Solar Cells and Their Applications [2 ed.] 0470446331, 9780470446331

A major update of solar cell technology and the solar marketplace Since the first publication of this important volume o

1,316 163 13MB

English Pages 618 Year 2010

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Solar Cells and Their Applications  [2 ed.]
 0470446331, 9780470446331

Citation preview

Copyright © 2010 by John Wiley & Sons, Inc. All rights reserved Published by John Wiley & Sons, Inc., Hoboken, New Jersey Published simultaneously in Canada No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/permission. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com. Library of Congress Cataloging-in-Publication Data: Solar cells and their applications / [edited by] Lewis Fraas, Larry Partain.—2nd ed. p. cm.—(Wiley series in microwave and optical engineering) ISBN 978-0-470-44633-1 (cloth) 1. Solar cells. I. Partain, L. D. II. Fraas, Lewis M. TK2960.S652 2010 621.31'244—dc22 2010000196 Printed in Singapore 10 9 8 7 6 5 4 3 2 1

Contents Preface

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

ix

Contributors .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

xiii

INTRODUCTION TO SOLAR CELLS

.

.

.

.

.

1

.

3

.

17

Solar Cells, Single-Crystal Semiconductors, and High Efficiency . . . . . . . . . . Lewis Fraas

43

PART I

.

Chapter

1

Solar Cells: A Brief History and Introduction . Lewis Fraas and Larry Partain

Chapter

2

Solar Cell Electricity Market History, Public Policy, Projected Future, and Estimated Costs . Larry Partain and Lewis Fraas

Chapter

Chapter

3

.

.

.

67

PART II TERRESTRIAL SOLAR CELL ELECTRICITY

.

.

111

Chapter

5

Crystalline Silicon Solar Cells and Modules Leonid Rubin

.

.

113

Chapter

6

Thin-Film Solar Cells and Modules . Robert Birkmire

Chapter

7

Terrestrial Module Fabrication and Assembly Technologies . . . . Christopher Bunner

Chapter

4

8

Solar Cell Device Physics . Larry Partain

Chinese Solar Cell Status. Wang Sicheng

.

.

.

.

.

.

.

.

.

.

.

137

.

.

.

.

159

.

.

.

.

171 v

vi

Chapter

CONTENTS

9

Tracking the Sun for More Kilowatt Hour and Lower-Cost Solar Electricity . . . . . . Ron Corio, Michael Reed, and Lewis Fraas

.

207

Solar Cell Systems: Definition, Performance, and Reliability. . . . . . . . . . Jason Strauch, Larry Moore, and Elmer Collins

.

219

Levelized Cost of Energy for Utility-Scale Photovoltaics . . . . . . . . . Matthew Campbell

.

.

251

.

.

271

Chapter 12

Low-Concentration Crystalline Silicon Systems . Lewis Fraas

.

273

Chapter

High-Concentration, III–V Multijunction Solar Cells . . . . . . . . . . Geoffrey Kinsey

.

.

293

High-Concentration Fresnel Lens Assemblies and Systems . . . . . . . . . . Gerhard Peharz and Andreas Bett

.

313

High-Concentration Cassegrainian Solar Cell Modules and Arrays . . . . . . . . Michael Ludowise and Lewis Fraas

.

337

Concentrator Solar Cell Installations at the University of Nevada, Las Vegas . . . . Suresh Sadineni and Robert Boehm

.

.

361

Concentrator Photovoltaic Field Installations . Francisca Rubio, María Martínez, and Pedro Banda

.

377

Chapter 10

Chapter

11

PART III TERRESTRIAL CONCENTRATOR SOLAR CELL SYSTEMS . . . . . . . . .

Chapter

Chapter

Chapter

Chapter

PART IV Chapter

13

14

15

16

17

SOLAR CELLS IN SPACE . 18

.

.

.

.

.

.

.

395

Space Solar Cells and Applications . Sheila Bailey and Ryne Raffaelle

.

.

.

.

397

CONTENTS

vii

PART V

OTHER ASPECTS AND CONSIDERATIONS .

Chapter

19

Solar Resource for Space and Terrestrial Applications . . . . Christian A. Gueymard and Daryl Myers

.

.

.

.

425

.

.

427

.

463

Chapter

20

Solar Energy Costs: The Solar Advisor Model . Paul Gilman, Nathan Blair, and Christopher Cameron

Chapter

21

Challenges of Large-Scale Solar Cell Electricity Production . . . . David Faiman

.

.

.

.

483

.

.

.

.

497

Market Overview of Flat Panel Detectors for X-Ray Imaging . . . . . . . . . Carl LaCasce, Larry Partain, and Chuck Blouir

.

499

Amorphous Silicon Transistors and Photodiodes . . . . . Robert Street

PART VI THIN FILMS AND X-RAY IMAGER TECHNOLOGIES . . . . . . Chapter

22

Chapter

23

.

.

.

.

.

.

511

Chapter

24

Amorphous Silicon Digital X-Ray Imaging . Richard Colbeth

.

.

533

Chapter

25

Photoconductor Digital X-Ray Imaging. George Zentai

.

.

.

559

PART VII Chapter

Index .

SUMMARY 26

.

.

.

.

.

.

.

.

.

581

Summary, Conclusions, and Recommendations. . . . Lewis Fraas and Larry Partain

.

.

.

.

.

.

583

.

.

.

.

.

.

613

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Preface This Second Edition is intended to be a comprehensive survey, review, and analysis of all the major factors related to the continuing technical development of solar cell electricity and its market development into a major worldwide source of electric power in response to powerful political and economic influences. It is divided into six major sections plus a Summary section including conclusions and recommendations. In contrast to the First Edition, Part I contains three initial chapters written so that nonspecialists and the more general readers and investors and policy makers can follow their contents without the need for specialized training or understanding. The goal is to allow a broad spectrum of readers to at least comprehend the market history, the influence of public policy, the likely costs of solar cell-generated electricity, and the special role that near-perfect, single-crystal semiconductor fabrication materials can have on overall performance. Chapter 4 in this part, on Solar Cell Device Physics (like the First Edition), is again aimed at advanced undergraduate and graduate college courses and other technical professionals involved in teaching, research, and commercial development. It not only covers the traditional abrupt p/n junction configuration of the First Edition but also expands into the very non-abrupt p-i-n geometries that characterize a whole new class of high-performance solar cells including interdigitated back-contact cells, point-contact cells, and heterojunction-with-intrisinsic-thin-layer (HIT) cells. It further addresses the special resistive restrictions that can limit p-i-n-type device performance as well as proposed paths to performance levels well beyond 50% efficiency levels. However, to maintain a reasonable length, this physics chapter does use the First Edition as a reference. Part II addresses the current state of terrestrial solar cell electricity technology and development programs. This includes the dominant crystalline silicon abrupt p/n junction devices and their large-scale fabrication and the emerging thin-film amorphous and polycrystalline semiconductor cells and modules. The amazing recent growth of the Chinese terrestrial solar cell program is presented in some detail. The potential advantages of tracking the sun are explored along with a detailed description of 3 years of field experience with fixed-axis crystalline silicon modules of 12% efficiency (under standard test conditions) in the Arizona desert. The emerging utility-scale installations are summarized along with their important cost-determining characteristics. Part III attempts to present a comprehensive overview of the terrestrial concentrator approach to solar cell electricity production and its special advantages ix

x

PREFACE

and challenges. This includes both low and high sunlight concentration levels with various system approaches as well as early results of small field tests at the University of Nevada and of substantial utility-scale field tests of multiple and varied concentrator systems in Spain. Part IV takes a broad look at space systems and all of the unique approaches, needs, accomplishments, plans, and future needs for space. Part V contains a chapter giving precise descriptions of the solar resource both terrestrially and in space. It also contains a chapter describing a sophisticated and detailed cost and performance model from the National Renewable Energy Laboratory (NREL). This Solar Advisor Model is reviewed and summarized. Finally, the special challenges of large-scale solar electricity production are explored. Part VI is a special four-chapter addition of the Second Edition that discusses how thin-film solar cells can be transformed into X-ray imaging devices when devices are reduced to submillimeter sizes and are aligned in columns and rows that are covered by a scintillator film that converts X-ray photons into visible light photons. If these are then attached to an array of the thin-film transistor switches, a flat-plate X-ray imager is produced. The market analysis of this whole X-ray imager field shows that its current market size of $2 billion per year should continuously evolve into a $15 billion per year wholesale market over the next 10 years or so as these devices continually improve in performance and drop in price. The final chapter summarizes the amazing growth of this solar cell electricity technology and market over the 15 years since the publication of the First Edition. It provides recommendations for how major countries and unions can play major roles from both technology and public policy perspectives and how continuing cost reduction and improved performance demands should be met under both near- and medium-term time frames. In summary, this book describes today’s baseline planar solar cell power systems as well as innovations in high-efficiency solar cells and concentrated sunlight systems that have occurred in the last 15 years, which now promise lower cost electricity competitive with other mainstream electric power sources. In addition to describing these technical breakthroughs in clear and simple terms, this book also describes the path from research breakthrough to high-volume production, emphasizing the cooperation required between government and private enterprise. Given this cooperation, solar cells can be a major contributor to the electric power production mix within the next 10 years. This book has been written for a large audience, not just a technical audience. It is hoped that any educated reader will find this book interesting, especially any reader who seeks to understand how the world’s energy supply problems can be increasingly addressed by exploiting direct solar energy resources available within a country’s borders. It further describes how most countries can start moving away from increasingly intense competition for decreasing depletable energy supplies and how they can continue moving toward a long-term, sustainable solution with inherently positive attributes.

PREFACE

xi

The thesis of this book is that solar energy can be cost competitive with other forms of electric power production and that the technical innovations required for this have already been made. Incentives for investment are needed to bring these innovations into high-volume production. It is hoped that this book will help educate the public, possible investors, as well as policy makers worldwide about the potential for a bright sunny energy future. Lewis Fraas Issaquah, WA Larry Partain Mountain View, CA

Contributors Sheila Bailey, NASA Glenn Research Center at Lewis Field, Space Environments and Experiments Branch, Cleveland, OH; email: [email protected] Pedro Banda, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Andreas Bett, Fraunhofer Institut für Solare Energiesysteme (ISE), Freiburg, Germany; email: [email protected] Robert Birkmire, Institute of Energy Conversion, University of Delaware, Newark, DE; email: [email protected] Nathan Blair, National Renewable Energy Laboratory, Golden, CO; email: Nate_ [email protected] Chuck Blouir, Varian Medical Systems, Cleveland, OH; email: chuck.blouir@ varian.com Robert Boehm, Center for Energy Research, Department of Mechanical Engineering, University of Nevada, Las Vegas, NV; email: rboehm@unlv. nevada.edu Christopher Bunner, Spire Corporation, Bedford, MA; email: cbunner@spirecorp. com Christopher Cameron, Sandia National Laboratories, Albuquerque, NM; email: [email protected] Matthew Campbell, Utility Power Plant Products, SunPower Corporation, Richmond, CA; email: [email protected] Richard Colbeth, Varian Medical Systems, Mountain View, CA; email: richard. [email protected] Elmer Collins, Sandia National Laboratories, Microsystems Science & Technology, Albuquerque, NM; email: [email protected] Ron Corio, Array Technologies, Inc., Albuquerque, NM; email: rcorio@wattsun. com David Faiman, Department of Solar Energy and Environmental Physics, Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Israel; email: [email protected] Lewis Fraas, President, JX Crystals Inc., Issaquah, WA; email: lfraas@jxcrystals. com Paul Gilman, Consultant, Chicago, IL; email: [email protected] Christian A. Gueymard, Solar Consulting Services, Colebrook, NH; email: [email protected] xiii

xiv

CONTRIBUTORS

Geoffrey Kinsey, Amonix, Inc., Seal Beach, CA; email: [email protected] Carl LaCasce, Varian Medical Systems, Salt Lake City, UT; email: carl.lacasce@ varian.com Michael Ludowise, SolFocus, Inc., Mountain View, CA; email: mike_ludowise@ solfocus.com María Martínez, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Larry Moore, Sandia National Laboratories, Microsystems Science & Technology, Albuquerque, NM Daryl Myers, National Renewable Energy Laboratory, Golden, CO; email: Daryl. [email protected] Larry Partain, Varian Medical Systems, Mountain View, CA; email: larry. [email protected] Gerhard Peharz, Fraunhofer Institut für Solare Energiesysteme (ISE), Freiburg, Germany; email: [email protected] Ryne Raffaelle, U.S. Department of Energy, National Center for Photovoltaics, National Renewable Energy Laboratory, Golden, CO; email: Ryne.Raffaelle@ nrel.gov Michael Reed, Array Technologies, Inc., Albuquerque, NM; email: mreed@ arraytechinc.com Leonid Rubin, Day4 Energy Inc., Burnaby, BC, Canada; email: lrubin@day4 energy.com Francisca Rubio, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Suresh Sadineni, Center for Energy Research, Department of Mechanical Engineering, University of Nevada, Las Vegas, Las Vegas, NV; email: [email protected] Wang Sicheng, Energy Research Institute, National Development and Reform Commision, Beijing, China; email: [email protected] Jason Strauch, Sandia National Laboratories, Integrated Microdevice Systems, Microsystems Science & Technology, Albuquerque, NM; email: jestrau@ sandia.gov Robert Street, Palo Alto Research Center, Palo Alto, CA; email: [email protected] George Zentai, Ginzton Technology Center, Varian Medical Systems, Mountain View, CA; email: [email protected]

PART I INTRODUCTION TO SOLAR CELLS

1 SOLAR CELLS: A BRIEF HISTORY AND INTRODUCTION LEWIS FRAAS1 AND LARRY PARTAIN2 1 JX Crystals Inc., 2Varian Medical Systems

1.1

BRIEF HISTORY

The history of the solar cell is really quite interesting [1]. In 1839, Edmond Becquerel found that two different brass plates immersed in a liquid produced a continuous current when illuminated with sunlight. We now believe that he had made a coppercuprous oxide thin-film solar cell. Later in the 1870s, Willoughby Smith, W. G. Adams, and R. E. Day discovered a PV effect in selenium. A few years later, an American named C. E. Fritts placed a sheet of amorphous selenium on a metal backing and covered the selenium with a transparent gold leaf film. He reported that this selenium array produced a current “that is continuous, constant, and of considerable force—with exposure to sunlight.” At the time, there was no quantum theory and there was considerable skepticism about his claim of converting sunlight into electricity. So he sent a sample to Werner Siemens in Germany, who was one of the most respected experts in electricity at the time. Siemens’s observation verified Fritts’s claims. However, the conversion efficiencies of both the thin-film cuprous oxide and the amorphous selenium solar cells were less than 1%. Around 75 years passed while quantum mechanics was discovered, the importance of single-crystal semiconductors was recognized, and p/n junction behavior was explained (see Chapter 3). By 1954, Chapin et al. [2] at Bell Labs had discovered, invented, and demonstrated the silicon single-crystal solar cell with 6% efficiency. Over the few following years, researchers brought the silicon solar cell efficiency up to 15%. The timing was fortunate because Sputnik was launched in 1957 and solar cells were the perfect lightweight low-maintenance remote electric power source. Today, silicon solar cells are being used to power the space station. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

3

4

A BRIEF HISTORY AND INTRODUCTION

The solar cell industry remained small until the first Arab oil embargo in 1973. Up until that time, the solar cell industry established a firm foothold with low-level but consistent cell and array production and performance. During those first 20 years, reliability was the driver and cost was not as important. After 1973, the flat-plate silicon module was brought down to earth and modified for weather resistance. This transition also included major improvements in cell and module fabrication that brought down costs dramatically (Fig. 2.3, chapter 2). Flat-plate “champion” silicon cell efficiencies (defined in Section 2.1, Chapter 2) have improved to values as high as 25%. Production module efficiencies have improved from around 10% for early modules to as high as 19% today (SunPower Corporation). Most important, annual production quantities have grown dramatically. Worldwide production exceeded 1 GW/year in 2002 and rose to over 3.8 GW/ year by 2006 (Fig. 2.1, Chapter 2). In the late 1970s, it was discovered that good cells could be made with multicrystalline wafers as long as the crystal size is at least 20 times larger than the optical absorption length [3]. Only those carriers within an optical absorption length from the crystal boundaries are lost. This is less than 5% of the carriers. Typical production quantity multicrystalline cell efficiencies are around 14%, whereas comparable single-crystal cells have efficiencies around 15%. By 2007, modules with multicrystalline cells accounted for about 45% of sales and modules with single-crystal cells accounted for about 40% of sales. Planar silicon cell modules dominated the market in 2007 because of their early well-funded foundation years for space satellites and their huge learning curve support (Fig. 2.3, Chapter 2) from single-crystal silicon and integrated circuit technology development. While silicon-based cells still dominate the solar cell electricity market today, several other cell types have now entered the market. (Solar cells are also known as PV cells.) These newer cell types have added diversity in potential applications as well as offered alternate paths to lower-cost solar electric power. These alternate cell types include hydrogenated amorphous silicon, cadmium teluride and CIGS thin-film cells (Chapter 6), as well as concentrator cells with efficiencies as high as 41% (Chapters 13–17).

1.2

APPLICATIONS AND MARKETS

In the late 1970s and early 1980s, the traditional solar cell electricity applications [4] were at remote locations where utility power was unavailable, for example, campers and boats, temporary power needs for disaster situations, and power for remote communication station repeaters. In the late 1980s and early 1990s, solar cells began to be routinely used to provide site-specific energy for urban and suburban homes, office buildings, and a multitude of other mainstream grid-connected applications. Also, solar cell electricity systems have become very important sources of energy in the developing world. Today, for an increasing number of power needs, solar cell electricity is the cheapest and best way to generate electricity.

APPLICATIONS AND MARKETS

5

In addition to the solar power arrays on space satellites, there are now many different types of PV systems used here on Earth including 1. 2. 3. 4. 5. 6.

remote stand-alone without battery storage, remote stand-alone with battery storage, small modules for calculators and toys, residential grid connected with DC to AC inverter, commercial grid connected with inverters, and PV fields for utility power generation.

Remote solar water pumping is a nice example of stand-alone solar cell electricity where batteries are not needed. Solar water pumping is very desirable for crop irrigation, livestock watering, and clean water for remote villages. Solar water pumping systems are now installed around the world. The nice thing about this application is that underground water is pumped when the sun is shining. It can be immediately used for crop irrigation. In other areas, it can be pumped into tanks for livestock to drink. In third world countries, pumping underground water for people to drink provides cleaner water than surface water thereby limiting disease. This application is quite economical because the system is simple. Battery storage or DC to AC conversion is not necessary. Simple solar trackers are used to maximize pumping time. The electric motors driving the pumps have a threshold current that must be provided before they will operate. By tracking the sun, this power is provided from dawn to dusk, not just at around noon as would be the case without tracking. Another application where there is a good match with demand is for air conditioning in developed countries like the United States. For many remote applications, storage is needed to store electric energy for when it is needed. Examples of these applications include off-grid cabins and remote communication repeater stations. For most solar cell applications where storage is needed, secondary or storage batteries are the best alternative. Generally, batteries should be deep discharge batteries such as marine batteries or motive power batteries. Forklift trucks and golf carts use large-capacity deep discharge batteries that are designed for long life and many discharge cycles. In addition to batteries, combination systems can be used to compensate for the fact that the sun does not always shine. A solar/wind combination is particularly good since quite often, either one or the other is available. Another combination system can be a solar–thermal cell electricity system. In this case, solar cells are located on your roof for generating electricity in the summer and infrared-sensitive PV cells (also known as TPV cells) are integrated into your heating furnace to generate electricity when it is cold and dark outside and you need heat to keep warm. In a TPV cell electricity system, a ceramic element is heated in the furnace flame and its glow in the infrared is converted to electricity by infraredsensitive TPV cells [5]. Solar-powered calculators are another familiar application for solar cells. While the efficiencies of amorphous silicon solar cells are much lower than either single or multicrystalline cells, an advantage for thin-film cells is that they can be made with cell interconnections built into the process. This means that for applica-

6

A BRIEF HISTORY AND INTRODUCTION

tions like powering calculators where voltage but little current is required to run the calculator, amorphous silicon circuits are preferred to save on the cost of interconnecting multiple cells to provide voltage. Credit is due to the Japanese for recognizing this advantage and to the inventors of the amorphous silicon solar cell for making solar cells a common household item [6]. Today, more and more homes on the grid are using solar cell arrays to generate electricity to save on costs of peak electric power. The passage of the PURPA by the U.S. Congress made it possible for a small producer to install generating systems and to sell the power to the utility at a favorable price without the enormous amount of red tape usually required of a new electric power producer. Most states have now also passed net metering laws that allow the electric meter at a home to run both directions. However, at least in California, the utility charge can at most be reduced to zero and they never pay any net money to their customers who produce more electricity than they consume. This allows homeowners generating solar cell electricity to send energy to the grid if they are producing excess electricity with a credit from the utility so that they can use electric power from the grid on days without sufficient sunlight. An example of real cost savings with a solar cell electricity installation for a homeowner in San Jose, California, is shown in Figure 1.1 [7]. Current Monthly KWH 2,500 2,000

Over 300% $0.26 300% Baseline $0.24

1,500

200% Baseline $0.19

1,000

130% Baseline $0.14

500

100% Baseline $0.13

e

Ju ly Au g Se u pt st em be r O ct ob N ov er em b D ec er em be r

ay

r il

Ju n

M

Ap

Ja nu a Fe ry br ua ry M ar ch



Monthly KWH With PV System 2,500 2,000

Over 300% $0.26 300% Baseline $0.24

1,500

200% Baseline $0.19

1,000

130% Baseline $0.14

500

100% Baseline $0.13

Ju ly Au g Se u pt st em be r O ct ob N ov er em be D r ec em be r

Ja nu a Fe ry br ua ry M ar ch Ap ril M ay Ju ne



Figure 1.1. When electric utility rates are staged, a homeowner with solar can displace electricity at the peak power rate as illustrated here. This example was originally presented by Akeena Solar on their web site in 2003 and then published in reference [7].

APPLICATIONS AND MARKETS

7

Figure 1.1 is for an actual case in 2003. Note in this figure that the utility electric rates are staged. While the homeowner pays a base rate of 13¢ per kilowatt hour that in itself is well above the national average. More importantly, the homeowner is paying twice that or 26¢ per kilowatt hour for his peak power. So his solar electric system is saving him money at the 26¢ per kilowatt hour rate. While the grid-connected solar cell electricity market started with residential customers, commercial customers are now starting to use solar arrays on their flat building rooftops. Figure 1.2 shows a photograph of two 1-kW solar cell arrays on a flat rooftop in Spokane, Washington. These arrays are mounted on carousel solar trackers (Chapter 9). People have been dreaming of the potential of solar cell electricity systems as a major electric power source for over 100 years. Now with the existence of solar power fields such as the one in China shown in Figure 1.3, this dream is becoming reality.

Figure 1.2. Two-kilowatt PV array from JX Crystals Inc on a commercial building flat rooftop.

Figure 1.3. Solar cell electricity generating field in Shanghai, China. System designed by JX Crystals Inc.

8

1.3

A BRIEF HISTORY AND INTRODUCTION

TYPES OF SOLAR CELLS AND MODULES

Unfortunately, solar cell electricity is still too expensive for widespread economical use (Section 2.4, Chapter 2). While it is hoped that traditional crystalline silicon module prices will continue to fall, there are other alternatives under development as shown in Figure 1.4. Figure 1.4 shows the three types of solar modules in use today [8]. The upper section (Figure 1.4a) of this figure shows the planar single-crystal silicon modules and fabrication procedure. This approach dominates the solar market today with over 85% of solar modules sold. As shown in Figure 1.5, retail module prices have

(a) Standard Silicon Single Crystal Module Fabrication Crystal to Ingot to Wafer to Module

(b) Concentrator Module Fabrication Smaller Single Crystal Cells With Mirrors (shown) or Lens Array

(c) Thin Film Module - Spray-on Successive Non-Crystalline Films

Figure 1.4. Alternate PV module types: (a) standard silicon single-crystal module fabrication, crystal to ingot to wafer to module; (b) concentrator module fabrication, smaller single-crystal cells with mirrors (shown) or lens array; and (c) thin-film module, spray-on successive noncrystalline films.

ARGUMENTS FOR SOLAR CELL ELECTRIC POWER

1.4

11

ARGUMENTS FOR SOLAR CELL ELECTRIC POWER

While solar cell electricity is still expensive today, there are three strong arguments for national programs to accelerate its transition into a mainstream power source. The first argument is that there is a logical path for future lower costs for solar electricity. There are three simple steps that will lead to lower cost given development and manufacturing scale-up. These steps are based on technical breakthroughs that have now been made. In step 1, given that the cost of solar electricity today (August 2009) is about 20¢ per kilowatt hour (Solarbuzz) for commercial-sized systems for fixed flat-plate systems in the sunny Southwestern United States, by implementing solar trackers where the modules continuously point at the sun, one can gain 1.3 times more kilowatt hour per installed kilowatt, reducing the cost of solar electricity to about 16¢ per kilowatt hour. This is already being done as evidenced in Figures 1.2 and 1.3 [10]. Step 2 is then to decrease the module cost while maintaining its performance by using lower-cost optical elements as shown in Figure 1.4b. This CPV approach by itself can potentially reduce the system cost for solar electricity to under 10¢ per kilowatt hour (see Chapter 12) [10]. In step 3, one then increases the module efficiency in the CPV approach to well over 20%. As described in Chapter 3, this should reduce the cost of solar electricity still further. “Champion” CPV module efficiencies as high as 31% [11] have now been demonstrated including the one shown in Figure 1.7. While logic suggests these lower costs, this will depend on funding for manufacturing scale-up and government top-down commitment. Actually, there are multiple approaches for CPV ranging from LCPV systems using linear mirrors with silicon cells as shown in Figure 1.4b to HCPV systems

Table III: Performance Summary Packaged Projected Measure Measure Cells at STC with at Module STC 90% Operate at STC Optical Temp (April 28) Effic (April 28) DJ Cell 17.4 W 15.7 W 14.4 W 15.1 W Power DJ Cell 31.5% 28.4% 26.1% 27.3% Effic. IR Cell 3.64 W 3.28 W 2.6 W 3.1 W Power IR Cell 6.6% 5.9% 4.7% 5.6% Effic. Sum 21 W 19 W 17 W 18.7 W Power Sum 38.1% 34.3% 30.8% 32.9% Effic. NIP DNI = 0.92; Area = 600 cm2; Input Power = 55.2 W

Figure 1.7. Prototype CPV module with demonstrated outdoor module efficiency of 31%.

12

A BRIEF HISTORY AND INTRODUCTION

with newer semiconductor materials [12] such as the one shown in Figure 1.7. These LCPV and HCPV systems will be described in more detail in Chapters 12–17. Of course, while the above three steps can be implemented, this still requires investment and political commitment. This leads us to our next two arguments in favor of national programs to accelerate the penetration of solar cell electricity into the mainstream energy mix. The second reason relates to the fact that oil and natural gas resources are being depleted. Quoting from Kenneth Deffeyes’s [13] book titled Hubbert’s Peak: The Impending World Oil Shortage, “In 1956, the geologist M. King Hubbert predicted that U. S. oil production would peak in the early 1970s. Almost everyone inside and outside the oil industry rejected Hubbert’s analysis. The controversy raged until 1970 when the U.S. production of crude oil started to fall. Around 1995, several analysts began applying Hubbert’s method to world oil production, and most of them estimated that the peak year for world oil will be between 2004 and 2008. These analyses were reported in some of the most widely circulated sources: Nature, Science, and Scientific American” [14]. The 2008 peaking of world oil prices to record levels above $140 per barrel seems to support these predictions. The war in Iraq that began in 2003 was likely influenced, at least in part by the shortage of proven U.S. oil and natural gas reserves that could only last 3.0 and 7.5 years, respectively, should the United States have to depend only on its own reserves [15]. The consequence of this “impending world oil shortage” is that electricity prices are going to be rising probably abruptly within the next 5–10 years. This affects the economics of solar cell electricity as solar modules based on semiconductor devices will last for 25 years or longer. Today’s cost competition calculations for solar cell electricity usually assume a short-term payback and nonescalating energy prices. The third argument in favor of bringing solar cell electricity into the mainstream is the environmental and moral argument. It is desirable to avoid global warming as well as oil related war. When one thinks about conventional electric power production, one thinks about oil, natural gas, nuclear, and coal as fuel sources. Solar cell electricity is not on this list because it is currently too expensive. However, these conventional fuel sources have hidden unintentional costs. For example, nuclear fuels are coupled with nuclear waste management and nuclear weapons. Then nuclear waste and nuclear weapons are coupled with the cost of homeland security and our fear of weapons of mass destruction. There are hidden costs involved in attempting to guarantee that nuclear materials do not find their way into the hands of terrorists. Another example of hidden costs is the world’s dependence on oil from the Middle East that is linked unavoidably, particularly in the United States and in other developed countries, with terrorists from the Middle East. It can arguably be claimed that wars have now been fought in the Middle East to secure oil supplies.

ABOUT THIS BOOK

13

In contrast to the unintended costs just enumerated, consider solar energy. Solar energy is inevitable on the larger scale of time. Solar energy is really already a primary energy source through wind and hydroelectricity. Solar energy generated our coal, oil, and natural gas via photosynthesis a hundred million years ago. Solar cells are very much more efficient than plants at converting sunlight to useful energy. Finally, solar energy is benign and will benefit the whole world.

1.5

ABOUT THIS BOOK

The first edition of this book [16] was published in 1995 and can serve as a reference for this second edition. This second edition is divided into four main parts. Part I is an introduction to the current markets, cell and module types, and the physics of solar cell operation. The solar cell electricity market has grown appreciably over the last 14 years as described in Chapter 2. The basics of solar cell operation are presented for single-crystal cells and for thin-film cells in Chapters 3 and 4. Part II of this book focuses on the status of solar cell systems today. Single and multicrystalline silicon and thin-film cells and modules are described in Chapters 5 and 6. Over the last 3 years, silicon module automated manufacturing is coming online with the promising major cost reductions. The traditional and currently dominant silicon module manufacturing, now with automation, is described in Chapter 7. Also, over the last 3 years, China has made a major commitment to solar module manufacturing, and the status of solar electricity in China is described in Chapter 8. A major cost reduction for solar cell electricity comes through the use of solar trackers as described in Chapter 9. Large multi-MW solar cell field installations are then described in Chapters 10 and 11. Part III of this book then describes newer concentrated solar cell and system developments. Chapters 12–17 describe various concentrator solar cell electricity (also known as CPV) modules and system types and installations. Major developments have been taking place here over the last 3 years and that potentially could lead to major cost reductions over the next 5 years. While it remains to be seen if thin-film solar modules can produce electricity at rates competitive with other mainstream electricity generating technologies, nevertheless, amorphous silicon thin-film panels have found a place in other applications and in major markets like flat panel displays and medical imagers. The fourth part of this book describes successful applications of thin film technology as a spin-off from solar cell electricity. Chapters 22–25 then discuss the newest and rapidly growing applications of amorphous silicon thin films in X-ray imaging. The issue of the cost of solar cell electricity, solar modules, and solar systems is a very important subject addressed from various points of view in Chapters 2, 3, 10, 11, 20, and 26. Many believe that solar electricity prices will drop as a result of the recent investments in thin-film module manufacturing, and it is true that thin-film module prices have fallen. However, conversion efficiencies for these commercially available thin-film modules are still under 10%, and this means that

14

A BRIEF HISTORY AND INTRODUCTION

2.5–3.0 times more module area needs to be deployed relative to modules with over 19% conversion efficiencies. So, costs need to be compared at the system level, not just at the module level. Furthermore, the focus needs to be on the cost of the electricity produced in cents per kilowatt hour, not just the hardware cost. A summary and conclusions are presented in Chapter 26. Features that distinguish this second edition from the first edition are the much larger number of solar field installations and the major advances in the concentrator arena using very high-efficiency cells as well as the advances in other novel uses of thin-film modules. The current magnitude and momentum of the solar cell electricity market development (as summarized in Chapter 2) makes its eventual success both inevitable and unstoppable. This is due to the certainty of its future development path with its inherent, major advantages along with the worldwide spread of the scientific knowledge and the manufacturing and engineering know-how (covered in Chapters 3–18), plus the national commitments (alluded to in the Public Policy section of Chapter 2) that will turn promise into reality. World fossil fuel energy production rates will decline in the near to medium time frames, and the current lifestyles of the developed world cannot continue without appropriate replacements. Thus, it is no longer a question of whether this solar cell electricity power transition will occur but one of who will lead this process, who will reap the most benefits, and on what time scale it will occur. ABBREVIATIONS AC—alternating current CIGS—copper indium gallium deselenide CPV—concentrating photovoltaic DC—direct current HCPV—high-concentration photovoltaic LCPV—low-concentration photovoltaic MW—megawatt PURPA—Public Utilities Regulatory Power Act PV—photovoltaic TPV—thermophotovoltaic

REFERENCES [1] [2] [3] [4]

J. Perlin. From Space to Earth, the Story of Solar Electricity. Ann Arbor, MI, AATEC Publications (1999). D. M. Chapin, C. S. Fuller, and G. L. Pearson. A new silicon p-n junction photocell for converting solar radiation into electrical power. J. Appl. Phys. 25, 676 (1954). H. C. Card, and E. S. Yang. IEEE-TED 24, 397 (1977). R. J. Komp. Practical Photovoltaics, 3rd edition, revised. Ann Arbor, MI, AATEC Publications (2001).

REFERENCES [5] [6] [7] [8] [9] [10] [11] [12]

[13] [14] [15] [16]

15

L. M. Fraas, J. E. Avery, and H. X. Huang. Thermophotovoltaic furnace-generator for the home using low bandgap GaSb cells. Semicond. Sci. Technol. 18, S247 (2003). D. E. Carlson and C. R. Wronski. Appl. Phys. Lett. 28, 671 (1976). L. M. Fraas. Akeena Solar example cited in Chapter 2, p. 15, in Path to Affordable Solar Electric Power and the 35% Efficient Solar Cell, Issaquah, WA, JX Crystals Inc. (2004). L. M. Fraas. Path to Affordable Solar Electric Power & The 35% Efficient Solar Cell. Available at http://www.jxcrystals.com/ (2004). M. A. Green. Solar cell efficiency tables (version 29), Prog. Photovolt. Res. Appl. 15, 15–40 (2007). L. M. Fraas, J. Avery, L. Minkin, H. X. Huang, A. Gehl, and C. Maxey. Carousel trackers with 1-Sun or 3-Sun modules for commercial building rooftops. Presented at the Solar 2008 Conference, May 6, 2008, San Diego, CA (2008). L. M. Fraas, J. Avery, H. Huang, L. Minkin, and E. Shifman. Demonstration of a 33% efficient Cassegrainian solar module. Presented at 4th World Conference on PV, May 7–12, Hawaii (2006). L. M. Fraas and R. C. Knechtli. Design of high efficiency monolithic stacked multijunction solar cells. In IEEE Photovoltaic Specialists Conference, 13th, Washington, D.C., June 5–8, 1978, Conference Record (A79-40881 17-44). New York, Institute of Electrical and Electronics Engineers, Inc., pp. 886–891 (1978). K. S. Deffeyes. Hubbert’s Peak: The Impending World Oil Shortage. Princeton, NJ, Princeton University Press (2001). C. A. Campbell and J. H. Laherrere. The end of cheap oil. Sci. Am. March, 78 (1998). www.BP.com web site has a section entitled “BP Statistical Review of World Energy 2003.” This site has country and regional proven reserves and consumption data for both oil and natural gas. L. D. Partain, ed. Solar Cells and Their Applications, 1st edition. New York, John Wiley & Sons (1995).

2 SOLAR CELL ELECTRICITY MARKET HISTORY, PUBLIC POLICY, PROJECTED FUTURE, AND ESTIMATED COSTS LARRY PARTAIN1 AND LEWIS FRAAS2 1 Varian Medical Systems, 2JX Crystals Inc.

2.1

MARKET HISTORY

The worldwide solar cell (or photovoltaic) electricity generation market has grown dramatically over the past 30 years, both in terms of gigawatts per year (Fig. 2.1) (1 GW = 1 billion watts) and in billions of U.S. dollars per year sales (Fig. 2.2) [1]. Worldwide production exceeded 1 GW/year in 2002 and rose to over 3.8 GW/ year by 2006, and worldwide sales increased from US$(2007)1.5 to 9.7 billion over this same time period. The US$(2007)390 billion sales of the world’s largest oil company, Exxon Mobile, are shown in Figure 2.2 for reference [2]. Since 1975, this progress has provided a 1000-fold increase in gigawatts per year produced and almost a 100-fold increase in billions of U.S. dollars per year in sales. From 1975 to about 1985, the gigawatt per year growth rate was about a factor of 2 every 2 years. From about 1985 to 1995, it dropped to about half that rate before returning to approximately doubling every 2 years from about 1995 through 2006. Another 1000-fold increase in gigawatts per year production would place the solar cell electricity generation capacity in the range of both the total 2006 U.S. installed electricity generation capacity of 1075 GW [3] and the installed 2005 world electricity generation capacity of 3889 GW [4] even after allowing for the “capacity factor” of terrestrial sunlight being available only about 15–22% of the time [5, 6], neglecting storage issues. Since most commercial solar cells for terrestrial use

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

17

18

SOLAR CELL ELECTRICITY

World Solar Cell Production (gigawatts)

10000

US PURPA

1000

Net Metering

JPL FSA

100

06 US Generating 05 World Capacity Generating Capacity

10 1

Rate To Double Every 2 Years

Cumulative

0.1 Annual

0.01 0.001 1970

1980

1990

2000

2010

2020

2030

2040

Year

Figure 2.1. The world solar cell production levels, on both an annual and cumulative basis, since 1975, with comparisons to the U.S. and world total electricity generating capacities in 2006 and 2005, respectively, and with support and public policy program time intervals for the JPL FSA, the U.S. PURPA, and the net metering programs. 1000 Exxon

World Solar Cell Market Size (billions 2007 US$/yr)

US PURPA

100

Net Metering

JPL FSA

Rate To Double Every 2 Years

10

1

0.1 1970

1980

1990

2000 2010 Year

2020

2030

2040

Figure 2.2. The world annual solar cell market size since 1975 with a comparison to the annual sales of the world’s largest oil company, Exxon Mobile, in 2007.

are designed for 20 years or greater lifetimes, the actual solar cell generation capacity lies between the annual and cumulative curves of Figure 2.1. The reason for the difference in the gigawatt and dollar growth rates is the constantly falling cost of terrestrial solar cells over the past 30 years as plotted in Figure 2.3. The 1975 technology was essentially slightly modified space satellite technology that cost about a hundred U.S. dollars per watt in 2007. By 2006, this had steadily dropped 25-fold to about four 2007 US$/watt. Like many high-technology and electronic products, this cost reduction roughly followed a learning

20

SOLAR CELL ELECTRICITY

the United States’ fourth place position was essentially shared with Taiwan, whose rate of growth was higher. As shown in Figure 2.5 the United States held the third position in the yearly installation of solar cells in 2007, barely ahead of the rest of Europe (i.e., all but Germany) and with the rest of Asia (all but Japan) quite close behind. By 2004, Germany’s installation rate surpassed that of Japan. There are three major options to try and resume the module cost reduction process plotted in Figure 2.3. The first is to fundamentally improve crystalline cell technology and performance well beyond that of the original 1970s “modified space cell” configuration. Two potential candidates for this are the commercial SunPower interdigitated back contact cell and the SANYO HIT cell (see Chapters 3 and 4) both with prototype modules of independently confirmed “champion” efficiencies over 20% [9]. The second is thin film technologies that require much less (typically 100X) material than crystalline silicon solar cells (see Chapter 6). By 2006, the thin film annual production rate had grown to within a factor of a 1000 of the mainly silicon, total solar cell production rate, led by the United States and Japan as shown in Figure 2.6. In 2007, thin films’ growth rate was greater that that of the total market for the first time (compare with Fig. 2.1). This latest growth was mainly due to thinfilm CdTe cells that have taken the thin film lead from amorphous silicon. The decreased thin film production in Japan was mainly due to reduced activity in amorphous silicon. The third major option, for continued cost per watt reduction, is concentrator solar cell systems where the sunlight is focused down to a small area so that much smaller amounts of expensive solar cell material are also required for electricity 10000 World Solar Cell Installations (megawatts per year)

Europe Feed-In Tariffs

1000

Germany 100,000 Roofs

Germany Total

Japan

US

100

Rest of Asia

10 Rest Of Europe

1 1998

2000

2002

2004

2006

2008

Year

Figure 2.5. The annual solar cell installation rate from 2000 through 2007, broken down by the United States, Germany, Japan, the rest of Europe, and the rest of Asia, with time interval comparisons to Germany’s 100,000 Roofs Program and Europe’s Feed-In Tariff Program.

MARKET HISTORY

21

World Thin Film Solar Cell Production (gigawatts/yr)

10000 1000 100 10 Rate To Double Every 2 Years

1 0.1

Europe Japan

Total US

0.01

Rest Of World

0.001 1994

1996

1998

2000

2002

2004

2006

2008

Year

Figure 2.6. The annual world production of thin-film solar cells from 2003 through 2006, broken down by the United States, Europe, Japan, and the rest of world, and with the world total extended through 2007. The dashed line indicates a growth rate to double every 2 years.

Cumulative World Solar Cell Sales (billions 2007 US$)

1000 Rate To Double Every 2 Years

100

$36 billion by 2006

10 Germany 100,000 Roofs

1 JPL FSA

0.1 1970

1980

1990

2000

2010

2020

Year

Figure 2.7. The world cumulative solar cell sales from 1975 through 2006, with comparisons to the total expenditures of the JPL FSA Program through 1985 and Germany’s 100,000 Roofs Program through 2003.

generation (see Chapters 12–17). The latter are in their earliest, larger-scale testing phases, with most installation efforts currently centered in Spain ([10]; Banda and Rubio, Chapter 17). Figure 2.7 shows the cumulative sales totals of solar cells that have grown to US$(2007)36 billion by 2006. This is a classic learning curve-type response to a total integrated investment, which has overwhelmingly been into crystalline silicon solar cells, over this 1975–2006 time period. This total magnitude serves

22

SOLAR CELL ELECTRICITY

as a significant and growing barrier to any new technology that essentially has to become competitive at a greatly accelerated growth rate (comparatively) but with much less total integrated investment. Until recently, the thin-film growth rate had not been able to duplicate that of crystalline silicon (Figs. 2.1 and 2.6). Compounding any new solar cell approach’s problems is crystalline silicon’s constantly improving efficiency, reliability, ruggedness, acceptance, and costs as it continues down its learning curve, with ever-growing totals of integrated sales investment. The classic response strategy for this situation is a disruptive one where the new technology enters a niche market where the dominant technology does not participate and proceeds independently down its own learning curve largely unchallenged until its new concepts suddenly become competitive [11]. A potential example of the latter is the very high-efficiency, but also very high-cost, triplejunction space solar cells, based on III–V compounds (e.g., GaAs) that now dominate the space solar cell market (see Chapters 13–17).

2.2

PUBLIC POLICY

Over the past 30 years, there have been multiple significant public policy decisions, which have directly driven and strongly influenced the rate of development and the geographical location of major segments in the world total solar cell electricity marketplace. Here, five primary ones are highlighted that can be readily correlated with major market responses evident in the historical statistical market data. The earliest is the JPL FSA from 1975 through 1985 [12]. It corresponds to the initial rapid rate of market growth as indicated in Figure 2.1. It transformed delicate and expensive space silicon cells into rugged, reliable, and affordable terrestrial ones through five increasingly large (hundreds of kilowatts) and rigorously competitive block purchases of silicon cell modules whose properties were monitored and meticulously tested and reported [13]. The last two of these blocks produced the two early peaks in market growth shown on the left side of Figure 2.2. Its US$(1985)235 million (US$[2007]403 million) [14] total program size was equivalent to about 30% of the total integrated total market sales between 1975 and 1983 as is evident from Figure 2.7. Both its magnitude and sharp focus (on transforming fragile, expensive space silicon cell technology into robust and affordable terrestrial products) made it the primary driver of this early market development period. At each stage, these modules had to pass increasing stiff environmental reliability, stability, and performance standards. By the end, the project had met all of its major goals except for the price of the modules. The emerging results of singlecrystal, sawed silicon wafers, screen printed with a top metal grid contact, sandwiched between a plastic back plane and a top glass plate with polyvinyl butyral adhesive or EVA, set the standard for the next 20 years of solar cell market leadership by the United States. Following World War II, the U.S. economy was the world’s largest and it could best afford such a market generating and leading investment into this new breakthrough renewable energy technology. According to Figure 2.7, the growth

PUBLIC POLICY

23

in the world market equaled the total US$(2007)403 million JPL investment by 1988, with geometrical larger annual revenues the following years into the 1990s when U.S. companies still led in the world marketplace. For the next 10 years (about 1985–1995), the growth rate dropped by about a factor of 2, supported but only rather weakly by the PURPA that had passed in 1978. PURPA allowed “independent” electricity producers to be paid utility wholesale “avoided costs” under long-term (15–30 years), fixed-price contracts at least early on [15]. This was strongly led by California Public Utility Commission’s interpretation and requirements that ordered utilities to offer standardized contracts, most notably Standard Offer No. 4, with fixed prices. In the 1980s, PURPA was estimated to have produced up to 12 GW in non-hydro renewable energy system installations in the United States. Over half (6.1 GW) were in California, but most (over 90%) were for wind, geothermal, and biomass. Unfortunately, PURPA was not restricted just to renewables. By the early 1990s, its support of renewables stagnated with most new contracts going for nonrenewable natural gas “cogeneration” plants (producing both electric and steam) due to changing conditions including relatively low natural gas costs. Nevertheless, a significant aspect of PURPA was its first demonstration of a “feed-in”-type tariff, where renewable energy sources received “above-market” prices for their power output under longterm, fixed-price contracts, resulting in a large market response and many installations, although the majority of these were not for solar cell electricity. The next big positive stimulus for solar cell electricity was net metering, which started in the early 1990s [15]. It became prevalent enough in the United States to restore the rapid growth of solar cell production by about 1995 (as indicated on Fig. 2.1). Net metering allows independent producers, such as rooftop systems on residences and commercial buildings, to be credited by their utility for the retail avoided cost of utility-delivered electricity to their site (as opposed to wholesale avoided costs of PURPA). Typically, the utility only credits for the siteproduced power, up to the net power delivered to the site from the utility each year. Thus, any excess site electricity production is not accrued and no actual payment goes from the utility to the customer. The benefit is that the utility customer, producing local power, can offset up to 100% of the retail value of the power delivered to his or her site each year. From Figure 2.4, it can be seen that the Japanese solar cell production exceeded that of the United States by 1999 and it has maintained this world leadership production position since that time. The cost of Japanese solar cell modules decreased by over a factor of 4 from 1994 to 2003 (from ¥1733 per watt to ¥583 per watt or from US$[2007]23.62 to 4.32 per watt [8, 14, 16]) and went from being not cost competitive to being very competitive. During this same period, the efficiency of their modules increased from 12% to the 15–17% range, while the thickness of the silicon wafers used decreased from 380 to 180 microns. The performance increase came from continuous incremental improvements in light trapping (textured surfaces, double antireflection coating, reduced contact shadow). It also came from lower losses (good-quality substrates, passivation, heterojunctions) and from lower resistance (thicker electrodes, fine contact pattern). Cost reductions came

24

SOLAR CELL ELECTRICITY

from thinner substrates, larger cell sizes, simpler and lower-temperature processes, non-vacuum technology, and from making substrates directly from molten silicon. This was all largely developed directly at large commercial Japanese companies, but in close coordination with Japanese universities and with the Japanese government agencies with a clear mandate and a government resources commitment sufficient for success. The latter program was essentially that of the NEDO as part of the New Sunshine Program under the MITI [8] and its successor organizations. Approximately 8% of Japanese corporate funding (including NEDO support) goes to basic research, and this has been enough to provide the high-performance crystalline silicon HIT cell from SANYO (see Chapters 4 and 5). Since the late 1980s, the Japanese economy has been one of the world’s largest. In 2007, it was the world’s second largest economy, exceeded only by that of the United States [17]. With its comparatively small commitment to military expenditures, Japan has been particularly free to focus heavily on its internal hightechnology industrial development. “… It was in the middle of that decade [the 1990s] that the country [Japan] realized the importance of innovation and redesigned its whole strategy” [18]. “The size of NEDO’s budget illustrates this point. Japan’s corporations spend a combined 111 billion [2008] US$ on research and development every year, but more than 90% of that actually goes to incremental improvements of existing products, with 7 billion [2008] US$ going to basic research. That makes the more than 1 billion [2008] US$ that NEDO pours into its 140 research projects equal to 16% of the total corporate [Japanese] investment into basic research … Take NEDO’s long-term commitment to solar cell research. On the one hand, Japanese manufacturers like Sharp and Kyocera have won a significant share of the global market using technologies developed through NEDO projects. … With … sinister place names like ‘Devil River’, ‘Valley of Death’ and ‘Darwinian Sea’, … the illustration on the table [of the] … director general of … NEDO looks like some kind of map … It’s a vivid visual metaphor created by the U.S. National Institute of Standards and Technology to express the pitfalls that lie waiting for inventors trying to turn their research into commercial products … In Japan … [the NEDO director general] is the man charged with helping the scientific community to catapult technology and know-how past these obstacles and into the marketplace. … Since being reorganized in 2003 … [NEDO has] complete control of the national projects … [to] cancel projects that aren’t going well and direct more money into promising-looking research.” There are no details of the amounts of NEDO funding of solar cell R & D programs over the 2000–2005 time period, but when combined with Japanese corporate R & D expenditures, these combined totals are plausibly on the order of 30% of the integrated cumulative sales of solar cells of ∼10 billion (US$[2007]) by that time (see Fig. 2.7). Such NEDO-coordinated and well-supported programs in Japan, centered in large commercial operations, are reminiscent of the scale and methods of the U.S. technology development efforts termed the military–industrial complex by U.S. President Eisenhower [19]. The latter evolved dramatically from the cauldron of pressures leading up to and into World War II [20], and their implementations can be reasonably credited with keeping the U.S. military technology more than com-

PUBLIC POLICY

25

petitive with Germany’s and Japan’s early on in the World War II years and later on with the former Soviet Union’s. Its time-tested and hard-proven U.S. Department of Defense’s Science and Technology program (of US$[2007]10.8 billion) is split with 13% to basic research, 41% to applied research, and 46% to advanced technology development [21]. The latter two guide new military technology through “the Valley of Death”-type obstacles, identified by the former Advanced Technology Program of the U.S. National Institute of Standards and Technology and as highlighted and accommodated in current NEDO programs. Survivors are then ready for standard procurement or sales processes. U.S. public policy since the 1980s has largely blocked comparable programs in the United States for any technologies and industrial developments other than those focused or directed at U.S. military applications. The U.S. military–industrial complex approach was a dramatic change and improvement in the R & D process and in the rapid translation of research success into large-scale production and availability [20]. Its only major missing part, relevant to the widespread commercial introduction of solar cell-based energy, is the economic viability and market evaluations like those of the former Advanced Technology Program of the U.S. National Institute of Standards and Technology. By the mid-1990s, the EU countries and their increasingly integrated industries and markets began to build an expanding and major role for themselves in the world economy, not only for the EU in general but for Germany in particular, which was the world’s sixth largest economy by 2007 [17]. By 2007, the combined economies of the EU actually exceeded that of the United States. From 1999 until 2003, the German government instituted the 100,000 Roofs Program, targeted only at solar cell electricity, with 1.7 billion euros (US$[2007]2.2 billion) [14, 16] in favorable loans and feed-in law tariffs of up to 0.574 euros per kilowatt hour (US$[2007]∼0.75 per kilowatt hour) for up to 20 years [22]. This US$(2007)2.2 billion level investment again represents about 30% of the total world integrated sales of solar cells up to 2003 as shown in Figure 2.7. This program generated 145 GW of installed solar cell panels in Germany in 2003 to make it the leading installer since that time (see Fig. 2.5). It is estimated that the Roofs Program generated 10,000 German jobs and created 800 million euros (US$[2007]1.0 billion) in German industry revenues in 2003 alone. With this rate of revenue growth, this total loan investment was duplicated in German sales revenues within about the three following years. In comparison, the size of the California economy ranks eighth in the world [23]. Its and Germany’s (ranked sixth) have both the ability and track record of early demonstration of breakthrough technology and market trends including ones in renewable energy. Expanding on the PURPA concept, Germany introduced its version of feed-in tariffs in 1990, which was refined into its successful form by 2000 when it became a federally managed program [24, 25]. With such feed-in tariff law programs instituted in multiple European Community nations, Germany, Spain, and Denmark each provide significant portions of their countries’ electricity totaling 5%, 9%, and 20%, respectively, in 2007 from renewable sources, which were under long-term (20 year) contracts in Germany and in Spain [26]. The 2007 solar cell

26

SOLAR CELL ELECTRICITY

electricity installation totals for Germany and Spain were 47% and 23% (for a combined 70%) of the world market. These latest feed-in tariffs are justified by rationale that can include the following [24]. Long-term contracts (up to 20 years) pay solar cell electricity “tariffs” up to five times the going commercial rate (e.g., US$[2007]0.75 per kilowatt hour) [22], as opposed to representative rates (of US$[2007]0.16 per kilowatt hour typical of Italy in 2006 [27]). When 3% of a country’s electric power is produced under such agreements and the costs are spread among all the users, the 500% premium becomes a ∼15% increase in each user’s electricity bill. The integrated production that supplies these countries’ installations drives down the cost of solar cell modules, according to its learning curve (see Fig. 2.3). For a learning curve of 2X cost reduction for every 10X increase in cumulative production, the cost premium would be reduced by a factor of 2 when the cumulative production grows 10-fold, and a factor of 4 for 100-fold. After that, the cost to the customers for the next 3% of a country’s solar cell electricity installation reduces the price premium by a factor of 2–4 down to 7.5–3.75%, respectively (compared to the initial 15%). With integrated solar cell production growth doubling every 2 years (see Fig. 2.1), these 10- and 100-fold increases would occur in about 3.3 and 13.0 years, respectively. Three percent of a large user market can be quite significant particularly in the early penetration stages of solar cells into the total electricity market. For instance, 3% of the 2007 U.S. installed electricity capacity would be over 32 GW, more than eight times the total size of the world solar cell production total in 2006. The benefits that could accrue from such a program are a nation’s electricity energy supply becoming substantially less dependent on imported oil and gas, from a domestic source of sunlight that never decreases and that is delivered on-site free of charge, where the electricity conversion process itself (excluding the feedstock refining, manufacturing, installation, recycling, and disposal of solar cells) produces no pollution and generates no greenhouse gases. As a widely distributed source, it would be less subject to disruption by terrorist or military acts or by natural disasters. An orderly transition into this major new energy resource could be made steadily over many years, before world oil and gas production peak and begin to decline, which otherwise could generate sharp price fluctuations and major delivery disruptions with little or no time for orderly change. The resulting industries generated within countries could earn sales revenues equal to such subsidies within about three or four following years (if historical trends repeat) while at the same time creating large numbers of new jobs and major positive contributions to each country’s international balance of trade. Early participation in the process could well lead to world economic market leadership positions and geographic shifts of the centers of activity that historical evidence suggests can last for many years. It is the opinion of many economists [28, 29] that China is well on its way to becoming the world’s largest economy within the next few years or decades, with India likely to vie with the United States for second place perhaps 10–20 years even further out. From indications like the very high rate of expansion of its solar cell production capacity (Fig. 2.4), China may well be in the best position to invest

PUBLIC POLICY

27

the tens to hundreds of billions of 2007 U.S. dollars that historical data indicate could thrust it into the world’s leading position, at least in the area of solar cell production. And if history repeats itself, China would rapidly equal this total investment amount within 3 or 4 years of sales from its industries while also benefiting from the jobs generated and national sales revenue as well as from further increases to its international balance of payments surplus. There have been many other government-sponsored programs, some involving even larger monetary investments at their time, than the five cited above, but none with the comparably large and demonstrable effects on the world’s solar cell electricity market development history. Hence, it is not only the magnitude of a government program’s investments that provides the impetus for whole new technology applications and for relocating the geographical centers of leading activity, but it also takes wise selection of program focus and strong program execution. It is instructive to note that only the first (the JPL FSA) and the third (the Japanese NEDO development of crystalline silicon) of the five government programs covered above was directly focused on a single technology approach. Most, if not all of the other three, were essentially solar cell technology area neutral and were economic and technology stimulus programs. To a large degree, these all have mainly made use of the original JPL crystalline silicon design, with only this technology’s steady and incremental improvements, developed mostly by the manufacturers themselves (with large government subsidies at least in Japan) in response to the growing market. Over the last 30 years, the U.S. government’s (and many other countries’) renewable energy programs have had their major focus on the development of new breakthrough technology. Unfortunately, to date, the results have had little measureable impact on the solar cell electricity market and its development trajectory. However, this may be about to change; as Figure 2.3 indicates, the last 10X increase in cumulative solar cell production has produced little or no decrease in module costs, essential to the strategies covered above, for making solar cell electricity into a major alternative to depletable oil, natural gas, and coal energy supplies. A relevant public policy success example for technology implementation is the U.S. Interstate Highway System that was initiated in the Eisenhower administration beginning in 1956 [30]. It was justified as a defense-related program and was ∼90% federally funded by highway user taxes because it was considered “vitally important to national goals.” The Interstate Highway segments cost many times that of the then typical highways, and most of the highway users taxed had no direct Interstate Highway access in the early years of the program. The eventual positive impact of this system on the employment and housing opportunities of a large fraction of the U.S. population and on the U.S. economy as a whole has arguably been even greater than that of the federally funded U.S. National Aeronautics and Space Administration and its space programs. An analogue for this in solar cell electricity would be a federal tax on the users of depletable energy, namely, oil, natural gas, and coal. For example, this could pay a fraction of the initial capital costs (starting at, say, 75%) of installed solar cell electricity generating systems until approximately 30% of the then current cumulative world solar cell electricity

28

SOLAR CELL ELECTRICITY

market has been subsidized. These depletable energy users would then be directly paying for the development and deployment of renewable energy systems “vitally important to national goals” to the point that the latter becomes self-sustaining. For scale and reasonableness comparisons to the above past and potential future government programs, the following historical cost numbers may be useful. The cost of petroleum and petroleum products imported into the United States was US$327 billion in 2007, an increase of US$27 billion from the 2006 costs [31]. The U.S. 3-month balance of payment (current account) deficit increased by US$(2007)10.2 billion largely due to petroleum and petroleum product imports, which brought the total U.S. deficit to US$(2007)176 billion by the end of March 2008 [32]. Should military actions be needed at some future time to avoid major disruptions in imported gas and oil to any country, the ongoing Iraq war could provide a rough order-of-magnitude cost estimate from its 5 year costs to the United States of US$450 billion through the end of September 2007 [33]. The total final estimated cost in 1991 of the 42,795 miles of the U.S. Interstate Highway System at that point in time was $128.9 billion with $114.3 billion (or 88.7%) paid for from U.S. government funds at an average cost of $3 million per mile [34].

2.3

PROJECTED FUTURE

The historical market data above allow analysis of known quantitative results and comparatively objective evaluations of probable causes of, and contributions to, major past events and developments. Future projections require extrapolations beyond what is presently known and require a greater degree of speculation and unavoidably involve more subjective judgments and opinions. Nevertheless, important and accurate parallels can often be drawn between past history and future results, when past lessons learned are well applied, using well-accepted scientific principles and limitations combined with sound mathematical engineering and market analyses as attempted below. A continued growth rate of doubling every 2 years of production (and appropriately lower sales growth due to learning curve price reductions in Fig. 2.3) for solar cell electricity products and systems could lead to a sizeable fraction of the total U.S. and world electricity generating capacity and energy sales being provided by solar cell electricity in the decade between 2020 and 2030, as seen from Figure 2.1. One risk that could impede this progress is continuing stagnation of the Figure 2.3 solar cell module cost reduction with cumulative production. Another risk is that reasonable approaches may not be readily found to match the production of solar cell electricity to the demand schedule of electricity customers (e.g., night use of electricity when the sun is not shining). An accelerating factor for an even faster rate of growth is the rise in the real cost of energy (after correction for inflation) derived from depletable fossil fuel sources (i.e., oil, natural gas, and coal). The growing fossil fuel energy demands of the rapidly growing economies of China and India, plus the steady growth from the other leading world economies, strongly suggest that this real price increase will continue inexorably

PROJECTED FUTURE

29

for the medium term and precipitously over the longer term when the world production rates of oil and natural gas energy supplies eventually peak and begin to fall. A parallel case of a similar solid-state electronic technology growth is the approximate doubling of the calculating capacity on computer-integrated circuit chips every 2 years, which is termed Moore’s law [35]. Although all agree that this computing power growth rate cannot continue indefinitely due to fundamental and mathematical limitations, this empirical Moore’s law progress rate has essentially remained on track for more than 40 years. In comparison to solar cell electricity’s potential, 2030 is less than 25 years away. The terrestrial solar resource available is not a fundamental limitation to producing solar cell electricity each year at the level of the world generation capacity in 2005 (shown in Fig. 2.1) or even higher. The intensity of sunlight striking the outer boundary of the earth’s atmosphere is 1367 W/m2 (±4%) 24 h/day and 365 days/year as shown in the First Edition, Append B [36]. The diameter of the earth (at the equator) is 12,756 km [37] and the intensity of sunlight striking the earth’s surface (through an average atmospheric path length—i.e., AM1.5G) is 0.704 of its intensity at the atmosphere edge as given in the First Edition, Append A [36]. If one combines these with the conservative assumptions that (1) only land-based systems will be used; (2) one-third of the earth’s surface is land; (3) the availability of sunlight only matches demand 20% of the time (i.e., the capacity factor = 0.2); (4) 1% of the land is appropriate for solar cell electricity generation; and(5) the average system conversion efficiency is 10%, then the worldwide solar cell electricity resource size is 8193 GW or more than twice the 2005 world electricity generation capacity of 3889 GW shown in Figure 2.1. As the world’s dominant energy sources have evolved over time, from wood to coal and then to oil and gas, the total energy market size has increased substantially with each transition due to inherent advantages following each step. The solar input resource is large, so that the major challenge is to match solar cell electricity supply with demand. With natural hydro capacities at 121 and 923 GW for the United States and the world, respectively, in 2005 [38], this renewable storage resource can only cover demand mismatches of up to these same orders of magnitude. At Figure 2.1 extrapolated growth rates, such world levels could be reached as early as 2025. Unless battery or some other storage alternatives become practical, available and cost-effective in the meantime, solar cell electricity could be limited to supplying a fraction of the total world electric energy demand. To become the major electric energy resource, some fundamental change would likely be needed in electric energy use patterns. One of the most important and valuable U.S. exports is agricultural crops. This crop production essentially occurs only in the daytime, peaks in the summer months, with a virtual shutdown for most of the winter. Time of day and time of year utility rates could start to encourage business and personal lifestyles to take full advantage of such time availability of solar cell-generated electricity. Electric energy intense industries, such as aluminum refining and others, might well be induced to shift major fractions of their yearly

30

SOLAR CELL ELECTRICITY

production toward the daylight hours of the summer, if there were financial incentives like lower electricity rates.

2.4

COST ESTIMATES

The levelized cost of energy, L, is a detailed calculation of the effective cost of energy provided, such as 2007 U.S. dollar per kilowatt hour, considering all of the life cycle factors and expenses that contribute to such effective costs [39]. The intent of this chapter is to be as free of involved mathematical expressions and derivations as possible. Hence, Table 2.1 is used to separately summarize simple and approximate expressions that can be readily used to estimate L values and what effects them most based on parameter input values in the range of those listed in Table 2.2. This starts with baseline estimates taken from recent field experience and then proceeds to projected future cost reductions in the near- and medium-term time frames. The less mathematically inclined can thus skip the next two paragraphs and proceed directly to the results that follow. A more sophisticated and rigorous cost analysis is the complete topic of Chapter 20 (Solar Advisor Model) of this book. As listed in Table 2.1, the levelized cost of energy (L) is obtained by taking all the capital costs incurred for installing a solar cell electricity system, converting these initial capital costs into an annualized charge using fixed charge rates, and then dividing the results by the electric energy produced by such a system in a year. The result is an L value in dollar per kilowatt hour. Here, the simple approach developed by the Electric Power Research Institute (EPRI) was used [39]. The annual energy produced per unit area is just the solar energy density SEY (in kilowatt hour per square meter) that strikes the surface of the system per year, which is then adjusted for all of the relevant conversion efficiencies. The derating efficiency accounts for the mismatch and interconnect losses inevitably encountered in system assembly. A major component of fixed charge rates is the loan amortization rate, which is the per dollar cost per year of obtaining a loan at a given rate for a given period of time. This is easily obtained from loan amortization tables or from use of the Microsoft Excel PMT function. In locations where there are tax write-offs for the interest particularly on home loans, this reduces the fixed cost rate, proportionally to the tax bracket B and the fraction f of the loan payment that is interest. Additional annual financing costs are covered in the factor Δ. In a residential home loan, this Δ includes the annual taxes and insurance cost (per total loan dollar value) that adds to the amortized “interest and principal” payment (i.e., the principal, interest, taxes, and insurance [PITI] payments). The L value is conveniently expressed in terms of area-related capital costs per unit area and power-related capital cost per unit power (including the DC to AC inverter costs for the latter). One just multiplies these by the area A and the peak power P needed to generate 1 kWh in 1 year, and this gives the expression for L. For the power part, one needs to realize that module efficiency values are based on the assumption of a standard peak value of solar energy striking a system

TABLE 2.2. Parameter Values for Baseline, Near-Term, and Medium-Term Time Frames Parameter

Symbol

Value

Units

Baseline

Near Term

Medium Term

Module efficiency

ηm

0.135 (13.5%)

0.16 (16%)

0.20 (20%)

— (%)

Module cost

CW

4

2.2

1.25

$/W

Sunlight energy density/year

SEY

2435

2435

2435

kWh/m2/year (Phoenix)

Area-related BOS

Cb

307

155

150

$/m2

DC to AC inverter cost

Ci

900

690

300

$/kW

Inverter efficiency

ηi

0.95 (95%)

0.96 (96%)

0.97 (97%)

System lifetime

LTS

30

35

35

Years

Inverter lifetime

LTi

5

10

20

Years

System derating efficiency

ηD

0.95 (95%)

0.95 (95%)

0.95 (95%)

— (%)

I

2.50

1.21

0.55

$/W

0.06 (6%)

0.06 (6%)

0.06 (6%)

1/year (%/year)

Government incentive Loan rate

— (%)

Loan amortization rate system

mS

0.0726 (7.26%)

0.069 (6.90%)

0.069 (6.90%)

1/year (%/year)

Loan amortization rate inverter

mi

0.237 (23.7%)

0.136 (13.6%)

0.0872 (8.72%)

1/year (%/year)

Interest fraction of amortization system

fS

0.826 (82.6%)

0.870 (87.0%)

0.870 (87.0%)

— (%)

Interest fraction of amortization inverter

fi

0.253 (25.3%)

0.442 (44.2%)

0.688 (68.8%)

— (%)

Tax bracket

B

0.28 (28%)

0.28 (28%)

0.28 (28%)

— (%)

Additional fixed charge rate

Δ

0.0442 (4.24%)

0.0442 (4.24%)

0.0442 (4.24%)

1/year (%/year)

Fixed charge rate system

FS

0.100 (9.82%)

0.0964 (9.64%)

0.0964 (10%)

1/year (%/year)

Fixed charge rate inverter

Fi

0.265 (26.3%)

0.163 (13.7%)

0.115 (11.8%)

1/year (%/year)

Indirect cost rate

r

0.225 (22.5%)

0.225 (22.5%)

0.225 (22.5%)

— (%)

Levelized cost of energy

L

0.32

0.15

0.09

$/kWh

COST ESTIMATES

33

is 1000 W/m2. It is a standard accounting technique to multiply direct costs by an indirect rate, r, to include overhead and general and administrative costs including yearly operations and maintenance costs. For the expressions used here, it was assumed that the majority of the balance-of-systems (BOS) costs (including installation costs but excluding inverter costs) were area related. Hence, any BOS given in power-related form was converted into area-related form using the conversion relationship that is four lines from the bottom of Table 2.1. It is noted that any first-year government incentives I offered in terms of dollar per watt rebates simply subtract directly from the capital cost of the modules CW in dollar per watt. The final L expression is given three lines from the bottom of Table 2.1. The module cost CW expression on the bottom line came just from solving the L expression for CW in terms of all the other parameters. The baseline calculated costs for residential solar cell system electricity of $0.32 per kilowatt hour in Phoenix, Arizona, are given on the third column bottom line of Table 2.2. The baseline input parameter values for module efficiency, module costs, BOS costs, inverter costs, inverter efficiency, system derating efficiency, system lifetime, and inverter lifetime came from the U.S Solar Energy Technology Multiyear Program Plan 2007–2011’s [40] 2005 benchmark numbers obtained from 200 residential installations between 2000 and 2005. The loan rate and tax bracket were taken from the 2008–2012 plan [41]. The indirect cost rate and the system fixed charge rate were taken from the EPRI values [39] and its 22.5% value includes the 0.5% operations and maintenance charge used in the Solar Advisor Model. The Δ value (adding to indirect charge rate) was chosen so that the calculated FS value equaled the EPRI value of 10%. The Phoenix annual solar energy density SEY was taken from Riordan’s Chapter 20 of the First Edition [36]. As a reality check, the fixed charge rate for a 2007 home loan in Mountain View, California, was calculated as shown in Table 2.3, and its value of 9.13% (for PITI) is in this general range. The government incentive of $2.50 per watt was chosen to give the same $0.32 per kilowatt hour L value as reported in the SETP 2007–2011 plan [40]. It is in the approximate range of the most beneficial rebates offered by the state of California.

TABLE 2.3. Fixed Charge Rate F for the Purchase of a $400,000 Home with 20% Down and a $320,000 30-Year Mortgage at a 6.375% Loan Rate in Mountain View, California, in 2007 Parameter

Monthly Cost ($)

Yearly Cost ($)

Normalized Yearly Cost per Loan ($)

1995.0

23,935.0

0.0748 (7.48%)

375.0

4500.0

0.0141 (1.41%)

766.0

0.0024 (0.24%)

Principal and interest Taxes Insurance Fixed charge rate F (total)

63.85

0.0913 (9.13%)

34

SOLAR CELL ELECTRICITY

Levelized Cost of Energy L ($/kW-hr)

One of the advantages of the simple Table 2.1 expressions is the ease with which they can be used for sensitivity analyses. Figure 2.8 shows such a sensitivity plot of the levelized cost L to different module efficiency values. For the baseline case with a module efficiency of 13.5% and an L value of $0.32 per kilowatt hour (shown by a vertical line), one can see the beneficial effects of greatly increased module efficiency rapidly saturate with a sharper increase in L costs with decreased module efficiency. Similarly, Figure 2.9 shows the sensitivity of allowed module 0.5

0.4

0.3 Base Line

0.2 Near Term

0.1 Medium Term

0 0

10

20

30

40

50

60

Module Efficiency ηm (%)

Figure 2.8. The levelized cost of energy L as a function of module efficiency ηm for the baseline, near term, and the medium term obtained from the expressions of Table 2.1 using the input parameter values of Table 2.2. 6

Base Line $0.32/kW-hr

Module Cost CW ($/W)

5 4 Near Term $0.15/kW-hr

3 2

Medium Term $0.9/kW-hr

1 0 -1 -2 0

10

20

30

40

50

Module Efficiency ηm (%)

Figure 2.9. The allowed module cost CW obtained by holding all the parameters constant, except for module efficiency ηm and module cost CW, using the other parameter values of Table 2.2 for the baseline, near term, and the medium term calculated with the expression at the bottom of Table 2.1.

COST ESTIMATES

35

costs in dollar per watt (around the baseline $4 per watt value indicated by the vertical line) with module efficiency but with all the other baseline parameters held constant including L at $0.32 per kilowatt hour. Here, higher module costs can be supported with higher module efficiencies, but again this rapidly saturates. At lower module efficiencies, the module cost that can be justified drops sharply. For module efficiencies below 5%, the constant energy cost value L cannot be supported even if the module costs went to zero. This effect is due to the BOS cost described by the Cb parameter in the Cw expression at the bottom of Table 2.1. One conclusion from these sensitivity analyses is that there is no one parameter whose improvement can make the cost of solar cell electricity competitive with current electricity rates. Hence, two more cases were considered and they are the “near-term” and “medium-term” cases shown in Table 2.2 and in Figures 2.7 and 2.8 that respectively correspond to L values of $0.15 per kilowatt hour and $0.09 per kilowatt hour, module efficiency values of 16% and 20%, and module costs of $2.20 per watt and $1.25 per watt plus all the other optimistic parameter values listed in Table 2.2. These parameter values were taken from the “2011” and “2020” projections of the SETP 2007–2011 tables [40] plus similar listed adjustments to the remaining parameters. It should be noted that the SETP 2008–2012 table and plans [41] accelerated such improvements in to the “2010” and “2015” time frames. In this context, “near term” corresponds to the 2010–2011 time frame and “medium term” corresponds to the 2015–2020 time frame. Again, the corresponding curves of Figure 2.8 illustrate the positive but saturating benefits of increased module efficiencies and the sharper penalties for lower module efficiencies. Here, Figure 2.9 shows the even stronger penalties for reduced module efficiencies with the L costs held constant at increasingly competitive values. For the specific $0.09 per kilowatt hour case considered here, module efficiencies below 8% could not be justified even if their costs were zero. Further, at a module price of $2 per watt, the calculated minimum module efficiency required to be competitive is 6% for the baseline case and 14% for the near-term cases but is not achievable for the medium-term case. Similarly, at a module price of $1 per watt, the calculated minimum efficiencies required to be competitive are 5%, 6%, and 15%, respectively, for the baseline, near-term, and medium-term cases. A bottom line here is that decreased module costs are essential for achieving competitive system performance but that lower-cost modules still have to have efficiencies relatively near the more expensive modules of higher performance to provide competitive, levelized cost values. All of these emphasize the importance of resuming the Figure 2.3 module cost learning curve reductions that most recently have become quite flat. The breakdown of the three major system components and their contributions to the levelized costs L are shown in Table 2.4 for the baseline, near-term, and medium-term cases. For all three, the module costs contribute in the 70s percent range; the area-related BOS contributions are in the mid-30s to low 40s percent range, with the inverter contributions in the 30s–40s percent range. The resulting L values are only possible due to government incentives that pick up 30–40% of the total component contributions. This is the case even though the Table 2.2

36

SOLAR CELL ELECTRICITY

TABLE 2.4. Individual Parameter Contributions to the Levelized Cost of Energy L Parameter

Symbol

Contribution to L ($/kWh and %) Baseline

Near Term

Medium Term

Module cost ($/W)

CW

0.235 (73.6%)

0.117 (77.8%)

0.066 (72.7%)

Area-related BOS ($/m2)

Cb

0.134 (41.8%)

0.052 (34.2%)

0.039 (43.6%)

DC to AC inverter cost ($/kW)

CI

0.098 (30.6%)

0.046 (30.8%)

0.014 (15.6%)

Government incentive ($/W)

I

–0.147 (−46.0%)

–0.064 (−42.8%)

–0.029 (−32.0%)

Levelized cost of energy

L

0.320 (100%)

0.150 (100%)

0.090 (100%)

parameter assumptions show the actual government contributions in dollar per watt falling by more than a factor of 3 over this period. The SETP 2008–2012 Plan [41] and its Solar Advisor Model actually have the government incentive going to zero by the medium term. If these government incentives were all zero, the respective L values would be $0.467 per kilowatt hour, $0.214 per kilowatt hour, and $0.119 per kilowatt hour for the baseline, near-term, and medium-term cases (from the addition of the top three contribution rows of Table 2.4). It is such amounts that would be appropriate for “feed-in tariff ” situations. A reality check of this zero-incentive “2005” baseline value is that it is on the order of the $0.75 per kilowatt hour long-term contracts that Germany’s programs signed in the early 2000s to make their “feedin” programs financially attractive. Indeed, it is such feed-in tariff scenarios that market history indicates have been the most effective in stimulating rapid adoption and in promptly moving new geographic regions into higher world market ranking positions. The above near- and medium-term “zero-incentive” L values seem particularly reasonable for this latter type of public policy support strategy.

2.5

CONCLUSIONS

The 1000-fold growth of the world’s production of solar cell electricity in the 30 years since 1975 resulted from a growth rate that doubled output every 2 years for the first and last 10 years of this period. The middle 10 years had a growth rate reduced by about a half. Another 1000-fold growth in production would provide an electricity production level equivalent to the world’s total installed generation capacity of 3889 GW in 2005. Should a growth rate of doubling every 2 years be maintained, the latter capacity levels would be reached in the decade between 2020 and 2030.

CONCLUSIONS

37

The technical and manufacturing expertise required to support such developments is now spread throughout the world. The challenges to achieving these results are likely on the same order as those faced by the Manhattan Project that produced the first nuclear weapons or by the National Aeronautics and Space Administration’s program that placed the first men on the moon. These challenges include matching the availability of solar cell electric power with demand, continued reductions in the costs of producing electricity from solar cell systems (probably by another factor of 4 from baseline values) in ways that do not deplete basic resources needed for their construction and deployment, that does not seriously degrade the environment, and in a time frame that meets the inevitable peaking and decline of energy provided from fossil fuels. The leadership position in this technology originated in the United States but has largely shifted to the European Community and to Japan and to China over the last 10–15 years. Although the remaining challenges are daunting, the expertise to overcome them is now developing throughout the world. Changes in the rankings of dominating regions leading this technology will get increasingly expensive as the cumulative world market continues to expand. Those with the wisest and boldest strategies will likely be the major benefactors. The political entities with the highest capabilities for playing major roles in the rapidly evolving world market for solar cell electricity are those backed by the largest economies. In current order of size, these economies are those of (1) the EU, (2) the United States, (3) China, (4) Japan, and (5) India [17]. The market history-demonstrated and the best-estimated public policy approaches for promptly changing rankings among the leaders in the world market, in descending order of effectiveness, are (1) feed-in tariffs; (2) depletable energy user taxes applied to renewable energy development and deployment; (3) government-funded incentives that pay a substantial portion of the capital costs in the first year of installation; (4) long-term (30 years), low-interest rate loans often funded with tax-free bonds; (5) net metering particularly when extended to total metering so that the electricity excess delivered to a utility is paid for at least at 30–40% of the retail delivered costs; and (6) other tax incentives that include loan interest write-offs at constant interest rates over 30-year time periods. Several of these six listed approaches contain features that overlap. Of the three major options for resuming module cost reductions with market growth, key to the large-scale adoption strategy described in this chapter, is the concentrator option that has been the least explored so far and that needs the greatest assistance in making the transition from research into large-scale commercial availability and into clear feasibility demonstrations. Concentrator technology’s failure so far to transition into commercial availability is likely due in part to the “Valley of Death” syndrome described above in the Public Policy section, and that is endemic to all new technologies trying to break into large established markets. The EU is beginning to demonstrate that different countries with adjacent borders, speaking various languages, and with a spectrum of individual cultures, economic development levels, and technology competencies can join into cooperative trade, economic, consumer, and political arrangements that thrust such unions

38

SOLAR CELL ELECTRICITY

into a world leadership position, as ranked by economic metrics in general and by renewable energy implementations and deployments in particular. At its peak, the British Empire demonstrated that a dominant economic power base and trading entity could be assembled from widely spread countries whose participations were established and maintained through military “support.” A key challenge of the twenty-first century may be whether other “adjacent” and “nonadjacent” but diverse countries can cooperate through similar unions motivated primarily by economic self-interest and mutual well-being. For the latter unions to remain competitive with the large and rapidly growing economies like China’s and India’s, the involved “union” populations likely need similar sizes and mobilities and similar lack of restrictions. This latter likely includes ready access to nearly “comparable” education and training for the involved populations plus the latter’s ability to promptly relocate to job and professional opportunities unfettered by overly restrictive or artificial constraints such as those based on nationality, class, race, religion, or sex. As a nation composed essentially of diverse immigrant populations, the United States may have some advantages in the latter areas but with a remaining question of how much immigration is needed, desirable, and politically acceptable. A key to competitive solar cells and systems manufacturing in more developed countries like the EU, the United States, and Japan is automation. The fabrication equipment will likely cost about the same no matter where it is located. A Pacific Rim advantage is lower-cost labor that automation can offset to a large degree. Competitive plant construction costs including land in developed countries can likely be accommodated by appropriate tax breaks and incentives. This would offer reduced employment opportunities in manufacturing, but it would retain the product sales and factory construction revenues so that the involved country could avoid balance of payment deficits that would otherwise develop. “Champion” cells and modules are very important demonstrations of what is possible with current technology, but their performance numbers must be used with caution. When a fabricator has made and tested 10,000 devices, the very “best” by definition occurs with a yield of one in 10,000. Such a performance extreme does not necessarily provide accurate estimates of what can actually be achieved on a large scale in the near term at reasonable costs. In the past, such differences between the average achieved in large installations and in solitary “champion” devices have been as high as a factor of 4X to 5X with the differences exacerbated with attempts to provide low costs using less understood or developed materials systems (see Chapter 1 in Reference 36). When the price paid for solar cells is high using very well understood and developed materials systems, such as with space solar cells that can range in price from $300 per watt to $1800 per watt, the difference between “champion” and average cell performance can drop into the 30% or even lower range (see Chapters 4 and 23 in Reference 36). The terrestrial challenge is to simultaneously provide relatively low costs and high performance. Concentrators are an approach to using “expensive” and well-developed cell materials technologies with potentially low-cost optics to provide such a simultaneous result. However, this approach has not yet been well developed or tested

ABBREVIATIONS

39

experimentally. Early data show that the output of a two-axis tracking module of a high 17% efficiency produces 99% more kilowatt hour per square meter per year in a sunny Palo Alto, California environment in a side-by-side comparison with a near horizontal (5 ° fixed tilt) 12% efficient module (Fraas and Partain, Chapter 26). Recent tests of a 3X concentrator, two-axis tracking module of crystalline silicon interdigitated back contact cells showed that side by side, it had comparable output in kilowatt hour per square meter to a fixed axis module of similar cells in Shanghai, China, but only in measurements for a single day (L. Fraas, pers. comm.). Unfortunately, most of the current concentrator demonstration sites do not include non-concentrator modules, for direct side-by-side comparisons in terms of kilowatt hour per square meter per year outputs. Theoretical calculations of the differences are beginning to become available (Gueymard, Chapter 19). However, such calculations have not yet been carefully validated with experimental data. Their accuracy is unlikely to be trusted until such validations have been completed. Sensitivity analysis illustrates that there is no one parameter whose improvement will lead to the widespread practical implementation of solar cell electricity. However, module cost reductions (in dollar per watt) and efficiency improvements (in percent) will play dominant roles. However, major reductions in inverter costs and improvements in their lifetimes (by factors of 3X to 4X according to the Table 2.2 estimates) will also be needed. Among the other major challenges are economical means for energy storage in combination with time-of-use schemes that will be required for large-scale penetration into the traditional electricity energy market over the longer term. Most of the basic scientific understanding and principles needed to accomplish this are essentially already known. It is the engineering applications and their optimizations and scale-up that remain Herculean in scope and in magnitudes. This a situation largely shared with that of the U.S. Manhattan Project and the U.S. Man-on-the Moon Program in their early days. However, with sufficient forethought, the extremely short time frames and major diversion of resources of the Manhattan Project should not be required. While never predictable, research breakthroughs will accelerate at least some small parts of this effort. It is no longer a question whether this new energy source transition will occur. It is only a question of who will lead the process and who will reap the most benefits.

ABBREVIATIONS AC—alternating electric current CdTe—cadmium telluride DC—direct electric current EIA—U.S. Energy Information Administration EU—European Union EVA—ethylene vinyl acetate FSA—Flat-Plate Solar Array Project

40

SOLAR CELL ELECTRICITY

GaAs—gallium arsenide HIT—heterojunction with intrisic thin-layer solar cell JPL—U.S. Jet Propulsion Laboratory MITI—Ministry of International Trade and Industry (Japan) NEDO—New Energy and Industrial Technology Development Organization (Japan) PMT—name of the Microsoft Excel function that calculates mortgage payments as a function of interest rates and mortgage periods for given loan amounts PURPA—U.S. Public Utility Regulatory Policies Act R & D—research and development SETP—U.S. Solar Energy Technology Mulityear Program X—multiplier REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

J. G. Dorn. Solar Cell Production Jumps 50 Percent in 2007, Earth Policy Institute. Available at http://www.earth-policy.org/indicatiors/solar/2007.htm. Accessed March 9, 2008 (2007). Exxon Mobile Corporation. Annual Report. Available at http://thomson.mobular. net/thomson/7/2677/3201/. Accessed September 6, 2008 (2007). Energy Information Administration, U.S. Government, Existing Generating Units in the United States. Available at http://www.eia.doe.gov/cneaf/electricity/epa/ epat2p2.html. Accessed September 6, 2008 (2007). Energy Information Administration, U.S. Government, Table H1, World Total Installed Generating Capacity. Available at http://www.eia.doe.gov/oiaf/ieo/ieoecg. html. Accessed September 6, 2008 (2008). L. Stoddord and R. Pletka. CEC Workshop on Renewable Energy. Available at http://www.energy.ca.gov/2004_policy_update/documents/2004_06_08_BLACK_ VEATCH.pdf. Accessed September 6, 2008 (2004). R. Wiser and M. Bolinger. Projecting the Impact of State Portfolio Standards on Solar Installaltions, slide 4. Available at http://www.cleanenergystates.org/library/ ca/CEC_wiser_estimates.0205.pdf. Accessed September 6, 2008 (2005). W. Wallace. Government terrestrial acceleration programs. In Solar Cells and Their Applications, L. Partain, ed., p. 495. New York, Wiley (1995). T. Tomita. Prog. Phototovolt Res. Appl. 13, 471–479 (2005). M. A. Green, K. Emery, and D. L. King. Solar cell efficiency tables (version 29). Prog. Photovolt. Res. Appl. 15, 15–40 (2007). V. Salas and E. Olias. Overview of the photovoltaic technology status and perspective in Spain. Renewable and Sustainable Energy Reviews. Elsevier, London. doi:10.1016/j.rser.2008.03.011 (2008). C. Christensen. The Innovator’s Dilemma. Boston, Harvard Business School (1997). E. Christensen. Flat plate solar array project. US Department of Energy Report, Jet Propulsion Laboratory, Pasadena, CA (1985). W. Callahan and R. McDonald. Flat-Plate Solar Array Project Final Report, JPL Publication 86-31, 5101-289, DOE/JPL 1012-125, Jet Propulsion Laboratory, Pasadena, CA (1986). Bureau of Economic Analysis, U.S. Government, Table 1.19 Implicit Price Deflators for Gross Domestic Product. Available at http://www.bea.gov/national/nlpaweb/ TableView.asp#Mld. Accessed September 8, 2008 (2008).

REFERENCES [15] [16] [17]

[18]

[19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

41

E. Martinot, R. Wiser, and R. Hamrin. Renewable Energy Policies and Markets in the United States. Available at http://www.resource-solutions.org/lib/librarypdfs/ IntPolicy-RE.policies.markets.US.pdf. Accessed September 13, 2008 (2005). New York Federal Reserve Bank, US$ Foreign Exchange Rates. Available at http:// www.ny.frb.org/markets/fxrates/historical/fx.cfm. Accessed September 10, 2008 (2008). Central Intelligence Agency, U.S. Government, The World Fact Book, Rank Order—GDP (purchasing power parity). Available at http://www.cia.gov/library/ publications/the-world-factbook/rankorder/2001rank.html. Accessed September 18, 2008 (2008). Fortune magazine. A catalyst for change, the Japan of the future, pp. S10-S11, Special Advertising Section, July 21. Available at http://www.timeinc.net/fortune/ services/sections/customprojects/sections/071126_Japan2.pdf. Accessed September 16, 2008 (2008). Military-Industrial Complex Speech, Public Papers of the Presidents, Dwight D. Eisenhower, 1960, pp. 1035–1040. Available at http://coursesa.matrix.msu. edu/∼hst306/documents/indust.html. Accessed September 27, 2008 (1961). D. E. Kash. Perpetual Innovation, Chapter 6. New York, Basic Books (1989). FY 2008 Budget Request for Defense S&T, FYI. AIP Bulletin of Science Policy News, February 13, 2007, No. 23, pp. 1–2 (2007). SolarBuzz. German PV Market. Available at http://www.solarbuzz.com/fastfactsgermany.htm. Accessed September 6, 2008 (2007). A brief history of the California economy. Available at http://www.dof.ca.gov/ HTML/FS_DATA/HistoryCAEconomy/index.htm. Accessed November 22, 2008 (2008). World Future Council, Feed-in Tariffs. Available at http://worldfuturecouncil.org/ fileadmin/user_upload/Maja/Feed-in_Tariffs_WFC.pdf. Accessed September 16, 2008 (2008). Wikipedia. Feed-in tariff. Available at http://en.wikipedia.org/wiki/Feed_in_Tariff. Accessed November 22, 2008 (2008). SolarBuzz. 2007 World PV Industry Report Highlights. Available at http://www. solarbuzz.com/Marketbuzz2008-Intro.htm. Accessed September 7, 2008 (2008). NEI Nuclear Notes, Italy Nuclear Update. Available at http://www.neinuclearnotes. blogspot.com/2006/02/italy-nuclear-update.html. Accessed September 15, 2008 (2006). China to become world’s largest economy by 2010, Finance Daily. Available at http://financedialy.co.uk/News/Chinatobe WorldsLargestEconomyby2010_448. html. Accessed September 27, 2008 (2008). The World’s Largest Economies in 2050, University of Phoenix, based on 2003 Goldman Sachs study. Available at http://www.everthing2.com/index.pl?node_ id=1756826. Accessed September 27, 2008 (2005). 50th Anniversary of the Interstate Highway System. Available at http://www.fhwa. dot.gov/interstate.history.htm. Accessed August 22, 2008 (2006). US oil import bill to top $400 billion this year. Available at http://www.reuters.com/ article/pressRelease/idUS236508+07-Mar-2008+BW20080307. Accessed September 7, 2008 (2008). Bureau of Economic Analysis, U.S. Government, U.S. International Transactions: First Quarter 2008. Available at http://www.bea.gov/newreleases/rels.htm. Accessed September 7, 2008 (2008). The Cost of Iraq, Afghanistan, and Other Global War. Available at http://zfacts.com/ metaPage/lib/CRS-Belasco-2006-09-Iraq-Costs-RP33110.pdf. Accessed September 7, 2008 (2008).

42 [34] [35] [36] [37] [38] [39] [40]

[41]

SOLAR CELL ELECTRICITY Interstate Cost. 50th Anniversary of the Interstate Highway System Frequently Asked Questons. Available at http://www.fhwa.dot.gov/interstate/faq.htm# question6. Accessed November 20, 2008 (2006). S. Thompson and S. Pathasarathy. Materials Today 9, 20–25 (2006). L. D. Partain, ed. Solar Cells and Their Applications. New York, Wiley (1995). Earth radius. Available at http://en.wikipedia.org/wiki/Earth-radius. Accessed November 22, 2008 (2008). Energy Information Administration, U.S. Government, Table H6, World Installed Hydroelectric and Other Renewable Generating Capacity. Available at http://www. eia.doe.gov/oiaf/ieo/ieoecg.htm. Accessed September 6, 2008 (2008). F. R. Goodman, J. C. Schaefer, and E. A. DeMeo. Terrestrial, grid connected systems. In Solar Cells and Their Applications, Chapter 16, L. D. Partain, ed. New York,Wiley (1995). US Department of Energy, Solar Energy Technologies Program, Multi Year Program Plan 2007-2011. Available at http://www.pv-era.net/doc_upload/documents/218_0 084aSolarEnergyTechnologiesProgram2007-2011_proof.pdf. Accessed October 12, 2008 (2007). US Department of Energy, Solar Energy Technologies Program, Multi Year Program Plan 2008-2012. Available at http://www1.eere.energy.gov/solar/pdfs/ solar_program_mypp_2008-2012.pdf. Accessed October 26, 2008 (2008).

3 SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS, AND HIGH EFFICIENCY LEWIS FRAAS JX Crystals Inc.

3.1

INTRODUCTION

How efficient can a solar cell be and how much power and energy can it produce and at what cost? These are the fundamental questions addressed in this chapter. In order to address these questions, it will be necessary to first describe the nature of sunlight and then the nature of semiconductors. It will be shown that singlecrystal semiconductors are necessary for high sunlight conversion efficiency and that high conversion efficiency is important for lower-cost solar electricity. In addition to silicon solar cells, a new class of semiconductors based on LED materials will also be discussed. The reader is familiar with LEDs and the fact that they come in different colors. It is noted herein that LEDs are just solar cells running in reverse. In other words, in a LED, electricity goes in and light comes out. In a solar cell, light goes in and electricity comes out. It will be shown that the LED class of semiconductors can be used to make multicolor or multijunction solar cells where the multicolor feature is critical for making solar cells with energy conversion efficiencies as high as 40%. While it is true that high-efficiency single-crystal solar cells are more expensive to make than amorphous or small grain size polycrystalline thin-film cells, fortunately, it is possible to use inexpensive mirror or lens materials to collect the sunlight and to focus the solar energy on these single-crystal converters thereby diluting their cost. This is the concentrated sunlight PV or CPV approach. The newer 40% cells are even more complex and expensive than single-crystal silicon Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

43

44

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

cells requiring higher-concentration HCPV systems, whereas the silicon singlecrystal cells can use lower concentration LCPV simpler systems. This chapter will then conclude with a discussion of the future cost potential for the various solar PV options comparing the potential future system-level cost for the conventional planar silicon cell and thin-film cell approaches with the concentrated sunlight approaches.

3.2

SUNLIGHT, RAINBOWS, AND PHOTONS

How much energy is in sunlight? The answer to this question requires a definition of some terms. The three terms are power, solar intensity, and energy. Power is in watts and solar intensity is in watts per unit area. Energy is in kilowatt hours. Homeowners pay for energy in kilowatt hours. Solar modules and arrays produce power in watts. The relevant measure of solar radiation is solar intensity. On a nice, sunny day at noon, the solar intensity is usually around 1 kW/m2. One square meter is close to 11 ft2. How does a scientist describe sunlight? The observation of rainbows proves that sunlight can be divided into different colors. Also, when a group of very fine parallel lines are scribed close to each other to make a grating, it is observed that the colors can be correlated with line spacing. This means that there is a wavelength connected to each color. So, light is an electromagnetic wave as shown in Figure 3.1, just like radio waves and microwaves with a wavelength, λ, and electric field, E. The history of the study of light is interesting. Actually, Newton thought of light as particles. However, with grating experiments in the 1800s, it was decided that light was an electromagnetic wave. Then, in about 1905, Einstein looked at the photoelectric effect and said that light comes in small energy packets called

E

Wave length

Figure 3.1. Electromagnetic wave.

SUNLIGHT, RAINBOWS, AND PHOTONS

45

quanta, which behave somewhat like particles. Einstein won the Nobel Prize for this work [1], not for his theory of relativity. Einstein noted that when the light of a certain color hits a metal, all of the electrons that escape the metal have the same peak energy and that when the light intensity is increased, the number of electrons increases, not the energy of each electron. It is like the energy in wave theory is E2, but the energy also equals nλ × eλ, where nλ is the number of photons and eλ is the energy of each photon. So, one can think of the electromagnetic wave intensity as getting smaller and smaller until one discovers that it is coming in discrete particles. It occurs in very fine steps. So now, the sun’s intensity spectrum can be divided into color or wavelength intervals as in Figure 3.2. To understand how solar cells work, first note that a photon’s energy is inversely proportional to its wavelength. This just means that shorter wavelength photons at the left in this curve have more energy than the longer wavelength photons at the right. Electric power is the product of voltage and current. In a solar cell, the power results when electrons pass through a voltage. It is the nature of semiconductors that each semiconductor material has a threshold absorption energy that then controls the voltage the electrons see. It is convenient then to talk in units called electron volts. An electron volt is the energy that an electron produces when it moves through a potential of 1 V. Photon energies and material threshold absorption energies are measured in electronvolt. For example in Figure 3.2, a red photon with an energy of 2 eV has a wavelength of about 0.6 microns, and the absorption threshold energy for silicon, called its bandgap energy, is at 1.1 eV, which equates to 1.1 microns (1 micron = 1 micrometer = 1 millionth of a meter = 1 μm). Now the efficiency limits can be seen. First, in Figure 3.2, all photons with energies less than 1.1 eV are lost; that is, all solar energy with wavelengths greater 1600

Watts/m2/micron

1400 1200 1000 800 600 400 200 0

0

0.5

1

1.5

2

2.5

Wavelength (microns)

Figure 3.2. Sun’s spectrum at AM1.5. The gray region is the usable photon energy for a silicon solar cell.

46

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

than 1.1 microns is lost because these photons are not absorbed. Next, for photons with energies large enough to be absorbed (left side of figure), each photon absorbed must create an electron that can move to generate voltage. The yield here should be one electron per photon, but the actual number is less and is called the quantum efficiency or QE. QEs of 90% occur in pure single-crystal materials. QEs in non-single-crystal materials are lower. So, single-crystal semiconductors are important for high efficiency as will be discussed in the next sections of this chapter. For now, high QEs in single-crystal materials are assumed. These moving electrons create the current. To calculate the current, the photons absorbed are simply counted. However, note that while a photon might have 2 eV of energy, it can only create an electron moving through a voltage of no more than the bandgap energy. For silicon, this is 1.1 eV. So, 2-eV photons loose at least 2.0 – 1.1 = 0.9 eV. The gray region in Figure 3.2 shows the energy that can be captured from the higher-energy photons. This now leads to a way to make higher-efficiency solar cells using multiple materials. Two stacked materials can absorb the lower-energy photons in one material generating a lower voltage and the higher-energy photons in a second material generating a higher voltage. The result is more photons better used. This concept is shown in Figure 3.3. The gray regions in Figure 3.3 show that more photon energy can now be captured. To implement this two-junction or two-color cell concept, one must choose materials wisely. The GaAs/GaSb two-color cell (2) demonstrated in 1989 uses two simple materials with bandgap energies of 1.4 eV (0.9 micron) and 0.7 eV (1.8 micron). Theoretically, the efficiency for this pair can be as high as 41%. Thirtyfive percent has been demonstrated [2].

1600

Watts/m2/micron

1400 1200 1000 800 600 400 200 0

0

0.5

1 1.5 Wavelength (microns)

2

2.5

Figure 3.3. GaAs/GaSb and solar spectrum. The GaAs absorbs higher-energy photons at the left, generating higher voltage. The GaSb absorbs lower-energy photons at the right (1 micron = 1 micrometer = 1 millionth of meter).

ELECTRONS IN ATOMS AS WAVES AND THE PERIODIC TABLE

47

For the more general reader who may not care about the chemical names, note that the top GaAs cell actually looks blue and the bottom GaSb cell actually looks red. So, one can also refer to the cells in multicolor cell stacks as blue cells glued on top of red cells. The chemical terms will be used through most of this book for the benefit of technical readers because the chemical names are more precise. More information on multijunction concentrator cells will be presented later in this chapter and in Chapter 13.

3.3 ELECTRONS IN ATOMS AS WAVES AND THE PERIODIC TABLE OF THE ELEMENTS There are several different types of solar cells made from materials ranging from single crystals to amorphous silicon. The goal here is to describe the different types of solar cells and their advantages and limitations. A fundamental description of the nature of semiconductors is presented, beginning with electrons in atoms as waves. The discussion of electrons as waves then leads to a description of semiconductors as single crystals. The theory of single-crystal semiconductors is then used to describe how diodes and solar cells work. A discussion of the effects of various defects in semiconductor materials on solar cell performance follows. The reader will see that the performances enumerated are consistent with the simple concepts presented. This chapter explains why high-efficiency cells require good single-crystal materials. In the last section, it was noted that the sun’s rays are really electromagnetic waves with varying wavelengths. Electromagnetic radiation includes radio waves, microwaves, and infrared, visible, and ultraviolet waves. When one thinks about longer wavelength radiation like radio waves, one always thinks about waves. However, for the shorter wavelengths associated with infrared and visible light, physicists start to talk about photons. A photon is like a particle or wavelet having a specific wavelength and energy. A photon is a quantum of energy or discrete packet of energy. Now, is radiation a wave or a particle? The answer is both! This is the wave–particle duality, a subject called quantum mechanics [3], a subject normally taught in graduate school physics classes along with a lot of mathematics. However, the key ideas can actually be described in simple nonmathematical terms, and these ideas are important to the understanding of solar cells. While electromagnetic radiation is normally thought of as waves, one generally thinks of electrons as particles circling an atomic nucleus just as planets circle the sun. However, an atom is really extremely small, so small that in crossing a human hair, one will pass by 200,000 atoms. Intuition based on everyday experience fails at this small size. It turns out that electrons around atomic nuclei are described by wave functions. Here is the wave–particle duality again. However, the rules that govern electrons in atoms and solids can be described in fairly simple terms. In Figure 3.4, start with the simple hydrogen atom [4, 5] with a single negatively charged electron and a single positively

48

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Nucleus

D Px, Py, Pz

4103 A 4342 A 4563 A

S

E X

6565 A

Figure 3.4. Left: potential well for an electron around the nucleus in an atom with energy level S, P, and D wave functions. Right: a spectral line sequence for hydrogen.

charged proton. The oppositely charged proton and electron attract each other and as they get closer and closer to each other, it is harder and harder to pull them apart. The electron is said to be in an energy well or a potential well as shown on the left of Figure 3.4. The question then is can the electron collapse down and sit on the proton? The answer is no. Why not? Scientists have observed the electromagnetic spectra emitted by atoms and find discrete wavelengths and energies as shown on the right in Figure 3.4. Not all energies are possible. How is this explained? Scientists hypothesize that the electron position is described by a wave function that then gives its probable position. Since one knows that the electron cannot be outside the potential well, one knows the wave functions have to be zero outside the well. Now, observe that the waves will have to have one, two, and three nodes as is shown in the wells at the left in Figure 3.4. For historical reasons, the state with one peak node is labeled S, and the states with two nodes are labeled Px, Py, and Pz. (x, y, and z are the three directions in three-dimensional space.) The next rule is that electrons can have positive and negative spins, and only one electron can occupy each state. So there will be two S states with opposite spins and two Px, two Py, and two Pz states for a total of eight state configurations possible. This wave hypothesis has proven to be very successful as it explains atomic spectra and the periodic table of the elements [6] and all of chemistry. The rule of eight including S and P orbits explains the second and third rows of the periodic table. Table 3.1 summarizes important features of the periodic table including the common commercial semiconductor materials. The D level transition metals are not shown since they are not relevant here. Compounds are formed from the elements in the periodic table. For example, table salt is a compound containing Na (sodium) and Cl (chlorine) and is written as NaCl. The two semiconductor

SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY

49

TABLE 3.1. Periodic Table of the Elements I

II

III

IV

V

VI

VII

H Hydrogen

VIII He Helium

Li Lithium

Be Berilium

B Boron

C Carbon

N Nitrogen

O Oxygen

F Fluorine

Ne Neon

Na Sodium

Mg Magnesium

Al Aluminum

Si Silicon

P Phosphorus

S Sulfur

Cl Chlorine

Ar Argon

Ga Gallium

Ge Germanium

As Arsenic

In Indium

Sb Antimony

compounds making up the two color solar cell stack described in Figure 3.3 are gallium arsenide (GaAs) and gallium antimonide (GaSb).

3.4 SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY Why is it important to know about electrons as waves? The answer is that waves are intrinsically periodic as are the atom locations in single crystals. It is this periodicity that makes semiconductors special. Historically, the semiconductor revolution started 80 years ago with the discovery of the importance of high-purity single crystals and the wave theory of solids. The first application of the electron wave theory was by A. Sommerfeld in 1928 [7]. As shown in Figure 3.5a, he described the motion of electrons in metals by assuming the electrons moved in a flat-bottom energy well bounded by the metal surfaces. Any wavelength would be possible, leading to a set of conduction band energy levels. The real breakthrough came in 1928 with F. Bloch [8], who then modeled a periodic single crystal with a set of periodic energy wells representing the atom positions within the larger energy well as shown in Figure 3.5b. He showed that there are then two or more energy bands separated by energy gaps. The states with wave functions centered on the atomic wells represent valence electrons, and the states with wave functions centered between the atomic wells represent conduction band electrons. A key point is that any electron can be near any atom in the crystal. A. H. Wilson [9] in 1931 then provided the elementary theory of semiconductors by noting that if the material is pure enough and if the valence band is completely full and the conduction band is completely empty, then one has a semiconductor. An intrinsic semiconductor is defined as having zero conductivity at zero temperature with increasing conductivity as the temperature increases. The increasing conductivity comes about as electrons are thermally excited into the conduction band. This is opposite from metals where the number of conduction

50

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Ec (a) Metal

Ec Ev (b) Semlconductor

Figure 3.5. Solid potential energy and energy band models for metal and semiconductor.

band electrons is fixed and collision frequency increases with temperature decreasing electrical conductivity as the temperature increases. A key point here is that there is an energy gap between the valence band and the conduction band and that this energy gap derives from the periodicity, which derives from the single crystal. This history is fascinating. However, the goal here is to explain the reasons why single crystals are important to solar cells and to probe the question of how pure and how perfect solar cell materials need to be. Most importantly, how can solar cells be integrated into systems to generate electricity on a large scale at a competitive price? Before describing semiconductors in more detail, let us return to the periodic table and contrast the semiconductors with metals and insulators to see why semiconductors are special and why they are needed to make solar cells. To preview the answer, note that in order to deliver electric power, a solar cell needs to generate both current and voltage. Generating current requires electron mobility and generating voltage requires a gap between electron energy states. Metals have electron mobility and insulators have gaps between energy states, but only semiconductors have both. The metals like sodium and magnesium are on the left in the periodic table. These atoms have only a few loosely bound electrons each, and they can be tightly packed with up to 12 nearest neighbors. Because the atoms are closely packed, the potential energy well for a metal looks like a flat-bottom well with the well extending to its surfaces. As shown in Figure 3.5a, the metal surfaces form the energy barriers confining the electrons. Because this well is so large compared to one atom, all electron wave function wavelengths and energies are possible. Electrons are then free to move around in the metal, but there are no energy gaps between energy states. Since the electrons hardly feel the metal atom core positions with the flat-bottom potential well, crystallinity is not important to metallic properties.

52

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

corresponding wavelength. Thus, the two types of states with bonding and antibonding electron locations between nearest silicon pairs or farthest silicon pairs are the only states allowed. There is an energy gap between these states because no other electron wave functions are allowed. The states representing the bonding states form what is called the valence band and the states representing the antibonding states form what is called the conduction band. Figure 3.6 also shows the energy potential and wave functions for a group III–V semiconductor. The III–V semiconductors are the materials used to make LEDs. In this case, a group III element like gallium can form tetrahedral bonds with a group V element like arsenic where the result is the sharing of four electrons per atom as in silicon. Group III–V is a rich class of semiconductors. It turns out that because of the crystal periodicity, there is both energy gap and electron mobility in semiconductors. Figure 3.7 allows one to visualize this more easily. In this figure, both connected bonded regions and open channels in between can be seen. One can imagine electrons traveling in the bonded regions or separately in more energetic states in the open channels. Propagating electrons in the bonded region have energies in a valence band, and propagating electrons in the open channels have energies in a conduction band. The separation between these regions provides the energy gap. Looking at Figure 3.7, one can also imagine a large foreign atom or a crystal boundary or defect interfering with flow in the channels or a total disruption of the channels smearing the two sets of energy states into each other.

Figure 3.7. A view of a channel open for conduction electron movement in a GaAs single crystal. Small sphere: gallium atom; large sphere: arsenic atom; white cylinders: valence bonds.

JUNCTIONS AND DIODES

53

Figure 3.7 suggests intuitively that electrons will have higher mobility in single crystals than in amorphous or small crystal size thin films. This is in fact true quantitatively. Electron mobility is easily and routinely measured. The electron mobility in single-crystal silicon is typically 1500 cm2/Vs, and in single-crystal GaAs, it is 4500 cm2/Vs. However, in amorphous silicon and copper indium diselenide (CIS), two common thin-film solar cell materials, it is only 4 cm2/Vs. This is a difference by a factor of 1000 consistent with our intuitive expectations based on Figure 3.7.

3.5

JUNCTIONS AND DIODES

It has now been established that carriers are mobile, allowing current to flow in solar cells. How does one use an energy gap to create voltage? A P/N junction (P = positive, N = negative) is needed. In the above description of electron movement in semiconductors, one now needs to note that it is important to count electrons. If the semiconductor is very pure (a state called intrinsic), then all of the bonding states will be occupied by electrons and there will be no electrons to move in the conduction band. Electrons cannot move in the valence band either because there are no empty spaces to move to. Substituting a small number of phosphorus atoms for silicon atoms can rectify this problem (one in a million). Since phosphorus is from group V, it has one more electron than silicon. The resultant material is labeled N-type because the extra electrons are negatively charged. Alternately, as a complement to the N-type material, one can substitute an aluminum atom for a silicon atom leaving the bonding or valence band one electron deficient because aluminum from group III has one less electron than a silicon atom. Now instead of thinking about a million electrons in the valence band, we talk about the missing electrons in the valence band. We call this a hole. It is like watching a bubble move in water. The hole has a positive charge and we call this material P-type. Now what happens when N- and P-type materials are brought together? The result is a P/N junction diode [10–12] as shown in Figure 3.8. The band edge diagrams at the bottom of this figure describe how a diode works. When the P and N regions first come together, the electrons and holes from each side diffuse together, eliminating each other, leaving an electric field region in the junction. This happens until the valence band edge (v) in the P material almost lines up with the conduction band edge (c) in the N material as shown on the left in this figure. At this point, the free electrons and holes on both sides of the junction have the same energy as shown by the dashed horizontal line. This is the zero voltage band diagram (Fig. 3.8A). Now notice that there is an energy hill for electrons to climb in order to move from the N to the P side of the junction. An applied voltage can either decrease this hill or energy barrier for forward bias (Fig. 3.8B) or increase it in reverse bias (Fig. 3.8C). If the hill is made small enough by a forward voltage about equal to two-thirds (67%) of the bandgap energy, Eg, then current starts to

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Current (Amps)

54

Current PN y

z

x

c Eg v E

c

P v

x (a) Zero voltage

c Eg v

(A)

(C)

Voltage (V)

Forward current

c Eg v

Metal Contacts

N

(B)

P

Electron c Applied N voltage v

P

N

c v

(b) Forward voltage

(c) Reverse voltage

Figure 3.8. Upper left: P/N junction diode; upper right: current versus voltage for P/N diode; lower left (A): conduction band minimum and valence band maximum positions through P/N junction at zero applied voltage. Lower middle (B): forward voltage band diagram—reduced barrier for high current flow. Lower right (C): reverse voltage—barrier blocks current flow.

flow. This corresponds to the knee in the diode current versus voltage curve shown at the top right in this figure. In reverse bias, no current flows because the barrier just gets bigger. Thus, a diode is a rectifier allowing current flow in only one direction.

3.6

SOLAR CELL BAND DIAGRAMS AND POWER CURVES

Referring now to Figure 3.9, a solar cell is just a large P/N junction diode with a metal grid on its front side facing the sun. A solar cell converts the energy in sunrays to electric power. Now we shall refer to the sunrays as photons. In Figure 3.9, the now familiar band edge diagrams are shown at the bottom. These band edge diagrams show how a solar cell works. First, a photon is absorbed exciting an electron from the ground state or valence band in the P material to an excited conduction band state. It is mobile in the conduction band and if it lives long enough in this excited state, it can diffuse to the junction and fall down the potential barrier. Another way of thinking about this potential barrier is simply that it represents an electric field region created by the initial separation of electron and holes when the junction was formed. Anyway, when an electron enters a field region, it gains electrical energy. This can be converted to a voltage and current to do work.

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

Sun ray Current (Amps)

Voltage (V)

P N

x

Eg

c c P

x

Max power point

Light generated current Forward dark current

c

E

Voc

Isc

Light generated current

v

55

Eg

N

Voc

v v

Short circuit current (Isc) at zero voltage

P Open circuit voltage (Voc) at zero current

Figure 3.9. Upper left: P/N junction solar cell with metal grid on top. Lower left: photon absorption excites electron into conduction band. Electron then falls through junction potential. Upper and lower right: current versus voltage curve for solar cell is diode I versus V curve moved down by light-generated current.

3.7

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

How efficient can a solar cell be and how do we achieve these high efficiencies? Theoretically, a solar cell efficiency of 70% is possible. However, no one believes that, in practice, this can be achieved. Still, a 35% efficient solar cell has been demonstrated and 45% is probably an achievable target. What needs to be done to achieve high efficiencies is a more interesting question. In fundamental terms, three things need to be done. First, for each photon absorbed, the excited state carrier generated needs to last long enough to be collected at the junction. Second, while the sun’s spectrum contains photons of different energies, the energy available in each photon must be used as wisely as possible. And third, the voltage a cell generates should be as close as possible to the bandgap energy. We will discuss each of these requirements in succession in the following paragraphs. The first requirement of one electron collected for every photon absorbed implies single-crystal material and high-purity material. The measure of electrons collected per photon absorbed is called quantum efficiency. Figure 3.10 provides a semiquantitative answer to the semiconductor purity question. To understand Figure 3.10, let us go back to the crystal channels shown in Figure 3.7. First, how far will an electron move through one of these crystal channels? The answer is about 100 atomic spaces. This is because the atoms are not really stationary but

56

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Random walk

Light absorption

Junction

Figure 3.10. A light-generated carrier diffuses to the junction in a random walk sequence.

are vibrating small distances around their home positions because they have thermal (heat) energy. This vibration energy is small, however, so that the excited electron does not return to the valence band but just gets deflected into another channel. We think of this deflection as a step in a random walk diffusion problem. This brings us back to Figure 3.10. The next question is how far is the excited-state electron away from the junction? This depends on the photon absorption distance. This absorption distance depends on the material and on the rules for photon absorption. Now we shall divert for a minute to the rules for photon absorption. This will be important because, as we will see, silicon is fundamentally different from the III–V semiconductors in its photoelectric properties. Looking back to the hydrogen atom in Figure 3.4, a rule for photon absorption is that the wave functions involved have to have different symmetries. For example, note that the S and D wave functions are symmetric around the position of the nucleus, while the P functions are antisymmetric. Thus, absorption between S to P and P to D are allowed, but S to D is not allowed. Now let us look at the wave functions for silicon and GaAs in Figure 3.6. Note that both wave functions for silicon are symmetric around the point between two silicon atoms. This means that photon absorption in silicon is not allowed to first order. In GaAs, however, photon absorption is allowed. So the photon absorption length in GaAs is about 10,000 atomic spaces. In reality, photons are also absorbed in silicon but in about 100,000 atomic spaces. This second-order absorption in silicon results because of atomic thermal vibrations. Now, returning to the purity question and the random walk diffusion problem, remember that a step length is about 100 atomic spaces. So a carrier in GaAs will

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

57

be about 100 steps away from the junction, and a carrier in silicon will be about 1000 steps away. However, in a random walk problem, the number of steps required to move N steps away from the start is N × N steps. So, the distance an excited electron must travel to the junction in GaAs will be 10,000 steps or one million atomic spaces. If it were to see a large impurity in a channel on this path, it could return to the valence band and be lost. So the purity requirement for GaAs is about 1 ppm. The analogous argument for silicon suggests a purity requirement of 10 ppb. This is 1/100,000,000. In fact, silicon solar cells lose performance given transition metal impurities in the range of several parts per billion. The above argument has been a little tedious, but the goal is to impress the reader with this purity requirement. By analogy, it should also be clear that good single-crystal quality without defects is as important as purity. The above purity specification is routinely met in commercial single-crystal silicon solar cells today as well as in various other single-crystal silicon-based devices that have revolutionized our lives over the last 50 years. While the reader is probably not aware of it, various single-crystal III–V devices have penetrated our everyday lives as well in the last 10 years. As the above argument about the difference in photon absorption for GaAs versus silicon suggests, III–V are often a better choice for photoelectric and optical-electronic applications. Referring to the periodic table, there are a large number of III–V materials available including GaAs, InP, InSb, and GaSb. Additionally, alloys of these materials are available including AlGaAs, GaAsP, GaInAs, InGaP, and InGaAsP. This makes a large set of bandgaps and electron mobilities available. Single-crystal III–V devices can now be found in cell phones, satellite receivers, CD music players, CD-ROMs in personal computers, taillights in cars, traffic stoplights, and military weapon systems. Single-crystal III–V devices are also key components in fiber optic phone communication and the Internet. In fact, the most efficient solar cells are made using III–V materials. This brings us back to our second requirement for making high-efficiency solar cells. We need to use the energy in the sun’s varied colored rays as efficiently as possible. A problem with sunlight is that the photons come in different colors with different associated energies. If we wanted to maximize the efficiency of a photodiode, we would illuminate it with only photons with a single energy with an energy equal to the bandgap energy Eg. Then, if the crystal quality and purity were sufficient, all of the excited carriers would be collected at the junction with 67% of the photon energy being delivered as a voltage. The energy conversion efficiency would be roughly 67%. However, referring to Figure 3.2, photons from the sun come with different energies. Some of the photons have too little energy to be absorbed, and some of the photons have energy considerable in excess of the bandgap energy. For the sun’s spectrum, this limits the single-junction solar cell efficiency to less than 30%. Fortunately, group III–V offers a solution because various materials with various bandgap energies are available. Specifically, one can stack a visible light-sensitive GaAs solar cell with metal grids on its front and back on an infrared-sensitive GaSb solar cell to arrive at the two-color or two-junction solar cell shown at the right in

58

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Figure 3.11. In this way, one absorbs the high-energy photons first in the top material generating a high voltage, while the low-energy photons pass through the top cell to be converted in the bottom cell. More photons are used and they are used more wisely. This then is the 35% efficient GaAs/GaSb two-color or two-junction solar cell demonstrated by L. Fraas et al. in 1989 [2].

Silicon Eg=1.1 eV

Ga Sb Eg=0.7 eV

Ga As Eg=1.4 eV

Figure 3.11. Left: for single-junction solar cell, sunlight contains high-energy photons with excess energy and low-energy photons with too little energy. Right: solar spectrum can be more efficiently utilized by stacking two different junctions together. Three junction cell

First Second Third

n+ n p p++ n n p+ p+ n n

Window InGaP GaInAs

Ge

p

From p. 888 (at 300 suns AM1.5) 13.3 mA/cm2 x Vop η = --------------------------------84 mW/ cm2 Vop= 2.55 V Efficiency = 40%

Active solar cell junction Shorting tunnel interconnect

Figure 3.12. Monolithic triple-junction InGaP/GaInAs/Ge CPV cell as first described by L. Fraas and R. Kinechtli in the 13th IEEE PV Specialist Conference, predicting 40% at 300 suns AM1.5 [13].

PV MODULE AND SYSTEM COST TRADES

59

Concentrating lens

Solar Cell Heat spreader back plane

Light Generated Current (Amps)

Sun light Sunlight concentrated

Sunlight not concentrated

Voltage (V)

Figure 3.13. Solar cells are more efficient with concentrated sunlight because both current and voltage increase.

As first predicted in 1978 [13], an efficiency of 40% has recently been demonstrated [14] for the monolithic three-junction InGaP/GaInAs/Ge CPV cell as shown in Figure 3.12. This brings us to the third way of increasing solar cell efficiency. For a given bandgap energy, we want to generate more voltage. Concentrating the sunlight onto the cell can do this. This is shown in Figure 3.13. Sunlight can be concentrated using a lens as is shown at the left in this figure. The resulting currents versus voltage curves with and without a lens are shown at the right. As is customary for solar cells, the diode curves here have been flipped over. Note that the higher current concentrator cell has a higher efficiency. This is because the diode is being driven harder to a higher current and voltage. In other words, if the light level goes up by 10, the current also goes up by 10 but at the same time, the voltage also goes up. In practice, the open-circuit voltage can go up from about two-thirds of Eg to about three-quarters of Eg under concentrated sunlight.

3.8

PV MODULE AND SYSTEM COST TRADES

So it is clear that single-crystal cells have higher efficiencies than amorphous or small grain size thin-film cells and that multijunction cells have still higher efficiencies. However, single-crystal cells are more expensive than thin-film cells and multijunction cells are very complex. It is also clear that CPV systems see only direct sunlight, which is less than global sunlight. So, it is necessary to look at the cost of complete PV systems. This is shown in Table 3.2. Note that the attached table has five columns comparing system-level costs measured in terms of simple payback times. The columns summarize future (ca.

60

TABLE 3.2. Comparison of Future Economics for Planar PV and Concentrated Solar PV Module Type

Thin Film

SC Silicon

LCPV

HCPV

HCPV JXC

Cell Efficiency @ temperature

9%

19%

22%

35%

44%

Annual available irradiancea

2336 kWh/m2 (fixed tilt)

2905 kWh/m2 (tracking 1-axis)

2382 kWh/m2 (1-axis; 87% global; 94% optical efficiency)

2178 kWh/m2 (2-axis; DNI = 78% global; 85% optical efficiency)

2178 kWh/m2 (2-axis; DNI = 78% global; 85% optical efficiency)

Annual kWh/m2 electricity

210 kWh/m2

552 kWh/m2

524 kWh/m2

762 kWh/m2

936 kWh/m2

Dirt penalty

–5% = >200 kWh/m2 –5% = >525 kWh/m2

–7% = >488 kWh/m2

–9% = >695 kWh/m2

–9% = >852 kWh/m2

Annual revenue at 10¢/kWh

$20/m2

$52.5/m2

$48.8/m2

$69.5/m2

$82/m2

Cell cost per m2 module

$50/m2

$300/m2

$100/m2

$100/m2

$137/m2

Module materialb

$50/m2

$60/m2

$67/m2

$67/m2

$67/m2

Optics

0

0

$57/m2

$65/m2

$65/m2

Module cost

$100/m2 ($1.11/W)

$224/m2 ($1.20/W)

$232/m2 ($0.92/W)

$360/m2 ($1.89/W)

$75/m (1-axis)

$150/m (2-axis rigid)

$150/m2 (2-axis rigid)

$25/m2

$25/m2

$25/m2

$40/m2

$40/m2

$160/m2

$460/m2

$324/m2

$422/m2

$459/m2

$35/m

Installation Total module area cost

2

2

2

$269/m2 ($0.85/W)

$75/m (1-axis)

Array support structure

2

62

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

2015) projected system costs for thin-film PV systems, single-crystal silicon flatplate systems, a JX Crystals 3-sun LCPV evolutionary system, an HCPV system using today’s 35% multijunction cells, and a JX Crystals HCPV Dual-Focus Cassegrainian system using a 44% efficient cell combination. These systems will be described in more technical detail in Chapters 12–17. Each column lists projected module outputs in annual kilowatt hour per square meter for Las Vegas, and each column summarizes projected system-level area-related costs in dollar per square meter. These costs include not just module cost but also support structure and installation costs. Given annual projected energy outputs and total area-related costs, it is then possible to calculate a simple system payback time given an assumed energy value of 10 ¢ per kilowatt hour. The good news is that all five columns for all PV technologies promise a simple payback time of less than 10 years. These systems should last for over 20 years. The first column is for the thin-film PV case. This case suffers from a low cell efficiency of 9%. Module costs are assumed to be those stated by First Solar and other thin-film PV companies of only $1.11 per watt. However, even given these low module level costs, because of the low cell efficiencies and lower-energy production per square meter, it takes a long time to pay off the cost of the array support structures and system installation. The projected simple payback time is 9.4 years. The second column is for the conventional single-crystal silicon planar module case. Here, the good news is the 19% cell efficiency, which then leads to 2.5 times more annual energy production per square meter of module area compared to the thin-film case. Simple one-axis tracking is one of the benefits here, giving more kilowatt hour per kilowatt installed. Trackers, unfortunately, are not affordable for the thin-film PV case. However, the cell cost is much higher than the thin-film case. Note that the cell cost assumed here is a future projection of $1.58 per watt. Given a higher-energy output, the field-related area costs are paid off faster than for the thin-film PV case, and the simple payback time works out to be 9.8 years. The third column is for an evolutionary LCPV case as will be described in Chapter 12. This case assumes the same 1-sun single-crystal cells and module manufacturing process as for the second column case. The three changes are the cut-up of the 1-sun single-crystal cells into thirds, the addition of low-cost linear mirrors, and the addition of a thin aluminum sheet heat spreader at the back of the laminated standard module. This approach has already been demonstrated in the field. The immediate benefit here is low risk, straightforward manufacturing, and a reduction of the cell cost by a factor of 3. Additional benefits are the ability to utilize more than just direct sunlight on less than perfect sunny days and higher optical throughput. The same simple one-axis trackers can be used as for the planar silicon module case. The reduction of the cell cost relative to column 2 with no additional penalties leads to a lower simple payback time of 7.8 years. The fourth column is for the HCPV case using today’s 35% multijunction cells. Here, of course, the 35% cell efficiency is spectacular. However, there are some negatives. One negative is that these HCPV systems only see direct sunlight

THE IMPORTANCE OF SINGLE CRYSTALS

63

(DNI) and not global illumination. While DNI can be as high as 90% of global illumination on the best blue-sky day, on an annual average even in Las Vegas or in Phoenix, it is only 78% of global illumination. A typical optical throughput of 85% for HCPV is also a negative. These two negatives make a 0.85 × 0.78 × 35% = 23% silicon planar cell competitive. However, the hoped-for good news for the HCPV approach is a lower module cost because of the cheaper optical materials relative to the high-efficiency single-crystal solar cells. If the low-cost modules ($0.92 per watt assumed here) and if the low-cost precision tracking ($150 per square meter assumed here) can be demonstrated in high-volume production, then this system will beat the planar system in sunny locations with a simple payback time of 7.2 years. However, this approach is more risky and will take longer to build up the manufacturing infrastructure than for the LCPV case. In the still longer term, column 5 presents an even higher-efficiency HCPV approach as will be described in Chapter 15. This approach uses a recently announced improved triple-junction monolithic cell (39% efficient) at the focus in the center of a Cassegrain optical configuration. A second infrared cell is also used behind a secondary mirror with a dichroic coating allowing the infrared to get to the second cell. The second GaSb infrared cell adds an addition 5% to give a cell combination efficiency of 44%. The GaSb cell, which is a diffused junction cell [16], is relatively inexpensive to make. Even with the cost of this second cell, its added efficiency leverages down all of the system cost to give an even more attractive payback time of 6.8 years. This approach, while all of the separate components have already been demonstrated, will require a lot of time and money to bring into high-volume production.

3.9

THE IMPORTANCE OF SINGLE CRYSTALS

From the above discussion of system-level cost trades, why have CPV systems not received more attention and why have thin film systems received so much attention? One answer is that searching for a 20% efficient low-cost thin-film solar cell is a very attractive dream. However, in this chapter, we have talked about electrons as waves and semiconductors as crystals to convey the message that this dream is not well founded on scientific principles. In fact, in graduate school solid-state physics classes, the bandgap in semiconductors is rigorously derived based on the assumption of the perfect periodic single-crystal lattice. However, the importance of single crystals to semiconductor devices is not generally conveyed in a simple understandable way. It is certainly not knowledge available to funding sources or the financial community. Figure 3.14 is an attempt to rectify this situation by making an analogy between an electron traveling in a solid and a car traveling through a forest. Organizing the atoms in single crystals is like removing the trees to make a road through a forest. Atoms out of place or atomic impurities are obstacles for the electron just like trees are obstacles for a car. Collisions with these obstacles force the electron (or the car) to lose energy. Efficiency is dramatically reduced.

64

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Figure 3.14. Single-crystal versus thin-film solar cells. Organizing the atoms in single crystals is like removing the trees to make a road through a forest. Atoms out of place or atomic impurities are obstacles for the electron just like trees are obstacles for a car. Collisions with these obstacles force the electron (or the car) to lose energy. If you were a car driving through the national forest, or an electron passing through a solar cell, which path would you rather take?

In any case, after 25 years of effort on thin-film solar cells, their module efficiencies are still low and they have not penetrated the mainstream electric power marketplace. Concentrator solar cells have not entered the marketplace yet in high volume either. There are several reasons for this, but it is not for lack of performance. The technology for solar concentrators is well founded on established scientific and engineering principles. One of the problems for concentrators is that a larger investment is required for hardware like lenses and trackers as well as for new solar cell manufacturing facilities. It is time for a serious top-level funding commitment. The technology is ready.

REFERENCES

65

ABBREVIATIONS AM1.5—1 and ½ Air-Mass (see Chapter 19) CPV—concentration photovoltaic DNI—direct normal incident sunlight E—electric field Eg—bandgap energy GaAs—gallium arsenide GaSb—gallium antimonide Group III–V—compounds consisting of group III and group V elements from the periodic table (LED semiconductor compounds) HCPV—high-concentration photovoltaic LCPV—low-concentration photovoltaic LED—light-emitting diode P/N junction—positive/negative junction PV—photovoltaic or solar cell QE—quantum efficiency (the number of electrons collected per incident photon) λ—wavelength

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]

A. Einstein. On the quantum theory of radiation. Phyikalische Zeitschrift 18, 121 (1917). L. M. Fraas, J. Avery, J. Gee, et al. Over 35% efficient GaAs/GaSb stacked concentrator cell assemblies for terrestrial applications. In 21st IEEE PV Specialist Conference, p. 190. Kissimmee, Florida, 1990. IEEE, New York (1990). R. P. Feynman, R. B. Leighton, and M. Sands. The Feynman Lectures on Physics, Volume III—Quantum Mechanics. Reading, MA, Addison Wesley (1965). N. Bohr. Radiation Spectra and the Hydrogen Atom. Philos. Mag. 25, 10 (1913). N. Bohr. The Theory of Spectra and Atomic Constitution. Fys. Tidsskr. 19, 153 (1921); 9, 1 (1922). R. P. Feynman, R. B. Leighton, and M. Sands. The Feynman Lectures on physics. In The Hydrogen Atom and the Periodic Table, Vol. 3, Chapter 19, Reading, MA, Addison Wesley (1965). A. Sommerfeld. Z. Phys. 47, 1 (1928). F. Bloch. Z. Phys. 52, 555 (1928). A. H. Wilson. Proc. R. Soc. A 133(458), 134, 277 (1931). J. M. Ziman. Principles of the theory of solids. In Electron States, pp. 72–74. Cambridge and London, Cambridge University Press (1964). C. Kittel. Introduction to Solid State Physics. 3rd Edition. New York, John Wiley & Sons (1967). S. M. Sze. Physics of Semiconductor Devices. New York, Wiley-Interscience (1969). L. M. Fraas and R. C. Knechtli. Design of high efficiency monolithic stacked multijunction solar cells. In 13th IEEE Photovoltaic Specialist Conference, Conference Record (A79-40881 17–44), Washington, D.C., June 5–8, pp. 886–891. Institute of Electrical and Electronics Engineers, Inc., New York (1978).

66 [14] [15]

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS R. R. King, D. C. Law, K. M. Edmondson, et al. 40% efficient metamorphic GaInP/ GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90 (18), p. 3516 (2007). L. M. Fraas, H. X. Huang, and J. E. Avery. Low cost high power GaSb photovoltaic cells. Thermophotovoltaic generation of electricity. In 3rd NREL Conference, T. Coutts, C. Allman, J. Benner, eds, p. 33. Woodbury, N.Y., AIP (1997).

4 SOLAR CELL DEVICE PHYSICS LARRY PARTAIN Varian Medical Systems

4.1 DEVELOPMENT OF QUANTUM MECHANICS AND SOLID-STATE ELECTRONICS Solar cells are likely the ultimate quantum mechanical and solid-state electronic devices. Although their behavior (photovoltaics) was observed in the early 1830s, their optimization and rapid advancement awaited the development of both these fields starting at the turn of the twentieth century. Many of these discoveries are summarized by the set of equations immediately below. The person most associated with each discovery is shown in parentheses along with the year of the discovery. Asterisks identify those advances that were awarded Nobel Prizes in physics [1, 2]. (Planck, 1900)*

E N = Nh f

(4.1)

(Einstein, 1905)*

E = hf

(4.2)

(Compton,1923)*

p = hf c

(4.3)

(De Broglie, 1923)

λ=h p

(4.4)

f ( ε ) = 1 [1 + exp {( ε − ε F kT )}]

(4.5)

(Fermi and Dirac, 1926) (Schrodinger, 1928)* (Shockley et al., 1947)*

[ − (h

2

8π m ) ∇ + V ( r )] FK ( r ) = EK FK ( r ) 2

2

ε*Fe − ε*Fh = qV and J n = qnμ n dε*Fe /dx.

(4.6) (4.7)

Equation 4.1 explained the shape of the black body light emission spectrum from hot objects including the sun. The energy at any emitted light frequency f consists Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

67

68

SOLAR CELL DEVICE PHYSICS

of N discrete (quantized) energy packets each with a small energy content proportional to f with a proportionality constant (Planck’s constant) value of h. Equation 4.2 more definitively established that the wave phenomena of light is actually made up of discrete photons each with a quantized energy value hf that explained the experimentally observed photoelectric effect of electron emission from metal surfaces in a vacuum illuminated with monochromatic light. Equation 4.3 established that each such photon possesses a quantized momentum value, p, which accurately explains the trajectories of photon “collisions” with electrons. Strangely enough, Equation 4.4 indicates that rest mass particles like electrons also behave like waves with wavelengths λ that are inversely proportional to their momentum value, p. This is one description of the wave–particle duality. In 1913, Neil Bohr’s [3] Nobel Prize-winning work (when combined with the De Broglie result of Eq. 4.4) explained the discrete electron energy levels of hydrogen atoms have just the right sizes to accommodate integer multiples of each electron’s wavelength, λ. Thus, the hydrogen atom has allowed electronic energy states with unoccupied energy gaps between each allowed state or energy “band.” It is reasonable to expect that when atoms are brought together to form solids, their atomic discrete levels are perturbed and broadened into allowed electronic energy state bands. Equation 4.5 describes the probability that any of a material’s allowed electronic states is actually occupied by an electron, in terms of an energy variable expression containing Boltzmann’s constant k and referenced to a special energy value εF (the Fermi energy). The latter has ultimately been found to be the electron’s potential energy value [4, 5]. 4.2 FUNDAMENTALS OF SOLAR CELL OPEN-CIRCUIT VOLTAGE The density of available quantum states to “free” electrons in a semiconductor can be approximated by solving Schrodinger’s wave equation for the case of a cube of semiconductor material where the states’ potential energy, V, is zero everywhere inside the cube and is infinite everywhere outside this cube [6]. All of the allowed solutions to the wave equation are sine waves whose wave numbers K (i.e., in the x direction Φ(x) ∼ sin(Kx)) are integer multiples of π/a where a is the dimension of one side of the cube equal to the atomic spacing of the semiconductor. If one then integrates such quantized K values in three-dimensional space, using a differential volume that is normalized by the volume (π3/a3) occupied by one individual quantum state and applying the Fermi–Dirac expression for the probability that any available state is actually occupied, then one obtains the well-known expressions for the volume density of electrons n in the conduction band (and similarly the volume density of holes p in the valence band) with the respective conduction and valence energy band edges specified by εC and εV [7, 8]: n = N c e −(εC − ε F ) kT

(4.8)

p = N v e −( ε F − ε V ) kT .

(4.9)

FUNDAMENTALS OF SOLAR CELL OPEN-CIRCUIT VOLTAGE

69

Here, NC = 2Mc(2πmdekT/h2)3/2 and Nv = 2(2πmdhkT/h2)3/2. mde and mdh are the density-of-states effective masses for “free” electrons and holes, respectively, and Mc is the number of equivalent minima at the conduction band edge [8, 9]. Each hole denotes the absence of an electron in the valence band. Note that the product of n × p gives np = N C N V exp[− ( ε C − ε V ) kT = N C N V exp ( − ε G kT ) = ni2 ,

(4.10)

which only varies with the bandgap and temperature regardless of the doping levels to make the semiconductor either n or p type or alternatively to characterize its intrinsic or “i” properties if undoped. Then ni is the electron and hole density in such “i” materials. While the density states derivation is based on the periodic atomic spacing a, any dependence on this a disappears in the total mathematical calculation process. The results, at least to the first order, appear to accurately describe the properties of even amorphous semiconductor materials with distributions of atomic spacings and not just a well-defined and single constant a as long as there is an approximately defined bandgap. Equations 4.8 and 4.9 are equilibrium expressions for materials in the dark. With light exposure, both conduction band electron and valence band hole concentrations increase above their equilibrium values. Absorbed photons with enough energy “pump” valence band electrons into the conduction band, leaving behind the valence band holes. The standard assumption is that Equations 4.8 and 4.9 still apply but with two different quasi-Fermi energy values, ε*Fe for the electrons and ε*Fh for the holes instead of the single-equilibrium Fermi energy εF0. For the latter, subscript “0” is added to denote its equilibrium position. These expressions clearly show that for n to increase, ε*Fe must move away from εF0 toward the conduction band edge εC. Similarly, ε*Fh must also move away from ε*F0 for p to increase but toward the opposite valence band edge εV. Since quasi-Fermi energies correspond to potential energies, the light-exposed semiconductor populations of electrons and holes actually possess measurably different potential energy values that, according to the first Shockely expression in Equation 4.7, is a measureable voltage, V. The magnitude of this voltage is expressed as V = [ Δε LFe + Δε LFh ] q ,

(4.11)

where Δε LFe = ε*Fe − ε F0 and Δε LFh = ε*F0 − ε Fh with the superscript “L” added to denote the nonequilibrium light exposure. It is this splitting of quasi-Fermi and potential energy values of electrons and holes that is the first fundamental step in the quantum mechanical conversion of the microscopic photon energy (typically expressed in electron volts per photon) into a macroscopic DC voltage, V, which can be physically measured and used to do work. The challenge is to include properly configured semiconductor materials into a device that allows such a macroscopic physical V measurement with a standard voltmeter.

70

SOLAR CELL DEVICE PHYSICS

A simple explanation of the underlying physics was presented in the First Edition (pp. 2–4 [5]). It details how concentration gradients (normalized by the total concentration value) produce forces per unit charge that are just as real as the forces that electric fields exert on such charges. Moving quantized particles through such concentration gradient force fields involves work that changes potential energies. This derivation is not repeated here.

4.3

SHOCKLEY DIODE MODEL OF SOLAR CELLS

In 1947, Shockley et al. [10] produced the Nobel Prize-winning discovery and demonstration of solid-state electronic transistors fabricated from single-crystal semiconductors with energy bandgap properties. Shockley’s breakthrough devices incorporated characteristics well summarized by the two expressions of Equation 4.7 (pp. 305–310, 463 [9]) that he used extensively in his work. The second Equation 4.7 expression also clearly implies quasi-Fermi level connections to potential energy because the gradient of the former provides a force that translates into velocity (and then to a current density, J) with a proportionality constant that is the electron mobility μn, where n is the density of conduction electrons in the semiconductor material. The energy gradient and resulting velocity prescribe how much of the starting potential energy and thus the efficiency is lost in establishing a DC terminal voltage, V, and a current flow as detailed below. Interesting enough, Shockley never clearly identified quasi-Fermi levels as measures of potential energy. However, there had long been hints of such a relationship, not the least being the Einstein relation that the ratio of diffusion coefficient to mobility is a known and calculable constant [8, 9]. In their bipolar forms, transistors consist of two back-to-back Shockley diodes. All of these quantum mechanics and solid-state electronic discoveries led directly to the detailed description of the Shockley diode solar cell behaviors and their controlling parameters needed for reproducible progress in this field. These preceding advances received a significant share of the Nobel Prizes in physics in the first half of the twentieth century, and their exploitation in the second half of the twentieth century, along with one more Noble Prize-winning discovery, has resulted in major increases in solar cell performance from ∼1% efficiency level in ∼1900 to over 40% currently. One wonders what the next century of Nobel Prizewinning discoveries may provide. In 1954, Chapin et al. [11] reported a 6% efficiency for an abrupt p/n junction diode single-crystal silicon solar cell conversion device. Seven years later, the derivation and treatment of the classic Shockley model was published by Shockley and Queisser [12] for the current–voltage properties and the efficiencies of this particular configuration. The corresponding geometry and nonequilibrium energy band structure for this type of solar cell exposed to light is shown schematically in Figure 4.1. This contact metallization partially covers its emitter to allow light entrance. A macroscopic and measureable voltage, V, is clearly indicated as being equal to

72

SOLAR CELL DEVICE PHYSICS

The quasi-Fermi level splitting in Figure 4.1 between ε*Fe and ε*Fh gives the highest potential energy difference available from the solar cell conversion process as characterized by setting the open-circuit voltage Voc to the voltage V of Equation 4.11. For low injection conditions, where the concentrations of the light-generated carriers ΔnL = n – n0 and ΔpL = p – p0 are much greater than that of the minority carriers but are much less than that of the majority carrier concentrations, the opencircuit expression simplifies to Voc = ε G q + ( kT q ) ln [ Δ nL po ( N C N V )]

(4.12)

for the p-type side of the solar cell and to Voc = ε G q + ( kT q ) ln [ Δ pL no ( N C N V )]

(4.13)

for the n-type side of the solar cell. For high injection where ΔnL and ΔpL become larger than the majority carrier concentration, these expressions respectively become 12 Voc = ε G q + ( 2 kT q ) ln ⎡⎣ Δ nL ( N C N V ) ⎤⎦ and

(4.14)

12 Voc = ε G q + ( 2 kT q ) ln ⎡⎣ Δ pL ( N C N V ) ⎤⎦ .

(4.15)

Only significant voltages are generated if ΔnL or ΔpL >> no or po, respectively (low injection), or is greater than both (high injection). In abrupt p/n junctions with significant doping levels and low injection, this means that voltage is only really generated by minority carriers. Note the twice higher temperature coefficient of the second term in the Voc expressions for high injection versus low injection. Also note that Voc approaches the εG/q bandgap value (in all four expressions) when the temperature T approaches absolute zero [12] or when the light-generated carrier concentrations approach 1018–1019 cm−3 concentrations (depending on the density-of-states effective mass values) where the arguments in the ln functions approach values of one. It is this latter case of high light-generated carrier concentrations giving bandgap Voc values that is another major point of this chapter for significantly increasing solar cell efficiencies. Given the absorption of a photon that produces an electron–hole pair, this specific optimization minimizes the loss of energy below the bandgap. This contrasts to minimizing the losses above bandgap in the multiple bandgap devices that have recently provided efficiencies exceeding 40% (Perharz and Bett, Chapter 14; [14]). The fundamental Shockley transistor assumption implies that any minority carrier that enters the depletion region exits the other side with relatively little change in the potential energy it had on entrance. To the degree that this assumption is accurate, it provides an efficient, low-loss process for transforming minority carriers into majority ones. This conversion process from minority to majority

SHOCKLEY DIODE MODEL OF SOLAR CELLS

73

carriers is the second major step of the abrupt p/n junction solar cell energy conversion process. It is instructive to note here the most efficient step theoretically possible for converting an absorbed photon’s energy in electron volts into a DC voltage, V. This occurs when all absorbed photons’ energies hf (i.e., monochromatic photons all of a single frequency f) just exceed the bandgap energy εG, and the open-circuit voltage Voc equals the bandgap value εG/q. This would only be possible if the quasi-Fermi levels ε*Fe and ε*Fh split all the way to the band edges εC and εV. It would be a perfect quantum mechanical conversion step with no direct loss. When the energy of an absorbed photon is greater than the bandgap, its energy excess greater than the bandgap creates “hot” electrons (and/or holes) whose energies are rapidly thermalized to steady-state values near the band edge by inelastic collisions with other charged carriers and/or with semiconductor lattice vibratrions (phonons). The latter quickly convert such extra energy into heat that no longer can contribute to steady-state carrier potential energy but only to the semiconductor’s temperature increase. The third major step in the energy conversion process is the extraction of the majority carriers’ light added energy from the semiconductor and the deposition of this energy into the metal contacts whose average electron potential energy values are also specified by their respective metal Fermi energy levels indicated by the arrows in Figure 4.1. Because of semiconductor band structure, mobile electrons can only occupy electronic states whose total energy is at or above the conduction band edge εC. They thus enter the metal with offset energy δεn as a kinetic energy (hot electron) that is again quickly thermalized by scattering down to the metal’s Fermi energy level. Since only the potential energy is available at the external metal contacts, the typical way to reduce this δεn loss is to raise the electron quasi-Fermi level by moderate to heavy doping at the semiconductor’s boundary. An analogous δεp loss process also occurs for majority carrier holes at the p-type interface with its metal contact. This is likewise typically reduced by moderate to heavy doping at the semiconductor’s boundary. For the highest conversion efficiencies, both these δεn and δεp need to be simultaneously reduced to as close to zero values as possible. The second expression of Equation 4.7 indicates that a loss of potential energy is required to transport charge from one part of a device to another as described by a current density, J. The latter is a fundamental and unavoidable potential energy loss even if the conversion step to open-circuit voltage is perfect and loss free. Despite the lack of band diagram clarity, it is the one-dimensional Shockley and Queisser [12] equations, as extended by Hovel [13], and only slightly extended more in the First Edition (chapter 1 [5]), that have formed the basis for most of the successful analysis and performance improvements that have eventually and recently led to cell efficiencies exceeding 40% ([14]; Perharz and Bett, Chapter 14). These well-known derivations will not be repeated here but will only be summarized by the resulting expressions that identify key parameters. In the Shockley abrupt p/n junction model of Figure 4.1, an electrostatically charged depletion region penetrates a depth, xE, into the emitter of thickness tE and

74

SOLAR CELL DEVICE PHYSICS

also penetrates a depth, xB, into the base of thickness tB. Outside the depletion region, there is no net charge, so any potential energy change, for the mobile electrons and holes there, cannot be due to an accumulation of net electrostatic charge. The majority carrier transport is simply described by the linear Ohm’s law, current–voltage relationships, specified by a single series resistance. The more important physics is contained in the nonlinear expressions (Eqs. 4.8 and 4.9) and relationships that describe the minority carrier transport, where the effects of concentration gradients are dominant. Combining the transport equations with the one-dimensional continuity equations and the standard boundary conditions, and the assumption of current dominated by minority carrier diffusion current at each edge of the depletion region [12, 13], one obtains the standard Shockley diode equations for current density J versus voltage V in an abrupt n/p junction solar cell as J = J 0 {exp [ q (V − JR ) kT ] − 1} − J L ,

(4.16)

V = ( kT q ) ln [( J J 0 ) + ( J L J 0 ) + 1] + JR,

(4.17)

or solving for V,

where Jo = (Joh + Joe) and J oh = q ( ni2 N D ) vdh f h and J oe = q ( ni2 N A ) vde f e and the light-generated current JL is ∞

J L = ∫0 qF ( λ ) [1 − r ( λ )] QE ( λ ) dλ.

(4.18)

Here, fh and fe are hyperbolic functions of the surface or interface recombination velocities, the diffusion velocities, the diffusion lengths, and the undepleted n and p layer thicknesses specified in the First Edition [5]. The R = RA is the specific series resistance, where R is the actual series resistance in ohms describing majority carrier transport and A is the cross-sectional area of the device. The ND and NA are the respective donor and acceptor doping concentrations of the emitter and base. The F(λ) is the photon flux density of light of wavelength λ, r(λ) is the light reflectivity of the emitter surface at λ, and QE(λ) is the quantum efficiency of the solar cell at the indicated wavelength. The open-circuit voltage Voc comes from setting J = 0 in Equation 4.17 to give Voc = ( kT q ) ln [( J L J 0 ) + 1] ,

(4.19)

which is equivalent to Equations 4.12 and 4.13’s low injection expressions applied to the boundaries of the depletion region. The short-circuit current density Jsc (where V = 0) is given from Equation 4.16 by the transcendental expression J sc − J 0 [ exp ( − qRJ sc kT ) − 1] = − J L .

(4.20)

SHOCKLEY DIODE MODEL OF SOLAR CELLS

75

If R is known, the light J–V characteristics are uniquely determined when any two of the three coefficients JL(∼ –Jsc), Jo, or Voc are known. Equation 4.19 gives the third when the other two are specified. The power density is the JV product (=J{[KT/q] In[(J/Jo) + (J/JL) + 1] + JR}) from Equation 4.17 that is maximum when the d(JV)/dJ = 0, at a maximum power current density Jm and voltage Vm. This Jm is negative, with a magnitude slightly less than JL (as shown in Fig. 4.2), and it is specified by the transcendental expression ln [( J L + J m + J 0 ) J 0 ] = − J m ( J L + J m + J 0 ) − 2qRJ m kT .

(4.21)

Substituting this Jm back into Equation 4.17 then gives Vm. The efficiency of the cell η is given by 2

1

0

Voc

J (A/cm2)

–1

–2 JL –3

R = 0.3 Ω-cm2 0.22

–4 0.0091

Jsc

0.15

–5

–6 –1

–0.5

0

0

0.06

0.5

1

1.5

2

V (V)

Figure 4.2. The light and dark J–V characteristics calculated with Equation 4.16 for Jo = 9.23 (10−20) A/cm2, JL = 5.17 A/cm2. The open-circuit voltage Voc and short-circuit current density Jsc are shown as a function of the specific series resistant R in ohm-square centimeter.

76

SOLAR CELL DEVICE PHYSICS

η = − Vm J m A Pin = − Voc J sc FF ( Pin A) = − [( ε G q ) VF][ J scmax JF] FF ( Pin A) ,

(4.22)

where Pin is the optical input power to the solar cell and A is the cell’s crosssectional area. The negative signs come from Jm and Jsc being negative. The fill factor FF is conveniently defined as the ratio FF = Vm J m Voc J sc = ( J m J L )(Vm Voc )( J L J sc ) ,

(4.23)

which specifies the “squareness” of the light J–V characteristics (see Fig. 4.2), and it has a maximum value of one and accounts for the unavoidable potential energy loss in transporting the charge to the solar cell boundaries for maximum power and the unavoidable “dark” current losses of forward biased diodes as well as losses due to high series resistance values. The (qVoc/εG) = VF loss fraction is the loss when the Voc is less than the bandgap. The loss fraction JF = J sc J scmax describes when all the incident photons with energies greater than the bandgap εG do not contribute to the Jsc (i.e., the external quantum efficiency is 40 mV/cm2) not only due to efficient light trapping and low shading but also due to better external quantum efficiency. The blue response is improved due to a highly n+-doped front diffusion and emitter localization on the back side, thus eliminating the front-side dead layer associated with a conventional PV cell structure with emitters localized on the front side. A long-wavelength response in the near IR spectral region has been improved as a result of a perfect rear-side passivation provided by the SiO2, which covers the entire rear-side area except for the contacting holes. Optimal front- and rear-side processing provides for an optical thickness for the PV cell about six times its actual thickness. Due to perfect frontand rear-side passivation, a Voc of about 700 mV has been achieved. An optimized screen printing process allows the FF to reach a value of >80%. In contrast to conventional crystalline silicon cells that demonstrate a Voc value decline of

128

CRYSTALLINE SILICON SOLAR CELLS AND MODULES Bus bar Wires

4

70

Bus bar

Bus bar

Fingers

Figure 5.5. Comparison between the conventional and D4 PV cell.

printed bus bars on the left in contrast to a D4 cell with a wire-tape electrode on the right. Referring to Figure 5.5 (left), the typical length that the electric current must flow through the screen printed finger before reaching the screen printed bus bars is not less than 35 or 20 mm in the case of a 6-in.2 conventional PV cell that utilizes two or three screen printed bus bars. In contrast, in the case of the D4 PV cell shown in Figure 5.5 (right), with the same size PV cell with the same grid pattern, the D4 electrode allows replacing two or three screen printed bus bars with 40 D4 electrode wires, thus reducing the current flow distance along the screen printed grids to only 2 mm. One may see that these wires are soldered to the bus bar provided for series interconnection with the following PV cell. Therefore, the resistive losses for the case of the conventional current collection through bus bar ribbons are at least 20 times higher when compared with the D4 current collecting technology. Employment of the D4 technology provides several additional advantages. First, it allows eliminating conventional screen printed bus bars from the front side, thus increasing the PV cell’s Rsh value and removing silver pads from the back side of PV cells, resulting in lowered recombination losses due to improved front- and rear-side passivation. Removal of silver/aluminum pads on the rear side allows eliminating one screen printing step and one drying step. It also decreases the silver paste consumption by 40% along with lowering production cost with no compromise in the cell efficiency. In fact, the D4 technology improves the PV cell efficiency by >0.1% absolute mainly due to higher Voc and FF values. In addition, it opens the possibility to introduce into mass production PV cells with substantially narrower screen printed fingers. It has been demonstrated that after industrial conventional multicrystalline (MC) PV cells are equipped with D4 electrodes, FF values of not less than 77% with screen printed fingers of 70-μm width and less than 6-micron height provide an improved cell efficiency up to 0.5% absolute due

PV MODULE

129

to lower shading and better front and rear-side passivation. Recent results demonstrated that production costs of the PV cells with narrow fingers may be substantially reduced due to >250% lower silver paste consumption relative to conventional PV cells with two bus bars. It should be appreciated that D4 technology opens a wide range of possibilities to introduce into mass production novel low cost and more efficient PV cells based on several promising technological concepts such as extremely narrow current collection metallic patterns produced either by screen printing or electroless plating, or aerosol, or ink-jet noncontacting direct printing, or laser-fired contacts that have been already developed but failed to find commercial application so far. There is realistic possibility to employ D4 technology for electric power collection and interconnection of the back contact PV cells, thus decreasing their production cost and simplifying PV module production.

5.6

PV MODULE

In contrast to the extensive progress in PV cell technology, the development of the production technology for PV modules with crystalline silicon PV cells has remained virtually unchanged for more than 30 years.

5.6.1

Conventional PV Module Production Technology

Since a PV cell is actually is a source of low-voltage ≈0.6 V with a DC electric current of about 34 mA/cm2 in sunlight, there is a need to interconnect a large number of PV cells in series in order to achieve a necessary and useful value for the DC voltage that in turn may be converted into AC by means of an inverter. Neighbored PV cells typically are interconnected in series by means of tinned copper tabs that are spot soldered onto the front side of the bus bars of a first PV cell and then onto the silver pads on the rear side of the next sequential PV cell. A certain number of PV cells interconnected in series are known as a PV string. Strings in their turn are interconnected in series by means of tinned copper bussing, thus producing a PV module layout. PV modules with series-interconnected PV cells perform optimally only when all the series-interconnected PV cells are illuminated with an approximately similar light intensity. However, if even one PV cell within the PV module layout is shaded while all other cells are illuminated, the entire PV module is adversely affected, resulting in a substantial decrease in power output from the PV module. In addition to temporary power loss, the module may be permanently damaged as a result of cell shading because when a PV cell is shaded, it starts to act as a large resistor rather than as a power generator. In this situation, the other PV cells in the PV string expose the shaded cell to a reverse voltage that drives electric current through this large resistor. This process may result either in the breakdown of a shaded PV cell or in heating it to such a temperature that it can destroy even the

130

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

entire PV module if a very high temperature persists. In order to eliminate the risk of a PV module damage in the event of shading, practically all PV modules employ BPD connected across each of the PV string and/or an entire module depending on specific PV module design and the quality of the input PV cells. The number of PV cells in a single PV string depends on the PV cell quality, namely, the ability to withstand back-voltage breakdown that each PV cell may be exposed to if one of them within the PV string is shaded. For example, for PV cells of good quality that can withstand back-voltage breakdown of 14 V and given that each PV cell generates a Vmax = 0.56 V, then the number of PV cells in one string should not exceed 24. Given that PV cells produced from metallurgical silicon typically have lower quality and back-voltage breakdown for these cells is not higher than 7 V, using them in PV strings comprising more than 12 cells is not recommended. Although employment of the BPD allows protecting the PV cells and PV strings against damage, it also causes substantial power losses of a PV module because the shading of just one PV cell results in an entire PV string switch-off. According to the industry estimate, almost 30% of kilowatt hour annual generation may be lost in field condition due to different sources of shade. Therefore, there is a need to optimize a PV module lay out in order to secure not only sufficient shading protection of PV cells but also to minimize annual kilowatt hour losses of PV systems as well. Since PV modules are generally expected to operate outdoors for typically 25 years without degradation, their construction must withstand various weather and environmental conditions. The front side of a typical PV module construction involves the use of a transparent sheet of low-iron tempered glass. The PV cell strings are sandwiched between sheets of polymeric encapsulant material, such as ethylene vinyl acetate, or thermal plastic material, such as polyvinyl butyral. An array of PV cells is placed onto the polymeric encapsulant material in such a way that the front sides of the cells face the transparent glass sheet. The back side of the array is covered with an additional layer of encapsulant material and a backsheet layer of weather protecting material, such as Tedlar® by DuPont, or a glass sheet. The additional layer of encapsulant material and the back-sheet layer typically have openings to provide for terminal electrical conductors to be passed from the PV cell strings through the back-side encapsulant layer and back sheet of weather protecting material to the outer surface for connecting with the electrical load through a junction box. For a PV module having an array of two PV strings, typically four conductors are arranged to pass through the openings so that they are all in proximity with each other so they can be terminated in a junction box mounted on the back-sheet layer. The glass, encapsulant layers, cells, and backsheet layer are typically vacuum laminated at about 14–160°C to build a free-of-air bubbles, strong bound structure that protects the PV cells from moisture penetration from the front- and back sides and also from the edges. The electrical interconnections of the PV strings and connections to BPDs are made in the junction box. The junction box is sealed on the back side of the PV module and is equipped with electric cables for neighbored PV module interconnection.

PV MODULE

131

An aluminum frame extends around the perimeter of the PV module and protects it against damage, provides mechanical strength against wind and snow loads, and facilitates mounting of the module to a support. At the same time, it is possible to employ PV modules without the aluminum frame if external supporting structures are sufficient to provide necessary mechanical strength. The fabrication of PV modules as described above with conventional technology is quite complicated and expensive. Layout of the PV module before lamination requires a separate step of “bussing” in which PV strings are electrically interconnected typically by means of tinned copper busses that increase the area that is occupied by the PV module, thus decreasing its conversion efficiency. Due to differences in thermal expansion coefficients between the copper and silicon and glass, there is a certain risk that the soldered spots between conventional tabs and front-side bus bars and rear-side silver pads may break causing PV module irreversible damage under inevitable changes of ambient temperature. There is an additional risk especially associated with utilization of thin PV cells. Exposing cells to spot soldering may result in PV cell breakage due to local heating and pressure. Once PV cells are interconnected in series by means of conventional technology, the series resistance of the produced PV module eventually exceeds the sum of all PV cell resistances due to the additional impact of soldering points, tabs, and bussing resulting in PV module FF value decline and corresponding efficiency losses when compared to the efficiency of the PV cells utilized in the module’s fabrication. For example, when PV cells with an average FF value of ≈76% are used to produce a 48-cell PV module, the module FF value typically will be lower than 73%. In other words, excessive series resistance associated with conventional PV module production technology is responsible for about 4% power loss when compared to the power that the input PV cells were capable of generating before being interconnected in the PV module. These insufficiencies become even more pronounced when modern highefficiency PV cells are utilized for PV module production. This is a clear demonstration of a growing conflict of interest between PV cell and PV module producers. In fact, PV cell producers are motivated to build PV cells with increased power output combined with decreased production cost. The easiest way to achieve this goal is to make PV cells thinner, thus securing economy of silicon material and to increase PV cell size, thus increasing the production capacity of existing manufacturing equipment with minimal additional investments. During the last 5 years, the size of PV cells has been increased from 100 × 100 cm2 to 125 × 125 cm2 and further to 156 × 156 cm2. Even a larger cell of 210 × 210 cm2 area has been developed and tested. PV module producers have been forced to buy these cells for an increased price based on dollar per Wp, but in return, they do not have sufficient benefit because thinner cells become more fragile and experience more pronounced bowing with corresponding higher-yield losses due to higher risk of breakage. To minimize this risk, there is a need for additional investment for more sophisticated production equipment for tab soldering and general handling. At the same time, conventional PV module fabrication technology does not allow increasing PV cell

132

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

power output at PV module level if the module series resistance is not decreased accordingly. Another challenge that PV module producers have to deal with is the necessity to achieve certain voltage values that a number of interconnected PV modules must generate per occupied area in order to secure the inverter’s most efficient operation. It is evident that although increased size of more efficient PV cells results in higher power output at the PV cell level, it eventually has a negative effect on the PV module performance due to the decreased voltage value since the number of PV cells per occupied area is diminished. Yet another challenge that PV module producers have experienced is the necessity to redesign the PV module layout to allow for an increasing number of PV strings now composed of a lower number of PV cells as required for cells made with metallurgical-grade silicon. Another example of a growing conflict of interest between PV cell and PV module producers is the employment of a PV cell with a selective emitter that demonstrates a higher efficiency in the blue spectral region. This advantage allows PV cell manufacturers to sell their cells at a higher price. At the same time, PV module producers do not observe the same efficiency gain on the PV module level not only because the glass and encapsulant materials substantially cut off the blue light but also due to the inevitable accumulation of dirt on the front side of PV modules. These general considerations reflect an urgent market request for developing novel, simpler, flexible, and cost-efficient technologies that are capable of utilizing more efficient PV cells in PV module production without loosing generated power.

5.6.2

Day4™ Technology for PV Module Production

One such promising technology has been recently developed at Day4 Energy Inc. [20, 21]. It was further demonstrated [22] that PV cells may be interconnected in series by means of electrically connecting the front-side D4 electrode of the first PV cell with the back-side D4 electrode of the sequential PV cell via the tinned copper bus bar. Since the thickness of this bus bar may be ≤100 μm, it securing extremely low series resistance of this interconnection. This approach allowed replacing conventional tabbing and stringing technology along with simplifying the PV module layout. New technology provides an elegant solution to produce U-type PV strings without conventional bussing by just turning each of two sequential PV cells with front and back electrodes by 90° before placing it on top of the rear-side electrode of the previous PV cell. It further allows eliminating conventional string interconnection by means of bussing, thus not only simplifying and optimizing PV module layout but also decreasing its production cost due to material and production step reduction. At the same time, PV module efficiency is increased due to its lower series resistance and economy of space occupied by conventional bussing. It is worth to notice that flexibility of the Day4 technology allows designing and producing PV modules with an increased number of PV strings, thus improving protection against shading and preserving annual generated kilowatt hour energy.

CONCLUSION

133

D4 PV module technology and specialized production equipment were introduced into mass production in 2006 after being UL and TUV certified. To date, about 40 MW of these PV modules have been installed on European, North American, and other markets and have received positive references from customers confirming excellent and reliable performance of installation. According to neutral party test results, installations with these PV modules demonstrate higher values of annually generated kilowatt hour in comparison with neighboring installation employing PV modules from competitive producers due to sufficiently higher shunt resistance of Day4 PV cells without bus bars, resulting in higher conversion efficiency at low light intensities. Depending on the quality of the input PV cells, the average power output of a mass-produced D4 PV module containing 48 MC PV cells is not less than 175 W. The efficiency of the best produced D4 module so far is 15%, while the average is typically >14% relative to the30 years of numerous numbers of PV systems’ reliable performance in different geographical locations. That is why the 25 years warranty on the crystalline silicon PV modules is based on a real solid ground. Therefore, we believe that the crystalline silicon PV industry will continue to be attractive for the long-term investments and to be the core basis for industrial solar electric energy generation. The material presented above illustrates the high potential for cost reduction and efficiency improvement practically at all steps of the crystalline silicon PV cell and module production that may be sufficient to make PV electric energy cost competitive with conventional nonrenewable sources. The challenge is how to convert this potential into sufficient cost reduction of kilowatt hour generated in-field conditions. It might sound strange but the current financial crisis in fact has created a moment of truth that is forcing the PV industry either to change its business model or to die. During the last 8 months, there has been an amazing price reduction for the PV cells (>40%) and PV modules (>50%). Keeping in mind that last year the Si-based PV cell contributed about 70% of the PV module cost and now it is lowered to 55%, there is sufficient potential for further PV module and kilowatt hour cost reduction. This trend must be combined with a modified PV industry business model that must change its current focus from dollar per Wp toward dollar per kilowatt hour and ROI. The selling policy must also be changed from the current 25-year guarantee for PV module Wp performance toward the 25-year guarantee of a certain kilowatt hour annual generation. We believe that the success and survival of the PV industry depends on the implementation of this ambitious but feasible program.

ABBREVIATIONS AC—alternating current AR—antireflective coating BPD—bypass diode BSF—back surface field CAD—computer-assisted drawing CZ—Czochralski process D4—Day4 Energy DC—direct current FF—fill factor Gen—generation HIT—heterojunction with intrinsic thin layer I—current IBC—interdigitated back contact Impp—current at maximum power point

REFERENCES

135

IR—infrared Isc—short-circuit current ISE—Institute for Solar Energy max—maximum mc—multicrystalline n—negatively or donor-doped semiconductor NREL—National Renewable Energy Laboratory p—positively or acceptor-doped semiconductor PERC—passivated emitter and rear cell PERL—passivated emitter, rear locally diffused Pmmp—power at maximum power point PV—photovoltaic R—resistance Rbl—bulk resistance Rem—emitter resistance ROI—return on investment Rs—series resistance Rsh—shunt resistance SiC—silicon carbide SiNx—silicon nitride antireflective coating SiO2—silicon dioxide TUV—Technischer Überwachungs-Verein (Technical Inspection Association) UL—Underwriters Laboratories V—voltage Vmmp—voltage at maximum power point Voc—open-circuit voltage Wp—watt peak η—efficiency 3D—three dimensional

REFERENCES [1]

[2] [3] [4] [5]

S. W. Glunz, A. Grohe, M. Hermle, M. Hofmann, S. Janz, T. Roth, O. Schultz, M. Vetter, I. Martin, R. Ferré, S. Bermejo, W. Wolke, W. Warta, R. Preu, and G. Willeke. Comparison of different dielectric passivation layers for application in industrially feasible high-efficiency crystalline solar cells. Presented at the 20th European Solar Conference and Exhibition, June 6–10, 2005, Barcelona (2005). S. W. Glunz. High-efficiency crystalline silicon solar cells. Article ID 97370. Review article. In Advances in Optoelectronics, doi:10.1155/2007/97370 (2007). A. Brenner and E. Riddell. Nickel plating on steel by chemical reduction. Journal of Research of the National Bureau of Standards, 37, 31–34 (1946). A. Brenner. Electroless plating comes of age. Metal Finishing 37, 61–68 (1954). M. V. Sullivan and J. H. Eigler. Electroless nickel plating for making ohmic contacts. Journal of the Electrochemical Society 104, 226–230 (1957).

136 [6] [7] [8]

[9] [10] [11]

[12]

[13] [14] [15] [16] [17] [18] [19]

[20]

[21]

[22]

CRYSTALLINE SILICON SOLAR CELLS AND MODULES L. F. Durkee. Method of plating by means of light. U.S. Patent No. 4 144 139, Solarex Corporation, Rockville, MD (1979). L. A. Grenon. Electroplating method. U.S. Patent No. 4 251 327, Motorola, Schaumburg, IL (1981). S. W. Glunz, J. Knobloch, D. Biro, and W. Wettling. Optimized high-efficiency silicon solar cell with Jsc = 42 mA/cm2 and η = 23.3%. In Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 392–395 (1997). K. F. Teng and R. W. Vest. Application of ink jet technology on photovoltaic metallization. IEEE Electron Device Letters 9, 591–593 (1988). T. Rivkin, C. Curtis, A. Miedaner, J. Perkins, J. Alleman, and D. Ginley. Direct write processing of photovoltaic cells. In Proceedings of the 29th IEEE Photovoltaic Specialists Conference, New Orleans, LA, pp. 1326–1329 (2002). M. F. Kaydanova, A. M. van Herst, A. Miedaner, C. Curtis, J. Alleman, M. S. Dabney, E. Garnett, S. Shaheen, L. Smith, R. Collins, J. I. Hanoka, A. M. Gabor, and D. S. Ginley. Direct write contacts for solar cell. In Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp.1305–1308 (2005). M. Hoerteis, P. L. Richter, and S. W. Glunz. Improved front side metallization by aerosol jet printing of hotmelt inks. Presented on the 23rd European Photovoltaic Solar Energy Conference and Exhibition, September 1–5, 2008, Valencia, Spain (2008). H. J. Gould. Method of manufacturing a back contact for semiconductor die. U.S. Patent 5,451,544 (1993). R. Preu, E. Schneiderlöchner, S. Glunz, and R. Loedeman. Method of producing a semiconductor-metal contact through a dielectric layer. U.S. Patent 6,982,218 (2001). M. A. Green and S. R. Wenham. Australian Patent No. 5703309, Australia (1984). M. Green, J. Zhao, A. Wang, and S. R. Wenham. Very high efficient silicon solar cell- science and technology. IEEE Transactions on Electron Devices 46(10), 1940– 1947 (1999). R. Swanson. Method of fabricating back surface point contact solar cells. U.S. Patent No. 4,927,770 (1990). L. Rubin and G. Rubin. Electrode for photovoltaic cells, photovoltaic cell and photovoltaic module. EP Patent 1 547 158 B1, 2002, August 29 (2007). A. Schneider, L. Rubin, A. Osipov, A. Smirnov, and P. Antipov. A new approach in solar cell module interconnection technique resulting in 5-10% higher PV module power output. Presented at the IEEE 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, May 8–12 (2006). A. Schneider, L. Rubin, and G. Rubin. Solar cell efficiency improvement by new metallization techniques—The Day4™ electrode concept. Presented at the IEEE 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, May 8–12 (2006). A. Schneider, L. Rubin, and G. Rubin. The Day4 electrode—A new metallization approach towards higher solar cell and module efficiencies. Presented at 21st European Photovoltaic Solar Energy Conference, September 4–8, 2006, Dresden, Germany (2006). L. Rubin and V. Nebusov. Photovoltaic module with edge access to PV strings, interconnection method, apparatus, and system. PCT/CA/2007/002301 (2007).

6 THIN-FILM SOLAR CELLS AND MODULES ROBERT BIRKMIRE Institute of Energy Conversion, University of Delaware

6.1

INTRODUCTION

Thin-film solar cell technologies originated in the 1960s with Cu2S/CdS, which was the first (flexible) thin-film solar cell and the first to achieve 10% efficiency in 1981 [1]. From 1970 through the 1980s, CuInSe2 [2], CdTe [3], and a-Si [4] became the solar cell materials of interest with all three achieving ∼10% efficiency. The mantra through this period was thin films will be in large-scale production and will cost ∼$1 per watt next year, which continued through the 1990s. In the 1990s, however, performance levels of the thin-film solar cells increased with CuInGaSe2 >19% [5], CdTe >16% [6], and a-Si >10% (stabilized) [7]. This, coupled with the expansion of c-Si technologies in the mid-1990s, has led to significant efforts to commercialize thin film technologies. At the turn of the century, PV began its rapid expansion with production increasing about 35%/year with most of the capacity in the c-Si module production. In 2008, worldwide solar cell production was ∼6.85 GW up from 3.44 GW in 2007 [8]. In the past few years, thin film manufacturing has undergone rapid expansion lead primarily by a-Si and CdTe technologies, and in 2008, thin film production was ∼0.89 GW up about 120% [8]. The emergence of the thin film technologies has been, in part, driven by the shortage of Si feedstock, which limited new production of c-Si and mc-Si modules. First Solar emerged as the largest thin-film PV manufacturer in the world and has expanded its manufacturing capacity of CdTe to 1 GW in early 2009. It is the first thin film company to demonstrate the “promise of thin-film PV” that high-throughput processing and large-scale facilities will significantly reduce the cost of PV modules and has achieved the lowest manufacturing cost per watt in the industry of less than $1 per Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

137

138

THIN-FILM SOLAR CELLS AND MODULES

TABLE 6.1. Best Verified Solar Cell Efficiencies for Thin-Film Solar Cells along with c-Si for Comparison under Standard Test Conditions Classification

Eff. (%)

Area (cm2)

VOC (V)

JSC (mA/cm2)

FF (%)

Test Center (Reference)

CdTe

16.5

1.03

0.845

25.9

75.5

NREL [6]

CuInGaSe2 CuInGaSe2

19.4

0.99

0.716

33.7

80.3

NREL [5]

20.0

0.419

0.692

35.7

81.0

NREL [5]

a-Si

9.5

1.07

0.859

17.5

63.0

NREL [7]

nc-Si

10.1

1.2

0.539

24.4

76.6

JQA [10]

a-Si/a-Si/a-SiGe

12.1

0.27

2.297

69.7

NREL [11]

Si (crystalline)

25.0

4.0

0.705

42.7

82.8

Sandia [12]

Si (multicrystalline)

20.3

1.002

0.664

37.7

78.9

NREL [13]

7.56

For a complete listing of the best cell efficiencies, see Green et al. [14].

watt [9]. a-Si manufacturing is rapidly increasing with Applied Materials, Oerlikon Corp, and Sharp leading the way. The production of CuInSe2-based modules has also increased substantially through efforts of Global Solar and Wurth Solar along with the entrance of Showa Shell and Honda into the market. This has spurred an infusion of venture capital for a large number of start-up companies in the United States and throughout the world in CuInSe2-, CdTe- and a-Si-based technologies as well as sparked the interest of several larger companies. The market share of thin-film PV modules is expected to increase substantially in the coming years. The performance of commercially available thin-film modules is less than crystalline Si (c-Si) modules that typically have efficiencies ranging from 12% to 19%. CuInSe2-based modules have the highest efficiency of thin-film modules in the range of 9–12% with CdTe modules from 9% to 11% and a-Si typically in the 5–8% range. The thin film technologies are at a similar development stage as c-Si was in the 1980s, and the next challenge is to increase module performance at the manufacturing level. A summary of the best cell and module performance measured at certified test laboratories is given in Tables 6.1 and 6.2 for a-Si-, CdTe-, and CuInSe2-based devices along with both single-crystal and multicrystal Si as reference points. It is interesting to note that both a-Si and CdTe, the most commercially advanced technology, have not recently increased the best performance value, mostly likely since the focus has been on increasing manufacturing capacity. In Figure 6.1, the best efficiency thin-film solar cell device is compared to the maximum obtainable single junction efficiency along with the best cell and module efficiency for single-crystal Si as a reference point. Single-crystalline Si is within about 85% of the efficiency limit, while the best CuInGaSe2 cells are ∼70% of the efficiency limit and can be directly compared

INTRODUCTION

139

TABLE 6.2. Best Verified Module Efficiencies for Thin Film along with c-Si for Comparison under Standard Test Conditions Classification

Eff. (%)

Area (cm2)

VOC (V)

ISC (A)

FF (%)

Test Center (Reference)

CdTe

10.9

4874

26.21

3.24

62.3

NREL [15]

31.2

2.18

68.9

NREL [16]

3.285

66.0

NREL [17]

CuInGa(Se,S)2

13.5

3459

a-Si/a-SiGe/a-SiGe

10.4

905

Si (large area)

20.1

16,300

66.1

6.35

78.7

Sandia [18]

Si (multicrystalline)

15.5

1017

14.6

1.37

78.6

Sandia [19]

4.353

For a complete listing of best module efficiencies, see Green et al. [14].

35 30

Cell Module Black-body limit

Efficiency (%)

c-Si 25 20

CuInGaSe2

AM0 CdTe

15 AM1.5 a-Si

10 5 0.5

1.0

1.5 Bandgap (eV)

2.0

2.5

Figure 6.1. Comparison of maximum achievable efficiency for single-junction solar cells to best-reported solar cell for different technologies. The best module performance is also included.

to the best multicrystalline Si cells since they have about the same bandgap and are polycrystalline in nature (not shown on graph). CdTe is about 55% of the efficiency limit, which indicates that advances in the laboratory are needed for the technology to reach its full potential. It is difficult to evaluate a-Si PV technology on this basis since the best devices are multijunction structures. The differential between the best cell and module can be used as a crude measure of the readiness of the technology for manufacturing. The efficiency of the CdTe and CuInGaSe2 modules is about 50–60% of the best cell, while multijunction a-Si and c-Si modules are about 80% of the best cell and mc-Si is 75%.

140

THIN-FILM SOLAR CELLS AND MODULES

The selection of PV modules will be based on annualized power output and levelized cost of energy for the specific location of the array and type of application. Figure 6.2 compares the annualized output for c-Si, mc-Si, a-Si, and CdTe at different geographic locations in the United States for residential rooftop arrays mounted at a fixed tilt for the local latitude. The graph was generated from the SAM [20], developed by DOE/NREL. Table 6.3 lists the module manufacturers and efficiencies used from SAM. The array size is ∼2.5 kW for all technologies and the SAM database is used for the modules. It is important to note that the lower-efficiency a-Si (5.7%) and CdTe (7.7%) arrays perform nearly as well as c-Si (19.3%) and better than mc-Si (13.3%) as the result of a better temperature coefficient. However, there is a large variation in temperature coefficient data for PV modules, and the effects of illu-

1500

1000

500

CdTe

0 AZ

mc-Si HI

MT Location

c-Si PA

pe

a-Si

Ty

Annual Output (kWh/kW)

2000

WA

Figure 6.2. Comparison of the annualized power output for deployment on a ∼2.5-kW roof residential rooftop array at different geographic locations for thin-film and Si modules. TABLE 6.3. Modules Used to Generate the Annualized Power from the Solar Advisor Model Database Module

Type

Eff. (%)

SunPower

c-Si

19.3

Kyocera

mc-Si

13.3

Uni-Solar

a-Si

5.7

First Solar

CdTe

7.7

THIN-FILM SOLAR CELLS AND MODULES

141

mination intensity on the module performance are not taken into account in the calculations. Thus, an accurate comparison requires a more detailed model that is validated by real data. Also, the lower-efficiency arrays require significantly more area, which increases the balance of system cost. The “figure of merit” will thus be the levelized cost of energy.

6.2

THIN-FILM SOLAR CELLS AND MODULES

The motivation for thin-film solar cells was, and still is, the potential for highspeed/high-throughput manufacturing and minimum material requirements to reduce cost. To meet this goal, large-scale manufacturing facilities are needed to obtain economies of scale, but until recently, the large up-front investment required inhibited the full-scale commercialization of thin film technologies. Thin-film modules all have several common components: (1) substrate; (2) a TCO and/or grid; (3) metal contacts; and in many cases, (4) a monolithic integration scheme. The substrate is either glass or a flexible web that is either plastic or metal foil, and the device configuration is either a superstrate or substrate design as shown in Figure 6.3. The superstrate is always glass for the thin film technologies and is coated with a TCO such as ITO, SnO2, or ZnO. Both CuInSe2- and a-Si-based modules are currently commercially being manufactured using glass and/or flexible substrates, while CdTe is always on a glass substrate. (There are several start-up companies exploring CdTe on a metal substrate.) For the substrate device configuration, glass is used for CuInSe2-based modules where Mo is used as the back contact, and flexible metal foils or polyimide substrates (web) are used for both a-Si- and CuInSe2-based modules. The attraction of the flexible substrate is the use

Figure 6.3. The two basic cell configurations for thin-film solar cells.

142

THIN-FILM SOLAR CELLS AND MODULES

of continuous roll-to-roll processing for manufacturing and was first proposed and demonstrated for Cu2S solar cells on a flexible Cu web [21]. Energy Conversion Devices and Iowa Thin Film Technologies adapted this for a-Si in the 1980s. In the 1990s, CuInSe2-based solar cells were developed on a polyimide and stainless substrate, and recently, several start-up companies are pursuing this approach on both polyimide and metal webs. A major advantage of thin film technologies is that large glass substrates or continuous webs can be coated as compared to c-Si, where individual wafer cells are fabricated and physically connected into modules using a “tab and stringing” process. For the glass and insulating webs, monolithic integration schemes are used for both superstrate and substrate modules to series connect individual cell segments, eliminating the need to assemble individual cells into the modules. Figure 6.4 illustrates the monolithic integration structure where the details of the processing steps depend on the device configuration and substrate/web. The cell segments and interconnections are defined by a series of laser and/ or mechanical scribes at different processing stages for glass substrates and are typically done as the last processing step for flexible insulating webs. For conductive webs used for both a-Si and CuInSe2, individual cells are prepared by cutting the web and applying a metal grid structure to provide the front contact to the cell and then are connected in series in a manner similar to assembling a c-Si module.

6.3

POLYCRYSTALLINE THIN FILM

Thin-film CdTe and CuInSe2 and related alloys are polycrystalline in nature with grain sizes of approximately 1–5 μm. Both CdTe- and CuInSe2-based solar cells

Figure 6.4. Typical monolithic integration scheme for thin-film modules.

POLYCRYSTALLINE THIN FILM

143

are heterojunction devices where CdTe is configured in a superstrate structure and CuInSe2 in a substrate structure. Both devices operate in a similar manner and have nearly 100% internal quantum efficiency and thus are generally not limited by Jsc. Approaches to improve device performance have primarily focused on Voc and FF, but the approach is different for each materials system. The primary mechanism limiting Voc, in both devices, is generally recombination in the space charge region, which is consistent with Shockley–Read–Hall recombination at defects within the bandgap [22, 23]. 6.3.1

CdTe Solar Cells and Modules

CdTe solar cell technology is leading the way in the development of thin-film PV in terms of cost and high-speed/high-throughput manufacturing. CdTe has a bandgap of 1.45 eV, which is a near perfect match to the AM 1.5 terrestrial spectrum for maximum solar cell efficiency (see Fig. 6.1). It is a direct bandgap semiconductor with an absorption coefficient >105 cm−1 at 700 nm and therefore only requires about a 1-μm thick film to absorb most of the incident light. All highefficiency CdTe solar cells have been made using a superstrate structure where the final processing steps, used to control the electronic properties and formation of a back ohmic contact, are carried out on a free CdTe surface. Efforts to use a substrate configuration where the CdTe is deposited on a conductive opaque substrate have not been successful, and thus CdTe solar cells are not compatible with a flexible substrate. Small area laboratory devices fabricated on a borosilicate glass substrate have achieved an efficiency of 1.3 eV, independent of the alloying materials, are less suitable for solar cells [41]. All high-efficiency CuInSe2-based devices are made using a substrate device structure where the substrate can be glass or a flexible plastic or metal foil. Efforts to use a superstrate configuration where the CuInSe2-based material is deposited on a transparent conducting substrate have not been successful [42]. The highest-efficiency solar cell, 20%, was made using a CuInGaSe2 film deposited by multisource evaporation [5], and the highest-efficiency module, 13.5%, was made using a CuInGa(Se,S)2 film grown by the reaction process [13]. CuInSe2-based solar cells are p/n heterojunction devices where a wide bandgap, n-type window layer such as CdS is used as the heterojunction partner with the p-type CuInSe2 base material. A schematic of the device structure is shown in Figure 6.7. Mo is used for the back ohmic contact and is deposited by sputtering. CuInSe2based thin films are generally deposited by the multisource elemental thermal evaporation or reaction of a precursor such as sputtered metals in a Se and/or S atmosphere. The remarkable property of CuInSe2 materials is the tolerance to

TABLE 6.4. CuInSe2 Alloy Materials Material

Eg (eV)

CuInSe2

1.0

CuGaSe2

1.68

CuAlSe2

2.72

CuInS2

1.53

CuGaS2

2.53

CuAlS2

3.5

CuInTe2

1.1

CuGaTe2

1.23

POLYCRYSTALLINE THIN FILM

147

Figure 6.7. Standard CuInSe2-based solar cell structure highlighting the individual layer in the device.

composition variations. The Cu concentration in the film can vary by 2–3% without affecting the device performance provided that the Cu-to-group III ratio is less than one. This is attributed to the broad single-phase region of CuInSe2 at the growth temperatures and may explain why, when forming a wider bandgap alloy, materials are less tolerant to composition. CuInSe2 may be the only I-III-V chalcopyrite material that has the broad single-phase regime. Another critical issue in the growth of CuInSe2-based materials is the incorporation of Na into the film that modifies the growth habit of the film and appears to reside at the grain boundaries and improves the properties of the solar cell device [43]. The soda lime glass substrate provides a source of Na, which diffuses from the glass thru the Mo contact into the growing CuInSe2-based film. However, for manufacturing, a diffusion barrier is, generally, used on the glass and the Na is controllably incorporated into the film. Two primary approaches have been used to deposit CuInSe2-based materials. Multisource evaporation has been used to deposit the CuInSe2-based thin films used for the highest-efficiency solar cells. Figure 6.8 is a schematic of the deposition system. The incident flux and substrate temperature are varied during the film growth to control the through-film composition, grain structure, and bandgap. Typically, for a CuInGaSe2 film, the through-film composition of Ga and In is controlled by varying the incident flux during film growth, creating a gradient in the bandgap through the film that depends on the Ga to (Ga + In) ratio. Variation in the Cu flux during growth can affect the structure of the film, but the Cu diffuses rapidly through the film, resulting in a uniform distribution of Cu. The best laboratory solar cell was fabricated using a three-stage process where the initial growth is just GaIn-Se followed by Cu-Se and then Ga-In-Se [5]. However, solar cells with efficiencies around 17% have been grown by simpler uniform growth and two-stage processes, which are more compatible with manufacturing [44].

148

THIN-FILM SOLAR CELLS AND MODULES Substrate

Se sparger tubes

θ Se Cu Se Ga Se

In

Se

Z Cu, Ga, In sources

Y X

Figure 6.8. Schematic of an elemental source thermal evaporation in-line system for depositing CuInGaSe2 films on a flexible web.

Figure 6.9. System for reacting precursor materials to form CuInGaSe2 films.

The second approach to growing CuInSe2-based materials is the reaction precursor films in a Se2 and/or S2 atmosphere, generally H2Se and/or H2S, and is referred to as selenization or surfurization. Figure 6.9 is a schematic of the reaction system. This approach was initially commercialized by ARCO Solar, Siemens Solar Industries, and Shell Solar Industries [45] and is the focus of many companies since the reaction process is simpler to translate to manufacturing. The simplest example of the reaction process is to sputter deposit Cu-In-Ga layers and to react these precursors in H2Se gas at atmospheric pressure. Typically, the reaction temperature is limited to less than 500°C since the H2Se will react with the Mo contact at higher temperature. Additionally, as a result of the reaction kinetics, the Ga resides near the back of the film, and thus the benefits of incorporating Ga to widen the bandgap and to increase Voc are not realized. Sulfur has been incorporated at the front surface where the junction is formed to increase the Voc, and a small area

POLYCRYSTALLINE THIN FILM

149

device of about 16% has been reported by Siemens [46]. Several companies are developing precursor processes that eliminate the metal sputtering process based on inks consisting of nanoparticles containing Cu-In-Ga-Se or electrodeposited metals. The n-type window layer for many CuInSe2-based devices is CdS and has been used for the highest-efficiency laboratory cells as well in commercial modules. The CdS is deposited by a wet chemical process, referred to as chemical bath deposition (CBD), and conformal CdS films can be deposited with a thickness of about 50 nm to minimize optical losses in the CdS layer. There has been a significant effort to replace the CdS with a non-Cd-containing window layer to make the modules more environmentally friendly. Some of the materials that have been evaluated are Zn(S,O,OH) [47], ZnSe [48], ZnIn2Se4 [49], In2S3 [50], and Inx(OH,S)y [51], which were deposited by a variety of techniques such as CBD, sputtering, MOCVD, and ALE. However, the device performance is, in general, not as good as devices using CdS. As with CdTe devices, a high-resistive TCO is deposited on the CdS, which is usually ZnO and has a resistivity from 1 to 100 Ω-cm with a thickness of ∼50 nm. Since there is a significant probability of pinholes in CdS, the highresistive TCO eliminates the possibility the conductive TCO from contacting the CuInSe2-based absorber, which can result in secondary diodes. Finally, the conductive TCO is deposited, which is usually doped ZnO or ITO having a resistivity of 10−3 to10−4 Ω-cm. Manufacturing of CuInSe2-based modules is at the early stages of commercialization and is being led by Showa Shell, Honda, Wurth Solar, and Global Solar, which have production capacity in excess of 15 MW along with numerous start-up companies throughout the world. Both multisource evaporation and reaction of precursors in a Se2 and/or S2 atmosphere are being used for the deposition of the CuInSe2-based absorber on glass, polymer, and metal foil where roll-to-roll processing is being developed using polymer and metal foil webs. Module efficiencies of close to 12% have been reported, and cell efficiencies on the steel substrate are about 10%. Recently, Ascent Solar reported monolithic integrated modules with efficiencies greater than 9.5%, verified by NREL using a roll-to-roll processing on a polyimide web. CuInSe2-based modules have the potential to achieve 15% at the commercial level based on laboratory results but lack fundamental knowledge to control properties, and the process has limited progress. The challenge with elemental multisource evaporation is the high source temperatures, required to obtain high film growth rates; typically, the Cu source operates at ∼1600°C, coupled with a Se environment [52]. Design and control of the system, particularly when depositing on a continuous web, requires in situ process control and diagnostics [53], which are being developed by several companies. For the reaction of precursors, the reaction time to form the CuInSe2-based films is limited by the reaction/diffusion rates. For a flexible web, the challenge is to react the precursors on a moving web where uniform delivery of the gas and reaction time become important issues. For modules fabricated on a glass substrate, a batch process is used where a large

150

THIN-FILM SOLAR CELLS AND MODULES

number of plates are reacted in a furnace. Developing a robust high-throughput manufacturing process needs to be demonstrated along with low cost comparable to that demonstrated for CdTe modules to accelerate the CuInSe2-based PV technologies.

6.4

HYDROGENATED a-Si, a-Si ALLOYS, AND nc-Si

Thin-film a-Si solar cells/modules have been commercially available starting with SANYO’s first commercial production facility in 1980. Today, a-Si-based modules are available from a number of companies and typically have efficiencies from 5 to 8%. The challenge for a-Si-based materials has been to improve the stabilized efficiency, which is less than their initial efficiency due to light-induced degradation, the Staebler–Wronski effect [54], which is a fundamental property of the an a-Si material. The basic device is a p-i-n structure where the n- and p-layers are 10–20 nm thick and the intrinsic i-layer is from 100 to 200 nm, depending on the device design. Engineering approaches to minimize the light-induced degradation have focused on multijunction device structures that minimize the thickness of the intrinsic layer in order to reduce the light-induced degradation. a-Si can be characterized as a direct bandgap material; however, the conduction and valence band edges are not well defined as in crystalline material as a result of band tail states. To compare the bandgap of different a-Si materials, the optical absorption coefficient as a function of energy is analyzed according to Tauc, where an optical or Tauc bandgap is rigorously defined [55]. The optical bandgap for a-Si varies from 1.7 to 1.8 eV depending on the H2 concentration in the film and has an absorption coefficient >105 m−1 at 550 nm. The bandgap can be modified from 1 to >2 eV by alloying with Ge to narrow the bandgap (a-SiGex) or with C (a-SiCx), O2 (a-SiOx), or N2 (a-SiNx) to widen the bandgap. The a-Si can be doped p-type with boron and n-type with phosphine and is used as a thin-film transistor in the display industry. The amorphous nature of the film can be transitioned to a nanocrystalline structure by controlling growth conditions and has been exploited for p-type window layers, μc-Si, to enhance optical transmission and for i-layers, nc-Si, to increase the absorption above 850 nm for use as a bottom cell in a tandem device structure [56]. The design of thin-film a-Si-based solar cells/modules is the most flexible of any of the thin film technologies. Both superstrate and substrate devices have been made on glass, foil, and plastic substrates, all based on a p-i-n structure as shown in Figure 6.10. To achieve the best performance, the device is illuminated through the p-layer, but the order of the deposition depends on the specific device design. The n-layer is deposited first for the substrate configuration, and the p-layer is deposited first for a superstrate structure. The contact to the n-layer is generally ZnO with a metal reflector, independent of the device structure, to facilitate light trapping in the device. The contact to the p-layer is a TCO where both SnO2 and ZnO have been used for modules.

HYDROGENATED a-Si, a-Si ALLOYS, AND nc-Si

151

Figure 6.10. Device structures used for a-Si-based solar cells.

Most a-Si, μc-Si, and nc-Si films used for commercial modules are grown by a PECVD process using RF (13.56 MHz) or VHF (40–100 MHz) excitation where SiH4 is the primary precursor gas and the substrate temperature is between 150 and 250°C. The growth rate for “device quality” a-Si films is from 1 to 5 A/s for the RF PECVD and is over 20 A/s for VHF PECVD [57]. However, VHF PECVD has the advantage of improved stabilized performance at the higher growth rates compared to RF PECVD, but designing the VHF electrode for uniform deposition over large areas is more challenging. Other growth processes have been used including microwave PECVD, photo- and catalytic (hot wire) CVD, and sputtering. Performance, especially stabilized, decreases as the growth rate increases regardless of the deposition method. The utilization of the SiH4 is, generally, less than 10% depending on the design equipment. Figure 6.11 shows a schematic of an RF–PECVD system. The intrinsic a-Si layer is grown using a gas mixture, SiH4 and H2, typically referred to as hydrogen dilution (ratio SiH4/H2) to control the bandgap and crystalline fraction in the film. Device quality nc-Si films have a crystalline fraction that is typically around 60% and consists of 10- to 20-nm crystallites that are encased by an a-Si:H “tissue.” The optical properties of the nc-Si:H are a mix of a-Si:H and c-Si, and thus the thickness of the intrinsic nc-Si:H layer needs to be about 1–2 μm to absorb a significant fraction of the long wavelength light. The conductivity of the a-Si is controlled by adding a dopant gas to the SiH4/H2 gas mixture, either diborane (B2H6), borntrifloride (BF3), or trimethylboron (B(CH3)3), for doped p-typing or phosphine (PH3) for n-type doping. The structure of a p-type a-Si window layer, which does not contribute to the current in the cell due to poor

152

THIN-FILM SOLAR CELLS AND MODULES Gas inlet

Substrate RF Electrode Substrate Transport RF

Plasma

Pump

Burn box

Figure 6.11. RF–PECVD system for growing a-Si films.

carrier properties, is grown using high H2 dilution to grow a μc-Si film that improves the short wavelength transmission and has a higher crystalline fraction than the nc-Si i-layer. Commercialization of a-Si-based module technology is undergoing a rapid expansion with several companies selling “turnkey” systems for single and multijunction modules based on RF or VHF PECVD. Module dimensions are increasing with Applied Materials supplying equipment to manufacture a 2.5 × 2.3 m module on a glass substrate. The Prometheus Institute projections for the thin film manufacturing capacity in 2012 is 9.6 GW with a-Si leading the way with a capacity of ∼5.2 GW. These projections are very conservative since 2008 thin-film capacity has already exceeded the projections in the report [58]. a-Si modules face several challenges to remain competitive with other thin film technologies and with c-Si wafer-based modules. The module efficiencies are low, and it is difficult to project efficiencies beyond 10%. However, a-Si-based modules have the lowest temperature coefficient of any PV technology, and thus the annualized power output can be comparable to or higher than modules with higher-efficiency ratings (see Fig. 6.2). For example, the annualized power output of a-Si modules exceeds that of c-Si by 5–15% in a large number of studies [59]. Two critical issues to improve the viability of a-Si technology are (1) improvement in stabilized performance, which is being addressed by going to a nc-S/a-Si tandem structure and (2) increased growth rate to reduce capital costs, which is particularly important for the nc-Si bottom cell where the i-layer is 1.5–2.0 μm. However, there appears to be a trade-off of performance and processing speed. a-Si technology also has several advantages: (1) flexibility in device design and structure; (2) pro-

ABBREVIATIONS

153

cessing is done at low temperature, less than 200°C; (3) material availability is not an issue even though SiH4 utilization is low; and (4) the flat panel display industry underpins the technology and scientific base.

6.5

OUTLOOK FOR THIN-FILM MODULES

The production of thin-film PV modules will continue to expand and probably, the manufacturing capacity will exceed c-Si in the next 5–10 years, primarily replacing mc-Si module market share. CuInSe2-based manufacturing challenges will be overcome and modules will reach or exceed 15% on glass and possibly flexible substrates utilizing high-speed processing technology to drive the manufacturing cost well below $1 per watt. New companies will enter the CdTe module market and module efficiencies will approach 13% with manufacturing cost less than $0.65 per watt. There will be a competition between the polycrystalline and a-Si technologies, which will be driven by the application and will levelize cost of power. As the production capacity expands, material availability of In and Te may become a critical issue, which would benefit the a-Si technology and be a driver for the organic PV technologies. The future of PV will consist of a mix of c-Si and thin film technologies where cost, application, location, and aesthetics will determine the choice of the PV technology. Thin film technologies will be the modules of choice for residential roof application and commercial build integrated applications. Medium-scale power- and utility-scale PV arrays will use both polycrystalline and high-efficiency c-Si modules. However, in hot arid locations, very high-efficiency concentrator PV arrays will also be used for utility-scale applications.

ABBREVIATIONS ALE—atomic layer epitaxy a-Si—amorphous silicon C—carbon CBD—chemical bath deposition Cd—cadmium CdS—cadmium sulfide Cd2SnO4—cadmium stannate CdTe—cadmium telluride Cl—chlorine c-Si—crystalline silicon CSS—close-spaced sublimation Cu—copper CuInGaSe2—copper indium gallium diselenide Cu2S—copper sulfide CuInSe2—copper indium diselenide

154

THIN-FILM SOLAR CELLS AND MODULES

DOE—Department of Energy Eff.—efficiency FF—fill factor Ga—gallium Ge—germanium H—hydrogen Hg—mercury I—undoped or intrinsic semiconductor III—elements in column 3 of periodic table In—indium ITO—In2O3:Sn Jsc—short-circuit current MBE—molecular beam epitaxy mc-Si—multicrystalline silicon Mo—molybdenum MOCVD—metal-organic chemical vapor deposition N—nitrogen n—negatively doped semiconductor Na—sodium nc-Si—nanocrystalline silicon NREL—National Renewable Energy Laboratory O—oxygen P—phosphorous p—positively doped semiconductor Pd—palladium PECVD—plasma-enhanced chemical vapor deposition PV—photovoltaics or solar cells PVD—physical vapor deposition RF—radio frequency S—sulfur SAM—Solar Advisor Model Se—selenium Si—silicon SiH4—silane Sn—tin SnO2—tin oxide TCO—transparent contact oxide Te—tellurium V—elements in column 5 of periodic table VHF—very high frequency Voc—open-circuit voltage VT—vapor transport Zn—zinc ZnO—zinc oxide μc-Si—microcrystalline silicon

REFERENCES

155

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24]

R. B. Hall, R. W. Birkmire, J. E. Phillips, and J. D. Meakin. Appl. Phys. Lett. 38, 925 (1981). R. A. Mickelsen and W. S. Chen. In Conf. Rec. 15th IEEE Photovoltaic Specialist Conf., pp. 800–803 (1981). Y.-S. Tyan and E. A. Perez-Albuerne. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 794–800 (1982). A. Catalano, R. V. D’Aiello, J. Dresner, B. Faughnan, A. Firester, J. Kane, H. Schade, Z. E. Smith, G. Swartz, and A. Triano. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 1421–1422 (1982). I. Repins, M. A. Contreras, B. Egaas, C. DeHart, J. Scharf, C. L. Perkins, B. To, and R. Noufi. Prog. Photovoltaics 16(3), 235–239 (2008). X. Wu, J. C. Keane, R. G. Dhere, C. DeHart, A. Duda, T. A. Gessert, S. Asher, D. H. Levi, and P. Sheldon. In Proceedings of the 17th European Photovoltaic Solar Energy Conference, October 22–26, 2001, Munich, pp. 995–1000 (2001). J. Meier, J. Sitznagel, U. Kroll, C. Bucher, S. Fay, T. Moriarty, and A. Shah. Thin Solid Films 451–452, 518–524 (2004). SolarBuzz Annual World Solar PV Industry Report. Available at http://www.solarbuzz.com/Marketbuzz2009-intro.htm (2009). First Solar Quarterly Report for 2009: www.firstsolar.com. K. Yamamoto, A. Nakajima, Y. Tawada, M. Yoshimi, Y. Okamoto, and S. Igari. In Proc. of the 2nd World Conference on Photovoltaic Energy Conversion, pp. 1284– 1290 (1998). J. Yang, A. Banerjee, S. Sugiyama, and S. Guha. In Conf. Rec. 26th IEEE Photovoltaic Specialist Conf., pp. 563–568 (1997). J. Zhao, A. Wang, M. A. Green, and F. Ferrazza. Appl. Phys. Lett. 73, 1991–1993 (1998). O. Schultz, S. W. Glunz, and G. P. Willeke. Prog. Photovoltaics 12, 553–558 (2004). A. M. Green, K. Emery, Y. Hishikawa, and W. Warta. Prog. Photovoltaics 17, 85–94 (2009). D. Cunningham, K. Davies, L. Grammond, E. Mopas, N. O’Connor, M. Rubcich, M. Sadeghi, D. Skinner, and T. Trumbly. In Conf. Rec. 28th IEEE Photovoltaic Specialist Conf., pp. 13–18 (2000). Y. Tanaka, N. Akema, T. Morishita, D. Okumura, and K. Kushiya. In Proc. 17th European Photovoltaic Solar Energy Conference, pp. 989–994 (2001). J. Yang, A. Banerjee, T. Glatfelter, K. Hoffman, X. Xu, and S. Guha. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 380–386 (1994). D. Rose, O. Koehler, N. Kaminar, B. Mulligan, and D. King. In Proc. 4th World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 2018– 2023 (2006). D. L. King, W. K. Schubert, and T. D. Hund. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 1660–1662 (1994). Available at http://www.nrel.gov/analysis/sam. A. M. Barnett, B. N. Baron, J. V. Masi, and T. W. F. Russell. Method for the continuous manufacture of thin film solar cells. U.S. Patent 4,318,938, March 9, 1982. S. Hegedus and B. McCandless. Sol. Energy Mater. Sol. Cells 88, 75–95 (2005). S. Hegedus and W Shafarman. Prog. Photovoltaics 12, 155 (2004). J. Jordan and S. Albright. Photovoltaic cell with thin CdS layer. U.S. Patent 5,279,678 (1994).

156

THIN-FILM SOLAR CELLS AND MODULES

[25]

T. Takamoto, T. Agui, H. Kurita, and M. Ohmori. Sol. Energy Mater. Sol. Cells 49, 219–225 (1997). B. McCandless and R. Birkmire. In Proceedings of the 28th IEEE PVSC, ed., pp. 491–494 (2000). B. McCandless. Unpublished work. X. Wu et al. In Conf. Rec. 28th IEEE Photovoltaic Specialist Conf., pp. 470–474 (2000). B. McCandless, L. Moulton, and R. Birkmire. Prog. Photovoltaics 5, 249–260 (1997). F. J. Kruger. Electrochem. Soc. 125, 2028–2034 (1978). Y. Tyan and E. Perez-Albuerne. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 794–800 (1982). R. Powell, G. Dorer, N. Reiter, H. McMaster, S. Cox, and T. Kahle. Apparatus and method for depositing a material on a substrate. U.S. Patent 5,945,163 (1999). B. McCandless, I. Youm, and R. Birkmire. Prog. Photovoltaics 7, 21–30 (1999). R. Wendt, A. Fischer, D. Grecu, and A. J. Compaan. Appl. Phys. 84, 2920–2925 (1998). R. Sudharsanan and A. Rohatgi. Sol. Cells 31, 143–150 (1991). J. Jordan. Photovoltaic cell and method. International Patent Application WO93/14524 (1993). S. Ikegami. In Technical Digest of the 3rd Photovoltaic Science and Engineering Conference, pp. 677–682 (1987). G. M. Hanket, B. E. McCandless, W. A. Buchanan, S. Fields, and R. W. Birkmire. J. Vac. Sci. Technol. A, 24(5), 1695 (2006). B. McCandless and J. Sites. In Handbook of Photovoltaic Science and Engineering, A. Luque and S. Hegedus, eds., pp. 633–644. John Wiley & Sons (2003). R. Birkmire. In Conf. Rec. 33rd IEEE Photovoltaic Specialist Conf., paper 370 (2008). W. Shafarman, R. Klenk, and B. McCandless. In Conf. Rec. 25th IEEE Photovoltaic Specialist Conf., pp. 763–768 (1996). T. Yoshida and R. W. Birkmire. In Proc. 11th European Photovoltaic Solar Energy Conference, pp. 811–816 (1992). M. Bodegard, K. Granath, A. Rockett, and L. Stolt. Sol. Energy Mater. Sol. Cells 58, 199–208 (1999). W. Shafarman and J. Zhu. Thin Solid Films 361–362, 473–477 (2000). R. Gay, M. Dietrich, C. Fredric, C. Jensen, and K. Knapp. In Proc. 12th European Photovoltaic Solar Energy Conference, pp. 935–940 (1994). D. Tarrant and J. Ermer. In Conf. Rec. 23rd IEEE Photovoltaic Specialist Conf., pp. 372–375 (1993). A. Ennaoui, U. Blieske, and M. Lux-Steiner. Prog. Photovoltaics 6, 447–451 (1998). Y. Othake, K. Kushiya, A. Yamada, and M. Konagai. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 218–221 (1994). Y. Othake, T. Okamoto, A. Yamada, M. Konagai, and K. Saito. Sol. Energy Mater. Sol. Cells 49, 269–275 (1997). J. Guillemoles, B. Canava, E. Yousfi, P. Cowache, A. Galtayries, T. Asikainen, M. Powalla, D. Hariskos, H. Schock, and D. Lincot. Jpn. J. Appl. Phys. 40, 6065–6068 (2001). D. Hariskos, R. Herberholz, M. Ruckh, U. Ruhle, R. Schaffler, and H. W. Schock. In Proc. 13th European Photovoltaic Solar Energy Conference, pp. 1995–1998 (1995). S. T. Junker, R. W. Birkmire, and F. J. Doyle III. AIChE J. 51(3), 878 (2005).

[26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52]

REFERENCES [53] [54] [55] [56] [57] [58]

[59]

157

K. Mukati, B. A. Ogunnaike, E. Eser, S. Fields, and R. Birkmire. Ind. Eng. Chem. Res. 48, 5975–5991 (2009). D. Staebler and C. Wronski. Appl. Phys. Lett. 31, 292 (1977). J. Tauc. In Optical Properties of Solids, F. Abeles, ed., pp. 277–313. Amsterdam, North-Holland (1972). A. Shah, J. Meier, P. Torres, U. Kroll, D. Fischer, and N. Beck. In Conf. Rec. 26th IEEE Photovoltaic Specialist Conf., pp. 569–574 (1997). H. Takatsuka, M. Noda, Y. Yonekura, Y. Takeuchi, and Y. Tamauchi. Solar Energy 77, 951–960 (2004). S. Grama and T. Bradford. Thin-Film PV 2.0: Market Outlook Through 2012: Executive Summary, the Prometheus Institute. Available at http://www.greentechmedia.com/GreentechMedia/Report/Thin-FilmPV20MarketOutlookThrough2012. html. S. Hegedus. Thin film solar modules: The low cost, high throughput and versatile alternative to Si wafers. Prog. Photovoltaics 14, 393 (2006).

7 TERRESTRIAL MODULE FABRICATION AND ASSEMBLY TECHNOLOGIES CHRISTOPHER BUNNER Spire Corporation

7.1

INTRODUCTION

The global desire to produce energy from various renewable energy options increases as political, environmental, and economic landscapes transform. As expected, the solar industry benefits from this renewed attention where governments recognize and support the need for renewable energy options through grants, low tariffs, and deferred taxes to diminish the consumption of fossil fuels. Unfortunately, environmental pressure remains tied to the cost of ownership. The regional desires for renewable solutions directly correlate with governmental financial incentives. Fortunately, the solar industry has attracted notice from complementing industries as their primary customers stagnate. This attention provides new ideas and initiatives that benefit the solar module manufacturing process. Overall, the goal of terrestrial module manufacturing remains the same: to consistently produce the least expensive and most reliable module. While the basic module construction and manufacturing processes have changed little, the methods and equipment available have matured. This chapter examines the available improvements to the module fabrication process through the evolution of materials and equipment.

7.2

MATERIALS

While the solar industry continues to grow and mature, the necessary materials to construct a silicon-based photovoltaic (PV) module continue to encompass over Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

159

BASIC MODULE ASSEMBLY

161

These thinner cells have presented special needs for the manufacturing process to achieve the required yields necessary to take advantage of the thinner material. These special requirements include low-impact cell handling, small stacks if cells are coin stacked, more stringent heating/cooling requirements, and a myriad of machine design features to maintain manufacturing process yield. Over the last decade, traditional silicon solar cell efficiency has increased from around 12% to 17%. While the individual PV cells become more efficient and evolve into larger formats, their versatility remains. Cells may be arranged in various series and parallel configurations to meet system requirements. Three years ago, a typical module produced 70 W through the interconnection of 36 125-mm cells in series. Today, the standard module design is 220 W using 60 156-mm cells in series. A larger cell size yields various benefits. First, by increasing cell area, cell power is increased. While each cell produces about 0.6 V, regardless of size, as surface area increases, cell current increases. Second, as the cell efficiency is improved, fewer cells are required to produce the same module power. Additionally, some cell manufacturers have taken additional paths to improving cell efficiency. Reducing or eliminating the top surface contacts by use of a rear-contact cell structure has gained more absorption area. Furthermore, the use of larger and more powerful cells reduces cell handling, increases solar cell production capacities, and reduces packaging and shipping costs. Cost reductions achieved in the production of cells directly relate to reduced module costs. While these wafers and cells will require additional handling and processing requirements, the efficiency improvements will likely benefit other solar applications. Cell testing and classification is an important process step for both the cell and module fabrication processes. While cell testing quantifies cell performance, it also enables cell sorting into performance bins, which allows module fabricators to maximize power output. Essentially, the lowest-performing cell within a series string will limit the output current and the subsequent power produced by the solar module. This limiting effect also extends to the installed system. Thus, system designers expect narrow module classes to reduce loses from module mismatching (Fig. 7.2).

7.3

BASIC MODULE ASSEMBLY

While alternative materials are available for flexible solar products, low iron tempered glass remains the standard load-bearing component for today’s silicon-based solar module. The glass must be fully cleaned to ensure adequate adhesion to the subsequent module layers. After the glass is permitted to dry, the first encapsulant layer is added. EVA continues to be the primary encapsulating plastic used in PV modules. Today’s faster formulations have cut lamination cycle times in half. Additionally, EVA has demonstrated improved reliability, and UV stabilizers have been added

162

TERRESTRIAL MODULE FABRICATION

Figure 7.2. Picture of cell testing machine. In addition to machine throughput improvements now with over 1000 cells/h, today’s cell test machines must be able to handle everthinning cell thicknesses.

to protect the laminate package against the sun’s spectrum and have prevented the degradation of EVA. Other encapsulant materials have continued to play an active role in the production process. Furthermore, some thin-film processes may be incompatible with the EVA process and have returned to using PVB—a material used in the 1970s before EVA became the prominent encapsulant. Next, solar cells are interconnected into series strings. While some module manufacturers use manual labor to interconnect solar cells, most new entrants to module building are utilizing available automated tabbing and stringing equipment. There are various soldering approaches, including conduction (hot bar), convection (hot air), and radiation (infrared lamps). The standardization of cell contact printing has enabled equipment manufacturers to standardize alignment and soldering equipment. Roughly 5 years ago, four stringing machines were required to support a 12-MW line. Today, the standard throughput exceeds 500 cells/h. Coupled with larger solar cells, one stringing machine can connect soldered strings to support a 12-MW production line (Fig. 7.3). The connection of cell strings (busing) remains similar. The strings are connected using a thicker and/or wider bus wire (ribbon) that brings the module current and voltage through an egress hole or slit in the back sheet ultimately terminated in a junction box. After the cell circuit is completed, an optional fiberglass layer is added to ensure cell ribbon isolation from the rear layer of the module. Additionally, this layer aids air evacuation and minimizes cell string movement during the lamination process.

BASIC MODULE ASSEMBLY

163

Figure 7.3. Photo of automated tabbing and stringing machine. Backsheet

Fiberglass

EVA

Cells

Glass

Figure 7.4. Picture of laminate layers.

A second sheet of encapsulant and the back-sheet layers are added to encapsulate the power-producing section of the module. While module designs may vary based on system requirements, all current designs require multiple power leads to egress the rear layer for string output monitoring. While PVF is still the dominant back-sheet material, increasing module production has stressed the available supply. This has encouraged additional material suppliers to explore other material options. These newer materials are beginning to supplement PVF shortages. Additionally, glass back sheets remain a viable option for the rear surface, especially in BIPV and in thin-film modules. Modern production equipment, such as laminators, must exhibit the flexibility to handle these various module constructions (Fig. 7.4). The layered package is placed into a lamination machine. The cycle time will depend on the curing time required for the encapsulant layers. Typically, the lamination cycle is divided into to two phases. First, the air is evacuated from the heated

164

TERRESTRIAL MODULE FABRICATION

processing chamber while the encapsulant layers melt. During the second phase, pressure is applied onto the laminate package to evacuate any remaining air. The total lamination time is approximately 10–20 min, depending on materials. After the laminate is permitted to cool, excess encapsulant and back-sheet material is trimmed; module frames, a junction box, and cables are added. Within the junction box, bypass diodes are incorporated to sense reverse current. A reverse current condition is possible if one or more cells are shaded or covered during daylight hours. Essentially, the cell would not pass the module current; the cell would overheat and potentially break the module glass. The installed bypass diodes protect the module until the shading is resolved. The final testing of modules can be a labor-intensive exercise. The module must be presented to a sun simulator. The module’s cables must be connected. Next, the module is transferred to a required high-voltage tester, where the module cables and frame are connected. The test results are used to rate the module and to monitor module line performance. Ultimately, after combining all of these module materials, the resulting module is expected to last in excess of 25 years.

7.4

EQUIPMENT OPTIONS

While module manufacturers endeavor to produce the best or the cheapest module to remain competitive, the equipment manufacturers must evolve to effectively handle new materials and larger module formats. The function of the equipment is to minimize handling and stress on the solar cells whenever possible. Some module manufacturers chose to establish or move factories to economically challenged areas, to take advantage of lower labor rates. Alternatively, more automation may be incorporated to reduce labor costs. The line concept has evolved to include powered conveyors and robotics. The addition of conveyors is a natural evolution as the solar industry matures. Conveyors are available to transport the glass through the factory as it transforms into a laminate and ultimately into a solar module. By utilizing conveyors, automation may be integrated at the various process steps. Essentially, by limiting operator contact, the resulting module can be produced faster and more consistently. Robots are now engaged throughout the module assembly process. The introduction of robotics was assisted by experience from other industries. While operators feed material into the work cells, robots perform the assembly steps and advance the module through the processes. For example, an automated 50-MW module assembly line utilizes robotics to load the module glass, to complete soldered cell string layup, and to install bus ribbon prior to the lamination process (Fig. 7.5). The fundamentals of busing and laminate assembly remain similar; however, the implementation of automation has allowed module manufacturers to manage these process steps more efficiently. For example, the utilization of robots transforms the soldering and assembly processes into a timed process step eliminating the dependence on an operator’s level of experience (Figs. 7.6 and 7.7).

EQUIPMENT OPTIONS

165

Figure 7.5. Photo of glass loading. A glass-loading robot eliminates labor from the beginning of the assembly process.

Figure 7.6. After cell soldering, the robot places completed strings onto the prepared glass.

Some module equipment is developed to reduce handling and others are developed to increase throughput. Larger lamination machines address both goals. Equipped with conveyors to automatically load and unload laminates, one laminator can now laminate four 220-W laminates per cycle. Coupled with the fastest cure EVA, just two laminators are necessary for a 50-MW fully automated module line (Fig. 7.8).

166

TERRESTRIAL MODULE FABRICATION

Figure 7.7. Robots with specially equipped soldering arms complete the module circuit.

Figure 7.8. Photo of larger area laminator. Larger area laminators integrated with conveyor systems provide higher throughput with minimal to no handling.

Additionally, robots perform the edge trimming, framing, and junction box placement process steps as the laminate travels along conveyors to the sun simulator (Figs. 7.9–7.11). Larger modules are very important to reducing system costs. While modules are purchased by module power (watts), the overall system costs are lower due to fewer junction boxes, cables, and frames. Additionally, fewer modules per installation reduce the number of DC connections and simplify system mounting requirements. This has led to the development of special equipment designed to handle the increased module sizes. Today, automation can eliminate manual lifting,

EQUIPMENT OPTIONS

167

Figure 7.9. Module frames are affixed in a robotic work cell.

Figure 7.10. Photos of framing and junction box application.

loading, and connecting from the operation. Automation can also apply the power label after testing. New ideas and initiatives that benefit the solar module manufacturing process will emerge from the increasing attention enjoyed throughout the renewable energy industries. Technology improvements will likely compliment other solar segments. For example, alternative and more efficient cell designs may be incorporated into concentrator and BIPV applications.

168

TERRESTRIAL MODULE FABRICATION

Figure 7.11. The completed module is conveyed over the sun simulator for electrical performance test results. Next, the module is transported through the high-voltage tester. Lastly, the performance and safety labels are affixed to the rear side of the module.

Figure 7.12. Picture of building integrated PV (BIPV) project at Denali National Park Visitor Center in Alaska.

Through additional governmental incentives, it becomes more popular to integrate solar products into new and existing structures. Attractive BIPV projects capitalize on the versatility of silicon cells, encapsulants, and other module materials (Fig. 7.12).

7.5

FUTURE

A combination of improved cell efficiency, reduced material costs, and governmental support shall increase the amount of electricity generated from silicon-

ABBREVIATIONS

169

based modules in the coming years. Ultimately, the renewed attention and global push for renewable energy solutions will invigorate the solar industry. The collaboration of ideas and people from other industries will further challenge today’s best practices and increase productivity while lowering system costs. While this growth is beneficial, in order for the industry to evolve, lower material costs are needed. The availability and quality of bus wire for cell-to-cell and circuit connections, encapsulating plastics, and composite back-sheet materials directly affect module manufacturing. While the basic module construction and manufacturing processes have changed little, the methods and equipment available have matured. ABBREVIATIONS BIPV—building integrated photovoltaics DC—direct current EVA—ethylene vinyl acetate PVF—polyvinyl fluoride PVB—polyvinyl butyral UV—ultraviolet

8 CHINESE SOLAR CELL STATUS WANG SICHENG Energy Research Institute, National Development and Reform Commission

8.1

INTRODUCTION (BY THE EDITOR)

Driven by the pressure of global warming, rising oil prices, and a deteriorating natural environment, clean renewable energy is receiving greater attention worldwide. China’s central government has stated its support for the development of a sustainable energy system that maximizes energy efficiency and the use of renewable energy sources. A key aspect of that initiative is the application of PV technology to convert light into electricity. The Chinese PV manufacturing industry has grown dramatically in recent years due to a strong demand from overseas markets. China’s production of solar cells and modules has grown at an average annual rate of 49.5% since 2002. By 2008, the production of solar cells reached over 2 GW, which was 33% of the global production. China now contributes a large part of the worldwide solar production and is now the largest producer of solar cells in the world. Intelligent and supportive Chinese central government policy has contributed to this rapid growth in the Chinese solar cell industry. This chapter provides an overview of the coordinated efforts in China to develop a strong domestic PV industry. This chapter was written by Wang Sicheng, who has contributed to the formation of China’s government policy. It was translated into English by Huang Han Xinag at JX Crystals Inc., and editorial comments are by Lewis M. Fraas. 8.2

CHINA’S SOLAR ENERGY RESOURCES AND DATA

8.2.1

Solar Irradiation Resources in China

PV power generation largely depends on the solar energy resources of the location. Therefore, it is necessary to know the distribution of the solar energy resources in Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

171

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

173

with 70 in the early 1950s. Of these, 2,300 are meteorological stations (county level) and 310 are observatories (including 122 standard observatories). Before 1993, there were 66 observatories gathering solar global, scatter, and direct irradiation data. After 1993, only 17 observatories kept the global and scattering data, while the others stopped measuring scattering and direct irradiation. As summarized in Table 8.1, currently only horizontal global irradiation data can be obtained from most observatories. 8.2.2

Solar Resource Overview and Global Comparison

In China, horizontal irradiation data can be obtained from local meteorological stations. Based on these data, the irradiation on slope solar modules can be computed. This irradiation is normally about 10–15% higher than the horizontal irradiation. The additional amount depends on the site, latitude, and the ratio of direct and indirect irradiation. With this input, the output power of a PV system can be derived from the irradiation on the slope PV arrays. The appendix at the end of this chapter presents tables for the solar radiation in various provinces in China. Table 8.2 presents solar resource data for three representative locations in China with comparison data from locations in the United States and Europe. Data are presented for modules mounted horizontally and tilted to the South. Data are also presented for modules mounted on 1-axis solar trackers. Note that the solar resource is always higher with tracking regardless of location, as discussed in Chapter 9 in this book. The eastern coastal region of China, which is depicted in Figure 8.1, has average solar irradiance levels (III). Shanghai is representative of this region in Table 8.2. While the irradiance in Shanghai is lower than most of the United States, including Boston, it is well above the irradiance levels in all of Germany. In contrast, the western region of China, as represented by Beijing (II), has higher irradiance levels. As Table 8.2 indicates, the irradiance level in Beijing is similar to that of Madrid, Spain, and Medford, Oregon, in the United States, although it is not as high as the southwestern United States. China also has areas where the solar resource is well above average. For example, in Tibet (I) the resource is higher than Phoenix, Arizona, which is located in the southwestern desert region of the United States. This higher solar irradiance in western China, together with China’s goal of promoting economic development in western China, motivates the development of solar energy.

8.3 PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA 8.3.1

Solar Cell R & D Level in the World

At present, the major restrictive factor for developing a PV power system is the high cost. Solar cells accounts for over 60% of the cost of a PV power system.

174

CHINESE SOLAR CELL STATUS

TABLE 8.2. Solar Availability in China Relative to United States and Europe Region

City and Latitude

China

United States

Europe

Annual Average Hours per Day on Horizontal Surface

Annual Average Hours per Day at Fixed Tilt

Tilt Angle

Annual Average Hours per Day 1-Axis Tracking

Beijing (39.6)

4.3

4.8

42

Shanghai (31)

3.6

4.0

35

Xizang (Tibet)

6.0

6.7

30

Medford, Oregon

4.4

4.9

42

6.5

Sacramento California

4.9

5.5

39

7.4

Boston, Massachusetts

3.9

4.6

42

5.7

Phoenix Arizona

5.7

6.5

33

8.6

Madrid (40), Spain

4.9

30

Lisbon (38), Portugal

5.4

30

Berlin (52), Germany

2.8

30

Therefore, developing low-cost, high-efficiency, high-reliability, high-stability, and long-lifetime solar cells becomes the highlight of the PV technology worldwide. Currently, the R & D of the solar cells mainly focuses on the commercialized Si crystalline solar cells, a-Si solar cells, CdTe solar cells, CIGS solar cells, and concentration solar cells. Attention is also given for next-generation solar cells, which have not been commercialized yet, including crystalline Si thin-film solar cells, dye-sensitized cells, organic thin-film cells, nanometer cells, and spectral absorption cells, through large finance and research efforts (Table 8.3).

8.3.2

PV Technology Innovations in China

In China, R & D of solar cells started in 1958. In 1971, the domestic solar cells were first used successfully on the Chinese satellite DFH-II. In 1973, solar cells were used in terrestrial applications. Since 1981, the R & D of solar cells and their applications have been listed on China’s national important scientific and technology plans. During the “Sixth to Eleventh 5-Year Plan” (2006–2010), solar cell components and application technologies have achieved great progress. Since

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

175

TABLE 8.3. World Leading Edge Levels of Solar Cells in Lab The World Highest Efficiency of Solar Cells in Lab Type of Cells

Efficiency (%)

Researcher

Note

Monocrystalline Si solar cell

24.7 ± 0.5

University of New South Wales, Australia

4 cm2

Back-leads concentration monocrystalline Si solar cell

26.8 ± 0.8

Sun Power, USA

96 X

GaAs multijunction cell

41.1 ± 0.5

Fraunhofer Institute, Germany

Concentration

Multicrystalline Si solar cell

20.3 ± 0.5

Fraunhofer Institute, Germany

1.002 cm2

InGaP/GaAs

30.28 ± 1.2

Japan Energy Co.

a-Si solar cell

14.5 (initial) ± 0.7

4 cm2

USSC, USA

0.27 cm2

12.8 (stable) ± 0.7 CIGS

19.5 ± 0.6

NREL, USA

0.410 cm2

CdTe

16.9 ± 0.5

NREL, USA

1.032 cm2

Multicrystalline Si thin-film cell

16.6 ± 0.4

University of Stuttgart, Germany

4.017 cm2

Nanocrystalline Si solar cell

10.1 ± 0.2

Japan

2-nm film

Dye-sensitized cell

11.0 ± 0.5

EPFL

0.25 cm2

HIT

21.5

Sanyo, Japan

2000, the national “863” project and “973” project have been initiated by the Science and Technology Ministry. The industrialization technologies and basic research of PV power have been supported from these projects. The important scientific achievements and technological innovations are described in the following sections. 8.3.2.1 Polysilicon Raw Material High-purity polysilicon raw material has always been a bottleneck for the development of China’s PV industry. Before 2005, all of the semiconductor-grade and solar cell-grade polysilicon material used in China was imported. In addition, import of related technologies has been limited under an embargo to China. Therefore, the development of high-purity polysilicon manufacturing technologies was listed in the national “Tenth 5-Year” Science and

176

CHINESE SOLAR CELL STATUS

TABLE 8.4. China’s Production Output and Requirement of High-Purity Polysilicon Year

2006

Solar cell production output (MW)

2007

2008

370

1088

2000

4000

11,000

20,000

300

1100

4500

Shortage (ton)

3700

9900

15,500

Self-fulfillment percentage

7.50%

High-purity polysilicon requirement (ton) High-purity polysilicon production output (ton)

10%

22.50%

Technology R & D Plan. Several refinement processes such as the Siemens method, flowing-bed method, and metallurgy method were investigated and brainstormed. At the same time, the R & D related to solving problems caused by waste materials and gases from production, and to protecting the environment, was also granted financial support. As a result, breakthroughs have been made on cleaner manufacturing technologies for high-purity polysilicon. For example, the production technology in some major high-purity polysilicon suppliers (Luoyang China Silicon, Sichuan Xinguang Silcon, Xuzhou China Energy, Emei Semiconductor, Zhongqing Daquan, etc.) has come up to the world average level. The production output has also been growing yearly. With even further scientific and technical efforts, the indexes of unit energy consumption, quality, purity, and clean production of China’s polysilicon manufacture should achieve advanced world standards. In addition, the quantity should fulfill the requirement of the country’s solar cell production (Table 8.4). 8.3.2.2 Crystalline Si Solar Cells Since the “Sixth 5-Year” Plan, crystalline Si solar cell technology has always been a key point for science and technology R & D. The technologies for monocrystalline solar cells, multicrystalline solar cells, high-efficiency solar cells, the industrialization for crystalline solar cells, and the development of special function solar modules (BIPV) have been given ample attention and funding and, therefore, have made encouraging progress. The efficiency of lab monocrystalline solar cells is up to 20.4% and the efficiency of production of multicrystalline solar cells is normally up to 16.5% (Fig. 8.2). In addition, China can now make its own 500-kg-grade polysilicon ingot furnaces (Fig. 8.3). The capacity of polysilicon ingot furnaces was up to 40-kg grade during the “Seventh 5-Year” period and to 160-kg grade in the “Tenth 5-Year” period.

8.3.3

Thin-Film Solar Cells

Given the relative low cost of thin-film solar cells, the development of these cells has always been highly supported by the Chinese government in PV power tech-

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

177

Figure 8.2. Monocrystalline and multicrystalline Si solar sells.

Figure 8.3. A 500-kg-grade polysilicon ingot furnace developed by China.

nologies. During the period of the “Seventh 5-Year” Plan, the Science and Technology Committee invested 20 million yuan to set up a 100-kW annual capacity thin film production line at the Beijing Non-Ferrous Metal Research Institute. Since the beginning of the twenty-first century, under the significant support of the government through the National Science and Technology R & D Plan, “863” project, “973” project, and other innovation projects, the R & D for industrialization technologies of thin-film solar cells such as amorphous, CdTe, CuInGaSe, dye sensitization, and micro/amorphous crystalline solar cells have shown remarkable progress. Thus there is now a solid foundation for the industrialization of the thinfilm solar cells in China. 8.3.3.1 CdTe Solar Cell The CdTe thin-film solar cell is a new type of solar cell. Compared with other types of thin-film cells, it has a lower cost and higher

178

CHINESE SOLAR CELL STATUS

Figure 8.4. A 300-kW CdTe solar cell pilot production line.

efficiency. The Science and Technology Ministry began supporting the research of CdTe thin-film solar cell by setting it as a subproject under the “Ninth 5-Year” Science and Technology R & D Plan. In 2006, under the support of the national “Tenth 5-Year 863” project, “the research and pilot production line of CdTe thinfilm solar cells,” a first CdTe thin-film solar cell pilot production line having a Chinese intellectual property right was built at the Solar Energy Material Research Institute of Sichuan University. Its annual production capacity is 300 kW for 40 × 40 cm, 8.25% efficiency cells. This achievement played a positive role in promoting the PV technologies in China (Fig. 8.4). 8.3.3.2 CIGS Solar Cells The CIGS solar cell is a thin-film solar cell with a promising future due to its low cost, high efficiency, high stability, and responsiveness to weak intensity. Under the support of the Science and Technology Ministry’s R & D project in the “Eighth 5-Year” to “Ninth 5-Year” Plans and the “863 key task” in the “Tenth 5-Year” Plan, the Nankai University has developed and built a CIGS experiment platform and technology transfer center. It has developed over 14% efficiency small-area CIGS cells on glass substrate and 9.2% and 10.6% efficiency CIGS cells on polyamide and on stainless steel substrates, respectively. The university has overcome several technical obstacles and completed a process flow for large-area cell experimental line and successfully made an 804-cm2 effective area, 7% efficiency cells. With these results, the university has leaped from a small-area lab technology to a large-area prototype technology and established a solid foundation for developing its own intellectual property right production lines (Fig. 8.5).

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

179

Figure 8.5. CIGS thin-film solar cells developed by Nankai University.

8.3.3.3 Si Substrate Thin-Film Solar Cells The Si substrate thin-film solar cell has the advantages of low cost, abundant raw material, no poison, no pollution, short energy payback time, and ease of large-scale (LS) continuous production. Under the support of the Science and Technology Ministry’s R & D schedule in four 5-Year Plans, from the “Fifth 5-Year” to “Ninth 5-Year” and the “793” task in the “Eleventh 5-Year” Plans, the Nankai University has made a breakthrough on a-Si/μc-Si tandem cell research—the initial efficiency of a 400-cm2 integrated tandem cell assembly reached 9.2%. As a result, the company Tianjin Jinneng Cell Science and Technology Ltd. has been established with the backing of the university. Under the support of the “973” task in the “Tenth 5-Year” and “Eleventh 5-Year” Plans, Nankai University carried out research for the next generation of Si thin-film cells—a-Si/μc-Si tandem solar cells—and developed a first μc-Si thinfilm solar cell experimental platform with advanced world standards. The efficiency reached 11.8% for small-area a-Si/μc-Si tandem cells and 9.7% for 10 × 10 cm integrated a-Si/μc-Si tandem cell assemblies. Recently, with the combination of VHF and HV techniques, at 12 A/s microcrystalline film deposition rate, the efficiency of a single-junction μc-Si cell has reached 9.36%, ranked in a world advanced level (Fig. 8.6). 8.3.3.4 Dye-Sensitized Solar Cell As early as 1994, the Plasma Institute of China Science Academy started researching and developing the dye-sensitized solar cells and achieved noticeable results. Especially, under the support of the Science and Technology Ministry’s projects in the “Tenth 5-Year” and “Eleventh 5-Year” Plans, the maximum efficiency of small-area solar cells reached 9% and large-area cells (40 × 60 cm2) reached 5.7%. This achievement is one of the better research results in the world at present. In addition, they also had some encouraging research results in the areas of dye synthetic technology, nanometer semiconductor thin film, and cell encapsulation and electrodes. Moreover, the institute has designed and synthesized many new dye light sensitization materials and investigated the materi-

180

CHINESE SOLAR CELL STATUS

Figure 8.6. PECVD system developed by Nankai University.

Figure 8.7. A 500-W solar power system with dye-sensitized solar cells.

als’ absorbability and sensitization behavior in space. With these key technologies, the solar modules were assembled using 15 × 20 cm2 cells. Under 1-sun intensity, the module’s efficiency was 5.9%. By the end of 2004, a 500-W dye-sensitized cell solar power model system was successfully built and placed in operation to test its stability. It made a solid foundation for future industrialization (Fig. 8.7). 8.3.3.5 Other Types of Solar Cells Under the support of the Science and Technology Ministry, encouraging progress has been made in research on other types of solar cells in China. Table 8.5 summarizes the lab efficiency of the solar cells in China.

PV SOLAR POWER SYSTEMS AND KEY EQUIPMENT

181

TABLE 8.5. The Max Efficiency of Other Types of Solar Cells in China Solar Cell Type

Max Efficiency (%)

Researcher

Area (cm2)

GaAs cell

29.25

Tianjing Electric Power Institute

1×1

CIGS

14.3

Nankai University

0.87

CdTe

13.38

Sichuan University

0.502

7.4

Plasma Institute, Science Academy

10.2 0.253

Dye-sensitized cell a-Si/µc-Si tandem solar cells

11.8

Namkai

HIT

17.27

Graduate School, Science Academy

8.4

1.2

PV SOLAR POWER SYSTEMS AND KEY EQUIPMENT

From the “Sixth 5-Year” to “Ninth 5-Year” Plans, the National Science and Technology R & D Projects focused on independent PV power systems. In this period, individual user power supplying systems, independent PV power stations, and related controllers and inverters were developed successively. Many solar power application model systems were built, spreading into PV power user homes, rural independent power stations and into the fields of broadcast, communication, oil, meteorology, PV pumping, and school PV education. In the period of the “Tenth 5-Year” to “Eleventh 5-Year” Plans, the focus of PV power R & D shifted to the grid-connected systems and relevant devices. Encouraging results have been achieved in grid-connected inverters, PV/building integrated engineering, LS gridconnected PV power stations, and PV array tracking systems. The following are some typical innovations in PV power systems and associated equipment.

8.4.1 Independent PV Power Stations and Solar/Wind Complementary Stations and Equipment From the “Seventh 5-Year” to “Tenth 5-Year” Plans, developing the technologies for independent PV power systems has received successive support from the country’s science and technology development plans. During this period, independent PV power systems from several dozen to hundreds of kilowatts, and the associated equipment of solar/wind complementary systems, were developed by mastering the technologies of integrated system design and key equipment production. Based on this development, during the “Tenth 5-Year” Plan, which was aimed at residents in off-grid remote western provinces, China popularized independent PV power application in LS. More than 1000 individual PV power stations and wind/solar complementary power stations were built. Hundreds of thousands of PV and wind/

182

CHINESE SOLAR CELL STATUS

solar home-using systems were promoted. The difficulties of living with electricity shortages has been resolved for about four million residents. As a result, the technology’s contribution to raising people’s living standard has been practically realized (Fig. 8.8). 8.4.2 Development of Serial Equipment for Grid-Connected PV Power Systems and Model Engineering During the “Tenth 5-Year” and “Eleventh 5-Year” Plans, China supported the development of equipment and model engineering for grid-connected PV power systems as a key point in its Science and Technology R & D Plan. Through the “Tenth 5-Year” Plan, breakthrough progress was made in developing grid-connected inverters, grid-connected PV/building integrations and LS grid-connected power systems in the desert. Larger power inverters were successively developed for grid-connected inverters from 3 to 100 kVA, mainly used for PV/building

Figure 8.8. Independent PV power station and control, and inverter equipment.

Figure 8.9. Grid-connected inverters (3 and 150 kVA) developed in the “Tenth 5-Year” Plan.

PV SOLAR POWER SYSTEMS AND KEY EQUIPMENT

183

integration, and small–medium inverters to 100–500 kVA, used in large desert power stations. In PV/building integration (BIPV), various integrated styles such as PV roof, PV wall and screen, PV fence, and PV shield were developed and successfully used on Beijing Olympic Studios and sports grounds. Many types of PV array trackers were also developed for large desert PV power systems. These successes have promoted China’s PV applications to advanced international level (Figs. 8.9–8.15).

Figure 8.10. A 500-kVA large power inverter developed in the “Eleventh 5-Year” Plan.

Figure 8.11. PV/building integration engineering completed in the “Tenth 5-Year” and “Eleventh 5-Year” Plans.

Figure 8.12. PV walls of the Olympic Studios and project lights at the Olympic Center.

184

CHINESE SOLAR CELL STATUS

Figure 8.13. Large desert power stations (100 and 500 kW) at Yangbajing, Tibet, and WuWei, Gansu province.

Figure 8.14. Single-axis and dual-axis trackers developed by the Electrical Engineering Institute (EEI), Science Academy of China.

Figure 8.15. A 50-kW horizontal tracker system developed by Kenuo Weiye Inc. of EEI.

THE GROWTH OF CHINA’S PV INDUSTRY

185

8.5

THE GROWTH OF CHINA’S PV INDUSTRY

8.5.1

China’s PV Power Industry Chain and Scale

The whole PV industry chain is composed of several production links. As shown in Figure 8.16, these links include high-purity Si material preparation, Si ingot and wafer production, solar cell fabrication, PV module assembly, and PV power system building. The output capacity of each link in China is listed on Table 8.6. The following are the key features of China’s PV industry:

Metallurgic silicon

Ultra-pure silicon Mono silicon ingots/wafers Multi silicon ingots/wafers Accessory material Mono-crystalline silicon cells Multi-crystalline silicon cells PV modules (Mono silicon, Multi silicon) Subsystems

PV power systems

Figure 8.16. PV industry chain.

TABLE 8.6. China’s PV Industry Status in 2008 Item

Year 2008

Solar cell production output

2000 MW (highest in world)

Capacities of industry chain

More than 40 Si material factories, including under construction (sum of capacity: 100,000 tons/ year); 60 Si ingot/wafer factories, total capacity: 3000 MW; 62 solar cell factories, total capacity 3000 MW; 330 solar module factories, total capacity: 4000 MW

PV industry output value

200 billion yuan (US$25 billion)

Employee

200,000

186

CHINESE SOLAR CELL STATUS

• • • • • •

8.5.2

Solar cell output has experienced rapid growth, reaching 2000 MW in 2008, the highest in world. In 2008, the output of high-purity Si material was 4500 tons in China. Seventy percent of the needs still have to be imported. The shortage problem should be relieved in 2010. Chinese PV enterprises now possess enough strength to enter into capital market. For example, by the end of 2008, 11 Chinese PV power enterprises emerged in overseas stock markets and 13 enterprises entered into domestic stock markets. Reliance on imported production equipment has changed. Now, the equipment of the whole production chain have been localized (including hightech equipment such as hydrogenation furnace for making high-purity Si, PECVD reactor, polysilicon ingot puller, and multiwire saw). China’s domestic PV market still remains stagnant. While in 2008 the output of solar cell production was 2000 MW, of this amount, only 40-MW cells were used domestically. Ninety-eight percent of the cells were exported. At present, more than 200 enterprises are making PV-related products, such as solar garden lamps, solar streetlights, traffic lights, solar energy toys, watches, and calculators. Ninety percent of the products are for exportation.

Current Status of China’s PV Industrial Chain

8.5.2.1 Polysilicon Raw Material To date, Si solar cells are the main products for commercial PV power systems. Worldwide, 98% of solar cells are made of high-purity polysilicon. As an essential material of solar cell production, highpurity polysilicon becomes the most important link in the PV industrial chain. Worldwide, in the production of high-purity polysilicon, a majority of the manufacturers adopted the “improved Siemens method—trichlorosilane reduction process”. This method accounts for 80% of the total output in the world. The total energy consumption of this high-purity polysilicon production is about 125– 170 kWh/kg Si. Before 2006, the polysilicon material market was dominated by 10 companies from developed countries. Table 8.4 shows their polysilicon production quantities from 2006 to 2008. The PV market stimulated rapid growth in the polysilicon industry. In addition to existing suppliers, who are expanding their production capacity with big steps, many companies are investing actively to build new polysilicon production lines. Being promoted by overseas PV markets, China’s polysilicon production and its technology are also growing and developing fast. By introducing, digesting, and absorbing advanced technologies of the improved Siemens method from Western countries, as well as innovations in the production by several enterprises, China’s polysilicon refining technology has caught up with the international advanced level. As an example, Xuzhou Zhongneng Silicon Business has realized an entire

THE GROWTH OF CHINA’S PV INDUSTRY

187

TABLE 8.7. China’s Polysilicon Production Capacity and Output from 2006 to 2008 Manufacturer

2006 (ton)

2007 (ton)

2008 (ton)

Capacity

Output

Capacity

Output

200

100

700

200

700

500

Luoyang China Si

300

200

1000

550

3000

1000

Sichuan Xinguang

1260

0

1260

210

1260

800

40

0

40

20

40

40

300

20

300

200

1500

100

Ermei Semiconductor Factory

Shanghai Lingguang Wuxi Zhongcai Xuzhou Zhongneng

4000

1800

2000

60

Others

8700

100

20,000

4500

3300

0

Output

Zhongqing Daquan Total

1500

Capacity

300

4550

1100

closed-loop and recycle, pollution-free production by using advanced processes including chlorine hydrogenation process, high-flow and fast-deposition regeneration, and highly effective recuperation. Its refining energy consumption index has been reduced to 127 kWh/kg, which is similar to international best level. The quality of the products reached the solar grade for its first-phase polysilicon and the electronic grade for the second-phase polysilicon. The polysilicon refining technologies actually have achieved a breakthrough progress. Table 8.7 shows the polysilicon output between 2006 and 2008 in China. Total output broke through the 1000-ton target in 2007 and reached 4500 tons in 2008. 8.5.2.2 Crystalline Si Ingot/Wafer Manufacturing crystalline Si ingots/wafers is the second link in the PV industry chain. Promoted by the PV market, crystalline Si ingot/wafer manufacture is growing fast. Up to now, there are 70 Si ingot/wafer factories in China. Estimated output of Si ingots/wafers in 2008 was 20,000 tons. The technologies in monocrystalline ingot production and its process have reached advanced world level. For some leading-edge production lines, the energy consumption index is 62 kWh/kg. Recently, the output of crystalline ingots has been growing quickly in China. Table 8.8 gives the production quantities of solargrade crystalline ingots for major production factories. The sum of the solar-grade crystalline ingot output was 2216, 4550, and 8070 tons for years 2005, 2006, and 2007, respectively. In China, up to the end of 2007, there were 2400 Si crystalline pulling machines; the total capacity was about 14,400 tons/year. The crystalline

188

CHINESE SOLAR CELL STATUS

pulling machine is the key equipment in crystalline Si ingot production. The technology for making crystal pullers is very mature in China. It has about 40 years of history in the production of pullers. Recently, the crystal puller production quantity has been increasing dramatically. Sales grew from 8 million yuan in 2004 to 800 million yuan in 2007. The quality of domestic pullers fully meets the requirements of the PV industry. The price is also only one-third to one-half that of imported counterparts. Now, in China, the total number of various crystalline pullers exceeds 2400 and most of them are domestic products. There are more than 10 enterprises manufacturing crystalline pullers in China. In 2006 and 2007, they sold 400 and 800 crystalline pullers, respectively, which not only satisfied the needs of the domestic market but also allowed for exportation in large quantities. For polysilicon ingot production, the technology is more sophisticated than for crystalline Si ingots. Therefore, there are less polysilicon ingot manufacturers in China. The energy consumption in polysilicon ingot production is around 33– 55 kWh/kg in China. Table 8.9 shows China’s major polysilicon ingot manufacturers and their output. The sum of polysilicon ingot output was 300, 1120, and 3700 tons in years 2005, 2006, and 2007, respectively. Up to the end of 2007, there were 230 polysilicon casting furnaces; total capacity was 7000 tons/year. In year 2008, the output was 20,000 tons (Tables 8.8 and 8.9). Promoted by market needs, the localization speed of the polysilicon casting furnaces is increasing, and the situation of importing the casting equipment is changing. Now, several enterprises, such as the forty-eighth Institute of China Electronic Science & Technology Group and Beijing Yuntong Inc., are developing the technologies for manufacturing polysilicon casting furnaces. They have developed 240–270 kg and 400–450 kg qualified serial furnaces and placed them in the market. Wafer sawing is a key process in Si wafer production. The quality of sawing directly affects the following processes in the production chain. As an advanced technology, the multiwire wafer saw is now gradually replacing the traditional inner wheel saw and has become the major technology in the production. Now, many factories in China are using a multiwire saw. The thickness of the cut wafer can be down to 180–200 μm, and the energy consumption is about 220–300 kWh/ kWp. With the most advanced multiwire saw, the wafers can be cut as thin as 180 μm; one 400-kg Si ingot can make 13,000 wafers plus. At present, in China, most multiwire saws are imported. Now, some Chinese companies are developing special wire saws for cutting solar-grade wafers. For example, a wire saw developed by a company in Shanghai is now under trial. Another wire saw made by a company in Guansu has passed product appraisal.

8.5.2.3 Solar Cell Manufacture 8.5.2.3.1 Crystalline Si Solar Cell Manufacture Crystalline solar cells are made by means of forming a diffused junction on the monocrystalline or multicrystalline wafers and some other related processes. The process flow for manufacturing com-

THE GROWTH OF CHINA’S PV INDUSTRY

189

TABLE 8.8. Crystalline Ingot Output in 2006 and 2007 (Ton) No.

Manufacturer

2005

2006

2007

Output

Output

Output

1

Nengjin Jinglong

1126

1250

1500

2

Zhejiang Zhihui

300

750

1100

3

Jinzhou Xinri

400

750

900

4

Yangzhong Huantai

350

900

5

Changzhou Tianhe

300

680

6

Yangchou Shunda

250

500

7

Shanghai Carmdanc

30

450

8

Changzhou Yijing

200

300

9

Jiangyin Hairun

40

200

100

150

100 80

10

Beijing Longfang

11

Zhejiang Jiashan

120

12

Neimongol Hushi

100

13

Zhejiang Kaihua

70

100

14

Shanghai Hejing

30

80

15

Shanghai Songjiang

30

70

16

Xinjiang Xinnengyuan

60

17

Huzhou Xinyuantai

50

18

Shanghai Jiujing

20

50

19

Tianjin Huanou

20

40

20

Others

210

360

710

21

Total

2216

4550

8070

TABLE 8.9. Polysilicon Ingot Output of China in 2006, 2007 (Ton) Manufacturer

2005

LDK Baoding Yingli

260

2006

2007

450

2300

550

1200

Changzhou Tianhe

120 80

80

Ningbo Jingyuan

Shaoxing Jinggong 40

40

40

Total polysilicon

300

1120

3740

190

CHINESE SOLAR CELL STATUS

mercialized crystalline solar cells similarly consists of mechanical damage removal, suede-surface formation, electrode printing and baking, and so on. Under the strong push from the international PV market, China’s solar cell manufacturing grew rapidly in 2006 and 2007. In spite of the restriction from the lack of raw material, there were still new companies successively taking part in the PV industry. Up to the end of 2007, there were 50 enterprises engaging in solar cell manufacturing. Based on the calculations from the numbers of the equipment being used, the capacity of the cell reproduction reached 2900 MWp (including 100 MWp of amorphous thin-film solar cells). As an example, the solar cell outputs of Wuxi Suntech were 157.5 MWp in 2006 and 327 MWp in 2007, and accounted for 35.9% and 30.1% of the total cell outputs of China. In 2006 and 2007, the total solar cell outputs in China were 438 and 1088 MWp, respectively. Among them, 426 and 1059.7 MWp were of crystalline solar cells. In 2008, China’s solar cell output was estimated at 2000 MWp. The LS solar cell production in China has been realized basically through importing production lines from overseas. Because the latest equipment and related technologies can be acquired from the international market, the production technologies for some leading enterprises in China can keep pace with the advanced countries. The average energy consumption from wafer to cell is 220 kWh/kWp, which is similar to the rest of world. Many enterprises and institutes have been attracted by the fast growth of the Chinese PV industry and developed a lot of advanced solar cell production equipment. For example, diffusion furnaces, plasma etchers, baking ovens, and wet cleaning benches have already been localized. Now, about half of the production lines in China have used homemade wet cleaning benches. And homemade semiautomatic screen printers have been used in large quantities. Automatic screen printers and automatic sorters are also under development and expected to be in the market in 1–2 years.

8.5.2.3.2 a-Si Thin-Film Cell Manufacture Recently, because of low cost and good performance under weak light intensity, amorphous thin-film solar cells have received significant attention and a-Si cell manufacture has grown rapidly. Before 2004, in China, a-Si cells were mainly single-junction cells. After Tianjing Jingneng Inc. introduced the 2.5-MWp dual-junction amorphous cell production line in 2004, amorphous dual-junction manufacturing developed quickly. Up to the end of 2007, in China, there were 10 enterprises producing a-Si cells with a total capacity of 80 MWp. In 2006 and 2007, the output of amorphous thin-film cells was 12 and 28.5 MWp, respectively. Since 2004, China’s a-Si cell manufacture has entered a rapid growth period. Three factors have encouraged the fast development of thin-film cell manufacture—prosperous world PV market, advances in technology for thin-film cell manufacturing, and relief from the shortage of polysilicon raw material, which had largely restricted the development of crystalline Si PV manufacture before the world economic crisis. Now, many Chinese enterprises are building and introducing more sophisticated a-Si/μc-Si tandem cell production lines. For example,

THE GROWTH OF CHINA’S PV INDUSTRY

191

Suntech Power (Shanghai) and Hebei Xinao Inc. are importing 50-MWp dualjunction a-Si/μc-Si cell production lines, respectively. Meanwhile, some enterprises are setting up their own R & D centers, manufacturing equipment by themselves, and expanding their production capacities. China’s thin-film cell industry should progress significantly after these new production lines having been completed and are in production. The thin-film cells are still facing restrictions on development from issues such as low efficiency, degeneration of performance, shorter lifetime compared with crystalline cell, and low market acceptability. At the moment, in the industrialization of thin-film cells, the technologies are still improving. The production equipment has not been finalized in design yet. And the initial investment is also very high (Table 8.10).

TABLE 8.10. a-Si Solar Cell Production Capacity and Output in Year 2008 (MWp) China’s a-Si solar cell production capacity and output (ton) No.

Manufacturer

Cell

Capacity

Output

1

Tuori New Energy, Shenzhen

a-Si

25

11

2

Suosaisi New Energy, Shanghai

a-Si

20

6

3

Chuangyi Tech, Shenzhen

a-Si

6

5

4

Shihua Chuangxin Tech, Beijing

a-Si

15

4

5

Jinneng Battery, Tianjin

a-Si

2.5

2.5

6

Qingfeng E-Light, Shenzhen

a-Si

1.5

1.5

7

Minghuan Solar, Shengzhen

a-Si

2

2

8

Golden Solar, Fujian

a-Si

12

2

9

Fusheng Solar Energy, Zhejinag

a-Si

1.5

1

10

Hengyang E-Light, Shenzhen

a-Si

1.2

1

11

Gerui Solar Energy, Harbin

a-Si

1

1

12

Hake New Energy, Heilongjiang

a-Si

1.2

13

Jiangsheng E-Light, Nantong

a-Si

25

Start 08

14

Furi PV, Shandong

CIGS

60

Start 09

0.9

15

Pule, Bangbu

a-Si

12

Start 08

16

Xinao Solar Energy, Hebei

a-Si

60

Start 08

17

SunTech Thin-Film, Shanghai

a-Si

40

Start 09

18

Best Solar (LDK)

a-Si

1000

09?

1174.9

35.9

Total

192

8.5.3

CHINESE SOLAR CELL STATUS

PV Module Assembly

The crystalline solar cells are finally connected to each other and encapsulated in a PV module form by lamination procedures in order to prevent the electrodes and cells from being eroded and cracked. The quality of lamination directly relates to the lifetime of PV modules. At the moment, in China’s PV industrial chain, the PV module manufacture is the link having the most mature production technology, highest equipment localization rate, lowest operation threshold, most enterprises, fastest expanding speed, and highest output. According to incomplete statistics, now in China, there are over 200 PV module manufacturers. The module lamination capacity was 3800 MWp for 2007 and 5000 MWp for 2008. Table 8.11 shows module production output in MW in China in 2006 and 2007. Because Chinese PV power market is still under development, most of the PV modules made in China have been exported, mainly to European and American countries. In 2007, the output of PV modules in China accounted for 28% of the world and ranked first in the world. As with solar cells, the latest equipment and related technologies for modules can be bought from the international market. The PV module assembly technologies in top Chinese manufacturers are now similar to advanced international manufacturers. The energy consumption index is 44 kWh/ TABLE 8.11. Module Production Output in MW in China in 2006 and 2007 No.

Manufacturer

2006

2007

No.

Manufacturer

2006

2007

5

20

1

Suntech, Wuxi

150

364

16

Erquan, Wuxi

2

Yingli, Baoding

55

150

17

Yinjing, Changzhou



20

3

Jingao

30

130

18

Zhongsheng, Jiangsu



15

4

Jiawei, Shenzhen

45

120

19

Rixing, Wuhan

10

15

5

Lingyang, Jiangsu

45

82

20

BP Jiayang

10

15

6

Adesi, Suzhou

25

82

21

Habo, Beijing

10

12

7

Tianhe, Changzhou

30

80

22

Guofei, Wuxi

8

10

8

Zhongdian, Nanjing

30

80

23

Huayuan, Donguan

5

10

9

Solar E, Ningbo

40

70

24

Junma, Xiameng

5

10

11

Shangpin, Wusi

25

50

25

Dongying, Shandong

5

10

10

SpaceTech, Shanghai

50

50

26

Chao, Shanghai

5

10

12

Tianda, Yunnan

7

30

27

Shunda, Jiangsu

5

10

13

Top Solar

20

30

28

Junxin, Jiangyin

5

8

14

Jingci, Tianjing

10

25

29

Zhongqing, Beijing

15

Sunflower, Zhejiang

5

25

30

Shanshan Youlika

78

150

Others

Total



7

3

7

721

1717

THE GROWTH OF CHINA’S PV POWER MARKET

195

TABLE 8.13. System Price as Influenced by Si Feedstock Price Si Material ($/kg)

Solar Module Price ($/Wp)

System Cost ($/Wp)

Gross Profit ($/Wp)

System Price ($/Wp)

250

4.00

6.05

1.15

7.20

200

3.44

5.39

1.03

6.42

150

2.97

4.81

0.92

5.73

100

2.47

4.24

0.81

5.05

50

2.00

3.69

0.70

4.39

TABLE 8.14. Projected Cost of Solar Electricity versus System Price and Annual Available Sunlight in Kilowatt Hour per Kilowatt Initial Invest. ($/W)

BIPV (1500 h)

BIPV (2000 h)

Electricity Price ($/kWh) $6

0.48

0.37

$4.5

0.36

0.28

$3

0.24

0.19

$2.25

0.18

0.14

$1.50

0.12

0.09

8.7

THE GROWTH OF CHINA’S PV POWER MARKET

Entering the twenty-first century, with the vigorous push and backing of the Chinese government, China’s PV power market achieved rapid growth. A series of state projects has been launched, including the “Tibet no-electricity counties renovation project,” “China bright project,” “Tibet Ngari PV power project,” “township electricity project,” and “the construction project in regions without electricity.” During the Ninth to Eleventh 5-Year Plan period, the Chinese government carried out a number of demonstration projects on city grid-connected PV systems and LS-PV systems in the desert. The Chinese government also missed a few opportunities to seek international assistance, greatly encouraging PV power in rural electrification. China’s renewable energy law took effect in 2006; the Chinese government departments (NDRC, Department of Science and Technology, Ministry of Construction, Ministry of Finance, Ministry of Information Industry, and Ministry of Agriculture) have actively promoted the PV power by launching

THE GROWTH OF CHINA’S PV POWER MARKET

197

8.7.1.2 BIPV Power System More than 70% of solar PV power systems are currently for on-grid power systems, primarily installed on buildings in the city (BIPV). According to long- and medium-term plans, the total installed capacity of China’s BIPV systems will be 50 MWp by 2010, and 1000 MWp by 2020. There are about 4 billion square meters of roof area in China, and about 5 billion square meters of south elevations. If 20% of this area is used to install solar systems, the total capacity would be 100 GWp. 8.7.1.3 LS-PV System in the Desert According to China’s long- and mediumterm plans for renewable energy development, by 2010, a number of 1∼10 MWp desert experimental power stations will be built in Gansu, Tibet, and Inner Mongolia (total installed capacity: 50 MWp). After further promotion during 2010–2020, by the end of 2020, the total installed capacity of LS-PV will reach 200 MWp. Twelve percent of China’s land area, or 1.2 million square kilometers, is nonfarmable desert and mudflat, mainly distributed in the rich solar radiation regions of northwest China. Its annual global solar radiation is over 1600 kWh/m2. If 1% of the desert area is used to install solar systems, the total capacity would be 1000 GWp, twice China’s national power installed capacity in 2006.

8.7.2 National Development and Reform Committee’s “Long- and Medium-Term Plans for Renewable Energy Development” The Chinese government attaches great importance to renewable energy development. On August 31, 2007, the National Development and Reform Committee released the “long- and medium-term plans for renewable energy development” (FGNY [2007] 2174); on March 3, 2008, the committee republished the Renewable Energy Development in the 11th Five-Year Plan (FGNY [2008] 610), which further defined the development objectives for renewable energy (Table 8.16).

TABLE 8.16. China’s Renewable Energy Development Plan Objectives Year

Wind power BGPG Solar power (thermal power)

2004

2010

2020

Capacity (10 MW)

126

1000

3000

Annual generation (TWh)

18.9

210

690

Capacity (10 MW)

200

550

3000

Annual generation (TWh)

51.8

212

835

Capacity (10 MW)

7.0

30 (5)

180 (20)

Annual generation (TWh)

0.84

3.9 (0.65)

23.4 (2.6)

198

CHINESE SOLAR CELL STATUS

TABLE 8.17. 2010 China’s PV Power Marketing Plan Objectives Market Classification

Grand Total Capacity (MWp)

Market Percentage (%)

Rural electrification

80

32

Communication and industry

40

16

PV products

30

12

BIPV

50

20

LS-PV

50

20

250

100

Total

TABLE 8.18. 2020 China’s PV Power Marketing Plan Objectives Market Classification

Grand Total Capacity (MWp)

Market Percentage (%)

Rural electrification

200

12.5

Communication and industry

100

6.25

PV products

100

6.25

BIPV

1000

62.5

LS-PV

200

12.5

Total

1600

100

China’s PV power development objectives for 2010 and 2020 are shown in Tables 8.17 and 8.18.

8.8

CHINA’S INCENTIVE POLICY FOR PV POWER

8.8.1

China’s Renewable Energy Law

On February 28, 2005, China’s renewable energy law was approved by the National People’s Congress and was supposed to be effective on January 1, 2006. Basically, China’s renewable energy law is similar to Germany’s “on-grid price” policy: the initial investment of the power system is borne by the developer, the developer builds the on-grid PV power system with the administrative licensing, recovers the initial cost, and makes a profit by selling the electricity produced by the PV system. Power grid utility companies offer fair prices (cost plus a reasonable profit) to buy the full output of PV electricity. The power companies will not bear the cost beyond the conventional price but will collect the electrovalence surcharge from the users of the national electricity grid systems.

CHINA’S INCENTIVE POLICY FOR PV POWER

8.8.2

199

Other Supporting Policies to Renewable Energy Law

In 2007, the State Electricity Regulatory Commission issued its twenty-fifth decree of the year, “supervisory method of power grid companies purchases the full quantity of renewable energy power” (effective on September 1, 2007), to emphasize that according to the renewable energy law, power grid companies must purchase the renewable energy power as a priority, and offer the grid integration services (the cost of grid integration services will be included in the electrovalence surcharge, apportioned by the users of the national electricity grid systems). On August 2, 2007, the General Office of the State Council transmitted the “energy-saving power dispatch regulation (draft)” (GBF [2007] 53), jointly issued by the National Development and Reform Committee, State Environmental Protection Administration, State Electricity Regulatory Commission, and State Energy Bureau, to emphasize that power grid enterprises should dispatch renewable energy power as a priority, as long as this could be done without prejudice to the adequate supply of electricity. On January 11, 2007, the National Development and Reform Committee issued the “interim provisions of deployment of renewable energy power surcharge” (FGJG [2007] 44). Renewable energy law defines that, when power grid enterprises purchase the renewable energy power, the cost difference between renewable energy power and conventional energy power will be allocated in the sales rate of electricity. The “interim provisions” clearly define the collection standard, collection scope and allocation, and balance decision of the surcharge of renewable energy power, as well as the procedure of applying for renewable energy power electrovalence subsidy, electricity price settlement, and surcharge quota allocation. To off-grid-independent power systems, the subsidies for O&M are also stipulated in “interim allocation plan for the price and cost of the renewable energy” (FGJG [2006] 7) and “interim provision of deployment of renewable energy power surcharge” (FGJG [2007] 44): the overportion of the operation cost of the independent renewable power systems, which are invested or supported by the country, to the average sale price of the electricity of the local utility will be compensated by the surcharge in the renewable energy electricity price. The application procedures and the method of compensation calculation are the same for the grid-connected power systems. From August 2006 to June 2008, a renewable energy surcharge of 0.1¢ per kilowatt hour was collected countrywide. In June 2008, this increased to 0.2¢ per kilowatt hour. This surcharge is used to subsidize renewable electricity (including wind electricity, bioelectricity, and PV electricity). The total amount of the collection is about $500 million. If $250 million of the surcharge funds are used to subsidize the grid-connected PV power and the annual generation of the PV power is 1300 kWh/kW, then 480-MW power will be supported for 4 yuan per kilowatt hour subsidy, or 960-MW power for 2 yuan per kilowatt hour subsidy, and or 1.92-GW power for 1 yuan subsidy. Unfortunately, up to now, 3 years after the renewable energy law took effect, because of the indecisiveness of the electricity price administration, PV power has not enjoyed its reasonable on-grid price.

200

CHINESE SOLAR CELL STATUS

ABBREVIATIONS a-Si—amorphous silicon BGPG—biomass power BIPV—building-integrated photovoltaics CdTe—cadmium telluride CIGS—copper–indium–gallium–selenium Co.—company CuInGaSe—copper indium gallium diselenide EPFL—Ecole Polytechnique Federale de Lausanne, Switzerland FGJG—Price Bureau of National Development and Reform Commission FGNY—Energy Bureau of National Development and Reform Commission GaAs—gallium arsenide GaP—gallium phosphide GBF—issued by the General Office of State Council GWp—gigawatt (billion watts) peak HIT—heterojunction with intrinsic thin layer HV—high voltage kWp—kilowatt peak LS—large scale max—maximum min—minimum MWp—megawatt peak NDRC—National Development and Reform Commission, China NREL—National Renewable Energy Laboratory, USA O&M—operation and maintenance PECVD—plasma-enhanced chemical vapor deposition p-Si—polycrystalline silicon PV—photovoltaic R & D—research and development Si—silicon u-Si—ultrapure silicon USSC—United Solar System Corporation VHF—very high frequency Wp—watt peak X—sunlight concentration ratio μc-Si—microcrystalline silicon

APPENDIX: CHINA SOLAR RESOURCE DATA Here, we defined the sum of irradiation on a tilted PV array as “annual usable hours.” The following tables show the average annual usable hours for each province of China.

APPENDIX: CHINA SOLAR RESOURCE DATA

201

TABLE 8A.1. Statistics of Solar Irradiation Resources of China’s 17 Southeastern Coastal Provinces and Cities Province

Annual Annual Irradiation Irradiation Max Min (MJ/m2) (MJ/m2)

Sum of Annual Horizontal Irradiation of the Capital (MJ/m2)

Horizontal Annual Usable Hours of the Capital (h)

PV Array Slope Angle (Degree)

Slope Annual Usable Hours of the Capital (h)

Heilongjiang

4683.69

4442.92

4683.69

1301.03

50

1496.179

Hebei

5008.89

5008.89

5008.89

1391.36

42

1558.323

Guangxi

4595.91

4294.11

4595.91

1276.64

25

1378.773

Jiling

5034.39

4640.64

5034.39

1398.44

45

1608.209

Guangdong

5161.46

4478.03

4478.03

1243.90

25

1343.41

Hubei

4312.92

4047.91

4312.92

1198.03

35

1377.74

Shendong

5123.01

4761.44

5123.01

1423.06

40

1636.516

Henan

5095

4764.36

4764.36

1323.43

40

1521.95

Liaoning

5068.67

4903.14

5067.41

1407.61

45

1618.757

Jiangxi

5045.26

4630.6

4832.08

1342.24

30

1449.624

Jiangsu

4855.49

4855.49

4855.49

1348.75

35

1483.621

Fujian

4410.74

4410.74

4410.74

1225.21

30

1323.221

Zhejiang

4751.82

4314.6

4314.6

1198.50

35

1318.349

Hinan

5125.1

5125.1

5125.1

1423.64

25

1494.82

Beijing

5620.01

5620.01

5620.01

1561.11

42

1748.448

Tianjin

5260.11

5260.11

5260.11

1461.14

42

1636.479

Shanghai

4729.25

4729.25

4729.25

1313.68

35

1445.049

Average

4985.16

4872.12

4912.75

1364.65

1502.04

202 TABLE 8A.2. Statistics of Solar Energy Resources of China’s Nine Western Provinces Province

Annual Irradiation Max (MJ/m2)

Annual Irradiation Min (MJ/m2)

Sum of Annual Horizontal Irradiation of the Capital (MJ/m2)

Horizontal Annual Usable Hours of the Capital (h)

PV Array Slope Angle (Degree)

Sum of Annual Irradiation on a Slope PV Array in the Capital (MJ/m2)

Slope annual Usable Hours of the Capital (h)

Xinjiang

6342.31

5304.84

5304.84

1473.57

45

6100.56

1694.601

Xizang

7910.65

6088.59

7885.99

2190.55

30

8832.31

2453.418

Inner Mongol

6195.18

5658.47

6041.35

1678.15

45

6947.56

1929.877

Qinghai

6951.76

6142.93

6142.93

1706.37

40

7064.37

1962.324

Gansu

6458.52

5442.78

5442.78

1511.88

40

6259.19

1738.665

Ningxia

5944.8

5944.8

5944.8

1651.33

42

6658.17

1849.492

Shanxi

5868.3

5513.84

5513.84

1531.62

40

6340.91

1761.365

Shaanxi

4730.51

4730.51

4730.51

1314.03

40

5440.08

1511.134

Yunnan

6156.72

4848.38

5182.78

1439.66

28

5597.4

1554.835

Average

6284.31

5519.46

5798.87

1610.80

38.89

6582.28

1828.41

TABLE 8A.3. Statistics of Solar Resources of China’s Lowest Irradiation in Four Provinces Annual Irradiation Max (MJ/m2)

Annual Irradiation Min (MJ/m2)

Sum of Annual Horizontal Irradiation of the Capital (MJ/m2)

Horizontal Annual Usable Hours of the Capital (h)

PV Array Slope Angle (Degree)

Slope Annual Usable Hours of the Capital (h)

Hunan

4212.60

4212.60

4212.60

1170.17

30

1287.19

Anhui

3792.51

3792.51

3792.51

1053.48

35

1211.50

Province

Sichuan

4229.74

3486.96

3792.51

1053.48

35

1179.89

Guizhou

4672.40

3471.07

3797.95

1054.99

30

1139.38

Average

4226.81

3740.79

3898.89

1083.03

1204.49

TABLE 8A.4. Comprehensive Efficiency for Different Power Systems PV Power System Type

Comprehensive Efficiency (%)

Individual PV power station

70

Grid-linked system on building

80

Grid-linked system on open field

85

TABLE 8A.5. Effective Annual Usable Hours for Different PV Power Systems in Each Province Region

Northwestern

Province

Annual Usable Hours (h)

Effective Hours per Year of Individual PV System (h/year)

Effective Hours per Year on Building a Grid-Linked System (h/year)

Effective Hours per Year on Open Field GridLinked System (h/year)

Xinjiang

1694.60

1186.22

1355.68

1440.41

Xizang

2453.40

1717.38

1962.72

2085.39

Inner Mongol

1929.90

1350.93

1543.92

1640.42

Qinghai

1962.30

1373.61

1569.84

1667.96

Gansu

1738.70

1217.09

1390.96

1477.90

Ningxia

1849.50

1294.65

1479.60

1572.08

Shenxi

1761.40

1232.98

1409.12

1497.19

Shaanxi

1511.10

1057.77

1208.88

1284.44

Yunnan

1554.80

1088.36

1243.84

1321.58

204

CHINESE SOLAR CELL STATUS

TABLE 8A.5. Effective Annual Usable Hours for Different PV Power Systems in Each Province (continued) Region

Eastern coastal

Poor resource region

Average

Province

Annual Usable Hours (h)

Effective Hours per Year of Individual PV System (h/year)

Effective Hours per Year on Building a Grid-Linked System (h/year)

Effective Hours per Year on Open Field GridLinked System (h/year)

Heilongjiang

1496.20

1047.34

1196.96

1271.77

Jilin

1608.20

1125.74

1286.56

1366.97

Liaoning

1618.80

1133.16

1295.04

1375.98

Hebei

1558.30

1090.81

1246.64

1324.56

Henan

1521.90

1065.33

1217.52

1293.62

Shendong

1636.50

1145.55

1309.20

1391.03

Guangdong

1343.40

940.38

1074.72

1141.89

Guangxi

1378.80

965.16

1103.04

1171.98

Hubei

1377.70

964.39

1102.16

1171.05

Jiangxi

1449.60

1014.72

1159.68

1232.16

Jiangsu

1483.60

1038.52

1186.88

1261.06

Fujian

1323.20

926.24

1058.56

1124.72

Zhejiang

1318.30

922.81

1054.64

1120.56

Hainan

1494.80

1046.36

1195.84

1270.58

Beijing

1748.40

1223.88

1398.72

1486.14

Tianjing

1459.70

1021.79

1167.76

1240.75

Shanghai

1445.00

1011.50

1156.00

1228.25

Hunan

1287.20

901.04

1029.76

1094.12

Anhui

1211.50

848.05

969.20

1029.78

Sichuan

1179.90

825.93

943.92

1002.92

Guizhou

1139.40

797.58

911.52

968.49

1551.2

1085.8

1241.0

1318.5

Note: Annual effective usable hours = annual usable hours × comprehensive efficiency of the system.

The average effective usage hours for different PV power system types: the following tables show the statistical results for China’s 26 provinces, regions, and cities, except for the four poor resources provinces.

APPENDIX: CHINA SOLAR RESOURCE DATA

205

TABLE 8A.6. The Annual Effective Usage Hours for Different System Types in China Annual Irradiation on Horizon (kWh/m2)

Annual Irradiation on Slope (kWh/m2)

Effective Usage Hours of Individual Power Systems

Effective Usage Hours of Grid-Linked on Building Systems

Effective Usage Hours of Grid-Linked on Field Systems

Northwestern

1610.80

1828.41

1250

1450

1540

Coastal

1364.65

1502.04

1000

1200

1250

Countrywide average

1487.73

1665.23

1100

1250

1350

Region

The following are the average effective usage hours for different regions: 1.

Northwestern region

TABLE 8A.7. The Average Effective Annual Usages Hours of Different Types of Systems in the Northwestern Region of China Classification

Resources Hours

Effective Usage Hours of Individual Power Systems

Effective Usage Hours of Grid Linked on Building Systems

Effective Usage Hours of Grid Linked on Field Systems

Year max

2453.4

1717.4

1962.7

2085.4

Year min

1511.1

1057.8

1208.9

1284.4

Year average

1828.4

1280.0

1462.7

1544.1

Day max

6.72

4.6

5.0

5.7

Day min

4.14

2.9

3.3

3.5

Day average

5.00

3.5

4.0

4.2

Based on the above statistics, the average annual effective usage hours in the northwestern region for different system types are the following: individual PV power system: 1200 h; on building grid-linked PV power system: 1400 h; and on field grid-linked PV power system: 1500 h. Editor’s note: sun tracking will increase these annual effective usage hours by about 30%.

206

2.

CHINESE SOLAR CELL STATUS

Eastern region

TABLE 8A.8. The Average Effective Annual Usage Hours of Different Types of Systems in the Eastern Region of China Classification

Resources Hours

Effective Usage Hours of Individual Power Systems

Effective Usage Hours of Grid Linked on Building Systems

Effective Usage Hours of Grid Linked on Field Systems

Year max

1748.4

1223.9

1398.7

1486.1

Year min

1318.3

922.8

1054.6

1120.6

Year average

1486.0

1040.2

1188.8

1263.1

Dairy max

4.79

3.4

3.8

4.1

Day min

3.61

2.5

2.7

3.1

Day average

4.07

2.8

3.3

3.5

Based on the above statistics, the average annual effective usage hours in the eastern region for different system types are the following: individual PV power system: 1000 h; on building grid-linked PV power system: 1200 h; and on field grid-linked PV power system: 1300 h. Editor’s note: sun tracking will increase these annual effective usage hours by about 30%.

9 TRACKING THE SUN FOR MORE KILOWATT HOUR AND LOWERCOST SOLAR ELECTRICITY RON CORIO,1 MICHAEL REED,1 AND LEWIS FRAAS2 1 Array Technologies, Inc., 2JX Crystals Inc.

9.1

INTRODUCTION

In the early history of terrestrial solar cells, the remote power applications required smaller expensive solar modules that were generally mounted in a fixed position. The up-front cost of the solar module in U.S. dollar per watt was the dominant economic metric, and the early government solar subsidies in California were based on this U.S. dollar per watt metric. Tracking the sun for these smaller systems simply added cost, and there was no financial incentive to do so. However, more recently, solar cell systems have gotten larger, and the economic metric has shifted to cents per kilowatt hour. By pointing the solar module at the sun all day instead of just at noon, the number of kilowatt hour produced by a given module typically increases by 25–40%. Today, the incremental cost penalty at the system level associated with tracking large arrays of 1 kW and more is now much less than this 25–40% benefit, making tracking now the preferred option for many applications. For ground-mounted solar PV systems generating electricity for peak power, solar tracked systems now account for over 50% of the new systems being installed. Solar tracking systems on commercial building flat rooftops are also now economically attractive. Fortunately, starting with the solar FIT in Germany, the government subsidies now encourage systems that maximize the kilowatt hour produced, thereby minimizing the cost of solar electricity in cents per kilowatt hour. Solar FITs are now available in Spain as well as in several states in the United States. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

207

208

9.2

TRACKING THE SUN

THE MODERN SOLAR TRACKER ERA

It can be argued that the modern solar tracker era began as a response to the “oil and energy crisis” of the early 1970s. In response, ARCO began manufacturing PV modules in the mid-1970s. By 1983, ARCO had built two grid-connected PV power plants on the Carrizo Plains in California. The first was a 1-MW plant and the second was a 5.2-MW plant. Neither plant was economically successful. Both plants were sold to the Carrizo Solar Corporation in 1990. Both ARCO plants used very large, computer-controlled D/A trackers that had been developed for pointing heliostats for solar thermal applications. The 5.2MW plant occupied 177 ac of land. The cost of installation was high due to the module production costs and the complex D/A trackers. The electricity produced was sold at 3–4¢ per kilowatt hour. The power would have had to be sold at five times that rate for the plant to be profitable. By the end of the 1990s, Carrizo Solar had shut the plants down and sold off the assets piece by piece. This first solar PV tracker experiment was then followed with the emergence of small solar tracker companies in the late 1980s and early 1990s. Robbins Engineering, Inc. and Zomeworks Corporation manufactured and sold passive, gas-based trackers, while Array Technologies, Inc. manufactured motorized active trackers that were controlled by a closed-loop, optical-based system. Both Zomeworks and Array Technologies exhibited slow but steady growth and primarily sold their products to an emerging remote home, “off-grid,” solar market. The Robbins Engineering trackers never gained the market place share that their counterparts did. In the United States, the “on-grid” or “grid-tie” market evolved in California during the 1980s and was driven by incentive-based programs offered by the utility districts. The SMUD was the most widely known early adopter. SMUD used an innovative S/A tracker design that proved to be much more affordable for utility applications than the early D/A tracker design used by ARCO. By 1998, over 1.9 MW of grid-tied PV had been installed. By 2008, programs characterized by the California Energy Commission and the California Public Utilities Commission had promoted 279.54 MW of installed grid-tied PV. Most of the California incentives were not “performance based.” They maximized the power in watts but not the energy produced in kilowatt hour. Maximizing the delivered AC energy fed back into the grid was not the primary goal. From a global perspective during the late 1990s, the concept of national or regional energy independence fostered FITs. A FIT is a guaranteed financial ROI based on the amount of energy in kilowatt hour delivered to the grid. Because of large FITs, utility scale-sized (over 1 MW) PV plants began to flourish in Germany and South Korea. In recent years, Spain, Italy, Greece, and Australia have enacted FIT-based renewable energy programs. FITs promote growth and a stable ROI, and encourage a reduction in installed PV system energy costs. Solar tracking PV marries well to the goal of getting the most power from utility-scale power plants. In an online June 1, 2009 Renewable Energy World article entitled “Solar Trackers: Facing the Sun,” it was suggested by an industry

TRACKER GEOMETRIES

209

analyst that “between 2009 and 2012, tracking systems will be used in at least 85% of all PV installations above 1 MW.” Certainly, there is now a serious interest in solar tracking as evidenced by the participation of over 30 solar tracking companies in San Francisco during the Intersolar North America 2009 exhibition.

9.3

TRACKER GEOMETRIES

The older traditional PV mounting structures for utility-scale PV generation are typically fixed at an optimum elevation tilt angle for power production without seasonal adjustment. A typical FR system is depicted in Figure 9.1. In most utility-scale planar PV plants today, the real question is not whether to track the sun but how to track the sun. Multiple solar tracking geometries are available to track the sun through its diurnal path. These can be first segregated into two basic categories: D/A and S/A. D/A trackers follow the sun in two axes to produce the greatest power gain from the PV module by keeping the sun’s rays perpendicular to the module surface. Many point-focus CPV technologies require D/A tracking to operate. D/A trackers, however, tend to be complex and costly, and need more area to deploy, thereby reducing their GCR. S/A trackers, depending on design, are simpler, require less area to deploy, and are generally less costly to purchase and O&M than D/A trackers. The geometric variations for the D/A trackers are shown in Figure 9.2. It turns out that tracking systems are actually more economical in most of the United States. For example, Table 9.1 presents a comparison for the economics for two systems producing similar amounts of energy in kilowatt hour per year in Madison, Wisconsin. Because the tracking system produces more annual kilowatt hour per installed kilowatt, fewer kilowatt and smaller numbers of PV modules are

W Tilt N

S

Azimuth E

Figure 9.1. Fixed-tilt rack system.

TRACKER GEOMETRIES

211 Sunny, total solar raditaion 5.14 kWh/m2 3000

Tracked Fixed

Watts Output

2500 2000 1500 1000 500 0 4:48 7:12 9:36 12:00 14:24 16:48 19:12 Time, PST (solar time, not local)

Figure 9.3. (a) Photograph of an Array Technologies D/A post-mounted PV tracking array. (b) Comparison of the power output on a sunny day for fixed and tracked PV arrays.

Figure 9.4. Rear view of a Wattsun AZ-225 D/A solar tracker. TABLE 9.2. Increase in Annual Kilowatt Hour for Various Tracker Geometries Relative to Fixed Tilt Tracking Geometry

Annual Increase in Solar Performance

Dual axis

29–40%

Single-axis azimuth tracker with array tilt of 45°

23–34%

Single-axis linear tracker with tilt of 20°

22–32%

Single-axis linear tracker with tilt of 0°

17–25%

It is now clear that for larger PV arrays, trackers produce more kilowatt hour relative to fixed-tilt PV arrays. However, how can the cost of tracking be reduced? S/A trackers are simpler because one of the drive elements is eliminated, and it turns out that they still generate a significant increase in the annual kilowatt hour produced as shown in Table 9.2.

212

TRACKING THE SUN

9.4 GROUND-MOUNTED TRACKERS FOR UTILITY PEAK POWER A simple S/A tracker moves the plane of a PV array around one central axis of rotation. The rotational axis can be vertical (points to the zenith), or the axis can be aligned with the N/S axis of the Earth. An azimuth tracker spins the array about a vertical axis and tracks the daily passage of the sun. Figure 9.5 showed several fixed-tilt azimuth trackers. Alternately, a linear axis tracker can “roll” an array of PV modules E to W. Figure 9.6 shows the concept for the simplest horizontal beam linear axis tracker. In the case shown in Figure 9.6, there are seven beams level with the ground aligned along the N to S direction. PV modules are mounted astride these beams.

Figure 9.5. Array Technologies, Inc. fixed-tilt azimuth tracker installation—2 MW, Spain.

Figure 9.6. Building block diagram of an Array Technologies, Inc. S/A HZ Tracking System. It is a utility-scale mechanically linked system.

AZIMUTH TRACKERS FOR COMMERCIAL BUILDING FLAT ROOFS

213

Figure 9.7. Array Technologies, Inc. S/A HZ trackers. Right 6.6 MW, Alamosa, Colorado.

There is a second drive beam running E to W with a single drive motor driving this beam such that the seven beams can rotate the module array from E to W over the course of the day. Larger versions of this S/A HZ tracking system are possible such that one motor can drive 300 kW of PV modules at one time. Figure 9.7 shows photographs of two large PV fields using these tracking systems.

9.5 AZIMUTH TRACKERS FOR COMMERCIAL BUILDING FLAT ROOFS As just shown, it is now clear that solar trackers are preferred for large solar PV utility fields producing peak power. However, of the potential markets for PV, commercial customers also represent an excellent opportunity. Commercial customers can buy and use larger PV systems; they presently pay retail prices for electricity; and they need electricity during the day when the sun is shining. There is an excellent opportunity to reduce the cost of solar electricity for commercial systems when trackers are mounting on commercial building flat rooftops. However, for commercial building flat rooftops, the design constraints are different. Solar trackers are needed that are sized to fit on the roof with no roof penetration and with low profiles for low wind resistance. The JXC fixed-tilt/ azimuth carousel is a low profile tracker designed to distribute weight evenly over a large area and for low wind resistance on the roof. As shown in Figure 9.8, it is designed as a compact prefabricated unit where the module support arms fold down for easy shipping. When these arms are folded down, the carousel is only 8 in. tall. Without mounted modules, it is slightly less than 8 ft wide and slightly less than 10 ft long so that multiple carousels can be stacked and easily shipped on a flatbed truck.

SOLAR TRACKER ECONOMICS

215

Figure 9.9. Carousel shipping and installation sequence. Top/left: carousel stack ships prefabricated on a flatbed truck. Top/right: hoisted to the roof. Bottom/left: deployed on roof. Bottom/right: modules mounted and operational.

Figure 9.10. JXC 1.2-kW S/A Azimuth Sun Tracker in operation in San Diego, California.

9.6

SOLAR TRACKER ECONOMICS

The economic viability of solar trackers is still often debated. However, there is now enough data to prove their economic benefit for the utility field case as shown in Table 9.3. The numbers listed in this table are specifically for the baseline c-Si module case. Tracking for commercial building rooftops is newer. Several questions are often asked. In the following, some of these questions are listed and answered specifically for the 1.2-kW carousel tracker.

216

TRACKING THE SUN

TABLE 9.3. Cost/Benefit Analysis for Solar Tracking for Utility Peak Power Fields Metric

Case 1: Historical

Case 2: Current

Case 3: Future

PV module cost

$4 per watt

$2 per watt

$1.50 per watt

PV system cost

$6.50 per watt

$4.50 per watt

$3.50 per watt

12%

16%

20%

FR cost

$0.45 per watt

$0.40 per watt

$0.25 per watt

D/A tracker cost

$1.25 per watt

$1.00 per watt

$0.75 per watt

D/A––FR

$0.80 per watt

$0.60 per watt

$0.50 per watt

(12.3%)

(13.3%)

(14.3%)

38%

38%

38%

S/A HZ tracker cost

$0.75 per watt

$0.55 per watt

$0.35 per watt

S/A––FR

$0.30 per watt

$0.15 per watt

$0.10 per watt

(4.6%)

(3.3%)

(2.9%)

24%

24%

24%

Efficiency

D/A penalty D/A gain

S/A HZ penalty S/A HZ gain

Question #1: How much energy is consumed for tracking and how does that compare with the energy produced? Answer #1: A solar tracker rotates one-half revolution during the day. It does this in very small steps. It returns to the starting point overnight. Specifically, for the rooftop carousel, the energy required per day is 6 Wh, and the energy produced per day is 6 kWh. So the energy required is 0.1% relative to the energy produced. Question #2: What is the cost per watt for a solar rooftop tracker? Answer #2: The answer to this question is a function of production volume and manufacturing setup. Production volume today for the carousels is small, and the manufacturing is done in a small prototype shop. However, a cost estimate in high-volume production can be made given its weight. The weight of the carousel without modules is about 200 lb. By analogy, a Kenmore washer weighs about 250 lb. A washing machine also has a drive motor and a controller just like a solar tracker, and it can be purchased and installed for about $400. This installed price for a 1.2-kW rooftop tracker translates to $0.33 per watt. Question #3: How much does the drive system cost? Answer #3: Again by analogy, one can purchase the drive with control for a satellite dish for about $85. For a solar tracker producing 1.2 kW, this cost will translate to about $0.09 per watt.

ABBREVIATIONS

217

Question #4: What about reliability and operation and maintenance cost? Answer #4: Nearly every home today has appliances with motors. Examples are washing machines and refrigerators. These systems are very reliable, and there is an infrastructure for O&M. Furthermore, solar trackers now have over a decade of operating experience. Note that solar trackers only turn one revolution per day. This equates to 7300 revolutions in 20 years. By analogy, with a simple wristwatch, this equates to the number of revolutions the second hand on a watch will make in 121.66 h or 5 days.

9.7

CONCLUSIONS

For larger PV plants today, the question is no longer whether or not to track the sun. The relevant question now is how to track the sun. Several tracker geometries and tracker types have been developed and described here. It is now time to scale up production in order to reduce costs. ABBREVIATIONS AC—alternating current ARCO—Atlantic Richfield Oil Company CPV—concentrating photovoltaic c-Si—crystalline silicon D/A—dual-axis DC—direct current E—east FIT—feed-in tariff FR—fixed rack GCR—ground coverage ratio HZ—horizontal N/S rotation axis JXC—JX Crystals N—north N/S—north/south O&M—operate and maintain PV—photovoltaic ROI—return on investment S—south S/A—single-axis SMUD—Sacramento Municipal Utility District W—west

10 SOLAR CELL SYSTEMS: DEFINITION, PERFORMANCE, AND RELIABILITY JASON STRAUCH, LARRY MOORE, Sandia National Laboratories

10.1

AND

ELMER COLLINS

INTRODUCTION

The previous chapters have presented the development of solar cell technology and physics and the basics of terrestrial solar cells and modules. The preceding chapter described sun trackers. Trackers are an additional hardware component used to increase the energy output from solar cell systems. This chapter will begin with a description of the complete set of components that make up a solar cell system. The basic components are the same for a smaller home-based system, a large-scale terrestrial field system, and even for space-based systems that generate AC power. After the section on system definition, a detailed description of the performance of an existing 3.51-MWDC solar cell system in a field in Springerville, Arizona, is presented. The Springerville Generating Station Solar System in eastern Arizona was built and is operated by TEP. The field experience with this system has provided a treasure of information that not only establishes a baseline for today’s state-of-the-art system capabilities but also has helped guide the development of a PV system technology for the future. These are the reasons that TEP and Sandia National Laboratories entered into a collaborative effort to track, analyze, and document the cost and field performance as well as the O&M experience associated with this system. The third section of this chapter will then present a reliability study for solar cell fields based on the data available from the Springerville system. The fourth section of this chapter will then summarize the economics for this Springerville solar cell system. This chapter ends with a summary and conclusions. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

219

220

SOLAR CELL SYSTEMS

This chapter is a condensation from papers presented by several authors from Sandia National Laboratories. The performance for the Springerville System was first described by Moore and Post [1], and the reliability study based on this system was presented by Collins et al. [2]. 10.2

SOLAR CELL SYSTEM DEFINITION

10.2.1

Basic Components

A number of components make up a PV power generating system and will be described from a system level down to the individual building blocks. At the highest level, a PV system produces electric power and, thus, is connected to an electrical load. This load can be DC batteries or equipment, or AC equipment including the electrical distribution grid. The primary focus in subsequent sections will be on grid-tied systems at larger scales, such as utility-level PV power generating fields. Many of the components and operating principles also apply to smallerscale residential systems that are grid tied, and when there is a substantial difference, it will be noted. A solar cell or PV system generates electrical power that is delivered to the load by wiring, typically transmission lines. In order to be grid connected and used by AC systems, the DC power must be converted to AC by an inverter. The inverter is a critical component to a PV system’s efficiency and reliability, as will be seen in subsequent sections. Figure 10.1 shows a schematic of a basic PV power system that is attached to the grid as depicted by the AC loads. As described previously, the most basic building block of a PV system is the P-N junction in the PV semiconductor material and cell. There are many different types of solar cells, and the focus in this section is on one of the most efficient and common, the c-Si cell. In a module, rows of cells are wired in series in what are known as strings. The strings are subsequently wired together and mounted to produce a module. These are shown in Figure 10.2. Modules are arranged into arrays, and, at larger scales, each array will typically deliver the power to an inverter. This is shown in Figure 10.3. The inverter is one of the most expensive components in the system, after the modules themselves, and is also responsible for up to 37% of system failures, as described in a subsequent section. These arrays and inverters can be grouped into larger systems that produce megawatt levels of power for the electric grid. Each array and inverter system is connected to the grid or other AC system with AC disconnects. These disconnects allow that part of the system to be isolated from the grid (or load) for servicing and repairs. There are additional components within the module that are needed to connect the modules together in an array and are used to increase the performance of the module and array. 10.2.2

Module Detail

In terms of mechanical construction, the module is physically made of a number of parts: the frame, the cell strings including cell interconnects, the junction box,

SOLAR CELL SYSTEM DEFINITION

221

Solar PV Electricity - PhotoVoltaic System

Buy from the Grid

Back-Up Battery Storage

PhotoVoltaic (PV) Flat Panel Collector

AC/DC Inverter

Import Meter Fusebox

Network / Grid

Export Meter

Lighting & Electricity - Sockets & Switches

Sell Back to the Grid

Block Diagram of PV System

Solar Irradiance

cell

Module

module array DC solar irradiance

AC AC loads

Battery

inverter

AC

+

Charge Controller

Inverter



engine-generator back-up (optional) battery storage (optional)

AC Loads DC Loads

Figure 10.1. Basic PV power system examples.

Figure 10.2. A single polycrystalline PV cell showing the diode grid lines, and a module cartoon showing the cell strings.

SOLAR CELL SYSTEM DEFINITION

223

seemingly simple as the top glass is highly optimized and serves multiple functions. First, it is a cover to protect the cells from environmental structural damage. The glass is also the primary structural support component of the cell string area, just like the frame is the structural support of the entire module. The glass also needs to be highly transparent in the convertible wavelength region of the cells, so that no reduction of available solar energy occurs in the glass. Furthermore, modern glass covers have special structure patterns on both sides of the glass. On the bottom surface of the glass, the texture allows for good lamination adhesion of the cell strings. The texture on the top surface is designed to help the glass minimize soiling, which obviously can reduce the convertible light transmission. The next layer in the stack is the EVA copolymer, which serves the function of being a bond layer between the glass and cells’ ARC. The ARC helps to trap incident photons within the cells so that they may be converted to electron–hole pairs and produce electricity. Next comes a second bond layer of EVA, and finally the back-side protective coating is made of Tedlar, a polyvinyl fluoride film with high emissivity for good radiative heat rejection and low UV degradation. Also within the lamination are the electrical interconnects between the cells, generally a tin-coated copper ribbon. As stated, PV cells are wired in series to form strings (of cells), which are then wired in series within an individual module. The next component that is needed to allow the modules to be wired together into an array is the junction box. The junction box is where all of the series-connected cells that are wired together using the tin-coated copper ribbon merge into a single positive and negative electrode. The junction box also represents some expense and is responsible for about 12% of the system failures, as will also be seen in a subsequent section. Since cells and modules are series connected, there is a potential for a low-performing part of the system to drag down the entire system performance, as each series-connected component is (electrically) current limited by the weakest performing component in the series. What this means in a practical sense is that the weakest link in the chain, the lowest performing cell, string, or module, can limit the electrical current and thus the system power of the entire system. A shaded or low-performing part or region can actually go into negative bias (negative voltage), which takes power away from the module in the form of heat generation in that cell or region and is much worse than if the cell were simply disconnected and bypassed. This is exactly what a bypass diode does. A bypass diode functions by allowing current to bypass the low-performing cell so that it does not reduce the system output as much. The system power is still decreased by the relative lack of contribution of the cell or region itself, but this is less than a reverse bias cell that is actually taking power out of the system when heating up. It is probably clear to the reader at this point that every aspect of a PV system has been optimized to increase performance and reduce cost. In the next section, the performance of a state-of-the-art utility-scale PV system is described. This system is operated by the TEP Company with the system performance analysis provided by a collaboration with Sandia National Laboratories.

224

SOLAR CELL SYSTEMS

10.3

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

10.3.1

Introduction

Over 12 MWDC of grid-connected PV systems are currently installed in Arizona, primarily by the state’s two largest investor-owned utilities: APS and TEP Company. The vast majority of the state’s installed generating capacity of utilityscale PV systems (100 kW and larger) utilizes flat-plate, c-Si collector technology. The APS experience has focused on one-axis, north–south-oriented, horizontal tracking arrays. The TEP systems incorporate standardized, fixed arrays. The 5 years of TEP operating experience with these systems, including performance, cost, maintenance, installation, and design, is the topic of this section.

10.3.2

TEP Company

TEP is the second largest investor-owned utility in Arizona, providing electricity to 392,000 residential, commercial, and industrial customers in Tucson and surrounding areas in southeastern Arizona. The company operates nearly 15,000 mi (24,135 km) of transmission and distribution lines throughout its service territory of 1155 mi2 (2991 km2). The utility is involved in a very active renewable energy program. The utility-scale PV generation effort is centered at the Springerville Generating Station Solar System in eastern Arizona. As shown in Figure 10.5, this facility, one of the largest PV generating plants in the world when installed, includes 4.6 MWDC of installed PV systems. Covering 44 AC (17.8 ha), this PV generating plant is grid intertied with a 34.5-kV TEP distribution line. Although the Springerville plant includes other collector technologies including amorphous

Figure 10.5. Springerville PV generating plant.

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

225

Si and cadmium telluride, c-Si accounts for nearly 80% of the plant’s capacity and is the focus of this section. The field experiences with these systems provide a treasure of information that not only establishes a baseline for today’s state-of-theart system capabilities but can also help guide the development of PV system technology for the future. These are the reasons that TEP and Sandia National Laboratories entered into a collaborative effort to track, analyze, and document the cost and field performance as well as the O&M experience associated with these systems.

10.3.3

Design and Installation Experience

The 26 c-Si systems at the Springerville plant are listed in Table 10.1. Each of these systems is an identical copy of a standardized array-field configuration that utilizes the same hardware components, wiring topology, and structural mounting plan.

10.3.4

Standard System Configuration

The standard system configuration includes ASE Americas (now Schott Solar) ASE-300-DG/50 modules and a Xantrex PV150 inverter. The arrays are mounted

TABLE 10.1. List of Springerville c-Si Systems with Each System Rated at 135 kWDC System

Installation Date

System

Installation Date

SGS-135C-1

July 13, 2001

SGS-135C-14

July 15, 2003

SGS-135C-2

July 13, 2001

SGS-135C-15

July 15, 2003

SGS-135C-3

August 17, 2001

SGS-135C-16

July 30, 2003

SGS-135C-4

October 2, 2001

SGS-135C-29

October 15, 2003

SGS-135C-5

October 23, 2001

SGS-135C-30

October 30, 2003

SGS-135C-6

December 14, 2001

SGS-135C-31

August 15, 2003

SGS-135C-12

May 30, 2002

SGS-135C-32

August 30, 2003

SGS-135C-7

August 1, 2002

SGS-135C-26

June 22, 2004

SGS-135C-8

August 1, 2002

SGS-135C-27

June 22, 2004

SGS-135C-9

August 1, 2002

SGS-135C-28

June 24, 2004

SGS-135C-10

September 17, 2002

SGS-135C-23

July 20, 2004

SGS-135C-11

June 24, 2002

SGS-135C-25

July 21, 2004

SGS-135C-13

June 15, 2003

SGS-13n-24

July 23, 2004

226

SOLAR CELL SYSTEMS

at a fixed latitude tilt of 34° facing due south with 450 modules per array. Based on each system’s areal footprint of 300 ft (91 m) north–south and 140 ft (43 m) east–west, the system power density is 110.6 kWAC/AC (273.3 kWAC/ha) of ground, which allows for generous access space for construction and maintenance activities. Each array string includes nine modules with two strings per row. The power per string is 2.7 kW, and the maximum string design voltage is 595 V at −22°F (−30°C). The operating voltage of each string is 380–430 V. The Xantrex PV150 inverter converts the variable voltage DC power to a 208-V three-phase AC power. The inverters have a maximum rating of 157 kVA, at which point they will limit output or come off-line, followed by an automatic restart. The maximum inverter rating was selected by TEP to accept increased output due to cloud-edge enhancement and cold temperatures. These enhancements have at times exceeded 160 kW. In addition, the higher rating allows for normal operation to be in the optimum area of the efficiency curve and allows the inverters to run cooler extending lifetime. Each unit has a DC disconnect, 150-kVA, 208-V to 480-V step-up/ isolation high-efficiency transformer, revenue meter, and AC disconnect. Groups of four units are connected in parallel to each of eleven 500-kVA, 480-V to 34,500-V high-efficiency step-up transformers. Each transformer has a continuous rating of 500 kVA and can accommodate up to 650 kVA for brief intervals. The high-voltage sides of the transformers are connected in parallel to a 34.5-kV underground distribution line, which connects to the overhead 34.5-kV distribution line that feeds the well field pumps of the nearby 1160-MW coal-fired Springerville Generating Station. The pumps operate continuously with an average total load of about 9000 kW.

10.3.5

Instrumentation/Testing

The Springerville PV generating plant is a normally unmanned site that is continuously monitored remotely via an Internet-based communications channel. Most operational functions such as inverter reconfiguration, fault resets, IV curve tracing, diagnostic testing, and performance analyses can be performed from the remote monitor site via the Internet communications channel. Fifty points of information are taken from each of the 26 inverters and the revenue meters on 10-s scan cycles, and averaged for both 10-s and 1-min archiving. Performance information is developed from the raw data for daily review and is archived in spreadsheet format by a control operator. Near real-time performance is available on the Internet. Alarm criteria have been developed for all operational parameters, and these alarms are logged and maintenance personnel are notified in case of operation of any array or inverter that is out of specifications. Spare parts are available on-site, and localbased service personnel are dispatched to the site to perform repairs in response to alarms. Both test equipment and trained personnel are used to diagnose and repair problems in the system in addition to continued support from the inverter and module vendors. Test equipment consists of mostly traditional utility items. For these systems, a clamp-on ammeter for DC and AC with a maximum range of

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

227

0–40 amp and a separate clamp-on ammeter for DC and AC with a range of 0–1000 amp are used. At least one, preferably two, voltmeter with a DC and AC range of 1000 V and an integral ohmmeter are essential. An optical temperature sensor, frequency counter, and a dual trace oscilloscope with a range to 100 kHz are needed. Harmonic meter measurements and a three-phase power factor meter are helpful as well. Specifications for the inverters included ±1% measurements of AC and DC current and voltage, ac frequency, and IGBT temperature, as well as calculated values for AC and DC power. The PV150 can acquire all AC, DC, temperature, state of operation, and power parameters at a maximum rate of one sample set per second. The first six inverters were installed, and data were taken to confirm the ±1% measurement accuracy. During the first year, these inverters were routinely checked for performance against calibrated test equipment. After the 1-year period, the PV150 power production data are validated using the revenue-grade utility meter on the output of the 480-V transformer. These kilowatt hour meters have a tested accuracy of ±0.25%. For each of the identical 26 individual systems, there are 50 strings of nine modules for a system STCs rated array nameplate power of 135 kWDC. Each string represents 2% of the rated individual system power. Individual string power degradation can be identified when a single module fails (i.e., no current in the string) or the module voltage falls to 90% or less of a good module value (i.e., low voltage cannot provide more than 50% expected current). A reduction in total array power output of 1% as compared with the output expected given solar insolation, temperature, wind speed, and wind direction can be accurately identified through the continuous performance evaluation monitors implemented in the software used for system supervision. Air temperature, wind speed, wind direction, and plane-ofarray solar insolation measurements are monitored and recorded on 10-s intervals from three locations within the plant footprint. The three temperature sensors are all NRG, model 110 s with radiation shield, with ±0.6°F (±0.33°C) accuracy. Wind speed is measured at the three locations with an NRG model #40 anemometer with ±0.2 mph (±0.32 kph) accuracy, and wind direction is measured at the three locations by NRG model #200p units, with an accuracy of ±1%. There are three plane-of-array solar radiation sensors used so that cross-comparison is possible. These instruments are MSX-01 reference cells each using a single square polycrystalline cell with a 1-Ω resistor in parallel with the cell output, resulting in a solar insolation measurement that exhibits very little temperature sensitivity.

10.3.6

Balance of System

The array configuration is designed to minimize the array-field BOS cost. The dual-stanchion array structural supports are fabricated steel with powder coating to minimize corrosion. The steel supports are staked to the ground to prevent windinduced uplift and sliding. The site preparation includes minimal surface disturbance/leveling of the natural terrain while retaining the native vegetation as much as possible to reduce surface erosion and to minimize dirt splash on the modules

228

SOLAR CELL SYSTEMS

Figure 10.6. (a) Typical 135-kWDC system. (b) Xantrex PV150 system inverter.

during rainstorms. Mounting of the arrays to the terrain may result in a slightly jagged array appearance along the row due to surface variations but the adverse PV output effects are near zero. The array electrical interconnection uses 600-V rated DC equipment and underground AC power distribution to minimize cost. Each system is installed exactly the same using a trained local labor pool. This standardized approach has resulted in a total system BOS cost of less than $1400 per watt direct current. The energy payback time for the Springerville BOS has been documented at 0.21 years, a significant improvement over previous central plant designs. A Springerville system is shown in Figure 10.6a. Note the white inverter enclosure at the back of the arrays near the center of the picture. A closeup photo of the Xantrex PV150 inverter and enclosure is shown in Figure 10.6b. 10.3.7

System Performance

To describe the system performance of the Springerville systems, the PV energy parameters that have been established by the IEA Photovoltaic Power Systems Program as described in the IEC standard 61724 are utilized. Three of the IEC standard 61724 system performance parameters—final yield, reference yield, and performance ratio—define the system field performance in terms of energy production, solar resource, and system losses. These provide an easily understood method not only to compare system performance with other system options but also to permit system owners/customers to determine if system performance is meeting expectations. This process has been proposed for widespread adoption here in the United States, and the author/editors certainly support this effort. Since all 26 Springerville systems were totally operational beginning in 2004, average performance results are presented based on data from 2004 to 2006. In addition, specific annual results are noted for each of these 3 years as well.

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

10.3.8

231

System Maintenance Experience

For the past several years, Sandia has been working to develop a comprehensive database model to track the life cycle costs of PV systems. This database, which continues to undergo improvements, was utilized to capture, document, and track schedule and unschedule maintenance service, repairs, replacements, and labor and travel costs associated with maintenance activities for these systems. Based on Microsoft Access, the database architecture is modular to support future additions, allows associations at the component level, allows multiple components to be tracked with a system, and provides for multiple failures to be documented as a result of a maintenance visit. Failure modes (what and why), activity dates (failure and repair), and costs (labor, parts, and travel) were captured and analyzed from system maintenance activity logs covering the period of mid2001 through 2006. From these data, analyses of failure modes and O&M costs were made. The Springerville systems provide a significant database for assessing the reliability and maintenance needs for a utility-scale generating plant operating in a utility environment. Altogether, a total of 11,700 identical PV modules and 26 identical inverters have been installed since mid-2001. Over the operating history from mid-2001 through 2006, a total of 156 unscheduled maintenance events were recorded for the Springerville systems. The events are grouped by categories including DASs, inverters, junction boxes, arrays (PV), systems, and AC disconnects. Figure 10.9a presents the breakdown of these events by component as a percentage of the total number of events. The unscheduled events resulted in a loss of generating capacity that affected one or more systems and required human intervention to restore the system(s) to full operational capacity. These events could be as simple as a manual restart of a tripped inverter or considerably more complex such as the repair of damage resulting from a lightning strike (the plant experienced strikes in years 2003–2005). An examination of maintenance events by category provides some insight into just how exemplary the maintenance experience at Springerville has been. Over half of the 32 AC disconnect events were associated with high contact resistance in the 480-V outdoor rated fused disconnects. In 2006, the contacts for all Siemens brand 480-V disconnects were changed, and the factory-installed contact grease was removed. The high-resistance problem appears to be due to dirt accumulation in the grease after 5 years of operation. Ten of the 11 failures associated with the DAS were all due to a severe 100-year lightning storm at the site in July 2003. That same storm caused 14 of the 58 problems with the inverters, 12 of which required replacing the PCU card. In addition, that same storm also accounted for 8 of the 13 system events that involved replacing damaged utility meters. Many of the other inverter events were associated with manually resetting trips, a problem since avoided in 2004 when autoreset capabilities were added to the inverters. As for junction boxes, 11 of the 18 events were associated with replacing failed blocking diodes. Eleven of the 24 PV array events were associated with damage from another lightning storm in July 2004. In general, the reliability of the systems has

232

SOLAR CELL SYSTEMS

Figure 10.9. (a) Unscheduled maintenance events by component. (b) Unscheduled maintenance events by category. (c) Inverter repairs by events and category.

been excellent, and the reliability of the primary PV components, modules, and inverters, has been impressive. Figure 10.9b presents a breakdown of unscheduled events by component as a percentage of the total unscheduled repair costs. As noted, the majority of the repair costs are associated with the inverters. A more detailed examination of the unscheduled inverter events provides a real-world perspective of the maintenance expectations in a utility environment. A breakdown of inverter events by repair category is presented in Figure 10.9c. The categories refer to the inverter operation. Controller includes those functions and circuit components necessary to control the power conversion and protec-

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

233

tive devices. Interlock refers to the DC disconnect and door interlock alarm circuitry. Design includes cabinet weather protection. Internal refers to fault conditions within the inverter not otherwise identified. Matrix is the power electronic inverter bridge and associated switching transistors, capacitor bank, heat sink, and cooling fans. Other captures those events such as wiring and contactor problems, inoperative switches, and unknown events that signal a fault condition. As noted earlier, 12 of the 20 controller events involved the replacement of PCU cards and the addition of enhanced lightning protection due to the July 2003 lightning strike. This enhanced protection involved the addition of lightning arrestors and associated surge-resistant components in many areas of the data collection system and on the 480 voltmeters of every inverter. The other eight events primarily involved PCU card replacements due to failures and/or intermittent problems. The interlock events were all associated with DC disconnect faults ranging from connector problems and nesting rodents to unknown causes. The five design events included faults due to one roasted spider, one roasted rodent, two cases of rain, and one case of blowing snow ingress into the cabinet. Improved gasket placement has solved those problems. In each case, the internal events involved an inoperative inverter with no obvious problem that responded to a restart. The six matrix events involved two cases of matrix failure and replacement, two cases of temperature sensor failure and replacement, one case of fan assembly and motor failure and replacement, and one unknown cause of high temperature alarm on the heat sink. The 10 other events involved loose connections, replacement of switches, a failed front panel, wiring problems, and unknown causes. Although not trouble free, the 26 inverters have provided an enviable maintenance record especially in the context of other documented inverter field performance problems. Through January 1, 2007, the 26 c-Si Springerville systems had provided 1206 system-months of continuous operation since installation. Over that same period, a total of 156 unscheduled maintenance events were recorded, which provides a mean time between unscheduled services per system of 7.7 months of operation. Scheduled maintenance was conducted on the plant each year. This included mowing the native vegetation as well as visual inspections of the arrays and power handling equipment. Table 10.2 lists the annual maintenance cost, both scheduled TABLE 10.2. Maintenance Cost as a Percentage of Capital Investment Year

Scheduled (%)

Unscheduled (%)

Total (%)

2002

0.08

0.01

0.09

2003

0.07

0.22

0.29

2004

0.06

0.04

0.10

2005

0.06

0.01

0.07

2006

0.04

0.03

0.07

SOLAR CELL SYSTEM RELIABILITY

235

predictive PV engineering. This system reliability analysis stems from Sandia’s role in engineering high-surety, high-consequence engineering systems and will be described in the next section.

10.4 SOLAR CELL SYSTEM RELIABILITY—USING THE PAST TO PREDICT THE FUTURE This section is a condensation of an IEEE PVSC34 conference paper submitted by Sandia National Laboratories by Collins et al. [2] on PV system reliability. It describes the application of a system reliability model developed for nuclear weapons that has now been applied to the TEP Company’s PV field. It is also more generally applicable to any PV system.

10.4.1

Introduction to Solar Cell System Reliability

System level reliability and availability estimates are required to facilitate cost trade-off studies associated with competing PV systems. Estimates of reliability are necessary in developing maintenance cost projections over the system’s lifetime. Availability estimates provide an input into annual energy generation projections. The reliability and availability of large PV systems have not been thoroughly investigated. Manufacturers in the PV industry are offering warranties of 20 years and better for PV modules with incomplete knowledge of their reliability in the diverse environments in which these modules are deployed. All stakeholders, including entrepreneurs, manufacturers, regulators, utility operators, and consumers, need a useful predictive model for reliability and availability. This predictive model can facilitate trade-offs in requirements and life cycle cost; identify improvements in design, manufacturing, and in situ operation; and provide estimates of performance over the system’s lifetime. This section describes a standard methodology used to characterize a system’s reliability and availability. The elements of a reliability/availability improvement program are illustrated in Figure 10.11. The three basic activities associated with a reliability/availability program are FMEA, system reliability/availability modeling, and accelerated life tests. System reliability/availability modeling allows quantification of system reliability and availability using multiple data inputs, such as field data, test data, and accelerated life test data. A system reliability model is a diagrammatic representation of all functions, in terms of subsystem or component events, that must occur for a successful system operation. This section describes a comprehensive approach to developing reliability and availability estimates for a large PV system. System reliability and availability were defined based on the operator’s expectations, and an RBD was developed to model system behavior. The RBD developed is a hierarchical reliability model.

SOLAR CELL SYSTEM RELIABILITY

237

An RBD is constructed by identifying all necessary functions and their associated components that must occur for the system to provide an output. The blocks are then arranged in a diagram in order of operation. Hierarchical system block diagrams are high-level diagrams that contain one or more subsystems or components within a block. Each block may have an associated RBD defined by lowerlevel functions. Accelerated life tests are tests that are run at elevated stress levels outside the stresses expected in normal operation. The objectives of these tests are 1.

identification of the life distribution parameters of time to failure for the applied stress, 2. identification of relationships (mathematical or physical/chemical) between time to failure and stress, and 3. evidence about whether failure mechanisms stimulated by the accelerated stresses are expected to occur at operational stress levels.

Accelerated life tests allow collection of time-to-failure information in situations where the time to test at normal stress levels is inordinately long, or the sample sizes required are too large. In this program, accelerated life tests will be used to estimate time to failure for failure mechanisms that require long periods of time to manifest. 10.4.2

Case Study: Reliability Model for TEP Springerville PV System

This reliability study focused on the c-Si PV portion of TEP Springerville, Arizona, grid-connected PV system. c-Si PV modules comprise approximately 80% of the PV generating system’s capacity. (Editor’s comment: very little, if any, field performance data and reliability data are available in the literature as of this book edition for thin film PV systems.) An RBD was developed for the c-Si portion of the PV system. To develop the RBD, a definition of system success was needed. A decision was made to have a simple definition for system success. Success was defined as the PV system delivering power to the grid. A hierarchical system model was developed describing the functional elements necessary to deliver power to the grid using a commercial software tool, ReliaSoft (Tucson, AZ) BlockSim 7™. The granularity of the model, the basic blocks in the model, was based on the major components of the system and the level of identification of field failures in the reporting process. The reader is referred to the original Collins et al. [2] paper for more details. To yield useful reliability and availability metrics, the RBD must be populated with both life distributions and repair distributions within each block. 10.4.3 Life Data to Populate the Reliability/Availability Model and Results of Analysis TEP shared detailed production and outage reports for the years 2003–2007. These reports were analyzed to extract data about failures, failure causes, times to failure,

SOLAR CELL SYSTEM RELIABILITY

239

Two basic approaches to data analysis are used for the components of the PV system. For the components that are replaced when they fail, life data analysis is employed. Data are fitted to common life distributions such as lognormal, exponential, two-parameter exponential, two-parameter Weibull, and three-parameter Weibull. Once data are fitted to a distribution, the parameters of that distribution are estimated using one of several available techniques. The “best” match life distribution is then selected based on weighing three criteria: 1.

goodness of fit (the p value from the Kolmogorov–Smirnov test), the maximum absolute difference between the hypothesized and empirical cdfs; 2. a likelihood ratio, the value of the log likelihood of the hypothesized distribution with the estimated parameters evaluated with the given data set; and 3. plot fit, the mean absolute difference between the hypothesized and empirical cdfs. Table 10.3 provides a summary of the distributions and estimated parameters used to model the components that are replaced when they fail. Scale and location parameters are in days or the applicable transformation. For components that are repaired rather than replaced, parametric recurrent data analysis is used. This approach is based on the GRP model. It is used when a component accumulates more than one failure over its service life. This model

TABLE 10.3. Summary of Life Distributions and Parameter Estimates for Replaced Components PV Component/ RBD Block

Distribution

Beta or Log SD (Shape)

Eta or Log Mean or Lambda (Scale)

AC disconnect

Weibull 3-RRX

0.35

Lightning

Exponential 1-RRX

Row box

Weibull 2-RRX

0.51

1.2E+06

PV module

Weibull 3-RRX

0.28

5.2E+12

480/34.5-kV transformer

Weibull 2-RRX

0.58

7100

208/480 transformer

Weibull 3-RRX

0.15

1.3E+10

Marshalling box

Lognormal 2-RRX

2.3

10

11,000

Gamma (Location) 3.9

0.00022

17

28

240

SOLAR CELL SYSTEMS

is particularly useful in modeling the failure behavior of a component and understanding the effects of repair on the age of the component. In order to obtain a virtual age, the exact occurrence time of failures should be available. However, the times are unknown until the corresponding event occurs. Therefore, the software uses Monte Carlo simulation to predict values for virtual time, failure number, MTBFs, and failure rate. Three coefficients are estimated by the software: two power law coefficients and a repair effectiveness coefficient. Parametric recurrent data analysis was used for the inverter only. Many inverters had multiple failures over the analysis time period 2003–2007. Life data analysis was used for all other components of the PV system. Analysis of the field data for inverters yielded a power law model with estimates of beta equal to 0.75 and lambda of 0.019 per inverter. The repair effectiveness q was estimated to be zero. A zero value means repair did not degrade the inverter’s reliability.

10.4.5

Maintenance Distributions and Parameter Estimates for Blocks

Repair times, time from failure to restoration, and other downtimes were identified for each component from the production and outage reports. For inverters, these downtime events were sorted into three categories: corrective maintenance, preventive maintenance (scheduled and unscheduled), and grid effects. Downtimes caused by grid effects are a special category. These downtimes are associated with inverters tripping and resetting due to voltage or frequency excursions in the external power grid connection that result in no physical damage to the system. A similar approach to life data analysis was taken for fitting repair distributions and estimating parameters. Data are fitted to common life distributions, such as lognormal, exponential, and Weibull. Once data are fitted to a distribution, the parameters of that distribution are estimated. Scale parameters are in days or the transformation. The results are provided in Tables 10.4 and 10.5.

TABLE 10.4. Summary of Repair Distributions and Parameter Estimates for the Inverter PV Component/RBD Block

Distribution

Corrective maintenance

Lognormal-RRX

Preventive maintenance

Exponential 1-RRX

Grid effects

Weibull 2-RRX

Beta or Log SD (Shape)

Eta or Log Mean or Lambda (Scale)

2.27

−4.25 2.62

1.07

0.16

SOLAR CELL SYSTEM RELIABILITY

241

TABLE 10.5. Summary of Repair Distributions and Parameter Estimates for Other Components PV Component/ RBD Block AC disconnect

Distribution

Beta or Log SD (Shape)

Eta or Log Mean or Lambda (Scale)

Weibull 2-RRX

0.71

1.4

Lightning

Weibull 2-RRX

0.73

10.8

Row box

Lognormal 2-RRX

2.07

−0.98

PV module

Lognormal 2-RRX

3.11

−1.37

480/34.5-kV transformer

Weibull 2-RRX

0.53

1.36

208/480 transformer

Lognormal 2-RRX

1.6

−2.33

Marshalling box

Weibull 2-RRX

0.35

3.55

10.4.6

Plots of Reliability and Availability versus Time

The life and repair distributions for all the components were imported into the BlockSim RBD for the PV System. A Monte Carlo simulation that sampled from the life distributions at discrete intervals and determined component and system state (operational, or down due to failure or maintenance) was executed. For availability, repair times are sampled for the various repair distributions associated with each block to determine component and system states. Results of these simulations at the system level were a straight-line plot indicating an availability of 100%. At this level, the model predicts at least one inverter will always supply some power to the grid. Figure 10.13 illustrates plots of reliability and availability for the inverter with PV array segment of the system. The reliability in Figure 10.13 reflects the effect of no repair or replacement of components. TEP’s rapid repair policy enables very high levels of availability with lower component reliability.

10.4.7

Expected Number of Failures in 20 Years

The expected number of failures as predicted by the model for each component for 5, 10, and 20 years are shown in Table 10.6. The number of failures predicted for the inverter assumes no additional reliability growth, a conservative assumption. The model does not currently incorporate the effects of degradation or wear out failure mechanisms. These data are not currently available for the TEP PV system. For the first 5 years, the inverter repair rate was 0.96 per inverter per year. For the PV modules, the replacement rate was approximately 5 in 10,000 PV modules per year.

242

SOLAR CELL SYSTEMS ReliaSoft BlockSim 7 - www.ReliaSoft.com

Availability and Reliability vs Time for 5 Years 1.000

Inverter With Array

Point Availability Line Point Reliability Line

A(t), R(t)

0.800

0.600

0.400

0.200

0.000 0.000

Mike Mundt Sandia Labs 5/2/2009 7:01:00 PM

365.000

730.000

1095.000

1460.000

1825.000

Time in Days, (t)

Figure 10.13. Plot of reliability and availability versus time for inverter with PV array.

TABLE 10.6. Predicted Number of Component Failures Component

Actual Number of Failures, 5 Years Cumulative

Expected Number of Failures, 5 Years Cumulative

Expected Number of Failures, 10 Years Cumulative

Expected Number of Failures, 20 Years Cumulative

125

132

231

429

PV module

29

26

31

38

AC disconnect

22

17

23

31

Lightning

16

10

20

41

PV 150 inverter (26 c-Si arrays)

208/480 transformer

4

3

3

3

34

25

35

50

Marshalling box

2

4

7

11

480/34.5-kV transformer

5

4

5

9

Row box

SYSTEM COST EXPERIENCE

243

The mean availability predicted by the model approaches 100% for the 5-year period. The effective availability reported by TEP for the system was 99.91% in 2007. Effective availability is defined as the actual power produced divided by the total power that could have been produced. Inverters are the most unreliable component in this system. Yet the availability of continuous power delivered to the grid is projected to be very high over the life of this system. However, an increase in inverter reliability can still lower corrective maintenance costs over the system’s life.

10.5

SYSTEM COST EXPERIENCE

10.5.1

Cost for TEP Springerville Solar Cell System

TEP is realizing significant cost benefits by incorporating standardized products, volume purchasing, and efficient array-field design and installation. The Springerville experience has documented some of the lowest installed system costs ever reported thereby establishing a benchmark for state-of-the-art utility-scale systems. A cost breakdown for systems installed in 2004 (the last year for system installations at Springerville) is presented in Table 10.7: 1.

Modules: the module price reflects a bulk purchase from the module manufacturer. 2. Array-field BOS: the site preparation cost includes ground leveling, fencing, and underground wiring. Structure cost includes mechanical mounting of the modules, support structure hardware, and staking. The electrical work includes module interconnect wiring, conduit, junction

TABLE 10.7. Cost Breakdown of Springerville Systems System Component

$/WDC

$/WAC

Modules

3.33

4.22

Array-field BOS

0.56

0.71

Inverter/transformers

0.40

0.51

Indirect/overhead

1.11

1.40

Total

5.40

6.84

Site preparation ($0.10/WDC) Structure ($0.15/WDC) Electrical ($0.30/WDC) AC Intertie $0.01/WDC)

244

SOLAR CELL SYSTEMS

boxes for both the string and row buses, disconnect switches, system protection and wiring on the AC side of the inverter to the 480-V transformer, and the DAS. The AC intertie cost includes the wiring and installation labor from the 480-V to the 34.5-kV transformer. 3. Inverter/transformers: this cost includes the purchase price of the Xantrex PV150 Inverter, the 150-kVA 208/480-V transformer for each system, and one-fourth of the 480/34.5-kV transformer cost (each 34.5-kV transformer gathers four of the systems). Installation labor for these components is included. 4. Indirect/overhead: indirect costs include system design, procurement, construction management, and project engineering. The overall project management for the Springerville installations was provided via contract by Tucson-based Global Solar Energy.

10.5.2

Energy Cost Figure of Merit

The true measure for comparing different PV system options is the cost of delivered kilowatt hour alternating current energy. To put the Springerville cost experience in perspective, Moore and Post [1] utilized an energy cost figure of merit defined as the average installed system cost (U.S. dollar per kilowatt direct current) divided by the energy output (kilowatt hour alternating current per kilowatt direct current) expected over a 30-year period. Although the resulting cost figure represents U.S. dollar per kilowatt hour alternating current, this figure does not include financing costs, the cost of capital, O&M costs, or any tax considerations and, thus, is not an LEC and is not portrayed as such (note that LEC for TEP is addressed in the next section). Using the 2004 system costs, this energy cost figure of merit is $0.10 per kilowatt hour alternating current for the Springerville systems. Interestingly, the Springerville energy cost figure of merit for fixed flat-plate systems is nearly identical to the energy cost figure of merit reported for one-axis, tracking horizontal flat-plate systems installed at Prescott, Arizona. It is also interesting to note the significant energy cost figure-of-merit difference between the utility-scale Springerville systems and residential systems. A total of 82 SunShare residential systems have been installed in the TEP service territory during the past few years and are being tracked by TEP and Sandia for performance, cost, and maintenance experience. These residential-size systems range from 1.2 to 5.9 kWDC. Using the average installed cost for these systems in 2004 of $7.32 per watt direct current and an average annual final yield of 1398 kWhAC/kWDC provides an energy cost of $0.175 per kilowatt hour alternating current for the residential systems, significantly higher than the utility-scale option. The energy cost of system O&M can also be described by a figure of merit defined as the annual cost of O&M divided by the annual energy output. As noted above, this is not an LEC, but it does provide a perspective on the cost impact of maintenance experience with the Springerville systems. Using the average annual

SYSTEM COST EXPERIENCE

245

maintenance cost of 0.12% of installed capital cost, the annual O&M energy cost is $0.004 per kilowatt hour alternating current. Including the expected inverter, rebuild costs increase the annual O&M energy cost to $0.007 per kilowatt hour alternating current. Considering the SunShare residential systems noted above and using the average annual maintenance cost of 1.7% of installed system cost provides a comparative annual O&M energy cost of $0.089 per kilowatt hour alternating current, an order of magnitude higher than the utility-scale systems.

10.5.3

Economic Perspective

The experience at Springerville provides a valuable utility perspective on the future use and needs of PV technology. These include actual utility-based energy generating costs, capacity factors, and operational aspects associated with solar electric generation. 10.5.3.1 Energy Cost The SEIA has provided a road map with established goals for expanding the use of solar power-generating capacity here in the United States. It is of interest to note that a photo of the Springerville systems is featured in this road map document. The road map goal over the next decade for PV systems is a selling price of $3.68 per watt alternating current in 2015 and a cumulative installed U.S. capacity of 9.6 GW. Coupling the TEP cost experience at Springerville with this SEIA cost goal provides an interesting perspective for the future of PV. Table 10.8 presents a comparison of today’s benchmark system costs for Springerville and a proposed breakdown of a 2015 utility-scale PV system meeting the road map goal in today’s dollars. Using the Springerville performance ratio of 0.79, the $3.68 per watt alternating current future system cost corresponds to an equivalent cost of $2.91 per watt direct current. The 2015 system cost components follow a proposed breakdown developed elsewhere for a c-Si system. The 2015 module cost is based on a manufacturing cost analysis for a c-Si production plant of 25 MW/ year developed by Spire Corporation. The proposed module cost is also consistent TABLE 10.8. System Costs for the Future System Component

Springerville System ($/WDC)

2015 System ($/WDC)

Modules

3.33

1.78

Array field

0.56

0.58

Inverter

0.40

0.25

Fixed

1.11

0.30

Total

5.40

6.84

246

SOLAR CELL SYSTEMS

with c-Si manufacturing cost projections developed through the U.S. DOE PVMaT program. While module costs and fixed costs require substantial cost reductions to achieve the 2015 goal, this comparison validates the creative system BOS approach developed by TEP at Springerville by already achieving the array-field BOS target projected for the next decade. As annual PV installation quantities increase in future years, it is expected that the fixed costs will be diluted over larger amounts of installed capacity and will be reduced on a U.S. dollar per watt direct current basis. The industry road map goal for 2015 is an LEC of $0.057 per kilowatt hour alternating current of PV generation. This compares to the TEP-calculated LEC in 2006 (pay as you go, no financing costs) of $0.062 per kilowatt hour alternating current for the Springerville PV generation. The TEP calculation, which includes both federal income tax credits and state property tax reductions for solar, already meets the road map baseline 2015 LEC of $0.115 per kilowatt hour alternating current. It is important to note again that the TEP strategic plan to incorporate solar generation in its service territory is focused on a pay-as-you-go funding to avoid the high costs of financing. The attractive TEP-calculated LEC for this facility is a direct result of this approach.

10.6

CONCLUSIONS

The energy data, maintenance experience, and costs with the Springerville c-Si systems provide a treasury of information that establishes a benchmark for current utility-scale fixed flat-plate PV systems technology. This operating assessment has identified a number of findings, including

• • • • • • • •

average annual AC system energy output is 1707 kWhAC/kWDC of array; average annual AC system power is 0.79 of the array DC nameplate rating; average annual O&M cost is 0.12% of initial system installed capital cost, not including rebuild/replacement cost of the inverter; the mean time between unscheduled maintenance services per system is 7.7 months of operation; innovative approaches including standardized array designs, low-cost array-field BOS, and bulk hardware purchases have resulted in an installed system cost of $5.40 per watt direct current; the average annual capacity factor for all systems was 19.5%; the LEC cost calculated by TEP (no financing costs) is $0.062 per kilowatt hour alternating current, which meets the 2015 SEIA baseline goal for PV generation; control instabilities associated with cloud passage over the PV plant require further efforts including possibly smart inverters and/or storage to smooth these generation transients and improve capacity credit for PV generation.

ABBREVIATIONS

247

ACKNOWLEDGMENTS The key contributions of Hal Post, Michael Dvorack, Jeff Mahn, Michael Mundt, and Michael Quintana are gratefully acknowledged.

ABBREVIATIONS AC—alternating current ACD—alternating current disconnect APS—Arizona Public Service ARC—antireflective coating A(t)—availability as a function of time BOS—balance of system cdfs—cumulative distribution functions c-Si—crystalline silicon DASs—data acquisition systems DC—direct current DOE—Department of Energy EVA—ethylene vinyl acetate FMEA—failure modes and effects analysis GRP—general renewal process IEA—International Energy Agency IEEE—Institute of Electrical and Electronic Engineers IGBT—insulated gate bipolar transistor IV—current voltage kWAC—kilowatt alternating current kWDC—kilowatt direct current kWhAC—kilowatt hour alternating current LEC—levelized energy cost ModJct—module junction boxes MTBFs—mean time between failures MWDC—megawatt direct current O&M—operation and maintenance PCU—power control unit PV—photovoltaic PVMaT—Photovoltaic Manufacturing Technology q—repair effectiveness RBD—reliability block diagram RGA—reliability growth analysis R(t)—reliability as a function of time SD—standard deviation SEIA—Solar Energy Industries Association Si—silicon STCs—standard test conditions

248

SOLAR CELL SYSTEMS

t—time TEP—Tucson Electric Power UL—Underwriters Laboratories, Inc. UV—ultraviolet WAC—watt alternating current WDC—watt direct current REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

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

L. M. Moore and H. N. Post. Five years of operating experience at a large, utilityscale photovoltaic generating plant. Progress in Photovoltaics: Research and Applications 16, 249–259 (2007). E. Collins, M. Dvorack, J. Mahn, M. Mundt, and M. Quintana. Reliability and availability analysis of a fielded photovoltaic system. In 34th IEEE Photovoltaic Specialists Conference, Philadelphia, June 7–12 (2009). L. Moore, H. Post, H. Hayden, S. Canada, and D. Narang. Photovoltaic power plant experience at Arizona Public Service—a 5-year assessment. Progress in Photovoltaics: Research and Applications 13, 353–363 (2005). L. Moore and H. Post. Photovoltaic power plant experience at Tucson Electric Power. Energy Conversion and Resources 2005, 387–394 (2005). Information available at www.tucsonelectric.com. Tucson Electric Power Company. Renewables Data for Year End 2006. Annual Report to the Arizona Corporation Commission. Available at http://www.greenwatts.com/Docs/ACCAnnual2006.pdf (2006). www.GreenWatts.com/pages/solaroutput.asp T. Hansen. The systems driven approach to solar energy: A real world experience. Proceedings of Solar Energy Systems Symposium, October 15–17, Albuquerque, NM (2003). Available at www.sandia.gov/pv (2003). J. E. Mason, V. M. Fthenakis, T. Hanse, and H. C. Kim. Energy payback and lifecycle CO2 emissions of the BOS in an optimized 3.5 MW PV installation. Progress in Photovoltaics: Research and Applications 14, 179–190 (2006). IEC. Photovoltaic system performance monitoring—Guidelines for measurement, data exchange, and analysis. IEC Standard 61724, Geneva, Switzerland (1998). B. Marion, J. Adelstein, K. Boyle, H. Hayden, B. Hammond, T. Fletcher, B. Canada, D. Narang, D. Shugar, H. Wenger, A. Kimber, L. Mitchell, G. Rich, and T. Townsend. Performance parameters for grid-connected PV systems. Proceedings of 31st IEEE Photovoltaic Specialists Conference, January 3–7, Lake Buena Vista, FL (2005). M. Thomas, H. Post, and R. DeBlasio. Photovoltaic systems: an end-ofmillennium review. Progress in Photovoltaics: Research and Applications 7, 1–19 (1999). L. Moore. Sandia’s PV reliability database: helping business do business, quarterly highlights of Sandia’s solar programs. Vol. 1. Available at www.sandia.gov/pv (2001). W. Bower. Inverters—Critical photovoltaic balance-of-system components: status, issues, and new millennium opportunities. Progress in Photovoltaics: Research and Applications 8, 113–126 (2000). A. Maish. Defining requirements for improved photo-voltaic system reliability. Progress in Photovoltaics: Research and Applications 7, 165–173 (1999).

REFERENCES [16] [17] [18] [19] [20]

[21] [22] [23]

249

R. West, K. Mauch, Y. C. Qin, N. Mohan, and R. Bonn. Status and needs of power electronics for photovoltaic inverters: summary document. SAND2002-1085, Sandia National Laboratories, Albuquerque, NM, April (2002). L. Moore, H. Post, T. Hansen, and T. Mysak. Residential photo-voltaic system experience at Tucson Electric Power. Sandia National Laboratories Internal Paper (2006). SEIA. Our solar power future: The U.S. photovoltaics industry roadmap through 2030 and beyond. January. Available at www.seia.org (2005). R. Little and M. Nowlan. Crystalline silicon photovoltaics: The hurdle for thin films. Progress in Photovoltaics: Research and Applications 5, 309–315 (1997). R. Mitchell, C. Witt, R. King, and D. Ruby. PVMaT advances in the photovoltaic industry and the focus of future photo-voltaic manufacturing R&D. In Proceedings of 29th IEEE Photovoltaic Specialists Conference, N.D. Stojadinovic, ed., pp. 1444– 1447 (2002). L. H. Crow. Reliability analysis for complex repairable systems. In Reliability and Biometry SIAM, Army Material System Analysis Activity Technical Report 138, pp. 379–410. Philadelphia (1974). M. Kijima and N. Sumita. Some results for repairable systems with general repair. Journal of Applied Probability 20, 851–859 (1989). R. Mettas and W. Zhao. Modeling and analysis of repairable systems with general repair. In IEEE Reliability and Maintainability Symposium Proceedings, pp. 176– 182 (2005).

11 LEVELIZED COST OF ENERGY FOR UTILITY-SCALE PHOTOVOLTAICS MATTHEW CAMPBELL SunPower Corporation

11.1

THE DRIVERS OF THE LCOE FOR UTILITY-SCALE PVs

PV power plants have emerged in recent years as a viable means of large-scale renewable energy power generation. A critical question facing these PV plants at the utility scale is the competitiveness of their energy generation cost with that of other sources. A common means of comparing the relative cost of electricity from a generating source is through an LCOE calculation. The LCOE equation allows alternative technologies to be compared when different scales of operation, investment, or operating time periods exist. This chapter reviews the LCOE drivers for a PV power plant and the impact of a plant’s capacity factor on the system LCOE. The impact of solar tracking to a plant’s capacity factor is reviewed as well as the economic trade-offs between fixed and tracking systems. From 2004 to 2008, the market for small (100×) optical concentration, III–V multijunction solar cells convert over 40% of sunlight into electrical power [1, 2]. The crossing of the 40% efficiency threshold in 2006 had more than just psychological value; it signaled that multijunction cells had reached a point at which their performance could more than offset their cost in high-concentration, two-axis tracking systems. Having served for more than a decade as the workhorse in space power applications, multijunctions are now being implemented for terrestrial, utility-scale solar power generation. III–V multijunction cells began displacing silicon single-junction cells in space applications in the 1990s. These structures are composed of compounds of elements in groups III and V of the periodic table. Though the cost of a III–V multijunction cell exceeds that of a silicon single-junction cell by at least an order of magnitude, this component cost is dwarfed by the cost of building and launching satellite payloads into space. In space, the higher performance of the multijunction means better economics of the overall system. For terrestrial applications, a similar situation exists with respect to high-concentration (over 100×) photovoltaic systems. High-concentration photovoltaic (HCPV) systems require relatively expensive optics and precise dual-axis tracking to maintain their focus on the sun throughout the day. The high-power densities (>10 W/cm2) incident on the cell surface require substantial investment in packaging and cooling design. Due to these higher system costs, the economics of the HCPV system favor an expensive, high-performance cell that then delivers higher energy output. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

293

294

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

A notion of the leveraging effect of cell performance on HCPV system economics is shown in Figure 13.1. The dependence of system cost on cell efficiency may be estimated for both fixed, flat-plate and concentrator system costs using reasonable assumptions for the various system component costs [3]. Component costs are taken from Swanson’s estimates for annual production volumes in the 10-kW to 10-MW range. For a 1× fixed, flat-plate system, the cell efficiency limit is assumed to be 30%. (The current record for single-crystal silicon is 25%.) Even assuming the cost of 500× cells is 300 times that for 1× cells, the low price for 1× cells is no match for the expensive, high-efficiency cells in a 500× HCPV system. The limiting case of zero-cost cells in a 500× system is also shown; it illustrates the relative insensitivity of an HCPV system to the component cell cost. Assuming modest production scale, the leveraging effect of high concentration will deliver a lower system cost in an efficiency range available with today’s III–V multijunction cells. In order to maintain high concentration on the multijunction cells, two-axis tracking is required. Though the need for this relatively expensive two-axis tracking is often cited as a disadvantage of the HCPV approach, it is worth reconsidering this viewpoint in light of the growing commitment to large-scale solar energy production. For solar energy to make a meaningful impact on the electrical grid, it needs to evolve into a large-scale, consistent energy resource. Solar panels deployed on fixed mounts or single-axis trackers are limited in this regard; they provide a peak power output that falls off when the panel surface is not directly facing the incident sunlight. By allowing panels to directly face the sun throughout the day and year, two-axis tracking provides the necessary consistency in power

Figure 13.1. 500× HCPV systems are relatively insensitive to cell cost, allowing highefficiency cells to deliver a low cost of power per system.

296

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

grown to contain any dislocations or other growth defects. A tunnel junction between the bottom and middle subcells passes current against the built-in field between them. The middle subcell is composed of Ga(In)As, an alloy of GaAs and a small percentage of indium to match the lattice constant of Ge. Following the second tunnel junction, the top GaInP subcell is grown. Both the top and middle subcells have BSFs to reduce carrier recombination losses at the back of the junction. Each of these subcells is also terminated with a front-side “window” layer. This consists of an alloy with a larger bandgap than the underlying emitter layer to reduce surface recombination losses in the emitter. Growth is completed above the top subcell with a thin “cap” layer of Ga(In)As that protects the underlying (aluminum-bearing) window layer from oxidation and serves as an ohmic contact for the front-side metal. During cell fabrication, the cap layer is etched away except in the vicinity of gridlines and an ARC is deposited on the window layer [4].

13.3

TUNNEL JUNCTIONS

The high efficiency of the III–V multijunction cell is due in part to its monolithic design. Since the layers are composed entirely of semiconductors, light is efficiently transmitted through intervening layers to its intended absorption layer. The monolithic design would not be possible without tunnel junctions. Due to the longer minority carrier lifetime of electrons than holes in most III–V materials, each subcell in the multijunction is doped as a (thin) n-type emitter on a (thick) p-type base. A complication in making a series connection of three such n-on-p diodes is that, at the interface of any two diodes, a p-on-n junction results. The presence of this diode of the opposite polarity would ordinarily serve to block current flow between subcells. To avoid this, a thin, degenerately doped p-on-n tunnel junction is inserted at the subcell interfaces. At relatively low current densities, the current–voltage characteristic of the tunnel junction is essentially ohmic, allowing current to flow between subcells (Fig. 13.3). Tunnel junctions make use of the quantum mechanical phenomenon that the positions of particles are not discrete but, rather, are defined by a probability distribution that falls off gradually on either side of their average positions. Such particles, placed near an energy barrier, would therefore have a nonzero probability of being on either side of the barrier: they can “tunnel” through the barrier. If the particles in question are charge carriers such as electrons or holes, and the barrier is sufficiently narrow for a large tunneling probability, large currents can be made to flow in a direction contrary to that indicated by the local electric field. The tunneling effect is strongest at voltages for which the alignment of the bandgaps at the junction is most favorable. In this region, the current–voltage response is essentially ohmic. As the current is increased, band alignment becomes less favorable and a falloff in current ensues. At higher voltages, the tunnel junction regains the characteristic of a typical p-n diode. A tunnel diode therefore has a peak tunneling current, above which the current will begin to fall off and return to that of the typical diode-like current–voltage

EFFICIENCY UNDER AM1.5

299

off at 500× concentration typically has an effective sheet resistance in the range of hundreds of ohms per square. In order to deliver maximum current under the terrestrial AM1.5 reference spectrum, the current balance between subcells must be readjusted with respect to the design for space (AM0). Absorption by the atmosphere, particularly in the near UV by the ozone layer, reduces the number of photons available to the top subcell. To compensate, the top subcell may be made thicker or with a lower bandgap in order to increase its current generation. When the middle subcell is GaAs or a low-indium Ga(In)As alloy, the bottom, Ge subcell will produce over 30% excess current with respect to the top and middle subcells. As a result, III–V multijunction cells are typically designed to produce equal amounts of current in the top and middle subcells to maximize the limiting current of the three-junction stack. The ARC is often designed so that this current-matched condition results when the cell is mated with a cover glass. 13.5

EFFICIENCY UNDER AM1.5

The cell efficiency measured under a particular reference spectrum is merely the first step in the prediction of the annual energy output of a III–V multijunction in terrestrial operation. However, it provides a convenient point for comparison of different technologies and provides guidance in cell optimization. Under standard test conditions, the cell is evaluated at a temperature of 25°C; illumination intensity is assumed to be 1000 W/m2; and the incident spectrum is the reference AM1.5, low AOD (G173-03 direct). To obtain a prediction of III–V multijunction efficiency, the current generated by three subcell materials over a range of bandgaps may be determined. Assuming each subcell of a given bandgap energy is able to convert available photons above this energy into an electrical current, it is possible to determine the limiting current for a set of bandgap combinations [5]. In the examples that follow, a typical external quantum efficiency of 94% is assumed for the portion of the incident spectrum with photon energy above the bandgap energy of each subcell. Wavelengths corresponding to energies below the bandgap energy are assumed to be transmitted to lower subcells or are lost. For the case where an upper subcell has available excess current with respect to the subcells below, the possibility of “leaking” excess photons from an upper subcell to a lower subcell is incorporated. In practice, this may be achieved either by thinning an upper subcell or via radiative coupling effects [6]. For the example shown in Figure 13.5, the bottom subcell is assumed to be Ge (with a bandgap energy of 0.67 eV). For ranges of top subcell and middle subcell bandgap energies, the resulting limiting current is thereby obtained. The standard diode equation may be solved for the cases where V = 0 and J = 0 to give an expression for the open-circuit voltage (VOC) as a function of shortcircuit current density (JSC): VOC =

nkT ⎛J ⎞ ⋅ ln ⎜ SC + 1⎟ , ⎝ ⎠ q J0

(13.1)

300

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

Figure 13.5. Isoefficiency plot for top and middle subcell bandgap ranges on a 0.67-eV bottom subcell at 500×. The bandgap ranges for several low lattice mismatch (100°C temperature excursions that occur in satellite orbit. Due to the often prolonged periods of storage prior to a launch, III–V multijunctions have passed space qualifications that included environmental testing for terrestrial conditions as well. Comparison with other terrestrial electronic applications also provides some insight. The compound semiconductors used in III–V semiconductors are widely deployed in military electronics, wireless handsets, and in long-haul (including undersea) fiber optic communications. High-brightness LEDs provide an instance where the operating current density exceeds that required in high-concentration photovoltaics [17]. Successful qualification testing and field performance in these industries give some confidence that sufficient reliability can be demonstrated in HCPV applications as well. If failure modes do exist, they are most likely to arise in the areas that are unique to multijunctions under high concentration. The lifetime of III–V multijunctions in the field will ultimately be determined by how they are integrated into specific HCPV systems. For example, unlike III–V multijunctions in space, cells for high-concentration photovoltaics must typically be mounted using solder to provide adequate thermal and electrical conductivity for their high current densities. Care must be taken to avoid the formation of voids in the solder under the cell. A solder void can trigger a thermal runaway condition in III–V multijunctions that will rapidly induce cell failure. A hot spot in the vicinity of a void, caused by nonuniformities in the concentrating optics, for example, leads to a reduction of the bandgaps of the subcells. This causes them to generate more current, which, in turn, makes the region still hotter; thermal runaway ensues. Though such failures will usually present themselves within a few minutes of on-sun operation, aggregation of small voids or stress cracking of the solder over time could bring about long-term degradation in this mode.

SUMMARY

309

One III–V compound known to oxidize readily is AlGaAs. Effective packaging allows for its use in various optoelectronic applications, but it has been substantially eliminated from most III–V multijunction cell structures. The top “window” layer of the cell is the one most exposed to potential environmental degradation. Typically composed of AlInP, it is less susceptible to oxidation than AlGaAs, but should still be encapsulated to retard long-term degradation. The use of ARC and a cover glass helps protect both the window layer and the front-side metallization from oxidation. The low bandgap energy of the Ge subcell makes it more susceptible to reverse bias breakdown. If strings of multijunction cells are connected in parallel strings, blocking diodes must be used so that strings with a higher voltage do not induce a reverse breakdown condition in an adjacent string. The economics of lowcost assembly processes must be balanced against such potential reliability issues. Rate of failure is often accelerated at higher temperatures. Since multijunctions deliver their highest efficiencies when they are cool, HCPV systems are designed so that the cell operating temperature remains relatively low, typically below 70°C. This confers on the system an added margin in terms of reliability. Within this temperature limit, long-term degradation mechanisms, such as lateral oxidation from the edges of the cells, can be substantially mitigated. This notwithstanding, considerable effort is still needed in order to fully vet III–V multijunctions for 25-year performance in HCPV operating conditions. The HCPV industry is proceeding to retire reliability risk by adapting the old aerospace maxim: test like you fly—fly like you test. In 2007, Spectrolab completed a celllevel qualification of multijunctions adapted from the IEEE 1513 standard for CPV receivers and modules. Ongoing qualifications draw on the more recent and more extensive IEC 62108 standard for modules. In parallel, HALT is being conducted by various HCPV groups to drive III–V multijunction packages to failure and thereby identifies potential failure modes in operation. Field testing, though slower, is the ultimate test. Multijunction HCPV systems have demonstrated 2-year operation without appreciable degradation [18]. As further test cycles are completed and more systems are deployed in the field, the necessary conditions to ensure a 25-year lifetime will be established.

13.9

SUMMARY

The 40% conversion efficiency of III–V multijunction cells under high concentration makes them an economical choice for use in HCPV systems operating at 500× and above. In order to realize this promise on a large scale, the design must be tailored to accommodate the cell’s response under operating conditions. As more experience is gained in the field, the virtuous cycle of deployment and further development will permit tuning of the design for long-term energy generation. Building on their legacy in space applications, III–V multijunction cells in HCPV systems will provide the world with large-scale, long-lasting, high-performance solar energy systems.

310

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

ABBREVIATIONS AlGaAs—aluminum gallium arsenide AlInP—aluminum indium phosphide AM—air mass AOD—aerosol optical depth ARC—anti-reflection coating BSF—back surface field CPV—concentrator photovoltaics DNI—direct normal irradiance Eg—semiconductor bandgap energy GaAs—gallium arsenide Ga(In)As—gallium arsenide, with trace amounts of indium GaInP—gallium indium phosphide Ge—germanium HALT—highly accelerated life testing HCPV—high-concentration photovoltaics IEC—International Electrotechnical Commission IEEE—Institute of Electrical and Electronic Engineers III–V—alloys using elements from columns III and V of the periodic table J—current density J0—reverse saturation current density Jmiddle—middle subcell current density (in a multijunction cell) Jsc—short-circuit current density Jtop—top subcell current density (in a multijunction cell) k—Boltzmann’s constant LED—light-emitting diode MOCVD—metal-organic chemical vapor deposition n—negatively doped semiconductor n—diode ideality factor NREL—National Renewable Energy Laboratory p—positively doped semiconductor q–charge of an electron SMARTS—Simple Model of the Atmospheric Radiative Transfer of Sunshine (http://www.nrel.gov/rredc/smarts/) T—absolute temperature UV—ultraviolet Voc—open-circuit voltage x—sunlight concentration ratio (direct sunlight is “1x”) γ—temperature coefficient for the ratio of semiconductor diffusion coefficients to minority carrier lifetimes

REFERENCES

311

REFERENCES [1] [2]

[3] [4] [5] [6]

[7] [8] [9] [10] [11]

[12]

[13] [14]

[15]

R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H. Yoon, R. A. Sherif, and N. H. Karam. 40% efficient metamorphic GaInP/GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90, 183516 (2007). J. F. Geisz, D. J. Friedman, J. S. Ward, A. Duda, W. J. Olavarria, T. E. Moriarty, J. T. Kiehl, M. J. Romero, A. G. Norman, and K. M. Jones. 40.8% efficient inverted triple-junction solar cell with two independent metamorphic junctions, Appl. Phys. Lett. 93, 123505 (2008). R. M. Swanson. The promise of concentrators. Prog. Photovolt. Res. Appl. 8, 93–111 (2000). D. Danzilio. Overview of EMCORE’s multi-junction solar cell technology and high volume manufacturing capabilities. Available at http://www.csmantech.org/ Digests/2007/2007%20Papers/01c.pdf (2007). Accessed April 3, 2009. M. W. Wanlass, K. A. Emery, T. A. Gessert, G. S. Horner, C. D. Osterwald, and T. J. Coutts. Practical considerations in tandem cell modeling. Solar Cells 27, 191–204 (1989). H. Yoon, R. R. King, G. S. Kinsey, S. Kurtz, and D. D. Krut. Radiative coupling effects in GaInP/GaAs/Ge multijunction solar cells. In Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Vol. 1, Osaka, Japan, May 11–18, pp. 745–748 (2003). G. S. Kinsey, P. Hebert, K. E. Barbour, D. D. Krut, H. L. Cotal, and R. A. Sherif. Concentrator multijunction solar cell characteristics under variable intensity and temperature. Prog. Photovoltaics Res. Appl. 16, 503–508 (2008). L. D. Partain. Solar cell fundamentals. In Solar Cells and Their Applications, 1st edition, L. D. Partain, ed., pp. 22–33. New York, Wiley (1995). W. Shockley and H. J. Queisser. Detailed balance limit of efficiency of p-n junction solar cells. J. Appl. Phys. 32, 510–519 (1961). Ioffe Physico-Technical Institute. Available at http://www.ioffe.rssi.ru/SVA/NSM/ Semicond/. Accessed January 21, 2009. C. H. Chen, S. A. Stockman, M. J. Peanasky, and C. P. Kuo. OMVPE growth of AlGaInP for high-efficiency visible light-emitting diodes. In High Brightness Light Emitting Diodes, Semiconductors and Semimetals, Vol. 48, G. B. Stringfellow and M. G. Craford, eds., pp. 97–144. San Diego, CA, Academic (1997). K. Nishioka, T. Takamoto, T. Agui, M. Kaneiwa, Y. Uraoka, and T. Fuyuki. Annual output estimation of concentrator photovoltaic systems using high-efficieny InGaP/ InGaAs/Ge triple-junction solar cells based on experimental solar cell’s characteristics and field-test meteorological data. Sol. Energy Mater. Sol. Cells 90, 57–67 (2006). G. S. Kinsey and K. M. Edmondson. Spectral response and energy output of concentrator multijunction solar cells. Prog. Photovoltaics Res. Appl. 17, 279–299 (2009) (in press). D. Aiken, M. Stan, C. Murray, P. Sharps, J. Hills, and B. Clevenger. Temperature dependent spectral response measurements for III-V multi-junction solar cells. In Proceedings of the 29th IEEE PV Specialists Conference, New Orleans, LA, May 21–24, pp. 828–831 (2002). W. E. McMahon, K. E. Emery, D. J. Friedman, L. Ottoson, M. S. Young, J. S. Ward, C. M. Kramer, A. Duda, and S. Kurtz. Fill factor as a probe of current-matching for GaInP2/GaAs tandem cells in a concentrator system during outdoor operation. Prog. Photovoltaics Res. Appl. 16, 213–224 (2007).

312

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

[16]

C. Gueymard. SMARTS, a simple model of the atmospheric radiative transfer of sunshine: Algorithms and performance assessment. Technical Report No. FSEC-PF270-95. Cocoa, FL: Florida Solar Energy Center (2005). M. Vazquez, C. Algora, I. Rey-Stolle, J. R. González. III-V concentrator solar cell reliability prediction based on quantitative LED reliability data. Prog. Photovoltaics Res. Appl. 15, 477–491. P. J. Verlinden, A. Lewandowski, H. Kendall, S. Carter, K. Cheah, I. Varfolomeev, D. Watts, M. Volk, I. Thomas, P. Wakeman, A. Neumann, P. Gizinski, D. Modra, D. Turner, and J. B. Lasich. Update on the two-year performance of 120 kWp concentrator systems using multi-junction III-V solar cells and parabolic reflective optics. Proceedings of the 33rd Photovoltaic Specialists Conference, San Diego, CA, May 12–16 (2008).

[17] [18]

14 HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS GERHARD PEHARZ AND ANDREAS BETT Fraunhofer Institut für Solare Energiesysteme (ISE)

14.1

INTRODUCTION

One of the most important benchmarks for energy technologies is the costs per generated kilowatt hour. Primarily, the levelized costs of electricity produced by PV technologies are dominated by the area-related costs of the solar cells, the conversion efficiency, and the lifetime. In the recent decades of PV development, it has been a challenge to improve all three of these cost factors simultaneously. Nevertheless, the levelized costs of PV electricity have decreased; however, the goal of grid parity has not yet been achieved. Grid parity means that the price per PV generated kilowatt hour equals the price per kilowatt hour paid in a household. Please note that grid parity is very country specific. A concept that could accelerate the cost reduction of PV electricity is the use of PV concentrators. The basic idea of concentrator PV is to use an optical element that focuses sunlight onto a relatively small solar cell. As a result, the solar cell-related costs in a PV concentrator system are of minor impact, which allows highly efficient cells, which are usually comparatively expensive, to be used. Today, the most efficient solar cells are III–V-based triple-junction cells. Under concentrated light, they achieve efficiencies of over 40%. However, their area-related costs are close to two magnitudes higher than those of bulk silicon solar cells, which clearly dominate the solar cell market today. For terrestrial applications, the highly efficient triple-junction solar cells are ideally used in combination with cheap optics, which collect the sunlight over an area that is several hundred times larger than the solar cell area and focus it on the solar cell. One of the cheapest known optical

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

313

314

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

elements for concentrating sunlight is a Fresnel lens. Hence, Fresnel lenses are often used for concentrator applications. Fresnel lenses have been used since the beginning of PV concentrator development. In this field, the first extensive experience was made by Sandia National Laboratories. Some of the key issues for PV concentrator design are addressed in an excellent work of Burgess et al. [1]. In contrast to non-concentrating PV systems, the solar cells in a concentrator have to operate under a much higher flux density. Consequently, it is of great importance that the solar cell and its related electrical assemblies have a low series resistance. Furthermore, the generated heat flux must be removed effectively; otherwise, the solar cell or assemblies can suffer damage. Thus, an effective cooling of the solar cells is required. This can be carried out passively by using air convection or actively by conducting a cooling fluid through a heat exchanger. From a designer’s point of view, passive cooling seems to be more desirable because it is less complex than active cooling. However, passive cooling requires relatively large areas to dissipate the heat. In a Fresnel lens concentrator, the solar cell is located beneath the Fresnel lens and thus at least, an area equivalent to the lens aperture is available for heat dissipation. This option is not possible for parabolic mirror concentrators where the solar cell is located in front of the optics and causes shading. One of the fist concentrator installations at Sandia National Laboratories used acrylic Fresnel lenses with an area of 912 cm2 and silicon solar cells with an area of 15.2 cm2 [1]. The ratio of lens area and cell area defines the geometric concentration ratio, which was 60X for this system. A two-axis tracking system was used to align the arrays to the sun and to keep the focal spots on the cells. The Fresnel lens concentrator concept of Sandia National Laboratories was adopted and further developed by other research institutes in the late 1970s. At the UPM, a PV concentrator using hybrid glass silicone–glass Fresnel lenses [2, 3] was installed. A photograph of this early Fresnel lens concentrator is shown in Figure 14.1. Basically, these silicone–glass Fresnel lenses were thought to be more reliable than acrylic lenses in terms of abrasion and UV degradation. Furthermore, the costs were expected to be lower than those of acrylic Fresnel lenses. Similar to the Sandia system, the Ramón Areces concentrator used silicon solar cells. The early concentrators were equipped with solar cells having a size of about 15 cm2. Considering a concentration ratio of 50X, the cells receive a light power of about 25 W, of which less than 10% was converted into electrical energy and the rest was mostly converted into heat. This heat needed to be dissipated, and early Fresnel lens concentrators used massive metal elements of several millimeters thickness. Furthermore, special attention must be drawn to thermal matching, when using solar cells with an area of several square centimeters. Differences in the thermal expansion coefficients must be taken into account when designing the interconnection of cell and heat sink. Consequently, thermal issues were of great importance for the early concentrators due to the relatively large cell area. In the 1990s, a Fresnel lens concentrator was also developed by Fraunhofer ISE, Germany, in cooperation with the Ioffe Institute in Russia [4–6]. In contrast

INTRODUCTION

315

Figure 14.1. The Ramón Areces installation at the Universidad Politécnica de Madrid (UPM). This PV concentrator used silicone–glass Fresnel lenses, silicon solar cells, and a two-axis tracking system. The cooling was managed passively by using massive metallic elements. The person in front of the concentrator is Prof. Gabriel Sala of the UPM. Courtesy of Prof. Gabriel Sala, UPM.

to the early Fresnel lens concentrators, the dimensions were downscaled by about a factor of 100. The solar cell area was reduced to several square millimeters and was made of III–V materials. This relaxes the requirements on the passive cooling. In particular, a copper heat sink with a thickness of several 100 μm is sufficient to keep the cell cool, even when operated under a concentration ratio of up to 1000X. Already in the year 1976, James et al. demonstrated that for these high concentration ratios, solar cells made from III–V semiconductors (in particular, AlGaAs/ GaAs heteroface single junction) work very well [7]. Furthermore, work on highly efficient III–V multijunction cells has been going on since the late 1980s [8, 9]. In 2001, Fraunhofer ISE introduced monolithic multijunction solar cells in concentrator modules. A concentrator module consisting of 40 × 40 mm2 Fresnel lenses and 13 mm2 metamorphic Ga0.35In0.65P/Ga0.83In0.17As dual-junction cells reached an efficiency of 24.8% measured under natural sunlight without temperature correction [6]. At this time, it was the highest reported efficiency of a Fresnel lens concentrator module. Motivated today by the fast increase of III–V multijunction efficiencies and therefore the potential to lower the cost per kilowatt hour, many start-up companies have sprung up in the field of PV concentrators. The main application for these

316

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

technologies is the production of electricity in large-sized power plants (several megawatt power scales) in sun-rich countries with a high share of direct sunlight. However, some companies are also aiming for rooftop applications. The following sections describe the most important components of a Fresnel lens concentrator and present some recent examples.

14.2

ELEMENTS OF FRESNEL LENS CONCENTRATORS

This section deals with the various elements found in every high-concentration Fresnel lens concentrator, for the most part. The Fresnel lens optics and the triplejunction cells are described in different subsections. Beside some basics, an overview is given about the manufacture as well as recent developments in the field. Please note that in particular, Fresnel lens systems with a two-dimensional concentration are considered here.

14.2.1

Fresnel Lens Optics

Basically, a Fresnel lens has optical properties that are similar to conventional lenses; however, they have lower material usages and therefore lower cost. This is due to the fact that a Fresnel lens is obtained by breaking a conventional lens into a set of concentric segments. Figure 14.2 shows a sketch of a conventional and planar Fresnel lens with the same size and focal distance. As a result of the faceted structure, a Fresnel lens offers an additional dimension in design space. The angle of each facet can be varied in order to decrease spherical aberration or the homogeneity of the light distribution in the focal spot [10]. In PV concentrators, Fresnel lenses with a flat surface are often used, similar to the one shown in Figure 14.2. They can be produced in a molding or stamping process, using acrylic or silicone materials. The relatively low manufacturing and material costs makes Fresnel lenses one of the cheapest available concentrating optics. This is the reason why they have been used from the beginning of PV

(A)

(B)

Figure 14.2. (A) A sketch of the cross section of a conventional lens and above a corresponding Fresnel lens with the same size and focal length. (B) The concentric segments of the Fresnel lens are sketched in a plan view.

318

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

However, when closed facet spacing is used, diffraction effects can decrease the optical efficiency of the Fresnel lens. These losses can become significant at facet diameters around 100 μm. In addition to the loss mechanisms specific to the Fresnel lens, the typical optical losses for lenses like aberration, absorption, scattering at surface imperfections, and surface reflection must be considered. Reflection losses at surfaces can be reduced with appropriate antireflection coatings; however, these coatings increase the cost of the Fresnel lens. For imaging optical elements, the maximal achievable concentration is determined by the ratio of focal length and diameter of the optics; this follows from the étendue. A flat Fresnel lens (as shown in Fig. 14.2) designed for high concentrations must refract light rays impinging the corner of the lens in a steep angle. This increases the chromatic aberration of the lens, and in extreme cases rays from the lens corners are lost due to total internal reflection. Thus, the maximal concentration of a Fresnel lens is limited by the chromatic aberration and reflection losses. The exact limit depends on a number of parameters (e.g., the refractive index). Typically, the maximum concentration limit is quoted to be about 500X for flat Fresnel lenses [11]. For geometric concentrations 40% under the reference spectrum. This was made possible by introducing subcell bandgaps that allow a current matching condition for the reference spectrum. However, the sun’s spectrum varies

326

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

Figure 14.8. The efficiency and the fill factor of a Ga0.35In0.65P/Ga0.83In0.17As/Ge triplejunction solar cell versus concentration. This cell was manufactured at Fraunhofer ISE and measured at the CalLab of Fraunhofer ISE. One can see that the efficiency increases with increasing concentration and peaks at about 454X. For higher concentrations the series resistance of the cell and the related drop in fill factor decreases the efficiency. However, at 800X the efficiency is still remarkably high (>40%).

during the course of the day and season and is usually not equal to the reference spectrum. Due to the resulting current limitations, it is expected that concentrator systems with triple-junction cells are more sensitive to changes in the solar spectrum than CPV systems using single-junction cells. The influence of the spectral variations on the electrical parameters of a triplejunction solar cell can be investigated in the laboratory by using a multi-light source simulator [34, 35]. Here, the electrical parameters of the triple-junction solar cell are measured, while the incident spectrum is changed systematically. Thus, on the cell level, the influence of the spectral changes is well investigated. Under varying sun spectra, the power of triple-junction solar cells can be expected to vary by about 20%. Unfortunately, the influence of the solar spectrum on the energy production of concentrator systems equipped with triple-junction cells is not well investigated at present. Long-term field experience is still lacking due to the rather young triple-junction technology and the lack of information on the changing solar spectrum combined with concentrator optics. However, some investigations of Fresnel lens concentrators equipped with triple-junction cells have been already performed and show the spectral impact in the morning and evening hours [36–38]. At these high air masses, spectral distribution of the sunlight is relatively “redrich.” As a result, the top cell current is significantly lower than the current of the other subcells and the efficiency of the concentrator system decreases.

RECENT EXAMPLES AND DEVELOPMENTS

327

Obviously, the annual energy production of concentrator systems using triplejunction solar cells is more affected by the solar spectrum than systems with singlejunction cells. In a number of theoretical studies, the annual energy production of multijunction solar cells with respect to the solar spectrum has been investigated [39, 40]. The basic conclusion from these theoretical investigations is that despite the penalty of current limitation, triple-junction solar cells outperform other solar cell concepts. This is not surprising when considering that air masses 30 kW) with a high stowed packaging efficiency. The SquareRigger solar array system is projected to achieve a specific power between 180- and 260-W/kg BOL, depending on the type of cells used. A SquareRigger system using thin-film cells is projected to offer an order-of-magnitude reduction in cost over conventional rigid panel systems. Ultraflex Solar Arrays were recently used to provide power to the Mars Phoenix Lander (see Fig. 18.9). Launched in August 2007, the Phoenix Mars Mission is the first in NASA’s Scout Program. Phoenix is designed to study the history of water and habitability potential in the martian arctic’s ice-rich soil. This was the first flight for this unique solar array technology developed by ATK’s Goleta, California facility. Each Ultraflex array unfolded like an oriental fan into a circular shape 2.1 m in diameter and will generate 770 W of power from sunlight at the distance Earth is from the sun. Since Mars is approximately 1.5 times farther from the sun, the solar arrays will produce less than half the power possible on

412

SPACE SOLAR CELLS AND APPLICATIONS

Earth. These same arrays have been projected as the solar arrays for use on Orion, the new proposed crew exploration vehicle to replace the shuttle.

18.2.5

Concentrating Arrays

Photovoltaic concentrating arrays have been proposed for missions to outer planetary missions, SEP missions, and missions that operate in high-radiation environments. These arrays are attractive for these missions because they have the potential to provide a high specific power, higher radiation tolerance, and improved performance in LILT environments. The technical issues in using concentrating arrays are precision pointing, thermal dissipation, nonuniform illumination, optical contamination, environmental interactions, and complexity of deployment. They can also decrease overall spacecraft reliability because a loss of pointing may cause significant power loss to the spacecraft. Reflective systems can have concentration ratios from 1.6X to over 1000X, with a practical limit of around 100X. Refractive designs are generally limited to the range of about 5X–100X, with a practical limit of around 20X. Solar energy may be focused on a plane, line, or point depending on the geometry of the concentrator design. These concentrators may be small and numerous if used in a distributed focus design, or they may be a single large concentrator as in a centralized focus design. AstroEdge™ array on the NRO STEX spacecraft, launched in October 1998, was the first spacecraft to use a concentrator as its main power source. This system used a reflective trough design with a nominal 1.5X concentration. The arrays were successfully deployed and cell currents were slightly higher than predicted. Thermal problems did occur on some of the panels owing to the higher concentrator operating temperature. The Deep Space 1 spacecraft launched in October 1998 used ScarletTM concentrator arrays to provide power to its ion propulsion engines [21]. Its two arrays were capable of producing 2.5 kW at 100 VDC. The ScarletTM array was developed by AEC-Able under a program sponsored by BMDO. The ScarletTM array had a refractive linear distributed focus with a 7.5X concentration ratio. The array has 720 lenses to focus sunlight onto 3600 solar cells. Deep Space 1 has two ScarletTM SAW assemblies. Each assembly is made up of a composite yoke standoff structure, four composite honeycomb panel assemblies, and four lens frame assemblies. High-efficiency TJ GaInP2/GaAs/Ge cells were used in this array. The first commercial concentrator array developed for space was the Boeing 702. It was used on the Galaxy XI spacecraft and was deployed on January 12, 2000. It had a reflective planar centralized focus concentrator design in which the sun’s rays were reflected onto a single rectangular plane of solar cells. It used thin-film reflectors and had a 1.7X concentration. It was designed for power levels of 7–17 kW over a 16+-year design life. The array deployed as expected and its initial power output was within the expected

ARRAYS

413

Figure 18.11. Flexible blanket version of the stretched lens array.

range. However, its concentrator surfaces degraded very quickly while in orbit. The specific power of this array was ∼60 W/kg using 24% efficient MJ solar cells. The similar Boeing 601 bus, which uses an ordinary planar solar array, is limited to about 15 kW of power owing to the array stowed volume limitations. Current efforts in concentrator arrays at NASA at the Department of Defense focus on the SLA (see Fig. 18.11) using a Square Rigger structure discussed above.

18.2.6

High-Temperature/High-Intensity Arrays

Missions to Mercury and other missions with close encounters to the sun (i.e., solar probe) have generated the need for cells and arrays that are capable of operating in high-light-intensity, high-radiation, and high-temperature environments. Two missions that had to contend with such an environment have already flown. Helios A, which reached 0.31 AU—the average earth-to-sun distance, was launched on December 10, 1974, and Helios B, which reached 0.29 AU, was launched on January 15, 1976. These spacecrafts used ordinary Si cells that were modified for high-intensity use and had second surface mirrors to cool the array. The remainder of their technology was very similar to what is used on standard arrays. In addition to these missions, the upcoming MESSENGER Discovery mission is planned for travel to 0.31 AU. Its solar array design is already under development. The current solar array technology can meet the needs of MESSENGER or other spacecrafts that approach the sun to about 0.3 AU, but with reduced performance and increased risk compared to other applications. Further progress is

414

SPACE SOLAR CELLS AND APPLICATIONS

required in both cell and array development for closer encounters to the sun. The common feature to the high-temperature and high-intensity solar arrays that have operated thus far is the replacement of a significant fraction of the solar cells by OSRs. These are mirrors that help control the array temperature near the sun at the cost of reduced power at larger distances. The MESSENGER design also off-points the array as the spacecraft nears the sun to keep the array below 130°C. The array is designed to tolerate pointing at the sun for a maximum of 1 h (probably much longer). However, it will be unable to function under this extreme condition (i.e., 260°C). The U.S. Air Force and BMDO also developed some high-temperature arrays in the late 1980s. The SCOPA and SUPER were designed to be capable of surviving laser attack. These were concentrator arrays that directed the incident laser light away from the solar cells. Although the laser light would not impinge directly, the arrays’ temperature would increase dramatically and thus, the arrays needed to withstand several hundred degrees Celsius. The high-temperature survivability of SCOPA and SUPER was achieved through changes to the contact metallization and through the use of diffusion barriers in the GaAs cells used. Both Tecstar and Spectrolab developed the cells in conjunction with this effort. Other smaller companies such as Astropower, Kopin, and Spire have also worked on developing high-temperature cells. GaAs cells reaching an AM0 efficiency of 18% were produced, which degraded less than 10% under 1 sun after annealing in vacuum for 15 min at 550°C. Concentrator cells that survived repeated 7-min excursions to 600°C were produced. These same cells exhibited only 10% loss with exposure to 700°C. NASA is also currently funding an effort to develop wide bandgap solar cells for high-temperature/high-intensity environments. Cells using materials such as SiC, GaN, and AlGaInP are being developed [23]. These cells may also benefit from high-emissivity selective coatings that will limit the unusable IR entering the solar cells and reduce their steadystate temperature.

18.2.7

Electrostatically Clean Arrays

There is an entire class of proposed missions designed to study the SEC. These spacecrafts typically measure the fields and particles associated with the solar wind. This requires that arrays be developed that do not distort the local environment or be electrostatically “clean.” These arrays must have their voltage separated from the space plasma and the array must be maintained at the same potential as the spacecraft. This is usually accomplished by coating the cell cover glass and arrays between the cells with a conductor. Since the coating for the cover glass must be transparent, a TCO such as indium tin oxide is used. The coatings between the cells must not short them out, so an insulating coating must first be applied to all of the interconnects before the conductive coating or “v” clips. All of this must be done within a thickness of ∼0.08 mm and within a width of about 0.8 mm.

INTEGRATED POWER SYSTEMS

415

Fabricating an electrostatically clean array presently costs three to six times as much as a typical array. This is due in large part to the hand labor involved in developing such arrays. These arrays are also less reliable due to the lack of robustness of the conductive coatings used to maintain the equipotential. In addition, these arrays are also generally body mounted, which cuts down on the available power to the spacecraft (i.e., pointing issues). The power is also limited due to the thicker cover glass that is employed owing to the high-radiation environment associated with SEC missions. Unfortunately, there is no a wide knowledge base on how to develop electrostatically clean arrays. This was demonstrated in the cost of developing the FAST solar array. The electrostatically clean body-mounted solar panels for FAST cost in excess of $7400 per test condition watt. The use of monolithic diodes on the latest generation of MJ solar cells could prove to be a tremendous advantage in developing electrostatically clean arrays. The presence of antennas, booms, and outcroppings from a body-mounted array requires that solar cells have bypass diodes to reduce the shadowing losses and potential damage to the arrays. The new built-in diodes will obviate the need and the expense involved in adding the diodes to the array circuitry. The NASA GSFC recently funded COI to study electrostatically clean arrays through the STP Program’s MMS and GEC projects. COI will be supplying the electrostatically clean solar panels for the CNOFS.

18.2.8

Mars Solar Arrays

Mars orbiters have used photovoltaic arrays that are quite similar to those used in Earth orbit with good results. However, Mars surface missions, in which the solar spectrum is depleted at short wavelengths, cause the efficiency of the cells to be lower than if the cells were operated above the atmosphere of Mars. The cell efficiency is reduced by about 8% (relative percent). In addition, the effect of dust accumulating on arrays was observed on the Mars Pathfinder mission by monitoring the JSC of cells exposed to the environment whose short-circuit current could be monitored on a routine basis. One cell indicated an increase in obscuration of about 0.3%/sol for the first 20 sols (note that a “sol” is a martian day of 24.6 h). The other cell indicated that over a longer period of ∼80 sols, the obscuration flattened out and seemed to be approaching an asymptote of around 20% obscuration [24]. Cells that are “tuned” to the martian solar spectrum and methods for mitigating dust obscuration will be necessary to produce efficient arrays for Mars surface power.

18.3

INTEGRATED POWER SYSTEMS

NASA has also been working to develop lightweight, integrated space power systems on small-area flexible substrates [24]. These systems generally consist of a high-efficiency thin-film solar cell, a high-energy density solid-state Li-ion

416

SPACE SOLAR CELLS AND APPLICATIONS

battery, and the associated control electronics in a single monolithic package. These devices can be directly integrated into microelectronic or MEMS devices and are ideal for distributed power systems on satellites or even for the main power supply on a nanosatellite. These systems have the ability to produce constant power output throughout a varying or intermittent illumination schedule as would be experienced by a rotating satellite or “spinner” and by satellites in a LEO by combining both generation and storage. An integrated thin-film power system has the potential to provide a low-mass and low-cost alternative to the current SOA power systems for small spacecrafts. Integrated thin-film power supplies simplify spacecraft bus design and reduce losses incurred through energy transfer to and from conversion and storage devices. It is hoped that this simplification will also result in improved reliability. The NASA Glenn Research Center has recently developed a microelectronic power supply for a space flight experiment in conjunction with the Project Starshine atmospheric research satellite (http://www.azinet.com/starshine/). This device integrates a seven-junction small-area GaAs MIM with an allpolymer LiNi0.8Co0.2O2. The array output is matched to provide the necessary 4.2-V charging voltage, which minimizes the associated control electronic components. The use of the matched MIM and thin-film Li-ion battery storage maximizes the specific power and minimizes the necessary area and thickness of this microelectronic device. This power supply was designed to be surface mounted to the Starshine 3 satellite, which was ejected into a LEO with a fixed rotational velocity of 5° per second. The supply is designed to provide continuous power even with the intermittent illumination due to the satellite rotation and LEO [25].

18.3.1

High Specific Power Arrays

To achieve an array-specific power of 1 kW/kg, a much higher cell specific power will be necessary. Similarly, the blanket specific power (i.e., interconnects, diodes, and wiring harnesses) must be over 1 kW/kg as well. The APSA assessment determined that the mass of the deployment mechanism and structure is essentially equal to the blanket mass for a lightweight system [26]. Therefore, a blanket specific power of approximately 2000 W/kg would be necessary to achieve a 1 kW/kg array. NASA is currently sponsoring an effort by AEC-Able Engineering to develop lightweight thin-film array deployment systems. Gains in array-specific power may be made by an increase in the operating voltage. Higher array operating voltages can be used to reduce the conductor mass. The APSA was designed for a 28-V operation at several kilowatts output, with the wiring harness comprising ∼10% of the total array mass, yielding a specific mass of ∼0.7 kg/kW. If this array was designed for a 300-V operation, it could easily allow a reduction of the harness specific mass by at least 50%. This alone would increase the APSA specific power by 5% or more without any other modification.

POWER SYSTEM FIGURE OF MERIT

417

The extremely high specific power arrays that need to be developed for SEP and SSP applications will require lightweight solar arrays that are capable of highvoltage operation in the space plasma environment. SEP missions alone will require 1000–1500 V to directly power electric propulsion spacecraft (i.e., no voltage step-up is required to operate the thrusters). NASA has proposed a thin-film stand-alone array-specific power that is 15 times the SOA III–V arrays, an area power density that is 1.5 times that of the SOA III–V arrays, and specific costs that are 15 times lower than the SOA III–V arrays [27].

18.3.2

High-Radiation Environment Solar Arrays

There are several approaches to mitigating the effects of a high radiation on a solar array. The simplest is to employ a thick cover glass (assuming that a commercial source could be developed). Thick cover glasses protect the cell from the highly damaging low-energy protons but will cause a significant decrease in the specific power of the array. However, this can be reduced if one adopts a concentrator design, assuming of course that the additional elements associated with the concentrator can withstand the high-radiation environment as well. A different approach is to try and develop cells that are more radiation resistant. Several of the materials that are being investigated for high-temperature/high-intensity missions have also shown good radiation resistance. However, these will not be suited to the high-radiation missions involving LILT. Many of the high-radiation NASA missions being considered occur at distances much greater than 1 AU. Thin films may offer a possibility since they have demonstrated some advantages with regard to radiation tolerance as previously mentioned, provided the problem of their low efficiencies are solved.

18.4

POWER SYSTEM FIGURE OF MERIT

There are many figures of merit that must be considered in developing an SSP system (i.e., specific mass, specific power, cost per watt, temperature coefficients, and anticipated radiation degradation of the solar cells used). The radiation hardness and the temperature coefficients for the III–V MJ cells are significantly better than Si cells, as previously discussed. This leads to a significantly higher EOL power level for an MJ cell as compared with a Si cell. This is shown in Table 18.4, where the BOL cell efficiencies at room temperature and the typical EOL cell efficiencies for LEO and GEO operating temperatures and radiation environments are presented. The difference in radiation degradation can have a huge impact on power system design. For example, if the area for a typical rigid panel is approximately 8 m2 and the area of a typical solar cell is 24 cm2, using a panel packing factor of 0.90 will allow the panel to have 3000 cells. Under GEO conditions, this panel populated with high-efficiency Si cells will produce 1.2 kW of EOL power. The

418

SPACE SOLAR CELLS AND APPLICATIONS

TABLE 18.4. A Comparison of the Relative Radiation Degradation of 75-μm Multijunction Cells and High-Efficiency Si under GEO and LEO Operation [24] Solar Cell Technology

BOL Efficiency at 28°C (%)

EOL Efficiency on Orbit (%)

GEO conditions (60°C)—1 MeV, 5E14 e/cm2 HES

14.1

12.5

2J III–V

20.9

20.0

3J III–V

23.9

22.6

LEO conditions (80°C)—1 MeV, 1E15 e/cm

2

HES

13.4

10.6

2J III–V

19.7

18.1

3J III–V

22.6

20.3

EOL power could almost be doubled to 2.2 kW if it were populated with SOA TJ cells. Alternatively, the solar arrays populated with high-efficiency Si cells would need to be 77% larger than arrays using TJ cells in order to deliver the equivalent amount of EOL power in GEO and 92% larger in LEO. The large difference in size between solar arrays populated with Si and MJ cells is very significant in terms of stowage, deployment, and spacecraft attitudinal control. This is especially true for very high-powered GEO communications satellites in which the Si solar array area can exceed 100 m2. The comparable array with TJ cells, although by no means small, would have an area of ∼59 m2. The array size will impact the spacecraft’s weight, volume (array stowage), and system requirements on spacecraft attitude control systems (additional chemical fuel). Three important figures of merit used in power system optimization are EOL area power density (watt per square meter), specific weight (watt per kilogram), and cost (dollar per watt). Representative values for the various SOA cell technologies are listed in Table 18.5. The EOL power per unit area for an MJ cell is significantly better than a Si cell. However, the EOL specific weight for Si is almost a factor of 2 greater than an MJ cell. This results in a slightly smaller EOL cost per watt for a highefficiency Si. This demonstrates the dramatic reduction in cost of MJ cells over the past few years. If one considers the mass of the necessary array components (i.e., panel substrate, face sheet, adhesive, hinges, insulators, and wiring) along with equivalent power per area for the different cell types, and also the cost involved in having the CIC and laid on rigid panels, then the cost for developing an array using MJ cells is slightly less than that for HES cells. The EOL specific weight values at the

POWER SYSTEM FIGURE OF MERIT

419

TABLE 18.5. EOL Area Power Density (W/m2), Specific Weight (W/kg), and Normalized (to HES) Cost ($/W) for High-Efficiency Si, Dual-Junction (2J), and Triple-Junction (3J) Bare Solar Cells [24] Solar Cell Technology

W/m2

W/kg

Normalized Cell Cost ($/W)

GEO conditions (60°C)—1 MeV, 5E14 e/cm2 75-μm HES

169

676

1.00

2J III–V

271

319

1.38

3J III–V

306

360

1.22

LEO conditions (80°C)—1 MeV, 1E15 e/cm2 75-μm HES

143

574

1.00

2J III–V

245

288

1.29

3J III–V

275

323

1.15

Table 18.6. EOL Specific Weight (W/kg) at the CIC and Panel Levels for 3-mil HighEfficiency Si, Dual-Junction (2J), and Triple-Junction (3J) Cells [24] Solar Cell Technology

CIC Sp. Power (W/kg)

Panel Sp. Power (W/kg)

Normalized Panel Cost ($/W)

75

1.00

GEO conditions (60°C)—1 MeV, 5E14 e/cm2 75-μm HES

261

2J III–V

219

95

0.9

3J III–V

248

108

0.8

LEO conditions (80°C)—1 MeV, 1E15 e/cm2 75-μm HES

221

63

1.00

2J III–V

199

86

0.84

3J III–V

223

97

0.75

CIC (with 100-μm ceria-doped microsheet cover glass) and the panel levels for these cells and the normalized cost per watt for the panels are shown in Table 18.6. A similar comparison with somewhat less expensive 100-μm HES cells (at the panel level) shows a slightly smaller cost advantage for the MJ cells. The 100-μm Si cells are less radiation-hard and have lower specific power (watt per kilogram) than 75-μm Si cells, but they cost about 35% less.

420

SPACE SOLAR CELLS AND APPLICATIONS

Currently, conventional space Si cells are less expensive than MJ cells at the panel level. However, their EOL power is much lower than either the high-efficiency Si cells or the MJ cells. The increased mass and area that their usage entails would have to be considered against the cost savings and other mission considerations in any comparative study. Engineers have worked on ways to improve space solar cells and arrays in terms of all the important figures of merit since the early days of our space program. Numerous mission studies have shown that even extremely high array costs can be worth the investment when they result in a lower array mass. In general, mass saving in the power system can often be used by payload. If the revenue generated by this payload (i.e., more transmitters on a communications satellite) is greater than the cost of higher-efficiency solar cells, the choice is rather an easy one to make. However, often more instrument capabilities will require more support from the spacecraft (e.g., command and data handling, structure, and attitude control) as well as more power. These additions can negate any apparent advantage to the overall spacecraft.

ABBREVIATIONS 2J—two junction 3J—three junction AFRL—Air Force Research Lab AIAA—American Institute of Aeronautics and Astronautics AlGaInP—aluminum gallium indium phosphide AM0—air mass zero AM1.5—air mass 1.5 APSA—advanced photovoltaic solar array a-Si—amorphous silicon a-Si:H—amorphous silicon-hydrogen alloy (same as a-Si) ASTM—American Society of Testing and Materials BMDO—Ballistic Missile Defense Organization BOL—beginning of life CdTe—cadmium telluride CIC—cells interconnected and covered CIGS—copper indium gallium diselenide CNOFS—Communication/Navigation Outage Forecast System COI—Composited Optics Incorporated c-Si—crystalline silicon DJ—double junction e—charge of an electron E—exponential ELO—epitaxial liftoff EOL—end of life FAST—fast auroral snapshot

ABBREVIATIONS

421

GaAs—gallium arsenide GaInAs—gallium indium arsenide GaInP—gallium indium phosphide GaN—gallium nitride GaSb—gallium antimonide Ge—germanium GEC—geospace electrodynamic connection GEO—geosynchronous GSFC—Goddard Space Flight Center HES—high-efficiency silicon HIHT—high intensity, high temperature HPSA—high-power solar array i—undoped or intrinsic semiconductor IB—intermediate band IEEE—Institute of Electrical and Electronic Engineers III–V—semiconductor alloys composed of elements from columns III and V of the periodic table IMM—inverted metamorphic InP—indium phosphide IR—infrared ISS—International Space Station JPL—Jet Propulsion Laboratory LEO—low earth orbit LILT—low intensity, low temperature LiNi0.8Co0.2O2—lithium-ion thin-film battery MEMS—microelectromechanical systems MEO—medium earth orbit MIM—monolithically integrated photovoltaic module MJ—multijunction MMS—magnetospheric multiscale n—negatively doped semiconductor NASA—National Aeronautics and Space Administration NREL—National Renewable Energy Laboratory OMVPE—organometallic vapor phase epitaxy OSR—optical solar reflector p—positively doped semiconductor PERL—passivated emitter, rear locally diffused PVSC—Photovoltaic Specialists Conference QD—quantum dot QE—quantum efficiency QW—quantum well SAW—solar array wing SCOPA—survivable concentrating photovoltaic array SEC—sun–earth connection SEP—solar electric propulsion

422

SPACE SOLAR CELLS AND APPLICATIONS

Si—silicon SiC—silicon carbide SLA—stretched lens array SOA—state of the art sol—martian day of 24.6 h Sp.—specific SSP—space solar power STP—solar terrestrial probe SUPER—survivable power system TCO—transparent conducting oxide TJ—triple junction TRMM—Tropical Rainfall Measuring Mission U—university UNSW—University of New South Wales USO—united solar ovonic UV—ultraviolet VDC—volts direct current XTE—X-ray timing explorer

REFERENCES [1] [2] [3] [4] [5]

[6] [7]

E. L. Ralph. High efficiency solar cell arrays system trade-offs. Paper presented at 1st World Conference on Photovoltaic Energy Conversion, December 5–9, Waikoloa, Hawaii (1994). J. E. Granata and J. Merrill. Mass and cost comparison of lightweight array and rigid array structures. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). M. Green, K. Emery, Y. Hishikawa, and W. Warta. Solar cell efficiency tables (version 33). Prog. Photovoltaics 17(1), 85–93 (2009). R. Tatavarti, G. Hillier, A. Dzankovic, G. Martin, F. Tuminello, R. Navaratnarajah, G. Du, D. P. Vu, and N. Pan. Lightweight, low cost GaAs solar cells on 4” epitaxial liftoff (ELO) wafers. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). J. Geisz, S. Kurtz, M. Wanlass, J. Ward, A. Duda, D. Friedman, J. Olson, W. McMahon, T. Moriarty, and J. Kiehl. High efficiency GaInP/GaAs/InGaAs triple junction solar cells grown inverted with a metamorphic bottom junction. Appl. Phys. Lett. 91, 023502 (2007). R. King, D. Law, K. Edmondsonh, C. Fetzer, G. Kinsey, H. Yoon, R. Sherif, and N. Karam. 40% efficient metamorphic GaInP/GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90, 183516 (2007). A. Barnett, D. Kirkpatrick, C. Honsberg, D. Moore, M. Wanlass, K. Emery, R. Schwartz, D. Carlson, S. Bowden, D. Aiken, A. Gray, S. Kurtz, L. Kazmerski, T. Moriarty, M. Steiner, J. Gray, T. Davenport, R. Buelow, L. Takacs, N. Shatz, J. Bortz, O. Jani, K. Goossen, F. Kiamilev, A. Doolittle, I. Ferguson, B. Unger, G. Schmidt, E. Christensen, and D. Salzman. Milestones toward 50% efficient solar cell modules. Paper presented at the 22nd European Photovoltaic Solar Energy Conference, September 3–7, Milan, Italy (2007).

REFERENCES [8]

[9]

[10] [11]

[12] [13] [14] [15] [16]

[17] [18] [19] [20] [21]

[22]

[23]

423

S. M. Hubbard, C. G. Bailey, C. D. Cress, S. Polly, J. Clark, D. V. Forbes, R. P. Raffaelle, S. G. Bailey, and D. M. Wilt. Short circuit current enhancement of GaAs solar cells using strain compensated InAs quantum dots. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). M. F. Führer, J. P. Connolly, M. Mazzer, I. M. Ballard, D. C. Johnson, K. W. J. Barnham, A. Bessière, J. S. Roberts, R. Airey, C. Calder, G. Hill, T. N. D. Tibbits, M. Pate, and M. Geen. Hot carriers in strain balanced quantum well solar cells. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). P. R. Sharps, B. Cho, A. Cornfeld, J. Diaz, F. Newman, M. Stan, and T. Varghese. Next generation high-efficiency space solar cells. Space Power Workshop Presentation, April 20–23, Manhattan Beach, CA (2009). R. Contini, E. Ferrando, D. Hazan, R. Romani, R. Campesato, M. C. Casale, and G. Gabetta. Comparison between CIGS and triple junction GaAs on thin Ge solar cell assemblies and related development strategies. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). R. Venkatasubramanian et al. Paper presented at the 26th IEEE PVSC, September 29–October 3, Anaheim, CA (1997). S. Bailey, D. Wilt, J. McNatt, L. Fritzenmeier, S. Hubbard, C. Bailey, and R. Raffaelle. Thin film poly III-V space solar cells. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). A. Luque and A. Marti. Increasing the efficiency of ideal solar cells by photon induced transitions at intermediate levels. Phys. Rev. Lett. 78, 5014–5017 (1997). A. Marti, N. Lopez, E. Antolin, E. Canovas, C. Stanley, C. Farmer, L. Cuadra, and A. Luque. Novel semiconducter solar cell structures: The quantum dot ingtermedial band solar cell. Thin Solid Films 511–512, 638–644 (2006). A. Marti, E. Antolin, C. Stanley, C. Farmer, N. Lopeq, P. Diaz, E. Canovas, P. Linares, and A. Luque. Production of photocurrent due to intermediate to conduction band transitions: A demonstration of a key operation principle of the intermediate band solar cell. Phys. Rev. Lett. 97, 247701 (2006). R. Laghumavarapu, A. Moscho, A. Khoshakhlagh, M. El-Emawy, L. Lester, and D. Huffaker. GaSb/GaAs type II quantum dot solar cells for enhanced intrared spectral response. Appl. Phys. Lett. 90, 173125 (2007). R. Raffaelle, S. Sinharoy, J. Anderson, D. Wilt, and S. Bailey. Multijunction solar cell spectral tuning with quantum dots. Paper presented at the 4th IEEE World Conference on Photovoltaic Energy Conversion, May 7–12, Waikoloa, HI (2006). D. Ferguson. Interactions between spacecraft and their environments. AIAA Paper #93-0705, NASA TM 106115, NASA Glenn Research Center, Cleveland, OH (1993). S. Bailey et al. Solar cell and array technology for future space science missions. Internal NASA Report to Code S (2002). M. Eskenazi, S. White, B. Spence, M. Douglas, M. Glick, A. Pavlick, D. Murphy, M. O’ Neill, A. J. McDanal, and M. Piszczo. Promising results from three NASA SBIR solar array technology development programs. In Proceedings of the Space Photovoltaic Research and Technology Conference, Cleveland, OH, September 16–18, pp. 59–94 (2003). A. Delleur and T. Kerslake. Electrical performance of the International Space Station US photovoltaic array during difacial illumination. In 37th Intersociety Energy Conversion Engineering Conference (IECEC), July 29–31, pp. 39–44, Washington, DC (2002). D. Scheiman, G. Landis, and V. Weizer. High-bandgap solar cells for near-sun missions. In Space Technology and Applications International Forum (STAIF-99),

424

[24] [25] [26] [27]

SPACE SOLAR CELLS AND APPLICATIONS January 30–February 4, Albuquerque, NM. AIP Conference Proceedings 458, 616– 620 (1999). G. Landis and P. Jenkins. Measurement of the settling rate of atmospheric dust on Mars by the MAE instrument on Mars Pathfinder. J. Geophys. Res., 105(E1), 1855– 1857 (January 25, 2000). R. Raffaelle et al. Integrated microelectronic power supply (IMPS). In Proc. 6th International Energy Conversion Engineering Conference, Vol. 1, July 29–August 2, Savannah, GA, pp. 239–242 (2001). P. Stella and R. Kurland. Latest developments in the advanced photovoltaic solar array program. In Proceedings of the 21st Photovoltaic Specialists Conference, May 21–25, Kissimimee, FL, pp. 569–574 (1990). S. Bailey, A. Hepp, and R. Raffaelle. Thin film photovoltaics for space applications. In Proceedings of the 6th International Energy Conversion Engineering Conference, July 29–August 2, Savannah, GA, pp. 235–238 (2001).

PART V OTHER ASPECTS AND CONSIDERATIONS

19 SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS CHRISTIAN A. GUEYMARD1 AND DARYL MYERS2 1 Solar Consulting Services, 2National Renewable Energy Laboratory

19.1

INTRODUCTION

Even if solar technologies, such as PV, were ideally 100% efficient, their absolute energy output would still be limited by the solar power they receive at any instant, or incident irradiance. The quantitative and qualitative aspects of the solar resource on planet Earth (or immediately around it) that are most pertinent to maximize the output and efficiency of PV applications and to determine their most appropriate location are described in this chapter. In the present context, the quantitative aspects will refer to the variations in total irradiance that is incident on a solar cell or PV collector, whereas the qualitative aspects will refer to its spectral and angular characteristics.

19.2

SOLAR RESOURCE IN SPACE

The very first application for solar cells was to provide power to satellites in space. PV panels are still the unique source of power for most, if not all, satellites in earth orbit. The principles and data provided here are essential for such applications but are also useful to expand the reach of PV applications to spacecrafts designed for more distant missions. The next section focuses on the quantitative aspects of extraterrestrial irradiance, whereas its qualitative aspects are discussed in Section 19.2.2.

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

427

SOLAR RESOURCE IN SPACE

429

normally outnumber sunspots, so that the TSI is at a relative maximum. The TSI’s periodicity, which is apparent in Figure 19.1, perfectly matches the well-known 11-year periodicity in solar activity (e.g., see Reference 5). From Figure 19.1, the daily value of TSI can be estimated as the combination of two distinct components: the “long-term” TSI with a relatively smooth 11-year periodicity and amplitude of less than 2 W/m2 and a highly variable daily perturbation caused by short-term phenomena, such as sunspots. The latter component is negligible during low-activity periods but can be abruptly high during the peaking phase of a cycle. Particularly noticeable is the record deficit of ≈5 W/m2 that occurred on October 29–30, 2003 when 167 sunspots were recorded [1]. With the composite PMOD dataset covering more than 30 years of TSI measurement, is it possible to evaluate the absolute uncertainty in TSI? Between 1996 and 2003, there was a good agreement—within about ±0.1%—between the various radiometers then in service. Things changed dramatically, however, when a new type of radiometer went to orbit on the SORCE satellite in March 2003. Its data stream2 shows a systematic difference in comparison with the PMOD data (Fig. 19.2). For the period March 2003–June 2009 (2259 common data days), this difference averages 4.43 W/m2. The low standard deviation of this difference (0.04 W/ m2) indicates that the two series of data have high combined precision. Nevertheless, the causes for the important systematic bias are not resolved and are still debated. Considering this state of affair, it is safe to assume that the true TSI value is somewhere between the PMOD and SORCE evaluations. Even though TSI is variable on a daily basis, daily excursions of less than about ±0.3% around the mean TSI are presumably of little importance in engineering. This is why the SC concept is still in use. The latest value proposed for SC is

Total Solar Irradiance (W m-2)

1368

Total Solar Irradiance 2003–2009

1367 1366 1365 1364

PMOD composite PMOD composite 27-day average SORCE 27-day average

1363 1362 1361 1360 2003

2004

2005

2006

2007

2008

2009

2010

Year

Figure 19.2. TSI from the PMOD composite and SORCE measurements, 2003–2009.

2

http://lasp.colorado.edu/sorce/data/tsi_data.htm.

430

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

1366.1 W/m2 [5, 6], based on the PMOD data. Since then, the PMOD dataset has been extended at both ends and has undergone slight corrections, while the sun itself has been through a period of exceptional inactivity and low TSI since 2007. Using the revised PMOD composite and extended dataset between 1976 and June 2009 (totaling 11,245 days with available data), the current determination of SC (defined as either the average of all daily values or the mean between the minimum and maximum values of the 27-day sliding average) would be 1365.9 W/m2, that is, a decrease of 0.2 W/m2 from the previous SC value. An even lower value, 1361.5 W/ m2, would result from rather using the absolute calibration of the SORCE data. All the irradiance values indicated above refer to the average sun–earth distance, R0 = 1.496 × 1011 m, or ≈1 AU. Due to the eccentric orbit of the earth around the sun, this distance varies by up to ±1.67% during the year. This induces a maximum variation of ±3.34% in TSI due to the inverse square law of radiation propagation. Approximate daily values of this sun–earth distance correction factor, S, can be found in textbooks (e.g., see Reference 7). More precise values can be obtained from S = [1.00014 − 0.01671 cos g − 0.00014 cos ( 2 g )]

−2

(19.1)

where g is the mean anomaly in radians, calculated as a function of D, the number of days elapsed since the current astronomical epoch (January 1, 2000 noon terrestrial time, internationally referred to as J2000.0), such that g = 6.24006 + 0.0172019699 D.

(19.2)

The instantaneous solar irradiance incident on other celestial objects (planets, satellites, etc.) can be obtained by multiplying the TSI or SC values described above by the factor (R0/R)2, where R represents the actual distance between the object and the sun. Two concluding remarks need be made about the angular characteristics of solar irradiance. First, TSI and SC are defined for a receiver (instrument or collector) continuously tracking the sun to maximize the collected energy by keeping solar rays at normal incidence. If the angle of incidence on the receiver is not 0 but θ, the collected irradiance decreases by a factor cos θ, from Lambert’s law. Finally, TSI and SC refer to the spatially averaged radiative output over the solar disk. The effect known as “limb darkening” makes the sun’s periphery significantly darker than its center. This is of practical importance only if a fraction of the solar disk is observed with, for example, a telescope, which is normally not the case in energy production.

19.2.2

Extraterrestrial Spectrum

The TSI, En0, discussed in the previous section represents the integration of the solar spectrum over all wavelengths:

SOLAR RESOURCE IN SPACE

431 ∞

En 0 = ∫0 Enλ d λ

(19.3)

where Enλ is the spectral irradiance at wavelength λ. The subscript n denotes that all irradiances are defined here at normal incidence. The same correction factors S and cos θ that apply to En0 also apply to Enλ. The limb-darkening function also applies, but it varies strongly with wavelength [8]. The disk-average spectral distribution of Enλ at the average sun–earth distance and normal incidence is discussed in what follows. This quantity is often referred to as “air mass zero” (AM0) irradiance. A first approximation of the spectral distribution of Enλ can be obtained from Planck’s law, with the simplifying assumption that the sun is a blackbody radiator. Its radiative temperature, T, can then be obtained by the Stefan–Boltzmann law, E = σT 4. Using the latest determination3 of σ = 5.670 400 × 10−8 W/m2/K4 and the volumetric mean solar radius, Rs = 6.96 × 108 m, the radiant output of the sun is obtained as E = SC (R0/Rs)2; hence, T = 5775.9 K for SC = 1366 W/m2. In reality, the sun is not a perfect blackbody radiator, so that its real spectrum must be determined by either modeling or measurement. A review of such historical determinations of the spectrum can be found elsewhere [1, 5]. An accurate experimental determination of Enλ is more difficult than that of Eñ0 because spectroradiometers are more complex than broadband radiometers and cannot be calibrated with as much absolute accuracy at all wavelengths. Moreover, a single detector cannot cover the whole range of wavelengths in the solar spectrum, so that many instruments are needed in parallel, which complicates the process of merging their data into a single composite spectrum. Finally, extra care is needed at short wavelengths: at the earth’s surface, the atmosphere is completely opaque below about 300 nm, so that spaceborne instruments need to be used to monitor the solar spectrum below this wavelength, while exposure to these wavelengths in space leads to instrument degradation, which needs to be corrected a posteriori. Most useful spectra are in practice obtained by combining partial spectra of various sources and by scaling them appropriately so that they integrate to the exact value of SC per Equation 19.3, as described elsewhere [5, 6]. These spectra have moderate resolution, which is normally sufficient for engineering applications. For other applications, higher resolutions are sometimes necessary in the ultraviolet because of the intense structure in that part of the spectrum. Good results have been obtained [10] by normalizing a high-resolution spectrum with the lowerresolution Gueymard spectrum.4 The latter is shown in Figure 19.3 in both irradiance units (watt per square meter per nanometer, left axis) and photon flux units

3

All physical constants used in this chapter are from CODATA 2006 [9], available from http://physics.nist.gov/cuu/Constants/papers.html. 4 This spectrum is available at http://www.solarconsultingservices.com/Gueymard_ET_ spectrum.txt.

432

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Figure 19.3. Planck radiation distribution and Gueymard’s extraterrestrial spectrum in irradiance units (left axis) and photon flux units (right axis).

(square centimeter per second per nanometer, right axis) for the 300- to 2500-nm spectral band, which is of most interest in PV applications. At any wavelength, the photon flux Fλ is related to Enλ, through Fλ = k Enλ λ ( h c ) ,

(19.4)

where h is the Planck constant (6.62606896 × 10−34 J s), c is the speed of light in vacuum (2.99792458 × 108 m s−1), and k is a numerical constant used to reconcile units. It equals 1 × 10−13 for λ in nanometer, Enλ in watt per square meter per nanometer and Fλ in square centimeter per second per nanometer. The Planck irradiance distribution at 5776 K is also indicated in Figure 19.3 for reference purposes. Its peak (at 502 nm) lies between those of the irradiance spectrum (451 nm) and of the photon flux (584 nm). There are differences in spectral irradiance between a quiet sun and an active sun (e.g., see Reference 1). However, most of these differences occur well below 400 m, where the sensitivity of solar cells is usually low. Therefore, despite the frequent variations in TSI noticed in Figures 19.1 and 19.2, a reference solar spectrum may be used without daily correction—except for the changing distance, per Equation 19.1.

ATMOSPHERIC EFFECTS

19.3

433

ATMOSPHERIC EFFECTS

The Earth’s atmosphere is a continuously variable filter. Broadly speaking, the changes in sun–collector geometry, weather, and local composition of the atmosphere constantly modify the spectral distribution and total magnitude of the incident solar radiation. 19.3.1

Basic Solar Geometry

The combination of the ≈23.45° tilt of the earth’s axis of rotation with respect to the plane of the earth’s orbit, in conjunction with the orbital period (≈365.25 days) of revolution around the sun, and the ≈24-hour planetary rotation period determines the varying solar position with respect to any location on the earth at a specified time. The local day length changes daily (except at the equator), and this can be predicted using standard astronomical data or simplified calculations. In order to calculate the position of the sun on the sky dome with respect to the local horizontal plane, the latitude and longitude of the place, the local standard time, solar declination and equation of time for the date are required (e.g., see Reference 7). Given these quantities, one can calculate the altitude of the sun above the horizon (solar elevation), or its complement, the angle between the zenith and the sun’s position in the sky (zenith angle) for a given time and place. The zenith angle, Z, will be used throughout this chapter. 19.3.2

Broadband Direct, Diffuse, and Global Radiation

The slender cone of radiation from the sun intercepted by the earth is called the direct beam, since it is well collimated. As the direct beam radiation passes through the earth’s atmosphere, two major extinction processes may affect the photons— absorption and scattering. Photons scattered by molecules and particles in the atmosphere that are not further absorbed contribute to uncollimated diffuse irradiance of the sky dome. Some radiation is reflected back into space by the atmosphere, clouds, or ground. Figure 19.4 shows these processes schematically. On a monthly average basis, very few areas on this planet experience a favorable solar resource representing more than 75% of the SC. In similarity with the concepts presented in Section 19.2, the collimated irradiance incident on a plane perpendicular to the sun’s direction is called DNI, denoted Ebn. DNI’s projection onto the horizontal plane is the direct horizontal irradiance, Eb, such that Eb = Ebn cos Z ,

(19.5)

where Z is also the angle of incidence of the sun rays (assumed to originate from the sun’s center) onto the horizontal plane. The sum of Eb and diffuse sky

ATMOSPHERIC EFFECTS

435

reflected from the ground in its field of view or specularly reflected by mirrors intentionally installed off its sides. This reflected irradiance is denoted Er. The total irradiance incident on the receiver, Es, becomes Es = Ebn cos θ + Rd Ed + Er ,

(19.8)

where Rd is the tilt factor for sky diffuse radiation. Whenever Es is not measured, Equation 19.8 must be used to evaluate it. Modeling is necessary for this task either to evaluate Es globally or each of the components in the rhs of Equation 19.8 separately: Ebn, Ed, Er, and Rd. Such models have been regularly proposed in the literature, and some of them are described in various textbooks (e.g., see References 7 and 11). If the sky radiance were ideally uniform or “isotropic,” Rd would be simply obtained from a pure geometrical expression: Rd = (1 + cos s ) 2 .

(19.9)

Many studies have shown that this simplification significantly underestimates the incident diffuse irradiance for tilted collectors facing the equator (i.e., facing south in the northern hemisphere or north in the southern hemisphere). Similar to Equation 19.9, it is frequently assumed that the ground is perfectly horizontal with an ideally isotropic reflectance, and with no shading, in which case one obtains Er = ρ E (1 − cos s ) 2 ,

(19.10)

where ρ is the ground reflectance, or “albedo,” whose value varies in practice between ≈0.1 for dark surfaces and ≈0.85 for fresh snow. Natural surfaces are generally not perfectly isotropic diffusers, so that their reflectance is enhanced in the direction of beam propagation, and therefore tends to increase with Z [12]. The addition of side mirrors or booster reflectors to a planar collector makes calculations even more intricate. This has been the topic of many investigations, but most of them either only address some specific reflector geometry or use intensive computer modeling such as ray tracing with no analytical solution. A general, simple, and accepted methodology that would encompass all possible reflector/collector geometries and various reflecting materials has yet to be proposed. The case of bifacial collectors, where reflectors redirect a fraction of the incident energy to the underside of a planar array, has also received some attention (e.g., see References 13 and 14). To optimize the performance of a PV module equipped with booster reflectors of any kind, it is important that the resulting enhanced irradiance is as homogeneous as possible over the area of the module. Since this is difficult to achieve all the time with any fixed module of large area, some compromise between module area and reflector enhancement is necessary. Only a few solutions have been proposed so far (e.g., see Reference 15).

436

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

The ideal and highly simplified cases described by Equations 19.9 and 19.10 almost never correspond to reality. This therefore introduces significant errors in the calculated POA irradiance, Es. Recent investigations have extensively discussed these sources of uncertainty using various models and methodologies and have found that current models had significant shortcomings under nonideal conditions [16]. It is therefore always preferable to measure Es with first-class instrumentation whenever the precise evaluation of the solar resource or the accurate monitoring of a solar installation is necessary. The relative magnitudes of the direct beam and sky diffuse radiation depend on the absorption and scattering properties of the atmosphere, which are discussed in what follows. 19.3.3

Absorption

The gases that constitute the earth’s atmosphere absorb solar radiation in specific spectral regions. These absorbers modify the solar spectral distribution of energy at the ground depending on their amount and path length of the beam. A convenient measure of this path length is the optical air mass (AM), generally denoted m, which reduces to ≈1/cos Z for Z less than about 75°. Thus, for a zenith sun, Z = 0° and AM = 1. For a rising or setting sun, Z = 90° and AM ≈ 38. Absorption processes remove energy from the incident irradiance and convert it to thermal energy in the atmosphere. Absorption is a strong function of air mass and of gas concentration. Most absorbing gases have a concentration that does not vary over time, and therefore their absorption can be easily predicted. Other gases are variable. It is well known that the concentrations of carbon dioxide and methane increase steadily due to human activity, but their effect in the shortwave spectrum is small. On a daily basis, the main variable gases are water vapor and ozone. The former has a significant effect on the spectrum above 700 nm, with strong absorption in the 920- to 980-nm, 1100- to 1200-nm, 1300- to 1500-nm, and 1750- to 1950-nm wave bands in particular. Ozone has a slight effect in the 550- to 750-nm wave band and has a strong effect below 350 nm. The ground-level ultraviolet irradiance is thus weak, although it can significantly affect the degradation of materials in particular. The amount of water vapor is highly variable over time and space. This amount is usually characterized in terms of precipitable water, w, which is defined as the height of water (usually expressed in centimeter) in a virtual column that would be obtained by condensing all the water vapor aloft. Typically, w varies between about 0.2 cm under cold, dry conditions and about 6 cm under hot, humid conditions. 19.3.4

Scattering

Scattering of radiation in the atmosphere is the redirection of photons by gas molecules (Rayleigh scattering) or particulates of various sizes (Mie scattering). With

ATMOSPHERIC EFFECTS

437

Rayleigh scattering, shorter wavelengths are scattered more strongly than longer wavelengths, explaining why the sky is blue. Besides gases, the atmosphere also contains larger particles, such as dust, sand, sea salt, smoke, or soot particles, which also scatter photons. The scattering process for these particles, referred to as “aerosols” because they are suspended in the atmosphere, is intricate. They also absorb radiation in small but variable proportions. The bulk measure of the total aerosol extinction effects due to both scattering and absorption is called AOD. The empirical Ångström’s law stipulates that, at any wavelength (between at least about 350 and 1000 nm), the total AOD by aerosols is generally proportional to a constant (β, the optical depth at 1 μm, also known as the Ångström turbidity coefficient) times the wavelength (expressed here in micrometer) raised to the inverse power of α, known as the wavelength exponent or the Ångström coefficient, such that τ aλ = β λ − α .

(19.11)

The value of α is an inverse function of the average size of the aerosols. Most aerosols are characterized by α between about 0.5 and 2.5. Volcanic and maritime aerosols tend to have a large size and therefore exhibit low α values, whereas the reverse is generally true for urban aerosols. Continental or rural aerosols stand in the middle, with α generally between 1 and 1.5. Accurate prediction of cloudless direct and diffuse irradiance cannot be done without a precise determination of AOD at all wavelengths, or at least of α and β. The main difficulty with aerosols is that both their type and total amount, and hence α and β, are highly variable over time and space. The actual AOD can be monitored with the help of ground-based multiwavelength sun photometers or of various spaceborne sensors. Sophisticated chemical transport models can also predict the worldwide distribution of AOD. Ground-based data are still by far the most accurate of these sources and should be preferred wherever possible. Under cloudless conditions, the amount of radiation removed from the direct beam and transformed in great part into diffuse sky radiation can be predicted accurately with appropriate radiative transfer models if AOD and some other atmospheric variables are known. The comparative performance of various broadband irradiance models has been investigated recently [17, 18] and shows the importance of using accurate AOD data. In addition to substantially affecting the magnitude of the direct and diffuse irradiance, AOD is also important with respect to changes in their spectral distribution (see Section 19.4.5). The two other most important variables that determine the magnitude of direct and diffuse irradiance are m and w. The relative effects of m, w, and β on DNI for a rural aerosol with α = 1.3 are shown in Figure 19.5. The thin vertical line indicates the turbidity conditions (β = 0.033 or τaλ = 0.084 at 500 nm) that have been chosen to develop the reference direct spectrum in standard G173 [19], which is further discussed in Section 19.4.6. The main other variables used for this standard spectrum are m = 1.5, w = 1.42 cm, and α = 1.2. The resulting DNI is Ebn ≈ 900 W/m2, which is indicated by a large gray dot on the vertical line. Figure

438

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS 1100

SMARTS Model Direct Normal Irradiance Sea level Rural aerosol, α = 1.3

1000 900

Precipitable water 0.5 cm 1.0 cm 2.0 cm 5.0 cm

Irradiance (W m–2)

800 700

m=1

600 500

m = 1.5

400 300

m=2 m=5

200 100

0

0.1

0.2 0.3 0.4 Ångström's Turbidity Coefficient β

0.5

Figure 19.5. Direct normal irradiance as a function of Ångström turbidity coefficient β, air mass m, and precipitable water w for sea-level conditions and rural aerosol. The vertical line indicates the turbidity conditions of ASTM G173. The large gray dot indicates the G173 DNI for m = 1.5 and w = 1.42 cm, that is, 900 W/m.

19.5 can be used to visually evaluate the change in DNI that would result from incremental changes in m, β, or w away from standard conditions. The aerosol amount in G173 is typical of clean sites at sea level. Extremely clean sites at high altitude often experience a 50% reduction in aerosol load (β ≈ 0.015). To the other extreme, the atmospheric turbidity may reach or may even exceed β = 0.5 under hazy conditions due to forest fires, desert dust storms, or heavy pollution. Section 19.4.5 further describes some useful sources of data for the main atmospheric variables to consider in irradiance predictions. 19.3.5

Cloud Attenuation

The solar resource is normally maximum under clear-sky conditions, but cloudy conditions are frequent at most locations. Clouds are constant “moving targets,” and studying their impact on the extinction of radiation is difficult. By far, the predominant attenuation process in clouds is scattering. Only very dense clouds absorb in a significant way. Attenuation effects depend on the type, microphysical

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

439

structure, thickness, and spatial distribution of clouds. Detailed microphysical properties, such as total liquid water path, water droplet size distributions, relative proportion of ice and water, and altitude and thickness of layers, have been found necessary to study cloud effects on solar radiation in detail [20]. At the other extreme, simple empirical results can be used to derive cloud attenuation factors for direct or global irradiance as a function of cloud fraction only (e.g., see Reference 21). The bulk transmittance of a cloud can also be modeled from its optical depth. These two essential quantities (cloud fraction and cloud optical depth) are now retrieved from various satellite sensors and are available on a worldwide basis in the form of gridded data, at more or less frequent time intervals. High thin clouds have low optical depths (typically 1–5), whereas dark dense low clouds have much larger optical depths (typically 50–90). Beam radiation is not transmitted by thick clouds, and only a fraction of it is transmitted by optically thin clouds. Clouds are near-neutral filters, so that their spectral transmittance is nearly flat in the visible.

19.4

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

The design and performance evaluation of any solar energy conversion system relies on information about the available solar resource. Measured data for a particular site are usually more accurate than model predictions. As described in Section 19.3.2, the basic radiation components at the ground are global, direct beam, and diffuse sky radiation. A PV system collects some combination of each of these components, possibly augmented by reflected radiation from the ground or booster reflectors. Solar radiation data and models for terrestrial applications are discussed below.

19.4.1

Broadband Global, Direct, and Diffuse Irradiance

Most commonly, radiation data are obtained from horizontal pyranometers, which sense global radiation. A pyrheliometer measures DNI with the help of a collimator to block most of the sky from its sensor (normal to the beam). The instrument’s field of view (about 6°) accepts some sky radiation around the ≈0.5° disk of the sun, termed circumsolar radiation (Section 19.4.8). DNI is less frequently measured than global radiation because of the need for a costly sun tracker. Diffuse radiation can be measured with a pyranometer using either a shadow band or a shading disk to mask the sun. At many sites, “direct irradiance” data are actually indirectly calculated using simultaneous measurements from a pair of pyranometers (one measuring global radiation and the other diffuse radiation) or from a rotating shadow band radiometer where global and diffuse radiation are measured in rapid succession with a single silicon cell sensor [22]. Such derived direct beam measurements are usually less accurate than pyrheliometer measurements [23, 24].

440

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

More generally, the best current practice in broadband irradiance measurement and related accuracy issues have been reviewed recently, showing that the basic accuracy of measured solar data is often not as good as one would expect [23, 25].

19.4.2

U.S. NSRDB

The NREL has established an NSRDB. The initial version of the database includes 30 years of hourly solar radiation and meteorological data for 239 sites in the United States [26]. In 2007, an update to the database was generated to include 1430 sites (including most of the original 239 sites) with hourly solar radiation and meteorological date for the period from 1991 to 2005 [24]. Both the original and updated NSRDB datasets contain more than 90% modeled solar radiation estimates from meteorological parameters, resulting in lower accuracy. The original NSRDB calculations were produced using the METSTAT model [27]. The 1991–2005 update required modifications to the original METSTAT model because of changes in the availability of certain input parameters. The update also includes hourly solar radiation estimates on a ≈10-km grid, derived from geostationary satellite data [28] for the period 1998–2005. Total global horizontal, direct beam, and diffuse irradiances for a particular cell in this latter dataset can be accessed interactively.5 Typical uncertainties for measured and modeled data in these databases is 10% for global horizontal irradiance and 15% for DNI but depends more specifically on the quality of input data and availability of some ancillary cloud cover data [24]. The objectives of the NSRDB were to provide consistent serially complete hourly time series solar data able to produce reasonably accurate (on the order of 5%) longer-term (e.g., monthly means) averaged data. Each individual hour of data could be wildly different from an actual measurement at that specific time because actual positions and geometry of clouds could not be accounted for. Statistical algorithms were developed and validated [27] to produce the correct statistical properties of frequency distributions for the long-term data. Data manuals containing computed irradiances, derived from the NSRDB basic radiation data, relative to various collector configurations such as latitude tilt, full tracking, and east–west tracking, were also prepared [29] (see http://rredc.nrel. gov/solar/ and the Appendix).

19.4.3

Solar Radiation Data for International Locations

Some national and international organizations collect and distribute radiation data from various networks for scientific research purposes. Examples include WRDC 5

http://www.nrel.gov/gis/solar.html.

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

441

[30], BSRN [31], ARM [32], and NOAA SURFRAD [33]. The data available from these organizations’ websites have varying resolution ranging from 1 min (ARM and SURFRAD) to daily totals and monthly totals (WRDC). Most nations have operational solar monitoring networks but have no public access to the data, although it may sometimes be purchased. The Appendix provides more details about available sources of measured and modeled data.

19.4.4 Irradiance for Fixed-Tilt, Tracking, and Concentrating Geometries When locally measured solar radiation data cannot be obtained, one falls back on solar radiation models to generate global, direct, and diffuse irradiance. Generally, one starts with clear-sky models that evaluate the radiation components from atmospheric input data, per the discussions in Sections 19.3.3 and 19.3.4. These clear-sky irradiances are then modified to take into account the effect of clouds (Section 19.3.5) to produce “all-sky” irradiance values. Various aspects of these methodologies are discussed in Reference 11, for instance. If only global horizontal irradiance is available, and one needs to estimate DNI, one of many global-to-direct conversion models of the literature needs to be used. Their performance is regularly evaluated against measured data (e.g., see References 16 and 34). Although the mean bias error of these estimates is generally reasonable on a long-term basis (depending on local climatic conditions), the hourly or sub-hourly random errors are considerable with all existing models. This important accuracy issue must be considered when using such modeled data for the design or performance simulation of CPV collectors, for instance. For fixed-tilt and tracking FP collectors, several models have been developed to solve Equation 19.8 and to produce estimates of POA irradiance, the most popular of which is the Perez model [35]. For instance, it is used in PV performance models such as PVFORM [36]. The performance of this type of models has been studied extensively, for a wide range of climatic conditions, instrumentation, local specificities, and experimental setups. Not surprisingly, these studies do not always concur in their conclusions. Recent findings [16] suggest that the accuracy of predicting the POA irradiance is seriously limited by a series of factors, including the high sensitivity of these models to uncertainties in direct irradiance, which is one critical input. Moreover, the usual complexity of local conditions (e.g., shading from obstacles and variance in ground reflection processes) is simply not considered in current models. It is therefore impossible to recommend a single model that would perform best under all possible conditions. As in the case for the globalto-direct conversion models, large random errors in the modeled irradiances (and even sometimes, significant bias errors for some POA geometries or locations) can be expected under realistic conditions. As overviewed in the Appendix, a wide diversity of modeled solar radiation products now exists. More products are actually described in the literature, mainly for climate research purposes. Although this diversity is a good thing, it also raises

442

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

the issue of which data source is preferable to obtain the needed information at any potential site in the world. This is currently an unresolved question because all models and their input data have weaknesses. Even though they might be validated in broad terms, it does not mean they are accurate under any condition, at any site, all the time. A few recent studies [37–39] have just scratched the surface of this practical problem and show significant differences from one dataset to the other. Many more in-depth studies will be necessary before it becomes possible to provide specific recommendations to all those who need highly accurate data to design or to simulate large and costly projects. Another issue of significant importance in the practice of designing PV systems is to rapidly compare the resource available for different possible collector geometries. This has received recent attention [40], based on the NSRDB modeled data discussed above. This study intercompared the annual and seasonal (summer and winter) resources for 14 different collector geometries and introduced the SREF method. A few results are outlined below and are presented in the form of questions and answers. Moreover, a numerical example is proposed at the end of this section. 1.

2.

If fixed planar collectors are to be used, what is their best tilt to maximize energy production during the summer, winter, or year? A conventional rule of thumb is that the optimal tilt angle is equal to latitude (L) for maximum annual energy production, L − 15° for maximum summer production and L + 15° for maximum winter production. Detailed calculations show that the L and L − 15° geometries both maximize the annual energy collection, therefore indicating a broad optimum. An equator-facing orientation is assumed throughout but is not too critical either [41]. At least for all the United States except Alaska, the L − 15° geometry is indeed ideal for summer, but only marginally better than a horizontal setup. For winter, the L + 15° geometry is the most favorable, but only 1–5% better than the latitude tilt setup. All these results, of course, are for ideal conditions and therefore do not consider the incidence of local microclimatic factors, partial shading, risks of snow accumulation on the collector, structural costs, aesthetics, and so on. Knowing that systems involving tracking devices cost more than a fixed design, how much more resource is available to a two-axis tracking FP than to a latitude tilt fixed collector? A two-axis tracking planar collector always benefits from the largest possible resource per collector unit area. (It is larger than what a two-axis concentrator receives at any time and location since the planar collector makes additional use of the diffuse irradiance from the sky and ground.) The annual energy enhancement of the two-axis planar geometry compared to the fixed-tilt geometry is from 25% (in cloudy climates) to 42% (in sunny climates) annually (Fig. 19.6). This range changes to 15–35% for winter operation or to 25–50% for summer operation.

444

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Example Evaluate the relative merits of several possible PV collector geometries to maximize annual collection at San Ramon, California (latitude L = 37.78°N, longitude 121.97°W), using the SREF method. Two possible sources of public domain long-term data are possible: the 1961–1990 NSRDB and the 1991–2005 NSRDB Updates. The closest location in the former dataset is Sacramento, 91-km NE. The closest locations in the latter dataset are Hayward, Oakland, and Livermore. The two first sites are directly on the Bay, whereas the latter site is in the same climatic environment as San Ramon and only 16-km SE, and is therefore preferable for this analysis. The monthly mean solar data for Sacramento are extracted from http://rredc.nrel.gov/solar/old_data/ nsrdb/1961-1990/dsf/data/23232.txt. Similarly, the data for Livermore are extracted from http://rredc.nrel.gov/solar/old_data/nsrdb/1991-2005/statistics/dsf/724927. dsf. The headers of these files are reproduced below. ID

City

State

Latitude (N)

Longitude (W)

23232

Sacramento

CA

38.52

121.50

724927

Livermore municipal

CA

37.70

121.82

Elevation (m)

Time Zone

Pressure (mb)

8

−8

1015

121

−8

999

The next step is to extract from these files the annual average daily mean values of global (H), direct normal (Hbn), diffuse (Hd), extraterrestrial on the horizontal (H0), and extraterrestrial normal (Hn0) irradiations, and report their values (in watt hour per square meter) in the next table. Sacramento, 1961–1990 H 4933

Livermore, 1991–2005

Hbn

Hd

H0

H n0

H

Hbn

Hd

5505

1556

8143

16405

4712

4892

1684

H0

Hn0

8142 16538

Using this table, the annual values of K = Hd/H, Kt = H/H0, and Kn = Hbn/Hn0 are calculated. The value of cos L is also needed to calculate the annual SREF factors (which provide the average irradiation on a collector relative to that incident on a latitude tilt planar collector or on a two-axis concentrator) using equations that are provided in the original publication [40] but are not repeated here for conciseness. The results are summarized in the following tables. Sacramento, 1961–1990 K 0.3154

Livermore, 1991–2005

Kt

Kn

cos L

K

Kt

Kn

cos L

0.3356

0.6058

0.7824

0.3574

0.2958

0.5787

0.7912

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

Geometry

445

SREF Sacramento

SREF Livermore

Fixed FP tilt = 0°

0.874

0.887

Fixed FP tilt = L−15°

0.985

0.987

Fixed FP tilt = L (reference)

1.000

1.000

Fixed FP tilt = L + 15°

0.960

0.959

Fixed FP tilt = 90°

0.643

0.632

Tracking FP one-axis N-S tilt = 0°

1.225

1.206

Tracking FP one-axis N-S tilt = L − 15°

1.315

1.290

Tracking FP one-axis N-S tilt = L

1.325

1.300

Tracking FP one-axis N-S tilt = L + 15°

1.293

1.267

Tracking FP two-axis

1.369

1.342

Tracking conc one-axis E-W tilt = 0°

0.656

0.624

Tracking conc one-axis N-S tilt = 0°

0.892

0.837

Tracking conc one-axis N-S tilt = L

0.974

0.914

Tracking conc two-axis

1.015

0.952

FP, flat-plate; conc, concentrator.

These results indicate that a tracking two-axis FP collector located in San Ramon would experience 34–37% more resource than the reference latitude tilt collector (indicated by italics in the above table). A two-axis concentrator (without consideration of the concentration factor, since only the outer collecting area of the receiver’s normal plane is considered here) would be nearly on par with the latter. The difference in SREF obtained at Sacramento and Livermore is not significant for FP collectors. For CPV collectors, however, the inland site (Sacramento) appears a “better” environment than Livermore. Results for San Ramon should be very close. To obtain absolute energy values, rather than SREF numbers relative to the latitude tilt FP, the annual irradiation incident on this reference collector is needed. Such data can be obtained from the 1961–1990 NSRDB6 or from an interactive 10-km resolution map7 for the period 1998–2005. For Sacramento, the first source provides a mean daily irradiation of 5.5 kWh/m2. For Livermore and San Ramon, the second source provides 5.542 and 5.483 kWh/m2, respectively. Using the Livermore data in the above table, for instance, the estimated annual resource for two-axis collectors would be 5.276 kWh/m2 for a concentrating geometry (irrespective of the concentration ratio) and 7.437 kWh/m2 for an FP geometry. Note the

6 7

http://rredc.nrel.gov/solar/old_data/nsrdb/1961-1990/redbook/sum2/state.html. http://mapserve2.nrel.gov/website/Resource_Atlas/viewer.htm.

446

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

lower concentrator resource in the former case reflects the inability to utilize diffuse radiation. Furthermore, whenever a CPV system is an option, the concentration ratio, engineering constraints, and associated risks—due, for instance, to the large interannual variability in DNI [42]—and costs significantly affect the selection of the most appropriate system.

19.4.5

Spectral Irradiance: Measurements and Modeling

Knowledge of the spectral distribution of the incident radiation on solar cells is important since their sensitivity is highly variable with wavelength. Spectral information is needed to optimize the deployment of PV technologies in different ways. For instance,





Multijunction cells can be tweaked to maximize their output under some specific spectral distributions rather than under a single reference spectrum. As a typical example, the effect of varying the incident spectrum on the performance of a tandem cell arrangement has been shown to be significant, therefore prompting adjustments to the top cell thickness so that the cell efficiency is maximized under each possible spectrum [43]. The actual performance of any solar cell under specific spectral irradiance conditions can be related to that under a reference spectrum, which is used for rating purposes. More generally, this leads to the notion of spectral mismatch, which is further discussed in Section 19.4.7.

Routine outdoor spectral measurements, using spectroradiometers, are conducted at only a few sites in the world, including NREL.8 Such measurements require sophisticated instrumentation, skilled personnel to perform regular calibrations, and delicate quality control. This explains their paucity and why most potential users prefer using modeled data. There exists a wide variety of radiative transfer models for atmospheric applications, as reviewed elsewhere [44]. However, they are complex to use and do not provide the spectral irradiance incident on tilted planes or the circumsolar contribution to direct irradiance (see Section 19.4.8). For solar engineering applications, the choice is therefore limited [45]. SMARTS has been used extensively in PV applications under cloudless conditions. Its algorithms are described thoroughly [44] and have been extensively validated against spectroradiometric measurements and advanced radiative calculations of direct and global irradiance on various tilts [46]. A comparison between predicted direct irradiance spectra and their measured counterpart at NREL (measurement uncertainty ±4% over most of the spectral range) appears in Figure 19.7 for two different air masses. The predicted spectra were smoothed here using a Gaussian filter tool in the SMARTS post-processor to 8

http://www.nrel.gov/midc/srrl_bms/.

448

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Figure 19.8. Comparison between the old (thick lines) and new (thin lines) reference AM1.5D and AM1.5G reference spectra.

systems, whereas the direct spectrum (“AM1.5D”) is used for CPV applications. These measurement principles are defined in standard IEC 60904 for planar systems, for instance. The rapid development of high-efficiency multijunction cells and optical concentrator devices since the late 1990s led to the realization that the earlier direct reference spectrum (ASTM E891) was not representative of the very clear/sunny conditions where such technologies would be deployed. A temporary “low-AOD” spectrum, obtained with SMARTS version 2.8, was then circulated [48]. Based on the findings in [47], the older reference AM1.5 spectra were eventually deprecated and replaced by those in a new standard, G173 [19]. The AM1.5D reference spectrum that it contains is based on SMARTS version 2.9.2, and is close to the “lowAOD” temporary spectrum. The new AM1.5G spectrum is not too different from the older one, besides an order of magnitude better spectral resolution. A comparison of the old and new direct reference spectra is shown in Figure 19.8. More recently, similar reference spectra, but for irradiance incident on vertical and 20°-tilt planes, have been standardized [49]. They are designed to provide spectral information that can be useful in BIPV applications, for instance.

19.4.7

Spectral Mismatch Calculations

The performance of solar cells is evaluated using a set of standard conditions, which specifies a reference spectrum. For example, the 2008 revision of IEC

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

449

60904-3 now specifies an AM1.5G spectrum that is equivalent to the G173 AM1.5G spectrum multiplied by 0.9971. This produces an integrated irradiance of exactly 1000 W/m2 for wavelengths between 0 and infinity. When the output of a new cell is measured by comparison to that of a reference cell in the laboratory (using a solar simulator) or outdoors (using the sun), there are intricate spectral effects to take into account due to the differing spectral responses involved and the difference between the actual test spectrum and the reference STC spectrum. These effects can be condensed into a spectral mismatch factor, whose calculation procedure is defined by standards (ASTM E973 and IEC 60904-7). The same procedure can be generalized, for instance, to evaluate the impact of the pure spectral effect on the performance of PV modules of different materials and spectral responses when subjected to non-STC spectral conditions at a specific location. Since the effective incident spectrum is rapidly changing over the day, it is convenient to use a spectral radiation model such as SMARTS to generate “typical” spectra at regular intervals (e.g., hourly) during an average day at a specific location for which atmospheric data are available. Hence, a DSEF can be defined as a generalized and time-averaged spectral mismatch factor [50], with no dependence on a reference cell in this case:

∫280 Sλ d λ ⎡⎣ ∫sr Ee (t ) Eeλ (t ) dt ∫sr Ee (t ) dt ⎤⎦ 4000

DSEF =

ss

ss

4000

∫280 Erλ Sλ d λ

(19.12)

where Erλ is the reference STC spectrum; Eeλ(t) is the effective incident spectral irradiance at time t; Ee(t) is its broadband (spectrally integrated) counterpart; Sλ is the cell’s spectral response; and sr and ss are the times of sunrise and sunset, respectively (assuming no horizon shading). The weighting scheme in the numerator ensures that the large spectral effects near sunrise and sunset are not overrated. The output of a PV module will appear to be enhanced (i.e., DSEF > 1) relative to its STC rating if the PV material’s spectral response is consistently closer to the average incident spectrum than to the STC reference spectrum. Typically for instance, DSEF is lower than 1 for crystalline Si and higher than 1 for amorphous Si due to their widely different spectral response. For CPV applications, AOD is found to have the largest effect on DSEF [50].

19.4.8

Angular Characteristics and Circumsolar Effect

For non-concentrating applications, the sun can be considered as a point source. Things are different for focusing solar systems with high-concentration ratios. The limb-darkening effect described in Section 19.2.2, augmented by various optical aberrations caused by reflecting surfaces, creates an image of the sun disk with a radial Gaussian distribution of its energy flux [51]. For all CPV systems (imaging or nonimaging), a related issue is the characterization of the large diffuse radiance

450

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

in the sun’s aureole, referred to as circumsolar radiation. Although it is diffuse in nature, it behaves essentially as direct radiation and can therefore be utilized by concentrators. The variation of radiance from the sun center to the edge of the circumsolar region intercepted by a concentrating collector is often referred to as “sunshape.” Earlier studies of the sunshape [51] have been based on experimental measurements of the radiance that were carried out at 11 U.S. locations during 1976–1981 [52]. This dataset has been quality controlled and is now available online.10 More recently, similar data have been obtained with different experimental setups installed at solar furnaces in Switzerland [53] and Germany [54]. Recent studies show to what extent these different results were in agreement, while providing simplified analytical modeling that might be used in concentrating systems (e.g., see Reference 55). The main shortcoming of the proposed empirical sunshape profile functions, however, is that they do not separate the clear from the cloudobscured line-of-sight cases and are not made dependent on the sun position (Z or m), nor on the atmospheric scattering properties (AOD and/or cloud optical depth), which are all key variables to describe the circumsolar effect. It is hoped that the needed refinements will result from the deployment of a recently introduced instrument, SAM [56]. A typical result of SAM measurements appears in Figure 19.9, showing radiance profiles at 670 nm under four different sky conditions, characterized by their optical depth11 between 0.046 and 0.578 (at 670 nm). These measurements12 have been obtained in Burlington, Massachusetts, and Greenbelt, Maryland, and show remarkable differences in radiance patterns depending on the sky condition. Since only sophisticated instrumentation can be used to automatically monitor the actual sunshape, it is unlikely that such measured data will become available on the geographic scale needed for CPV applications, for instance. Various theoretical studies have provided descriptions of the spectral or broadband circumsolar radiance under specific atmospheric conditions of sun position and aerosol type and amount. Some of these studies also analyzed the sky obscuration by thin clouds, demonstrating the considerable increase in circumsolar radiation in such cases. As a general result, whenever the direct irradiance from the sun decreases, due to either increasing air mass or increasing optical depth from aerosols or clouds, the circumsolar effect increases. This additional resource recoups only a fraction of the lost direct irradiance, however. The complexity of such theoretical studies precludes their use for repetitive calculations. A simplified clear-sky aureole model has therefore been derived [57] for applications where both the solar and circumsolar irradiance are needed, and has been integrated into the SMARTS model described in Section 19.4.3. This

10

http://rredc.nrel.gov/solar/old_data/circumsolar/. This is the total optical depth for both aerosol and cloud (if present). 12 These data have been extracted from Visidyne’s public repository, ftp://ftp.visidyne.com/, and have been kindly processed by their scientists at our request. 11

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

451

Figure 19.9. Circumsolar radiance at 670 nm as measured by the SAM instrument at Burlington, Massachusetts, and Greenbelt (NASA’s Goddard Space Flight Center), Maryland. The vertical line indicates the edge of the solar disk.

simplified aureole model has been validated against theoretical results from the literature. It allows the prediction of the circumsolar radiance and irradiance up to 10° from the sun center, when the aerosol type is known or can be inferred from spectral measurements. Assuming that the circumsolar spectral radiance at a distance ξ from the sun center, Nc(ξ, λ), has radial symmetry, the corresponding broadband irradiance, Ec(ξ0), within the acceptance angle ξ0 is calculated numerically by SMARTS from λ2

ξ0

1

s

Ec ( ξ 0 ) = 2π ∫λ d λ ∫ξ N c ( ξ, λ ) sin ξdξ,

(19.13)

where ξs is the angular distance from the sun’s center to its outer edge (0.2665° at mean sun–earth distance), and λ1 and λ2 are the wavelength limits of the shortwave spectrum (e.g., 280 and 4000 nm as used by SMARTS). Typical results are shown in Figure 19.10 for a rural aerosol model (which should be representative of most likely cases in CPV applications), varying air mass (m = 1–3), and typical AOD for very clear (β = 0.025), average (β = 0.15), and hazy (β = 0.45) conditions. It is obvious that, for these conditions, the circumsolar contribution increases almost linearly with the acceptance angle of the receiver.

452

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Figure 19.10. Circumsolar irradiance (in percent of the broadband direct irradiance emanating from the sun’s disk) calculated by the SMARTS model as a function of the receiver’s acceptance angle, ξ0, air mass, m, and aerosol optical depth at 1 μm, β. Each of the three groups of β results includes four curves, in ascending order of air mass from bottom to top. The acceptance angle of the most common pyrheliometer is indicated by a vertical line, showing by how much actual direct irradiance measurements can be overestimated in practice. The large gray dot on this line indicates the circumsolar contribution (about 0.25%) for the conditions of ASTM G173.

For a point-focus concentrator, the relationship between acceptance angle ξ0 and geometrical concentration C can be defined [58] by C = 1 sin 2 ξ 0 ,

(19.14)

so that C = 33.2 for ξ0 = 10°. The results in Figure 19.10 can therefore be considered as typical for focusing systems with concentration ratios above 30, which is the case for CPV power plants. These results can also be used to evaluate how the DNI measured with a regular “normal incidence pyrheliometer” should be corrected13 to obtain the actual irradiance received within the field of view of a specific concentrator for any C above 30. 13 The measured DNI should be corrected downward if ξ0 < 2.9° or upward otherwise with this type of pyrheliometer. This 2.9° limit might be slightly different from other pyrheliometer models.

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

453

In addition to CPV power plants, there is a new segment of the CPV market that makes use of low concentration ratios for residential and commercial applications, with the help of simple low-cost concentrators or booster reflectors. In such cases, C is typically between 1.4 and 5.0, resulting in large acceptance angles of up to 45° (assuming line-focus concentration, for which Eq. 19.14 must be replaced by C = 1/sin ξ0). The curves in Figure 19.10 cannot be extrapolated to 45° because multiple scattering processes significantly modify the sky radiance patterns beyond about 10° of the sun center. Broader measurements or calculations are therefore needed in this case. Various experiments have been conducted to measure the distribution of sky radiance across the whole sky hemisphere with sky scanners of different designs, and some empirical models have been derived from these measurements. Modeling the instantaneous radiance under partly cloudy skies is obviously most challenging, and only limited information is available [59]. So far, no comprehensive intercomparison of all the existing radiance models has been done. This is unfortunate because the limited results shown in Figure 19.11 indicate that the cloudless radiances predicted by some of these models widely disagree in the solar aureole, where accuracy matters most in CPV applications. This is a disturbing finding because cloudless conditions are supposed to be the easiest to model. Consequently,

Figure 19.11. Cloudless-sky radiance distributions predicted by various models for moderately clear summer conditions (Z = 30° and β = 0.15, left) and very clear winter conditions (Z = 60° and β = 0.025, right). These distributions are along the sun’s principal plane, that is, the vertical plane containing the zenith and the sun. The position of the sky element is defined by its zenith angle, counted here positively in the sun quadrant and negatively in the opposite direction. The vertical line indicates the sun’s position. The sky radiance is normalized by the horizontal diffuse irradiance. This relative radiance is also corrected so that its spatial integration over the sky hemisphere equals 1.

454

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

none of the existing models can be recommended for the task of predicting the radiance between 10° and 45° of the sun’s center until a detailed assessment is conducted conclusively. In the mean time, two of the possible models [60, 61] are suggested for approximate evaluations.

19.5

CONCLUDING REMARKS

The information discussed in this chapter provides the basis for what PV system designers need to quantify or qualify the solar resource available for either space or terrestrial applications. High-accuracy data can be obtained for space applications, with typical uncertainties of ±0.3% for broadband irradiance and ±5% for spectral irradiance. For terrestrial applications, the situation is more complex since it depends on the type of measurement or model used, quality of input data for models, spatial or temporal extrapolation methods used, and so on. On-site measurements with first-class thermopile instrumentation, optimal maintenance and quality control, and state-of-the-art calibration and correction methods constitute the ideal case. Uncertainties of about ±3% in DNI, ±5–7% in global irradiance, and ±5–10% in diffuse irradiance are then achievable. At the other extreme, when irradiance needs to be modeled using atmospheric data or needs to be highly extrapolated, much larger bias and random errors can be expected. On a monthly average basis, uncertainties can reach ±10–15% in global irradiance and ±15–20% in DNI, or more at cloudy sites. The latter figures will have to be improved in the future to meet the requirements of so-called bankable data needed for megawattscale solar power plant projects, for which a precise evaluation of the solar resource is essential. Solar radiation modeling is doing rapid progress, while better worldwide input data sources from, for example, spaceborne sensors become regularly available, so that “high-end” data users who need the best accuracy possible should stay regularly informed. Long-term variations in solar radiation for the next 20–50 years, in relation with climate change, should also be considered for realistic power output predictions.

APPENDIX: SOURCES OF MEASURED OR MODELED SOLAR RADIATION DATA A large number of data sources are listed below. The geographic area they cover is indicated by an acronym: United States (USA), Europe (EU), other parts of the world (PW), or worldwide (WW). A dollar sign ($) indicates data for purchase. Measured Data ARM: http://www.arm.gov/data/datastream.php?id=sirs (USA) BSRN: http://www.bsrn.awi.de/en/home (WW)

APPENDIX

455

IDMP: http://idmp.entpe.fr/ (WW) NOAA historical data: http://rredc.nrel.gov/solar/old_data/noaa/ (USA) NOAA/ISIS: http://www.srrb.noaa.gov/isis/index.html (USA) NOAA/SURFRAD: http://www.srrb.noaa.gov/surfrad/ (USA) NREL’s Renewable Resource Data Center (RredC) Solar Data: http://www.nrel.gov/rredc/solar_data.html (USA) http://rredc.nrel.gov/solar/new_data/confrrm/ (USA) http://rredc.nrel.gov/solar/old_data/hbcu/ (USA) http://rredc.nrel.gov/solar/old_data/circumsolar/ (USA) http://rredc.nrel.gov/solar/old_data/spectral/ (USA) http://rredc.nrel.gov/solar/old_data/semrts/ (USA) http://rredc.nrel.gov/solar/old_data/wa/ (USA) http://www.nrel.gov/midc/ (USA) http://rredc.nrel.gov/solar/new_data/Saudi_Arabia/ (PW) SDW: http://solardatawarehouse.com/ (USA) ($) University of Oregon: http://solardat.uoregon.edu (USA) University of Texas: http://www.me.utexas.edu/∼solarlab/tsrdb/tsrdb.html (USA) WRDC: http://wrdc-mgo.nrel.gov/ or http://wrdc.mgo.rssi.ru/ (WW) Mostly Modeled Data and Solar Resource Maps 3Tier: http://firstlook.3tier.com/solar/ (WW) ($) AWS Truepower: http://www.awstruepower.com/solutions/solar/resourceassessment/ (WW) ($) Clean Power Research: http://www.cleanpower.com (USA) ($) DAYMET: http://www.daymet.org/ (USA) DLR/SOLEMI: http://www.solemi.de/home.html (EU, PW) ($) DLR/ISIS: http://www.pa.op.dlr.de/ISIS/ (WW) EnMetSol: http://www.energy-meteorology.de/ (WW) ($) FocusSolar: http://www.focussolarusa.com/ (WW) ($) GeoModel: http://geomodel.eu/ (WW) ($) Green Power Labs: http://greenpowerlabs.com/solar_GIS.html (PW) ($) Helioclim/ESRA: http://www.helioclim.net/esra/ (EU, PW) ($) IrSOLaV: http://www.irsolav.com/ (WW) ($) MeteoControl: http://www.meteocontrol.de/ (WW) ($) Meteonorm: http://www.meteonorm.com/pages/en/meteonorm.php (WW) ($) NASA/POWER: http://power.larc.nasa.gov/ (WW) NASA/SSE: http://eosweb.larc.nasa.gov/sse or http://power.larc.nasa.gov/ (WW) NREL’s Solar Spectral Distributions: http://rredc.nrel.gov/solar/spectra/ NSRDB: http://rredc.nrel.gov/solar/old_data/nsrdb (USA) PVGIS: http://re.jrc.ec.europa.eu/pvgis/index.htm (EU) (PW) Satel-Light: http://www.satellight.com/indexgS.htm (EU) SoDa: http://www.soda-is.org/eng/index.html (WW) ($) SolarGIS: http://solargis.info/ (WW) ($) SWERA: http://swera.unep.net/ (PW)

456

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Typical Meteorological Years (TMY) These are datasets consisting of hourly data over 365 days, normally including solar radiation, temperature, humidity, wind speed and direction, cloud cover, as well as additional pertinent information for the energy simulation of solar systems or complete buildings. Various methodologies have been proposed to develop such synthetic years, including the TMY2 and TMY3 methods used at NREL. http://rredc.nrel.gov/solar/old_data/nsrdb/tmy2/ (USA) http://rredc.nrel.gov/solar/old_data/nsrdb/1991-2005/tmy3/ (USA) http://apps1.eere.energy.gov/buildings/energyplus/cfm/weather_data.cfm (WW).

ABBREVIATIONS AM—air mass AOD—aerosol optical depth ARM—Atmospheric Radiation Measurement Program ASTM—American Society for Testing and Materials BIPV—building integrated photovoltaic BSRN—Baseline Surface Radiation Network c—speed of light (2.99792458 × 108 m/s) C—concentration ratio of a solar cell or system CPV—concentrating photovoltaic D—number of days elapsed since noon, January 1, 2000 DNI—direct normal irradiance DSEF—daily spectral enhancement factor E—global irradiance incident on a horizontal surface (W/m2) Eb—direct irradiance incident on a horizontal surface (W/m2) Ebn—DIRECT irradiance incident on a surface normal to sun rays (W/m2) Ec—circumsolar irradiance incident on a surface normal to sun rays (W/m2) Ed—diffuse sky irradiance incident on a horizontal surface (W/m2) Ee—effective incident irradiance (W/m2) Eeλ—effective incident spectral irradiance (W/m2 nm) En0—total solar irradiance at the limits of Earth’s atmosphere (W/m2) Enλ—spectral solar irradiance at the limits of Earth’s atmosphere (W/m2 nm) Er—reflected irradiance incident on a tilted receiver (W/m2) ERλ—reference STC spectral irradiance (W/m2 nm) Es—total irradiance incident on a tilted receiver (W/m2) Fλ—photon flux at the limits of Earth’s atmosphere (cm−2 /s nm) FP—flat plate g—mean sun anomaly (rad) h—Planck constant (6.62606896 × 10−34 J s) H—daily global irradiation on a horizontal surface (Wh/m2)

ABBREVIATIONS

457

Hbn—daily direct normal irradiation (Wh/m2) Hd—daily diffuse irradiation on a horizontal surface (Wh/m2) Hn0—daily extraterrestrial direct normal irradiation (Wh/m2) H0—daily extraterrestrial direct irradiation on a horizontal surface (Wh/m2) H—daily global irradiation on a horizontal surface (Wh/m2) IEC—International Electrotechnical Commission K—ratio Ed/E or Hd/H Kn—ratio Ebn/En0 or Hbn/Hn0 Kt—ratio E/(En0 cos Z) or H/H0 L—site latitude (degree) m—air mass Nc—circumsolar spectral radiance (W /m sr nm) NOAA—National Oceanic and Atmospheric Administration NREL—National Renewable Energy Laboratory NSRDB—National Solar Radiation Data Base PMOD—Physikalisch-Meteorologisches Observatorium Davos POA—plane of array PV—photovoltaic or solar cell R—actual sun–earth distance (m) R0—average sun–earth distance (1.496 × 1011 m) Rd—ratio of sky diffuse irradiance on a tilted receiver to that on the horizontal Rs—volumetric mean solar radius (6.96 × 108 m) s—receiver’s tilt angle from the horizontal S—sun–earth distance correction factor Sλ—solar cell’s spectral response SAM—(instrument for) sun and aureole measurement SC—solar constant SMARTS—Simple Model of the Atmospheric Radiative Transfer of Sunshine SORCE—Solar Radiation and Climate Experiment sr—time of sunrise SREF—solar radiation enhancement factor ss—time of sunset STC—standard test conditions SURFRAD—Surface Radiation Network t—time T—temperature (K) TMY—typical meteorological year TMY2—TMY version 2 TMY3—TMY version 3 TSI—total solar irradiance w—precipitable water (cm) WRDC—World Radiation Data Center Z—solar zenith angle α—aerosol wavelength exponent or “Ångström coefficient” β—aerosol optical depth at 1000 nm, or “Ångström turbidity coefficient”

458

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

γ—solar azimuth from North θ—angle of incidence of the sun’s rays on a surface or receiver λ—wavelength (nm) ξ—angular distance from the sun center (degree) ξ0—radial acceptance angle of a receiver (degree) ξs—angular distance from the sun’s center to its outer edge (degree) ρ—ground reflectance for shortwave radiation, or “albedo” σ—Stefan–Boltzmann constant (5.670 400 × 10−8 W/m2 K4) τaλ—aerosol optical depth at wavelength λ ψ—receiver azimuth from North (degree)

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

C. A. Gueymard. Reference solar spectra: Their evolution, standardization and comparison to recent measurements. Adv. Space Res. 37, 323–340 (2006). D. Labs and H. Neckel. The solar constant. Sol. Phys. 19, 3–15 (1971). C. Fröhlich. Solar irradiance variability since 1978: Revision of the PMOD Composite during Solar Cycle 21. Space Sci. Rev. 125, 53–65 (2006). C. Fröhlich and J. Lean. The sun’s total irradiance: Cycles, trends and related climate change uncertainties since 1976. Geophys. Res. Lett. 25, 4377–4380 (1998). C. A. Gueymard. The sun’s total and spectral irradiance for solar energy applications and solar radiation models. Sol. Energy 76, 423–453 (2004). ASTM. Standard solar constant and zero air mass solar spectral irradiance tables. Standard E490-00, American Society for Testing and Materials, West Conshohocken, PA. Available at http://www.astm.org/ (2000). M. Iqbal. An Introduction to Solar Radiation. Toronto, Academic Press (1983). H. Neckel. Analytical reference functions F(λ) for the sun’s limb darkening and its absolute continuum intensities (300 to 1100 nm). Sol. Phys. 229, 13–33 (2005). P. J. Mohr, B. N. Taylor, and D. B. Newell. CODATA recommended values of the fundamental physical constants: 2006. Rev. Mod. Phys. 80, 633–730 (2008). J. J. Michalsky and P. W. Kiedron. Comparison of UV-RSS spectral measurements and TUV model runs for clear skies for the May 2003 ARM aerosol intensive observation period. Atmos. Chem. Phys. 8, 1813–1821 (2008). V. Badescu, ed. Modeling Solar Radiation at the Earth Surface. Berlin, Springer (2008). C. A. Gueymard. An anisotropic solar irradiance model for tilted surfaces and its comparison with selected engineering algorithms. Sol. Energy 38, 367–386 (1987); Erratum, Sol. Energy 40, 175 (1988). B. Robles-Ocampo, E. Ruiz-Vasquez, H. Canseco-Sanchez, et al. Photovoltaic/ thermal solar hybrid system with bifacial PV module and transparent plane collector. Sol. Energ. Mat. Sol. C. 91, 1966–1971 (2007). K. J. Weber, V. Everett, P. N. K. Deenapanray, et al. Modeling of static concentrator modules incorporating lambertian or v-groove rear reflectors. Sol. Energ. Mat. Sol. C. 90, 1741–1749 (2006). S. Hatwaambo, H. Hakansson, J. Nilsson, and B. Karlsson. Angular characterization of low concentrating PV– CPC using low-cost reflectors. Sol. Energ. Mat. Sol. C. 92, 1347–1351 (2008). C. A. Gueymard. Direct and indirect uncertainties in the prediction of tilted irradiance for solar engineering applications. Sol. Energy 83, 432–444 (2009).

REFERENCES [17] [18] [19]

[20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30]

[31] [32] [33] [34] [35] [36]

459

C. A. Gueymard. Direct solar transmittance and irradiance predictions with broadband models. Pt 2: Validation with high-quality measurements. Sol. Energy 74, 381–395 (2003); Corrigendum, Sol. Energy 376, 515 (2004). C. A. Gueymard and D. R. Myers. Validation and ranking methodologies for solar radiation models. In Modeling Solar Radiation at the Earth Surface, V. Badescu, ed., pp. 479–509. Berlin, Springer (2008). ASTM. Standard tables for reference solar spectral irradiances: Direct normal and hemispherical on 37° tilted surface. Standard G173-03, American Society for Testing and Materials, West Conshohocken, PA. Available at http://www.astm.org/ (2003). A. A. Kokhanovsky. Cloud Optics. Berlin, Springer (2006). J. A. Davies and D. C. McKay. Evaluation of selected models for estimating solar radiation on horizontal surfaces. Sol. Energy 43, 153–168 (1989). J. J. Michalsky, J. L. Berndt, and G. J. Schuster. A microprocessor-based rotating shadowband radiometer. Sol. Energy 36, 465–470 (1986). C. A. Gueymard and D. R. Myers. Evaluation of conventional and high-performance routine solar radiation measurements for improved solar resource, climatological trends, and radiative modeling. Sol. Energy 83, 171–185 (2009). S. Wilcox, M. Anderberg, W. Beckman, et al. National Solar Radiation Database 1991–2005 Update: User’s manual. Golden, CO, NREL, Rep. TP-581-41364. Available at http://www.nrel.gov/docs/fy07osti/41364.pdf (2007). C. A. Gueymard and D. R. Myers. Solar radiation measurement: Progress in radiometry for improved modeling. In Modeling Solar Radiation at the Earth Surface, V. Badescu, ed., pp. 1–27. Berlin, Springer (2008). E. L. Maxwell, W. F. Marion, D. R. Myers, et al. National Solar Radiation Data Base—Final Technical Report Vol. 2. Rep. TP-463-5784, National Renewable Energy Laboratory, Golden, CO (1995). E. L. Maxwell. METSTAT—The solar radiation model used in the production of the National Solar Radiation Data Base (NSRDB). Sol. Energy 62, 263–279 (1998). R. Perez, P. Ineichen, K. Moore, et al. A new operational model for satellite derived irradiances, description and validation. Sol. Energy 73, 307–317 (2002). W. Marion and S. Wilcox. Solar radiation data manual for flat-plate and concentrating collectors. Rep. TP-463-5607, NREL, Golden, CO (1994). A. Tsvetkov, S. Wilcox, D. Renné, and M. Pulscak. International solar resource data at the World Radiation Data Center. In Proceedings of the Solar ’95 Conference, Minneapolis, MN, R. Campbell-Howe and B. Wilkins-Crowder, eds., pp. 216–219. Boulder, CO, American Solar Energy Society (1995). A. Ohmura, E. G. Dutton, B. Forgan, et al. Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate research. B. Am. Meteorol. Soc. 79, 2115–2136 (1998). T. P. Ackerman and G. Stokes. The Atmospheric Radiation Measurement Program. Phys. Today 56, 38–45 (2003). J. A. Augustine, J. DeLuisi, and C. Long. SURFRAD—A national surface radiation budget network for atmospheric research. Bull. Am. Meteorol. Soc. 81, 2341–2357 (2000). P. Ineichen. Comparison and validation of three global-to-beam irradiance models against ground measurements. Sol. Energy 82, 501–512 (2008). R. Perez, R. Seals, P. Ineichen, et al. A new simplified version of the Perez diffuse irradiance model for tilted surfaces. Sol. Energy 39, 221–231 (1987). D. F. Menicucci and J. P. Fernandez. User’s manual for PVFORM: A photovoltaic system simulation for stand-alone and grid-interactive applications. Rep. SAND850376, Sandia National Laboratories, Albuquerque, NM (1988).

460

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

[37]

D. R. Myers. General cloud cover modifier for clear sky solar radiation models. In Proceedings of the Optical Modeling and Measurements for Solar Energy Systems Conference, Vol. 6652, D. R. Myers, ed., pp. 6652–08. Bellingham, WA, SPIE, International Society for Optical Engineering (2007). D. R. Myers. Relative performance of multiple solar radiation resource assessment data sources. In Proceedings of the Solar 2007 Conference, R. Campbell-Howe, ed., paper 031A. Boulder, CO, American Solar Energy Society (2007). M. Suri, J. Remund, T. Cebecauer, et al. First steps in the cross-comparison of solar resource spatial products in Europe. In Proceedings of the Eurosun 2008 Conference Freiburg, Germany, ISES (2008). C. A. Gueymard. Fixed or tracking solar collectors? Helping the decision process with the solar resource enhancement factor. In Optical Modeling and Measurements for Solar Energy Systems II Conference, Vol. 7046, B. K. Tsai,, ed., doi: 10.1117/12.795506. Bellingham, WA, SPIE, International Society for Optical Engineering (2008). C. B. Christensen and G. M. Barker. Effects of tilt and azimuth on annual incident solar radiation for United States locations. In Proceedings of the Solar Forum 2001 Conference, S. J. Kleis and C. E. Bingham, eds., pp. 225–232. New York, NY, American Society of Mechanical Engineers (ASME) (2001). C. A. Gueymard and S. M. Wilcox. Spatial and temporal variability in the solar resource: Assessing the value of short-term measurements at potential solar power plant sites. In Proceedings of the Solar 2009 Conference, R. Campbell-Howe, ed., paper 118. Boulder, CO, American Solar Energy Society (2009). W. E. McMahon, K. Emery, D. J. Friedman, et al. Outdoor testing of GaInP2/GaAs tandem cells with top cell thickness varied. Proceedings of the International Conference on Solar Concentrators for the Generation of Electricity or Hydrogen, Golden, CO, National Renewable Energy Laboratory, Rep. No. CP-520-37819, http://www.nrel.gov/docs/fy05osti/38172.pdf (2005). C. A. Gueymard and H. D. Kambezidis. Solar spectral radiation. In Solar Radiation and Daylight Models, T. Muneer, ed., pp. 221–301, Amsterdam, Elsevier (2004). C. A. Gueymard. Advanced solar irradiance model and procedure for spectral solar heat gain calculation (RP-1143). ASHRAE Trans. 113(1), 149–164 (2007). C. A. Gueymard. Prediction and validation of cloudless shortwave solar spectra incident on horizontal, tilted, or tracking surfaces. Sol. Energy 82, 260–271 (2008). C. A. Gueymard, D. Myers, and K. Emery. Proposed reference irradiance spectra for solar energy systems testing. Sol. Energy 73, 443–467 (2002). K. Emery, D. Myers, and S. Kurtz. What is the appropriate reference spectrum for characterizing concentrator cells? In Proceedings of the 29th Photovoltaic Specialists Conference, pp. 840–843. New Orleans, LA, IEEE (2002). ASTM. Table for reference solar spectral distributions: Direct and diffuse on 20° tilted and vertical surfaces. Standard G197-08, American Society for Testing and Materials, West Conshohocken, PA. Available at http://www.astm.org/ (2008). C. A. Gueymard. Daily spectral effects on concentrating PV solar cells as affected by realistic aerosol optical depth and other atmospheric conditions. Proceedings of the Optical Modeling and Measurements for Solar Energy Systems II Conference, Vol. 7410, B. K. Tsai, ed., paper 7410–07. San Diego, CA, SPIE, International Society for Optical Engineering (2009). C. N. Vittitoe and F. Biggs. Six-Gaussian representation of the angular brightness distribution for solar radiation. Sol. Energy 27, 469–490 (1981). J. E. Noring, D. F. Grether, and A. J. Hunt. Circumsolar radiation data: The Lawrence Berkeley Laboratory reduced data base. Rep. TP-262-4429, NREL, Golden, CO. Available at http://rredc.nrel.gov/solar/pubs/circumsolar/ (1991).

[38] [39] [40]

[41]

[42]

[43]

[44] [45] [46] [47] [48] [49] [50]

[51] [52]

REFERENCES [53] [54] [55] [56] [57] [58] [59] [60] [61]

461

M. Schubnell. Influence of circumsolar radiation on aperture, operating temperature and efficiency of a solar cavity receiver. Sol. Energ. Mat. Sol. C. 27, 233–242 (1992). A. Neumann, A. Witzke, S. A. Jones, and G. Schmitt. Representative terrestrial solar brightness profiles. J. Sol. Energ.-T. ASME 124, 198–204 (2002). D. Buie and A. G. Monger. The effect of circumsolar radiation on a solar concentrating system. Sol. Energy 76, 181–185 (2004). J. G. DeVore, A. T. Stair, A. LePage, et al. Retrieving properties of thin cloud from solar aureole measurements. J. Atmos. Ocean. Technol. 26, 2531–2548 (2009). C. A. Gueymard. Parameterized transmittance model for direct beam and circumsolar spectral irradiance. Sol. Energy 71, 325–346 (2001). A. Rabl. Comparison of solar concentrators. Sol. Energy 18, 93–111 (1976). M. A. Rosen, F. C. Hooper, and A. P. Brunger. The characterization and modelling of the diffuse radiance distribution under partly cloudy skies. Sol. Energy 43, 281–290 (1989). A. P. Brunger and F. C. Hooper. Anisotropic sky radiance model based on narrow field of view measurements of shortwave radiance. Sol. Energy 51, 53–64 (1993); Erratum, Sol. Energy 51, 523 (1993). A. W. Harrison and C. A. Coombes. Angular distribution of clear sky short wavelength radiance. Sol. Energy 40, 57–63 (1988).

20 SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL PAUL GILMAN,1 NATHAN BLAIR,1 AND CHRISTOPHER CAMERON2 1 National Renewable Energy Laboratory, 2Sandia National Laboratories

20.1

INTRODUCTION

No discussion of solar cells is complete without an examination of their role as part of a complete PV system in the context of an energy project. From a technical perspective, solar cells are just one component in an electricity-generating system that consists of PV modules, inverters, and BOS components. From economic, financial, and policy perspectives, solar cells are just one element in a valuegenerating system whose worth depends on the quantity of energy the system produces, the system lifetime, installation costs, operating and maintenance costs, a range of financial options, and incentives from a variety of institutions and organizations. Computer models can help analysts understand how solar cells perform in a system and how they contribute to the cost of energy produced by the system. In this chapter, we describe and present sample analyses for one computer model designed specifically for solar energy projects: the U.S. Department of Energy’s SAM, developed by NREL, Sandia National Laboratories, and other partners. SAM calculates life cycle economic metrics such as the LCOE by integrating the results from hour-by-hour performance models with those from a cash flow cost model. SAM facilitates system-based analysis, which is an alternative to conventional approaches to valuing and comparing PV technologies. While conventional approaches focus on metrics that emphasize solar cells, such as cell efficiencies and manufacturing costs, a system-based approach focuses on energy production metrics, such as the LCOE. System-based analyses account for parameters determined by the economic context of projects in the residential, commercial, and utility electricity markets. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

463

464

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

The SAM analysis examples described later in this chapter illustrate how system-based analysis can help ensure that important considerations in technology selection are included in decision making for research management, policy making, system design, and other efforts to develop the full potential of PV technologies. The system-based approach offers the following advantages:







Find the Low-Hanging Fruit. The location of a PV project plays an important role in both its technical feasibility and its economic attractiveness to investors. Using a systems approach on a large regional scale can help analysts discover unexpected opportunities for PV project development. For example, analysts have been quick to identify the development potential of obvious locations in sunny regions of the United States. However, they have only relatively recently shown that for some less sunny locations, factors such as the coincidence of the solar resource with demand peaks and transmission constraints limiting the availability of power from central generating plants offer opportunities for largescale implementation of distributed PV projects. A conventional analysis might focus only on annual electricity production per unit of available solar energy, and fail to capture the value of the solar resource’s seasonal or daily coincidence with high electricity prices. Example 3 described below is one example of how SAM can be used for location-specific analysis, in this case comparing system output and LCOE for identical systems operating in a hot and cold climate. Take Off the Blinders. Sometimes, analysts fall into the trap of focusing on a particular area of interest and miss the big picture. For example, by focusing exclusively on reducing PV module or inverter installation costs, an analyst may miss the effect of lower-cost component performance on the system life cycle cost. For example, a lower-cost module might degrade more rapidly than a more expensive one, or a lower-cost inverter may require more frequent repair and replacement, ultimately making the electricity generated by a system using those components more expensive in spite of the lower capital costs. Examples 4 and 5 described later in this chapter show how SAM can help with analyses of system and inverter lifetimes. Make Good Business Decisions. Looking at how components operate as part of a system and how they contribute to the system’s life cycle cost can help managers to choose the most promising direction to take research and development programs or to choose products for specific projects. Although choosing PV system equipment to minimize installation costs may seem like an obvious good business decision, in reality, other factors may come into play that cause unexpected choices to make the best business sense. Example 7 shows how the type of financing of a utility-scale PV project is just as important a consideration as the cost and efficiency of the PV modules.

LCOE

465



Compare Options on a Level Playing Field. Comparing projects based on different energy technologies, particularly capital-intensive ones like PV systems and operating cost-intensive ones like fuel-based systems, requires metrics that accurately express the value of the different options. Obviously, basing a technology choice between a utility-scale PV project and a natural gas plant on the installation cost alone would make the gas plant seem like the most cost-effective choice but would not capture the true cost of electricity generated by the two systems. Similarly, evaluating distributed generation projects requires a metric that is comparable to the cost per kilowatt hour of retail electricity against which the projects compete. All of the analyses in this chapter use one suitable metric for such comparisons, the LCOE, which is described next.

20.2

LCOE

The LCOE represents a project’s TLCC per unit of energy. The TLCC is the present value of the sum of the installation and project cash flows Cn over project life N, assuming a discount rate of d: TLCC =

N

Cn

n=0

(1 + d )n



.

By definition, multiplying the total annual energy production Qn by the LCOE for each year in project life N and discounting the resulting stream of values to year 1 results in a value equal to the TLCC: N

Qn × LCOE

n =1

(1 + d )n



= TLCC.

The LCOE is therefore mathematically equivalent to the TLCC divided by the “discounted” annual energy production of the system: LCOE =

TLCC . Qn ∑ n n =1 (1 + d ) N

The LCOE is typically expressed in cents per kilowatt hours to make it comparable to cost of energy values for projects based on different types of technologies. To calculate the LCOE of a PV system, SAM first uses an hourly simulation engine to calculate the system’s annual electric output based on a set of parameters describing the physical system, then calculates a series of annual cash flows based on a set of cost and financial assumptions, and finally calculates the LCOE using

466

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

the appropriate discount rates. The annual cash flows include the income and costs shown in Table 20.1. The LCOE is useful for comparing alternative options for projects based on different technologies or operating in different markets. For example, the LCOE would be useful to compare the true cost of a residential rooftop PV system with residential utility rates, or the cost of a utility-scale PV system with electricity market rates. The LCOE is also an appropriate metric for comparing different kinds of energy projects. Although this chapter focuses on the LCOE, SAM also generates other useful metrics shown in Table 20.2. The metrics depend on the type of financing.

TABLE 20.1. Annual Cash Flow Income and Cost Categories in SAM Income

Electricity sales Incentive payments Tax deductions, including for depreciation Debt

Costs

Installation Operation and maintenance Equipment replacement Debt payments and interest

TABLE 20.2. Economic Metrics Available in SAM for Each Type of Financing Type of Financing

Available Metrics

Residential cash Residential loan or mortgage Commercial cash Commercial—standard loan

LCOE Net present value (NPV) Simple payback kWh/kW—year 1 Capacity factor System performance factor Annual output—year 1

Utility and IPP Commercial third-party ownership

LCOE Internal rate of return Minimum DSCR First-year PPA PPA escalation rate Debt fraction kWh/kW—year Capacity factor System performance factor Annual output—year 1

THE SAM SOFTWARE

467

Although some metrics included in the list such as internal rate of return and simple payback do not reflect the true value of a project, they are included as available options because many analysts are familiar with them.

20.3

THE SAM SOFTWARE

SAM models a range of solar energy technologies for electricity generation, including PV arrays, solar thermal troughs, power towers, and dish–Stirling systems. SAM also includes a simple thermal model for comparisons between solar energy systems and fossil fuel-based thermal power plants. For PV systems, SAM first calculates the system’s total electricity production in kilowatt hours for the first year based on weather data for a particular location and physical specifications of the array and inverter. SAM calculates the total production for subsequent years based on an annual degradation factor, and annual cash flows based on financial and economic inputs to determine the LCOE and other economic metrics.









Weather Data. SAM’s simulation engine uses data from hourly weather files to model the solar resource at a given location. The weather file supplies radiation data for energy calculations, and wind speed, temperature, snow cover, and other data to model the effects of temperature and ground reflectance on system performance. Financial and Economic Inputs. SAM’s cash flow calculations depend on a set of input variables describing the project’s finances, such as the analysis period equivalent to the system lifetime, discount rate, inflation rate, loan amount, and loan rate. SAM provides different financing options that use different sets of input variables. Table 20.3 shows the financing options available to the different project types. Incentives. SAM models two types of incentives: tax credits and incentive payments. Tax credits can be provided by a state or federal government. Incentive payments can be provided by a state or federal government, utility, or other entity. An investment-based incentive or tax credit is a one-time payment to the project made in year 1 of the project cash flow that is either a fixed amount, a percentage of total installed costs, or a function of system size. A production-based incentive or tax credit is an annual payment based on the amount a of energy produced by the system in each year. Incentive payments may or may not be taxable by either the federal or state government. System Performance. SAM simulates the hourly performance of one or more PV modules in an array, calculating an hourly DC output value that it passes to the inverter performance model, which in turn calculates the system’s AC output (Fig. 20.1). Table 20.4 describes the three module performance models available in SAM, and Table 20.5 describes the two inverter models available in SAM.

468

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

TABLE 20.3. Financing Options for Different Project Types Market

Financing Options

Residential

Commercial

Central generation



Description

Cash

The owner pays cash in the amount of the total installed cost in year 0 of the project cash flow.

Loan or mortgage

The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes a principal and interest payment in subsequent years.

Cash

The owner pays cash in the amount of the total installed cost in year 0 of the project cash flow.

Loan

The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes an interest and principal payment in subsequent years.

Third-party ownership

The project earns revenues through electricity sales to cover project costs. The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes an interest and principal payment in subsequent years.

Utility and independent power producer (IPP)

The project earns revenues through electricity sales to cover project costs. The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes an interest and principal payment in subsequent years.

Performance Adjustments. SAM uses a set of derating factors to convert the predicted output of a single module into an array output value. A set of DC derating factors accounts for module mismatch, diodes and connections, DC wiring, shading, soiling, and, if applicable, tracking losses. A second set of derating factors accounts for post-inverter losses from AC wiring and transformers. An availability factor accounts for system outages for maintenance. To calculate the system output in years after the first year, SAM applies an annual degradation factor to account for the annual reduction in system output due to equipment aging (Fig. 20.1).

THE SAM SOFTWARE

469

TABLE 20.4. Module Performance Model Descriptions Sandia PV array performance model

An empirically based model that consists of a set of curve-fit equations for each module in a database derived from outdoor module testing conducted on a two-axis tracker

California Energy Commission (CEC) performance model

A five-parameter model developed by the University of Wisconsin’s Solar Energy Laboratory, which uses a database of standard rating condition data from the manufacturer’s module specifications or independent test laboratories and a set of semiempirical correlation equations to predict the module I–V curve. SAM assumes that modules operate at their maximum power point.

Simple efficiency model for flat-plate modules

A simple model that calculates module output by multiplying the total global solar radiation incident on the array by the module’s efficiency value and area and then by adjusting the performance through the use of a temperature coefficient, assuming a normal cell operating temperature of 25°C. A different efficiency value can be entered for up to five incident radiation values.

Simple efficiency model for concentrating modules

A simple model that calculates module output by multiplying the direct normal component of the solar radiation by the module’s efficiency value and area. A different efficiency value can be entered for up to five incident radiation values

PVWatts Solar Array

An implementation of NREL’s web-based PVWatts model that represents the complete system using a single DC-toAC derating factor. As of October 2009, this model includes preliminary battery storage and load models that will be improved and will be eventually available with the other module model options.

TABLE 20.5. Inverter Performance Model Descriptions Sandia performance model for gridconnected inverters

The model uses a set of parameters for commercially available inverters in a set of equations to calculate the AC power output based on DC power input and voltage values generated by a PV module model. The parameter sets are maintained by Sandia National Laboratories and are primarily based on a list of laboratory performance measurements maintained by the California Energy Commission.

Single-point efficiency model

A simple model that represent the inverter performance with two parameters: the AC nominal capacity rating and an efficiency value. This simple model multiplies the PV array’s DC output by the efficiency value to calculate the AC output.

ANALYSIS EXAMPLES

20.4

471

ANALYSIS EXAMPLES

The following examples illustrate SAM’s use in analyses based on the system approach described above. The analyses start with assumptions similar to those of the analysis described in Chapter 2, which is for a 4-kW grid-connected residential system with net metering, and using module and inverter efficiency and system cost assumptions reasonable for a representative system in the United States. Example 1 replicates the Chapter 2 analysis, and Examples 2–7 show additional analyses illustrating some of the insights that a computer model like SAM can add.

Example 1: Module and Inverter Conversion Efficiency Improvements and Cost Reductions over Time This analysis replicates the “simple approach” analysis described in Chapter 2. SAM’s cash flow method, described earlier in this chapter, replaces the fixedcharge rate method of the simple approach. In place of the solar energy density and efficiency factors used in the simple approach to predict the system’s total annual output, SAM uses the hour-by-hour performance simulation described above to calculate the system’s hourly output. The near-term and medium-term systems use assumptions reasonable for 5 and 15 years in the future with the expectation of the following improvements over time, assuming that the area remains constant:

• • • •

increases in module and inverter efficiency, increases in inverter lifetime, reduction in capital costs, and reduction in government incentives.

Table 20.7 summarizes the performance parameters that describe the physical system. The analysis uses solar radiation and position, ambient temperature, wind speed data from the TMY2 weather file for Phoenix, Arizona, and assumes a fixed array tilted at 18°. The analysis also assumes that the inverter capacity increases with the array capacity in the near and medium terms to match the array capacity, which increases with the array efficiency as the area remains constant. Table 20.8 summarizes the cost, financial, and incentive assumptions of the analysis that SAM uses to calculate the cash flow and economic metrics. In addition to the values in the table, the analysis assumes a 5.5% discount rate (to discount future values to present values) and a loan fraction of 100% for all three cases. The following adjustments were required to account for differences in measurement units used by the two methods:



The simple approach includes a BOS cost expressed in dollars per square meter that includes installation costs, while in SAM, the BOS cost is

472

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

TABLE 20.7. Summary of Performance Parameter Inputs for SAM Analysis Parameter

Baseline

PV array



Module efficiency Array capacity Array area Derating factor System lifetime Inverter

Near Term

Medium Term

Units







13.5

16.0

20.0

%

4.0

4.7

5.9

29.63

29.63

29.63

m2

0.95

0.95

0.95



Same as analysis period

kWDC-peak (STC)

Years









Efficiency

0.95

0.96

0.97



Total capacity

4.0

4.7

5.9

kWAC

Lifetime

5







10

20

Years

expressed in dollars, and the installation and BOS costs are in separate categories. The BOS cost in Table 20.8 is the product of the area-related BOS (Cb) cost in Table 2.2 and the array area in Table 20.7. The simple approach includes an indirect cost rate of 22.5%, of which 0.5% accounts for both operation and maintenance and inverter replacements. In SAM, the indirect cost is a fixed amount and is a component of the total capital cost. The indirect cost in Table 20.8 is 22% of the total direct capital cost. SAM calculates an annual operation and maintenance cost as a function of up to three operation and maintenance cost components: a fixed annual cost, a fixed annual cost per DC kilowatt of installed capacity, and a variable annual cost per AC megawatt hour generated per year. This analysis assumes an annual operation and maintenance calculated as 0.5% of the total installed cost for the first year, and escalated at the inflation rate of 2.5%. SAM calculates the inverter replacement cost based on the inverter capital cost, inflation rate, and replacement interval in years. SAM only applies the inverter replacement costs at the replacement intervals.

Tables 20.9 and 20.10 show a summary of the results, including LCOE values comparable to those from the simple method used in Chapter 2. The results show a trend in LCOE reductions as module and inverter efficiencies improve and as capital costs drop over time, even as government incentives decline. The magnitude of the trend shown in Table 20.10 is similar for both the SAM and simple approach methods. Note that the system output in the first year for each case increases

ANALYSIS EXAMPLES

473

TABLE 20.8. Summary of Cost, Financial, and Incentive Inputs for SAM Analysis Baseline

Near Term

Medium Term

PV module

4.00

2.20

1.25

$/WDC

Inverter (= Ci × inverter capacity)

3600

3280

1800

$/unit

Balance-of-system (BOS) cost (= Cb × array area)

9096

4592

4445

$

Installation labor (included in BOS cost)

0

0

0

$

28,696

23,965

13,652

$

Indirect (miscellaneous) (= 0.22 × A)

6313

5272

3003

$

Total installed cost (C)

8.75

4.97

2.81

$/WDC

First-year inverter replacement cost

3600

3280

1800

$/year

2.5

2.5

2.5

%/year

Total direct costs (A)

Inflation rate Inverter replacement interval

Units

5

10

20

175.00

116.80

82.90

Analysis period

30

35

35

Years

Loan rate

6.0

6.0

6.0

%/year

Loan period

30

35

35

Years

Capacity-based incentive

2.50

1.21

0.55

$/WDC

Federal tax rate

28.0

28.0

28.0

%

Scheduled and unscheduled maintenance (= 0.005 × C × array capacity)

Years $/kWDC-year

Notes: The inverter cost is based on the DC–AC inverter cost (Ci) in Table 2.2 and on the inverter capacity in Table 20.7. The BOS cost is based on the area-related BOS cost (Cb) in Table 2.2 and on the array area in Table 20.7. The table shows first-year values for scheduled and unscheduled maintenance and is based on the array capacity in Table 20.7.

because the array area is held constant as the module efficiency increases, thereby increasing the array’s peak rated capacity. Figure 20.2 shows the results generated by SAM of an analysis similar to the sensitivity analysis described in Chapter 2. The analysis varies the module efficiency for each of the three cases while holding the other assumptions constant. The inverter efficiency is held constant over the range of module

474

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

TABLE 20.9. Summary of Results for SAM Analysis Baseline Levelized cost of energy

34.7

First-year output

Near Term

Medium Term

15.9

7550

9040

Simple approach results





Levelized cost of energy

32.0

15.0

Units

10.1

Cents/kWh

11,300

kWh/year



— 9.0

Cents/kWh

TABLE 20.10. Reduction of Levelized Cost of Energy from the Baseline Baseline

Near Term

Medium Term

Units

SAM

0

54

71

%

Simple approach

0

53

72

%

Levelizec Cost of Energy (¢/kWh)

80 70 60 Base Line Near Term Medium Term

50 40 30 20 10 0

0

5

10

15 20 25 Module Efficiency (%)

30

35

Figure 20.2. Levelized cost of energy as a function of module efficiency.

efficiencies, but the number of inverters is adjusted to ensure that the inverter capacity is sufficient to handle the array’s output for each point in the sensitivity analysis. Again, the results show similar trends to the results based on the simple approach in Chapter 2. The advantage of using a computer model to perform this kind of analysis over the simple approach is that more parameters can be adjusted to explore the impact of changes in design, costs, incentives, and financial assumptions. The following examples explore some of the possibilities based on the baseline analysis presented above.

ANALYSIS EXAMPLES

475

Example 2: Module Degradation In addition to system lifetime, SAM can model the change in performance of PV modules over time. PV module output typically decreases over time as module components age. Analysts often use an annual degradation rate between 0.5% and 1.0%. Module manufacturer lifetime warranties typically guarantee that the module often will generate no less than 80% of the nameplate output at the end of the warranty period, which is usually in the 25- to 30-year range. Figure 20.3 shows the effect in SAM of changing the system degradation for the baseline case described above while holding all other assumptions constant. SAM compounds the degradation rate annually: there is no loss of output in year 1; in year 2, the system produces 99% of the year 1 output; in year 3, it produces 99% of the year 2 output, and so on. The analysis shows that each 1% change in annual degradation rate increases the system’s LCOE by about 10% (Fig. 20.3).

Example 3: Temperature Correction

Multiple of Levelized Cost of Energy (1 = No Degradation)

Module performance is affected by the cell temperature, which SAM calculates as a function of the ambient temperature, ambient wind speed, and the solar radiation incident on the array. SAM models this effect using the module’s temperature coefficient and information from the weather file. In a relatively hot climate such as Phoenix, the effect of a −0.5%/°C maximum power temperature coefficient, which is typical for multicrystalline silicon modules, on a rack-mounted module is to reduce its total annual energy output by about 10% from that of a theoretical module that does not respond to cell temperature (represented in SAM by a temperature coefficient of zero) (Fig. 20.4).

1.6 1.5 1.4 1.3 1.2 1.1 1 0

0.5

1

1.5 2 2.5 3 3.5 System Degradation (%/yr)

4

4.5

5

Figure 20.3. Effect of system degradation on levelized cost of energy.

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL Fraction of Annual Output (1 = No Temperature Correction)

476 1 0.95 0.9 0.85 0.8

0.75 –1

–0.8 –0.6 –0.4 –0.2 Temperature Coefficient (%/°C)

0

Multiple of Levelized Cost of Energy (1 = No Temperature Correction)

Figure 20.4. Effect of module temperature coefficient on annual output for the relatively hot climate of Phoenix.

1.3 1.25 1.2 1.15 1.1 1.05 1 –1

–0.8 –0.6 –0.4 –0.2 Temperature Coefficient (%/°C)

0

Figure 20.5. Effect of module temperature coefficient on levelized cost of energy.

The temperature-related reduction in output results in a corresponding reduction in the LCOE, as shown in Figure 20.5. We used the results of SAM’s hourly simulation to further explore the relationship between module temperature and output. Figures 20.6 and 20.7 show hourly conditions for a hot day in Phoenix and for a cold day in Casper when the incident solar radiation peaks at 1 kW/m2 for both locations. At 1 p.m. on August 17 in Phoenix, the cell temperature peaks at about 62°C, while the incident solar radiation is 1 kW/m2 and the ambient temperature is about 40°C. In Casper at 1 p.m. on March 18, the total incident radiation is also 1 kW/m2, but the ambient temperature is about −5°C and the cell temperature is 13°C, well below the module’s

477

0.5

25 5

°C

°C

45

kW/m2

1

65

–15

0 Noon Total Incident Radiation (kW/m2)

65

4

45

3

25

2

5

1

–15

kW

ANALYSIS EXAMPLES

0 Noon Array Output (kW-DC)

Ambient Temperature (°C)

Cell Temperature (°C)

5 –15

0 Noon

°C

0.5

25

kW/m2

°C

45

4 3 2 1 0

65 45

1

65

25 5 –15

kW

Figure 20.6. Effect of ambient temperature on system output for a day in August of a hot climate (Phoenix, Arizona).

Noon

Total Incident Radiation (kW/m2)

Array Output (kW-DC)

Ambient Temperature (°C)

Cell Temperature (°C)

Figure 20.7. Effect of ambient temperature on system output for a day in March of a cold climate (Casper, Wyoming).

normal operating cell temperature of 25°C. The resulting output of the 4-kW array is 3.2 kW for the system in Phoenix and 4.0 kW for the system in Casper. Note that these output values are for the array of 40 modules and account for array losses. Example 4: System Lifetime PV systems are typically assumed to have a lifetime of between 25 and 30 years, often based on the expected lifetime of the modules. Module manufacturers typically provide a 25- or 30-year warranty for crystalline silicon modules. The lifetimes of modules using newer technologies are uncertain. In this analysis, we explored the relationship between the LCOE for the baseline system in Phoenix with the system lifetime. Figure 20.8 shows the LCOE for the baseline case described above for a range of system lifetimes. In order to isolate the effect of the system lifetime, we held all other assumptions, including the module efficiency and cost, constant over the range of system lifetimes.

Multiple of Levelized Cost of Energy (1 = No O&M Cost)

ANALYSIS EXAMPLES

479

2 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0

0.5 1 1.5 2 2.5 3 3.5 4 4.5 O&M Cost as Percent of Total Installed Cost (%)

5

Figure 20.10. Effect of operation and maintenance costs on levelized cost of energy.

decreases as the inverter lifetime approaches the system lifetime of 30 years. This type of graph also makes it possible to evaluate different inverter options when the price depends on the inverter lifetime. For example, in this analysis, the graph shows that even with an installation cost as high as $1 per watt, an expensive inverter with a 15-year lifetime is more cost-effective than a 5-year inverter with an installation cost as low as 40 ¢ per watt. Example 6: Operation and Maintenance Expenses When evaluating technology options for energy projects, investors sometimes consider only the project’s total capital cost. In this analysis, we explore the sensitivity of the LCOE to project operation and maintenance costs, which would not be accounted for an evaluation based only on capital cost. Figure 20.10 shows that for the baseline residential 4-kW system in Phoenix, each 1% increase in operating and maintenance cost increases the LCOE by just less than 10%. Example 7: Debt Fraction and Debt Rate The examples so far are based on the 4-kW residential baseline system analyzed above and in Chapter 2. The following example analyzes a 10-MW utility-scale system. The analyses for the residential system assume that the project sells electricity through a net metering agreement with an electric service provider and is financed by a loan. The electricity sales price is an input to the model and can be either a fixed rate with annual escalation or based on a time-of-use schedule. For the residential system, the LCOE represents the cost of installing and operating the system over its lifetime and includes annual loan principal and interest payments, tax payments, and benefits from incentives and electricity sales.

482

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

While this chapter focused on PV systems, SAM can also model concentrating solar power systems, including parabolic trough, dish–Stirling, and power tower systems. SAM also includes a simple thermal model that can be used to compare solar systems to fossil fuel-based power plants. SAM provides a useful platform for comparing different technology options using a consistent modeling framework. ABBREVIATIONS A—total direct costs AC—alternating current BOS—balance of system C—total installed cost Cb—area-related BOS costs CEC—California Energy Commission Ci—inverter cost Cn—project cash flows d—discount rate DC—direct current DSCR—debt service coverage ratio IPP—independent power producer I–V—current–voltage characteristics or curve LCOE—levelized cost of energy N—project life NPV—net present value NREL—National Renewable Energy Laboratory PPA—power purchase agreement PV—photovoltaic or solar cell Qn—total annual energy production SAM—Solar Advisor Model STC—standard test condition TLCC—total life cycle cost TMY2—typical meteorological year data

21 CHALLENGES OF LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION DAVID FAIMAN Ben-Gurion University of the Negev

21.1

INTRODUCTION

The basic problem with large-scale solar power production stems from the diluteness of solar energy compared to other more readily available sources (such as barrels of oil). This necessitates extremely large sunlight capture areas and their associated infrastructure costs. Coupled to this is the relative remoteness of the richest terrestrial solar capture regions—the world’s deserts—from today’s population centers. In the past, the space requirement problem had been relatively muted compared to the problem of developing solar technologies that could compete economically with fossil fuel. And even the latter problem has been confused by fluctuating fuel prices and debates over a variety of geological, environmental, and political issues. However, now that the advent of CPV appears to be opening an avenue to really cost-effective solar power [1], other practical issues related to its large-scale implementation come to the forefront. Chief among these issues are storage, transmission, and financing. PV-generated electricity is less dispatchable than ST electricity, as far as utilities are concerned, because there is no intermediate thermal energy stage. Thermal energy is relatively easy to store, enabling the electricity from an ST plant to be generated when needed. In the case of PV, and particularly CPV, it is the electricity that must be stored and, as will be discussed below, this creates problems of a qualitatively different kind.

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

483

484

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

Electrical transmission is also a problem over large distances. In conventional grid networks, generating stations are typically interconnected by AC transmission lines of 100–200 km in length. For such networks, average transmission losses are found to be in the neighborhood of 7%. However, transmission lines of thousands of kilometers in length will be needed for connecting future desert-based solar power plants to distant cities. Lastly, financing takes on a new dimension when the payback times on initial investments may be tens of years. Several of these issues have been addressed in a series of volumes published by the International Energy Agency [2–4], whose Task 8 Photovoltaic Specialists Working Group coined the term VLS-PV for solar PV plants on the gigawatt scale. However, those volumes were more concerned with the problem of how to get a VLS-PV program started rather than that of implementing it on a worldwide scale, which is the concern of the present chapter. In the following sections, we shall attempt to discuss, in a structured manner, the scale of the problem and the areas in which a certain amount of R & D is still desirable. Our principal conclusion will be that technology is already sufficiently advanced and cost-effective to enable a massive VLS-PV program to go ahead.

21.2

THE SCALE OF THE PROBLEM

There are two aspects to implementing what has been termed the “grand solar plan” [5]. One is to realize the scale of investment that will be needed to replace a fossilfueled world with one whose energy needs are largely met by solar power. The other is to pinpoint the holes in our present knowledge in order to be able to create a viable solar technology in all of its details. Figure 21.1, for example, indicates that present world electricity generation is in excess of 20,000 TWh/year [6]. If we take a nominal 10% efficiency as being typical of what desert regions can provide via the employment of PV technology [7], and 2000 kWh/m2 as being a typical value for the solar irradiance in a desert region, then present world electricity production is equivalent to about 100 B square meters of solar panels and three to four times as much land area [7] for their placement. Thus, in order to meet only the present annual rise in electricity production of 610 TWh/year, seen in Figure 21.1, it would be necessary to erect 3 B square meters of PV panels each year on about 10,000 km2 of land. There are also likely to be significant political complications to a worldwide grand solar plan. In the United States, a single political system would be involved in providing solar-generated electricity to the entire population, but elsewhere, the situation is more complicated. For example, countries far removed from the world’s deserts would need to come to agreement with one another regarding the location and cost of power lines and the ensured supply of power. In addition, Europe, for example, even if treated as a single political entity, would need to reach agreement with the various Saharan states on whose territories “their” future VLS-PV plants would stand.

SPACE AND TIME REQUIREMENTS

485

22000

Annual Electricity Generation [TWh]

21000

World Electricity Production 1998-2007 (Source: BP Statistical Review 2008)

20000

y = –1.2011e+6 + 608.19x R^2 = 0.982 3 % per annum rise (base 2008)

19000 18000 17000

1% reference error bars

16000 15000 14000 1996

1998

2000

2002

2004

2006

2008

2010

Year

Figure 21.1. World electricity production during the past 10 years [6].

21.3

SPACE AND TIME REQUIREMENTS

The 10,000 km2 of additional annual land requirement that conventional PV would need in order to level off the slope of the graph in Figure 21.1 could be significantly reduced via the employment of CPV technology. This is because, under concentrated sunlight, CPV cells operate at considerably higher efficiency than 1-sun cells. Specifically, for CPV cells with a nominal (STC) efficiency of 32%, only 6 km2 of land would be needed for the generation of each 1 TWh/year [1]. By comparison, conventional fixed flat-panel PV would require 17 km2 of land [7]. Furthermore, with CPV cells already having been demonstrated with efficiencies greater than 40%, the land requirement is reduced proportionately. However, if we remain with the initial assumption of 32% cell efficiency [1], if the 610 TWh/year rise in world electricity requirements will continue unchanged, then in order for VLS-PV, in its CPV variety, to halt this rise, it will be necessary to generate an additional 610 TWh of solar electricity (from 305 GWp of CPV generating capacity) each year. In the first years, this would require 3660 km2 of new land annually, but as the technology improves (i.e., as the efficiency of massproduced CPV cells rises above the nominal 32%), the annual land requirement might eventually be reduced by perhaps 50%. Since the Sahara desert alone is approximately 10,000,000 km2 in area, this presents well over 2000 years of breathing space.

486

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

On the other hand, if we were to replace all of the world’s present 20,000 TWh of electricity production by CPV technology, it would require the immediate assignment of 120,000 km2 of land (cf. 360,000 km2 for conventional stationary flat-panel PV [7]), that is, slightly more than 1% of the Sahara. Therefore, other than the political need for the various countries that share the Sahara to reach agreement on such a project, there would appear to be no serious land limitations for replacing all of the world’s electrical generating capacity by CPV and for continuing the present electrical growth rate, in like manner, for a great many years. There is also one other time requirement that should not be neglected. It stems from the fact that no PV plant on a gigawatt scale has ever been built. Raviv has estimated [1] that, taking into consideration the requirement of a certain amount of additional R & D, it would take 4 years to construct a VLS-PV manufacturing facility capable of a throughput of 1 GWp/year. Thus, this is the time that would need to elapse, after a political decision had been taken to go ahead with such a program, before construction could commence on the first VLS-PV plant. The latter would then commence operation a year later.

21.4

COST REQUIREMENTS AND FINANCING

21.4.1

The Raw Costs

Let us first address the scenario in which we were to construct VLS-PV plants at a rate that would enable the world to freeze its conventional generating capacity at its present level. As we have seen, the world would need to generate an additional 610 TWh/year, each year. If we take a 60% capacity factor as being representative of modern fossil-burning power plants, then failure to implement a renewable energy program will, in all probability, require the world to install conventional plants at a rate not less than 116 GW/year. Since the cost of conventional generating capacity is around the US$1 per watt mark, we may infer that the world will in any event be prepared to pay US$116 B per year for new generating capacity. This then is the cost goal that VLS-PV will need to reach if conventional market drivers will be prepared to finance it. For the purpose of comparison, we have already seen that 3 B square meters of conventional PV panels could perform this task. At present-day costs of US$670 per square meter of panel area, for large PV systems [7], we are talking about a global infrastructure cost of US$ 2 T per year. This is already a factor of 17 larger than the expected cost of continued conventional growth, without including the addition of storage, transmission, and (probably rising) real estate costs. By contrast, Raviv [1] has addressed the large-scale introduction of VLS-PV from the direction of CPV technology. His basic argument is that CPV cells enable units that could operate at a minimum of 25% system efficiency to be constructed, at a cost of less than $1 per Wp (excluding the storage costs, which Raviv does include in his plan). The reason that such low costs are possible is that the increased cell efficiency, coupled to the three-orders-of-magnitude reduction in PV materials

COST REQUIREMENTS AND FINANCING

487

that are brought about under very high solar concentration levels (500–1000X), reduces the expensive PV material to an almost insignificant part of the total system cost. In a series of previous publications [1, 3, 4, 8, 9], Raviv’s scheme was applied to a number of individual country case studies. For each case, it was assumed that VLS-PV plants at the GWp scale would be constructed on an annual basis, the electricity would be sold at whatever the going rate is for that country, and the steadily increasing revenues from the plants would be used to pay off the initial infrastructure costs and eventually to fully fund the continued construction of such plants. Roughly speaking, a 1-GWp VLS-PV plant would cost approximately $1 B and, at a nominal tariff of 10¢ per kilowatt hour, it would generate $200 M in annual revenues. Therefore, ignoring for a moment all of the additional costs, once five VLS-PV plants are constructed, their combined annual revenue should suffice to fund the construction of each of the subsequent plants. The fact that CPV technology can enable VLS-PV plants to be cost-effective leads to two unexpected benefits, referred to as “Type 2” and “Type 3” sustainability. The Type 2 stage is reached when all accrued costs are paid off (bank loans, interest, factory, etc.). At this stage, the electricity tariff can be dropped to something like US5¢ per kilowatt hour and the revenue from the existing VLS-PV plants is still sufficient to continue the annual plant construction program without the need for any further external financing. Type 3 sustainability is achieved when, after the construction of a sequence of 29 plants, it becomes necessary to construct two new plants per year thereafter: one to continue the ever-increasing demand for electricity and the other to replace a 30-year-old VLS-PV plant, which, by assumption, would by then have reached the end of its useful working life. Type 3 sustainability generally requires a slight adjustment of the electricity tariff in year 30 (downward in the case of California [8]) in order to allow the continued annual construction of two VLS-PV plants without the need for any external financing. These economic forecasts for CPV were based on an assumed STC cell efficiency of 32% (25% in the field) and an annual VLS-PV plant output of 2 TWh from 12 km2 of land. Many of the world’s deserts have annual DNIs higher than the 2222 kWh m2/year assumed for Raviv’s calculations [1]. This fact, together with steadily improving CPV cell efficiencies (the 41% mark having recently been exceeded [10]), means that in all probability, VLS-PV plants will occupy considerably less land than 12 km2/GWp. To generate 610 TWh/year (i.e., to freeze the continued worldwide growth of fossil fuel consumption for electricity production), it would be necessary to construct 305 GWp/year of VLS-PV plants. If we assume a nominal electricity selling price of US10¢ per kilowatt hour and 5% real interest rate, then the necessary credit line would reach a maximum value of US$1.91 T in year 12 (a figure that compares most favorably with the US$1.39 T the world would in any event have spent on new conventional generating capacity, excluding the additional fuel costs). From then onward, the credit line would decrease and would become fully paid off in year 20. At this stage, there would be 4880 GWp of installed CPV capacity, and the electricity selling price could be reduced to US4.04¢ per kilowatt hour for Type

488

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

2 sustainability. These plants would be generating 9760 B kilowatt hours of electricity per year, that is, roughly 30% of the entire global projected generation at that time. At the end of year 35, the first set of VLS-PV plants would be 29 years old, and it would become necessary from then on to replace this set as well as to construct a new set each year. That is to say, it would be necessary to construct twice as many plants per year after the initial 35 years. This would require a slight upward adjustment of the electricity selling price—to US4.28¢ per kilowatt hour—in order to achieve Type 3 sustainability. Note also that at year 36, there would be 10,980 GWp of installed solar capacity, generating 21,960 TWh annually, which is about 50% of the world’s projected total electrical generation at that time. All this from infrastructure that had not cost anything and without the need for any fuel. This argument demonstrates that cost should not be an issue. Admittedly, a 20-year payback period is not very attractive for venture capital, but there are many investment portfolios, such as pension funds, that seek long-term secure investments. Clearly, electricity-generating infrastructure that does not depend on the availability or price of fuel constitutes an excellent investment for such purposes.

21.4.2

Feed-In Tariffs

A number of European countries have passed laws that require utilities to purchase PV energy from clients who wish to sell it—at rates that are greatly in excess of the amount it would cost the utility to generate the electricity itself via conventional means. The purpose of such so-called feed-in tariffs is to encourage the public to purchase PV systems that would not otherwise be a cost-effective investment. Spain is a good example, where PV electricity may be sold to utilities for approximately 50¢ per kilowatt hour. To illustrate in a quantitative way the manner in which such feed-in tariffs could help promote a more rapid introduction of a VSL-PV program, it is instructive to take the specific example of Spain and to examine two extreme case scenarios. The first is where a VLS-PV program is introduced and the electricity is sold at the minimum possible tariff that would ensure Type 3 sustainability. The second scenario is where the electricity is sold at the 50¢ per kilowatt hour “feedin” tariff. Because Spain’s electricity production has been growing at an annual rate of 10.6 TWh/year [6], it would be necessary to install 5.3 GWp/year in order to enable the country to freeze its fossil fuel consumption for electricity production. The minimum tariff that could enable this to be carried out would be 6.01¢ per kilowatt hour. Under this scenario, the entire cost of the program would be returned in year 34 (the final year in the effective life of VLS-PV plant no. 1). At that stage, the tariff could be lowered to 4.31¢ per kilowatt hour for the subsequent annual construction of 10.6 GWp (i.e., two sets of VLS-PV plants) of solar generating capacity each year—without the need for any external funding.

TECHNOLOGICAL CHALLENGES

489

Now under this first scenario, it would be necessary to establish a credit line that would reach a maximum of $66.3 B in year 20, before falling to zero in year 34. Clearly, a larger starting tariff would enable the project costs to be returned in a shorter time span. For example, a 10¢ per kilowatt hour starting tariff would result in the required credit line reaching a maximum of only $34.3 B, in year 12, before falling to zero in year 20. However, most interesting of all would be the situation of a 50¢ per kilowatt hour starting tariff (via “feed-in” legislation). Under this scenario, the maximum credit line needed for such a program would be only $12.0 B, which would be reached in year 6, and which would drop to zero by year 8 of the program. At that stage, the cost of continuing to construct 5.3 GWp of VLS-PV plants per year could be fully borne by an electricity tariff of 14.92¢ per kilowatt hour (a very reasonable tariff by today’s European standards). If the feed-in tariff were to be discontinued at this stage and the electricity sold at the new rate, then Type 2 sustainability would continue until year 35 of the program, at which point it would be possible to lower the tariff to 4.31¢ per kilowatt hour for subsequent Type 3 sustainability. One sees, therefore, that even though CPV technology when implemented on the GW per year scale, as envisaged by Raviv [1], can be fully cost-effective without the need for subsidies, the entire program could be speeded up by a factor of approximately 4 via the skillful employment of feed-in tariffs on a scale that is already in practice in Europe.

21.5

TECHNOLOGICAL CHALLENGES

21.5.1

Storage

Raviv’s original VLS-PV scheme recognized the need for storage in order to guarantee dispatchability of the CPV-generated power, and indeed, a certain amount of storage was incorporated into his program. Specifically, his suggested prescription was to incorporate 2 GWh of storage for each GWp of solar generating capacity. However, a moment’s thought suffices to enable one to appreciate that this amount of storage must be significantly increased in order truly to guarantee all-day electrical availability from a VLS-PV plant. This can be seen as follows. At a typical desert latitude of 30°, the shortest day of the year has 10 hours of daylight and 14 hours of night. If such a day (or sequence of days) were to be fully overcast, it would be necessary to charge the batteries with nighttime fossilgenerated energy. It would therefore require 14 GWh of storage capacity (at a nominal 70% round-trip efficiency) in order for “the plant” to be able to guarantee the delivery of 10 GWh of daytime energy: that is, 1 GW of power during the entire 10-hour day. On the other hand, in mid-summer there are only 10 hours available for nighttime battery charging and a daytime power requirement for 14 hours. Therefore,

490

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

in the event that the following day were to be fully overcast, a 1 GWp VLS-PV plant would need a 2 GW battery in order to absorb 10 hours of nighttime charge and, assuming the same battery efficiency, to deliver 14 GWh during the following day. This argument is obviously a gross over-simplification because totally overcast days are a rarity in deserts in mid-summer. This means that a 20 GWh battery would seem to be greatly oversized for summer usage. Furthermore, a fully charged battery would be unable to accept solar input the following day. However, this simplified argument does indicate that somewhere between 14 and 20 GWh of storage per GWp of generating capacity would be needed to ensure full solar dispatchability under all weather conditions. This is an order of magnitude larger than the 2 GWh in the original Raviv model. We shall return to this matter again when storage is discussed in greater detail below. Naturally, the strategy for using such large storage would need to be configured to take into account weather forecasts and electricity tariffs that vary throughout the day. But it could provide the utility with the opportunity of using part of its nighttime base-line generation more efficiently than is generally possible without large amounts of storage. Therefore, the total cost of such a storage facility should not simply be added to the cost of its associated VLS-PV plant. Thus far, the discussion of storage obviously implies the need for low-cost, high-efficiency, long-life batteries. All of these properties will require further research and development (R&D). But there is another aspect of storage that has not hitherto received much attention and that will also require intensive R&D. It turns out that the inability of conventional electricity-generating plants to switch on and off on the same rapid time scale at which clouds obscure and permit solar irradiation limits the useful size of VLS-PV plants if large amounts of solargenerated energy are not to be wasted [11–15]. As an example, Solomon [12] and Solomon et al. [13–14] studied the compatibility of hourly solar irradiance data from various parts of the Negev Desert with the corresponding load requirements of the Israel electricity grid for the year 2006. They simulated, inter alia, the output of a fixed flat-panel PV system and found that if the Israeli grid could have had a flexibility of 100% (i.e., perfect ability to ramp up and down in synchronization with its varying PV input), then a VLS-PV system as large as 5.4 GWp could have inputted all of its generated energy that year without the need to dump any surplus, and without the need for storage. Figure 21.2 shows the hourly output of such a 5.4 GWp VLS-PV plant (dotted line), together with the corresponding output of the IEC (continuous line) during that week (March 20–26) in which the “no-dump” hour would have occurred, that is, when solar output would have precisely met the total grid requirements (on March 25) without the former having exceeded the latter at any other hour in the year. Such a VLS-PV plant would have made a substantial contribution—17.4% of the annual energy needs—given that the total, fossil-fueled, generating capacity of the grid was 10.5 GW that year. However, the same IEC data indicated that the actual grid flexibility during 2006 was only 65%. This means that only the difference between a day’s total load

492

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

allowing approximately 10% of its total output to be dumped rather than used by the grid [12, 13]. But this is about as far as one can go without the use of storage. However, as first pointed out by Denholm and Margolis [11] and elaborated in greater detail by Solomon [12] and Solomon et al. [15], in addition to the energy capacity of storage one must also take into consideration its power capacity, namely the rate at which it can be charged and discharged. In the case of the Israeli grid, Solomon et al. [15] found that by employing appropriately sized storage with optimally chosen properties, and changing the strategy for operating the various conventional plants (including those of the base load variety!) it could be possible to increase the contribution of VLS-PV plants to approximately 60% of the annual grid requirements. Today, most utilities naturally operate their various plants in a manner that balances security of supply and profitability. With the expected advent of cost-competitive solar plants, the detailed strategy for grid management may be expected to change, particularly if conventional fuel sources begin to dwindle. The above discussion implies that special storage systems may need to be developed if VLS-PV plants of the CPV variety are to play an important role in large-scale power production. This is because of the intermittent nature of their power output. Because even fast-ramping turbines have ramp-down times of order 10–15 minutes, relatively sudden bursts of PV power would not lead to any fuel saving unless appropriately fast-acting storage were to be available. Supercapacitors could have the necessary speed for absorbing the energy, but only on a one-time basis. Depending upon the frequency of such cloud events, the supercapacitor may not have sufficient time to trickle its charge into the grid before being called upon again in order to absorb a second transient charge of energy. With appropriate meteorological data (i.e., DNI measurements on a fast time scale, and taken over a range of distances comparable to the size of a VLS-PV plant) it would be possible to assess the frequency of such events and to ignore them if the energy involved were to be a small fraction of the total annual solar availability. But such measurements have yet to be made. Furthermore, all of the simulations discussed above were performed with data on an hourly basis. Therefore, in addition to meteorological data on an appropriately fast time scale, it will be necessary to study grid output data with a time resolution that enables plant ramp-up and ramp-down properties to be quantified with greater detail.

21.5.2

Transmission

Long distance AC power transmission is conventionally carried out at high voltages in order to reduce joule losses as much as possible. On the other hand, it is not feasible to raise the voltage too far because corona discharges to the ground may dominate the joule losses. Typical voltages employed today for such purposes, depending on country, are accordingly found in the range 115–765 kV. As a simple estimate of the joule losses to be expected from the wide-scale implementation of a VLS-PV program, consider a line of 1000 km in length with an AC resistance (for 1272 AWG [16]) of 0.054 Ω/km. Suppose 1216 A (the

TECHNOLOGICAL CHALLENGES

493

maximum current in this manufacturer’s catalog) is injected at one end of the line at a voltage of 765 kV, which corresponds to an injected power of 930 MW, roughly the peak output of a 1-GWp VLS-PV plant. The joule loss would accordingly be 80 MW, or nearly 9%. On the other hand, a recently published announcement about an ultrahighvoltage 800-kV DC power line, of 2000-km length, that will deliver 6400 MW of power at a loss less than 7% [17] implies an effective line resistance less than 7 Ω. From this, one may infer that for such a line, operating at a similarly high DC voltage, the loss in power from a 930 MW VLS-PV plant over a distance of 1000 km would be less than 1%. Joule losses can, in principle, be reduced to zero via the employment of superconducting cables. However, at present, even so-called high-temperature superconductors require cryogenic cooling to liquid nitrogen temperatures. There is thus a parasitic energy requirement for liquefying the appropriate quantities of nitrogen and maintaining the liquid at approximately −200°C. Therefore, unless truly ambient-temperature superconductors are discovered, UHVDC will probably be the technology of choice for transmitting the power from VLS-PV plants over long distances. These, of course, are only the joule losses. Other losses are to be expected from the various electronics, connections, switches, and so on, needed to get the power into and out of the lines. Nevertheless, the above estimates would seem to indicate that UHVDC technology is sufficiently advanced to allow low-loss power transmission over distances comparable to those necessary to connect the Sahara to Europe, or the deserts in the SW United States to the population centers in the NE.

21.5.3

Smart Grids

The possibility of long distance, low-loss electrical transmission can, in principle, solve more problems than merely shunting power from deserts to large population centers. This is because such a grid would encompass regions having considerably different weather conditions and load requirement patterns. With an appropriately enlarged control system, load sharing could occur on a scale of complexity that was never possible in the past. To a certain extent, this would also alleviate the dispatchability problem because different cloud conditions at different locations would smooth out the combined solar inputs to such a grid from widely located VLS-PV plants. Weather forecasting would obviously also play an important role. In addition, each 15° of longitude between VLS-PV plants would extend solar availability by 1 h. Thus, the continental United States and the Sahara desert would each provide a “solar day” extended by several hours. The random on/off output of a CPV plant on cloudy days bears a certain similarity to the uncertainty in output from a wind plant. But the former is more manageable. This can be seen as follows. On a cloudless day, the output of a CPV plant depends almost entirely on geometric factors (time of year, time of day,

494

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

geographic latitude). Clouds diminish the clear day pattern in a somewhat random manner. However, because, unlike the wind, clouds are visible and their velocities measurable, an appropriately automated observation system associated with each VLS-PV plant could provide advance warning to a smart grid about the expected performance of each plant during the coming hour or so [4]. Hence, the increasingly popular term “smart grid” would ultimately include input from “smart PV plants,” and probably too, “smart storage systems.”

21.6

CONCLUSIONS

In this chapter, the argument has been made that CPV technology is the key to a massive replacement of fossil fuel on a worldwide scale. High cell efficiencies make it less land intensive than the other solar technologies and, because of high levels of optical concentration (typically around 1000X), only relatively small amounts of costly PV material are needed. These two factors hold out the hope that VLS-PV plants of the CPV variety could be economically competitive with fossil-powered plants. It was shown that the world’s deserts, specifically the Sahara and the SW deserts of the United States, are more than large enough to encompass VLS-PV plants of a scale necessary to halt the present growth rate of fossil-fueled plants in Europe and in the United States, and similar results can be shown for the rapidly developing economies of China and India [9] via the Gobi and Thar deserts, respectively. Cost estimates based on the Raviv model [1] indicate that the economics of such a program are favorable even without the need for subsidies. Moreover, the timescale of investments required is of a kind that should be attractive to pension funds and other portfolios that require longer-term solidity. Storage continues to a be an area in which further R & D is needed— particularly high-speed storage for smoothing out the rapid on/off effect that passing clouds would have on a CPV plant. However, the urgent need for efficient storage is alleviated to a certain extent by the possibility of combining a VLS-PV program with a UHVDC smart grid system. Therefore, in conclusion, the technology is ripe for a cost-effective VLS-PV program to go ahead. In the United States, China, and India, such a program could commence at once, because each country enjoys a single political system. In Europe/Africa, a number of international agreements would need to precede the implementation of such a program.

ABBREVIATIONS AC—alternating current AWG—American wire gauge B—billion CPV—concentrator photovoltaics DC—direct current

REFERENCES

495

DNI—direct normal irradiance GWp—gigawatt peak IEC—Israel Electric Corporation M—million MWp—megawatt peak NE—northeast PV—photovoltaic or solar cell R & D—research and development ST—solar thermal STC—standard test condition SW—southwest T—tera (1000 billion or 1 trillion) UHVDC—ultrahigh-voltage direct current VLS-PV—very large scale photovoltaic Wp—watt peak

REFERENCES [1]

[2] [3] [4] [5] [6] [7] [8] [9] [10]

D. Raviv and R. Rosenstreich. Powering Europe from the sun: The scenario of largescale centralized CPV power stations. In 19th European Photovoltaic Solar Energy Conference, W. Hoffmann, J.-L. Bal, H. Ossenbrink, et al., eds., pp. 3745–3750. Munich: WIP and Florence: ETA (2004). K. Kurokawa, ed. Energy from the Desert: Feasibility of Very Large Scale Photovoltaic Power Generation (VLS-PV) Systems. London, James & James (2003). K. Kurokawa, K. Komoto, P. van der Vleuten, et al., eds. Energy from the Desert: Practical Proposals for Very Large Scale Photovoltaic Systems. London, Earthscan (2007). K. Komoto, M. Ito, P. van der Vleuten, et al. eds. Energy from the Desert: Very Large Scale Photovoltaic Systems for Socio-Economic Development. London: Earthscan (2009). K. Zweibel, J. Mason, and V. Fthenakis. A solar grand plan. Scientific American January: 48–57 (2008). BP Statistical Review for 2008. Available at http://www.bp.com/productlanding.do? categoryId=6929&contentId=7044622 (2008). L. M. Moore and H. N. Post. Five years of operating experience at a large, utilityscale photovoltaic generating plant. Progress in Photovoltaics: Research and Applications 16, 249–259 (2008). D. Faiman, D. Raviv, and R. Rosenstreich. Using solar energy to arrest the increasing rate of fossil-fuel consumption: The southwestern states of the USA as case studies. Energy Policy 35, 567–576 (2007). D. Faiman. Implementing VLS-PV on a worldwide scale, in a cost-effective manner. In Renewable Energy 2006 Proceedings, October 9–13, 2006, Chiba, Japan, CD-ROM, pp. 63–68. K. Schneider. World record 41.1% efficiency reached for multi-junction solar cells at Fraunhofer ISE. http://www.ise.fraunhofer.de/press-and-media/press-releases (January 14, 2009).

496

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

[11]

P. Denholm and R. M. Margolis. Evaluating the limits of solar photovoltaics (PV) in traditional electric power systems. Energy Policy 35, 2852–2861 (2007). A. A. Solomon. Matching the intermittent output of renewable energy systems to the needs of an electricity grid: A mathematical study with important policy implications. PhD thesis, Ben-Gurion University of the Negev, Beersheba, Israel (2010). A. A. Solomon, D. Faiman, and G. Meron. An energy-based evaluation of the matching possibilities of very large photovoltaic plants to the electricity grid: Israel as a case study. Energy Policy (2010), doi: 10.1016/j.npol.2009.12.024. A. A. Solomon, D. Faiman, and G. Meron. The effects on grid matching and ramping requirements, of single and distributed PV systems employing various fixed and sun-tracking technologies. Energy Policy (2010), doi: 10.1016/j.npol.2010.02.056. A. A. Solomon, D. Faiman, and G. Meron. Properties and uses of storage for enhancing the grid penetration of very large photovoltaic systems. Energy Policy (2010), doi: 10.1016/j.npol.2010.05.006. Phelps Dodge Intl. Corp. On-line spec sheets available at http:www.pdic.com. T. F. Armistead. Longest, most powerful line will apply new technology. http://enr. construction.com/news/powerindus/archives/080109d.asp (January 9, 2008).

[12] [13] [14] [15] [16] [17]

PART VI THIN FILMS AND X-RAY IMAGER TECHNOLOGIES

22 MARKET OVERVIEW OF FLAT PANEL DETECTORS FOR X-RAY IMAGING CARL LACASCE, LARRY PARTAIN, AND CHUCK BLOUIR Varian Medical Systems

22.1

INTRODUCTION

Thin-film solar cells for terrestrial electric power production are of interest for their potential lower cost, but they have other characteristics that make them of value for other applications. They can be deposited on large-area glass substrates and devices can be integrally interconnected. These traits are valuable for medical imaging devices. While thin-film semiconductors have large electronic defect levels limiting their efficiencies for energy conversion applications, this trait makes them tolerant to X-ray exposures for medical X-ray imaging. This and the next three chapters address the application of advanced thin film solar cell technology to the field of flat panel digital electronic X-ray imaging. The present chapter provides an introductory overview from a market perspective that is quite positive. Chapter 23 covers the device physics of the dominant amorphous silicon semiconductor thin film technology (on glass substrates) that enables largearea flat plate electronics at much lower costs than traditional crystalline semiconductors. This includes the p+-i-n+ structures required for solar cell behavior plus the alternate n+-i-n+ configurations (plus field effect electrodes) that provide the TFTs essential for imaging. Chapter 24 describes the placement of scintillator thin films on the solar cell thin films that convert X-ray photons into the visible light photons detected by solar cells. It further details the separation of the solar cell into the millions of “photodiodes” laid out in rows and columns on glass panels that give X-ray images as each photodiode “pixel” is electronically switched Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

499

500

MARKET OVERVIEW OF FLAT PANEL DETECTORS

3,500,000

Market Size ($000)

3,000,000

Mammo

2,500,000

Dynamic 2,000,000

RAD

1,500,000 1,000,000

CR

500,000 0 2007

2008

2009

2010

2011

2012

2013

Year

Figure 22.1. The wholesale flat panel market.

by its TFT into a digital readout amplifier. It covers the critical low-noise performance essential for high-quality imaging. Finally, Chapter 25 describes the replacement of both the scintillator and the photodiodes in the X-ray imager with photoconductor films, dense and thick enough to directly absorb X-ray photons in a manner that provides potentially higher X-ray image quality at even lower costs. Medical X-ray imaging, much like the photography industry, has been undergoing a transition: a transition from analog-based technologies to solid-state digital technologies. Since the earliest days of X-ray, analog devices such as film and image intensifiers were used to convert X-ray information into a visual media. As the demands on hospital resources have increased, the need for more effective imaging devices has also increased. Hospitals require cost-effective imaging equipment capable of handling higher patient throughput, easy integration with modern picture archive solutions (PACS), and high reliability (Fig. 22.1). While human diagnostic imaging needs remain the main catalyst for digital detector design and evolution, other X-ray-based imaging markets have also been monitoring and transitioning to their use. These markets include veterinary imaging, security and inspection systems, and NDT systems. Essentially, all film-based imaging modalities are either in the transition phase or in the investigation phase to flat panel digital imaging technology. The worldwide FPD market has grown dramatically since the late 1990s. In 2009, the market was estimated to be more than $2 billion per year and is growing at more than 15% per year. Medical imaging professionals believe this trend will continue as the performance of FPDs improve, FPD costs are reduced and new applications are created.

MARKET HISTORY: FILM TO DR

22.2

501

MARKET HISTORY: FILM TO DR

The transition from film to digital imaging began in the human diagnostic imaging departments of many hospitals. The first technology was called CR and was introduced in 1983 by Fuji Photo Film Co. CR utilizes a photostimulable phosphor composed of europium-activated barium fluorohalide or europium-activated barium fluoride bromide. One of these mixtures is coated on a screen similar to a standard intensifying screen. This screen is then placed in a special cassette holder that functions like a standard film/screen cassette. The appeal of CR technology is ease of integration as CR works with standard X-ray imaging equipment. The downside to CR technology has been the lack of any significant workflow improvement in the radiology department. Cassettes still have to be manually processed similar to the conventional film cassette process. So, the patient still has to wait in the procedure room until after the technologist can verify that the proper imaging has been obtained. FPDs date back to the early 1970s when scientists at Xerox PARC began development work on amorphous silicon. Over the next 20 years, Varian Medical Systems, Xerox, GE Company, Thompson, and many other companies invested in the technologies that are utilized in modern FPDs. This technology was dubbed DR, indicating its ability to directly capture and output an image without manual operator intervention. By the late 1990s, many companies had established commercial enterprises for the manufacture of FPDs for DR applications, but the digital imaging industry remained largely dominated by CR technology until the early 2000s. Amorphous selenium was the basis for the first full-size DR image detectors offering a compact footprint, high resolution images, and fast image display. Amorphous selenium detectors were referred to as “direct” capture detectors because the X-ray photons that hit the selenium layer were directly converted to electrons for storage. Unfortunately, amorphous selenium also required very high bias voltage on the photoconductor and the selenium layer was very sensitive to temperature and humidity, all of which lead to premature failure of the detectors and poor reliability. With the exception of breast imaging, amorphous selenium detectors have been largely replaced today in most markets by the amorphous silicon detectors sold by all the major OEMs. In today’s marketplace the most common implementation of DR detector technology is based on large amorphous silicon photodiode arrays. These arrays are made up of several million TFTs coupled to photodiodes which store charge generated by light produced by X-rays absorbed in the conventional gadolinium oxysulfide or cesium iodide scintillation layers. This design is referred to as an “indirect” capture device because X-ray photons are converted to visible light photons by the scintillator before being converted to electrons for storage. Indirect capture DR products are used almost exclusively by all major diagnostic imaging OEMs and system integrators. This design provides instant imaging with excellent diagnostic quality and little operator intervention. Wait times for patients is dramatically reduced while the workflow in the radiology department is streamlined.

502

MARKET OVERVIEW OF FLAT PANEL DETECTORS

General radiography has been most impacted by the use of flat panels. Dose reduction, gains in patient throughput, and ease of use are just some of the benefits realized in radiography. The latest developments include battery-powered detectors with wireless image transmission capability. The first to announce a wireless capable detector was Trixell in 2006. The latest and most revolutionary development in this market segment was introduced at the RSNA 2008 by Carestream Health. Carestream Health’s introduction of a 14 × 17 in. (36 × 43 cm) wireless detector in a package the same size as a conventional film cassette will solve many of the problematic mechanical upgrade issues generally associated with upgrading existing film- or CR-based radiographic equipment to digital technology. In the 1990s, digital fluoroscopy systems began to appear in hospitals. Early digital systems were based on traditional image intensifiers coupled to analog then to CCD cameras and digital imaging workstations. Fluoroscopic studies as well and digital photo spot images were captured in the workstation and could be viewed on monitors, saved to other media such as videotapes, or could be used to create prints or films. This technology quickly became the industry standard, and most new systems marketed by all major OEMs included digital imaging. Following the movement of film-based general radiographic imaging to FPD technology, the next evolution for DR was to address the unique needs of fluoroscopy. The development of readout electronics capable of scanning the detector at up to 30 fps facilitated the use of these new compact imaging devices in modalities requiring live fluoroscopic imaging. While image quality was acceptable with image intensifier-based products, their huge footprint and short life span made them ripe for replacement with DR’s compact new amorphous silicon photodiode technology. 22.3

APPLICATIONS

Conventional diagnostic imaging applications such as radiography, fluoroscopy, angiography, cardiology, and urology now utilize DR technology. DR has also facilitated the development of several new imaging capabilities. Starting with NDT using slow acquisition rates and inanimate objects and rapidly progressing to realtime high-speed imaging of human anatomy, CBCT has made it possible to image a tumor during radiation treatment in order to more accurately plan and target the treatment beam. Smaller field-of-view detectors with CBCT capability have revolutionized the market for dental implants, making it possible to precisely place dental implants around critical nerves in the jaw. Ongoing research is moving these specialized DR detectors ever closer to performing actual diagnostic quality CT as well as many other imaging tasks such as cargo container inspections, bomb disposal, and food inspection, just to name a few. 22.3.1

Cardiovascular

FPDs for use in cardiovascular applications were introduced in early 2000 by GE. Doctors quickly recognized the benefits: larger field of view, high resolution, and

APPLICATIONS

503

better patient positioning as compared to image intensifier-based systems. The market quickly expanded the use of flat panels as Philips, Siemens, and Toshiba introduced flat panels into their product lines. By 2008, all the major medical OEMs had removed image intensifiers from their cardiovascular products.

22.3.2

Angiography and Other Applications

Angiography and other specialized applications followed closely behind cardiology in the transition from image intensifier-based imaging to FPDs. As the field of view, image acquisition speeds, and overall image quality capabilities increased, OEMs began adopting flat panels for use on standard C-arm-based angiography systems and on general radiographic and other diagnostic imaging systems. Many procedures such as peripheral run-off studies were dramatically simplified with the use of square or rectangular detectors, and overall patient access has dramatically improved along with a much longer life expectancy (Figs. 22.2 and 22.3). One other huge benefit of amorphous silicon flat panel image detectors over their image intensifier-based predecessors is the elimination of the need to constantly increase the radiation dose over time to maintain image quality.

Figure 22.2. Interventional R/F table system (Hitachi).

504

MARKET OVERVIEW OF FLAT PANEL DETECTORS

Figure 22.3. Multipurpose C-arm R/F table for angiography and GI (Hitachi).

Figure 22.4. Dental cone beam CT system, Imaging Sciences International.

APPLICATIONS

22.3.3

505

Dental CBCT

Flat panels have virtually created a new standard of care within the dental imaging community. Cone beam CT systems dedicated for use in dental implants are now in use throughout the world. Dentists are now able to image their patients in-office, use dedicated software customized for implant procedures, and, in some parts of the world, can bill for a CT exam (Fig. 22.4).

22.3.4

Mammography

Many modalities have attempted to address the huge volume of diagnostic images required to screen for breast cancer globally. While X-ray film-based mammography systems still dominate this procedure, flat panel-based systems are gaining market share. First introduced in early 2000 digital mammography systems have evolved rapidly, and new study results affirm that digital imaging speeds the screening process and makes it possible to use computer-aided diagnosis to reduce the tremendous workload this procedure places on radiologists. Currently, both amorphous silicon and amorphous selenium detectors are used in this modality.

22.3.5

Medical CBCT

CBCT remains in its infancy but is gaining momentum as a diagnostic tool for whole body imaging. Several systems based on mobile C-arms are showing promise along with a unique O-arm system sold by Medtronic. The O-arm system sets the standard currently for flat panel-based CBCT during surgery. Several hurdles remain to be addressed before FPDs can perform at ultrahigh speeds such as that required for diagnostic CT of the heart, but image quality is excellent for the less dynamic areas of the body (Fig. 22.5).

Figure 22.5. O-arm cone beam CT system (Medtronic).

506

MARKET OVERVIEW OF FLAT PANEL DETECTORS

Figure 22.6. Varian Oncology Systems treatment platform with IGRT.

22.3.6

Portal Imaging

Portal imaging started a revolution in oncologic imaging during radiation treatment. By providing the ability to create an image from the exit beam during treatment, oncologists were able to begin the process of seeing what they were treating during the treatment. Unfortunately, due to the megavoltage X-ray beam used in this procedure, image quality was poor. Rapid evolution, however, quickly resulted in a low-energy X-ray source and detector being mounted orthogonally to the treatment beam. This revolution, called OBI, brought near diagnostic quality imaging to the oncologist for tumor position verification during treatment. Ongoing improvements to this new technology, now termed IGRT, have made FPDs an integral part of, and the standard of care on, all new radiation treatment systems (Fig. 22.6).

22.4

FLAT PANEL MARKET EVOLUTION

The early market landscape included many companies who were either developing the basic technology or finding ways to enhance and productize core technologies

INDUSTRY LANDSCAPE

507

designed by others. Some designs attempted to tile large numbers of CMOS detectors together to form a large detector array, but major assembly and reliability problems proved too difficult to overcome. Physically large detectors incorporating one or more CCD cameras along with an elaborate lens system appeared on the market from several companies, but this design was quickly dubbed an interim technology and sales rapidly declined. The modern FPD market is dominated by amorphous silicon photodiodebased detector designs for all modalities except for mammography. The high resolution and DQE of amorphous selenium at mammography energies have become more desirable for breast imaging, but these detectors continue to be plagued by poor reliability. GE continues to successfully market an amorphous silicon photodiode-based detector for mammography with excellent image quality and reliability.

22.5

INDUSTRY LANDSCAPE

Although not a major player today, Analogic has been a long-time participant in the digital imaging system business. In June 1999, Analogic purchased the medical detector and pure metal assets of Noranda, marking its entry into the flat panel market with amorphous selenium-based radiographic detector technology. In today’s marketplace, there are some smaller companies marketing selenium-based products, but Hologic’s products designed for mammography make up the largest portion of systems based on this photoconductor technology. In April 1999, Canon formed Canon Medical Systems by combining their radiography business and a floundering PACS operation that they owned. Canon is focused on amorphous silicon panels for radiographic applications, and currently, Canon has several panel sizes on the market. Their approach has been to provide both the DR detector and the image processing software designed specifically to maximize the image quality of their panels. Canon panels use a Schottky barrier photodiode with slower speed operation than the pin photodiodes used by most other flat panel manufacturers. Thus, the Canon panels are only capable of conventional radiographic imaging. GE, along with their partner and supplier PerkinElmer, has been one of the top suppliers of DR-based diagnostic imaging products for the human medical marketplace. GE’s Revolution XQ/I dedicated chest system was first introduced at RSNA 1998. In March of 2000, GE also received FDA clearance to market its Senographe 2000D digital mammography system and their Innova 2000 flat panel-based cardiac system. GE followed these products with many other systems to date, and they remain one of the top OEMs using DR technology. Kodak entered the CR market in 1994 with the Ektascan Imagelink. Kodak sold off their healthcare imaging division, which became Carestream Health in 2006. Carestream Health is focused on digital imaging systems that include CR and DR technologies. Carestream Health recently revolutionized the general radiographic market with the introduction of a film cassette-sized wireless DR detector.

508

MARKET OVERVIEW OF FLAT PANEL DETECTORS

In March 1997, Philips Medical Systems, Siemens Medical Systems, and Thomson Tube & Electroniques (now Thales) joined forces in a consortium to develop flat panel products by creating a company called Trixell S.A.S. Trixell would be the source of DR technology for most products sold by Siemens and Philips. Varian Medical Systems was the first to market in the late 1990s with amorphous silicon photodiode-based detectors capable of real-time imaging for fluoroscopy. This was followed closely by products from GE and then by Trixell. Varian remains the largest independent supplier of FPD products in the world. Varian also pioneered CBCT technology using FPDs and remains the leader in this emerging diagnostic imaging application.

22.6

MARKET SUMMARY

Over the past 10 years, FPDs have revolutionized the radiology department in the modern hospital. The radiology department is more efficient through higher patient throughput; the dose per given X-ray is lower than with analog technology; images can be easily transmitted electronically to radiologists at remote locations; and the doctors have more imaging information available to them via advance applications. During the same period of time, the costs of FPDs have reduced and, as with all technology-based products, will continue to see price erosion as more and more players enter the market and as overall product volumes grow. Some new Schottky barrier photodiode designs allow the detector arrays to be built alongside the mega-volume LCD displays built for flat screen TVs and for computer monitors. These evolutions will drive costs down while allowing suppliers and OEMs to evolve the technology and at the same time return acceptable margins to their shareholders. The opportunity for smaller hospitals and healthcare facilities in third world countries to purchase FPD-based equipment will increase, making it very possible that all X-ray images in the future will be done with digital FPDs.

22.7

FUTURE TRENDS

The costs of digital technology, including large-area flat panel displays and flat panel X-ray imagers, continually decrease as their capabilities dramatically increase. The resulting growth with new applications is so large that the dollar size of the total world wholesale X-ray flat panel market, presently approaching $2.5 billion, continues to rise attractively at double-digit rates. This is somewhat analogous to Moore’s law like trajectory for digital computers that play an integral role in this whole imaging technology base. As with computers, the underlying imaging technology must continually evolve to keep pace on such a technology advancement treadmill. While the standard now is higher-speed pin photodiodes covered

ABBREVIATIONS

509

with scintillators, simpler and lower-cost Schottky barrier technology is beginning to replace the pin devices in the lower-performance, lower-speed, but larger-area radiology segment of this market. CR is a comparatively mature technology in terms of further cost reductions and improved performance, and its market share is expected to decrease continually. The highest-resolution performance is currently provided commercially with amorphous selenium-based photoconductor imagers. However, laboratory examples are now being reported for photoconductors of even higher sensitivity and more robust characteristics and with a potential for much lower costs (Chapter 25). In addition, low-cost, nonphotolithographic fabrication of laboratory prototypes with modest spatial resolution is emerging even on flexible plastic substrates (Chapter 23). Only time, the market, and continuous technical advancements will determine the winning technology champions of the future. ABBREVIATIONS C-Arm—support structure shaped like a “C” CBCT—cone beam computed tomography CCD—charge-coupled device CMOS—complementary metal oxide semiconductor CR—computed radiography CT—computed tomography DQE—detective quantum efficiency DR—digital radiography FDA—U.S. Federal Drug Administration FPD—flat panel detector fps—frames per second GE—General Electric GI—gastrointestinal i—undoped or intrinsic semiconductor region IGRT—image-guided radiation therapy LCD—liquid crystal display n+—heavily negatively doped semiconductor region NDT—nondestructive testing O-arm—support structure shaped like an “O” OBI—on-board imaging OEM—original equipment manufacturer p+—heavily positively doped semiconductor region PARC—Palo Alto Research Center PACS—picture archival and communications system RAD—general radiographic X-ray imaging market R/F—radiography and fluoroscopy RSNA—Radiological Society of North America TFT—thin-film transistor

23 AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES ROBERT STREET Palo Alto Research Center

23.1

INTRODUCTION

The interest in a-Si started as a basic research activity aimed at understanding how atomic disorder in solid-state materials has affected the quantum electronic properties of a semiconductor [1] and has grown into a $100 B industry making displays, solar cells, and X-ray imagers. Solar cells were the first application [2], and this industry has grown slowly but steadily since about 1980 and is perhaps now at a cusp in the growth with the current renewed interest in renewable energy. The display application of a-Si TFTs grew more rapidly to the present $100 B LCD business. Digital X-ray imagers are a smaller market than either displays or solar cells but are driving a technology revolution, replacing X-ray film [3]. The common property of these applications is the ability to fabricate TFTs and p-i-n photodiodes in a way that can be scaled to a very large area. This chapter discusses the electronic and material properties of a-Si and how these determine the structure and performance of the devices, and what issues limit their performance. 23.2

PROPERTIES OF a-Si

a-Si and the various alloys and doped layers that form the technology are deposited in a PECVD reactor [4]. Silane, SiH4, is the usual source gas, but there are a wide variety of other gases for the different compounds. The energy provided to the plasma dissociates the gases, and the resulting chemical radicals deposit on the surface and form the film. The growth surface and the resulting film are shown schematically in Figure 23.1. As a result of the hydride gas sources, the a-Si film contains about 10 atomic % of hydrogen (Fig. 23.1), which proves to be critical Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

511

512

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES a-Si growth surface H

H

H

H

H

H2 H

H

H H

H

H

H

H

H

H

H H H

H H

H H

H

H

H

H H

Log(density of states) (cm–3eV –1)

Figure 23.1. Left: illustration of the surface of an a-Si film during PECVD growth, showing the attachment of various Si-H radicals. Right: illustration of the bulk atomic structure of a-Si showing the atomic disorder and the bonded hydrogen.

valence band

mobility edge

1022 1020

extended states

1018 1016

band tail localized states

conduction band

extended states

defect states 0

1 Energy (eV)

2

Figure 23.2. The density of states of amorphous silicon showing the conduction and valence bands with extended state conduction, the localized states of the band tails, and the dangling bond defect states near the middle of the bandgap. The mobility edge divides extended from localized states.

for the electronic properties but which is also the source of the device instability problems that are discussed below. One of the main reasons that a-Si is in largescale production for TFT arrays is that the PECVD process is scalable to a very large area, and the deposition process also provides the doped contacts needed for TFTs, solar cells, and photodiodes, along with the dielectrics needed for TFT backplanes.

23.2.1

Electronic Properties of a-Si

The amorphous atomic structure has a direct effect on the electronic properties of a-Si. Figure 23.2 shows the electronic density of states, which is divided into three

PROPERTIES OF a-Si

513

distinct regions [1, 5]. The conduction and valence bands contain extended electron and hole states allowing for electronic conduction. The atomic disorder and consequent scattering reduces the carrier mobility compared to crystalline silicon by ∼100X to a value of about 10 cm2/Vs. At the edge of the bands are the band tails comprising localized states, which also originate from the atomic disorder and can easily trap mobile carriers. As a result, carrier transport is by a trapping and release process between the band tails and the extended states of the bands, and transport occurs near the mobility edge (see Fig. 23.2), which is the energy separating these two groups of states. The valence band tail turns out to be wider than the conduction band tail—a result largely related to the particular bonding properties of silicon—so that the effective electron mobility is about 1 cm2/Vs and the hole mobility is two to three orders of magnitude lower [5]. Therefore, any device that needs high mobility is restricted to electron transport. Within the bandgap are defect states identified as silicon dangling bonds— silicon atoms with three rather than four bonds. Good quality undoped a-Si has less than 1016 cm−3 defects, which is sufficiently low to allow good electronic transport and photoconductivity properties. The low defect density is the main consequence of the hydrogen that is retained in the film, without which the defect density would be at least 1000 times higher. As we shall see, the defect density is rather easily increased, either by strong illumination to the detriment of solar cells or by charge induced in the bands, which affects the transistor stability. The design of devices and circuits must take these instabilities into account. The density of states distribution in Figure 23.2 is known reasonably accurately for the valence and conduction bands, but the shape of the defect band is known only approximately. a-Si is readily doped by the addition to the plasma deposition of a gaseous hydride containing either a group III (i.e., boron) or group V (i.e., phosphorus) during the plasma growth [6]. The temperature dependence of the conductivity in heavily doped a-Si is illustrated in Figure 23.3 and indicates that there is a range of conductivity values depending on the deposition conditions and the treatment of the films. The increase in conductivity by doping is constrained by the band tails, which prevents the Fermi energy from moving close to the band edge. The wider valence band tail makes p-type doping less effective than n-type. The temperature dependence of the conductivity is thermally activated and reflects the energy separation of the Fermi energy from the mobility edge, although the activation energy decreases at lower temperature; near room temperature, the energy is 0.1–0.2 eV for electrons and ∼0.3 eV for holes. Doping also increases the defect density substantially [5] by a mechanism related to the solar cell instability (see Section 2.2). The result is that doping is effective in moving the Fermi energy, but the conductivity is modest compared to crystalline silicon, and the minority carrier lifetime of doped a-Si is poor because of recombination with the extra defects. The electronic properties of the doped film determine the design of solar cells and photodiodes, as is discussed later. The high defect density of the doped film is useful as it enables the formation of thin tunnel junctions in a-Si tandem cells.

514

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

1

Conductivity (Ω–1cm–1)

n-type 1% PH3

10–2

p-type 1% B2H6

–4

10

10–6 2

3

4

5

Temperature 1000/T (K)

Figure 23.3. Conductivity of heavily doped n-type and p-type a-Si, corresponding to a deposition gas with 1% dopant during the PECVD deposition of material. The range of values resulting from different growth conditions and sample treatment is indicated.

While mobility is the most important parameter for a TFT, the mobilitylifetime product, μτ, determines the performance of the solar cell or photodiode, because the distance that mobile change travels before deep trapping is μτE, where E is the electric field. For a charge, Q0, induced near one side of a thin a-Si film, in a uniform electric field, the fraction of the charge that is collected by the external circuit is given by [7] Q Q0 = (μτV d 2 )[1 − exp ( − d 2 μτV )] ,

(23.1)

where d is the film thickness and V is the voltage. Figure 23.4 shows measurements of the electron and hole μτ product for doped and undoped a-Si [8]. The different μτ values for the undoped material reflect different defect densities ranging from about 3 × 1015 cm−3 for the upper right data in Figure 23.4 to 1018 cm−3 for the lower left. The value of μτ in undoped a-Si is larger for electrons than for holes, which is important for the design of photodiodes and solar cells. The μτ values for doped samples reflect the increasing defect density with doping and that only minority carriers are trapped.

23.2.2

Hydrogen and Defect Stability

One of the important limitations of a-Si is the instability of the a-Si material to the formation of additional dangling bond defects. Both solar cells/photodiodes and

PROPERTIES OF a-Si

515

Figure 23.4. Electron and hole mobility-lifetime products measured on doped and undoped a-Si. The diagonal data correspond to undoped a-Si with a range of defect densities. The doped samples have a range of doping levels [8].

1018 Saturated value

A44

NS (cm–3)

1017

HI

MI

LI

1016

1015 10–5

10–3

10–1 101 Time (hr)

103

Figure 23.5. Defect creation kinetics resulting from prolonged illumination at different intensities: high (HI), medium (MI), and low (LI), showing the power law creation kinetics leading to an eventual steady state [10].

TFTs are affected, and the defect creation imposes constraints on the design and performance of the devices. The metastable defects in solar cells [9] are created by the recombination of electron–hole pairs. Figure 23.5 shows the time dependence of the defect density at different illumination levels [10]. The defect density increases by 100X to about 1017 cm−3, where it then tends to stabilize. The increase

516

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

in defect density is sublinear in time, roughly following a t1/3 dependence, and also sublinear in illumination intensity G. Hence, the defect density increases quickly at the start of exposure and then progressively more slowly. Stutzmann et al. [11] proposed that defects are created by the bimolecular recombination of electrons and holes in the bands, while the dominant recombination channel was by monomolecular recombination through the defect states in the gap. This model leads to a defect creation rate given by N D = const G 2 3t 1 3,

(23.2)

which fits the data well [11]. According to this model, the energy released by the electron–hole recombination breaks silicon bonds and creates the dangling bond defects. The hydrogen contained within the a-Si film is closely involved in the process—either a Si–H bond is broken or the hydrogen stabilizes the bond breaking of a Si–Si bond. Metastable defects are removed by annealing at 150–200°C, with an activation energy of ∼1.5 eV that corresponds to the hydrogen diffusion energy. A similar defect creation mechanism applies to TFTs where the source of the energy is the charge induced into the conduction band when the TFT is turned on. To summarize, undoped a-Si has a low density of deep gap states and has a carrier mobility of about 1 cm2/Vs for electrons and considerably lower for holes. n-type doped a-Si has room temperature conductivity near 0.1 S/cm, with p-type 10–100 times lower, and both have a much higher density of deep recombination centers than undoped a-Si. Additional defects are induced in undoped material by strong illumination or by charge accumulation. The TFTs and photodiodes discussed next are designed specifically to take into account these properties. 23.3

a-Si TFT

The first a-Si TFTs were reported by Spear’s group at the University of Dundee [12]. Along with the semiconductor layer, a TFT requires a low leakage gate dielectric and preferably doped ohmic source and drain contacts. The discovery of doping in a-Si provided the ohmic contact material and the PECVD deposition of silicon nitride provided the gate dielectric, so that the different layers of the TFT can be made by the same deposition process. The TFT operates in the n-channel accumulation mode. n-channel operation is used because electrons have a much higher carrier mobility than holes, and the high defect density of doped a-Si:H means that the channel must be undoped and therefore operating in accumulation mode. 23.3.1

TFT Structure and Fabrication

Figure 23.6 shows the structure of the two most common alternative a-Si TFT designs. Both have a bottom-gate structure and the same silicon nitride dielectric

518

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

ohmic contact but are invariably coated with a metal electrode because the n+ layer by itself has a very high sheet resistance. Both processes result in the source and drain contacts overlapping the TFT channel, which is necessary to ensure low-resistance contacts, but provides some parasitic capacitance to the gate. The parasitic capacitance has a detrimental effect on the speed and performance of TFT arrays and on the noise of image sensor arrays. Self-aligned TFTs minimize the capacitance and have been demonstrated [13] but are not currently used in general display manufacturing. Partial selfalignment of the island etch TFT can be carried out using a lithography step to pattern the passivation layer, with exposure through the substrate using the gate as the mask.

23.3.2

TFT Characteristics

The TFT transfer characteristics, shown in Figure 23.7, contain three regions— above threshold, subthreshold, and the off-current. The TFT is turned on at positive gate voltages above the threshold voltage of typically 1–2 V; the leakage current when the TFT is turned off at negative gate voltages is usually below 1 pA and can be much lower. The subthreshold region in between describes the transition between the on and off states. In the subthreshold region, the current increases exponentially due to the movement of the Fermi energy through the localized states. The subthreshold slope is typically ∼0.5 V/decade.

Source-drain current (A)

10–6 10–8 W/L = 2 Vds = 5 V

10–10 10–12

above threshold

off 10–14 10–16 –10

subthreshold

0

10

20

Gate Voltage (V)

Figure 23.7. Example of the transfer characteristics for an island etch amorphous silicon TFT.

a-Si TFT

519

The TFT ON characteristics are described by the usual MOS transistor relations [14]. In the linear regime, at a small source–drain voltage, VDS, the current, IDS, is given by I DS = CG μ FE (VG − VT − VDS 2)VDSW L ,

(23.3)

where VG is the gate voltage, VT is the threshold voltage, W is the TFT width, and L is the length. A 300-nm nitride dielectric has a gate capacitance, CG, of 5 × 10−8 F/ cm2, and a field effect mobility, μFE, of 0.8 cm2/Vs yields a current of 6 μA in a device with W/L = 10 operating at 15 V above threshold. At a high VDS, the current saturates at I SAT = CG μ FE (VG − VT ) W 2 L . 2

(23.4)

The transfer characteristics of an a-Si TFT are shown in Figure 23.7, for the linear regime. The minimum leakage current is typically when VG is about −5 V, and the small off-current gives an on/off ratio of 106–108, which is sufficient for an LCD backplane or an image sensor. The n-type contacts are essential to provide a barrier to hole conduction at negative gate bias. The leakage current may be due to bulk thermal generation current in the channel or due to injection through the barrier formed by the n-type source and drain contacts. Generally, the leakage is lower in the island etch structure compared with the BCE structure, possibly because the island protects the channel from plasma and etching damage.

23.3.3

The Bias Stress Effect in a-Si TFTs

a-Si TFTs have a threshold voltage instability that is closely related to solar cell degradation [15]. Figure 23.8 illustrates the change in TFT transfer characteristics when the device has been turned on for a period of time. The threshold voltage increases with time, with a corresponding decrease in the current, but the mobility and other parameters are essentially unchanged. The change in VT depends on gate voltage and time according to the following description [16]: ΔVT = const. t α VGγ ;

α ∼ 0.33, γ ∼ 2 − 4.

(23.5)

Measurement of the VT shift is usually done to the linear regime largely because the main application of TFTs is active matrix backplanes, which operate in the linear regime. The VT shift is smaller in saturation [17] primarily because the channel charge is lower, and it is the channel charge that determines the VT shift. There are two mechanisms contributing to the VT shift in a-Si [18]. The gate dielectric, usually silicon nitride, contains traps into which electrons can tunnel

520

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES Ids (A) 10–4 10–5

Initial

10–6

After stress e.g. Vg = 20V for 2 hours

10–7 10–8 10–9 10–10

Vds = 20V

10–11 –5

0

5

10

15

20

Vqs (V)

Figure 23.8. Threshold voltage shift of an a-Si:H TFT as a result of extended gate bias stress at a positive gate voltage. The threshold voltage shift is due to defect generated in the channel.

from the channel. The second mechanism is charge creation in the a-Si layer, and it is this process that is related to defect creation in a solar cell. The energy associated with an electron occupying a conduction band state (in the solar cell, it is the energy of an electron–hole pair) induces a bond-breaking process and creates dangling bond defects. The process is catalyzed by the presence of hydrogen, but the exact reaction has never been made completely clear. The two bias stress mechanisms both follow Equation 23.5, although with different parameters, α and γ, and so are not very easy to distinguish without careful analysis. Tunneling into the gate dielectric tends to be a stronger function of gate voltage and hence predominates at a high gate bias. In a good quality a-Si TFT of the usual structure made for displays, defect creation typically dominates up to a gate voltage of about 20 V and tunneling into the dielectric at higher gate voltage. The threshold shift is not too much of a problem for image sensors and displays because the TFTs are turned on with a low duty cycle, and it is sufficient to design the circuit so that it can accommodate a VT shift of perhaps 2–4 V over the life of the device. However, the stress effect makes it difficult to use a-Si TFTs for analog devices and when the gate bias is applied continuously for a long time.

PHOTODIODES

23.4

521

PHOTODIODES

The a-Si:H p-i-n photodiode is used as the light detector in X-ray image sensor arrays because of its high quantum efficiency and low reverse-bias leakage current. Low leakage is important because the illumination level is very weak and high signal to noise is critical. The same p-i-n structure is used for a solar cell where the illumination is many orders of magnitude larger. The photodiode structure is illustrated in Figure 23.9 and usually consists of 100 nm–1 μm of undoped a-Si between much thinner p- and n-doped layers. The doped layers provide rectifying contacts but do not contribute to the light sensitivity because doping causes a high density of charged dangling bond defects in a-Si:H as discussed earlier. The minority carrier lifetime in doped a-Si:H is therefore so small that most photogenerated carriers recombine in the doped layers before they reach the i-layer. On the other hand, the high defect density also results in a narrow depletion width, so that a doped layer thickness of 10 nm or less is sufficient to form the junction. The barrier height for the p-i-n device is larger than for a metal Schottky diode [19] even with high work function metals such as Pd or Pt, and so makes a better photodiode with a lower reverse-bias leakage current. Since charge collection of holes is less efficient than for electrons (see Fig. 23.4), the photodiode operates best when the p-doped layer faces the illumination. Electron–hole pairs are preferentially created near the illuminated surface and so holes have a smaller distance to travel for collection. The p-i-n diode must be fabricated with metal contacts on both top and bottom because the thin doped layers are not sufficiently conductive by themselves. A 10-nm-thick p-type a-Si:H layer has a sheet resistance of ∼1011 Ω per square, and the more conductive n-type layer is ∼108 Ω per square. An ITO transparent conductor is generally used for the illuminated surface, and any convenient metal serves as the reflecting back contact. The photocurrent, IPC, is given by I PC = β ηQE G e,

(23.6)

illumination

ITO transparent contact p-doped layer ~10 nm undoped a-Si ~1 μm n-doped layer ~10 nm metal contact

Figure 23.9. Structure of the p-i-n photodiode showing the thin n+ and p+ contacts, the thicker undoped layer, the transparent top electrode, and bottom metal contact. Illumination is through the top surface.

522

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

where G is the incident photon flux, ηQE, is the internal quantum efficiency, and β is a factor containing other loss mechanisms described below. Unlike the n-i-n diode, which can have photoconductive gain, the internal quantum efficiency of the reverse-bias p-i-n diode cannot exceed unity because there is no significant charge injection. ηQE approaches close to unity when the charge collection length, μτE, is much larger than the thickness, d, as described by Equation 23.1. This condition applies in a p-i-n diode with low-defect-density a-Si and a moderate applied field. A typical external quantum efficiency spectrum is shown in Figure 23.10 and is governed by several loss factors. The decrease at long wavelength is due to the rapidly decreasing optical absorption coefficient of a-Si:H below the bandgap energy. The short wavelength decrease mostly arises from absorption in the upper doped layer, along with reflection and absorption in the transparent metal conductor. The peak quantum efficiency of 80–95% occurs in the range 500–600 nm, which is suitable for the X-ray imager, because the usual X-ray converter phosphor emits at ∼550 nm. The peak quantum efficiency can be increased by using lowabsorption doped layers and by designing the transparent metal as an antireflection layer.

23.4.1

Comparison of p-i-n Solar Cells and Photodiodes

The a-Si solar cell and the photodiode are similar devices, both being p-i-n structures. However, there are design differences that reflect the different operating conditions of the two devices. The solar cell is made substantially thinner than the

1.0 Quantum efficiency

Solar cell 0.8 0.6

Photodiode

0.4 0.2 0 400

500

600

700

800

Wavelength (nm)

Figure 23.10. Typical spectral response of a photodiode showing sensitivity across the visible region. The maximum quantum efficiency is limited by reflection at the ITO contact, absorption in the p-layer, and incomplete charge collection in the bulk.

PHOTODIODES

523

photodiode (typically 100–300 nm vs. 1 μm), primarily to allow for the high density of light-induced defects induced by the illumination, which reduces the value of μτ. The solar cell operates with a smaller built-in potential because it is in slight forward bias, also making charge collection more difficult. Hence, both the solar cell and the photodiode need a high-quality i-layer to allow high charge collection, but optimization of the charge collection is difficult in the solar cell. High quantum efficiency is important for both devices but is more critical for the solar cell. Figure 23.10 indicates that the solar cell has been engineered to have a better spectral response than the typical photodiode by texturing the film to increase optical absorption, optimizing the contacts for low absorption, and often grading the i-layer bandgap and doping for increased charge collection. The p- and n-doped contact layers are typically either microcrystalline or a wide bandgap alloy in the solar cell to provide the low optical absorption while also minimizing the series resistance when a large current flows. The X-ray imager photodiode passes a much smaller current (picoampere vs. milliampere) so the highly conductive contacts are less critical, but optimizing the p/i junction for very low reverse leakage is more important, and this is one reason why the photodiode is typically about 1 μm thick. One additional source of leakage current in the photodiode is down the sidewall of the sensor, which therefore has to be well passivated, particularly for small sensors (∼100 × 100 μm) where edge effects can dominate. The effect is not significant in a solar cell because the surface area of the device is large.

23.4.2

Diode Characteristics

The forward and reverse characteristics of the p-i-n diode are illustrated in Figure 23.11. The forward current follows the expected exponential current–voltage relation I = I 0 exp ( eV nkT ) .

(23.7)

The ideality factor n is about 1.5. The ideality factor is closer to unity in an a-Si metal Schottky diode [19], reflecting a thermionic emission process. The larger junction barrier of the p-i-n diode reduces thermionic emission and causes the generation–recombination mechanism to dominate, which gives a larger ideality factor and is also consistent with the reverse current mechanism, as discussed below. In reverse bias, a depletion layer forms with width given by WD = [ 2εε oVBI eN DI ] , 12

(23.8)

where NI is the density of ionized gap states, assumed to be constant for convenience. When the deep states are broadly distributed across the middle of the bandgap, depletion only occurs down to the middle of the bandgap because deeper

524

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES 1 10–2

Sensor Current (A/cm2)

Dark Forward Bias 10–4 Light Reverse Bias

10–6 10–8 10–10

Dark Reverse Bias

10–12 0

1

2

3

Voltage (V)

Figure 23.11. a-Si p-i-n diode characteristics showing the exponential diode region in forward bias followed by a current limited by series resistance. The dark reverse current is limited by thermal generation.

states are filled from the valence band faster than the electrons are emitted to the conduction band. This mechanism of thermal generation creates electron–hole pairs, which are the origin of the reverse leakage current [20]. Since a-Si:H is weakly n-type, the depletion layer forms from the p-type contact by the emission of electrons. Measurements find an ionized state density for undoped a-Si:H of about 5 × 1014 cm−3, which gives a depletion layer of about 1 μm for a voltage of 1 V. Hence, for a 1-μm photodiode with a more typical reverse bias of ∼5 V, the depletion width is larger than the film thickness and the internal field is approximately uniform. The reverse leakage current in the photodiode is about ∼10−11 A/cm2 (see Fig. 23.11) and is thermally activated with an energy of about 0.9 eV, consistent with thermal generation from the middle of the bandgap. Figure 23.12 shows that the leakage current decays slowly after the application of a reverse-bias voltage, over a timescale of minutes at room temperature. The transient component arises from the depletion of charge from the deep states as the depletion layer is formed, and the steady-state current, JSS, is reached when the rate of emission of electrons and holes from deep traps is equal. The thermal generation mechanism gives a steadystate current density [20], I SS = NkT ω 0 exp [ − ΔE kT ] ξ ( F ) d ,

(23.9)

PHOTODIODES

525

Figure 23.12. Transient dark current of a p-i-n photodiode showing the slow decay to a steady state over a timescale of minutes [21].

where N is the density of states near midgap, ΔE is half the bandgap energy, ω0 is ∼1012 s−1, ξ(F) represents the weak dependence on the electric field, and d is the sample thickness. Both the transient current decay and the steady-state reverse-bias thermal generation current are directly related to the defect density and are consistent with other measurements of the defect distribution of a-Si:H. Contact injection and edge leakage are other contributions to the leakage current in some photodiodes [21]. Contact injection occurs when electrons tunnel across the p–i interface leading to an increase in the leakage current, particularly at high applied bias and in samples with a high defect density. The exposed periphery of the photodiode provides a different source of leakage current, which is particularly significant in small devices where the ratio of edge to area is large. In the presence of both bulk and edge components, the total current in a square diode of edge length dS, is given by I total = I B dS2 + 4 I P dS ,

(23.10)

where IB cm−2 and IP cm−1 are the bulk and peripheral conductance. There is considerable variability in the leakage current when the bias is large, due to these mechanisms.

526

23.5

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

TFTs AND PHOTODIODES IN SENSOR ARRAYS

The important attributes for sensors and TFTs in an X-ray detector are radiation hardness, response linearity, low image lag, and low electronic noise. a-Si turns out to be a radiation hard material, and exposures of up to 50 Mrad of gammas do not seriously degrade the photodiode or TFT performance [22, 23]. Part of the explanation is that the devices are thin and silicon has low atomic number, and so the interaction with high energy radiation is weak. Radiation does increase the defect density, but the effects are slow to appear in view of the significant intrinsic defect density in the material. Response linearity and image lag in the photodiode are closely related. An ideal p-i-n diode has unity gain and hence a perfectly linear response. However, incomplete charge transfer of the generated electrons and holes reduces the gain and can cause a departure from linearity. The trapped charge is typically released at a later time when it can create ghosts in subsequent images, the effect known as image lag. Charge trapping is governed by the mobility-lifetime product according to Equation 23.1, and when trapping is weak, the fraction of charge trapped is approximately FT = d 2 2μτV ,

(23.11)

assuming a uniform electric field. Trapping is therefore minimized in thin samples with a high bias voltage and large values of μτ. In a good quality a-Si:H photodiode of 1 μm thick with 5-V bias, 1–3% of the electrons are trapped and 5–10% holes, because of their smaller value of μτ. As noted above, charge trapping, and therefore image lag, is lowest when the sensor has light incident on the p-doped layer because holes have a shorter distance to travel to the contact. The nonlinearity in the response of an X-ray detector comes about from the way the detector operates. The bias voltage across the sensor decreases as the light flux and exposure time increase because the sensor charge is stored on the capacitance of the diode until it is read out. From Equation 23.11, the charge trapping increases as the sensor approaches voltage saturation, and hence the amount of charge trapping increases with the illumination. The collected signal charge when there is an illumination intensity, G, can be expressed as a function of voltage and time: t

QS (V , t ) = ∫ [1 − FT (V )] G dt .

(23.12)

0

Figure 23.13 shows the response of an a-Si X-ray detector under radiographic (single shot) and fluoroscopic (continuous measurement) mode [24]. The difference in the mode response is that fluoroscopic measurements are not subject to image lag because the charge trapping is exactly offset by charge release from previous frames. The radiographic mode data are subject to charge trapping and show increasing nonlinearity of response as saturation is approached. A fit to

NEW DEVELOPMENTS IN a-Si DEVICES

527

5 Fluoroscopic mode

QPIX (pC)

4

VBIAS = - 6V

Radiographic mode

3 VBIAS = - 4V

2 VBIAS = - 2V

1 0

0

200 400 600 800 1000 Incident Light Signal (# LED flashes)

Figure 23.13. The measured response of an amorphous silicon X-ray detector in radiographic and fluoroscopic modes and at different bias voltages [24].

Equation 23.12 gives a value of FT of 2–4% at 5-V bias, corresponding to an average value of μτ of ∼10−7 cm2/V. Several sources of electronic noise occur in an X-ray detector and from the associated readout electronics. Some of the noise sources are the materials and devices of the backplane, and some are external to the array. From the point of view of the pixel circuit, the most important are the resistance noise in the pixel TFT and shot noise in the photodiode current. The current noise is generally negligible in sensors at low bias voltage because of the very low leakage currents. Thermal noise of the TFT ON resistance generates the familiar kTC noise component, which is proportional to the pixel capacitance. Additional noise arises from 1/f noise in the TFT, the magnitude of which is sensitive to the device material and fabrication.

23.6

NEW DEVELOPMENTS IN a-Si DEVICES

23.6.1

TFTs and Sensors on Flexible Substrates

The ability to make a-Si TFTs and image sensors on flexible substrates is a recent development [25, 26], but a-Si solar cells on flexible substrates have been available for several years [27, 28]. The possible advantages for an X-ray detector include a more robust device, since breakage of the glass substrates is a significant practical issue. A curved sensor might be interesting for dental X-ray or for computed tomography detectors. Also, the fabrication of a sensor array on a thin X-ray transparent substrate opens the possibility to image X-rays that are incident from

528

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

Figure 23.14. Photograph of an amorphous silicon TFT backplane fabricated on a plastic substrate.

the substrate side of the detector, which enhances the spatial resolution when used with a phosphor converter in the indirect detection mode. Either plastics or metal foils are potential flexible substrates for TFT backplanes. Plastics have the advantages of being lighter, of being insulating rather than conducting, and of generally having a better surface quality. The challenge is that many of the most suitable plastics cannot be heated above 200°C, while the standard a-Si TFT deposition occurs at up to 350°C, and p-i-n sensors up to 250°C. The effort to make low-temperature a-Si devices compatible with plastic substrates has been quite successful [25, 29]. The TFT characteristics are almost unchanged compared to the high-temperature growth. The low-temperature deposition does increase the defect density, but this effect is minimized by hydrogen dilution of the deposition gas and other changes in the deposition conditions. The increased defect density causes a small (∼1 V) increase in threshold voltage, and about a 20% increase in the bias stress threshold voltage shift. Low-temperature p-i-n photodiodes also have similar characteristics to the high-temperature devices. The increased defect density affects both the charge collection and the reverse-bias leakage current. Since both of these parameters are critical to the performance of an X-ray imager, further optimization of the growth and device configuration is probably still needed. Figure 23.14 shows an image of an a-Si TFT backplane fabricated on PEN plastic [26]. 23.6.2

Digital Lithography Processing of TFTs and Sensor Arrays

Much of the expense in the fabrication a-Si TFT detector arrays is in the photolithography processes that are used to pattern the device. Photolithography involves

530

23.7

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

SUMMARY

a-Si devices form the basis of three significant technologies: displays, solar cells, and X-ray image sensors. These applications share a requirement of large area fabrication, which is enabled by the PECVD method of depositing a-Si, along with the various doped layers, alloys, and dielectrics. The properties of a-Si follow from its noncrystalline atomic structure and from the small amount of hydrogen that is incorporated during deposition. The magnitude of the mobility, the defect and doping properties, and the stability of the devices can all be related back to these two factors. Once the material properties were understood, remarkably good devices could be made, which suit the applications well. Displays need the high dynamic range and low threshold voltage of the TFT. Solar cells benefit from the ability to engineer the bandgap by alloying, and the tunnel junction that is formed at the p–n interface, but suffer significantly from the instability induced by the hydrogen. Photodiodes have high gain and high dynamic range. ACKNOWLEDGMENTS Many scientists, both past and present, have contributed to the development of a-Si technology at the Palo Alto Research Center, supported by funding from PARC, Xerox, and several government agencies. The contributions of these people and organizations are gratefully appreciated. It has also been a pleasure to work with several external collaborators in the basic understanding and the various applications of a-Si. ABBREVIATIONS a-Si—amorphous silicon B—billion BCE—back-channel etch C—capacitance CG —gate capacitance of a TFT d—semiconductor film thickness dS —edge length of a square diode e—electronic charge E—electric field strength f—frequency FT —fraction of charge trapped in a-Si bandgap states G—photon illumination intensity H—hydrogen i—undoped or intrinsic semiconductor I—current IB —bulk current conductance per square centimeter IDS —source–drain current of a TFT

REFERENCES

531

III—elements in the third column of the periodic table IP—peripheral current conductance per centimeter IPC —photocurrent ISAT —saturation current of a TFT ISS —diode reverse-bias saturation current ITO—indium tin oxide k—Boltzmann’s factor L—length of a TFT LCD—liquid crystal display Mrad—mega (million) rads where rad is a unit of x-radiation dose of ionization energy MOS—metal oxide semiconductor transistor configuration n—negatively doped semiconductor n—diode ideality factor N—density of trapping states near midgap ND—light-generated defect density in the bandgap NI —density of ionized gap states in depletion region of a diode p—positively doped semiconductor PECVD—plasma-enhanced chemical vapor deposition PEN—polyethylene naphthalate Q—charge magnitude Si—silicon TFT—thin-film transistor t—time T—absolute temperature V—elements in the fifth column number of the periodic table VBI—built-in barrier height voltage of semiconductor diode VDS —source–drain voltage of a TFT VG—gate voltage of a TFT VT —threshold gate voltage of a TFT W—width of a TFT WD—width of a diode depletion layer β—photodiode loss mechanisms factor ΔE—energy of a-Si trap from the band edge ε—relative dielectric constant of a semiconductor εo—dielectric constant of free space ηQE—internal quantum efficiency of a photodiode τ—electron or hole lifetime in a semiconductor μ—electron or hole mobility in a semiconductor μFE—field effect mobility ω0—attempt to escape frequency for transitions from bandgap trap states REFERENCES [1]

N. F. Mott and E. A. Davis. Electronic Processes in Non-Crystalline Solids. Oxford, Clarendon Press (1979).

532 [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

[15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30]

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES D. E. Carlson and C. R. Wronski. Appl. Phys. Lett. 28, 671 (1976). R. A. Street. Large area image sensor arrays. In Technology and Applications of Amorphous Silicon, R. A. Street, ed., pp. 147–221. Berlin, Springer-Verlag (2000). J. C. Knights and G. Lucovsky. CRC Crit. Rev. Solid State Mater. Sci. 21, 211 (1981). R. A. Street. Phil. Mag. B63, 1343 (1991). W. E. Spear and P. G. LeComber. Solid State Comm. 17, 1193 (1975). K. Hecht. Z. Phys. 77, 235 (1932). R. A. Street, J. Zesch, and M. J. Thompson. Appl. Phys. Lett. 43, 672 (1983). D. L. Staebler and C. R. Wronski. Appl. Phys. Lett. 31, 292 (1977). H. R. Park, J. Z. Liu, and S. Wagner. Appl. Phys. Lett. 55, 2658 (1989). M. Stutzmann, W. E. Jackson, and C. C. Tsai. Phys. Rev. B32, 23 (1985). A. J. Snell, K. D. Mackenzie, W. E. Spear, P. G. LeComber, and J. A. Hughes. Appl. Phys. 24, 357 (1981). Y. Kuo. a-Si TFT structures. In Thin Film Transistors; Materials and Processes, Volume 1, Chapter 4, Y. Kuo, ed. Dordrecht, Kluwer (2004). A. Nathan, P. Servati, K. S. Karim, D. Striakhilev, and A. Sazonov. Device physics compact modeling and circuit applications of a-Si TFTs. In Thin Film Transistors; Materials and Processes, Volume 1, Chapter 3, Y Kuo, ed. Dordrecht, Kluwer (2004). M. J. Powell. IEEE Trans. Electron Devices 36(12), 2753 (1989). T. Tsukada. Active matrix liquid crystal displays. In Technology and Applications of Amorphous Silicon, R. A. Street, ed., pp. 7–93. Berlin, Springer-Verlag (2000). K. S. Karim, A. Nathan, M. Hack, and W. I. Milne. IEEE Electron Device Lett. 24(9), 583–585 (2004). C. van Berkel and M. J. Powell. Appl. Phys. Lett. 51(14), 1094–1096 (1987). M. J. Thompson, N. M. Johnson, R. J. Nemanich, and C. C. Tsai. Appl. Phys. Lett. 39, 274 (1981). R. A. Street. Appl. Phys. Lett. 57, 1334 (1990). R. A. Street. Hydrogenated Amorphous Silicon. Cambridge, Cambridge University Press (1991). I. D. French, A. J. Snell, P. G. LeComber, and J. H. Stephan. Appl. Phys. A31, 19–22 (1983). J. M. Boudry and L. E. Antonuk. Med. Phys. 23, 743 (1996). L. E. Antonuk, Y. El-Mohri, J. H. Siewerdsen, J. Yorkston, W. Huang, V. E. Scarpine, and R. A. Street. Med. Phys. 24, 51 (1997). H. Gleskova and S. Wagner. IEEE Trans. Electron Devices 48, 1667 (2001). T. N. Ng, R. A. Lujan, S. Sambandan, R. A. Street, S. Limb, and W. S. Wong. Appl. Phys. Lett. 91, 063505 (2007). S. Guha. Multijunction solar cells and modules. In Technology and Applications of Amorphous Silicon, R. A. Street, ed., pp. 252–305. Berlin, Springer-Verlag (2000). F. R. Jeffery, D. P. Grimmer, S. Brayman, B. Scandrett, and M. Noak. PVMaT improvements in monolithic a-Si modules on continuous polymer substrates. AIP Conf. Proc. 394, 451 (1996). W. S. Wong, R. Lujan, J. H. Daniel, and S. J. Limb. J. Non-Cryst. Solids 352, 1981 (2006). W. S. Wong, S. E. Ready, J. P. Lu, and R. A. Street. Electron Device Lett. 24, 577 (2003).

24 AMORPHOUS SILICON DIGITAL X-RAY IMAGING RICHARD COLBETH Varian Medical Systems

24.1

WHY a-Si?

a-Si X-ray detectors for medical, dental, veterinary, and nondestructive test applications have seen explosive growth, with the annual number of units nearly doubling in each of the last 5 years. This growth has resulted from two decades of research and refinement by numerous academic and industrial organizations. This chapter explores the design issues unique to large-area X-ray sensors and answers the question, why a-Si? From the beginning, a-Si technology has had a number of compelling advantages that still elude competing technologies. First and foremost, a-Si is the technology of large-area electronics, from solar cells to flat panel displays to X-ray detectors. The lack of an effective lens for X-rays requires an X-ray detector to be at least the size of the object of interest. Because of image magnification, shown schematically in Figure 24.1, it is often required that the detector capture an image projection that is double the object size in medical procedures. Today, it is common to find a-Si X-ray detectors for chest radiography that have a 43 × 43 cm (17 × 17 in.) active area. When research into a-Si X-ray imagers began in the late 1980s, work on a-Si for solar cells and displays had been in progress for decades. The idea then, as now, was to build on the huge investments in flat panel display production capacity, fabrication tools, and basic research. Because of the large and growing flat panel industry, a-Si detectors are cost-effective and are becoming more so with the increasing annual volume of both detectors and displays. a-Si devices offer unique advantages not readily available in other X-ray detectors. Because there is very little long-range order, a-Si is not easily damaged Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

533

DETECTOR CONSTRUCTION

535

eliminating sources of image distortion. With the advent of a-Si detector technology, X-ray imaging was able to take an immediate leap forward in DQE, resolution, small area contrast and uniformity.

24.2

DETECTOR CONSTRUCTION

A picture of a 14 × 17″ portable radiography detector is shown in Figure 24.2 (Varian Medical Systems 4336R). This particular FPD has the same footprint and form factor as the traditional 14 × 17″ X-ray film cassette, and so provides a dropin upgrade to digital. The detector front cover is made of a low X-ray attenuation carbon fiber. X-rays exiting the patient are captured and converted to digital data, which are then transmitted over a high-speed link to a workstation. In this particular case, the imager link is Gigabit Ethernet, but fiber optics, LVDS, and Cameralink connections are also common. In addition to transmitting the digital image data, the Gigabit Ethernet link also delivers all the control data to the receptor. Figure 24.3 shows the internal construction typical of the indirect FPDs. Here, indirect refers to the fact that X-rays deposit energy in a scintillator screen generating visible photons, which are in turn captured and converted to electron– hole pairs in the photodiode array. In contrast, a direct FPD converts the X-ray photon energy directly to electron–hole pairs. The relative advantages of these approaches will be discussed in a later section and in Chapter 25. Typically, the

Figure 24.2. Radiographic flat panel detector (Paxscan 4336R). The active area is approximately 43 × 36 cm with 139-μm pixel pitch. The form factor matches the standard X-ray film cassette, including the thickness of 15.5 mm.

536

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

Figure 24.3. Typical internal architecture for the indirect-type FPD.

scintillator screens are made from GOS or CsI(Tl). With minor modification, both GOS and CsI were adapted from the technologies that FPDs were targeted to replace. GOS “intensifying” screens have been in use with X-ray film for a very long time, and CsI is a key component in IITs, which are the front-end capture device in traditional X-ray video imaging. The core of the detector is an a-Si photodiode array. Like the TFT arrays used in flat panel displays, the photodiode array is fabricated on a glass substrate using standard semiconductor processing. A rigid baseplate is used to support the glass and to mount the rest of the electronics. Figure 24.4 shows a Varian Medical Systems 4030CB with the back cover removed, revealing the boards and the TAB chips at the periphery. Since the mobility of electrons in an a-Si is on the order of 1 cm2/Vs, it is difficult to build quality, high-speed, low-noise amplifiers using the a-Si itself. As a consequence, the drive and charge-to-voltage amplifiers are generally external to the photodiode array. All the rows and columns of the array are brought to landing pads at the periphery of the glass in groups (e.g., 128), where they are connected to TAB packaged chips. TAB bonding technology is used extensively in flat panel displays. Because the pitch of the connections coming off the array is comparable to the pixel pitch, there is a mismatch between the connection density achievable on the array (∼100 μm) and what is readily available on the printed circuit boards (∼300 μm). To resolve this, the first level of drive and readout electronics is mounted in flexible TAB packages having glass side connections for every row (or column) and board side connections at a lower density, but sufficient to bring in power, control, and serial data signals. On the glass side, the TAB package is bonded to the TAB landing patterns by means of an ACF using a combination of heat, pressure, and ultrasonic energy. On the board side of the TAB connection, ACF, solder, or connectors are used. Figure 24.4’s inset shows a close-up view of the TAB connected readout and gate driver chips. Notice that the readout chips are connected to the board with solder, while the gate drivers are connected with ACF.

THE p-i-n/TFT PIXEL

537

Figure 24.4. Electronics of a Paxscan 4030CB real-time detector for radiography, fluoroscopy, and vascular applications. This FPD has split data lines, with readout electronics on two sides of the panel. The close-up view of the TAB bonded readout and gate driver chips are shown on the lower right.

24.3

THE p-i-n/TFT PIXEL

While there are a number of pixel architectures in use, the p-i-n/TFT architecture is found in the largest number of products and in the most varied applications. This architecture, first proposed for X-ray sensors by Antonuk and Steet in 1989 [1], is shown in Figure 24.5. Each pixel consists of just two elements, a TFT and a p-i-n photodiode. A picture of the pixel is shown in Figure 24.6. The light-sensitive p-i-n photodiode occupies the bulk of the pixel real estate. Its transparent ITO top electrode allows visible light to penetrate the diode. In addition, the top-side p-layer is made thin enough to allow most of the visible photons to be absorbed in the thick intrinsic layer of the diode. The typical quantum efficiency is shown in Figure 23.10 of Chapter 23. The peak absorption is in the green at 550 nm, which is well matched to the light output from GOS and CsI(Tl) screens. There are a few things to notice in the pixel photograph. First is that the schematic of the pixel is easily recognizable. Second is that the TFT switch and p-i-n photodiode are coplanar. The ratio of light-sensitive pixel area (p-i-n area) to the total pixel area is known as the FF. A finite amount of the pixel area must be devoted to the TFT switch, so as the pixel size shrinks beyond a certain size, the FF drops off rapidly. Broadly speaking, this type of pixel architecture is used for pixel sizes down to 100-μm pitch. Below 100 μm, there are alternative architectures that overcome the FF limitation. The third takeaway from this picture is that the p-i-n photodiodes are mesa isolated. This is quite different than the situation in crystalline CCD or CMOS detectors. The mesa isolation assures that there

538

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

Figure 24.5. Organization of the a-Si array and connected electronics.

Figure 24.6. Photograph of a p-i-n/TFT pixel.

is little cross talk or blooming of signal from one pixel to the next, even in situations where the light exposure is very high. The effect of saturating the pixel is to forward bias the photodiode, in which case the excess current is shunted on to the array bias line. For a moment, consider the signal chain. X-rays are absorbed by the scintillator producing light proportional to deposited energy. For CsI(Tl), it is typical that one visible photon will be generated for each 26 eV of energy. In diagnostic radiography, the range of energy used is roughly 10–150 keV. A single photon of 70 keV will produce approximately 2700 visible photons. This light is emitted in all directions, only half of it downward toward the photodiode array. However, it is common to use an efficient light reflector at the X-ray entrance side of the scintillator. So in the best case, all the visible photons are directed toward the array. Assuming the light is at 550 nm, 80% of these photons will be absorbed by the

THE p-i-n/TFT PIXEL

539

photodiode. For the 127-μm pixel shown, the FF is 57%. Therefore, a single 70keV X-ray photon will generate 2700 × 0.80 × 0.57 = 1230 electrons. This gain in signal from X-rays to electrons resulting from the intimate contact between scintillator and photodiode is the reason that FPDs have excellent quantum efficiency. In contrast, an X-ray detector based on CCDs may use the same CsI scintillator but collect the light through a mirror and lens system focusing the visible photons onto the CCD, which is preferably located outside of the X-ray beam. Because of the limited solid angle over which the lens collects photons, there is a quantum sink [2] or severe gain loss at this point in the signal chain. The result is that CCD-based detectors have comparatively poor quantum efficiency and also suffer from veiling glare and cross talk due to scattered light. At the pixel level, the readout timing is very straightforward. During integration, the TFT switch is open (OFF). Light absorbed by the p-i-n photodiode creates electron–hole pairs in the intrinsic layer. The internal field of the photodiode under reverse bias separates the electrons and holes forcing them to opposite electrodes, where the charge is stored on the capacitance of the p-i-n diode. Charge stored on the pixel capacitance causes the voltage on the floating node (node a in Fig. 24.7)

Figure 24.7. Equivalent circuit and timing of the front-end electronics connected to the a-Si array.

540

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

to move negatively in the direction reducing the bias on the photodiode. During readout, the TFT switch is closed (ON), connecting node a of the pixel to the data line, which is held at a virtual bias by a charge integrating amplifier. Node a is then discharged onto the data line and into the feedback capacitance of the charge integrating preamplifier. Figure 24.7 shows schematically the front-end circuitry of the readout chip, its relation to a single pixel, and the timing needed to capture the signal from one pixel. The charge integrating preamplifier is reset, and then the TFT gate is pulsed ON to discharge the pixel. The signal charge is accumulated on the feedback capacitance of the preamp, converting the signal to a voltage. Then the voltage is captured by an S&H. Chapter 23 discusses in detail the electronic properties of the TFT and p-i-n photodiode. Here we consider the key parameters in the design of the sensor, which are the following: Parameter

Significance

TFT ON/OFF resistance

Pixel discharge time/signal loss rate

p-i-n photodiode capacitance

Signal capacity Pixel discharge time kTC noise of the pixel

p-i-n diode leakage

Limits length of the integration time and dynamic range Source of shot noise

Parasitic capacitance on the data line (column)

Increases the noise gain for pixel noise and correlated line noise

Resistance of the data line

Source of pixel noise

Deep-level traps

Image lag Gain effect/ghosting

The ON resistance of the TFT in combination with the pixel capacitance sets the discharge time constant, τ. TFT τ = RON × Cpix

A typical value for τ is 3 μs. Ideally, the TFT access time would be on the order of seven time constants to assure complete discharge of the pixel. In practice, it is possible to use just a few time constants, although the shorter discharge time results in a gain loss and in an increase in the image lag due to the charge left behind on the pixel. The OFF resistance of the TFT determines how much charge will be lost from each pixel during one integration time, which can vary from 33 ms to several seconds. Also note that all the pixels hanging on a given data line will be leaking charge to the amplifier during the time in which the pixel of interest is discharged, which typically varies from 10 to 100 μs. This is a source of nonlinearity since the

ELECTRONICS ARCHITECTURE

541

amount of leakage charge will depend on the signal levels of the other pixels on the column. The problem is exaggerated by the fact that signal levels in the darkest areas of the image behind the patient can be 1000 times smaller than pixels exposed to the unattenuated beam. The data line capacitance and, to a lesser extent, the data line resistance are key parameters in the detector noise, which is discussed in more detail below. One other key photodiode parameter is the amount of deep-level traps. Traps create image lag and gain variations. A detailed description of traps in a-Si devices can be found in Chapter 23, and their effect on detector performance is explored further in a later section of this chapter.

24.4

ELECTRONICS ARCHITECTURE

The basic electronics architecture for an a-Si detector is shown in Figure 24.8. Groups of rows, typically 128 or 256, are connected to gate driver chips. The gate driver chip is a shift register with high-voltage outputs. A bit is input to the shift register and clocked at the line frequency down through one gate driver to the next until all the rows in the panel have been selected. The individual outputs of the gate driver swing between the TFT OFF voltage when deselected (∼–5 V) to the TFT ON voltage when selected (∼+15 V). On the column side of the array, the electronics is more extensive. ASICs are connected to groups of columns, typically groups of 120 or 128. In this chapter, the terms ASIC and readout chip are used interchangeably. Each column is

Pixel

128 channels

14 bit ADC

ASIC

ADC output

FPGA

Gain selection: 1,2,4,5

Off-Panel Transmission

128 channels

14 bit ADC

ASIC Gain selection: 1,2,4,5

a-Si array 16 x

Figure 24.8. Architecture of panel readout electronics.

542

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

connected to a signal acquisition channel inside the ASIC, consisting of a charge integrating preamplifier, an S&H, and an output multiplexor. The 128 voltage signals are serialized by the multiplexor into a single data stream. The data stream out of the ASIC is a series of voltage levels, output at the pixel rate, which are fed into a voltage amplifier, which in turn feeds an ADC chip. The purpose of the voltage amplification stage is primarily to match the output signal of the ASIC to the input requirement of the ADC. The output of the ADC is a digital word, typically 14 or 16 bits wide. The digital data from all the ASIC/ADC channels are connected via a data bus to an FPGA, which drives the data through the off panel transmission medium. With this background, we can visualize how a panel is read out or scanned. Each row is selected in sequence. When a row is selected, all the pixels in that row are connected to the columns (data lines) of the array. The charge signals from each pixel are converted to a voltage and are captured on the S&H capacitances, in parallel, at one moment in time. Once the voltage signals from each pixel in the row are stored on the thousands of S&H circuits, the signals are then multiplexed out to the ADCs and are converted to digital signals for transmission on the digital data bus. Very often it is possible to “pipeline” the capture process, which means that the signal acquisition at the front-end preamp can take place during the multiplexing and transmission process. A pipelined architecture can significantly speed up the frame rate by decreasing the time needed to acquire, process, and transmit each row of data. The readout method described above is called progressive scanning. Each row is read in sequence with the data from each pixel in the row coming out in sequence. One variation on this method is to use split data lines, so that the top and bottom half of the panel can be read out in parallel. In this case, data from the top half of the panel are output directly, while data from the bottom half of the panel are stored in memory until the second half of the frame time. This architecture requires double the electronics but has the advantage of faster frame rates and larger X-ray windows. Most FPDs are used in pulsed applications where the X-ray beam is pulsed ON for a fixed duration. This is very common in fluoroscopy (video X-ray) applications. And radiography is by nature a pulsed application since the objective is to acquire a single snapshot image. In pulsed applications, if the beam pulse arrives on the panel during the readout sequence, the result is a signaldependent offset shift due to capacitive coupling of the generated photocurrent signal into the preamp. In order to avoid this nonlinearity, it is best to deliver the X-ray beam pulse during a non-scanning period of the frame time, referred to as the X-ray window. With split data lines, the panel scanning is confined to half of the frame time, opening up an X-ray window at least half the frame period. The relationships between the detector readout, the X-ray beam pulse, and the digital video data output are shown in Figure 24.9. Another common variation on the readout sequence is to create super-pixels though binning. Here, multiple rows are selected simultaneously, while at the same time the signals from multiple columns are combined, either in the ASIC or digitally. For example, sending two shift-in bits to the gate driver chips and combining

NOISE

24.5.4

547

Input-Referred Thermal Noise

Thermal noise at the preamp inputs connected to each column of the imager is usually the largest source of electronic noise in FPDs. Thermal noise is generated at resistances in the circuit. Besides the preamp itself, the data line’s own resistance is the other major thermal noise source. The thermal noise itself is not particularly high compared to other types of detectors, but the noise gain at this point in the circuit is significant due to the large parasitic data line capacitance. Any thermal noise present at the input to the preamp will have the following noise gain: Anoise =

Cdata + Cin . Cfb

The feedback capacitance on the preamp, Cfb, is matched to the pixel capacitance or set at a lower value in order to get higher charge-to-voltage gain. A typical value for Cfb is 1 pF. On the other hand, Cdata, the data line parasitic capacitance, is on the order of 50–100 pF depending on the size of the a-Si array and on the number of pixels (resolution). The data line length can be up to 43 cm long, having more than 3000 crossover points with underlying structures and a gate–drain overlap capacitance of thousands of TFTs. Also substantial is the input capacitance to the preamp, Cin, which can be 10–15 pF. For the best noise performance, the preamp input transistors are ideally large-area PMOS devices, minimizing the 1/f noise of the amplifier, at the expense of increasing the total data line capacitance. So it is very common in FPDs that the input-referred thermal noise will be gained up by a factor 100, and unfortunately, the signal charge does not see this gain. Just 50 μV of noise at the preamp input can result in over 30,000 electrons of noise, which is roughly the average signal in low-dose fluoroscopy. Clearly, the noise performance of the ASIC preamp is critical to image quality in FPDs. In the discussion above, the noise bandwidth was ignored. It is reasonable to ask why the TFT OFF resistance, which can be mega ohms, is not a major source of thermal noise. The answer is in that the TFT OFF resistance, in combination with the data line capacitance, creates a low-pass filter with a bandwidth below the pass band of the ASIC electronics (∼1–250 kHz), so this noise is effectively rejected. In contrast, the data line resistance is on the order of 1 kΩ, so its thermal noise tends to be in the pass band of the ASIC electronics. At the higher end of the frequency spectrum, the input noise is rejected by the bandwidth limitations of the preamp and S&H circuits. The preamp bandwidth needs to be relatively high in order to hold the input node at “virtual ground,” maintaining the linearity of the charge-to-voltage conversion. However, the only speed requirement on the S&H bandwidth is that the signal settles at a rate sufficient to maintain the line rate. Reducing the bandwidth prior to the signal capture can dramatically reduce the acquired noise. For a given input noise spectral density, in units of V Hz , Qnoise = NSD × NEB1 2 × (Cdata + Cin + Cfb ) ,

548

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

where

NSD = the input-referred noise spectral density and NEB = the noise equivalent bandwidth.

24.5.5

Line Noise

Global noise sources, such as the gate driver power supplies and the array photodiode bias, create correlated line noise because all pixels in a row are sampled at the same moment in time. Noise from the power supplies coupled into the preamp front end sees the same noise gain as the input-referred thermal noise. And making the situation worse, the acceptable level of line noise is on the order of 10 times less than that of the input-referred thermal noise. To meet this condition with just power supply regulation would require power supplies with 5 μV of noise or less, which is extremely difficult, particularly in compact portable systems. Figure 24.12 shows two images of FPD electronic noise, with and without line noise rejection circuitry. In the left-hand image, the line noise variation is one-half the pixel noise, which means that the SNR degradation due to line noise is only 20%. However, it is clear from this image that the line noise dominates the visual appearance. This is a well-known phenomenon related to the human visual system’s excellent ability to distinguish horizontal lines. This problem was an issue in the early days of analog TV. At least one paper from that time proposed that the random noise in the image would need to be 15 times greater than the correlated line noise, in order for the line noise to be invisible. Here, let us define the line

Detector Electronic Noise

σ line ≈

1 × σ pixel 2

Detector Electronic Noise with Line Noise rejection

σ line ≈

1 × σ pixel 10

Figure 24.12. Two images of FPD electronic noise, with (right) and without (left) line noise rejection circuitry.

NOISE

549

noise ratio or LNR as the ratio of the total noise in the image divided by the electronic line noise: LNR =

σ total , σ line

2 2 where σ total = σ X-ray . + σ 2pixel + σ line In FPDs, the ratio of electronic pixel noise to electronic line noise should be a minimum of 10, because at the X-ray quantum-limit, where σX-ray = σelectronic ≈ σpixel, the X-ray noise contributes another factor of 1.4 to the LNR, bringing the total to approximately 15. In FPDs operating above the quantum-limited dose, noise in the X-ray beam dominates the total noise and the LNR will increase with increasing dose, making the line noise invisible at higher signals. Line noise is primarily visible in the lowest part of the signal range, that is, the darker areas of an image where the signal level has fallen below the quantum-limited dose.

24.5.6

kTC Noise

At any point in the signal acquisition process where a voltage level is set on a capacitor through a switch, the thermal noise from the resistance in the switch is captured along with the signal voltage. This noise is known as kTC noise because the magnitude of the switch resistance is irrelevant and the noise depends only on Boltzmann’s constant, the temperature in degree kelvin, and the capacitance. The reason that the resistance drops out of the noise equation, as explained in Figure 24.13, has to do with the self-filtering of the passive resistance, capacitance

A( f ) = v in

1 1 + j 2pfRC

f −3db = v2

R C

v out

1 2πRC

R

v in

C

v out

v 2 = 4kTRΔf Noise Equivalent Bandwidth (NEB) = f −3db × 2 vout = 4kTR × NEB =

π 2

=

1 4 RC

kT C

Output noise is independent of R ! Figure 24.13. The equivalent circuits explaining the resistance-independent kTC noise.

552

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

frequencies above Nyquist are aliased into the pass band of the imager (0 to fNyquist) as noise. We can estimate the aliasing contribution to the NPS of the detector by dividing the integral of the MTF2 above Nyquist by the integral of the MTF2 over all frequencies. In the case of the 150-μm CsI on a 127-μm array, 17% of the signal power is aliased into the pass band as noise. Photoconductor-based detectors are discussed in Chapter 25. This type of detector can achieve unparalleled resolution approaching the theoretical limit imposed by the pixel pitch (the Sinc function). However, this resolution comes at the cost of noise performance. For general radiography and fluoroscopy, the desirable pixel pitch based on the imaging task is between 120 and 200 μm. Using smaller pixels comes at the cost of SNR (i.e., increased dose) because the number of X-ray photons per pixel shrinks and, as discussed, the SNR is the square root of the number of absorbed photons. Using photoconductor-based detectors with a 127-μm pixel pitch leads to large aliasing noise (>40%), which is one of the reasons that this technology has not succeeded in general radiography or fluoroscopy. However, in mammography where the imaging task requires a pixel pitch between 50 and 100 μm, the Nyquist frequency is sufficiently high that aliasing is reduced to an acceptable level. Today, photoconductor-type FPDs based on amorphous selenium have a large share of the full-field digital mammography market.

24.6

LAG AND GAIN EFFECT

As discussed in Chapter 23 and alluded to earlier, the a-Si in the p-i-n photodiodes has a large density of deep-level traps, which modulate the signal applied to the array. At the beginning of an X-ray pulse, some of the generated signal charge goes into filling these traps. The result is that the sensitivity of the array increases as the beam ON time progresses because less and less of the generated signal charge is lost to traps. This sensitivity modulation is known as the gain effect. When the beam shuts off, the trapped charge is emitted, contributing to the signal of later frames. This is known as image lag, since a shadow of the previous image will appear in the postexposure frames. In real-time medical imaging, there are many applications where the FPD must switch from a high-dose radiograph (highresolution snapshot) to low-dose fluoroscopy (video) in less than 1 s. Because the radiography dose can be 100–1000 times larger than the fluoroscopy dose, the radiographic image lag is often comparable to the average signal in the fluoroscopy sequence. There are several strategies used to mitigate the gain and lag effects. The first is to correct the images through an algorithmic approach [4], applied after the image acquisition. Although computation resources are required, this approach works well because, to first order, the trapping and emission processes are well behaved and consistent over large variations in signal. The other methods involve prefilling the traps prior to signal acquisition, either by using a bias light or by forward biasing the p-i-n photodiode [5]. Figures 24.15 and 24.16 show the dramatic reduction in gain and lag effect possible using these techniques.

101.00% 100.00%

Sensitivity

99.00% 98.00% Standard Mode Reference Mode Forward Bias Mode

97.00% 96.00% 95.00% 94.00% 0

20

40

60

80

100

Frame Number

Figure 24.15. The strong variation in the sensitivity or gain with frame number for the standard mode and its substantial reduction with the forward bias mode. 0.09%

2.50%

0.08% 0.07% 0.06% 1.50%

0.05%

Standard Mode Forward Bias Mode

0.04%

1.00%

0.03%

Lag with Forward Bias

Lag in Standard Mode

2.00%

0.02%

0.50%

0.01% 0.00% 0

20

40

60

80

0.00% 100

Fram e Num ber

Standard mode

Forwardbias mode

Figure 24.16. Lag effect with and without the forward bias mode (top) and the corresponding postexposure lag images (bottom).

554

24.7

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

ALTERNATIVE ARCHITECTURES

There are a number of other alternative detector architectures briefly summarized in Table 24.2. These concepts span the gamut from current products to pure research & development.

TABLE 24.2. Alternative Detector Architectures Key Features Photoconductor

X-rays converted to charge directly

Potential Advantages Highest resolution

Charge is electrostatically focused on to pixel electrode.

MIS photodiode

MIS-type photodiode No p-type a-Si is needed.

Obstacles Manufacturability of available photoconductor materials High applied bias voltage leads to noise and reliability issues

Compatible with standard LCD processing

Requires a specific reset cycle, so pixel is not continually integrating Difficult to use in real-time (video) imaging

Maxfill p-i-n/ TFT [6]

p-i-n photodiode is on top of TFT. Photodiode is still mesa isolated to avoid signal blooming in saturation.

Organic detectors

Ink-jet printed devices using organic polymers

Increased FF If a low k dielectric is used, Cdata can be reduced.

Low cost

Increased processing complexity, lower yield

Maturity of the technology

Flexible substrates Active pixel sensors

Incorporate more circuitry (TFTs) into the pixel itself.

Good low-dose performance Extended dynamic range

Increased processing complexity, lower yield Same performance may be achievable at a lower cost using standard architecture.

ABBREVIATIONS

24.8

555

SUMMARY

a-Si FPDs are enabling a digital revolution in X-ray imaging, similar to the sea change seen in commercial photography and video. The inherent radiation hardness and large area capability of a-Si detectors are clearly enabling attributes. But the excellent linearity, wide dynamic range, and edge-to-edge uniformity of a-Si FPDs provide high image quality and support advance applications like cone beam CT. Two other obvious advantages are size and weight. The IIT, used in traditional X-ray video, is roughly the size of a kitchen garbage can. An FPD with a 12″ × 16″ active area (20″ diagonal) consumes less than 25% of the volume of a 12″ IIT and less than 15% that of a 16″ IIT. In addition, the FPD takes the place of not only the IIT but also the attached image recording devices, including the CCD camera, the 35-mm cine camera, and the spot film device. The result is vastly improved access to the patient in interventional procedures. In addition to the reduction in size, the weight of a flat panel imager is less than 60% of the IIT-based imaging chain. Despite the advantages of a-Si in large area detectors, clearly their size and limited on-detector processing capability create unique design constraints. While most of the core design challenges have been overcome, developers continue to push toward lower electronic noise for better low-dose performance, improved linearity to reduce image artifacts, higher frame rates to support advance applications like dual energy imaging, and expanded dynamic range for improved cone beam CT.

ABBREVIATIONS ACF—anisotropic conductive film ADC—analog-to-digital converter, converts an input voltage to digital data Anoise—noise gain a-Si—amorphous silicon ASIC—application-specific integrated circuit; refers to custom readout chips connected to the columns of the array C—capacitance Cdata—dataline capacitance CCD—charge-coupled device CDS—correlated double sampling Cfb—preamp feedback capacitance Cin—preamp input capacitance CMOS—complementary metal oxide semiconductor; refers to a photodiode-based silicon imaging chip Cpix—pixel capacitance Cs&h—sample and hold capacitance CsI, CsI(Tl)—thallium-doped cesium iodide, a type of X-ray scintillator CT—computed tomography

556

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

DQE—detective quantum efficiency EPID—electronic portal imaging detector f—frequency FF—fill factor fNyquist—Nyquist frequency FPD—flat panel detector FPGA—field-programmable gate array GOS—gadolinium oxysulfide, a type of X-ray scintillator Gy—Gray, a unit of radiation dose i—undoped or intrinsic semiconductor IIT—image intensifier tube ITO—indium tin oxide k—Boltzmann’s constant LNR—line noise ratio LVDS—low-voltage differential signaling Mrad—mega (1 million) rads, where rad is a unit of X-ray dose describing the ionization energy deposited in an irradiated object MTF—modulation transfer function n—negatively doped semiconductor NEB—noise equivalent bandwidth NPS—noise power spectrum NSD—noise spectral density p—positively doped semiconductor PMOS—p-type metal oxide semiconductor devices Qnoise—electronic noise in coulombs R—resistance RH—relative humidity S—signal intensity S&H—sample and hold circuit SNR—signal-to-noise ratio T—absolute temperature TAB—tape-automated bonding TFT—thin-film transistor σ—signal standard deviation, i.e., noise in the signal τ—pixel discharge time constant

REFERENCES [1] [2]

L. E. Antonuk and R. A. Steet. Multi-element-amorphous-silicon-detector-array for real-time imaging and dosimetry of megavoltage photons and diagnostic X rays. U.S. Patent # 5,079,426, September (1989). P. Munro, J. A. Rawlinson, and A. Fenster. Therapy imaging: A signal-to-noise analysis of a fluoroscopic imaging system for radiotherapy localization. Med. Phys. 17, 673–772 (1990).

REFERENCES [3] [4] [5]

[6]

557

M. H. White, D. McCann, I. Mack, and F. Blaha. Coherent sampled readout circuit and signal processor for a charge coupled device array. U.S. Patent 3,781,574, October (1972). L. D. Partain, I. Mollov, C. Tognina, and R. E. Colbeth. Method and apparatus for correcting excess signals in an imaging system. U.S. Patent 7,208.717, October (2002). I. Mollov, C. Tognina, and R. E. Colbeth. Photodiode forward bias to reduce temporal effects in a-Si based flat panel detectors. In Medical Imaging 2008, Proceedings of SPIE, Vol. 6913-13, J. Hsieh and E. Samei, eds., pp. 69133S-1 to 69133S-9. Bellingham, WA, SPIE (2008). R. L. Weisfield, W. Yao, T. Speaker, K. Zhou, R. E. Colbeth, and C. Proano. Performance analysis of a 127-micron pixel large-area TFT/photodiode array with boosted fill factor. Medical Imaging 2004: Physics of Medical Imaging, Vol. 536842, M. J. Yaffe and M. J. Flynn, eds., pp. 338–348. Bellingham, WA, SPIE (2004).

25 PHOTOCONDUCTOR DIGITAL X-RAY IMAGING GEORGE ZENTAI Varian Medical Systems

25.1

INTRODUCTION

Conceptually, photoconductor detectors initially seem quite different from solar cells. They consist of a highly resistive semiconductor layer placed between two metal electrodes to efficiently collect any mobile electric charge generated by incident photons based on a voltage applied to the electrodes. However, the a-Si solar cell structure shown in Figure 23.9 actually shows this exact same structure. The a-Si solar cell does have thin, highly doped p+- and n+-layers next to the metal electrodes. A standard way of producing ohmic metal contacts to semiconductors is to heavily dope the regions just adjacent to the contacts. For solar cell energy conversion, the doping needs to be of opposite types at each electrode. Solar cells are typically reverse biased when used as photon detectors and imagers so that no net positive energy conversion is produced. In essence, solar cell detectors and photoconductor detectors are fundamentally quantum converters of light, absorbing as many incident photons as possible. Ideally, every absorbed photon contributes to a net current signal that is linearly related to the intensity of the photon flux striking the detector’s surface with a minimal noise contribution. In this mode, the behavior and principles of operation of photoconductors are very nearly identical to those of solar cells that contain a very high-resistance photon active region. The latter includes point contact crystalline concentrator solar cells in addition to the a-Si ones. For solar cell energy conversion, the device bandgap needs to be in the same range as the incident photon energies. For X-ray detectors, there is no such restriction so that photoconductors are often chosen with bandgaps larger than most visible light photons because this gives the lowest dark currents important to Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

559

560

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

X-ray imaging. Furthermore, the a-Si TFTs, used in X-ray imagers, have the same basic structure with a layer of high-resistance semiconductor contacted by two metal contacts made ohmic by highly doped n-type interface layers (see Fig. 23.6). Here, the device current is controlled not by absorbed photons but by a third intervening field-effect contact separated from the a-Si by an insulating layer. Traditionally, there were two ways of X-ray imaging to capture information in medical practice: 1. radiography, a snapshot of overlapped internal organs of the imaged patients, and 2. fluoroscopy, a continuum of images or a series of snapshot images of the imaged objects. In the early 1970s, X-ray CT was invented and provided medical practitioners with an unparallel view of patient cross-sectional views without surgery. Many other X-ray imaging techniques are just slight variations of the above X-ray imaging methods. For all these medical X-ray imaging modalities, high performance at low X-ray exposure and at high spatial frequencies is demanded. Most AMFPIs currently available commercially are indirect in that they use a two-stage process, where X-ray photons are first converted into visible light in a top scintillator layer, and light then (in the second step) is transported down into an underlying layer array of photodiodes that convert part of this light into detectable electronic charge. This is in contrast to the D-AMPFI that eliminates the need for a scintillator and replaces the photodiode layer with a photoconductor, which directly converts the X-ray photons into a detectable electronic charge in a single step. In addition to simplicity and potentially lower costs, D-AMFPIs provide other major advantages of higher spatial resolution and better contrast sensitivity than the indirect configurations. This is due to the fundamental limitation of scintillators that emit light uniformly in all 4π geometric directions so that much of the light is lost, and the fraction that does transport down excites not only the pixel just below its generation position but also multiple adjacent pixels as shown in Figure 25.1. This fundamentally limits the spatial resolution, particularly in indirect devices with thick enough scintillators to efficiently absorb most incident X-ray photons. The greater the scintillator thickness, the greater image blurring and the lower associated spatial resolution (see Fig. 25.2). In direct imagers, a voltage is applied to a photoconductor layer to collect the charge, generated by the absorbed X-ray photons, so that almost all of the charge transports to the pixel underneath the generation positions due to the field focusing caused by the applied electric field as shown in Figure 25.3. This focusing effect exists even in thick photoconductor layers so the resolution does not degrade with increasing detector thickness and increasing sensitivity (Fig. 25.4). This is especially advantageous at imaging with high-energy X-rays. Such advantages are dramatically demonstrated in a-Se D-AMPFI (the single D-AMPFI currently available commercially), whose spatial and contrast resolution are unprecedented and unmatched by any indirect imager. These advantages become particularly valuable

INTRODUCTION

561

Figure 25.1. Scattered light generates charge not only in the pixel where the X-ray hits the imager but also in neighboring pixels.

Figure 25.2. Thicker scintillator—better absorption—worse resolution (MTF).

562

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

Figure 25.3. Charge generated by an X-ray beam moves perpendicular to the electrodes— no charge spreading.

Bias Bias

El. field

X-ray

Charge movement

Lower sensitivity

Good resolution

El. field

X-ray

Top electrode

Charge movement

Higher sensitivity

Pixel array

Good resolution maintained at thicker layers

Line spread function

Figure 25.4. Thicker layer—better absorption—no degradation in resolution (MTF).

PHOTOCONDUCTOR IMAGERS

563

in mammography, where the high spatial resolution requirements (all the way out to the Nyquist sampling frequency limit) are closely coupled to the additional need for resolving subtle contrast variations. The a-Se D-AMFPIs are beginning to dominate the commercial flat panel mammography market.

25.2 GENERAL REQUIREMENTS FOR PHOTOCONDUCTOR MATERIALS USED FOR X-RAY DETECTION AND IMAGING Direct detector materials preferably exhibit several attributes. These include high X-ray absorption, high charge collection, low dark current, and good uniformity. These are difficult to achieve in a single material. The most important physical parameters of the popular X-ray conversion photoconductor materials are summarized in Table 25.1 [1]. For efficient X-ray absorption at clinical exposure energies, we need high Z materials. In addition, the X-ray energy required to generate an electron-hole (e-h) pair, designated by the parameter W, should be low. The lower the W value, the higher the number of e-h pairs liberated by the interacting X-rays and the higher the X-ray sensitivity. As Table 25.1 indicates, the parameter W is much smaller for HgI2, PbI2, and CdZnTe than for a-Se. The larger the mobility-lifetime product (μτ), the greater the distance the electrical charge can move in the detector without recombination. Greater distances result in higher sensitivity due to better charge collection. The high Z number, low W, and high μτ result in a very high signal, which helps overcome noise sources in fluoroscopic (low dose) modes of operation. The ability of the imager to operate at low bias voltages allows low-voltage electronic design. The a-Se requires minimum 10 V/μm electrical field, while other photoconductor materials need only a 1 V/μm field to work at decent sensitivity levels. Another key feature of the materials is the processing temperatures they require to be fabricated into the desired form. The AMFPI panels have an a-Si switching matrix TFT layer on a glass substrate, and the highest temperature these TFTs can tolerate without damage is about 220–240°C. From this point of view, CdZnTe is not a good candidate because it requires a minimum of 600°C substrate fabrication temperature, which is way over the limit of the TFTs. There have been experiments to deposit CZT on a separate substrate, but on high-pixel-count largearea arrays, the connections to the pixellated TFT matrix readout array rapidly become unmanageable, limiting the maximum imager size only to about 9 in.2 [2]. CdZnTe, PbI2, and HgI2 are all much more stable in terms of phase transitions at elevated and at low temperatures than a-Se. a-Se imagers require temperature control not only during operation but also during shipment.

25.3

PHOTOCONDUCTOR IMAGERS

At present, only a-Se imagers are commercially available; the other imagers mentioned in this chapter are only experimental devices.

564

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

TABLE 25.1. Comparison of a Few Photoconductor Candidates for X-Ray Conversion Poly-HgI2

Poly-PbI2

a-Se

Poly-CdZnTe

80, 53

82, 53

34

48, 30, 52

Energy bandgap (Eg) (eV)

2.1

2.3

2.2

1.5–1.7

Charge pair formation energy (W) (eV)

5

5.5

50 (eff.)

4.5

Lower W—higher gain

Mobilitylifetime product (μτ) (cm2/V)

10−5

h: 1.8 × 10−6 e: 7 × 10−8

10−6–10−5

e: 8 × 10−3 h: 3 × 10−5

Higher μτ—better charge collection

Operational electric field (E) (V/mm)

0.2–1.0

0.2–1.0

10

1–2

Lower E— reduced risk of electrical breakthrough

Processing temperature (°C)

120

220

100

600

Lower temperature— simpler process

Stability (°C)

>100

>200

60oC) and a temperature expansion–coefficient mismatch delamination at lower temperatures (100 kW

Figure 26.7. Comparison of installed costs for crystalline versus thin-film systems.

TABLE 26.4. System Level (Dollar per Watt) from Solar Buzz July 2009 Solar I

Solar II

Installed Home System

Solar III

Installed Commercial System Installed Industrial System

On grid or off grid (2 kW)

On grid (50 kW)

On grid (500 kW)

Customer price

$17,394

Customer price

$327,438

Customer Price

$2,373,260

$/W

$8.70/W

$/W

$6.55/W

$/W

$4.75/W

600

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

Leveled Cost Of Electricity L = (1+r)(Cm+Cb)Fs/ηsSha + (1+r)CiFi / ha + O&M The 9 key variables are: 1. ηs = PV system conversion efficiency 2. Cm = PV module cost ($/m2) 3. Cb = area related BOS including installation ($/m2) 4. S = Site specific solar intensity (kW/m2) 5. ha = Annual solar hours for PV system (tracking) 6. Ci = Inverter cost in $/kW 7. F = Fixed charge rate (converts initial investment into annualized charge) 8. r = Indirect cost rate (permitting, NRE) 9. O&M

Figure 26.8. An LCOE expression and input parameter summary.

a large number of diverse groups in order to bring down the price of solar electricity in terms of cost per kilowatt hour. For example, TF companies emphasize low-cost modules, but this is just the Cm term in the LCOE equation. The module supports, field wiring, and installation costs can be higher when the module efficiency is low because more modules need to be installed. This is the Cb term in this equation. The sunlight, S, available at the location is certainly important. Following the sun by tracking the modules will increase the number of hours per year of operation, which is the ha term. For comparison, the product Sha equals SEY, the annual sunlight energy density of a site from Chapter 2. Increasing the annual hours of operation will reduce the impact of the inverter cost, Ci. Note also that the cost of the hardware and installation are not the only costs. The projects have to be financed by the banks, and this is the finance (F) term. Government permitting is also required, and this can cause delays increasing costs, and this is part of the project-specific overhead (r) term. Finally, the system will need some maintenance over time. In chapter 16 of the First Edition, the EPRI authors gave a table of the LCOE equation parameter values for their projections for the future TF case. These projected values are repeated here in the left number column in Table 26.5 along with our current projected values for the c-Si and low concentration cases. This allows a comparison of the projected LCOE values achievable for the cost of solar cell electricity in the fairly near term. The results are exciting with the low concentration option’s projected cost of solar electricity at 8.6¢ per kilowatt hour given volume scale-up. Notice that the BOS (BOS = Cb) number is lower for the TF case than for the c-Si or low concentration cases. This is because one-axis tracking is assumed for the c-Si and low concentration cases. Because of the lower efficiency for the TF case, tracking is not affordable. Tracking for the latter two cases then gives more annual operating hours.

CURRENT STATUS

601

TABLE 26.5. Realistic EPRI Equation Parameters for Thin-Film, c-Si, and Low Concentration Systems Variable

Module cost (Cm) BOS (Cb) (Cm + Cb) Module efficiency (η) Module power Annual hours (ha) S haη

Value

Comment

Thin Film

c-Si

$133/m2

$400/m2

$224/m2

100 MW/year

2

$100/m2



2

2



$29/m

2

Low Concentration

$100/m

2

$162/m

$500/m

$324/m

10%

19%

19%

100 W/m2

190 W/m2

190 W/m2

2321

3000 2

— 2

570 kWh/m

570 kWh/m

0.698

0.877

0.568

$234/kW

$234/kW

$234/kW

0.10

0.078

0.078



Fixed charge rate (F)

10.3%

10.3%

10.3%

EPRI 1995

Indirect cost (r)

0.225

0.225

0.225

EPRI 1995

F(1 + r)

0.126

0.126

0.126



System cost (¢/kWh)

10.0

12.0

8.1



O&M (¢/kWh)

0.5

0.5

0.5



¢/kWh

10.5

12.5

8.6



(Cm + Cb)/S haη Power-related BOS (Ci) Ci/ha

232 kWh/m

3000 2

S at 1 kW/m2

per year — EPRI 1995

In Table 26.5, the area-related cost variables for solar cell systems are logically given in dollar per square meter and the power converter cost is given in dollar per kilowatt. However, the solar sales and customer communities are accustomed to quotes in dollar per watt. Table 26.6 translates the key parameter assumptions given in Table 26.5 into dollar per watt to facilitate comparisons in that customary terminology. The numbers in this table are revealing. For example, the $3.52 per watt price for an installed c-Si solar cell system is within sight in the relatively near future from today’s $4.75 per watt large system price. It is also quite straightforward to see the evolutionary developments associated for the low concentration implementation to arrive at a system price of $2.39 per watt. This system capital equipment price then implies an LCOE price for solar electricity of 8.6¢ per kilowatt hour, which would be competitive for commercial customers now paying retail prices for electricity at over 10¢ per kilowatt hour.

602

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

TABLE 26.6. Translation of Key Parameter Assumptions of Table 26.5 into Customary Terminology

Module cost (Cm)

Thin Film ($/W)

c-Si ($/W)

Low Concentration ($/W)

1.33

2.10

1.18

BOS (Cb)

0.29

0.53

0.53

(Cm + Cb)

1.62

2.63

1.71

(Cm + Cb + Ci)

1.85

2.86

1.94

(1 + r) (Cm + Cb + Ci)

2.22

3.52

2.39

Note that it is the metric AC kilowatt hour per square meter per year that is required for accurate LCOE calculations regardless of whether 1 sun or concentrating systems are used in fixed tilt or with one- or two-axis tracking (see row 6 in Table 26.5). This metric is easily available for any site that records the AC kilowatt hour produced in a year. This is then divided by the area of the system, which is the external dimension of the modules multiplied by the number of modules in the installation. Site design- and system design-specific factors, like ground coverage ratios (see Chapter 11), are included in the area-related BOS cost factor. Procedures for calculating LCOE have been discussed in several chapters in this book. While the various authors’ terminology changes, the concepts and required inputs and derived results are similar to those from the EPRI equation.

26.2.7

Applications and Markets

In the summary discussion in this chapter so far, the focus has been on solar cell modules for the various electric power markets. In the last section, it was noted that this market can be segregated into the residential, commercial, and industrial or utility sectors. However, for the TF systems, interesting spin-off applications have developed, such as for medical X-ray imaging and video displays (LCDs). The X-ray applications were described in more detail here in Chapters 22 through 25. Two properties of amorphous silicon solar cells make them well suited for X-ray imaging. First is their lack of crystal structure, which makes them very resistant to X-radiation damage (up to megarad doses). The second is their relatively high bandgap and accompanying low dark current, which is particularly valuable in radiography. Devices dominated by resistive effects (photoconductors) can also offer particular advantages for such applications. A third and major advantage is their ability to be fabricated as large-area electronic imagers on glass substrates at commercially competitive costs. The 2009 digital X-ray flat panel imager market was $2 billion and growing at a 15% per yearly rate. Continued growth at the rate for the next 15 years would correspond to a $16 billion annual market size.

RECOMMENDATIONS

603

26.2.8 Government Program Investments and Solar Cell Electricity Growth The macroeconomics of solar cell electricity and its timing are becoming increasingly attractive. The historical government program sizes needed to change world solar cell market leadership positions are an estimated ∼30% of the cumulative market size, soon approaching the $30–$100 billion levels (see Fig. 2.7). Fortunately, the increased revenues produced typically equal such investment magnitudes over the following 3 or 4 years. The time required to reach the 100- to 1000-fold increase in market size and to supply substantial fractions of the world energy markets is on the order of 15–20 years (see Fig. 2.1), should current trends continue. For magnitude and timing comparisons, the costs of the Alaskan oil pipeline exceeded 23 billion (in 2007 U.S. dollar [9]), and the first oil flow through the pipe in 1977 was 9 years after the first gusher oil well demonstration in Prudhoe Bay, Alaska [10]. This project created on the order of 10,000 jobs in Alaska and produced a long-lasting increase in the Alaskan economy. Its environmental impacts were substantial, including the loss of at least 32 lives, but the disruptions and responses to the 1973 oil embargo made its construction possible, all with private financing in a very mature industry. However, the 35-year construction of the U.S. Interstate System completed in 1991 cost $129 billion with $114 billion (all in then current U.S. dollar) paid by the U.S. Federal Government from highway user taxes [11]. It proceeded because its construction was deemed “vitally important to national goals.” For further magnitude comparisons, the cost of oil and oil product imports to the United States in 2007 was $327 billion [12]. Should future oil embargos or other supply disruptions encourage direct military interventions to preserve deliveries, recent U.S. experience in Iraq indicates 5-year costs can be on the order of US$450 billion [13] with strong environmental impacts including loss of life exceeding 10,000. There are at least five world economies large enough to make these levels of investment into solar cell electricity technology and infrastructure development. Several appear ready, willing, and able to proceed with such government support.

26.3

RECOMMENDATIONS

In spite of the remarkable accomplishments of the solar industry to date, unfortunately, costs are still too high for really large-scale use on par with fossil fuels today. Government policy makers interested in bringing revolutionary innovations into the market place should understand that there are four stages that an innovation moves through on its way to large-scale commercialization. These stages are 1. invention and first prototype benchtop demonstration; 2. integration of the prototype into a system and qualification testing; 3. manufacturing technology and production; and 4. market introduction, acceptance, scale-up, and cost reduction.

604

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

The c-Si and TF solar modules and systems have now moved to stage 4 in this process. The technical expertise and worldwide market for the manufacture, distribution, installation, development, and exploitation of these solar cells for large-scale electricity production are so widespread and so pervasive that its continued growth and prosperity is well beyond the scope, control, or even the need of participation from any one country. It is recommended that a country or economic union that wants to move up or to remain in the top rankings in this inevitable process adopt one or more of the six identified public policy programs. History has shown them to be the most effective processes for achieving shifts in rankings—with feed-in tariffs being the leading example. History also indicates that the level of investment required to accomplish market shifts is on the order of 30% of the total integrated market investment worldwide up to that point. Hence, the more time that passes, the more difficult it becomes to arrange such investments for market shifting success, plus the more restricted such opportunities become to just the largest world economies. Fortunately, history also indicates that the investment payback time for the involved economies is on the order of 3 years in terms of contributions of products sold by these economies. Furthermore, the process itself generates a large number of jobs, economic regional growth and prosperity, and even reversals in a country’s or union’s balance of international payments where all of these benefits can last for multiple decades. However, there are now opportunities for newer solar cell technologies to move through these stages with government assistance and with the tremendous promise of lower cost and the associated benefits to society. The various concentrator module options represent an example. They are now emerging out of stage 1 and are entering stage 2. There is a need for demonstration projects for these newer solar options. Certification for safety and durability can be carried out in parallel with these demonstration projects. The lack of demonstration project funding in the United States as shown in Table 26.7 is one of the reasons the United States lost its early leadership in the solar industry. Without these funded demonstration and certification projects, the newer technologies are simply stuck in the laboratory with no clear path to commercialization. In the United States and in Europe, there is also a need for government investment in automated manufacturing as required in stage 3. Without automated

TABLE 26.7. Breakdown of Annual Solar Cell Development Budgets, 2001 Japan

United States

Germany

Research and development

51.0

35.0

26.7

Demonstration

16.5

0.0

5.5

188.4

84.6

29.6

Market stimulation

Source: International Energy Agency Photovoltaic Power Systems Programme.

RECOMMENDATIONS

605

manufacturing (see Chapter 7 for the current state of the art), low-cost labor wins. This is one of the reasons for China’s recent success in the solar market. The value of forcing the cost of solar cells down toward the $1 per watt level, currently being led by TF CdTe, cannot be overestimated. However, it is not just the initial purchase price of, say, a megawatt of modules that is important. It is the total life cycle cost in cents per kilowatt hour that will ultimately determine the large-scale technology competitiveness plus a technology’s staying power as the market size grows 100-fold and then 1000-fold. It involves the whole host of parameters that determine the LCOE, which is the most comprehensive integrated performance metric for the systems. There is no single characteristic that can, for the long term, make solar cell electricity economically competitive with current utility-provided electric power. Focus on just one isolated parameter can be very misleading. One of the strongest recommendations related to that is the large-scale field testing of the various available technologies, in a single location at the same time, side by side, under identical conditions. This independent field testing should start with sunny locations and then should spread to any other less favorable localities believed conducive to solar cell electricity conversion. The key parameters that should be monitored in such installations are the kilowatt hour per square meter per year produced by the competing technologies, the integrated solar insulation energy in kilowatt hour per square meter per year striking the systems installed at this site, and thus their field efficiencies as well as the installation costs and the degradation and failure rates plus the maintenance costs per year. To the best of the authors knowledge, there are no such independent test sites collecting such comprehensive but critical data at the present time anywhere in the world. These hard data are going to be crucial for justifying the advantages and disadvantages of the new technologies of TFs (including CdTe and CIGS) and of concentrators that are struggling to make inroads into the strong and long-standing dominance of c-Si. The test conditions should include one- and two-axis tracking versus no tracking. Standard c-Si technology (without concentration) is approaching its limits for price reduction as its material costs are approaching 50% of it total manufacturing costs. This fundamentally limits future price reductions. Without the competition of new “threats” like TF CdTe and concentrators, single-crystal silicon innovations like HIT and IBC cells would not likely occur, at least in a timely fashion. Only time and the market will determine ultimate shifts in technology dominance and market leadership. Another strong recommendation is for substantial government or union support program funding for the transition of new technologies into the market place—a challenge often referred to as the “valley of death” thwarting new and novel approach introductions. The barriers to market entry grow more formidable daily for the newer technologies against the currently dominant c-Si technology. TF CdTe has apparently broken through at least initially. However, the total range of creditable competition, including concentrators, needs to be brought at least to the large-scale field testing levels to establish its true long-term potentials and pitfalls.

606

26.4

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

SUMMARY

The solar cell market has grown 100-fold in yearly power products sold since the First Edition 15 years ago with (non-concentrating) c-Si still supplying the dominant share by far. However, c-Si is having a challenging time trying to reach “grid parity,” and a niche market has developed for TF solar cells that can provide lower initial purchase prices per kilowatt relative to c-Si. An alternate life cycle analysis of equivalent cost of power produced (in cents per kilowatt hour) shows that TFs may not be economically competitive long term if their efficiencies remain low or if their fundamental compositions contain elements of comparatively low abundance or until side-by-side field evaluations can establish relevant performance, maintenance, degradation, and other cost elements versus the competition. For example, TF efficiency values are still lower by up to a factor of two. Concentrator solar cell systems represent another alternative to 1-sun c-Si and TF systems. These concentrator systems are experiencing their first large-scale demonstrations in major installations in Spain. A key metric that is being determined but is still not accurately known is the kilowatt hour per square meter per year produced by concentrator systems versus 1-sun flat plates in their various installation conditions (of fixed angle and one- and two-axis tracking; for example, see Table 26.2 Palo Alto data). However, there are early indications that concentration (at least at low concentration values) may be one of the first to really compete with the traditional fixed-tilt c-Si and maintain the pressure for reductions in the most important metric, which is the life cycle cost in cents per kilowatt hour. Unfortunately, for the United States, its market lead held until the early 1990s has successively slipped country by country to the point that four other countries, Japan, China, Germany, and Spain, now occupy the leading positions. Their positions have advanced due to one or more of the six identified effective public policy programs where feed-in tariffs have been the most dominant and effective. Multiple such programs are largely in place in most of the current market leaders. China’s success, in particular, appears to have been centered around low-cost loans for building and equipping factories, low or zero land costs, plus favored tax treatments, but most recently, it has also started including feed-in tariffs. One advantage that the United States has is the great solar resource in the Southwestern region. This is an ideal environment for the low concentration option. The latter promises a straightforward and immediate path to an LCOE of under 10¢ per kilowatt hour for customers paying retail prices for electricity. One major U.S. opportunity could be to exploit the U.S. current battery-powered cars lead by closely coupling it to solar cell power production with the corresponding major reductions in oil imports and in the carbon footprint from personal transportation based on gasoline fuels. The latest device physics analysis indicates that the steady, continuing, and substantial increases in the performance of solar cells to well above 50% and on into the over 70% efficiency range will continue for many years before approaching fundamental limits, particularly in devices intended for concentrator systems. The latter will continually demand the fabrication of solar cell structures from increas-

FINAL OVERVIEW

607

ingly near-perfect crystalline semiconductor materials. Applications of TF solar cells to digital X-ray imaging are serving a multibillion dollar per year market that could realistically grow to well beyond $10 billion per year over the next 15 years.

26.5

FINAL OVERVIEW

The solar cell or PV effect was first reported by Bequerel in 1839 (see Reference 15, chapter 1). Development of practical sunlight energy conversion devices awaited the development of quantum mechanics and solid-state electronics in the first half of the twentieth century. Applications of these were pioneered by Shockley’s development of bipolar junction transistors in the mid-twentieth century, consisting essentially of two abrupt p/n junctions formed back to back using c-Si [15]. Chapin et al. [16] adapted this technology to demonstrate an abrupt p/n junction c-Si solar cell with a 6% conversion efficiency. Shockley and Queisser’s [17] classic modeling and analysis of abrupt p/n junction solar cells followed shortly thereafter. The ultimate optimization of such an abrupt p/n junction concept is perhaps the 25% efficient PERL cell of Chapter 4. The latter utilized light trapping, surface passivation, and limited-area ohmic contact features incorporated by the use of comparatively expensive photolithography fabrication processes. The cost-performance trade-offs of the terrestrial solar cell electricity market are currently dominated by abrupt p/n junction, c-Si solar cells of about 12–14% efficiencies where the expensive photolithography process has been replaced with the much lower-cost screen printing process. Over the last 4 years or so, the continual price reductions of these types of cells have stagnated (Chapter 2). This has slowed the overall adoption rate of solar cell-generated electric power. However, polycrystalline TF cells (Chapter 6) containing up to 100 times less materials but with 9–10% conversion efficiencies have just begun to make significant market penetration to resume the learning curve price reduction trajectory (see Fig. 2.3). The Nobel Prize-winning development of heterojunctions in III–V semiconducting single crystals in the 1960s and 1970s allowed the fabrication of ohmic contacts for high-current solar cell operation simultaneously with excellent surface passivation. As this was applied to high-current level concentrator solar cells, the champion GaAs single-crystal, single abrupt p/n junction solar cell efficiency of 28% was achieved at a sunlight concentration ratio of ∼240X (Chapter 4). Such high light concentration can offset to a degree the high costs of GaAs heterojunction solar cells that do involve photolithography fabrication. The versatility of III–V semiconductor materials has allowed the extension of this heterojunction technology to solar cells of three different abrupt p/n junctions stacked inside a single device. With each junction optimally selected with different bandgaps, the broadband light from the sun is much better matched for greatly reduced “above bandgap” efficiency losses. A major result is the concentrator champion cell with a 41% efficiency as described in Chapters 13 and 14. In a parallel path, very non-abrupt p-i-n junction solar cell development began in the late 1960s with the pioneering work of Kim and Schwartz [18] in

608

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

single-crystal Ge that was first adapted to amorphous silicon in the mid-1970s by Carlson and Wronski [19]. Decades of dedicated and painstaking work ultimately lead to the p-i-n IBC c-Si cell of Swanson et al. described in Chapter 4. Replacement of photolithography with lower-cost screen printing was one of several innovations that enabled this cell to become commercially successful. Expansion of this p-i-n concept in Japan to include an amorphous silicon-c-Si heterojunction resulted in the commercially successful HIT cell also covered in Chapter 4, which also uses screen printing fabrication. A key aspect of both the IBC and HIT p-i-n c-Si cells is that both are providing commercial quantities of modules with efficiencies in the 18–19% range in contrast to the commercial abrupt p/n junction c-Si devices with efficiencies stuck down in the ∼12–14% range. Even though they are somewhat higher in price, these better performance p-i-n solar cells appear to remain commercially competitive in the market place. As evident from the single-crystal silicon PERL cell, high performance alone does not provide commercial competitiveness. As further illustrated by the diminishing market role of amorphous silicon solar cells, lowest price per watt alone also does not make a solar cell technology competitive (see Fig. 2.9). Market winners provide the best trade-offs and compromises between the total system price and performance. These total system components and cost elements vary dramatically between installations with the thin modules lying flat on a building rooftop versus the much more complex two-axis tracking systems using high sunlight concentration and somewhat thicker modules. A metric that attempts to properly adjust for these different factors is the LCOE. LCOE is the system lifetime estimate of the energy a system produces in dollars (or euros or whatever currency) per kilowatt hour. For systems with dramatically different performance levels (e.g., a factor of 2), a very important performance input parameter is the kilowatt hours produced by the system per year and per square meter of the system facing the sun. This has to be combined with a ground coverage ratio (Chapter 11). This ratio equals one for thin modules covering the flat roof of a building. However, the capacity factor for horizontal flat roof installations is low. Capacity factor is a measure of how well the time variation of the system relative output power matches the time variation of the load that the system supplies. This capacity factor is even lower for fixed mounting on vertical structures like building walls. Typically, thin solar modules are mounted tilted at an optimum fixed angle facing south to increase the capacity factor to around 20% in sunny locations for utility loads (Chapter 10) as well as to increase the energy produced per square meter per year. For this, the ground coverage ratios fall to the order of 50% to avoid excessive shading by adjacent modules. The highest possible capacity factor and energy produced per square meter per year comes with two-axis tracking where the ground coverage ratio can be on the order of 15% depending on location and system design details. The values for all of these parameter inputs for accurate LCOE estimates are just becoming widely available. One has to be careful in comparing results that were not all taken at a single site, but the Table 26.2 results for sunny locations roughly indicate up to a factor of 3 difference in the annual

ABBREVIATIONS

609

average kilowatt hours per square meter per day. For low 6% efficiency performance systems mounted at a fixed tilt, this metric can be as low as 0.35. For a ∼12% fixed tilt, it can rise to ∼0.5–0.6 levels and on up to ∼0.9–1.1 values for two-axis tracking systems with higher ∼17–19% system efficiency values. It is of particular interest that the 1-sun, IBC two-axis trackers can produce this metric’s values of almost 1, in the same general range as two-axis trackers that use the very high-performance three-junction III–V cells that involve the comparatively thicker modules required in high concentration operation. A challenge for such tracking IBC systems is to address the low prices currently approaching $1 per Wp for TF cells albeit at lower 9–10% efficiency values. One immediate and attractive trade-off opportunity is the use of the low 3X concentrator system with the IBC cells to reduce prices down toward the dollar per Wp range if the IBC’s overall high ∼17–19% efficiency performance can be maintained over a 20- to 30-year field installation lifetime. The high-concentration, three-junction, two-axis tracking systems have just begun their learning curve development process. Both their performance and prices should improve substantially with growing experience over the coming years. MJ cells with both heterojunctions and p-i-n features have not yet been reported, but theoretical models suggest such configurations should provide device efficiencies well above 50% if junction open-circuit voltage values can be made to approach bandgap values. This would be accomplished by reducing the “below-bandgap” efficiency losses and would likely require high concentration ratios and “i” regions of near-perfect single-crystal semiconductor material quality and with near-ideal passivation of all its boundaries.

ABBREVIATIONS AC—alternating current AM1.5—air mass 1.5 Au—gold BOS—balance of system Cb—area-related balance of system costs Cd—cadmium Ci—inverter cost CIGS—copper indium gallium diselenide CM—concentrating module Cm—module cost CPU—central processing unit of a computer c-Si—crystalline silicon DVD—digital video disk EPRI—Electric Power Research Institute F—fixed charge rate GaInAs—gallium indium arsenide Ge—germanium

610

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

ha—hours per year of operation HIT—heterojunction with intrinsic thin layer i—intrinsic of undoped semiconducting material IBC—interdigitated back contact III–V—material compounds from columns 3 and 5 of the periodic table InGaP—indium gallium arsenide ISE—Fraunhofer Institute for Solar Energy Systems ISFOC—Institute for Solar Photovoltaic Concentrators kWp—kilowatt peak measured under STC LCD—liquid crystal display LCOE—levelized cost of energy (electricity) MJ—multijunction n—negatively doped semiconductor material NRE—nonrecurring engineering cost NREL—National Renewable Energy Laboratory O&M—operation and maintenance p—positively doped semiconductor material PERL—passive emitter, rear locally diffused PV—photovoltaic or solar cell PVSC—Photovoltaic Specialists Conference PVUSA—a U.S. government and utility-sponsored “Photovoltaics for Utility Scale Applications” test site where solar cell system performance was related to prevailing environmental conditions using regression analysis r—indirect cost S—sunlight intensity at a given site in kilowatt per square meter SAM—Solar Advisor Model SEY—sunlight energy density at a given site in kilowatt hour per square meter per year STC—standard test condition (25°C cell temperature, 1 kW/m2 AM1.5 global sunlight) Te—tellurium TF—thin film USGS—U.S. Geological Survey VF—voltage factor Wp—watt peak under STC η—module efficiency ηF—field efficiency fraction with values between 0 and 1

REFERENCES [1] [2]

G. P. Willeke. Progress in industrial crystalline silicon solar cell technology. In 33rd IEEE Photovoltaic Specialists Conference Record, May 11–16, 2008, San Diego, CA. New York, IEEE (2008). Solarbuzz. Available at http://www.solarbuzz.com/. Accessed July 15, 2009 (2009).

REFERENCES [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]

611

P. Fath. The status of c-Si. Presented at the 34th IEEE Photovoltaic Specialists Conference, June 7–12, 2009, Philadelphia, Penn., N.Y. IEEE (2009). Central Intelligence Agency, U.S. Government, The World Fact Book, Rank Order— GDP (purchasing power parity). Available at www.cia.gov/library/publications/theworld-factbook/rankorder/2001rank.html. Accessed 2008 Sept. 18. S. Hester and T. Hoff. Long term PV module performance. In 22nd IEEE Photovoltaic Specialist Conference Record, pp. 198–202. New York, IEEE (1985). G. B. Haxel and M. Diggles. Rare Earth Elements—Critical Resource for High Technology. Available at http://pubs.usgs.gov/fs/2002/fs087-02/ (2002). Tellurium. Available at http://www.en.wikipedia.org/wiki/Tellurium. Accessed May 7, 2010 (2010). Scott Roberts. Reality check: First Solar and tellurium. Available at http:// seekingalpha.com/article/55845-reality-check-first-solar-and-tellurium Accessed May 7, 2010 (2010). BEA. Bureau of Economic Analysis, U.S. Government, Table 1.19, Implicit Price Deflators for Gross Domestic Product. Available at http://www.bea.gov/national/ nlpaweb/TableView.asp#Mld. Accessed September 8, 2008 (2008). Wikipedia. Trans-Alaska Pipeline System. Available at http://en.wikipedia.org/wiki/ Trans-Alaska_Pipeline_System. Accessed September 26, 2009 (2009). Interstate. 50th Anniversary of the Interstate Highway System. Available at http:// www.fhwa.dot.gov/interstate/faq.htm#question6. Accessed August 22, 2008 (2006). Reuters. US oil import bill to top $400 billion this year. Available at http:// www.reuters.com/article/pressRelease/idUS236508+07-Mar-2008+BW20080307. Accessed September 7, 2008 (2008). zfacts. The Cost of Iraq, Afghanistan, and Other Global War. Available at http:// zfacts.com/p/447.html. Accessed September 7, 2008 (2008). L. D. Partain. Solar Cells and Their Applications, 1st Edition, Chapter 1, L. D. Partain, ed. New York, John Wiley & Sons (1995). W. Shockley. Electrons and Holes in Semiconductors. 1st Edition, New York Van Nostrand Reinhold (1950). D. Chapin, C. Fuller, and G. Pearson. J. Appl. Phys. 25, 676–677 (1954). W. Shockley and J. Queisser. J. Appl. Phys. 32, 510–519 (1961). C. Kim and R. Schwartz. IEEE Trans. Electron Devices 16, 657–663. D. Carlson and D. Wronski. Appl. Phys. Lett. 28, 671–673 (1976). R. Wiser, G. Barbose, C. Peterman, and N. Darghouth. Tracking the Sun II: The Installed Cost of Photovoltaics in the U.S. from 1998–2008. Lawrence Berkeley National Laboratory Report LBNL-1516E, LBNL, October (2009).

INDEX Note: f represents a figure and t represents a table. Abengoa, 274, 275f AC (alternating current), 256, 256f, 386–88, 492–93 accelerated life tests, 237 active pixel sensors, 554 ADCs (analog-to-digital converters), 541f, 542 aerosol optical depth (AOD), 437, 447–49 aerosol printing, 121–22 aliasing, 546, 550–52, 551f aluminum, 10, 53, 118–19, 276 American Society for Testing and Materials (ASTM), 383 AMFPIs (active matrix flat panel imagers), 560, 561f, 562f, 563 Amonix Inc. current status of, 586 HCPV (high-concentration photovoltaic) Fresnel concentrator systems, 365–71, 366f, 367f, 367t, 368f, 369f, 370f, 371f, 372f, 373, 373t MegaModule, 373–75, 374t, 375f amorphous selenium (a-Se). See a-Se (amorphous selenium) amorphous silicon (a-Si). See a-Si (amorphous silicon) Analogic, 507 analog-to-digital converters (ADCs), 541f, 542 AOD (aerosol optical depth), 437, 447–49 application-specific integrated circuits (ASICs), 541–43, 541f

ARCO (Atlantic Richfield Oil Company), 208, 274 Arizona capacity factor in, 234–35, 234f instrumentation/testing in, 226–27 system configuration in, 225–26 system design, installation, 225, 225t system maintenance in, 231–34, 232f, 233t system performance in, 228–30, 229f, 230f TEP Company in, 224, 224f Array Technologies, 212, 212f, 213f arrays Abengoa, 275f body-mounted, 407, 407f concentrating, 412–13, 413f electrostatically clean, 406, 414–15 flexible foldout, 409–10, 409f, 410f flexible roll-out, 410–12 high specific power, 416–17 high-radiation environment solar, 417 high-temperature/high-intensity, 413 photodiodes, TFTs and, 526–28, 527f, 528f photodiodes and, 536, 538f, 539f, 551, 551f rigid panel planar, 408–9, 408f See also space solar arrays a-Se (amorphous selenium) history of, 3 imaging and, 501–2 X-ray imagers, 564–68, 566f, 567f, 568f

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

613

614 a-Si (amorphous silicon) about, 503, 559–60 amorphous tandem cell research, 179, 181t applications for, 5–6, 527–30, 528f, 529f, 602 detectors, 541–43, 541f, 543f, 544f efficiency of, 140, 140f, 140t, 401t HIT cell and, 91–92, 91f, 124 hydrogenated, 150–53, 151f, 152f, 514–16 manufacture of, 188–91, 191t market history of, 501 photodiode arrays and, 536, 538f, 539f, 551, 551f p-i-n photodiodes and, 521–25, 521f, 522f, 524f, 525f p-i-n solar cells and, 77, 96–98, 96f, 97f properties of, 89, 89t, 511–20, 512f, 514f, 515f, 517f, 518f, 520f, 602 SCLI (space charge-limited current) model and, 94 space solar cells and, 404–5 TFTs (thin-film transistors) backpane, 528, 528f X-ray detectors and, 526–27, 527f, 533–35, 534f X-ray imaging and, 9–10, 10f, 555 See also bandgap energy (Eg); FPDs (flat panel detectors); TF (thin-film) solar cells Asia market history, 20f ASICs (application-specific integrated circuit), 541–43, 541f See also readout chips ASTM (American Society for Testing and Materials), 383 AstroEdge arrays, 412 Atlantic Richfield Oil Company (ARCO), 208, 274 atmospheric effects, 433–39, 434f automation and assembly line, 164–68, 165f, 166f, 167f, 168f automobiles, 592–93, 592f azimuth trackers for commercial building flat roofs, 213–14, 214f, 215f geometries for, 210f, 211t structure of, 367–68, 368f See also fixed-tilt trackers; trackers

INDEX band edges, 53, 54, 68–69, 86–87, 103–4 bandgap energy (Eg) definition of, 45–46 efficiencies and, 59, 100f, 299–301 lattice constants and, 324–25, 325f multijunction cells and, 402, 402t open-circuit voltage and, 68–70, 73 See also a-Si (amorphous silicon); CuInSe2-based solar cells bandgap voltage, 98–102, 100f, 101f, 104 batteries, 5, 220, 221f, 489–90 benefit/cost analyses, 216t See also SAM (solar advisor model) BIPV (building integrated photovoltaic), 167–68, 168f, 195t, 197 body-mounted arrays, 407, 407f Boeing 702, 412–13 BOL (beginning of life) efficiencies, 417, 418t BOS (balance of system), 227–28, 228f BPDs (bypass diodes), 130 cadmium telluride (CdTe) solar cells. See CdTe (cadmium telluride) solar cells calculators, 5–6 Canon Medical Systems, 507 carousels, 213–16, 214f, 215f Carrizo Solar Corporation, 208 cars, 592–93, 592f case studies dual-cell Cassegrainian system, 353–54, 353f, 354t on raw costs, 487 reliability model, 237 Cassegrainian HCPV (high-concentration photovoltaic) systems about, 337–38, 357 cell selection for, 356–57 construction of, 347–48 design, operation of, 338–43, 339f, 340f, 341f, 342f dual-cell concentrators, 352–54, 352f, 353f, 354t optical design of, 344–46, 345f, 346f, 347f panels, 350–52, 351f, 354, 355f, 356 two-axis tracking systems, 349–50, 350f See also concentrating optical systems

INDEX CBCT (cone beam computed tomography), 505, 505f, 506f CDS (correlated double sampling), 550 CdTe (cadmium telluride) solar cells about, 142–45, 143f, 144f China and, 177–78, 178f, 181t detectors, 568 efficiencies of, 138–40, 138t, 139t, 140t market history of, 20 space solar cells and, 404–5 Te for, 597–98, 597f X-ray imagers, 569 See also TF (thin-film) solar cells CdZnTe (CZT), 563, 564t, 568, 569 cell efficiencies. See efficiencies central receiver reflectors, 343–44 champion cells, 38, 78–83, 79f, 80t champion GaAs, 88–89, 89t China average annual usable hours, 200, 201t, 202t, 203t, 204t, 205t, 206t CdTe (cadmium telluride) solar cells and, 177–78, 178f, 181t CIGS (copper-indium-gallium-selenium) solar cells and, 175t, 178, 179f, 181t equipment in, 181–83, 181t, 182f, 183f future investment by, 26–27 incentive policy, 198–99 industry chain, 185–91, 185f, 185t, 187t, 189t, 191t industry in, 171–73, 172f, 172t, 174t market growth in, 195–98, 196f, 196t, 197t, 198t market history of, 19–20, 19f module assembly, 192, 192t module prices, 193, 193f, 194f, 194t, 195t R&D, 173–80, 175t, 176t, 177f, 178f, 179f, 180f solar irradiation in, 171–73, 172f, 172t CIGS (copper-indium-gallium-selenium) solar cells, 175t, 178, 179f, 181t, 404–5 circumsolar radiation (CSR), 320–21, 449–54, 451f, 452f, 453f cloud attenuation, 438–39 CMOS (complementary metal oxide semiconductor), 574–75

615 combination systems, 5 commercial azimuth trackers, 213–14, 214f, 215f compound parabolic concentrators (CPCs), 321–22 Compton, 67–68 concentrating arrays, 412–13, 413f concentrating optical systems, 341–44, 342f, 343–44 See also Cassegrainian HCPV (highconcentration photovoltaic) systems concentrating photovoltaic (CPV) costs and, 11–12, 11f definition of, 10 efficiencies of, 340–41, 341f, 389, 390f field installations, 377–80, 379f junctions and, 58f, 59 See also rooftop CPV (concentrating photovoltaic) systems concentrating photovoltaic (CPV) systems about, 361, 365, 606 early development of, 273–74, 274f, 275f, 276 rating standards for, 383–85, 384t, 386f See also power plants concentrator point contacts, 89–90, 90f concentrators costs of, 273, 290–91, 290t, 342–43, 342f current status of, 585, 585f design of, 298–99, 298f efficiency vs., 340–41, 341f market history of, 38–39, 64 multijunctions and, 304, 305f reflective, 361–63, 362f, 433–36, 434f refractive, 363–65, 363f, 364f See also crystalline silicon (c-Si) modules; Fresnel lens concentrators; TF (thin-film) modules Concentrix Solar, 327–30, 328f, 329f conduction band junctions, diodes and, 53, 54f open-circuit voltage and, 68–70 power curves and, 54, 55f valence bands and, 512f, 513 wave theory and, 49–50, 51f, 52 cone beam computed tomography (CBCT), 505, 505f, 506f correlated double sampling (CDS), 550

616 cost/benefit analyses, 216t See also SAM (solar advisor model) costs concentration and, 273, 290–91, 290t, 342–43, 342f for CPV (concentrating photovoltaic), 11–12, 11f current status of, 590–91 economic modeling, 586–87 efficiencies and, 262–63, 262f of FITs (feed-in tariffs), 36 for large-scale solar power production, 486–89 power plants, 265–67, 266f for PV (photovoltaic) modules, 193, 193f, 194f, 194t, 195t for PV (photovoltaic) systems, 59, 60–61t, 62–63, 596 for SAM (solar advisor model), 465–67, 466t, 470, 470t of solar cell modules, 8–9, 9f, 17–20, 19f, 23–24 for solar trackers, 215–17, 216t for Springerville Generating State Solar System, 243–46, 243t, 245t See also LCOE (levelized cost of energy) CPCs (compound parabolic concentrators), 321–22 CPV (concentrating photovoltaic). See concentrating photovoltaic (CPV) CPV (concentrating photovoltaic) systems. See concentrating photovoltaic (CPV) systems CR technology, 501 crystalline silicon (c-Si) about, 102–4, 113–14, 133–34 China and, 176, 177f concentrator point contacts, 89, 90f efficiency limitations of, 116–20 HIT cell, 91–92, 91f ingots, wafers, 187–88, 189t manufacture of, 188–90 market history of, 21–22 nc-Si (nanocrystalline silicon), 150–53, 151f, 152f PERL cell, 90 point contact crystalline silicon cells, 83–89, 84f, 85f, 86f, 88f, 89t

INDEX PV (photovoltaic) cells and, 114–16, 116f technologies using, 120–29, 123f, 126f, 127f, 128f See also low-concentration crystalline silicon systems crystalline silicon (c-Si) modules about, 606 current status of, 584–85, 584f, 585f opportunities for, 590–91, 590f PV (photovoltaic) modules, 129–33 See also concentrators; TF (thin-film) modules crystals, 63–64, 64f CSR (circumsolar radiation), 320–21, 449–54, 451f, 452f, 453f CuInSe2-based solar cells, 142–43, 146–50, 146t, 147f, 148f See also bandgap energy (Eg) current collecting technologies, 120–22 CZ (Czochralski process), 114–16, 116f CZT (CdZnTe), 563, 564t, 568, 569 D/A (dual-axis) trackers, 208–11, 210f, 210t, 211f Day4, 126–29, 126f, 127f, 128f, 132–33 DC (direct current) ratings, 386, 387f deep discharge batteries, 5 demonstration plants. See ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) dental procedures, 504f, 505 Department of Defense, U.S., 25 detectors. See flat panel detectors (FPDs); photoconductors; X-ray detectors D4, 126–29, 126f, 127f, 128f, 132–33 digital imaging, 501–2 See also portal imaging digital lithography, 528–29, 529f diodes, 53–54, 54f, 130 See also p-i-n photodiodes; P/N junction diodes; Shockley diodes direct beams, 371, 433, 434f, 439–40 direct current (DC) ratings, 386, 387f DNI (direct normal irradiance) future comparisons, 60–61t, 62–63 rating standard and, 383–84, 384t

INDEX scattering and, 437–38, 438f SMARTS model and, 306–7, 307f, 437, 438f See also irradiance double heterostructures, 79–80, 79f DR technology, 501–6, 503f, 504f, 505f, 506f dual-axis (D/A) trackers, 208–11, 210f, 210t, 211f dye-sensitized solar cells, 179–80, 180f, 181t efficiencies under AM1.5, 299–301, 300f of a-Si (amorphous silicon), 140, 140f, 140t, 401t bandgap energy (Eg) and, 59, 100f, 299–301 BOL (beginning of life), 417, 418t China and, 180, 181t costs and, 262–63, 262f of CPV (concentrating photovoltaic), 340–41, 341f, 389, 390f history of, 4 increasing, 55–59, 55f, 56f, 58f, 59f inverter conversion, 471–74, 472t, 473t, 474t limitations of, 116–20 measured field performances, 593–96, 594t PV (photovoltaic) cells and, 115–20, 400–401, 401t QEs (quantum efficiencies), 46, 55, 522 space solar cells, 417–18, 418t for TF (thin-film) solar cells, 138–40, 138t, 139t, 140t of triple-junction solar cells, 324–25, 326f 28.2%, 78–83, 79f, 80t Eg (bandgap energy). See bandgap energy (Eg) Einstein, Albert, 44–45, 67–68 electricity applications for, 6–7, 6f, 7f arguments for, 11–13, 11f market history of, 17–22, 18f, 19f, 20f, 21f, 36 projected future of, 28–30

617 electroless plating, 120–21 electromagnetic radiation, 47–48 electromagnetic waves, 44–45, 44f, 45f electron volts, 44–47 electron wave theory, 49–53, 50f, 51f, 52f electronic portal imaging detectors (EPIDs), 534 electrons, 47–48, 48f, 50–51, 55 electrostatically clean arrays, 406, 414–15 Emcore, 401–3 energy, definition of, 44 ENTECH, 274, 275f EOL power, 406, 417–20, 418t, 419t EPIDs (electronic portal imaging detectors), 534 EUCLIDES, 377–78, 378f European market history, 19–20, 19f, 20f, 21f EVA (ethylene vinyl acetate), 161–62, 163f, 222f, 223 extraterrestrial irradiance, 427–32, 428f, 429f, 432f extraterrestrial spectrum, 430–32, 432f Exxon Mobile, 17, 18f failure modes and effects analysis (FMEA), 236 feed-in tariffs (FITs) about, 207, 208 costs and, 36 countries and, 25–26, 589, 589f, 606 installation rates and, 20f for large-scale solar power production, 488–89 PURPA (Public Utilities Regulatory Power Act) and, 23, 25–26 See also governments Fermi energy, 71, 513 Fermi levels ohmic contacts, heterojunction interfaces and, 78 p/n, p-i-n crystalline silicon cells and, 86–87 Shockley diode model and, 70–71, 71f, 73 FF (fill factor), 76–77, 77f, 116, 326f field performances, 370, 370f, 593–96, 594t

618 figures of merit, 417–20, 418t, 419t film imaging, 501–2 First Solar, 145, 597–98 FITs (feed-in tariffs). See feed-in tariffs (FITs) fixed-tilt trackers about, 209f, 263, 595 D/A (dual-axis) trackers vs., 210f, 210t irradiance and, 441–46, 443f linear axis trackers vs., 212, 212f, 213f See also trackers flat panel detector (FPD) market about, 499–500, 500f, 508 evolution of, 506–7 future of, 508–9 history of, 500, 501–2, 507–8 flat panel detectors (FPDs) about, 555 applications for, 502–3, 503f, 504f, 505–6, 505f, 506f construction of, 535–36, 535f, 536f, 537f future trends for, 508–9 noise from, 544–52, 545f flat panels, 361, 491f, 502, 574–75 FLATCON modules, 327–30, 328f, 329f Flat-Plate Solar Array Project (FSA), 18f, 21f, 22 flat-plates (FPs) collectors, 433–34, 434f, 441, 445 modules, 4, 460 flexible foldout arrays, 409–10, 409f, 410f flexible roll-out arrays, 410–12 flexible substrates, 527–28, 574, 575 Florida Power and Light (FPL), 251, 252f FMEA (failure modes and effects analysis), 236 FPDs (flat panel detectors). See flat panel detectors (FPDs) Fraunhofer Institute for Solar Energy, 314–15, 327 Fresnel lens concentrators about, 313–16, 315f, 331, 343–44 applications for, 330, 330f developments of, 327–30, 328f, 329f elements of, 316–22, 316f, 317f, 320f, 321f

INDEX HCPV (high-concentration photovoltaic) Fresnel concentrator systems, 365–71, 366f, 367f, 367t, 368f, 369f, 370f, 371f, 372f, 373, 373t history of, 274, 274f, 275f refraction of light rays and, 364, 364f triple-junction solar cells, 322–27, 323f, 325f, 326f, 403 FSA (Flat-Plate Solar Array Project), 18f, 21f, 22 Fuji Photo Film Co., 501, 565–66, 566f GaAs (gallium arsenide) electron movement in, 52–53, 52f InGaP and, 352, 352f, 353, 356–57 junctions and, 46–47, 46f, 607 p-on-n, n-on-p, 78–83, 79f, 80t wave functions for, 56–58 gain effect, 552, 553f GaSb (gallium antimonide) IR solar cells, 353–54, 357 junctions and, 46–47, 46f single crystals and, 63 wave functions for, 57–58 GE (General Electric), 507 Germany 100,000 Roofs Program, 21f, 25–26 market history of, 20, 20f solar cell budget of, 604t ghosting, 567–68, 568f global comparisons, 173, 174t governments economic development programs by, 591 incentive policies, 33, 35–36, 36t, 467 investments by, 587, 591, 603, 603–4 public policies and, 22–28, 37, 113, 198–99 subsidy programs and, 133 See also feed-in tariffs (FITs); specific countries grid-connected PV power systems, 182–83, 182f, 183f, 184f HCPV (high-concentration photovoltaic) systems comparison of, 60–61t, 62–63

INDEX Fresnel concentrator systems, 365–71, 366f, 367f, 367t, 368f, 369f, 370f, 371f, 372f, 373, 373t multijunctions in, 293–95, 294f, 297f, 373–75, 374t, 375f See also Cassegrainian HCPV (highconcentration photovoltaic) systems; III-V multijunctions heterojunctions, 78, 607–9 HgI2 imagers, 569–74, 570f, 571f, 572f, 573f high-concentration Cassegrainian solar cells. See Cassegrainian HCPV (high-concentration photovoltaic) systems high-concentration Fresnel lens systems. See III-V multijunctions HIT (heterojunction with intrinsic thin-layer) cells, 89t, 91–92, 91f, 124, 608 hospitals DR technology applications in, 502–3, 503f, 504f, 505–6, 505f, 506f imaging history of, 501–2 market history of, 500, 500f See also X-ray detectors Hubbert’s Peak (Deffeyes), 12 hydrogenated a-Si (amorphous silicon), 150–53, 151f, 152f, 514–16 IBC (interdigitated back contact) cells about, 83–89, 84f, 85f, 86f, 88f, 89t, 90f novel types of, 124–26 p-i-n solar cells and, 608–9 See also Shockley diodes IEC (International Electrotechnical Commission) standard 60904, 448–49 standard 61724, 228 standard 62108, 380–82, 390, 391, 392 TC82 Working Group 7, 383–85, 384t, 386f IEC (Israel Electric Corporation), 490, 491f image lag about, 552, 553f ghosting vs., 567–68, 568f of PbI2, HgI2 imagers, 571, 571f TFTs (thin-film transistors) and, 526, 540–41

619 image sensors, 527–28 imaging, 501–2 See also AMFPIs (active matrix flat panel imagers); portal imaging; X-ray imaging IMM (inverted metamorphic) multijunction cells, 301–3, 302f, 303f, 324, 402–3 incident irradiance, 427 India market history, 19–20, 19f industry about, 133–34 in China, 171–73, 172f, 172t, 174t equipment and, 164–68, 165f, 166f, 167f, 168f future of, 168–69 growth in, 113–14, 159–61 initial investment in, 261–65, 262f, 264f, 265f landscape of, 507–8 recommendations for, 603–5, 604t industry chain, 185–91, 185f, 185t, 187t, 189t, 191t See also polysilicon infrared-sensitive PV (photovoltaic) cells, 5, 63 ingots, 176, 177f, 187–88, 189t ink-jet printing, 121 Institute for Solar Energy (ISE) Fraunhofer, 121–22 Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC). See ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) insulators, 50 integrated power systems, 415–17 interdigitated back contact (IBC) cells about, 83–89, 84f, 85f, 86f, 88f, 89t, 90f novel types of, 124–26 p-i-n solar cells and, 608–9 See also Shockley diodes International Electrotechnical Commission (IEC). See IEC (International Electrotechnical Commission) International Space Station (ISS) Si solar cells, 397–98, 398f, 399f, 400–401, 400f Interstate Highway System, U.S., 27–28 inverted metamorphic (IMM) multijunction cells, 301–3, 302f, 303f, 324, 402–3

620 inverters conversion efficiencies of, 471–74, 472t, 473t, 474t lifetime analysis of, 478–79 performance models for, 469t irradiance Chinese statistics, 201t, 202t, 203t, 204t, 205t, 206t extraterrestrial, 427–32, 428f, 429f, 432f for fixed-tilt, tracking, concentrating geometries, 441–46, 443f solar, 171–73, 172f, 172t spectral, 306, 446–47, 447f terrestrial applications and, 439–40 See also DNI (direct normal irradiance); TSI (total solar irradiance) ISE (Institute for Solar Energy) Fraunhofer, 121–22 ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) demonstration plants, 391f future work of, 391 IEC 62108 tests and, 381–82 lessons learned, 390–91 plant acceptance by, 382–83 plant installations, 377–80 radiation results, 388, 388f, 389f, 390f rating standards for, 383–88, 387f technologies selected by, 379f Israel Electric Corporation (IEC), 490–91, 491f ISS (International Space Station) Si solar cells, 397–98, 398f, 399f, 400–401, 400f Italian Space Agency, 405 Japan budget of, 604t market history of, 19–20, 19f, 20f, 21f production in, 23–25 JPL (U.S. Jet Propulsion Laboratory), 18f, 21f, 22–23 junctions CPV (concentrating photovoltaic) and, 57–59, 58f GaAs (gallium arsenide) and, 46–47, 46f, 607 open-circuit voltage and, 80

INDEX P/N junction diodes, 53–54, 54f See also heterojunctions; multijunction (MJ) solar cells; III-V multijunctions; tunnel junctions JX Crystals Inc., 60–61t, 62, 213, 214f, 215f See also 3-sun LCPV (low-concentration photovoltaic) concept Kodak, 507, 565, 566f Kroemer, Herbert, 79–80 kTC noise, 549–50, 549f large-scale solar power production about, 25, 483–84, 494–95 costs for, 486–89 problem with, 484, 485f space, time requirements for, 485–86 technological challenges of, 489–94, 491f Las Vegas, Nevada, 60–61t, 62, 594t, 595 See also University of Nevada, Las Vegas laser buried grid PV cells, 122–23, 123f laser-fired contact, 122 laws. See public policies LCOE (levelized cost of energy) about, 252–53, 269 drivers of, 251–52 estimation of, 30, 31t, 32t, 33–36, 33t, 34f, 36t forecasting tool, 268, 268f inputs for, 253–54 SAM (solar advisor model) and, 465–67, 466t sensitivity of, 255, 255f for solar cell systems, 599, 600f system lifetime and, 477, 478f, 479, 479f utility-scale systems and, 255–60, 256f, 258f, 259f, 260f See also costs LCPV (low-concentration photovoltaic) Abengoa arrays, 275f comparison of, 60–61t, 62 future of, 285, 287–89, 287f, 287t, 288f, 288t, 289f

INDEX mirror module development, 279–83, 279t, 280f, 281f, 282f, 283f, 284f, 285, 285f, 286f predictions vs. goal comparison, 290t See also 3-sun LCPV (low-concentration photovoltaic) concept levelized cost of energy (LCOE). See LCOE (levelized cost of energy) light history, 44–45 light rods, 321–22 light-induced plating, 120–21 line noise, 548–49, 548f linear axis trackers, 212, 212f, 213f lithography, 528–29, 529f Livermore, California, 444–46 low-concentration crystalline silicon systems about, 273, 290–91 early development of, 273–74, 274f, 275f, 276 future cost of, 290, 290t future manufacture of, 285–89, 286f, 287, 287f, 287t, 288f, 288t, 289f mirror module development for, 279–83, 279t, 280f, 281f, 282f, 283f, 284f, 285, 285f See also crystalline silicon (c-Si); nc-Si (nanocrystalline silicon); point contact crystalline silicon cells; 3-sun LCPV (low-concentration photovoltaic) concept LS-PV systems, 197, 198t mammography, 505, 507, 566–68, 566f, 568f manufacturers current status of, 587, 588t, 589 history of, 137, 140, 140t initial investment of, 261–65, 262f, 264f, 265f markets a-Si (amorphous silicon) history, 501 changes in, 583 of concentrators, 38–39, 64 demand by, 589, 589f history of, 17–22, 18f, 19f, 20f, 21f, 36 for solar cells, 4–7, 589, 589f See also China; flat panel detector (FPD) market

621 Mars, 407, 407f, 411–12, 415 See also space solar arrays mc (monocrystalline). See crystalline silicon (c-Si); crystalline silicon (c-Si) modules; low-concentration crystalline silicon systems; nc-Si (nanocrystalline silicon); point contact crystalline silicon cells measured field performances, 593–96, 594t MegaModules, 366–67, 367f, 367t, 369f, 373–75, 374t, 375f MESSENGER, 413–14 metals, 50 Middle East, 12 Mie scattering, 436 military, 24–25 models inverter performance, 469t reliability/availability, 237–38, 238f, 239t, 240t, 241t, 242f, 242t, 243 SCLI model, 93–98, 93f, 96f, 97f See also Shockley diodes; SMARTS (Simple Model of the Atmospheric Radiative Transfer of Sunshine) model module performance models, 469t modules about, 140t, 220, 222–23, 222f assembly of, 161–64, 162f, 163f, 192, 192t costs, prices of, 193, 193f, 194f, 194t, 195t Day4 technology and, 132–33 degradation of, 475, 475f, 476f, 477f equipment for, 164–68, 165f, 166f, 167f, 168f future of, 168–69 materials for, 159–61, 160f production technology for, 129–32 monolithic multijunction solar cells, 338–41, 339f, 340f, 341f Moore’s law, 29 MTF (modulation transfer function), 551–52, 551f, 565, 566f multijunction (MJ) solar cells concentrators and, 304, 305f design of, 295–96, 295f

622 multijunction (MJ) solar cells (cont’d) in HCPV (high-concentration photovoltaic) systems, 293–95, 294f, 297f, 373–75, 374t, 375f inverted metamorphic (IMM), 301–3, 302f, 303f, 324, 402–3 monolithic, 338–41, 339f, 340f, 341f operating condition performance, 304–8, 305f, 307f reliability of, 308–9 in space, 402–4, 402t See also Cassegrainian HCPV (highconcentration photovoltaic) systems; III-V multijunctions; triple-junction solar cells Nankai University, 178–79, 179f, 180f nanostructured solar cells, 404 NASA. See space solar cells National Development and Reform Committee, 197–99, 197t, 198t National Solar Radiation Data Base, U.S., 440 natural gas depletion, 12 nc-Si (nanocrystalline silicon), 150–53, 151f, 152f See also crystalline silicon (c-Si); low-concentration crystalline silicon systems; point contact crystalline silicon cells NEDO (New Energy and Industrial Technology Development Organization), 24–25 New South Wales, University of, 122–23, 123f noise, 544–52, 545f, 548f, 549f NREL (National Renewable Energy Laboratory), 124, 306, 324, 440 NSRDB (National Solar Radiation Data Base), 440 nuclear fuels, 12 ohmic contacts, 78 oil depletion, 12 100,000 Roofs Program, 21f, 25–26 open-circuit voltage, 68–70, 73–74, 75f, 80–81, 82f

INDEX optical losses, 117 See also Fresnel lens concentrators organics, 405 ORNL (Oak Ridge National Laboratory), 283–84, 285f, 286f Pacific Gas & Electric Company (PG&E), 251–52 PbI2 imagers, 569–74, 570f, 571f, 572f, 573f PERC (passivated emitter and rear cell) PV (photovoltaic) cells, 123 periodic table, 48, 49t, 50–52 PERL (passivated emitter, rear locally diffused) PV (photovoltaic) cells, 89t, 90, 123–24 See also Shockley diodes PG&E (Pacific Gas & Electric Company), 251–52 photoconductors detectors, 559, 563, 564t imagers, 563–75, 566f, 567f, 568f, 570f, 571f, 572f, 573f See also X-ray detectors photodiodes about, 521–25, 521f, 524f, 525f arrays, 536, 538f, 539f, 551, 551f TFTs (thin-film transistors) and, 526–28, 527f, 528f photoinduced metal plating, 120–21 photolithography, 528–29, 529f photons, 45–47, 45f, 46f photovoltaic (PV) solar cells. See PV (photovoltaic) solar cells Physikalisch-Meteorologisches Observatorium Davos (PMOD) dataset, 428–30, 428f, 429f p-i-n photodiodes about, 521–25, 521f, 522f, 524f, 525f lag, gain effect and, 552, 553f p-i-n solar cells about, 77, 102–4, 607–8 advanced concept, 81–83, 82f a-Si (amorphous silicon) and, 77, 96–98, 96f, 97f concentrator point contacts, 89–90, 90f HIT cell, 91–92, 91f PERL cell, 90 photodiodes vs., 522–23

INDEX point contact crystalline silicon cells, 83–89, 84f, 85f, 86f, 88f, 89t p-i-n/TFT pixel architecture, 537–41, 538f, 539f See also TFTs (thin-film transistors) planar collectors, 434, 441–46, 443f planar PV (photovoltaic), 60–61t Planck’s law, 431–32, 432f PMOD (Physikalisch-Meteorologisches Observatorium Davos) dataset, 428–30, 428f, 429f P/N junction diodes about, 53–54, 54f, 55f, 102–4 concentrator point contacts, 89–90, 90f HIT cell, 91–92, 91f PERL cell, 90 p-i-n solar cells and, 77 Shockley diodes and, 70–74, 88–89 point contact crystalline silicon cells, 83–89, 84f, 85f, 86f, 88f, 89t See also crystalline silicon (c-Si); lowconcentration crystalline silicon systems; nc-Si (nanocrystalline silicon) politics, 484 See also public policies polycrystalline CuxS/CdS, 93f, 94 Si (silicon) TFTs, 574 TJ (triple junction) cells and, 403 types of, 142–50, 143f, 144f, 146t, 147f, 148f polysilicon, 175–76, 176t, 186–87, 187t, 261–62 portable radiography detectors, 535–36, 535f, 536f, 537f portal imaging, 506, 506f, 534 See also digital imaging; X-ray imaging power, definition of, 44 power plants acceptance of, 382–83 expenses of, 265–67, 266f initial investment in, 261–65, 262f, 264f, 265f installation procedures for, 378–80, 379f power stations, 181–83, 182f, 184f power systems, 415–20, 418t, 419t power transmission, 492–93 prices current status of, 598–602, 598t, 599f, 599t, 600f, 601t, 602t

623 for PV (photovoltaic) modules, 193, 193f, 194f, 194t, 195t See also costs printing, 118–19, 121–22 progressive scanning, 542–43, 543f, 544f prototype tests, 382 public policies, 22–28, 37, 113, 198–99 PURPA (Public Utilities Regulatory Power Act), 6, 18f, 23, 25–26 PV (photovoltaic) solar cells c-Si (crystalline silicon) and, 114–16, 116f current collecting technologies for, 120–22 efficiencies of, 115–20, 400–401, 401t infrared-sensitive, 5, 63 novel types of, 122–29, 123f, 126f, 127f, 128f See also LCPV (low-concentration photovoltaic); solar cells; specific types PV (photovoltaic) systems components of, 220, 221f, 222–23, 222f costs of, 59, 60–61t, 62–63, 596 history of, 5–7, 6f, 7f, 8f lifetime analysis of, 477, 478f, 479f reliability of, 235–41, 236f, 238f, 239t, 240t, 241t, 242f, 242t, 243 residual value of, 267 See also China; concentrating photovoltaic (CPV) systems quantum efficiencies (QEs), 46, 55, 522 quantum mechanics, 47–48, 67–68 quantum well (QW) solar cells, 98–104, 100f, 101f quasi-Fermi energy, 69 quasi-Fermi levels, 70–77, 71f, 87, 97–99 Queisser, 70–77, 75f, 77f radiation electromagnetic, 47–48 ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) results for, 388, 388f, 389f, 390f solar, 433, 434f, 440–41, 454–56

624 radiography detectors, 535–36, 535f, 536f, 537f rainbows, 44–47 rating standards, 383–88, 384t, 386f, 387f Raviv model, 486–87, 489–90, 494 Rayleigh scattering, 436 RBD (reliability block diagram), 235–37 readout chips, 536, 537f, 541f See also ASICs (application-specific integrated circuit) recombination losses, 118 reference spectra, 447–48, 448f reflective concentrators, 361–63, 362f, 433–36, 434f refractive concentrators, 363–65, 363f, 364f reliability block diagram (RBD), 235–37 reliability/availability model, 237–38, 238f, 239t, 240t, 241t, 242f, 242t, 243 renewable energy development, 197–98, 197t, 198t resistivity losses, 118–20 rigid panel planar arrays, 408–9, 408f Robbins Engineering, Inc., 208 robotics and assembly line, 164–68, 165f, 166f, 167f, 168f rooftop CPV (concentrating photovoltaic) systems about, 7, 7f, 216, 593 applications for, 330, 330f on commercial buildings, 213–14, 214f, 215f comparison of, 140, 140f See also 3-sun LCPV (low-concentration photovoltaic) concept S/A (single-axis) trackers, 209–11, 211t, 212, 212f, 213f, 215f Sacramento, California, 444–46 SAM (solar advisor model) about, 463–65, 481–82 analysis examples of, 471–81, 472t, 473t, 474t, 475f, 476f, 477f, 478f, 479f, 480f, 481f costs for, 465–67, 466t, 470, 470t software for, 467, 468t, 469t, 470, 470f, 470t SAM (sun and aureole measurement), 450, 451f

INDEX Sandia National Laboratories Fresnel lenses at, 314 performance models for, 469t solar cell system reliability, 234–41, 236f, 238f, 239t, 240t, 241t, 242f, 242t, 243 SANYO Electric Company, 124, 287–88, 287t, 288t satellites, 427–32, 428f, 429f, 432f See also space solar arrays SC (solar constant), 428–30, 428f, 429f SCARLET concentrator arrays, 412 scintillators, 535–36, 560, 561f SCLI (space charge-limited current) model, 93–98, 93f, 96f, 97f SCOPAs (survivable concentrating photovoltaic arrays), 414 screen printing, 118–20 SEIA (Solar Energy Industries Association), 246–47 selective emitters, 119–20 semiconductors, 49–53, 50f sensor arrays. See arrays Shockley diodes about, 70–77, 71f, 75f, 77f PERL cells and, 90 P/N junction diodes and, 70–74, 88–89 See also IBC (interdigitated back contact) cells; PERL (passivated emitter, rear locally diffused) PV (photovoltaic) cells short-circuit current (Isc), 74, 75f, 116, 117 shot noise, 550 Siemens, Werner, 3 silicon about, 3–4, 51–52, 51f PV (photovoltaic) modules, 159–61, 160f single-crystal solar cells, 57, 63–64, 64f substrate thin-film solar cells, 179 TFTs (thin-film transistors), 574 See also a-Si (amorphous silicon) simple efficiency model for flat-plate modules, 469t single-axis (S/A) trackers, 209–11, 211t, 212, 212f, 213f, 215f single-point efficiency model, 469t smart grids, 493–94

INDEX SMARTS (Simple Model of the Atmospheric Radiative Transfer of Sunshine) model CSR (circumsolar radiation) and, 451, 452f DNI (direct normal irradiance), 306–7, 307f, 437, 438f terrestrial applications of, 446–47, 447f solar arrays. See arrays; space solar arrays solar cell efficiencies. See efficiencies solar cell electricity. See electricity solar cell systems. See PV (photovoltaic) systems solar cells applications for, 602 current status of, 584, 584f history of, 3–4, 607–9 installation rate of, 20f manufacturing companies of, 587, 588t, 589 market and, 4–7, 589, 589f modules, 8–9, 9f, 17–20, 19f, 23–24 opportunities for, 590–93, 590f photoconductor detectors vs., 559 prices for, 598–602, 598t, 599f, 599t, 600f, 601t, 602t space satellites and, 427–32, 428f, 429f, 432f types of, 8, 8f See also multijunction (MJ) solar cells; PV (photovoltaic) solar cells; space solar cells; triple-junction solar cells solar constant (SC), 428–30, 428f, 429f Solar Energy Industries Association (SEIA), 246–47 solar intensity, 44 solar irradiation, 171–73, 172f, 172t solar power production. See large-scale solar power production solar radiation, 433, 434f, 440–41, 454–56 solar trackers. See trackers solar water pumping, 5 solar/wind complementary stations, 181–82 SolFocus Cassegrainian concentrator panels, 348, 348f solid-state electronics, 29, 67–68, 70–77, 71f, 75f, 77f Sommerfeld, A., 49

625 SORCE (Solar Radiation and Climate Experiment) measurements, 429, 429f space charge-limited current (SCLI) model, 93–98, 93f, 96f, 97f space satellites, 427–32, 428f, 429f, 432f space ships, 3 space solar arrays about, 406, 407t body-mounted, 407, 407f concentrating, 412–13, 413f electrostatically clean, 414–15 flexible foldout, 409–10, 409f, 410f flexible roll-out, 410–12 high-temperature/high-intensity, 413 Mars solar arrays, 415 rigid panel planar, 408–9, 408f See also arrays space solar cells about, 397–98, 398f, 399f, 400–401, 400f CIGS, a-Si, CdTe, 404–5 current status of, 586 efficiencies of, 417–18, 418t III-V space cells, 401–2 integrated power systems and, 415–17 MJ (multijunction) cells, 402–4, 402t nanostructured solar cells, 404 organics and, 405 See also solar cells spectral irradiance, 306, 446–47, 447f See also irradiance spectral mismatch calculations, 448–49 Spectrolab, 401–2 split data lines, 542–43, 543f, 544f Springerville Generating State Solar System assessment of, 246 BOS (balance of system), 227–28, 228f capacity factor at, 234–35, 234f case study on, 237 configuration of, 225–26 costs for, 243–46, 243t, 245t design, installation of systems, 225, 225t instrumentation/testing at, 226–27 maintenance of, 231–34, 232f, 233t performance of, 228–30, 229f, 230f TEP Company and, 224, 224f Sputnik, 3 SquareRigger, 410–11 Staebler-Wronski effect, 150 STC efficiencies, 593–96, 594t

626 storage challenges, 489–92, 491f SunPower Corporation IBC (interdigitated back contact) cells, 124–26 initial investment of, 261–63, 264f LCOE (levelized cost of energy) forecasting tool, 268, 268f system performance prediction and, 260 trackers, 257, 257f, 258f, 261f sunshape, 450 SUPERs (survivable power systems), 414 survivable concentrating photovoltaic arrays (SCOPAs), 414 system-based analyses, 464–65 TAB (tape-automated bonding), 536, 537f Taiwan market history, 19–20, 19f tellurium, 597–98, 597f temperature degradation, 475–77, 476f, 477f TEP (Tucson Electric Power) Company. See Springerville Generating State Solar System terrestrial solar cells. See solar cells terrorists, 12 tests, 380–82 TF (thin-film) arrays, 410–12 TF (thin-film) modules, 584–85, 584f, 585f, 590–91, 590f, 606–7 See also concentrators; crystalline silicon (c-Si) modules TF (thin-film) solar cells China and, 176–80, 177f, 178f, 179f, 180f future of, 153 history of, 137–41, 138t, 139t, 140f, 140t single-crystal solar cells vs., 63–64, 64f types of, 141–42, 141f X-ray imaging and, 499–500 See also a-Si (amorphous silicon); CdTe (cadmium telluride) solar cells TFTs (thin-film transistors) a-Se X-ray imagers and, 564–65 a-Si (amorphous silicon), 516–20, 517f, 518f, 520f CdTe (cadmium telluride), CZT X-ray imagers, 569 future of, 574 image lag and, 526, 540–41 PbI2, HgI2 imagers and, 571–73

INDEX sensor arrays and, 526–28, 527f, 528f See also p-i-n/TFT pixel architecture thermal noise, 547–48 thermophotovoltaic (TPV) cells, 5 3-sun LCPV (low-concentration photovoltaic) concept about, 276, 276f, 277f, 278, 278f, 279 future of, 285, 287–89, 287f, 287t, 288f, 288t, 289f mirror module development, 279–83, 279t, 280f, 281f, 282f, 283f, 284f, 285, 285f, 286f III-V multijunctions design of, 295–99, 295f, 297f, 298f efficiency of, 299–301, 300f in HCPV (high-concentration photovoltaic) systems, 293–95, 294f, 297f history of, 607 performance of, 304–8, 305f, 307f reliability of, 308–9 tunnel junctions in, 296–97, 297f See also multijunction (MJ) solar cells III-V space cells, 401–2 tilted collectors, 433–36, 434f TIR (total internal reflection), 345, 356f total incident radiation, 468, 470f total solar irradiance (TSI), 428–30, 428f, 429f TPV (thermophotovoltaic) cells, 5 trackers about, 5, 207, 608–9 costs of, 215–17, 216t CPV (concentrating photovoltaic) and, 10 D/A (dual-axis), 208–11, 210f, 210t, 211f geometries for, 209–11, 209f, 210f, 210t, 211f, 211t HCPV (high-concentration photovoltaic) Fresnel concentrator systems, 367–69, 368f linear axis, 212 manufacturing companies of, 208–9 S/A (single-axis), 209–11, 211t, 212, 212f, 213f, 215f SunPower and, 257, 257f, 258f, 261f two-axis Cassegrainian systems, 349–50, 350f See also azimuth trackers; fixed-tilt trackers

INDEX transmission challenges, 492–93 triple-junction solar cells, 322–27, 323f, 325f, 326f, 403 trucks. See automobiles TSI (total solar irradiance), 428–30, 428f, 429f Tucson Electric Power (TEP) Company. See Springerville Generating State Solar System tunnel junctions, 296–97, 297f, 339, 340f See also multijunction (MJ) solar cells; III-V multijunctions Ultraflex Solar Arrays, 411–12 United States advantages of, 606 market history of, 19–20, 19f, 20f, 21f public policies in, 22–28 solar cell budget of, 604, 604t University of Nevada, Las Vegas CPV (concentrating photovoltaic) systems at, 361 MegaModule at, 374, 375f reflective concentrators at, 361–63, 362f refractive concentrators at, 363–65, 363f, 364f 3-sun modules, 285, 285f, 286f University of New South Wales, 122–23, 123f U.S. Department of Defense, 25 U.S. Interstate Highway System, 27–28 U.S. Jet Propulsion Laboratory (JPL), 18f, 21f, 22–23 U.S. National Solar Radiation Data Base, 440 utility-scale systems BOS (balance of system), 227–28, 228f capacity factor of, 234–35, 234f configuration of, 225–26 design, installation of systems, 225, 225t instrumentation/testing of, 226–27 maintenance of, 231–34, 232f, 233t performance of, 228–30, 229f, 230f TEP Company and, 224, 224f

627 valence bands, 68–70, 512f, 513 Van Allen belts, 397–98 Vanguard cells, 397–98 Varian, 275f, 276 Varian Medical Systems 4030CB, 536, 537f voltage. See bandgap voltage; open-circuit voltage wafers, 187–88, 189t Wakondo Technologies, 403–4 wars, 12 water pumping, 5 wave theory, 49–53, 50f, 51f, 52f wave-particle duality, 47–48 Webster effect, 88 Wilson, A. H., 49 wind/solar complementary stations, 181–82 X-ray detectors about, 533–35, 534f construction of, 535–36, 535f, 536f, 537f electronics architecture of, 541–43, 541f, 543f, 544f photoconductor material requirements for, 563, 564t solar cells vs., 559 TFTs (thin-film transistors), photodiodes and, 526–27, 527f See also flat panel detectors (FPDs); photoconductors X-ray image lag / gain effect, 552, 553f X-ray imaging amorphous selenium (a-Se) and, 564–68, 566f, 567f, 568f a-Si (amorphous silicon) and, 9–10, 10f, 555 CdTe (cadmium telluride) solar cells and, 569 history of, 501–2, 560, 561f, 562f, 563 TF (thin-film) solar cells and, 499–500 See also digital imaging; portal imaging X-ray photons, 544–52, 545f Zomeworks Corporation, 208