution of the critical current density of Nb3 Sn, which we believe is the material that will provide the next step in

Fig. 1. Evolution of the critical current of industrially produced Nb–Ti (left) and Nb3 Sn wires (right). The critical current of Nb–Ti is quoted at 5 T and 4.2 K, for Nb3 Sn at 20 T and 4.2 K. (Data courtesy of J. Parrell, Oxford Superconducting Technology, Carteret, NJ, USA.)

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

superconducting accelerator magnets, measured at 20 T and 4.2 K. We note a jump in Jc in the early 2000s, which is the result of the Conductor Development Program (CDP) supported by DOE since the year 1999 [32]. This improvement was achieved thanks to the choice of optimal Sn content and nonCu fraction. The record current density achieved in commercial wires in 2005 is still unsurpassed, while work focusses presently on other properties such as copper RRR, magnetization, stable production, and yield. This vigorous R&D is still ongoing to prove that Nb3 Sn is mature for HEP applications in the range of 15 T [33–35]. Results are expected by 2015, based upon which the production of the first Nb3 Sn accelerator magnets could start, for installation in the LHC in the early 2020s. The classic low-temperature superconductors (LTSs) are only a part of the present landscape of commercial superconductors. Figure 2 shows the state-of-the-art current density in LTS and HTS materials of interest for magnet construction [36]; here the authors report the engineering current density JE obtained as the ratio of the critical current Ic to the total cross-section of the conductor (strand or tape) Aconductor, including stabilizer, barriers, and structural or buffer materials: JE =

Ic . Aconductor

(2)

27

HTS materials have exceedingly high critical fields at low temperature, 4.2 K in Fig. 2, and they possess great potential as high-field superconductors. This potential can be exploited to push further the field in accelerator magnets, beyond the limit of about 18 T for Nb3 Sn. This target is being pursued by a number of companion programs worldwide [37, 38], targeting specifically the Bi2 Sr2 CaCu2 O8+x (Bi-2212) and YBa2 Cu3 O7−δ (YBCO) class of cuprates. Results are expected on the horizon of 2015, by which time further demands from HEP may call for higher particle energies and larger accelerators. In the meantime, the HTS materials have found a first large-scale use in the current leads of the LHC (Bi-2223). In the near future, HTS materials, as well as MgB2 , are considered for applications in superconducting bus bar cables in the LHC [40]. These superconducting links would allow placing the powering equipment, including power converters, control electronics, and possibly protection diodes, in zones that are well shielded from radiation. This application is a major scale-up of a concept realized using LTS materials, such as the bus powering the J-PARC combined function magnets at KEK [41]. With material demands in the range of thousands of kilometers, these power transmission applications provide a definite push for the industrialization of novel materials, which are still in their industrial infancy compared to Nb–Ti and Nb3 Sn.

Fig. 2. State-of-the-art engineering current density for technical superconductors (at 4.2 K except for Nb–Ti, which is quoted at 1.9 K). (Data courtesy of P. J. Lee, Applied Superconductivity Center, NHMFL, Tallahassee, FL, USA.)

December 24, 2012

14:49

28

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

2. Relevant Parameters for Application in Accelerator Magnets A good superconducting material must have a number of specific properties that make it suitable for practical accelerator magnet applications. They are listed and discussed below, with typical order-ofmagnitude estimates based on common practice in accelerator magnet design and construction.

2.1. Critical current density A practical and economical superconductor has a high critical current density and operating margins under the desired operating conditions of field and temperature. An efficient winding requires a coil current density Javg , averaged over all coil components, in the range of 400–500 A/mm2 . Significantly smaller values result in coils that are too large and too costly [23, 42]. Much larger values in magnets producing the strong fields of interest here would yield mechanical stresses in the winding exceeding allowable levels for the superconductor, or other coil components such as the insulation, and may be difficult to protect in case of quench. The above target on Javg can be translated in a typical requirement for

the engineering current density in the superconducting wire or tape, i.e. JE (defined earlier), or in the non-Cu fraction, i.e. Jc , by considering the processing of wires and the amount of material in a coil cross-section. Figure 3, which is an example for a Nb3 Sn cable, shows schematically the various material distributions. Cabling a superconducting wire, e.g. in the form of a Rutherford cable described later, can induce a degradation of the critical current caused by heavy deformation of the filaments. Acceptable degradation is in the range of 5–10%. The void fraction of a Rutherford cable, combined with the space occupied by the cable insulation, further reduces the achievable current density in the windings by typically 30% for the wire in the cable, which means that, overall, the current density in the windings is reduced by about 35% when compared to the round wire before cabling. A required winding current density Javg of about 400–500 A/mm2 needs therefore an engineering current density JE of about 600–800 A/mm2 in the wire. In a typical accelerator wire, the non-Cu fraction (i.e. the superconductor, as well as barriers, substrates, and residual from possible heat reactions) constitutes 50% of the wire cross-section. The

Fig. 3. Rutherford cable, wire, a typical filament cross-section of the non-Cu fraction, and a detail of the fine-grain volume in a powder-in-tube Nb3 Sn wire. The approximate cross-sectional fractions in the non-Cu area are: fine-grain Nb3 Sn — 40%; large-grain Nb3 Sn — 10%; core — 25%; and unreacted Nb or Nb–Ta — 25%. (Reproduced from Ref. 39; wire and filament images courtesy of P. J. Lee and C. M. Fisher, Applied Superconductivity Center, Tallahassee, FL, USA; cable image courtesy of L. Oberli, CERN, Geneva, Switzerland.)

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

requirement for Jc , which is defined here over the wire cross-section excluding the stabilizer fraction in accordance with the common convention, is thus about 1200–1600 A/mm2 . Finally, as shown in Fig. 3, only a fraction of the non-Cu cross-section may be superconducting, depending on the specific material. In the example shown there, a powder-in-tube (PIT) Nb3 Sn wire, it is seen that the fine-grain Nb3 Sn volume, which carries the superconducting current, amounts to about 40% of the non-Cu cross-section, and thus only about 20% of the total wire crosssection. The requirement for the current density in the superconducting fraction, or JSC , is therefore of order 3000–4000 A/mm2 , which is available in modern Nb3 Sn wires up to about 15 T at 4.2 K [43]. The above ranges for JE and Jc are good guidelines for estimating the magnetic field reach of a given material. As an example, we see, examining the plot in Fig. 2, that Nb–Ti is suitable for applications up to 10 T at 1.9 K, while Nb3 Sn extends this reach to 16 T at 4.2 K, which agrees with the achieved dipole magnetic field records [44, 45]. It should be noted that in large-scale applications the magnets will not be operated so close to their limits, and that a typical margin of about 20%, measured along the magnet load line, is commonly observed. HTS materials are approaching the range of practical interest, but still require an improvement of JE by a factor of 3–4 (Bi-2212; Sec. 4), or an increase in the superconducting fraction and/or a reduction in the anisotropy (YBCO; Sec. 4) to become effective in a compact magnet, and enable the magnetic field range to be extended beyond the performance of the classic LTS conductors. 2.2. Stabilizer The superconducting material must be compatible with a high-electrical-conductivity material (the most common material is Cu, but Al and Ag are also used), to be placed in intimate electrical and thermal contact with the superconductor itself. This so-called stabilizer is necessary for both stabilization and protection, carrying current when a (local) partial transition to the normal state occurs inside the superconducting filament. In a superconducting wire or tape for HEP magnets, the typical ratio λ of stabilizer to superconductor is in the range of 1–2. Lower values of λ are usually not feasible for reasons of protection, given the large energy stored per unit volume

29

of conductor in accelerator and detector magnets. Larger values of λ, as could be demanded for protection, are not practical as the conductor manufacturing may become problematic, and the conductor’s efficiency decreases. The stabilizer must maintain its properties through the wire and coil processing, and in particular achieve a high residual resistivity ratio (RRR), ideally combined with a low magnetoresistance. For the high-purity copper commonly used in LTSs, a good target is an RRR of about 100 or higher, which corresponds in zero magnetic field to a resistivity in the range of 1.6 × 10−10 Ω m or less. It should be noted that the RRR reduces owing to the magnetoresistance of the material, which has a substantial effect especially on high-purity materials. 2.3. Magnetization, flux jumps, AC loss A magnetic field change induces shielding currents in a superconductor that do not decay. For this reason these currents are referred to as persistent. The magnetic moment per unit volume M associated with persistent currents is proportional to the current density of the shielding currents, Jc , and the characteristic size of the superconductor, D: M ≈ Jc D.

(3)

The magnetization can attain large values at low background field, where Jc is large, and when the superconductor has a large dimension exposed to the field change. In LTSs, µ0 M at the typical injection fields of modern synchrotrons (e.g. 0.5 T for the LHC) is in the range of several tens of mT (Nb–Ti) to hundreds of mT (Nb3 Sn). Such a magnitude is sufficient to perturb the field generated by the magnet, and hence requires careful control and compensation. Furthermore, if the magnetic energy stored in the shielding currents becomes sufficiently large compared to the heat capacity of the superconductor, the magnetization can collapse suddenly in a process referred to as a flux jump [46]. Flux jump instabilities plagued early magnets built with monofilamentary wires and tapes, whose characteristic dimension was of the order of a millimeter. Flux jumps can be controlled by subdividing the superconductor into small filaments, which both decreases the magnetization and improves the magnetic and thermal stabilizing effect of the low-resistance matrix. As with the bulk behavior described above, field variations induce shielding currents between

December 24, 2012

14:49

30

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

the superconducting filaments. These currents couple the filaments electromagnetically by finding a return path crossing the conductor matrix. The amount of filament coupling depends on the resistivity of the matrix, which has to be low for good protection, and the geometry of the current loop. In the extreme case of wires and tapes with untwisted filaments, coupling currents could travel along large lengths (such as the kilometer length in a magnet) and find a low crossresistance. The net effect would be that the multifilamentary matrix would respond to field changes as a single bulk filament, losing the advantage of fine subdivision. Decoupling of the filaments is achieved by shortening the current loop, by twisting the wire with a typical pitch on the order of a few millimeters. 2.4. Mechanical properties Superconducting wires in an accelerator magnet operate under large deformations as a result of the thermal contraction differences during cooldown and the high Lorentz forces that develop during charging of the magnet. The strain on the superconductor that is generated by the thermal contraction differences is predominantly longitudinal, and typically on the order of a few tenths of a percent compressive strain, since the wire matrix and the magnet construction materials mostly exhibit a larger thermal contraction than the superconducting materials. The Lorentz loads are predominantly transverse and result in operating stresses in the range of 50– 100 MPa in the present accelerators, and up to projected values of 150–200 MPa for high-field magnets made of Nb3 Sn. An obvious way to reduce the strain due to the difference of thermal contraction is to choose magnet construction materials that match the thermal contraction of the superconducting material. In practice, this is not always possible due to conflicting demands on construction materials. As to transverse loading, the stress generated by the Lorentz load depends on the current density and the thickness of the winding pack [23]. A thin winding pack with high Javg makes very effective use of the superconductor, but experiences higher stress than a thicker winding pack with a lower Javg . On the other hand, there are limits to the thickness of the winding given by available space, and a thick winding becomes ineffective from the point of view of the field generated, the amount of material required, and the cost. In

practice, Javg and winding thickness are a compromise between the effective use of the conductor, and acceptable Lorentz loads. The role of the Lorentz loads in this compromise becomes more important in high-field magnets, where the use of brittle Nb3 Sn and HTS conductors results in magnet designs that are stress-limited to the maximum acceptable by the material. Among the possible mitigation measures, the winding of Nb3 Sn high-field magnets is precompressed at room temperature and during cooldown, so that the peak field (and minimum margin) region unloads when the magnet is powered [47, 48]. Understanding the superconductor response to mechanical loads is of paramount importance, especially in the perspective of the use of superconductors for the next generation of high-field accelerator magnets. The behavior of superconductors under stress and strain is mainly determined through axial strain experiments on the wires and tapes [49–58], and through transverse pressure experiments on Rutherford cables [33, 59–65]. In the sections below we review the main results obtained to date. 2.4.1. Axial strain sensitivity A summary of the typical behavior of the critical current as a function of axial strain for the various superconductors is presented in Fig. 4. The behavior of Nb–Ti under axial strain is omitted, since Nb– Ti only exhibits relevant sensitivity to axial strain above 1% [66], which is beyond the strain range that is acceptable in magnets. Nb3 Sn and YBCO both

1

?

0.8

0.6

0.4

3

0.2

0

0

0.2

0.4

0.6

Fig. 4. Typical variations of the critical current as a function of axial strain for Nb3 Sn [49, 50], Bi-2212 [51, 52], and YBCO [55–58].

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

exhibit a reversible sensitivity to axial strain. Nb3 Sn can be axially compressed to beyond 1% strain, and the Ic will fully recover when the axial compressive strain is released. In axial tension, the reversibility only occurs up to the so-called irreversibility strain limit (irr ), beyond which cracks start to form in the Nb3 Sn volume. The magnitude of irr depends on the wire layout, and appears also strongly related to the inner architecture of the strand (e.g. the filament or subelement diameter) as well as the use of dopants (e.g. Ti appears to yield a higher irr than Ta). The typical (reversible) reduction in Ic in Nb3 Sn amounts to −50% at −0.5% compressive strain at 4.2 K and 16 T (Fig. 4). The sensitivity of the Ic of Nb3 Sn to axial strain, i.e. the steepness of the reduction, increases with the magnetic field and with the temperature, i.e. when approaching the field– temperature superconducting phase boundary [49]. The strain sensitivity of YBCO was initially dominated by crack formation in the YBCO layers, and it is only recently that reversible behavior has been observed [54]. The observation of reversibility in modern YBCO conductors is partly a result of improvements in measurement techniques, and likely also due to improvements in conductor fabrication, where a homogeneous YBCO layer is well bonded to a strong substrate that prevents local stress concentrations. Since YBCO was until recently mainly developed for the electric utility industry, most measurements were so far performed at liquid nitrogen temperature, and the material exhibits a sensitivity to strain at 76 K that is comparable to Nb3 Sn at 4.2 K (see Refs. 55 and 56, Fig. 4). More recent reversible data at 4.2 K and medium magnetic field [57, 58] show a striking reduction in the sensitivity to strain, perhaps due to the fact that at these magnetic fields and temperatures the material is very far away from the field–temperature superconducting phase boundary. A virtual absence of strain sensitivity over an axial strain range that is about 1% wide holds significant promise for the application of YBCO in high-field magnets. The behavior of Bi-2212 under strain is, in contrast to Nb3 Sn and YBCO, to a large extent dominated by irreversible reductions in the critical current, which can be attributed to crack formation in the Bi-2212 [52, 67]. For many years, the behavior of the Ic of Bi-2212 under axial strain could be described according to the following model that

31

was introduced in the mid-1990s [51]. A compressive strain, such as that introduced by the larger thermal contraction of the Ag–alloy matrix compared to the Bi-2212, causes an initial, irreversible reduction of Ic (point a to point b in Fig. 4). A further increase in the compressive strain causes a further irreversible reduction of the Ic (point b to point b’). When tensile strain is applied (point b to point c), the critical current reduces only slightly compared to the strong initial irreversible reduction, until at point c cracks appear and the Ic collapses irreversibly. The plateau between points b and c is quasireversible: When the strain is cycled along this plateau, a slight initial further irreversible reduction of Ic occurs, until about 10–50 cycles, after which the Ic reduction stabilizes [68]. In more recent work [52], increased reversibility was observed, and it is suggested that the irreversibility is a result of local stress concentrations around the porosity and voids that plague the present Bi-2212 wires (see Sec. 4), and that the behavior on the plateau is in principle intrinsic and related to strain-induced changes in the field–temperature phase boundary [53], similar to Nb3 Sn [49], and as recently suggested for YBCO [55–57]. Overall, these trends hold promise for when Bi-2212 wires can eventually be made dense, which is required to increase the current density (Sec. 4), even though Bi-2212 might still be hindered by the fact that a brittle web of Bi-2212 is present in a soft matrix, and local stress concentrations will inevitably develop under load. Nevertheless, ongoing developments on round wire Bi-2212 will likely change its behavior under stress and strain. As it is, even with the present material, and accepting an initial loss of 20% of Ic , it seems that Bi-2212 can accept the strain range of 1% relevant for magnets.

2.4.2. Transverse pressure on cables The superconducting cables must be capable of tolerating, reversibly, transverse stress up to 150– 200 MPa, and the associated deformations, including all stresses at intermediate steps during magnet manufacturing and cryogenic operation. Such large loads are not problematic in Nb–Ti Rutherford cables, due to the limited sensitivity to strain in combination with the ductility of the alloy. However, for Nb3 Sn, with its brittleness and large sensitivity to strain,

December 24, 2012

14:49

32

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

large transverse loads can be an issue. A large number of transverse pressure experiments have been performed on Nb3 Sn Rutherford cables during the 1990s at the University of Twente in The Netherlands in an 11 T solenoid [59–62], and by Lawrence Berkeley National Laboratory, Berkeley, CA, USA (LBNL), at the National High Magnetic Field Laboratory in Tallahassee, FL, USA in an 11 T split pair magnet [63, 65]. These measurements were performed to establish the acceptable load limits for the first Nb3 Sn dipole magnets poised to surpass the 10.5 T magnetic field record of dipoles made using Nb–Ti technology [44, 69], namely Twente’s 11 T MSUT magnet [70] and LBNL’s 13.5 T D20 [71]. The data from the different transverse pressure systems are comparable [33], and established the 150– 200 MPa load limit for Nb3 Sn accelerator magnets which is still used. The benchmark 150–200 MPa safe limits were established from measurement on medium-currentdensity wires. The current density in Nb3 Sn increased by more than a factor of 2, from 2000 to 2005 (Fig. 1), and to determine whether similar transverse pressure limits were valid for modern high-current-density wires, new transverse pressure experiments were performed. The more recent measurements were performed on wires to try to establish a more economical alternative to complex and expensive full-size cable experiments. The results from transverse pressure experiments on wires [72–76], however, showed a significantly larger reduction in the critical current with transverse pressure, which triggered worries as to whether high-current Nb3 Sn Rutherford cables would be more sensitive to transverse pressure, and therefore unsuitable for veryhigh-field magnet designs. These results contrasted with the successful test results from magnets, which were loaded to the 150–180 MPa region [77]. It was found that the larger Ic reduction observed in singlewire experiments was due to insufficient side support, as well as the absence of epoxy impregnation in some cases, which were shown to be defining factors in earlier cable experiments [59]. The intuitive explanation is that transverse loading in one direction, such as loading a round wire between parallel plates, causes stress concentration at the contact and strong shear stresses inside the wire. Local stress concentrations and shear stresses are strongly reduced once the wire has a lateral support, as provided by the

neighboring wires of a cable. In addition, the epoxy that impregnates the wire tends to redistribute the local transverse force and produce a stress state that approaches hydrostatic conditions. This has been proven recently on modern, high-Jc Nb3 Sn wires [78], confirming that the large observed degradation in single-wire measurements quoted earlier is a measurement artifact. A defining test of the US LHC Accelerator R&D Program (LARP) TQS03 magnet at CERN [79], in which the magnet was preloaded up to 260 MPa in the windings while experiencing very limited permanent degradation, re-established Nb3 Sn firmly as the conductor of choice for future upgrades of the LHC. In contrast to the well-established data on LTS Rutherford cables under transverse loads, little is known for the HTS conductors. To the authors’ knowledge, there exists only one transverse pressure experiment on Bi-2212 Rutherford cables [64], while the transverse pressure effects on YBCO Roebel cables are, so far, not documented in the available literature. The limited data on Bi-2212 Rutherford cables suggest a less favorable sensitivity to transverse pressure, and indicate irreversible damage above perhaps already 60 MPa. It should be noted that these data stem from older-generation Rutherford cables, in which the Bi-2212 has perhaps a high void fraction (see Sec. 4). Nonetheless, even at 100% dense Bi-2212, the prospect of high transverse loads on a brittle web of Bi-2212 in a soft matrix does not seem favorable, but without available data such statements are highly speculative. New characterizations of further-optimized Bi-2212 conductor, in which the Bi-2212 ideally approaches a 100% density, should determine what loads are acceptable, and to what extent internal reinforcements in the windings should be included (at the cost of current density) to intercept the accumulation of Lorentz loads. 2.5. Manufacturing properties A good superconductor must be easy to manufacture and be cost-effective. An efficient manufacturing process, associated with high yield, is a prerequisite to achieving low cost. For this reason a good indicator is the piece length, which should be in the range of 1 km and longer for HEP applications. Processing after magnet winding, and material compatibility, are additional parameters to be considered in the

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

final cost. One such example is Nb3 Sn, which needs a high-temperature heat treatment (> 600◦ C) for the formation of the superconducting phase, which is not compatible with organic insulators. Further, HEP magnets put large values on homogeneity of performance, as well as consistent production over a typical project time span, which can be of several years from inception to operation in an accelerator. Indeed, as shown by the example of the Nb– Ti production for the LHC [85], homogeneity of the critical current, stabilizer fraction, magnetization, wire geometry and mechanical properties can have a vital impact on the final performance of the magnet system. 3. State-of-the-Art Conductors for HEP 3.1. Nb–Ti 3.1.1. Discovery and beginnings The alloy Nb–47%Ti [80] is undoubtedly the most successful practical superconductor to date. Nb–Ti rapidly followed Nb–Zr as one of the first commercial superconductors. These two alloys were mechanically very tough, as they were originally developed for high-strength rivets. Among the two, Nb–Ti is easier to manufacture, possesses superior mechanical strength, and has a 2 T advantage in the upper critical field. For these reasons Nb– Ti became the dominating conductor used in the magnets built at that time, and long after [12]. Fabrication of Nb–Ti became an industrial-grade process with the advent of the multifilamentary conductor made by Prof. R. Rose and his MIT group in 1964 [81, 82] using a process developed by F. Levi [83] and first commercially produced by Imperial Metal Industries, Ltd. 3.1.2. State of the art The initial industrial productions, for example used for the construction of the Tevatron, had a specified current density Jc of 1800 A/mm2 at 5 T and 4.2 K, and a filament diameter of about 10 µm, which at the time was a significant production challenge. The understanding of the physical mechanisms of pinning, and in particular the key role of high homogeneity in the initial Nb–Ti alloy in combination with advanced processing to cause a fine distribution of precipitated normal-conducting

33

α-Ti pinning centers, led to a dramatic jump of Jc in the years 1980–1985. This performance improvement was the fruit of a number of activities within the R&D program coordinated by the SSC Central Design Group [84]. The resulting material, dubbed Hi–Ho Nb–Ti, has become the present industrial standard. Nb–47%Ti has a critical temperature of 9.2 K, and a critical field of 14.5 T. At 4.2 K and 5 T the non-Cu critical current density Jc is approximately 3000 A/mm2 , while at 7.5 T and 4.2 K Jc drops to 1500 A/mm2 . Cooling to 1.9 K results in this point being shifted in field by 3 T, up to 10.5 T. This field range represents the upper (quench) limit for the use of Nb–Ti in accelerator dipole magnets, leading to a safe 8 T operational limit in large-scale accelerator applications. Nb–Ti is easily available in long lengths (a few km piece length) in the form of multifilamentary wires where the superconductor is dispersed in a copper matrix of high purity, with an RRR comfortably in the range of 200. A thin Nb barrier separates the Nb–Ti alloy from the copper, a heritage of initial developments that wished to avoid the formation of inter-metallics of Cu and Ti during high-temperature extrusion and annealing heat treatments. One example of industrial wire, a double-stack LHC inner strand, is shown in Fig. 5. It contains approximately 9000 filaments of 7 µm diameter in a matrix with an outer diameter of 1.065 mm and a Cu:Nb–Ti ratio of 1.65. Standard industrial production yields filament sizes of a few µm (5–10), which is mandatory for reducing the field perturbations induced by persistent current magnetization. A typical value of magnetization due to persistent currents is shown in Fig. 6 for an LHC Nb–Ti strand with a filament diameter of 6 µm. The magnetization at the LHC injection field, around 0.5 T, is approximately µ0 M ≈ 10–15 mT, which is a representative value for this product. Filaments in this range of dimensions are also stable against flux jumps, which make the behavior reproducible and easier to control. Finally, homogeneity of Nb–Ti production is excellent, at the level of a few percent for key parameters such as critical current, magnetization, wire composition, and geometry [85]. Approximately 2000 tons per year are fabricated worldwide, mainly for MRI applications. The typical cost for the HEP-grade Nb–Ti described here is in the range of around 1–2 EUR/kAm (evaluated at 5 T and 4.2 K).

December 24, 2012

14:49

34

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

Fig. 5. One of the multifilamentary Nb–Ti strands used in the LHC. It has a diameter of approximately 1 mm, and each Nb–Ti filament (shown in the detail micrograph) has a diameter of 7 µm. The matrix is pure copper.

Fig. 6. Magnetization loops measured on an LHC Nb–Ti strand for the outer layer of the dipole magnets (LHC reference strand). The Nb–Ti strand has filaments of 6 µm geometric diameter. (Data courtesy of D. Richter and B. Bordini, CERN, Geneva.)

3.1.3. Challenges The research on basic understanding of the mechanisms of pinning in Nb–Ti, an activity that unfolded in the late 1980s, during the years of the SSC R&D, was naturally associated with the attempt to exploit the full potential of the material in terms of critical current. The best results were obtained using artificial pinning centers (APCs) and ternary alloying with Ta. APC Nb–Ti, where the additional pinning centers are provided by a Nb or Cu structure in the Nb–Ti, has achieved critical current density in excess of 10,000 A/mm2 at 4.2 K and 5 T [86, 87]. The APC samples are in general layered structures, fabricated by sputtering and photolithography, i.e. a process that provides highly anisotropic properties,

and far from industrial practice. The addition of Ta in the range of 15–25% produces a ternary alloy that has a critical field increased by approximately 1 T, at the loss of some of the α-Ti-precipitate pinning centers [84]. These R&D results show that there may still be room for improvement. None has become an industrial standard, however — possibly influenced by the much greater potential of Nb3 Sn. A second field of active research is the application of Nb–Ti to fast-cycled accelerator magnets [88]. This is a relatively old target, already pursued in the mid-1970s when superconducting cycled magnets were developed at Rutherford Laboratory [89], CEA, and the Kernforschungszentrum Karlsruhe [90]. Among the prospective applications at the time was a superconducting option for the CERN Super Proton Synchrotron (SPS) [91], then built using resistive magnets. The main focus of research at the time was on how to reduce the ac loss associated with field sweeps in the superconducting filaments (hysteresis loss), as well as in the wires and cables (coupling loss) on cycled magnets, and this work boosted the understanding of the role and means of twisting and transposition. The development of low-loss Nb–Ti strands with ultrafine filaments (0.1 µm) and highly resistive barriers (Cu–Ni) received a great impulse in the following years as part of the R&D on electrical machines and ac applications at 50/60 Hz [92]. This program was abandoned given the weak economic case, but some of the results were exploited in the construction of a low-loss Nb–Ti conductor used to wind a demonstration poloidal coil [93] for high-voltage

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

35

pulsed operation in the Tore Supra tokamak. More recently, given the success of the slowly ramped synchrotron quoted earlier, low-loss Nb–Ti strands and cable are receiving renewed interest for fast-cycled machines such as the SIS-100 and SIS-300 accelerators which are part of the FAIR complex, under construction at GSI [94]. The main argument for the use of superconducting magnets in this case is the installed electrical power and the cost of energy, which translates into the need to minimize the ac loss caused by the magnetic field cycle. To reduce the ac loss, it is necessary to: • Reduce the size of the filaments, which decreases the magnetization and the hysteresis loss; • Use small-diameter strands, with a tight twist pitch, which reduces the voltage driving coupling currents across the interfilamentary matrix; • Use resistive matrices or resistive barriers within the stabilizer matrix, to increase the resistance of the path of coupling currents.

Fig. 7. Scatter plot of hysteresis loss per unit superconductor volume for a bipolar ± 1.5 T cycle at 4.2 K versus Jc at 4.2 K and 5 T of the same wire, for a selected number of standard and low-loss wires produced in the past 15 years. The lines represent computed loss using a fully penetrated filament model, and filament diameters of 1, 2, 5, and 10 µm.

In the range of wire diameters of interest, 0.5– 1 mm, Nb–Ti filaments of 1–2 µm seem to be a practical lower limit. The fabrication of such assemblies becomes challenging and, with on the order of 100,000 filaments in the cross-section, single stacks are no longer feasible. A double stack of subelements can be effective, but the inhomogeneities among the materials can result in severe deformation of the fine filaments, causing loss of Jc , and a filament size that is larger than the ideal one. Furthermore, fine and closely spaced filaments tend to couple with each other, either through direct contact or because of electrical proximity. The resulting magnetization is larger than that expected from the single-filament geometry, as if it were produced by an effective filament of increased diameter. This effect, measured in Cu/Nb–Ti strands with filaments below 4 µm, can be mitigated by adding small percentages of magnetic materials such as Ni or Mn in the stabilizer matrix. Unfortunately, magnetic materials have an adverse effect on Jc , and in general fine filament wires fall short of the high Jc standard quoted earlier. The interplay of filament diameter and critical current density can be seen in Fig. 7, which shows a scatter plot of hysteresis loss for a ±1.5 T bipolar cycle, proportional to the magnetization, and critical current density at 4.2 K, 5 T for a selection of

standard (LHC, ITER) and R&D Nb–Ti strands produced in the past 15 years. The plot also contains lines of hysteresis loss at constant filament diameter, and shows that high Jc can be achieved with an effective filament diameter down to about 3 µm. Smaller values of the filament diameter are feasible, but this seems difficult to achieve without a significant loss of Jc . The use of resistive matrices is of special importance for reducing the coupling loss due to currents that tend to shield the interior of a strand by flowing in the superconducting filaments at the strand surface, and closing through the stabilizer. It is not possible to increase at will the electrical resistivity of the matrix, as low resistivity is required for stabilization and protection. An alternative is to introduce resistive barriers in the cross-section, thus decreasing detrimental transverse conductivity while maintaining beneficial longitudinal conductivity. This is done by introducing materials such as Cu–Ni as a spacer among filaments, subelements, and around the multifilamentary region. Using such resistive barriers, the time constant of the strand coupling currents can be reduced from the order of magnitude of tens of ms (e.g. 25 ms for an LHC inner layer strand) to below 1 ms, with a proportional reduction of ac loss.

December 24, 2012

14:49

36

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

3.2. Nb3 Sn 3.2.1. Discovery and beginnings The intermetallic compound Nb3 Sn [95] is the second LTS material that has found its way from material research to large-scale applications. Nb3 Sn is a brittle compound, and all modern manufacturing routes involve assembly of the precursor elements Nb, Sn and Cu additions (necessary for reducing the temperature at which Nb3 Sn is formed; see Ref. 96 and later discussion) into large-size billets that are extruded and/or drawn to the final diameter wire. The Nb3 Sn is then formed by a solid state diffusion reaction that is induced by a heat treatment at high temperature. Various manufacturing routes have been established industrially, resulting in wires with different critical current density and filament size. As reported earlier, the initial manufacture of technical Nb3 Sn was achieved by filling Nb tubes with crushed powders of Nb and Sn. The tube was sealed, compacted, and drawn to long wires. This primitive Powder-in-Tube (PIT) technique required reaction at high temperature, in the range of 1000–1400◦C, to form the superconducting phase. The high temperature, which causes excessive grain growth and low vortex pinning, is required as a result of the presence of two very-low-Tc Sn-rich line compounds (NbSn2 and Nb6 Sn5 ) that are stable in the binary compound below approximately 930◦ C [97]. An early alternative to the PIT was the fabrication in the form of tapes, by passing a Nb tape through a bath of molten Sn, and reacting the coated tape to form Nb3 Sn [98]. Although successful in demonstrating the use of Nb3 Sn in high-field magnets, neither technique was practical. The large filaments in the case of the PIT wire, and the inherently large aspect ratio of the tape, invariably result in large trapped magnetization and flux jump instabilities. In the late 1960s, Tachikawa introduced an alternative concept based on solid state diffusion. In his original work on V3 Ga (another superconducting intermetallic of the same family of materials with the A15 crystal structure), he used small filaments of V surrounded by a Cu matrix alloyed with Ga [99]. Solid state diffusion at high temperature mobilizes one of the two components (Ga), which reaches the filament (V) and reacts to form the superconducting phase. The same principle has been exploited to fabricate Nb3 Sn wires by the so-called bronze route,

which is today one of the leading techniques for manufacturing Nb3 Sn.

3.2.2. State of the art Industrial Nb3 Sn is presently produced by one of the following three manufacturing techniques: bronze route, internal tin, and PIT (see the architecture schematics shown in Fig. 8). A bronze route wire is made up of a large number of Nb or Nb-alloy filaments assembled in a Snrich bronze matrix. The composite is usually inserted in a can of high-purity copper stabilizer, with a thin barrier of material chemically inert to Cu, such as Nb, Ta, or Va. The assembly is then extruded and drawn to the final wire diameter. The superconducting phase is formed by submitting the wire to a heat treatment at a temperature in the range of 600– 700◦ C, i.e. significantly lower than the temperature required for the formation of Nb3 Sn from a binary mixture of Nb and Sn. This is enabled by the fact that the presence of Cu destabilizes the Sn-rich line compounds [96, 97]. The lower temperature prevents the excessive grain growth that is inevitable in binary systems, thereby increasing the pinning efficiency. At sufficiently high temperature, Sn diffuses in the Cu– Sn matrix and reacts with the Nb filaments to form the superconducting Nb3 Sn phase. The bronze route is a very-well-established process, which is used at present to produce most industrial Nb3 Sn. The main drawback of the bronze route comes from the limit of Sn solubility of 9.1 at.% (or 15.8 wt.%) in the ductile α-bronze. In fact, not all the Sn content in the bronze can be mobilized for reaction with the Nb filaments, and the Sn-depleted bronze matrix left after heat treatment takes a significant portion of the total non-Cu cross-section of the wire. Although beneficial for keeping the superconducting filaments physically decoupled, the matrix detracts from the real estate available in the wire for the superconducting phase. The limited Sn source further results in the formation of a relatively large fraction of off-stoichiometric niobium–tin [101] (which is stable from 18 to 25 at.% Sn [97]) with a reduced field–temperature phase boundary [102] that is not superconducting at higher magnetic fields. The relatively small fraction of Sn-rich Nb3 Sn, combined with the lost real estate that is occupied by the bronze, results in an upper limit to the

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

37

Fig. 8. Schematic overview of Nb3 Sn wire manufacturing processes, reproduced from Ref. 39 and based on Ref. 100. The schematic arrangement of materials and subelements is only for illustration purposes.

non-Cu critical current density Jc in the range of 1000 A/mm2 at 4.2 K and 12 T in optimized wires. An alternative to the bronze route wire consists in assembling a large number of Nb or Nb-alloy filaments and pure Sn or Sn-alloy rods in a Cu matrix. Such an assembly can be surrounded by a barrier that prevents diffusion of Sn, and further enclosed in a high-purity copper can. The stacked assembly is drawn to the final size of the wire. Restacking of assemblies is commonly done to decrease the final size of the subelements. Upon heat treatment, a cross diffusion process between Sn and the Cu matrix takes place, resulting in a Sn flow toward the multifilamentary region and subsequent reaction of Nb and Sn into Nb3 Sn. Several variants of this process have been devised, differing mainly in the way the various initial components are arranged. Because of the presence of the Sn source internal to the assembly, they are generally referred to as internal tin techniques. By comparison with the bronze route, an internal tin wire avoids the limitation on the amount of Sn inherent to its solubility in bronze, and increases the freedom in the layout. High non-Cu critical current is achieved by maximizing the amount of Nb in the matrix, keeping the Cu matrix fraction to a minimum, and introducing the quantity of Sn required for complete reaction to close to stoichiometry. This

apparently trivial task is in reality a tantalizing balance of cross-section optimization, diffusion and reaction kinetics, and practical manufacturing issues. The optimization of the internal tin technique, fostered by the US DOE Conductor Development Program (CDP), launched in 1999, produced the spectacular jump in Jc visible in Fig. 1. One of the most successful high-Jc internal tin processes developed within the scope of the CDP is the Restacked r ) of Oxford Superconducting Rod Process (RRP Technology, which regularly achieves non-Cu Jc values in excess of 3000 A/mm2 at 4.2 K and 12 T. To reach such high values of Jc , both the quantity (the amount of superconductor that is formed in the non-Cu fraction) and the quality (grain refinement, Sn content, and ternary element addition) of the Nb3 Sn must be optimized. This is possible by reducing the fraction of Cu in the matrix — also referred to as the local area ratio (LAR) — to a practical manufacturing minimum in the range of 0.1–0.3, introducing alloying additions such as Ta or Ti, and by an optimized heat treatment schedule. Furthermore, the barrier that separates the multifilamentary region from the high-purity Cu is made of Nb and partially reacted during heat treatment, thus adding to the final superconducting crosssection. After heat treatment, the tightly packed

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

38

Nb filaments and the reacted portion of the barrier grow into a completely connected cross-section of Nb3 Sn, fully coupled, whose characteristic dimension is hence approximately the size of the stacked subelement, which can be relatively large (50–100 µm). Finally, high-Jc Nb3 Sn wires can be manufactured using an evolution of the PIT process originally devised by Kunzler. The idea developed in the mid1970s at the Netherlands Energy Research Foundation (ECN) consisted in stacking tubes of Nb, filled with crushed powders of NbSn2 and a small percentage of Cu additive (necessary for the destabilization of the Sn-rich line compounds), in a high-purity Cu matrix. The stacked assembly is drawn or extruded to the final diameter and heat-treated, where attention should be paid to preventing the reaction front reaching the outer boundary of the Nb tube. Initial PIT wires produced by ECN, with 18–36 tubes in the Cu matrix, achieved a non-Cu Jc of about 500 A/mm2 at 4.2 K and 12 T. These values were consistently improved in the 1980s, until the 1990s production wire with 192 tubes reached a non-Cu Jc of 1700 A/mm2 at 4.2 K and 12 T. Further optimization of the layout, powders, the use of Ta as the alloying element in the Nb tube, and industrialization took place in the 1990s and 2000s, first at ShapeMetal Innovation (SMI) in Enschede, The Netherlands, then at European Advanced Superconductors in Hanau, Germany, presently Bruker-EAS (see also Ref. 103). Industrial PIT wires from Bruker regularly achieve non-Cu Jc in excess of 2500 A/mm2 at 4.2 K and 12 T. In practice, the OST RRP and Bruker PIT are at present the only two options of Nb3 Sn with sufficiently high Jc for HEP applications, and available in large quantities from industry. In Fig. 9 we show a cross-section of two standard layouts: a 0.7 mm RRP

Fig. 9.

108/127 stack and a 1 mm 192-tube PIT. Figure 10 shows typical critical current values of two such strands, heat-treated using the recommended schedule from the manufacturer. It is to be noted that the difference in Jc of the two processes, which is significant at 12 T, tends to decrease above 15 T and the two curves cross around 20 T. The current-carrying capability at medium (i.e. 12–15 T) versus high field (i.e. 20 T and higher) can be manipulated in various ways. The upper critical field (Hc2 ) (and to a lesser degree the critical temperature Tc ) can be increased by varying the amounts of alloying elements such as Ta or Ti [104], which will increase specifically the high-field performance. The upper critical field maximizes for 1.5 at.% Ti and for 4 at.% Ta addition. The difference is due to the fact that Ta replaces Nb and Ti replaces Sn in the Nb3 Sn lattice (suggested in Ref. 105 and confirmed by Ref. 106). Commercial Nb–7.5 wt.%–Ta alloy, for which the Ta content cannot be readily varied, is used in most RRP and PIT wires, and Hc2 cannot be varied by changing the Ta content. Recent RRP wires combine pure Nb rods with commercially available Nb–47 wt.%–Ti rods, and can vary the amount of Ti by varying the ratio between the Nb and Nb-alloy rods, and therefore manipulate Hc2 through compositional variations. A second way by which the Jc at medium fields can be balanced against high-field performance is through the reaction temperature. More aggressive reactions at higher temperature increase the Sn activity, and therefore generate a Sn-richer Nb3 Sn with a higher Hc2 [97, 102], and thus increased performance at higher field. This goes, however, at the cost of pinning efficiency, since the grain dimension of the reacted Nb3 Sn is a power function of the reaction temperature [103], which reduces the pinning

Layouts from leading manufacturers of Nb3 Sn strands for HEP applications.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

Fig. 10. Typical non-Cu Jc for the strands in Fig. 9, based on measurements at 4.2 K.

force [97], and therefore the performance at medium fields. Lower-temperature reaction results in Nb3 Sn that is less rich in Sn, with reduced Hc2 and highfield performance, but can retain smaller grains and thus increased pinning and performance at medium magnetic fields. The conductors can hence be optimized for medium- or high-field performance, and the difference between the RRP wire and the PIT wire in Fig. 10, as well as the cross-over at higher field, can be attributed to such optimizations. 3.2.3. Challenges The spectacular increase of Jc achieved over the past 10 years is a great success, but has also brought a number of riddles. In some cases, magnet performance was found to be below expectations, affected by instabilities that could be reproduced in single strands and cables both experimentally and theoretically [107, 108]. The basic explanation lies in the combination of the well-known effect of flux jumps at low field, and a newly defined self-field instability at high field [109]. The performance limits are shown schematically in Fig. 11, which reports measured critical and quench currents of a 0.8 mm wire fabricated by the OST RRP process, consisting of a stack of 54 subelements of 80 µm diameter in a Cu matrix with an RRR of 80. The upper line in the graph is the critical current Ic as obtained by measurements at high field, and extrapolated to low field. What is observed experimentally (symbols in Fig. 11) is that the wire reaches Ic at high field (above 10 T in Fig. 11). At low and medium field the wire has sudden resistive transitions

39

Fig. 11. Critical current and quench values as a function of the magnetic field, depicting high-field Jc values, and low- and medium-field quench values that are due to instabilities and measured using field sweeps at constant current.

well before reaching the critical current. This can be interpreted by an instability induced by the dissipation of the energy stored in the magnetization associated with the distribution of persistent current in the filaments (magnetization flux jump instability) or with the distribution of transport current among filaments (self-field instability). The magnetization flux jump instability dominates the behavior at very low field (up to 2 T in Fig. 11), and is usually evidenced by sweeping the field at constant current, a so-called V (H) measurement. The severity of the flux jump instability depends on the size of the superconducting filaments. The maximum current that the wire can reach in this regime usually dips at field ranges of the order of the penetration field, when the magnetization also reaches its maximum. At intermediate field (the range of 2–10 T in Fig. 11), the dominating effect is caused by the potential collapse of the self-field magnetization associated with the current distribution in the wire, which is usually concentrated in the external skin of superconducting filaments. The self-field magnetization is proportional to the size of the multifilamentary region in the wire. The collapse is triggered by a small perturbation that can have external origins, e.g. mechanical energy release. At moderate field (up to 8 T in Fig. 11), where Jc and the magnetic moment associated with the current distribution are large, the perturbation that is required to trigger the instability is small (the so-called energy regime in Fig. 11). At increasing field Jc decreases, the self-field magnetization also decreases, and the magnitude of the perturbation that triggers the instability increases.

December 24, 2012

14:49

40

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

This is visible in Fig. 11 as a transition region — the perturbation region from 8 T to 10 T. Eventually, the self-field instability is no longer triggered by the energy spectrum associated with the specific operating environment, and the wire reaches the critical current (above 10 T in Fig. 11). In practice, a very high Jc , in the range of 3000 A/mm2 , is accessible only in strands of modest diameter (typically 1 mm and smaller) if the filament diameter is small (typically below 50 µm) and the RRR is large (typically above 100). Achieving simultaneously a high Jc with small filaments and high RRR is challenging for any of the leading wire manufacturing routes. The reason is that to achieve a high Jc , the filament cross-section must be reacted almost completely, with the risk of a Sn leak in the stabilizer matrix and a catastrophic drop of RRR. This is particularly true in Rutherford cables, in which the diffusion barrier is significantly distorted at the edges of the cable. In practice, a fixed thickness of the Nb barrier is left unreacted (a few µm), which is essentially a lost percentage of the filament cross-section. A demand for high RRR hence limits the maximum achievable Jc . Reducing the filament diameter while maintaining the thickness of the unreacted barrier also reduces the real estate available for reaction, and causes a reduction of the final Jc . In summary, the critical current density Jc , effective filament diameter, and RRR have a simple but very delicate interplay, which requires a careful compromise in the strand design. The values of Jc reported earlier have been achieved with RRP and PIT wires of 50 µm subelement diameter or larger, and have resulted in RRR values in the range of 50–250, with a better average for the PIT wire at the cost of a reduced Jc . Such filaments are sufficiently small to avoid lowfield instabilities, but still result in large magnetization and partial flux jumps. It is for this reason that present R&D is mostly focused on a reduction of the magnetization through assemblies of higher subelement count. The absolute value of magnetization, and the effect of the reduction of the subelement dimension, can be appreciated through Fig. 12, where we compare measurements of RRP 0.8 mm wire with a 54/61 stack (70 µm subelement), and 0.7 mm wire with a 108/127 stack (40 µm subelement), to results obtained on PIT 0.7 mm wire

Fig. 12. Magnetization of Nb3 Sn wires of different architecture, external, and subelement diameter. (Data courtesy of B. Bordini and D. Richter, CERN, Geneva.)

with 114 tubes (45 µm subelement), and PIT 1 mm wire with 192 tubes (50 µm subelement). We observe the expected reduction of magnetization and instability in RRP wires from 70 µm to 40 µm. In addition, the fact that PIT wires have a lower Jc at low field when compared to RRP gives an additional benefit to the PIT. At comparable, or even slightly larger subelement, the low-field magnetization, and flux jumps are further suppressed. The compromise among the demands for high Jc , high RRR, and small subelement diameter can be combined in a target performance specification for large-scale HEP applications reported in Fig. 13. The targets given there are based on data available on stability current, magnetization, and estimates for flux jump onset, and refer to strands with a diameter of 1 mm and smaller. At this diameter, an RRR of

Fig. 13.

Performance target space for HEP Nb3 Sn.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

100 is considered sufficient to avoid significant degradation due to self-field instability. A subelement of 20 µm would lead to a magnetization width at 1 T (a sensible projected injection field) of approximately 150 mT, which is still a few times larger than that obtained in the accelerator superconducting magnets built with Nb–Ti, but within a reasonable correction range. Most important, with a subelement of this size we would expect no flux jumps at any operating temperature, including 1.9 K. We report in the same schematic three-axis representation present typical values for the RRP and PIT technologies, which show how the subelement diameter is indeed the most challenging among the three parameters. For this reason, present R&D is mostly devoted to reducing the subelement diameter, while still preserving Jc and RRR. Specifically, OST is developing an RRP assembly of 217 subelements which has already been tested in prototype lengths and yielded promising results [110]. Similarly, Bruker-EAS is working on PIT R&D material made with 192 tubes, drawn to small diameters. Measurements of short lengths of this PIT wire drawn to 0.6–0.7 mm have recently shown that the critical current density can be preserved at marginal RRR loss, and that the magnetization exhibits no flux jumps at 4.2 K [111]. The R&D mentioned above will lead to similar subelement diameters for both routes, in the range of 30–35 µm. A further reduction to 20 µm would require assemblies of 500 subelements, which is a large number for the precise and clean conditions required for successful area reduction. Further advances will hence depend on novel fabrication technologies, as for example fostered by the US Small Business Innovation Research (SBIR) program in the US. Finally, and most important, Nb3 Sn for HEP is at present still an expensive superconductor. With a price in the range of 10 EUR/kAm (evaluated at 12 T and 4.2 K), additional effort to reduce the price is mandatory in order to make a large-scale accelerator application a viable option. This depends, among other things, on the material yield and the capability to manufacture long piece length, in the range of several kilometers. Present production of HEPgrade material is limited — an estimate of 2 tons per year, a small quantity when compared to the ITER Nb3 Sn production, averaging 100 tons per year over

41

the past four years. This gives confidence that production capacity will not be an issue, and a scale effect could be expected once the demand increases for the first accelerator applications. 4. Advances in HTS Materials The maximum field that can be attained using Nb–Ti and Nb3 Sn is intrinsically limited by the field–temperature superconducting phase boundary. In Nb–Ti, µ0 Hc2 (0) amounts to 14.5 T, whereas for Nb3 Sn, µ0 Hc2 (0) has a maximum of 30 T in wires [102]. The above values must be reduced because of the need for a significant Jc and operating margin. In dipoles, Nb–Ti reaches a maximum of 80% of its Hc2 (1.9 K), whereas Nb3 Sn reaches 65% of its Hc2 (4.2 K) [112]. The lower percentage for Nb3 Sn is a result of the fact that the grains are approximately a factor of 10 too large to achieve optimal pinning at high magnetic fields [33]. In contrast, the upper critical field of HTSs at low temperature (e.g. 4.2 K) is 100 T or higher. As such, HTS materials, and specifically YBCO and Bi2212, do not have an intrinsic limitation in terms of the achievable magnetic field, but they are rather limited by the stress that would be produced in the winding. This is why YBCO and Bi-2212 are presently receiving much attention from the highfield community. In the following subsections, we will report on the state of the art of these two materials. 4.1. Bi-2212 The main attractiveness of Bi-2212, besides the highfield properties recalled above, is that it is available as a round wire, and can thus be formed into a Rutherford cable, as has been consistently demonstrated since the 1990s [113–116]. The record JE is around 500 A/mm2 at 20 T, 4.2 K in a 1-m-long round wire [117] (Fig. 2), and the present longlength performance is around 200–250 A/mm2 at 20 T, 4.2 K, and thus a factor of 3–4 too low for effective magnet windings (see Subsec. 2.1). Bi-2212 requires, as with Nb3 Sn, a so-called wind-and-react magnet fabrication process, as a result of the brittleness of the Bi-2212 in combination with the small bending radii that will be required in insert coils for hybrid Nb3 Sn/Bi-2212 magnets. Bi-2212 has to be reacted in a 1 atmosphere oxygen pressure environment around 890◦ C, with an accuracy of ± 1◦ C,

December 24, 2012

14:49

42

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

which places stringent requirements on the construction and insulation materials, as well as on the furnace control. Nonetheless, small solenoid demonstration coils have been fabricated and successfully tested at various institutes [118–120], demonstrating that the use of Bi-2212 for high-field magnets is indeed feasible. A recent breakthrough within the US Very High Field Superconducting Magnet Collaboration (VHFSMC) [37], which focused on determining the feasibility of Bi-2212 for very-high-field magnet applications, highlighted the main current-blocking mechanism in round wire Bi-2212. It was long thought that the grain boundaries, and the formation of undesired phases, were the main currentblocking mechanisms, but it was recently pointed out that large voids, dubbed “bubbles,” form inside the Bi-2212 fractions and block the current [121]. These bubbles form for two reasons. First, a residual void fraction of about 25% is required inside the Bi-2212 in wires, to allow wire drawing of the hard Bi-2212 particles inside the soft Ag matrix. This distributed void fraction agglomerates into bubbles during the partial melt reaction. Second, the bubble formation is amplified by the sudden release of oxygen from the powder during the melting of the Bi-2212, and the reaction of oxygen with contaminants, such as carbon and hydrogen, to form CO2 and H2 O. It is assumed that pure oxygen can quickly diffuse through the Ag matrix, but the CO2 and H2 O cannot, causing internal pressure in the wires that amplifies the bubbles. The internal pressure can even cause rupture of the soft Ag-alloy matrix that is close to its melting point, with leakage of the core constituents as a result. The Bi-2212 Strand and Cable Collaboration (BSCCo) [122] in the US, which is a continuation of the VHFSMC is, in close collaboration with industry, trying to mitigate the formation of bubbles through densification, and removal of the contaminants. Bi-2212 densification studies have increased the JE in short wire lengths to beyond the record levels, thereby approaching the 600–800 A/mm2 range that is needed for successful application in high-field magnets [118, 123]. Densification could, by removal of the local stress concentrations due to the presence of voids, also positively alter the behavior of Bi-2212 under mechanical loads (see Subsec. 2.4). Overall, the progress in Bi-2212 wire and magnet technology has been substantial

over the last decade, and Bi-2212 appears promising once the formation of bubbles can be mitigated in long lengths of conductor. 4.2. YBCO YBCO is available as a tape for which the Jc strongly depends on the direction of the applied field (Fig. 2), and it requires near-perfect texture to achieve a high Jc . The Jc anisotropy is a result of an anisotropy in Hc2 , in combination with a reduced pinning efficiency for magnetic fields that are perpendicular to the tape face, or H  c. The upper critical magnetic field for fields perpendicular to the tape face µ0 Hc2 (0)  c ≈ 120 T, whereas µ0 Hc2 (0) ⊥ c ≈ 250 T [124–127]. The introduction of, for example, self-assembled BaZrO3 into the YBCO layer [128] results in a nanoscale distribution of so-called BZO nanodots and nanorods that form pinning centers, specifically for field applied in the c direction, i.e. perpendicular to the tape width. This increases the Jc for this field direction, but has, so far, mainly been successful for higher temperatures [129], although the BZO does increase the overall pinning efficiency at 4.2 K [130]. The different efficiency at lower temperatures stems from the fact that the coherence length, and therefore the diameter of the flux lines, which is twice the coherence length, reduces at lower temperatures and results, in combination with reduced thermal activation of the flux lines, in different pinning interactions to dominate at 4.2 K compared to 77 K [129, 131]. As a consequence of the Jc anisotropy and the inevitable magnetic field components that are parallel to the c axis in magnets, it is, for now, more realistic to observe the current-carrying capacity in the “bad” field direction, i.e. with the field perpendicular to the broad side of the tapes. The current-carrying capacity of YBCO with the magnetic field applied in the c direction is on the order of 400 A/mm2 at 20 T and 4.2 K (Fig. 2). This is approaching the required current density levels, specifically since only 1% of the cross-section carries all the superconducting current. Increasing the JE by increasing the YBCO layer thickness is difficult, since misalignment becomes more significant when the layer thickness is increased, but recent efforts to increase the layer thickness to 2 µm have been successful, and the current densities in the required 600– 800 A/mm2 levels are within commercial reach [132].

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

A second route to gain JE is to reduce the substrate thickness, which now constitutes roughly 50% of the cross-section, but this means a substantial change in the delicate optimizations for the fabrication processes. Nevertheless, with presently only 1% of the cross-section being superconductor, it is clear that, as with Bi-2212, the potential of YBCO is significant. A major obstacle to the application of YBCO in HEP magnets is that it is difficult to form a transposed cable that enables currents in the tens-of-kA region. Roebel-type cables are considered [133–135], but are still in their relative infancy. Recent tests at CERN on two Roebel cables made by Karlsruhe Institute of Technology (KIT) and General Cable Superconductors (GCS), assembled from 10 and 15 12-mm-wide tapes respectively, showed critical current at 4.2 K in excess of 10 kA in a parallel 10 T background field, and approximately 4 kA in a 10 T perpendicular field [136]. The critical current of both cables is in agreement with the value expected from single tapes, once the contributions of background and self-field are properly taken into account. With the quoted performance, these Roebel cables approach the desired target for a high-field insert, and are hence promising alternatives to Bi2212 Rutherford cables. Many issues are still open, such as insulation, winding quality, and the mechanics of the relatively loose assembly of tapes. Recently proposed alternatives [137] provide a higher mechanical stability, but so far do not retain sufficient overall current density. 4.3. Further challenges Beyond the quest for higher Jc , and the difficulty of making high-current and compact cables discussed above, Bi-2212 and YBCO share a number of common challenges. The first is the management of mechanical and thermal stresses, with acceptable and reversible critical current degradation (see also our earlier discussion). This will require fundamental work on understanding and improving the strain sensitivity of Bi-2212, and most likely the choice of a magnet design with features that limit the strain and stress in the high-field, HTS-based portion. Material compatibility with structural alloys and insulation fibers is another matter of concern, especially for Bi-2212, which requires a high-temperature heat reaction in an oxygen atmosphere. YBCO, although not requiring a heat treatment, has been

43

found to suffer irreversible degradation and delamination, most likely caused by differential thermal contraction with respect to impregnation resins. Quench detection and protection is the next concern for HTS materials. At high temperatures, such as 77 K, the temperature margin is in principle sufficiently low to cause a fast normal zone propagation (NZP). On the other hand, the operating current at these high temperatures is low, and the NZP is severely hindered by the lack of driving force (I 2 R) and by the large heat capacity of the materials. At low temperatures, by contrast, the heat capacity is low, and with a large Jc the operating current, and Joule heating, can be high enough to increase the NZP. On the other hand, the large temperature margin tends to slow the NZP. Another aspect that should be considered is that for quench propagation the relevant quantity is not Hc2 (T ), but rather the irreversibility field Hirr (T ), at which the flux lines become depinned. For round wire Bi-2212 the irreversibility line is low, and crosses 20 K at 10 T [127], therefore reducing the temperature margin for high-magnetic-field applications at 4.2 K to levels that are comparable to those of Nb3 Sn. This suggests that with an appropriate choice of the superconductor operating point and sufficient current density to drive the NZP, the NZP velocities might become high enough to allow the detection of quenches in magnet systems. In this respect, the fact that HTS will mainly be used as the high-field insert of a background magnet will be beneficial. Note, however, that for YBCO, a similar appreciable lowering of the temperature margin with an increasing magnetic field is not observed, as a result of a much higher Hirr (T ) [127]. Recent experiments on small-scale coils built with Bi-2212 and operated at 40–80% of Ic up to 20 T have shown an NZP as small as a few cm/s [138], conflicting with the promise of a low irreversibility line. For now, at least in small Bi-2212 coils, voltage– current transitions can be readily measured [116], rendering the requirement of a high NZP for quench detection less urgent, but this might change once conductors become more homogeneous and the coils larger. The above discussion leads to the conclusion that it is not clear at forehand whether quench detection and protection will be a major issue for high-field HTS magnets, but the limited data available on

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

44

high-field NZP seems to suggest that the situation is unfavorable. A wider experimental database and improved understanding is definitely a high priority. YBCO and Bi-2212 have in common the fact that they do not have well-separated, transposed filaments. This results in large magnetic moments that are detrimental to the field quality of an accelerator magnet. Secondary to the issues above, but mandatory for an application to the high accuracy of an accelerator magnet, the matter of filament diameter and control of coupling will have to be addressed in HTS wires, tapes, and cables. Finally, with a cost in the range of 200– 400 EUR/kAm (evaluated at 20 T and 4.2 K), there is clearly a large optimization work required on the production chain before HTS materials can be used on a large scale. 5. Superconducting Cables Wires and tapes manufactured with the LTS and HTS materials listed above carry currents in the range of a few hundred amperes, and are appropriate for winding small magnets, where the magnet inductance and stored energy are not an issue. On the other hand, the large-scale dipole and quadrupole magnets of an accelerator are connected in kilometerlong strings, and the stored energy can reach hundreds of MJ, up to the 1 GJ of an LHC dipole sector powered at nominal current. To decrease their inductance and limit the operating voltage, it is mandatory to use cables made up of several wires in parallel that are able to carry much larger currents, typically in the range of 10 kA. Such cables must ensure good current distribution through transposition, combined with precisely controlled dimensions necessary for obtaining coils of accurate geometry, as well as good winding characteristics. These properties are the characteristic of

Fig. 14. A Rutherford cable for the inner layer of the LHC dipoles, showing the Nb–Ti filaments in a few etched strands.

the flat cable invented at the Rutherford Laboratory in England [139]. A typical Rutherford cable, shown in Fig. 14, is composed of fully transposed twisted wires (Nb– Ti in the figure). The rectangular geometry of the cable provides high strand packing and is flexible enough to wind magnet coils of various geometries. The transposition length, also referred to as the twist pitch, is usually kept short, of the order of a few centimeters. To improve the winding properties, the cable is slightly keystoned, i.e. the cable thickness is not constant from side to side. The angle formed by the planes of the cable upper and lower faces is called the keystone angle, which is usually in the range of 1◦ –2◦ . A summary of cable characteristics for the major superconducting accelerator projects is given in Table 1. The cabling process is invariably associated with large deformations at the edges of the cable, where the wires are plastically deformed. This is necessary for achieving mechanical stability of the cable, but can lead to degradation of the critical current

Table 1. Main characteristics of the (bare) superconducting cables used to wind the dipoles for the four superconducting colliders. Name

Tevatron HERA RHIC LHC inner LHC outer

Strand diameter (mm)

Thickness (mm)

Width (mm)

Twist pitch (mm)

Keystone angle (degree)

0.68 0.90 0.65 1.065 0.825

1.26 1.48 1.17 1.90 1.48

7.8 10.0 9.7 15.1 15.1

66 95 94 115 100

2.1 2.2 1.2 1.2 0.9

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

of the superconductor, as well as significant distortion of the diffusion barriers, or their breakage. The degradation can be divided into two origins: degradation associated with the intrinsic properties of the strand, and degradation due to the choice of cabling parameters. The intrinsic properties of wires vary greatly, depending on the superconductor, as well as on the specific architecture. As an example, Nb–Ti strands are known to have good tolerance to deformation and cabling, while Nb3 Sn strands are less forgiving. Among strands based on the same superconducting material, architecture details such as filament size, position in the strand, and spacing are of the utmost importance. The intrinsic tolerance to cabling can be verified by tests such as the sharp bend, in which the wire is bent back sharply on itself in a fixture under controlled conditions, etched, and examined for broken filaments. Cabling parameters that affect the amount of critical current degradation are mainly the amount of overall compaction, as well as other finer details, such as the cabling tension, the angle and shape of the cabling mandrel, the stability of the tooling, and the use of lubricants. Cable compaction depends on the thickness and width of the cabling cavity formed by the rollers placed at the cabling point of a typical machine. Following Ref. 140, we can define a thickness and width compaction (respectively ct and cw ) as follows: ct =

t − 1, 2d

cw =


Nd 2 cos(θ)

45

it has been found empirically that the cable compaction should be limited to the range of 86% or larger. This should be considered as both an average and a local limit, to be respected specifically at the thin edge of a keystoned cable. Such a limit imposes very tight constraints on the range of feasible keystone angles, especially when producing wide cables for small-aperture magnets. In addition, and most important (as stressed in Ref. 140), compaction by itself does not guarantee a good final result. Indeed, the same compaction can be achieved by reducing either the width or the thickness, but the final result is very different in terms of cable degradation. It was found, again empirically, that for Nb3 Sn the preferred parameter range for cabling is ct ≈ −5% to −10% and cw = −3% to 0%. Note how in this case the final width of the cable can exceed the theoretical cable dimensions, which is done intentionally to avoid excessive deformation at the thin edge, where the strands that transit from the upper to the lower face of a Rutherford cable are subjected to large deformations. This is in general satisfactory from the point of view of critical current degradation, which is typically limited to 5–10% of a virgin strand, but the resulting cable may not be sufficiently compacted to achieve the mechanical stability that is necessary for winding. Despite the many years of experience, cabling is still a delicate balance between limited wire deformation and desired cable compaction. It depends on the specific material that is cabled as well as the

(4) w 

− 1,

(5)

+ 0.72d

where t is the cable thickness, w the cable width, d the strand diameter, N the number of strands in the cable, and θ the twist pitch angle. In addition, it is useful to define the overall cable compaction as follows: c = (ct + 1)(cw + 1).

(6)

An acceptable compaction range for cabling of Nb– Ti strands is ct ≈ −10% to −15%, cw ≈ −5% to −10%, and c ≈ 80–85%. Typical cabling degradation achieved in Nb–Ti cables manufactured in the above range is less than 5%. Such deformation, acceptable for Nb–Ti, is excessive for Nb3 Sn. For Nb3 Sn,

Fig. 15. Compilation of computed width and thickness compaction values for a number of cables manufactured in the past 10 years for Nb3 Sn applications, compared to the compaction of the LHC Nb–Ti cables. (Data for D20, RD3, HQ, LQ courtesy of D. Dietderich, LBNL; data for SMC, FRESCA2, DS 11 T courtesy of L. Oberli, CERN.)

December 24, 2012

14:49

46

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

Fig. 16. Detail of the edge of prototype Rutherford cables built with PIT Nb3 Sn, 1-mm-diameter wires, with slightly different width (as indicated), identical thickness (1.82 mm), and twist pitch (120 mm). The area placed in the rectangles of the lowmagnification micrographs is shown in the high-magnification images below. (Micrographs courtesy of A. Bonasia and L. Oberli, CERN.)

desired cable geometry, and the simple rules given above provide only a starting point for empirical optimizations. A collection of width and thickness compaction is reported in Fig. 15, showing how the realized Nb3 Sn cables tend to cluster in an area of reduced width deformation when compared to the LHC Nb–Ti cables. To illustrate the difficulty of such optimization, we show in Fig. 16 an example of cabling trials that have taken place during the development of a large-size Rutherford cable for the EuCARD magnet FRESCA-2. The cable, built with 40 PIT strands of 1 mm diameter, has a nominal dimension of 20.9 mm, a width of 1.82 mm, and a twist pitch of 120 mm. Micrographs of this cable, before heat treatment, are shown on the right hand side of the figure. They demonstrate that the local deformation at the most compacted location, the cable edge, is controlled to a tolerable level. Specifically, we observe no merging of subelements, and an acceptable reduction in the Nb thickness which guarantees that the Sn leakage and associated Cu poisoning during the heat treatment is small. On the other hand, as shown on the left hand side of the figure, a small reduction with respect to these optimized dimensions (a 2.5% reduction in the width) causes considerable deformation in the filaments and merging, which leads to degradation of Jc , a very low local RRR, and a large effective filament dimension.

The concept of Rutherford cables can be easily applied to any material that comes in the form of round wires, and it has been extended to round Bi2212 HTS wires [113–116]. Further cable concepts, such as the Roebel bar, or tape assemblies around a stabilizer core or tube, discussed earlier, are in an early stage of development and have not yet found an accelerator application.

6. Summary High-performance superconducting materials and conductors are the basic ingredient of the magnets for the large-scale accelerators that have been pushing the frontier of particle physics in the past 30 years. As we have discussed in this article, high critical current density is the principal, but not exclusive, target of such optimized conductors. While the presence is still largely dominated by the widespread use of Nb–Ti, literally the workhorse for all HEP applications to date, we predict that the next five years will be decisive for Nb3 Sn. A decennium of preparation has resulted in Nb3 Sn wires that are approaching the maturity necessary for HEP application, and the baton is now in the hands of the magnet builders whose task is to engineer solutions for the use of this upgraded material. At the same time, we hear HTS materials knock at the door. While priority is naturally, and rightly, given to Nb3 Sn, we believe that it is

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

important to continue the technological exploration of YBCO and Bi-2212. Fundamental questions need to be answered on the basic properties (critical current, mechanics), and the application in magnets (cables, protection) of these materials. Given the long time for such research, we stress the importance of national and international programs that are presently pursuing the construction of simple but very meaningful small-scale demonstrator magnets. In summary, the landscape of superconductors for HEP, and more generally accelerator applications, is varied and most interesting, with a number of opportunities and critical decision points approaching in the coming years. Superconductivity remains a fascinating field, from the enchantment of quantum physics on the microscopic scale, to the engineering challenge of the largest instrument ever built on the macroscopic scale. Acknowledgments The authors are grateful for the material and expertise provided by D. C. Larbalestier of the Applied Superconductivity Center at the NHMFL, Tallahassee, FL, USA; A. Ballarino, B. Bordini, L. Oberli, and D. Richter of CERN, Geneva, Switzerland; D. R. Dietderich of LBNL, Berkeley, CA, USA; and J. Schwartz of NCSU, Raleigh, NC, USA. This work was partly supported by the Director, Office of Science, High Energy Physics, US Department of Energy, under contract No. DE-AC02-05CH11231.

[7]

[8] [9] [10]

[11] [12] [13] [14] [15]

[16]

[17]

[18]

[19]

[20]

References [1] H. Kolm, B. Lax, F. Bitter and R. Mills (eds.), High Magnetic Fields: Proc. 1st Int. Conf. on High Magnetic Fields (Cambridge, MA, USA; Nov. 1–4, 1961), p. 1662. [2] J. E. Kunzler, IEEE Trans. Magn. 23, 396 (1987). [3] C. Laverick and G. M. Lobell, Rev. Sci. Instrum. 36, 825 (1965). [4] J. R. Purcell, The 1.8 tesla, 4.8 m I.D. bubble chamber magnet, in Proc. 1968 Summer Study on Superconducting Devices and Accelerators (Brookhaven National Laboratory, June 10– July 19, 1968), pp. 765–785. [5] K. Jaeger and J. R. Purcell, Operation and accuracy of the 12-foot bubble chamber magnet, in Proc. 4th Int. Conf. on Magnet Technology (MT4) (Brookhaven National Laboratory, 1972), pp. 328–338. [6] D. P. Brown, R. W. Burgess and G. T. Mulholland, The superconducting magnet for the Brookhaven

[21]

[22] [23] [24] [25] [26] [27] [28] [29] [30]

47

National Laboratory 7-foot bubble chamber, in Proc. 1968 Summer Study on Superconducting Devices and Accelerators (Brookhaven National Laboratory, 1969), pp. 794–814. E. U. Haebel and F. Wittgenstein, Big European Bubble Chamber (BEBC) magnet progress report, in Proc. 3rd Int. Conf. on Magnet Technology (DESY, Hamburg, 1970), pp. 874–895. G. B. Yntema, Phys. Rev. 98, 1197 (1955). G. B. Yntema, IEEE Trans. Magn. 23, 390 (1987). J. E. Kunzler, E. Buehler, F. S. L. Hsu, B. T. Matthias and C. Wahl, J. Appl. Phys. 32, 325 (1961). J. E. Kunzler, E. Buehler, F. S. L. Hsu and J. H. Wernick, Phys. Rev. Lett. 6, 89 (1961). T. G. Berlincourt, Cryogenics 27, 283 (1987). Rutherford Laboratory Superconducting Applications Group, J. Phys. D 3, 1517 (1970). G. H. Morgan, J. Appl. Phys. 41, 3673 (1970). R. R. Wilson et al., The energy doubler design study: A progress report. Technical report (Fermilab, Batavia, IL, USA, 1974). A. Tollestrup, The Tevatron hadron collider: A short history, in Int. Conf. on the History of Original Ideas and Basic Discoveries in Particle Physics (Ettore Majorana Centre for Scientific Culture, Erice, Sicily, Italy; Jul. 29–Aug. 4, 1994). F. E. Mills, Isabelle design study, in Proc. 1973 Particle Accelerator Conference (1973), pp. 1036– 1038. Central Design Group, SSC conceptual design report, Technical report (Central Design Group, 1986). D. E.-S. (Center), HERA: A Proposal for a Large Electron–Proton Colliding Beam Facility at DESY (Deutsches Elektronen-Synchrotron, 1981). G. Gurov, UNK status and plans, in Proc. 1995 Particle Accelerator Conference (Dallas, TX, USA, 1995), pp. 416–419. M. Anerella et al., Nucl. Instrum. Methods Phys. Res. Sect. A: Accelerators, Spectrometers, Detectors and Associated Equipment 499, 280 (2003). L. R. Evans and P. Bryant, J. Instrum. 3, S08001, 164 (2008). L. Rossi and L. Bottura, Reviews of Accelerator Science and Technology. H. Desportes, IEEE Trans. Magn. 17, 1560 (1981). M. A. Green et al., IEEE Trans. Magn. 15, 128 (1979). J. Benichou et al., IEEE Trans. Magn. 17, 1567 (1981). R. Wands et al., IEEE Trans. Magn. 19, 1368 (1983). J. M. Baze et al., IEEE Trans. Magn. 24, 1260 (1988). R. Apsey et al., IEEE Trans. Magn. 21, 490 (1985). H. H. J. ten Kate, Physica C: Supercond. 468, 2137 (2008).

December 24, 2012

14:49

48

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

[31] A. Herve, IEEE Trans. Appl. Supercond. 10, 389 (2000). [32] R. M. Scanlan, IEEE Trans. Appl. Supercond. 11, 2150 (2001). [33] D. R. Dietderich and A. Godeke, Cryogenics 48, 331 (2008). [34] S. A. Gourlay, IEEE Trans. Appl. Supercond. 16, 324 (2006). [35] G. de Rijk et al., IEEE Trans. Appl. Supercond. 22, 4301204 (2012). [36] P. J. Lee, private communication, http://magnet. fsu.edu/∼lee/plot/plot.htm (2012). [37] Very High Field Superconducting Magnet Collaboration: A collaboration of groups at Brookhaven National Laboratory, Fermilab, Lawrence Berkeley National Laboratory, Los Alamos National Laboratory, the National High Magnetic Field Laboratory, North Caroline State University, and Texas A&M University seeking to understand and apply round wire Bi-2212. [38] L. Bottura, L. de Rijk, G. Rossi and E. Todesco, IEEE Trans. Appl. Supercond. 22, 4002008 (2012). [39] A. Godeke, Performance boundaries in Nb3 Sn Superconductors. Ph.D. thesis (University of Twente, Enschede, The Netherlands, 2005). [40] A. Ballarino, IEEE Trans. Appl. Supercond. 21, 980 (2011). [41] T. Ogitsu et al., IEEE Trans. Appl. Supercond. 19, 1081 (2009). [42] L. Rossi and E. Todesco, Phys. Rev. ST Accel. Beams 10, 112401 (2007). [43] P. J. Lee and D. C. Larbalestier, IEEE Trans. Appl. Supercond. 15, 3474 (2005). [44] D. Leroy, L. Oberli, D. Perini, A. Siemko and G. Spigo, Design futures and performance of a 10 T twin aperture model dipole for LHC, in Proc. 15th Int. Conf. on Magnet Technology (Beijing, China, 1998), eds. L. Lianghzhen, S. Guoliao and Y. Lugang (Science Press, Beijing, China, 1998), Vol. 1, pp. 119–122. [45] A. F. Lietzke et al., IEEE Trans. Appl. Supercond. 14, 345 (2004). [46] M. N. Wilson, Superconducting Magnets (Oxord Science Publications, 1983). [47] P. Ferracin et al., IEEE Trans. Appl. Supercond. 15, 1132 (2005). [48] P. Ferracin et al., IEEE Trans. Appl. Supercond. 16, 378 (2006). [49] A. Godeke, B. ten Haken, H. H. J. ten Kate and D. C. Larbalestier, Supercond. Sci. Tech. 19, R100 (2006). [50] D. Arbelaez, A. Godeke and S. O. Prestemon, Supercond. Sci. Tech. 22, 025005 (2009). [51] B. ten Haken, A. Godeke, H. J. Schuver and H. H. J. ten Kate, IEEE Trans. Magn. 32, 2720 (1996). [52] N. Cheggour et al., Supercond. Sci. Tech. 25, 015001 (2012).

[53] X. F. Lu et al., IEEE Trans. Appl. Supercond. 22, 8400307 (2012). [54] N. Cheggour et al., Appl. Phys. Lett. 83, 4223 (2003). [55] D. C. van der Laan et al., Supercond. Sci. Tech. 23, 014004 (2010). [56] D. C. van der Laan et al., Supercond. Sci. Tech. 23, 072001 (2010). [57] J. S. Higgins and D. P. Hampshire, IEEE Trans. Appl. Supercond. 21, 3234 (2011). [58] M. Sugano, K. Shikimachi, N. Hirano and S. Nagaya, Supercond. Sci. Tech. 23, 085013 (2010). [59] H. Boschman, A. P. Verweij, S. Wessel, H. H. J. ten Kate and L. J. M. van de Klundert, IEEE Trans. Magn. 27, 1831 (1991). [60] H. H. J. ten Kate, H. W. Weijers and J. M. van Oort, IEEE Trans. Appl. Supercond. 3, 1334 (1993). [61] J. M. van Oort, R. M. Scanlan, H. W. Weijers, S. Wessel and H. H. J. ten Kate, IEEE Trans. Appl. Supercond. 3, 559 (1993). [62] J. M. van Oort, R. M. Scanlan, H. W. Weijers and H. H. J. ten Kate, Adv. Cryog. Eng. (Materials) 40, 867 (1994). [63] D. R. Dietderich, R. M. Scanlan, R. P. Walsh and J. R. Miller, IEEE Trans. Appl. Supercond. 9, 122 (1999). [64] D. R. Dietderich et al., IEEE Trans. Appl. Supercond. 11, 3577 (2001). [65] P. Bauer et al., IEEE Trans. Appl. Supercond. 11, 2457 (2001). [66] J. W. Ekin, IEEE Trans. Magn. MAG-23, 1634 (1987). [67] B. ten Haken and H. H. J. ten Kate, Physica C 270, 21 (1996). [68] B. ten Haken, A. Beuink and H. H. J. ten Kate, IEEE Trans. Appl. Supercond. 7, 2034 (1997). [69] D. Dell’Orco et al., IEEE Trans. Appl. Supercond. 3, 637 (1993). [70] A. den Ouden, W. A. J. Wessel and H. H. J. ten Kate, IEEE Trans. Appl. Supercond. 7, 733 (1997). [71] A. D. McInturff et al., Test results for a high field (13 T) Nb3 Sn dipole, in Proc. 1997 Part. Accel. Conf. (Vancouver, Canada, 1998), Vol. 3, pp. 3212– 3214. [72] E. Barzi, T. Wokas and A. V. Zlobin, IEEE Trans. Appl. Supercond. 15, 1541 (2005). [73] B. Seeber et al., IEEE Trans. Appl. Supercond. 17, 2643 (2007). [74] B. Seeber, A. Ferreira, V. Ab¨ acherli and R. Fl¨ ukiger, Supercond. Sci. Tech. 20, S184 (2007). [75] E. Barzi, D. Turrioni and A. V. Zlobin, IEEE Trans. Appl. Supercond. 18, 980 (2008). [76] B. Seeber et al., IEEE Trans. Appl. Supercond. 18, 976 (2008). [77] P. Ferracin et al., IEEE Trans. Appl. Supercond. 19, 1240 (2009).

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230002

Superconducting Materials and Conductors: Fabrication and Limiting Parameters

[78] G. Mondonico, B. Seeber, A. Ferreira, B. Bordini, L. Oberli, L. Bottura, A. Ballarino, R. Flkiger and C. Senatore, Supercond. Sci. Tech. 25, 115002 (2012). [79] H. Felice et al., IEEE Trans. Appl. Supercond. 21, 1849 (2011). [80] J. K. Hulm and R. D. Blaugher, Phys. Rev. 123, 1569 (1961). [81] H. E. Cline, R. M. Rose and J. Wulff, Research on a superconducting Nb–Th eutectic alloy, Technical report: NASA Report CR 54103 (1964). [82] H. E. Cline, B. P. Strauss, R. M. Rose and J. Wulff, J. Appl. Phys. 37, 5 (1966). [83] F. P. Levi, J. Appl. Phys. 31, 1469 (1960). [84] D. C. Larbalestier and P. J. Lee, New developments in niobium titanium superconductors, in Proc. 1995 Particle Accelerator Conference and International Conference on High Energy Accelerators (1996), pp. 1276–1281. [85] J. D. Adam et al., IEEE Trans. Appl. Supercond. 12, 1056 (2002). [86] J. D. McCambridge et al., IEEE Trans. Appl. Supercond. 5, 1697 (1995). [87] E. Kadyrov, A. Gurevich and D. C. Larbalestier, Appl. Phys. Lett. 68, 1567 (1996). [88] M. N. Wilson, Cryogenics 48, 381 (2008). [89] M. N. Wilson et al., “AC3” — a prototype superconducting synchrotron magnet, in Proc. 1972 Appl. Supercond. Conf. (May 1–3, 1972; Annapolis, MD, USA), eds. H. M. Long and W. F. Gauster, pp. 277–282. [90] H. Brechna and M. A. Green, Pulsed superconducting magnets, in Proc. 1972 Appl. Supercond. Conf. (May 1–3, 1972; Annapolis, MD, USA), eds. H. M. Long and W. F. Gauster, pp. 226–238. [91] G. Bronca, G. Neyret, J. Parain and J. Perot, Consequences of replacing conventional magnets by superconducting magnets in an existing synchrotron, in Proc. 1972 Appl. Supercond. Conf. (May 1–3, 1972; Annapolis, MD, USA), eds. H. M. Long and W. F. Gauster, pp. 283–287. [92] T. Verhaege, Y. Laumond and A. Lacaze, Handbook of Applied Superconductivity (1998), p. 415. [93] H. Krauth, IEEE Trans. Magn. 24, 1023 (1988). [94] G. Moritz, in Proc. 2007 Particle Accelerator Conference (Albuquerque, New Mexico, USA, 2007), p. 3745. [95] B. T. Matthias, T. H. Geballe, S. Geller and E. Corenzwit, Phys. Rev. 95, 1435 (1954). [96] K. Tachikawa and P. J. Lee, History of Nb3 Sn and Related A15 Wires, in 100 Years of Superconductivity, eds. H. Rogalla and P. H. Kes (CRC Press, 2012), Chap. 11.3. [97] A. Godeke, Supercond. Sci. Tech. 19, R68 (2006). [98] M. G. Benz, IEEE Trans. Magn. 2, 760 (1966). [99] K. Tachikawa and Y. Tanaka, Japan. J. Appl. Phys. 5, 834 (1966).

49

[100] M. T. Naus, Optimization of internal-Sn Nb3 Sn composites. Ph.D. thesis (Univ. of Wisconsin– Madison, 2002). [101] V. Ab¨ acherli et al., IEEE Trans. Appl. Supercond. 15, 3482 (2005). [102] A. Godeke et al., J. Appl. Phys. 97, 093909 (2005). [103] A. Godeke, A. den Ouden, A. Nijhuis and H. H. J. ten Kate, Cryogenics 48, 308 (2008). [104] M. Suenaga, D. O. Welch, R. L. Sabatini, O. F. Kammerer and S. Okuda, J. Appl. Phys. 59, 840 (1986). [105] R. Fl¨ ukiger et al., Supercond. Sci. Tech. 21, 054015 (2008). [106] M. G. T. Mentink et al., Physics Procedia 36, 491 (2012). [107] V. Kashikin and A. Zlobin, IEEE Trans. Appl. Supercond. 15, 1621 (2005). [108] B. Bordini, E. Barzi, S. Feher, L. Rossi and A. Zlobin, IEEE Trans. Appl. Supercond. 18, 1309 (2008). [109] B. Bordini and L. Rossi, IEEE Trans. Appl. Supercond. 19, 2470 (2009). [110] J. Parrell et al., IEEE Trans. Appl. Supercond. 19, 2573 (2008). [111] M. Thoener, V. Abaecherli, A. Szulczyk, B. Lindenhovius, J. Sailer and K. Schlenga, presented at ICEC 24–ICMC 2012 (May 14–18, 2012; Fukuoka, Japan). [112] A. Godeke et al., IEEE Trans. Appl. Supercond. 17, 1149 (2007). [113] R. M. Scanlan et al., IEEE Trans. Appl. Supercond. 9, 130 (1999). [114] T. Hasegawa et al., IEEE Trans. Appl. Supercond. 11, 3034 (2001). [115] T. Hasegawa et al., IEEE Trans. Appl. Supercond. 12, 1136 (2002). [116] A. Godeke et al., Supercond. Sci. Tech. 23, 034022 (2010). [117] H. Miao et al., IEEE Trans. Appl. Supercond. 17, 2262 (2007). [118] H. Miao, Y. Huang, M. Meinesz, S. Hong and J. Parrell, Adv. Cryog. Eng. (Materials) 58, 315 (2011). [119] H. Weijers et al., IEEE Trans. Appl. Supercond. 20, 576 (2010). [120] A. Godeke et al., Physics Procedia 36, 812 (2012). [121] F. Kametani et al., Supercond. Sci. Tech. 24, 075009 (2011). [122] Bi-2212 Strand and Cable Collaboration: A collaboration of groups at Brookhaven National Laboratory, Fermilab, Lawrence Berkeley National Laboratory, and the National High Magnetic Field Laboratory, seeking to understand and apply round wire Bi-2212. [123] J. Jiang et al., Supercond. Sci. Tech. 24, 082001 (2011). [124] T. Sekitani et al., Physica C 392–396, 116 (2003).

December 24, 2012

14:49

50

WSPC/253-RAST : SPI-J100

1230002

L. Bottura & A. Godeke

[125] T. Sekitani, N. Miura, S. Ikeda, Y. H. Matsuda and Y. Shiohara, Physica B 346–347, 319 (2004). [126] T. Sekitani, Y. H. Matsuda and N. Miura, New J. Phys. 9, 47 (2007). [127] D. C. Larbalestier, The long road to high current density superconducting conductors, in 100 Years of Superconductivity, eds. H. Rogalla and P. H. Kes (CRC Press, 2012), Chap. 11.1. [128] J. L. MacManus-Driscoll et al., Nat. Mater. 3, 439 (2004). [129] A. Xu et al., Supercond. Sci. Tech. 23, 014003 (2010). [130] V. Braccini et al., Supercond. Sci. Tech. 24, 035001 (2011). [131] A. Gurevich, Supercond. Sci. Tech. 20, S128 (2007). [132] Bruker-EST, YBCO Coated Conductor Data Sheet (2012). [133] W. Goldacker et al., Supercond. Sci. Tech. 22, 034003 (2009). [134] S. Schlachter, W. Goldacker, F. Grilli, R. Heller and A. Kudymow, IEEE Trans. Appl. Supercond. 21, 3021 (2011).

[135] N. Long, R. Badcock, K. Hamilton, A. Wright, L. Jiang and Z. Lakshmi, J. Phys. Conf. Ser. 234, 022021 (2010). [136] J. Fleiter, A. Ballarino, L. Bottura and P. Tixador, Supercond. Sci. Tech., submitted for publication (2012). [137] D. C. van der Laan, X. F. Lu and L. F. Goodrich, Supercond. Sci. Tech. 24, 042001 (2011). [138] L. Ye, F. Hunte and J. Schwartz, Stability and quench behavior of Bi2 Sr2 CaCu2 Ox coils at high magnetic fields. Private communication (2012). [139] G. Gallagher-Daggit, Superconductor cables for pulsed dipole magnets. Technical report (Rutherford Laboratory Memorandum No. RHEL/M/A25, 1973). [140] D. R. Dietderich, N. L. Liggins and H. C. Higley, Development of high current Nb3 Sn Rutherford cables for NED and LARP. Private communication (2008).

Arno Godeke was educated in Mechanical Engineering and Applied Physics in The Netherlands. He has broad multidisciplinary experience in mechanical, electrical, and cryogenic design, fabrication, and implementation; and, in materials science, large-scale applications and the physics of superconductors. He has worked in superconductivity in various positions at the University of Twente, The Netherlands, from 1992 through 2005. In 1998, he visited the NHMFL, Tallahassee, FL, USA, on a sabbatical, and worked at the University of Wisconsin, Madison, WI, USA, from 2002 to 2003 to obtain his PhD in Applied Physics from Twente in 2005. Since 2006, Dr. Godeke is at LBNL, Berkeley, USA, where he is responsible for the conductor support for LBNL’s high-field magnets and LARP, the development of new high-field technologies, and fundamental research through the education of young talent.

Luca Bottura is a Nuclear Engineer at the Engineering Faculty of the University of Bologna (Italy), and has received a PhD from the University College of Swansea (Wales, UK) for the physical modeling, scaling and numerical analysis of quench in large forceflow cooled superconducting coils. After nine years of experience in the design and testing of superconducting cables and magnets for fusion (NET and ITER), he joined CERN in 1995, where he initially supervised field mapping activities for the LHC magnets, and devised the Field Description for the LHC (FiDeL), an embedded system of the LHC controls. As of July 2011, he is the leader of the MSC group in the CERN Technology Department, in charge of the resistive and superconducting magnets for the CERN accelerator complex, the associated manufacturing and test technologies, and installations.

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Reviews of Accelerator Science and Technology Vol. 5 (2012) 51–89 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300034

Superconducting Magnets for Particle Accelerators Lucio Rossi CERN, Technology Department, Gen` eve 23, CH-1211, Switzerland [email protected] Luca Bottura CERN, Technology Department, Magnet Group Leader, Gen` eve 23, CH-1211, Switzerland [email protected]

Superconductivity has been the most influential technology in the field of accelerators in the last 30 years. Since the commissioning of the Tevatron, which demonstrated the use and operability of superconductivity on a large scale, superconducting magnets and rf cavities have been at the heart of all new large accelerators. Superconducting magnets have been the invariable choice for large colliders, as well as cyclotrons and large synchrotrons. In spite of the long history of success, superconductivity remains a difficult technology, requires adequate R&D and suitable preparation, and has a relatively high cost. Hence, it is not surprising that the development has also been marked by a few setbacks. This article is a review of the main superconducting accelerator magnet projects; it highlights the main characteristics and main achievements, and gives a perspective on the development of superconducting magnets for the future generation of very high energy colliders. Keywords: Superconductivity; magnets; accelerators; large scale applied superconductivity.

1. Introduction

and the unveiling of the new world beyond the Standard Model — make the LHC the crossroads between past and future R&D.

In the same year in which he received the Nobel Prize for the investigation of properties of matter at low temperature (we can certainly say superconductivity is the most striking part of it!), K. H. Onnes [1] was dreaming of a 100,000 G (10 T) magnet. However, it took much more time than he thought for the dream to become a reality, and only at the end of the 1970s did superconducting magnet technology really take off. From then on, magnets have been the most important application of superconductivity, and accelerators can be credited with being among the drivers of the development of this technology. One characteristic of accelerators is that they are pursued by large laboratories, with programs over a long time, which allows investigation and R&D to be done in a fruitful way. The latest example, the LHC, is the summit of over 30 years of development of superconducting magnets (SCMs). Its giant size and its outspoken goals — the Higgs particle, whose recent discovery [2] has been heralded worldwide,

2. Main Characteristics of Superconducting Magnets for Accelerators Magnets are different according to the type of accelerators they are intended for. Colliders or synchrotrons for high energy physics (HEP) rings are the most challenging and they are superconducting with a few exceptions, such as fast cycling machines (for example, the J-PARC main ring). Magnets for superconducting linear colliders, such as the ILC or X-FEL, are also superconducting, a choice mainly motivated by the advantage of integration with the superconducting radio-frequency (SCRF) cavities, rather than by technical gain. Finally, there is a considerable effort in the development of SCMs for low energy accelerators, like cyclotrons or synchrocyclotrons. These were used in the past mainly for nuclear research but are now mainly for medical 51

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

52

applications. This type of SCM is akin to magnets for particle detector spectrometry or for MRI. In this article we will consciously restrict the scope to SCMs for HEP colliders. SCMs become of interest when the required field is above the iron saturation limit, i.e. approximately 2 T. In some instances, however, SCMs are becoming the preferred choice for their compactness and their low energy consumption, a topic that is increasingly important in the design of new accelerators.

2.1. Functions of superconducting magnets The first function of a magnet is to guide and steer the particle, i.e. to keep it in orbit in a circular accelerator or to just bend in a transfer line. The second main function is focusing the beam, thus providing it with the necessary stability in the plane perpendicular to the trajectory. Except when the beam has a very low energy, a domain where solenoids are suitable, the adequate force can be given only by a transverse field, i.e. a magnetic field perpendicular to the particle trajectory. In the case of high energy accelerators, like synchrotrons and colliders, the beam region is a cylinder that follows the beam path and has the smallest practical dimension, as shown in Fig. 1. As will be discussed later, the cost and technical complexity of the magnetic system are proportional to the energy stored in the magnetic field, which explains why it is important to minimize the size of the magnet bore.

Fig. 1.

Schematic of an accelerator dipole.

Despite the fact that static magnetic fields do not accelerate, in circular accelerators the bending (dipole) field eventually determines the final energy reach. In relativistic conditions, the relation between the beam energy, Ebeam (in TeV), the dipole field B (in T) and the radius of the beam trajectory inside the bending field R (in km) takes a very simple form: Ebeam ∼ = 0.3BR.

(1)

Since the dipole field typically covers two-thirds of the accelerator, R is about two-thirds of the average radius of the ring. Equation (1) shows clearly the interest in the highest possible field for a given tunnel. The other important function of magnets is, in general terms, to assure the stability of the beam in the transverse space. This is accomplished by quadrupoles, sextupoles and octupoles. While the main quadrupoles are usually of similar — although reduced — size and complexity with respect to the main dipoles, sextupoles and octupoles are of much smaller size and field. In this context it is worth mentioning the quadrupoles that are just before the collision points. These quadrupoles, usually called low-β, provide the optical manipulation to focus the beam to the smallest feasible dimension (the minimum β function) at the collision point. This set of quadrupoles features an aperture significantly larger than the lattice quadrupoles, and requires a large field gradient, which results in a peak field that can be close to that of the main dipoles. A key demand on accelerator magnets is good field quality: the harmonic content of main magnets and low-β triplet quadrupoles needs to be designed, controlled during production and corrected in operation to a precision of 10–100 ppm. Given the unavoidable manufacturing tolerances and uncertainties, this implies extensive use of corrector magnets. Small dipole correctors are needed to maintain the beam on the desired orbit or to generate local bumps. Higher order corrector magnets, of small size and in great numbers, are distributed all over the ring to trim the linear beam parameters such as tune and chromaticity, and insure nonlinear stability. As an example, in the LHC there are almost 8000 superconducting corrector magnets from dipoles up to dodecapoles (12-pole), distributed all over the ring.

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

2.2. Magnet design The design of superconducting accelerator magnets is largely concerned with the optimal distribution of compact superconductors around the beam aperture. In fact, in contrast to classical electromagnets, the field in a superconducting accelerator magnet is mainly produced by the current in the conductor, rather than the magnetization of an iron yoke. Very schematically, an SCM for large scale accelerators consists of a coil wound with highly compacted cables, tightly packed around the bore which delimits the vacuum chamber hosting the beam. The coil shape is optimized to maximize the bore field and achieve acceptable field quality, as described later. The large forces that are experienced by the coil (several tens to hundreds of tons per meter) cannot be reacted by the winding alone, which has the characteristic shape of a slender racetrack (see

Fig. 2.

53

Fig. 1), and hence the force is transferred to a structure that guarantees mechanical stability and rigidity. The iron yoke that surrounds this assembly closes the magnetic circuit, shielding the surroundings from stray fields and providing a marginal gain of magnetic field in the bore. In addition, it can have a structural function in reacting or transferring the Lorentz forces from the coil to an external cylinder. Finally, the magnet is enclosed in a cryostat that provides the thermal barrier features necessary for cooling the magnet to the operating temperature, which is in the cryogenic range (1.9–4.5 K for accelerators built to date). Various implementations of this basic concept are discussed in the review of the historical development of accelerator magnets (see Sec. 3), while in Fig. 2 the cross section of the LHC dipole in its cryostat illustrates this principle.

Cross section of the LHC dipole in its cryostat.

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

54

2.2.1. Electromagnetic design The prime purpose of the electromagnetic design is to obtain a multipolar field with the quality demanded by beam physics requirements. For long and slender HEP magnets, the magnetic field is 2D and is best represented using complex multipoles [3]. Defining the complex variable z = x + iy, where the plane (x, y) is that of the magnet cross section, the function By +iBx of the two components of the magnetic field is expanded in series: By + iBx =

∞


(Bn + iAn )z n−1 .

(2)

n=1

The coefficients Bn and An of the series expansion, called normal and skew components respectively, are the multipoles of the field, and determine the shape of the field lines. A pure multipolar field has only one nonzero Bn (or An ) for a given value of n, which is called the main order of the field. As an example, n = 1 is a pure dipole and n = 2 is a quadrupole. Nonzero Bn and An for n other than the main order are usually referred to as field harmonics or field errors. Several ways can be found to generate perfect multipolar fields required by HEP accelerator magnets. As an example, we show in Fig. 3 a number of arrangements of current distribution that generates a perfect transverse dipole field. It has been demonstrated by I. Rabi [4] that two uniform current density cylinders with opposite current polarity generate a perfect dipole field in the region of current overlap, i.e. in the internal current-free region. The total surface current, Js = J · t, J being the current density and t the current thickness at the midplane, is maximum at the midplane and zero at the vertical axis, following a cos ϑ behavior. More generally, the uniform dipole field may be generated by a constant current density with a geometry given by opposite intersecting ellipses. The current

Fig. 3.

distribution obtained with two overlapping ellipses, rotated by 90◦ , produces a quadrupole field. In the above cases the region of the intersection cannot be a circle. As an alternative, a perfect dipole field can be generated by a shell of current, of constant thickness t, in which the volume current density is maximum at the midplane and vanishes toward the pole region with a J = J0 cos ϑ dependence. In this case the inner region can be perfectly circular. A J0 cos ϑ not only generates a perfect dipole field, but is the most efficient current distribution, i.e. any other distribution requires more total current (Amp`ere turns) to generate a given central field; it produces more magnetic flux and has higher stored energy. This consideration is very important for SCMs for which the cost of the conductor is one of the dominant cost factors. It is instructive to compare the central field of the dipole generated by the ideal cos ϑ distribution, B0 = 1/2µ0 J0 t, to that of an infinitely long solenoid of the same current density and thickness, i.e. B0 = µ0 J0 t. In practice, the same current density and thickness generates only half the field of the solenoid configuration: what in a solenoid is a moderate field level can be considered a great achievement for transverse field magnets! As with the dipole, a perfect quadrupole field can be generated by a thin shell of constant thickness with a current density varying as J = J0 cos 2ϑ. In fact, a thin shell configuration can generate any multipole field of order n with a current distribution J = J0 cos nϑ. In practice, SCMs are constituted by shells of constant current density, with spacers, in such a way as to mimic the cos ϑ distribution, i.e. they are a mix of the two concepts mentioned above. A practical coil cross section can be approximated as sectors of uniform current density shown in Fig. 4. The configuration shown in the figure (lef t) generates an

Uniform dipole and quadrupole fields generated by intersecting circular and elliptical conductors.

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

Fig. 4.

55

Principle of sector coils that generate an approximate dipole field (lef t) and quadrupole field (right).

approximate dipole B1 , with higher order field errors. Because of symmetry, the only field errors produced (allowed multipoles) are normal multipoles of order 2n + 1, i.e. B3 , B5 , B7 , . . . . Similarly, the configuration in Fig. 4 (right) produces an approximate quadrupole B2 with normal higher order multipoles of order 2(2n + 1), i.e. B6 , B10 , B14 , . . . . In Tables 1 and 2, a set of practical formulae for the main field and errors are reported for dipoles and quadrupoles. Examining the equations in Table 1 for the field and field errors in the sector coil dipole, we see that a choice of ϕ = 60◦ cancels the sextupole error B3 . The first nonzero multipole error, the decapole B5 , is a few percent, i.e. much larger (two orders of magnitude) than is acceptable in an accelerator magnet. Better field quality can be obtained by segmenting the sectors using insulating wedges, and using two (or more) nested layers. This adds degrees of freedom that can be used to improve the field homogeneity,

at the cost of an increased complexity of the winding. In Fig. 5 we show the coil cross sections of the four large scale superconducting synchrotrons. It is evident how the coils have evolved in complexity to follow the increased demand of field quality. In analogy to the dipole, a choice of a sector angle ϕ = 30◦ in a quadrupole cancels the first allowed multipole, the dodecapole B6 . Tables 1 and 2 report other key quantities for the design of an accelerator magnet, namely the resultant forces in a coil quadrant (or octant), the midplane stress and the energy per unit length. An additional quantity of relevance is the coil radial width, w = Rout − Rin , which is used to estimate the overall coil volume, mass and material cost. Tables 1 and 2 can be used to study the functional dependencies of these quantities on the main design inputs, namely the desired field, the magnet aperture and the average current density. As an example,

Table 1. Practical analytical formulae for the dipole field and field errors for the dipole sector coil configuration in Fig. 4 (lef t). The force in a coil refers to a quadrant. The azimuthal stress is intended as average on the coil midplane. B1 =

Main field Field errors n = 3, 5, . . . , 2i − 1

An

Stress in the mid-plane

Energy per unit length

− R2−n 2µ0 R2−n in J out sin(nϕ) π n(2 − n) =0

Bn =

# √ ! √ « „ 4π + 3 π 3µ0 J 2 2π − 3 3 3 Rout + Rout + ln R3in − Rout R2in π 36 12 Rin 36 6 √ « – » „ « „ 1 3µ0 J 2 1 3 Rin 1 − R ln R3in + Fy = π 12 out 4 Rout 12 – » 3µ0 J 2 1 4 2 1 Fz = Rout − Rout R3in + R4in π 6 3 2 « – » „ „ « 2 2 6µ0 J 1 1 1 5 3 Rin + Rout + ln R3in − Rout R2in σθ = 4π 36 6 Rout 3 4 Rout − Rin " « # „ πB 2 R2in 2 Rout − Rin 1 Rout − Rin 2 1+ E/l = + µ0 3 Rin 6 Rin √

Force per coil quadrant

2µ0 J(Rout − Rin ) sin(ϕ) π

Fx =

"

√

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

56

Table 2. Practical analytical formulae for the field gradient and field errors for the quadrupole sector coil configuration in Fig. 4 (right). The force in a coil refers to an octant. The azimuthal stress is intended as average on the coil midplane. G = B2 =

Main field

« „ 2µ0 Rout sin(2ϕ) J ln π Rin

2−n 2−n 4µ0 Rout − Rin J sin(nϕ) π n(2 − n) An = 0 √ « – » « „ „ 4 2 1 3µ0 J 1 12Rout − 36R4in Rin + R3in + ln Fx = 6π 72 Rout Rout 3 # " ! √ √ √ « √ „ 4 2 1 Rin 3µ0 J 3−4 5−2 3 3 Rin 2− 3 Fy = + Rout + ln R3in + π 36 12 Rout 6 Rout 18 « – » « „ „ 1 3µ0 J 2 1 4 Rout + R R4in − ln Fz = 4π 4 out Rin 4 √ « – » „ „ « 4 1 3µ0 J 2 1 7R4out + 9R4in 1 Rin + ln R3in + σθ = π 36 Rout 3 Rout 3 Rout − Rin " «# „ πG2 R4in 1 1 R4out − R4in Rin ln E/l = − “ “ ””2 8 R4in 2 Rout 2µ ln Rin

Field errors n = 6, 10, . . . , 2(2i − 1)

Bn =

Force per coil octant

Stress in the mid-plane Energy per unit length

0

Rout

Fig. 5. Coil cross section, to scale, for the dipole magnets of (left to right) Tevatron, HERA, RHIC and LHC (for the LHC, only one of the two twin coils is shown).

for a dipole of a given field B and aperture Rin we have the approximate relations 1 1/2 1 , F ≈ const, σ ≈ J E ≈ ; J J for a given current density J and aperture Rin we have w≈

w ≈ B,

F ≈ B2,

σ ≈ B,

E ≈ B 5/2 ,

and for a given field B and current density J we have w ≈ const,

F ≈ Rin ,

σ ≈ Rin ,

3/2

E ≈ Rin .

Given a design field, the radial width of the coil (and thus the mass and cost of the magnet) scales inversely proportional to the current density in the winding. In addition, at constant coil width, an increase in the current density allows a higher field to be reached. In other words, the current density is the most important design parameter of accelerator magnets. In Table 3, the typical overall current density

of various systems is reported: accelerator magnets work by far at the highest value, so it is no surprise that they give the strongest reason for the highest current density. The benefit of superconductivity is evident from the previous consideration. To quantify this benefit, it suffices to note that the LHC ring is 26.7 km and requires some 130 tons of LHe inventory to operate its 8.3 T SCMs. The power at the plug of the refrigeration system is about 45 MW. By contrast, if the LHC had been built with classical resistive magnets at 1.8 T, the circumference would have been more than 100 km long and would have required about 900 MW of installed power (the power output of a large nuclear power plant unit). This would have led to prohibitive construction and operational costs, and in addition it would have had an unacceptable impact on the environment since the 900 MW power is rejected as warm water (at a temperature which is too low for any use).

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators Table 3.

Current density Joverall and other characteristics of different types of large magnetic systems.

Magnetic system (only dc) Resistive-air cooled Resistive-water cooled SC magnets for particle detectors SC Tokamaks for fusion∗ SC magnets for MRI SC laboratory solenoids SC accelerators ∗ Top

57

Current density Joverall (A/mm2 )

Operating current (kA)

Typical field range (T)

System stored energy (MJ)

1–5 10–15 20–40

1–2 1–10 2–20

2700 A/mm , 5 T–4.2 K, in conjunction with high quality fine filaments of 5– 6 µm and the new cable insulation based on full polyimide. However, a number of technical choices were probably not fully optimized — contributing, with bad management and adverse circumstances, to the decision to cancel the project. Among them one has to mention the use of single bore magnets which

The LHC was starting R&D at the end of the 1980s, with very low profile and funding. It relied a lot on the SCC R&D. However, to take full advantage of the existing 26.7-km-long LEP tunnel, it pushed the Nb– Ti magnet technology to its extreme. Design innovations were [26]: (i) use of the two-in-one design proposed first by BNL and dismissed for the SCC and RHIC; actually, the LHC went further by using for the dipole the original “twin” variant, where the two channels are fully coupled both magnetically and mechanically; (ii) cooling to 1.9 K to boost Nb–Ti performance and make use of the superior conductivity and heat transfer properties offered by superfluid helium. In fact, up to 1989 the use of Nb3 Sn was still considered, with R&D that was marked by the first break of the 10 T wall by a dipole coil [27]; however, cost and industrial maturity were in favor of Nb– Ti and HeII, and the performance of Nb3 Sn had to improve significantly over that of Nb–Ti, which happened only many years later, as shown in the graph of Fig. 16 and discussed in Ref. 10. Another significant innovation of the LHC SCMs was in the control of the interstrand resistance of the cable. Also called contact resistance, Rc , this critical parameter that controls interstrand coupling currents was still an open point of the SSC design at its closure (at least for the SC booster); the large LHC cable is vulnerable to values below 10 µΩ, while the high Jc calls for collective cable stabilization, which can disappear for values beyond 100–200 µΩ. Interstrand resistance values in the right window, 10 < Rc < 200 µΩ, were obtained by carefully

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

69

Fig. 16. Critical current density of Nb–Ti, at 4.2 and 1.9 K, compared with that of Nb3 Sn at 4.2 K, this last at the time of the LHC decision on superconductors and today.

controlling the thickness of the Sn–Ag coating layer of the strands, followed by a customized (according to actually measured thickness) air heat treatment of the whole cable to induce surface oxidation [28]. 3.5.3. LHC dipole magnet design The LHC main dipoles (or main bends, MBs) have been designed for a nominal operation at 8.33 T, 86% of the maximum current Imax , or l = 0.86 (see Subsec. 2.2.4), which is the critical current on the load line. The mechanics was, however, designed for the “ultimate” field of 9 T, corresponding to l = 0.93. The design of the LHC MBs has gone through about 10 years of evolution, with three generations of design, all featuring two coil layers wound with different cables [29]. The three generations differ in the coil layout, in the collar design and in how the coil–collar assembly interferes with the yoke-skin assembly. Making reference to Fig. 17, in which the details of a quadrant of one aperture are visible, the basic design characteristics of the third — final — generation are: Coil layout. With a 56 mm free bore, it is based on six-conductor blocks. After an unsuccessful attempt in 1992–98 to work with five coil blocks of the second generation, the final design has been based on an optimized six-block layout, where the conductors are as radial as possible and the shear forces among conductors are minimized. Coils. The two conductor shells are wound with different cables whose margins in critical current are very similar, so it is an optimized grading. This

Fig. 17. Quadrant of the LHC dipole, with all mating surfaces indicated.

feature has improved the central field per unit current but has reduced the margin to quench, and it implies that imperfections in the winding of the second layer are as important as those of the inner layer, despite the considerable number of turns grouped in the two blocks of the outer shell. The coils are composed of poles of two layers each. The necessity to avoid sorting in order not to slow the production requires that each pole, and even each layer, be identical within 100 µm, which corresponds to a variation of about 0.1% of the main field, 3.5 and −0.4 units (10−4 ) of the main harmonics, sextupole and decapole respectively, and to

December 29, 2012

17:30

70

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

about 12 MPa in azimuthal coil prestress. The necessity of top–bottom and left–right symmetry means that coils must all be similar, even better than the stated figures. Collars. These are of the twin type, i.e. the two magnetic channels are fully coupled, a novelty that was not without risk. Made out of special austenitic steel with very low magnetization under operating conditions, collars are obtained by fine blanking according to a shape that ensures the wanted coil cavity under stress and cold conditions, and for this reason the collars are slightly elliptical (ε = 0.1 mm) when punched. The choice of stainless steel was introduced relatively late in the Project, after a long period when an aluminum alloy was preferred. With the aluminum collar the prestress induced by thermal contraction would have been good for 5 T or so, like for HERA. The additional restraining was coming from the external shrinking cylinder through the yoke. For this reason the yoke had an open vertical gap; this gap and the weld-induced shrinking proved to be very difficult to control at the desired level of accuracy, especially in long magnets industrially built. The first long magnet of the LHC first generation, with six coil blocks [30], had aluminum collars and in 1994 reached 9 T in two quenches and no retraining; see Fig. 18. The approval of the LHC was based on that magnet; however, this success was not always reproduced by subsequent magnets. So, austenitic steel for collars was introduced as the last big modification, allowing for a more comfortable margin in the construction and assembly tolerances. Also, thanks to its higher rigidity, the use of austenitic steel helps

Fig. 18. Training quench of the first full prototype of the LHC dipole (CERN–INFN collaboration, 1994).

limit (but cannot avoid) conductor movements, as discussed in Ref. 14. Iron. The presence of an iron insert at the vertical symmetry plane in between collar and iron yoke (see Fig. 17) helps the assembly accuracy and the transmission of vertical force from the shell to the collars, through the iron yoke, in a position that is critical for the twin design. Indeed, the lack of left-to-right symmetry in twin collars is one of the main disadvantages with respect to the single collaring coil assembly. The inclined surface of the iron insert is meant just to compensate for the reduced rigidity in the central arm of each aperture of the collars. The interference between the iron yoke and the collars is also situated at the midplane, the outer arm of each aperture, and at two different positions along the collar outer arm. The iron insert proved to be useful for fine-tuning the field quality for even harmonics. Cold mass assembly. As previously stated, the coil– collar assembly is surrounded by the magnetic circuit contained by a shrinking cylinder, formed by welding two half-shells made out of austenitic steel. The carefully controlled welding shrinkage provides the necessary rigidity to the whole magnet. The forces are transmitted by interference among very rigid pieces (collars and yoke). Therefore, not only is the precision of the single pieces high (typically ± 20 µm for the collars and ± 50 µm for the yoke) but the assembly must ensure this precision as well over the 15 m magnet length. The typical azimuthal stress history (at the coil– collars interface, near the pole region, at 90◦ ; see Fig. 17) for the LHC dipoles is shown in Fig. 19, where there can be noted relatively low prestress after cool-down, about 20% of the peak stress seen by coils during manufacturing, maybe the least satisfactory feature of the LHC dipole structural design. Magnet-to-magnet variation can be easily of the order of 15–20%. One should also note that the upper limit of the magnet powering is not necessarily zero stress, which implies detachment of the coil from the pole: if the friction associated with the movement does not generate excessive heat, the magnet can indeed reach a field in the range of 9.5 T to 10 T, corresponding to Imax , as some models and prototypes did. However, this usually occurs with many training quenches and the reproducibility of this result would

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

71

Fig. 19. Compressive stress at the coil–pole interface (see Fig. 17) for the LHC dipole during construction, cool-down and energization.

be very poor and not suitable for a reliable operating machine. The magnet is curved, with a sagitta of about 9 mm, corresponding to a radius of curvature of 2812.36 m. This curvature has a tolerance of ± 1 mm, with the exception of the extremities of the magnet, where the tolerance is very tight: ± 0.3 (systematic) and 0.5 mm rms in order to keep the corrector magnets centered with respect to the beam tube, so as to avoid harmonic feed-down (detrimental to beam optics.)

3.5.4. LHC main and insertion quadrupoles The main dipole system constitutes the backbone of the LHC and it alone accounts for 85% of the cost of the magnetic system, i.e. for almost 50% of the accelerator. The quadrupoles are of course the other main part of the magnet system, the last one being the corrector magnets, which will not be discussed in this article. There are various types of large quadrupoles: the main quadrupoles for a regular lattice and part of the dispersion suppressors, special quadrupoles for the rest of the dispersion suppressors, the matching sections between experimental regions and the regular arc, and the low-β quadrupoles just before the collision points. The main quadrupoles (MQs) have been designed by the CEA–Saclay (France) team in collaboration with CERN [31]. The design was based on the same conductor as the main dipole outer layer, for cost and risk reduction — although, from the e.m. design point of view, the 15-mm-wide cable is really big for the 56 mm aperture of the magnet. For the same reason no grading was applied for the two layers, with some loss of design efficiency, but a gain in construction efficiency (use of the double pancake

technique with no joint between the inner and the outer layer). The ratio of peak to useful field (λ; see Subsec. 2.2.4) is highly optimized. The magnet is a classical two-in-one, where the two coils are lodged in one iron yoke which has a central ridge — a central iron arm that magnetically decouples the two apertures. The forces are such that free-standing collars can comfortably retain the stress, the only coupling between collars and iron being the pins to insure straightness and proper alignment. The matching section quadrupole (MQM) has many features similar to those of the MQ; however, the design was achieved with a much smaller cable, the objective being to reduce the powering current (6 kA, versus the 12 kA required by the MQ). Since all these magnets are individually powered [32], this choice greatly simplifies the cold powering. The coil “efficiency,” i.e. the quantity of the superconductor for a given field/gradient, is higher than in the MQ. This is, however, at the expense of smaller margins for quench and stability. Very interesting is the case of the MQY: these quadrupoles, also of classical two-in-one design with amperage of 6 kA or less, feature a large aperture — 70 mm. They operate at 4.2 K; however, thanks to grading of the superconductor in the two double layers and also to a special grading (the transition from the inner to the outer type of cable happens inside the second layer), they reach a peak field similar to the MQ and MQM (more than 6 T in operation, with 7.5 T as the magnet critical field) [32]. The low-β quadrupoles feature a single wide aperture of 70 mm and operate at 1.9 K. Here the two teams that shared the construction ended up with two very different designs. The Fermilab MQXB [33] was based on large cables (12 kA), with grading among the two layers, and free-standing collars inside

December 29, 2012

17:30

72

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

the iron yoke. The KEK MQXA [34] was based on small cables (7 kA) with four layers. The coils are assembled using 10-mm-wide spacer-type, nonsupporting collars. The prestress in the coils and their rigidity are provided by the yoke structure, which consists of horizontally split laminations keyed at the midplane. The locking of the yoke is similar to that of the collar for dipoles, so deviation from fourfold symmetry is a possible risk, which was finally successfully mastered. The MQXA design uses 40% more superconductor than the MQXB, and therefore has a higher peak field and gradient, which translate into a bigger margin, but it was possibly more complex to manufacture than the MQXB. Both designs proved to be successful and reached the goals, showing that optimization is not a rigid and unique concept. A more detailed comparison of the e.m. design of these magnets, and others as well, can be found in Ref. 35; in Fig. 20 the cross section of the MQXA magnet is shown.

4. Experience from LHC Magnets The LHC is such a large project with such a large number and so many types of accelerator magnets that the feedback from the construction and commissioning has great value for design and future projects.

Fig. 20. Cross section of the MQXA, the low-β quadrupole designed by KEK in collaboration with CERN.

4.1. Construction Design alone, good as it could be, cannot guarantee the accuracy and uniformity required of accelerator magnets: component construction and assembly are critical, too. The construction has been described elsewhere [36, 37], so here we recall that in the LHC the most important components that could influence the performance of the magnets, and that required uniformity all along the production, were kept under the control of CERN: (i) All components of the coils from superconducting cables to copper spacers; (ii) The main components of the 2D cross section: austenitic steel collars, iron yoke laminations, external cylinders; (iii) Most mechanical components of the magnet, e.g. cold bore tube (beam pipe), HeII heat exchanger tube, end covers; (iv) Most electrical components, e.g. quench heaters (for pre-series magnets), superconducting bus bars, protection diodes, instrumentation feedthrough. All these components were designed and purchased by CERN separately and given to the magnet assemblers. In certain cases CERN even purchased the raw materials, which were then given to a company for transformation into finished components before delivery to magnet assemblers. This was the case when the properties of the raw materials were critical for performance: (a) the superconductor (Nb–Ti alloy) of the superconducting strands; (b) austenitic steel for collars; (c) low carbon steel (iron) of the yoke lamination. In this way the risk of being “supplier of its own suppliers” was great and occasionally generated tension and organizational problems, and in a few cases some extra cost. However, in this way the final product was guaranteed and basically no surprises were experienced at the test and measurement of a magnet. Somehow CERN moved quality checks to the earliest possible moment, rather than waiting to check the final result and eventually rejecting what was not conforming. This was the key to the fact that all contracts for the main dipoles were completed according to the schedule, with negligible extra cost. Since the coil is compressed in a cavity; given by the locked collars (see Figs. 14 and 17), the

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

Fig. 21. Variation of the shim thickness for the two coil layers of LHC dipoles during manufacturing.

azimuthal prestress and the azimuthal length of the coils, this last determining the harmonic content, are not independent. As discussed earlier, the coil size determines the harmonic content. A 0.12 mm variation of the azimuthal length of the coil package (i.e. about 0.25% of the developed length) causes a variation of the sextupole b3 of more than four units. On the other hand, if the coil is forced at a constant azimuthal dimension, the prestress at room temperature varies by 15 MPa. It was decided to steer production for field quality, i.e. the coil geometry was fixed irrespective of the induced stress variation, provided that the coil size — measured under 50 MPa stress — was not deviating by more than ± 0.12 mm. A larger variation triggered a fine-tuning of suitable shim positioned between coil upper end and pole; see Fig. 17. The shim variation from nominal size of the entire 1278-dipole production is reported in Fig. 21, from Ref. 38, where it can be noticed that adjustment

73

was rare and mainly at the beginning. Variation of the coil’s azimuthal length along the 15 m length of a similar coil was negligible, with σ ∼ 20 µm. A key step in the quality control was the use of extensive magnetic measurements at low current, 8 A, at room temperature, carried out during construction on each magnet both as collared coil and then as “yoked” cold mass. Measurements were introduced first for steering production toward the field quality allowed window. This procedure triggered two fine-tunings of the cross section, done via change of the spacers inside the coil, to stay inside the target (see Fig. 22, from Ref. 38), where the results of the two interventions, carried out without stopping production, are clearly visible in the measured data. Dipoles with the first two cross sections were allocated almost all in one sector of the machine, compensating locally for the differences in the measured field components (see the next section). In addition, the system proved to be a useful tool for detecting serious hidden defects that could impact on the magnet quench or electrical performance. Intercepting defects at the earliest stage, by any means, including complex magnetic measurements, has been a guiding star for the LHC project. Its usefulness also for correcting weak procedures, besides finding single large mistakes, is summarized in Fig. 23, where the number of coil disassemblies triggered by magnetic measurements on the production premises is plotted as a function of production. Mistakes were numerous at the beginning; then, after a quiet period when we thought that production was well under control, we had a sudden increase, and we

Fig. 22. Sextupole components of all dipoles measured at warm at the three manufacturing sites. The data on the left of the dashed vertical line are the first 30 dipoles with the cross section before the first fine-tuning (see text for details).

December 29, 2012

17:30

74

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura Table 4.

Fig. 23. Number of defects serious enough to trigger a coil disassembly during LHC dipole production.

were on the verge of blocking production. The sudden increase in the number of mistakes was correlated with massive injection of new personnel, necessary for reaching the required rate for mass production; review of procedures, tighter QA and careful training of personnel were among the remedies that eventually made the two-thirds remaining of production very stable, with very few deviations observed. The production of the LHC magnets lasted seven years. Preseries contracts (7% of the total) were signed in 2000, series contracts in 2002, and the last dipole and quadrupole were delivered perfectly on time, on 7 November 2006. We had less than 2% of magnet rejections, and only two (less than 0.2%) could not be repaired. This success demonstrated the achieved maturity of the SCM technology based on Nb–Ti, eventually completing the route initiated by the Tevatron.

Magnet sorting criteria.

Type of magnet

Number

Sorted quantity

Arc dipole (MB)

1232

Arc quadrupoles (SSS) Dispersion suppressor quadrupoles

360 64

Matching sections quadrupoles Cold low-beta quadrupoles (Q1. . .Q3) Separation and recombination cold dipoles (D1. . .D4) Warm quadrupoles (MQW)

50

Geometry Field (b1 , a2 , b3 ) Training Field (b2 ) Training of trim quadrupoles Geometry Geometry

24

Geometry

16

Geometry

48

Field geometry

based on the magnet performance (mainly geometry and field quality, but also electrical and quench issues). The objective was to preserve and, if possible, optimize the machine performance as originally projected by the design studies. Sorting followed the criteria listed in Table 4. Some of the motivations and results of sorting are shown by two representative plots in Figs. 24 and 25. In Fig. 24 we show the maximum deviation of the mechanical axis of the dipole for all magnets installed (two bores, V1 and V2), on top of the class definition for installation in the dispersion suppressor region (golden), any location in the arc (silver) and the middle of a regular cell (midcell). In spite of

4.2. Installation and sorting Well-performing magnets are a prerequisite for a good accelerator, but beam quality and reliable operation require additional considerations of sorting and optimal placement, as well as precise and reproducible control. This was done to a high degree for the LHC. A total of approximately 1800 large SCMs had been produced, tested and installed during the four years of construction of the LHC. Each magnet was examined from the viewpoints of electrical performance, magnet protection, field quality and magnet alignment, by a Magnet Evaluation Board [39] charged with magnet acceptance and sorting for the whole ring. The installation and tunnel location of the magnets in the LHC accelerator were optimized

Fig. 24. Measurements of the geometry of the LHC main dipole and sorting class (see text for details).

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

Fig. 25. Rms b3 for the two beam tubes of the LHC: with no sorting (higher columns) and with sorting.

a significant fraction of magnets exceeding the manufacturing specification of 1 mm on the axis deviation from the nominal orbit, sorting of dipoles by their geometry has allowed us to preserve the accelerator mechanical aperture to values exceeding initial specifications, and with no bottleneck in the arc. A similar benefit of sorting is observed in Fig. 25, where we plot the rms normal sextupole b3 for the arcs of each of the eight sectors into which the LHC is subdivided, for both of the apertures V1 and V2. Two columns report rms b3 as would be obtained by a random installation, and the two smaller columns represent its effective value when pairing and compensation are taken into account. Typically, the values of b3 fall within the manufacturing specification, with a marginal exception for sector 78, which contained the initial production with a different cross section. After sorting, the effective rms b3 is reduced by a factor of 3, with the benefit of a strong reduction of the third order resonance driving term. 4.3. The 2008 incident On 19 September 2008, just nine days after the spectacular first circulating beam in the LHC, the magnet system experienced a very grave incident. During the current ramp-up of the last sector which was not yet brought to its maximum (indeed, start-up was done with two sectors not yet fully tested), a faulty electrical interconnection, carrying about 9 kA, melted away, generating an electric arc that perforated the helium vessel. There were many consequences, with huge collateral damage due to the helium pressure rise (above 8 bar, instead of the 1.5 bar of design). Consequently, 53 main magnets were removed and

75

only 16 could be reinstalled after minor interventions; 37 magnets, a more than 500 m length of the accelerator, had to be replaced with spares, which were barely sufficient in number. The finding and recommendation can be found in Ref. 40, while detailed descriptions and considerations of the reasons have been discussed in Ref. 41. The incident was triggered by the only completely faulty interconnection (now we know this for sure); however, the electrical splice between magnets is done according to a weak design and even apparently good splices have a nonnegligible probability of undergoing a thermal runaway at high current because of discontinuity of the copper stabilizer. For this reason the accelerator performance is today limited to 60% in terms of the magnetic field and then of the beam energy. Only a campaign of consolidation, foreseen in 2013–14 and aimed at assuring stabilizer continuity in all splices and at fixing the defectives ones, will allow the LHC to reach its nominal design parameters. It is, however, useful to comment that the incident happened because an interface was not given the necessary attention. Considered mainly as a mechanical problem, there was insufficient analysis of the superconducting behavior of the bus bar joints, which eventually resulted in lack of continuity of the stabilizer and in lack of diagnostic and consequent protection. It is also interesting to note that the electrical arc alone would have seriously damaged the two adjacent magnets, with some important but not severe collateral damage (pollution of the beam pipe). What made the incident so grave was the lack of protection against a release of liquid into the warm cryostat that was ten times larger than expected in the considered fault scenario. The magnet, cryostat and cryogenic teams did not discuss the possible worst scenario together, in this way missing implementation of a few simple measures that would have greatly reduced the consequences. The lesson is somehow simple and complex at the same time: while the “superconducting magnet” is by far the most complicated part, the magnetic system is the realm of complexity. System integration is much more than simply fitting the components: it is the discipline that makes a laboratory apparatus a working device. In the case of the LHC magnet system, the superconducting bus bars, and its joints, earlier considered an “easy” component, are actually

December 29, 2012

17:30

76

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

a critical one that, together with the protection diode and the coil quench heaters, will probably be a source of concern during all of the LHC life. System integration is a view that should be considered at the early stage of a project, and even discussed together with the project of the main components, especially when dealing with a phenomenally complex machine like the LHC. It is also important to note that recovery was very fast: indeed, the huge work of repair and reinstallation was carried out in less than nine months. In Fig. 26 one can see the LHC magnets installed in the tunnel. 4.4. Operation 4.4.1. Magnetic model The installed magnets are presently operated by a complex control system that translates the desired optics into powering currents on a circuit-by-circuit basis. The key elements of this conversion are

Fig. 26. LHC magnets in the LHC tunnel during electrical checks: in blue are the main dipoles, while the U-shaped tube visible in the middle connects a quadrupole to the cryogenic line.

the circuit transfer functions, based on measured field quality data that have been parametrized and synthesized into simple but complete fitting functions that form the Field Description of the LHC (FiDeL) [42]. FiDeL is a much-evolved form of the Tevatron feedforward and of the RHIC magnetic model. It allows dispensing of the reference magnets originally planned for the LHC, as was done for HERA, with a nonnegligible saving of cost and logistics. In addition to the transfer functions, FiDeL provides a parametrization of the field errors in the main magnet circuits (dipoles and quadrupoles), which is used to forecast currents in correction circuits. The commissioning tests of FiDeL, during the first injections into the LHC, have shown that the field model has a predictive capability of better than 10 units on the integrated dipole field (0.1% in energy), better than 25 units on the integrated quadrupole gradient (0.2 units of tune), and better than 0.5 units on the integrated sextupole (20 units of chromaticity). These results, achieved blindly (i.e. without beam feedback), are quite spectacular. Trimming the model using the accumulated beam-based measurements, and adopting strict precycling procedures, has reduced the range for day-to-day operation by one order of magnitude. The experience from the operation of all previous SC accelerators has led to a very good understanding of a number of key issues, the most outstanding being chromaticity control and beta-beating [43]. The stability of the magnetic machine, also thanks to the carefully studied precycle, is one of the characteristics most appreciated by the operation crew and one of the key elements of the success of LHC operation. Chromaticity correction, which has puzzled operators at the Tevatron and plagued those at HERA, is now possible at the LHC to within the equivalent of a few units of chromaticity (i.e. a few ppm of the sextupole field error) to be compared to 10 units (1000 ppm) of the sextupole error generated by persistent current in the dipoles (see Fig. 14). This includes a practically lossless compensation for the infamous decay and snapback responsible for up to 30% beam loss at the Tevatron. Beta-beating, an indicator of the goodness of the local optics, has been found in the range of 30– 40% for the bare machine at injection (i.e. before corrections are applied). This becomes 10% at the end of the ramp to 4 TeV, i.e. within the extremely

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

tight beam specification. Local corrections reduce the above values to 5–10% at most throughout the ramp.

4.4.2. Quench and powering Accelerator magnets exhibit training, i.e. they initially quench at a current level well below Imax , defined in Subsec. 2.2.4, but at each successive current ramp they tend to improve the current level at which spontaneous quench occurs. Eventually, the quench training attains a plateau, at which point the quench current does not improve any longer, as shown in the training curve of Fig. 18. The plateau current is ideally Imax . In practice, however, for accelerator magnets the quench plateau is frequently around l ∼ 90%. This is not surprising, recalling our discussion on margins and stability. On one hand, when approaching Imax the reduction of the temperature margin and the limited amount of stabilizer cause a strong reduction of the stability margin, vanishing at l = 1. At the same time, the spectrum of mechanical perturbations increases in amplitude as the field increases, also because the mechanical structure may be designed for an operation current smaller than Imax (as previously mentioned, the LHC dipole has been mechanically designed for the ultimate operation at 9 T, i.e. l = 93%). After a full warm-up of the magnet, the training generally restarts, with a new cycle of training quenches. The training memory is defined as the difference between the level of the last quench before a thermal cycle and the level of the first quench after the thermal cycle. Good memory means little loss of quench level, of the order of 5% or less, like the one shown in Fig. 18. Bad memory is sometimes referred to as “detraining.” All LHC magnets have been cold-tested, at the final acceptance test at CERN, which has allowed their mechanical and electrical integrity and their quench performance to be assessed [44]. The 1232 LHC dipoles required on average just one quench per magnet to go beyond the nominal operating field of 8.3 T, at 11,850 A. The quench tests (carried out directly at 1.9 K without a previous test at 4.2 K) are important not only to assess if a magnet reaches and passes the nominal operating field but also to train the magnet in such a way that once it is installed in the tunnel it should not require much retraining. Because of the small stability margin, the memory between thermal

77

cycles is far from being assured, and so the initial test strategy was to push all magnets to the ultimate field of 9 T in the first training curve and submit to a thermal cycle and retraining curve only the magnets with sluggish training behavior (those that needed more than nine quenches to reach 9 T). After some initial delay, the rate of magnet delivery ramped up so fast that it was not possible to execute all the planned tests within the allocated time. The magnet test sequence was then redefined and as a consequence the majority of the magnets were tested only up to the nominal field of 8.4 T or quenched not more than twice before they attained 8.6 T. Only about 10% of dipoles have been submitted to a thermal cycle, either for verification purposes or because they were underperforming with respect to the above criteria. For this sample, the number of quenches per magnet to reach the nominal field at the second thermal cycle dropped, as expected, to an average of 0.10–0.15. The quadrupoles, having lower e.m. forces and stored energy, and a larger margin than the dipoles, needed on average less than 0.5 training quenches per magnet to pass the nominal current level of 11,850 A, and their memory was found sufficient to assure practically zero quench below nominal current in a training performed after the thermal cycle. During commissioning of the installed magnets, in summer 2008, we found that a family of dipoles presented a loss of memory higher than expected. In Fig. 27, the forecast of quenches in the tunnel based on the single acceptance tests is reported as a continuous curve for all three dipole families (each

Fig. 27. Quench of LHC dipoles in the tunnel. Continuous curves represent the forecast based on the acceptance test, separated by manufacturer. Dot and cross markers are actual quenches that occurred during hardware commissioning (HC).

December 29, 2012

17:30

78

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

family corresponds to a different manufacturer); the spread in the curves indicates the uncertainty in the analysis. On the same plot, we report using dot and cross markers the quenches actually experienced during the commissioning campaign, before the startup of the accelerator. Because of problems linked to the quench detection, the dipole current during commissioning was limited to the equivalent of 6.5 T (about 5.5 TeV in terms of beam energy). Out of the eight sectors, each comprising 154 dipoles, only one sector — sector 5–6 — was pushed up with the intention to reach nominal operation at 8.3 T. However, as can be noticed in Fig. 27, a number of quenches occurred, and the training slope is rather flat, more than expected from the extrapolation of data from series magnet tests. The training campaign of sector 5–6 was stopped at 7.9 T (6.6 TeV), because of the incident in sector 3–4, and it will be resumed only after interconnection consolidation in 2015; see Subsec. 4.3. Sector 5–6 contains a predominance of firm 3 magnets, and for them a larger loss of memory was observed. It can be seen from the plot that the forecast for firm 3 (black curves) was slightly worse than for the other two families. It was, however, surprising to notice that the quenches during commissioning (indicated as “firm 3 HC” cross markers) stay consistently well below the corresponding forecast from the acceptance test at 1.9 K. For the moment the reason for this unexpected behavior is not known and no correlation could be reasonably established with factors that may have influence on training behavior. A study on this effect is reported in Ref. 45. This effect, if confirmed, will probably require an estimated four months to train the whole LHC to reach 8.3 T, corresponding to the 7 TeV nominal beam energy. This is a tantalizing time, especially because there is no guarantee that the training memory will be retained upon a thermal cycle of an LHC sector. However, we also see that it should be very fast to reach a field corresponding to beam energies in the range of 6.5–6.7 TeV. This energy is short of the nominal goal, but should be a safe value to be reached rapidly after the 2013–14 shutdown. For a hadron collider the discovery potential is not a threshold function of the beam energy; the slight loss in terms of beam energy will be partly compensated for by more integrated luminosity, and hence is not a big issue.

5. Magnets for Pulsed Synchrotrons The main challenge for fast-cycled accelerator magnets is not so much the bore field and aperture, which fall in the typical range of feasibility already demonstrated, but rather to achieve them with the required repetition rate, economically and reliably. A number of specific issues can be mentioned: • AC loss. The control and reduction of AC loss in the cold mass has the utmost importance in reducing cryoplant investment and operation cost, and in limiting the temperature excursions in the conductor. • Cooling. The heat loads on the magnet, especially those originating from the AC loss and beam heating, must be removed efficiently to warrant a margin sufficient for stable operation. • Quench detection and protection. Protection of SCMs is especially demanding in the case of fastramping machines, due to the high inductive voltages in comparison with the voltage developed by a resistive transition. Voltage compensation and magnet protection must be proven in the presence of an inductive voltage during ramps that can be as large as 1000 times the detection threshold. • Field quality. The contribution of coupling currents in the superconductor and eddy currents in the iron yoke is difficult to predict, control and measure at the desired resolution during fast ramps. • Material fatigue, over several hundred million cycles, influencing material selection and, possibly, requiring dedicated testing. An impulse to this line of research was given recently by the German laboratory GSI in Darmstadt, which is pursuing the construction of a new Facility for Antiprotons and Ion Research (FAIR) [46]. The central part of this complex is the two rings SIS100 and SIS300, which will be built in the same tunnel and will have magnetic rigidity Bρ = 100 Tm and Bρ = 300 Tm, respectively. To achieve this magnetic rigidity, the dipoles of SIS100 will have a bore field of 2 T, with a window frame geometry mentioned in Subsec. 2.2.2 (see Fig. 8), providing a rectangular bore of 130 mm × 65 mm. The dipoles of SIS300 have a classical cos ϑ layout, and they provide a peak field of 4.5 T in a round bore with a diameter of 100 mm. The magnets for these two rings are

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

especially challenging, because the operation mode of the complex foresees fast ramping of the energy. SIS100 should undergo a full cycle in 1 s, corresponding to a ramp rate of 4 T/s. The ramp rate requirements for SIS300, which will operate as a storage ring, are softer, but still the aim is to ramp the ring at 0.5 to 1 T/s. The SIS100 R&D at GSI is supported by activities at the JINR laboratory in Dubna (Russia). A synchrotron similar to SIS100, the Nuclotron, has been in operation at JINR since 1994 [47]. The Nuclotron dipole magnets are operated in the accelerator at a peak field of 1.5 T, ramping at 0.6 T/s, and have achieved a peak field of 2 T, ramping at 4 T/s. For SIS300, initial work has been performed in collaboration with BNL in the USA. At the time the accelerator had lower beam energy and dipole field requirements (it was actually named SIS200). A prototype magnet, GSI001 with a single layer coil and similar in construction to the RHIC dipole, was built and tested successfully at BNL, demonstrating operation up to a 4 T bore field in pulsed conditions up to 4 T/s. The magnet sustained short pulse sequences between 2 T/s (500 repeated cycles) and 4 T/s (3 repeated cycles) without quenching [48]. Since end-2006, the Italian INFN, in collaboration with GSI, has launched a prototype design and

79

construction activity to demonstrate the feasibility and test the performance of a dipole for SIS300 [49]. The INFN program, dubbed DiSCoRaP, has focused on the design and construction of a dipole prototype with peak field of 4.5 T, curved with a sagitta of 28 mm over a 4 m length and with a ramp rate of 1 T/s, i.e. the present parameters for the SIS300 dipoles. The dipole (see Fig. 28) is at present under test. An additional interest of the above R&D, beyond the construction of FAIR, is that the range of design parameters considered for SIS100 and SIS300 is the same as would be necessary for an upgrade of the PS and SPS injectors at CERN. The range of parameters reported above is relatively large, and spans different technologies. We have nonetheless tried to find a common denominator among the various options by plotting in Fig. 29 the required field ramp rate (dB/dt) versus the peak field in the bore B. We notice an interesting feature of the scatter plot, namely that most of the points for fast-ramped magnets are clustered around a curve B × (dB/dt) = const. In fact, the product Π = B × (dB/dt) is proportional to the power per unit volume released in the magnet. Hence, for a given magnet design, an increasing value of Π is associated with higher terminal voltage and AC loss, two of the main issues for ramped magnets listed

Fig. 28. Cross section of the cold mass and picture of the first curved coil of the DiSCoRaP dipole for FAIR-SIS300 (courtesy of G. Volpini, INFN–Milano).

December 29, 2012

17:30

80

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

Fig. 29. B–dB/dt chart for SCMs: the ones for pulsed operation accumulate around the upper trend line.

earlier. Neglecting the large range of designs reported in Fig. 29, which is a conscious oversimplification, we can thus use Π as an indicator of the ramped performance: magnets with the same Π are assumed to be equally difficult to design and build, with respect to the pulsed mode. Most magnets presently in design or prototyping for ramped applications are aimed at a target value of Π = 7 T2 /s, which also covers the range of parameters considered for both a PS and an SPS upgrade. The plot in Fig. 29 also reports the computed value of Π for the four large scale superconducting synchrotrons, to demonstrate the jump in performance sought after. 6. Next Generation High Field Magnets 6.1. High luminosity LHC Although beam energy, which determines the discovery reach, is the first objective of a collider, luminosity, proportional to the instantaneous rate of collisions, is a very close second. Even more important is the time integral of the luminosity, i.e. the total number of collisions that can be recorded by the detectors. Once a collider has pushed the performance of the dipole magnets to its maximum practical limit, reaching the highest possible beam energy, the next, unavoidable step is to increase the luminosity, in order to improve the number and statistical relevance of the data from the detectors. If we

take the example of the LHC, the collider presently operates at 4 TeV, which is approximately 60% of the nominal energy. The beam luminosity achieved so far is 0.7 · L0 , where L0 is the nominal design luminosity, L0 = 1034 · cm−2 · s−1 . This is an excellent result, considering that the planned increase in energy will yield additional luminosity. An indication of the good performance is the integrated luminosity (proportional to the total number of collisions), which amounts today to 15 fb−1 (inverse femtobarns), which is already more than the total accumulated during the seven years of Tevatron Run II. After the 2013–14 shutdown, when the main circuits splices will be consolidated (see Subsec. 4.3), the LHC is expected to reach an operating energy approaching the nominal 7 TeV. At this energy the luminosity will most likely exceed the nominal value L0 , thus further increasing the rate of data production. After a few more years, however, an increase in luminosity is required to further extend the physics reach of the collider. Indeed, around 2022 the present plan foresees carrying out an upgrade of the LHC to increase its peak luminosity by a factor of 5 and possibly multiply by a factor of 10 its integrated luminosity [50]. To reach this goal, at least two magnet systems, the collimation region and the low-β quadrupoles, will require a substantial upgrade, well beyond the limit of the present LHC technology. Especially for the low-β quadrupoles, the required performance is near the limit of the performance of Nb3 Sn (see Ref. 51 for a thorough discussion on limits of SC material for a high field). The work described above is within the scope of the broader High Luminosity LHC upgrade (HL-LHC). The High Luminosity machine is the next frontier for accelerator magnets: while the LHC has been the peak of 30 years of Nb–Ti magnet development, the HL-LHC gives the opportunity to prove on an existing accelerator the suitability of high field magnet technology, on a limited number of magnets. If successful, this could open the door to another, much bigger project based on SCMs.

6.1.1. The 11 T two-in-one dipole The increased luminosity will require that additional collimators be placed in the dispersion suppressor (DS) region, which is at 1.9 K, to protect from the increased rate of particle loss. The DS region does

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

not have built-in space for the new collimators, which are 3–4 m long. One possibility of creating this space is to substitute a standard dipole (8.3 T × 14.3 m ∼ = 120 Tm) with a dipole of equivalent integral bending strength, but producing a stronger field (11 T) over a shorter length (11 m). The bore field is well beyond the reach of Nb–Ti technology and therefore Nb3 Sn technology must be used for such magnets: Fermilab and CERN collaborate closely on this project. Such a two-in-one dipole has many severe constraints. It must be powered in series with the other LHC main dipoles (i.e. the given operating current) and should be practically identical in bending strength (i.e. the given integral field) with harmonic content not too far from that of the LHC dipoles. In addition, the distance between the two apertures and the outer diameter of the cold mass must be the same as for a standard dipole (i.e. broad geometry fixed). The design adopted is a classical cos ϑ; see Ref. 52 and Fig. 30. The coil is based on cable built using a 0.7 mm strand with a minimum current 2 density of 2750 A/mm at 12 T. The nominal copper content is 53% and the effective filament diameter is presently in the range of 40–50 µm. The coil is double-layer, like the LHC dipole, but without superconductor grading, which would have required splicing between cables. The electromagnetic design is further complicated by two issues: (i) the superconducting persistent currents, due to high Jc and a relatively large filament size, generate a b3 harmonic of ∼ 45 units (10−4 of the main field) at the injection field, six times larger than that of the LHC main dipole; (ii) the 30% higher field in an iron yoke geometry that is essentially the same as for the

81

LHC dipoles results in large saturation of the transfer function, and unacceptably high b3 at the flat top field — 6.6 units. Reduction of the persistent current sextupole can be obtained by means of passive magnetic shims near or in the coils. However, a further effort in reducing the filament size of the Nb3 Sn conductor from the present 50 µm to the range of 30 µm is pursued to bring the residual effect to within an acceptable range of correction of ± 10 units. The saturation effects are strongly reduced by shaping the internal iron profile and by a set of three saturation control holes — a well-known technique also used in the RHIC and LHC. The mechanical design to withstand the forces, ∼ 70% higher than in the LHC dipole, relies on clamping by austenitic steel collars and by a lineto-line fit between collars and iron yoke: the iron yoke and outer shell are assembled with interference, a procedure that avoids excessive stress during collaring but requires very tight tolerances and careful assembly. In this way the transverse stress, a constant concern with fragile Nb3 Sn, is kept below 150 MPa under all conditions, while the pole–coil interface remains always under compressive stress. The plan for the LHC is to manufacture the 11-m-long dipole by joining in the same cryostat two straight magnets of length 5.5 m, placed such that the empty zone for the collimators is in the middle of the cryostat length. This configuration has the advantage that the equivalent kick is identical to that of a standard LHC dipole, without the need for further corrections. While the total number of

Fig. 30. Cross section of the 11 T two-in-one dipole and picture of the finished 2-m-long cold mass (single bore) ready for testing at Fermilab.

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

L. Rossi & L. Bottura

82

such dipoles is still under discussion, pending a better understanding of the present performance and future needs of the collimator system, the demonstration and prototyping works are in full swing. The plan foresees two short (∼ 2 m) single bore dipoles, with the final cross section to be manufactured by end of 2012. The test of the first one is underway, with encouraging results. A full size, 5.5-m-long prototype is expected by 2015. 6.1.2. The low-β quadrupoles and shell–bladder structure The keyword for the magnets needed for the upgrade in the collision region is “large aperture.” The goal is to be able to further squeeze the beam in the interaction regions below the 55 cm nominal value of the betatron function at the interaction point (IP), the so-called β ∗ . The plan is to reduce β ∗ by a factor of 4, down to about 15 cm, so the aperture of the quadrupoles has to double from 70 mm to ∼ 150 mm, the beam size being proportional to (β ∗ )−1/2 . To keep the quadrupole triplet as compact as possible, the required gradient and the very large aperture result in high fields in the coil, in the range of 12 T. The baseline option is hence to procure Nb3 Sn quadrupoles. The technology is not yet fully validated for use in an accelerator, but remarkable progress has been achieved within the scope of the 10-year-long US-LARPb effort, through which several 90-mm-aperture quadrupoles and one 120mm-aperture quadrupole have been built and tested in multiple variants and conditions.

Fig. 31.

b LARP

1230003

In the first R&D phase, dealing with 1-m-long model magnets, two different mechanical designs have been pursued and evaluated: the classical collar-type structure and the so-called shell–bladder structure. The shell–bladder structure, first developed for accelerator magnets at LBNL by S. Caspi [53], consists in precompressing the coils against an external restraining cylinder, or shell. We use the example of Fig. 31, where we consider for simplicity two racetrack coils, and substitute the collars with two thick plates, to illustrate the concept. Bladders in the interface between plate and iron yoke are pressurized to reach the final prestress plus an amount to compensate for spring-back. Permanent keys are inserted with minimal tolerance in slots between the plate and the yoke, after which the bladders are depressurized and removed. The coils are hence left in the desired situation of initial prestress. When the magnet is finally powered, the e.m. forces act on the coils, and release a portion of the prestress with ideally no movement of the coils; see Fig. 31. With respect to the classical collar system, the shell–bladder concept directly controls stress, rather than relying on the effect of interference between collar and coil. The prestress is thus less sensitive to the actual size and rigidity of the coil. In addition, since the cylinder can be in a material of high thermal contraction such as aluminum, there is the benefit of additional strain during cool-down. The latter effect can in principle be obtained also using aluminum collars (as in HERA and the early LHC design), but so far designers have

Prestress based on the bladder-and-keys concept applied to a racetrack coil system.

(LHC Accelerator Research Program) is a US collaboration sponsored by DOE. The laboratories working on the LARP magnet R&D are BNL, Fermilab and LBNL.

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

Fig. 32. Training curve of the Nb3 Sn LARP quadrupole LQS01b with the indicated equivalent Nb–Ti performance. The line indicated with SSL (short sample limit) is the magnet Imax .

considered impractical the use of aluminum collars for high field magnets. The shell–bladder design has been so successful that in 2008 it was selected by the LARP collaboration as the baseline structure for the low-β quadrupole for the LHC upgrade, which at present is based on cos ϑ layout. The successful test of the first “long” Nb3 Sn magnets — the so-called LQS, whose coils are 3.6 m long — has marked the definitive affirmation of this structure that has gone beyond expectation, attaining 90% of Imax with the version LQS01b [54]; see Fig. 32. The structure is evolving further, adding features to assure alignment and field quality, with the design and construction of HQ, a 1-m-long model with a large coil aperture, of 120 mm — a further

Fig. 33.

83

step toward the final requirement. Adding these features, which are now almost provided with a collar structure, makes the shell–bladder structure a little bit more complex (see Fig. 33), still retaining a few advantages with respect to collars. However, more designs and tests are necessary, especially in long magnets, before we can state that this shell–bladder structure constitutes a full alternative to collars for accelerator quality magnets. The HQ quadrupole poses formidable challenges, given the jump in stresses and in stored energy per unit length. HQ has successfully reached 80% of its Imax , a threshold that is critical to qualifying a design for operation, and that for this quadrupole means a record 12 T peak field in the coils [55]. Nonetheless, a number of key issues, mainly of electrical insulation reliability, large strand and cable magnetization and repeatability of results, still need to be resolved. For conductors, the target is Jc in the range of 1500 A/mm2 at 15 T (for comparison, the ITER 2 Nb3 Sn strand is based on Jc of 1000 A/mm at 12 T, 2 i.e. approximately 500 A/mm at 15 T) and filament size of 50 µm or less. The two high current options (RRP from OST, USA; and PIT from Bruker, EAS, Germany) are both viable, with a clear advantage at this moment for the US product, which has benefitted from the long term conductor development program guided by DOE. The main issues that still have to be resolved and on which the community is concentrating the effort are: • Performance. Magnets still have to fully prove reliable operation at 80% of the short sample, and should be fully free from conductor instability.

Cross section and inside view of the LARP HQ quadrupole (courtesy of G. Sabbi, LBNL).

December 29, 2012

17:30

84

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

• Field quality. Coil geometry reproducibility, which is related to the random component of the field harmonics, although encouraging and improving, suffers still from poor statistics. A cored cable is almost certainly needed to avoid ramp rate effects due to the potential of low interstrand resistance, which can result from sintering. The next generation of LARP coils will make use of a cored cable, to gain experience on this issue, which is also relevant to the 11 T dipole project. • Radiation resistance. All materials have to withstand an extremely high radiation load — to reach the final target of 3000 fb−1 , one has an accumulated dose that, in present estimates, could reach ∼ 10–100 MGy. A systematic program has been launched by CERN in collaboration with a few European institutes, as well as KEK and J-PARC in Japan. • Length. With 150-mm-aperture magnets providing a 140 T/m operational gradient, the machine optics needs magnets of 7 m and 9 m length. So far, Nb3 Sn dipoles and quadrupoles exist as 1-mlong models, and a few 3.6-m-long quadrupoles. Replacing 9 m by two 4.5-m-long units is becoming the baseline, with moderate impact on luminosity performance.

6.2. The high energy LHC and the HTS frontier The possibility of increasing the beam energy of the LHC has been considered at CERN in 2010 [56]. The project appears feasible; the most critical issue is the maximum field attainable by the main dipoles, which determines the final performance of the machine, according to (1). The minimum goal of the high energy machine, the HE-LHC, is to double the present LHC design energy, but a more ambitious target has actually been set at 33 TeV of center-of-mass (collision) energy. A proton beam energy of 16.5 TeV requires operation of the main dipoles at 20 T, with a huge jump beyond the state of the art, as can be seen from the plot of the historical evolution of the dipole field for hadron colliders shown in Fig. 34, where the range of interest for the HL-LHC and HE-LHC is indicated, together with the domain accessible by various superconductors; see also Fig. 11.

Fig. 34. Dipole field versus time for the main past projects and the region of interest for the LHC upgrades.

6.2.1. Generic high field dipoles: R&D For high field magnets, the stresses are such that the shell–bladder concept previously mentioned looks very attractive. It can be applied to quadrupoles and dipoles, to cos ϑ and to block coil layouts, i.e. to coils rectangular in shape, like the one in Fig. 31. The fact that Nb3 Sn coils have a very high modulus, more than 20 GPa rather than the 5–10 GPa common for NbTi, makes controlling stress via collars more difficult and favors the shell–bladder structure. The quest for a high field dipole is at present underway via four main programs: (i) LBNL, with a long historical record, is pushing the limit of the rectangular block coil with the shell–bladder structure, with a series of magnets called HDs. HD2, which is the first to feature a free bore obtained with flare coil ends, (see Fig. 35), reached 13.8 T, which is about 78% of Jc [57], while HD3, with a larger bore, has experienced some electrical problems that are temporarily delaying the project; (2) the LD1 program — Large Dipole 1 — is a 13 T dipole with a large bore (> 100 mm) for a high field US cable test facility on the horizon of 2015; (3) the EU program EuCARD is aiming at producing first a large bore (> 100 mm) 13 T dipole, for the CERN cable test facility called Fresca2, by 2013, and then at reaching a total field of 19 T by adding a small HTS racetrack without a free bore; (4) the EU program EuCARD2 — just approved to start in 2013 — aims at developing a 10 kA class HTS cable and at designing and manufacturing a 5 T, 40 mm bore dipole of accelerator quality, wound with the cable mentioned above. The scope is to eventually insert the 5 T HTS

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

85

Fig. 35. HD2 dipole of LBNL: detail of the coil with arrow indicating a conductor displacement in the corner (left); FE model of the coil showing race-track structure with flare ends (center); picture of the cold mass before testing (right).

dipole into a large Nb3 Sn dipole to prove that HTS can enhance the field to > 15 T with a useful bore. 6.2.2. Magnets for the collider: HE-LHC The possibility based on a dipole having an operating field of 24 T for an energy upgrade of the LHC was already considered in 2006 [58]. This was based, 2 however, on a current density Je of 600–800 A/mm at 24 T, still far from being achievable. Recently, at CERN, a study has been carried out, and the target field for the main dipoles, the main driver of the entire project, has now been set to 20 T (operational) in a 40 mm bore, which would enable the HE-LHC to reach 33 TeV center-of-mass energy for proton collisions [59]. A prestudy clearly identified the following critical points: • The required margin is set to about 20%, i.e. l = 0.8, which is large in absolute terms, being 5 T. The possibility of designing for a large l, and a lower margin, must be thoroughly investigated. • The overall current density of the coil should be 2 around 400 A/mm , at the design field, as it is for dipoles of previous accelerator magnets [60]. This requires that the engineering current density Je of the basic element, strand or tape, be substantially 2 higher than the overall 400 A/mm in the operating coil. In order to generate 20 T in the bore, the coil width is about 80 mm, almost three times the width of the LHC coil. • Bore size has been fixed at 40 mm, and the outer diameter of the iron flux return yoke must

not exceed 800 mm (compared to 570 mm in the present LHC dipoles), which is a tough constraint considering the amount of magnetic flux that needs to be intercepted. Based on previous observations, a preliminary magnet layout was designed, using Nb–Ti, Nb3 Sn and HTS conductor. Based on rectangular coil blocks (see Fig. 35), for its better suitability for shell–bladder force retention (anyway, given the size of the coil with respect to the bore, the gain of cos ϑ versus block is negligible), it features an unprecedented superconductor grading, for cost reduction and volume containment, while gaining maximum performance. The possibility of separate powering of the coil sections is being considered. Although this configuration complicates the circuitry and the interconnections, it can offer us some key advantages: • It allows separate optimization of cable size and amperage for the three materials. Moreover, while Nb–Ti and Nb3 Sn can be manufactured in very large cables (15–20 kA); this is not at all possible for HTS. • Coil segmentation will favor magnet protection, a technique largely employed in the large solenoid magnets (working at ∼ 1 kA rather than > 10 kA as accelerator magnets do), which is probably needed with an amount of stored energy and inductance ∼ 15 times higher that of the LHC. • Dynamic compensation of the field harmonics. This is extremely important, since it is very unlikely that Nb3 Sn and in HTS will feature the

December 29, 2012

17:30

86

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

Fig. 36. Cross section of the preliminary magnet lay-out of the 20 T dipole for HE-LHC. The upper expanded quadrant shows the different superconductors used in each coil block, while the lower quadrant reports the peak field in each block.

5–7 µm filament size developed for the SSC and LHC Nb–Ti. We have to live with filaments in the range of 25–50 µm for Nb3 Sn, and most probably in the range of 50–100 µm for the HTS part, with sextupole components coming from persistent current of 50–100 units. Although the large ratio of coil width to bore size makes it easier to reach the necessary field quality, use of passive shims is unlikely to fully compensate for these large effects. Separate powering, first proposed for the SSC [61], would facilitate compensation of these effects as well as other dynamic effects due for example to interstrand resistance that are difficult to fully control at the cable level. The project presents immense challenges, the first one being to make available the necessary superconductors and then to make them the required conductors. The total quantity of superconductor is three times that used for the LHC, i.e. about 3000 tonnes of finished strands (or tapes), containing about 40% of superconductor and 60% of stabilizer. Nb3 Sn certainly needs further improvement; however, it is on the good route, thanks to the

HL-LHC–driven R&D. The biggest uncertainty concerns the HTS materials, which are still far from being ready for this type of practical application. The candidates are only Bi-2212 round wire, which has better 4.2 K transport properties than Bi-2223, and YBCO tape, whose performance is reported in Fig. 11. Bi-2212 has the right topology for compact cables and possesses isotropic properties [62]. However, it has strong strain dependence, requires high temperature heat treatment in an O2 atmosphere, and has been plagued, so far, by bubble formation and ceramic leakage through the silver barrier. Current density Je is not far from the target; basically, we would need a moderate, but not modest, improvement of 50% over that of Fig. 11. Cost remains an issue, as well as little support from the rest of the superconducting magnet and conductor development community. YBCO has the desired current density and it is mechanically robust. However, it is strongly anisotropic (see for example the two curves in Fig. 11), and is available only in the form of tapes.

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

Even ignoring the problem of magnetization, tapes are not suitable to be assembled in compact flat cable. Solutions like the Roebel bar and other types are under investigation. Indeed, one main aim of the above-mentioned EuCARD2 program is to address this issue, complementing the ongoing program in the USA, more focused on Bi-2212. One issue, unfortunately equally shared by Bi2212 and YBCO, is their very high cost: at least five times that of Nb3 Sn, which in turn is five times that of Nb–Ti. Any large scale high field magnet program is hence completely dependent on the success of the superconductor R&D and cost reduction program, in particular for HTS. The basic R&D study on HTS for the HE-LHC must be carried out in the next 4–5 years, since by 2016 or 2017 a credible and substantiated design must be available. Should HTS not meet the very demanding requirements of the HE-LHC, the door to the 16–20 T region will be closed, at least for the objectives of the HE-LHC. The HE-LHC magnets will then be based on Nb–Ti and Nb3 Sn technology, with the goal being a maximum operating field of 15.5 T, a figure that still enables a respectable 26 T center-of-mass (collision) energy. 7. Conclusions Accelerators are a fascinating domain for superconducting magnets because they constitute a very demanding application, with a spectrum of properties required for the magnets that are unmatched by other applications. We have limited our review to HEP accelerator magnets, which are by far more numerous and the ones where the technological advance is more evident: other types of accelerators require SCMs that are more similar in technology to solenoids and detector magnets. For accelerator magnets, quench level is a critical issue, of course, but to enable their use in accelerators many other properties have to be controlled to a narrow window, simultaneously requiring a rigorous and integrated approach. The success of the technology and its enabling character is shown by the fact that four very large projects based on SCMs have been built in the last 30 years: Tevatron, HERA, RHIC and LHC. In this article we have discussed the continuation of the success story, namely the planned upgrades of the LHC and the first large pulsed synchrotron, FAIR, to be built at GSI. One

87

key ingredient of the success is the capability of the accelerator community to provide a common forum for materials scientists, superconducting technologists and magnet engineers to discuss and to have mutual feedback. The global performance of an SCM — quench level, field quality, uniformity and reliability — depends in a decisive manner on the superconducting cable. New high field territory exploration critically depends on new materials — like HTS — growing from materials science to technical conductors. In a sentence: A superconducting magnet cannot be better than its conductor — but it could be much worse! if not well designed, carefully manufactured and thoroughly tested. Acknowledgments The authors thank Ezio Todesco of CERN for the fruitful discussions about magnet design, as well as for the material provided in the many years of analysis. The magnet scaling equations in Tables 1 and 2 are largely the fruit of the work of, and discussion with Paolo Ferracin of CERN. References [1] H. K. Onnes, Communication at the Third International Congress of Refrigeration (Chicago, Sep. 1913). [2] Last Update in the Search for the Higgs Boson, seminar held at CERN, 4 July 2012; available at: https://indico.cern.ch/conferenceDisplay.py?confId= 197461. [3] R. A. Beth, Complex representation and computation of two-dimensional magnetic fields, J. Appl. Phys. 37(7), 2568 (1966). [4] I. I. Rabi, A method of producing uniform magnetic fields, Rev. Sci. Instrum. Vol. 5, Feb. 1934, pp. 78–79. [5] E. Fischer, P. Schnizer et al., Design and test status of the fast ramped superconducting SIS100 dipole magnet for FAIR, IEEE Trans. Appl. Supercond. 21, 1844–1848 (2011). [6] A. M. Akhmeteli, A. V. Gavrilin and W. S. Marshall, Superconducting and resistive tilted coil magnets for generation of high and uniform transverse magnetic field, IEEE Trans. Appl. Supercond. 15, 1439–1443 (2005). [7] L. Rossi and E. Todesco, Electromagnetic design of superconducting dipoles based on sector coils, Phys. Rev. ST Accel. Beams 10, 112401 (2007). [8] L. Rossi and E. Todesco, Electromagnetic efficiency of block design in superconducting dipoles, IEEE Trans. Appl. Supercond. 19, 1186–1190 (2009).

December 29, 2012

17:30

88

WSPC/253-RAST : SPI-J100

1230003

L. Rossi & L. Bottura

[9] S. Caspi and P. Ferracin, Limits of Nb3 Sn accelerator magnets, in Proc. 2005 Particle Accelerator Conference (Knoxville, Tennessee), pp. 107–111. [10] L. Bottura and A. Godeke, Superconducting materials and conductors, fabrication, and limiting parameters; this issue, 2012. [11] M. N. Wilson, Superconducting Magnets (Oxford University Press, 1983). [12] K. H. Mess, P. Schmuser and S. Wolf, Superconducting Accelerator Magnets (World Scientific, 1996). [13] M. N. Wilson, NbTi superconductors with low ac loss: A review, Cryogenics 48, 381–395 (2008). [14] C. Lorin, P. P. Granieri and E. Todesco, Slip-stick mechanism in training the superconducting magnets in the Large Hadron Collider, IEEE Trans. Appl. Supercond. 21, 3555–60 (2011). [15] A. V. Tollestrup, Superconducting magnets, in Physics of High Energy Accelerators, eds. R. A. Carrigan, F. R. Hudson and M. Months (AIP Proceedings, 87, 1979), pp. 699–804. [16] S. Wolff, Superconducting Hera magnets, IEEE Trans. Magn. 24, 719–722 (1988). [17] M. Anerella et al., The RHIC magnet system, Nucl. Instrum. Methods A 499, 280–315 (2003). [18] A. Tollestrup and E. Todesco, The development of superconducting magnets for use in particle accelerators: from Tevatron to the LHC, in Reviews of Accelerator Science and Technology, eds. A. Chao and W. Chou, Vol. 11, pp. 185–210 (2008). [19] M. A. Green, GESSS Machine Design Committee Reports, Report KFK 1764, 1972 (Kernforschungs Zentrum Karlsruhe). [20] R. Perin, T. Tortschanoff and R. Wolff, Magnetic design of the superconducting quadrupole magnets for the ISR high luminosity insertion. Internal report CERN ISR-BOM 79-02, 1979. [21] H. T. Edwards, The Tevatron energy doubler: A superconducting accelerator, Annu. Rev. Nucl. Part. Sci. 35, 605–660 (1985). [22] P. F. Dahl et al., Superconducting magnet models for Isabelle, in Proc. PAC 1973, pp. 688–692. [23] A. V. Zlobin, UNK superconducting magnets development, Nucl. Instrum. Methods A 333, 196–203 (1993). [24] C. Peters et al., Use of tapered key collars in dipole models for the SSC, IEEE Trans. Magn. 24, 820– 822 (1988). [25] J. Strait et al., Tests of full scale SSC R&D dipole magnets, 25, 1455–1458 (1989). [26] R. Perin, The superconducting magnet system for the LHC, IEEE Trans. Appl. Supercond. 3 (1991). [27] A. Asner, R. Perin, W. Wenger and F. Zerobin, First Nb3Sn superconducting dipole model magnets for the LHC break the 10 tesla field threshold, in Proc. MT-11 Conference (Tsukuba, 1989) (Elsevier Applied Science, 1990), pp. 36–41. [28] D. Richter, J. D. Adam, D. Leroy and L. R. Oberli, Strand coating for the superconducting cables of the

[29]

[30]

[31]

[32] [33]

[34]

[35]

[36]

[37]

[38] [39]

[40]

[41]

[42] [43]

[44]

[45]

LHC main magnets, IEEE Trans. Appl. Supercond. 9, 734–741 (1999). L. Rossi, The LHC main dipoles and quadrupoles toward series production, IEEE Trans. Appl. Supercond. 13, 1221–1228 (2003); errata corrige in IEEE Trans. Appl. Supercond. 13, 3874–3877 (2003). E. Acerbi, M. Bona, D. Leroy, R. Perin and L. Rossi, Development and fabrication of the first 10 m long, IEEE Trans. Magn. 30, 1793–1796 (1994). M. Peyrot, J. M. Rifflet, F. Simon, P. Vedrine and T. Tortschanoff, Construction of the new prototype of main quadrupole cold masses for the arc short straight sections of LHC, IEEE Trans. Appl. Supercond. 10, 170–173 (2000). R. Ostojic, The LHC insertion magnets, IEEE Trans. Appl. Supercond. 12, 196–201 (2002). R. Bossert et al., Test results from the LQXB quadrupole production program at Fermilab for the LHC interaction regions, IEEE Trans. Appl. Supercond. 14, 187–190 (2004). Y. Ajima et al., The MQXA quadrupoles for the LHC low-beta insertions, Nucl. Instrum. Methods In Phys. Res. A, 550, 499–513 (2005). L. Rossi and E. Todesco, Electromagnetic design of superconducting quadrupoles, Phys. Rev. ST Accel. Beams 9, 102401 (2006). L. Rossi, Experience with LHC magnets from prototyping to large-scale industrial production and integration, in European Particle Accelerator Conference, EPAC2004 118–122 (Jacow website), pp. 141–145. L. Rossi, The Large Hadron Collider and the role of superconductivity in one of the largest scientific enterprises, IEEE Trans. Supercond. 17, 1005–1014 (2007). E. Todesco, Report on field quality in the main LHC dipoles, CERN report, EDMS No. 807803. P. Bestmann et al., Magnet acceptance and allocation at the LHC Magnet Evaluation Board, in Proc. PAC07 (Albuquerque, 2007), pp. 3739–3741. M. Bajko et al., Report of the Task Force on the Incident of 19th September 2008 at the LHC, CERN LHC Project Report 1168, 31 Mar. 2009. L. Rossi, Superconductivity: Its role, its success and its setbacks in the Large Hadron Collider of CERN, Supercond. Sci. Technol. 23, 034001 (2010). N. Sammut et al., Phys. Rev. ST Accel. Beams 9, 012402-1-12 (2006); 10, 082802 (2007). E. Todesco et al., The magnetic field model of the Large Hadron Collider: Overview of operation at 3.5 TeV and 4 TeV, in Proc. IPAC 2012, pp. 2194– 2196. P. Pugnat and A. Siemko, Review of quench performance of LHC main superconducting magnets, IEEE Trans. Appl. Supercond. 17, 1091–1096 (2007). C. Lorin, A. Siemko, E. Todesco and A. Verweij, Predicting the quench behavior of the LHC dipoles

December 29, 2012

17:30

WSPC/253-RAST : SPI-J100

1230003

Superconducting Magnets for Particle Accelerators

89

during commissioning, IEEE Trans. Appl. Supercond. 20 (2010). P. Spiller et al., Status of the Fair SIS100/300 Synchrotron design, in Proc. PAC07 (Albuquerque, New Mexico), pp. 1419–1421. A. D. Kovalenko, Nuclotron: Status and future, in Proc. EPAC 2000 (Vienna, Austria), pp. 554– 556. G. Moritz et al., Recent test results of the fast-pulsed 4 T cos Θ dipole GSI001, in Proc. 2005 Particle Accelerator Conference (Knoxville, Tennessee). P. Fabbricatore et al., Development of a curved fast ramped dipole for FAIR SIS300, IEEE Trans. Appl. Supercond. 18, 232–235 (2008). L. Rossi, LHC upgrades plans: Options and strategy, presented at the 2011 Int. Particle Accelerator Conference, IPAC 2011 (San Sebastian, Spain), pp. 908–912. E. Todesco and P. Ferracin, Limits to high field magnets for particle accelerators, IEEE Trans. Appl. Supercond. 22, (2012). A. Zlobin et al., Status of a single-aperture 11 T Nb3Sn demonstrator dipole for LHC upgrades. Presented at the 2012 Int. Particle Accelerator Conference, IPAC 2012 (New Orleans, USA); published on the Jacow website. P. A. Bish, S. Caspi et al., A new support structure for high field magnets. Lawrence Berkeley National Laboratory report, LBNL-47796SC-MAG 738.

[54] G. Ambrosio et al., Progress in the long Nb3Sn quadrupole R&D by LARP, IEEE Trans. Appl. Supercond. 21, 1858–1862. [55] S. Caspi et al., Test results of HQ01, a 120 mm bore LARP quadrupole magnet for the LHC luminosity, IEEE Trans. Appl. Supercond. 21, 1854–1857 (2011). [56] R. Assmann et al., First thoughts on a higher-energy LHC, CERN-ATS-2010-177 (2010). [57] P. Ferracin et al., Assembly and test of HD2, a 36 mm bore high field Nb3Sn dipole magnet, IEEE Trans. Appl. Supercond. 19, 1240–1243 (2009). [58] P. McIntyre and A. Sattarov, On the feasibility of a tripler upgrade for the LHC, in Proc. PAC’05, pp. 634–636. [59] E. Todesco and F. Zimmermann (eds.), The HighEnergy Large Hadron Collider, Report CERN2011-03 (2011). [60] L. Rossi and E. Todesco, Conceptual design of 20 T dipoles for high-energy LHC, in The HighEnergy Large Hadron Collider, report CERN-201103 (2011), pp. 13–19. [61] F. R. Huson et al., Superferric magnet option for the SSC, IEEE Trans. Nucl. Sci. NS-32, 3462–3465 (1985). [62] A. Godeke et al., Wind-and-react Bi-2212 coil development for accelerator magnets, Supercond. Sci. Technol. 23, 034022 (2010).

Lucio Rossi obtained his PhD in physics from the University of Milan in 1980 and then started his researches on applied superconductivity for accelerators in 1981, at the University of Milan, where he became Professor of Experimental Physics in 1992. He worked on the Milan K800 Superconducting Cyclotron (5 T, 40 MJ), on the ZEUS detector thin superconducting solenoid for HERA at Desy, on the first superconducting dipole prototypes for the LHC, and on the 25-m-long superconducting toroidal magnet of the ATLAS experiment. In May 2001 Prof. Rossi joined CERN, where he led the construction of the superconductor and magnets of the LHC Project. Since 2011 he has been in charge of the High Luminosity LHC project and of the High Energy LHC study.

Luca Bottura is a Nuclear Engineer at the Engineering Faculty of the University of Bologna (Italy), and has received a PhD from the University College of Swansea (Wales, UK) for the physical modeling, scaling and numerical analysis of quench in large forceflow cooled superconducting coils. After nine years of experience in the design and testing of superconducting cables and magnets for fusion (NET and ITER), he joined CERN in 1995, where he initially supervised field mapping activities for the LHC magnets, and devised the Field Description for the LHC (FiDeL), an embedded system of the LHC controls. As of July 2011, he is the leader of the MSC group in the CERN Technology Department, in charge of the resistive and superconducting magnets for the CERN accelerator complex, the associated manufacturing and test technologies, and installations.

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53]

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

This page intentionally left blank

1203001˙book

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Reviews of Accelerator Science and Technology Vol. 5 (2012) 91–118 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300046

Superconducting Magnets for Particle Detectors and Fusion Devices Akira Yamamoto High Energy Accelerator Research Organization (KEK), 1-1 Oho, Tsukuba, Ibaraki 305-0801, Japan [email protected] Thomas Taylor AT Scientific LLC, 11 Chemin du Gamay, 1233 Bernex, Switzerland [email protected] The application of superconductivity to the large magnets required for charged particle spectroscopy in high energy physics experiments, and for plasma containment in fusion experiments, has resulted in a spectacular leap in the efficiency of these devices. First applied in the late 1960s, the technology has progressed to meet increasingly demanding goals of the experiments and has stimulated important development of the associated conductors. In this article we describe briefly the basic requirements that determine the design of the different types of magnets. This is followed by descriptions of examples of representative working and projected magnets, as well as essential auxiliary equipment. An overview is provided of ongoing development that may impact on the design of future magnets. Keywords: Superconductor; magnet; detector; accelerator; collider; fusion; tokamak; stellarator.

1. Introduction

years. In order to ensure this cryogenic stability, a large fraction of stabilizing copper was incorporated into the matrix of the conductors used to build coils for bubble chamber magnets [4–6], at Argonne and Brookhaven National Laboratories in the US, and at CERN in Europe. Although they required a certain dexterity with the operation to eliminate problems due to eddy currents etc. the magnets achieved the desired fields and worked reliably, and the magnet for the Big European Bubble Chamber (BEBC) held the world record for stored energy of a single device for many years. These magnets used primitive conductors with large filaments of Nb–Zr or Nb–Ti which were made before the findings in the late 1960s of another fundamental study that indicated the twin requirements of small filament size and twisting [7]. This would enable the design of magnets with more compact coils using conductors requiring far less stabilizing material, thus heralding the modern era of superconductivity as applied to magnets. The magnets for spectrometers in physics experiments (i.e. “detector” magnets) operate under conditions different from those for fusion science and this has led to a different approach to their design

The appearance of practical type 2 superconductors in the early 1960s sent a shock wave through the communities of experimentalists involved in the megascience disciplines of high energy and fusion physics. Limits on the size — imposed by the cost of power, and on the field level — imposed by the saturation of iron, of the large magnets required for the spectroscopy of charged particles and for the confinement of plasma, would be swept away. Initial hopes were dashed when it was found that even small trial magnets were not stable due to flux jumps in the primitive conductors, but in 1965, when the criterion for full stabilization was first described [1–3], reliable devices could be envisaged. The concept is simple: incorporate sufficient normal-conducting material to provide a bypass for the current in the event of a transient loss of superconductivity, the resistance being low enough to ensure that with helium bath cooling the temperature does not exceed the critical temperature of the superconductor. The reaction was immediate, and very large spectrometer magnets that took advantage of the new technology were designed, built and running within a few short 91

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

92

and operation. Whereas detector magnets are usually operated at constant field, fusion magnets are either pulsed or coupled to pulsed fields: losses that occur in the superconductor when the field changes create heat that must be quickly removed, and this has a profound effect on how the cooling is provided. Thus, since pulsed magnets must be cooled directly via a bath or flow of helium, the indirect cooling method that has become standard for detector magnets cannot be conveniently applied. The corresponding conductor technology has also progressed and diverged to meet the goal of providing the maximum magnetic field with minimum material for high energy physics, and the maximum peak field for fusion. Despite this divergence in requirements, due to the similar scale of the devices there is much that the two communities can share, including certain aspects of the peripheral equipment, such as insulation, mechanical support and feeders. 2. Magnets for Particle Detectors 2.1. Requirements Magnetic fields provide a convenient method to analyze the momentum of charged particles. In order to extend the energy reach of detectors, magnetic fields are required on a large scale. The basic relation between magnetic field and charged particle trajectories is described by p = mv = qρB,

(1)

where p is the momentum, m the mass, q the charge, ρ the bending radius, and B the magnetic flux density. The deflection (bending) angle, θ and sagitta, s, of the trajectory are determined by θ≈

L L = qB · , ρ p

s≈q·B·

L2 , 8p

where c1 and c2 are constants [9]. The first term represents the error contributed by the Coulomb scattering (for a homogeneous medium). In addition, many particles are also short-lived, so there can be a definite advantage in going to high fields. A possible drawback is that in high fields the tracks of particles having low momentum curl up and can confuse the read-out [10]. So the precise goals of the experiment, in addition to the inevitable cost/benefit analysis, have a strong impact on the design of the magnet. The first large superconducting magnets referred to in the introduction were used in spectrometers for fixed target experiments. The fully stabilized conductors of Nb–Zr and Nb–Ti alloy embedded in copper were open-wound in pancake coils to provide a dipole field perpendicular to the emanating shower of particles, cooling being provided by immersion in a bath of liquid helium. As an example, the Big European Bubble Chamber (BEBC) magnet is shown in Fig. 1. An example of another approach is the dipole magnet for the Omega experiment [11] at CERN that was cooled by flowing helium through a hollow channel in the conductor, much as resistive magnets are cooled using water (and similar to what is now done for fusion magnets). However, from the early 1970s high energy physics experiments have increasingly concentrated on colliders, for which solenoid geometry is usually favored, and this has had a strong impact on the evolution of the associated detector magnets. A general-purpose detector at a colliding beam accelerator consists of three major stages: a central tracker close to the beam intersection, intermediate electromagnetic and hadronic calorimeters, and an

(2)

where L is the path length in the magnetic field [8]. So, with regard to resolution in practical measurements in particle detectors, it would appear to be more effective to enlarge the magnetic volume than the field strength. However, in a large device the particle must traverse more material and will be subject to multiple Coulomb scattering, and a more complete expression for resolution is
c1 c2 p2 ∆p = + 2 4, (3) 2 p B ·L B ·L

Fig. 1.

The BEBC coil.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

external muon detector system. Momentum measurements are carried out in the regions of the tracker and of the muon detector, where a strong magnetic field is required for high resolution. Such fields may be produced in various configurations, such as solenoids or toroids, or a combination of the two, and only in specific cases by dipoles. 2.1.1. Solenoids Solenoid fields have been widely used in many collider experiments. These feature a quasi-uniform field in the axial direction and benefit mechanically from being self-supporting. The coil of the solenoid is thus an integral part of the detector system through which energetic charged particles will pass. This means that an effort should be made to minimize the material used in the coil and the cryostat — especially if the coil is located in front of the electromagnetic calorimeters — to reduce the level of Coulomb scattering. The magnetic flux must be returned outside the solenoid coil, and in most cases this is done via an iron yoke. Momentum analysis is usually performed by measurement of particle trajectories within the solenoid, and the resolution (ignoring Coulomb scattering) is expressed as p p ∆p ∝ ∼ , p B · L2 B · R2

(4)

where R is the radius of the solenoid. Both large coil radius and high field are efficient for achieving good momentum resolution. 2.1.2. Toroids Toroidal coils provide the unique feature of a closed magnetic field without the necessity of an iron flux return yoke. Because no field exists at the collision point and along the beam line, there is no impact on the beam — which is convenient for the accelerator. Within the toroid the field is inversely proportional to the distance from the axis. The particle momentum can be derived from measurements of the deflection angle combined with the sagitta of the trajectory. The deflection (bending) power, B ·L, is given by  B · L ≈ Bi · Ri · (R · sin θ)−1 dr   Ro Ri · ln , (5) = Bi · sin θ Ri

93

where Ri is the inner and Ro the outer radius of the coil, Bi the field at Ri , and θ the angle between the particle trajectory and the beam line. Momentum resolution due to the deflection can be expressed as p p(sin θ) ∆p ∝ ≈ , p B·L Bi · Ri · ln(Ro /Ri )

(6)

showing conveniently better resolution in the forward and backward (smaller θ) directions [10]. A comfortable solution is obtained with Ro /Ri = 3–4. This is conceptually ideal, as the forward particles carry higher momentum, but there are three problems associated with using a practical toroid for a superconducting detector magnet. First, the particles have to cross the coil to get into the field. While in principle it is possible to locate a (thin) continuously wound toroidal coil around the circle, in practical designs the coil is separated into 6–12 lumped coils, which provides reasonable acceptance and accessibility. The design of the magnet is nevertheless complex: as well as the direct force tending to push the inside leg toward the axis and the return (outer) leg away from the axis, the mechanical structure needs to sustain a decentering force between two adjacent coils. The second problem is the opacity of a large fraction of the magnet due to the structure required to support these forces as well as the lumped nature of the coils. The third problem is that with Ro /Ri = 3–4 the peak field in the inside leg of the coil is 3–5 times higher than the useful magnetic field for momentum analysis. This is clearly not an efficient use of the superconductor given its sensitivity to the magnetic field, so if a toroid is to be used the ratio Ro /Ri should be small and the inner volume provided with a different magnet for momentum analysis of particles close to the vertex (as for the ATLAS experiment at the LHC; see Subsec. 2.2.1). Toroids can be interesting for providing magnetic analysis in the forward direction, where solenoids are less efficient and the polar uniformity of the field gives better coverage than a dipole. An example of this is the CLAS magnet [12] at CEBAF (JLab), which is shown in Fig. 2. 2.1.3. Dipoles Dipole fields for spectrometers are normally provided in a gap between two iron pole pieces linked to an iron flux return yoke. As in the case of the toroid,

December 29, 2012

15:27

94

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

Fig. 2. The CLAS magnet consists of six flat, kidney-shaped coils arranged as a toroid for analysis in the forward direction.

the electromagnetic forces are not fully supported by the coil itself and require additional structure. In large-scale magnets, the coil design becomes complex. Although superconducting dipole spectrometer magnets have been proposed, resistive magnets are often more practical for collider applications when realistic assumptions are made regarding the duty factor, overhead for cryogenics and integrated power consumption. 2.1.4. Open axial field magnets A variant of the solenoid, the open axial field magnet provides an axial field with completely open access from the sides and field-shaping poles. Besides eliminating the coil from the trajectories of particles over a wide range of transverse angles, this geometry is very favorable for the installation and maintenance of detector equipment, as evidenced in a resistive version for an experiment at the CERN ISR [10]. A major drawback is that the attainable field level is relatively low, and while an imposing (superconducting) evolution of the design was proposed for the GEM experiment at the SSC [13], regular thin solenoids are now generally preferred [14]. 2.2. Practical magnets for collider detectors Experimental particle physics at the highest energies is now performed at colliding beam accelerators. Most superconducting magnets for such collider

detectors are solenoids. The evolution of the design of this type of magnet has been reported by several authors [15–17] and, as seen in Table 1 [18–39], a large number of such magnets have been successfully designed and constructed. In this table X refers to the thickness of the coil in radiation lengths (X0 ), which is a convenient way to express transparency [40]. The significance of E/M is explained in Subsec. 2.3.2. The length of the solenoids is generally about twice the diameter of the bore. The technology of this type of magnet has converged on the use of indirectly cooled coils made from aluminumstabilized superconductor. The twin constraints of coil transparency and cost drove the designs in this direction from the outset, and led to the successful development of the coextrusion process of Nb–Ti (in its copper matrix) with pure aluminum. Besides being far better that copper with regard to transparency, very pure aluminum performs better as a stabilizing agent than copper, thanks to its lower resistivity at low temperature. Magnetoresistance is also less of a problem. However, the mechanical properties of pure aluminum are poor, and with the relentless striving for higher fields (and thinner coils) recent development work has concentrated on overcoming this drawback. The two superconducting solenoids installed at the Large Hadron Collider (LHC) at CERN, for the CMS and ATLAS detectors, both rely on reinforcement of the aluminumstabilized conductor. It is also standard practice to wind the solenoids on the inside of a thin mandrel that provides longitudinal support. The mandrel is not subject to buckling forces during winding, and in operation hoop stress is shared with the conductor. While these magnets must be designed to be sufficiently robust to survive quenching (i.e. transition to the normal resistive state), they are also designed with a large margin, the conductor operating at typically less than 50% (along the load line) of its short sample current at nominal field. Care is also taken to avoid crack-prone pockets of unfilled epoxy in the winding that could release sufficient energy to provoke a quench, so this should be rare. Passive protection is ensured by spreading the quench quickly to the whole winding using pure Al strips [30] and/or inductive heating of the mandrel (quench-back) [21], often aided by active protection with energy extraction into an external resistor. The magnets are very reliable.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices Table 1.

95

Summary listing of superconducting solenoids constructed for high energy physics experiments.

Experiment

Laboratory

PLUTO ISR point 1 CELLO PEP4/TPC CDF TOPAZ VENUS AMY CLEO-II ALEPH DELPHI ZEUS H1 BESS WASA BABAR D0 BELLE ATLAS-CS BESS-polar CMS BESIII CMD-3

DESY CERN Saclay/DESY LBL/SLAC KEK/FNAL KEK KEK KEK Cornell Saclay/CERN RAL/CERN INFN/DESY RAL/DESY KEK KEK/Uppsala INFN/SLAC FNAL KEK KEK/CERN KEK CMS/CERN IHEP (China) BINP

R (m)

B (T)

I (kA)

X (X0 )

E/M (kJ/kg)

E (MJ)

Year

Ref.

0.75 0.85 0.85 1.1 1.5 1.45 1.75 1.2 1.55 2.75 2.8 1.5 2.8 0.5 0.25 1.5 0.6 1.8 1.25 0.45 3.0 1.45 0.35

2.2 1.5 1.5 1.5 1.6 1.2 0.75 3 1.5 1.5 1.2 1.8 1.2 1.2 1.3 1.5 2.0 1.5 2.0 1.0 4.0 1.0 1.5

1.3 2 3 2.27 5 3.65 4 5 3.3 5 5 5 5 0.38 0.9 6.83 4.85 4.16 7.8 0.48 19.5 5 1

4.0 1.1 0.6 0.83 0.84 0.70 0.52 N/A 2.5 2.0 1.7 0.9 1.8 0.2 0.18 0.5 0.9 N/A 0.66 0.156 N/A N/A 0.085

2.3 1.8 5.0 7.6 5.4 4.3 2.8 N/A 3.7 5.5 4.2 5.2 4.8 6.6 6 N/A 3.7 5.3 7.1 9.2 12 2.6 8.2

4.1 3.0 7.0 11 30 19 11.7 40 25 136 110 10.5 120 0.25 0.12 27 5.6 37 38 0.34 2600 9.5 0.31

1972 1977 1978 1983 1984 1984 1985 1985 1988 1987 1988 1988 1990 1990 1996 1997 1998 1998 2001 2005 2007 2008 2009

18 19 20 21 22 23 24 25 26 27 28 29 28 30 31 32 33 34 35 36 37 38 39

2.2.1. SC detector magnets at the LHC It is appropriate to look in more detail into the superconducting detector magnets at the LHC, because these embody the science and experience derived from all the previous work done, and as such represent the present state of the art [41]. CMS The CMS detector is designed around a single solenoid magnet surrounded by a return yoke acting also as the external muon spectrometer: an end view of the coil is shown in Fig. 3. The solenoid is designed to provide a field of up to 4 T in a warm bore 6 m in diameter and a magnetic length of 12.5 m [37]. The maximum stored energy is 2.65 GJ at its test induction of 4 T, making it the single superconducting magnet with the most stored energy in the world. The cold mass is 225 t. At the nominal field of 4 T the current rating is 19.5 kA. The operating field is 3.8 T. The design [42] features a self-supporting coil, made from hybrid reinforced conductor, and an assembly of five transportable modules of four-layer coils.

Fig. 3.

End view of the CMS coil.

The aluminum-stabilized superconductor [43] is reinforced with high-strength aluminum alloy flanges attached to the soft aluminum by electron beam welding, as shown in Fig. 4. A major part of the

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

96

Fig. 4. The CMS conductor. The coextruded Rutherford cable consists of 32 strands of Nb–Ti/Cu, diameter 1.26 mm.

hoop stress is constrained in the alloy flanges, and due to this innovative technique the coil could be designed to produce a central field of 4 T on such a scale. The coil was wound using the inner winding technique mentioned above and further discussed in Subsec. 2.3.3; it is cooled indirectly by thermosiphoninduced flow of two-phase helium, in cooling tubes welded to the support cylinder. The flux return yoke consists of a 10,000 t barrel structure made up of flat steel plates between which are installed chambers for detecting traversing muons. ATLAS The ATLAS detector is optimized by combining an axial magnetic field in the central detector region with a surrounding azimuthal magnetic field to provide muon spectroscopy [44]. This field configuration

Fig. 5.

requires a sophisticated magnet system consisting of a solenoid coil to provide the axial magnetic field, and toroidal coils to generate the azimuthal magnetic field. The ATLAS magnet system thus consists of four major components: a central solenoid, a barrel toroid and a pair of end-cap toroids. A general view of the system is shown in Fig. 5. A particular challenge of the ATLAS magnet system is its size and the combined configuration of the solenoid and toroidal coil systems, to accommodate the physics requirements of a light and open structure that keeps muon scattering to a minimum. The central solenoid magnet provides an axial magnetic field of 2 T in a 2.3-m-diameter warm bore in the central tracker region, providing a deflection power of 4.6 T·m [35]. The coil is located in front of the liquid argon calorimeter, so it must be both thin and as transparent as possible to ensure good calorimeter performance, with minimum interaction of particles. The design features: (i) a high-strength aluminum-stabilized conductor, (ii) a pure aluminum strip to induce uniform energy absorption in the thin coil and (iii) a common cryostat, together with that of the liquid argon calorimeter, to minimize wall material. An extensive effort was put into achieving homogeneous reinforcement of the aluminum stabilizer while keeping sufficiently low electrical resistivity, as

Schematic view of the ATLAS magnet system.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

97

The two other major experiments at the LHC, ALICE and LHCb, feature new large dipole magnets. It was initially envisaged to equip these experiments with superconducting magnets, but after analysis of initial cost, risk and duty factors, it was decided to adopt resistive technology. Fig. 6.

Cross-section of the BT conductor.

2.2.2. Detector magnets for astrophysics

characterized by its residual resistivity ratio (RRR). An optimum solution has been found that uses a combined process of microalloying and cold-work hardening [45]. This is discussed in Subsec. 2.3.1. As a result of the development, the radiation thickness is 0.66 X0 . The ATLAS barrel toroid (BT) is optimized with eight coils assembled symmetrically around the beam axis. It has an inner free bore of 9.5 m, an outer diameter of 20 m and a length of 26 m [44]. It provides a deflection power (BL) of 2–6 T·m, increasing with the forward angle. A cross-section of the BT conductor is shown in Fig. 6. Muon chambers are installed between the individual cryostats housing the coils to give a combined measurement of sagitta and deflection, and thus achieve good momentum resolution. The ATLAS end-cap toroids (ECT) also feature eight coils, with the azimuthal position rotated by 22.5◦ with respect to the BT to provide a radial overlap, optimizing the bending power in the interface region of the coil systems [46]. In contrast with the BT, the eight coils are assembled as a single cold mass. While this simplifies the mechanical support of the magnetic forces, it does not allow installation of muon chambers within the toroidal field. The deflection measurement is made with a deflection power (BL) of 4–6 T·m at each end. The three toroids are powered in series. Table 2 shows the parameters of the ATLAS system.

Table 2.

An example of a superconducting magnet for astrophysics research is that provided for the BESS-polar experiment. This experiment was made to study cosmic ray antiparticles as observed from a balloon flying over Antarctica [36]. To analyze the momentum of particles, a uniform field is provided in the detector system by means of an extremely thin superconducting solenoid. The coil is wound with highstrength aluminum-stabilized superconductor using microalloying, discussed in Subsec. 2.3.1. The conductor is sufficiently strong to support the electromagnetic forces when the coil is tested at 1 T. In operation the central field is 0.8 T, with uniformity ∆B/B ≤ ±9% in a warm bore tracking volume 0.75 m in diameter and 1 m in length. The maximum field in the coil is 1.1 T at a current of 358 A; the magnet must operate in ballooning conditions with acceleration loads of up to 10 g. The coil thickness is 3.4 mm, giving a radiation thickness of 0.056 X0 . The coil (shown in Fig. 7) is assembled into a cryostat with a warm bore and is indirectly cooled by thermal conduction via aluminum shells/strips connected to a liquid helium reservoir at one end. Including cryostat material, the surface density of the wall is 2.3 g/cm2 ,

General parameters of the ATLAS system.

Inner diameter (m) Outer diameter (m) Length (m) Mass (t) Operating current (kA) Stored energy (MJ) Peak field on winding (T)

CS

BT

ECT

2.44 2.63 5.3 5.7 7.6 38 2.6

9.4 20.1 25.3 830 20 1080 3.9

1.65 10.7 5 239 20 206 4.1

Fig. 7.

The BESS-polar coil.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

98

operation of the cryogenics in the space station it was decided to install a permanent magnet system. 2.3. Specific technology for detector magnets 2.3.1. Aluminum-stabilized superconductor

Fig. 8.

The AMS magnet system.

corresponding to a radiation thickness of 0.1 X0 . The magnet performed according to expectations. The completely open 0.87 T dipole field of the superconducting magnet for the Alpha Magnetic Spectrometer (AMS) [47] is provided by Helmholtz coils with active flux return via sets of racetrack coils, as shown in Fig. 8. The coils are connected in series and all feature the same aluminum-stabilized conductor. Being destined for installation in the space station, the stray field had to be low and the magnetic moment canceled. The warm bore is completely open, but not being self-supporting the system requires additional structure. It is evidently more complicated and its mass is much greater than that of an equivalent thin solenoid. The magnet was constructed, but due to concerns over the long-term

Fig. 9.

Aluminum stabilization of the superconductor is a key technology in modern detector magnets. It contributes to the stability of the superconductor with minimum material and weight. The progress of aluminum-stabilized superconductors for detector magnets is illustrated in Fig. 9. The NbTi/Cu superconductor strand/cable is coextruded with aluminum stabilizer: this process provides reliable diffusion bonding [43, 45]. Major design parameters of some recently developed aluminum-stabilized superconductors are given in Table 3. Over the last two decades, aluminum-stabilized superconductor has undergone a number of improvements with regard to mechanical strength. One approach was to provide reinforcement of the stabilizer itself; another was to work with a hybrid of soft high-conductivity material with strong alloy. Homogeneous reinforcement has been achieved by combining microalloying and cold-work hardening [45]. Microalloying provides mechanical strength and stability against annealing during the process of coil curing; cold-work hardening with area reduction also contributes to the reinforcement. Metals such as Mg, Si, Cu and Zn have the common feature of contributing little to electrical resistivity under the condition of so-called “solid–solution” [48]. On

Evolution of conductors for detector magnets.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices Table 3. Type

Relevant parameters of high-strength conductors. Composition

Yield strength (MPa) Al

Full conductor

RRR

ATLAS

Ni(0.5%)Al

110

146

590

CMS

Pure Al & A6082-T6

26 428

258

1400

Next

Ni–Al & A6082-T6

110 428

∼300

∼ 300

99

In contrast, the small-sized aluminum-stabilized superconductor was developed for thin solenoids for astrophysics detectors, due to the need for lightweight devices having low operating current with persistent mode operation in an environment of limited power and cryogenic resources. In a further development of the hybrid approach it is envisaged to replace the pure aluminum by high-strength, high RRR material. This is to ensure that the conductor does not migrate under electromagnetic forces when the magnet operates at more than 4 T (being considered for future detectors). It thus combines the two methods, coextruding the conductor with the microalloyed material followed by e.-b. welding of the tough alloy flanges [50]. 2.3.2. E/M ratio and transparency

Fig. 10. Plot showing the progress made in the development of high-strength aluminum having good electrical conductivity.

the other hand, an additive metal such as Ni has a uniquely lower solubility: beyond a low threshold, it is crystallized and precipitated. The composite with aluminum, Al3 Ni, is uniformly distributed in the form of filaments in the pure aluminum base metal [49], contributing to mechanical strength while allowing the aluminum to keep reasonably low electrical resistivity. Figure 10 shows the progress of the mechanical strength of aluminum stabilizer as a function of electrical resistance at 4.2 K, compared with typical oxygen free copper (OFC) as stabilizer. It is seen that the strength of the aluminum stabilizer is comparable with that of copper, while maintaining its high RRR and the all-important advantage of lightness. For the CMS solenoid project, a different technical approach for the reinforcement was to use the hybrid configuration, shown in Fig. 4. This configuration is very effective for large conductors where it can be used because it is possible to weld without overheating the superconductor. It allows a hoop strain of 0.15% induced by a hoop stress of 105 MPa, which is required for the 4 T CMS solenoid.

Compactness and transparency of the magnet are important in order to create a magnetic field with minimum disturbance for the particles and having maximum detector acceptance. For these reasons, the ratio of stored energy to effective coil cold mass, called the E/M ratio, is a useful parameter for scaling the lightness, and compactness (or efficiency) of the magnet [45]. The E/M ratio in the coil is approximately equivalent to its enthalpy, H, and it determines the average temperature rise of the coil after a quench, as follows: E (7) = H(T2 ) − H(T1 ) ≈ H(T2 ), M where T1 is the initial temperature and T2 is the average coil temperature after full energy absorption in a quench. E/M ratios of 5, 10 and 20 kJ/kg correspond to ∼65-, ∼80- and ∼100 K. In the case of a solenoid, E/M can be expressed as the ratio of stress, σh , to equivalent density, d, of the coil, and is given by  B2 E σh 2µ0 dv = . (8) ≈ M d · Vcoil 2d The necessary coil thickness is determined by t=

B2R . 2µ0 σh

(9)

The E/M ratios in various detector solenoid magnets are shown in Table 1 and Fig. 11. For early generations of thin magnets, E/M ≈ 5 kJ/kg. Based on the development of high-strength aluminum stabilizer, ∼10 kJ/kg was achieved for the SDC prototype. Using similar high-strength

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

100

cooling (see Subsec. 2.3.4). The technique was first applied to the winding of the TOPAZ coils [23] and has since become standard practice. 2.3.4. Indirect cooling

Fig. 11.

Plot showing the progression to greater E/M .

aluminum stabilizer, the ATLAS central solenoid reached 8.1 kJ/kg at its test field of 2.1 T. The CMS solenoid achieved an E/M ratio of 12 kJ/kg at its nominal field of 4.0 T. It was not required to be a thin solenoid but there was a strong incentive to moderate the mass of the coil for reasons of physical size as well as cost. During testing, a prototype magnet for the BESS-polar program achieved E/M ≈ 13 kJ/kg without damage [51]. The CMD magnet for a detector at the VEPP accelerator at BINP incorporates a novel protection scheme where groups of turns are shunted with a brass bridge, thereby avoiding the situation where all the energy is deposited in a small quenching section of the coil [39]. By using this technique it was possible to achieve a very transparent magnet with E/M = 9. 2.3.3. Inner winding technique The traditional way of winding a solenoid is to apply the conductor under tension to the outside of a mandrel or bobbin. The tension should be such as to ensure sufficient compressive prestress between winding and bobbin that the glue between the components is not put under undue tension when the hoop stress increases in the conductors due to the electromagnetic forces. This requires the bobbin to be thick enough to avoid buckling. Conversely, if the winding can be done inside the mandrel the compressive force between the components increases when the current is increased, and there is no bucking force during the winding. In addition, having the ground insulation between coil and mandrel under pressure ensures the good thermal conduction required for indirect

The first superconducting detector magnets were cooled by immersion in a bath of liquid helium. This method of cooling provides a stable temperature and is efficient for evacuating heat produced in the winding by eddy currents and other losses, particularly those associated with changing the current setting. For large magnets, however, the helium containment vessel is complicated and the quantity of helium required to fill the vessel can be very large. But detector magnets are run at constant current, and it was soon realized that it should be sufficient to cool the windings indirectly via flowing helium in a system of pipes. This leads to a much simpler design and is now universally accepted as the preferred method of cooling. The pipes must be sized to allow the cold mass to be cooled down, and insulated electrically from the winding to avoid having to use insulating breaks. In the case of solenoids the pipes are usually welded to the external mandrel, which provides a large area heat sink that is bonded with epoxy resin to the coil via ground insulation composed of glass-fiber-reinforced epoxy (GFRP) or tape combining polyimide film with GFRP. The coils of the toroids are not self-supporting: they are potted in rigid aluminum box structures, and the pipes are welded (or glued) to the boxes. Depending on the layout, the cooling is achieved by pumping either supercritical or two-phase helium through the pipes, or (in the case of solenoids) by the thermosiphon technique. Examples are shown in Fig. 12; for more details see Ref. 14. 2.4. Future plans for detector magnets The International Linear Collider (ILC) is proposed as a future major facility for high energy physics research. Important R&D is being carried out on both the accelerator and the two detectors, and the Technical Design Report will be issued at the end of 2012. The detectors feature large high-field solenoids having parameters, as summarized in Table 4. These magnets will require conductors which combine the technologies used for the ATLAS CS and CMS solenoids at the LHC [50].

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

Fig. 12.

Examples of winding and cooling arrangements.

Table 4. Tentative parameters of solenoid detector magnets for the ILC compared with those used at CERN’s LHC. LHC

ILC

A-CS

CMS

ILD

SiD

Requirement: Clear bore rad. (m) Central field (T)

1.18 2

2.95 4

3.44 4.0 (design) 3.5 (nominal)

2.6 5

Design parameters: Coil inner rad. (m) Half-length (m) No. of coil layers Full thickness (m) Max. field (coil) (T) Nom. current (kA) Stored energy (GJ) Cold mass (t) E/M (kJ/kg)

1.23 2.7 1 0.045 2.6 7.73 0.04 5.7 7

3.25 6.25 4 ∼0.3 4.6 20 2.6 220 12.3

3.61 3.67 4 ∼0.3 4.3 21.7 2.2 183 11.9

2.5 2.5 6 ∼0.4 ∼5.8 20 ∼1.7 ∼80 ∼12

3. Magnets for Fusion Devices Thoughts toward the harnessing of fusion power were first expressed in the 1920s when it was found that the mass of four hydrogen atoms was greater than that of one helium atom. This explained the origin of the power of the sun, and toward the end of the 1930s physicists in the US started to study how to confine a hot plasma using magnetic fields. Following WW2 there was understandably much interest in controlling nuclear fusion, but the work continued secretly until the 1958 Atoms for Peace conference in Geneva [52]. At this conference, scientists from the US, the UK and the USSR revealed some of the

101

results of their fusion programs, and it was realized that extracting power from the fusion process should be possible — but that learning how to control the plasma would be a great challenge. Conferences were organized to enable the exchange of information on progress in solving the scientific and technological problems, and in Europe a regional collaboration was set up, headed by EURATOM. Due to its extreme temperature, the reacting plasma needs to be confined in a virtual container: this can be done by shaping magnetic fields [53] to create a “magnetic bottle.” The simplest form of this is produced by two circular coils, and was the basis of the first experiments, but the bottle is not perfect, and for long pulses plasma can escape from the ends. There are two solutions to this topological problem. The first is to create a three-dimensional quadrupole field where the coil geometry resembles that of a baseball seam, and the second is to bend the solenoid into a torus so that escaping plasma from one end goes into the other end [54]. The torus is conceptually pleasing but there is a problem: the field in a standard torus is inversely proportional to the radius, so that plasma can leak out from the side. The first approach to addressing this problem was proposed in the early 1950s by Lyman Spitzer at Princeton Laboratory of Plasma Physics, who recognized that the toroidal geometry would incur rapid loss of particles by curvature drifts and by E×B losses resulting from charge separation, but that by twisting the torus to form a figure of eight in which the effects of particle drifts in the two resulting U-bends are canceled. In addition to such physical twisting, the required field pattern can also be created by combining different sets of coils (toroids, helices, etc.) or by shaping the coils, and machines based on this concept are called stellarators. Another approach, developed at the Kurchatov Institute of Atomic Energy in Moscow in the 1960s, was to use a regular toroidal magnet, but to create the rotational transform [54] by inducing a strong current in the plasma itself. In such devices, called tokamaks, the current is induced in the plasma by ramping the field in a strong solenoid at the center of the torus. This implies that the tokamak is essentially a pulsed device, whereas the stellarator can be run for much longer periods (limited by field quality) [55]. Spectacular results from early tokamaks have led to enduring popularity of this device.

December 29, 2012

15:27

102

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

Between the mid-1950s and the 1980s, much progress was made in fusion research, and plasma confinement had improved to the point where the Lawson criterion [56] for break-even conditions was close to being met. However, it was also clear that for such a fusion reactor to be viable (defined as being capable of providing more than 1 GW of continuous power) it would have to be large — and equipped with a very large high-field magnet system. Such magnets would be expensive to build and run, so the concurrent advent of practical superconductivity gave rise to many proposals for fusion devices, at the same time as those for spectrometers for particle physics. However, initial experiments using ad hoc apparatus in the USSR and the US revealed major problems other than that of providing a magnetic field (some devices were aptly named “perhapsatrons”). The magnets are more complicated than those used for accelerator-based physics experiments and the exploratory nature of plasma research required that the field geometry be adaptable. Both resistive and superconducting magnets were used, but while the recent generation of machines is superconducting, experimental plasma physics has to this day relied more on machines featuring resistive magnets (T-10 in the USSR, JT-60 in Japan, DIII-D in the US, and JET in Europe). The first superconducting devices were built using primitive conductor and as much magnet experiments as apparatus for plasma physics. Nevertheless, in the 1980s a very large mirror device and several tokamaks were built featuring large superconducting coils. From the mid-1970s there was a growing consensus that (1) a very large device would be needed to produce reasonable quantities of energy, (2) there were many technological problems that needed to be addressed, which would take time and (3) beyond a certain level the experiments required to perform the tests would be larger than any one country would be prepared to finance via its research budget. So while individual countries continued their investigations using midsize national machines, it was decided to create an international collaboration for developing the larger components. The first step in this direction was the Large Coil Task (LCT), completed in 1988, to be followed by the International Tokamak Experimental Reactor (ITER), presently being constructed, and eventually DEMO, bridging the way to commercial reactors. These experimental reactors, as

well as the present generation of national facilities that are being built, feature sophisticated superconducting magnet systems. 3.1. Tokamaks The principle of the magnet system for a tokamak is illustrated in Fig. 13. In this figure the coils of the toroid are circular, which is the case for some tokamaks (including Tore Supra; see below), but the natural cross-section of the plasma is D-shaped and most machines feature the familiar D-shaped toroidal coils. Such a coil can be conveniently shaped to feature constant tension around the curved part of the D, which is usually exploited in the design [57]. When the toroidal field (TF) coils are assembled, the straight parts of the windings lie on a cylinder, and the inward forces that appear when the coils are excited can be conveniently reacted by wedging. Within the cylinder is installed the central solenoid (CS), which is pulsed to induce a current in the plasma. The CS is divided into sections that are powered separately to control the plasma. Control is also provided by separately powering the poloidal field (PF) coils, which are also pulsed. It has been found that the plasma is very sensitive to the accuracy of the field, leading to a requirement for tight tolerances on conductor placement, and also to the need for numerous correction coils, in addition to the segmentation of the main coils, to compensate for unavoidable errors. While the toroidal magnet is constructed from about 20 planar modules, and the toroidal field itself only produces radial forces on the coils, the other fields give rise to strong out-of-plane

Fig. 13.

The principle of the tokamak.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

forces that must be taken into account in the design of both the coil casing and the supporting structure. During the learning process, the correction of tokamak fields involved “tweaking” the magnet system, so it not surprising that accessible resistive coils were favored in the past. However, a lot of experience has been gained and recent large fusion machines are superconducting with an eye to developing in parallel such systems, which will in any case be necessary for building reactors of a scale sufficient to produce power economically.

3.1.1. The large coil task A multinational program of cooperative R&D on superconducting magnets for fusion was initiated in 1977. The goal of the Large Coil Task (LCT) was to prove the design principles and fabrication techniques for SC coils being considered for toroidal

Table 5. Coil

magnets of tokamaks. Under the agreement the US was to supply the test facility at Oak Ridge National Laboratory (ONRL) and three of the six coils, the other three being supplied by the EU, Japan and Switzerland. The coils were D-shaped, measuring about 2.5 m × 3.5 m. The last coil was completed in 1985 and the toroidal array underwent extensive testing in 1986–1987. All six coils exceeded the design performance, reaching peak fields of up to 9 T. While the coils were specified to ensure compatibility, the design of the conductor, winding and structure was left to the design teams. The main features of these coils are summarized in Table 5. The conductors chosen for the coils are shown in Fig. 14. The variety of approaches is striking, as is the fact that they all produced satisfactory results. The Westinghouse coil featured a cable-in-conduit conductor (CICC) [58], marking the entry of this technology into the magnet designer’s kit.

The main features of the six coils for the Large Coil Task (LCT).

US 1

US 2

US 3

Cooling Conductor Supplier Ides (kA) A/mm2 Winding Style

Pool boiling Nb–Ti IGC/GD 10.2 27.4 Edge 14 layers

Pool boiling Nb–Ti IGC 10.5 24.7 Flat 6 double pancakes

Forced flow Nb3 Sn Airco 17.76 20.1 Grooves 24 plates

Case

SS Weld General Dynamics

SS Bolt + weld General Electric

Al Bolt Westinghouse

Manufacturer

103

Fig. 14.

EU Forced flow Nb–Ti Vacuum-schmelze 11.4 24.1 Flat 7 double pancakes Impregnated SS Bolt + seal Siemens

Conductors used for the LCT, shown to scale.

JA Pool boiling Nb–Ti Hitachi 10.22 26.6 Edge 20 double pancakes SS Bolt + weld Hitachi

CH Forced flow Nb–Ti SASM/BBC 13.0 30.7 (square) 11 double pancakes Impregnated SS Bolt BBC

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

104

A very complete report was written at the conclusion of the LCT [59], and it is evident that both the technical findings and the experience with the international collaboration have provided useful input for the much larger task of ITER. This international program was nevertheless complemented by national tokamak initiatives to push ahead with fusion science research (as opposed to technical development of large magnet components). Some of these machines (in the US, Japan and the UK) were based on the use of resistive magnets (TFTR, JT60 and JET), but in parallel with this activity the Soviet Union and France decided to build national tokamaks with superconducting coils (T-7, T-15 and Tore Supra). The next generation of superconducting tokamaks has already appeared in the form of EAST in China, K-STAR in South Korea and SST-1 in India, and is to be joined in the near future by an upgrade of JT-60 to a superconducting version (JT-60SA) with large in-kind contributions from Europe via an initiative entitled “The Broader Approach.” These tokamaks are being configured to complement ITER by providing an opportunity for scientists to better prepare for experiments on the large machine when it is finished in the early 2020s. 3.2. Stellarators Stellarators are toroidal devices for magnetic confinement that do not rely on the magnetic field produced by current induced into the plasma. This is done by creating surfaces of magnetic flux formed by the field as it wraps around the plasma on the torus by using

Fig. 15.

currents flowing in a combination of coils. Whereas the first stellarator designs consisted of combining a regular toroid, having planar coils with helical and poloidal windings, present day devices are based on either helical windings supplemented by poloidal coils (often referred to as heliotrons [60]) or specially shaped nonplanar windings (referred to as torsatrons [61]). Thanks to modern three-dimensional field computation techniques, it is possible to create coil shapes that provide both the necessary field shaping with well-supported conductor, and the possibility of accessing the vacuum chamber. Figure 15 shows schematically the arrangements used for the Large Helical Device (LHD) and the W7-X designs, which are to be compared with the tokamak design in Fig. 13. Stellarator magnets are more complicated than those of a tokamak, making it harder to produce devices having good closed magnetic surfaces and similar levels of confinement, but their potential for achieving long burns is alluring [62]. 3.3. Magnetic mirror devices Up to the mid-1980s, the effort was commendably broad-based, in that magnetic confinement using both mirror and toroidal devices, many of which were superconducting, was pursued vigorously. In the US in particular, magnetic mirror devices received a lot of attention, through the construction of the massive superconducting yin–yang coils, and culminating in the assembly and testing of the Mirror Fusion Test Facility (MFTF-B) [63] in 1986. This facility, installed at Lawrence Livermore National Laboratory (LLNL), used the yin–yang coils to close

Examples of coils for stellarators.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

the ends of a linear device built up from a set of superconducting solenoid coils. Shortly after its completion, the facility was closed due to financial constraints, and the program was recentered on the use of toroidal magnetic fusion facilities. While the major effort worldwide is presently focused on large tokamaks and (to a lesser extent) stellarators, there is still ongoing activity on more compact variants of the devices, e.g. MAST (Mega-Ampere Spherical Tokamak [64]), which do not require superconducting technology. 3.4. Listing of superconducting magnets for fusion devices

Baseball II and MFTF-B The interesting attribute of a current flowing along a line in the form of the baseball seam is that progressing from its center in any direction the magnetic field

Device Baseball II IMP SUMM LIN-5B T-7 TESPE MFTF-B LCT (6) T-15 TRIAM-1M Tore Supra DPC-EX, U POLO LHD, HC LHD, PC CSMC (ITER) TFMC (ITER) EAST, TF EAST, PF KSTAR, TF KSTAR, PF SST-1 W7-X JT-60SA, TF JT-60SA, CS ITER, TF ITER, CS ITER, PF

increases, making it a convenient bottle for plasma. The Baseball II magnet, shown in Fig. 16, was the first of a series of similarly shaped superconducting magnets for fusion experiments at LLNL, which culminated, via the Mirror Test Facility (MTF) and Tandem Mirror Experiment (TMX) programs, in the massive MFTF-B facility. As seen in Fig. 17, the last variants constitute a linear, solenoid-wrapped section, between the yin–yang mirror coil pairs. The yin–yang magnets shown in Fig. 18 were wound using stabilized Nb–Ti/Cu of square cross-section; the high-field solenoid units were wound using prereacted Nb3 Sn tape soldered to copper stabilizer. IMP

The salient features of superconducting magnets for fusion devices of the past, the present and the near future are listed in Table 6 [65–90]. The information in the list is complemented in the text that follows.

Table 6.

105

The classical mirror arrangement is shown in Fig. 19. The quadrupole coils were made using Nb3 Sn tape consisting of five prereacted ribbons on stainless steel substrate interleaved with copper and soldered. The conductor was cowound with high-purity Al ribbon to damp sudden changes due to flux jumps. The solenoid coils were wound using 15 untwisted Nb–Ti filaments embedded in copper.

Salient features of superconducting fusion magnets.

Location

SC type

Cooling

Bpk (T)

Iop (kA)

E (MJ)

Year

Ref.

LLNL, US ORNL, US Lewis/NASA Kurchatov, USSR Kurchatov, USSR FzK, D LLNL, US ORNL, US Kurchatov, USSR AFRC, JP CEA, Cadarache JAERI, JP FzK, D NIFS, JP NIFS, JP JAERI, JP FzK, D ASIPP, CN ASIPP, CN KBSI, KR KBSI, KR IPR, India IPP, Greifswald, D JAEA, JP JAEA, JP IO, Cadarache, F IO, Cadarache, F IO, Cadarache, F

Nb–Ti Nb–Ti/Nb3 Sn Nb–Ti/Nb3 Sn Nb–Ti Nb–Ti Nb–Ti Nb–Ti/Nb3 Sn Nb–Ti/Nb3 Sn Nb3 Sn Nb3 Sn Nb–Ti Nb–Ti/Nb3 Sn Nb–Ti Nb–Ti Nb–Ti Nb3 Sn Nb3 Sn Nb–Ti Nb–Ti Nb3 Sn Nb–Ti/Nb3 Sn Nb–Ti Nb–Ti Nb–Ti Nb3 Sn Nb3 Sn Nb3 Sn Nb–Ti

Bath Bath Bath Bath Forced flow Bath Bath Bath/Forced flow Forced flow Bath Bath, 1.8 K Forced flow Forced flow Bath, 3.8 K Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow Forced flow

7.5 8.5 10.3 8 5 7 12.7 8 9.3 11 9 6.7 1.6 6.9 6.5 13 9.5 6 4.5 12.3 8/9.7 5 6.7 6.4 9 10 13 7

2.4 0.4, 0.8 0.4 2.8 6 7 5.9 10–18 5.6 6.2 1.4 17 15 13 30 46 80 14.5 14.5 35 22 10 18 25 20 64 45 55

17 2.4 18 6.5 20 8.8 1000 750 790 76 600 10 1.9 920 845 640 80 320 60 470 200 74 920 1500 300 41,000 6400 3000

1970 1971 1971 1973 1975 1983 1987 1987 1987 1987 1988 1990 1992 1999 1999 2000 2002 2006 2006 2008 2008 2012 2013 2016 2016 2020 2020 2020

65 66 67 68 69 70 63 59 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

A. Yamamoto & T. Taylor

106

Fig. 16.

1230004

Fig. 19.

The baseball magnet (photo courtesy of LLNL).

The IMP classical mirror system.

flexibility in shaping the field. The inner windings were Nb3 Sn ribbon on stainless steel and stabilized with copper, and all the other coils used Nb–Ti with 15 filaments embedded in square section copper, similar to the IMP conductors. It was the largest facility of its type at the time. One coil had to be rebuilt; all coils trained. LIN-5B

Fig. 17.

MFTF-B mirror facility layout.

This was a mirror device having a baseball-seamshaped winding. The conductor was copper-stabilized Nb–Ti with cross-section 6.2 mm × 6.2 mm and Cu:Sc = 4:1. There was significant training. T-7 This was the first tokamak with a superconducting toroidal field coil. It was also the first fusion magnet to use forced flow cooling. The 48-coil device was used in parallel with T-10, a normal-conducting tokamak with similar parameters, until the late 1980s, when it was shipped to China. It has been upgraded to become EAST at ASIPP, Hefei. TESPE

Fig. 18.

The yin–yang magnet (photo courtesy of LLNL).

SUMM The facility consisted of four solenoidal coils, each having an inner, middle and outer winding, all of which could be powered separately for maximum

The TESPE facility at KfK was built specifically to study the fabrication and operation of superconducting toroidal field magnet systems. Six D-shaped coils were mounted in a circle to represent a typical tokamak. The facility was used to study various technical and safety issues associated with large superconducting magnets. LCT The coils were subjected to a comprehensive barrage of tests both individually and connected in a

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

107

coils. Nb3 Sn conductor was chosen over Nb–Ti, because of its larger temperature margin. Tore Supra

Fig. 20.

The LCT test setup at ORNL.

torus at ORNL (see Fig. 20). All the coils performed as specified, and the exercise was deemed a success. It showed that such coils can be built in different countries in a variety of ways and work when assembled together, giving some confidence that such a device can be successfully produced by an international collaboration (such as ITER). It is interesting to note that only one of the six coils, that made by Westinghouse, featured the cable-in-conduit conductor (CICC). T-15 Building on the experience gained from the construction of T-7, the Kurchatov Institute for Atomic Energy undertook the construction of a larger tokamak with superconducting TF coils. The major radius of the toroid was 2.43 m, and there were 24 coils wound from Nb3 Sn conductor. The conductor consisted of an 11-strand, flat, cored cable fastened to lateral copper cooling tubes by a thick galvanic coating of copper. The 1.5-mm-diameter strands contain 14,641 filaments, diameter 4.5 µm, in a bronze matrix and are twisted with a pitch of 25 mm. Coils were made using a react-and-wind process that was developed using trial windings. Every coil was tested before assembly into the tokamak. Some degradation occurred during winding, but the magnet was endowed with a very large margin and performed reasonably well. The forced cooling method provided sufficient stability.

The toroidal magnet of Tore Supra (major radius 2.25 m) consists of 18 circular coils of Nb–Ti conductor stabilized with copper and bath-cooled to 1.9 K with pressurized superfluid helium — the first largescale device to take advantage of the thermal conductivity of superfluid helium to stabilize the winding, as well as to take advantage of the increased temperature margin. The maximum field on the winding is 9 T. The other (pulsed) coils are resistive. The cooling works well, but there were initially some issues with insulation due a metal chip in the winding, bringing into evidence the risk of relying on liquid helium for the insulation between the spacers providing the cooling channels. The magnet has been in operation, giving good service, for over 20 years. DPC DPC-EX was a 1-m-diameter demonstration poloidal field coil made to test the viability of Nb3 Sn for this application. The test was performed using two background field coils (DPC-U1 and DPC-U2) made with Nb–Ti conductor. The test coil consisted of two flat pancakes wound from the prereacted flat CICC, as shown in Fig. 21. The conductor consisted of 153 multifilamentary Nb3 Sn strands wound around a stainless steel core and flanked by subchannels connected hydraulically at 300 mm intervals, encased in stainless steel sheet welded along both sides. The strands were chromium-plated to reduce coupling losses. To minimize bending strain during

TRIAM-1M The toroidal field of this high-field tokamak (major radius 0.8 m) is produced by 16 superconducting

Fig. 21. 0.4 mm.

DPC-EX conductor. The insulation thickness is

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

108

winding the heat treatment was performed on a 1.4m-diameter mandrel. The series-connected coils performed as expected, ramping up to 17 kA in 1 s. POLO This was a superconducting 3-m-diameter prototype of a poloidal field coil (nominal current 15 kA), which was constructed and tested according to a typical tokamak specification as concerns fast ramping (2 T/s), plasma disruption (80 T/s, 10 ms) and high voltage withstand (23 kV). The coil was wound with Nb–Ti conductor (Nb– Ti/Cu/Cu–Ni = 1/5/3), and strands of 1.25 mm diameter in the CICC formed, as shown in Fig. 22. The test was made in the TOSCA facility at FzK, and showed that the technology for making CICC Nb–Ti poloidal coils was now fully developed.

Fig. 23.

The LHD HC conductor.

LHD The magnet system of the Large Helical Device (LHD), a heliotron type of stellarator developed in Japan to confine current free steady state plasma, consists of a pair of helical coils (HCs) together with three pairs of poloidal coils (PCs), as shown in Fig. 15. The LHD is at present the largest superconducting fusion machine, with a major radius of 3.9 m, and was the first to feature a fully superconducting magnet system. The HCs were designed to provide 3 T (6.9 T at the conductor). A cross-section of the conductor is shown in Fig. 23. The complex Fig. 24.

Fig. 22.

Conductor for POLO.

Cross-section of an LHD HC coil in its casing.

helical coils were wound directly into machined stainless steel casing, as shown in Fig. 24. Constraints due to the proximity of the plasma led to the need for a relatively high average current density of 40 A/mm2 in the HCs, and although the conductor was designed to be cryostable according to the Maddock criterion [91], it was found to be necessary to restrict operation to 91% of the nominal 13 kA, at which level operation has been extremely reliable. The coils require some subcooling to achieve nominal current [92]. The PCs are designed to permit operation in slow-pulse mode. The CICC features 486 strands of uncoated Nb– Ti/Cu conductor designed to work at up to 30% of short sample current.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

109

The machine was constructed in the period 1990–1997 and has been operating extremely reliably for 15 campaigns of experiments since 1998, with over 1450 powering cycles of the superconducting magnet system and less than 1% downtime. ITER model coils (TFMC and CSMC) Models were manufactured and tested to validate designs of the ITER toroidal field (TF) and central solenoid (CS) conductors and windings, as well as to initiate the international collaboration. With an outer diameter of 3.6 m and a height of 1.8 m, the CS model coil was only slightly smaller than an ITER CS module (4.1 m/1.9 m). It consisted of an outer module made in Japan and an inner module made in the US, together with an insert coil for extensive tests on the square thick-walled CICC. The TF model consisted of a 4.6-m-long racetrack coil with radial plate support of the round, thin-walled CICC, and was wound in Italy and assembled in France by a consortium of European companies. The CS model testing was done at JAERI in Naka, Japan and the TF model was tested at FzK, Karlsruhe, Germany. The Nb3 Sn stranded CICCs were very similar to those presently being produced for ITER (see below). The model coil program has been extremely useful in providing an opportunity to develop the techniques for producing and winding the unprecedentedly large conductors, and demonstrating the viability of the associated wind-and-react process, with insulation of the reacted conductor prior to insertion into the casing.

Fig. 25.

View of the EAST tokamak.

Fig. 26.

The KSTAR tokamak.

EAST The Experimental Advanced Superconducting Tokamak (EAST) at ASIPP, Hefei, China comprises 16 D-shaped TF coils. Its major radius is 1.75 m, the minor radius being 0.4 m. The superconducting coils can create and maintain a toroidal magnetic field of up to 3.5 T. Other key superconducting components of EAST are 14 superconducting poloidal field (PF) coils, HTS current leads and superconducting buslines. A bird’s eye view of EAST is shown in Fig. 25. The machine was commissioned in 2006 and is now in regular use. KSTAR The Korea Superconducting Tokamak Advanced Research (KSTAR) device is fully superconducting and includes 16 TF coils, four pairs of CS coils, and

three pairs of PF coils (see Fig. 26). The superconducting coils were manufactured without internal joints. All TF coils and CS coils, and one pair of PF coils, are made using Nb3 Sn superconductor. The large PF coils are made from NbTi superconductor. CICC was used throughout and after reaction the coils were taped and vacuum-pressure-impregnated (VPI). In order to confirm the design and manufacturing procedures, two kinds of superconducting coils were fabricated and tested in the superconducting magnet test facility at the Korea Basic Science Institute (KBSI). The prototype TF coil was full-size. Due to the appearance of leaks during testing, the original GFRP insulation breaks were replaced with ceramic components [93]. The machine was commissioned in 2008 and is now in regular use.

December 29, 2012

15:27

110

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

SST-1

JT-60SA

The magnet system of SST-1 comprises 16 superconducting D-shaped TF coils, 9 superconducting PF coils and a pair of resistive PF coils inside the vacuum vessel. It was first assembled in 2009, but there were problems due to leaks and excessive heat load, so it was decided to make a thorough overhaul and restart the tests in 2012.

At the Japan Atomic Energy Agency (JAEA), the JT-60 tokamak is being converted to a fully superconducting device known as JT-60 Super Advanced (JT-60SA). This is being constructed in the framework of the Japan–EU “Broader Approach” projects, which complement the ITER program. In JT-60SA, the 1400-ton magnet system consists of 18 TF coils, 4 CS modules and 7 plasma equilibrium field (EF) coils. The TF coil uses Nb–Ti CICC; the conductor for the CS is Nb3 Sn CICC; the maximum magnetic field on this pulsed magnet is 10 T. The inner legs of the TF coils are wedged together to withstand the inward electromagnetic forces, and the whole toroidal structure is mounted on articulated struts that ensure that the body remains centered during cooldown (see Fig. 28). The CS is hung from the toroid. TF coils and support structure are supplied by the EU. Commissioning is planned for 2016.

W7-X In order to overcome deficiencies of previous stellarators, IPP conducted a systematic search for the optimum magnetic field. In order to achieve the helical twisting of field lines solely with external coils, the latter have to be twisted and the magnet coils and plasma have a complicated shape. The outcome is the optimized shape illustrated in Fig. 15. The quality of plasma equilibrium and confinement is expected to be on a par with that of a tokamak. The conductor is CICC (see Fig. 27) where the jacket is an aluminum alloy (AlMgSi) coextruded around a bundle of 243 Nb–Ti/Cu strands (Cu:SC = 2:7) with a void fraction of 37%. Although the nonplanar coils are difficult, the machine has five-fold symmetry, so besides being convenient for providing access to the plasma vacuum vessel, there is a corresponding reduction in the number of types of coils. The magnet project has suffered delays due to insulation and leak problems; it is expected to commission the system in 2013. For a fusion power plant, stellarators could provide a technically simpler solution than tokamaks: the goal of the Wendelstein experiments at IPP is to clarify experimentally this thesis.

Fig. 27.

The Wendelstein magnet (W7-X) conductor.

ITER There have been three generations of experimental fusion devices. Two further generations of devices are planned in order to achieve the goal of commercial power generation by fusion. The fourth-generation experimental fusion device is the ITER experiment currently being constructed at the Cadarache site in the south of France. The ITER reactor is designed

Fig. 28.

JT-60SA, showing details of supporting structure.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

to operate at a nominal thermal power of 500 MW and should take a step toward the goal of building fusion power stations. The construction of ITER began in 2008. The facility is designed to produce thermal fusion with a power amplification of 10, to test long-duration burning plasma, aiming at continuous operation, and to test technologies and components essential for a commercial fusion reactor. The ITER machine relies on a massive superconducting magnet system (see Fig. 29) that will be by far the largest of its type ever built [94, 95]. The TF coil consists of 18 D-shaped coils (9 m × 14 m, compared to 3 m× 4.2 m for K-STAR) connected in 9 circuits. The maximum field on the winding is 11.8 T, and it is foreseen to operate the magnet DC for periods of about a week. The CS consists of a vertical stack of six winding modules that is hung from the TF coil assembly. The modules are powered independently with pulsed currents of up to 45 kA, the maximum field on the windings being 13 T. The six PF coils are also attached to the TF coil assembly and powered independently with pulsed currents of up to 45 kA. The magnet system is completed by three independent sets of six correction coils, connected in nine independent circuits at currents of up to 10 kA. All the coils use CICC connected hydraulically in multiple parallel circuits and cooled by a flow of supercritical helium with an inlet temperature

111

of 4.7 K. The superconductor for the high-field coils (TF and CS) is multifilament Nb3 Sn, and is Nb–Ti for the PF and CC coils [96]. The toroidal coils feature a conductor having a round cross-section. After shaping the coil it is reacted, insulated and transferred to radial plates that are fitted into the rigid stainless steel casing, which take the out-of-plane forces and contribute to the general support of the tokamak. In this way the forces on the conductor do not pile up. A crosssection of the winding in its case at the position of the inside leg is shown in Fig. 30. The CS is wound from conductor of thick square cross-section that supports the hoop stress. The cross-sections of the Nb3 Sn conductors for the CS and TF are shown in Fig. 31. The central hole is there to avoid excessive pressure buildup in case of a quench far from an end. The ITER tokamak is very large, but it has to be assembled very precisely, typical tolerances being between 1 mm and 3 mm. Installation is therefore the subject of an important ongoing study. As an illustration, Fig. 32 shows the installation, of the TF coils around the CS. 3.5. Specific technologies The realization of large superconducting magnets for fusion devices has spawned significant R&D.

Fig. 29. Cut view of the ITER magnet system. The central solenoid, poloidal and correction coils are supported from the toroidal coil system.

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

112

Fig. 30.

View of the TF winding at the inside leg. Fig. 32.

Fig. 31.

CS and TF conductors.

The magnets need to run at high current and the windings are subjected to large intermittent forces, with implications for the choice of conductors, the method of cooling, and the design of the supporting structures. The windings need to be well-insulated between turns and to ground; to avoid arcs in case of vacuum leaks, the insulation also needs to be Paschen-tight. 3.5.1. Conductors The present consensus favors CICC, with bundled Nb–Ti superconductor for fields of up to about 7 T and Nb3 Sn for higher fields, of up to about 13 T. The strands are chromium- or nickel-plated to reduce ac losses, and copper wires are added to the bundles, to give an overall Cu/SC ratio of about 3,

Installing the TF coils.

deemed sufficient for stabilization conjunction with forced flow cooling of supercritical helium. Nb3 Sn was favored from the earliest times because of its greater temperature margin, but poor experience with degradation due to strain when winding after reaction has led to the preference for a wind–react– transfer process. Nb3 Al is more strain-tolerant, and an ITER insert wound using this material worked perfectly, but the conductor is only produced on a small scale and is expensive. Until recently bronze route Nb3 Sn conductor was prescribed for tokamaks, but RRP (Restacked Rod Process [96]) conductor developed for use in accelerator magnets has been found to work well [97], and a much larger (up to ×3) current density makes it more economical to use, with copper strands replacing some superconducting strands. Coils are designed to work at less than 50% of the short sample as measured along the load line. The reader is referred to an excellent review of conductors for fusion magnets by Bruzzone [98]. 3.5.2. Cooling Bath cooling is out of favor following insulation, problems in Tore Supra and unexpected stability issues in the LHD, and with the large wetted perimeter in a typical CICC, forced flow supercritical

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

helium looks like a good solution — especially for the pulsed CS and PF coils. However, there is a drawback: while the coils are excited in series, the cooling must be provided in parallel circuits, and so there need to be insulating breaks in the pipes. Such breaks, which must work reliably after heat cycling between room and liquid helium temperature, are not produced industrially. For coils that are not ramped it may be worth reconsidering some variant of bath cooling to reduce the number of breaks. 3.5.3. Insulation Wherever possible, vacuum pressure impregnation (VPI) with resin-filled glass fiber, the industry norm, is used [99]. Where there are high voltages, layers of polyimide film are introduced into the mix. The resin presently favored by the fusion community is cyanate ester [100], which is more radiation-resistant than conventional epoxy. Industrially produced polyimide–glass (PG) and glass– polyimide–glass (GPG) insulating tape has been found to have remarkable electrical and mechanical properties [101]. This material was developed for the accelerator and detector magnet community some time ago, and has now gained acceptance with builders of fusion magnets. 3.5.4. Structures The electromagnetic forces in large toroidal magnet systems, especially tokamaks, are formidable, both in their strength and in their variety, and present mechanical engineers with a correspondingly important challenge. The toroidal coils are subjected to both in-plane and out-of-plane forces which determine the required thickness of the coil casing to limit movements and preserve the field quality. It is now standard practice to rely on the TF as the supporting structure for the other windings with the PF coils attached to consoles and the CS hung from the top, all the gravity forces being transferred to the ground via pedestals. The solution found for the gravity support of JT-60SA is particularly noteworthy, in that it maintains the central axis of the magnet system during the cooldown without resorting to the use of sliding or bending plates, as shown in Fig. 26. This approach is analogous to that applied to the horizontal support of the CS in the ATLAS experiment at

113

CERN, where it was also necessary to keep the coil accurately centered during cooldown [35].

3.5.5. Feeders In order to contain the inductance, to limit voltages and to facilitate protection, the large fusion magnets are wound using conductor of large cross-section, enabling them to be powered at high current — up to 30 kA for the LHD and 68 kA for ITER. The associated heat leak into the cryostat must be minimized, added to which it is convenient to locate the power converters, which require servicing, in a separate building. This means that careful attention should be paid to the design of the feeder system. While it is customary to mount the current leads on the cryostat and use water-cooled cables to connect to the power converters, NIFS innovated in providing the LHD with a superconducting bus, allowing the leads to be located in the power converter hall. The nine bus-lines were installed in semiflexible transfer line cryostats built up of 5 concentric corrugated tubes, as shown in Fig. 33. In this way it was possible to manufacture and test the equipment at the factory, and install it rapidly on-site using the technique for laying cables [102]. The system has performed reliably since its installation in 1998. ITER will use a mixed system, with the lead boxes installed around the periphery of the tokamak [103]. Between the magnets and the lead boxes the current will be transferred via Nb–Ti CICC featuring S-bends to allow for thermal contraction. Between the lead boxes and the power converters there will be water-cooled aluminum busbars. In

Fig. 33.

Superconducting bus-line for the LHD.

December 29, 2012

15:27

114

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

order to reduce the heat leak into the cryostat, the system will feature leads with an HTS lower section [104, 105], a solution also adopted by W7-X and JT-60SA [106]. 4. Conclusions The first applications of superconducting magnets to be considered when type 2 superconductors were rendered practical by the work in the mid-1960s were for research in particle and plasma physics, where the potential payoff was immediately recognized. Fortunately for the application of superconductivity, funding for physics research was sufficient at the time to warrant extensive experimentation and finance rapid evolution in the fabrication of conductors. Physics research continues to be a driving force in the quest for better conductors, and in the meantime the availability of reliable material has led to large-scale industrial and medical applications of superconductors, in particular for nuclear magnetic resonance (NMR) and magnetic resonance imaging (MRI). Demand for MRI magnets has created a substantial commercial market for superconductors, which can be justifiably regarded as a useful spinoff from the work done on the more esoteric magnets required for physics research. It is the continuing pursuit of large-scale applications to magnets for physics research instruments that drives the search for better-performing superconductors — which will one day appear in more advanced commercial applications such as high-field MRI [107]. The work done on the development of superconducting magnets for particle detectors and fusion devices is not confined to the development of conductors. The very large magnets call for sophisticated mechanical analysis, for the development of materials capable of withstanding the electromagnetic forces in a hostile radiation environment, and for the development of robust electrical insulation capable of withstanding multiple heat cycling to cryogenic temperatures — development that also presents high potential for spinoffs. There is not space to go into the fine details of these aspects of the magnets. Likewise, we have not addressed the details of standard magnet protection and instrumentation, for which the reader can consult the excellent treatment of the subject in the book by Wilson [108]. With regard to the types of magnets discussed here, it may be that some technologies have been

hastily abandoned on the basis of a few failures, and opportunities lost by slavishly following an accepted “fashion.” For example, one could keep the door open to liquid helium bath cooling (pool boiling) as well as indirect cooling technology for those fusion magnets that are not pulsed. Similarly, “react-and-wind” technology should not be abandoned, as conductor degradation need not be worse than for the windand-react process. Future use of HTS material will also require this approach. It is clear that despite the evident differences in the two species of magnets, the communities benefit from a good exchange of information and by observing developments fostered by a third community making magnets for accelerators. As an example, the fusion community is now benefiting from the development of high-current-density Nb3 Sn conductor required for accelerator magnets. A curious counterexample is the tradition of avoiding the use of MLI (superinsulation) in some fusion devices, while its use in detector magnets (and accelerators) is regarded as essential for providing stable cryogenic conditions, besides reducing the complication of cooling the heat shields. Finally, these magnets are essentially one-of-akind, and failure at any step can lead to long delays and substantial cost overruns. Formal quality control and close technical follow-up on the manufacturer’s premises is vital, and should continue through to testing, installation and operation of the device. The recent tribulations with the manufacture of the coils for the W7-X stellarator highlight the importance of tight quality assurance, as well as inclusion in the design of features that facilitate associated quality control. Acknowledgments We are indebted to N. Mitchell, P. Libeyre, B. Lim and T. Mito for their advice on the preparation of the section on magnets for fusion devices, and to F. Kircher for carefully reading the text and providing many useful suggestions. References [1] A. R. Kantrowitz and Z. J. J. Stekly, Appl. Phys. Lett. 6, 56 (1965). [2] C. Laverick, Cryogenics 5, 152 (1965). [3] Z. J. J. Stekly and J. L. Zar, IEEE Trans. Nucl. Sci. NS-12, 367 (1965).

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

[4] K. Jaeger and J. Purcell, in Proc. 4th Int. Conf. Mag. Tech., MT-4 (1972), pp. 328–332. [5] A. G. Prodell, J. Appl. Phys. 40, 2109 (1969). [6] F. Wittgenstein, in Proc. 4th Int. Conf. Mag. Tech., MT-4 (1972), pp. 295–300. [7] M. N. Wilson, C. R. Walters, J. D. Lewin and P. F. Smith, J. Phys. D 3, 1517 (1970). [8] http://pdg.lbl.gov/2012/reviews/rpp2012-rev-particle-detectors-accel.pdf. [9] M. Derrick, Science 158, 325 (1967). [10] T. Taylor, Phys. Scrip. 23, 459 (1980). [11] M. Morpurgo, Part. Accel. 1, 255 (1970). [12] J. O’Meara, J. Alcorn, P. Brindza, M.-S. Chew, G. Doolittle, M. Fowler, B. Mecking, C. Riggs, D. Tilles and W. Tuzel, IEEE Trans. Magn. 25(2), 1902 (1989). [13] B. C. Barish, AIP Conf. Proc. 272, 1829 (1992). [14] A. Yamamoto, Nucl. Instrum. Meth. A 453, 445 (2000). [15] H. Hirabayashi, IEEE Trans. Magn. 24(2), 1256 (1988). [16] H. Desportes, IEEE Trans. Magn. 30(4), 1525 (1994). [17] D. E. Baynham, IEEE Trans. Appl. Supercond. 16(2), 493 (2006). [18] C. P. Parsch, E. Bartosch, F. B¨ ohm, G. Horlitz, G. Kunst, S. Wolff, A. Stephan and D. Hauck, in Proc. 4th Int. Conf. Mag. Technol. MT-4 (1972), pp. 275–286. [19] M. Morpurgo, Cryogenics 17(2), 89 (1977). [20] H. Desportes, J. Le Bars and G. Mayaux, Adv. Cryogen. Eng. 25, 175 (1980). [21] M. A. Green, Cryogenics 24(3), 659 (1984). [22] H. Minemura, S. Nori, M. Noguchi, R. Yoshizaki, K. Kondo, R. Fast, R. Kephart, R. Wands, R. Yamada, K. K. Aihara, K. Asano, I. Kamishita, I. Kurita, H. Ogata, R. Saito T. Suzuki and T. Yamagiwa, Nucl. Instrum. Meth. A 238, 18 (1983). [23] A. Yamamoto, H. Inoue and H. Hirabayashi, J. de Phys. 45, C1-337 (1984). [24] A. Arai, O. Araoka, Y. Doi, T. Haruyama, N. Ishihara, M. Kawai, Y. Kondo, T. Matsui, T. Mito, M. Sakuda, S. Suzuki, S. Tadano, M. Wake, M. Kuramoto, K. Ishibashi and K. Maehata, Nucl. Instrum. Meth. A 254, 317 (1987). [25] Y. Doi, T. Haruyama, H. Hirabayashi, S. Ishimoto, A. Maki, T. Mito, T. Omori, S. Terada and K. Tsuchiya, Nucl. Instrum. Meth. A 274, 95 (1989). [26] D. M. Coffman, R. D. Ehrlich, S. W. Gray, M. E. Nordberg and M. Pisharody, IEEE Trans. Nucl. Sci. 37(3), 1172 (1990). [27] J. M. Baze, H. Desportes, R. Duthil, J. M. Garin, Y. Pabot, J. Heitzmann, M. Jacquemet, J. P. Jacquemin, F. Kircher, J. C. Languillat, J. Le Bars, A. Le Corroller, C. Lesmond, J. C. Lottin, J. P. Lottin, C. Mathis, C. Meuris, J. C. Sellier

[28] [29] [30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40] [41] [42] [43]

115

and C. Walters, IEEE Trans. Magn. 24(2), 126 (1988). P. T. M. Clee and E. Baynham, in Proc. 11th Int. Conf. Magn. Technol. MT-11 (1989), pp. 206–211. A. B. Oliva, O. Dormicchi, M. Losasso and Q. Lin, IEEE Trans. Magn. 27(2), 1954 (1991). Y. Makida, Y. Doi, T. Haruyama, H. Inoue, N. Kimura, S. Saitou, K. Tanaka, H. Yamaoka and A. Yamamoto, Adv. Cryogen. Eng. 37, 401 (1991). H. Yamaoka, A. Yamamoto, Y. Makida, N. Kimura, K, Tanaka, H. Hirabyashi, R. Ruber, H. Calen, N. Takasu and T. Doi, S. Abe, IEEE Trans. Magn. 32(4), 2151 (1996). P. Fabbricatore, S. Farinon, R. Parodi, E. Baynham, S. F. Carr, D. A. Cragg, R. J. G Greenhalgh, P. L. Sampson, T. G. O’Connor, O. Fackler, B. Bell, W. Burgess, W. Craddock, R. F. Boyce, J. Krebs and N, Yu, IEEE Trans. Magn. 32(4), 2210 (1996). R. P. Smith, H. E. Fisk, N. Krempetz, D. Markley, R. Rucinski, B. Squires, R. Yamada, S. Mine, N. Kimura, T. Kobayashi, H. Kozu and W. Odashima, Adv. Cryogen. Eng. A 43, 229 (1998). Y. Makida, H. Yamaoka, Y. Doi, J. Haba, F. Takasaki and A. Yamamoto, Adv. Cryogen. Eng. A 43, 221 (1998). A. Yamamoto, Y. Makida, R. Ruber, Y. Doi, T. Haruyama, F. Haug, H. Ten Kate, M. Kawai, T. Kondo, J. Metselaar, S. Mizumaki, G. Olesen, O. Pavlov, S. Ravat, E. Sbrissa, K. Tanaka, T. Taylor and H. Yamaoka, Nucl. Instrum. Meth. A 584(1), 53 (2008). Y. Makida, A. Yamamoto, K. Yoshimura, K. Tanaka, J. Suzuki, S. Matsuda, M. Hasegawa, A. Horikoshi, R. Shinoda, K. Sakai, S. Mizumaki, R. Orito, Y. Matsukawa, A. Kusumoto, J. Mitchell, R. Streitmatter, T. Hams, M. Sasaki and N. Thakur, IEEE Trans. Appl. Supercond. 19(3), 1315 (2009). A. Herv´e, D. Campi, B. Cure, P. Fabbricatore, A. Gaddi, F. Kircher and S. Sgobba, IEEE Trans. Appl. Supercond. 18, 346 (2008). L. Zhao, Z. Zhu, K. Wang, J. Zhou, Z. Hou, J. Hu and Z. Liu, IEEE Trans. Appl. Supercond. 20(3), 156 (2010). A. V. Bragin, L. M. Barkov, S. V. Karpov, V. S. Okhapkin, Y. S. Popov, A. A. Ruban and V. P. Smakhtin, IEEE Trans. Appl. Supercond. 18(2), 399 (2010). http://en.wikipedia.org/wiki/Radiation length. T. Taylor, IEEE Trans. Appl. Supercond. 10(1), 342 (2000). A. Herv´e, IEEE Trans. Appl. Supercond. 10(1), 389 (2000). B. Blau, D. Campi, B. Cur´e, R. Folch, A. Herv´e, I. L. Horvarth, F. Kircher, R. Musenich, J. Neuenschwander, P. Riboni, B. Seeber, S. Tavares,

December 29, 2012

15:27

116

[44] [45]

[46]

[47]

[48] [49]

[50]

[51] [52] [53] [54] [55]

[56] [57] [58]

[59] [60] [61] [62]

[63]

[64] [65]

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

S. Sgobba and R. P. Smith, IEEE Trans. Appl. Supercond. 12(1), 345 (2002). H. Ten Kate, IEEE Trans. Appl. Supercond. 10(1), 347 (2000). A. Yamamoto, Y. Makida, K. Tanaka, Y. Doi, T. Kondo, K. Wada and S. Meguro, Nucl. Phys. B 78, 565 (1999). D. E. Baynham, F. C. Carr, E. Holton, J. Buscop, A. Dudarev, P. Benoit, G. Vandoni, R. Ruber, R. Pengo and H. H. J. Ten Kate, IEEE Trans. Appl. Supercond. 17(2), 1197 (2007). B. Blau, S. M. Harrison, H. Hofer, I. L. Horvarth S. R. Milward, J. Ross, S. C. C. Ting, J. Ubricht and G. Viertal, IEEE Trans. Appl. Supercond. 12(1), 349 (2002). R. P. Reed, Cryogenics 12(4), 259 (1972). K. Wada, S. Meguro, H. Sakamoto, T. Shimada, Y. Nagasu, I. Inoue, K. Tsunoda, E, Endo, A. Yamamoto, Y. Makida, K. Tanaka, Y. Doi and T. Kondo, IEEE Trans. Appl. Supercond. 10(1), 373 (2000). A. Yamamoto, T, Taylor, Y. Makida and K. Tanaka, IEEE Trans. Appl. Supercond. 18(2), 362 (2008). A. Yamamoto and Y. Makida, Nucl. Instrum. Meth. A 494, 255 (2002). www.iter.org/newsline/47/680. A. H. Boozer, Rev. Mod. Phys. 76, 1071 (2004). Y. V. Gott, M. S. Ioffe and V. G. Telkowskii, Nucl. Fusion Suppl. 2, Part 3, 1045 (1962). C. D. Beidler, E. Harmeyer, F. Herrnegger, J. Kisslinger, Y. Igitkhanov and H. Wobig, ITC12, Toki Conf. (2001). J. D. Lawson, Proc. Phys. Soc. B 70, 6 (1957). J. File and G. W. Sheffield, in Proc. 4th Int. Conf. Magn. Technol., ed. Winterbottom (1972). M. O. Hoenig, Y. Iwasa and D. B. Montgomery, in Proc. 5th Int. Conf. Magn. Technol., ed. Sacchetti (1975), pp. 519–523. D. S. Beard, W. Klose, S. Shimamoto and G. V´ecsey (eds.), Fus. Eng. Des. 7, 1 (1988). K. Uo, Plasma Phys. 13, 243 (1971). D. Marty, E. K. Maschke, J. Touche and C. Gourdon, Nucl. Fusion 12, 367 (1972). V. Erckmann, H.-J. Hartfuss, M. Kick, H. Renner, J. Sapper, F. Schauer, E. Speth, F. Wesner, F. Wagner, M. Wanner, A. Weller, H. Wobig and the W7-AS teams at IPP Garching, FZK Karlsruhe and IPF Stuttgart, Fusion Eng., Proc. 17th IEEE/NPSS Symp. 1, 40 (1997). T. A. Kozman, S. T. Wang, Y. Chang, E. N. C. Dalder, C. L. Hansen, R. E. Hinkle, J. O. Myall, C. R. Montoya, E. W. Owen, R. L. Palasek, D. W. Shimer and J. H. van Sant, IEEE Trans. Magn. MAG-19(3), 859 (1983). http://www.ccfe.ac.uk/MAST.aspx. C. D. Henning and C. E. Taylor, Cryogenic Engineering: Present Status and Future Development

[66]

[67]

[68]

[69]

[70]

[71] [72] [73]

[74]

[75]

[76]

[77]

[78]

[79]

(Heywood-Temple Industrial Publications, 1968), pp. 139–132. K. R. Efferson, D. L. Coffey, R. L. Brown, J. L. Dunlap, W. F. Gauster, J. N. Luton and J. E. Simkins, IEEE Trans. Nucl. Sci. 18(4), 265 (1971). E. J. Lucas, Z. J. J. Stekly, T. A. deWinter, J. C. Laurence and W. D. Coles, Adv. Cryogen. Eng. 15, 167 (1970). V. E. Keilin, N. A. Chernoplekov and I. A. Chukhin, Problems of Nuclear Physics and Engineering (IAE, Moscow ) 17, 8 (1981) (in Russian). D. P. Ivanov, V. E. Keilin, E. Y. Klimenko, I. A. Kovalev, S. I. Novikov, B. A. Stavissky and N. A. Chernoplekov, IEEE Trans. Magn. MAG-13(1), 694 (1977). K. P. J¨ ungst, G. Friesinger, W. Geiger, A. Gr¨ unhagen, H. Kiesel, P. Komarek, M. Kutschera, G. Obermaier, E. Seuss and L. Yan, in Proc. 9th Int. Conf. Magn. Technol., ed. Marinucci (1985), pp. 36–39. N. A. Chernoplekov, Fus. Eng. Des. 20, 55 (1993). E. Jotaki, S. Itoh and K. Nakamura, Fus. Eng. Des. 36, 289 (1997). R. Aymar, G. Bon Mardion, G. Claudet, F. Disdier, R. Duthil, P. Genevey, J. Hamelin, P. Libeyre, G. Mayaux, C. Meuris, J. Parain, P. Seyfert, A. Torossian and B. Turck, IEEE Trans. Magn. MAG-17(5), 1911 (1981). Y. Takahashi, N. Koizumi, Y. Wadayama, K. Okuno, M. Nishi, T. Isono, K. Yshida, M. Sugimoto, T. Kato, F. Hosono, T. Ando and H. Tsuji, IEEE Trans. Appl. Supercond. 3(1), 610 (1993). M. Darweschsad, G Friesinger, R. Heller, M. Irmisch, H. Kathl, P. Komarek, W. Maurer, G. N¨ other, G. Schleinkofer, C. Schmidt, C. Sihler, M. S¨ usser, A. Ulbricht and F. W¨ uchner, Fus. Eng. Des. 36, 227 (1997). S. Imagawa, S. Masuzaki, N. Yanagi, S. Yamaguichi, T. Satow, J. Yamamoto and O. Motojima (LHD Group), Fus. Eng. Des. 41, 253 (1998). T. Satow, S. Imagawa, N. Yanagi, K. Takahata, T. Mito, S. Yamada, H. Chikaraishi, A. Nishimura, S. Satoh and O. Motojima, IEEE Trans. Appl. Supercond. 10(1), 600 (2000). N. Martovetsky, P. Michael, J. Minervini, A. Radovinsky, M. Takayasu, C. Y. Gung, R. Thome, T. Ando, T. Isono, K. Hamada, T. Kato, K. Kawano, N. Koizumi, K. Matsui, H. Nakajima, G. Nishijima, Y. Nunoya, M. Sugimoto, Y. Takahashi, H. Tsuji, D. Bessette, K. Okuno, N. Mitchell, M. Ricci, R. Zanino, L. Savoldi, K. Arai and A. Ninomiya, IEEE Trans. Appl. Supercond. 12(1), 600 (2002). A. Ulbricht, J. L. Duchateau, W. H. Fietz, D. Ciazynski, H. Fillunger, S. Fink, R. Heller, R. Maix, S. Nicollet, S. Raff, M. Ricci, E. Salpietro,

December 29, 2012

15:27

WSPC/253-RAST : SPI-J100

1230004

Superconducting Magnets for Particle Detectors and Fusion Devices

[80]

[81] [82]

[83]

[84]

[85] [86]

[87]

[88]

[89]

[90]

[91] [92]

G. Zahn, R. Zanino, M. Bagnasco, D. Besette, E. Bobrov, T. Bonicelli, P. Bruzzone, M. S. Darweschsad, P. Decool, N. Dolgetta, A. della Corte, A. Formisano, A. Gr¨ unhagen, P. Hertout, W. Herz, M. Huguet, F. Hurd, Yu. Ilyin, P. Komarek, P. Libeyre, V. Marchese, C. Marinucci, A. Martinez, R. Martone, N. Martovetsky, P. Michael, N. Mitchell, A. Nijhuis, G. N¨ other, Y. Nunoya, M. Polak, A. Portone, L. Savoldi, Richard, M. Spadoni, M. S¨ ußer, S. Turt` u, A. Vostner, Y. Takahashi, F. W¨ uchner and L. Zani, Fus. Eng. Des. 73, 189 (2005). Y. N. Pan, P. D. Weng, Z. M. Chen, B. Z. Li, S. T. Wu, W. Y. Wu, B. J. Ciao, I. Yu, D. Wu, X. B. Wu, Q. Chen and W. G. Chen, IEEE Trans. Appl. Supercond. 10(1), 628 (2002). W. Y. Wu and B. Z. Li, Fus. Eng. Des. 58–59, 277 (2005). H. J. Ahn, Y. W. Lee, T. H. Kwon, S. C. Lee, C. H. Choi, Y. K. Oh, D. K. Lee, J. S. Lee, D. S. Kim and C. D. Hong, IEEE Trans. Appl. Supercond. 12(1), 492 (2002). Y. K Oh, C. H. Choi, J. W. Sa, K. I. You, D. K. Lee, M. Kwon, G. S. Lee, H. J. Ahn, T. H. Kwon, Y. W. Lee, S. C. Lee and C. D. Hong, IEEE Trans. Appl. Supercond. 12(1), 615 (2002). S. Pradhan, Y. C. Saxena, B. Sarkar, G. Bansal, A. N. Sharma, K. J. Thomas, V. Bedakihale, B. Doshi, C. P. Dhard, U. Prasad, P. Rathod, R. Bahl, A. Varadarajulu, A. Mankani and SST-1 Team, in Proc. 20th IAEA Fusion Energy Conf. (2004), pp. FT/3-4Rb. J. Sapper and the W7-X team, Supercond. Sci. Technol. 13, 516 (2000). V. Tomarchio, P. Barabaschi, A. Cucchiaro, P. Decool, A. Della Corte, A. Di Zenobio, D. Duglue, L. Meunier, L. Muzzi, M. Nannini, M. Peyrot, G. Phillips, A. Pizzuto, C. Portafaix, L. Reccia, K. Yoshida and L. Zani, IEEE Trans. Appl. Supercond. 20(3), 572 (2000). K. Tsuchiya, K. Kizu, T. Ando, H. Tamai and M. Matsukawa, Fus. Eng. Des. 82(5–14), 1519 (2007). C. Sborchia, Y. Fu, R. Gallix, C. Jong, J. Knaster and N. Mitchell, IEEE Trans. Appl. Supercond. 18(2), 463 (2008). P. Libeyre, C. Beemsterboer, D. Bessette, Y. Gribov, C. Jong, C. Lyraud, N. Dolgetta, N. Mitchell and T. Vollmann, IEEE Trans. Appl. Supercond. 20(3), 398 (2010). B. Lim, F. Simon, Y. Ilyin, C. Y. Gung, J. Smith, Y. H. Hsu, C. Luongo, C. Jong and N. Mitchell, IEEE Trans. Appl. Supercond. 21(3), 1918 (2011). T. Ito and H. Kubota, Cryogenics 29(6), 621 (1989). S. Imagawa, N. Yanagi, S. Hamaguchi, T. Mito, K. Takahata, H. Chikaraishi, H. Tamura, A.

[93]

[94] [95]

[96] [97]

[98] [99] [100] [101]

[102]

[103]

[104]

[105]

[106] [107]

[108]

117

Iwamoto, S. Yamada and O. Motojima, IEEE Trans. Appl. Supercond. 16, 755 (2006). K. Kim, H. K. Park, K. R. Park, B. S. Lim, S. I. Lee, Y. Chu, W. H. Chung, Y. K. Oh, S. H. Baek, S. J. Lee, H. Yonekawa, J. S. Kim, C. S. Kim, J. Y. Choi, Y. B. Chang, S. H. Park, D. J. Kim, N. H Song, K. P. Kim, Y. J. Song, I. S. Woo, W. S. Han, S. H. Lee, D. K. Lee, K. S. Lee, W. W. Park, J. J. Joo, H. S. Park, G. S. Lee and KSTAR Team, in Proc. 20th IAEA Fusion Energy Conf. (2004), pp. FT/P7-16. N. Mitchell, D. Bessette, R. Gallix and C. Jong, IEEE Trans. Appl. Supercond. 18(2), 435 (2008). N. Mitchell, A. Devred, P. Libeyre, B. Lim, F. Savary and ITER Magnet Division, IEEE Trans. Appl. Supercond. 22(3), 4200809 (2012). http ://www . oxford - instruments . com / products/ superconducting-wires/documents. M. K. Sheth, P. J. Lee, D. M. McRae, C. M. Sanabria, W. L. Starch, R. P. Walsh, M. C. Jewell, A. Devred and D. C. Larbalestier, IEEE Trans. Appl. Supercond. 22(3), 4802504 (2012). P. Bruzzone, IEEE Trans. Appl. Supercond. 16(2), 839 (2006). D. Evans, J. Knaaster and H. Rajainm¨ aki, IEEE Trans. Appl. Supercond. 22(3), 4202805 (2012). J. A. Rice, K. R. Gall and G. Vos, IEEE Trans. Appl. Supercond. 13(2), 1476 (2003). R. Prokopec, K. Humer, R. K. Maix, H. Fillunger and H. W. Weber, J. Knaster, IEEE Trans. Appl. Supercond. 22(3), 7700604 (2012). T. Uede, S. Yamada, T, Mito, H. Hiue, I. Itoh and O. Motojima, IEEE Trans. Appl. Supercond. 11(1), 2563 (2001). Y. Song, K. Lu, X. Huang, Y. Chen, T. Zhou, S, Liu, P. Bauer, Y. Bi, Y. Chen, A. Devred, K. Ding, E. Niu, C. Y. Gung, N. Mitchell, G. Shen and Z. Wang, IEEE Trans. Appl. Supercond. 22(3), 4800404 (2012). R. Heller, D. Aized, A. Akhmetov, W. H. Fietz, F. Hurd, J. Kellers, A. Kienzler, A. Lingor, J. Maguire, A. Vostner and R. Wesche, IEEE Trans. Appl. Supercond. 14(2), 1774 (2004). A. Ballarino, P. Bauer, Y. Bi, A. Devred, K. Ding, A. Foussat, N. Mitchell, G. Shen, Y. Song, T. Taylor, Y. Yang and T. Zhou, IEEE Trans. Appl. Supercond. 22(3), 4800304 (2012). R. Heller, W. H. Fietz, A. Kienzler and R. Lietzow, Fus. Eng. Des. 86, 1422 (2011). P. Vedrine, G. Aubert, F. Beaudet, J. Belorgey, C. Berriaud, P. Bredy, A. Donati, O. Dubois, G. Gilgrass, F. P. Juster, C. Meuris, F. Molinie, F. Nunio, A. Payn, T. Schild, L. Scola and A. Sinanna, IEEE Trans. Appl. Supercond. 20(3), 696 (2010). M. N. Wilson, Superconducting Magnets (Oxford University Press, 1987).

December 29, 2012

15:27

118

WSPC/253-RAST : SPI-J100

1230004

A. Yamamoto & T. Taylor

Akira Yamamoto has been working for many years in the field of applied superconductivity for particle physics and accelerators. His particular interest has been in the development of superconducting magnets for particle detectors, a domain in which he is highly respected as a world expert. Recently he has also been contributing to the development of superconducting rf technology for the International Linear Collider project, in the context of its realization via global cooperation.

Thomas Taylor moved from R&D on microwave tubes in the UK to participate in the ISR project at CERN. There he worked on the design and procurement of magnets for accelerators and experiments, beam optics, and system design of low-beta insertions. Introduced to superconductivity during a stay at BNL in 1974, he has maintained a keen interest in this subject, designing and building superconducting devices and later chairing the group reporting to the LHC experiments committee on the detector magnets. He is presently collaborating with ITER on design aspects of the feeder system.

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Reviews of Accelerator Science and Technology Vol. 5 (2012) 119–146 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300058

Superconducting Radio-Frequency Fundamentals for Particle Accelerators Alex Gurevich Department of Physics and Center for Accelerator Science, Old Dominion University, 4600 Elkhorn Avenue, Norfolk, VA 23529, USA [email protected] An overview of fundamentals of superconductors under radio-frequency electromagnetic fields in particle accelerators is given, with emphasis on intrinsic physics and materials mechanisms which limit the performance of the superconducting radio-frequency (SRF) resonator cavities. Multiscale mechanisms which control the surface resistance and the quality factor of the SRF cavities at low and high rf fields are discussed. We also discuss possible ways of pushing the limit of the SRF performance by materials impurities and multilayer nanostructuring which may open up opportunities of using materials other than Nb to significantly increase the maximum accelerating fields and improve the performance of the SRF cavities operating at 4.2 K. Keywords: Superconducting cavities; Meissner state; surface resistance; quality factor; breakdown field.

1. Introduction

is quantified by the quality factor [4],
ωµ0 V |H(r)|2 dV , Q=  2 A Rs |H(r)| dA

An ideal particle accelerator should provide high accelerating electric fields Eacc at minimum power consumption in a radio-frequency (rf) resonator cavity structure. The typical values of Eacc ∼ 10−102 MV/m in existing accelerators are mostly limited by the cavity materials parameters, so pushing the limit of energies of accelerated particles inevitably requires increasing the number (up to tens of thousands) of cavities. The necessity to reduce the power consumption in big machines brought about the idea of using superconducting rf resonating (SRF) cavities, proposed more than 50 years ago [1–3]. This idea appeared at the height of the excitement triggered by the breakthrough in the understanding of the microscopic mechanism of superconductivity, just four years after the development of the Bardeen–Cooper–Schrieffer (BCS) theory. The BCS theory, among many other things, explained the extremely low power loss in superconductors under low-frequency electromagnetic fields, providing the theoretical background for the emerging SRF technology. The SRF cavities have much lower dissipation than the cavities made of non-superconducting metals, such as Cu. The rf dissipation in a cavity

(1)

proportional to the ratio of the mean electromagnetic energy stored in the cavity to the mean dissipated power, where the integration of the rf magnetic field H(r, ω) for the excited rf mode with the circular eigenfrequency ω = 2πf goes over the cavity volume V and the surface A. Here the surface resistance Rs , caused by the rf dissipation in the cavity wall, defines one of the main parameters of merit of the SRF cavities. Generally, Rs (H(r), r) depends on the rf field amplitude H(r) and may also vary along the cavity wall due to the surface defects, as will be discussed below. Since ω ∼ c/L is inversely proportional to the characteristic size of the cavity L, it is convenient to write the quality factor in the form Q(Bp ) = G/Rs , where Bp = µ0 Hp , Hp is the magnitude of the rf magnetic field, G = cµ0 α, c is the speed of light, · · · means averaging of Rs over the cavity surface, as defined by Eq. (1), and the dimensionless factor α depends on the cavity geometry. The vacuum impedance G0 = µ0 c = 377 Ω sets the scale of G to the accuracy of the cavity-specific factor α ∼ 1. For instance, for a pillbox cavity, α = 0.68 and G = 257 Ω [5, 6]. 119

January 10, 2013

13:7

120

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

It is instructive to compare the values of Rs and Q for the normal and the SRF cavities. For a clean Nb with the normal state resistivity ρn = 10−8 Ω m at T  10 K, the surface resistance for the normal skin effect Rs = (πµ0 f ρn )1/2  7 mΩ yields Q ∼ Q0 /Rs ∼ 105 at the rf frequency f = 1.3 GHz. However, the cavity-grade superconducting Nb at 2 K has Rs  20 nΩ [6], which results in a five orders of magnitude gain in Q ∼ Q0 /Rs  2 × 1010 . Such a small Rs at temperatures T well below the superconducting transition temperature Tc and frequencies f  kB Tc /h is due to the rf heating of the exponentially small density of unpaired electrons (quasiparticles) resulting from thermal dissociation of the Cooper pairs [8–11]:    2 1.76Tc Af exp − , (2) RBCS = T T where the factor A depends on the purity of the material, and h and kB are the Planck and Boltzmann constants, respectively. The BCS surface resistance decreases greatly as T decreases, suggesting that the efficiency of the SRF cavities is best at the lowest temperature defined by the balance of power losses and operational costs. Meanwhile, the low-field Rs (T, f ) observed on many superconducting materials generally follows a somewhat different temperature dependence, Rs (T ) = RBCS (T ) + Ri , for which Rs (T ) does not vanish at T → 0 but tends to a temperature-independent residual resistance Ri [12, 13]. The functional form of RBCS (T, f ) imposes restrictions on the range of T and f in which the SRF cavities outperform the normal metal resonator cavities, given that the parameters of merit of accelerators also include the materials and operational costs, particularly the cost of refrigeration. For dc or long-pulse operations, the cost of the rf power and refrigeration for normal cavities becomes prohibitive if Eacc exceeds a few MV/m, yet the normal cavities can provide higher peak gradients ∼ 100 MV/m for short(millisecond)-pulse low-duty applications than the SRF cavities [5, 6]. In turn, the quadratic dependence of RBCS ∝ f 2 makes the SRF cavities more efficient at low frequencies, typically 50 MHz < f < 3 GHz, where 50 MHz is limited by the cavity diameter L ∼ c/f . The higher-frequency (0.3
f
3 GHz) SRF cavities are mostly used for electrons with β = v/c → 1 at 1.8 − 2 K. At such low T , the

exponential drop of RBCS (T ) compensates for the increase of RBCS at higher f , while heat transfer from the cavities to the helium coolant becomes much more efficient below the transition to the superfluid state at Tλ = 2.17 K. At the same time, a smaller RBCS at lower frequencies, 50
f
500 MHz, allows using bigger SRF cavities (mostly in low-β heavy ion accelerators) at the boiling temperature of liquid helium, T0 = 4.2 K, for which the cryogenic costs are much lower than at 2 K [5, 6]. Based on the exponential decrease of RBCS with Tc in Eq. (2), one may think that a superconductor with the highest critical temperature would be best for the SRF cavities. However, the current SRF material of choice, Nb, has a rather modest Tc = 9.2 K as compared to many superconductors, like Nb3 Sn with Tc = 18.2 K, MgB2 with Tc ≈ 40 K [14, 15], high-Tc cuprates with Tc up to 120 K or the recently discovered large family of Fe-based pnictides with Tc up to 55 K [16]. The reasons which make Nb special are both fundamental and technological, the critical temperature being not the only important parameter here. As far as the cavity applications are concerned, the right SRF material should provide: (1) Low surface resistance, including low residual resistance at T → 0; (2) An s-wave Cooper pairing state with a full superconducting gap on the entire Fermi surface; (3) A high lower critical magnetic field Hc1 at which the weakly dissipative Meissner state is destroyed due to penetration of vortices; (4) A high superheating magnetic field which defines the theoretical limit of the SRF breakdown; (5) High thermal conductivity to transfer the rf dissipated power through the cavity wall; (6) Grain boundaries transparent to high rf screening currents in polycrystalline cavities; (7) Comparatively simple chemical composition, so that the material is not contaminated by nonsuperconducting second phases, and the superconducting properties are not degraded by local chemical nonstoichiometry; (8) Good mechanical properties and malleability to minimize crack formation during cavity manufacturing (forging, deep drawing, etc.) Here only points 1–4 pertain to the fundamental superconducting properties, some of these requirements being mutually exclusive. For instance,

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

materials with Tc higher than Tc = 9.2 K of Nb are the type-II superconductors with Hc1 much smaller than Hc1 of Nb, which has the highest lower critical field among all superconductors. Thus, Nb turns out to be best-protected against highly dissipative penetration of vortices which degrade Q(Bp ) at higher rf field amplitudes and can cause cavity quench. In turn, high-Tc cuprates, in addition to their very low Hc1 values, do not satisfy point 2 because they are the so-called d-wave superconductors for which the superconducting gap vanishes along certain directions on the Fermi surface. This results in the power law temperature dependence of the surface resistance Rs ∝ T n , with n = 2−4 [13], which can hardly compete with the exponentially small Rs (T ) for the swave superconductors at T  Tc . Non-superconducting properties of the SRF cavities are also important. For instance, Nb3 Sn has high Tc and low Rs in weak rf fields [17–20], but its thermal conductivity κ is some three orders of magnitude lower than κ of clean Nb at 2 K. Thus, despite its better SRF performance at low fields, Nb3 Sn is more prone to the rf overheating which degrades Q(Bp ) at higher fields (even a several micron thick Nb3 Sn film on the inner surface of the Nb cavity can significantly increase the thermal impedance of the cavity wall). As a result, it was Nb which provided the best compromise for all these conflicting physics and materials requirements. This situation is typical of other power and magnet applications of superconductors which also involve compromises between conflicting requirements, depending on the operating conditions and on the specific application [21, 22]. The progress in the Nb technology over the last four decades has resulted in a significant increase of the accelerating gradient in the SRF cavities, from  3 MV/m to  40−50 MV/m [5–7]. For instance, Fig. 1 illustrates one of the main SRF parameters of merit: the quality factor Q as a function of the peak rf field Bp for the high-performance Nb cavities [23]. These Q(Bp ) curves can also be replotted as functions of the peak electric field Ep (MV/m) = 0.43Bp (mT) or the accelerating field Eacc (MV/m) = 0.29Bp (mT), using the conversion coefficients for the particular cavity geometry. The breakdown fields Bb , above which the cavities become nonsuperconducting, are the end points of the Q(Bp ) curves, as shown by the arrow in Fig. 1 (for this particular cavity, Bb = 179 mT and Eacc ≈ 52 MV/m).

121

The state-of-the-art Nb cavities have reached even higher breakdown fields and accelerator gradients (up to 59 MV/m) [24–26]. The Q(Bp ) curves have three distinct regions, evident in Fig. 1: (1) low fields in which Q(Bp ) stays flat or increases as Bp increases; (2) intermediate fields in which Q(Bp ) gradually decreases by a factor of ∼2 or so (medium-field Q-slope); (3) high fields around the breakdown field Bb in which Q(Bp ) can either continue the gradual decline or start decreasing much faster as Bp increases (highfield Q-slope). The Q(Bp ) curves shown in the figure demonstrate several important points. First of all, the best SRF cavities can achieve very high quality factors, Q ∼ 1010 , and accelerating fields up to Eacc  50 MV/m. This is rather remarkable given that this field corresponds to the peak magnetic fields Bp  200 mT close to the thermodynamic critical field Bc of Nb at 2 K [27]. At Bp  Bc the screening rf current density flowing at the inner cavity surface is close to the depairing current density — the maximum current density a superconductor can carry in the Meissner state [31–33]. Thus, the breakdown fields of the best SRF cavities have nearly reached the theoretical limit defined by the superheating field Bsh  Bc [34–39]. These advances of the Nb cavity technology indicate that further increase of Bb well above 200 mT may be unlikely, yet Fig. 1 shows that proper materials treatments can significantly improve the SRF performance. In particular,

Fig. 1. The quality factor as a function of the peak magnetic field Bp for the fine grain ILC single-cell cavity “AES001” after the baseline test (squares) and after the heat treatment (triangles) at 800◦ C/3 h and 400◦ C/h. The circles are the data after additional in situ baking at 120◦ C for 24 h. (Reproduced from Ref. 23.)

January 10, 2013

13:7

122

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

the high-temperature (600–800◦C) annealing for a few hours combined with the low-temperature (100– 120◦ C) baking for 24 h not only increases Q(Bp ) and the breakdown field but also eliminates the high-field Q-slope [23, 40–44]. There is no general agreement in the literature on the mechanisms of this effect, but this issue remains an area of active investigation, as will be discussed below. Although the breakdown fields of the best Nb cavities are close to the theoretical limit, the question about the fundamental lower limit of the surface resistance and how it could be reduced by impurities and surface modifications remains open. The obvious answer that the maximum Q at Bp  Bb is just limited by the low-field BCS surface resistance Q = G/RBCS may be oversimplified, because the observed increase of Q(Bp ) like the one shown in Fig. 1 does not follow from the conventional BCS model [8, 46]. Understanding the physics behind this increase of Q(Bp ) and how it could be extended to higher fields remains an important problem to be addressed in order to access the ultimate SRF performance limits. In turn, increasing the SRF breakdown fields above 200 mT would require breaking the Nb “monopoly” and using superconductors with higher Tc , which implies resolving the low Bc1 problem of these materials. It was recently suggested to use nanoscale multilayer coating of the inner surface of the Nb cavities to increases Hc1 and push the breakdown field up to the superheating field of the coating material [47]. This approach may open up an opportunity to turn many high-Tc superconductors into high-performance SRF materials, which could potentially push the maximum accelerating field above 100 MV/m. This article gives a brief overview of the basic physics and materials mechanisms which may limit the quality factor of the SRF cavities at low and high rf fields, and also discusses possibilities of using materials other than Nb to increase the SRF performance. Here we focus only on the SRF fundamentals for single-cell, β = 1 cavities, leaving aside technical details of the electrodynamics of different cavity geometries [5–7], as well as such extrinsic SRF performance-limiting factors as field emission and multipacting [5, 6, 48–50]. The latter two issues have become of lesser importance in recent years, after the effective techniques of cleaning the inner surface of the Nb cavities and of high-power processing

were developed. These techniques basically remove or deactivate “dust” microparticles or other local sources of field emission of electrons, which are then further accelerated by the rf field and repeatedly impact the cavity wall. The field emission causes a strong Q(Bp ) slope above a threshold field and is accompanied by bursts of X-rays produced by the accelerated electrons. It can cause multipacting — an avalanche of secondary electrons produced by the electrons hitting the cavity wall. The multipacting can be effectively suppressed by changing the shape of the SRF cavities [6, 49, 50]. In this article we mostly discuss the physics and materials mechanisms behind the performance limits of high-gradient SRF cavities for which the filed emission and multipacting are suppressed by the appropriate cavity treatments [7]. 2. Electrodynamics of Superconductors The SRF cavities usually excite the TM010 resonance mode in which the accelerating electric field is maximum at the cavity axis while the tangential rf magnetic field is maximum at the cavity surface. In turn, the surface magnetic field induces the screening current density J(z, t) flowing along the inner cavity surface. The electrodynamics of these currents is governed by the Meissner effect, which gives rise to the expulsion of the magnetic field from the bulk of a superconductor. The Meissner effect is one of the most fundamental properties of superconductivity, resulting from the Bose condensation of the Cooper pairs (electrons with antiparallel momenta and spins glued together by lattice vibrations — phonons) and the appearance of the energy gap 2∆ in the spectrum of quasiparticles. In the BCS theory ∆ = 1.764kB Tc is proportional to the binding energy of the Cooper pairs, where kB is the Boltzmann constant [31]. The surface resistance results from the rf dissipation in a thin, ∼100 nm layer determined by the London penetration depth λ, over which a low-frequency magnetic field penetrates in a superconductor (for the cavity-grade niobium, λ ≈ 40 nm at 2 K). It is the materials and superconducting properties of this thin layer (much thinner than the 2–3-mm-thick cavity wall) at the inner cavity surface which control the SRF performance. The atomic and chemical structure of this surface layer has been studied by different techniques, including transmission electron microscopy (TEM) [51], X-ray photoelectron

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

(a)

123

(b)

Fig. 2. Transmission electron microscopy (TEM) images of the surface layer of BCP (a) and BEP (b)–treated Nb samples. (Reproduced from Ref. 51.)

spectroscopy [52–54] (XPS), atom probe tomography [56, 57] and electron energy loss spectroscopy (EELS) [58]. For instance, the TEM shown in Fig. 2 reveals a complex structure consisting of a 2–4-nm-thick layer of niobium oxides and suboxides followed by a several-nm-thick transitional zone and disordered region with enhanced concentration of atomic defects such as vacancies, or oxygen, carbon or hydrogen impurities. The detailed physics and materials mechanisms by which the atomic structure of this surface layer affects superconducting properties and supplies impurities which diffuse to a thicker, ∼100 nm layer of the rf field penetration [59] are not well understood. These mechanisms may be tuned by the cavity heat treatments, such as low-T baking, [23, 40–44] to improve the SRF performance. The electromagnetic response of superconductors in weak fields is described by the following relation for the Fourier components of the current density J(k, ω) induced by the magnetic vector potential A(k, ω) (in the gauge divA = 0) [8, 9]: J(k, ω) = −K(k, ω)A(k, ω),

(3)

where B = ∇ × A, and the complex electromagnetic kernel K(k, ω) depends on the circular frequency ω = 2πf and the wave vector k of the rf field. The real part of K describes the Meissner effect caused by the frictionless dynamics of the superconducting condensate, so that K(0, 0) = 1/µ0 λ2 in the static limit. The imaginary part of K describes dissipative processes due to the rf heating of quasiparticles, and the generation of quasiparticles by the rf

field. The BCS theory gives a general formulas for K(k, ω, T, ) as functions of k, ω, T and the mean free path  due to scattering of electrons on impurities [8, 9]. Although the BCS model gives a correct quantitative description of the electromagnetic response, it can hardly be used for quantitative calculations of K(k, ω, T, ) in such materials as Nb or Pb, for which the electron–phonon coupling responsible for the Cooper pairing is not weak [27]. The properties of superconductors with strong electron– phonon interaction are described by the Eliashberg theory [27], for which formulas for K(k, ω, T, ) were also obtained [10]. The analysis of microwave conductivity for Nb and Pb in the framework of the Eliashberg theory was performed in Refs. 28 and 29. The function K(k, ω) can be used to calculate the surface impedance:  ∞ J(z, ω)dz. (4) Z(ω) = Rs + iXs = E(0, ω) 0

Here the surface resistance Rs can be expressed in terms of K(k, ω, T, ) by certain integral relations derived in Refs. 8–11, 30, which also depend on the way the electrons are scattering by the surface (specular or diffusive). Using these general relations, Z(ω, T ) can be calculated numerically for arbitrary temperatures, rf frequencies and concentrations of impurities for particular materials like Nb [11, 28]. The situation simplifies at low temperatures and frequencies hf  kB Tc for which Rs  Xs . Since RBCS (ω) decreases quadratically with ω, the SRF cavities, unlike the normal cavities, are most efficient

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

124

at low frequencies, f ∼ 50 MHz − 3 GHz much lower than the gap frequency fg = 2∆/h or fg (GHz) = 74Tc (K) = 680 GHz for Nb at T  Tc [1–3]. Such low-frequency rf quanta cannot break the Cooper pairs and generate new quasiparticles, so a weak rf field in the SRF cavities mostly results in the Joule heating of equilibrium quasiparticles. The surface impedance depends on several important lengths: the London penetration depth λ; the coherence length ξ, which quantifies the size of the Cooper pairs; and the mean free path , which affects both λ() and ξ(). We discuss here the most transparent case of λ ξ and  < ξ, for which the quasi-stationary rf magnetic and electric fields are confined in the layer of thickness λ at the surface: B(z, t) = Bp e−z/λ sin ωt,   Bp e−z/λ sin ωt, Js (z, t) = µ0 λ E(z, t) = ωλBp e−z/λ cos ωt.

(5) (6)

the limit of weak ohmic dissipation, σ1  σ2 :  1/2 σ2 µ0 ω iµ0 ω µ0 ωσ1 +i Z=  . σ1 − iσ2 2σ22 µ0 ω σ2 (10) Substituting here Eq. (9) yields 2Rs = µ20 ω 2 λ3 σ1 (ω),

Xs = µ0 ωλ.

(11)

The effect of scattering of electrons by impurities on Rs depends on the relation between the penetration depth λ and the mean free path , as illustrated in Fig. 3. The upper panel shows a trajectory of a quasiparticle for a dirty material with   λ, in which a quasiparicle propagating toward the surface undergoes multiple scattering on impurities while interacting with the rf field. The lower panel shows the clean case of λ  , for which an incoming quasiparticle propagates ballistically during interaction with the rf field and the subsequent

(7)

Here λ is essentially the London penetration depth for a quasistatic magnetic field for which the Joule rf power q = J(t)E(t) calculated from Eqs. (5)–(7) averages to zero. This is because Eqs. (5)–(7) only describe undamped oscillations of the superconducting current density Js described by the first London equation, dJs /dt = (e2 /m)E, which is nothing but the second Newton law for the superfluid condensate of Cooper pairs. Dissipation comes from a small “normal” component of the current density Jn which oscillates in phase with the electric field. Both in-phase and out-of-phase components of J(r, t) can be calculated from the Maxwell equations combined with Eq. (3), which generally gives a rather complicated nonlocal relation between J(r, ω) and A(r, ω) in the coordinate space [8]. In the limit of λ ξ, Eq. (3) simplifies to the local “ohmic” form, J(rω) = (σ1 − iσ2 )E(r, ω),

(8)

with a frequency-dependent complex conductivity, σ(ω) = σ1 (ω) − iσ2 (ω). The reactive part σ2 can be expressed in terms of λ using Eqs. (6)–(8): σ2 =

1 . ωµ0 λ2

(9)

The surface impedance is then calculated using the standard formula of the electromagnetic theory [4] in

Fig. 3. Sketch of the trajectory of a quasiparticle shown by the dashed line for the normal (top) and anomalous (bottom) skin effects. Here the black dots show the scattering impurities, and the color map and horizontal arrows show the distributions of the Meissner screening currents confined in the surface layer of thickness of the London penetration depth λ.

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

reflection off the surface. These two cases are similar to the normal and anomalous skin effects in nonsuperconducting metals [45]. For the normal skin effect (upper panel), Rs depends on the mean free path and thus can be tuned by the redistribution of impurities at the surface. By contrast, Rs in the anomalous skin effect limit (lower panel) becomes independent of the materials purity. The dirty limit  < ξ0 may be more relevant to the actual Nb surface shown in Fig. 2. Indeed, if  is of the order of the spacing between impurities, the dirty limit occurs for the atomic concentration of impurities x exceeding a rather small value, xd ∼ (a/ξ0 )3  6 × 10−5 at. %, where a = 0.33 nm is the lattice spacing in Nb and ξ0 = 40 nm. 2.1. BCS surface resistance The BCS theory gives explicit formulas for Rs and λ at low temperatures and frequencies relevant to the SRF cavities. The penetration depth for T = 0 and an arbitrary mean free path  is defined by [10]

cos−1 (α/2) 1 π 2 , α < 2, (12) − = λ2 αλ20 2 1 − α2 /4

1 2 cosh−1 (α/2) π , α > 2, (13) = −  λ2 αλ20 2 α2 /4 − 1 where λ0 = (m/µ0 ne2 )1/2 is the London penetration depth in the clean limit at T = 0, and n and m are the electron density and effective mass, respectively. The parameter α = πξ0 / quantifies scattering on impurities, so that α  1 and α 1 correspond to the clean and the dirty limit, respectively. Here ξ0 =
vF /π∆ is the coherence length in the clean limit, where vF is the Fermi velocity. As  decreases, λ() increases from λ0 in the clean limit,  ξ0 , to λ()  λ0 (ξ0 /)1/2 in the dirty limit,  < ξ. The dependence of λ() is often  approximated by the interpolation formula λ  λ0 1 + ξ0 /. In turn, the coherence √ length ξ() decreases from ξ0 at  ξ0 to ξ  ξ0 for   ξ0 [31]. The BCS temperature dependence of ∆(T ) can be approximated by [10]  2 πT 1/2 , (14) ∆(T ) = ∆0 cos 2Tc2 where ∆0 = ∆(0). In the BCS model, weak electron scattering on nonmagnetic impurities does not affect

125

thermodynamic properties of superconductors with a spherical Fermi surface (Anderson theorem) [31], so ∆(T ), Tc and Bc are independent of the material purity. This may not be the case for the actual atomic impurities (like oxygen, carbon or hydrogen), which can reduce ∆ in a more realistic case of strong scattering and anisotropic Fermi surface [87], consistent with experiments on the Nb cavities [43]. The dissipative part σ1 (ω) of the complex conductivity which defines Rs in Eq. (11) at f < fg and  < ξ0 is given by the Mattis–Bardeen theory [8, 10]:  2σn ∞ (2 + ∆2 +
ω)[f () − f ( +
ω)]  √ d , σ1 =
ω ∆ 2 − ∆2 ( +
ω)2 − ∆2 (15) where σn is the normal state residual conductivity and f () = (e/kB T + 1)−1 is the Fermi distribution function. For T  Tc , the main contribution to the integral in Eq. (15) comes from a narrow range of quasiparticle energies,  − ∆ ∼ kB T  ∆, for which 2 − ∆2 ≈ 2∆( − ∆), and f () − f ( +
ω) ≈ [1 − e−
ω/kB T ]e−/kB T . Then the integration can be performed exactly (see e.g. Ref. 60):    
ω σ1
ω 4∆ sinh K0 e−∆/kB T , (16) = σn
ω 2kB T 2kB T where K0 (x) is a modified Bessel function. The conductivity σ1 (T ) decreases exponentially as T decreases, similar to the density of normal quasiparticles neq = n(πkB T /2∆)1/2 e−∆/kB T . For very low temperatures, T  hf /2kB ∼ 10−3 K at f ∼ 1 GHz, the asymptotic expansion of K0 (x)  (π/2x)1/2 e−∆/kB T yields σ1  2σn ∆(πkB T )1/2 (
ω)−3/2 e−∆/kB T . In the opposite limit of T hf /2kB (relevant to the SRF cavities), the use of K0 (x)  ln(2/x) − C at x  1 yields σ1  (2σn ∆/kB T )[ln(4kB T /
ω)−C]e−∆/kB T , where C = 0.577 is the Euler constant. Substituting this into Eq. (11) gives     C1 kB T ∆ µ2 ω 2 λ3 σn ∆ ln exp − , (17) Rs = 0 kB T
ω kB T where C1 = 4/eC = 2.246. Equation (17) reduces to Eq. (2), in which A depends weakly on ω, while the dependence of Rs on the mean free path  is determined by the factor λ3 σn , where λ() is defined by Eqs. (12) and (13), and σn is given by the Drude

January 10, 2013

13:7

126

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

formula σ = ne2 /pF , where pF =
(3π 2 n)1/3 is the Fermi momentum. As a function of , the product σλ3 ∝ (1 + ξ0 /)3/2 has a minimum at  = 0.5ξ0 , where ξ0  40 nm for a clean Nb at T = 0. However, this estimate was based on Eq. (17), derived for the local limit under the normal skin effect. As the material gets cleaner and  increases, the transition from the normal to the anomalous skin effect occurs and Eq. (17) becomes invalid as the electromagnetic response becomes nonlocal. In this case the exact Rs approaches a finite value at  → ∞ while Eq. (17) suggests that Rs → ∞ as σn ∝  → ∞. Yet a shallow minimum of Rs () at  ∼ ξ was also obtained from numerical calculations of Rs using the full BCS electromagnetic kernel K(k, ω, T, ) [11]. The minimum in Rs () suggests that the SRF cavity performance could be optimized by tuning the density of impurities at the cavity surface. For instance, different heat treatments can redistribute impurities at the surface to provide the optimum condition of   ξ. A minimum in the surface resistance as a function of the low-T baking temperature was indeed observed [61]. In the clean limit  λ, the surface resistance is given by an expression similar to Eq. (17) in which σn is replaced by an effective conductivity, σeff ∼ ne2 λ/pF [8]. Here σeff resembles the Drude formula, in which the mean free path is replaced with the field penetration depth, similar to the anomalous skin effect in normal metals [45]. 2.2. Residual resistance The temperature dependence of Rs (T ) observed on the Nb cavities at low rf fields Bp  Bc generally follows Eq. (17) of the BCS theory except at very low temperatures. Numerous experiments have shown that Rs (T ) can be described by the following semiempirical relation (see e.g. Refs. 12 and 13):    2 βTc Af exp − + Ri , Rs = (18) T T where β ≈ 1.9 for Nb is slightly higher than the BCS prediction, β = 1.76, and the constant term Ri is the so-called residual resistance. While the larger value of β is consistent with the strong electron–phonon coupling in Nb [27], the appearance of a finite Ri is more subtle since the BCS theory gives Rs (T ) → 0 at T → 0. Yet the residual resistance has been observed in numerous measurements on both the SRF cavities

Fig. 4. Arrhenius plot of the surface resistance before and after baking. The solid lines show the fit to Eq. (18). (Reproduced from Ref. 61.)

and superconducting thin films. For instance, the Arrhenius plot of ln Rs versus 1/T in Fig. 4 clearly shows a deviations from the straight line at low T due to a finite Ri in the Nb cavities. In the literature Ri was attributed to dissipation caused by trapped vortices oscillating under the rf field [62–65], lossy oxides or metallic hydrides on the Nb surface [6, 7, 59, 66–69], grain boundaries [70–73], generation of hypersound [74] or localized electron surface states [75]. A significant contribution to Ri caused by eddy current losses in metallic hydrides and the degradation of the Q-factor (the so-called Q-disease) has been well documented for the Nb cavities [66–69] and films [76–78]. Formation of metallic hydride precipitates from over-saturated solid solution of H interstitials is characteristic of Nb [79]. Hydrogen in the Nb cavities is usually absorbed during the standard buffer chemical polishing or electropolishing or machining, which dissolves or damages the passivating layer of dielectric pentoxide, Nb2 O5 , on the Nb surface shown in Fig. 2 [7, 80]. Experimental evidences of the contribution of grain boundaries to Ri have been inconclusive: in Refs. 70–72 the residual surface resistance of smallgrain Nb films was attributed to grain boundaries, while in Refs. 76–78 it was shown that under proper heat treatment of sputtered Nb films, Ri can be reduced down to the level of ∼1−10 nΩ, comparable to Ri of high-performance polycrystalline Nb cavities, and no significant dependence of Ri on the grain size was observed. It was also pointed out that Ri can

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

be increased by the surface roughness and absorption of noble gases on the Nb surface [76–78]. A correlation between Ri and crystalline disorder has been observed; for instance, Ri can be as low as a few nΩ in cavity-grade Nb with RBCS = 10−20 nΩ at 2 K and 1.3 GHz [6], while dirty Nb films can have much higher Ri [76–78, 81]. Overall, there are abundant experimental evidences that the intrinsic factors discussed above can significantly increase Ri , and yet one can pose a question whether there is a lower limit of Ri if all these factors have been eliminated by proper cavity treatment, for example by high-temperature (600–800◦C) annealing which removes hydrogen from Nb [7, 80]. The issue of the fundamental lower limit of Ri is important in order to understand how far the SRF performance of the best Nb cavities can be further improved or to evaluate other promising SRF materials, like Nb3 Sn [18–20]. The fact that Ri > 0 has been observed on other materials very different from Nb [13] suggests that the residual surface resistance may be a general feature of superconductors, yet the conventional model of weak and spatially uncorrelated impurity scattering in the BCS theory cannot explain Ri [8–11]. Indeed, Eq. (17) takes into account scattering on nonmagnetic impurities in the BCS model which

127

preserves the key property of the electronic density of states: N () = 0 for all energies || < ∆, resulting in Ri = 0. An insight into the physics behind Ri may come from the tunneling measurements of N () in superconductors [82–85]. These experiments have shown that N () differs from the density of states of the BCS model in which N () diverges at  = ∆ and vanishes at  < ∆, as shown by the black line in Fig. 5(a). In the observed N () the singularities are smeared out and subgap states with finite N () appear at energies below ∆, as shown by the red line. Such N () is often described by the phenomenological Dynes formula [82, 83]: Nn ( − iγ) . N () = Re  ( − iγ)2 − ∆2

(19)

Here γ is a damping parameter, which results in a finite density of states, N (0)  γNn /∆, at the Fermi level. Furthermore, N () extracted from the tunneling conductance measurements on the cavitygrade Nb [84, 85] also exhibits finite N () for  < ∆, as shown in Fig. 5(b). Interestingly, the lowtemperature baking reduces N () at  < ∆ and sharpens the peaks at  = ∆. The physics of the subgap states is not well understood; however, any finite density of states N (0) at the Fermi level ( = 0) gives rise to a residual

4

3.5

3

n

2.5

N(ε)/N

January 10, 2013

2

1.5

1

0.5

0 0

0.5

1

1.5

2

ε/∆

(a)

(b)

Fig. 5. (a) N () described by Eq. (19) for γ/∆ = 0.2 (red). The black line shows N () of the BCS model. (b) Tunneling spectra on cavity-grade Nb. The red line is a fit to the Shiba model [87], which assumes the presence of magnetic impurities at the surface. The spectra correspond to vacuum-backed (lower curve) and unbaked (upper curve) Nb (shifted for greater clarity). (Reproduced from Ref. 84.)

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

128

surface resistance, irrespective of particular microscopic mechanisms. In this case Ri may be estimated from Eq. (11) with σ1 of the order of the conductivity of a normal metal with the density of states N (0) on the Fermi level: σi ∼ γσn /∆. Hence, µ20 ω 2 λ3 σn γ . (20) ∆ Here the values Ri  10–20 nΩ observed on the large-grain niobium cavities at f = 1.47 GHz [43], λ = 40 nm and σn = 109 Ωm would correspond to γ/∆ ∼ 10−2 . Several models of the subgap states in N () at || < ∆ have been suggested in the literature. For instance, inelastic scattering of electrons on phonons [88] in the Eliashberg theory gives rise to an imaginary part of ∆ which plays a role similar to that of the parameter γ in Eq. (19). Other mechanisms include strong Coulomb correlations [89], local variations of the BCS coupling constant by impurities [90] or pairbreaking magnetic impurities taken into account in a more rigorous way than in the original Abrikosov–Gor’kov theory [87]. In other theories the tail in N () at  < ∆ results from spatial correlations in impurity scattering [90, 91]. The model of localized subgap states caused by magnetic impurities [86] was used to interpret the tunneling data shown in Fig. 5(b) and to calculate Rs , assuming strong spin flip scattering in the Shiba model. Of these mechanisms the inelastic electron–phonon scattering [88] decreases strongly at T  Tc , so it may be of lesser importance for the SRF cavities than the temperature-independent impurity scattering. As was discussed above, other models of Ri invoke extrinsic mechanisms such as trapped vortices, lossy hydrides or grain boundaries. Trapped vortices, which appear due to the cavity cooldown from the normal to the superconducting state, can result in a temperature-independent part of Rs and hotspots on the cavity surface [6, 61–65]. The contribution of trapped vortices can be reduced by better magnetic shielding of the cavity or by optimizing the cavity cooldown [64]. The grain boundaries in polycrystalline cavities could, in principle, contribute to Ri ; however, experimental data indicate that the effect of clean grain boundaries on the SRF performance of the Nb cavities is inessential. Indeed, the best SRF Nb cavities can almost reach the ultimate SRF breakdown field close to the thermodynamic critical field Bc ≈ 200 mT [24–26], at which the Ri ∼

density of screening current flowing across the grain boundaries on the cavity surface approaches the pairbreaking current density Jd [31]. This is inconsistent with the Josephson weak link model [70–73], in which the grain boundaries strongly depress the tunneling current density Jc as compared to Jd . Also, there seems to be no apparent correlation between the grain size and the SRF performance; some of the best-performing Nb cavities with the record high breakdown fields Hb  200 mT have small grains [25]. This, however, does not mean that the grain boundaries in Nb are always transparent to strong rf currents, because the current-blocking propensity of grain boundaries can be increased by segregation of impurities and nonsuperconducting precipitates (hydrides, or others), and thus would depend on the heat treatment of the cavities, as will be discussed below. The above consideration suggests that Ri in the best-performing Nb cavities may be dominated by bulk impurity scattering. Indeed, had Ri been dominated by islands of subgap states, the tunneling measurements would have shown a highly nonuniform N () with localized regions of enhanced N () at trapped vortices, grain boundaries, patches of metallic NbO or NbO2 suboxides beneath the pentoxide layer or hydride precipitates situated between regions with the ideal NBCS (). Instead, in addition to enhanced density of states in rf hotspots at structural defects [85], the subgap states always form a significant uniform background in N () due to either localized magnetic impurities in the surface oxide layer [84, 85] or the bulk contributions proposed in Refs. 89–91. The scenario that the lower limit of Ri is controlled by electron scattering on point atomic defects in the bulk is consistent with the results of irradiation of the Nb cavities with the 175 MeV protons, which significantly increased Ri but reduced RBCS [92]. The irradiation was performed in situ in a vacuum liquid helium cryostat, which apparently excluded intrinsic factors like hydride formation upon the proton irradiation. Understanding the physics of Ri is important for evaluating the performance limits of the SRF cavities.

3. Quality Factor Q(B) In the previous sections, the surface resistance at weak fields Bp  Bc was discussed. This section

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

outlines mechanisms which may contribute to the dependence of the quality factor Q(Bp ) on the rf field amplitude Bp (see Fig. 1). We start with the higher field part of the Q(Bp ) curve and consider what may control the maximum rf field which a superconductor can withstand. 3.1. Superheating field As Bp increases, the rf currents start breaking the Cooper pairs and driving quasiparticles out of equilibrium, which makes the electromagnetic response nonlinear and Rs dependent on Bp . The rf Meissner state then becomes unstable above the superheating field Bsh (T ) at which a transition to a dissipative normal or vortex states occurs, depending on the Ginzburg–Landau (GL) parameter, κGL √= λ/ξ. For type-I superconductors with κGL < 1/ 2 (like in Pb), the transition from the Meissner state to the normal state occurs at B = √Bsh . For type-II superconductors with κGL > 1/ 2, there are three transition fields: the lower critical field Bc1 (T ) = (φ0 /4πλ2 )[ln κGL + 0.5], the superheating field Bsh (T ) and the upper critical field Bc2 (T ) = φ0 /2πξ 2 , where φ0 = π
/e = 2.068 × 10−15 Wb is the magnetic flux quantum. Nonmagnetic impurities increase Bc2 but decrease Bc1 [31]. Clean Nb is a marginal type-II superconductor with λ ≈ ξ ≈ 40 nm and the highest Bc1 (0) ≈ 170 mT. Another important SRF field is the thermodynamic critical field Bc , which defines the condensation energy of the superconducting state Fn − Fs = At T ≈ Tc the GL theBc2 /2µ0 per unit volume. √ ory gives Bc = φ0 /2 2λξ, while at T = 0 the BCS theory gives Bc = (µ0 Nn )1/2 ∆. The field Bc can be extracted from the measurements of the specific heat jump at Tc , which yield Bc (0) ≈ 200 mT for Nb [27]. Weak nonmagnetic impurities do not affect Bc , according to the Anderson theorem [31, 87]. At B > Bc1 type-II superconductors undergo a transition from the Meissner state to the mixed state of quantized magnetic vortices, each vortex having a core of diameter  2ξ where ∆(r) is suppressed, surrounded by circulating supercurrents decaying over the length λ. Vortices carry the magnetic flux quantum φ0 and are spaced by a = 1.08(φ0 /B)1/2 in a hexagonal lattice with the areal vortex density nv = B/φ0 at Bc1 < B < Bc2 . Penetration and oscillation of vortices under the rf field gives rise to strong dissipation and the surface resistance of

129

the order of Rs in the normal state [93]. However, the Meissner state can remain metastable at higher fields, B > Bc1 up to the superheating field Bsh at which the Bean–Livingston surface barrier [34] for penetration of vortices disappears and the Meissner state becomes unstable. Thus, Bsh is the maximum magnetic field at which a type-II superconductor can remain in a true nondissipative state not altered by dissipative motion of vortices. Calculations of Bsh and the depairing current density Jd have a long history, starting from the pioneering works [36, 37] using the GL theory, which is applicable only at T ≈ Tc . The GL theory has shown that, as B reaches Bsh , the screening current density at the surface reaches Jd  Bc /µ0 λ, and the Meissner√state becomes unstable. It was found that Bsh = ( 5/3)Bc if κGL 1, Bsh = 2−1/4 κ−1/2 Bc if κGL  1, and Bsh ≈ 1.2Bc at κGL ≈ 1. Figure 6 shows two regions of the screening current density J(z): the nonlinear region of high J ∼ Bsh /µ0 λ at the surface where the instability due to pairbreaking effects occurs, and the region  λ away from the surface where J(z) becomes so small that the linear London equations (5)–(7) are valid. In the following we will see how this narrow nonlinear region (3–20 nm for Nb) can determine Bsh and, under certain conditions, the high-field Q-slope. To evaluate the performance limits of the Nb cavities, the GL results for Bsh (T ) have often been extrapolated to low T  Tc , where the GL theory becomes invalid. However, unlike Bsh (T ) near Tc , calculation of Bsh (T ) at T  Tc requires solving the

1 0.8

J(z)/J(0)

January 10, 2013

0.6 0.4 nonlinear region

0.2 0

0

0.5

1

1.5

2

2.5

z/λ Fig. 6. Distribution of screening current density J(z) at the surface at high rf fields Bp  Bsh . The pairbreaking effects which cause the instability of the Meissner state are most pronounced in the narrow nonlinear region at the surface.

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

130

BCS or Eilenberger equations [94], which are valid for all T but are much more complicated than the GL equations. The latter is important because manifestations of the pairbreaking effects in clean superconductors at T  Tc can be rather different from T ≈ Tc , as was shown long ago [32, 33]. The first calculation of Bsh (T ) for the entire temperature range 0 < T < Tc in the clean limit and κGL → ∞ was done by Galaiko [35], √ who obtained Bsh = 0.84Bc at T → 0 and Bsh = ( 5/3)Bc = 0.745Bc at T → Tc . Recent calculations of Bsh (T ) from the Eilenberger equations in the clean limit at κ 1 revealed a maximum in Bsh (T ) at low T [38]. This nonmonotonic dependence of Bsh (T ) shows that the behavior of Bsh (T ) at low T can hardly be extrapolated from the GL results near Tc . The effect of impurities on Bsh (T ) at T  Tc is unusual, because the clean limit in the s-wave superconductors (all SRF materials) at T  Tc is a rather singular case: for T = 0, the Meissner currents do not affect the superfluid density ns until the superfluid velocity vs = J/ns e reaches the critical value, vs = vc = ∆0 /pF [32, 33]. The latter results from the energy spectrum (p) of quasiparticles in the presence of a superfluid flow induced by external magnetic field or current [32, 33]: (p) = ±



(p2 /2m − EF )2 + ∆2 ∓ pF vs ,

(21)

where EF = p2F /2m is the Fermi energy, p is the quasiparticle momentum and ± correspond to the electron and hole branches, respectively. The last term of Eq. (21) describes the Doppler shift which tilts the spectrum (p) as shown in Fig. 7. As a result, the energy gap becomes anisotropic and no longer equal to ∆: the minimum gap g = ∆ − vs pF for the quasiparticles moving along the superflow is reduced, while the gap ∆ + vs pF for the quasiparticles moving against the superflow is increased. Here g (vs ) vanishes at the critical velocity vc = ∆/pF , so above the critical current density J > Jc = envc = en∆0 /pF , the superfluid density ns (vs ) rapidly drops to zero in a narrow region, vc < vs < 1.08vc , due to the pairbreaking effects [32]. Unlike the d-wave superconductors, the s-wave superconductors at T  Tc do not exhibit the nonlinear Meissner effect caused by the dependence of ns (J) on J [81, 95–97], and the superheating field is reached at vs > vc , which corresponds to a gapless state [98]. For κGL 1, the

1.5

1

0.5

ε(p)/EF

January 10, 2013

εg

0

−0.5

−1

−1.5 −1.5

−1

−0.5

0

0.5

1

1.5

p/pF Fig. 7. The quasiparticle spectrum defined by Eq. (21) in a current-carrying state. The red and the blue curves show electrons and hole branches for the positive (red) and the negative (blue) direction of current at vs pF . The minimum quasiparticle gap g = ∆ − vs pF is reduced as compared to ∆.

field Bg at which the gap g closes at T = 0 is given by [32, 39]  1/2 2 Bc ≈ 0.816Bc. (22) Bg = 3 Since Bs = 0.84Bc [35], the gapless state in the clean, large-κ limit occurs in a narrow field range, 0.97Bsh
B < Bsh . In the rf field, the spectrum (p, t) rocks between the blue and the red curves in Fig. 7. The surface resistance Rs ∝ exp[−g (J)/kB T ] is mostly due to thermal activation of quasiparticles across the minimum gap g (Bp ); thus, Rs becomes strongly dependent on the rf field amplitude at Bp  T Bc /Tc [100, 101], which in turn can give rise to a high-field Q-slope. Moreover, because g (Bp ) in the clean limit closes at the field Bg smaller than Bsh , the surface resistance at Bp > Bg greatly increases and becomes of the order of Rs in the normal state. Thus, the dc superheating field in the clean limit cannot be regarded as the maximum rf field at which the Meissner state exhibits weak rf dissipation even at T  Tc . However, nonmagnetic impurities can reduce the field dependence of Rs , which may allow one to eliminate the high-field Q-slope by proper materials modifications. Recently the effect of both nonmagnetic and magnetic impurities on Bsh (T ) for κGL 1 in the

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

entire temperature range 0 < T < Tc was addressed by solving the Eilenberger equations [39]. The results show that nonmagnetic impurities do not affect Bsh near Tc where Bsh = 0.745Bc and Bc is independent of the impurity concentration, in accordance with the Anderson theorem. However, at low temperatures Bsh (α) now has a maximum as a function of the nonmagnetic scattering parameter α = πξ0 /, and the maximum in Bsh (α) washes out as T decreases, as shown in Fig. 8. The overall effect of impurities on Bsh is weak, Bsh varying from 0.84Bc in the clean limit (α = 0) to Bsh  0.812Bc in the dirty limit (α = 20). By contrast, magnetic impurities cause the pairbreaking spin flip scattering, giving rise to strong suppression of both Tc and Bsh [39].

1.05 1.04 1.03 1.02 1.01 1 0.99

0

0.5

1

1.5

2

2.5

3

0.88 0.87 0.86 0.85 0.84 0.83 0.82 0.81

0

5

10

15

131

The nonmonotonic dependence of Bsh (α) in Fig. 8 results from a peculiar effect of screening current on the quasiparticle density of states N () shown in Fig. 9. In the clean limit α = 0 [Fig. 9(a)], the Meissner currents turn the singularity at  = ∆ into a finite peak and reduce the gap g in the quasiparticle spectrum. Here g is defined as the maximum energy at which N () = 0, where g is different from ∆ in the presence of current and in the absence of subgap states. In the clean limit, ∆2 is proportional to the superfluid density, which at T = 0 is independent of B up to Bg , defined by Eq. (22) [32, 33]. By contrast, g decreases as B increases and vanishes at B = Bg < Bsh , and so a clean superconductor at B = Bsh is in the gapless state with a finite density of states at the Fermi level, as shown by the red line in Fig. 9(a). This picture was illustrated above by Fig. 7 based on Eq. (21). Impurities change the effect of the magnetic field on N (, B), as shown in Fig. 9(b). Here the gap g also decreases as B increases, yet g remains finite at B = Bsh , as is evident from N (, Bsh ), shown by the red line. In this dirty limit impurities preserve the superconducting gapped state all the way to the superheating field. Thus, there is a critical density of impurities or, in other words, the scattering parameter αc above which the gap g at B = Bsh opens, so that a superconductor at B = Bsh is in the gapless state for α < αc and in a gapped state at α > αc . The gap g at B = Bsh opens at αc = 0.36, which means that the mean free path is smaller than c = πξ0 /αc ≈ 8.73ξ0 [39]. Notice that the appearance of the gap at B = Bsh is not accompanied by the reduction of Bsh , in fact, Bsh is slightly increased as compared to the clean limit (see Fig. 8). The effect of current flow on N () was first addressed by Fulde [98] and observed by tunneling measurements on Al nanowires [99]. The dependence of g on α at α > 0.36, shown in Fig. 9(c), can be approximated by the formula g = 0.56∆0 [tan−1 (0.63α + 1.38) − 1] to an accuracy better than 1.2%. The gap g (α) increases monotonically with α and approaches the maximum value g (∞) = 0.323∆0 for α → ∞.

20

Fig. 8. The superheating field as a function of the nonmagnetic scattering parameter α = πξ0 /. The top panel shows the evolution of the peak in Hsh (α) = Bsh /µ0 as the temperature changes. The bottom panel shows Hsh (α) in the extended range of α at T = 0. (Reproduced from Ref. 39.)

3.1.1. Nonlinear surface resistance, the high-field Q-slope and the baking effect The disorder-induced transition from the gapless to the gapped state at B = Bsh can have important

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

132

2

1.5

1

0.5

0 0

0.5

1

1.5

2

2.5

(a) 2 0.3

1.5

0.2 0.2

1

0.1 0.1

0.5

0 0 0

1

2

3

0

0

0.5

1

1.5

2

2.5

0

20

40

60

80

100

(b) Fig. 9. Effect of Meissner current on the normalized density of states ν() = N ()/Nn at different ratios of B/Bsh for: (a) the clean limit at α = 0; (b) the moderately dirty limit at α = 3.6. The red lines show the density of states at the superheating field. Figure 3(c) shows the gap g (α) at B = Bsh as a function of α. The gap g (α) opens at α > 0.36, as shown in the inset. (Reproduced from Ref. 39.)

implications for Rs at high rf fields Bp ∼ Bc . Here the Meissner rf currents reduce the quasiparticle gap g (J), giving rise to a nonlinear dependence of the surface resistance on Bp :   g (Bp ) . (23) Rs ∝ exp − kB T As the field increases, Rs (T, Bp ) becomes dependent on Bp at Bp  T Bc /Tc in the clean limit [100, 101]. Therefore, the dc superheating field in the clean limit has no relevance to the maximum rf breakdown field of the Meissner state. Calculation of the nonlinear Rs (B) from the microscopic theory requires solving complicated equations of nonequilibrium superconductivity which take into account

current pairbreaking, the effect of the rf field on the quasiparticle distribution function for the rocking quasiparticle spectrum shown in Fig. 7, and collisions of electrons with impurities and phonons [46]. Very little has been done in this direction. Qualitatively, the results of Ref. 39 suggest that nonmagnetic impurities can reduce the field dependence of Rs because they restore the gap g in the quasiparticle spectrum at B = Bsh , as shown in Fig. 9. In the dirty limit, α > 1, the gap g  0.3∆00 may therefore be big enough to ensure both the exponentially small Rs in Eq. (23) and the lack of quasiparticles generated by the rf field with
ω < g . As a result, alloying the material with nonmagnetic impurities could be beneficial for the

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

SRF performance since the impurities reduce the high-field Q(Bp )-slope. Moreover, Bsh ≈ 0.83Bc at α > 1 can now be regarded as a true maximum field amplitude at which the Meissner state can survive under the rf field. The above results may also pertain to the baking effect which significantly reduces the high-field Q-slope, as shown in Fig. 1. It has been observed that the high-field Q-slope is very sensitive to the properties of a thin, ∼20 nm surface layer. The Qslope can also be tuned reversibly by first anodizing the Nb cavity to restore the high-field Q-slope by the subsequent baking which suppresses the highfield Q-slope [42, 43]. Many groups have pointed out that the diffusive redistribution of impurities, particularly interstitial oxygen or hydrogen in this layer [55], may be behind the baking effect [23, 40–44]. The length L = (Dt)1/2 over which impurities diffuse from the oxide surface layer to the bulk during the time t gives L  17 nm for the interstitial oxygen at 120◦C and t = 48 h, taking the value of diffusivity D from Ref. 102. The baking effect was attributed to local suppression of ∆ by oxygen (the oxygen pollution model reviewed in Ref. 41). Detailed profiles of oxygen concentration for different baking temperatures and times were measured using depth-resolved X-ray scattering [103]. It was recently suggested, based on the positron annihilation spectroscopy data, that the baking effect may result from dissociation of hydrogen–vacancy complexes [44]. The extensive experimental data on the baking effect do not unambiguously point at a microscopic mechanism which would reduce the high-field Q-slope, not least because it is not quite clear why the high-field Q-slope appears in the first place. An insight into the baking effect may come from the results of Ref. 39, which suggest that the fielddependent Rs is determined by a thin (as compared to 2λ  80 nm) nonlinear region where the quasiparticle gap g in Eq. (23) is reduced by the rf currents, as shown in Fig. 6. If baking pollutes a thin (2–20 nm) nonlinear surface layer adjacent to the pentoxide layer with impurities, the gap g increases, reducing the field dependence of Rs and the high-field Qslope. Several features of this scenario are essential: (1) the increase of g by atomic crystalline disorder in the Meissner state at B ∼ Bsh is not really sensitive to the particular impurities, so diffusion of both likely candidates — interstitial oxygen and

133

hydrogen — could contribute; (2) the critical atomic concentration of impurities above which the gap g at Bp = Bsh opens, xg ∼ (0.33/8 ∗ 40)3 ∼ 10−7 at. %, is extremely small, so the nonlinear layer at the real Nb surface (like the one shown in Fig. 2) is at least in the moderately dirty limit for which g > 0 is still smaller than the maximum g = 0.33∆0 in the dirty limit   ξ0 . In this case pollution of this layer with a very small concentration, x ∼ 10−2 − 10−1 at. %, of any impurities would increase g and reduce the high-field Q-slope. Even such small concentrations of impurities diffusing in the nonlinear surface layer could contribute to the baking effect, although they can hardly be detected by the depth-resolved X-ray scattering [103]. Yet this technique revealed a significant (7–8 fold) increase of the concentration of interstitial oxygen in a 5–10 nm thick surface layer after 145◦ C baking for 4 h. The maximum concentration of O reached about 7–8 at % [103]. 3.2. Thermal feedback model of the SRF breakdown Small Rs (T ) in the SRF cavities ensures very low dissipation, but strong rf fields drive superconductors out of equilibrium. The simplest manifestation of nonequilibrium effects is heating due to transfer of the rf energy absorbed by quasiparticles to the crystalline lattice. Heating can make the Meissner state unstable because the exponential temperature dependence of Rs provides a strong positive feedback between the rf Joule power and heat transfer to the coolant, resulting in thermal runaway [104]. This mechanism has been incorporated in the thermal feedback model, which can describe the decrease of the quality factor Q(Bp ) with Bp and the cavity quench above the breakdown field Bb (see e.g. Refs. 6, 105 and 106). A simple analytical thermal feedback model [101] was proposed for a superconducting slab of thickness d with one side exposed to the rf field and the other side cooled by liquid helium at T = T0 = 2 K, as shown in Fig. 10. A steadystate temperature distribution T (z) across the slab is determined by the heat diffusion equation Bp2 dT d κ(T ) + 2 Rs (Tm , Bp )δ(z) = 0, dz dz 2µ0

(24)

where κ(T ) is the thermal conductivity across the cavity wall, and the second term describes the rf

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

134

heating source. The delta function δ(z) accounts for the rf heating localized in a narrow surface layer of thickness λ  d, much smaller than the cavity wall thickness  2−3 mm, so Rs (T ) is taken at the surface temperature Tm . The solution of Eq. (24) depends on two surface temperatures, Tm and Ts , determined from the boundary conditions: κdT /dz = −Rs Hp2 /2 at z = +0 and κdT /dz + αK (Ts , T0 )(Ts − T0 ) = 0 at z = d, where αK (Ts , T0 ) is the Kapitza thermal conductance between the outer cavity surface and the liquid helium. These boundary conditions give the following self-consistent equations for Tm and Ts :

E, J

T Tm

Ts T0

H(t)

0 0.25

d

z

 αK (Ts , T0 )(Ts − T0 )d =

1

2

Bp < Bb

Bp/B0

0.1

0.05

1.2

1.4

1.6

T /T m

1.8

0

11

10

10

10

9

10

0

0.2

κ(T )dT,

Bp2 Rs (Tm , Bp ) = αK (Ts , T0 )(Ts − T0 ). 2µ20

0.15

0 1

Tm

(25)

Ts

0.2

Q

January 10, 2013

0.4

0.6

0.8

1

B /B p

b

Fig. 10. A summary of the thermal feedback mode. Top panel: Temperature distribution across the cavity wall. Red depicts the thin, ∼ λ  d layer of rf dissipation. The temperature jump Ts − T0 at z = d is due to the Kapitza thermal resistance between the superconductor and the He coolant. Middle panel: The graphic solution to Eq. (28) for different ratios of Ri /RBCS (T0 ) = 0.1, 1, 5 (from top to bottom). The points 1 and 2 correspond to the stable and unstable solutions, respectively, and the circles show the breakdown fields. Here B02 = 2µ20 T0 καK /(κ + dαK )RBCS (T0 ). Bottom panel: The field-dependent quality factor Q(Bp ) = G/RBCS (Tm ), where Tm (Bp ) was calculated from Eq. (28) with Ri = 0, and Q(0) = 2 × 1010 . In this model Q(0)/Q(Bp ) = e = 2.718.

(26)

Equation (25) is obtained by integrating the constant heat flux q = −κ(T )dT /dz from z = 0 to z = d, and then using the boundary condition q = αK (Ts − T0 ) at z = d. Equation (26) is the conservation law: the rf power equals the heat flux to the coolant. The quality factor Q(Bp ) = G/Rs (Tm ) is determined by the dependence of the surface temperature Tm (Bp ) on the rf amplitude Bp , which can be obtained by numerically solving Eqs. (25) and (26). Here Rs (T ) has the strongest (exponential) dependence on T , while the Kapitza conductance αK (T, T0 ) is power laws of T and T0 , with αK ∝ T0n with n = 3−5 if T ≈ T0 , while κ(T ) for clean Nb has a phonon peak around 2 K [107–109]. Calculations of Q(Bp ) by solving Eq. (24) numerically and taking into account the temperature dependencies of the materials parameters of Nb and liquid helium are described in Refs. 6 and 106. For the Nb cavities operating at low temperatures, Eqs. (25) and (26) simplify significantly, so that explicit formulas for Q(Bp ) and the breakdown field Bb can be obtained. As will be clear below, Tm (Bp ) increases weakly with Bp and so the maximum overheating even at the breakdown field is much smaller than the coolant temperature: Tm − T0  T02 /βTc  T0 , since T02 /βTc = 0.18 K for T0 = 1.8 K, β = 1.9 and Tc = 9.2 K. In this case κ(T ) and αK (T ) in Eqs. (25) and (26) can be taken at T = T0 . Solving then Eq. (25) for Ts , and substituting the result into Eq. (26), yields the following

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

equation for Tm : Bp2 (Tm − T0 )καK Rs (Tm , Bp ) = , 2µ20 κ + dαK

(27)

This thermal balance equation simplifies for weak fields Bp  Bc at which Rs is independent of Bp . Using Rs from Eq. (18), we rewrite Eq. (27) in terms of Bp as a function of Tm : Bp2 =

2µ20 καK Tm (Tm − T0 ) . [Aω 2 exp(−∆/kB Tm ) + Tm Ri ](κ + αK d) (28)

The dependence of Tm (Bp ) can be understood from the graphical solution to Eq. (28) shown in the middle panel of Fig. 10. Here point 1 corresponds to a stable solution for which Tm (Bp ) increases as Bp increases, while point 2 corresponds to an unstable thermal equilibrium. The stable and the unstable solutions merge as Bp (Tm ) reaches the maximum at Tm = Tb , defining the breakdown field Bb = Bp (Tb ), above which thermal runaway occurs. Here Bb can be obtained from Eq. (28) and ∂Bp /∂Tm = 0. For Rs (T ) given by Eq. (18), the breakdown occurs at small overheating, θ = (Tm − T0 )/T0  1, for which RBCS (T ) can be approximated by RBCS (T ) = R0 exp(θ∆/kB T0 ), where R0 = RBCS (T0 ). Then Eq. (28) and ∂Bp /∂θ = 0 can be solved exactly for Ri = 0, giving θ = kB T0 /∆  1. If Ri < R0 , the maximum overheating Tp − T0 and the breakdown field Bb can be calculated taking into account the first-order correction in Ri /eR0  1:   Ri kB T02 1+ , (29) Tb − T0 = ∆ eR0 

2καK kB T02 Bb = µ0 e∆(κ + dαK )R0   Ri , × 1− 2eR0

1/2

(30)

where e = 2.718. In this model the field-dependent quality factor Q(Bp ) = G/Rs (Tm ) can be calculated substituting Tm (Bp ) into Rs (Tm ). An example of Q(Bp ) for the case Ri = 0 is shown in the bottom panel of Fig. 10. Here the Q(Bp ) curve has an infinite slope at the breakdown field, and Q(0)/Q(Bp ) = e. Equations (29) and (30) give explicit dependencies of the breakdown field on T0 and the rf frequency, for instance Bb ∝ 1/ω for the BCS surface resistance R0 given by Eq. (17), to the accuracy of

135

a slowly varying logarithmic factor ln1/2 (kB T0 /
ω) and a small correction due to the second term in the brackets. Equations (29) and (30) also show which parameters are most crucial for the SRF thermal breakdown. First of all, it is the strong temperature dependence of the BCS surface resistance which drives the SRF breakdown, the maximum uniform overheating being ∼0.1−0.2 K. Here Ri does not affect Bb and Tb much: for Ri = 5 nΩ and R0 = 20 nΩ for good Nb cavities at 1.8 K and 1.3 GHz [61], the residual resistance in Eq. (30) only reduces Bb by 5%. Notice that the cold state 1 in Fig. 10 at Bp < Bb is metastable and can be destroyed by a thermal fluctuation of finite magnitude δT  (2) (1) Tm − Tm which drives the state 1 to the unstable state 2 and eventually to the cavity quench. This behavior is a manifestation of the thermal bistability which can result in dynamic pattern formation during the cavity quench, such as propagation of thermal switching waves [104] along the cavity, uniform flux jumps [110] or dendritic propagation of hot filaments of magnetic flux [111–113] well known for superconducting wires and thin films under dc or ramping magnetic fields. Another feature of the thermal SRF breakdown is that the interplay of different temperature dependencies of Rs (T0 ), κ(T0 ) and αK (T0 ) in Eq. (30) can result in a nonmonotonic behavior of Bp (T0 ). Here Rs (T0 ) is given by Eq. (18), while αK (T0 )  200 · T0n W/m2 K, where n = 4.65 and T0 in K [107, 108] is controlled by emission of thermal phonons to the superfluid He. Thermal conductivity κ(T ) generally decreases as T0 decreases, but high-purity Nb exhibits a phonon peak κ  20−30 W/mK around 2 K [109] due to the interplay of phonon and electron contributions [114]. Below 2 K, κ(T0 ) is mostly dominated by scattering of phonons on grain boundaries and other crystalline defects, resulting in κ(T0 ) ∝ T03 . The temperature dependence of Bb (T0 ) can be calculated using a more general equation obtained from Eq. (28) for an arbitrary ratio Ri /R0 :  2 Ri Bb Bb0 = 2 ln , (31) eR0 Bb0 Bb where Bb0 (T0 ) is the breakdown field at Ri = 0 given by Eq. (30). Shown in Fig. 11 is an example of Bb (T0 ) in the case of high thermal conductivity for which the thermal impedance of the cavity wall is limited by the Kapitza resistance, κ αK d.

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

136

1.1 1

Bb/B max

January 10, 2013

0.9 0.8 0.7 0.6 0.5

1

1.5

2

2.5

3

T0, K Fig. 11. Temperature dependence of Bp (T0 ) calculated from Eq. (31) for Nb with Ri = 0.1R0 at 2 K and assuming that κ  hαK , and αK (T0 ) ∝ T0n with n = 4.65.

Here the drop in Bb (T0 ) for T0 < 1.5 K results from the effect of the residual resistance, Ri > RBCS (T0 ). For T > 2 K, the BCS resistance dominates so that Bb  µ0 [2αK kB T02 /e∆RBCS (T0 )]1/2 ∝ (3+n)/2 exp(βTc /2T0 ). The latter expression is minT0 imum at Tm = βTc /(3 + n) ≈ 2.3 K (for β = 1.9, n = 4.65, Tc = 9.2 K), consistent with what is shown in Fig. 11. In this case the cavity would be more stable at the temperature T0 ≈ 1.5 K, lower than the conventional T0 ≈ 1.8−2 K. Notice that, for the Nb cavities operating at 2K, the overheating by only 0.17 K drives the superfluid He through the lambda point Tλ = 2.17 K, above which the thermal Kapitza resistance increases. This can further increase the overheating and give rise to the high-field Q-slope or ignite thermal breakdown. We now evaluate Bb for cavity-grade Nb at 2 K, taking αK = 5 · 103 W/m2 K, κ = 20 W/mK, ∆/kB = 17.5 K, Rs (2 K) = 20 nΩ, and d = 3 mm. Then Eq. (30) gives Bb ≈ 200 mT, close to Bc of Nb at 0 K. This indicates that, for the chosen materials parameters, thermal breakdown does not play a major role, and the Q(Bp ) curve is mostly determined by the isothermal surface resistance Rs (Bp ) in the Meissner state. The purity of Nb, the conditions of the surface facing the coolant and the thickness of the cavity wall are also important [109]. For the high-purity Nb at 2 K, the term αK d = 15 W/mK in the denominator of Eq. (30) is not much smaller than κ = 20 W/mK, so the thermal impedance of the cavity wall is limited by both the Kapitza resistance and the thermal conductivity. If Bb  Bc , the SRF performance of the best Nb cavities is primarily

determined by the fundamental physics of the surface resistance in strong rf fields, while thermal breakdown plays a lesser role. In this case the intrinsic field dependencies of Rs due to the kinetics of nonequilibrium quasiparticles and rf pairbreaking manifest themselves at fields lower than the fields at which the global heating becomes important. The situation can change for dirtier Nb or other superconductors like Nb3 Sn. For low-purity Nb, the surface resistance may not change much, but the thermal conductivity can drop to κ  5 W/mK [109]. In this case heat transfer through the cavity wall is mostly limited by thermal conductivity, and Bp  132 mT becomes smaller than Bc . As a result, the rf heating can degrade the Q-factor; so the use of highpurity Nb significantly improved the SRF cavities [6]. As far as the thermal stability is concerned, the Nb purity is beneficial, but Rs is affected by impurities in a 40 nm surface layer which does not change the thermal impedance of the cavity wall. Thermal stability can become a problem for “better” superconductors like Nb3 Sn. For instance, Eq. (18) suggests that higher Tc ≈ 18 K of Nb3 Sn would yield much lower Rs as compared to Nb. Measurements of Q(Bp ) have shown that the Qfactor at low field and 2 K is indeed some 2–5 times higher than Q of Nb (see Fig. 12), indicating that Rs ∼ 5 nΩ. However, because κ  10−2 W/mK of Nb3 Sn at 2 K [17] is some three orders of magnitude lower than for Nb [17], the thermal breakdown field Bb ∼ 10 mT for bulk Nb3 Sn cavities turns out to be much lower than both Bc  540 mT of Nb3 Sn and Bb  200 mT of the high-purity Nb. Moreover, even

Fig. 12. The quality factor Q as a function of the peak electric field for Nb3 Sn for 2 K and 4.2 K. (Reproduced from Ref. 18.)

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

a 2–3-µm-thick Nb3 Sn film deposited on the inner surface of the Cu cavity can significantly increase the thermal impedance RT = αK + dNb3 Sn /κNb3 Sn + dCu /κCu of the cavity wall. For instance, if dNb3 Sn = 3 µm, κNb3 Sn = 10−2 W/mK [17], dCu = 3 mm and κCu = 103 W/mK at 2 K, the thermal impedance of the Nb3 Sn film dNb3 Sn /κNb3 Sn = 3 × 10−4 W/K is ∼ 102 higher than for the Cu wall. For the Nb3 Sn on the Nb cavity, with dNb = 3 mm and κNb = 20 W/mK, the thermal impedance of the Nb3 Sn film is twice that of the Nb cavity wall. Thus, the rf overheating in Nb3 Sn can become a more serious problem than for Nb, suggesting that effective heat transfer management is required to reveal the potential of Nb3 Sn in the SRF cavities. Some possibilities of doing so will be discussed below. 3.3. Hotspots on the cavity surface The uniform thermal SRF breakdown is an idealization of a more complicated behavior of the SRF cavities which often exhibit localized sources of the rf dissipation (hotspots). Such hotspots have been observed in many thermal map measurements using arrays of thermometers placed on the outer cavity surface [61, 115]. An example of such thermal maps shown in Fig. 13 reveals characteristic scales and magnitudes of hotspots: local overheating

by δT  0.1−0.7 K spread over ∼1 cm along the cavity surface. The thermal map experiments pose the following questions: (1) What is the effect of hotspots on the global Q(Bp ) curve? (2) To what extent do hotspots reduce the breakdown field by igniting local thermal quench which then propagates along the cavity surface? (3) What are the physical and materials mechanisms behind the hotspots and how do they affect the size and magnitude of the hotspots? Temperature and spatial scales of hotspots can be described by the 2D thermal diffusion equation for T (x, y) along the cavity surface: dκ∇2 T − γ(T − T0 ) + q(T, r) = 0.

(32)

Here T (x, y) is averaged over the cavity thickness, γ = αK /(1 + dαK /κ) is the effective thermal impedance of the cavity wall, and the rf power q = Rs (T, Hp , r)Hp2 /2 can depend not only on T but also on the position r along the surface. Equation (32) reduces to the heat balance equation (27) if T is uniform, but it also describes distribution of T (x, y) around a region of enhanced q(r) by a defect. Such defects could be grain boundaries, which facilitate local vortex penetration, lossy precipitates like hydrides, nonuniform patches of suboxides or surface steps which can cause local field enhancement [6]. Here q(r, T ) = q0 (T ) + δq(r, T ), where q0 (T ) is the uniform rf power and δq(r, T ) is an extra power caused by a defect. The solutions T (x, y) of the highly nonlinear equation (32) exist only below the local breakdown field Bbl smaller than the uniform breakdown field Bb . If Bp > Bbl , the hotspot ignites thermal quench [104] propagating along the cavity, as was observed in Ref. 116. Numerical calculations of T (x, y) around localized heat sources with a lower breakdown field and the thermal instabilities described by the nonlinear heat diffusion equation similar to Eq. (32) were performed in Ref. 117. It is easier to understand the behavior of weak hotspots for which T (x, y) = Tm + δT (r), where the background temperature Tm is given by Eq. (28), and a small temperature disturbance δT (r) < T02 /Tc, is described by the linearized equation (32): dκ∇2 δT − (γ − q0 )δT + δq(Tm , r) = 0,

Fig. 13. Hotspots on the surface of a Nb cavity imaged by temperature mapping. (Reproduced from Ref. 61.)

137

(33)

where q0 = Hp2 Rs /2, the prime denotes a derivative over T , and all parameters are taken at T = Tm (Bp ). The radially symmetric solution to Eq. (33) for a

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

138

small defect, r0  Lh = (dκ/γ)1/2 , is

r Γ K0 , r > d. (34) δT (r) = 2πdκ L Here Γ is the extra power produced by a defect. The maximum δT (0) on the outer cavity surface is estimated by cutting off δT (r) in Eq. (34) at r ∼ d: δT (0) 

κ Γ ln , 4πdκ d(γ − q0 )

r < d.

(35)

At large distances r > L from a hotspot δT (r) drops exponentially over the length L: 1/2  Lh dκ 2 L=  = + d . (36) , L h αK 1 − q0 /γ Here the thermal diffusion length Lh gives the size of the lateral temperature distribution from a localized heat source in the absence of the rf heating. For d = 3 mm, κ = 20 W/mK and αK = 1 kW/m2 K, Eq. (36) yields Lh = 8.3 mm. Because of the field-dependent factor 1 − q0 /γ, which accounts for uniform rf heating, the radius of a hotspot L(Bp ) increases as the rf field amplitude increases, L(Bp ) diverging if Bp → Bb . For Bp ≈Bp , it follows from Eq. (36) that L(Bp )  Lh / 1 − (Bp /Bb )2 . The hotspot radius L(Bp ) can therefore be much greater than either the defect size r0 or the thermal length Lh . For example, L = 3.2Lh  26.6 mm for Bp = 0.95Bb and the parameters used above, so even a small defect can cause a hotspot much greater than the wall thickness d, consistent with the thermal maps shown in Fig. 13. The expansion of hotspots as the rf field increases was also observed in thermal map experiments [43]. The hotspot expansion and the local temperature increase as Bp increases can result in stronger rf dissipation at higher fields [101], which manifests itself in a significant high-field Qslope as compared to Q(Bp ) of the uniform thermal feedback model represented in the bottom panel of Fig. 10. 3.4. Mechanisms of hotspots Hotspots are usually ascribed to certain structural features of the SRF cavities which cause enhanced local dissipation. Microstructural features are associated with such defects as grain boundaries in polycrystalline Nb cavities, surface topography which causes local rf field enhancement, and lossy nonsuperconducting precipitates, particularly hydrides

or metallic suboxides. Hotspots can also be caused by bundles of trapped vortices oscillating under the rf fields. Because of the extremely high quality factors, Q ∼ 1010 −1011 , the Nb cavities become susceptible to defects present in any solid but usually negligible in other power applications of superconductors in dc magnets, motors, generators or transmission lines [21]. In the following we discuss how these defects can result in hotspots in the SRF cavities.

3.4.1. Grain boundaries Grain boundaries (GBs) are very common planar defects which subdivide the material into misoriented crystalline regions (see e.g. the review Ref. 118). They can impede current flow, particularly in such materials as the high-Tc cuprates in which the critical current density through the GB drops exponentially as the misorientation angle θ increases [118], so a typical spread of θ ∼ 40o in polycrystals can reduce Jgb by 2–3 orders of magnitude. Recent experiments revealed similar weak-linked GBs in pnictides [118]. The current-limiting GBs in cuprates are a serious obstacle to power applications because, instead of flowing along the wire, current breaks into disconnected loops circulating in the grains. GBs can potentially be serious performancelimiting factors for the SRF cavities, given that the magnitudes of screening currents at the surface approach the depairing limit, Jd  Hc /λ. Thus, if GBs in a polycrystalline Nb are the Josephson weak links for which Jgb  Jd , the extremely high Meissner current densities flowing along the inner cavity surface cannot pass through GBs, in which case the oscillating rf field would result in magnetic flux penetrating in and out of the GB network, which is accompanied by significant rf losses. Thus, the assumption that GBs in Nb behave as weak links [70–73] seems inconsistent with the fact that the high-performance Nb cavities can operate almost at the depairing limit with the quality factors Q ∼ 1010 (see Fig. 1). Moreover, no significant effect of the grain size on Q(Bp ), Ri and Bb has been observed, some of the best-performing Nb cavities being smallgrain polycrystalline cavities [24–26]. The fact that GBs in polycrystalline Nb seem transparent to the extremely high rf currents Jgb  Jd may result from the large coherence length, ξ = 40 nm, of Nb, which also lacks the features

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

of cuprates causing the weak-linked GBs: low carrier density, strong sensitivity to the oxygen nonstoichiometry, large Thomas–Fermi screening length and proximity of the superconducting state to nonsuperconducting (antiferromagnetic) phases which precipitate on GBs [22]. However, segregation of oxide or hydride nanoprecipitates, or impurities on GBs (the so-called Cottrell atmospheres) [118, 119], could degrade Jgb , turning GBs in Nb into weak links, although this seems not to be the case for the best small-grain Nb cavities. The situation may change for other SRF materials like Nb3 Sn, in which GBs may be less transparent to strong rf currents than in Nb. Indeed, GBs in Nb3 Sn are known to be effective pinning centers for vortices [21, 120], which implies local suppression of superconductivity at the GBs and thus Jgb  Jd . For strong-linked GBs, dissipation can come from penetration of the mixed Abrikosov–Josephson vortices at a field not much lower than Hc1 [121]. 3.4.2. Hydride and oxide precipitates, dislocations The complex structure of the Nb surface shown in Fig. 2 can contribute to the surface resistance [59], because islands of thin metallic suboxides NbO and NbO2 could potentially be a source of ohmic losses in strong rf fields. However, because the thickness of the suboxide layer, 1 nm, is much smaller than both ξ and the proximity length ξN ∼ ξTc /T , the suboxides become superconducting due to the proximity effect [31]. This greatly reduces the rf dissipation [122] and increases the local superheating field nearly to its bulk value [123]. Hotspots can also result from clusters of lossy niobium hydride precipitates (with typical lateral sizes ∼1–10 µm) [124, 125] which can form during chemical polishing or electropolishing of the Nb cavities. The metallic hydrides cause the so-called Q-disease — a significant reduction of Q and the breakdown field [66–69]. The treatment usually involves high-temperature (600–800◦C) annealing followed by a quick cavity cooldown [6, 7]. Recently it was suggested that hotspots can result from networks of dislocations at the surface [126]. Usually individual dislocations do not locally suppress superconductivity in clean Nb [120], for the same reasons that GBs (which can be regarded as a chain of edge or screw dislocations [118]) are not

139

weak links. It is, however, possible that dislocation networks which appear due to local plastic deformation of the Nb cavities can be contaminated by the Cottrell atmospheres of segregated impurities or preferential diffusion of impurities (such as oxygen, hydrogen or carbon) along the dislocation cores [118]. This can result in stronger local suppression of superconductivity, turning dislocations into sites which can produce extra rf dissipation or trap oscillating vortices, as will be discussed below. 3.4.3. Surface topography and pits The typical surface roughness of the Nb cavities on the scale of a few microns [127, 128] is macroscopic, as far as superconductivity is concerned. In this case the screening sheet current density I(r) = H (r) follows the local surface profile, resulting in local enhancement of the parallel magnetic field H (r) at the tips of the surface protuberances [6]. For instance, a semispherical protuberance on the cavity surface increases the local magnetic field at the tip by 50% [4], which implies a local reduction of the rf breakdown field for a given Eacc . In this macroscopic approach, the local magnetic field H(r) = µ−1 0 ∇ × A outside the superconductor is described by the Laplace equation ∇2 A = 0 for the magnetic vector potential A with the boundary condition of zero normal field component B⊥ = ∇⊥ × A = 0 at the surface. The topography studies have revealed a spectrum of surface features with a broad distribution of scales, from tens of microns to tens of nanometers [127, 128]. Here the Laplace equation cannot be used near sharp corners or protuberances with the curvature radius of the order of (or smaller than) the size of the Cooper pair, ∼ ξ = 40 nm. In this case the general equation (3) yields a nonlocal integral relation between J(r) and A(r), which cuts off the singularities in J(r) and H(r) at sharp corners. Surface defects of another type, which have recently attracted much attention as sources of local rf dissipation, are pits on the cavity surface (usually near the weld edges). There are experimental evidences that pits with typical sizes of ∼200−500 µm can result in hotspots which can trigger the cavity quench [116, 129]. Yet the solution to the Laplace equation [4] shows that the local field inside an ideal semispherical pit is reduced as compared to the applied field, except for the mathematically sharp

January 10, 2013

13:7

140

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

edges where J(r) and H(r) are singular. If, however, the edges of a pit on the real Nb surface are rounded on the scale of λ  40 nm, much smaller than the pit size, the local J(r) and H(r) are reduced everywhere in the pit, which thus becomes a coldspot. It is quite possible that pits can become hotspots because of networks of dislocations caused by local plastic deformation at the inner surface, or contamination with impurities. The latter seems consistent with the tunneling studies [85]. In a way, this situation may be similar to the grain boundaries: noncontaminated grain boundaries and pits in Nb seem to be not causing extra rf dissipation. However, segregation of impurities or non-superconducting precipitates (like hydrides) at pits and grain boundaries can turn them into hotspots which can degrade the performance of the Nb cavities.

3.4.4. Vortex hotspots Vortices in the SRF cavities are mostly generated by temperature gradients or external fields B > Bc1 (T ) during the cavity cooldown through Tc when the lower critical field Bc1 (T ) vanishes. For instance, the unscreened Earth field HE  0.4 Oe can generate vortices spaced by a  (φ0 /µ0 HE )1/2  7 µm. Since Bc1 (T )  Bc1 (0)(1−T 2 /Tc2) increases as T decreases (Bc1 (0)  170 mT for Nb), the subsequent cooldown to T = 2 K at which H  Hc1 (T ) makes vortex lines thermodynamically unstable, forcing them to escape through the cavity surface. However, in doing so a fraction of vortices can get trapped by nonsuperconducting precipitates (hydrides or oxides) or networks of dislocations or grain boundaries, giving rise to pinned vortex bundles. Dissipation due to trapped vortices in the Nb cavities was first addressed in Ref. 62, where it was also pointed out that the vortex dissipation can result in a residual resistance. The subsequent studies have shown that the performance of the Nb cavities and Ri is indeed improved by screening the Earth magnetic field [6, 65, 78]. However, vortices generated by thermal gradients during cavity cooldown are more difficult to control. The model of Ref. 62 pointed out the importance of vortex dissipation in the SRF cavities but did not take into account some essential features of trapped vortices which were addressed recently in Ref. 63. Some of these features are illustrated by

Fig. 14. Configurations of trapped vortices: vortex line going through the cavity wall (1); vortex semiloop at the inner surface (2); part of the vortex loop going in the bulk (3) [63].

Fig. 14, which shows two types of trapped vortices: (1) vortex lines going through the cavity wall; (2) vortex loops starting and ending at the inner cavity surface exposed to the rf field; (3) vortex segments at the inner surface. Here, only tips of vortices coming out from the inner surface or fraction of vortex loops spaced by
2λ from the surface are exposed to the rf field. Calculation of the rf power qv using a solution to a dynamic equation for a pinned flexible vortex segment driven by the rf currents gives the frequency dependence of qv (ω), which depends on the orientation and the length of these segments and the distance from the surface, as shown in Fig. 14. The hotspots caused by trapped vortices can be distinguished experimentally from the hotspots due to fixed materials features like suboxide patches or arrays of lossy hydride precipitates, or any defects or impurities which cause local variation of the BCS surface resistance. Vortices can be depinned by currents which exert the Lorentz force fL = −φ0 [n × J] or by thermal gradients ∇T which exert the force fT = −s∗ ∇T per unit length, where s∗ (T )  −φ0 ∂Hc1 /∂T is the transport entropy carried by quasiparticles in the vortex core [120]. The vortex hotspot can therefore be moved or changed by applying thermal gradients caused by heaters placed on the outer cavity surface [130] or by laser-beamscanning the inner cavity surface [132]. Vortices get depinned if fT exceeds the pinning force fp = φ0 Jc , where Jc is the critical current density. The first experiments [130, 132] have shown that thermal maps do change after the application of thermal gradients ∼1 K/mm, which proves that some of the

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

hotspots are due to trapped vortices even though the cavities were screened from the Earth magnetic field. It should be noted that the thermal gradient cannot eliminate all vortex hotspots; for example, it can depin and annihilate the loop 2 in Fig. 14, but it can only shift the vortices 1 and 3 along the cavity surface. Yet, identifying vortex hotspots suggests that the cavity performance can be improved by reducing the density of defects (particularly at the inner cavity surface) which can trap vortices as they exit the cavity upon the cooldown. Generation of vortices by thermal gradients is a complex process which cannot be ascribed to the magnetic fields produced by thermoelectric currents. The reason is that thermoelectric effects vanish in the Meissner state [133] and yet magnetic flux can be generated in bimetallic superconducting/normal structures which can be relevant to the Nb on Cu cavities. In addition to the trapped vortices, significant rf dissipation can occur at high fields Bp  Bc1 as vortices start moving in and out of a superconductor by breaking through places on the surface where the Bean–Livingston barrier [34] oscillating with the rf field is locally weakened by defects. This mechanism can produce a high-field Q-slope, and since the peak current density at the surface J  Hc1 /λ is of the order of Jd , the velocity vp ∼ 10 km/s of penetrating vortices exceeds the speed of sound. Thus, the characteristic penetration time τ0 ∼ vp /λ ∼ 10−11 s is much shorter than the rf period [63]. Such superfast penetration of single vortices can result in dynamic instabilities [131] and strong overheating, which in turn can trigger irreversible dendritic penetration of hot filaments of magnetic flux through the Meissner state at the breakdown field [63]. Such macroscopic magnetothermal flux jumps determined by diffusion of heat and magnetic flux are much slower than the superfast penetration of single vortices. For flux jumps and dendritic flux penetration in bulk superconductors, the propagation time τf ∼ 1−102 µs is much larger than the rf period, although τf can be shorter in thin films [110–113]. 4. Other SRF Materials and Multilayer Coating The development of the Nb cavities operating almost at the depairing limit [24–26] poses the question whether further significant increase of the SRF accelerating gradients is possible. This task

Table 1. Material Nb Nb3 Sn NbN MgB2 B0.6 K0.4 BiO3 Ba0.6 K0.4 Fe2 As2

141

Possible SRF materials.

Tc (K)

Bc (mT)

Bc1 (mT)

λ (nm)

9.2 18 16.2 40 31 38

200 540 230 320 440 500

170 40 20 20–60 30 30

40 85 200 140 160 200

obviously requires superconductors with the superheating fields and critical temperatures higher than for Nb. Many such materials exist, but all of them are type-II superconductors with Bc1 lower than for Nb, as illustrated in Table 1. Thus, Nb turns out to be best-protected against dissipative penetration of vortices, indicating that other superconductors with lower Bc1 may not allow the achievement of accelerating gradients and high-field quality factors better than Nb. Figure 12 shows the Q(Eacc ) curves for a 1.3 µm thick Nb3 Sn film deposited on the inner surface of a high quality Nb cavity [18], which has Tc and Bc more than twice higher than for Nb. Here Q(Eacc ) at low fields is higher than for Nb (see Fig. 1), consistent with a lower RBCS of Nb3 Sn due to its higher Tc which is nearly twice of Tc of Nb. However, as Eacc increases, Q(Eacc ) drops at the peak magnetic field close to Bc1 for Nb3 Sn. At higher Eacc , the Q-factors are at least by an order of magnitude worse than for the Nb cavities. This suggests that vortices start penetrating in and out of the Nb3 Sn cavities at Bp  Bc1 in a few places of the cavity surface where the Bean–Livingston barrier [34] is locally reduced by structural defects [63], as was discussed above. Since single vortices penetrate almost instantaneously over the rf period [63], the onset of the Q drop in Fig. 12 appears to be consistent with the condition that the peak magnetic field exceeds the dc lower critical field: Bp  Bc1 . It was suggested [47] that the low Bc1 problem for non-Nb, high-gradient SRF accelerators could be addressed if the Nb cavities are coated with multilayers consisting of thin layers of high-Bc superconductors sandwiched between thin dielectric layers, as shown in Fig. 15. Here the thickness of high-Bc superconducting layers is smaller than λ of that material, and the thickness of the dielectric can be a few nm, to suppress the Josephson coupling between the superconducting layers. This idea was based on

13:7

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

142

20

Hc1/Hc1b

15

Nb

10

5

0

0

1

2

3

4

5

d/λ Fig. 16. Parallel Hc1 as a function of the film thickness d calculated for κGL = 20 using the results of Refs. 134 and 135. The dashed line shows Hc1 (d) given by Eq. (38).

1 ds

0.8

B/Ba

January 10, 2013

di

0.6

Bi/Ba

0.4 0.2 0

S I S 0

1

Bulk

I 2

3

4

5

6

x/λb Fig. 15. Top: Multilayer coating consisting of thin (d < λ) layers of superconductors with high Bsh separated by thin dielectric layers, on the inner surface of the Nb cavity. Bottom: Profile of a screened magnetic field in a multilayer.

the work of Abrikosov, who showed that the parallel Bc1 in a film of thickness d < λ is greatly increased as compared to the bulk Bc1b as the vortex currents get squished by the film surfaces [134, 135]:   λ φ0 + 0.497 , d λ, (37) ln Bc1b = 4πλ2 ξ   d 2φ0 Bc1 = − 0.07 , d  λ. (38) ln πd2 ξ Here Bc1 at d  λ depends weakly on the coating material, so getting Bc1 > 200 mT requires d
102 nm if ξ = 5 nm. A multilayer with d
30−50 nm could screen the peak rf field of 500 mT and Eacc > 100 MV/m in such a way that the local field at the interface with Nb would be smaller than the bulk Bc1 of Nb, as shown in Fig. 16. The significant increase

of the parallel Bc1 in thin films has been observed on Nb films [81, 136]. The Nb cavity plays a significant role here since the thick Nb substrate suppresses nucleation of short perpendicular vortices in the thin coating layers, which otherwise could appear due to very low per⊥ because of a big demagnetizing factor pendicular Bc1 [137]. Nb also provides good heat transfer through the cavity wall, reducing the rf overheating due to the poor thermal conductivity of materials like Nb3 Sn or NbN. In this regard the multilayer coating [47] is principally different from the Cu cavities coated with thick (a few µm) Nb films [76–78] or thick Nb3 Sn films [6, 19] which do not increase the parallel Bc1 as compared to its bulk value. Moreover, thicker Nb3 Sn films with d > 1 µm can significantly increase the thermal impedance of the cavity wall, deteriorating thermal stability and reducing the breakdown field, as was discussed above. In that regard, thin film multilayers with d = 10–100 nm may be more beneficial as that they would cause weaker overheating than thick film Nb3 Sn coating. The maximum Bp at which the multilayers remain in the Meissner state is determined by the condition that the screening current density in the layers reaches the pairbreaking limit. Thus, the maximum breakdown field is limited by the superheating field of the coating material, which potentially opens up an opportunity of reaching Eacc  100 MV/m with Nb3 Sn multilayers, as it is evident from Table 1. The choice of d for the multilayer can be made using a general dependence of Bc1 (d) shown in Fig. 16.

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

Notice that, while even moderately thick films, d  λ, can increase Bc1 for a particular material, the main purpose of the multilayers is to increase Bc1 (d) above 200 mT. The latter can only be done with thin (d < λ) layers given that Nb has the highest Bc1 . Achieving Eacc  100 MV/m thus requires even thinner layers.

143

or superconducting pnictides [16]. Developing highquality films and addressing the physics and materials challenges in understanding the role of impurities at high fields, the effect of grain boundaries (particularly in Nb3 Sn) or band decoupling in the two-band MgB2 at high fields [141] will be required to realize the full potential of the multilayer technology in particle accelerators.

5. Conclusions and Outlook After more than 50 years of research and development, the Nb technology has been able to produce the SRF cavities operating at the theoretical limit, yet many physics and materials science mechanisms behind the field dependence of the quality factor of the Nb cavities are still not well understood, including theories of nonequilibrium superconductivity and the role of impurities at high rf fields. Several materials treatments of the Nb cavities, such as low-temperature baking, increasing the purity of Nb, and reducing the rf dissipation caused by hydrides or trapped vortices, turned out to be very successful in boosting the SRF performance [6, 7]. Yet understanding the physics and materials science of the 100-nm-thick surface layer can bring new ways of decreasing Rs by the proper management of surface impurities and oxide layers, since the optimum SRF performance may require a certain degree of materials disorder. The scientific challenges are quite formidable, given the goal of improving the already extremely high quality factors, Q > 1010 , of the Nb cavities. This requires manipulation of a delicate materials nanoscale structure in the 100 nm surface layer while dealing with the materials defects which can turn their SRF behavior from beneficial to benign or deadly, depending on the heat treatment of impurity management. The use of multilayer coating may offer an opportunity to break the “Nb monopoly” in the SRF cavities by taking advantage of the many available superconductors with higher Bc and lower Rs . Developing the in situ coating technology requires new approaches, such as atomic layer deposition, which has already demonstrated conformal coating of the inner surface of the Nb cavity with a thin Al2 O3 layer which improved the SRF performance [139, 140]. Among possible multilayer materials are Nb3 Sn, which has been studied in cavity applications [18–20], or NbN or MgB2 [14, 15, 138],

Acknowledgments I am grateful to G. Ciovati and C. Antoine for useful discussions and comments. References [1] A. P. Banford and G. H. Stafford, Plasma Phys. (J. Nucl. Energy, Part C ) 3, 287 (1961). [2] W. M. Fairbank, J. M. Pierce and P. B. Wilson, in Proc. 8th Int. Conf. on Low Temp. Phys. (London, 1962) (Butterworths, Washington DC, 1963), p. 324. [3] P. B. Wislon and H. A. Schwettman, IEEE Trans. Nucl. Sci. 12, 1045 (1965). [4] J. D. Jackson, Classical Electrodynamics, 3rd ed. (Wiley, 1989). [5] D. Proch, Rep. Prog. Phys. 61, 431 (1998). [6] H. Padamsee, J. Knobloch and T. Hays, RF Superconductivity for Accelerators, 2nd ed. (Wiley, 2007). [7] C. Z. Antoine, Materials and Surface Aspects in the Development of SRF Niobium Cavities (EuCARDBOO-2012-001). [8] D. C. Mattis and J. Bardeen, Phys. Rev. 111, 412 (1958). [9] A. A. Abrikosov, L. P. Gorkov and I. M. Khalatnikov, JETP 9, 636 (1959). [10] C. B. Nam, Phys. Rev. 156, 470, 487 (1967). [11] J. P. Turneaure, J. Halbritter and H. A. Schwettman, J. Supercond. 4, 341 (1991). [12] J. P. Turneaure and I. Weissman, J. Appl. Phys. 39, 4417 (1968). [13] M. A. Hein, Microwave Properties of Superconductors. NATO ASI Series 375, 21–53 (2002). [14] X. X. Xi, Rep. Prog. Phys. 71, 116501 (2008). [15] E. W. Collings, M. D. Sumption and T. Tajima, Supercond. Sci. Technol. 17, S595 (2004). [16] D. C. Johnston, Adv. Phys. 59, 803 (2010). [17] G. D. Cody and R. L. Cohen, Rev. Mod. Phys. 36, 121 (1964). [18] G. Muller, P. Kneisel, D. Mansen, H. Piel, J. Pouryamout and R. W. Roeth, in Proc. 5th EPAC (London, 1985), p. 2085. [19] A. Andreone, A. Cassinese, A. Di Chiara, M. Iavarone, F. Palomba, A. Ruosi and R. Vaglio, J. Appl. Phys. 82, 1736 (1997).

January 10, 2013

13:7

144

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

[20] S. M. Deambrosis, G. Keppel, V. Ramazzo, C. Roncolato, R. G. Sharma and V. Palmieri, Physica C 441, 108 (2006). [21] D. Larbalestier, A. Gurevich, M. Feldmann and A. Polyanskii, Nature 414, 368 (2001). [22] A. Gurevich, Nat. Mater. 10, 255 (2011). [23] G. Ciovati, G. Myneni, F. Stevie, P. Maheshawari and D. Griffis, Phys. Rev. ST Accel. Beams 13, 022002 (2010). [24] F. Furuta, K. Saito, T. Saekia, H. Inouea, Y. Morozumia, T. Higoa, Y. Higashia, H. Matsumotoa, S. Kazakova, H. Yamaokaa, K. Uenoa, Y. Kobayashia, R. S. Orra and J. Sekutowicz, in Proc. 2006 European Particle Accelerator Conference (Edinburgh, Scotland, 2006), eds. C. Biscari (INFN-LNF), H. Owen (CCLRC DL), Ch. Petit-Jean-Genaz (CERN), J. Poole (CERN) and J. Thomason (CCLRC RAL), paper MOPLS084. [25] R. L. Geng, Physica C 441, 145 (2006). [26] R. L. Geng, G. V. Eremeev, H. Padamsee and V. D. Shemelin, in Proc. PAC07 (1997) p. 2337. [27] J. P. Carbotte, Rev. Mod. Phys. 62, 1027 (1990). [28] F. Marsiglio, J. P. Carbotte, R. Akis, D. Achkir and M. Poirier, Phys. Rev. B 50, 7204 (1994). [29] T. Mori, E. J. Nicol, S. Shiizuka, K. Kuniyasu, T. Nojima, N. Toyama and J. P. Carbotte, Phys. Rev. B 77, 174515 (2008). [30] W. Lee, D. Rainer and W. Zimmermann, Physica C 159, 535 (1989). [31] M. Tinkham, Introduction to Superconductivity (McGraw-Hill, 1975). [32] R. H. Parmenter, RCA Rev. 26, 323 (1962). [33] J. Bardeen, Rev. Mod. Phys. 34, 667 (1962). [34] C. P. Bean and J. D. Livingston, Phys. Rev. Lett. 12, 14 (1964). [35] V. P. Galaiko, JETP 23, 475 (1966). [36] J. Matricon and D. Saint-James, Phys. Lett. A 24, 241 (1967). [37] S. J. Chapman, SIAM J. Appl. Math. 55, 1233 (1995). [38] G. Catelani and J. P. Sethna, Phys. Rev. B 78, 224509 (2008). [39] P.-J. Lin and A. Gurevich, Phys. Rev. B 85, 054513 (2012). [40] B. Visentin, in Proc. 11th Workshop on RF Superconductivity (Travemunde, 2003), ed. D. Proch (DESY, Hamburg, Germany, 2004), TuO01. [41] G. Ciovati, Physica C 441, 44 (2006). [42] G. Eremeev and H. Padamsee, Physica C 441, 62 (2006). [43] G. Ciovati, P. Kneisel and A. Gurevich, Phys. Rev. ST Accel. Beams 10, 022002 (2007). [44] B. Visemtin, M. F. Barthe, V. Moineau and P. Desgardin, Phys. Rev. ST Accel. Beams 13, 052002 (2010). [45] G. E. H. Reuter and E. H. Sondheimer, Proc. R. Soc. London A 195, 336 (1948).

[46] N. B. Kopnin, Theory of Nonequilibrium Superconductivity (Oxford University Press, 2001). [47] A. Gurevich, Appl. Phys. Lett. 88, 012511 (2006). [48] H. A. Schwettman, J. P. Turneaure and R. F. Waites, J. Appl. Phys. 45, 914 (1974). [49] C. M. Lyneis, H. A. Schwettman and J. P. Turneaure, Appl. Phys. Lett. 31, 541 (1977). [50] R. L. Geng, in Proc. 2003 Particle Accelerator Conference (Portland, Oregon, USA, 2003), p. 264. [51] A. T. Wu, D. Gu and H. Baumgart, in Proc. SRF 2009 (Berlin, Germany), TUPPO041, p. 300. [52] H. Hahn and H. J. Halama, J. Appl. Phys. 47, 4629 (1976). [53] A. Dacca, G. Gemma, L. Mattera and R. Parodi, Appl. Surf. Sci. 126, 219 (1998). [54] Q. Ma, P. Ryan, J. W. Freeland and R. A. Rosenberg, J. Appl. Phys. 96, 7675 (2004). [55] M. Delheusy, A. Stierle, N. Kasper, R. P. Kurta, A. Vlad, H. Dosch, C. Antoine, A. Resta, E. Lundgren and J. Andersen, Appl. Phys. Lett. 92, 101911 (2008). [56] J. T. Sebastian, D. N. Seidman, K. E. Yoon, P. Bauer, T. Reid, C. Boffo and J. Norem, Physica C 441, 70 (2006). [57] K. E. Yoon, D. E. Seidman, C. Antoine and P. Bauer, Appl. Phys. Lett. 93, 132502 (2008). [58] R. Tao, R. Todorovic, J. Liu, R. J. Meyer, A. Arnold, W. Walkosz, A. Romanenko and L. D. Cooley, J. Appl. Phys. 110, 124313 (2011). [59] F. L. Palmer, R. E. Kirby, F. K. King and E. L. Garwin, Nucl. Instrum. Mater. Phys. Res. A 279, 321 (1990). [60] J. Zmuidzinas, Rev. Cond. Mat. Phys. 3, 169 (2012). [61] G. Ciovati, J. Appl. Phys. 96, 1591 (2004). [62] M. Rabinovitz, J. Appl. Phys. 42, 88 (1971). [63] A. Gurevich and G. Ciovati, Phys. Rev. B 77, 104501 (2008) [64] O. Kugeler, A. Neumann, W. Anders and J. Knobloch, Rev. Sci. Instrum. 81, 074701 (2010). [65] S. Aull, O. Kugeler and J. Knobloch, Phys. Rev. ST Accel. Beams 15, 062001 (2012). [66] S. Isagawa, J. Appl. Phys. 51, 6010 (1980). [67] B. Bonin and R. W. R¨ oth, in Proc. 5th Workshop on RF Superconductivity, ed. D. Proch (Hamburg, Germany, 1991), p. 210. [68] K. Saito and P. Kneisel, in Proc. 5th Workshop on RF Superconductivity, ed. D. Proch (Hamburg, Germany, 1991), p. 665. [69] R. W. R¨ oth, in Proc. 6th Workshop on RF Superconductivity (SEBAF, Newport News, 1993), p. 1126. [70] B. Bonin and H. Safa, Supercond. Sci. Technol. 4, 257 (1991). [71] C. Attanasio, L. Maritato and R. Vagio, Phys. Rev. B 43, 6128 (1991).

January 10, 2013

13:7

WSPC/253-RAST : SPI-J100

1230005

Superconducting Radio-Frequency Fundamentals for Particle Accelerators

[72] A. Andreone, A. Cassinese, M. Iavarone, R. Vaglio, I. I. Kulik and V. Palmieri, Phys. Rev. B 52, 4473 (1995). [73] J. Halbritter, J. Supercond. 10, 91 (1997). [74] K. Scharnberg, J. Appl. Phys. 48, 3462 (1977). [75] J. Halbritter, J. Appl. Phys. 58, 1320 (1985). [76] C. Benvenuti, S. Calatrone, I. E. Campisi, P. Darriulat, M. A. Peck, R. Russo and A.-M. Valente, Physica C 316, 153 (1999). [77] C. Benvenuti, S. Calatrone, P. Darriulat, M. A. Peck, A.-M. Valente and C. A. Van ’t Hof, Physica C 351, 421 (2001). [78] S. Calatroni, Physica C 441, 95 (2006). [79] H. K. Birnbaum, M. L. Crossbeck and M. Amano, J. Less-Common Metals 49, 357 (1976). [80] R. E. Ricker and G. R. Myneni, J. Res. Natl. Inst. Stand. Technol. 115, 353 (2010). [81] N. Groll, A. Gurevich and I. Chiorescu, Phys. Rev. B 81, 020504(R) (2010). [82] R. C. Dynes, V. Narayanamurti and J. P. Carno, Phys. Rev. Lett. 41, 1509 (1978). [83] R. C. Dynes, J. P. Carno, J. P. Hertel and T. P. Orlando, Phys. Rev. Lett. 53, 2437 (1984). [84] T. Proslier, J. F. Zasadzinskii, J. Moore, L. Cooley, C. Antoine, M. Pellin, J. Norem and K. Gray, Appl. Phys. Lett. 92, 212505 (2008). [85] T. Proslier, J. F. Zasadzinskii, L. Cooley, M. Pellin, J. Norem, J. Elam, C. Z. Antoine, R. A. Rimmer and P. Kneisel, IEEE Trans. Appl. Supercond. 19, 1404 (2009). [86] M. Kharitonov, T. Proslier, A. Glatz and M. J. Pellin, Phys. Rev. B 86, 024514 (2012). [87] A. V. Balatskii, I. Vekhter and J.-X. Zhu, Rev. Mod. Phys. 78, 373 (2006). [88] T. P. Deveraux and D. Belitz, Phys. Rev. B 44, 4587 (1999). [89] D. A. Browne, K. Levin and K. A. Muttalib, Phys. Rev. Lett. 58, 156 (1987). [90] A. I. Larkin and Yu. N. Ovchinnikov, JETP 34, 1144 (1972). [91] J. S. Meyer and B. D. Simons, Phys. Rev. B 64, 134516 (2001). [92] H. J. Halama, Appl. Phys. Lett. 19, 90 (1971). [93] M. W. Coffey and J. R. Clem, Phys. Rev. B 45, 9872 (1992). [94] G. Eilenberger, Z. Phys. 214, 195 (1968). [95] D. Xu, S. K. Yip and J. A. Sauls, Phys. Rev. B 51, 16233 (1995). [96] T. Dahm and D. J. Scalapino, J. Appl. Phys. 81, 2002 (1997); Phys. Rev. B 60, 13125 (1999). [97] M.-R. Li, P. J. Hirschfeld and P. Wolfle, Phys. Rev. Lett. 81, 5640 (1998); Phys. Rev. B 61, 648 (2000). [98] P. Fulde, Phys. Rev. 137, A783 (1965). [99] A. Anthore, H. Pothier and D. Esteve, Phys. Rev. Lett. 90, 127001 (2003). [100] I. O. Kulik and V. Palmieri, in Proc. 1987 Workshop on RF Superconductivity (Padova, Italy, 1987), p. 309.

145

[101] A. Gurevich, Physica C 441, 38 (2006). [102] R. W. Powers and M. V. Doyle, J. Appl. Phys. 30, 514 (1959). [103] M. Delheusy, Ph.D. thesis (Paris-Sud XI and Stuttgart Universities, 2008). [104] A. V. Gurevich and R. G. Mints, Rev. Mod. Phys. 59, 941 (1987). [105] B. Visentin, in Proc. 11th Workshop on RF Superconductivity (Lubeck/Travemunde, 2003). [106] R. R¨ oth et al., in Proc. 1992 European Particle Accelerator Conference, eds. E. H. Henke et al. (Editions Frontiers, 1992), p. 1325. [107] H. Mittag, Cryogenics 13, 94 (1973). [108] S. Van Sciver, Helium Cryogenics (Plenum, New York, 1986). [109] A. Boucheffa, M. X. Fran¸cois and F. Koechlin, Cryogenics 34, 297 (1994). [110] R. G. Mints and A. L. Rakhmanov, Rev. Mod. Phys. 53, 551 (1981). [111] I. Aranson, A. Gurevich and V. Vinokur, Phys. Rev. Lett. 87, 0670031 (2001). [112] I. S. Aranson, A. Gurevich, M. S. Welling, R. J. Wijngaarden, V. K. Vlasko-Vlasov, V. M. Vinokur and U. Welp, Phys. Rev. Lett. 94, 0370021 (2005). [113] E. Altshuler and T. H. Johansen, Rev. Mod. Phys. 76, 471 (2004). [114] N. W. Ashkroft and N. D. Mermin, Solid State Physics (Brooks/Cole, Belmont, USA, 1976). [115] J. Knobloch, H. Miller and H. Padamsee, Rev. Sci. Instrum. 65, 3521 (1994). [116] M. S. Champion, L. D. Cooley, C. M. Ginsburg, D. A. Sergatskov, R. L. Geng, H. Hayono, Y. Iwashita and Y. Tajima, IEEE Trans. Appl. Supercond. 19, 1184 (2009). [117] H. Padamsee, IEEE Trans. Magn. 19, 1322 (1983). [118] J. H. Durrell, C.-B. Eom, A. Gurevich, E. E. Hellstrom, C. Tarantini, A. Yamamoto and D. C. Larbalestier, Rep. Prog. Phys. 74, 124511 (2011). [119] X. Song, G. Daniels, D. M. Feldmann, A. Gurevich and D. Larbalestier, Nat. Mater. 4, 470 (2005). [120] A. M. Campbell and J. E. Evetts, Adv. Phys. 21, 199 (1972). [121] A. Gurevich, Phys. Rev. B 65, 214531 (2002). [122] G. Fischer, R. Klein and J. P. McEvoy, Phys. Rev. Lett. 19, 193 (1965). [123] A. L. Fauch´ere and G. Blatter, Phys. Rev. B 56, 14102 (1997). [124] A. Romanenko, G. Wu, L. D. Cooley and G. Ciovati, in Proc. SRF2011, THPO008 (2011). [125] A. Romanenko and F. Barkov, SRF Materials Workshop (Jefferson Lab., 2012). [126] A. Romanenko and H. Padamsee, Supercond. Sci. Technol. 23, 045008 (2010). [127] H. Tian, G. Rubeill, C. Xu, C. E. Reece and M. J. Kelley, Appl. Surf. Sci. 257, 4781 (2011). [128] C. Xu, H. Tian, C. E. Reece and M. J. Kelley, Phys. Rev. ST Accel. Beams 15, 043502 (2012).

January 10, 2013

13:7

146

WSPC/253-RAST : SPI-J100

1230005

A. Gurevich

[129] M. Ge, G. Wu, D. Burk, J. Ozelis, E. Harms, D. Sergatskov, D. Hicks and L. D. Cooley, Supercond. Sci. Technol. 24, 035002 (2011). [130] G. Ciovati and A. Gurevich, Phys. Rev. ST Accel. Beams 11, 122001 (2008). [131] A. I. Larkin and Yu. N. Ovchinnikov, in Nonequilibrium Superconductivity, eds. D. N. Langenberg and A. I. Larkin (North-Holland, Amsterdam, 1986), p. 493. [132] G. Ciovati, S. Anlage, C. Baldwin, G. Cheng, R. Flood, K. Jordan, P. Kneisel, M. Morrone, G. Nemes, L. Turlington, H. Wang, K. Wilson and S. Zhang, Rev. Sci. Instrum. 83, 034704 (2012). [133] D. J. Van Harlingen, Physica B 109 & 110, 1710 (1982). [134] A. A. Abrikosov, JETP 19, 988 (1964). [135] G. Stejic, A. Gurevich, E. Kadyrov, D. Christen, R. Joynt and D. C. Larbalestier, Phys. Rev. B 49, 1247 (1994).

Alex Gurevich is currently a Professor of Physics at Old Dominion University. He got his Ph.D. degree from the USSR Academy of Sciences in 1984. He has held staff research positions at the Institute of High Temperatures in Moscow, the University of Wisconsin–Madison and the National High Magnetic Field Laboratory in Tallahassee, Florida. In 1989– 1990 he worked for two years as a Humboldt scholar at the Nuclear Research Center in Karlsruhe, Germany. Prof. Gurevich has been working on different theoretical aspects of condensed matter physics, superconductivity, materials science and rf properties of superconductors in particle accelerators. He is the author of two books and more than 150 journal papers, and also a Fellow of the American Physical Society.

[136] C. Z. Antoine, S. Berry, S. Bouat, J.-F. Jacquot, J.-C. Villegier, G. Lamura and A. Gurevich, Phys. Rev. ST Accel. Beams 13, 121001 (2010). [137] E. H. Brandt, Rep. Prog. Phys. 58, 1456 (1995). [138] T. Tajima et al., in Proc. Linear Accelerator Conf. LINAC 2010 (Tsukuba, Japan, 2010), p. 854. [139] M. J. Pellin, T. Proslier, J. Norem, J. Zasadzinski, P. Kneisel, R. Rimmer, L. Cooley and C. Antoine, in Proc. Int. Conf. on RF Superconductivity (Beijing, 2007). [140] T. Proslier, J. Zasadzinski, J. Moore, M. Pellin, J. Elam, L. Cooley, C. Antoine, J. Norem and K. E. Gray, Appl. Phys. Lett. 93, 192504 (2008). [141] A. Gurevich and V. M. Vinokur, Phys. Rev. Lett. 97, 137003 (2006).

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Reviews of Accelerator Science and Technology Vol. 5 (2012) 147–184 c World Scientific Publishing Company
DOI: 10.1142/S179362681230006X

Superconducting Radio-Frequency Systems for High-β Particle Accelerators Sergey Belomestnykh Collider-Accelerator Department, Brookhaven National Laboratory, Upton, NY 11973-5000, USA [email protected] This article addresses the physics and engineering of the superconducting radio-frequency systems for high-β particle accelerators. I consider different geometries for cavities, discuss criteria for optimization, and options for higher-order mode damping. In reviewing recent progress in the field, SRF systems are classified according to the functions they perform. These systems are fundamental RF accelerating systems, deflecting/crab cavities, harmonic RF systems, and SRF photoemission electron guns. Keywords: RF superconductivity; accelerating structure; higher-order mode; deflecting cavity; SRF photoinjector.

1. Introduction

material for making SRF cavities. Other articles in this volume of Reviews of Accelerator Science and Technology address the fundamental principles of RF superconductivity [5], SRF technology R&D [4], and applications of SRF to low-β accelerators [6]. Much of the progress in the field of SRF for highβ was driven by two “frontiers”: the high-energy frontier and the high-luminosity frontier [7]. International R&D on a high-energy electron–positron linear collider led to the development of the highly successful nine-cell superconducting TESLA cavities [8] and their associated technology. Coordinated efforts first on the TESLA and then on the International Linear Collider (ILC) enabled major advances in understanding the requirements for reaching very high accelerating gradients, and in improving the cavity fabrication and preparation techniques. This work culminated in achieving surface fields close to the fundamental limit of the critical RF magnetic field in re-entrant single-cell cavities [9] and secured steady improvement in the effort of attaining high gradients in the production of multicell cavities. On the other hand, application of SRF cavities to high-beamcurrent storage rings (colliders and light sources) [10] required addressing several different issues, viz. delivering very high RF power to beams with amperelevel average currents and ensuring strong damping of the parasitic higher-order modes (HOMs). Both challenges were met, so that nowadays SRF is the

Over the last three decades, the science and technology of applying radio-frequency (RF) superconductivity to particle accelerators has progressed, evolved, and matured tremendously. Employing superconducting RF (SRF) improved the performance of many accelerators and supported the building of new ones by enabling applications that were not previously possible. The advantages of the SRF technology have been discussed many times before (see e.g. Refs. 1 and 2). Here I want to emphasize that the most attractive features of applying this technology to particle accelerators lie in its high accelerating gradient, Eacc , achievable in continuous wave (CW) and long-pulse operating modes, and extremely low RF losses in the cavity walls at cryogenic temperatures. This article focuses on reviewing recent progress in applying SRF to near–velocity-of-light, or high-β, particle accelerators. More specifically, I have limited this review to β > 0.8. For a historical overview and details on older installations, readers are referred to the excellent previous review article [3], the textbooks [1, 2], and the bibliographies therein. The SRF structures discussed below are made either of bulk niobium or of copper with a thin niobium film sputtered on to it. While many laboratories are exploring other superconducting materials, none of the alternatives can compete yet with niobium as the best 147

December 29, 2012

17:31

148

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

technology of choice for many new storage rings. Another important factor in convincing the accelerator community of the practicality of SRF machines was the achievement of high reliability in long-term CW operation of large SRF installations — CEBAF and LEP. The number and the variety of high-β applications are growing rapidly. As new types of applications appear, they impose different requirements and, consequently, often present new challenges. Previous solutions do not always work. Examples of such new uses of SRF at high-β are compact deflecting/crab cavities and SRF photoinjectors: both are very active fields of research, where new ideas have been proposed in the last few years. The SRF system of a particle accelerator encompasses many subsystems and components tightly coupled together and to the accelerator as a whole. The machine’s parameters determine most of the system’s higher-level requirements. However, there are other factors, technology-dependent ones and otherwise, that influence the system design: optimizing the accelerator’s cost, maximizing achievable RF power through a fundamental input coupler, meeting the requirements for extracting HOM power, etc. In many cases the requirements and imposed limitations compete and a systems engineering approach must be used to generate a successful, efficient design. The following are among the subsystems and components of an SRF linac: cavities (both RF and mechanical aspects), fundamental RF input power couplers (FPCs), HOM dampers, frequency tuners, cryostats, high- and low-level RF systems, instrumentation, vacuum system, and cryogenics. As it is practically impossible to consider all aspects of the SRF systems in one article, I have chosen to discuss only two of them: cavity geometries and optimization criteria; options for higher-order mode damping. Reviews of various subsystems and components can be found in the proceedings of the SRF, LINAC, and Particle Accelerator Conferences, available on the JACoW website [11]. In particular, the fundamental power couplers were reviewed in Refs. 12–18. Also, the article [4] provides a comprehensive review of bulk niobium material R&D, cavity fabrication, surface treatments, and cavity diagnostics. It discusses alternative materials to bulk Nb. Research on SRF fundamentals is covered in Refs. 4 and 5.

Before proceeding further, I remind readers of the definitions of several important parameters, or figures of merit, that I use throughout this article. The cavity accelerating voltage Vc is the ratio of the maximum energy gain that a particle moving along the cavity axis can achieve to the charge of that particle. As all existing high-β multicell SRF structures operate in a π standing wave mode, the optimal length of the cavity cells (active length) is βλ/2. Then, the accelerating gradient is defined as the ratio of the accelerating voltage per cell to the cell length, or Eacc = Vc /(βλ/2). Accelerating cavities have beam tubes attached to the end cells. Also, there might be some other elements between the cavities and between the cryomodules, such as HOM dampers, vacuum gate valves, and bellows. Thus, the total length of the SRF linac is much greater than its active length. Then, the real estate gradient is defined as the ratio of the linac’s energy gain to its total length, and the fill factor is the ratio of the accelerating gradient to the real estate gradient (or the linac’s active length to its total length). The cavity quality factor Q0 tells us how many RF cycles (multiplied by 2π) are required to dissipate the energy stored in the cavity. The quality factor can be derived as a ratio of two values: the geometry factor, G, and the surface resistivity, Rs . As its name suggests, the geometry factor is determined only by the shape of the cavity, and hence is useful for comparing cavities with different shapes. Surface resistivity depends only on material properties and frequency. The cavity’s shunt impedance determines how much acceleration a particle can get for a given power dissipation in the cavity Pc : Rsh = Vc2 /Pc . A related quantity is the geometric shunt impedance Rsh /Q0 , or simply R/Q, which depends only on the cavity’s shape. Finally, there are two other important figures of merit: the ratios of the peak surface electric and magnetic fields to the accelerating gradient — Epk /Eacc and Bpk /Eacc . A high surface electric field can cause field emission of electrons, thereby increasing the heating of the cavity wall and prematurely quenching superconductivity, and generating dark current in the accelerator. A high surface magnetic field may limit the cavity’s performance at high gradients if it approaches the fundamental limit of the critical RF magnetic field.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

This article is organized as follows. In Secs. 2 and 3, I discuss two issues: cavity geometries and optimization; options for HOM damping. In subsequent sections, I review recent progress in the field. For this I classify SRF systems according to their performed functions: fundamental RF systems; deflecting/crab cavities; harmonic systems for manipulating the beam in phase space; SRF photoemission electron guns. I conclude with a brief summary. Tables 1 and 2 each compile a list of high-β SRF systems: decommissioned; operational; under construction or active R&D (i.e. cavities are being fabricated and tested). 2. Cavity Geometries Used in High-β SRF Structures The evolution of high-β superconducting accelerating structures resulted in cavities with elliptical cell shapes [19], which are dominating the field. The design and limitations of such cavities are well understood. However, recently, several new, nonelliptical, cavity shapes have been proposed for high-energy proton storage rings, deflecting/crab systems, compact electron linacs, and SRF electron guns. First, I consider elliptical cavities in detail, and then I review other structures.

A typical SRF accelerating structure for linacs consists of a chain of resonant cells coupled via iris apertures, forming a standing wave structure. The beam tubes attached to the end cells allow particles to pass through the structure and also couple RF power to the cavity at the fundamental-mode frequency and extract higher-order mode power that the beam excites. For high-current storage rings, where requirements to damp HOMs are especially demanding, single-cell cavities are often used. Figure 2 illustrates single-cell and multicell elliptical geometries. Figure 3 shows fields along the cell profile line and the locations of peak surface fields for the inner half-cell of a multicell cavity. Since cavities are designed for different applications, they are optimized to satisfy different criteria, so necessitating tradeoffs in their design. Consider, for example, two cavities operating in the CW regime: the 1497 MHz seven-cell low-loss (LL) cavity for the CEBAF 12 GeV Upgrade [24] and the 500 MHz single-cell CESR superconducting B-cell cavity [25]. The first was designed to improve upon the original Cornell-shape (OC) five-cell CEBAF cavities and was optimized to assure low cryogenic losses. The RF power loss in the cavity walls is given by Vc2 · Rs , (1) G · R/Q where Vc is the accelerating voltage, Rs the surface resistivity, G the cavity geometry factor, and R/Q the geometric shunt impedance. The product of the latter two parameters was used for optimizing the cavity because both parameters depend only on the cavity’s shape. The resulting geometry has a noticeably different equatorial region from the OC cavity, viz. more vertical walls and a smaller iris. As it is evident from Table 3, the parameters of the new design are better than those of the original CEBAF cavity. Two more cells were added to the LL cavity to improve the linac fill factor; this was possible due to low beam current in CEBAF and, hence, its relaxed requirements for HOM damping. The second cavity, the CESR B-cell, was optimized to provide very strong HOM damping as required to support high beam currents in a storage ring. It has the same wall slope angle of 105◦ as the OC cavity, but to allow all higher-order modes to propagate out of the cell and achieve exceptional damping, the beam pipe was made very large, with Pc =

2.1. Elliptical cavities The terms “elliptical cavity” and “elliptical shape” are descriptions of the cavity’s profile line, which consists of several (typically two) elliptical arcs with half-axes A, B, a, and b connected by a straight line l (Fig. 1). Req is the radius at the cavity’s equator, and Ra is the radius of the iris aperture. The cell shape becomes re-entrant when the wall’s inclination angle α is less than 90◦ . The elliptical shapes were developed to address several specific limitations facing SRF technology. An equatorial arc serves two purposes. First, this shape suppresses multipacting [21], which limited the performance of pillbox cavities. Later [22] it was understood that using an arc of optimal shape assures uniformity of the distribution of the magnetic field along the surface, and thus reduces its peak value. This, in turn, leads to higher accelerating gradients and lower RF losses. Using elliptical arcs at the cavity’s iris lowers the peak surface electric field [23], so alleviating possible field emission.

149

Country

Accelerator type Linac, CW Storage ring/collider Storage ring/collider Storage ring/collider

S-DALINAC CEBAF JLab FEL FLASH A0 Photoinjector ELBE ERL Injector ALICE SNS KEK-B CESR TLS CLS BEPC-II Diamond SSRF JAEA ERL-FEL LHC SOLEIL

Darmstadt U. JLab JLab DESY Fermilab HZDR Cornell U. ASTeC ORNL KEK Cornell U. NSRRC Canadian Light Source IHEP Diamond Light Source SINAP JAEA CERN Synchrotron SOLEIL

Germany USA USA Germany USA Germany USA UK USA Japan USA Taiwan Canada China UK China Japan Switzerland France

Recirculating linac Recirculating linac Energy recovery linac Linac, pulsed Linac, pulsed Linac, CW Linac, CW Energy recovery linac Linac, pulsed Storage ring/collider Storage ring/collider/light source Storage ring/light source Storage ring/light source Storage ring/collider/light source Storage ring/light source Storage ring/light source Energy recovery linac Storage ring/collider Storage ring/light source

CEBAF Upgrade European XFEL Compact ERL PKU-SETF ASTA R&D ERL CeC PoP at RHIC PLS-II NSLS-II TPS RHIC

JLab DESY KEK Peking U. Fermilab BNL BNL PAL BNL NSRRC BNL

USA Germany Japan China USA USA USA Korea USA Taiwan USA

Recirculating linac Linac, pulsed Energy recovery linac Energy recovery linac Linac, pulsed Energy recovery linac Linac, CW Storage ring/light source Storage ring/light source Storage ring/light source Storage ring/collider

Cornell ERL ILC Project X SNS Upgrade Project X

Cornell U. — Fermilab ORNL Fermilab

USA — USA USA USA

Energy recovery linac Linear collider Linac, pulsed Linac, pulsed Linac, CW

6 32 16 288

1300 509 500 352

19.4 47 19.2 491

Decommissioned Decommissioned Decommissioned Decommissioned

12 338 26 56 1 6 5 4 48 (β = 0.81) 8 4 1 1 2 2 3 4 16 4

2998 1497 1497 1300 1300 1300 1300 1300 805 509 500 500 500 500 500 500 500 400 352

10.3 169 14.6 58 1.0 6.2 1.2 4.2 43 2.4 1.2 0.3 0.3 0.6 0.6 0.9 3.6 6 1.7

Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational Operational

+80 640 2 2 24 1 1 2 2 (4) 2 1

1497 1300 1300 1300 1300 704 704 500 500 500 56

56 664 2.1 2.1 25 1.1 1.1 0.6 0.6 (1.2) 0.6 —

Construction Construction Construction Construction Construction Construction Construction Construction Construction Construction Construction

384 15,764 224 +36 152 (β = 0.9)

1300 1300 1300 805 650

313 16,360 232 40 158

R&D R&D R&D R&D R&D

1230006

USA Japan Germany Switzerland

WSPC/253-RAST : SPI-J100

Stanford U. KEK DESY CERN

Status

S. Belomestnykh

SCA TRISTAN HERA LEP

Number of cavities Frequency (MHz) Active length (m)

17:31

Laboratory

Fundamental SRF at high-β accelerators around the world.

December 29, 2012

Accelerator

150

Table 1.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators Table 2. Accelerator

Laboratory

Deflecting/crab cavities: PS CERN KEK-B KEK APS ANL LHC Upgrade CERN Harmonic RF systems: FLASH DESY Elettra Sincrotrone Trieste SLS PSI NSLS-II BNL SRF photoinjectors: ELBE PKU-SETF NPS-FEL R&D ERL CeC PoP at RHIC BERLinPro NPS-FEL WiFEL

HZDR Peking U. NPS BNL BNL HZB NPS U. of Wisconsin

151

Other SRF systems at high-β accelerators around the world. Country

Accelerator type

Number of cavities

Frequency (MHz)

Status

Switzerland Japan USA Switzerland

Beamline Storage ring/collider Storage ring/light source Storage ring/collider

2 2 8 4 (6)

2865 509 2815 400

Decommissioned Operational R&D R&D

Germany Italy

Linac, pulsed Storage ring/light source

4 2

3900 1500

Operational Operational

Switzerland USA

Storage ring/light source Storage ring/light source

2 1 (2)

1500 1500

Operational R&D

Germany China USA USA USA Germany USA USA

Linac, CW Energy recovery linac Linac, CW Energy recovery linac Linac, CW Energy recovery linac Linac, CW Linac, CW

1 1 1 1 1 1 1 1

1300 1300 500 704 112 1300 700 200

Operational Operational Operational Construction Construction R&D R&D R&D

Fig. 1. Geometry of elliptical cells showing the non-reentrant (left) and re-entrant (right) shapes [20]. The cell shape is re-entrant if the wall’s inclination angle α < 90◦ . The cell profile line comprises two elliptical arcs with half-axes A, B, a, and b connected by a straight line l.

a diameter of 24 cm. As a result, some other parameters were compromised. This is acceptable in a storage ring operation as there are typically very few cavities and they operate at low (5–10 MV/m) accelerating gradients. For comparison, the dimensions of the B-cell in Table 3 are scaled to 1497 MHz. Another example is the cells proposed for the International Linear Collider (ILC). The original

nine-cell 1300 MHz TESLA cavity shape [8] was developed in 1992. The walls of the cavity have a slope angle of 103.2◦, and the shape is welloptimized, with an emphasis on reducing Epk /Eacc . However, with the evolution of SRF technology, the emphasis of the optimization changed. Better cavity preparation techniques reduced significantly the probability of field emission, and thus achieving a lower peak surface magnetic Bpk field became more important than reducing the peak surface electric field Epk at the same accelerating gradient [26]. Also, it was shown that, with re-entrant cell shapes, cavities can be cleaned to reach very high gradients. Several alternative shapes were proposed for the ILC. Here I compare two of them with the original TESLA shape following the approach used in Ref. 27. The two proposed cavities, shown in Fig. 4, are of a re-entrant and an LL shape. Both cavities were optimized for a better Bpk /Eacc ratio, with the aim of assuring the maximum accelerating gradient. Table 4 compares the parameters of the three cavities. In addition, the TESLA and LL shapes are compared in Fig. 5. As the table shows, the two alternative ILC designs have an iris with a smaller radius than the TESLA cavity design and all their parameters are significantly better except for Epk /Eacc .

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

152

Fig. 2.

Generic single- and seven-cell elliptical cavities.

Fig. 3. (a) Geometry of an inner half-cell of a multicell cavity, and (b) field distribution along the profile line [19]. Hpk and Epk are the peak surface magnetic and electric fields. (Courtesy of V. Shemelin.)

Table 3. Comparison of the CEBAF OC and LL center cells and the CESR B-cell (scaled to 1497 MHz). Parameter Req (mm) Ra (mm) G (ohm) R/Q (ohm) G · R/Q (ohm2 ) Epk /Eacc Bpk /Eacc [mT/(MV/m)]

OC

LL

B-cell

93.5 35.0 273.8 96.5 26,422 2.56 4.56

87.0 26.5 280.3 128.8 36,103 2.17 3.74

91.5 40.1 265.7 89.0 23,647 2.50 4.16

Interestingly, it was found that optimization for maximal acceleration (Bpk /Eacc ) and for minimal losses (G · R/Q) essentially leads to the same shape of the cavity [20]. With standing wave SRF structures reaching fields close to the theoretical limit and with no

alternative material that could replace niobium, yet another approach to achieving high accelerating gradients was proposed recently [29]. Traveling wave structures are well utilized in normal-conducting linacs, but all SRF accelerators to date operate in standing wave mode owing to their simpler geometries, less material needed to fabricate them, and more straightforward cavity processing. However, superconducting traveling wave structures with a small phase advance per cell (Fig. 6) may significantly increase (up to 46% [30]) the accelerating gradient for the ILC. This approach is still in its early development stages, and only single-cell prototypes have been built and tested so far [31]. One of the cavities reached a gradient equivalent to 31 MV/m in a TESLA-shaped cavity. A three-cell structure is in preparation for testing.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

Fig. 4.

153

Magnetic field contours in the inner cells of (a) TESLA, (b) re-entrant, and (c) low-loss cavities proposed for the ILC [27].

Table 4. Comparison of the center cells for TESLA and two alternative designs of the ILC cavity.

2.2. New geometries for high-β applications

Parameter

TESLA

Re-entrant

LL

Ra (mm) G (ohm) R/Q (ohm) G · R/Q (ohm2 ) Epk /Eacc Bpk /Eacc [mT/(MV/m)]

35 271 113.8 30,840 1.98 4.15

30 284.3 135 38,380 2.30 3.57

30 284 133.7 37,970 2.36 3.61

For certain applications, elliptical cavities are not very suitable. For example, due to very long bunches, low-frequency cavities must be used in hadron circular accelerators and storage rings like the RHIC. Alternatively, a large aperture might be required to accommodate beams with big emittances (superconducting linacs for the neutrino factory/muon collider), which also leads to low-frequency cavities. Making elliptical cavities below ∼300 MHz is hardly feasible, due to their large size, as is evident from Fig. 7. Some applications (crab cavities for the LHC) impose very strict requirements on the cavity’s dimensions, making it difficult to fulfill with elliptical geometries. On the other hand, the dearth of available beamline space or the requirements of fast acceleration preclude the use of normalconducting structures. Thus, developing new superconducting cavity shapes is the only option. In this section, I consider two examples of new TEM-class SRF cavity geometries. Such cavities are typically employed in low- and medium-β accelerators, but were recently adapted to high-β. Another novel structure under development is an SRF photonic band gap (PBG) accelerating cell [33–35]. More

Fig. 5. Comparison of the half-cell geometry of the low-loss shape and the TESLA (TTF) shape [28].

Fig. 6.

A traveling wave structure with a feedback waveguide [30].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

154

Fig. 7.

A 200 MHz superconducting niobium-on-copper cavity on a horizontal supporting frame [32].

alternative cavity shapes will be discussed in later sections. 2.2.1. Quarter-wave resonators A quarter-wave resonator (QWR) is a section of the coaxial transmission line shorted at one end and open at the opposite end at a distance equal to a quarter of the fundamental mode wavelength. The resonator can be “closed” at the open end by an extension of the outer conductor with an end wall, thus forming an accelerating gap of length g, as shown in Fig. 8. The gap structure forms a loading capacitor C, which shortens the length of the coaxial part. The horizontal orientation of the QWR is common for normalconducting cavities, but was not utilized in SRF accelerators until recently [36]. Superconducting QWRs are known more for their use in low-β linacs, where they are oriented vertically, with the beam coming through a hole in the inner conductor serving as a drift tube. Thus, low-β structures have two accelerating gaps. Another possible use of the QWR is for beam deflection [36]. In this case particles pass through the end capacitor gap at 90◦ relative to the axis of symmetry as indicated in Fig. 8. The shunt impedance of the capacitively loaded QWR is given by Z0 1 , (2) Rsh = 0 2 αl 1 + ( ωCZ αl )

Fig. 8. Schematic of a quarter-wave resonator: l ≈ λ/4 is the transmission line’s length; a and b are, respectively, the radii of the inner and outer conductors; and g is the accelerating gap. The electrically shorted end is on the right [36].

where Z0 is the characteristic impedance of the coaxial line, α the attenuation coefficient, α=

Rs a−1 + b−1 , 2η ln(b/a)

(3)

Rs the surface resistivity, and η the impedance of free space. Due to the loading capacitance, the length of the coaxial line is shortened by λ − l = cZ0 C. (4) 4 The main advantage of the QWR is its compactness, allowing its usage at very low frequencies. Other advantages are its very high mechanical stability, making it insensitive to pressure changes in ∆l =

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

Fig. 9.

155

A 500 MHz SRF two-spoke cavity (courtesy of Old Dominion University).

the helium bath, and wide separation of the lowest higher-order mode from the fundamental mode. As I will discuss later, QWR structures are gaining popularity and are considered for fundamental RF systems, deflecting/crab cavity applications, and SRF electron guns. 2.2.2. Spoke resonators A spoke resonator consists of a cylindrical vessel, coaxial with the beam’s axis, loaded by λ/2-long structures oriented perpendicular to the beam axis. There are single- and multi-spoke resonators. In the latter, the loading elements are rotated by 90◦ from one to another. The distance between the centers of the spokes is βλ/2. A β = 1 version of a multispoke resonator was recently proposed as an alternative structure for small electron linacs [38]. Since the diameter of the spoke cavity is approximately half that of an elliptical cavity, it supports either a more compact design at the same frequency or an operation at lower frequency for the same transverse dimension. The lower frequency allows a 4 K operation, which might be attractive for small installations, where operating at 2 K may not be practical. Figure 9 shows the design of the 500 MHz two-spoke cavity developed recently for electron linacs [39]. 2.2.3. SRF PBG cavity A photonic bandgap accelerating structure is a periodic lattice of macroscopic components. In a twodimensional SRF PBG, such components could be metallic rods. A defect in the lattice (for example, no rod on the axis) can trap an accelerating mode, while other, higher-order, modes could still propagate to the periphery of the structure. This creates a unique capability of a PBG cell to serve as a HOM damper and an accelerating cell simultaneously [33, 34]. While the PBG cell with round rods has Bpk /Eacc

Fig. 10. Conceptual drawing of an SRF accelerator section incorporating a PBG cell with HOM couplers [33].

almost twice worse that that of the TESLA cell, it can complement elliptical cells in an accelerating section, as shown in Fig. 10. This will result in reducing long-range wakefields and increasing beam current thresholds in high-current SRF accelerators. However, optimization of the rod shape and spacing can significantly improve the peak magnetic field in a PBG cell [33]. Two 2.1 GHz SRF PBG single-cell resonators were fabricated and tested, reaching an accelerating gradient of 15 MV/m, with Q0 close to its predicted value [34, 35]. A photograph of the PBG niobium resonator is shown in Fig. 11.

3. Options for HOM Damping Extremely low RF losses, which make SRF cavities so attractive in the first place, are a handicap when we consider higher-order modes. The parasitic interaction of a bunched beam with HOMs can excite multibunch instabilities (longitudinal and transverse in storage rings, beam breakup in linacs

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

156

Fig. 11.

Table 5. Accelerator type

Photograph of the 2.1 GHz PBG resonator [34].

Typical requirements for high-β SRF cavities for fundamental RF systems. Example

Accelerating gradient

RF power

HOM damping

High (>20 MV/m)

High peak (>250 kW), low average (∼10 kW)

Moderate (Q = 104 –106 )

Pulsed linacs

ILC, XFEL

CW low-current linacs

CEBAF, ELBE

Low-to-moderate (8–20 MV/m)

Low average (5–15 kW)

Relaxed

High-current ERLs

BNL R&D ERL, BERLinPro

Moderate (15–20 MV/m)

Low average (5–15 kW)

Strong (Q = 103 –104 )

CW high-current injectors

Cornell ERL injector

Low-to-moderate (5–20 MV/m)

High average (50–1000 kW)

Strong (Q = 103 –104 )

High-current storage rings

CESR, KEKB, LHC

Low (5–10 MV/m)

High average (hundreds of kW)

Very strong (Q = 102 –103 )

and ERLs), lead to emittance growth and bunch-tobunch energy spread, and cause additional cryogenic losses [41]. Depending upon the type of accelerator, different levels of damping HOM quality factors are required: from 102 · · · 103 for storage rings to 104 · · · 106 for some linacs (Table 5). To keep the impedance of higher-order modes under control, special HOM dampers are attached to the beam tubes of SRF cavities. A HOM damper is nothing more than an RF/microwave absorber terminating a transmission line coupled to the superconducting cavity [40]. The

coupling is achieved via either an aperture or a coupling circuit consisting of lumped elements. There are designs using different transmission lines and coupling circuits. Aperture-coupled designs employ beam pipe HOM absorbers (wherein the beam pipe serves as a circular waveguide), rectangular waveguides, and radial lines. Coaxial line filters can be coupled via antennae, loops, or choke joints. Waveguides cannot support TEM modes, and electromagnetic waves can propagate in them only above the cutoff frequency of the lowest mode. Therefore, they act as natural high-pass filters and do

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

not need additional circuits to reject the fundamental RF. They are inherently more broadband than other types of HOM dampers. The cavity beam pipe frequently is enlarged to facilitate propagation of the lowest-frequency HOMs toward an absorber. The absorber is a section of the beam pipe with a layer of microwave absorbing material (lossy ferrites or ceramics, for example). Coaxial and radial lines can transmit TEM waves (hence no cutoff frequency). Therefore, HOM couplers based on these lines need special means to reject the fundamental RF. Coaxial antenna or loop couplers are more compact than other types of HOM damping schemes and can be tuned to the frequencies of trapped modes. HOM dampers are usually placed on the beam pipes next to the cavity. If a higher-order mode is trapped inside the cavity, it will not be coupled to the damper as there will be no electromagnetic field in the beam pipes [41]. The reasons for mode trapping are weak cell-to-cell coupling, and a big difference in HOM frequencies between inner cells and end cells (Fig. 12). A solution for avoiding trapped modes is always a compromise. Opening the iris increases cell-to-cell coupling (Fig. 13), but lowers R/Q of the fundamental mode. Changing the shape of the end cells alters the resonant frequencies of HOMs and keeps the fundamental R/Q high, but the frequency change is not the same for different HOMs and can

157

create a new trapped mode. Reducing the number of cells per cavity also helps [42], but entails a worse real estate gradient. It is also possible to trap modes inside a cryomodule if the neighboring cavities have different frequencies for a particular HOM [43]. Managing HOM impedance must be an integral part of any cavity design, with the possible exception of low-beam-current machines. It was recently argued [44] that in a low-beam-current CW SRF linac of Project X with a sparse beam spectrum, damping of HOMs may not be necessary. On the other hand, sometimes HOM damping alone is not enough to suppress beam instability and additional measures might be necessary. For example, to increase the relative cavity-to-cavity HOM frequency spread, several cavity “classes” were introduced for the future Cornell ERL by making small changes to the baseline inner cell shape [45].

3.1. Beam pipe absorbers Originally, beam pipe HOM absorbers or loads were developed at Cornell University and KEK for the high-beam-current colliders, CESR and KEKB. The two HOM loads are similar conceptually, but have different designs [46, 47]. Both use lossy ferrite materials to absorb microwave power of higherorder modes excited by beams. The basic idea is to design the cavity geometries, in which all HOMs

Fig. 12. The contour of the electric field’s amplitude: an example of mode trapping in a 13-cell cavity. End cells and inner cells have different frequencies for this resonant pattern [43].

Fig. 13. The contour of the electric field’s amplitude: a five-cell 704 MHz structure with high cell-to-cell coupling for the R&D ERL at BNL [43].

December 29, 2012

17:31

158

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

can propagate out of the accelerating structure into large-diameter beam pipes [25, 48]. Thus, the water-cooled absorbers do not have to be close to the cavities and are placed outside the corresponding cryomodules at room temperature. The HOM loads were designed to handle high power and in operation reached 5.7 kW at CESR and 16 kW at KEKB. The CESR and KEKB cavities equipped with these absorbers achieved very good damping: the loaded quality factors for HOMs ranged from 102 to 103 . The cavities of this type are often called HOM-damped or even single-mode cavities. In addition to high-current storage rings, the CESR-type HOM loads are installed in the R&D ERL at BNL to damp HOMs on a five-cell 704 MHz cavity, where the achieved damping is excellent as well [49]. Ferrite materials are brittle and, if cracked, can contaminate SRF cavities with particles and degrade the cavity’s performance. A HOM load with a ceramic break was proposed as a way to eliminate potential exposure of cavities to ferrite particles [49]. A concept of such a load is shown in Fig. 14, and its implementation for the SRF gun at BNL is shown in Fig. 15. While it is acceptable to have beam pipe HOM loads outside cryomodules for standalone units in storage rings or small R&D machines, it would be a very inefficient use of space in long linacs. To improve the real estate gradient, Cornell University

Fig. 15. The ceramic-ferrite damper with some ferrite tiles removed [49].

and KEK are developing beam pipe loads operating at cryogenic temperatures [50, 51]. The latest Cornell design is shown in Fig. 16. It is based on the ERL injector HOM beamline load [52], but with significant simplifications. The load has a graphiteloaded SiC microwave absorber cylinder brazed onto a metal sink, and stainless steel bellows. The cylinder is extended along the bellows to shield them from the beam and to damp high-frequency trapped modes in the bellows sections. Graphite-loaded SiC material has strong, broadband losses and sufficient DC conductivity to avoid its being charged up by the beam. Dampers of this type are probably the most efficient, but their main disadvantage for linacs is that they occupy significant real estate along the beam axis and thus reduce the fill factor. However, even in pulsed linacs a few beam pipe absorbers might be necessary to complement the loop HOM couplers at very high frequencies, as is planned for the European XFEL [41]. 3.2. Rectangular waveguide HOM couplers

Fig. 14. A ferrite HOM damper with a 10-cm-inner-diameter ceramic break. The plates with ferrite tiles are held in place by a clamshell holder (dimensions are in cm) [49].

The first rectangular waveguide HOM coupler was developed at Cornell University for what later became five-cell CEBAF cavities. New designs of such couplers are being worked on primarily at Jefferson Lab in the context of developing high-current CW cryomodules [53]. A “standard” configuration uses a Y-shaped arrangement of the waveguides on the cavity beam tubes. The latest applications

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

Fig. 16.

159

The Cornell ERL main linac HOM beamline absorber. Left: CAD model. Right: Prototype [50].

Fig. 17. A seven-cell cavity with waveguide HOM dampers for BERLinPro (courtesy of HZB).

adopting this concept are the SPX crab cavities for the APS Upgrade, and the ERL main linac cavities at BERLinPro (Fig. 17). While they provide very efficient damping in a broad frequency range and do not overly compromise the fill factor, waveguides significantly complicate the design of the cavity and cryomodule. 3.3. Coaxial antenna/loop couplers Coaxial dampers can offer strong damping, depending on a particular RF design. Their main disadvantage is their fundamental RF rejection filters, which must be carefully tuned. Couplers of this type were proposed and developed for HERA and LEP in the mid-1980s. Their construction and location (couplers were immersed in the liquid helium bath) supported CW operation at moderate gradients. LEP2 cavities had two HOM couplers [55]. The cavity’s higher-order modes were coupled via the hook-shaped inner conductor protruding into

the beam pipe, which was below the cutoff for the fundamental RF. The inductance of the hook and its capacitance to the wall of the beam port formed a notch filter at the fundamental RF frequency. The hook was filled with liquid helium for cooling. The notch filter could be adjusted after installation. LEP-type couplers were later adapted for the SOLEIL, LHC, and Super-3HC cryomodules at SLS and ELETTRA. In addition to the LEP2 dipole mode couplers, LHC cavities have broadband couplers (Fig. 18). SOLEIL and Super-3HC cavities have two types of HOM couplers, for dipole and monopole modes with a much higher coupling factor than in LEP cavities. TESLA HOM couplers are derived from HERA couplers with a simplified RF design due to more moderate damping specifications (QHOM ∼ 105 ). The TESLA couplers (Fig. 19) are located outside the helium vessel to optimize cost. This was possible because of the negligibly small heating in pulsed operation (1% duty factor). These couplers are successfully used on all TESLA cavities at high gradients in pulsed mode [43]. Later, these couplers were scaled for use in other applications, in both pulsed and CW regimes. At SNS, the couplers were scaled to 805 MHz with some minor modifications. At Fermilab, they were scaled for use on the 3.9 GHz cavities designed for FLASH (DESY). In both cases, difficulties were encountered with multipacting, overheating and failure of the couplers. The 3.9 GHz version had to be redesigned to shift the multipacting zones above the operating

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

S. Belomestnykh

160

(a) Fig. 18.

Fig. 19.

1230006

(b)

LHC HOM couplers: (a) narrowband dipole coupler; (b) broadband coupler [56].

Cross-section of the TESLA HOM coupler and photograph of its can and inner part, both made of bulk niobium [43].

gradients [57]. The CW operation of TESLA-style HOM couplers at relatively high gradients (about 20 MV/m) is planned for 1.5 GHz cavities for the 12 GeV CEBAF Upgrade [58] and 1.3 GHz cavities of the Compact ERL at KEK [59]. Enabling CW operation required improving cooling and making some modifications to the design. To avoid the difficulties in tuning narrowband notch filters, BNL is developing two new designs. A HOM coupler for the high-current 704 MHz SRF cavity [60] has a two-stage niobium band-stop filter (Fig. 20) tuned to reject the fundamental RF, but to have good transmission for the first HOM at 817 MHz. Three such couplers will be attached to the cavity at each end. The three large inductors efficiently conduct heat to the outside and could be cooled with either liquid or gaseous helium. The second design is for a 56 MHz quarter-wave resonator. The resonator, described later, will have four

magnetically coupled HOM dampers [61] equipped with high-pass discrete element filters.

4. Recent Progress with Fundamental SRF Systems The main purpose of the fundamental SRF systems is to provide energy to charged particle beams at a fast acceleration rate. However, operating cavities at the highest achievable gradient is not always optimal for an accelerator. Machine-dependent and technology-dependent factors determine the operating gradient of RF cavities and influence the cavity’s design. Based on the accelerating gradient, RF power and HOM damping requirements, fundamental SRF cavities are divisible into five types (Table 5). Figure 21 shows photos of different ones. Since niobium is the material of choice for SRF accelerating cavities, all those depicted are made of bulk sheet

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

Fig. 20.

161

Layout of the HOM coupler with a two-stage band-stop filter [60].

(a)

(d)

(b)

(e)

(c) Fig. 21. SRF cavities for fundamental RF systems: (a) 1300 MHz nine-cell TESLA cavity (courtesy of DESY); (b) 1500 MHz five-cell CEBAF cavities (courtesy of Jefferson Lab); (c) 704 MHz five-cell BNL1 cavity for ERL; (d) 1300 MHz two-cell Cornell ERL injector cavity (courtesy of Cornell University) and (e) 400 MHz single-cell LHC cavity (courtesy of CERN).

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

S. Belomestnykh

162

niobium, except for the LHC cavity, wherein a thin film of niobium is sputtered onto the inner surface of the cavity fabricated out of copper. 4.1. Success of the TESLA technology The international TESLA collaboration was formed in the early 1990s to develop a TeV-scale electron– positron linear collider based on SRF accelerating structures. A nine-cell cavity of optimized shape was designed [8] and many cavities were fabricated and tested. A photograph of the TESLA cavity is shown in Fig. 21(a). In working to improve the performance of TESLA cavities, many technological problems were solved and challenges overcome, including perfecting the fabrication techniques, developing new methods of cavity processing, and designing and building cryomodules, frequency tuners, and fundamental power couplers. A TESLA Test Facility (TTF) was built at DESY to demonstrate the feasibility of a high-gradient SRF linac as a base for a future linear collider and to advance the science of SASE FELs. It was a success. Later, this facility was renamed FLASH and became a user facility. It is difficult to overestimate the influence of the TESLA technology on many other projects around the world, and the reader will see references to it not only in this section but throughout the article. In this section I review projects that directly use the TESLA technology and/or further develop it. The FEL-focused branch of the technology has developed into the European XFEL project, while the high-gradient efforts continue toward the future International Linear Collider (ILC). 4.1.1. FLASH and European XFEL FLASH is a superconducting linac with an RF photo cathode gun driving a SASE FEL in the XUV and soft X-ray regime [62]; Fig. 22 is a schematic layout of

Fig. 22.

1230006

it. It has seven accelerating cryomodules, each with eight nine-cell 1.3 GHz niobium cavities. The theoretical maximum energy gain per module ranges from 180 MeV (the older modules) to 240 MeV (the latest XFEL prototype). A superconducting 3.9 GHz thirdharmonic cryomodule with four cavities produced at Fermilab is used for linearizing bunch compression. Table 6 lists some key parameters of the FLASH accelerator. The European X-ray free electron laser (XFEL) [63] is based on the TESLA superconducting linac. According to the latest design change, which reduced the accelerator energy from 17.5 GeV to 14 GeV, the linac will have 80 accelerator modules, each housing eight TESLA-type cavities [64]. The cavity’s design gradient is 23.6 MV/m. Construction of this project started in early 2009. 4.1.2. International Linear Collider The International Linear Collider is planned as a next high-energy-frontier electron–positron collider. The main linac design is based on TESLA cavities and will accelerate electron and positron beams to 250 GeV. The design is described in the ILC Reference Design Report and is progressing through the technical design (TD) phase [65]. The TD phase consists of four major steps: realization of a successful cavity production yield of 50% as an interim milestone and 90% by the end of the TD phase (at the design gradient of 35 MV/m); achievement of a 31.5 MV/m gradient in a cavity string in one cryomodule; performing system tests with beam acceleration; and preparing for industrialization of the ILC technology. The collaboration established a standard process for fabricating the cavity and for its surface preparation (Table 7). Due to continued improvement of cavity processing and better understanding of the gradient limit, the interim gradient goal of a 50% yield at

Overview of FLASH from the RF photocathode gun (left) to the experimental hall (right) (courtesy of DESY).

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators Table 6.

FLASH accelerator parameters [62].

Normalized emittance at 1 nC (90% rms) Charge per bunch Beam energy Bunches per train Bunch spacing Train repetition frequency

1.4 mm·mrad 0.15–1.5 nC 375–1250 MeV 1–500 1–25 µs 10 Hz

163

progressing well at three facilities: TTF/FLASH at DESY, the New Muon Laboratory (NML) at Fermilab, and the Superconducting Test Facility (STF) at KEK. The cavity-string performance was demonstrated at all three locations. 4.1.3. Application of TESLA technology to CW linacs

Table 7. The fabrication and surface preparation process for the ILC nine-cell cavity [65]. Fabrication

Material purchasing Subcomponent fabrication Cavity assembly using EBW technology Acceptance inspection

Surface Preparation Process

Electropolishing: EP1 (∼150 µm) Cleaning with ethanol or detergent High-pressure pure water rinsing (field flatness tuning) Electropolishing: EP2 (∼20 µm) Cleaning with ethanol or detergent High-pressure pure water rinsing Antenna assembly Baking at 120◦ C

(First & second pass allowed)

Cold test

Performance test with temperature and radiation monitoring, and RF mode test

35 MV/m and Q0 ≥ 8 · 109 was achieved (Fig. 23). Also, at least one demonstration of a 90% production yield by one cavity manufacturer (RI) at 38 MV/m was demonstrated at Jefferson Lab [66]. The production batch consisted of ten cavities. Integrating the cavity/cryomodule string and the cold tests are

Fig. 23.

The TESLA technology has also found applications in CW linacs. The ELBE linac at HZDR uses four TESLA cavities in two cryostats of the Stanford/Rossendorf design to provide multiple secondary beams, both of electromagnetic radiation and particles. The superconducting linac has been operating since 2001. Performance of the cavities is limited to ∼10 MV/m by field emission [68]. Two ELBE-type cryomodules are installed and operating at the energy recovery linac ALICE [69]. Other examples of application of the TESLA technology to CW linacs are described in the ERL subsection. 4.2. CEBAF 12 GeV Upgrade CEBAF is currently the largest SRF installation in the world. It is a recirculating linac with a top energy of 6 GeV. There are 40 cryomodules, each containing eight five-cell 1497 MHz cavities of the OC design. Their performance has gradually improved with time, from the original accelerating gradient of 5 MV/m to ∼7.5 MV/m. The accelerator is undergoing an upgrade project

ILC cavity gradient performance progress as of July 2012 [67].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

164

Fig. 24. Maximum voltage from C50 cryomodules before and after reworking [70].

to boost its energy to 12 GeV [70]. To achieve this goal, ten new cryomodules (C100) will be installed. Each will consist of eight seven-cell cavities of the LL design [24] and will deliver acceleration of 100 MeV. The LL cavities will operate at an accelerating gradient of 19.2 MV/m. In addition, to provide a reliable 6 GeV base from the old linac, a cryomodule refurbishing project was launched to rework the ten weakest cryomodules. The objective was to reach a 50 MeV energy gain per cryomodule, corresponding to an accelerating gradient of 12.5 MV/m. Hence, the name of the cryomodules is C50. While most of the cavities can support higher fields, a few

Fig. 25.

even to 20 MV/m, the CEBAF RF systems support only up to 13.5 MV/m. There are other operational constraints [70]. Figure 24 shows the demonstrated operational beam voltage capacity of the ten C50 cryomodules before and after reworking them. The CEBAF 12 GeV Upgrade project is well underway. To provide an additional performance margin, all cavities are receiving a light electropolish as the final surface treatment. The performance of the cavities in individual acceptance testing is very good (Fig. 25). To avoid taking nonbeneficial risks, their performance testing often stopped at 27 MV/m even though a cavity limitation has not been encountered. The test results show that technological methods applied to fabricating and preparing SRF cavities allow predictably to exceed gradient requirements. As stated in Ref. 70, this longsought-after goal has largely been achieved for CW applications. Six C100 cryomodules are already installed in the CEBAF tunnel, and one more has been completed. Two of the cryomodules were tested with beam prior to the machine upgrade shutdown. 4.3. ERL main linacs Energy recovery linacs (ERLs) are among the fastevolving applications of the SRF technology [71, 72].

Performance of C100 cavities in individual acceptance testing [70].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

In them, SRF cavities operate at moderate accelerating gradients: future machines plan for 15–20 MV/m, near the optimum for CW linacs. Currently, operational ERL cryomodules are at JLab FEL, ALICE, and BNL’s R&D ERL. The Jefferson Lab FEL was the first ERL to demonstrate high-current energy recovery [73]. Studies of beam breakup, performed at this machine, are seminal to the field. Presently, three cryomodules are installed in the FEL accelerating particles from 10 MeV to 135 MeV [72]. Two are of the CEBAF design with an improved HOM damping, and one is of the CEBAF Upgrade design. The cryomodule at BNL’s R&D ERL is of an original design developed specially for very-highcurrent application. It houses one five-cell 704 MHz cavity [Fig. 21(c)] [74]. As the cavity was designed to support the ERL beam current up to 500 mA, the requirements for HOM damping are on a par with those of the storage ring cavities and so its design somewhat resembles their designs. The cavity has a very large iris diameter of 17 cm and an even larger beam pipe diameter of 24 cm. This allows using the CESR-type beam pipe HOM loads for very efficient HOM damping. More SRF systems are in construction or R&D stages at KEK for the compact ERL (cERL), Cornell University (Cornell ERL), Peking University (PKU-SETF), and HZB for BERLinPro. All four

Fig. 26.

165

systems will operate at 1300 MHz and are closely related to the TESLA technology. All four aim to demonstrate 100 mA beam operation. For the initial stage of the cERL, one cryomodule with two nine-cell cavities is under construction [75]. A schematic view of the cryomodule is shown in Fig. 26. A slide-jack tuner incorporating piezo elements is used for frequency tuning. Beam pipe ferrite absorbers are attached to the cavities for HOM damping. Two cavities were fabricated and reached 25 MV/m with good Q-values. Conditioning the input couplers up to 40 kW CW traveling wave is complete. The cryomodule assembly is in progress. The main linac cryomodule of the Cornell ERL will house six seven-cell cavities operating at 16.2 MV/m [50]. The first prototype of a niobium cavity was fabricated and recently tested in a horizontal test cryostat. Its performance significantly exceeded the specification of Q0 = 2 · 1010 at Eacc = 16.3 MV/m (Fig. 27). The Saclay I frequency tuner and beam pipe HOM load will be used on this cavity. The Peking University Superconducting ERL Test Facility (PKU-SETF), will have two TESLAtype nine-cell cavities housed in a main accelerator cryomodule [124]. The maximum design energy of the machine is 30 MeV. Finally, the BERLinPro main linac will have three seven-cell cavities with very strong HOM damping using rectangular waveguide couplers

Schematic view of the cERL main linac cryomodule [75].

December 29, 2012

17:31

166

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

Fig. 27. Intrinsic quality factor versus accelerating field at 1.6 K and 1.8 K for the prototype Cornell ERL main linac cavity installed in the test cryomodule [50].

(Fig. 17) [54]. The current design features CornellERL-like center cells and the Y-shaped waveguide damping scheme originally developed at Jefferson Lab. One of the waveguides is replaced by the TTF-III fundamental power coupler. 4.4. Other SRF linacs, including ERL injectors There are several more existing and planned SRF linacs which I would like to mention here. SDALINAC is a unique S-band SRF linac which has been operating in Darmstadt since 1987. It consists of ten 20-cell niobium cavities operating at 2997.5 MHz. The accelerating gradient is 5 MV/m, allowing it to operate with beam energy up to 130 MeV after two recirculations. An upgrade of the machine is underway [76]. The JLab FEL injector employs two five-cell SRF cavities of the OC type. A DC gun operating at 325 keV provides up to 10 mA beam current. The cavities accelerate the beam further to 10 MeV for injection into the main linac. A prototype ERL injector has been operational at Cornell University for several years [77]. This fivecavity 1300 MHz SRF linac is designed to deliver high brightness and high average beam current beam to a future ERL. The cavities are two-cell elliptical structures equipped with two fundamental power couplers each. Recently, the injector has reached a CW beam current of 50 mA, setting a new record for this type of SRF linacs [78]. During this operation, 50 kW of RF power were coupled into each cavity and transferred to the beam. It is planned to further increase the

beam current in the Cornell ERL injector prototype toward the 100 mA specification. An injector cryomodule for the cERL comprises three two-cell SRF cavities [79] with an operating gradient of 14.5 MV/m. As with the Cornell ERL injector, each cERL injector cavity has two fundamental power couplers capable of delivering 167 kW of RF power per coupler. The cryomodule has been assembled and installed in the machine for testing and subsequent beam operation. At BNL, a Coherent electron Cooling Proof-ofPrinciple (CeC PoP) experiment is in preparation [80]. It will employ a short SRF linac, delivering a low-current electron beam with 1–5 nC charge per bunch. A 112 MHz SRF photoemission gun will generate electrons (see the description of the gun in Subsec. 7.3.1). An energy boost of 20 MeV will be provided by a 704 MHz five-cell BNL3 cavity [81]. A full-scale copper model and the first niobium version of the BNL3 cavity were fabricated by AES, Inc. (Fig. 28). The niobium cavity is in preparation for vertical testing. The second niobium cavity and cryomodule will be fabricated at Niowave, Inc. Upon vertical testing of both cavities, one will be chosen for integration in the cryomodule. The high-β section of the Project X CW linac will comprise 152 five-cell 650 MHz β = 0.9 cavities [82], operating at an accelerating gradient of 17.5 MV/m. The RF and mechanical design of the cavities is complete and the first prototypes have been fabricated (Fig. 29). 4.5. High-current storage rings High-current storage rings impose very demanding requirements on HOM damping and fundamental RF power couplers. Application of SRF cavities is very beneficial and is covered in great detail in the publications [1–3] and the bibliographies therein. Pioneering application of SRF at TRISTAN (KEK), HERA (DESY), and LEP (CERN) paved the way for the high-current electron–positron colliders CESR (Cornell University) and KEKB (KEK). The demonstrated advantages of superconducting cavities and the successful, reliable operation of SRF cavities in CESR [83] and KEKB [48] encouraged other laboratories to consider SRF for their projects. The cavities are single-cell elliptical structures with large apertures allowing easy coupling of HOMs to the dampers, which are either beam pipe loads

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

167

As already mentioned, elliptical cavities are not always an optimal choice. In accelerators like the heavy ion collider RHIC, where injected beams have large longitudinal emittance, using a low-frequency, high-voltage cavity can help to keep the ions from spilling into neighboring buckets. In the RHIC, it was proposed to use a 56 MHz quarter-wave SRF resonator as a “storage cavity,” meaning that the cavity is detuned and damped during the injection and energy ramp, and becomes active only at the energy of the experiment [36, 86]. The cavity will operate in a heavy beam loading regime with the beam-induced voltage of 2 MV. A low-power (1 kW) RF system will be used only to improve the phase and amplitude stability of the cavity’s field. The cavity will have four loop-type coaxial HOM couplers and a fundamental mode damper. The latter will be inserted to lower the cavity loaded Q to about 300 during injection and acceleration, making the cavity “invisible” to the beam. The fundamental damper is withdrawn from the cavity once the beam is at the energy of the experiment. The 56 MHz cavity in the helium vessel is shown in Fig. 30. 5. Deflecting/Crab Cavities

Fig. 28. A BNL3 five-cell cavity for the CeC PoP experiment (courtesy of AES, Inc.).

(CESR- and KEKB-type cavities) or coaxial couplers (LHC- and SOLEIL-type cavities). Nowadays many new high-current colliders and light sources operate or plan to operate single-cell HOM-damped SRF systems [84, 85] (Table 1).

Fig. 29.

With contemporary high energy particle colliders employing collisions at large crossing angles, the use of crab-crossing schemes becomes essential. As space in the interaction regions is limited, SRF cavities, with their ability to provide high transverse kicks, have a natural advantage. The first superconducting crabbing system was implemented at KEKB; more details about it, its commissioning, and its operation are given in Refs. 3, 87 and 88. Crab cavities also can assist in generating subpicosecond X-ray pulses in storage rings [89]. On the other hand, SRF deflecting structures can serve

The HE650 cavity prototype for Project X (courtesy of Fermilab).

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

168

Fig. 30.

Layout of the 56 MHz RHIC storage cavity.

for beam separation. The first such structure was designed and fabricated at KfK, Karlsruhe [90] and operated at CERN between 1977 and 1981. SRF separators are considered for the CEBAF 12 GeV Upgrade at Jefferson Lab [91] and for Project X at Fermilab. Similar to the fundamental RF cavities, the first crab cavities were elliptically shaped, but operating in a TM110 -like mode. To avoid degeneracy between two polarizations of this mode, the cavities are made non-axisymmetric (squashed). Because their operating mode is a higher-order mode, the cavities are significantly larger than accelerating cavities operating in the fundamental TM010 -like mode. Also, damping of the unwanted modes is more difficult in elliptical crab cavities as, in addition to damping HOMs, one must damp lower-order modes (LOMs) and sameorder modes (SOMs). There are applications wherein the dimensional requirements for crab/deflecting cavities are very tight and could not be met by the elliptical cavities of a “standard” design. One such application is the LHC [92, 93]. In response, several compact crab cavity shapes were proposed recently [94, 95]. In this section I consider two examples of crab cavity systems under active development.

chosen for the SPX crab cavity system [96]. The cavities and cryomodules are being developed in collaboration with Jefferson Lab. To generate 2 ps X-ray pulses, a total deflecting voltage of 2 MV is required from 2815 MHz single-cell SRF cavities. As with the fundamental RF cavities used in storage rings, strong damping of parasitic modes is necessary in order to to avoid multibunch instabilities. Two versions of the crab cavity, Mark I and Mark II, were designed and prototyped (Fig. 31). Both have elliptical squashed cell shapes and use waveguides for coupling HOMs and LOM out from the cells. Both use a Y-shaped end group, similar to that developed at Jefferson Lab for high-current cryomodules, to damp HOMs. The lower-order mode is dealt with differently, with a waveguide damper either on the beam pipe (Mark I) or on the cell body (Mark II). The latter has a more compact

5.1. SPX system for the APS Upgrade The APS Upgrade at ANL will include generating short-pulse X-rays (SPX) [89]. An SRF option was

Fig. 31. SPX deflecting cavity designs: Mark I (left) and Mark II (right) [98].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

Fig. 32. Mark I cavity Q versus deflecting voltage measured at ANL [97].

geometry and was chosen as the SPX baseline design. One Mark I and one Mark II cavity were vertically tested at Jefferson Lab, surpassing the deflecting voltage specification. However, the Mark II prototype had low Q. Further processing and testing is underway to improve the cavity Q [96]. Meanwhile, ANL is developing the capability of on-site cavity testing. The first vertical test of the Mark I cavity, prepared at Jefferson Lab and shipped to ANL under vacuum, was successfully completed [97]. The test showed no field emission, no multipacting, and exceeded the specified gradient and Q for SPX. The results, shown in Fig. 32, are consistent with the vertical test measurements at Jefferson Lab. Four deflecting cavities will be placed in a single cryomodule delivering the required 2 MV transverse kick voltage. A second cryomodule with four more cavities will completely reverse the crabbing “chirp” induced by the first set of cavities. It will be installed downstream of the beamlines utilizing SPX setup.

169

400 MHz was chosen because, on the one hand, it is low enough so that the nonlinearity of the transverse kick does not noticeably reduce luminosity, and, on the other hand, it is high enough so that cavities might be designed to be sufficiently compact to be compatible with the IP space restrictions. A major effort to develop compact crab cavities resulted in several TEM-like deflecting mode geometries [92–95]. Three primary concepts satisfying the key physical and RF constraints were chosen and are being developed toward a final design. These concepts are a quarter-wave resonator [99], an RF dipole cavity [100], and a four-rod cavity [101]. Table 8 compares parameters of the three resonators. Their geometries are shown in Figs. 33–35. One of the main advantages of the quarter-wave and RF dipole designs is that there are no lower- or sameorder modes and the frequency of the nearest HOM

Table 8. Parameters of the three TEM-type 400 MHz crab cavities for the LHC, at 3 MV.

Parameter

RFdipole

Fourrod

Quarterwave

Cavity radius (mm) Cavity length (mm) Aperture (mm) Epk (MV/m) Bpk (mT) R⊥ /Q (ohm) Next mode (MHz)

147.5 597.2 84 33.8 61.5 336.4 581.8

143/118 500 84 32 60.5 915 317–318

142.5 336.5 84 32.3 57.3 384 582

5.2. Compact crab cavities for the LHC luminosity upgrade The LHC luminosity upgrade aims at reducing the interaction point (IP) beta functions by a factor of two to three. Consequently, the crossing angle must be increased to retain normalized beam separation in the common focusing channel [92]. Increasing the crossing angle would diminish the benefit of reducing the beta function unless crab cavities are used to recover head-on collisions at the IP. Depending on the IP’s optics, the required transverse kick can be achieved with either two or three 400 MHz crab cavities operating at 3 MV [93]. The RF frequency of

Fig. 33.

The RF dipole crabbing cavity for the LHC [102].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

170

Fig. 34.

The four-rod crab cavity for the LHC [93].

particle beams. Sometimes there is a need to modify the longitudinal phase space. In linacs, bunches are typically accelerated on the crest of the RF wave. If the bunches are long, the cosine-shaped wave introduces nonlinear energy spread, which may affect bunch properties later, for example, after bunch compression. In storage rings, particles are accelerated near zero crossing, where the RF wave is linear. The slope of the RF affects the bunch length. Often, it is beneficial to make the bunches either longer or shorter. Doing this with the fundamental RF is inefficient. In all cases, adding a harmonic system can be a powerful way to modify the longitudinal phase space and improve the beam properties. Recently, it was pointed out that crab-crossing schemes in colliders also can benefit from introducing harmonic cavities to linearize the transverse kick along the bunches and avoid significant loss in luminosity, especially in the case of long hadron bunches [103]. In the following, I consider several examples of SRF harmonic systems. 6.1. Third-harmonic system at FLASH

Fig. 35. The double quarter-wave resonator for the LHC upgrade crabbing scheme.

is well separated from the operating frequency. Also, it is important to note that all three cavities are at least three times smaller than an elliptical crab cavity at 400 MHz. Prototyping of the three compact designs is underway to understand the challenges related to their fabrication and surface treatment, and to validate the RF designs and demonstrate a kick voltage of 3 MV per cavity. Thereafter, a final technology choice will take place to design and construct a cryomodule hosting the prototype cavity, which foreseeably will be tested in the SPS.

6. RF Harmonic Systems for Manipulating Beams in Longitudinal Phase Space The shape of the longitudinal phase space formed by the fundamental RF wave is not always optimal for

At the FLASH facility at DESY, to reduce the space charge effects in the low-energy regime, longer bunches are generated from the RF gun. These long bunches sample the nonlinearity of the sinusoidal RF wave and the result is a banana-shaped profile in the longitudinal phase space. After strong bunch compression, this generates a high spike in the charge density, which negatively affects lasing of the SASE FEL and tunability of the machine. A third-harmonic system (3.9 GHz) linearizes bunch compression for long bunches. The third-harmonic cavity/cryomodule was designed, fabricated, and assembled at Fermilab [104]. The nine-cell elliptical cavity shape was scaled from TESLA, but reoptimized for a lower surface field in the end cells. The inner cells have an iris diameter of 30 mm, which is increased to 40 mm in the end cells for better coupling with the fundamental power coupler. Two HOM couplers are mounted at each end of the cavity. The cavities were fabricated from 2.8-mm-thick bulk niobium (Fig. 36). Eight cavities were fabricated and tested in a vertical cryostat. All tested cavities have reached a gradient of at least 18 MV/m. Four of them were chosen for installation into the cryomodule; all reached at least 22 MV/m.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

Fig. 36.

171

The final weldment of the nine-cell cavity (courtesy of Fermilab).

Fig. 37.

Layout of the ACC39 cryomodule (courtesy of Fermilab).

The cryomodule, named ACC39 and shown in Fig. 37, was shipped from Fermilab to DESY, installed in FLASH during the upgrade in 2010 [62], and commissioned soon after. It is capable of providing a total energy gain of 20 MeV.

6.2. Bunch length manipulation in storage rings Having the ability to manipulate bunch length may significantly improve the performance of high-current storage rings. In low- to medium-energy storage ring light sources, the beam lifetime is often limited by the Touschek effect (large-angle intrabeam scattering). One of the methods of improving the Touschek lifetime is to reduce the charge density by lengthening the bunches. A particularly attractive

option is to lengthen the bunches by using harmonic RF cavities [105]. Passive harmonic cavities are effective instruments for bunch length manipulation and are in use at a number of light sources. Superconducting cavities, having a high quality factor and low R/Q, enjoy a number of advantages over normalconducting ones, when used in an idle (passive) regime [106, 107]. Among the advantages are: negligible beam energy loss; operating far from resonance, where the harmonic phase is close to an optimum; less sensitivity to an ion cleaning gap [108]. There are two currently operational third harmonic systems, at ELETTRA and SLS, and one in the R&D stage, for NSLS-II at BNL [109]. The bunch lengthening is accompanied by an increase in the spread of synchrotron frequencies

December 29, 2012

17:31

172

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

within the bunch. This enhances damping of the longitudinal coupled-bunch instabilities (so-called Landau damping). In colliders, it is important for the bunch length to match the beta function at the interaction point so as to avoid luminosity decrease because of an “hourglass” effect. In this case, short bunches are required and can be obtained by adjusting the phase of the harmonic cavity voltage appropriately.

6.2.1. Bunch shortening experiments with passive cavities at CESR A proof-of-principle experiment was performed at CESR to check the feasibility of using passive SRF cavities for bunch shortening [110]. For the experiment, one of the four CESR cavities was switched to a passive mode. The cavity was detuned far from resonance until the beam current reached 100 mA. Then, the tuner feedback loop was activated to keep the cavity voltage at 0.9 MV. It was possible to store a beam current of 400 mA. As there was no instrumentation available to measure the bunch length directly, it was calculated from the measured synchrotron frequency. The measured dependence of the synchrotron frequency on the beam current was in good agreement with the theoretical predictions (Fig. 38). The observed increase in the synchrotron frequency with the passive cavity voltage confirmed that the bunches were indeed shortened. More experiments followed, with the cavity external Q-factor adjusted to 106 from its nominal value of 2 · 105

Fig. 38. Comparison of the calculated dependence of the synchrotron frequency on the beam current (red curve) with the data measured during the passive cavity experiment at CESR (black squares) [110].

using a waveguide transformer [111]. A trial highenergy physics run showed that it is possible to reach luminosity comparable with that reached in normal operating conditions. 6.2.2. Super-3HC third-harmonic cryomodules at SLS and ELETTRA Two identical third-harmonic superconducting systems are operational at SLS and ELETTRA [112]. The systems are scaled-to-1.5-GHz versions of the SOLEIL cryomodule. This Super-3HC cryomodule (Fig. 39) was developed through collaborative efforts of CEA-DAPNIA-Saclay, PSI, and Sincrotrone Trieste [113]. Sputtered-niobium-on-copper cavities were made by CERN and tested in a vertical cryostat. Then they were transported to Saclay, where they were assembled in cryomodules and tested. Both cryomodules exceeded their design goal of a Qfactor of 108 at an accelerating gradient of 5 MV/m and a temperature of 4.5 K [114]. Utilization of the third-harmonic SRF cavities allowed both SLS and ELETTRA to improve the beam lifetime by factors of 2 to 3.5, depending on machine parameters, vacuum conditions, etc. An additional benefit of the bunch lengthening is the suppression of the longitudinal coupled-bunch instabilities by Landau damping. 7. SRF Photoemission Electron Guns Application of the SRF technology to photocathode electron injectors is a very active field of research [116, 117]. SRF has advantages over other electron gun technologies (DC or normalconducting RF) in the continuous wave mode of operation, where it can potentially provide a higher acceleration rate, and generate high-bunch-charge and high-average-current beams. SRF photoemission guns merge three sophisticated technologies: highquantum-efficiency photocathodes, superconducting RF, and high-repetition-rate synchronizable lasers. Among the challenges imposed by these technologies are maintaining an ultrahigh-vacuum (UHV) environment for the cathodes, maintaining the cleanliness of the cavity RF surfaces while allowing operation and replacement of the cathodes, designing the low-RF-loss and low-heat-leak interface between the cold cavities and the warmer cathodes, and synchronizing high-repetition-rate lasers with RF.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

Fig. 39.

173

The super-3HC cryomodule at SLS [115].

The first SRF photoemission electron source was proposed in the late 1980s [118, 119], and the first experiments were performed in the early 1990s [120]. However, the idea did not catch on until about ten years ago, when the first electron beam was obtained from a half-cell SRF gun at FZD [121]. This gun was followed by a 3.5-cell gun, the first to inject a beam into an accelerator. By accomplishing this, the HZDR/ELBE gun demonstrated the feasibility of the SRF gun concept with a normal-conducting Cs2 Te cathode. Several more guns have produced electron beams recently. While SRF guns have made excellent progress, many issues still need to be addressed. To produce low-emittance beams from SRF guns we must: reach a higher acceleration rate than has been achieved so far; have RF focusing near the cathode and/or the first solenoid as close to the cavity as possible; develop precise synchronization of a laser with RF; and develop proper transverse and temporal bunch shaping. To generate high bunch charges at high repetition rates, one needs high quantum efficiency (QE) photocathodes with a long lifetime. Metal cathodes are robust, but have low QE and are suitable only for use in the initial phases of SRF gun development,

when a high beam intensity is not required. A special coating of metal cathodes can increase the QE up to 7 · 10−3 [122]. Superconducting (niobium or lead) photocathodes have been used in small R&D guns. Semiconductor photocathodes are the preferred option for many projects as they can provide very good QE — 10% and higher. However, these cathodes are very sensitive to contamination and require UHV conditions for operation. The most developed semiconductor photocathodes are GaAs(Cs), Cs2 Te, and CsK2 Sb. Gallium arsenide is the only one of the three suitable for producing polarized electrons, but it is the most sensitive to ion back bombardment, requires an extremely good vacuum, and has a short lifetime. Cesium telluride is the most robust and has demonstrated a very long lifetime in an SRF gun, but requires the use of ultraviolet lasers. This makes it more difficult to use for high bunch charge: more laser power is needed at the same QE than at longer wavelengths, optics and pulse shaping are more difficult as well. Cesium potassium antimonide can be used with green lasers and is the most preferred option at present. Finally, diamond can be used to boost the photoemission current by a factor of ∼100

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

174

and diamond-amplified photocathodes are very promising for high-charge applications, though still require more R&D. For more details, I refer readers to the recently published review of photocathode R&D for future light sources [122]. Preparation of cavities for SRF guns has proven to be difficult so far. The cavities have either a very small opening for inserting a cathode or no opening at all in the case of superconducting cathodes. This makes chemical etching and cleaning of cavities challenging. The effect of normal-conducting cathodes on the SRF cavity performance is not clear yet, and the SRF guns still have to demonstrate stable operation in accelerators. This includes handling high average RF power, managing parasitic kicks from input power couplers, and effective damping of higher-order modes. The first SRF guns utilized elliptical cavity geometries, conventional for the high-β SRF cavities. Recently, several new guns were designed using a quarter-wave resonator geometry. QWRs are especially well suited for generating beams with high charge per bunch. They can be made sufficiently compact even at low RF frequencies (long wavelengths). The long wavelength allows generation of long electron bunches, thus minimizing space charge effects and enabling high bunch charge. Also, such guns should be suitable for experiments requiring high-average-current electron beams. In the following, I review active SRF gun projects.

Fig. 40.

7.1. Hybrid DC–SRF photoinjector at Peking University A hybrid DC–SRF gun has been developed for the Superconducting ERL Test Facility (SETF) at Peking University. The unique design combines a compact 100 kV Pierce DC gun with a 3.5-cell, 1300 MHz SRF cavity (Fig. 40). This simplifies the cavity design and somewhat decouples it from the photocathode. The gun is designed to produce a 5 MeV beam with a bunch charge of 100 pC, rms emittance of 1.2 mm·mrad, and repetition rate of 81.25 MHz. The gun cavity demonstrated good performance during a vertical test, reaching a quality factor of >1010 at an accelerating gradient of 23.5 MV/m. The test was performed at Jefferson Lab. During a horizontal cold test at Peking University, the field was up to 11.5 MV/m (limited by available RF power) at Qext = 6·106 [123]. The first beam measurements were performed with an ∼50 µA beam current generated by a Cs2 Te photocathode at the beam energy of 2.5 MeV. The initial emittance was measured to be 5–7 mm·mrad. The gun operated at an accelerating gradient of ∼6 MV/m [124]. 7.2. Elliptical cavity SRF guns 7.2.1. ELBE at HZDR ELBE is a multipurpose facility based on a CW SRF linac [68]. It has several modes of operation

The DC–SRF gun at Peking University [124].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

with different beam parameters: a high-peak-current operation with a bunch charge of 80 pC and a repetition rate of 13 MHz; a high-bunch-charge (1 nC), low-repetition-rate (∼1 MHz) operation; and a lowemittance, medium-charge-per-bunch (100 pC) operation with short pulses. An SRF gun will have to provide beams satisfying these machine requirements in order to eventually replace an old thermionic electron gun. The design energy of the ELBE 3.5cell, 1300 MHz SRF gun is 9.4 MeV. The three full cells are of the TESLA shape. The half-cell is velocity-optimized. The cathode insertion allows one to easily exchange and precisely position a liquidnitrogen-cooled, electrically isolated semiconductor photocathode. The gun cavity is shown in Fig. 41 and its parameters are listed in Table 9. The cavity performance in the cryostat is not very good: the quality factor is ∼3 · 109 , almost ten times lower than in a vertical test, with field emission starting at a peak field of 13 MV/m. The cavity operation is limited by field emission and liquid helium consumption. However, there has been no Q degradation for the last four years and it performs similarly with and without a cathode. The ELBE gun is the first SRF gun in the world to inject a beam into an accelerator. It provides ELBE with CW and pulsed beams. The maximum bunch charge injected into the linac is 120 pC at a repetition rate of 50 kHz with some beam losses, and 60 pC at 125 kHz with 100% beam transmission. The beam kinetic energy is limited to 3 MeV, due to field emission in

Fig. 41.

175

Table 9. Design parameters of the 1300 MHz, 3.5-cell, TESLA-shaped SRF gun cavity. Stored energy Quality factor Dissipated power Maximum beam power Geometry factor Accelerating voltage Accelerating gradient R/Q Epk /Eacc Bpk /Eacc

32.5 J 1010 25.8 W 9.4 kW 241.9 Ω 9.4 MV 18.8 MV/m 166.6 Ω 2.66 6.1 mT/(MV/m)

the SRF cavity in CW mode. Pulsed mode operation allows energy increase to 4 MeV. The Cs2 Te photocathode demonstrated a lifetime of ∼1 year with QE of 1% and a total extracted charge of 35 C. Several upgrades are planned for this gun [125, 126]. The new laser with a repetition rate of 13 MHz and better stability of the optical pulse amplitude will allow high-average-current operation (1 mA). Two new cavities have been constructed and tested at Jefferson Lab. One cavity is made from RRR300 fine-grain niobium, and the other from large-grain niobium (Fig. 42). The fine-grain cavity reached a peak surface field of 38 MV/m. If this performance is preserved, it will allow operation at a kinetic energy of 6–7 MeV. 7.2.2. BERLinPro at HZB BERLinPro at HZB is a demonstration energy recovery linac, which will generate and accelerate

The 3.5-cell SRF gun cavity for ELBE [116].

December 29, 2012

17:31

176

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

Fig. 43. The 1.6-cell SRF gun cavity after fabrication at Jefferson Lab [128].

SRF gun cavity was brought to HZB, where it was assembled into a cryostat and successfully commissioned at the HoBiCaT facility in 2011 [129, 130]. The cavity was operated in CW mode at peak field levels between 10 MV/m and 20 MV/m and loaded quality factors from 1.4 · 107 to 6.6 · 106 . The maximum kinetic energy was 1.9 MeV. In Fig. 44, the beam kinetic energy is shown as a function of the injection phase at 12 MV/m and 20 MV/m field at the cathode. Fig. 42. Fine-grain (left) and large-grain (right) SRF gun cavities [125].

a 100-mA, 50-MeV, low-emittance beam with a 1.3 GHz repletion rate (77 pC per bunch) [127]. To achieve these rather demanding parameters, a threestage program to develop an SRF gun was initiated. At the first stage, a 1.6-cell gun (Fig. 43) with a superconducting lead cathode is used as a beam demonstrator [128]. A thin film of lead, deposited onto the back wall of the cavity, is used to take advantage of an order of magnitude better QE of lead over niobium. At later stages, two additional prototypes will follow, the first being in the planning stage. It will include a high-QE CsK2 Sb photocathode to provide the full BERLinPro bunch charge at an average current of less than 5 mA. The second prototype will add high-power RF couplers to handle the full 100 mA beam loading. After the 1.6-cell cavity preparation and successful testing at Jefferson Lab, the all-superconducting

Fig. 44. Measured beam kinetic energy for 12 and 20 MV/m peak field versus launch phase compared to ASTRA and longitudinal tracking simulations [129].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

7.2.3. R&D ERL at BNL

7.3. Quarter-wave resonator SRF guns

The R&D ERL at BNL is a facility dedicated to developing accelerator technologies for future ERL applications in RHIC and eRHIC colliders with an emphasis on achieving high average beam currents and high bunch charges [131]. A 704 MHz half-cell SRF gun was designed by BNL [132, 133] and fabricated by AES, Inc. It has two RF input power couplers allowing delivery of 1 MW of RF power to a 500 mA electron beam. HOM damping is provided by an external beamline ferrite load with a ceramic break, described in Subsec. 3.1. A high-temperature superconducting (HTS) solenoid is located inside the cryomodule near the beam exit. The gun cryomodule was designed and fabricated by AES. Fundamental power couplers were manufactured by CPI/Beverly and conditioned with the maximum RF power of 250 kW pulsed and 125 kW CW in a full standing wave mode at BNL [134]. The gun cavity was chemically etched, cleaned and vertically tested at Jefferson Lab. The cryomodule was assembled at BNL, including its hermetic string (Fig. 45) preparation in the cleanroom. The cryomodule is installed in the ERL blockhouse for the first cold test and subsequent beam operation. This gun will use a CsK2 Sb photocathode and is expected to operate with a gap voltage up to 2.5 MV and will be capable to generate a bunch charge up to 5 nC and a beam current of 500 mA at 2 MV.

7.3.1. 112 MHz SRF gun for the CeC PoP experiment at BNL

Fig. 45.

177

A superconducting 112 MHz quarter-wave resonator was developed by collaborative efforts of BNL and Niowave, Inc. [135]. A low frequency was chosen for this gun to take full advantage of QWR benefits. Its ability to generate long bunches will reduce space charge effects. A short accelerating gap (relative to the wavelength) makes the transit time factor close to unity and the gap field practically constant. The center conductor geometry naturally accommodates the choke joint and allows mechanical decoupling of the cathode assembly from the niobium cavity. The gun cryomodule is shown in Fig. 46. The first cold test of the gun was successfully performed at Niowave in December of 2010 [135]. At present, the gun cryomodule is undergoing modifications for compatibility with installation in the RHIC tunnel, where it will produce electron bunches for the Coherent electron Cooling Proof-of-Principle (CeC PoP) experiment [80]. In addition, it will be used for studies of different types of photocathodes. LowRF-loss, low-heat-load stalks with a load-lock system for multialkali and diamond-amplified photocathodes have been designed and are being fabricated. The cryomodule will be equipped with a doublepurpose FPC/frequency-tuner assembly [136]. The first cooldown of the gun at BNL is scheduled for

Hermetic string assembly of the 704 MHz half-cell SRF gun for R&D ERL at BNL.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

178

Fig. 46.

The 112 MHz QWR SRF gun cryomodule.

2013. The gun is expected to operate at accelerating voltages up to 2 MV and generate electron beams with a charge per bunch from 1 nC to 5 nC at a repetition rate of 78 kHz. 7.3.2. 500 MHz SRF gun at NPS A 500 MHz SRF gun has been developed for the future FEL at NPS [137]. This gun, built and beamtested recently [138], is shown in Fig. 47. A niobium cathode on a copper stalk was used in the initial test. The following beam parameters were achieved: a beam energy of >460 keV, a bunch charge of 78 pC, and an rms emittance of 5 mm·mrad. NPS is working on a new, 700 MHz gun design, which will incorporate a number of improvements, enhancements and problem fixes over the 500 MHz gun design.

Fig. 47.

7.3.3. 200 MHz WiFEL SRF gun at the University of Wisconsin To support a future seeded VUV/soft X-ray free electron laser facility, WiFEL [139], a high-repetitionrate, VHF superconducting RF electron gun is under development at the University of Wisconsin [140]. A QWR geometry was chosen to reduce the size of the cavity at low frequency (199.6 MHz), which allows operation at 4.2 K. The goal is to operate at a cathode field of 40 MV/m and produce 200 pC bunches with a kinetic energy up to 4 MeV and a normalized transverse emittance of less than 1 mm·mrad. The design average beam current is 1 mA. The cavity was fabricated by Niowave, Inc. and processed at Jefferson Lab prior to the first cold test in a temporary cryostat at Niowave in February of

Layout of the 500 MHz NPS QWR gun [138].

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

179

Fig. 48. Mechanical layout of the cavity, He vessel, HTS solenoid, cathode holder, RF coupler, and mirror box of the 200 MHz WiFEL SRF gun [140].

2012. The main goal of the test was to determine the low-field Q0 , which was measured to be 3 · 109 , in agreement with simulations. After the cold test, a titanium helium vessel was attached to the cavity and the assembly shipped to Wisconsin for integration into the cryostat (Fig. 48). The cavity tuner uses a scissors jack driver similar to the one used by Michigan State University and Jefferson Lab. The loadlock/cathode holder is based on the HZDR system. Finally, an HTS solenoid provides emittance compensation for the gun. 8. Summary Superconducting radio frequency is now a core technology for a wide variety of accelerator applications worldwide, such as high-energy and nuclear physics experiments, generation of X-rays from undulators and free electron lasers, neutron sources, and production of rare isotopes. The technology is rapidly extending its functionality from the fundamental RF systems to deflecting/crab cavities, harmonic RF systems, and SRF photoinjectors. Several new large high-β installations are underway, either in a construction phase (CEBAF 12 GeV Upgrade, European XFEL) or in an R&D phase (Cornell ERL, Project X, ILC). In addition, there are many more, smaller-scale projects. The diversity of new applications and the large number of operating and planned SRF installations around the world indicate that the technology is considered essential by many laboratories. Acknowledgments I would like to thank many colleagues who provided updates on their work and figures. I am extremely

grateful to Avril Woodhead (BNL) for help with editing the text. This work is supported by Brookhaven Science Associates, LLC under Contract No. DEAC02-98CH10886 with the US DOE. References [1] H. Padamsee, J. Knobloch and T. Hays, RF Superconductivity for Accelerators, 2nd edn. (WileyVCH, 2008). [2] H. Padamsee, RF Superconductivity: Science, Technology and Applications (Wiley-VCH, 2009). [3] T. Furuya, Development of superconducting RF technology, Reviews of Accelerator Science and Technology 1, 211 (2008). [4] C. E. Reece and G. Ciovati, Superconducting radio-frequency technology R&D for future accelerator applications, Reviews of Accelerator Science and Technology, this issue. [5] A. Gurevich, Superconducting radio-frequency fundamentals for particle accelerators, Reviews of Accelerator Science and Technology, this issue. [6] M. Kelly, Superconducting radio-frequency for lowbeta particle accelerators, Reviews of Accelerator Science and Technology, this issue. [7] H. Padamsee, An overview of RF superconductivity research, IEEE Trans. Appl. Supercond. 5, 828 (1995). [8] B. Aune et al., Superconducting TESLA cavities, Phys. Rev. ST Accel. Beams 3, 092001 (2000). [9] R. L. Geng et al., High gradient studies for ILC with single-cell re-entrant shape and elliptical shape cavities made of fine-grain and largegrain niobium, in Proc. PAC’07 (Albuquerque, NM, USA, 2007), pp. 2337–2339. [10] J. Kirchgessner, The use of superconducting RF for high current applications, Part. Accel. 46, 151 (1994). [11] http://www.jacow.org.

December 29, 2012

17:31

180

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

[12] I. Campisi, Fundamental power couplers for superconducting cavities, in Proc. SRF2001 (Tsukuba, Japan, 2001), pp. 132–143. [13] I. Campisi, State of the art power couplers for superconducting RF cavities, in Proc. EPAC’02 (Paris, France, 2002), pp. 144–148. [14] Proc. Workshop on High-Power Couplers for Superconducting Accelerators (Jefferson Lab, Newport News, VA, USA, Oct. 30–Nov. 1, 2002), http : //http ://www . jlab . org / intralab / calendar/ archive02/HPC/papers.htm. [15] T. Garvey, The design and performance of CW and pulsed power couplers — A review, in Proc. SRF2005 (Ithaca, NY, USA, 2005), pp. 423–427. [16] A. Variola, High power couplers for linear accelerators, in Proc. LINAC’06 (Knoxville, TN, USA, 2006), pp. 531–535. [17] S. Belomestnykh, Overview of input power coupler developments, pulsed and CW, in Proc. SRF2007 (Beijing, China, 2007), pp. 419–423. [18] H. Sakai, Overview of CW input couplers for ERL, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 951–955. [19] S. Belomestnykh and V. Shemelin, High-β cavity design — A tutorial, in Proc. SRF2005 (Ithaca, NY, USA, 2005), pp. 2–19. [20] V. Shemelin, Optimal choice of cell geometry for a multicell superconducting cavity, Phys. Rev. ST Accel. Beams 12, 114701 (2009). [21] P. Kneisel et al., Nucl. Instrum. Meth. Phys. Res. 188, 669 (1981). [22] V. Shemelin, H. Padamsee and R. L. Geng, Optimal cells for TESLA accelerating structure, Nucl. Instrum. Meth. Phys. Res. A 496, 1 (2003). [23] M. M. Karliner et al., On the problem of comparison of accelerating structures operated by stored energy, Preprint INP 86-146 (Novosibirsk, 1986) (in Russian). [24] J. Sekutowicz et al., Cavities for JLab’s 12 GeV upgrade, in Proc. PAC2003 (Portland, OR, USA, 2003), pp. 1395–1397. [25] H. Padamsee et al., Design challenges for high current storage rings, Part. Accel. 40, 17 (1992). [26] K. Saito, Critical field limitation of the niobium superconducting RF cavity, in Proc. SRF2001 (Tsukuba, Japan, 2001), pp. 583–587. [27] J. Sekutowicz, TM-class cavity design: Tutorial. Presented at the 15th International Conference on RF Superconductivity (Chicago, IL, USA, 2011), http://twindico.hep.anl.gov/indico/getFile.py/access ? contribId = 1 & resId = 0 & materialId = slides & confId=631. [28] R. L. Geng, Review of new shapes for higher gradients, in Proc. SRF2005 (Ithaca, NY, USA, 2005), pp. 175–180. [29] P. Avrakhov, A. Kanareykin and N. Solyak, Traveling wave accelerating structure for a

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

superconducting accelerator, in Proc. PAC’05 (Knoxville, TN, USA, 2005), pp. 4296–4298. P. Avrakhov et al., Progress towards development of an L-band SC traveling wave accelerating structure with feedback, in Proc. EPAC’08 (Genoa, Italy, 2008), pp. 871–873. P. Avrakhov et al., Three-cell traveling wave superconducting test structure, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 940–942. R. L. Geng et al., 200 MHz Nb-Cu cavities for muon acceleration, in Proc. SRF2003 (Luebeck/Travemuende, Germany, 2003), pp. 531– 534. E. I. Simakov et al., Pushing the gradient limitations of superconducting photonic band gap structure cells, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 2801–2803. E. I. Simakov et al., An update on a superconducting photonic band gap structure resonator experiment, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 2140–2142. E. I. Simakov et al., First high power test results for 2.1 GHz superconducting photonic band gap accelerator cavities, Phys. Rev. Lett. 109, 164801 (2012). I. Ben-Zvi, Quarter wave resonators for beta ∼1 accelerators, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 637–645. I. Ben-Zvi, The quarter wave resonator as a crabbing cavity, LHC-CC10, 4th LHC Crab Cavity Workshop (2010), http://indico.cern.ch/getFile. py/access?contribId=52&sessionId=1&resId=1& materialId=slides&confId=100672. J. R. Delayen, S. U. De Silva and C. S. Hopper, Design of superconducting spoke cavities for highvelocity applications, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 1024–1026. D. Gorelov et al., Development of a superconducting 500 MHz multi-spoke cavity for electron linacs, in Proc. IPAC’12 (New Orleans, LA, USA, 2012). E. Haebel, Wakefields — Resonant modes and couplers, in Proc. Joint US–CERN–Japan International School “Frontiers of Accelerator Technology” (World Scientific, 1999), p. 490. J. Sekutowicz, HOM damping and power extraction from superconducting cavities, in Proc. LINAC’06 (Knoxville, TN, USA, 2006), pp. 506– 510. R. A. Rimmer, Higher-order mode calculations, predictions and overview of damping schemes for energy recovering linacs, Nucl. Instrum. Meth. Phys. Res. A 557, 259 (2006). J. Sekutowicz, Multi-Cell Superconducting Structures for High Energy e+ e− Colliders and Free Electron Laser Linacs (EuCARD, 2007). V. Yakovlev et al., Resonance effects of longitudinal HOMs in Project X linac, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 991–993.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

[45] N. Valles and M. Liepe, Designing multiple cavity classes for the main linac of Cornell ERL, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 937– 939. [46] E. Chojnacki et al., Beam line RF load development at Cornell, in Proc. PAC’99 (New York, NY, USA, 1999), pp. 845–847. [47] T. Tajima et al., Development of HOM damper for B-factory (KEKB) superconducting cavities, in Proc. PAC’95 (Dallas, TX, USA, 1995), pp. 1620– 1622. [48] K. Akai et al., RF systems for the KEK B-factory, Nucl. Instrum. Meth. Phys. Res. A 499, 45 (2003). [49] H. Hahn et al., Higher-order-mode absorbers for energy recovery linac cryomodules at Brookhaven National Laboratory, Phys. Rev. ST Accel. Beams 13, 121002 (2010). [50] M. Liepe et al., Progress on superconducting RF work for the Cornell ERL, in Proc. IPAC’12 (New Orleans, LA, USA, 2012). [51] M. Sawamura et al., Cooling properties of HOM absorber model for cERL in Japan, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 350– 352. [52] V. Medjidzade et al., Design of the CW Cornell ERL injector cryomodule, in Proc. PAC’05 (Knoxville, TN, USA, 2005), pp. 4290–4292. [53] R. Rimmer et al., Concepts for the JLab ampereclass CW cryomodule, in Proc. PAC’05 (Knoxville, TN, USA, 2005), pp. 3588–3590. [54] W. Anders et al., RF and SRF components for BERLinPro, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 53–55. [55] A. Butterworth et al., The LEP2 superconducting RF system, Nucl. Instrum. Meth. Phys. Res. A 587, 151 (2008). [56] D. Boussard and T. Linnecar, The LHC superconducting RF system, Adv. Cryogen. Eng. A 45, 835 (2000). [57] T. Khabiboulline et al., New HOM coupler design for 3.9 GHz superconducting cavities at FNAL, in Proc. PAC’07 (Albuquerque, NM, USA, 2007), pp. 2259–2261. [58] C. Reece et al., Diagnosis, analysis, and resolution of thermal stability issues with HOM couplers on prototype CEBAF SRF cavities, in Proc. SRF2007 (Beijing, China, 2007), pp. 656– 659. [59] K. Watanabe et al., New HOM coupler design for ERL injector at KEK, in Proc. SRF2007 (Beijing, China, 2007), pp. 531–535. [60] Wencan Xu et al., New HOM coupler design for high current SRF cavity, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 925–927. [61] E. Choi, H. Hahn and I. Ben-Zvi, Design of the fundamental mode damper and the HOM dampers for the 56 MHz SRF cavity, in Proc. PAC’09 (Vancouver, BC, Canada, 2009), pp. 900–902.

181

[62] M. Vogt et al., Status of the free-electron laser FLASH at DESY, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 3080–3082. [63] The European X-Ray Free-Electron Laser : Technical Design Report, DESY 2006-097 (2007). [64] H. Weise, Status of the European XFEL, in Proc. LINAC’10 (Tsukuba, Japan, 2010), pp. 6–10. [65] A. Yamamoto, Advances in SRF development for ILC, in Proc. SRF2011 (Chicago, IL, USA, 2001), pp. 7–12. [66] R. L. Geng, Overview of ILC high gradient cavity R&D at Jefferson Lab, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 74–78. [67] C. Ginzburg, ILC Cavity Database Update, Joint ACFA Physics/Detector Workshop and GDE Meeting on Linear Collider KILC12 (2012), http://ilcagenda.linearcollider.org/conferenceOther Views.py?view=standard&confId=5414. [68] H. Buettig, CW operation of the SRF linac ELBE. Presented at the CWRF2008 Workshop, http://indico.cern.ch/getFile.py/access?contribId= 20 & sessionId = 8 & resId = 1 & materialId = slides & confId=19621. [69] P. McIntosh et al., SRF system operation on the ALICE ERL facility at Daresbury, in Proc. SRF2009 (Berlin, Germany, 2009), pp. 34–40. [70] C. E. Reece, SRF challenges for improving operational electron linacs, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 14–19. [71] M. Liepe and J. Knobloch, Superconducting RF for energy-recovery linacs, Nucl. Instrum. Meth. Phys. Res. A 557, 354 (2006). [72] G. Neil, Energy recovery linacs for light source applications, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 761–765. [73] G. Neil et al., Sustained kilowatt lasing in a freeelectron laser with same-cell energy recovery, Phys. Rev. Lett. 84, 662 (2000). [74] R. Calaga et al., Ampere class linacs: Status report on the BNL cryomodule, Nucl. Instrum. Meth. Phys. Res. A 557, 243 (2006). [75] K. Umemori et al., Status of main linac cryomodule development for Compact ERL Project, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 67–69. [76] R. Eichhorn et al., The upgraded injector cryostat module and upcoming improvements at the SDALINAC, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 100–102. [77] B. Dunham et al., Performance of the Cornell high-brightness, high-power electron injector, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 20–22. [78] M. Liepe et al., High current operation of the Cornell ERL superconducting RF injector cryomodule, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 2378–2380.

December 29, 2012

17:31

182

WSPC/253-RAST : SPI-J100

1230006

S. Belomestnykh

[79] K. Umemori et al., Construction of cERL cryomodules for injector and main linac, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 956–961. [80] I. Pinayev et al., Status of proof-of-principle experiment for coherent electron cooling, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 400– 402. [81] Wencan Xu et al., High current cavity design at BNL, Nucl. Instrum. Meth. Phys. Res. A 622, 17 (2010). [82] S. Barbanotti et al., Status of the mechanical design of the 650 MHz cavities for Project X, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 943–945. [83] S. Belomestnykh and H. Padamsee, Performance of the CESR superconducting RF system and future plans, in Proc. SRF2001 (Tsukuba, Japan, 2001), pp. 225–233. [84] M. Pekeler, SRF in storage rings, in Proc. SRF2005 (Ithaca, NY, USA, 2005), pp. 98–103. [85] S. Belomestnykh, Superconducting RF in storagering-based light sources, in Proc. SRF2007 (Beijing, China, 2007), pp. 19–23. [86] I. Ben-Zvi, Superconducting storage cavity for RHIC. BNL note C-A/AP/337 (2009), http:// www . c - ad . bnl.gov/ardd/ecooling/docs/ PDF/56 MHz/ap note 337%20(2).pdf. [87] K. Hosoyama et al., Construction and commissioning of KEKB superconducting crab cavities, in Proc. SRF2007 (Beijing, China, 2007), pp. 63–69. [88] K. Akai et al., Commissioning and beam operation of KEKB crab RF system, in Proc. SRF2007 (Beijing, China, 2007), pp. 632–636. [89] A. Zholents et al., Generation of subpicosecond Xray pulses using RF orbit deflection, Nucl. Instrum. Meth. Phys. Res. A 425, 385 (1999). [90] A. Citron et al., Nucl. Instrum. Meth. Phys. Res. 164, 31 (1979). [91] S. Ahmed et al., Beam dynamics studies for transverse electromagnetic mode type RF deflectors, Phys. Rev. ST Accel. Beams 15, 022001 (2012). [92] R. Calaga, Crab crossing for LHC Upgrade, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 988– 993. [93] R. Calaga, Crab cavities for the LHC upgrade, LHC Performance Workshop (Chamonix, 2012), http://indico.cern.ch/getFile.py/access?contribId= 64 & sessionId = 9 & resId = 0 & materialId = paper & confId=164089. [94] G. Burt, New cavity shape developments for crabbing applications, in Proc. SRF2009 (Berlin, Germany, 2009), pp. 479–484. [95] J. Delayen, Compact superconducting cavities for deflecting and crabbing applications, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 631–636. [96] A. Nassiri et al., Status of the Short-Pulse X-Ray Project at the Advanced Photon Source, in Proc.

[97]

[98]

[99]

[100]

[101]

[102]

[103]

[104]

[105]

[106]

[107]

[108]

[109]

[110]

[111]

[112]

IPAC’12 (New Orleans, LA, USA, 2012), pp. 2292– 2294. J. D. Fuerst et al., Tests of SRF deflecting cavities at 2 K, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 2300–2302. A. Nassiri et al., Status of the Short-Pulse X-Ray (SPX) Project at the Advanced Photon Source (APS), in Proc. PAC’11 (New York, NY, USA, 2011), pp. 1427–1429. R. Calaga et al., A quarter wave design for crab crossing in the LHC, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 2260–2262. D. Gorelov et al., Engineering of a superconducting 400 MHz crabbing cavity for the LHC HiLumi Upgrade, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 2411–2413. B. Hall et al., Analysis of the four rod crab cavity for HL-LHC, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 2275–2277. S. U. De Silva and J. R. Delayen, Design and development of superconducting parallel-bar deflecting/crabbing cavities, in Proc. IPAC’12 (New Orleans, LA, USA, 2012). V. N. Litvinenko, Some thoughts on “crabbing” Presented at eRHIC Meeting (May 18, 2011); not published. E. Harms et al., Third harmonic system at Fermilab/FLASH, in Proc. SRF2009 (Berlin, Germany, 2009), pp. 11–16. J. M. Byrd and M. Georgsson, Lifetime increase using passive harmonic cavities in synchrotron light sources, Phys. Rev. ST Accel. Beams 4, 030701 (2001). S. Belomestnykh, Operating experience with β = 1 high current accelerators, in Proc. SRF2003 (Luebeck/Travemuende, Germany, 2003), pp. 225– 233. P. Marchand, Superconducting RF cavities for synchrotron light sources, in Proc. EPAC’04 (Lucerne, Switzerland, 2004), pp. 21–25. G. Penco, Experimental studies on transient beam loading effects in the presence of a superconducting third harmonic cavity, Phys. Rev. ST Accel. Beams 9, 044401 (2006). J. Rose et al., Design of a 1500 MHz bunch lengthening cavity for NSLS-II, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 210–212. S. Belomestnykh et al., Superconducting RF system upgrade for short bunch operation of CESR, in Proc. PAC2001 (Chicago, IL, USA, 2001), pp. 1062–1064. S. Belomestnykh et al., Using passive cavities for bunch shortening in CESR, in Proc. PAC2003 (Portland, OR, USA, 2003), pp. 1306–1308. M. Pedrozzi et al., First operational results of the third harmonic superconducting cavities in SLS and ELETTRA, in Proc. PAC2003 (Portland, OR, USA, 2003), pp. 878–880.

December 29, 2012

17:31

WSPC/253-RAST : SPI-J100

1230006

Superconducting Radio-Frequency Systems for High-β Particle Accelerators

[113] M. Svandrlik et al., The SUPER-3HC Project: An idle superconducting harmonic cavity for bunch length manipulation, in Proc. EPAC’2000 (Vienna, Austria, 2000), pp. 2052–2054. [114] P. Bosland et al., SUPER-3HC cryomodule: layout and first tests, in Proc. EPAC’02 (Paris, France, 2002), pp. 2217–2219. [115] P. Bosland et al., Third harmonic superconducting passive cavities in ELETTRA and SLS, in Proc. SRF2003 (Luebeck/Travemuende, Germany, 2003), pp. 239–243. [116] A. Arnold and J. Teichert, Overview on superconducting photoinjectors, Phys. Rev. ST Accel. Beams 14, 024801 (2011). [117] S. Belomestnykh, Survey of SRF guns, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 23–26. [118] H. Chaloupka et al., A proposed superconducting photoemission source of high brightness, in Proc. SRF1989 (Tsukuba, Japan, 1989), pp. 583–588. [119] A. Michalke et al., Photocathodes inside superconducting cavities, in Proc. SRF1991 (Hamburg, Germany, 1991), pp. 734–742. [120] A. Michalke et al., First operation of high-quantum efficiency photocathodes inside superconducting cavities, in Proc. EPAC’92 (Berlin, Germany, 1992), pp. 1014–1016. [121] J. Janssen et al., First operation of a superconducting RF-gun, Nucl. Instrum. Meth. Phys. Res. A 507, 314 (2003). [122] D. H. Dowell et al., Cathode R&D for future light sources, Nucl. Instrum. Meth. Phys. Res. A 622, 685 (2010). [123] F. Zhu et al., Status of the DC-SRF photoinjector for PKU-SETF, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 973–976. [124] K. Liu, DC-SRF photocathode injector for ERL at Peking University. Presentation at the ERL2011 Workshop at KEK (Oct. 17–21, 2011), http://kds. kek . jp/getFile . py/access ? contribId = 15 & session Id=23&resId=0&materialId=slides&confId=7855. [125] A. Arnold et al., Fabrication, tuning, treatment and testing of two 3.5 cell photo-injector cavities for the ELBE linac, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 405–407. [126] P. Murcek et al., Modified SRF photoinjector for the ELBE at HZDR, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 39–42.

183

[127] J. Knobloch et al., Status of the BERLinPro energy recovery linac project, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 601–603. [128] A. Neumann et al., Photoinjector tests at HoBiCaT, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 962–968. [129] A. Neumann et al., First characterization of a fully superconducting RF photoinjector cavity, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 41–43. [130] T. Kamps et al., First demonstration of electron beam generation and characterization with an all superconducting radio frequency (SRF) photoinjector, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 3143–3145. [131] D. Kayran et al., Status of high current R&D energy recovery linac at Brookhaven National Laboratory, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 2148–2150. [132] R. Calaga et al., High current superconducting gun at 703.75 MHz, Physica C 441, 159 (2006). [133] A. Burrill et al., BNL superconducting RF guns — Technology challenges as ERL sources, Nucl. Instrum. Meth. Phys. Res. A 557, 75 (2006). [134] W. Xu et al., Design, simulations, and conditioning of 500 kW fundamental power couplers for a superconducting RF gun, Phys. Rev. ST Accel. Beams 15, 072001 (2012). [135] S. Belomestnykh et al., Superconducting 112 MHz QWR electron gun, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 223–225. [136] T. Xin et al., Design of the fundamental power coupler and photocathode inserts for the 112 MHz superconducting electron gun, in Proc. SRF2011 (Chicago, IL, USA, 2011), pp. 83–86. [137] K. L. Ferguson et al., NPS BPL and FEL facility update, in Proc. FEL2010 (Malmo, Sweden, 2010), pp. 30–32. [138] J. Harris et al., Design and operation of a superconducting quarter-wave electron gun, Phys. Rev. ST Accel. Beams 14, 053501 (2011). [139] J. Bisognano et al., Progress toward the Wisconsin free electron laser, in Proc. PAC’11 (New York, NY, USA, 2011), pp. 2444–2446. [140] R. Legg et al., Status of the Wisconsin SRF gun, in Proc. IPAC’12 (New Orleans, LA, USA, 2012), pp. 661–663.

December 29, 2012

17:31

184

WSPC/253-RAST : SPI-J100

S. Belomestnykh

Sergey Belomestnykh is Head of the Superconducting RF (SRF) group in the Collider-Accelerator Department of the Brookhaven National Laboratory (BNL) and an Adjunct Professor of Physics at Stony Brook University. He started his scientific career at Budker Institute of Nuclear Physics in Novosibirsk, Russia, after graduating from Novosibirsk State Technical University. At Budker Institute, he has worked on normal conducting RF systems for circular accelerators. In 1994 he joined Laboratory for Elementary-Particle Physics at Cornell University and began working on SRF technology. He was involved in developing the CESR superconducting RF cavities. After becoming the CESR RF group leader, he was responsible for installation, commissioning and operation of the CESR SRF system. He has participated in the Cornell Energy Recovery Linac (ERL) program, leading R&D and commissioning efforts for several subsystems of the ERL injector. Since 2010, he has been working at BNL, pursuing his interests in developing new superconducting RF cavities and photoemission electron guns for future particle accelerators.

1230006

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Reviews of Accelerator Science and Technology Vol. 5 (2012) 185–203 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300071

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators Michael Kelly Argonne National Laboratory, 570 South Cass Avenue, Argonne, IL 60439, USA [email protected] High-power proton and ion linac projects based on superconducting accelerating cavities are driving a worldwide effort to develop and build superconducting cavities for beta < 1. Laboratories and institutions building quarter-wave, halfwave and single- or multi-spoke cavities continue to advance the state of the art for this class of cavities, and the common notion that low-beta SRF cavities fill a need in niche applications and have low performance is clearly no longer valid. This article reviews recent developments and results for SC cavity performance for cavities with beta up to approximately 0.5. The considerable ongoing effort on reduced beta elliptical cell cavities is not discussed. An overview of associated subsystems required to operate low-beta cavities, including rf power couplers and fast and slow tuners, is presented. Keywords: Superconducting cavity; low-beta.

1. Introduction

However, over the past decade there has been widespread interest in accelerating ions up to higher velocities (β > 0.1) for a variety of applications. Consequently, new structures have been designed and prototyped, extending upward the range of beta for TEM cavities and extending downward the range for elliptical cavities. In fact, mature geometries for SC accelerating structures spanning the full velocity range exist and some of these are shown in Fig. 1. Reference 6 contains a good discussion on the merits of rf superconductivity using low- and mediumvelocity cavities.

Two distinct classes of superconducting (SC) structures were developed as accelerating cavities in the 1960s and 70s by largely independent groups of scientists and engineers. For particles with high velocities (β ∼ 1), elliptical cell structures were developed at Stanford University for CW acceleration of electron beams [1]. Electromagnetic (EM) modes of these cavities resemble pillbox modes and typically use the TM010 pi mode for particle acceleration. For relatively-low-velocity acceleration of protons and heavy ions, useful cavity geometries were qualitatively much different. The most successful shapes were some variation of a quarter-wave resonator operated in the lowest EM resonant mode, resembling a transverse electromagnetic (TEM) mode of a shorted coaxial transmission line. This class of cavities is sometimes referred to as TEM cavities. Accelerators based on TEM cavities were mostly used for low-energy nuclear physics studies and accelerated ions up to about 10% the speed of light (β ∼ 0.1). Previous reviews [2–5] provide a comprehensive discussion on the development of early structures for ion acceleration.

2. Applications Practically all of the early linacs based on lowbeta SRF cavities are relatively small machines with a dozen to a few dozen cavities used to accelerate moderate beam currents (microamps or less) for heavy ion nuclear physics or related studies. With a couple of exceptions, cavity frequencies are close to 100 MHz, and the maximum beta is around β ∼ 0.1. A list of the existing linacs using low-beta SC quarter-wave cavities is shown in Table 1. The list is time-ordered from top to bottom, and shows 185

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

M. Kelly

186

Fig. 1. Practical superconducting cavities spanning the full range of low and intermediate values of beta.

ATLAS, the first SC ion linac (commissioned in 1978), and the recently commissioned SARAF halfwave cryomodule at Soreq. SARAF [7] is the first accelerator to use a half-wave structure (coaxial halfwave or spoke) to accelerate ion beams. Several other proposed or planned SRF linacs, many of them large and including several different cavity geometries, will use quarter- and Table 1. Facility

Argonne ATLAS Stony Brook Florida State JAERI INFN Legnaro Canberra TRIUMF New Delhi SARAF Michigan State

half-wave structures to accelerate ions in the intermediate-velocity region of 0.1 < β < 0.5. Examples of projects proposing to use half-wave and spoke cavities are shown in Table 2. The present state of the art for low-beta SRF linacs in routine operations is represented by the ATLAS intensity upgrade cryomodule at Argonne (commissioned in 2009) [8] and the ISAC-II [9] heavy ion linear accelerator (in operation at TRIUMF since 2006). The TRIUMF accelerator consists of 40 SC quarter-wave cavities similar to the INFN-Legnaro design and is based on relatively simple coaxial niobium tubes. Cavities have two beta values, β = 0.057 and 0.071, with each cavity providing approximately 1 MV of accelerating potential. A Phase II expansion of ISAC-II, completed in 2010 and using β = 0.11 quarter-wave cavities, added 20 MV of the accelerating gradient using 20 cavities. The ISAC-II cryomodules were assembled in a cleanroom, in order to reduce particulate contamination on the rf surfaces. This permitted operation with fairly high values of the electric field on the cavity rf surface. The average value of Epeak ∼ 35 MV/m has been stable

Operational superconducting cavity linacs for low beta.

Cavity type (s)

Frequency (MHz)

Beta (v/c)

Number of cavities

First operation

Split-ring, QWR Split-ring, QWR Split-ring QWR QWR Split-ring, QWR QWR QWR HWR QWR

48–109 150.4 97 130, 260 80, 160 150.4 106, 141 97 176 80.5

0.01–0.15 0.07–0.1 0.07–0.1 0.1 0.05–0.13 0.09–0.11 0.057, 0.071 0.08 0.08 0.045

64 40 15 46 74 14 40 18 6 6

1978 1983 1987 1994 1994 1996 2006 2007 2009 2011

Table 2.

Proposed SRF projects based on λ/2 cavities.

Facility

Cavity type (s)

Frequency (MHz)

Beta (v/c)

Number of cavities

SARAF Phase II MSU FRIB RISP Project X EURISOL ESS IFMIF China CAS

QWR HWR HWR HWR, spoke HWR, spoke 2-spoke HWR HWR

176 322 162.6, 325 162.5, 325 176, 352 352 175 162.5

0.08–0.12 0.28, 0.53 0.12, 0.3, 0.53 0.10–0.47 0.09, 0.15, 0.3 0.5 0.094 0.09

28 224 238 61 108 32 14 16

1.0–2.1 0.8–3.7 0.5–0.8 1.7–3.8 0.6–2.0 4.4 0.7 0.4

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

from commissioning up until the present. Other features and techniques common, for example, in elliptical cell linacs, such as highly optimized geometries, baking, electropolishing, separate cavity rf space and cryogenic vacuum space and 2 K operation, provide opportunity for improvements for this range of beta. Some of these newer techniques will be used with the SPIRAL-2 [10] SC linac currently under construction at GANIL. The linac will use two cryomodule types: a set of low-beta modules each housing one β = 0.07 cavity, and “high-beta” modules housing two β = 0.12 cavities. The design accelerating field of the SPIRAL-2 β = 0.12 quarter-wave cavities is EACC = 6.5 MV/m (Leff = nβλ/2 = 0.41 m), which corresponds to an accelerating voltage of more than 2.5 MV per cavity. The relatively high voltage/cavity compared to operating cavities is due largely to substantially reduced surface fields from electromagnetic design optimizations. The cryomodule design also incorporates clean assembly and separate cavity and insulating vacuum spaces.

187

The energy upgrade of the ATLAS heavy ion linac at the ANL in 2009 used a single 5-m-long cryomodule containing seven 109 MHz β = 0.15 SC quarter-wave cavities. Design features to support high performance included cleanroom assembly, separate cavity and insulating vacuum spaces, low peak surface fields based on an optimized geometry, and rf surfaces prepared using electropolishing on cavity subcomponents rather than the more common buffered chemical polishing (etching). In addition, the long cryomodule has a high packing factor and results in a relatively high “real estate” gradient of 3 MV/m. As tested inside the cryomodule, cavities have an average maximum gradient of EACC = 7.6 MV/m (Leff = nβλ/2 = 0.39 m), providing an average of 2.9 MV per cavity. Operations in ATLAS are at 2.1 MV per cavity due to the finite energy handling capability of the VCX fast tuner. The large voltage per cavity and the high real estate gradient constitute an important advance for this class of cavity; however, it also highlights the need for fast tuning systems that are compatible with

Fig. 2. Quarter-wave cavities. Clockwise from top left: ANL 97 MHz β = 0.07, 0.1, Stony Brook 150 MHz β = 0.06, 0.1, New Delhi 97 MHz β = 0.08, ANL 425 MHz β = 0.15, INFN-LNL 160 MHz, ANL 109 MHz β = 0.155.

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

188

1230007

M. Kelly

Fig. 3. Half-wave cavities. Clockwise from top left: ANL 162 MHz β = 0.28, MSU 322 MHz β = 0.28, INFN-LNL 352 MHz β = 0.17, 31, ANL 350 MHz β = 0.12, SARAF 176 MHz β = 0.08.

the large cavity stored energies inherent in high-field operation. At Soreq NRC, the Soreq Applied Research Accelerator Facility (SARAF) [7] has been commissioned with an SC cryomodule housing six β = 0.09 half-wave resonators (HWRs) and three SC solenoids. The second phase of the project will include additional SC modules to bring the final deuteron beam energy to 40 MeV. The SARAF and SPIRAL-2 linacs represent the future trend toward high-intensity ion beam accelerators. The choice of the SARAF half-wave cavity was based on the absence of magnetic steering

fields on the beam axis, an intrinsic feature of the HWR geometry. It is noted that solutions to the effect of magnetic steering in quarter-wave cavities have been developed [11]. The SARAF cryomodule system, designed and built by ACCEL (now Research Instruments), uses clean assembly and a separated cavity vacuum space. The planned accelerating voltage/cavity of 0.86 MV has not yet been achieved, due primarily to the large sensitivity of this particular half-wave design to pressure fluctuations in the 4 K helium bath and the difficulty of phase-locking cavities at higher fields.

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

189

Fig. 4. Spoke cavities. Clockwise from top left: ANL 805 MHz β = 0.28, IPN-Orsay 352 MHz β = 0.35, Julich 760 MHz β = 0.2, ANL 350 MHz β = 0.63, LANL 350 MHz β = 0.17.

2.1. Future applications New SC accelerating cavities do, in many cases, provide the best solution for the next generation of ion linacs for basic and applied science and technology. Since SC cavities are the only demonstrated technology compatible with the production and acceleration of continuous (CW) ion beams at very high currents (up to several milliamperes), they are an attractive option for applications requiring accelerated beams with high power of a megawatt or more. High-power proton beams can be used to produce intense beams of spallation neutrons for energy production. Stated advantages are inherent safety and higher efficiency for conversion of stockpiles of nuclear waste from existing reactors (“ADS”) [12]. Other applications include interrogation of shipping containers and the irradiation of actinide targets to produce large quantities of isotopes for medicine. The International Fusion Materials Irradiation Facility (IFMIF) [13] plans to use two high-power CW accelerator drivers based on β = 0.094 HWRs, each delivering deuteron beams to a final energy of 40 MeV. The proposed beam current of 125 mA per driver section would eventually require the demonstration of an rf power coupler capable of transmitting up to 200 kW of forward power per device for a cavity accelerating gradient of EACC = 4.5 MV/m.

This power is well beyond the few kilowatts of rf power used with any TEM cavities to date. In basic research, a high-intensity SC ion linac is the basis of Project X [14] at Fermilab, and also for a multi-ion driver for the Facility for Rare Isotope Beams (FRIB) [15] for research in nuclear physics. Plans for the Project X low-beta cavities include three types of λ/2 structures (one half-wave and two spokes). Four types of TEM cavities (two quarterwaves and two half-waves) are under development at Michigan State University for use in FRIB. The FRIB SC low-beta linacs require a total of 330 cavities. Here, MSU is pursuing an approach using moderately optimized resonator designs, but with a somewhat lower voltage per cavity than what is possible, with the goal of reducing cavity fabrication costs and achieving the lowest overall cost. The only presently planned use of the multi-spoke geometry is for the European Spallation Source (ESS) at Lund [16], where the intermediate-velocity section of the linac will use 32 double-spoke cavities at 352 MHz optimized for β = 0.5. 3. Electromagnetic Design It has been conventional wisdom that the relatively low gradients for low beta are intrinsic and due to a necessarily complicated design with

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

M. Kelly

190

inherently less favorable EM parameters. Peak surface magnetic fields in early low-beta cavity designs were, indeed, relatively high, with a value of Bpeak ∼25 mT/(MV/m) for the ATLAS split-ring, for example. This is particularly large when compared with the well-known TESLA elliptical cell cavity shape with Bpeak = 4.2 mT/(MV/m). Other important cavity properties, including surface electric field, shunt impedance and geometrical factor, were likewise not optimized in most early TEM cavity designs. More recent TEM cavities are readily designed using PC-based fully three-dimensional simulation codes such as MAFIA, CST Microwave Studio, HFSS and ProEngineer/ANSYS. These codes enable a careful parametric study of the cavity geometry and have resulted in major improvements in cavity EM parameters. For a pedagogically correct tutorial on low- and intermediate-beta cavity design, the reader is referred to Ref. 17. Generally, parameters of the cavity design, such as the optimum beta, the frequency and the beam aperture, require a detailed consideration of global accelerator parameters like beam current and beam loss requirements, transition energies from one cavity frequency to another, the frequency of other components such as an RFQ injector, and so on. However, nearly all new designs may make use of design features discussed here. One example of each of the three common TEM-class cavities is presented in the following sections. These are not intended to be the “best” that may be achieved but are intended to highlight generally applicable design features for modern TEM cavities. 3.1. Quarter-wave resonator In order to maximize voltage gain per cavity and accelerator “real-estate” gradient generally, and at the same time have reasonably low cryogenic losses, the EM design should be optimized with respect to four primary parameters. These are: (1) the peak surface electric field, Epeak , which will have a one-to-one correspondence with the onset of field emission, other contributions being equal. (2) The peak magnetic field, Bpeak , has a fundamental limit of ∼150–200 mT [18] and is often linked to the performance-limiting phenomenon, referred to as “quench” in both lowa Alternatively,

1230007

and high-beta cavities. (3) The geometrical shunt impedance,a RSH /Q = V 2 /ωU , where V is the accelerating voltage, ω the cavity frequency and U the stored energy, should be maximized to provide the maximum accelerating voltage at a minimum value of the cavity stored energy. Finally, there is (4) the geometry factor, G = RsQ, where Rs is the total cavity surface resistance and Q is the unloaded cavity quality factor. This is a measure of the effectiveness of the cavity geometry (without respect to specific material losses) in providing accelerating voltage, and clearly should be maximized. An excellent reference on general scaling rules for these parameters as a function of cavity beta is available [19]. The peak surface fields depend on the shape and dimensions of both the center conductor base and the aperture region. While both areas affect the peak electric and magnetic fields, the shape and dimensions of the center conductor base have the primary effect on the surface magnetic fields while the aperture region has the most influence on the peak surface electric field. Key design details for the geometry shown in Fig. 5 [20] include a tapered center conductor, with the effect of distributing the magnetic field almost uniformly over the quarter-wave stem. The conical outer conductor shape increases the total volume near the top of the cavity, reducing the peak surface magnetic field. Note that generally the real estate gradient is not sacrificed by a modest taper because ∼6 cm or more is required to make standard connections between cavity and solenoid beam aperture. The minimum values of the peak electric field are achieved, similarly, by tailoring the donut-shaped “drift tube” at the end of the center conductor. The fields should be distributed relatively uniformly over the surface. Likewise, the magnitude of the electric field on the “re-entrant nose” surfaces, where the beam enters and exits the cavity, is similar to that on the drift tube surface. EM parameters for this cavity geometry, scaled to an accelerating gradient of 1 MV/m, are shown in Table 3.

3.2. Half-wave resonator Half-wave resonators (HWRs) are under development for the acceleration of high-intensity proton

shunt impedance is sometimes defined as V 2 /2ωU , as in circuit theory.

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

Fig. 5.

191

Quarter-wave cavity for β = 0.077.

Table 3. Electromagnetic parameters for cavities in Figs. 5–7; fields for EACC = 1 MV/m; Leff = nβλ/2. Cavity

Beta (v/c)

Epeak (MV/m)

Bpeak (mT)

RSH /Q(Ω)

G(Ω)

QWR HWR Spoke

0.07 0.10 0.21

3.2 4.7 3.8

5.8 5.0 5.8

548 272 242

40 48 84

Fig. 6.

and heavy ion beams in the 0.1 < β < 0.5 velocity range. It is common to design HWR geometries using a pair of essentially coaxial cylinders and a center conductor with a “squashed” racetrack shape near the beam aperture [21]. The design shown in Fig. 6, intended for use in Project X at a frequency of 162.5 MHz, has additional design features, including

Half-wave cavity for β = 0.1.

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

M. Kelly

192

Fig. 7.

Fermilab single-spoke cavity for β = 0.2.

a ring-shaped center conductor instead of the common racetrack shape [22]. It has been pointed out recently that the intrinsic asymmetry of the racetrack section of the center conductor produces a quadrupole electric field perpendicular to the beam axis [23], which can, in the aggregate, cause a noticeable transverse emittance growth. The use of the ring shape practically eliminates the effect. There appear to be no significant negative consequences from the ring shape; in fact, shunt impedance is measurably improved. Other design features are similar to those of the quarter-wave resonator (QWR) and include tapered inner and outer conductors to reduce the surface magnetic field. As with the QWR, the angle of the taper is limited such that the real estate gradient is not sacrificed. It is also noted that the addition of coupling/rinsing/chemistry ports at the ends of the cavity can cause local enhancements to the surface magnetic field and increase the cavity Bpeak . This problem is mitigated by using a sufficiently large blend radius at the intersection, with a value of ∼1.2 cm for a 5-cm-diameter port in this case. EM parameters for the geometry shown in Fig. 6 are listed in Table 3. 3.3. Spoke resonator SC TEM-class spoke cavities have been an area of active research during the past decade, with application to CW and pulsed ion linacs required for proposed facilities worldwide. Single- and multi-spoke geometries have been developed for use with ions

over the full mass range and for ion velocities 0.15 < v/c < 0.6. Recently, optimized designs up to β ∼ 1 have been presented [24]. The geometry shown in Fig. 7 was optimized at Fermilab for 325 MHz and intended for use in the low-energy part of the Project X H-linac [25]. The transverse size of the spoke geometry is largely determined by the frequency and, as with quarter- and half-wave cavities, the distance between the accelerating gaps is mostly determined by the desired beta. The racetrack shape at the aperture region is generally more effective than a circular or elliptical cross section for the purpose of reducing Epeak . The large spoke base and the highly re-entrant end wall spread out the surface currents and provide a large volume for the magnetic field in order to minimize Bpeak . A further reduction of Bpeak is shown to be possible by introducing a racetrack shape near the base of the spoke [24]. However, reducing Bpeak by making the end walls more re-entrant (i.e. making the cavity longer) is eventually counterproductive, since this also requires additional space along the beam axis. 4. Mechanical Design The material of choice for high-performance SC cavities is a ∼3-mm-thick high-purity niobium sheet, typically formed or machined to produce niobium subcomponents and then electron-beamwelded together to form the complete niobium cavity. Other materials, such as sputtered niobium or niobium explosively bonded onto copper, are used

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

today with some success for low-beta cavity fabrication [26]; however, these are far less common. For an excellent review of alternative technologies, the reader is referred to Ref. 27 and the references therein. This overview presumes the choice of a thin wall niobium sheet as the starting point for mechanical design. Published details on low-beta cavity mechanical design and fabrication are difficult to find in the literature, since they are oftentimes left to the industrial supplier. However, in all cases the niobium cavity must be housed in a cryogenic helium vessel and, generally, there are common design criteria that should be satisfied. Three common techniques exist for housing cavities in a helium vessel: • Titanium joined to niobium using an intermediate niobium–titanium alloy transition. • Stainless steel brazed to niobium using a copper alloy. • A niobium helium vessel directly welded to the niobium cavity. In all cases, the niobium cavity and the helium vessel form an integral mechanical system that determines nearly all important cavity performance characteristics. The multiphysics capability of ANSYS is often used to analyze the complete niobium cavity and helium vessel for the following sensitivities: • The maximum stress levels in the niobium cavity and the helium vessel under anticipated pressure/temperature conditions. • The mechanical eigenfrequencies of the system. • The response of the cavity frequency to changes in the helium bath pressure. • The response of the cavity frequency to changes in the cavity field (Lorentz detuning). • The response of the cavity frequency to mechanical tuners or coarse tuning during fabrication. Allowable material stresses are, in most regions of the world, determined by local pressure vessel requirements; however, typical maximum von Mises stress values are roughly 50 MPa for a high-purity fine-grained niobium sheet, 124–138 MPa for 304/316 stainless steel and 275 MPa for grade 2 titanium, all at room temperature. Real world values are typically

193

higher [28, 29]; however, the use of the larger value of the stress, particularly as resulting naturally from cryogenic temperatures, is generally not permitted by the pressure code requirements. The lowest mechanical eigenfrequency for an ∼100 MHz thin wall QWR, such as for the geometry shown in Fig. 5, will typically be around 50 Hz and may be increased somewhat using niobium reinforcing ribs. Resonant vibrations of the center conductor can be effectively damped using a design such as the passive mechanical damper developed at INFN-Legnaro [30]. Another extremely simple and effective technique that is generally applicable, but not often used, is to electrically center the quarterwave stem, by mechanically bending during fabrication, for example. With precise centering (+/−100 microns), the frequency shifts resulting from typical vibrations can be practically eliminated [31]. Half-wave and spoke geometries of the type discussed in this review will have central conductor eigenfrequencies >100 Hz. This would appear to be a favorable situation when one is considering unwanted microphonic-induced frequency shifts; however, operational data are scarce [32]. The response of the cavity to external changes in the helium pressure, ∆f /∆p, can vary by orders of magnitude depending on the details of the mechanical design. With careful mechanical design it has been shown that ∆f /∆p can essentially be nulled [31]. In mechanical designs in which the effect is not explicitly considered, such as with spoke cavities, values of several hundred hertz per torr are common. A reasonable rule of thumb for TEM cavities operated at 4 K and with light beam loading is that a value of (∆f /∆p)/f ∼ 2–10 × 10−8 is achievable through design and will lead to cavity frequency shifts similar to or smaller than typical frequency shifts from resonant vibrations. Machines operating at 2 K or with a very large beam loading may relax requirements for ∆f /∆p. In noisy environments, such as from a liquid helium refrigerator where pressure fluctuations exceed ∼1 mbar, requirements may be more stringent.

5. Fabrication High-purity bulk niobium, available from several vendors, is presently the material of choice for fabrication of both low- and high-beta SC cavities. It

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

194

has good SC properties and is formable, machinable and weldable. Fabrication starts with a niobium sheet and bar stock, and the steps include machining, forming, rolling, welding and final surface processing, all having an impact on cavity performance. Lead/copper cavities have been built and operated at several laboratories, albeit with considerably lower performance than for bulk niobium cavities. Hydroforming or deep-drawing techniques may be used to produce complex niobium shapes and in large quantities if needed. Formed niobium parts are welded together under high vacuum, typically 2 × 10−5 torr or better, using an electron beam. Demountable joints are less common in new lowbeta TEM cavities, due to the difficulty in producing low-loss rf joints. Welded cavities intended for operations are housed inside an integral helium jacket constructed from stainless steel, titanium or “reactor grade” niobium. Recently, wire electric discharge machining (EDM) techniques have been developed together with industry and may be used to perform essentially all of the trimming on niobium cavity subcomponents before electron beam welding [33]. The primary technical benefit is that there is essentially no possibility for foreign material inclusions as compared to traditional machining. End milling, still by far more common, requires particular skill and expertise with niobium. Hydrogen contamination from EDM is not a major drawback, since cavities require degassing anyway to achieve optimal performance at 2 K. Even for 4 K operation, hydrogen degassing is recommended to avoid the possibility of hydrogen Qdisease. The recast layer left over after EDM, a combination of niobium- and brass-oxides, is 5 microns thick and should be removed using standard buffered chemical polishing for 5–10 min prior to electron beam welding or electropolishing. Generally, TEM cavity fabrication and quality assurance techniques have improved over the past decade, leading to improved cavity performance; however, the strict techniques now used to achieve surface magnetic fields with Bpeak > 120 mT in many elliptical cell cavities are not yet uniformly and systematically employed for low-beta TEM cavities. 5.1. Tuning during fabrication Simulation codes are commonly used to predict cavity frequencies to an accuracy that is certainly

1230007

M. Kelly

better than 1% of the nominal frequency. This is still ∼10 times larger than the typical slow tuner range when combined with fabrication uncertainties; the result is that essentially all lowbeta cavities require tuning adjustments during fabrication. A common technique for low-beta QWR and HWR cavities is to include a straight (nontapered) section of excess niobium material on the center conductor specifically to facilitate tuning during fabrication. This coarse tuning is performed using either standard mill machining or wire EDM to cut lengths of niobium from the straight section of the cavity center conductor. It is noted that for QWR and HWR geometries both the inner and outer conductors may need to be cut in order to keep the beam apertures on the same axis. At least a couple of techniques permit finer QWR tuning after essentially all components except the bottom of the cavity have been welded (or clamped) on. At the ANL, the dome of the QWR is trimmed in order to bring the frequency to within a few kHz of the desired value, and then electron-beam-welded. At TRIUMF, buffered chemical polishing has been used to preferentially remove niobium material from the rf surface in order to achieve the desired final frequency. Finally, after processing of the rf surface using electropolishing or buffered chemical polishing, the cavity can be squeezed or stretched inelastically at the beam ports to bring the frequency within the window of the slow tuner system. In this case, the cavity must be designed with sufficient strength in the aperture region to permit squeezing or stretching.

5.2. Chemical processing The present state of the art in surface preparation of superconducting rf cavities calls for the removal of at least a 100-micron-thick damaged layer resulting from niobium forming and machining. Electropolishing [34] (EP) has been employed with SC cavities since the early 1970s and is widely used to produce smooth surfaces free of major defects. The other common technique for damaged layer removal is buffered chemical polishing (BCP), which produces a manifestly rougher surface than EP when used on finegrained niobium material.

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

Fig. 8.

195

Electropolishing system for low-beta cavities with a coaxial geometry.

In elliptical cell cavities treated with BCP, the limiting accelerating fields are somewhat lower than those achieved with EP [35]. Similarly, most TEMclass cavities treated with EP exhibit a smaller Q-slope and lower rf losses, particularly at high gradients [36], than those treated with BCP. It is noted, however, that for moderate fields (Bpeak < 100 mT) many well-performing TEM cavities are still being produced using only BCP. At ANL, EP was used with the early split-ring cavities in ATLAS. A variation on the EP technique was proposed and applied at ANL as required by new designs with no demountable joints and with relatively small (5 cm), access ports at the ends of the cavity. This meant that the inside rf surface was no longer easily accessible for EP after the final electron beam welding. Instead, the niobium subcomponents were heavily electropolished individually and then welded together. An ∼5–10-min standard BCP was used to remove any residues from welding. The final surface was smoother than was possible using only a heavy BCP treatment [37]. However, this technique has the drawback that it reverses at least some of the smoothing from EP and the region of the final closure weld does not receive a heavy chemical treatment. A new EP system designed for low-beta SC cavities with a coaxial geometry was recently built and operated at ANL [38]. Conceptually, the design was based on that used for the EP of 1.3 GHz nine-cell elliptical cavities for the worldwide International Linear Collider development effort. The low-beta system incorporates the useful features from the elliptical cell systems including the capability for acid circulation and hydrogen gas removal during EP

and, crucially, for repeated polishing on the cavity, if needed, without producing the progressively rougher surface associated with BCP. Unlike the elliptical cell systems, the low-beta system allows polishing of the finished cavity and direct water cooling of the niobium cavity throughout the process using circulating water in the liquid helium jacket. The temperature of the entire niobium surface is controlled to an accuracy of about 1◦ C, which has been found to be critical for achieving optimal results [39]. These techniques are still new and evolving rapidly, particularly for low-beta cavities, and the practical benefits for new ion linacs are just being realized. 5.3. High-pressure rinse and clean assembly A major advance in SC cavity performance was achieved in the early 1990s at laboratories such as KEK, DESY and JLab with the introduction of high-pressure water rinsing and cleanroom assembly. These techniques are being adapted by many laboratories and institutions for low-beta TEM cavities. High-pressure water rinsing using filtered, 0.1 µm or better, deionized water in an ISO class 5 (class 100) or better clean area for at least 1 h is now standard and required practice for reliably achieving cavity surface electric fields of Epeak = 35 MV/m. Similarly, final assembly of the cavity together with the power coupler and the vacuum system hardware must also be properly performed in a suitable, ISO class 5 or better, clean area. An example of a typical single-cavity clean assembly is shown

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

196

1230007

M. Kelly

with dozens of clean SC elliptical cavities starting around the year 2001 with surface gradients exceeding Epeak > 50 MV/m (EACC = 25 MV/m) and little or no performance degradation. The low-beta SC cavities at TRIUMF and ANL, commissioned in 2006 and 2009 respectively with Epeak ∼ 35 MV/m, are presently running at field levels similar to those at the time of commissioning. 6. Recent Developments 6.1. Field performance

Fig. 9. Cleanroom assembly of a half-wave resonator at Michigan State University.

in Fig. 9. Cleanroom particle counters showing particulates at a range of sizes, such as 0.3, 0.5 and 1 micron, are now used regularly to help ensure clean assembly. Subsystems such as cavity power couplers or tuners that share the rf volume with the cavity must also be carefully cleaned. Michigan State University has used diagnostic tools [40], including surface and liquid particle counters, to assess the effectiveness of ultrasonic cleaning and water rinsing at low pressures ∼0.2 MPa. At ANL, couplers are manually high-pressure-rinsed with a nozzle pressure of ∼10 MPa in order to remove particulates prior to final assembly onto the cavity. Relatively little data exists on the maintenance of cleanliness for long-term linac operations. The best demonstration to date is FLASH at DESY [41] (formerly the TTF-VUV FEL), which was among the first to use separate cavity and cryogenic vacuum systems and modern cleaning techniques to operate with high surface electric fields. The linac has operated

Remarkable test results for several types of TEM cavities have been obtained recently. Sixteen β = 0.12 88 MHz QWR cavities from thin-walled highRRR niobium were recently built for the second section of the SPIRAL-2 linac [42]. All cavities were treated using standard BCP for 4 h, taking care to cool the cavity uniformly using water in the integral helium vessel. The cavity was polished in the vertical orientation and flipped once to increase polishing uniformity. Notably, after the BCP, these cavities showed a measureable benefit from in situ 110◦ C baking for 48 h, with a factor-of-2 drop in surface resistance at 4 K at the nominal operating field of EACC = 6.5 MV/m (Bpeak = 65 mT). Test results for individual cavities are shown in Fig. 10, where in several cases Bpeak > 100 mT. Cavities are presently being installed into the SPIRAL-2 cryomodules with the production couplers and tuners, with efforts focused on maintaining similar performance as in singlecavity tests. At FNAL, single-spoke cavities are being prototyped and tested for use in the front end section

Fig. 10. 4 K test results for 16 88 MHz β = 0.12 QWR cavities at IPNO for SPIRAL-2. Leff = βλ = 0.41 m. Note the high Q values.

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

of a high-intensity proton linac [25]. Fabrication includes the use of die forming, machining and electron beam welding starting with high-RRR (purity), 3-mm-thick niobium. A pair of bare niobium cavities optimized for β = 0.2 have been purchased by FNAL from industrial suppliers. BCP on both cavities was performed with standard 1:1:2 BCP at temperatures from 15◦ C to 17◦ C. For the second cavity, an improved system for BCP was used to provide better uniformity of the acid circulation over the cavity surface and, thus, more uniform material removal. Final surface preparation was performed using highpressure deionized water for rinsing and cleanroom assembly at the joint Argonne/Fermilab cavity processing facility. Cold test results are shown in Fig. 11. At the maximum gradient (EACC = 22 MV/m; Leff = 0.19 m), the total voltage is 4.5 MV. With an overall physical length along the beam axis of only 0.33 m, the achievable “real estate” gradient of ∼10 MV/m is comparable to values for higher-beta elliptical cell cavities. This performance is due both to the optimized shape and to the absence of significant fabrication defects. This is evidenced by the high-peak magnetic field, Bpeak ∼ 125 mT. In terms of real-world performance, this would correspond to an increase of roughly a factor-of-2 relative to previous cavities tested for this velocity region. Cold test results at both 2 and 4 K for a set of production 72 MHz β = 0.077 quarter-wave cavities for ATLAS were completed recently [43], and an example is shown in Fig. 12. Processing included

197

Fig. 12. 2 K and 4 K test results for 72.5 MHz β = 0.077 QWR cavities at ANL for ATLAS. Leff = βλ = 0.32 m.

heavy, 150 µm EP on the complete cavity with the helium vessel, the first time this technique has been used on a low-beta cavity, as discussed in Sec. 5. The maximum 4 K cavity accelerating gradient of EACC = 13.7 MV/m is approximately twice that planned for ATLAS operation and was measured following approximately 3 h of low-level multipacting conditioning and 1 h of short-pulse conditioning with the use of 4 kW of peak power. At 4 K the measured low-field quality factor of 5 × 109 gives a residual surface resistance of 2.6 nΩ, a value that is rarely achieved even for the best elliptical cell cavities. rf power to reach EACC = 7.9 MV/m was 5 W. The maximum accelerating potential, 4.3 MV, is 2–4 times higher than for quarter-wave cavities in routine operations today. The 2 K limit of EACC = 15.5 MV/m (Bpeak = 110 mT) was limited by available X-ray shielding.

6.2. 2 K operation

Fig. 11. 2 K and 4 K test results for 325 MHz β = 0.21 spoke cavity at FNAL. Leff = βλ = 0.19 m.

In addition to the high-field performance discussed above, it has also been demonstrated that low surface resistances of 5 nΩ or less can be achieved in lowand medium-beta TEM cavities when operating at 2 K. This is due to the substantially reduced Q-slope typically observed for 2 K operation, as is evident for data shown in Fig. 12. So far, at high-field levels (∼14 MV/m, 100 mT) the lowest rf losses have been achieved through the use of EP of the cavity rf surface and hydrogen degassing. This is consistent with results from elliptical cell cavities, where the low rf losses at high

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

198

gradients are best achieved when cavities are electropolished and then baked at 600◦ C–800◦C. Nearly all low-beta quarter-wave, half-wave and spoke cavities built to date were initially intended for 4 K operation; however, the results for low-beta quarter-wave cavities and medium-beta triple-spoke cavities (β = 0.5 and β = 0.62) after hydrogen degassing at 600◦C show an rf surface resistance 10 times lower in 2 K operation than in 4 K [44]. Even with the increased cost of 2 K refrigeration of roughly ∼3.5 times (both capital and operating), this nonetheless provides a major opportunity for cost savings for future CW ion linacs. Operation at 2 K is now planned even for the lowest-beta SRF cavities at FRIB, Project X and ESS. Finally, it appears to be well within reach to develop useful low-beta cavities with Bpeak ∼ 120 mT and Rs ∼ 5 nΩ at T = 2 K. This would provide a CW accelerating potential of VACC ∼ 5 MV per cavity with β ∼ 0.1. When combined with long cryomodules with high packing factors, this technology would provide an attractive option for megawatt-class CW linacs in areas of applied science and technology such as for medical isotope production and acceleratordriven systems.

1230007

M. Kelly

with intersecting spokes supported at both ends, has excellent mechanical stability [44] and may be designed so that it is nearly insensitive to helium pressure fluctuations. The coaxial half-wave geometry has similarly good mechanical properties. Quarter-wave cavities have a loading element which is unattached at one end and may be susceptible to vibrations; however, recently, a passive mechanical vibration damper, developed first at INFN-LNL [30] to directly damp microphonics, can be used with excellent results. Piezoelectric and magnetostrictive transducers, still under development, have been operated with TEM cavities in a few cases in order to compensate for the effects of microphonics. Coupler and tuner design for SRF cavities offers the opportunity for creativity, since reasonable options are often more varied than for the SRF structures themselves. However, the design for ancillary systems should be guided by some general principles. First, the device should provide ample range or capability, obviously at a reasonable cost. Second, the device should not sacrifice or jeopardize cavity performance. A brief overview of solutions is provided here, some of these proven and others yet to be fully demonstrated.

7. Microphonics, Tuners and Couplers 7.1. Microphonics

7.2. Couplers

Low-beta TEM cavity linacs, such as for the proposed rare isotope production facilities in Table 3, with moderate beam currents (∼0.5 A), will generally have loaded cavity bandwidths of a few hertz. These values are smaller-than-usual microphonic(vibration)-induced frequency shifts. For many other higher-intensity linacs with beam currents of 5 milliamps or more, the bandwidth of the cavity/coupler system is several tens of hertz or larger and is comparable to or larger than the magnitude of the frequency shifts. This presumes that proper design considerations, such as those discussed in Sec. 5, have been made. For modern low-beta TEM cavities, the problem of microphonics is best mitigated by mechanical design. Measurements indicate that these cavities are naturally stiffer than the elliptical cell geometries constructed from a similar 3 mm niobium sheet. For example, the spoke cavity geometry, composed essentially of a cylindrical outer housing

The various combinations of inductive and capacitive, fixed and variable rf power couplers have been developed recently for low- and mid-beta cavities, suitable for 1–40 kW rf power. These couplers often provide the primary means for performing cavity faster tuning by overcoupling. Previously, couplers for low-beta cavities transmitted only several hundred watts of forward power and were used to accelerate low beam currents. These couplers required only modest development effort. Generally, the choice of inductive or capacitive couplers must be made based on the application and cavity type. Capacitive coupling is generally favored for higher-power SRF applications. With clean cavities, coupling from the top of the cavity is also avoided. This leads naturally to the choice of a capacitive coupler for use with a bottom-coupled quarter-wave resonator. For half-wave geometries it is increasingly common [13] to orient the entire cavity horizontally inside the cryomodule in order to

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

199

40 kW in traveling wave mode and are being installed into the SPIRAL-2 cryomodules. Figure 14 shows a set of capacitive couplers for a 72 MHz β = 0.072 cavity/cryomodule upgrade of ATLAS [47]. This design tested with up to 4 kW forward power in full reflection uses a pair of ceramic windows, one at room temperature and one actively cooled to 80 K, along with a 7 cm variable bellows to adjust the external Qext in the range between 106 and 1010 . The two-window design provides a relatively simple and compact assembly for handling while in the cleanroom. Fig. 13.

Fixed capacitive power coupler for SPIRAL-2.

7.3. Tuners insert a capacitive coupler up from the bottom into the high electric field region of the cavity. The choice of fixed or variable coupler is also application-dependent and should consider both the rf conditioning and operational requirements, since the former is also critical to real-world cavity performance. Presently most low- and mid-beta cavities benefit substantially from high-power pulse conditioning with several kilowatts of rf power into the cavity, because field emission is rarely eliminated completely. The variable coupler may also offer the possibility to effectively condition low-level cavity multipacting. Considering the high cost of rf power, the increased complexity of the variable coupler may be justified. A recent example of a fixed capacitive coupler for SPIRAL-2 [45, 46] is shown in Fig. 13. This design is based on a roughly 4-cm-diameter coaxial line with a single warm 6-mm-thick ceramic window. The couplers have been tested with forward power as high as

Fig. 14.

Important considerations for slow tuner design include tuning range, sensitivity, response time, physical size, serviceability, reliability (related to the number and complexity of moving parts), impact on cavity performance and cost. TRIUMF reported on an operational cavity tuner mechanism based on a motor-driven linear actuator [48] with a penetration through the lid of the cavity cryostat, extending downward to a flexible tuning plate on the bottom of the cavity. It is noteworthy that this single device can be operated so as to compensate for frequency deviations at rates up to 30 Hz. Michigan State University uses a related system for the ReA3 accelerator and the planned FRIB. This tuner has a penetration up from the bottom of the cryostat, driving a tuning plate with both a motor and a fast mechanical piezoelectric element [49]. Motor-driven systems must, in general, be very carefully designed in order to avoid the inducing microphonics by operation of the tuner itself.

4 kW capacitive power couplers for ATLAS (left) with 7 cm variable bellows (right).

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

200

For slow tuning in SPIRAL-2, a pair of highRRR 3-cm-diameter niobium plungers have been developed and are inserted into the coupling ports near the high magnetic field region of the quarterwave cavity [50, 51]. One fixed plunger provides a coarse adjustment, while the second moveable plunger provides smaller dynamic tuning. The plungers are hollow and filled with liquid helium. When used with the SPIRAL-2 β = 0.12 cavity, a pair of plunger provides 54 kHz of the total tuning range. Both the variable and fixed geometries have been successfully tested. This technique has the benefit that it is generally applicable and requires only suitable ports on the cavity through which the plunger can be inserted. An important caveat is that a low-loss rf seal is required at the cavity port. At Argonne, slow tuners for quarter-wave cavities use a helium-gas-driven pneumatic bellows [52] to squeeze the cavity near the beam axis. The design can provide large compressive forces (>20 kN), is compact, sensitive and reliable, and has only one moving part (the bellows). Pneumatic actuation also helps to avoid the introduction of microphonics from the tuner itself.

1230007

M. Kelly

demonstrated, but is the subject of ongoing studies for Project X at Fermilab and ESS at Lund. In SRF cavities of all types, magnetic fields exert an outward force on the cavity surface, while electric fields exert an inward force on the cavity walls. Deflections from both of these cause the cavity frequency to shift downward in response to the rf field. Values of the Lorentz detuning coefficient vary widely for low-beta cavities in the range of 1–6 Hz/(MV/m)2 . The value can be reduced somewhat by increasing the overall stiffness of the cavity or the cavity/tuner combination. The most detailed experimental study to date on pulse operation in a low-beta cavity has been performed at FNAL for a single-spoke cavity [53]. The technique uses a lever tuner mounted near one of the beam port flanges of the helium vessel and uses a set of four piezoelectric actuators to push on the beam flange. The total tuning range of 240 kHz was more than sufficient to compensate for pressure changes in the helium bath. With a millisecond-scale time constant, the tuner was also used to actively compensate for Lorentz detuning shifts up to 2 kHz with the cavity operated in pulsed mode up to accelerating gradients of EACC = 20 MV/m (Leff = nβλ/ 2 = 0.19 m).

7.3.1. Fast tuners Fast tuners are generally required for linacs with low-to-moderate beam currents or when overcoupling is not used. They are less well developed for present low- and mid-beta cavities. The variable reactance (VCX) fast tuner at ANL is reliable and cost-effective; however, it requires a separate penetration into the cavity rf space and has limited switching power and frequency range and is thus not planned for high-gradient cavities. Other mechanical fast tuners based on piezoelectric or magnetostrictive elements developed at places such as MSU, SARAF/RI and ANL appear promising but are yet to be demonstrated in routine operations. In low-beta cavities intended for pulsed operation, the dynamic Lorentz detuning will typically be the dominant source of cavity frequency shifts. Since the physical process is fundamentally different from the pressure-induced frequency shifts, a different approach to cavity design and tuning is required. Pulsed operation of TEM cavities has yet to be

8. Conclusion Low- and medium-beta cavities and subsystems are a relatively mature technology, with a unique set of features including high gradients, large acceptance, low rf losses and good mechanical properties. Together with elliptical cell cavities, reliable superconducting cavities are available for the entire velocity range. New projects using designs for 0.1 < β < 0.5 are proposed or under construction worldwide. Scientists and engineers working on low- and mediumvelocity cavities continue to advance the state of the art. The field is driven by goals for new capabilities such as very high beam currents and the potential for large cost savings. Areas where major gains (factors of 2–3) are barely being realized include increased gradients through better EM design and cavity processing. The employment of the latest techniques for projects under construction will surely drive new applications such as large hadron linacs for basic science, proton linacs for accelerator-driven systems, and radioisotope production for medicine.

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

Acknowledgments The author is grateful to K. Shepard, J. Delayen, A. Facco, G. Olry, R. Laxdal, S. Bousson, V. Palmieri and others for earlier excellent presentations, discussions and papers on low-beta rf superconductivity. References [1] H. A. Schwettman et al., Proc. 5th Int. Conf. High Energy Accelerators (Frascati, 1965), p. 690. [2] T. Furuya, Development of superconducting rf technology, in Reviews of Accelerator Science and Technology, Vol. 1 (2008), pp. 211–235. [3] H. Padamsee, K. W. Shepard and R. Sundelin, Physics and accelerator applications of rf superconductivity, Annu. Rev. Nucl. Part. Sci. 43, 635–686 (1993). [4] D. Storm, Review of low-beta superconducting rf structures, in Proc. SRF 1993 (Newport News, Virginia, 1993), p. 216. [5] L. M. Bollinger, Low-β SC linacs: Past, present and future, in Proc. LINAC 1998 (Chicago, IL, 1998). [6] A. Facco, Low and medium beta SC cavities, in Proc. EPAC 2004 (Lucerne, Switzerland, 2004). [7] C. Piel, K. Dunkel, F. Kremer, M. Pekeler, P. Stein, B. Gladbach and D. Berkovits, Phase 1 commissioning status of the 40 MeV proton/deuteron accelerator SARAF, in Proc. EPAC08 (Genoa, Italy, 2008). [8] M. P. Kelly, J. D. Fuerst, S. Gerbick, M. Kedzie, K. W. Shepard, G. P. Zinkann and P. N. Ostroumov, Superconducting quarter-wave resonators for the ATLAS energy upgrade, in Proc. LINAC08 (Victoria BC, Canada, 2008). [9] R. E. Laxdal, R. J. Dawson, M. Marchetto, A. K. Mitra, W. R. Rawnsley, T. Ries, I. Sekachev and V. Zvyagintsev, ISAC-II superconducting linac upgrade — Design and status, in Proc. LINAC08 (Victoria, BC, Canada, 2008). [10] R. Ferdinand, T. Junquera, P. Bosland, P. E Bernaudin, H. Saugnac, G. Olry and Y. G´ omezMart´ınez, The spiral-2 superconducting linac, in Proc. LINAC08 (Victoria, BC, Canada, 2008). [11] P. N. Ostroumov and K. W. Shepard, Correction of beam steering effects in low velocity superconducting quarter-wave resonators, PRST Accel. Beams 4, 110101 (2001). [12] H. A¨ıt Abderrahimh, J. Galambos, Y. Gohar, S. Henderson, G. Lawrence, T. McManamy, A. C. Mueller, S. Nagaitsev, J. Nolen, E. Pitcher, R. Rimmer, R. Sheffield and M. Todosow, Accelerator and target technology for accelerator-driven transmutation and energy production, DOE White Paper, Sep. 17, 2010. [13] F. Orsini et al., Preliminary results of the IFMIF cavity prototype tests in vertical cryostat and cryomodule development, in Proc. SRF 2011 (Chicago, Illinois, 2011).

201

[14] S. D. Holmes, S. D. Henderson, R. Kephart, J. Kerby, I. Kourbanis, V. Lebedev, S. Mishra, S. Nagaitsev, N. Solyak and R. Tschirhart, Project X functional requirements specification, in Proc. IPAC 2012 (New Orleans, LA, 2012). [15] A. Facco et al., Superconducting resonator development for the FRIB and ReA linacs at MSU: Recent achievements and future goals, in Proc. IPAC 2012 (New Orleans, LA, 2012). [16] H. Danared, Design of the ESS accelerator, in Proc. IPAC 2012 (New Orleans, LA, 2012). [17] J. R. Delayen, Medium-β superconducting accelerating structures, in Proc. SRF 2001 (Tsukuba, Japan, 2001). [18] H. Padamsee, J. Knobloch and T. Hays, RF Superconductivity for Accelerators, 2nd edn. (Wiley-VCH, 2008), p. 101. [19] J. R. Delayen, Low and intermediate beta cavity design — A tutorial, in Proc. SRF 2003 (Travemunde, Germany, 2003). [20] B. Mustapha, P. N. Ostroumov, Electromagnetic optimization of a quarter-wave resonator, in Proc. LINAC 2010 (Tsukuba, Japan, 2010). [21] J. R. Delayen, superconducting accelerating structures for high-current ion beams, in Proc. LINAC 1988 (Newport News, VA, 1998). [22] B. Mustapha, P. N. Ostroumov and Z. A. Conway A ring-shaped center conductor geometry for a halfwave resonator, in Proc. IPAC 2012 (New Orleans, Louisianna, 2012). [23] I. G. Gonin et al., Effects of the rf field asymmetry in SC cavities of the project X, in these proceedings, paper WEPPC047. [24] J. R. Delayen, S. U. De Silva and C. S. Hopper, Design of superconducting spoke cavities for high velocity applications, in Proc. PAC 2011 (New York, NY, 2011). [25] I. G. Gonin et al., Single spoke cavities for the lowenergy part of the CW linac of Project X, in Proc. IPAC’10 (Kyoto, Japan, 2010). [26] S. Calatroni et al., Superconducting sputtered Nb/Cu QWR for the HIE-ISOLDE project at CERN, in Proc. LINAC 2010 (Tsukuba, Japan, 2010). [27] V. Palmieri, New developments in low beta superconducting structures, in Proc. SRF 2005 (Gif-surYvette, France, 2005). [28] R. P. Walsh, R. R. Mitchell, V. T. Toplosky and R. C. Gentzlinger, Low temperature tensile and fracture toughness properties of SCRF cavity structural materials, in Proc. SRF 1999 (Santa Fe, NM, 1999). [29] T. J. Peterson, H. F. Carter, M. H. Foley, A. L. Klebaner, T. H. Nicol, T. M. Page, J. C. Theilacker, R. H. Wands, M. L. Wong-Squires and G. Wu, Pure niobium as a pressure vessel material, in Proc. 2009 Cryogen. Eng. Conf. (Tucson, AZ, 2009).

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

202

[30] A. Facco, K. W. Shepard and G. P. Zinkann, A vibration damper for a low velocity four gap accelerating structure, in Proc. SRF 1999 (Santa Fe, NM, 2009). [31] Z. A. Conway, M. P. Kelly, P. N. Ostroumov and K. W. Shepard, Tuning techniques for low-beta SRF cavities, in Proc. SRF 2011 (Chicago, IL, 2011). [32] M. P. Kelly, J. D. Fuerst, M. Kedzie and K. W. Shepard, Cold tests of a superconducting co-axial half-wave cavity for RIA, in Proc. LINAC 2004 (L¨ ubeck, Germany, 2004). [33] M. P. Kelly, Z. A. Conway, S. M. Gerbick, M. Kedzie, R. C. Murphy, B. Mustapha, P. N. Ostroumov and T. Reid, SRF advances for ATLAS and other β < 1 applications, in Proc. SRF 2011 (Chicago, IL, 2011). [34] H. Diepers, et al., A new method of electropolishing niobium, Phys. Lett. A 34 (1971). [35] L. Lilje, Nucl. Instrum. Meth. A 516, 2–3 (2004). [36] M. P. Kelly, Processing and test results for SC drift-tube cavities, in Proc. RFSC-Limits Workshop (Argonne, IL, 2004). [37] R. L. Geng, in Proc. 9th Workshop on RF superconductivity (Santa Fe, New Mexico, 1999). [38] S. M. Gerbick, M. P. Kelly, R. C. Murphy and T. Reid, A new electropolishing system for low-beta SC cavities, in Proc. SRF 2011 (Chicago, IL, 2011). [39] H. Tian and C. E. Reece, Quantitative EP studies and results for SRF Nb cavity production, in Proc. SRF 2011 (Chicago, IL, 2011). [40] L. Popielarski, L. Dubbs, K. Elliott, I. Malloch, R. Oweiss and J. Popielarski, Cleanroom techniques to improve surface cleanliness and repeatability for SRF coldmass production, in Proc. IPAC 2012 (New Orleans, LA, 2012). [41] L. Lilje, Experience with the TTF, in Proc. 2005 Part. Accel. Conf. (Knoxville, TN, 2005). [42] P.-E. Bernaudin, P. Bosland, R. Ferdinand and Y. Gomez-Martinez, Status of the SPIRAL 2 superconducting LINAC, in Proc. IPAC 2010 (Kyoto, Japan, 2010). [43] Z. A. Conway et al., to be published in Proc. LINAC 2012.

1230007

M. Kelly

[44] M. P. Kelly, Status of superconducting spoke cavity development, in Proc. SRF 2007 (Beijing, China 2007). [45] E. Rampnoux et al., RF power coupler development for superconducting spoke cavities at Nuclear Physics Institute in Orsay, in Proc. PAC09 (Vancouver, BC, Canada, 2009). [46] Y. G´ omez Mart´ınez, T. Cabanel, J. Giraud, R. Micoud, M. Migliore, J. Morfin, F. Vezzu, P. Bosland, P.-E. Bernaudin, R. Ferdinand and G. Olry, Power couplers for SPIRAL 2, in Proc. SRF 2011 (Chicago, IL, 2011). [47] M. P. Kelly, S. M. Gerbick, P. N. Ostroumov, J. D. Fuerst and M. Kedzie, A new fast tuning system for the ATLAS intensity upgrade cryomodule, in Proc. LINAC 2010 (Tsukuba, Japan, 2010). [48] R. E. Laxdal, R. J. Dawson, M. Marchetto, A. K. Mitra, W. R. Rawnsley, T. Ries, I. Sekachev and V. Zvyagintsev, ISAC-II superconducting linac upgrade — Design and status, in Proc. LINAC08 (Victoria, BC, Canada, 2008). [49] J. Wlodarczak, P. Glennon, W. Hartung, M. Hodek, M. J. Johnson, D. Norton and J. Popielarski, Power coupler and tuner development for superconducting quarter-wave resonators, in Proc. LINAC08 (Victoria, BC, Canada, 2008). [50] D. Longuevergne, S. Blivet, G. Martinet, G. Olry and H. Saugnac, A novel frequency tuning system based on movable plunger for SPIRAL2 high-beta superconducting quarter-wave resonator, in Proc. LINAC08 (Victoria, BC, Canada, 2008). [51] S. Bousson, Advances in SRF for low-β ion linacs, in Proc. SRF 2011 (Chicago, IL, 2011). [52] G. P. Zinkann, S. Sharamentov and B. Clifft, An improved pneumatic frequency control for superconducting cavities, in Proc. PAC05 (Knoxville, TN, 2005). [53] Y. Pischalnikov, E. Borissov, T. Khabiboulline, R. Madrak, R. Pilipenko, L. Ristori and W. Schappert, Tests of a tuner for a 325 MHz SRF spoke resonator, in Proc. PAC 2011 (New York, NY, 2011).

January 10, 2013

13:13

WSPC/253-RAST : SPI-J100

1230007

Superconducting Radio-Frequency Cavities for Low-Beta Particle Accelerators

Michael Kelly has been a superconducting radiofrequency scientist at Argonne National Laboratory since 1999. Following his early work as a graduate student on the SRF booster accelerator at the University of Washington Nuclear Physics Laboratory, he began work at Argonne to develop the first systems to process, clean and assemble modern low-beta superconducting cavities. He led developments on several new SRF structures required to bridge the velocity region between early low-beta cavities and elliptical cell cavities for electron accelerators. Dr. Kelly has also been the principal investigator for SRF activities at Argonne for the International Linear Collider (ILC).

203

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

This page intentionally left blank

1203001˙book

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Reviews of Accelerator Science and Technology Vol. 5 (2012) 205–226 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300083

Cryogenic Technology for Superconducting Accelerators Kenji Hosoyama Accelerator Division, National Laboratory for High Energy Physics, Oho, Tsukuba-shi, Ibaraki-ken, 305 Japan [email protected] Superconducting devices such as magnets and cavities are key components in the accelerator field for increasing the beam energy and intensity, and at the same time making the system compact and saving on power consumption in operation. An effective cryogenic system is required to cool and keep the superconducting devices in the superconducting state stably and economically. The helium refrigeration system for application to accelerators will be discussed in this review article. The concept of two cooling modes — the liquefier and refrigerator modes — will be discussed in detail because of its importance for realizing efficient cooling and stable operation of the system. As an example of the practical cryogenic system, the TRISTAN cryogenic system of KEK Laboratory will be treated in detail and the main components of the cryogenic system, including the high-performance multichannel transfer line and liquid nitrogen circulation system at 80 K, will also be discussed. In addition, we will discuss the operation of the cryogenic system, including the quench control and safety of the system. The satellite refrigeration system will be discussed because of its potential for wide application in medium-size accelerators and in industry. Keywords: Cryogenics; cooling; refrigerator; satellite refrigerator; superconducting magnet; superconducting cavity.

1. Introduction

in the accelerator field, compared with other wellestablished LTS superconducting devices. We will not include discussion of new superconductor types in this article. The Nb–Ti and Nb3 Sn conductors, two notable LTSs, are used for SCMs. The critical temperatures of Nb–Ti and Nb3 Sn are about 10 K and 18 K, respectively. Pure Nb is used for accelerator SCCs and its critical temperature is 9.25 K. For cooling the normal-conducting magnets and cavities operated at room temperature, water is commonly used as coolant and the specific heat of the water is utilized for cooling. But, for the cooling of LTS devices, the latent heat of vaporization of liquid helium or the specific heat of cold supercritical helium is used. The use of helium, which is the only cryogenic to remain unsolidified below 14 K, is mandatory. The helium refrigerators used for cooling the superconducting devices can be classified into large, medium, small, and very small refrigerators (socalled “cryocoolers”), depending on the cooling power. Large helium refrigerators have been designed and constructed for large accelerators like Tevatron,

1.1. Cooling of superconducting magnets and cavities Superconducting devices such as magnets (SCMs) and cavities (SCCs) are nowadays widely used as the main components of accelerators, because they can create higher magnetic and electrical fields than normal-conducting devices, at reduced electrical power consumption. On the other hand, they must be cooled and their temperature kept lower than the critical temperature Tc in order to operate in the superconducting state. Owing to the discovery of new high-temperature superconductors (HTSs) like yttrium barium copper oxide (YBCO) and bismuth strontium calcium copper oxide (BSCO), critical temperatures Tc can reach 90 K and 110 K, respectively. For the family of low temperature superconductors (LTSs), magnesium diboride (MgB2 ) has the highest Tc — 29 K. Expectation of wide application of these new superconductor types is increasing with the possibility of operation at higher temperature with a higher performance potentiality. However, practical applications of these HTSs are limited, especially 205

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

206

HERA, RHIC, CEBAF, TRISTAN, KEKB, LEP or LHC, and operated for a long time. The LHC 2 K helium refrigeration system presently in operation is the largest refrigeration system ever built in the world. Larger helium refrigeration systems are under study for future projects like the International Linear Collider (ILC) or the High-Energy LHC (HE-LHC). Small and medium helium refrigerators have been constructed in large numbers and extensively operated in the accelerator laboratories for the cooling and testing of superconducting devices. Recently, cryocoolers have been used for cooling various kinds of superconducting magnets, like NMR magnets for medical application, silicon crystal growth magnets for industrial application, and smallsize beam-line magnets for accelerator application. 2. Helium Cryogenic System

to be recovered by the liquefied cycle. A liquefier therefore has to supply nonisothermal refrigeration between 4.5 K and 300 K. The helium refrigerator produces liquid helium which has been vaporized by the superconducting heat loads, but cold saturated vapor is recovered by the refrigerator cycle, i.e. a refrigerator has to supply isothermal refrigeration at 4.5 K. Usually, a helium liquefier can be used as a refrigerator and vice versa. We want to discuss the operation modes of liquefiers and refrigerators in some detail, because this concept is important for designing the system. As an example of a practical cryogenic system, the cryogenic system for the TRISTAN [1] superconducting RF cavity will be discussed in detail. The important issues in realizing stable and reliable operation of the cryogenic system in terms of the instrumentation, safety, and device protection will also be discussed.

2.1. Cooling principle

2.2. Simple helium cryogenic system

We will discuss the cooling principle of helium liquefiers and refrigerators in this section. The helium liquefier produces liquid helium which has been vaporized and warmed up to room temperature by the superconducting device’s heat loads, and has

A typical cryogenic system for cooling a superconducting device is shown in Fig. 1. The cryogenic system consists of a helium liquefier/refrigerator composed of a cold box and a helium gas circulation compressor unit, a liquid helium storage dewar, a Helium Gas Storage 150 bar Pure Helium

Liq. Nitrogen Storage Cooling Tower Helium Gas Storage

Impure Helium

LN2

Helium Purifier

Gas Bag

Recovery/Purifier Compressor

Cooling Water

Liquid Transfer Line

Cold Box

SCM/SCC

Helium Gas Circulation Compressor Liquid Helium Dewar

Fig. 1.

Typical helium cryogenic system.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

helium gas recovery and purifier system, a helium gas storage tank, a liquid nitrogen storage tank, a liquid helium distribution system sending liquid helium to the device, and a cooling water system. The helium gas is compressed by the circulation helium gas compressor unit from 1 bar to about 16 bar. The compression heat is removed by the cooling water system. Before entering the cold box, as the compressor unit is oil-lubricated, oil has to be removed by dedicated filters. The vacuum-insulated cold box contains heat exchangers, turbo expanders, piping, and flow control valves. The cold box heat exchangers recover the enthalpy of the different cold return flows and precool the high-pressure supply helium gas. An 80 K liquid nitrogen precooler could be added to the heat exchange process. Additional cooling capacity is produced at different temperature levels by turboexpanders in which the enthalpy of the gas is adiabatically extracted via an expansion with production of mechanical work. At the outlet of the heat exchanger train, the high-pressure helium flow is finally expanded to about 1 bar in a Joule–Thomson (J–T) valve. This final expansion produces a mixture of saturated liquid and vapor. The fraction of liquid produced directly depends on the temperature of the helium before the final expansion. At 16 bar, the J–T expansion starts to produce saturated liquid at 1 bar only if the temperature before the J–T valve is below 7.7 K. The saturated liquid helium is then separated and stored in the liquid helium dewar. The saturated helium vapor is sent back to the cold box, where its enthalpy is recovered in the low-pressure heat exchanger stream.

Fig. 2.

207

The liquid helium stored in the dewar is supplied through the vacuum-insulated helium transfer line to the SCM or SCC, where it is vaporized by the heat loads. Depending on the way to recover the vaporized helium, via the gas bag after warmup to room temperature or directly via the cold box in a saturated condition, the system will work respectively in “liquefaction” or “refrigeration” mode. We will discuss these modes in detail later, as they definitely impact the design of the cryogenic system. 2.3. Main components of the helium refrigerator The major components of the helium liquefier/ refrigerator include the helium compressors, the turboexpanders, the heat exchangers, and the liquid helium transfer lines.

2.3.1. Oil-lubricated screw compressors Screw compressors are used widely for the helium liquefier/refrigerator system. Although their efficiency, i.e. isothermal efficiency, is not so high, about 50%, such compressors are often used because of their high reliability for long-term operation and their easy maintenance compared to piston compressors, which are more mechanically complex. The screw compressor consists of two cylindrical rotors with multiple helical lobes or grooves installed in a casing, as shown in Fig. 2. The helium gas enters the suction side into the volume between the rotors and the casing, moves through the threads as the screws rotate, and is compressed at the end of the

Screw-type helium compressor.

December 24, 2012

14:49

208

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

screws. Oil is injected into the rotors for lubrication, sealing, and helium cooling during compression. Two screw compressors could be combined in series, in a so-called two-stage arrangement, in order to reach the required pressure ratio with a single unit. Oil contamination of the gas during compression must be fully removed prior to being sent to the cold box; otherwise the remaining oil will stick on the surfaces of heat exchangers and piping, creating a degradation of the heat exchanger performance and plugging of the pipes. The oil removal system usually consists of a primary liquid oil separator, followed by three or four stages of oil-coalescing filters, and completed by an active charcoal absorber trapping the oil’s volatile components.

2.3.2. Gas-bearing turboexpander Historically, reciprocating expansion engines were used in the early stage of the helium liquefier/ refrigerator. George Claude of France developed the liquefier [17] and later Dr. Samuel C. Collins of MIT developed a helium liquefier with high-efficiency reciprocation expanders. Such reciprocating expansion engines were adapted in 24 satellite refrigerators for the Tevatron cryogenic system, demonstrating high efficiency during the more than 10 years of operation. However, regular replacement of the wear parts was needed during periodic maintenance to guarantee stable and efficient operation. Today, modern turboexpanders are using static or dynamic gas bearings requiring no wear parts, making them maintenance-free. The first effective turboexpander was designed by P. Kapiza. Later the turboexpanders were used for air separation plants. Commercially available helium liquefiers with high-performance turbo expanders are mainly manufactured by two companies, Air Liquide in France and Linde in Switzerland. Hitachi and Kobe Steel in Japan have developed helium liquefiers with turboexpanders which were supplied to Japanese universities and research institutes. Figure 3 shows a turboexpander made by Hitachi, used for a helium refrigerator cooling KEK superconducting RF cavities. The high-pressure helium gas is throttled by the nozzles and its enthalpy is changed to kinetic energy, i.e. it forms a high-speed gas stream which is injected into the turbine wheel. The momentum of the gas is transferred

Fig. 3.

Turboexpander for the helium refrigerator.

to the wheel, creating the turbine wheel rotation. The gas in the turbine wheel is further expanded in the turbine vane, imparting rotating force to the wheel. Then the expanded gas is sent to the diffusor, in which the velocity is converted to pressure. The work extracted at the turbine wheel is used to compress the gas at the brake compressor and is absorbed as heat by the cooling water. The highspeed rotating part of the turbine, i.e. the turbine wheel, the shaft, and the brake compressor wheel, is supported without contact by radial and thrust gas bearings. 2.3.3. Heat exchanger The heat exchanger is also a key component of high-performance liquefier/refrigeration systems. It is designed to effectively exchange the heat between two gas flows: high-pressure flow gas and lowpressure return gas. The heat transfer surface area of the heat exchanger must be large so as to increase

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

Fig. 4.

209

Aluminum plate-fin–type heat exchanger.

the efficiency, and at the same time the pressure drop in the heat exchanger, especially in the return line, must be minimized. There are many different types of heat exchanger configurations. Aluminum platefin heat exchangers are usually employed for practical helium liquefier/refrigeration systems. Typical structures of aluminum plate-fin counterflow heat exchangers are shown in Fig. 4. A unit cell of the heat exchanger is seen in the figure (left). A gas flow through the channel surrounded by separation plates and side bars and a corrugated aluminum plate is set in the gas flow channel to increase the heat transfer surface. Through stacking the cells, the heat exchanger with three flow channels is assembled by brazing, as shown in Fig. 4 (right).

negligible compared to the SSM/SCC heat loads. We need a low-heat-loss distribution system consisting of a transfer line and cold control valves. This issue will be discussed in more detail later. 2.4. Efficiency of helium refrigerator COP Usually, the efficiency of a refrigerator can be expressed by its coefficient of performance (COP), which corresponds to the ratio of cooling power Qc to the required input power W . For an ideal refrigerator operating with a reversible Carnot cycle, the COP is given by the equation [16, 18, 19] COP =

2.3.4. Liquid helium transfer line The liquid helium transfer line is also an important component of the helium liquefier/refrigerator system. The liquid helium produced at the J–T valve in the cold box is transferred to the liquid helium dewar and finally sent to the SSM/SCC cryostat for cooling the superconducting devices. In the case of the refrigerator mode, the vaporized helium cold gas in the SSM/SCC cryostat must be sent back to the cold box through the transfer line. In Fig. 1, for simplicity, only one cryostat is shown, but in a practical helium cooling system multiple SSM/SCCs must be cooled. In this case we need a liquid helium distribution system with transfer lines and flow control valves. The helium dewar is usually placed near the cold box but the SSM/SCC cryostat is placed far away from it. Generally, in practical cryogenic systems, the heat losses at the distribution system are not

T2 Qc = . W T1 − T2

The COP of an ideal refrigerator depends only on the cooling temperature T2 and the compressor operation temperature T1 . The ratio of the COP of practical refrigerator to the COP of an ideal one is called the %Carnot. Table 1 gives the COPs of ideal and practical refrigerators for cooling temperature corresponding to liquid nitrogen, liquid hydrogen, and liquid helium. In the case of an ideal refrigerator working at 300 K, to get a cooling power Qc of 1 W at 77 K or 4.2 K, we need an input power W of about 3 W or 70 W, respectively. Table 1.

N2 H2 He

COPs of ideal and practical refrigerators.

T (K)

COF (ideal)

COF (practical)

%Carnot

77.4 20.4 4.2

0.35 0.073 0.014

0.1–0.16 0.013–0.025 0.0013–0.0058

28–45 17–34 9.3–30

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

210

The %Carnot of practical helium refrigerators is not so high — about 10% for a medium-size refrigerator (100 W at 4.5 K) and about 30% for a large refrigerator (10 kW at 4.5 K), mainly due to the low efficiency of the helium compressor. This means that for a cooling power of 1 W at liquid helium temperature, we need about 230 W, even with the most efficient helium refrigerator, such as the helium refrigerators for the LHC cryogenic system. We need more than 500 W/W in the mediumsize refrigerators (300 W at 4.4 K or 100 L/h).

2.5. Component size of the cryogenic system The density ratio between saturated liquid helium at 4.2 K and gaseous helium (300 K and 1 bar) is about 750. Consequently, the size of components, depending on their operating temperature and pressure, can vary to large extents. As an example, a dewar of about 400 L will be required to store a helium inventory of 50 kg. The same inventory will require 20 m3 helium gas storage (300 K, 16 bar) or a 400 m3 gas bag (300 K, 1 bar). In a 300 W and 4.4 K cooling power refrigerator, about 400 L of liquid helium is liquefied at a rate

Fig. 5.

of about 0.1 L/s. This corresponds to the filling of a typical glass of water in about 1 s. This flow rate is relatively modest and the pipe diameter required for carrying this amount of liquid helium at 1 bar in the transfer line is usually small.

3. Practical Cryogenic Systems As an example of a practical cryogenic system for accelerator application, we will discuss the cryogenic system of superconducting RF cavities for the TRISTAN electron–positron collider at KEK, which was designed and operated for more than 20 years. 3.1. Cryogenic system for the TRISTAN superconducting RF cavity A cryogenic system for the TRISTAN superconducting RF cavity [5–7] was designed and constructed for cooling the 32 five-cell superconducting RF cavities (508 MHz) in 16 cryostats required to upgrade the electron and positron beam energy of TRISTAN from 27 GeV to 32 GeV. Figure 5 shows the TRISTAN superconducting RF cavities installed in a horizontal cryostat.

TRISTAN superconducting RF cavities in the cryostat.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

Operation of the cryogenic system was started in 1988 and continued until the shutdown of the TRISTAN in 1995, following an integrated operation time of 38,000 h in 7 years [11]. After the termination of the TRISTAN project, construction of a high-luminosity double-ring asymmetric electron–positron (8 GeV × 3.5 GeV) collider, the KEK B-Factory (KEKB) [3], started in 1994 as a five-year project in the TRISTAN tunnel. Eight single-cell higher-order modes (HOMs)–damped superconducting acceleration RF cavities integrated in eight cryostats were installed in the high-energy electron ring (HER). The operation of this new equipment started in 2000. The cryogenic system, which was designed and constructed and operated for the TRISTAN superconducting RF cavity, was reused for cooling the KEKB superconducting RF cavities. In 2007 two superconducting crab cavities [4, 14] were added and installed near the KEKB superconducting RF cavities. The operation of the cryogenic system continued until 2010, the year of the KEKB shutdown. Its total operation time was about 100,000 h, over more than 20 years. The superconducting RF cavities for KEKB and the cryogenic system will be reused for the Super-KEKB project in the future. 3.2. Design heat load of the cryogenic system The heat load of the TRISTAN cryogenic system consists of the static and the RF heat load, as shown in Fig. 6. The total heat load of about 4 kW at 4.4 K comes from about 1 kW of static heat loss of 16 cryostats and the transfer line, which is to be added about 3 kW of the RF heat load of the superconducting cavities under design conditions with an intrinsic quality factor Q0 of 109 and an acceleration field Eacc of 5 MV/m. The RF loss of the cavity during the RF oper2 and 1/Q0 , where the ation is proportional to Eacc intrinsic quality factor Q0 of the cavity is defined as the ratio of the stored energy and power loss per RF cycle. In the design of the cryogenic system, we decided to apply a safety margin of 50% for taking into account a possible degradation of the Q0 value over several years of operation and consequently to install a nominal cooling capacity of 6.5 kW at 4.4 K. The superconducting RF cavities were operated at the

211

Fig. 6. Heat load of the TRISTAN superconducting RF cavities.

designed acceleration field on average for about seven years, with heat loads corresponding to about half of the design value. The average quality factor of the RF cavities was better than expected and we had enough safety margins for the operation of the cryogenic system. 3.3. Overview of the cryogenic system A schematic flow diagram of the cryogenic system for the TRISTAN superconducting RF cavity is shown in Fig. 7. The cryogenic system consists of: • A large helium refrigeration system with a helium gas compressor unit and cold boxes; • A liquid helium dewar; • 16-cryomodule superconducting RF cavities; • A liquid nitrogen circulation system with a compressor unit and a cold box; • A large-scale liquid helium and nitrogen distribution system with a long transfer line; • A helium gas recovery and purification system; helium gas storage tanks; • A large-capacity liquid nitrogen storage tank.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

212

Fig. 7.

Cryogenic system for the TRISTAN superconducting RF cavity.

The main parameters of these components are shown in Table 2. The operation of the cryogenic system is controlled by a process control computer installed near the helium compressor units [8]. The cold boxes and compressor units are installed in the building at ground level. The liquid helium and nitrogen produced by the cold boxes at ground level are transferred through the long transfer line system (total length about 380 m) to the cryomodules of superconducting RF cavities in the underground tunnel. 3.3.1. Helium compressor unit The helium compressor unit consists of six oillubricated compressors (made by Mayekawa) and four stages of an oil filter, as shown in Fig. 8. Oil-lubricated screw compressors were chosen to attain reliability for long-term operation. The process helium gas is compressed from 1.05 bar to 5 bar by the first-stage compressors. The oil in the compressed gas is removed by the oil separator and cooled by the water heat exchanger at the

Table 2. system.

The main parameters of the TRISTAN cryogenic

He Ref. cold box Refrigeration capacity Number of truboexpanders Bearing type

Hitachi 8 kW at 4.4 K 5 Gas bearing

Compressor unit Number of compressors Type Mycom pressure (Mpa) Flow rate (Nm3 /h) Motor power (kW)

Maekawa 6 Oil-flooded screw 320 L × 3 + 320 S × 1 1.9/0.4/0.105 14,600/2300/12,200 373 × 3 + 127 + 423

LN2 circulation system Refrigeration capacity Number of turboexpanders Number of compressors Type Kobe Steel Pressure (Mpa) Flow rate (Nm3 /h)

Kobe Steel 6.5 kW at 80 K 1 gas bearing 2 Oil-flooded screw KTS75A × 2 0.7/0.12 700 × 2

Storage and buffer tank Medium-pressure tank (1.97 MP) High-pressure vessel (15 Mpa) Liquid helium dewar Cryostat LN2 storage tank

100 m3 × 9 0.5 m3 × 18 × 4 12,000 L 830 L ×16 50,000 L

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

Fig. 8.

213

Helium compressor unit for the TRISTAN cryogenic system.

intercooler. Then the compressed helium gas is sent to the second-stage compressor and compressed further, up to 17 bar. The lubrication oil is pressurized by mechanical pumps (not shown in the figure) and injected into the compressors. The oil is recovered in the oil separator and sent back to the pump. The total required electric power for this compressor unit is 2.6 MW, including 130 kW for cooling water and oil pumps. The motors for the compressors are dipole-type motors driven by a 6.6 KV (50 Hz) power supply line. 3.3.2. Cold boxes The vertical cold boxes (4 m D × 6 m H and 2 m D × 3 m H) have six aluminum plate-fin heat exchangers, five turboexpanders (T1–T5), and control valves. A supercritical turboexpander, T3, was installed just before the J–T valve to increase the cooling capacity from 4 kW to 6.5 kW. A cooling capacity of about 8 kW at 4.4 K could be obtained by this system without liquid nitrogen precooling. The turboexpanders T4 and T5 for 80 K precooling are installed in an auxiliary cold box. In addition, an 80 K charcoal filter is installed in another auxiliary cold box to purify the process helium gas during cooldown of the system. After the J–T expansion, the liquid helium is separated and stored in a 12,000 L dewar. The liquid helium level in the dewar is automatically controlled by an electric heater. 3.3.3. Distribution system and transfer line Liquid helium produced in the 12,000 L dewar is distributed to the 16 cryostats installed in the

straight section of the underground TRISTAN tunnel through a multichannel transfer line. Cold gas vaporized in the cryostats returns to the cold box through the return line of the same transfer line. Figure 9 shows a cross-section of this multichannel transfer line. The helium channels for liquid helium supply and cold gas return are thermally insulated by an aluminum 80 K shield which is cooled by the liquid nitrogen supply and return lines. The multichannel transfer line works as a main header line for the supply and return lines, and has 16 connection boxes along the line. The pipe diameter of the helium return line is designed to guarantee a small pressure drop and a large cold buffer volume to suppress the pressure fluctuation in the cavities’ cryostats. Each connection box has control valves for helium and nitrogen lines, and bayonet connection ports for subtransfer lines which are connected to the cavity cryostat. Bayonet connection ports are used for easy connection and disconnection of the cavity cryostats to the connection boxes. At both ends of the multichannel transfer line, the supply and return lines of helium and nitrogen are connected by the so-called“cold end valves.” Bypass flows through these cold end bypass valves keep the temperatures of helium and nitrogen lines in the multichannel transfer lines constant and prevent thermal oscillation in the horizontal flow line even with the cooling of a small fraction of the cryostats. 3.3.4. Liquid nitrogen circulation system A liquid nitrogen circulation system cools the 80 K thermal shield of 16 cryostats and eases the flow control in the 16 parallel cooling loops. A turboexpander is used to reduce the liquid nitrogen consumption.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

214

Fig. 9.

Multichannel transfer line for the TRISTAN cryogenic system.

We will discuss the liquid nitrogen circulation system in greater detail later. 3.4. Helium gas handling The total amount of liquid helium handled in the whole system during steady-state operation is about 16,500 L, including about 2500 L of liquid helium stored in the 12,000 L dewar. For gas helium storage during the system shutdown, nine 100 m3 medium-pressure (19 bar) storage tanks are available. An off-line helium gas recovery and purification system consists of a fivestage air-cooled oil-lubricated reciprocating compressor (150 bar, 150 Nm3 /h), a high-pressure helium gas purifier (80 K, 150 bar, 150 Nm3 /h), and highpressure storage cylinders (4 × 1350 Nm3 ).

Before the commissioning of KEKB in 2001, we washed the high-pressure line of the first heat exchanger for the helium refrigerator to remove oil contamination accumulated during the first 10 years of operation. We also overhauled the turboexpanders, the screw compressors, and the electric motors. The cooling water towers for the helium gas compressor unit, the electric motors of oil pumps, relay switches, and sequencers of the helium compressors were replaced by new ones during shutdown of the cryogenic system. The electropneumatic converters of the control valves on top of the header connection boxes in the tunnel, which were damaged by high radiation doses, were also replaced.

4. Cavity and Cryostat

3.5. Maintenance and overhaul

4.1. Superconducting cavity and cryostat

The cryogenic system operated for more than 20 years following the commissioning of the TRISTAN superconducting RF cavity in 1984. Inspections and maintenance were carried out regularly, once a year, during the summer shutdown of the system. During this period, the machines with wearing parts, such as the helium compressors, including the air compressor unit, and valves were inspected carefully and repaired if necessary. Many components, especially electrical parts, wore out during the long running time and were replaced by new ones to maintain the required performance.

The cooling flow diagram of the TRISTAN superconducting accelerator RF cavities is shown in Fig. 10. Each cryostat contains two five-cell 508 MHz superconducting RF cavities made of 2-mm-thick pure niobium sheets produced by forming (see Fig. 5). The cold mass is mainly composed of a 1,200 kg stainless steel vessel filled with 830 L of liquid helium. The heat load to 4.4 K without RF input of each cryostat is about 30 W. An 80 K copper thermal shield is cooled by liquid nitrogen. A vertical coaxial input coupler is attached to each five-cell cavity to supply the 100 kW RF power required for beam

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

Fig. 10.

215

Cooling scheme of the TRISTAN superconducting RF cavity.

operation. Inner conductor is cooled by water and outer conductor is cooled by vaporized helium gas flow. This gas flow is controlled by valves to avoid subcooling. 4.2. Instrumentation and control valves All sensors and control valves for cooling the cavity are also shown in Fig. 10. The thermosensors are installed inside the helium vessel to measure the temperatures of the helium vessel, cavity, and input coupler, respectively. The pressure in the helium vessel is measured by a pressure sensor attached to the port of the cryostat. The liquid level in the helium vessel is measured by a superconducting level sensor installed inside the cryostat. Two redundant sensors are installed in each cryostat: one is used for operation and the other is for spares to avoid disassembly of the cryostat in case of a sensor failure. To control the mass flow of cold helium gas during cooldown of the cavities, mass flow meters were installed on each helium supply line, but we did not need to use them because we could control the cooling speed of each cavity by knowing the temperature of the cavity. Five thermosensors (Q1–Q5) installed inside the helium vessel were used to measure the temperature profile. During the cooldown we took care to keep the temperature difference small so as to prevent helium leakage at the seal connection part due to thermal shrinkage.

The cavity and insulation vacuum are carefully monitored during operation, especially during cooldown operation to check for helium leakage. The design pressure of the TRISTAN superconducting RF cavity is 1.3 bar, because a smaller wall thickness is preferable from the viewpoint of cooling. To prevent buckling of the cavity by pressure from outside due to the thin cell structure, we must take care of the pressure of the helium vessel, which has to remain below 1.2 bar during operation. 4.3. Safety protection of the cryostat Two types of protection, referred to as “hard” and “soft” protections, are prepared for pressure rises in the helium vessel, for example by quench. Quench phenomena will be discussed in greater detail in a later section. When the pressure in the helium vessel rises over the setting value of 1.3 bar, a mechanical safety valve on top of the cryostat opens automatically and the helium gas flows outside the tunnel through largesize piping. In parallel with this mechanical safety valve, a pneumatic bypass valve is installed. This valve opens at 1.25 bar by an electric signal from the pressure sensor. In case these two safety valves are not working, a rupture disk installed on top of the cryostat opens to release the helium gas into the tunnel. The opening or leaking of the safety valves discussed above can be detected by the thermosensors attached to the discharge pipe of the safety valves.

December 24, 2012

14:49

216

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

5. Operation of the Cryogenic System 5.1. Cooldown of the cavities The 32 five-cell superconducting RF cavities in 16 cryostats were cooled from room temperature down to 150 K, in parallel with 80 K helium gas precooled by liquid nitrogen in the helium refrigerator, and then the turboexpanders T1 and T2 were started for further cooldown. The cooling helium gas is sent to the cryostat and sprayed inside the helium vessel from both ends of the pipe set horizontally in the upper part of the vessel, as shown in Fig. 10. The cavity and the helium vessel are mainly cooled by the convection of the cold helium gas flow inside the vessel developing a small temperature difference inside the vessel. By using this cooling method we can cool down the cavity uniformly from room temperature. After reaching 5 K in the cavities, we stopped the supply of the cold gas to the cryostats and started the liquefaction in the helium dewar and the filling of the cavity cryostats. It takes three days to cool down from room temperature to liquid helium temperature and one day to fill the 16 cryostats with liquid helium. The cooldown speed of the cavities of about 10 K/h is limited by a requirement concerning the cryostats and cavities, i.e. avoiding leaks at the seals due to large thermal shrinkage. Before cooldown of the cavities, about 8000 L liquid helium was liquefied in the 12,000 L dewar. The flows of the liquid helium and nitrogen to the cryostats are controlled by the valves at connection boxes. 5.2. Steady-state operation of the cavities According to the beam operation pattern of the TRISTAN, i.e. injection, acceleration, and collision modes, the heat load of each cryostat increases from 30 W (static loss) to about 120 W (static loss plus RF heat load) in about 2 min, and 150 W for about 1.5 h. The electric heaters in the helium vessels are used to compensate for this large dynamic range by keeping a constant flow which minimizes the pressure fluctuations and eliminates the need to control the return valves. Figure 11 shows the heat load compensation by a PID-controlled heater in the helium vessel.

Fig. 11.

Compensation of the heat load by the heater.

The liquid helium levels in the cryostats are PIDcontrolled by the supply valves at the connection boxes. 6. Quench in Superconducting Magnets and Cavities 6.1. Quench in superconducting magnets In superconducting magnets, the resistive transition or “quench” is what happens when any part of the conductor in a magnet goes from the superconducting to the resistive state. The superconducting magnet is excited by the current from the DC power supply, and the magnetic energy is stored in the magnet as a magnetic field. The total energy available for dissipation as heat is the stored energy in the magnet. For high-field high-current magnets, the stored energy is large. For example, the stored energy of a twin-aperture LHC dipole is about 7.0 MJ at a current of 11.8 kA. Usually, after a quench, when possible, the stored energy could be partly extracted through the superconducting bus-bar and current leads and dumped into external resistance situated outside the cryostat to limit energy dissipation in the coil and consequently the temperature and pressure rises and liquid helium vaporization. 6.2. Quench in superconducting cavities In superconducting RF cavities, the quench phenomenon is also possible even if the stored energies are very small (less than 10 J). Figure 12 shows a simplified setup of the RF system for superconducting RF cavity and RF power distribution. The RF power generated at the

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

6.3. Quench protection

Circulator Input Coupler

RF Generator

RF Load

Superconducting RF Cavity Quench Normal Area Beam Loading Cavity Loss

Fig. 12.

217

RF system for the superconducting RF cavity.

high-power RF source is supplied to the superconducting RF cavity through a circulator and an input coupler. Some parts of the RF power enter the cavity through the input coupler and excite the acceleration field in the cavity. The remaining part of the RF power is reflected and absorbed in a dummy load. The coupling of the input coupler to the cavity is expressed by the coupling coefficient β corresponding to the ratio of the inflow RF power entering the cavity to the RF power loss in the cavity. In the case of β = 1, all the input RF power is consumed as RF loss in the cavity with no reflection. In the TRISTAN superconducting RF cavity, the RF loss Pwall is small (50 W for Q0 = 109 and Eacc = 5 MV/m) in comparison with the required RF power Pbeam for beam acceleration of TRISTAN (about 100 kW). Corresponding to energy gain of the beam, the coefficient β of the input coupler of the TRISTAN superconducting RF cavities must be set larger than 104 . In this case, if there is no beam in the cavity, almost all RF power is reflected at the input coupler. When the beam injection into the TRISTAN ring starts, the beam loading of the cavity increases according to the beam current in the cavity, i.e. Pbeam increases automatically and the reflecting power at the input coupler decreases because of the very large β coefficient. In this overcoupling situation, if some part of the cavity turns to the resistive state, the cavity RF power is supplied automatically from the input coupler toward the beam loading power Pbeam and a large power Pnorm is lost in the cavity. There is absolutely no difference between the power loss by beam loading and the power loss in the normal part of the cavity. This loss mechanism in the superconducting RF cavity causes the quench in the cavity.

The quench in the magnet and cavity causes a sudden pressure rise in the helium vessel. We need quench protection to make the energy loss in the helium vessel as small as possible. To minimize the heating time of the quench, we need an efficient quench detection system. A magnet quench can be detected by measuring via voltage taps the normal resistive voltages appearing in the coil. If a quench is detected in the magnet coil, the magnet power supply is disconnected and the stored energy in the magnet is extracted outside the cryostat and dissipated in a dumped resistor at room temperature. A cavity quench can be detected by measuring precisely the RF power flow into and from the cavity, i.e. the power loss in the cavity. In the superconducting cavity system, the quench detector measures the RF power loss in the cavity. When an abnormal power loss rise is detected, it gives a quench signal to the RF source, which quickly cuts the RF power. At the same time the liquid helium supply valves at the connection box are closed by the quench signal. 7. Operation Mode of the Helium Liquefier/Refigerator 7.1. Liquefier and refrigerator modes Usually a helium liquefier, which is designed and used to liquefy helium, can also operate as a refrigerator. By using the electrical heater located in the liquid helium dewar, it is possible to vaporize all the liquid produced by the cold box. In this case, the level in the dewar will remain constant (no liquefaction) and the cold box will recover all the delivered flow in saturated vapor. The system will operate in refrigeration mode. 7.1.1. Liquefaction and refrigeration capacities Figure 13 shows the liquefaction rate of a typical 100 L/h helium liquefier as a function of the heating power. The liquefaction rate is maximum (100 L/h) without heat input, decreases with the heating power almost linearly, and finally reaches zero at a maximum heating power of 300 W, which corresponds to the maximum refrigeration power. Of course, the liquefier/refrigerator can also operate in a mixed mode, i.e. operate as a refrigerator in parallel with

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

218

300

Enthalpy [ kJ/L ]

Liquefaction Rate [ L/hr ]

100

50

Nitrogen 200

100 0 0

100

200

Helium

300

Refrigeration Power [W]

Fig. 13.

0

Liquefaction rate versus cooling power.

0

100

200

300

Temperature [ K ]

consuming the liquid helium as heat load, for example when cooling the current leads of the SCM and the input coupler of the SCC. In these cases vaporized helium gas is warmed up to room temperature and returned to the suction of the helium compressor. As a rule of thumb, because of the thermodynamic properties of helium gas, a liquefier with a 100 L/h liquefaction capacity has a cooling power of about 300 W at 4.4 K, i.e. the liquefaction rate of 1 L/h corresponds to cooling power of about 3 W.

Fig. 14. volume.

Enthalpies of helium and nitrogen of unit liquid

the vaporized gas, i.e. the sensible heat of cold helium gas, is not recovered by the system. In the refrigerator mode, the vaporized cold gas is returned to the cold end of the refrigerator (through path A) and the cooling power of the gas is recovered.

7.1.2. Cooling

7.1.3. Refrigerator mode and return transfer line

The vaporization heat of liquid helium at 1 L/h consumption corresponds to a cooling power of about 0.7 W. In the case of the liquefaction mode, only this vaporization heat, i.e. the latent heat of liquid helium, is used for cooling. Figure 14 shows the enthalpies of helium and nitrogen of unit liquid volume as a function of temperature. The enthalpy gaps at 4.2 K and 77 K correspond to the latent (vaporization) heat of the liquid helium and nitrogen, respectively. The latent heat of helium gas is very small compared to that of nitrogen, but enthalpy dependence on temperature is large. This means that the vaporized helium cold gas has a large cooling power. Figure 15 shows the difference between the liquefier mode and the refrigerator mode schematically. In the liquefier mode, cooling the SCM/SCC is simple and straightforward, only a one-way supply transfer line is needed and vaporized helium gas used for cooling the SCM/SCC is sent back to the helium compressor suction through the room temperature piping (path B), and the cooling power of

The cooling power can be increased more than three times by adopting the refrigeration mode, but in this refrigeration mode we need an additional return line to recover cold return helium from the cryostat to the cold box. In this case the pressure drop in the return line must be considered, because the cooling temperature is determined by the helium pressure in the cryostat and for high-performance and stable operation of superconducting devices lower-temperature and lower-pressure operation is preferable. The operation pressure in the cryostat must be higher than the cold end pressure of the return line in order to send back the cold vaporization gas in the cryostat to the helium refrigerator cold box through the return transfer line. The cold end pressure of the return line is usually designed around 1.1–1.2 bar due to the large pressure loss of about 0.1 bar at the first heat exchanger return line and the suction pressure of about 1.05 bar at the helium compressor. Not only must the return line have a low heat loss, but its pipe must be large enough to reduce the pressure loss.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

219

Gas Bag

Heat Exchanger

Cold Box

Heat Exchanger

A

Cold Vaporized Gas Return to Cold Box

B

Cold Vaporized Gas Send to Gas Bag

J–T Valve

Liq. He Liq. He Liq. He

SCM / SCC Fig. 15.

Liquefier and refrigerator modes.

This means that, for designing an efficient cooling scheme for the refrigeration mode cooling scheme, use of the high-performance transfer line system is essential. 8. Transfer Line 8.1. Multichannel main transfer line Figure 16(a) shows a cross-section of the vacuuminsulated multichannel transfer line [10] — referred to as the main transfer line — developed at KEK for the cold test stand of the KEKB superconducting RF cavity [3]. The main characteristics of the multitransfer-line system developed at KEK include the following: • • • •

Low heat loss; 80 K aluminum thermal shield; Small cold mass; Easy fabrication and low cost.

The most important performance requirement for the transfer lines, i.e. low heat loss to the helium lines, is fulfilled by using a well-designed 80 K thermal shield to block the heat inleaks from the outside room temperature vacuum pipe. 8.1.1. Structure of the multichannel transfer line Two helium channels, i.e. the supply and return helium lines, are guarded by the aluminum 80 K

thermal shield, which is cooled by liquid nitrogen in the stainless steel cooling pipe. The shield is supported by four G-10 support plates at every ∼ 1 m interval in the vacuum stainless steel pipe. The support plates are fixed by stainless steel bolts to the shield through a G-10 support block. The helium pipes and the shield are wrapped with 10 and 30 layers respectively of aluminized polyester film with a polyester interlayer net to reduce heat inleaks by radiation as well as by convection in the case of a degraded insulation vacuum. The helium flow and return pipes are fixed by a fiberglass-reinforced polyester collar and supported in the aluminum 80 K thermal shield by G-10 plates. 8.1.2. Aluminum 80 K shield A half aluminum 80 K thermal shield is made from aluminum alloy by extrusion and assembled into a full shape. Aluminum alloy is used for this material because of its light weight and because it is easy to extrude even for a complex shape cross-section like this. The aluminum alloy used in the design does not have good thermal conductivity at very low temperature but is sufficient at the 80 K region and has high mechanical strength to support the piping structure. A thin stainless steel pipe for the liquid nitrogen line is inserted into the groove outside the half aluminum shield. The precise shape and smooth surface of the groove make the pipe fit tightly. Good thermal

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

220

Aluminum 80K Shield

Stainless Steel He Return φ 34

LN2 Flow

φ14

G-10 Support Plate G-10 Support Block Polyester Callor

He Flow φ17.3 Stainless Steel LN2 Flow φ14

LN2 Return

φ14

Stainless Steel He Flow φ17.3 Screw Stainless Steel Clip

Stainless Steel LN2 Return φ14

0

He Return φ 28 5

10

15

scale (cm)

Fig. 16.

High-performance multichannel transfer lines.

contact between the liquid nitrogen pipe and the aluminum shield is guaranteed by this structure. 8.1.3. Small cold mass, thin stainless pipe Thin stainless steel pipes are used for the helium and nitrogen lines to reduce the weight of the transfer lines for ease of handling and to reduce the cold mass of the transfer line. Making the cold mass compact is very important for operation of the transfer line system. In the cooling operation, including the initial cooldown, large cold mass of the transfer line does not affect the operation too much. But, in the case of a “transient”operation, i.e. a restart after a stop of the helium flow, small cold mass is preferred, because the cooldown of the large cold mass due to the stopped flow causes a large pressure rise. This pressure rise during transient operation increases the recovery time of the transfer line system to the steady state. 8.1.4. Assembly of the transfer line In the multichannel transfer line discussed above, not only its high performance but also its ease of

assembly must be considered in the design. The aluminum 80 K thermal shield with thin wall thickness could be made by extrusion very precisely and at a reasonable price. The half shield has a circular groove for installing the thin stainless steel pipe for liquid nitrogen, two shallow rectangular grooves for the support posts, and at both ends female and male connections for assembly into a full shield. Using these grooves and connections we can assemble the whole structure precisely without soldering.

8.2. Multichannel subtransfer line The main transfer line and the cryostats for the SCM/SCC are connected by using the subtransfer lines. Usually, the helium line of the subtransfer lines is designed short and not protected by the 80 K thermal shield. But the heat loss of this “simple” structure transfer line with multilayer vacuum insulation is not small. To reduce the heat loss, newtype subtransfer lines, which have an 80 K thermal shield, were developed at KEK. Figures 16(b) and 16(c) show cross-sections of the subtransfer lines for supply and return lines. The structure is similar to that of the main transfer line discussed before. End

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

221

9.2. 80 K shield cooling by liquid nitrogen

Fig. 17.

Joint part of the subtransfer line.

connection parts of the transfer line are terminated by bayonet-type joints. Figure 17 shows the joint part of the subtransfer line. The helium line is connected directly by a bayonet-type joint but the nitrogen line is connected via a liquid nitrogen bypass line. The subtransfer lines, part A and part B, can rotate around the rotation axis shown in Fig. 19, i.e. the axes of the bayonet joints of the helium and nitrogen lines. We can change the effective length, i.e. the distance between the two end connection points of the subtransfer line, by adding a midbayonet joint and adjusting the rotation angle. By this adjustable mechanism we can eliminate the flexible bellow section, and handling of the subtransfer system for the installation becomes very easy.

In the liquid nitrogen cooling scheme, the large latent heat of liquid nitrogen is used for 80 K shield cooling. The simplest way of liquid nitrogen cooling is direct cooling, in which liquid nitrogen stored in the storage tank is supplied directly for 80 K shield cooling. Usually, this storage tank is built outside of the building far away from the helium refrigeration system and needs a long liquid nitrogen transfer line. The heat loss of long transfer lines is not negligible and it is difficult to economically cool several loads in parallel. These problems can be solved by using the liquid nitrogen circulation system for cooling the 80 K thermal shield, i.e. installing the circulation system near the helium refrigeration system. 9.3. Liquid nitrogen circulation system 9.3.1. Liquid nitrogen pump For the liquid nitrogen circulation, one possibility is to use liquid nitrogen pumps. There are two types of liquid nitrogen pumps: piston-type and turbo-type. The piston pump is better from the point of view of high compression rates and small mass flow rates, which are suitable for cooling long-path piping of the 80 K thermal shield. But pulsations of the liquid nitrogen flow and a reduced lifetime are problematic. The turbopump is better for long and reliable operation but usually it is not easy to provide the high compression rate. 9.3.2. Liquid circulation by warm compressors

9. Liquid Nitrogen Circulation System 9.1. 80 K shield cooling by helium gas For the 80 K thermal shield, there are two cooling methods: one uses liquid nitrogen and the other cold helium gas flow supplied from the helium refrigerator. The 80 K shield cooling using cold helium gas has the big advantage of not using liquid nitrogen for operation, and there are many cryogenic systems in the world which have adopted this method. In this cooling scheme, the specific heat of helium gas is used to extract heat from the 80 K shield, and the main mass flow of high-pressure cold helium in the refrigeration cycle must be circulated through the transfer line and cryostat 80 K shield. The operation pressure of the 80 K piping in the multitransfer line and cryostat must be designed higher than 10 bar.

Another possibility is to use a liquid nitrogen circulation system consisting of warm compressors and heat exchangers which operate by consuming liquid nitrogen supplied from outside the system. Room temperature compressed gas is cooled through heat exchangers by using vaporized cold gas of supplied liquid nitrogen from the outside liquid nitrogen tank. In this system the cooling power comes from both the latent heat of vaporization of the liquid nitrogen and the sensible heat of the produced vapor. The amount of liquid nitrogen consumption can be reduced from 23 L/h to 17 L/h for 1 kW cooling power. 9.3.3. Practical liquid nitrogen circulation system This type of liquid nitrogen circulation system has been studied extensively at KEK and a prototype

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

222

From Liq. Nitrogen Storage Tank 0.7 MPa

Liq. Nitrogen Supply to SC Cavities

Liq. N2

1250 Nm3/hr Oil Filters A

75 kW

Liq./Gas Nitrogen Return

0.12 MPa Liq. N2

Screw Type Compressor B After Cooler Air Cooled

0.12 MPa 7 m3

75 kW

Vent to Atmosphere Fig. 18. Table 3. Cooling capacity Turbine Number of turbines Flow rate Inlet pressure Outlet pressure Inlet temperature Isentopic efficiency Rotor speed Rotor diameter Gas bearing thrust, journal Fabricator

Subcooler

Turbo-Expander

Ballast Tank 6.5 kW at 80 K Flow diagram of the liquid nitrogen circulation system.

The main parameters of the liquid nitrogen circulation system.

6.5 kW

(Nm3 /h) (kPa) (kPa) (K) (%) (rpm) (mm) Static Kobe Steel

at 79 K

1 950 670 120 123 75 45,000 80

EXT 8D

with 1 kW at 80 K cooling power was constructed and tested successfully. A large liquid nitrogen circulation system with 6.5 kW at 80 K has been designed and constructed for the TRISTAN cryogenic system and operated stably for more than 20 years. The liquid nitrogen circulation system in the TRISTAN cryogenic system for cooling the 80 K thermal shield for transfer lines and SCC cryostats was shown in Fig. 7, and its detailed flow and parameters are shown in Fig. 18 and Table 3. The circulation system consists of two compressors and a cold box with heat exchangers and liquid, a nitrogen storage vessel, and a turboexpander. The turboexpander is used for standalone operation to reduce the liquid nitrogen consumption. This system has three operation modes: (1) Direct cooling mode. Liquid nitrogen is directly supplied to the 80 K thermal shield of transfer

Cooling capacity Compressor Number of compressors Flow rate Inlet pressure Outlet pressure Rotor diameter Rotor length Rotor speed Input power Type Fabricator

6.5 kW

(Nm3 /h) (kPa) (kPa) (mm) (mm) (rpm) (kW) Oil-flooded screw Kobe Steel

at 79 K

2 540 × 2 105 700 147 238 5335 55 × 2

KST55A-E

lines and cryostats from the liquid nitrogen storage tank outside. This mode is used for initial cooldown and in case of failure of the two compressors. (2) Circulation mode. This mode is used in case the turboexpander fails. It needs only one compressor for operation. (3) Refrigerator mode. This mode operates the turboexpander to reduce the liquid nitrogen consumption.

10. Satellite Refrigerator 10.1. Satellite refrigeration system The Tevatron cryogenic system [12, 13] is a hybrid system consisting of a large liquefier — the central helium liquefier (CHL) — and 24 satellite refrigerators arranged around the 6 km circumference of the

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

Tevatron ring. The CHL liquefied about 5400 L/h of helium and supplied it to the 24 satellite refrigerators through the high-performance liquid helium transfer line located around the ring. The satellite refrigerators receive the liquid helium from the CHL and produce the subcooled liquid helium to be delivered to the magnet strings in the tunnel for cooling. They return the warm helium gas to the CHL compressor suction. The satellite refrigerator acts as an amplifier of cooling power by using the enthalpy of the helium supplied by the CHL as liquid at about 130 L/h and produces the cooling power of about 1 kW.

10.2. Merit of the satellite refrigerator system As discussed previously, for efficient and economic cooling of the SCM/SCC, the refrigerator mode is better, and to realize this, a high-performance transfer line system (in addition to the supply line, and the return line with larger-pipe-size piping, i.e. low heat loss and low pressure drop lines) will be required. In the satellite cooling scheme, the liquid helium produced at the central liquefier is sent to the satellite refrigerator installed near the cooling load through the one-way liquid helium transfer line, and vaporized warm gas is sent back to the central liquefier through room temperature piping. This cooling pattern is very similar to the liquefaction mode. But in this scheme the cold cooling power, i.e. the sensible heat, of vaporized cold gas is recovered by the heat exchanger in the satellite refrigerator cold box and the supply of helium gas to the satellite cold box is liquefied for cooling the SCM/SCC. The satellite refrigerator is very simple, consisting of a heat exchanger and a Joule–Thomson valve. It does not need the expander to get cooling power. The satellite cooling scheme has the good advantages of the liquefier and refrigerator modes — simplicity of cooling structure and high cooling efficiency.

10.3. Application of the satellite cooling scheme As for application of satellite refrigerators, we will discuss a medium-size case which is separated far away from the central liquefier, which has enough excess cooling power.

223

10.3.1. Satellite refrigeration system for the KEKB crab cavity The installation of four superconducting crab cavities, with a total heat load of about 800 W, was planned at the collision point of the KEKB to improve the luminosity. For cooling this, the satellite refrigeration cooling scheme was proposed. The reasons for adopting the satellite refrigeration scheme for the KEKB crab cavity cryogenic system are as follows: (1) There is surplus cooling power, about 5 kW at 4.4 K, of the existing cryogenic system, which operated for KEKB superconducting accelerator cavities. (2) In this case, the four crab cavities were connected directly to the existing Nikko helium cryogenic system for its cooling; a high-performance long transfer line with a large pipe size cold return line must be constructed to keep the pressure drop small. (3) In the satellite cooling scheme, only a oneway supply helium transfer line is required, and return helium gas is sent back to the Nikko cryogenic system as room temperature gas. The heat loss and the construction cost of this long transfer line can be reduced and system operational flexibility increased. Figure 19 shows a schematic flow of the satellite system for KEKB superconducting crab cavities [9]. The satellite refrigerator installed near the crab cavity in the tunnel can produce about 400 W at 4.4 K cooling power by using about 150 L/h liquid helium supplied from the Nikko cryogenic system without a turboexpander. The cooling power of the two satellite refrigerators at Tsukuba is produced and supplied by the large refrigerator at Nikko, which has high efficiency and very high reliability. This satellite refrigerator scheme is attractive compared to the independent refrigeration systems, even if the heat load of about 200 W (0.2 W/m) from the 1-km-long transfer line is taken into account, because the required helium gas flow rate is nearly half that of an independent helium refrigerator, since it has no turboexpander, and its system configuration is very simple. The liquid nitrogen used in this system can be supplied from the Nikko cryogenic

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

K. Hosoyama

224

Fig. 19.

Satellite refrigeration system for the KEKB crab cavities.

system through 80 K radiation shield piping of a 1-km-long transfer line. This satellite cooling scheme for the KEKB superconducting crab cavity was not realized because a new crab cavity scheme, the so-called “global crab crossing scheme,” was adopted for KEKB [14]. In this crossing scheme the existing cryogenic system can be used for cooling, because two crab cavities — one in the electron ring the other in the positron ring — were installed in the Nikko straight section of the KEKB near the KEKB acceleration superconducting cavities.

Central Liquefier

10.3.2. Small satellite refrigeration system Recently, many kinds of SCMs have been used for medical and industrial applications, and also in the fields of high-energy physics and accelerators. By the development of cryocooler technology, many SCMs are cooled by cryocoolers. The motivation for using cryocoolers to cool the SCMs is the elimination of complicated helium refrigeration systems, especially their transfer lines, which have large heat losses. On the other hand, from the economic point of view, the efficiency of cryocoolers is not so good.

Gas Helium Circulation Pine Line

Liq. Helium Distribution Line

Liq. Helium Storage

Satellite Ref. #1

Fig. 20.

Satellite Ref. Satellite Ref. #2 #3

Satellite Ref. #N

Satellite refrigeration system for magnet cluster cooling.

December 24, 2012

14:49

WSPC/253-RAST : SPI-J100

1230008

Cryogenic Technology for Superconducting Accelerators

A satellite refrigerator can be used for the cooling of a magnet cluster, as shown in Fig. 20, for industrial and accelerator applications, such as SCMs for silicon crystal growth and beam lines of medium-size accelerator facilities. The attractive points of the satellite refrigerator scheme are the following: (1) The coupling between the central liquefier and the satellite refrigerator is not so strong and each satellite can operate independently, including the operation temperature. (2) The cooling capacity of each satellite can be changed by adjusting the liquid helium supply and the mass flow of the supplied gas. With this property we can get enough redundancy of cooling power even if there is not a large safety margin for each refrigerator. (3) There is no need for a continuous liquid helium supply to each refrigerator, by having the liquid helium reservoir inside the satellite refrigerator cold boxes. The intermittent liquid helium supply, for example twice a day, can reduce the heat loss at the transfer line to a very low level, practically zero. To realize this kind of small satellite refrigeration system, the main components of the system, such as the high-performance transfer line, compact heat exchanger, small J–T valve, and reliable compact compressor, will be necessary and must be available at a reasonable price. In view of the global energy saving, a substantial reduction in the operation costs of the magnets for accelerators and industrial applications is urgent. To achieve this, effective utilization of the superconducting devices is very important and a high-performance (i.e. with high efficiency, good reliability, and easy operation) cryogenic system is essential. References [1] Y. Kimura, TRISTAN project and KEK activities, in Proc. XIII Int. Conf. on High Energy Accelerators (Novosibirsk, USSR, 1986). [2] KEK B-Factory Design Report, KEK Report 95–7 (1995).

225

[3] T. Furuya et al., Beam test of a superconducting damped cavity for KEKB, in Proc. Particle Accelerator Conference (Vancouver, Canada, 1997). [4] K. Hosoyama et al., Crab cavity for KEKB, in Proc. 7th Workshop on Superconductivity (1988). [5] K. Hara et al., Cryogenic system for TRISTAN superconducting RF cavity, in Advances in Cryogenic Engineering, Vol. 33 (Plenum, New York, 1988), p. 615. [6] K. Hosoyama et al., Cryogenic system for the TRISTAN superconducting RF cavities: Performance and status, in Advances in Cryogenic Engineering, Vol. 35 (Plenum, New York, 1992), p. 615. [7] K. Hosoyama et al., Cryogenic system for TRISTAN superconducting RF cavity: upgrading and present status, in Advances in Cryogenic Engineering, Vol. 37 (Plenum, New York, 1992), p. 683. [8] A. Kabe et al., Cryogenic system for KEKB superconducting RF cavity, in Advances in Cryogenic Engineering, Vol. 45 (Plenum, New York, 2000), p. 1347. [9] K. Hosoyama et al., Design and performance of the KEKB superconducting cavities and its cryogenic system, in Advances in Cryogenic Engineering, Vol. 43 (Plenum, New York, 1998), p. 123. [10] K. Hosoyama et al., Development of a high performance transfer line system, in Advances in Cryogenic Engineering, Vol. 45 (Plenum, New York, 1992), p. 123. [11] K. Hosoyama et al., Cryogenic system for TRISTAN superconducting RF cavities, description and operation experience, ICEC16/ICMC proc. (Elsevier Science, New York, 1992), p. 183. [12] C. Rod et al., Energy doubler refrigeration system, IEEE Trans. Nucl. Sci. NS-27(3), 1328 (1977). [13] P. C. Vander Arend, Helium refrigeration system for Fermilab energy doubler, in Advances in Cryogenic Engineering, Vol. 23 (Plenum, New York, 1977), p. 420. [14] K. Hosoyama et al., Development of the KEK-B superconducting crab cavity, in Proc. EPAC08 (Genoa, Italy, 2008), pp. 2927–2931. [15] K. Hosoyama, Cryogenic systems, in Proc. Asian Accelerator School: Physics and Engineering of High-Performance Electron Storage Rings and Superconducting Technology (Beijing, 1999) (World Scientific, 2002), pp. 394–456. [16] K. Mendelssohn, The Quest for Absolute Zero (Weidenfeld and Nicolson, 1966). [17] R. F. Barron, Cryogenic Systems (Oxford University Press, New York, Clarendon, Oxford, 1985). [18] R. B. Scott, Cryogenic Engineering (van Nostrand, Princeton, New Jersey, 1966).

December 24, 2012

14:49

226

WSPC/253-RAST : SPI-J100

K. Hosoyama

Kenji Hosoyama received his Bachelor of Physics degree from Niigata University in 1970, and his Master’s degree and Ph.D. from Tohoku University in 1972 and 1975, respectively. He became a postdoc at the KEK Laboratory to work in superconducting RF cavity research till 1978. He became a Professor at KEK in 2002. His contributions included design, construction, and operation, as well as the R&D of the superconducting magnet, superconducting RF, and the cryogenic systems for the TRISTAN, AR, and KEKB projects. Prof. Hosoyama pioneered the R&D, construction, and installation of the superconducting RF crab cavity for the KEKB project. He also joined the R&D effort on the magnets for the Superconducting Super Collider project in the US. He was Visiting Scholar at Lawrence Berkeley Laboratory in 1983–1984.

1230008

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Reviews of Accelerator Science and Technology Vol. 5 (2012) 227–263 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300095

Superconductivity in Medicine Jose R. Alonso Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA [email protected] Timothy A. Antaya Ionetix Corporation, 8 Merrill Drive, Hampton, NH 03842, USA [email protected] Superconductivity is playing an increasingly important role in advanced medical technologies. Compact superconducting cyclotrons are emerging as powerful tools for external beam therapy with protons and carbon ions, and offer advantages of cost and size reduction in isotope production as well. Superconducting magnets in isocentric gantries reduce their size and weight to practical proportions. In diagnostic imaging, superconducting magnets have been crucial for the successful clinical implementation of magnetic resonance imaging. This article introduces each of those areas and describes the role which superconductivity is playing in them. Keywords: Superconducting magnets; cyclotron; hadrontherapy; gantry; MRI.

1. Introduction

of SQUIDs (Superconducting QUantum Interference Devices) as ultrasensitive pickups of electromagnetic signals. Substantial progress in biological imaging is being made using these devices, and readers are directed to an excellent treatment of this subject in a recently published book, 100 Years of Superconductivity, edited by Horst Rogalla and Peter H. Kes (CRC Press, 2011). The chapter on medical applications has a thorough treatment of SQUIDs. Another area of superconductivity which is quite active in the accelerator field but which has not yet seen application in the medical area is superconducting radio frequency. Accelerating structures based on SCRF have performance specifications far exceeding the current requirements for existing medical applications, but nothing says this will not change in the future. This article will address the use of superconducting magnets in therapy and diagnostic applications. It will provide a brief introduction to each field, describe the current state of the art and the impact that superconductivity is having or is expected to have in the near future.

As new technological developments emerge, a logical move is to search for applications in medicine and healthcare. Superconductivity has been no exception. The primary application in this area has been high-field, very efficient magnets, with noteworthy examples in cyclotrons in radiotherapy and isotope production applications, beam delivery for reducing the size and operating cost for proton and ion beams for therapy, and most notably for magnetic resonance imaging (MRI) magnets. In the last case, currently the largest commercial application of superconductivity, these magnets have for all practical purposes been the principal enabler of this new imaging modality, which is having a tremendous impact not only on the quality of diagnostic information now available, but also in unraveling function in biological systems, and studying biochemical processes. One particular medical application of superconductivity will not be addressed, as it is quite far afield from the main focus of this journal; namely, the use of Josephson junctions, primarily in the form

227

December 29, 2012

15:48

228

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

2. Superconducting Cyclotrons as Sources of Medically Relevant Particle Beams 2.1. Overview of particle beam therapy Few people have not heard of the Bragg peak as an argument why protons (or beams of other ions) offer substantial advantages for “external beam” radiation therapy. Figure 1 schematically compares the dose distribution for such ions with similar beams of X-rays (or neutrons). As energy loss (related to energy deposition in the surrounding medium) varies as 1/E, charged particles stiff enough to travel in a straight line (which electrons aren’t!) will lose most of their energy as they stop at the end of their range. Multiple scattering does broaden the peak a bit, so heavier ions will show sharper Bragg peaks, but for the most part protons already can concentrate a stopping dose in a tumor volume, and provide very significant sparing of normal tissue compared to X-rays. This translates into fewer side effects, greater tolerance to the treatment and, particularly important for pediatric

Fig. 1. Comparative dose distributions for various therapy beams. Protons and carbon exhibit the “Bragg peak” dose enhancement in the vicinity where the particles stop, compared to the basically exponential dose distribution for X-rays. High-energy neutrons (∼60 MeV) have similar energy deposition to the photons from a 7–8 MV source. The carbon Bragg peak is sharper owing to less multiple scattering. The dose for carbon beyond the end of the range is from lighter (longerrange) fragments from nuclear reactions of the slowing carbon ions.

cases, a much lower chance of secondary tumor generation many years after the treatment. We are currently seeing a tremendous growth in the number of centers offering proton therapy, and now overseas (but not yet in the US [1]) a growing number of hospitals and research institutions delivering carbon ions. Not only is the Bragg peak sharper for carbon, allowing finer control of field edges and distal falloff, but the higher charge of the carbon ion induces greater biological damage at the stopping point. This factor, termed LET,a varies roughly as the square of the charge of the ion, so each carbon ion (Z = 6) deposits about 36 times the energy at its stopping point as a proton. Clinical practice for the treatment of cancer has developed around the ability of tumor tissue to repair from radiation “insults,” and the large number of fractions usually given in a course of radiation treatments (as many as 48 daily fractions) arises from the slightly different repair rate between normal and cancer tissue (cancer normally repairs more slowly). Carbon ions change this paradigm. By introducing more ionizing radiation into the cells, more damage is done and much less repair is possible. In fact, Japanese researchers [2] are now treating liver and lung sites with four or even one fraction with success equal to that of treatment with the normal fractionation schemes. LET can play a very important role in the future of external beam therapy! Neutrons have been used for radiotherapy for almost 50 years as well and have shown effectiveness in certain sites [3]. A fast-neutron radiation field is usually produced by sending 60 MeV protons or deuterons onto a beryllium target, and the resulting neutron depth–dose curve is approximately equivalent to that of a 7–8 MeV photon beam. The dose is delivered by the neutrons transferring energy to protons though nuclear collisions; the recoiling (low-energy) protons break molecular bonds and do the biological damage. In fact, the LET of these recoiling protons is substantially higher than X-rays, so there was a high degree of optimism in the early days that neutrons could provide substantial improvements over X-rays for cancer treatment [4]. Figure 2 shows neutrons as

a “Linear Energy Transfer. Conjugate of dE/dx, or instead of the energy lost by a particle per increment of length, it is the energy deposited in the surrounding medium by the slowing or attenuating beam. LET is determined by the slowing of secondary electrons (“delta” rays), and as these are mostly forward-peaked the LET distribution is usually displaced downstream of the primary interaction. It is a good measure of biological damage done by ionizing radiation such as X-rays, protons or ion beams.

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

Superconductivity in Medicine

Fig. 2. Comparison of X-rays, protons, neutrons and carbon ions for relative effectiveness: X-rays have low LET and poor dose distribution, and protons also low LET but good dose definition; neutrons have poor dose definition but higher LET, and carbon offers both high definition and high LET.

having higher biological effectiveness, but from Fig. 1 we see that the dose distribution is similar to X-rays. In fact, the high expectations for neutrons did not materialize [5]; the clinical results were dominated by complications owing to the lack of good dose localization, and excessive damage to surrounding normal tissue. So why this long introduction to neutron therapy? Simply that the first use of superconductivity for clinical therapy was a neutron-producing cyclotron designed and built by Henry Blosser [6] and installed at the Harper-Grace Hospital in Detroit — presented at greater length below. 2.2. Clinical requirements for charged particle beams Table 1 details the three clinical parameters most relevant for the provider of ion beam acceleration and beam delivery.b Table 1. Selected clinical requirements for ion beams for external beam therapy [7, 8]. Beam range in tissue Beam current (average) Delivery orientation

b Note

30 cm Adequate to treat 1 L volume to 2 Gy in 1 min Isocentric around patient

1230009

229

Range is determined by need to reach any anatomical site, and as the target is usually treated with overlapping fields, beam must enter, and reach the target volume, from several orientations. To guarantee 30 cm in the patient, the usually specified energy for protons is 250 MeV. While this corresponds to a range of almost 37 cm, energy is lost in the devices used for forming the large field, particularly if passive scattering is used [9]. Scanning systems have less material in the beam, and a lower top energy could be provided from the accelerator and still meet the 30-cm-range requirement in the patient. The accepted beam energy for carbon ions is 400 MeV/amu. Beam current is specified in terms of treatment time. Delivering a fraction in a minute or less is desirable from a patient comfort standpoint — immobilization devices are rarely conducive to a patient remaining still, as well as for promoting efficient patient throughput for the treatment room. It is difficult to specify an actual beam current, as the particle flux required for the prescribed dose depends on the size, shape and depth of the field, even for targets of equal volume. For a proton beam interfaced with a scanning system, an average current of a few nanoamperes is usually adequate. However, a cyclotron is a fixed energy machine; beam is always extracted at its top energy.c To place stopping particles at different depths in the target, the energy needs to be varied. With beam extracted from a cyclotron this is done with a wedge-degrader and energy selection magnet system. In addition, a set of jaws selects the beam of correct emittance to pass through the gantry. Because of multiple scattering, greater energy reduction leads to large beam losses on the jaws, and so while at full energy most all of the beam is acceptable for downstream transport, at 100 MeV this efficiency could be a few percent or less. Thus, to meet the treatment time specification the beam current from the cyclotron must be increased, to preserve the “brightness” of the beam as it is transported to the gantry. To complicate matters, substantial beam loss can occur in the process of converting the very tight beam entering the treatment room to the radiation field covering the size (the maximum field can be

that this is a very small subset of the full clinical requirements for proton or ion beam therapy facilities. References 7 and 8 present a much more complete set of requirements. c Variable-energy cyclotrons have been built, but require complex RF systems and trim coils to keep the magnetic field tuned for stable beam operation. All medical cyclotrons today are fixed energy machines.

December 29, 2012

15:48

230

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

20 × 20 cm or greater) of the target. Passive scattering, for instance, involves at best only 40% transport efficiency, again increasing the need for more particles from the accelerator to meet the treatment time requirement. In all, cyclotron currents for a viable therapy application must approach the microampere level. Treatments are normally given with patients lying on their backs on a treatment couch. This position usually provides the greatest comfort for the patient, but more importantly corresponds to the usual position in which diagnostic information is obtained (from CT or MRI; see Sec. 4). With highprecision beams, having very accurate information of the actual coordinates of the target volume is essential. Unfortunately, the human body is far from being a rigid structure,d and the best chance of knowing where a target is located will occur if treatment is performed in the same position as the diagnostic scan was made. The implication for beam delivery is that beam should be able to enter the patient from any angle over a full 4π sphere. The most logical way of achieving this is to bring the (horizontal) beam into the treatment room, and run it through a “gantry,” a rotating transport line that can put the beam at any angle in a vertical plane with the target volume

located at the “isocenter.” Oblique entry ports are obtained by rotating the patient couch in the horizontal plane. X-rays for therapy are produced now by compact electron linacs that can easily be accommodated on such a gantry, or in any event the electron beams of 20 MeV or less can be bent in ∼ 10 cm radii by reasonable magnets. Achieving “isocentric delivery” with electron beams is quite practical. However, the magnetic rigidity (Bρ) of a 250 MeV proton beam is about 2.5 tesla-meters (T-m), so gantry sizes with conventional magnets become very large. 400 MeV/amu carbon beams have a Bρ of 6.6 T-m, making isocentric delivery even more difficult. 2.3. Role for superconductivity The first machines used for investigation of protons and ions for therapy were accelerators in nuclear physics laboratories where, though not under ideal conditions, the groundwork was laid for transferring proton- and ion-beam therapy to the hospital environment. The first hospital-based proton facility was installed in 1990 at Loma Linda California [10], and consisted of a Fermilab-built 250 MeV synchrotron and three large gantries (see Fig. 3).

Fig. 3. Model of the Loma Linda proton therapy facility. The overall facility dimensions are approximately 90 ×40 m. A compact, ∼ 7-m-diameter weak-focusing synchrotron (lower left corner) provides 250 MeV proton beams to three gantry rooms and two fixed beam rooms. Over 15,000 patients have been treated since 1990. (Courtesy of Loma Linda University Medical Center.) d Studies

of abdominal sites have shown shifts of many centimeters in organ location and thickness of intervening tissue (determines the range of treatment beam) from scans taken with patients in supine versus erect positions.

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

The diameter of each gantry is 13 m. In 1993 IBA won the contract for providing a 235 MeV isochronous cyclotron with two gantries plus two fixed beam lines to the Massachusetts General Hospital. This configuration — an accelerator (normalconducting) and a backbone beam transfer line to several treatment rooms, most of them provided with gantries — has become the standard for essentially all of the new treatment centers (see Fig. 4). In the US alone, IBA and other vendors have installed eight facilities modeled after the first MGH unit (the number is growing almost monthly), and growth in other parts of the world has been equally prolific [11]. The rationale for this configuration is that the accelerator is the most expensive component, and the use of beam in each treatment room is very low (most of the time it is used in patient setup and alignment), so a single accelerator can service several — as many as five — treatment rooms. The corollary of this, though, is that a center for delivering proton (or other ion) treatments will be very large and very expensive. Costs for new proton facilities of this ilk are now on the order of US$150 million. While an argument can be made that this is not an unreasonable price for building a new radiation therapy center capable of treating over 1000 patients per year, it is nonetheless a substantial hurdle for the vast majority of hospitals that might be interested in providing proton therapy and that cannot justify a facility of this size.

Accelerator 230 MeV Energy Slits

Degrader

Fig. 4. IBA’s four-gantry layout. The cyclotron vault is at the far left; the graphite degrader and energy selection spectrometer (highlighted in the inset) passes beam into the long transport line, where it is peeled off into the appropriate treatment room. Each gantry weighs about 100 tons, and is 12 m in diameter. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

231

The drive, then, is to reduce the cost of hardware, delivery systems, and facilities to make protons more affordable. If inexpensive sources of 250 MeV protons could be built, and gantry size reduced to not require a huge vault, then one- or two-room systems could be within the reach of many more customers. The ultimate goal would be to reduce the size to where an existing vault currently housing an electron linac could be upgraded to contain a proton system. The most effective way of achieving a more compact size is to increase the magnetic field in the accelerator and delivery systems, using superconducting magnets. 2.4. Cyclotrons for therapy Isochronous cyclotrons work correctly when the magnetic field design is valid for a given ion and final energy. Since their introduction in the late 1950s, however, resistive-magnet-based cyclotrons designs have been challenged by the nonlinearity of the pole and return yoke steel. In a typical project, a succession of model magnets would be built to validate and optimize the magnetic field configuration before the design of the actual machine could be finished. It was realized in the early 1970s that, by substitution of the standard hollow copper conductor-based resistive coils in isochronous cyclotrons with superconducting coils having higher current density and substantially more ampere turns for a given size, the iron in the poles would saturate and become computationally linear [12]. This meant that for the first time one could design computationally an isochronous cyclotron, and build it with the expectation that it would work as designed. Quickly, it was shown that one had to make a choice between high bending strength and high flutter — the former yielding heavy ion machines and the latter leading to energetic proton machines [13]. The engineering superconductor of choice at that time being NbTi led immediately to cyclotrons in the 3–5 T range. The first of these was the variable energy K500 heavy ion cyclotron at MSU [14]. At 5 T, the K500 is roughly one-tenth the overall size and mass of an equivalent resistive-magnet-based cyclotron, and consumes roughly one-third the wall plug power, including the cryogenic systems. In addition to making the machines more compact, there were operational advantages to going

December 29, 2012

15:48

232

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

superconducting: reduced thermal cycling, less magnet charging hysteresis, smaller/lower power RF systems, and wider tuning ranges in the variable energy machines. However, this came with new engineering challenges: complex cryostat designs, higher magnetic forces, handling cryogens and quench protection systems. But, once commissioned, these cyclotrons have lifetimes measured in decades; all have met their performance goals, and cryogen plants have been replaced by mechanical cryocoolers in the latest systems. The K500 produced its first extracted beam in 1982, and is still in routine operation at the present time at MSU. The emphasis here on the MSU K500 is not an accident. All of the superconducting cyclotron designs now being developed for cancer therapy share most of its overall features. An attractive configuration at the present time has the cyclotron closely coupled to the gantry, providing beam to a single or just a pair of treatment rooms. This significantly reduces the cost of a therapy facility and is seen as a key to wider deployment of proton therapy systems. The Mevion Monarch (superconducting synchrocyclotron), the Varian Medical Systems PROSCAN (superconducting isochronous), both in operation, and the upcoming IBA S2C2 (superconducting synchrocyclotron) are all based on this configuration. The Varian ProScan, first built by ACCEL, now a division of Varian Medical Systems in Germany, has also been installed in the until-now-standard multiroom configuration. But, before describing these systems, the first superconducting machine to deliver therapeutic beams will be described. 2.4.1. Neutrons: MSU/Harper-Grace In the 1980s, it was not as clear as now as to whether particle therapy should be done with protons, neutrons or heavy ions. Some medical groups in fact favored neutron therapy. In 1984, Henry Blosser started the design of a superconducting 50 MeV deuteron cyclotron, under a commission from Dr. R. L. Maughan, as a neutron therapy instrument for the Wayne State University Harper-Grace Hospital in Detroit, Michigan [15, 16]. (See Figs. 5 and 6.) First operated in 1992 with patients, it ran routinely until about 2007, when it was shut down for programmatic reasons. An attempt was made to recommission it in 2010, but this effort was scuttled by a hard-to-diagnose vacuum leak that had

Fig. 5. Henry Blosser and Dr. William Powers with the completed neutron source cyclotron mounted on a rotating gantry. (Courtesy of National Superconducting Cyclotron Laboratory, Michigan State University.)

Fig. 6. Patient’s eye view of the neutron source. All equipment is hidden behind the cowling. Cyclotron rotation covers 360◦ ; movable floor planks rotate out of the way when the cyclotron is below floor level. (Courtesy of National Superconducting Cyclotron Laboratory, Michigan State University.)

developed in the cryostat. To date, this machine remains the only superconducting cyclotron to be developed for neutron cancer therapy. This compact superconducting isochronous cyclotron established many key features now seen in new machines. (See Fig. 7.) It was gantry-mounted, providing a collimated secondary neutron beam to a single treatment room. Since this machine was cooled with both liquid nitrogen and helium, a novel cryostat design was developed — and is now patented [17] — which would insure that the cryostat boiloff vent lines were always above the level of the cryogens

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

233

counterbalancing the transport magnets bringing the beam to the patient. Systems from Varian and IBA are now realizing almost exactly this same picture. Both isochronous cyclotrons and synchrocyclotrons have been developed for this application, and will be discussed below. 2.4.2.1. Varian Medical Systems/ACCEL PROSCAN isochronous cyclotron

Fig. 7. The upper yoke and poles for the 50 MeV deuteron cyclotron, without the cryostat. The neutron exit channel is at the top. The maximum field is 5.5 T, the average field is 4.6 T, and the beam radius when it strikes the target is 45 cm.

in the rotating magnet. It was a Q/A = 0.5 cyclotron, accelerating deuterons to 50 MeV total energy (25 MeV/amu), which impinged on an internal beryllium target (thus sidestepping the need for an extraction system) to generate a secondary neutron beam. Magnet steel helped harden the exiting neutron spectrum, and tungsten collimators in the yoke concentrated the neutron beam on the isocenter. The high radiation generated inside the cyclotron required the selective use of tungsten shielding in the cryostat to protect the superconducting coils. The field design was similar to that of the K500, with a 5.5 T maximum field on the hills and a 4.6 T average accelerating field. With a compact high field central region, it employed a 10 µA internal Penning ion source with separate cathode feeds on opposite sides of the median plane [18]. This ion source concept is still in use in the Varian PROSCAN cyclotron. 2.4.2. Protons The focus of development in recent years has been on compact cyclotrons for delivery of proton beams. In his 1988 report describing his deuteron machine [16], Blosser already envisioned a superconducting machine mounted on a gantry

Long-term studies of the efficacy of proton therapy by the Radiation Oncology Department at the Massachusetts General Hospital (MGH), Boston, conducted at the Harvard Cyclotron Laboratory’s (HCL) old 165 MeV synchrocyclotron, provided strong motivation [19] to acquire a dedicated proton therapy facility at MGH. The National Cancer Institute (NCI) funded a study grant, jointly conducted by MGH, HCL and the Lawrence Berkeley National Laboratory (M. Goitein and J. R. Alonso, PIs), which led to development of very complete clinical specifications [7, 8]. NCI, as a condition for providing initial construction funding for a facility at MGH, stipulated that the machine and technical components for the system would be procured from the private sector, not from a national laboratory. In response to a call issued by MGH, proposals were received for both cyclotron- and synchrotronbased systems. One of the proposals included a superconducting isochronous cyclotron design developed by Blosser and his MSU team [20]. While MGH selected the normal-conducting C235 isochronous cyclotron from IBA, work continued on Blosser’s concept [21]. The overall magnet design followed the K500 example, except that the peak field was limited to 3 T in order to have sufficient vertical focusing. At 3 T, this machine is about half the mass of the IBA C235 proton therapy cyclotron, so it is taking advantage more of the other features of superconducting cyclotrons: linear iron, minimal thermal cycling and low electrical wall power. ACCEL GmbH began to work with MSU on the machine in 2001 [22], with construction of the first system starting in 2003 as the centerpiece of the PROSCAN Project at the Paul Scherrer Institute (PSI) in Villigen, Switzerland, aimed at providing a dedicated source of protons for the PSI medical programs. Prior to this, protons were obtained by slicing off a tiny fraction of the 2 mA beam from their megawatt-class 590 MeV ring

December 29, 2012

15:48

234

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

Fig. 9. Enlarged view of superconducting coils for the PROSCAN cyclotron. NbTi coils. The current is 160 A, corresponding to about 106 ampere turns.

Fig. 8. CAD cutaway of the ACCEL PROSCAN isochronous 250 MeV proton cyclotron with the pole elevated [26].

cyclotron, and degrading them to 200 MeV at the entrance of the medical beam area. In conjunction with PSI, and with support from MSU, ACCEL began installation of the first production system in Villigen at the end of 2004. A key innovation was the elimination of the cryogen plant, in favor of a closed loop liquid helium refrigeration system operating in the thermal siphon mode with a set of four dedicated cryocoolers. An additional single-stage cryocooler provided cooling for the intermediate temperature thermal shield in the cryostat. The PROSCAN facility uses an energy degrader and energy selection system in the transport line to the treatment area to provide variable energy from 70 to 250 MeV. Intensity variations, to control the dose rate, are accomplished by means of a vertical deflector plate at the center of the cyclotron. This accurate intensity modulation coupled with a gantry with transverse beam steering [23] allows precision spot scanning, resulting in a precise 3D dose distribution in the tumor while sparing nearby normal tissue. The pioneering scanning development work done at PSI has stimulated

most system manufacturers to develop their own scanning systems for proton therapy. The switch of the clinical programs to the PROSCAN system occurred in late 2006, with the first patients treated with proton beams from the superconducting cyclotron in February 2007 [24–26]. While the first PROSCAN system was being fabricated for PSI, ACCEL contracted to build a second identical machine for Munich’s Reineker Center. The simultaneous assembly of two systems led to numerous challenges, with the Munich system lagging behind the Villigen system significantly in final acceptance. The first patient treatments occurred in Munich in March 2009. Varian Medical Systems acquired all of ACCEL in 2007, and sold the research instruments component of ACCEL to Brucker in 2008, in order to concentrate on the development of proton therapy systems. In 2012, a new dedicated Varian proton beam therapy fabrication center is fully operational in Bergish Gladbach, and five additional cyclotrons are in various stages of assembly there, with unit No. 3 recently installed at the six-treatment-room proton therapy facility under construction at the Scripps Medical Center in San Diego.

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

Superconductivity in Medicine

1230009

235

Fig. 10. Midplane cut and central region of the ACCEL cyclotron: (1) iron yoke ring, (2) cryostat containing a superconducting coil, (3) extraction electrode, (4) retracted radial beam probe, (5) extracted beam, (6) accelerating dee, (7) field-shaping magnet pole (“hill”), (8) ion source. In red are the first spiral orbits in the central region and the beam extraction path [26].

2.4.2.2. Mevion/Still River synchrocyclotron The Harvard Cyclotron was in fact a synchrocyclotron, which is weak-focusing like the original Lawrence cyclotrons, but addresses the relativistic mass increase of the proton by letting the resonance frequency fall synchronously during acceleration. The main guide magnetic field is obtained from an azimuthally symmetric pole that creates the required weak focusing to stabilize ion motion during acceleration. In contrast to isochronous cyclotrons, which rely on strong focusing, derived from the “flutter” created by an azimuthally varying magnetic field, no flutter is required. This is important since, to zero order, flutter scales inversely with the average field. Note that the long acceleration cycle, on the order of microseconds, means that intrinsically synchrocyclotrons are low-duty-factor and hence lowintensity accelerators. However, the experience at Harvard showed that the proton current that could be extracted from a synchrocyclotron was more than adequate for proton therapy. Using the magnet technology developed for the superconducting isochronous cyclotrons by the mid-1980s, one could get to a 3 T superconducting isochronous cyclotron, with the limit principally being creating sufficient flutter for stable vertical

motion. This leads to cyclotrons of between 50 and 90 tons at 250 MeV. To make them smaller and more compact, one has to operate at even higher magnetic fields. To overcome this limitation Blosser and his team at MSU looked at the possibility of reintroducing the synchrocyclotron, first in a set of studies leading to a patent for a 5 T synchrocyclotron in 1985 [27], and later in the thesis study by Xaio Yu Wu [28], which addressed in more detail fundamental acceleration issues which demonstrated that such a superconducting synchrocyclotron was indeed feasible. Little changed with respect to the feasibility of compact high-field superconducting synchrocyclotrons until Ken Gall and Timothy Antaya met in January 2003. Gall, who had earlier been a radiotherapy postdoc at Harvard, was looking for a way to realize a single-treatment-room concept, like the neutron therapy installation at Harper-Grace Hospital discussed above, but for protons at 250 MeV. To make a cyclotron on a gantry compact enough to fit in an affordable treatment room, Gall thought that he needed a synchrocyclotron above the 5.5 T limit set by the previous MSU studies. Antaya, who had just arrived at MIT, had been working on large-scale superconducting magnet systems for fusion and other applications, and had developed a broad array of new

December 29, 2012

15:48

236

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

engineering techniques for the use of advanced superconductors in difficult high-field magnets. Antaya thought that he knew how to approach the problem of scaling both the beam dynamics and the magnetic field design of synchrocyclotrons to very high fields. Gall and Antaya began a collaboration in 2003. By July 2004, Antaya and his team at MIT had demonstrated the feasibility of synchrocyclotrons at 8 T and 9 T, while Gall had formed Still River Systems to take on the challenge of the design and commercialization of a single-treatment-room proton therapy system based on the MIT synchrocyclotron design. The higher-field synchrocyclotron, at about 9 T, was chosen as the baseline. It required the use of an advanced superconductor, Nb3 Sn, which is brittle in the state where it is superconducting, but was a specialty of the Magnet Group of Joe Minervini at MIT [29]. The effort to develop the 9 T synchrocyclotron moved to detailed engineering in 2005–2006. Still River Systems began fabrication of the first synchrocyclotron in 2007, with the first extracted beam achieved at clinically required levels in May 2010. By late 2011, the first system had been installed at the Barnes-Jewish Hospital in St. Louis, which was FDAcleared for patient treatment in Spring 2012. Still River Systems became Mevion in 2011. At present, multiple systems are under fabrication at Mevion facilities, and it is possible that current orders will quickly result in a doubling of the number of superconducting cyclotrons operating worldwide. The 9 T MIT synchrocyclotron, shown in Fig. 11, includes a number of design innovations required to achieve the high-field particle accelerator, making it

Fig. 11. Ken Gall and the Monarch 250 8.9 T synchrocyclotron. (Courtesy of Mevion Medical Systems.)

the most compact 250 MeV proton accelerator built to date by any accelerator technique. The overall properties of this cyclotron are: a magnet diameter of 2 m, a magnet height of 1.5 m, a mass of less than 20 tons; the final energy is approximately 254 MeV and the RF spans 100–150 MHz. Ions are accelerated to full energy in about 16,000 turns taking 200 ms, and a repetition rate of up to 1000 MHz was planned. The central magnetic field is of order 8.9 T and the peak field in the conductor is about 11 T, making this the highest-field circular particle accelerator for protons. The coils can be dry — no cryogens — cooled by a set of mechanical Gifford–McMahan cryocoolers. Ions reaching the full radius have a small momentum spread, allowing self-extraction via a passive fixed magnetic field perturbation, with full energy being achieved at greater than 90% of the maximum pole radius of around 30 cm [30]. Figure 12 shows the single-room Mevion gantry configuration. The synchrocyclotron, known as the Monarch 250, is mounted so that the beam plane is vertical at the center of the “barrel” structure between the end arms, and the extracted beam is aimed directly at the isocenter. The energy-selection, collimation and dosimetry devices are mounted on a separate, substantially less massive rotating structure, rotating in synchronization with the main gantry. This lighter inner gantry allows greater beampositioning accuracy by avoiding the slight deflection errors introduced by rotating the outer gantry.

Fig. 12. Mevion gantry system. The cyclotron is mounted directly in line with the patient couch, and the extracted beam points to the isocenter. The gantry rotates through 190◦ . (Courtesy of Mevion Medical Systems.)

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

Superconductivity in Medicine

1230009

237

2.4.2.3. IBA S2C2 synchrocyclotron In 1996, IBA began looking at superconducting synchrocyclotrons for proton radiotherapy independently of Still River Systems and MIT. In this case a fixed machine was desired, so high compactness was not required. In addition, the term of the Blosser–Milton superconducting synchrocyclotron patent [27], which set a field threshold for new noninterfering machine design of higher than 6.5 T, had expired. With these considerations in mind, IBA settled on a synchrocyclotron concept (Fig. 13) with a field of about 6 T, which would be significantly more compact, lower-cost and lowerpower than any proton therapy cyclotron existing at the time, including its own C235, developed for MGH, which by then was installed in a number of proton centers around the world. IBA used its own development team for the overall system design, magnet and the RF system for the S2C2 shown in Figs. 14 and 15, and contracted Pierre Mandrillon’s AIMA company in Nice, France to complete the design of the synchrocyclotron acceleration mode and ion source. Design work was completed in 2010. To set the scale of this compact superconducting synchrocyclotron, the completed magnet is shown in Fig. 16, while Fig. 17 shows for size comparison the pole for the C235, IBA’s normal-conducting isochronous cyclotron. As of early 2012, the magnet for the first system has been successfully operated at full field, while other components are well advanced, and the cyclotron is expected to produce

Fig. 14. Schematic of IBA’s S2C2 synchrocyclotron. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

Fig. 15. RF system for the S2C2 synchrocyclotron, showing the single dee. Frequency variation is accomplished with a rotating capacitor in the square box off the left side. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

beam by early 2013. The full clinical system, including pencil beam scanning and a compact gantry (but with normal-conducting magnets), is called Proteus One, and is shown schematically in Fig. 18. The first installation of this system will be at Centre Antoine Lacassagne in Nice, France, with patient treatments expected to start in 2015. 2.4.3. Ions

Fig. 13. 2009 concept for IBA’s 250 MeV synchrocyclotron. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

Pioneering Bragg peak therapy with ions heavier than protons was done in the 1970s, and ’80s at the Lawrence Berkeley National Laboratory, first with

December 29, 2012

15:48

238

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

Fig. 18. Schematic drawing of IBA’s single-room Proteus One system. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

Fig. 16. Yves Jongen with the magnet of the S2C2 250 MeV IBA synchrocyclotron. Inset: ∼ 1-m-diameter pole piece. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

Fig. 17. Steel yoke for IBA’s C235 normal-conducting 230 MeV proton isochronous cyclotron. The diameter is almost twice that of the S2C2, and the weight four times as much. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

the 184” synchrocyclotron and 225 MeV/amu helium ions, then with a variety of ions from carbon to argon at the Bevalac [31]. This accelerator had capabilities for achieving a 30 cm range in tissue for any of these ions, but the heavier ions experienced large losses due to nuclear reactions, so the beam — contaminated with nuclear fragments — was anything but

pure when it reached the distal end of long-range fields. However, for shallower fields the Bragg peak was quite clean, and for ions up to neon, the fragmentation tail seen in Fig. 1 at full range was acceptably small. The rationale for the very heavy ions, proposed by Tobias [32], was that biology experiments showed the so-called oxygen enhancement ratio (OER)e to be unity for silicon and argon. The medical team opted to do the majority of the Bevalac treatments [33] with neon ions. Although the clinical results for the 433 patients treated between 1976 and 1992 were quite good, late effects seen in tissue along the entrance channel (the so-called “plateau” region upstream of the Bragg peak) suggested that neon ions may have been too heavy. This has led to carbon being the ion of choice for all follow-up investigations, now occurring in Japan and at several centers in Europe. A total of almost 8000 patients have now been treated with carbon ions at these centers. To date, accelerators for carbon ions have all been quite large synchrotrons, following the example of the Berkeley Bevalac and the GSI SIS-18 in Darmstadt. The operating facilities in Japan (HIMAC, Hyogo and Gunma) and in Europe (Heidelberg, Pavia) all have synchrotrons of ∼ 20 m or larger diameter, and transport lines bringing beam to three or more treatment rooms. Clinical research is driving the requirement that treatments with carbon be closely cross-compared with protons, so the new facilities all have the requirement of an accelerator capable of delivering both 400 MeV/amu carbon (fully stripped 6+ ions)

e Many “radio-resistant” tumors require a factor-of-3 (OER = 3) greater X-ray dose to be controlled, owing to lack of oxygen, which promotes damage-causing free radicals in the area irradiated. This comes about in rapidly growing tumors that have outstripped vascular development and so receive less blood flow. Heavier ions, with higher ionization density and less chance for repair, do not rely so much on free radicals to inhibit repair function.

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

Superconductivity in Medicine

and 250 MeV protons. The very large difference in rigidity of these beams creates an interesting challenge to the accelerator designer.f In all, the size and cost of these facilities are substantially greater (at least a factor of 2) than for a comparable proton-only facility, making the financial barrier even more daunting for those centers interested in exploring carbon as a therapy beam. Decreasing the size and cost of the technical components is clearly a highly desirable goal. While the beam delivery challenges of ions, in particular the isocentric requirement in Table 1, are substantial (and will be discussed in the following section), strong incentives exist for developing more compact sources of ion beams of suitable quality. The first design effort for a superconducting neon cyclotron, namely the 1985 EULIMA project spearheaded by Mandrillon [34], was not successful in obtaining funding support owing to its size and complexity. It was also substantially before its time, coming when the radiation oncology community was only learning about the possibilities of using ions for therapy, and cost–benefit analyses could not be favorable, because benefits were still uncertain.

2.4.3.1. IBA C400: Isochronous carbon/proton cyclotron In the past ten years, studies on using cyclotrons to produce carbon and proton beams suitable for therapy have been conducted at Catania [35] and IBA in partnership with JINR Dubna [36], both studies concluding that such cyclotrons were indeed practical, and that the two beam species could be produced in the same machine. The IBA C400 has reached a stage of maturity where funds are being sought to build the first unit. The single-stage IBA C400 isochronous cyclotron utilizes axial injection of fully stripped carbon (up to 3 emA of C6+ ), from ECR sources marketed by Pantechnik in Bayeux France [37], and of molecular hydrogen in H+ 2 form, from a multicusp source [38]. The axial injection line can accommodate more than these two sources, so any ion species that can be produced with a Q/A of 0.5 can be given

1230009

239

to the cyclotron for acceleration. As all injected ion species have the same charge-to-mass ratio, the basic acceleration parameters are similar, and only a slight change in the 70 MHz (fourth harmonic of cyclotron frequency) is needed to accelerate any of these species. The cyclotron, shown schematically in Fig. 19, has a 6.4 m outer diameter, is 3.4 m high and weighs about 650 tons. While quite large, it is still substantially smaller than the existing synchrotrons used for producing therapy beams of carbon. The pole radius is 1.87 m; the hill field is 4.5 T, the valley field is 2.45 T and fields at the coil are well within NbTi capabilities. There are plans to cool the coils with two cryocoolers running a closed loop helium thermosyphon system. Figure 20 shows the four highly curved hill segments, with RF cavities filling two of the valleys. The fully stripped species (e.g. C6+ ) reach a maximum energy of 400 MeV/amu, and are extracted via an electrostatic deflector; extraction efficiency of 80% or better is anticipated. The H+ 2, on the other hand, is given to a stripper foil placed where the beam reaches 260 MeV/amu. The breakup of the H+ 2 molecule produces two protons that spiral inward (instead of outward, as is the case with H− cyclotrons), but the azimuthal field variation can be used to produce orbits that will allow the protons to escape. Figure 21 shows that at the indicated foil

Fig. 19. IBA’s C400 isochronous superconducting cyclotron for 400 MeV/amu carbon ions and 260 MeV protons. The outer diameter of the steel is 6.3 m. (Courtesy of IBA, Louvain-laNeuve, Belgium.)

4+ , the ion for carbon and protons into the synchrotron is facilitated by using H+ 2 ions, which have the same Q/A as C usually produced from the ECR sources employed. Thus, two ion source lines merge into a single injector linac. Beam is stripped (to bare protons or C6+ ) at a few MeV/amu for injection into the synchrotron, where then the difference in rigidities must be accommodated. Though the beams are not run together, the ranges of operating parameters for these two ions are very different. f Injection

December 29, 2012

15:48

240

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

Fig. 20. Cutaway of the C400 showing four spiraled hills and two RF cavities. Two extraction lines are seen; the outer one captures protons while the inner one transports Q/A = 0.5 ions. The last dipole brings the two beams into a common transport line. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

systems, as shown in Fig. 4. Calculations indicate that degrading carbon beams is not unacceptable; the nuclear breakup into three alpha particles, while substantial, has very few of the alphas passing through the tight collimation slits in the energy selector location. The proton gantries seen in Fig. 4 will not transport the much stiffer carbon beam, but at least one treatment room can be outfitted with the proton gantry for those patients receiving proton treatments from this cyclotron. In the following section, the issues associated with gantries suitable for carbon beams will be discussed. By 2010, the detailed design and beam dynamics modeling of the C400 had sufficiently advanced to the level where vendors were asked to quote on key subsystems including the magnet and cryogenics, and a contract was in place for the deployment of the C400 prototype to the French company CYCLHAD (un CYCLotron pour l’HADron-th´erapie). CYCLHAD is a joint venture between IBA, SAPHYN (SAnt´e et PHYsique Nucl´eaire), a semipublic company in Caen, France, and financial partners. This collaboration has as a goal a construction start in 2012, in Caen, France, with the first carbon therapy treatments foreseen in 2015.

2.5. Superconducting technology in isotope cyclotrons

Fig. 21. Proton extraction from the C400. The stripper foil reduces H+ 2 to two protons that spiral in the azimuthally inhomogeneous field to the extraction point at the top of the figure. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

placement the protons will undergo two spirals and exit at the same place as the carbon beam does. The angle at the extraction point is not exactly the same, so a chicane (seen in Fig. 20) is employed to bring the two beams into a common transport line. The plan is to use the same energy selection and transport system which IBA employs with its proton

Radioisotopes have been used for many years as tracers. An amusing anecdote concerning the first recorded use — in 1911, long predating accelerator production — is told about the young Hungarian chemist (and future Nobel Prize winner for his work on tracers) George de Hevesy, who, during on a stay in England at Ernest Rutherford’s laboratory, spiked his dinner with radium to confirm that his boarding house matron was using leftovers instead of fresh food [39]. But, with the advent of artificially produced isotopes, radioactive isotopes came into widespread use both as diagnostic agents in medicine and industry (isotopes emitting low-energy gammas), and for therapeutic purposes (with alpha or long-range beta emitters). For medical applications, a key consideration was the isotope lifetime, and the distribution network to deliver the isotope to the end-user without excessive decay losses. Isotopes with several-day halflives could be produced in concentrated centers and

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

distributed to use sites. The workhorse of this production center became the ∼ 30 MeV cyclotron; IBA’s Cyclone 30 is a prime example. PET scanning (described in Sec. 4) uses positron emitters, which typically have short half-lives — too short to permit the delivery time to be more than a few hours. For this application, having production capability as close to the use site as possible is very important; most desirable is being directly in the hospital or research center where the clinical studies are performed. Many small cyclotrons are today commercially available: compact, self-shielded 10– 15 MeV machines, with normal-conducting magnets and automated systems that cycle the target material (usually in liquid or gas form) through the accelerator beam to autochemistry systems to produce a labeled pharmaceutical ready for clinical administration. The largest market is for 18 F (110 min half-life) attached to deoxyglucose, a metabolic fuel that is selectively absorbed from the bloodstream in areas exhibiting high glucose uptake, such as cancer cells. But other light isotopes are also of interest: 11 C (20 min T1/2 ), 13 N (10 min T1/2 ) and 15 O (2 min T1/2 ), probing different metabolic processes and targeting specific sites and functions.

Fig. 22.

241

These systems, provided by IBA, Sumitomo, ACSI or other commercial suppliers, accelerate H− ions and extract via stripping. The stripped ion, now a proton, bends in the opposite direction in the magnetic field, and so is easily and efficiently extracted. This technique has vastly improved the usefulness of small cyclotrons: extraction of positive ions from a cyclotron is difficult and lossy, and performance has been limited by heat and activation in the extraction region. H− extraction is essentially 100% efficient and avoids all these issues. Another advantage is that beam can be shared between two strippers, splitting the heat load on foil and target (quite often, target heating is the limiting factor), and essentially doubling the production rate. Though compact, the shielding clamshell for these self-contained cyclotrons is typically 3–4 m across. There is definitely room for a more compact solution! 2.5.1. Oxford instruments’ OSCAR superconducting cyclotron The first superconducting isotope cyclotron was OSCAR, developed by Oxford Instruments in 1989 [40]. This unit, shown in Fig. 22, was an

Oxford Instruments’ OSCAR 12 MeV superconducting yokeless isochronous cyclotron [40].

December 29, 2012

15:48

242

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

amazingly sophisticated instrument. It produced external beams of about 150 µa at 12 MeV. Superconducting coils provided an average field of 2.36 T, with no return yoke. A thin steel sleeve surrounded the coils, with a set of superconducting bucking coils to channel flux through the steel sleeve, and provided essentially complete cancelation of any stray field, thus accounting for the very low weight of the total system — less than 2 tons. All the superconducting coils were run in persistent mode. Cooling was provided by a liquid helium bath, but a cryocooled 20 K heat shield kept helium boiloff to where refilling was needed only once every two months. Three sets of sector steel poles in the 500-mmdiameter warm bore of the magnet provided the vertical focusing, and isochronicity; the valleys were open to allow full room for the 108 MHz resonators acting as terminated transmission lines. H− ions were produced from an external source, easily accessible for maintenance, and axially injected into the plane of the cyclotron. A stripper, at 220 mm radius, provided extracted proton beams with high efficiency. A total of nine production units were known to have been built. Today, several are still in operation.

Fig. 23. Tim Antaya with steel pole and yoke pieces for the Ionetix, a 12.5 MeV cyclotron called the Isotron. (Courtesy of Ionetix, Hampton, NH.)

2.5.2. Ionetix isotope production cyclotrons Ionetix has recently demonstrated a compact highfield (6 T) H+ cyclotron — named the Isotron — for ammonia 13 N (10 min T1/2 ) production, with protons at 12.5 MeV striking an internal target. Ammonia 13 N is a superior cardiac diagnostic agent but has not yet been used outside of research hospitals. This extremely compact Nb3 Sn-based system (pole shown in Fig. 23; assembled cryostat in Fig. 24) has a mass of less than one ton and consumes less than 7 kW of power during operation. It is optimized to produce unit doses of the ammonia tracer on demand. Two unit-dose systems are presently under construction for deployment at leading medical centers in 2012 and it is expected that commercial sales will begin in 2013. 2.6. Summary Superconductivity is certainly playing a key role in the medical application of particle beams. IBA, Mevion and Varian are forging a new technology path where superconductivity will be making proton therapy more and more a mainstream modality for

Fig. 24. Cutaway of the cryostat and magnet for the Isotron. (Courtesy of Ionetix, Hampton, NH.)

treating cancer. Their compact cyclotrons improve system reliability and reduce the size and cost of accelerators. These systems are in the field today, and their rapid expansion will speed the deployment of proton therapy facilities in the coming years. IBA’s new C400 cyclotron for producing carbon ions — as well as protons — will also play an important role in improving access to ion beam therapy, especially when coupled with superconducting gantries discussed in the following section, by reducing the size and cost of centers capable of treatment with heavier ions closer to the size of present-day proton facilities. Supercompact cyclotrons producing isotopes for PET and other radioisotopes can have a very large

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

Superconductivity in Medicine

impact on nuclear medicine by improving access to and convenience in obtaining valuable isotopes. 3. Beam Delivery and Compact Gantries 3.1. Gantry considerations Implementing the requirement for isocentric delivery drives the size of the room in which the patient is treated. Proton gantries, shown schematically and photographically in Figs. 25 and 26, are usually on the order of 12 m in diameter, and 10 m long. Considering that the entire room must be adequately shielded — 250 MeV protons require about 2 m of

1230009

243

concrete — the cost for enclosing a large gantry is very high. As the roof must span the entire width, too, structural integrity requirements add to the cost. The mechanical structure of the gantry is also a considerable expense. The requirement for precision of the beam at the isocenter, often expressed as a “circle of confusion,” is on the order of a millimeter radius. That is to say, the on-axis beam must pass through this millimeter sphere regardless of the orientation angle of the gantry. This places severe limits on allowed deflection of the structure, and considering the weight of the magnets requires a massive, rigid structure. The most common proton gantries weigh about 100 tons.

Fig. 25. Plan view schematic of the IBA proton gantry. To keep the magnet size and weight down, the beam-spreading system is placed after the last magnet, but this does increase the swept radius of the gantry because of the need for drift distance for the beam to reach the required field size. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

Fig. 26. Photograph of the IBA gantry support structure. The heavy, 100-ton structure is needed to preserve stiffness and alignment accuracy of beam through all rotation angles. (Courtesy of IBA, Louvain-la-Neuve, Belgium.)

December 29, 2012

15:48

244

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

Considerable cost savings could result from developing an effective “compact” gantry. Various beam line configurations have been considered [41, 42], and in some cases constructed, for reducing the gantry diameters such as the off-center configuration where the patient couch and 90◦ magnet rotate about a common center. While saving space, medical personnel have not been enthusiastic about this design because of inconvenient patient access. Nonorthogonal configurations — beam swept around a cone of, say, 60◦ instead of the full vertical plane — also save space, but lead to treatment configuration compromises that detract from their usefulness (for one thing, no vertical beam). Clear are the advantages of increasing the field strength in the last magnet, achievable with superconductivity — not only in decreasing the bending radius with higher fields, but also in decreasing the overall weight of the transport system and so reducing the demands on the structural elements of the gantry. However, the problem is far from simple. Gantry size is not solely driven by magnet strength. Upon emerging from the last magnet, and directed toward the patient, the beam must be spread laterally to cover the full size of the field in both transverse planes. If the aperture of the last bending magnet is small, then the spreading system must introduce the required angular divergence, and the drift distance from magnet to patient must be large enough to achieve the largest field size specified. The spreading system (along with field-definition and beam-monitoring instrumentation), sometimes referred to as the beam “nozzle,” plus the required drift distance could add as much as 3 m to the radius of the gantry. The gantry diameter can be reduced by placing at least part of the beam-spreading system upstream of the last bend, which increases the effective drift distance. This has the advantage, as well, of making the beam more parallel, and increasing the SAD (source–axis distance).g However, the cost of this is that the aperture of the last magnet must be large enough to accommodate the spreading beam. Again, superconductivity can make this requirement much less onerous.

One problem that must be addressed, though, is energy variability. Each treatment involves particles with a wide range of energy, corresponding to the depth and thickness of the tumor in the beam direction. For instance, a tumor 10 cm thick located between 10 and 20 cm inside the body will require protons of 120 MeV at the front (proximal) edge, and 175 MeV at the back (distal) edge (not counting energy loss in the spreading system); 34% and 83% of the maximum rigidity, respectively. As energy selection is done shortly after the accelerator, the entire beam line must accommodate the range of rigidities. And, to meet the treatment time requirement, the variation in magnet settings must be accomplished in no more than a few seconds. Ramping superconducting magnets at this rapid rate will present a considerable challenge. As we will see, various approaches are being taken to address this issue. Observe that the simplest technique, namely running the full energy beam through the gantry and performing the energy degrading just after the last magnet, is possible but carries the large drawback of substantially increasing the neutron flux the patient is exposed to from the inevitable loss of particles in this degrading system. Studies are being performed of the effect of this whole-body neutron exposure at proton facilities that use passive scattering, and where heavy collimators close to the patient are used for lateral field shaping [43]. 3.2. Proton gantries In essentially all commercially supplied proton gantries today, the beam-shaping system, whether scanning or passive scattering, is located after the last magnet. As mentioned earlier, this adds as much as 6 m to gantry diameter. For 2.5 T-m rigidity, and the 2 T magnet, the proton orbits have a radius of 1.25 m, contributing 2.5 m to the diameter. Steel yoke and gantry support structure can add another 2– 3 m, making it difficult to design a gantry smaller than 11 or 12 m in diameter. Use of superconducting magnets, with 4 T or higher fields, would save 1.25 m from beam bending, and the smaller magnets might save another 1–2 m in the gantry diameter, bringing

g A small SAD implies a very high divergence angle for the beam, and the field size at the surface of the patient will be smaller than at the treatment depth, leading to a higher dose for skin and intervening tissue (dose ∼ flux/area).

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

245

a superconducting proton gantry to about 8 or 9 m in diameter. 3.2.1. ProNova Solutions/ProCure Sketched in Fig. 27 [44] is the concept proposed by Ionex and ProCure, and being developed by ProNova Solutions for a light proton gantry based on superconducting magnets. The configuration, using combined function (dipole–quadrupole) windings into an achromatic configuration, and magnet apertures allowing up to about a 9% momentum acceptance with high transmission efficiency. The beam scanning system is located in line with the patient, and the gantry diameter is a bit over 8 m. The total mass of magnets is expected to be less than 5 tons, which will substantially decrease the total weight of the gantry. Further details are not yet available, but the goal is to have a system in the field in 2015.

Fig. 28. Carbon gantry at the Heidelberg ion therapy center. The massive steel structure supporting the magnets has been cut away in this schematic. Scanning magnets are located before the last 90◦ bend. Compare the size of the patient on the couch with the cross section of the last magnet. (Courtesy of T. Haberer, HIT, Heidelberg, Germany.)

3.3. Ion beam gantries Where superconductivity can have the largest impact is on carbon beam delivery systems [45]. All facilities treating patients today with carbon ions utilize fixed beam orientations — either horizontal, vertical or oblique (45◦ ). While robotic patient-positioning systems can provide some flexibility in beam port orientation, the clinical world still calls for isocentric delivery. The first and only existing carbon gantry has been built at Heidelberg (Fig. 28). This gantry, weighing about 600 tons, has scanning magnets

Fig. 27. Proton gantry being developed by ProNova Solutions and ProCure. Based on sets of achromatic superconducting magnets, even with the scanner mounted in line with the patient, this gantry still offers size and particularly weight advantages over conventional proton gantries [44].

upstream of the last bending dipole, which as a consequence has a bore of 20 cm and a gap of 20 cm. But this does reduce the distance between the end of the magnet and the patient. Still, the diameter of the gantry is 13 m, and its overall length is 25 m. This gantry has just completed its commissioning and treated its first patient in November 2012. 3.3.1. NIRS/HIMAC After the closure of the Bevalac, the torch for ion beam therapy was picked up by HIMAC at the National Institute for Radiological Studies in Chiba, Japan. This large complex, with two synchrotrons and three treatment rooms, started treating patients in 1994. The treatment rooms are configured with static ports — one room with a horizontal beam, one with vertical delivery and the third with both a horizontal and a vertical port. A new project, initiated in 2006, will build a second therapy complex [46] with three treatment rooms — two with both horizontal and vertical beam lines, and the third with an isocentric gantry [47]. This gantry [48], schematically shown in Figs. 29 and 30, also places the scanning magnets before the last 90◦ bend, requiring large apertures in the last magnets. The overall dimensions of the volume swept by the gantry are 5.5 m radius and 13 m length. This

December 29, 2012

15:48

246

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

Fig. 29. Layout of the HIMAC gantry, showing the two sets of small-bore (60 mm dia.) magnets BM1-6 and the last 90◦ set with the maximum bore diameter of 290 mm. (Courtesy of Y. Iwata, NIRS, Chiba, Japan.)

Fig. 30. Schematic of the new HIMAC superconducting gantry. (Courtesy of Y. Iwata, NIRS, Chiba, Japan.)

makes it approximately the size of a current proton gantry, and it is estimated that the weight of this gantry will also approximate that of the current proton installations. The first two sets of magnets, labeled BM1-6 in Fig. 29, have the same cross section and bending radius, with two sets of 70◦ bends (18◦ , 26◦ and 26◦ for BM1,2,3, and inverse for BM4,5,6) and a

30 mm bore radius. The last four magnets each provide a 22.5◦ bend, and have bore radius values of 85, 120, and the last two 145 mm to accommodate the spreading beam. The field size at the isocenter is 200 × 200 mm. The maximum dipole field in the small-bore magnets is 2.88 T; in the large-bore magnets, 2.37 T. Operation is planned to coordinate with the recently developed multienergy flat-top mode for extracted beam from the synchrotrons [49]; the small energy changes between slices can be accommodated by the superconducting magnets in about 200 milliseconds, to change current and allow for settling, then a few hundred milliseconds to deliver beam to that slice through the scanning system. All magnets are “combined-function,” with two sets of NbTi coils, the inner 8 layers wound in a cos(2θ) configuration to provide the quadrupole focusing field, the outer 26 layers in a cos(θ) configuration to provide the dipole field. Figure 31 shows the profile of this magnet, with the warm beam tube, a superinsulation layer to the stainless coil base, coils and the cold steel yoke. The total diameter of the cold mass is 500 mm. Figure 32 shows the two magnets (BM2 and BM10) that have been built and are under test. 3.3.2. CEA/ETOILE The ETOILE project, aimed at carbon/proton therapy in Lyon, France, has been under study for about 10 years [50]. A recent effort has been a joint CEA

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

Fig. 31. Cross section of the small-bore magnet. Cold steel encases the coils. (Courtesy of Y. Iwata, NIRS, Chiba, Japan.)

(Saclay)/IBA study for the design of a gantry [51]; the concept is for a “classical” Pavlovic layout [52] (a layout similar to Fig. 30), with normal magnets except for the last 90◦ dipole (Fig. 33). For a radius

247

of curvature of 2 m, the maximum bending field will be 3.2 T, while the highest field in the coils will be 5.4 T, comfortable for the NbTi conductor. The right image in Fig. 33 shows the 12 planar coils, which include 6 coils for active shielding. With no iron, the weight is substantially reduced. The planned slew rate (dB/dtmax ) is 0.04 T/s, adequate to treat one slice of a carbon field in about 1.5 s. Cooling is planned with no liquid reservoir; the 10 Sumitomo cryocoolers, capable of operating at any orientation (necessary to accommodate gantry rotation), provide adequate cooling, though the cooldown time is long (about 20 h for quench recovery from 70 K, 1 week from Troom ). The study has concluded that the concept is entirely feasible, and the project is awaiting funds for further development of the design, and for building of prototypes. 3.3.3. Tilted double-helix coil Also aiming to produce a more compact and lighter gantry, a team at LBNL is developing a novel magnet

Fig. 32. Coil packages and finished magnets for the small-bore (BM2) and large-bore (BM10) combined-function HIMAC gantry magnets. The length of the magnets is about 1 m in both cases, the outer diameter of BM2 is 690 mm, and BM10 1.24 m. (Courtesy of Y. Iwata, NIRS, Chiba, Japan.)

December 29, 2012

15:48

248

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

Fig. 33. CEA/ETOILE large-bore 90◦ bend dipole. The left most image shows the outer vacuum vessel with cryocoolers; the other two images show the coil packages, plus bucking coils for active shielding, saving weight of steel [51].

design for the final 90◦ large-aperture toroidal bend based on tilted double-helix (also referred to as canted cosine theta, CCT) windings [42]. Proposed in 1970 by Meyer and Flasck [53], a dipole field can be generated by overlaying solenoidal coils that are tilted in opposite directions (Fig. 34). This configuration cancels the solenoidal field and sums the dipole component of both. In addition, quadrupole and higher-order fields can be obtained with appropriate winding schemes, as shown for the quadrupole case in Fig. 35. Thus, with appropriate windings, a combined-function magnet, with appropriate corrections, can be produced in a compact, lightweight package.

Fig. 34. Double-helix canted coil windings showing the current path. The solenoid component is canceled, while the dipole components add. The configuration can be easily wound on a toroidal mandrel to make effective gantry magnets [42].

Fig. 36. Combined-function configuration proposed by Goodzeit et al. [54], obtained by varying the spacing of each winding around the circumference of the mandrel.

In fact, Goodzeit et al. [54] have shown that the multipole components can be introduced into a single double-helix winding by appropriately shaping the path of each winding, as illustrated in Fig. 36. The LBNL group has calculated the effect on the field of windings along a toroidal path, as would be needed for the last 90◦ gantry bend, and determined that field errors can be corrected with appropriate sextupole compensation. Small prototypes for both dipole and quadrupole magnets have been built with NbTi, and tested at LBNL. The design simplicity and low construction costs give confidence that this technique can find very effective application in compact gantries. 3.3.4. FFAG

Fig. 35. [42].

Tilted coil configuration for a pure quadrupole field

Fixed field alternating gradient (FFAG) lattices have been proposed for many different applications [55]; their particular advantages are strong focusing and very high momentum acceptance without having to change the magnet settings. Figure 37 illustrates the basic concepts for scaling and nonscaling configurations, showing alternating inward and outward bending dipoles, with also alternating

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

249

Fig. 37. Schematic of FFAG “scaling” and “nonscaling” configurations. A “scaling” lattice preserves the same tune for different momenta, not important for single-pass transport lines. The configuration can be designed to maximize momentum acceptance in a minimum aperture. FFAG basically offers excellent strong-focusing lattices [55].

gradients corresponding to each dipole direction. For a simple transport line (as opposed to an accelerator), transit times and orbit tunes are unimportant, and the lattice parameters can be optimized for maximum momentum acceptance in a minimum aperture. Trbojevic [56, 57] has designed FFAG gantries suitable for carbon beams with momentum acceptance δp/p of ± 20%, so a single field setting of the gantry magnets could transport carbon beams between 200 and 400 MeV/amu (see Fig. 38). This acceptance covers carbon ranges from about 9 cm to over 25 cm, quite adequate for most treatments. Note that the outward-bending elements of the lattice will increase the radius of the gantry, but strong superconducting magnets can compensate for this; Trbojevic’s gantry designs have radii of about 4.5 m, with magnet apertures of around 4 cm. The magnets are challenging, but concepts are believed possible using winding configurations capable of creating all the necessary field configurations, producing the required fields and gradients, and still

Fig. 38. Nonscaling FFAG gantry for carbon ions, showing trajectories for 400 MeV/amu and 200 MeV/amu. Both sets easily fit inside a 4-cm-diameter bore tube. The radius of the full gantry is no more than 4.5 m. Following the exit shown, scanning magnets and a special FFAG cell will be designed to bring the beam to the perpendicular orientation and allow for the required 20 × 20 cm scanned field [56].

remaining within the technical limits for the NbTi conductor. 3.4. Summary The potential impact of superconductivity for gantries and beam delivery can be enormous. Very significant cost savings can be realized by decreasing the weight and swept volume in gantries. Novel ideas and new technologies are emerging to effect these changes, but still need to be tested in actual clinical deployment; however, the pressing need for these new developments will ensure that they are given every opportunity to succeed. In fact, it is probably fair to say that the future of carbon ion therapy will depend on the success of these endeavors. 4. MRI Magnets 4.1. Advances in imaging Imaging quality has made huge strides in recent times. X-ray CT (computerized tomography) scanning, MRI (magnetic resonance imaging) and PET (positron emission tomography) are providing detailed spatial and functional information that has totally revolutionized diagnostic capabilities and is in addition enabling pinpoint precision for target definition in radiation therapy. Each has particular strengths, and the greatest diagnostic value is being obtained by combining information from all three. The greatest challenge comes in overlaying or matching the images from each, known as “fusion,” and is now being addressed by combining modalities into the same apparatus; so, for instance, a patient is scanned with CT, and in the same setup a PET scan is performed, thus assuring alignment of the coordinate systems for both diagnostic procedures. PET– CT devices are now available [58], and PET–MRI is

December 29, 2012

15:48

250

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

close to commercial reality. Software fusion of CT and MRI images is quite well developed [59].

4.1.1. Computerized tomography CT reached commercial viability in the 1970s. It relies on passing a thin beam of X-rays (typically 80 kVPh ) through the object being imaged, measuring the attenuation of the X-ray beam in a detector beyond the object [60]. This beam is rotated through 360◦ along with the detector, collecting attenuation information at each angle. The full data set is passed through a back-projection reconstruction algorithm to produce an image of tissue density in the plane of the slice. The patient is moved to scan subsequent slices, collecting data to provide a full three-dimensional image of the region of interest. Figure 39 shows a representative scan with a modern CT instrument. By using fan beams and arrays

of detectors, and more recently spiral motion of the rotating mechanism, the efficiency of data collection has significantly improved, reducing scanning times to a few seconds. Image quality is also being enhanced, by using dual-energy scan systems (e.g. 80 and 140 kVP). Image resolution is determined by the size of the X-ray focal spot and the detector size, with pixel sizes now in the submillimeter range. Attenuation at this energy is dominated by Compton scattering off electrons in the tissue, which in turn is highly dependent on atomic mass; so bone, with a large amount of calcium, is more easily imaged than soft tissue. The reconstruction algorithm assigns to each pixel a “Hounsfield number”i related to the electron density in that pixel. As electron density is closely correlated with energy loss or attenuation of the beam, CT is the primary tool for treatment planning for therapy with external beam radiation. 4.1.2. Positron emission tomography

Fig. 39. CT image showing high contrast between bone (skull) and soft tissue, but differentiation of soft tissue structure is difficult. (Image credit: zhuravliki/123RF Stock Photo.) h This

PET also relies on detection of photons, but in this case the photons are 511 keV back-to-back gamma rays emitted when a positron annihilates with an electron. The positrons are emitted by a radioactive nucleus disintegrating inside the subject; these positrons slow down and encounter an electron typically within a 1–2 mm distance from the parent nucleus. The annihilation gammas are given off almost perfectly aligned, so detecting the two photons allows one to draw a line defining the path along which the decaying nucleus will be located. By observing many such coincident pairs, a density map of the radioactivity can be generated [61]. The radioactive material is attached to a substance that is selectively absorbed in the tissue being imaged, so the PET spectrum will identify these tissues. As discussed in Subsec. 2.5, a common system is to attach 18 F to deoxyglucose, which is absorbed from the blood to fuel metabolic activity. PET can be used for studies of activity in the brain, or inversely areas of the brain that have been inactivated by disease such as Alzheimer’s.

stands for “kV peak,” referring to the high-voltage terminal potential, which is also the maximum X-ray energy. In this case, 80 keV electrons strike a heavy (usually tungsten) target, producing a bremsstrahlung spectrum with a maximum X-ray energy of 80 keV. This spectrum is “hardened” by filters that absorb lower-energy components, producing a spectrum that might have peak intensity at around 60 keV. i Named after Sir Godfrey Hounsfield, who shared the 1979 medicine Nobel Prize for development of the CT reconstruction algorithm.

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

Identification of metastatic lesions, also characterized by higher metabolic activity, makes PET of extreme value for locating targets for therapy. Figure 40 shows a whole-body PET scan, identifying locations of metastatic disease, secondary tumors located far from the primary mass. We have seen in Subsec. 2.5 that the production of the short-lived positron emitters is a major application for small accelerators. Desirable half-lives for these isotopes should be short enough to render insignificant the dose of radiation to the patient following the procedure, and so are less than an hour or two. Consequently, isotopes that are not available from a “generator”j should be produced very close to the use point to minimize the loss of activity due to long delivery times. Large PET centers tend to have their own cyclotrons, with automated chemistry to transport and process the target material directly into the pharmaceutical to be used for the procedure.

Fig. 40. Whole-body PET scan showing metastatic tumors, identified by high metabolic activity and the uptake of FDG (18 F-deoxyglucose). (Courtesy of Blanchard Valley Health System, http://www.bvhealthsystem.org.)

251

As stated previously, this places great emphasis on development of accelerator and chemistry systems that are compact, reliable and inexpensive. 4.1.3. Magnetic resonance imaging MRI became a practical diagnostic tool in the late 1980s, and has overtaken CT for image resolution and quality [62, 63]. Furthermore, since it does not involve ionizing radiation, the not-inconsequential dose to the patient from high-quality CT scans can be avoided. The subject is placed in a strong, uniform magnetic field, and orthogonal gradient coils adjust the field at each location within the scanned volume to bring hydrogen nuclei (protons) into resonance with an applied radio-frequency signal. A complex series of RF pulses and sequence of time variation for the gradient coils, and a sensitive pickup antenna, detect relaxation of the proton spin orientation in the magnetic field at each {x, y, z} voxel coordinate to produce exquisitely sharp images of the anatomical structures. A combination of proton density and different relaxation times in different tissues can be used to adjust contrasts, and tease out very detailed information on tissue types (see Fig. 41). The “traditional” MRI system today uses either 1.5 T or 3.0 T solenoid fields, with superconducting coils operating in persistent mode. As will be discussed, development is proceeding on systems at both higher and lower fields, for different clinical and research objectives. While image quality and information for diagnostic purposes is excellent, absolute coordinate accuracy is more difficult to establish than in the case of CT scans. The coordinates of the voxel being examined are obtained from the magnetic field map summing the main field and the instantaneous settings for the three gradient coils, and any effect that can distort the magnetic field at the location of the resonating nucleus will result in errors in coordinate location. The currents in the gradient coils predict the field at a voxel to be a given value, B(x, y, z), but a chemical shift (caused by screening of the field at the nucleus due to orbital electrons or molecular orbitals), paramagnetic susceptibility of tissues (such

j Some radionuclides valuable for imaging, such as 99m Tc (T 1/2 = 6 h), are obtained by the decay of a long-lived parent, in this case 99 Mo (T1/2 = 66 h). The parent is produced and transported (mostly, for 99 Mo, from overseas) to the use site, and becomes the “generator” of the daughter that is collected for the imaging study.

December 29, 2012

15:48

252

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

Fig. 41. MRI images of the brain, with different weightings, corresponding to different timing sequences in the pulse train. PD is proton density, a basic map of hydrogen distribution. T1 and T2 reflect different relaxation and dephasing times. Tissues affect these relaxation times in different ways, so for instance cerebrospinal fluid has a much longer relaxation time and is highlighted in the T2 image [63].

as iron in blood), or any magnetic material in close proximity to the scanned volume can introduce a ∆B that will make the nucleus resonate at a slightly different frequency, and so appear to be displaced from its actual site according to δB/δr of the gradient coil configurations. For diagnostic purposes this is usually of little consequence, but it is important for treatment planning with external radiation beams. MRI and CT images are fused to compensate for this: the fused image displays soft tissue which cannot be seen in CT, while the more accurate dimensional information from CT is used to more reliably fix the target coordinates. 4.2. Historical development of MRI The first whole-body scan was performed by R. Damadian in 1977 in the Downstate Medical School, State University of New York in Brooklyn [64] (see Fig. 42). The superconducting magnet with a 53 in. (1.4 m) warm bore was built in the Medical School shop, using NbTi wire. The coil was bathed in liquid helium, with two separate (concentric) 77 K toroidal heat shields. Designed to run at 0.5 T, most of the R&D for their MRI imaging was performed at 0.1 T. While the resolution was quite poor (see Fig. 43), it did demonstrate the viability of imaging with nuclear resonance. Though this was the first MRI image obtained, techniques developed by P. Lauterbur [65] and P. Mansfield [66] were credited with showing the path to the technology in use today. While eyebrows were raised, Damadian was passed over when the 2003 Nobel Prize in Physiology or Medicine was

Fig. 42. R. Damadian, L. Minkoff, M. Goldsmith and the “Indomitable,” the world’s first MRI scanner which they developed [64].

awarded to Lauterbur and Mansfield of laying the foundation of modern-day MRI. A recently published book, 100 Years of Superconductivity [67], provides in its chapter on medical applications an excellent summary of the commercial

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

253

Fig. 43. The first MRI image of human anatomy, obtained with the “Indomitable” device: L. Minkoff’s chest [64].

development of the field, and the succession of companies associated with development of the large and challenging magnets. Challenging because of the requirements for extreme field uniformity, bore sizes suitable for accommodating the patient to be scanned, very high fields, and the need for deployment in clinical environments with extremely high reliability to be operated by personnel not specialized in cryogenics or superconductivity. Initial developments toward commercialization of MRI were focused on relatively-low-field, mainly resistive magnets, of 0.5 T or below. However, in 1980, Oxford Instruments convinced GE that it could provide a superconducting solenoid with a bore large enough for a patient to fit in (∼ 65 cm), with a field of 2 T. The first magnet delivered to GE at its Niskayuna, New York research laboratory exhibited a substantial downward drift in the persistent current at high fields, but at 1.5 T it did run stably and reliably. The magnet was delivered in 1982, and the GE team began assembling a system capable of imaging. Jumping to this high field was considered at the time a high risk, because it was believed that the RF for proton resonance at 1.5 T (63 MHz) would be attenuated in the body and so insufficient signals would be obtained. Fortunately, this turned out not to be the case, and the first brain image at 1.5 T was obtained in 1983. The success of this endeavor led to rapid development and deployment of MRI systems, with the 1.5 T magnetic field becoming the de facto standard for these clinical instruments. In addition to GE, units were built and marketed by Siemens, Philips, Toshiba, Hitachi and several firms. Today, more than 10,000 units are operating in the field. Figure 44

Fig. 44. Typical MRI scanner in clinical use: a GE Brivo MR355 1.5 T scanner. (Courtesy of GE Healthcare. http:// www3 . gehealthcare.com/en/Products/Categories/Magnetic Resonance Imaging/Brivo MR355)

shows a modern clinical instrument; the bore is of the order of 70 cm, and the length of the magnet from cover to cover is about 1.5 m. The units weigh several tons. Initially cryostats contained up to 1000 L of liquid helium, but the development of effective cryocoolers by Sumitomo Heavy Industries in the early 1990s enabled substantial cost reductions by eliminating the need for liquefaction plants and the large inventory of helium. In the early 2000s, scanners operating at fields of 3 T were available for clinical use, with the appearance and specifications essentially identical to those of the 1.5 T units except for higher resolution, greater flexibility of imaging options and, of course, higher costs. 4.3. Safety considerations While MRI scanners are free of the ionizing radiation associated with CT and PET, other hazards related to the very large magnetic field must be considered. 4.3.1. Ballistic hazards Any magnetic object — such as a key chain, screwdriver, metal cart or other object — accidentally brought into the room containing the scanner could become a projectile capable of doing very considerable damage to the equipment, or should a patient be in the scanner, causing possibly life-threatening injuries. Mitigation of this requires detailed logging of all materials in the scanner room, and very tight

December 29, 2012

15:48

.

254

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

discipline for personnel entering the room, especially regarding the contents of their pockets. As the magnets are run in persistent mode, i.e. are energized around the clock, this discipline extends to the offhours janitorial crew and the mops and pails they use to clean the room. To minimize the stray fields, and help with mitigating this projectile problem, manufacturers now provide magnetic shielding, either in the form of iron often built into the room because of the large bulk required, or through active compensating coils — separate superconducting windings that buck the fringe field. 4.3.2. Patient screening The high magnetic fields can have disastrous effects on individuals unfortunate enough to not have been carefully screened prior to placement in the MRI magnet. Some implants, such as pacemakers, cardiovascular catheters and surgical clips, may be severely affected by the fields. A very thorough and detailed website, http://www.MRISafety.com, maintained by Frank G. Shellock, categorizes hundreds of devices according to classification of “safe,” “conditional” or “unsafe” in the default field of 1.5 T. At higher fields many of the “conditional” ratings will probably become “unsafe.” One very serious consideration that must be explored during screening is whether the patient has received injury from an explosion or impact that may have left (magnetic) metallic fragments imbedded in tissue, organs or, worse, the brain or eye. The process of inserting the patient into the scanner is likely to cause these objects to move, with possibly serious injury to the patient. At the very least, severe distortions of the image will result. (See Fig. 45.) 4.3.3. Operational housekeeping Obtaining good images requires having very strong and efficient RF transmitters and sensitive pickups. It is often not enough to use coils located at the edge of the inner bore, which would be used for a wholebody scan. To obtain the highest resolution and contrast for a particular anatomical site, very specialized coils are fabricated that are placed in close proximity to the site, such as wrapping around a knee or elbow, or close to the skin for certain sites or organs. (See Fig. 46.) After the patient is placed on the couch, the

Fig. 45. Distortion caused by a metallic fragment, in this case a small piece of shrapnel from a grenade explosion. The magnetic fragment has affected the magnetic field in its vicinity, producing an uncorrectable artifact [63].

coil is put at the appropriate location and the patient is then translated into the magnet bore. Great care must be taken by the technician that the cables to the RF coil are dressed properly along the patient, and specifically that there are no loops in the cables. Cases of severe burns have been recorded where inductive energy absorbed by such loops caused rapid heating of the cable. While already a problem with the 1.5 T magnets, the effects become considerably more serious in the new research scanners operating at 7 T, or even up to 12 T.

4.3.4. Psychological impact The environment inside the scanner is not exactly the most pleasant for many patients, in particular those with tendencies toward claustrophobia. Although the technology is improving, scan times are still relatively long (several minutes at least), and the pulsing sequence of the gradient coils and RF system produces substantial acoustic noise. Tolerance to this environment should be explored during the screening process, and patients likely to experience problems discouraged from undergoing an MRI procedure. Manufacturers are addressing this particular issue by technological improvements that decrease

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

255

Fig. 46. Collection of specialized RF coils tailored to fit closely to the anatomical site being scanned. The products shown marketed by Hitachi. (http://www.hitachimed.com/products/mri/oasis/CoilGallery/index.html)

scanning time, reduce acoustic noise and prevent the patient from feeling like they are being stuffed into a torpedo-launching tube. Specifically, Toshiba claims to have improved the structural containment of the gradient coils to produce significantly quieter scans [68]. Hitachi and now others are developing “open” scanners (see Fig. 47) which do not confine the patient in a narrow tube. The configuration of the field is a vertical dipole field instead of the customary solenoid field, but to provide the largest possible opening for the patient (in the Hitachi case, the pole separation is around 80 cm) the maximum field strength is only about 1.2 T. Providing the field uniformity over a suitably large FOV (field of view) is also quite a challenge.

Fig. 47. The Hitachi OASIS system, a split-coil vertical field 1.2 T system with a very open configuration. (http://www. hitachimed . com / products / mri / oasis / specifications / index . html)

4.4. Research directions R&D is proceeding in two directions: toward higher fields for better resolution and faster scan times, and toward lower fields for various clinical reasons. 4.4.1. Lower fields Though it is known that higher fields produce better results, significant effort is being devoted to development of scanners operating at lower fields. The reasons for this are mostly clinical, and the challenge for the industry is to improve image quality to acceptable levels at these low fields. Two main reasons driving technology in this direction are: to increase the accessibility of MRI to those patients who cannot tolerate the high fields due to implants or other at-risk materials that cannot be separated from the patient; and to open up the area containing the magnetic volume to reduce claustrophobia or make MRI accessible to heavier patients. An unfortunate demographic trend toward very large individuals, particularly in North America, is excluding an ever-increasing fraction of the population from MRI procedures with industry-standard devices. The preceding section mentioned the Hitachi OASIS model (Fig. 47), with the split-coil vertical field configuration, which allows mainly unrestricted open spaces for larger patients and a more comfortable environment. While these coils are superconducting, engineering work is being done by several manufacturers on permanent magnet designs that

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

256

could significantly reduce cost and simplify operations by reducing any need for cryogenic systems. 4.4.2. Higher fields Increasing the magnetic field offers advantages of substantial increase in signal strength, which in turn can be converted into very high spatial resolution and/or decreases in scanning times. Basic signal strength is related to polarization of the protons in the medium being scanned, i.e. to the number of protons contributing to the resonant signal. This is in turn given by a Boltzmann distribution of the form 
γhB0 N+ − N− , (1) = tanh P = + N + N− 4πkT

Fig. 48. Estimated performance of MRI scanners for higher magnetic fields [69]. Note the logarithmic vertical scale.

where P is the net polarization, γ the gyromagnetic ratio, h Planck’s constant, B0 the applied magnetic field, and kT the Boltzmann thermal distribution factor. At room temperature in a 1.5 T field, P = 3 × 10−6 , quite a small number, but still adequate for generating a good signal. For very small values of x, tanh{x} expands as x3 , (2) 3 so polarization goes roughly linearly with the applied field B0 . By increasing the field from 1.5 T to 3 T, the polarization factor increases by a factor of 2; at 10 T it is almost a factor of 7 higher. This will in itself have a huge effect on the quality of images produced, but the story does not stop there. Image quality is related to the signal-to-noise (S/N) ratio. Noise factors can also be affected quite beneficially by an increased field, and detailed analyses [69, 70] estimate that overall improvement in image quality can be as high as ∆B 2.5 . Figure 48 [69] summarizes the benefits of higher-field MRI. Figure 49 illustrates quite spectacularly the improvements in image resolution in increasing the B field from 1.5 T to 7 T. In addition, high-field magnets create the opportunity to do very elegant/exotic scans (for example, diffusion-weighted scans and others where one can actually look at the behavior of bundles of neurons) and spectroscopy that permits a determination of whether radiation/chemotherapy has actually controlled a tumor. Higher fields improve the sensitivity for fMRI (functional MRI) which measures brain activity by observing changes in associated blood flow (Fig. 50) [71]. tanh{x} ∼ x −

Fig. 49. MRI images taken in 1.5 T and 7 T scanners. The improvement in image quality is quite apparent [63].

Improved resolution for spectroscopy research is seen in Fig. 51, which highlights the increased quality of chemical shift data at higher field strength. A chemical shift provides information on the molecular bonds of the resonating protons, and so can be an indicator of the structural environment. Increased signal strength also enables detection of otherwise weak signals from other chemical species such as 13 C, also shown as a curve in Fig. 48. For basic scans, improved S/N can be utilized in two different directions: increased resolution, as a smaller volume (voxel size) is needed to obtain data of adequate quality; or increased scanning speed, as the rate of data acquisition is higher. Figure 48 shows both these effects. The respective curves are not independent: the same voxel size leads to faster scan speeds, or smaller voxels scanned for the same time. A spatial resolution of tens of microns is viewed as possible in 10 T fields, while at normal resolutions scan speeds can be high enough to capture normal

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

Fig. 50.

257

fMRI (functional MRI) image of brain activity responding to different stimuli [71].

Table 2. Examples of MRI magnets in service or contemplated. Low-field magnets are usually configured with split coils to allow unrestricted access to the scan plane. 1.5 T and 3 T systems are the workhorse units for clinical applications. Research units at 7 T and 9.4 T are operational today at several centers. The Iseult project is building an 11.7 T, 90-cm-bore unit. Higher-field magnets are planned for animal studies, but will have smaller bores.

Fig. 51. Spectra for proton resonance spectroscopy (MRS) for γ aminobutyric acid (GABA) at 1.5 T and 7 T, also showing the significant improvement in the chemical shift data quality [69].

biological motions such as the heart rate without the need for gating or sampling. Table 2 summarizes the magnet configurations in use today or contemplated for future use. At the highest fields, not only are the magnets exceedingly challenging, but adverse physiological effects of magnetic interactions with tissue or moving fluids (e.g. blood) could become important. In addition, the very high resonant frequencies will have attenuation lengths short enough that penetration of the RF signal into the subject will be a limiting factor in signal strength. For example, the attenuation length of

Field (T)

Proton frequency (MHz)

Bore size (cm)

0.5 1.5 3 7

21 63 125 300

Open, split coil ∼ 70 ∼ 70 ∼ 70

9.4 11.7 14.1 17.4

400 500 600 750

∼ 70 90 40 25

Application

Clinical Routine clinical Advanced clinical Human research, near-term clinical Research Advanced research Animal studies Small animal studies

600 MHz RF in tissue is about 7 cm, so imaging large objects in 14 T fields will lose quality. Magnets up to 9.4 T are uniformly made with NbTi conductor; however, for fields of 11.7 T and higher, critical fields and currents make this conductor problematical. The Iseult project utilizes NbTi,

December 29, 2012

15:48

258

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

but requires cooling with superfluid He at 1.8 K, substantially affecting the construction and operation cost of the unit. Nb3 Sn would be required to operate the 11.7 T magnet at 4.2 K, or to reach higher fields.

4.5. High-field MRI research centers Numerous research centers and projects are operating today. Two will be highlighted: the Center for Magnetic Resonance Research (CMRR) at the University of Minnesota, and Iseult/INUMAC at CEA Saclay.

4.5.1. CMRR Specializing in research with high-field systems, this center operates numerous scanners, from a commercial 3 T Siemens Trio unit to a 9.4 T, 65-cm-bore system provided by Magnex Scientific (now Agilent). This unit (Fig. 52) saw the first research with human patients at this field, published in 2006 [72]. Two 7 T units (Magnex — passively shielded; and Siemens — actively shielded) and one 4 T unit (Oxford), all with 90 cm bores, complete the instruments on hand. A research project with Agilent is underway to develop and install a 10.5 T, passively shielded, 88-cm-bore system. This magnet will use NbTi cooled to 3 K.

Fig. 52. Photo of the 9.4 T whole-body scanner at the University of Minnesota’s Center for Magnetic Resonance Research, site of the first scan with a human subject at this high-magnetic-field value [72].

4.5.2. Iseult/INUMAC The centerpiece for the French NeuroSpin center at Saclay will be an 11.7 T whole-body scanner, which is at present under construction. Currently operating are a 3 T “trio TIM” and a 7 T research, whole-body scanner, both from Siemens (90-bore); a 7 T “Pharmascan” unit (active shielding, 16 cm bore) and 17.2 T “BioSpec 170/25” (active shielding, 25 cm bore) units from Bruker. The latter uses NbTi at 2 K. The 11.7 T, 90-cm-bore scanner is being developed as a joint French–German project, under the leadership of Pierre Vedrine, CEA Saclay. Magnet design details are extracted from Ref. 73: NbTi (cryostable — flat Rutherford cable) in direct contact with supercooled (1.8 K) helium is wound into 170 double-pancake coils of 1 m inner diameter and 1.9 m outer diameter, for a total solenoid length of 3.8 m. Concentric shielding coils, outside the main solenoid, reduce the stray field to acceptable levels. Figure 53 shows a cross section of the cryostat with solenoid and shielding windings; Fig. 54 is a cutaway projection of the scanner. The current status is that the Technical Design Report has been completed, prototype tests have been concluded and construction contracts are underway. Delivery of the system, for commissioning, is scheduled for 2014.

Fig. 53. Cross section of the Iseult 11.7 T cryostat, showing the main field pancake coils and the two (larger-radius) active shielding coils. The bore is 90 cm, and the total length of the warm tube is 3.8 m [73].

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

1230009

Superconductivity in Medicine

Fig. 54. Cutaway of the Iseult 11.7 T magnet and cryostat assembly [73].

4.6. Summary Superconductivity is certainly playing a crucial role in modern diagnostic imaging. In fact, MRI magnets are now the largest commercial application for superconducting material, and represent a level of technological maturity that is truly remarkable. That so much stored energy, at cryogenic temperatures, can be safely installed and accepted in clinical environments is a concept which pioneers in the field would have believed optimistic, if not unthinkable. Building on this remarkable achievement, further developments are proceeding. Higher-field magnets will lead to continued improvements in image quality and, through novel techniques and high-precision imaging, will become powerful tools for unraveling the complexities of human diseases and abnormalities. Development of new materials that remain superconducting at higher temperatures and higher fields will further expand capabilities, and decrease the hardware and operations costs of units, bringing MRI within substantially greater universal reach for universal healthcare. 5. Future Directions With the advent of cryocoolers and careful 3D mechanical–thermal designs, cryogen plants have been eliminated from the operation of superconducting magnets for particle accelerators and beam transport systems, and this translates into more compact

259

systems that can be more widely distributed — away from research institutions to clinical and industrial environments. These magnets are still based on low-temperature superconductors, principally NbTi and Nb3 Sn, and require two-stage cryocoolers that are capable of only a few watts of heat removal at 4 K. In addition, a two-stage system requires a more complex cryostat with a more expensive thermal design. There are two developments coming in the near term, which will alter this landscape. First, operation at up to 6 T at 10–12 K with Nb3 Sn coils will simplify and lower the cost of two-stage cryocoolers and cryostats significantly, as well as reducing the wall plug power. This requires only the further development of Nb3 Sn wind-and-react coil technology, now in progress at a number of institutions and companies around the world. An even larger gain is to be made by shifting to MgB2 conductors at 20–30 K for fields of 2–4 T. This will enable the use of single-stage cryocoolers and greatly simplified cryostats. These advances will allow large-aperture superconducting magnets for scanning radiotherapy gantries. Such gantries will have higher momentum acceptance, affording faster tumor scanning and allowing more patients to receive ion beam radiotherapy treatments. They will also be more compact, enabling wider distribution of lower-cost treatments. These advances will also mean more compact devicelike particle accelerators for a broad array of applications in medicine, security, industry and basic science. The final frontier will be the use of the other HTS conductors, principally PBCO, at 70 K or above. At present, BSCCO and YBCO conductors are cost-prohibitive, and are not available in forms that can be readily engineered into superconducting coils other than racetrack configurations. However, progress with HTS-conductor-based magnets for medicine will most likely be possible soon, and will lead to significant breakthroughs: simultaneously reducing cost and complexity, simplifying operation and improving functionality. Acknowledgments Writing this article would not have been possible without the input from the many friends and colleagues. Many will be mentioned, with apologies for

December 29, 2012

15:48

260

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

any omissions. Steve Gourlay at LBNL pointed one of us (J.R.A.) to the recently published review he had contributed to: 100 Years of Superconductivity [67]. This book provided very valuable background material, particularly in the MRI area. Deep gratitude goes to Jeff Masten, chief medical physicist at the Aspirus Regional Cancer Center in Wausau, Wisconsin, for background and information regarding MRI. His thoughtful reading and comments on the MRI section give confidence that it is fairly well on track. Joe Minervini of MIT helped with general information about superconducting magnets. Regarding material for the therapy and beam delivery sections, thanks go to many individuals: ProCure’s John Cameron, IBA’s Yves Jongen, Varian’s Marcel Marc, Michael Schillo and Peter vom Stein; Yoshihiro Iwata for recent information on the HIMAC gantry, and Dave Robin at LBNL for general information on the status of compact gantry research, and particularly his novel double-helix magnet work. Dejan Trbojevic provided much information about the FFAG gantry, his specialty. We are also particularly indebted to our friends at Michigan State University, namely Felix Marti, John Brandon and John Vincent, for background information and recent developments relating to this institution as the birthplace of superconducting cyclotrons. Our only regret is that we were not able to consult Henry Blosser, the true father of the field; his state of health precluded a direct visit. We wish him well! References [1] J. R. Alonso, Accelerators for America’s future workshop: Medicine and biology, J. Health Phys. 103, 667–673 (2012). [2] T. Kamada and H. Tsujii, HIMAC: A new start for heavy ions, in Ion Beam Therapy, ed. U. Linz (Springer, 2012), pp. 611–621. [3] G. Laramore, J. Krall, T. Griffin, W. Duncan, M. Richter, K. Saroja, M. Maor and L. Davis, Neutron versus photon irradiation for unresectable salivary gland tumors: Final report of an RTOG-MRC randomized clinical trial, Int. J. Radiat. Oncol. Biol. Phys. 27, 235–240 (1993). [4] M. Catterall, The results of randomized and other clinical trials of fast neutrons from the medical research council cyclotron, London, Int. J. Radiat. Oncol. Biol. Phys. 3, 247–253 (1977). [5] B. Jones, The neutron-therapy saga: a cautionary tale. http://medicalphysicsweb.org/cws/article/ opinion/32466, 17 Jan. 2008 (link verified 28 June 2012).

[6] H. G. Blosser, J. Dekamp, J. Griffin, D. Johnson, F. Marti, B. Milton, J. Vincent, G. Blosser, E. Jemison, R. Maughan, W. Powers, J. Purcell and W. Young, Compact superconducting cyclotrons for neutron therapy, IEEE Trans. Nucl. Sci. NS-32, 3287–3291 (1985). [7] W. Chu, J. Staples, B. Ludewigt, T. Renner, R. Singh, M. Nyman, J. Collier, I. Daftari, H. Kubo, P. Petti, L. Verhey, J. Castro and J. Alonso, Performance specifications for proton medical facility. Lawrence Berkeley Laboratory report LBL-33749 (1993). [8] K. Gall, L. Verhey, J. Alonso, J. Castro, J. Collier, W. Chu, I. Daftari, M. Goitein, H. Kubo, B. Ludewigt, J. Munzenrider, P. Petti, T. Renner, S. Rosenthal, A. Smith, J. Staples, H. Suit and A. Thornton, State of the art? New proton medical facilities for the Massachusetts General Hospital and the University of California Davis medical center, Nucl. Instrum. Methods Phys. Res. B 79, 881– 884 (1993). [9] W. Chu, B. Ludewigt and T. Renner, Instrumentation for treatment of cancer using proton and light ion beams, Rev. Sci. Instrum. 64, 2055–2122 (1993). [10] J. Slater, D. Miller and J. Archambeau, Development of a hospital-based proton beam treatment center, Int. J. Radiat. Oncol. Biol. Phys. 14, 761–775 (1988). [11] PTCOG (Particle Therapy Co-operative Group), founded in the mid-1980s, provides a focus for the ion beam therapy community. A table of currently operating therapy centers is given at http://ptcog. web.psi.ch/ptcentres.html (link verified 29 June 2012). [12] J. S. Fraser and P. R. Tunnicliffe, A study of a superconducting heavy ion cyclotron as a post accelerator for the CRNL MP Tandem, Chalk River National Laboratory report CNRL-1045 (1974). [13] H. Blosser and D. Johnson, Focusing properties of superconducting cyclotron magnets, Nucl. Instrum. Methods 121, 301–306 (1976). [14] H. Blosser and F. Resmini, Progress report on the 500 MeV superconducting cyclotron, in Proc. Part. Accel. Conf., IEEE NS-26 (1979), pp. 3653–3658. [15] R. Maughan, H. Blosser, W. Powers, E. Blosser, G. Blosser, J. Vincent, G. Ezzell, C. Orton and D. Ragan, Progress with the superconducting cyclotron neutron therapy facility for Harper-Grace hospitals, Radiat. Prot. Dosim. 23, 357–360 (1988). [16] H. G. Blosser, Superconducting cyclotrons for neutron therapy. Michigan State University Cyclotron Laboratory report MSUCP-52 (1988). [17] H. G. Blosser, G. F. Blosser, E. Jemison and J. Purcell, Vented 360 degree rotatable vessel for containing liquids. US Patent 4633125 (1986). [18] T. A. Antaya, M. Mallory and J. Kuchar, Operating experience with an ion source in a superconducting

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

Superconductivity in Medicine

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27] [28]

[29]

[30]

cyclotron, in Proc. Part. Accel. Conf. IEEE NS-30 (1983), pp. 2716–2718. P. J. Kim and H. Shih, The place of ion beams in clinical applications, in Ion Beam Therapy, ed. U. Linz (Springer-Verlag, Berlin, Heidelberg, 2012), pp. 17–29. H. G. Blosser, Proposal for a manufacturing prototype superconducting cyclotron for advanced cancer therapy. MSUCL-874 (1993). J. Kim and H. Blosser, Optimized magnet for a 250 MeV proton radiotherapy cyclotron, in Proc. Cyclotrons and Their Applications, 16th Int. Conf. (2001), pp. 345–347. M. Schillo, A. Geisler, A. Hobl, H. U. Klein, D. Krischel, M. Meyer-Reumers, C. Piel, H. Blosser, J. Kim, F. Marti, J. Vincent, S. Brandenburg and J. P. M Beijers, Compact superconducting 250 MeV proton cyclotron for the PSI PROSCAN proton therapy project, in Proc. Cyclotrons and Their Applications, 16th Int. Conf. (2001), pp. 37–39. E. Pedroni. T. B¨ ohringer, A. Coray, G. Goitein, M. Grossmann, A. Lomax S. Lin and M. Germann, A novel gantry for proton therapy at the Paul Scherrer Institute, in Proc. Cyclotrons and Their Applications, 16th Int. Conf. (2001), pp. 13–17. A. E. Geisler, J. Hottenbacher, H. U. Klein, D. Krischel, H. Røocken, M. Schillo, T. Stephani, J. H. Timmer and C. Baumgarten, Commissioning of the ACCEL 250 MeV proton cyclotron, in Proc. Cyclotrons and Their Applications, 18th Int. Conf. (2007), pp. 9–14. M. Schippers, J. Duppich, G. Goitein, M. Jermann, A. Lomax. E. Pedroni, H. Reist, B. Timmermann and J. Verweij, The use of protons in cancer therapy at PSI and related instrumentation, J. Phys.: Conf. Ser. 41, 61–71 (2006). H. U. Klein, A. Geisler, A. Hobl, D. Krischel, H. R¨ ocken and J. Timmer, Design, manufacturing and commissioning of compact superconducting 250 MeV cyclotrons for proton therapy: A short report from the field, IEEE/CSC & ESAS European Superconductivity News Forum 2, 1–6 (2007). H. G. Blosser and B. Milton, Superconducting synchrocyclotron. US Patent 4,641,057 (1985). X. Y. Wu, Conceptual design and orbit dynamics in a 250 MeV superconducting synchrocyclotron, Ph.D. dissertation (Michigan State University, East Lansing, MI 1990). B. J. Senkowicz, M. Takayasu, P. J. Lee, J. V. Minervini and D. C. Larbalestier, Effects of bending on cracking and critical current of Nb3 Sn ITER wires, in Proc. Appl. Supercond. Conf. (2004), 2MG09. T. A. Antaya, J. Feng, High field synchrocyclotron extraction proof of principle demonstration with a small compensated axial bump. Plasma Science and Fusion Center, MIT, Report PSFC/RR-07-11 (2007).

1230009

261

[31] J. Castro and J. Quivey, Clinical experience and expectations with helium and heavy ion irradiation, Int. J. Radiat. Oncol. Biol. Phys. 3, 127–131 (1977). [32] C. Tobias, E. Blakely, E. Alpen, J. Castro, E. Ainsworth, S. Curtis, F. Ngo, A. Rodriquez, R. Roots, T. Tenforde and T. Yang, Molecular and cellular radiobiology of heavy ions, Int. J. Radiat. Oncol. Biol. Phys. 8, 2109–2120 (1982). [33] J. Castro, W. Saunders, C. Tobias, G. Chen, S. Curtis, J. Lyman, J. Collier, S. Pitluck, K. Woodruff, E. Blakely, T. Tenforde, D. Char, T. Phillips and E. Alpen, Treatment of cancer with heavy charged particles, Int. J. Radiat. Oncol. Biol. Phys. 8, 2191–2198 (1982). [34] P. Mandrillon, C. Carli, G. Cesari, F. Farley, N. Fietier, R. Ostojic, M. Pinardi, N. Postiau, C. Rocher, G. Rickewaert and J. Tang, Feasibility studies of the EULIMA light ion medical accelerator, in Proc. 3rd Eur. Part. Accel. Conf. (Berlin, 1992), pp. 179–181. [35] M. Maggiore, L. Calabretta, M. Camarda, G. Gallo, S. Passarello, L. Piazza, D. Campo, D. Garufi and R. LaRosa, Design study of medical cyclotron SCENT300, in Proc. Conf. Heavy Ion Accel. Technol. (Venice Italy, 2009), pp. 79–83. [36] Y. Jongen, M. Abs. W. Beeckman, A. Blondin, W. Kleeven, D. Vandeplassche, S. Zaremba, V. Aleksandrov, A. Glazov, S. Gurskiy, G. Karamysheva, N. Kazarinov, S. Kostromin, N. Morozov, E. Samsonov, V. Shevtsov, G. Shirkov, E. Syresin and A. Tuzikov, IBA C400 Cyclotron project for hadron therapy, in Proc. 18th Conf. Cyclotrons and Their Applications (2007), pp. 151–153. [37] ECR ion source manufacturer. www.pantechnik.fr/ (site accessed 18 July 2012). [38] K. N. Leung, D. Bachman and D. McDonald, RF driven multicusp ion source for particle accelerator applications, in Proc. Eur. Part. Accel. Conf. (1992), pp. 1038–1040. [39] P. W. Frame, Tales from the atomic age. http:// www.orau.org/ptp/articlesstories/hevesy.htm (accessed 18 July 2012). [40] R. Griffiths, A superconducting cyclotron, Nucl. Instrum. Methods Phys. Res. B 40/41, 881–883 (1989). [41] E. Pedroni, Gantries, in PTCOG49 (Chiba, Japan, 2010). http://ptcog.web.psi.ch/PTCOG49/presentationsEW/18-1 gantries.pdf (link verified 8 July 2012). [42] D. S. Robin, D. Arbelaez, S. Caspi, C. Sun, A. Sessler, W. Wan and M. Yoon, Superconducting toroidal combined-function magnet for a compact ion beam cancer therapy gantry, Nucl. Instrum. Methods Phys. Res. A 659, 484–493 (2011). [43] D. Brenner and E. Hall, Secondary neutrons in clinical proton radiotherapy: a charged issue, Radiothe. Oncol. 86, 165–170 (2008). [44] J. Cameron, private communication.

December 29, 2012

15:48

262

WSPC/253-RAST : SPI-J100

1230009

J. R. Alonso & T. A. Antaya

[45] M. Pullia, Carbon ion gantries, in Proc. PARTNER Course on Accelerators and Gantries, CERN Conf. ID 174714 (2012). http://indico.cern.ch/contributionDisplay.py?contribid=13&confid=174714 (link verified 19 July 2012). [46] K. Noda, T. Furukawa, T. Fujimoto, T. Inaniwa, Y. Iwata, T. Kanai, M. Kanazawa, S. Minohara, T. Miyoshi, T. Murakami, Y. Sano, S. Sato, E. Takada, Y.Takey, K. Torikai and M. Torikoshi, New treatment facility for heavy-ion cancer therapy at HIMAC, Nucl. Instrum. Methods Phys. Res. B 266, 2182–2185 (2008). [47] T. Furukawa, T. Inaniwa, S. Sato, Y. Iwata, T. Fujimoto, S. Monohara, K. Noda and T. Kanai, Design study of a rotating gantry for the HIMA new treatment facility, Nucl. Instrum. Methods Phys. Res. B 266, 2186–2189 (2008). [48] Y. Iwata, K. Noda, T. Shirai T. Murakami, T. Furukawa, S. Mori, T. Fukuta, A. Itano, SK. Shouda, K. Mizushima, T. Fujimoto, T. Ogitsu, T. Obana, N. Amemiya, T. Orakasa, S. Takami, S. Takayama and I. Watanabe, Design of a superconducting rotating gantry for heavy-ion therapy, Phys. Rev. ST Accel. Beams 15, 004701 (2012). More recently, private communication with Y. Iwata. [49] Y. Iwata, T. Kadowaki, H. Uchiyama, T. Fujimoto, E. Takada, T. Shirai, T. Furukawa, K. Mizushima, A. Takeshita, K. Katagiri, S. Sato, Y. Sano and K. Noda, Multiple-energy operation with extended flattops at HIMAC, Nucl. Instrum. Methods Phys. Res. A 624, 33–38 (2010). [50] J. Conto, ETOILE: the hadrontherapy project for Lyon (France), in Proc. Eur. Part. Accel. Conf. (2002), pp. 2724–2726. [51] F. Kirchner, Superconducting gantries, in Proc. PARTNER Course on Accelerators and Gantries, CERN Conf. ID 174714 (2012). http://indico.cern. ch/contributionDisplay.py?contribid=11&confid=17 4714 (link verified 19 July 2012). [52] M. Pavlovic, E. Griesmayer and R. Seemann, Beam-transport study of an isocentric rotating ion gantry with minimum number of quadrupoles, Nucl. Instrum. Methods Phys. Res. A 545, 412–426 (2005). [53] D. I. Meyer and R. Flasck, A new configuration for a dipole magnet for use in high energy physics application, Nucl. Instrum. Methods 80, 339–341 (1970). [54] C. Goodzeit, R. Meinke and M. Ball, Combined function magnets using double-helix coils, in Proc. Part. Accel. Conf. (2007), pp. 560–562. [55] M. Craddock, New concepts in FFAG design for secondary beam facilities and other applications, in Proc. Part. Accel. Conf. (2005), pp. 261–265. [56] D. Trbojevic, FFAGs as accelerators and beam delivery devices for ion cancer therapy, Rev. Accel. Sci. Technol. 2, 229–251 (2009).

[57] D. Trbojevic, Non-scaling fixed field alternating gradient gantries, in Proc. 2nd Workshop on Hadron Beam Therapy of Cancer (Erice, 2011). http:// erice2011.na.infn.it/TalkContributions/Trbojevic. pdf (link verified 9 July 2012). [58] E. Lin and A. Alavi, PET and PET/CT: A Clinical Guide (Thieme, New York, 2009). [59] G. Sannazzari, R. Ragona, M RuoRedda, F. Giglioli, G. Isolato and A. Guarneri, CT–MRI image fusion for delineation of volumes in three-dimensional conformal radiation therapy in the treatment of localized prostate cancer, Br. J. Radiol. 7, 603–607 (2002). [60] E. Seeram, Computed Tomography: Physical Principles, Clinical Applications, and Quality Control (Saunders Elsevier, St. Louis, 2001). [61] G. Saha, Basics of PET Imaging (Springer, 2004). [62] K. Westbrook, C. Kaut Roth and J. Talbot, MRI in Practice (Wiley–Blackwell, 2011). [63] E. Blink, Basic MRI physics. http://mri-physics. net/textuk.html (link verified 24 June 2012). [64] R. Damadian, M. Goldsmith and L. Minkoff, NMR in cancer: XVI. FONAR image of the live human body, Physiol. Chem. Phys. 9, 97–107 (1977). [65] P. C. Lauterbur, Image formation by induced local interactions: Examples employing nuclear magnetic resonance, Nature 242, 190–191 (1973). [66] P. Mansfield and A. Maudsley, Medical imaging by NMR, Br. J. Radiol. 50, 188–194 (1977). [67] H. Weinstock, Medical applications, in 100 Years of Superconductivity, eds. H. Rogalla and O. Kes (Taylor & Francis Group, 2011), pp. 5598–5626. [68] http://www.toshiba-medical.eu/en/Our-ProductRange/ MRI/Technologies/Pianissimo (link verified 24 June 2012). [69] H. Wada, M. Sekino, H. Ohsaki, T. Hisatsune, H. Ikehira and T. Kiyoshi, Prospect of high-field MRI, IEEE Trans. Appl. Supercond. 20, 115–122 (2010). [70] D. Hoult, Sensitivity of the NMR experiment, Wiley Encyclopedia of Magnetic Resonance 7, 4256–4266 (1996). [71] http://en.wikipedia.org/wiki/Functional magnetic resonance imaging (link verified 8 July 2012). [72] T. Vaughan, L. Delabarre, C. Snyder, J. Tian C. Akgun, D. Shrivastava, W. Liu, C. Olson, G. Adriany, J. Strupp, P. Adersen. A. Gopinath, P. F. van de Moortele, M. Garwood and K. Ugurbil, 9.4 T human MRI: preliminary results, Magn. Reson. Med. 56, 1274–1282 (2006). [73] P. Vedrine, G. Aubert, F. Beaudet, J. Belorgay, C. Berriaud, P. Bredy, A. Donati, O. Dubois, G. Gilgrass, F. P. Juster, C. Meuris, F. Molinie, F. Nunio, A. Payn, T. Schild, L. Scola and A. Sinanna, Iseult/INUMAC whole body 11.7 T MRI Magnet status, IEEE Trans. Appl. Supercond. 20, 696–701 (2010).

December 29, 2012

15:48

WSPC/253-RAST : SPI-J100

Superconductivity in Medicine

Jose Alonso helped pioneer the development of radiotherapy with ion beams during his tenure as head of Bevalac operations at LBNL during the 1970s and ’80s. He has continued to be active in the field of ion beam therapy in advisory and consulting roles. In addition, he helped launch the SNS at Oak Ridge by managing the collaborative efforts of the various national laboratories in this project through the CD0 and CD-1 stages. After retirement from LBNL in 2002, he assisted in the assembly of the SCT and Pixel trackers for ATLAS, served as Director of the Sanford Underground Lab at Homestake, and is now working with his alma mater, MIT (where he received his Ph.D. in 1967), developing compact neutrino sources based on megawatt-class cyclotrons.

1230009

263

Timothy Antaya received his Ph.D. in Accelerator Physics from Michigan State University in 1984. His dissertation involved the beam formation processes at the center of the K500 cyclotron, the first superconducting cyclotron and the first accelerator of any type to employ a superconducting magnet. At MSU he contributed to the development of all the MSU SC cyclotrons and led the ECR Ion Source Program. After seven years at Babcock & Wilcox, designing large superconducting tokamak and detector magnets, he joined MIT in 2002, where he introduced a new class of compact high-field superconducting cyclotrons. The first of these, a 9 T synchrocyclotron, is now being commercialized for single-treatment-room proton beam radiotherapy. In late 2011, Dr. Antaya left MIT to join Ionetix, a company he had founded earlier for the commercialization of device-like compact cyclotrons for medicine, security and basic science.

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

This page intentionally left blank

1203001˙book

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Reviews of Accelerator Science and Technology Vol. 5 (2012) 265–283 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300101

Industrialization of Superconducting RF Accelerator Technology Michael Peiniger,∗ Michael Pekeler† and Hanspeter Vogel‡ RI Research Instruments GmbH, Friedrich-Ebert-Straße 1, 51429 Bergisch Gladbach, Germany ∗[email protected] †[email protected] ‡[email protected] Superconducting RF (SRF) accelerator technology has basically existed for 50 years. It took about 20 years to conduct basic R&D and prototyping at universities and international institutes before the first superconducting accelerators were built, with industry supplying complete accelerator cavities. In parallel, the design of large scale accelerators using SRF was done worldwide. In order to build those accelerators, industry has been involved for 30 years in building the required cavities and/or accelerator modules in time and budget. To enable industry to supply these high tech components, technology transfer was made from the laboratories in the following three regions: the Americas, Asia and Europe. As will be shown, the manufacture of the SRF cavities is normally accomplished in industry whereas the cavity testing and module assembly are not performed in industry in most cases, yet. The story of industrialization is so far a story of customized projects. Therefore a real SRF accelerator product is not yet available in this market. License agreements and technology transfer between leading SRF laboratories and industry is a powerful tool for enabling industry to manufacture SRF components or turnkey superconducting accelerator modules for other laboratories and users with few or no capabilities in SRF technology. Despite all this, the SRF accelerator market today is still a small market. The manufacture and preparation of the components require a range of specialized knowledge, as well as complex and expensive manufacturing installations like for high precision machining, electron beam welding, chemical surface preparation and class ISO4 clean room assembly. Today, the involved industry in the US and Europe comprises medium-sized companies. In Japan, some big enterprises are involved. So far, roughly 2500 SRF cavities have been built by or ordered from industry worldwide. Another substantial step might come from the International Linear Collider (ILC) project currently being designed by the international collaboration GDE (‘global design effort’). If the ILC will be built, about 18,000 SRF cavities need to be manufactured worldwide within about five years. The industrialization of SRF accelerator technology is analyzed and reviewed in this article in view of the main accelerator projects of the last two to three decades. Keywords: Superconducting RF; cavities; accelerator technology; industrialization; technology transfer.

1. Introduction

2. SRF Accelerator Technology

The realization of worldwide accelerator projects calls for manufacturing capabilities in industry. During the R&D and early prototyping phase of these projects, industry has often been involved at a job shop level in manufacturing single components of SRF cavities, and no responsibility was taken by industry for the final cavity. In this article we will not report on these job shop activities but will focus on work where the contribution of industry is at a somewhat higher level, meaning that complete SRF cavities and/or SRF accelerator modules meeting a clearly defined technical specification are produced.

SRF cavities [1] have advantages allowing the design of new accelerators which cannot be realized with conventional room temperature cavities. 2.1. Advantages of SRF cavities 2.1.1. Continuous wave operation at high accelerating gradients The benefit of superconductivity is the loss of electrical resistance at temperatures below the critical temperature. In contradiction to the DC resistance, the RF resistance is not completely zero when the superconducting stage is reached but still

265

December 29, 2012

17:32

266

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

many orders of magnitude lower than in the normalconducting stage. The quality factor Q0 of the cavity scales inversely with the RF resistance and with the power consumption of the cavity. Normal-conducting cavities reach quality factors of 104 and superconducting cavities 109 –1011. These remarkable high quality factors allow continuous wave (cw) operation of SRF cavities with little power consumption. Only about 20 W RF power is needed to operate a 1-mlong 1.3 GHz SRF cavity at a 20 MV/m accelerating gradient. The same cavity operated at 20 MV/m made from copper would dissipate roughly 20 MW and can thus only be operated either in pulsed mode or at much lower accelerating gradients to avoid melting of the cavity by RF heating. As the SRF cavity must be operated at liquid helium temperatures, the 20 W power loss of the cavity is dissipated into a surrounding helium bath, and a cryogenic plant with wall plug power of some 10 kW is needed for the cavity operation.

To not degrade the purity of the niobium, cleanliness and specific quality control steps through the whole production process are mandatory. Moreover, before electron beam welding, the individual niobium parts have to be chemically cleaned by a mixture of hydrofluoric, nitric and phosphoric acid (so-called buffered chemical polishing, BCP), followed by rinsing with deionized water. During the electron beam welding, a vacuum better than 5 × 10−5 mbar must be maintained; all required to not risk any contamination inside the welding seam with foreign material or residual gases which would result in degradation of the niobium purity. Electron beam welding puts low heat into the material, causing a small distortion from welding. To achieve the correct cavity resonant frequency and to avoid excitation of unwanted HOMs, manufacturing of the cavity with highest-dimensional precision is necessary but not sufficient. Intermediate RF testing of single components prior to the final welding must be done as well.

2.1.2. Higher order mode (HOM) damping Due to the low power dissipation, SRF cavities do not necessarily have to be designed with the highest shunt impedance. This allows cavity designs with large beam pipe diameters and basically higher order mode (HOM) free cavity designs can be achieved. No HOMs are trapped inside this type of cavity but they leave — if excited by the beam — the cavity through the large beam pipes and are damped outside the cavity by resistive HOM loads. The nonexistence of HOMs for these types of SRF cavities allows stable acceleration of very high beam current without beam breakup or degradation of beam emittance, making SRF cavities very attractive for high current accelerators. 2.2. Cavity manufacturing An SRF cavity is a metallic structure normally made out of high purity niobium, a metal that becomes superconducting at temperatures below 9.2 K. In order to achieve the highest cavity performance, the high purity of the niobium has to be maintained through the complete manufacturing process. The cavity niobium parts are produced using standard metal forming and machining techniques, and are joined by means of electron beam welding.

2.3. SRF cavity surface preparation and cold RF testing High accelerating voltages and quality factors are reached by the SRF cavity only when the inner surface is absolutely free of dust or foreign materials. Otherwise, effects like thermal breakdown or unwanted high field emission loading might limit the cavity performance one or two orders of magnitude below the design values [1]. The steps required to achieve a clean inner surface and to prepare the cavity for the cold RF test are: • Removal of about 150 µm in total from the inner cavity surface by buffered chemical polishing (BCP) or electropolishing (EP). • Hydrogen degassing of the cavity in an all-metal vacuum furnace at temperatures between 600◦ C and 800◦C (hydrogen can enter the bulk niobium during the BCP or EP process). • Immediate rinsing of the cavity volume with DI water. • Transfer of the cavity into clean room class ISO4 (equivalent to US FED class 10). • High pressure rinsing of the inner cavity surface with particle-filtered DI water for several hours.

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology

• Drying and assembly in an ISO4 clean room, and sealing of the cavity volume with all-metal seals and RF antennas. • Pump-down and leak check with an oil-free pumping station and leak checking system. This recipe is valid only for cavities manufactured from bulk niobium. As will be explained later, CERN developed a cavity manufacturing process by sputtering a thin niobium film onto copper cavities. For the Nb/Cu cavities no chemical treatment or vacuum furnace annealing is done after niobium coating, but rinsing with DI water and assembly in an ISO4 clean room are accomplished in the same way. It is standard practice that the individual cavity will undergo a cold RF test before it is assembled to the module and used in the accelerator. For the RF test, the cavity is completely immersed in liquid helium using normally a bath cryostat. The cavity quality factor is measured in dependence of the accelerating gradient as a proof of functionality before operation in the accelerating module. 2.4. SRF accelerator modules A cut view of a typical SRF accelerator module, in this case a Cornell-designed 500 MHz single cell module, is shown in Fig. 1. The cavity is placed inside a helium vessel, allowing operation at liquid helium temperatures. The helium vessel is surrounded by a thermal shield cooled by either liquid nitrogen or gaseous helium and a vacuum vessel, both needed to reduce the static heat load of the SRF module into the helium bath. As described above, little RF power is needed to excite the RF field inside the cavity. Much more power (up to several 100 kW) is needed to accelerate the beam, especially at high beam currents. The RF power is transferred from the power source to the cavity by the RF input coupler (either a coaxial or waveguide coupler). The design of the coupler needs to provide a vacuum barrier via an RF window and must be optimized for low RF losses and low static heat load to the helium system. Next to the cavity, beam pipe sections are mounted and terminated with all-metal gate valves. Special care must be taken during the SRF module assembly, so that the cavity inner surface will not be contaminated. Therefore, the module assembly has to be partially performed in an ISO4 clean

267

room. In addition, leak tightness of the helium vessel, of the thermal shield, of the vacuum vessel and especially the cavity vacuum must be achieved. 3. Required Manufacturing Capabilities and Involved Industries The manufacture of SRF cavities requires at least state-of-the-art milling and turning machines, metal forming, electron beam welding, and chemical treatment facilities, as well as a high level of quality control. At the beginning of SRF cavity industrialization some 30 years ago, all this was found at companies involved in the manufacture of nuclear power plants. As an example, the German company Interatom GmbH, a daughter company of Siemens AG, could use the existing machining, electron beam welding, furnace and chemical surface treatment installations for the production of SRF cavities. Some 10 years later, in 1994, ACCEL Instruments GmbH became the successor for this Interatom activity after a management buyout and maintained these infrastructures. For companies not already having some of the equipment available in-house which is specific to the manufacture of SRF cavities, a significant investment of several million euros would be required to start from scratch. It should be noted that due to extensive R&D at laboratories the production quality of SRF cavities has been increased during the last 20 years. Cavities for electron linacs produced in the 1980s were designed for accelerating gradients of about 5 MV/m, whereas today these gradients are exceeding 20 MV/m. As a consequence, the quality standards in industrial production had to be improved accordingly. 3.1. Interface between laboratory and industry For each accelerator project, the best interface between the laboratory and industry has to be defined. There are basically three different levels for delivering complete SRF cavities. 3.1.1. Bare cavity For most projects, the SRF cavities are built to print and according to manufacturing specifications written by the laboratory. For some projects the helium

December 29, 2012

17:32

268

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

Fig. 1. Cut view of a typical SRF accelerator module with the main components highlighted, in this case a 500 MHz single cell module designed by Cornell University.

vessel welded to the cavity may also be part of the delivery. The acceptance of the product is given after dimensional control, RF control of the fundamental frequency and leak checking of the cavity. Basically no guarantee for later cavity RF performance is given by industry in this case. A guarantee is given only if cavity nonperformance can be traced back to obvious manufacturing errors. The cavity surface preparation, the cavity cold testing and the module assembly are under the responsibility of the laboratory. 3.1.2. Cavities ready for testing For some SRF accelerator projects, not only are the cavities mechanically manufactured but also the surface preparation is done by the industry. The

cold RF testing itself and the module assembly are still tasks of the laboratory. In this case specified accelerating gradients and quality factors have to be achieved by the cavity during the cold RF test. Either a guarantee on those parameters is given by industry or a guarantee is given that the work is performed according to a defined recipe ensured through an extensive quality assurance program. 3.1.3. Performance guarantee on SRF accelerator modules In this case, the complete manufacture and assembly of all components of an SRF accelerator module is done by the industry accepting a performance guarantee on the assembled SRF module. Eventually, intermediate testing of cavities and couplers is

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology Table 1.

269

Major companies involved in the production of SRF accelerator technology.

Company

Country

Large scale accelerator projects

Active

Babcock & Wilcox Dornier CERCA Ansaldo Mitsubishi Electric Company Mitsubishi Heavy Industries Zanon AES Roark Niowave PAVAC SDMS RI (formerly ACCEL, Siemens/Interatom)

USA Germany France Italy Japan Japan Italy USA USA USA Canada France Germany

CEBAF CEBAF, HERA, FLASH LEP200, FLASH LEP200, FLASH JAERI, KEK-B TRISTAN; KEK-B, ILC R&D FLASH, ISAC, XFEL, GANIL ILC R&D FRIB ILC R&D, FRIB ILC R&D GANIL CEBAF, LEP 200, JAERI, FLASH, LHC, SNS, GANIL, CEBAF upgrade, XFEL, CESR, ILC R&D

Early 1980s–mid-1990s Early 1980s–mid-1990s End-1980s–end-1990s End 1980s–mid-1990s Since mid-1980s Since mid-1980s Since mid-1990s Since end-1990s Since mid-2000s Since mid-2000s Since end-2000s Since end-2000s Since mid-1980s

accomplished at the laboratory using the existing test infrastructure. Today, only a few companies can offer the complete delivery of turnkey SRF modules. 3.2. Involved industries The companies known by us to be involved so far in the production of SRF accelerator technology are listed in Table 1. Several early involved companies, like Babcock & Wilcox (USA), Dornier (Germany), CERCA (France) and Ansaldo (Italy), left the business. In the last few years, new companies like Niowave (USA), Roark (USA), PAVAC (Canada) and SDMS (France) have entered the field. Research Instruments (RI) (Germany) is the successor for the cavity production activities of ACCEL Instruments, which was the successor to Interatom/Siemens for these activities. AES (US) is a successor to Northrop Grumman. To our knowledge, Zanon (Italy) has been producing SRF cavities for about 20 years. Mitsubishi Electric Company and Mitsubishi Heavy Industries (both of Japan) have been supporting and supplying Japanese laboratories with SRF cavities for a long time. In addition, Toshiba and Hitachi recently entered the field of superconducting cavity production in all cases in close collaboration with KEK [2]. 4. Large Scale SRF Accelerator Projects 4.1. S-DALINAC at the University of Darmstadt The first superconducting accelerator with a significant industrial contribution was the S-DALINAC

[3] accelerator at the University of Darmstadt, which was built from 1981 to 1987. The accelerator was designed by a collaboration of the Universities of Wuppertal and Darmstadt. The cavities were 20-cell structures with a resonance frequency of 3 GHz, which made handling, tuning and assembly of the structures challenging. The ten 20-cell cavities were built by the German company Dornier. SDALINAC (Fig. 2) is still in operation, producing beam for nuclear physics experiments and for driving an FEL.

4.2. CEBAF, LEP200, TRISTAN and HERA In the 1980s and early 1990s, large scale accelerators were built using a large number of SRF cavities at different frequencies. The continuous electron beam accelerator facility (CEBAF) at Jefferson Laboratory in the USA needed 360 five-cell cavities (Fig. 3) at 1.5 GHz resonance frequency [4]. Those cavities were produced at Interatom within a time period of three years. The surface preparation, testing and module assembly was done at Jefferson Laboratory, although this work were also performed in industry on some cavities during the prototyping phase of the project. The assembled accelerator modules were installed in the recirculating CEBAF linac to drive nuclear physics experiments. In 2007 an upgrade program of CEBAF was approved and 86 more 1.5 GHz seven-cell SRF cavities (Fig. 4) were ordered from RI. The cavities

December 29, 2012

17:32

270

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

Fig. 2. S-DALINAC in Darmstadt was one of the first linear accelerators using SRF technology with cavities to be built by industry (courtesy of the University of Darmstadt).

Fig. 3. SRF cavities and cavity components for CEBAF at Jefferson Laboratory. Waveguide input and HOM couplers are used for these cavities.

were built within a time frame of 1.5 years in 2009 and 2010 [5]. The work at RI included also the removal of 120 µm from the inner surface by BCP. During the series production, RI ran the workshop

Fig. 4. SRF cavities for the CEBAF upgrade project just coming out of the EB welding machine, after the last weld of the cavities has been performed.

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology

in a two-shift operation and could achieve a delivery rate of up to eight cavities per week. The design value of the accelerating gradient for the CEBAF upgrade cavities is roughly four times higher than the design value of the original CEBAF cavities of 5 MV/m. The upgrade of the LEP electron–positron collider at CERN to higher energies was possible only with the installation of superconducting cavities which were designed to operate at a factor of 10 higher acceleration gradients in cw operation than the previous copper cavities. The chosen resonant frequency of 352 MHz results in quite large cavities. Each used LEP four-cell cavity is about 2.5 m long. The necessary investment of high purity niobium needed to manufacture the cavities was judged by CERN to be too high. In order to reduce costs, an extensive R&D effort was started at CERN to develop a procedure allowing the manufacture of SRF cavities by sputtering a 1–2 µm thin niobium film onto a copper cavity. After that method was developed successfully, CERN transferred this technology within complete module supply contracts to three industrial partners, namely Ansaldo, CERCA and Interatom [6]. In total, 72 SRF modules (Fig. 5), each about 13 m long and housing four four-cell cavities, were produced within four years. For the SRF module assembly CERN supplied the couplers, tuners and instrumentation. Also, the intermediate cold RF test of the coated cavities was performed by CERN. In the 1980s, two more big circular electron collider accelerators using SRF technology — TRISTAN [7] at KEK (Fig. 6), Japan and HERA [8] (Fig. 7) at DESY, Germany — were built. The two

271

projects used a similar frequency (HERA 500 MHz, TRISTAN 508 MHz). For HERA 16 four-cell cavities and for TRISTAN 32 five-cell cavities were needed. The TRISTAN and HERA cryostats are both housing two cavities and using coaxial input couplers. The cavities for HERA were manufactured between 1987 and 1990 by Dornier, and the cavities for TRISTAN between 1986 and 1989 by Mitsubishi Heavy Industries [9]. The surface preparation, cavity testing and module assembly were done by the laboratories. 4.3. CESR and KEK-B The success of HERA and TRISTAN led in the late 1980s and early 1990s to the development of 500 MHz SRF modules housing one single cell cavity for high energy electron/positron colliders at Cornell and KEK. In both colliders — CESR at Cornell and the B-factory (Fig. 8) at KEK — the acceleration of very high current made the usage of SRF technology very favorable [10]. The cavities for CESR were produced by ACCEL and those for KEK-B by Mitsubishi Heavy Industries. The 10 SRF modules at KEK and the five SRF modules at Cornell are still in operation, with high reliability. As will be explained later, this type of SRF module developed through these projects is today supplied by industry and regularly used in fourth-generation light sources. 4.4. JAERI FEL At JAERI in Japan, an FEL was designed in the early 1990s using SRF technology in a more unusual way. In order to avoid the investment of a big

Fig. 5. A 13-m-long SRF accelerator module housing four pieces 352 MHz four-cell cavities after production and assembly in industry just before shipment to CERN.

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

272

Fig. 6.

Fig. 7.

The SRF accelerator module installation at TRISTAN (courtesy of KEK).

A straight section at HERA for the 500 MHz SRF system (courtesy of DESY).

cryogenic plant, the cavities should be operated in pulsed mode, reducing the overall cryogenic load from RF and standby losses to below 4 W per module and allowing one to recondense the evaporating helium with cryocoolers normally used on superconducting magnet systems. The complete SRF module design and manufacture was contracted to the German company Siemens/Interatom [11]. The four produced SRF modules (Fig. 9), two of them housing a single cell

and the other two housing a five-cell 500 MHz cavity each, reached the design specification after shipment from Germany to Japan and installation into the FEL at JAERI. 4.5. FLASH, European XFEL, TESLA and ILC In 1991 the international TESLA (Tera Electronvolt Superconducting Linear Accelerator) collaboration

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology

Fig. 8.

273

The 508 MHz single cell SRF accelerator module developed at KEK for the KEK-B-factory (courtesy of KEK).

Fig. 9.

One of the 500 MHz SRF modules after final assembly at Siemens and prior to shipment to JAERI.

was founded, with the aim of carrying out the design and prototyping of a large 500 GeV–1 TeV electron– positron linear collider. The aim of the TESLA collaboration was to boost the achievable gradients and quality factors in SRF modules and to find an economical SRF module design with which the large collider could finally

be built within a reasonable budget. An accelerator module design using eight 1.3 GHz nine-cell cavities [12] was chosen. In order to reduce the investment in the cryogenic plant, the cavities were designed to function in pulse operation at about 1.5% duty cycle. The collaboration was led by DESY, and DESY decided to build the TESLA Test Facility (TTF),

December 29, 2012

17:32

274

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

which is a 1 GeV SRF linac driving a SASE FEL. Meanwhile this TTF linac has been in full operation since 2003 and renamed to FLASH. In 2004 an international advisory committee decided that the next linear collider — when built — should be based on SRF technology, and the international linear collider (ILC) collaboration led by the Global Design Effort (GDE) replaced the TESLA collaboration. In parallel since 1997, the European XFEL linac, a 20 GeV SRF linac, was designed at DESY based on the technology developed for TTF/FLASH. The European XFEL construction was approved in 2006 and construction began in 2009. Since 1991 more than 250 of these 1.3 GHz cavities have been manufactured by a total of 10 different companies in all regions. Four hundred more such cavities each have been under construction at RI and Zanon since 2010, the last ones to be delivered in 2014. With the realization of the European XFEL, more than 1000 1.3 GHz nine-cell cavities (Fig. 10) will be manufactured and the European XFEL linac will be the accelerator using the largest number of installed SRF cavities. Achieving an accelerating gradient of 35 MV/m in such 1.3 GHz nine-cell cavities is recognized today as the highest standard and possible only if manufacture and surface preparation are performed without any compromise in view of quality. Should the ILC be realized, 18,000 more such cavities are foreseen to be built within a time frame of about five years. For the European XFEL [13], the cavities will be manufactured by RI and Zanon “ready for cold test.” At both companies the cavities will be completely manufactured (including the helium vessel) and also receive their surface preparation. To achieve this, a detailed specification and quality assurance program was written by DESY, major new infrastructure installations like clean rooms, high pressure water

rinsing systems and vacuum furnaces were realized at the companies, and the technology transfer of the surface preparation will be carried out by the laboratory experts at the companies.

Fig. 10. The superconducting 1.3 GHz nine-cell cavity used for the European XFEL project.

Fig. 11. A 400 MHz single cell copper cavity for the LHC before sputtering the 1–2-µm-thick niobium film.

4.6. LHC The SRF system of the LHC [14], the largest accelerator built so far, consists of only 16 single cell 400 MHz SRF cavities (Fig. 11). The cavities were again produced by the Nb/Cu technology and 21 cavities were manufactured at ACCEL ready for cold RF test between 1997 and 1999. The assembly of five (four installed in the LHC, one spare) SRF modules, each housing four cavities, was done at CERN. 4.7. SNS at Oak Ridge National Laboratory For the Spallation Neutron Source (SNS), a total of 109 SRF cavities (Fig. 12) were produced at ACCEL within a three-year period between 2001 and 2004

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology

275

4.8. ISAC, SARAF, SPIRAL2 and FRIB

40 MeV SARAF linac, consisting of an ion source, an RFQ and a prototype SRF module housing six SRF half wave resonators, was completely designed, manufactured, tested and assembled by ACCEL and finally delivered in 2006. The prototype SRF module showed satisfactory RF parameters; however, the beam operation with deuterons was not fully possible, mainly owing to overheating of the room temperature RFQ. SPIRAL 2 at GANIL, France is a deuteron and heavy ion accelerator with a high beam current using in total about 26 QWRs [18]. The cavities with a helium vessel (Fig. 13) were built between 2007 and 2009 at ACCEL, Zanon and SDMS. Cavity testing and module assembly were done by the French laboratories IPN-CNRS and CEA. Currently these laboratories are performing the module assembly. The so-far-largest accelerator for heavy ions using SRF technology, FRIB, is currently under construction at Michigan State University (MSU) [19]. In total, about 340 HWR and QWR cavities are planned to be used. Ten early QWR cavities were produced by Niowave. They are utilized in the MSU Re-Accelerator project, RεA [20]. The first 160 HWR cavities for FRIB were contracted in 2011 to Roark, USA. The contract is split into three phases. During the currently running first phase, two prototype cavities have been built following surface preparation and RF testing by MSU. After that a preseries phase of 12

For acceleration of particles (heavy ions or low energy protons) at velocities well below the speed of light, other cavity designs than the above-shown elliptical cavities are needed. Quarter wave (QWR), half wave (HWR) and different kinds of spoke cavities are chosen. The 20 cavities for the ISAC-II accelerator [16] at TRIUMF were produced between 2002 and 2005 by Zanon. The cavities are quarter wave resonators (QWRs) for acceleration of heavy ions. These were the first purely industrially produced cavities for heavy ions. Previously, heavy ion accelerator cavities had been produced by job shopping or by the laboratories, like the QWRs used at the ATLAS facility (Argonne, US) or in the Indus accelerator project (India). In 2002 Israel’s Soreq institute contracted a design study to ACCEL for a 40 MeV proton/ deuteron linac based on SRF accelerator technology [17]. After that design study the first phase of the

Fig. 13. Four of a total of 16 medium beta QWR cavities equipped with a helium vessel after delivery from ACCEL to IPN, France and prior to assembly to the SRF modules (courtesy of IPN).

Fig. 12. For the Spallation Neutron Source at Oak Ridge, 35 low beta (top) and 74 high beta (bottom) SRF cavities (805 MHz) were built.

[15]. The cavities were delivered from ACCEL to Jefferson Laboratory, which was responsible for the cavity treatment and module assembly. The assembled modules were then shipped from Jefferson Laboratory to Oak Ridge, where the SNS has been operating since then without any problem with the SRF system.

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

276

cavities followed by the series production is planned. The last series cavity should be delivered in 2016. The remaining cavities for FRIB are foreseen to be ordered soon. 5. SRF Technology Transfer The development of mature SRF accelerator modules requires not only deep knowledge of beam dynamics and mechanical and cryogenic engineering, but also the manufacture and testing of prototype cavities, couplers and cryostats as well as operation of the SRF modules with beam. In summary, such SRF module development is a multi-man-year effort for at least 3–5 years. 5.1. License agreements 5.2. SRF for high current storage rings As stated above, it is very hard, if not impossible, for a laboratory with little or no knowledge of SRF technology to design, build and test an SRF system by them. In 1998 the Taiwanese laboratory NSRRC decided to upgrade their existing light source, TLS (Taiwan Light Source), with SRF accelerator modules in order to achieve higher beam currents. NSRRC looked for an existing SRF module design which could be used at TLS without a major redesign which would incorporate the possible risk of “surprises” after installation. Two existing RF

Fig. 14.

module designs were identified which could be used at TLS: the SRF module developed at KEK for their B-factory and the SRF module developed at Cornell University for CESR. Both modules had been working successfully with beam for several years. Finally, NSRRC had chosen the Cornell technology and was looking for an industrial partner able to manufacture the SRF modules “turnkey” with a performance guarantee within a fixed price contract. In 1999 ACCEL concluded with Cornell a technology transfer agreement through which all existing know-how of the SRF module design and manufacture was given to ACCEL, allowing ACCEL on the other hand the marketing of the Cornell SRF module design to other potential customers/laboratories [21]. ACCEL concluded a contract with NSRRC in 2000 and manufactured two SRF modules until 2004, when one module was installed in the storage ring (Fig. 14), whereas the second serves as a hot spare. The installed SRF module has been in operation at TLS since then [22]. Very soon after the NSRRC decision for usage of SRF, other light sources decided to go the same way, purchasing turnkey accelerator modules from industry. Even Cornell University ordered two more SRF modules from industry. Since then, on average, one module has been contracted every year, as can be seen from Table 2. Most customers decided to go with the Cornell design (Cornell, Canadian Light Source

Installation of a turnkey 500 MHz SRF module into the TLS storage ring at NSRRC, Taiwan.

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology Table 2.

Overview of large-scale SRF accelerator projects worldwide.

Project

Laboratory

Country

Product

Type

No. of Cavities

MHz

Production time (SRF; estimate)

S-DALINAC TRISTAN HERA CEBAF LEP 200

Darmstadt KEK DESY JLAB CERN

Germany Japan Germany USA Switzerland

Cavity Cavity Cavity Cavity Module

20-cell 5-cell 4-cell 5-cell (4 × 4)-cell

10 32 16 360 238

3000 508 500 1500 352

1981–1985 1985–1989 1986–1990 1988–1992 1990–1995

JAERI CESR KEK-B TTF/FLASH

JAERI Cornell KEK DESY

Japan USA Japan Germany

Module Cavity Cavity Cavity

4-cell 1-cell 1-cell 9-cell

4 5 10 160

500 500 509 1300

1991–1993 1993–1995 1993–1998 1993–2005

LHC SNS ISAC-II SARAF GANIL

CERN ORNL TRIUMF Soreq GANIL

Switzerland USA Canada Israel France

Cavity Cavity Cavity Module Cavity

1-cell 6-cell QWR HWR QWR

21 109 20 6 16+12

400 805 84 172 88

1997–1999 2001–2004 2002–2005 2002–2009 2007–2009

CEBAF upgrade XFEL FRIB

JLab DESY MSU

USA Germany USA

Cavity Cavity Cavity

7-cell 9-cell HWR

86 800 160

1500 1300 322

2009–2010 2010–2014 2011–2016

Table 3.

277

Manufacturer

Dornier MHI Dornier Interatom Interatom, CERCA, Ansaldo ACCEL ACCEL MHI Dornier, CERCA, Ansaldo, Zanon, Siemens, ACCEL ACCEL ACCEL Zanon ACCEL ACCEL, Zanon, SDMS RI RI, Zanon Roark

Turnkey SRF modules for light sources produced by industry.

Project

Laboratory, country

No. of SRF modules

Production time

Design

Industry

CESR TLS CLS BEP-II DLS

Cornell, USA NSRRC, Taiwan UOS, Canada IHEP, China DIAMOND, UK

SSRF SOLEIL PAL-II TPS NSLS-II

SSRF, China SOLEIL, France PAL, S. Korea NSRRC, Taiwan BNL, USA

2 2 2 2 3 1 3 1 3 3 2

1999–2002 1999–2004 2000–2003 2003–2006 2003–2006 2012–2014 2005–2007 2005–2008 2010–2013 2010–2013 2011–2014

Cornell Cornell Cornell KEK Cornell Cornell Cornell LEP/LHC Cornell KEK Cornell

ACCEL ACCEL ACCEL MHI ACCEL RI ACCEL ACCEL RI MHI AES

[23], DIAMOND Light Source [24], Shanghai Light Source [25] (Fig. 15) and Pohang Light Source). The KEK-B style SRF module (slightly redesigned from 508 MHz to 500 MHz) was installed at the BEPC-II collider in Beijing [26] (Fig. 16), and has been chosen for the new Taiwan Photon Source at NSRRC as well. The KEK-B style SRF modules can be purchased from Mitsubishi Heavy Industries with support from the KEK experts. When Brookhaven National Laboratory decided to use SRF accelerator technology for their new light source NSLS-II, AES also decided to make a technology transfer agreement with Cornell University and finally won the contract on the delivery of two

turnkey SRF modules of that design. These modules are currently under production and will be delivered in 2014. It should also be mentioned that the SRF system of LEP and LHC triggered the development of another SRF module design for light sources. A collaboration of CERN and CEA developed a 352 MHz SRF module housing two single cell cavities produced by the Nb/Cu technique. A first module was built by this collaboration in the early 2000s and installed into the SOLEIL Light Source in France [27]. A second module was contracted to ACCEL in 2005. ACCEL built the module using the know-how

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

278

Fig. 15.

Two 500 MHz SRF accelerator modules during installation into the storage ring at SSRF, Shanghai.

Fig. 16. Two 500 MHz SRF accelerator modules produced by Mitsubishi Heavy Industries and KEK for the BEP-II collider at IHEP, China (courtesy of KEK).

gained through the LEP and LHC contracts. The module was delivered in 2008 and has been in operation at SOLEIL since then (Fig. 17) [28].

5.3. SRF for low energy FELs A second, similar way of marketing SRF modules developed at one laboratory through industry for

other laboratories could be accomplished for the SRF modules developed at Forschungszentrum Dresden (FZD) for the ELBE FEL accelerator [29]. The ELBE module uses two 1.3 GHz nine-cell cavities developed by the TESLA/TTF collaboration in one cryostat. However, the cavities are operated in cw mode (and not in pulsed operation). Consequently, all other auxiliary systems, like the input coupler,

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology

Fig. 17.

Fig. 18.

279

The SOLEIL 352 MHz SRF module allows maintenance through holes in the vacuum vessel.

Illustration of the SRF module developed at FZD for the ELBE FEL accelerator, housing two XFEL-type cavities.

tuner and helium vessel, had to be redesigned for cw application. This work was done at FZD with the help of the Stanford HEPL group in the 1990s (Fig. 18).

After the first two SRF modules were equipped with cavities from ACCEL and installed in ELBE, other laboratories are now evaluating the use of this technology for FEL application. ACCEL has

December 29, 2012

17:32

280

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

established a license/technology transfer agreement with FZD on the marketing of those SRF modules. Two modules were already produced for Daresbury Laboratory between 2004 and 2007 [30] (Fig. 18). 5.4. SBIR In the USA, laboratories and industry can work together within the framework of SBIR (Small Business Innovative Research) on the development of SRF technology. AES and Niowave are doing SRF module development for BNL [31, 32], Jefferson Laboratory, FermiLab and Old Dominion University through this program. In addition, they are active in developing SRF technology for the Navy. 6. Current Market Situation and Future Perspectives 6.1. Market situation of the last few decades The demand for SRF technology has been almost stable through the last three decades. Despite all the very large accelerators using this technology, only about 2500 cavities have been ordered since the mid1980s, representing an average production of roughly 100 cavities per year. Estimating the value of one cavity including niobium material (which is often supplied by the laboratory) at 100 kEUR, the annual turnover for SRF cavity production worldwide is about 10 million EUR. This value alone is certainly not enough volume for the globally involved industry, keeping in mind that the production of SRF cavities not only requires high engineering and quality control standards but also expensive manufacturing installations like chemical plants, class ISO4 cleanrooms and electron beam welders. As has been mentioned before, the cavity is only a small part of the SRF accelerator. Most work following the cavity production, like cavity treatment and testing as well as SRF module assembly, is normally still done by the laboratories. Estimating the value of a turnkey SRF accelerator module to be more than a factor of 10 higher than the bare cavity, the potential business of the last few decades would have been more than 100 MEUR per year. The establishment of the state-of-the-art surface preparation technique in industry within the European XFEL project might be a big boost for accepting that surface preparation of SRF cavities could and should be

done in industry in the future. Of course, industry is also interested in the assembly of full-performing accelerator modules on a large scale. 6.2. Future perspectives 6.2.1. Funded new accelerator projects For the next 5–7 years, the SRF accelerator market will be dominated by the construction of the European XFEL (800 cavities), the FRIB project at MSU (340 cavities) and the ESS project at Lund (240 cavities). All these projects are already funded. South Korea is planning a facility for nuclear physics research [33] using an SRF linac of similar size to FRIB. Approval of the project might come next year. In China, the construction of a new Spallation Neutron Source using a superconducting linac (100 cavities) is planned. Final approval of the project is expected within the next five years. 6.2.2. Energy recovery linacs While the demand for cavities has been relatively stable in the last few decades, a new kind of accelerator is currently being developed which could increase the market volume. Several laboratories (Cornell [34, 35] (Fig. 19), BNL [36], KEK [37]) are at the final design stage of an energy recovery linac using SRF technology. Such an ERL could be the drive source of a much-improved light source. Each ERL would require probably 200 SRF cavities and, if proven to be a reliable technology, at least one new ERL project might be funded every year. 6.2.3. SRF technology for energy application (ADS) There are more projects planned that might also have an impact on the future industrialization of SRF technology. In China, Japan, India and Europe (Myhrra, Belgium [38]), SRF linacs are being designed which should serve as a driver for a power plant converting long-living isotopes from nuclear waste into shorter-living isotopes. The process will lead to generation of energy and can take place only when the beam hits the target, making this technique fail-safe. The linac of such an “accelerator-driven system” (ADS) would require about 150 SRF cavities.

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology

Fig. 19.

281

A prototype ERL SRF injector module under testing at Cornell University (courtesy of Cornell).

It could be the first application for SRF technology outside the scientific market. 6.2.4. International linear collider As stated above, the next linear collider [39] would be based on SRF cavities and might be built within the next 10 years. It would require some 18,000 1.3 GHz nine-cell cavities, which probably would be produced equally in the three involved regions: Asia, the Americas and Europe [40]. The cavities will be assembled into 1750 cryomodules. The project would establish additional manufacturing capabilities for SRF technology, and industry would then be ready for application, hopefully also outside the scientific market. 6.3. Commercialization So far no commercial application for an SRF module has been identified, allowing a long term and substantial investment for the complete development and full testing of a mature SRF accelerator module by industry alone without a funded project. 7. Conclusion Based on the successful development and prototyping of superconducting RF accelerator technology by the accelerator laboratories worldwide, qualified industry could prove to be a reliable and mandatory

partner for the delivery of complete superconducting cavities and modules for all such small and large scale accelerator projects in the scientific market within the last three decades. With some 2500 cavities produced by industry so far at a relatively constant rate of some 100 pieces per annum, the demand from the scientific market is still small. This may change if the ILC with a demand for some 18,000 cavities is decided to be built within the next 10 years, or with the use of this challenging technology in the commercial market. Acknowledgments All the three authors worked as scientists in the field of SRF technology for a relatively long time before they moved to industry more than two decades ago (Hanspeter Vogel and Michael Peiniger) and about 15 years ago (Michael Pekeler). They are deeply grateful for the faithful and fruitful partnership with the people of all the worldwide leading laboratories and universities in this technology from that time until today. Special thanks go to Professor Emeritus Helmut Piel of the University of Wuppertal for inspiring Hanspeter Vogel and Michael Peiniger to bring SRF to industry, and to Professor Emeritus Peter Schm¨ user of the University of Hamburg/DESY and Professor Emeritus Hasan Padamsee of Cornell University for teaching Michael Pekeler SRF during the early days of the TESLA collaboration.

December 29, 2012

17:32

282

WSPC/253-RAST : SPI-J100

1230010

M. Peiniger, M. Pekeler & H. Vogel

References [1] H. Padamse, J. Knobloch and T. Hays, RF Superconductivity Accelerators, Wiley Series in Beam Physics and Accelerator Technology (1998). [2] Y. Yamamoto et al., Recent results of 1.3 GHz 9-cell superconducting cavities in KEK-STF, in Proc. LINAC10 (Tsukuba, Japan, 2010). [3] A. Richter, Operational experience at the SDALINAC, in Proc. EPAC96 (Sitges, Barcelona, Spain, 1996). [4] P. Kneisel et al., RF superconductivity at CEBAF, in Proc. SRF91 (DESY, Hamburg, Germany, 1991). [5] F. Marhauser et al., Fabrication and testing of CEBAF 12 GeV upgrade cavities, in Proc. IPAC11 (San Sebasti´ an, Spain, 2011). [6] E. Chiaveri, Large scale industrial production of superconducting cavities, in Proc. EPAC96 (Sitges, Barcelona, Spain, 1996). [7] S. Mitsunobu et al., Superconducting actvities at KEK, in Proc. SRF91y (DESY, Hamburg, Germany, 1991). [8] D. Proch et al., Laboratory Report DESY, in Proc. SRF93 (Newport VA, USA, 1993). [9] T. Furuya, private communications (KEK). [10] J. Kirchgessner, The use of superconducting RF or high current applications, Part. Accel. 46, 151–162 (1994). [11] D. Dasbach et al., Industrial production of superconducting accelerators, in Proc. PAC 93 (Washington DC, USA, 1993). [12] B. Aune et al., Superconducting TESLA cavities, Phys. Rev. ST Accel. Beams 3, 092001 (2000). [13] H. Weise, The European XFEL based on superconducting technologies, in Proc. SRF09 (Berlin, Germany, 2009). [14] D. Boussard et al., The LHC superconducting cavities, in Proc. PAC99 (New York, NY, USA, 1999). [15] M. Pekeler et al., Fabrication of superconducting cavites for SNS, in Proc. LINAC04 (L¨ ubeck, Germany, 2004). [16] R. E. Laxdal et al., Recent progress in the superconducting RF program at TRIUMF/ISAC, in Proc. SRF07 (Ithaca, NY, USA, 2007). [17] I. Mardor et al., The SARAF CW 40 MeV proton/deuteron accelerator, in Proc. SRF09 (Berlin, Germany, 2009). [18] R. Ferdinand and P. Bertrand, Status and challenges of the spiral 2 facility, in Proc. LINAC10 (Tsukuba, Japan, 2010).

[19] R. York et al., Status and plans for the facility for rare isotope beams at Michigan State University, in Proc. LINAC10 (Tsukuba, Japan, 2010). [20] M. Leitner, private communication (MSU). [21] M. Pekeler et al., Production of superconducting accelerator modules for high current electron storage rings, in Proc. EPAC04 (Lucerne, Switzerland, 2004). [22] C. Wang et al., Operational experiences of the superconducting RF module at TLS, in Proc. SRF05 (Ithaca, NY, USA, 1955). [23] R. Tanner et al., Canadian light source storage ring RF system, in Proc. SRF05 (Ithaca, NY, USA, 2005). [24] M. Jensen et al., Operational experience of the DIAMOND SCRF system, in Proc. SRF07 (Beijing, China, WEP73, 2007). [25] J. F. Liu et al., Operation of SRF in the storage ring of SSRF, in Proc. IPAC10 (Kyoto, Japan, 2010). [26] G. Wang et al., The commissioning of BEPC-II RF system, in Proc. SRF07 (Beijing, China, 2007). [27] P. Marchand, Superconducting RF cavities for synchrotron light sources, in Proc. EPAC04 (Lucerne, Switzerland, 2004). [28] P. Marchand, private communication (SOLEIL). [29] A. B¨ uchner et al., The ELBE project at DresdenRossendorf, in Proc. EPAC2000 (Vienna, Austria, 2000). [30] P. von Stein et al., Fabrication and installation of superconducting accelerator modules for the ERL prototype (ERLP) at Daresbury, in Proc. EPAC06 (Edinburgh, Scotland, 2006). [31] I. Ben-Zvi, private communication (BNL). [32] S. Belomestnykh, private communication (BNL). [33] S.-K. Ko, Conceptual design of Korea rare isotope accelerator, in Proc. 14th Int. Conf. Accelerator and Beam Utilization (2010). [34] C. Mayes et al., Cornell ERL Research and Development, in Proc. PAC11 (New York, NY, USA, 2011). [35] G. Hoffst¨ atter, private communication (Cornell). [36] I. Ben-Zvi, ERL prototype at BNL, in Proc. SRF09 (Berlin, Germany, 2009). [37] S. Sakanaka et al., Status of ERL and cERL projects in Japan, in Proc. LINAC10 (Tsukuba, Japan, 2010). [38] J.-L. Biarrotte, High power hadron accelerator of nuclear energy, Fusion Sci. Tech. 61, 15–20 (2012). [39] www.linearcollider.org [40] A. Yamamoto, private communication (KEK).

December 29, 2012

17:32

WSPC/253-RAST : SPI-J100

1230010

Industrialization of Superconducting RF Accelerator Technology

Michael Peiniger received his Ph.D. in Physics from the University of Wuppertal in 1989. In 1987 he joined the Siemens daughter company Interatom GmbH, for building up a new business of RF accelerators and special manufacturing projects for the worldwide research labs and advanced technology industry. In 1994 he cofounded ACCEL Instruments GmbH and bought out the above business, and since then he has been developing this activity as an entrepreneur and managing director. In 2007 ACCEL Instruments was sold to Varian Medical Systems, Inc. Since 2009 the business of RF accelerators and special manufacturing projects has been run by RI Research Instruments GmbH, which is majorityowned by Bruker Energy and Supercon Technologies Inc., with Dr. Peiniger and other management holding equity stakes. Michael Pekeler received his Ph.D. in Physics at the University of Hamburg (DESY) in 1996. During his thesis work he was a visiting scientist at Cornell University for nine months and studied the field-limiting mechanisms on superconducting RF (SRF) cavities within the framework of the TESLA collaboration. He was a postdoc at DESY, working on SRF R&D and commissioning of the first cavities and modules for the TESLA Test Facility, now called FLASH. In 1999 he joined ACCEL Instruments GmbH (now RI Research Instruments GmbH) as Project Manager and was responsible for the technology transfer and industrialization of Cornell/CESR SRF cavity and module technology. Today Dr. Pekeler is Director and head of business SRF, cryogenics and special manufacturing projects at RI.

283

Hanspeter Vogel received his Diploma in Physics from the University of Wuppertal in 1984. From 1984 to 1986 he worked in the superconducting RF (SRF) group at DESY, Hamburg. After that he joined the Siemens daughter company Interatom GmbH in 1986, contributing to building up a new business of RF accelerators and special manufacturing projects for the worldwide research labs and advanced technology industry. In 1996 he joined ACCEL Instruments GmbH, and he took responsibility as director in 2003. In 2007 ACCEL Instruments was sold to Varian Medical Systems, Inc. Since 2009 the business of RF accelerators and special manufacturing projects has been run by RI Research Instruments GmbH, which is majority-owned by Bruker Energy and Supercon Technologies Inc., with Vogel as Managing Director since 2012.

January 11, 2013

15:23

WSPC/253-RAST : SPI-J100

This page intentionally left blank

1203001˙book

January 10, 2013

13:17

WSPC/253-RAST : SPI-J100

1230011

Reviews of Accelerator Science and Technology Vol. 5 (2012) 285–312 c World Scientific Publishing Company
DOI: 10.1142/S1793626812300113

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications Charles E. Reece∗ and Gianluigi Ciovati† SRF R&D Department, Thomas Jefferson National Accelerator Facility, 12000 Jefferson Avenue, Newport News, Virginia 23606, USA ∗[email protected] †[email protected] Superconducting rf (SRF) technology is evolving rapidly, as are its applications. While there is active exploitation of what one may call the current state-of-the-practice, there is also rapid progress in expanding in several dimensions the accessible and useful parameter space. While state-of-the-art performance sometimes outpaces thorough understanding, the improving scientific understanding from active SRF research is clarifying routes to obtain optimum performance from present materials and opening avenues beyond the standard bulk niobium. The improving technical basis understanding is enabling process engineering to improve both performance confidence and reliability and also unit implementation costs. Increasing confidence in the technology enables the engineering of new creative application designs. We attempt to survey this landscape to highlight the potential for future accelerator applications. Keywords: Superconducting rf; niobium; rf structures; superconducting surface treatment; accelerator applications.

1. Introduction

accelerators, questions regarding optimal niobium material characteristics, performance-limiting mechanisms, chemical and mechanical forming and treatment processes, and other material systems that may someday surpass standard bulk niobium with performance and/or cost-effectiveness. Finally, we briefly review candidate future applications of SRF technology that are at various stages of conception.

The press for ever more powerful, economical, efficient and versatile particle accelerators continues to motivate pursuit of innovative solutions involving superconducting materials, particularly those which can support high electromagnetic fields within rf accelerating structures. The superconducting rf (SRF) technology which fills this need now offers robust solutions within a particular parameter space and offers prospects for yet more significantly improved cost per unit energy gain. Such gains make credible a variety of new types of accelerator application solutions. Other contributions to this volume address current operating systems and rf structure design elements in detail. Here, we provide a high-level characterization of the technological domain that is confidently deployable, as demonstrated by operating systems. We follow this with a summary of the best-demonstrated performance of serviceable SRF devices, although such performance may not yet be attained with desirable confidence or affordable cost. Next, we review current research that seeks to answer key questions that stand in the way of maximal exploitation of SRF for particle

2. Summary of the “State-of-the-Practice” for SRF-Based Accelerator Technology The current “state-of-the-practice” of SRF technology applied to accelerators is represented by those performance levels established as specifications for funded and launched projects and this on the foundation of long-running installations. The European–XFEL project is fully underway. The performance levels required of this facility have been well demonstrated. The 928 cavities of the full design will operate at a 23.6 MV/m accelerating gradient in pulsed mode, 1.4 ms with a 10 Hz repetition rate, with an anticipated Q0 of 1010 at 2.0 K [1, 2]. The largest current implementation of SRF technology is the CEBAF recirculating electron linac at Jefferson Lab, with over 350 operational cavities. 285

January 10, 2013

13:17

286

WSPC/253-RAST : SPI-J100

1230011

C. E. Reece & G. Ciovati

This facility is presently installing a major upgrade to 12 GeV, based on the addition of 10 new highperformance cryomodules [3, 4]. During the final week of the physics run in the 6 GeV era, one of the new upgrade cryomodules demonstrated full design requirements of 460 µA CW, providing 108 MeV gain per pass [5]. The eight seven-cell cavities provided the project-specified average 19.2 MV/m with a Q0 > 7 × 109 at 2.07 K. The ISAC-II accelerator at TRIUMF is operational with 40 quarter-wave resonators (QWRs) of three types. The ISAC-II performance goal is to operate CW at a gradient of 6 MV/m, corresponding to a peak surface field of 30 MV/m and a peak magnetic field of 60 mT with a cavity dissipated power ≤7 W [6]; while the heavy ion accelerator ALPI at INFNLegnaro sustains operation of 64 QWRs running at 3–6 MV/m using a mix of SRF material technologies [7]. The SPIRAL 2 accelerator at GANIL has twelve β = 0.07 and fourteen β = 0.12 88 MHz QWRs operating at 4.2 K. The cavities operate in CW at a gradient of 6.5 MV/m with less than 10 W dissipated power per cavity. During vertical tests, the cavities achieved peak surface electric fields up to ∼56 MV/m and peak surface magnetic fields up to ∼100 mT with Q-values of ∼ 1 × 109 at 4.2 K. The sixteen 400 MHz LHC SRF cavities are running now, each capable of providing up to 2 MV for beam operations with 5.5 MV/m, having been tested to 8 MV/m [8, 9]. SRF cavities are serving dutifully in electron storage ring applications around the world. In such high beam loading applications, the chief challenges tend to be input power couplers and convenient damping of higher-order-mode (HOM) power, in contrast to the SRF cavity gradient and Q [10]. The ATLAS Energy Upgrade cryomodule containing seven β = 0.15 QWR cavities resonating at 109 MHz is operating with beam, providing 14.5 MV of accelerating voltage in 4.5 m. This is a record for this beta range and represents a factor-of-3 performance gain over prior ATLAS technology [11]. The seven cryomodules of FLASH at DESY are operating reliably at 20–25 MV/m with 1.4 ms pulses, 10 Hz rep rate [12]. A third-harmonic cryomodule for phase space linearization was contributed by FNAL to FLASH. It contains four 3.9 GHz SRF cavities produced

by US members of the Tesla Technology Collaboration (TTC). The performance of these cavities at DESY comfortably exceeds the design gradient of 14 MV/m, with each being clean of fieldemission-induced X-rays to 20 MV/m and quench fields greater than 24 MV/m [13]. The Spallation Neutron Source (SNS) has seven years of commissioning and operation experience and is now delivering 1 MW of beam power on target with 80 SRF 805 MHz cavities. This was the world’s first superconducting linac for pulsed proton beams. Although the design gradients were 10.2 MV/m and 15.9 MV/m for the β = 0.61 and β = 0.81 cavities, respectively, the operational gradient distributions are essentially the same for the two sets: ∼10– 15 MV/m, with the dominant performance limitation being field-emission-derived heating effects [14]. 3. SRF State-of-the-Art in 2012 3.1. Best β = 1 structure performances The technologically most demanding SRF challenge recently has been the R&D push to make an International Linear Collider (ILC) technically viable and credibly affordable. Such a facility would dwarf all other SRF applications in scale. The push to make 35 MV/m confidently attainable has met with significant success. The challenge is principally one of quality control in materials and processes, and secondarily one of cost optimization by process and material optimization. The international R&D collaboration has yielded several multicell cavities with performance exceeding 40 MV/m, and yield of 90% with performance >35 MV/m has been demonstrated by at least one laboratory [15–17]. The present best nine-cell gradient performance reported has been from two cavities fabricated from ingot Nb by Research Instruments GmbH. These cavities were quench-limited in the 25–29 MV/m range after standard acid etching treatment, but quench fields increased to 31–45 MV/m after subsequent electropolishing, which corresponds to peak magnetic surface fields of up to 192 mT [18]. One of the CEBAF upgrade prototype sevencell cavities, LL002, a fine-grain, high RRR Nb with a history of >250 µm etch removal, reached 43.5 MV/m in a 2010 test, limited by available rf power, after a 34 µm electropolish. For this “lowloss” cell shape design, this corresponds to peak

January 10, 2013

13:17

WSPC/253-RAST : SPI-J100

1230011

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications

Fig. 1.

287

Best CEBAF upgrade cavity performance test.

magnetic surface fields of 162 mT. One of the few 12 GeV upgrade production cavities which was actually tested to its limits, C100-RI-006 quenched at 41.6 MV/m (Fig. 1). 3.2. Best performance from β < 1 structures Investments in careful process analysis and engineering are also yielding performance progress in Nb cavities with the more complicated geometries required for β < 1 applications. ANL recently set a new record for an accelerating gradient with a 72 MHz β = 0.077 cavity: Eacc = 12 MV/m and Q0 of 6 × 109 [19]. Similarly, the revised designs of two 80.5 MHz QWRs and two 322 MHz half-wave resonators (HWRs) required for FRIB have recently been tested and exceeded all requirements. The residual resistance measured in the three prototype families was below 5 nΩ up to about 100 mT in QWRs, and about 80 mT in HWRs [20]. FNAL is developing a β = 0.22 325 MHz singlespoke resonator (SSR1) for a possible Project X front end. A recent test demonstrated the project requirement of Eacc = 12 MV/m and Q0 of 5 × 109 [21]. 4. Current SRF Research 4.1. Bulk Nb material R&D The technical specifications for bulk Nb material commonly used for the fabrication of SRF cavities are listed in Table 1 [22]. Parameters other than RRR and impurity content are important for good formability of the material. The RRR value has been

Table 1. Specifications for Nb bulk material used for the fabrication of SRF cavities. RRR Grain size (µm) Yield strength, σ0.2 (MPa) Tensile strength Elongation at fracture Vickers hardness Content of main impurities (wt. ppm)

>300 ≈50 50 < σ0.2 < 100 >100 30% ≤50 Ta ≤ 500; O ≤ 10; N ≤ 10; C ≤ 10; H ≤ 2

specified in order to assure good thermal conductivity, κ, of the material, as κ at 4.2 K is directly proportional to RRR. High thermal conductivity allows thermal stabilization of defects. A simplified model shows that the quench field, Hquench, due to a normal-conducting defect of surface resistance Rn and radius a, is proportional to the square root of κ [23]:
κ Hquench ∝ . (1) aRn Specifications for the content of the main impurities in Nb were dictated by the need to assure a high RRR value and to limit the concentration of impurities such as hydrogen and oxygen, which have a well-known negative impact on the superconducting properties of Nb [24, 25]. The understanding of various factors which contribute to Nb RRR continues to mature [26]. Few companies throughout the world can provide the Nb material with the specifications listed in Table 1, and that at a cost which is significantly higher than that of lower-purity Nb and which

January 10, 2013

13:17

288

WSPC/253-RAST : SPI-J100

1230011

C. E. Reece & G. Ciovati

has increased rather sharply in the past five years. Elaborate manufacturing steps which include forging, grinding, rolling and annealing contribute to the cost of fine-grain Nb sheets for SRF cavity applications. In addition, each of these manufacturing steps can potentially introduce foreign inclusions into the material, which can subsequently result in a reduced cavity quench field. Quality control methods, such as the use of eddy current scanning systems to detect inclusions of size of the order of 50–100 µm, have been employed by some laboratories to inspect Nb sheets received from the manufacturer [22]. In addition to cost, the availability of low-Ta ingot and the sheet niobium production throughput by experienced industrial vendors impose bottlenecks for producing large quantities of niobium sheets envisioned in large projects. In the following section we will discuss a new type of material, referred to as “ingot Nb,” which has proven to be a viable alternative to the standard fine-grain Nb. 4.1.1. Ingot Nb In 2005, a collaboration between CBMM, Brazil, and Jefferson Lab [27] demonstrated that the mechanical properties of niobium directly sliced from an ingot were adequate for forming into cavity half-cells and that cavities built from such material achieved performance levels comparable to those of cavities built from standard fine-grain Nb [28, 29]. Nb disks are typically sliced from an ingot to the required thickness by wire electrodischarge machining and have a nonuniform crystal structure, with grains of typical size ranging from a few to several centimeters. Studies at DESY proved that the cavity cell shape could be kept within specified tolerances by proper tooling during the forming and machining of the half-cells, in spite of the anisotropic mechanical properties of the disks [30]. Sharp steps at grain boundaries, which also result from nonuniform mechanical properties, can be easily eliminated by local grinding or centrifugal barrel polishing. (See Subsec. 4.3.1.) The ingot Nb technology for SRF cavity production was fully proven at DESY, where two nine-cell cavities built from this type of material have been installed in a cryomodule in the FLASH accelerator [30]. In addition, DESY qualified eight additional nine-cell cavities, which will be installed in the first

cryomodule for the XFEL project. Several of these cavities reached accelerating gradients greater than 40 MV/m, with one of them establishing a recordsetting gradient for these nine-cell TESLA-style cavities of 45 MV/m [18]. Studies at Jefferson Lab and DESY also indicated that 20–30% lower surface resistance was obtained in multicell cavities built from ingot Nb, compared to cavities built from fine-grain Nb, for the same treatment procedure, cavity frequency and operating temperature [29, 31, 32]. A reduction of the surface resistance by a factor of ∼4 was recently obtained at Jefferson Lab with an ingot Nb single-cell cavity heat treated at high-temperature in a vacuum furnace [33]. Investigation into the specific material characteristics which produce this nice result is underway. The obvious reduction in the cost of fabrication of ingot Nb disks, along with the reduced need for their quality control compared to fine-grain Nb sheets, makes them a cost-effective solution for the fabrication of SRF cavities. Many companies worldwide are able to offer ingot Nb material, and the introduction of multiwire slicing techniques makes the mass production of ingot Nb disks possible. The above-mentioned evidence for better performance of ingot Nb cavities compared to fine-grain cavities, which has emerged in the last few years, adds to the attractiveness of ingot Nb material for SRF cavity fabrication. Current R&D topics related to ingot Nb material include measurements of superconducting, thermal and mechanical properties on samples. Measurements of dc critical fields and magnetization done at Jefferson Lab indicated a lower pinning current density for ingot Nb samples than fine-grain ones [34]. This result, together with the reduced flux trapping efficiency of ingot Nb compared to fine-grain Nb, as it was measured on samples at BESSY [35], might explain the lower rf residual resistance of ingot Nb cavities. Studies on the mechanical properties of ingot Nb are discussed in Subsec. 4.1.4. Low-temperature thermal conductivity measurements done on ingot Nb samples at several laboratories showed that, unlike for fine-grain Nb where κ is limited at ∼2 K by phonon scattering at grain boundaries, κ has a pronounced “phonon peak” at ∼2 K [36–38]. Although the phonon peak is reduced by plastic deformation of

13:17

WSPC/253-RAST : SPI-J100

1230011

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications

TTF, 800°C+BCP

Eacc, MV/m

the samples, it can be restored by a subsequent hightemperature heat treatment [36]. This result would imply better thermal stabilization of the cavity inner surface, which results in a reduced dependence of the surface resistance with an increasing rf field, which is often observed for such heat-treated ingot Nb cavities at 2 K. 4.1.2. Single-crystal Nb Ingot Nb disks with a large (∼20 cm)-diameter central single-crystal allow the fabrication of singlecrystal cavities. Two ∼2.3 GHz single-cell prototypes were built and tested at Jefferson Lab and both reached a peak surface magnetic quench field of ∼160 mT at 2.0 K [39]. A method based on rolling and annealing steps was developed at DESY to enlarge a single-crystal disk to a diameter of ∼23 cm, suitable for 1.3 GHz cavity fabrication, while maintaining a single-crystal structure [40]. It was also shown that, by properly matching the orientation of two half-cells, the singlecrystal structure was preserved after electron-beamwelding the half-cells. RF tests of 1.3 GHz single-cell prototypes at 2.0 K reached quench field values of ∼160 mT [41]. The main attraction of single-crystal cavities over typical ingot Nb is the uniformity of mechanical properties and therefore better forming. So far, there seems to be no significant cavity performance improvement compared to cavities made of “standard” ingot Nb, with few-cm-sized grains. Efforts to produce large-diameter single-crystal Nb ingots have been pursued by KEK in collaboration with Tokyo Denkai in Japan [42].

35 30 25 20 15 10 5 0 150

LG, 800°C+BCP

350

450

550

Fig. 2. Maximum accelerating gradient as a function of RRR for 1.3 GHz nine-cell cavities made of fine-grain (blue diamonds) and ingot Nb (red squares), heat-treated at 800◦ C and etched by BCP. (Figure taken from Ref. 30.)

was found at KEK, for RRR values ranging between 130 and 800 [44]. Additional evidence for the reduced impact of high RRR on the quench field was provided by ingot Nb cavities: as shown in Fig. 2, there is no dependence of the quench field (corresponding to Eacc values > 25 MV/m) on RRR, ranging between 150 and 500, as measured on nine-cell ingot Nb cavities treated by BCP [30]. These findings suggest that normal-conducting inclusions of size of the order of tens of micrometers are not anymore the dominant cause of quench in SRF cavities. In the case of ingot Nb material, the possible presence of a phonon peak at ∼2 K breaks the connection between thermal conductivity and RRR. As is shown in Fig. 3, for example, single-crystal samples Heraeus Large Grain RRR477 Heraeus Single Crystal RRR438 Fine Grain RRR1200 Wah Chang Fine Grain RRR500

4.1.3. Niobium purity and tantalum content As mentioned in Subsec. 4.1, the requirement of high-purity (RRR > 300) Nb was dictated by the need for high thermal conductivity to avoid premature quenches. A correlation between higher κ and increased quench field was indeed noticeable when looking at a data set of nine-cell cavities measured at DESY and made of fine-grain Nb, treated by BCP. Nevertheless, when including data from nine-cell cavities made of fine-grain Nb and treated by electropolishing, the correlation between κ and Hquench is lost, as shown in Fig. 5.61 of Ref. 43, for RRR values ranging from ∼200 to ∼750. The same conclusion

250

289

RRR

1000

κ (W/mK)

January 10, 2013

100

10

1 1

T (K)

10

Fig. 3. Thermal conductivity as a function of temperature of fine-grain and ingot Nb (large-grain or single-crystal) samples. (Figure adapted from Ref. 36.)

January 10, 2013

13:17

290

WSPC/253-RAST : SPI-J100

1230011

C. E. Reece & G. Ciovati

of lower RRR values have higher κ (2 K) than finegrain samples of higher RRR. As mentioned before, high-temperature heat treatments provide a method for enhancing κ (2 K) after forming. These recent findings suggest that, for cavities with an Eacc specification below ∼30 MV/m, Nb material with RRR greater than ∼150 is acceptable. Additional accumulated experience with cavities built from this “medium-purity” material could further strengthen this statement. Tantalum is a “substitutional” impurity in Nb and is therefore difficult to separate from the Nb itself. In 1996 it was found that a premature quench in a cavity was due to a specific Ta cluster [45]. In order to reduce the possibility of such “clusters” being present in the Nb sheets, the concentration of Ta in Nb was specified to be ≤500 wt. ppm. This specification imposes a significant increase in the cost of the Nb sheets and precludes the use of Nb extracted from pyrochlore ore, which is a significant fraction of the sources of Nb production. The mechanism for creating such clusters and their frequency of occurrence are unclear. Studies on single-cell cavities and samples have been done in recent years to re-evaluate the role of Ta. Magnetization measurements show no significant difference among ingot Nb samples with Ta content between 600–1300 wt. ppm [31, 33, 36, 46]. Tests of single-cell cavities made of fine-grain Nb showed no correlation between Ta content, 600–1300 wt. ppm, and maximum Eacc up to ∼30 MV/m [47]. Peak surface magnetic field quench values of ∼100 mT have been obtained in multicell cavities made of ingot Nb with Ta content up to 1500 wt. ppm [30]. These results suggest that the Ta content specification can also be relaxed without significant loss in cavity performance. If the presence of Ta clusters is feared, eddy current scanning can be used to exclude Nb disks with such defects from the cavity fabrication process [24]. The potential reduction in the cost of the material which can be obtained by relaxing the RRR and Ta content specifications may be very attractive, particularly for small accelerators with “moderate” gradient specifications for research at universities or for industrial applications. These results suggest that the Ta content specification can also be relaxed for cavities requiring only medium gradients of less than 30 MV/m.

4.1.4. Mechanical properties Renewed efforts to better understand the mechanical properties of both fine-grain and ingot Nb and their implications for cavity performance and formability have been undertaken by several laboratories and universities. One of the final steps in the production of finegrain Nb sheets is a 2% thickness reduction by rolling. This process causes higher strain to be localized near the surface, smaller grain size and different texture at the surface, compared to deeper in the bulk. Orientation image maps (OIMs) obtained by electron backscattered diffraction (EBSD) showed cross-sectional microstructure and texture gradients to vary greatly among samples cut from fine-grain Nb sheets [48]. This results in variability in forming half-cells and in the metallurgical state after hightemperature heat treatments or electron beam welding (EBW). The yield strength of fine-grain niobium decreases by up to ∼40% after heat treatment up to ∼1200◦C. It also depends on purity (lower-purity Nb has higher yield strength) and grain size [49]. For example, as-received ingot Nb samples showed yield strength values ranging between 30 MPa and 50 MPa [45, 50], compared to ∼55–80 MPa for asreceived high-purity fine-grain Nb [51]. Issues with batches of fine-grain Nb sheets not fully recrystallized (nonuniform grain size throughout the thickness) have occurred in the past and resulted in unacceptably low yield strength of the material after heat treatment at 800◦ C [52]. As a result, the heat treatment temperature for hydrogen degassing for SNS and CEBAF upgrade cavities at Jefferson Lab was 600◦ C. The 600◦ C recipe was also used for ILC R&D cavities initially at JLab; due to excessive stiffness issues, an 800◦ C recipe replaced the 600◦ C recipe [53]. An 800◦ C bake recipe has been used for a long time at DESY. In the past few years, there has been increasing evidence for the metallurgical state of the material influencing not only cavity formability and robustness but also cavity performance: • A correlation between high dislocation content and enhanced losses at high (>90 mT) rf field has been reported [54]. • A correlation between etch pits, crystal defects and enhanced RF losses at medium (20 mT < Bp < 90 mT) field has been reported [55].

January 10, 2013

13:17

WSPC/253-RAST : SPI-J100

1230011

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications

• EBW of cold-worked fine-grain Nb can induce pitting in the zone where thermal stress accumulates (the so-called “heat-affected zone”) [56]. It was shown that pits in this zone caused premature quenches in some 1.3 GHz nine-cell cavities [57, 58]. All the above considerations suggest the need for optimization studies to identify the proper sequence of processing steps and heat treatment parameters, for each type of material, which gives the best compromise between mechanical properties and cavity performance. For ingot Nb material, it would be important to identify crystal orientations which do not behave well during forming and should therefore be avoided during the ingot production by the manufacturer. Such study is ongoing at MSU [59, 60]. 4.2. Bulk Nb cavity fabrication The techniques most commonly used for the fabrication of bulk Nb cavities are deep-drawing of Nb sheets and electron beam welding of the formed parts. Electron beam welding of Nb is a critical process which requires machine-specific parameters to be mastered in either an industry or laboratory setting. Occasionally, issues with defects in the welds or in the heat-affected zone limit the cavity performance to Eacc ∼ 15–20 MV/m [30, 54, 55, 61]. Forming processes which could avoid most of the welds in a multicell cavity are hydroforming, which has been pursued mostly at DESY, and spinning, which has been used at INFN-Legnaro. Initial studies on single-cell cavities made by either hydroforming [62] or spinning showed that Eacc ∼ 40 MV/m was achievable [63] — comparable to the best results obtained from cavities made by standard deep-drawing and welding. The hydroforming machine at DESY allows the fabrication of up to three-cell 1.3 GHz cavities, and three full nine-cell prototypes were built by welding three-cell units. The nine-cell cavities were treated by EP and achieved Eacc ∼ 30–35 MV/m. A 1.3 GHz nine-cell cavity was also built by spinning [63], but a hole was made in one of the irises after centrifugal barrel polishing (CBP) because of the nonuniform thickness of the material. Both the hydroforming and the spinning of cavities require a seamless Nb tube of the appropriate

291

diameter and uniform fine-grain structure, usually made “in-house” by deep-drawing, flow-forming, extrusion or a combination of these techniques. The use of “single-crystal” seamless Nb tubes has also been proposed recently [64]. In addition, the hydroforming technique was applied to fabricate cavities of Nb/Cu-clad material. These could have the advantages of reduced cost (because of less Nb) and fewer welds, and improved thermal conductivity and stiffness from the backing Cu. Single-cell cavities have been built at both DESY and KEK, reaching Eacc values of up to 40 MV/m [58]. Temperature gradients across the cavity have to be minimized near Tc , in order to avoid trapping the magnetic field attributed to thermoelectric currents. Reduction of Q0 values after quenches has been measured and attributed to this same effect [58]. The starting Nb/Cu-clad tube had been made by different techniques, such as explosion bonding, back extrusion and hot rolling. There is yet no clear indication of a preferred tube-forming method, and there has not been significant progress toward fabrication of cavities longer than two cells in the past few years. 4.3. Surface finishing — for quality control and economy A very thin layer of material in the cavity surface determines the response to rf fields. The superconducting penetration depth for Nb is only ∼40 nm. Efforts to obtain optimum performance for accelerator applications, then, focus very strongly on controlling the morphology and structure of that thin layer of material. 4.3.1. Centrifugal barrel polishing Initially started as an effort to obtain a standard surface finish and reduce the required quantities of hazardous processing chemicals, the use of centrifugal barrel polishing (CBP) on Nb SRF cavities was reported by Higuchi, Saito et al. at KEK in 1995 [65]. The first batch of single-cell cavities to attain consistent ≥40 MV/m performance was prepared by CBP, followed by light EP [66, 67], in 2006. Recently the CBP treatment process has been picked up and extended at FNAL and JLab in an effort to produce consistently smooth surfaces, somewhat forgiving of the details of the prior

January 10, 2013

13:17

292

WSPC/253-RAST : SPI-J100

1230011

C. E. Reece & G. Ciovati

fabrication processes [68, 69]. The goal of this work is to find a reliable “low-tech” and inexpensive process that transforms the high-field surfaces into a smooth, uniform surface that subsequently requires only a minimum of chemical processing to leave a “crystallographically clean” surface exposed. While further optimization is underway, examples have demonstrated that otherwise flawed and otherwise unacceptable cavities can be brought to peak performance via inclusion of CBP in their treatment [70]. 4.3.2. Acid etching of Nb — BCP Long the standard way of exposing fresh material in bulk Nb cavities, etching with HF:HNO3 :H3 PO4 reagent acids mixed in a ratio of either 1:1:1 or 1:1:2 is commonly referred to as buffered chemical polishing (BCP). The exothermic chemical reaction requires thermal management to obtain uniform results. Nb cavities are typically etched with 15◦ C circulating acid. The details of the technique for realizing predictable etching removal rates over increasingly complicated geometrical shapes continue to be explored [71]. As the target operational surface fields of SRF structures have been pushed up in recent years, the natural residual surface roughness of fine-grain Nb after BCP treatment is found to induce limiting effects, apparently due to local field enhancements that subsequently lead to nonlinear losses. Such roughness is attributed to variation in the local etch rate of different exposed crystal planes and to residual stresses from dislocations and other defects [55]. The roughness and the associated nonlinear losses are greatly reduced in heat-treated ingot Nb surfaces. 4.3.3. Standard electropolishing of Nb The highest field performance of Nb SRF cavities so far is achieved by those with a surface smoothened by electropolishig, in contrast to etching. While this has now been unambiguously demonstrated for finegrain Nb, it also appears to be the case for ingot Nb, although the incremental benefit is less and conditions of improvement are less well resolved. The standard Nb electropolishing (EP) that has evolved from the Siemens recipe involves HF:H2 SO4 reagent acids in a ratio of ∼1:9 as the electrolyte, with a Nb anode and a high-purity Al cathode.

The typical physical arrangement, mounted first by KEK for the TRISTAN cavities, has the cavity and collinear cathode horizontal, with the electrolyte filling approximately 60% of the cavity [72]. Recent involvement of electrochemical insights and methods in Nb EP research has yielded a clarified understanding of the polishing mechanism and in turn provided pointers for process optimization [73–75]. Identification of the importance of managing process temperature control for both uniform removal and optimum surface leveling, together with recognition of the source and methods for reducing and removing precipitating sulfur, have borne fruit in increased cavity performance for the ILC R&D effort [17], the JLab 12 GeV upgrade cavities [76, 77], the ANL QWR cavities [19] and the high-grade cavity set for the XFEL [18], among others. The 80 CEBAF upgrade cavities, for example, were given a consistent 30 µm temperature-controlled EP following a 160 µm BCP etch provided by the cavity manufacturer. The resulting performance was sufficiently reliable that most of the cavities were given their cryogenic acceptance test only after their helium vessel had been welded on, and testing was administratively limited to 27 MV/m [78]. (No more than 23 MV/m will be useful in CEBAF, due to other limitations in the accelerator.) 4.3.4. Vertical electropolishing While Nb EP yields smoother surfaces and in many cases surfaces that can support higher fields than the relatively easy BCP etching process, the commonly used methods are rather complex and labor-intensive compared with what one would like. Because of this, efforts are underway at several labs to develop cavity EP with the cavity filled with electrolytes and oriented with the central axis vertical (VEP). Such an arrangement would have several advantages, not least of which would be its greater prospect for process automation. Cornell has recently reported successful VEP of a nine-cell Tesla-style cavity with field performance exceeding 35 MV/m [79]. CERN and Saclay are investing in new VEP systems, and JLab has repurposed a closed chemistry cabinet for VEP R&D [80, 81]. In addition to temperature regulation, the vertical orientation presents additional challenges to managing the hydrogen bubbles evolved from the cathode and discerning and providing the

January 10, 2013

13:17

WSPC/253-RAST : SPI-J100

1230011

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications

appropriate level of electrolyte circulation for predictable, uniform removal. 4.4. Surface cleaning for field emission control Surface cleanliness is a persistent requirement of SRF accelerating structures. The best-prepared low-loss surface can be made irrelevant by a small amount of contaminating particulates. Such particulates easily become sources of field emission when they find their way to high-surface-electric-field regions of a cavity. Adequately clean Nb surfaces have been demonstrated to be field-emission-free to ≥150 MV/m [82], supporting the notion that field emission from typical technical Nb surfaces may be attributed to either mechanical damage or extrinsic contamination. Residues from chemical processing steps must be removed, else they concentrate when drying. Generous dilution and removal by ultrapure water rinsing in the context of a controlled cleanroom environment is considered necessary. Careful contamination control technique is essential to obtaining and retaining adequately clean high-field surfaces. Highpressure (>100 bar) rinsing with ultrapure water (HPR) is routinely applied as a final preassembly step. The actual protocols for such HPR have little demonstrated basis, however. Techniques at each different laboratory have evolved via loose empiricism rather than quantitative process development. Water pressure, nozzle geometry, flow rate, nozzleto-surface distance and surface sweep rate can all reasonably be expected to affect particulate removal efficiency, so the domain is yet ripe for careful study and improvement [83, 84]. Ethanol or methanol rinse is employed at some labs to provide assurance of removing precipitated sulfur; others, such as Jefferson Lab, rely on ultrasonic agitation with a detergent for this purpose. CO2 snow has been demonstrated to be highly effective at cleaning Nb surfaces [85]. This has not yet, however, been deployed in a way suitable for multicell cavities. 4.5. Thermal treatments It has been clearly established empirically that low-temperature (120–160◦C) baking of etched or electropolished Nb surfaces in ultrahigh-vacuum or atmospheric noble gas [86] for several hours (LTB)

293

significantly reduces the high-field Q-slope, allowing achievement of Bp -values of up to ∼190 mT at 2.0 K [87]. In addition to that, LTB of such cavities causes a reduction of RBCS and often an increase in the residual resistance, Rres , so that a moderate 10–20% increase of Q0 at 2.0 K, at low field, is obtained [88]. LTB protocols vary significantly in duration and hold temperature, with 120◦ C for 24–48 h just prior to cold rf testing being most commonly used. Discussion of the theoretical understanding of the phenomenon is presented in Subsec. 5.1 below. While vacuum baking at 600–800◦C of near-final Nb cavities is routinely applied for the reduction of interstitial hydrogen to avoid hydride precipitation and its attendant “Q-disease,” current research is examining more subtle residual surface composition effects that may be influenced by postchemistry thermal treatments [89]. Early indications from JLab suggest opportunities for reduction of both residual and BCS rf surface resistance with carefully managed surface chemistry during and immediately following ≥1200◦ C baking of ingot Nb cavities [33, 90]. 4.6. Alternative materials to bulk Nb 4.6.1. Motivations for alternatives to bulk Nb With the real prospect of the community being able, in the not-so-distant future, to specify preparation methods which will reliably yield nearly the theoretical best performance in both supported fields and resistive losses of Nb, one naturally asks the following question: How may accelerator systems be fielded yet more efficiently than Nb will allow? Several dimensions merit consideration. If one could realize the same rf characteristics from a thin layer of Nb (less than 1 µm is required) on a less expensive, easily formable and high-thermalconductivity substrate, then significant cost savings may be accessible in materials. Also, one may have access to more convenient cooling configurations employing, for example, channel cooling rather than bath cooling, which in turn opens up new possibilities in the engineering design of accelerator cryomodules. Alternatively, if materials with higher Tc and adequately high critical fields could be made to fulfill their apparent theoretical potential, then huge savings in capital and operational cryogenic costs might be realized. Nb3 Sn, NbTiN and MgB2 are among the materials receiving such attention.

January 10, 2013

13:17

294

WSPC/253-RAST : SPI-J100

1230011

C. E. Reece & G. Ciovati

Further down the road there remains the prospect of hybrid multilayer films which may support dramatically higher fields with very low losses. 4.6.2. Magnetron-sputtered Nb films The largest fielded application of sputtered Nb film was on the 352 MHz LEP2 cavities. For 272 of these cavities, Nb film on copper was chosen to save cost. These cavities fulfilled their mission for LEP2, but were very hard pressed to perform with adequate Q at 7 MV/m [91]. The best-demonstrated performance of Nb film technology cavities dates from the post-LEP development work for LHC cavities at CERN [92, 93]. 1.5 GHz cavities demonstrated 15 MV/m with Q0 of 1010 at 1.7 K. The magnetron-sputtered cavities show a characteristic nonlinear loss. Such a limiting Q-slope persists in the operating LHC cavities today [8]. The RRR of such sputtered Nb films was found to be ∼30. Nb/Cu medium-β QWR cavities have recently been developed for ALPI [94], and development of sputtered Nb/Cu 101 MHz QWRs for HIE-ISOLDE is underway at CERN [95]. 4.6.3. Energetic condensation of Nb When one is considering the gas phase deposition of niobium, it may be helpful to recognize that its very high melting temperature implies very low mobility of atoms near the growth surface at temperatures tolerable to desired application substrates such as copper and aluminum. To realize high crystallinity with acceptably low defect density, additional local energy is required. “Energetic condensation” provides this as kinetic energy of the arriving ions. A collaboration led by JLab has been characterizing Nb film growth on a variety of model substrates as a function of substrate preparation and energy of arriving Nb ions [96–105]. The methods of creating the energetic ions vary from ECR-plasma-biased extraction to cathodic arc discharge plasma, to highpower impulse magnetron sputtering (HiPIMS) [106] championed by Andre Anders of LBNL. Nb films grown by energetic condensation show an epitaxial character with low defect density, as measured by RRR quite comparable to that of highpurity bulk Nb now used for cavity fabrication (200– 400). While attaining the maximum RRR is not itself

the goal of deposited Nb films, learning how highquality Nb crystals develop under controlled conditions on well-known substrates gives insight into tailoring growth conditions for eventual technical applications. Candidate growth conditions have been identified which yield very attractive Nb films grown on fine-grain Cu, such as would make for a fine cavity substrate. Attempts to realize such conditions on test elliptical cavities are planned for the coming year. 4.6.4. Nb3 Sn Nb3 Sn continues to receive low-level attention as a material which might exceed the performance of Nb. With Tc ∼ 18.2 K there is the attractive prospect that high-frequency structures made with Nb3 Sn might have negligible losses at the relatively more convenient operating temperature of 4.5 K. Because Nb3 Sn is very brittle and has low thermal conductivity, it must be formed in shape as a film on a suitable substrate. To obtain good stoichiometry from gas phase supplied Sn diffusion into Nb, a reaction temperature above 930◦C is required. The most thorough work on Nb3 Sn for SRF was carried out at Universit¨ at Wuppertal in the 1980s and 1990s [107]. Recently, work with this material has resumed at Cornell University [108]. Careful material science is yet required to understand and confidently control the Nb3 Sn crystal growth dynamics so as to produce low-loss surfaces to fields corresponding to Eacc > 15 MV/m. 4.6.5. NbTiN Although A15 compounds such as Nb3 Sn have a higher Tc , the Nb B1 compounds are less sensitive to radiation damage and crystalline disorder. NbN and NbTiN are the B1 compounds with the highest critical temperature 17.3 K and 17.8 K, respectively. The ternary nitride NbTiN presents all the advantages of NbN and exhibits increased metallic electrical conduction properties with a higher titanium percentage. Ti is a good nitrogen getter, so the higher the Ti composition, the lower the number of vacancies. For these reasons, work at Jefferson Lab is targeting NbTiN together with the dielectric AlN for development of SRF multilayer structures [109, 110]. Figure 4 shows a TEM image of an FIBcut cross-section of a recent multilayer film formed by dc reactive sputtering.

13:17

WSPC/253-RAST : SPI-J100

1230011

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications

295

Xi et al. of Temple University are developing the growth of MgB2 via hybrid physical–chemical vapor deposition (HPCVD) [118]. Superconductor Technologies Inc., has deposited MgB2 using a reactive evaporation (RE) technique at 550◦ C [119]. The field-dependent rf surface impedance of singlecrystal MgB2 samples has recently been measured at 7.5 GHz [120]. The road to understanding and exploiting this material still lies ahead. 5. Research on Fundamentals 5.1. Nonlinear losses

Fig. 4. TEM image of an FIB-cut cross-section Cu/Nb/ AlN/NbTiN film structure formed by dc reactive sputtering.

4.6.6. MgB2 MgB2 is a binary compound that contains hexagonal boron layers separated by close-packed magnesium layers. It has a high Tc , ∼39–40 K. In SRF applications, a MgB2 film with large grain size (>100 µm) is preferred, because grain boundaries in the film could cause strong pinning which would contribute to a high-field Q-drop [111]. RF penetration depth is 100–200 nm at temperatures below 5 K in the GHz range. So a film with a severalhundred-nm thickness is required. Compared to Nb, MgB2 has a smaller lower critical field Hc1 and a larger upper critical field Hc2 , determined principally by its penetration depth and coherence length, respectively. The superheating critical field of MgB2 , √ calculated from Hsh = 0.75 Hc1 Hc2 [112, 113], is 170–1000 mT depending on the field’s direction [114], which suggests that it might be possible to achieve a 200 MV/m gradient [115]. MgB2 is also an attractive choice for multilayer film coatings to benefit from the lower surface resistance, which was proposed by Gurevich [116]. The SRF properties of MgB2 films formed by two different processes are being characterized [117].

The typical SRF performance of a cavity is expressed in terms of Q0 as a function of the rf field, either Eacc on-axis or peak surface electric (Ep ) or magnetic fields (Bp ). RF tests of bulk Nb cavities at or below 2.0 K show several nonlinearities, referred to as Q-slopes, occurring typically at low (Bp < ∼30 mT), medium (∼30 < Bp < ∼90 mT) and high field (Bp > ∼90 mT), whose origins are not well understood. Figure 5 shows a typical Q0 (Bp ) plot for a bulk Nb cavity. Understanding the origin of these nonlinearities is important not only from a scientific point of view but also to help identify better treatments to sustain high Q0 values at high rf fields. Low-β cavities and elliptical-type cavities tested at 4.2 K commonly show a monotonic decrease in Q0 by about a factor of ∼3–10 up to ∼80–90 mT. This effect is attributed mainly to a thermal feedback in which Rs (Bp ) increases because of the warming of the inner cavity surface with an increasing rf field due to the exponentially increasing BCS surface resistance and the poor heat conductivity of He I [121].

T = 1.7 K

1E+10

Q0

January 10, 2013

before bake 120C/9h bake

Quench

1E+09 0

20

40

60

80

100 120 140 160 180

Bp (mT) Fig. 5. Typical Q0 (Bp ) curves for a 1.5 GHz bulk Nb cavity before and after LTB (adapted from Ref. 122).

January 10, 2013

13:17

296

WSPC/253-RAST : SPI-J100

1230011

C. E. Reece & G. Ciovati

and by low-temperature (120–160◦C) baking in ultrahigh vacuum or atmospheric noble gas [86] for several hours (LTB). In a model discussed in Ref. 125, this effect is related to nonequilibrium superconductivity caused by suboxide precipitates near the Nb surface. An alternative explanation, proposed in Ref. 88, relates it to the detrapping of vortices which are within a depth of the order of the rf penetration depth from the surface, due to the Lorentz force of the rf field. According to this model, the Q0 (Bp ) curve should be hysteretic at low field, with a constant, high Q0 value obtained after reducing the rf power back to low field. An attempt at a systematic investigation of this phenomenon was published in Ref. 126, but further dedicated studies are necessary for a better understanding. Fig. 6. State-of-the-art performance of a 1.5 GHz Nb/Cu thin film cavity at 1.7 K. (Figure taken from Ref. 123.)

Thin film Nb magnetron-sputtered cavities exhibit a monotonic decrease in Q0 by about one order of magnitude up to ∼90 mT at or below 2.0 K, and such strong degradation of Q0 has precluded their application to accelerators which require Eacc > ∼15 MV/m [123]. The Rs (Eacc ) plot for a state-ofthe-art Nb/Cu thin film cavity is shown in Fig. 6. In spite of significant R&D efforts, mainly at CERN, there is no clear explanation for the origin of such losses. Possible hypotheses include suppression of the lower critical field, Hc1 , and of the critical superfluid velocity, because of the low mean free path of the normal electrons in the films, or additional losses due to “weak links” because of the granularity of the films [123], and fine-scale surface roughness. A recent analysis of field-dependent rf losses measured on magnetron-sputtered Nb/Cu thin film samples at CERN [124] showed good agreement with a model based on electric-field-driven losses due to quasiparticle tunneling into localized states in the surface oxide (referred to as the “interface tunnel exchange” model, discussed in Ref. 125 and references therein). 5.1.1. Low-field Q-slope An increase in the Q0 value by up to ∼50% between ∼2 mT and ∼20 mT is sometimes observed. This effect seems to be enhanced at lower temperatures

5.1.2. Medium-field Q-slope A reduction of Q0 by about 20–50% between ∼20 mT and ∼90 mT is typically observed. This phenomenon seems to be enhanced by EP and LTB and at lower temperatures while it is reduced in large-grain cavities. The Rs field dependence is often well described as being quadratic, although a linear term was sometimes included, particularly to fit data after LTB [43]. The linear term was related to losses caused by Josephson fluxons at “strong” superconducting links near the Nb surface, consisting of oxide-filled grain boundaries [127]. The Rs (Bp ) dependence predicted by the thermal feedback model is not as strong as is often measured experimentally, particularly for cavities at gigahertz frequency for which RBCS is relatively low compared to S-band cavities [121]. The convenient availability of large Nb crystals has enabled a more precise characterization of the Kapitza resistance at the niobium–superfluid-helium interface [128]. An intrinsic nonlinearity of RBCS due to the increase in thermally activated quasiparticles by the rf field was calculated in the clean limit in [129]. The model provides a good fit of the experimental data in the gigahertz range and predicts a stronger slope at lower temperatures, as has been measured. Although some common trends in the mediumfield Q-slope for different surface treatments and

January 10, 2013

13:17

WSPC/253-RAST : SPI-J100

1230011

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications

materials have been identified experimentally, a good understanding of which parameters are affected by such treatments/materials, in relation to the medium-field Q-slope, is still missing. 5.1.3. High-field Q-slope A sharp, exponential drop of Q0 starting at ∼90 mT in the absence of field emission characterizes the performance of bulk Nb cavities tested after typical active chemical processing. Temperature mapping of the outer cavity surface during rf tests shows nonuniform losses (“hot spots”) occurring in the high-magnetic-field region of the cavities’ surface. Since this phenomenon was first reported in 1997, many models have been proposed to explain the origin of the anomalous losses. They include defective oxides, high interstitial oxygen near the surface, a reduced field of the first flux penetration due to lattice defects or impurities near the surface, and local quenches at sharp surface features (such as steps at grain boundaries). Extensive reviews of experimental data and models related to the high-field Q-slope have been presented over the years in many publications [87, 130–132]. The most recent model proposes a field-dependent expression for the surface resistance which can describe the Q-slopes in any field region. At the origin of Rs (Bp , T ) are small (compared to the rf penetration depth and coherence length) isolated “defects” which suppress the surface barrier for flux penetration. Rs is factorized in a temperature-dependent term, resulting from the increase in the density of normal-conducting electrons above a certain percolation temperature, and a field-dependent term resulting from the growth of the normal-conducting volume with the applied rf magnetic field [133]. While some of the models provide a good fit of the experimental data in the high-field Q-slope region, none of them has clearly provided a physical explanation for the phenomenon which is widely accepted, although hydrogen vacancy complexes in the near surface have been implicated [134]. It was found empirically that EP plus LTB significantly reduces the high-field Q-slope of fine-grain Nb cavities (ingot Nb may not need EP), allowing attainment of Bp values of up to ∼190 mT at 2.0 K [87]. In addition to that, LTB causes a reduction of RBCS and often an increase in the residual resistance, Rres , so that a moderate 10–20% increase in Q0 at

297

2.0 K, at low field, is obtained [88]. Lower Rres values can be restored by rinsing the cavity with HF (49%) after baking [135]. A well-accepted explanation for the baking effect is also missing. That the beneficial effect of LTB on the high-field Q-slope is found to be dependent on the material (fine-grain or ingot Nb) and treatment combination (EP, BCP or postpurification) [54] suggests that multiple mechanisms are involved in what is generally described as a “highfield Q-slope.” 5.1.4. Nonlinear BCS losses Xiao has recently derived a field-dependent extension of the Mattis–Bardeen theory of a BCS superconductor’s surface impedance [136]. Prior description of the response of a superconducting surface to rf fields derived from BCS theory has been in the lowfield limit. Xiao finds, as a result of the modified density of states distribution with flowing Cooper pairs, that the surface resistance of such a superconductor initially decreases with an increasing field, before increasing at further higher fields. Such behavior has not been previously predicted. An intriguing correspondence of this prediction with the frequently observed “low-field Q-slope” is noted. Analysis will continue. 5.2. RF critical field Theoretically, the amplitude of the maximum rf magnetic field which can be applied at the surface of a type II superconductor before the surface barrier to flux penetration vanishes is the so-called superheating field, Hs . The ratio of Hs divided by the thermodynamic critical field, Hc , has been calculated near Tc using the Ginzburg–Landau (GL) theory in the limits of the small and large GL parameter, κGL [137]. The calculation of Hs at low temperature is quite complex, and it has recently been done in the clean limit [138] and as a function of impurities [139] in the limit of large κGL . The results show that Hs = 0.84Hc

T → 0,

clean limit,

(2)

whereas Hs changes by less than ∼3% in the dirty limit, in the presence of nonmagnetic impurities. Magnetic impurities, on the other hand, significantly suppress Hs . An interesting result of the calculation was that, unlike in the clean limit, the presence of nonmagnetic impurities maintains an energy gap

January 10, 2013

13:17

298

WSPC/253-RAST : SPI-J100

1230011

C. E. Reece & G. Ciovati

in the quasiparticle spectrum near Hs , which would result in reduced Rs nonlinearity. To the authors’ knowledge, there exists no calculation of Hs (T  Tc ) as a function of κGL in either the clean or the dirty limit. The temperature dependence of Hs can be approximated as   2  T . (3) Hs (T ) ≈ Hs (0) 1 − Tc Experimentally, the superheating field has been achieved in SnIn and InBi alloy samples with different κGL values and type I superconductors such as Sn, In and Pb near Tc [140]. However, measurements of Hs in Nb and Nb3 Sn cavities as a function of temperature showed Hs values lower than predicted by the theory for T  Tc . The cause of this discrepancy is not clear [141]. The highest Bp value measured on bulk Nb cavities at 2.0 K is 210 mT [142], greater than Hc (2.0 K) = 190 mT. While some measurements confirm the possibility of reaching the superheating field, as predicted theoretically, Hs might still be of limited practical significance for SRF cavities: with a typical surface area of several meters square, it is easy to imagine locations with a reduced surface barrier because of, for example, roughness or clusters of impurities. Furthermore, operating cavities close to Hs will be impractical as Rs approaches the normal-state surface resistance. Nevertheless, understanding Hs from both the experimental and the theoretical point of view is important, particularly for materials alternate to Nb with high κGL , which have Hc values a factor of ∼2 greater than Nb. 5.3. Surface topography Surface treatments on bulk Nb cavities and substrate and deposition methods in thin film cavities determine the topography of the outermost ∼100 nm layer which carries the rf current. A comprehensive way to look at surface roughness over different length scales (“macroroughness” vs. “microroughness”) is to plot the power spectral density (PSD) as a function of the spatial frequency. The PSD is calculated from the surface profiles measured with different instruments, such as an atomic force microscope or a stylus profilometer, over different length scales. PSD data can

be compared with different surface structure models to understand how processes determine the surface topography [143]. Data analysis clearly indicates that BCP, unlike EP, produces surface structures dominated by step edges, because of differential grain boundary etching. Smoothing by EP occurs mostly in the spatial frequency range corresponding to 1–10 µm [66]. Removal of only ∼15 µm by EP is already sufficient to smooth sharp edges in BCP-treated samples, whereas the overall root-mean-square (rms) roughness decreases steadily with increasing material removal by EP, up to ∼50 µm [144]. The region of the cavity with the largest rms step height is the equatorial weld, which is also the region where the surface magnetic field is close to the peak value. Measurements of samples show that ∼90 µm of material removal by EP is required to reduce the rms step height in the weld region to a value comparable to areas not affected by the welding [145]. Replica techniques have been developed by several laboratories to extract the topography of outstanding features on the inner surface of cavities which are often associated with quenches [146]. The resolution of the technique is ∼1 µm and typical defects are “pits,” ∼100 µm in diameter and ∼100 µm deep, usually found in the EBW heataffected zone, as mentioned in Subsec. 4.3 Systematic studies at both DESY [147] and KEK [42] show that the quench field of a cavity treated by EP and LTB decreases with increasing material removal by BCP, followed by LTB. This trend is inverted if additional material removal by EP is done [148]. While these studies help to make the case why “smoother is better,” there exist examples of cavities reaching Eacc ∼ 40 MV/m having a “rough-looking” surface [149, 150]. Furthermore, no significant dependence in the performance of multicell ingot Nb cavities at DESY was found, whether grain boundary steps were mechanically polished prior to BCP [30]. What remains elusive is clear discrimination of what scale roughness “matters.” Efforts have begun to model specific nonlinear losses due to specific representative topographies [151, 152]. Finally, it should be mentioned that changes in nonlinear losses and the quench field associated with EP or BCP might also be related not only to topography but also to different concentrations of interstitial impurities, such as hydrogen and oxygen, or lattice

January 10, 2013

13:17

WSPC/253-RAST : SPI-J100

1230011

Superconducting Radio-Frequency Technology R&D for Future Accelerator Applications

defects near the surface introduced by the two processes, but specific evidence is lacking. 5.4. Grain boundaries The influence of grain boundaries on cavity performance is another controversial topic under investigation. Measurements on three single-cell cavities of the same shape made of fine-grain, large-grain and single-crystal Nb and treated in the same way did not show significant differences of performance [153]. As mentioned in the previous section, sharp steps at grain boundaries cause a geometric local magnetic field enhancement which can lead to a premature quench, if the direction of the rf field is nearly normal to the grain boundary plane [154]. Magneto-optical [155] and flux flow [156] studies on Nb samples showed that: • Grain boundary steps of height greater than ∼10 µm found on the fine-grain-welded samples can induce preferential flux penetration in those regions. • Preferential flux penetration at a grain boundary is clearly observed in bicrystal samples when the grain boundary plane is close to parallel with the applied magnetic field. In this configuration, the dissipation due to flux flow is the highest. There exist no direct measurements of the grain boundary depairing current density, Jb , in bulk Nb. However, it can be estimated from the following formula, assuming that the grain boundary makes a superconductor–insulator–superconductor junction: Jb =

π∆ , 2eGgb

(4)

where ∆ is the energy gap at 0 K, e is the electron’s charge and Ggb is the grain boundary specific resistance. Taking Ggb ∼ = 2 × 10−13 Ωm2 [157] and ∆ = 1.55 meV [158], one obtains Jb ∼ = 1.2 × 1010 A/m2 . The critical depinning current density, Jc , obtained from magnetization measurements of cavity-grade Nb samples was at least a factor of 10 lower than Jb [29]. This would suggest that losses due to oscillations of pinned vortices under an rf field might be more significant for a material such as bulk Nb, compared to high-temperature superconductors with stronger pinning. As mentioned in Subsec. 4.2.1, the efficiency of trapping the residual magnetic field when cooling

299

Nb samples below Tc was found to be lower in large-grain than fine-grain Nb. This effect, combined with the reduced grain boundary resistance contribution, because of fewer boundaries, might explain the lower Rres often measured for ingot Nb cavities, compared to that of fine-grain cavities. In addition, the onset of the high-field Q-slope is typically higher in ingot Nb than fine-grain cavities for the same surface treatment [43]. Oxide-filled, small-size (