Centrifugal Separations in Biotechnology [2 ed.] 008102634X, 9780081026342

Table of contents :
Cover
Centrifugal Separations in Biotechnology
Copyright
Contents
In God, I trust
Preface to Second Edition (2019)
Preface to First Edition (2007)
1 Introduction
1.1 Introduction
1.1.1 Common Host Cells for Secreting Recombinant Protein
1.1.2 Platform for Protein Expression
1.1.3 Extracellular Protein
1.1.4 Intracellular Protein—Liquid
1.1.5 Intracellular Protein—Inclusion Body
1.2 Centrifugal Separation and Filtration
1.2.1 Sedimenting Centrifuge
1.2.2 Filtering Centrifuges
1.3 Pros and Cons of Filtration Versus Centrifugation
1.4 Generic Flow Sheet for Biopharmaceutical Process
1.5 Other Centrifugal Separations
1.6 Inputs and Outputs of Centrifuge
1.7 Separation Metrics
1.7.1 Protein Yield
1.7.2 Centrate Suspended Solids
1.7.3 Throughput Rate
1.7.4 Cell Viability
1.8 Text Organization
1.9 Summary
References
Problems
2 Principles of Centrifugal Sedimentation
2.1 Introduction
2.2 Nonintuitive Phenomena
2.2.1 Pressure Gradient in a Fluid Under Centrifugal Acceleration
2.2.2 Combined Centrifugal and Gravitational Accelerations
2.2.3 Coriolis Effect
2.3 Intuitive Phenomena
2.3.1 Centrifugal Acceleration
2.3.2 Fluid in a Centrifuge Bowl Not at Solid-Body Rotation
2.3.3 Regimes of Sedimentation
2.3.4 Stokes’ Law
2.3.5 Settling With Concentrated Solids
2.4 Process Functions
2.5 Summary
References
Problems
3 Batch and Semibatch Centrifuges
3.1 Spintube
3.2 Centrifugal Filter
3.3 Ultracentrifuges
3.3.1 Analytical Ultracentrifuge
3.3.2 Preparative Ultracentrifuge
3.3.3 Centrifugal Elutriation
3.4 Tubular Centrifuge
3.4.1 General Tubular Bowl Geometry
3.4.2 Ribs and Solids Scraper
3.4.3 Automatic Plunger Cake Discharge
3.5 Summary
References
Problems
4 Disk Centrifuge
4.1 Lamella/Inclined Plate Settler
4.1.1 Inclined Plate Settler Principle
4.1.2 Complications in Inclined Plate Settler
4.2 Disk-Stack centrifuge
4.2.1 General Disk Geometry
4.2.2 Disk Angle
4.2.3 Disk Spacing
4.2.4 Process Functions of Disk Centrifuge
4.2.5 Feed Solids
4.2.6 Manual Disk Centrifuge
4.2.6.1 Clarification of the Light Phase
4.2.6.2 Separation
4.2.7 Intermittent Discharge
4.2.7.1 Two Intermittent Discharge Designs
4.2.7.2 Angle of Cone for Discharge
4.2.7.3 Discharge Frequency
4.2.7.4 Concentrate Solids Discharge
4.2.7.5 Intermittent Discharge Interruption
4.2.8 Chamber Bowl
4.2.9 Continuous Concentrate Discharge
4.2.9.1 External Nozzle discharge
4.2.9.2 Small-Diameter Concentrate Discharge
4.2.9.3 Internal Vortex Nozzle Discharge
4.2.9.4 Applications of Different Concentrate Discharge Designs
4.2.9.5 External Nozzle Designs
4.2.10 Liquid Discharge
4.2.10.1 Centripetal Pump for Liquid Discharge
4.2.10.2 Hermetic Seal Design at Liquid discharge
4.2.11 Solution to Adverse Heating Effect
4.3 Feed Inlet and Accelerator
4.3.1 Introduction to Low Shear
4.3.2 Hydro-Hermetic Feed Design
4.3.3 Power Loss
4.3.4 Feed Acceleration Visual and Quantitative Testing
4.3.5 Improved Feed Accelerator
4.3.5.1 Improved Accelerator Without Smoothing Disk Section
4.3.5.2 Improved Accelerator With Smoothing Disk Section
4.4 Other Considerations
4.4.1 Materials of Construction
4.4.2 Clean-in-Place
4.4.3 Sterilization-in-Place
4.4.4 Containment
4.4.5 Surface Finish
4.4.6 Temperature Control
4.4.7 Water Requirements
4.4.8 Noise Level
4.4.9 Explosion Proof Design
4.5 Examples of Commercial Disk-Stack Centrifuge
4.6 Summary
References
Problems
5 Decanter Centrifuge
5.1 Solid Bowl or Decanter Centrifuge
5.2 Feed Rate
5.3 Pool Depth
5.4 Rotation Speed and G-Force
5.5 Differential Speed
5.6 Sedimentation Enhancement Using Chemical
5.7 Three-Phase Separation
5.8 Cake Conveyance
5.8.1 Dry Beach
5.8.2 Hydraulic Assist
5.9 Summary
References
Problems
6 Commercial Applications of Centrifugation in Biotechnology
6.1 Generic Flow Sheet of Biopharmaceutical
6.2 Mammalian Cell
6.3 Yeast Processing
6.4 Hormones Processing
6.5 Insulin Production
6.6 Biotech Separation of Inclusion Bodies
6.7 Vaccines Processing
6.7.1 Concentrated Cell-Based Product
6.7.2 Serum Product
6.8 Enzymes Processing
6.8.1 Extracellular Enzymes
6.8.2 Intracellular Enzymes
6.9 Probiotic Processing
6.10 Aquaculture
6.11 Alternative Meat
6.12 Baker Yeast Processing
6.13 Omega-3 From Microalgae
6.14 Ethanol Production
6.15 Other Biotech Processing
6.15.1 Recovery of Coagulation Factors From Blood Plasma
6.15.2 Tissue From Animal Cells
6.15.3 Laboratory Concentration and Buffer Exchange Using Centrifugal Filter
6.16 Summary
References
Problems
7 Concentrating Solids by Centrifugation
7.1 Introduction
7.2 Concentrating Underflow
7.3 Compaction
7.4 Expression or Percolation
7.5 Compaction Testing
7.6 Compaction Pressure
7.6.1 Test-Tube Compaction
7.6.2 Decanter Compaction
7.6.3 Considerations of Cake Compaction
7.7 Recommendations for Increasing Solid Concentration in Underflow
7.8 Summary
References
Problems
8 Laboratory and Pilot Testing
8.1 Process Objectives
8.2 Solid, Liquid, and Suspension Properties
8.2.1 Solids Properties
8.2.2 Mother Liquid Properties
8.2.3 Feed Slurry Properties
8.3 Bench-Scale Testing
8.3.1 Separability
8.3.2 Flocculant and Coagulant in Bench Tests
8.3.3 Test Variables
8.3.4 Material Balance
8.3.4.1 Material Balance Consideration for Bench Scale
8.3.5 Acceleration and Deceleration Time Duration
8.3.6 Settling Velocity
8.3.6.1 Apparatus for Visualizing Sedimentation Behavior
8.3.6.2 Sedimentation Behavior for Mondispersed Suspension
8.3.6.3 Sedimentation Behavior for Polydispersed Suspension
8.3.6.4 Sedimentation Behavior for Monodispersed Suspension With Hindered Settling
8.4 Centrifugal Filter Testing
8.4.1 Steady-State Membrane Centrifugal Filtration to Determine Protein Diffusivity and Solubility
8.4.2 Transient Membrane Centrifugal Filtration to Determine Protein Osmotic Pressure and Membrane Resistance
8.5 Pilot Testing
8.5.1 Material Balance Consideration for Pilot/Production Scale
8.5.1.1 Material Balance by Volume Fraction
8.5.1.2 Material Balance by Mass Fraction
8.5.2 Product (Protein) Yield
8.5.3 Pilot Test Factors
8.5.3.1 Monitored Variables in Pilot Tests
8.5.3.2 Metrics of Pilot Tests
8.5.3.3 Flocculant and Coagulant in Pilot Tests
8.6 Summary
References
Problems
9 Selection and Sizing of Centrifuges
9.1 Selection
9.1.1 Introduction
9.1.2 Tubular Centrifuge Selection
9.1.3 Disk Centrifuge Selection
9.1.4 Centrifuge Comparison
9.2 Centrifuge Sizing
9.2.1 Sizes and Rates
9.2.2 Dimensionless Le Number
9.2.3 Spintube (Bottle) Centrifuge
9.2.4 Sizing for Disk Centrifuge
9.2.4.1 Efficiency η in Le Number
9.2.5 Sizing for Tubular, Chamber, and Decanter Centrifuge
9.3 Feed Particle Size Distribution
9.4 Performance of Tubular Centrifuge
9.5 Summary
References
Further Reading
Problems
10 Troubleshoot and Optimization
10.1 Troubleshooting
10.1.1 Timescale of Occurrence
10.1.2 Mechanical or Process Problem
10.1.3 Process Problems
10.1.3.1 High Centrate Turbidity
10.1.3.1.1 High Feed Solids Throughput Causing High Centrate Turbidity
10.1.3.1.2 Finer Feed Solids Causing High Centrate Turbidity
10.1.3.1.3 Concentrate Not Discharging Causing High Centrate Turbidity
10.1.3.1.4 Lower Process Temperature and Higher Viscosity Causing High Centrate Turbidity
10.1.3.2 Wet/High-Moisture Concentrate
10.1.4 Mechanical Problem
10.1.4.1 High Vibration
10.1.4.2 Other Mechanical Problems
10.2 Optimization
10.2.1 Separation Metrics
10.2.2 Monitored Variables
10.2.3 Controlled Variables
10.2.4 Simple Optimization Scheme
10.2.4.1 Optimizing Disk, Tubular, Decanter, and Chamber Bowl Centrifuge
10.2.4.2 Optimizing Centrifuge (Increase Viscosity μ)
10.3 Summary
Problems
11 Visualization and Modeling of Flow and Separation in Tubular Centrifuge
11.1 Flow Visualization
11.2 Improved Moving Layer Flow Model
11.3 Effect of Velocity Profile
11.4 Effect of Friction Within the Flow Layer
11.5 Dimensionless Le Parameter
11.6 Quantitative Prediction
11.6.1 Total Solids Recovery in Cake
11.6.2 Total Solids Recovery in the Centrate
11.6.3 Particle Size Distribution of Supernatant/Overflow
11.6.4 Cumulative Size Recovery
11.7 Sedimentation Tests
11.7.1 Experiments on Sedimentation in Rotating Bowl Centrifuge
11.8 Summary
References
Problems
12 Disk-Stack Modeling
12.1 Disk Model
12.1.1 Continuum Phase
12.1.2 Dispersed Phase
12.2 Model Validation
12.3 Complications
12.4 Summary
References
Problems
13 Performance Projection of Centrifuges in Bioseparation
13.1 Disk Centrifuge
13.1.1 Baseline Case (400-mm Disk)
13.1.2 Effect of Fine Size Distribution (400-mm Disk)
13.1.3 Effect of G-Force (580-mm disk)
13.1.4 Efficiency η in Le Number (580-mm Disk)
13.1.5 Disk Centrifuge for Yeast Processing (500-mm Disk)
13.1.6 Disk Centrifuge for Inclusion Body Separation (260-mm Disk)
13.1.7 Enzymes (580-mm Disk)
13.2 Tubular Centrifuge
13.2.1 High-G Tubular (150- and 300-mm Tubular)
13.2.2 Low-G Tubular (150- and 300-mm Tubular)
13.3 Decanter
13.4 Spintube
13.5 Further Discussion on Numerical Simulations
13.6 Summary
References
Problems
14 Rotating Membrane in Bioseparation
14.1 Membrane
14.1.1 Osmotic Pressure Resistance
14.1.2 Gel Resistance
14.1.3 Membrane Fouling and Cake Formation
14.1.4 Two Scenarios on Rotation
14.2 Rotating Disk Membrane With Surface Parallel to the G-Force
14.2.1 Dimensionless Numbers
14.2.2 Governing Equations and Solutions
14.2.2.1 Model
14.2.2.2 Approximate Analytical Solution
14.2.2.2.1 Mass Boundary Layer
14.2.2.2.2 Membrane Flux
14.2.3 Gel Concentration
14.2.4 Determining Diffusivity
14.2.5 Parametric Effects
14.2.5.1 Effect of Reynolds Number
14.2.5.2 Schmidt Number Effect
14.2.5.3 Effect of Bulk/Feed Concentration
14.3 Rotating Membrane With Membrane Perpendicular to the G-Force
14.3.1 Spintube Equipped With Membrane Module—Centrifugal Filter
14.3.2 Model on Swinging Bucket Equipped With Ultrafiltration Membrane
14.3.3 Comparing Test Results With Predictions
14.4 Summary
References
Problems
15 Flocculation With Decanter Centrifuges
15.1 Introduction
15.1.1 Coagulation and Flocculation
15.1.2 Decanter Centrifuge
15.1.3 Problems
15.2 Monotonic Size Distribution Model
15.2.1 Moving Layer
15.2.2 Floc Model
15.2.3 Exponential Floc Size Distribution
15.2.4 Leung Number Calculation
15.2.5 Model Solution
15.3 Field Test
15.3.1 Two Decanter Tests at Wastewater Treatment
15.3.2 Determining In Situ Floc Size
15.3.2.1 Matching Data to Model Prediction
15.4 Prediction
15.5 Scale-Up
15.5.1 Le scale-Up
15.5.2 Sigma Scale-Up
15.5.3 G/g-Volume Scale-Up
15.5.4 Surface Area Scale-Up
15.6 Summary
References
Further Reading
Problems
16 Case Studies of Monotonic and Unimodal Size Distribution Models
16.1 Introduction
16.2 Monotonic Model Equivalent for Disk-Stack and Tubular Centrifuges
16.3 Disk-Stack Centrifuge for Processing Protein From Mammalian Cell Culture
16.3.1 Low Cell Density Cell Culture
16.3.2 High Cell Density Cell Culture
16.3.3 Centrate Solids and Turbidity
16.4 Tubular Centrifuge for Separating E. coli Lysate
16.5 Tubular Centrifuge for Separating S. pneumoniae Flocculate
16.6 Unimodal Size Distribution Model
16.6.1 Unflocculated Suspension
16.6.2 Flocculation in Disk-Stack Centrifuge
16.7 Comparing the Solids Recovery Between Monotonic and Unimodal Size Distributions
16.8 Summary
References
Problems
17 Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification
17.1 Introduction
17.2 Mammalian Cells With Cell Debris
17.3 Processing Hybridoma Cell Broth
17.4 Bimodal Size Model
17.5 Application of Bimodal Model on Hybridoma Cell Separation
17.6 Variation in Fine Fractions From Debris
17.7 Size of Whole Cells
17.7.1 Whole Cells Average Size
17.7.2 Size Range on Whole Cells
17.8 Effect of Smaller Size (The Debris)
17.9 Size Recoveries
17.9.1 Size Recovery of Small (S) Particles in Centrate
17.9.2 Size Recovery of Large (L) Particles in Centrate
17.9.3 Example on Classification
17.9.4 Smaller Size Fraction Further Apart From Larger Size Fraction in Bimodal Feed
17.9.5 Smaller Size Fraction Closer to the Larger Size Fraction in Bimodal Feed
17.10 Classification of Inclusion Bodies
17.10.1 Conventional Inclusion Bodies Processing
17.10.2 New Inclusion Bodies Processing
17.11 Centrifuges for Inclusion Bodies processing
17.11.1 Disk-Stack Centrifuge
17.11.2 Tubular Centrifuge
17.11.3 Spintube Centrifuge
17.12 Separation by Size and Density Difference
17.13 Summary
References
Problems
18 Integration of Unified Modeling With Practice in Centrifugal Separations
18.1 Introduction
18.2 Unified Modeling to Centrifugal Separation
18.3 Applications of the Unified Separation Models
18.3.1 Analysis of Test Data
18.3.2 Prediction/Forecast
18.3.3 Guiding Testing
18.3.4 Optimization
18.3.5 Troubleshooting
18.3.6 Scale-Up/Scale-Down
18.4 Integration of Unified Separation Models With Practice
18.5 Summary
Appendix A: Nomenclature
Subscripts
Symbols
Appendix B: Buckingham-π Analysis for Decanter and Tubular, Disk-Stack, and Spintube Centrifuges
B1 Decanter and Tubular Centrifuge (Separation/Clarification)
B2 Disk-Stack Centrifuge (Separation/Clarification)
B3 Spintube Centrifuge (Separation/Clarification)
Appendix C: Centrate or Concentrate Discharge Through Rotating Impeller
Appendix D: Answers to Problems in Chapters 2–17
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Chapter 6
Chapter 7
Chapter 8
Chapter 9
Chapter 10
Chapter 13
Chapter 14
Chapter 15 Solutions
Chapter 16 Solutions
Chapter 17 Solutions
Index
Back Cover

Citation preview

Centrifugal Separations in Biotechnology

Centrifugal Separations in Biotechnology

Second Edition Wallace Woon-Fong Leung

Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom 50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States Copyright © 2020 Elsevier Ltd. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress ISBN: 978-0-08-102634-2 For Information on all Elsevier publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Susan Dennis Acquisitions Editor: Kostas Marinakis Editorial Project Manager: Sara Pianavilla Production Project Manager: Joy Christel Neumarin Honest Thangiah Cover Designer: Greg Harris Typeset by MPS Limited, Chennai, India

Contents

In God, I trust Preface to Second Edition (2019) Preface to First Edition (2007) 1

xvii xix xxiii

Introduction 1.1 Introduction 1.1.1 Common Host Cells for Secreting Recombinant Protein 1.1.2 Platform for Protein Expression 1.1.3 Extracellular Protein 1.1.4 Intracellular Protein—Liquid 1.1.5 Intracellular Protein—Inclusion Body 1.2 Centrifugal Separation and Filtration 1.2.1 Sedimenting Centrifuge 1.2.2 Filtering Centrifuges 1.3 Pros and Cons of Filtration Versus Centrifugation 1.4 Generic Flow Sheet for Biopharmaceutical Process 1.5 Other Centrifugal Separations 1.6 Inputs and Outputs of Centrifuge 1.7 Separation Metrics 1.7.1 Protein Yield 1.7.2 Centrate Suspended Solids 1.7.3 Throughput Rate 1.7.4 Cell Viability 1.8 Text Organization 1.9 Summary References Problems

1 1 6 8 9 9 10 12 14 15 16 18 19 19 20 20 20 21 21 22 23 23 25

v

vi 2

CONTENTS Principles of Centrifugal Sedimentation 2.1 Introduction 2.2 Nonintuitive Phenomena 2.2.1 Pressure Gradient in a Fluid Under Centrifugal Acceleration 2.2.2 Combined Centrifugal and Gravitational Accelerations 2.2.3 Coriolis Effect 2.3 Intuitive Phenomena 2.3.1 Centrifugal Acceleration 2.3.2 Fluid in a Centrifuge Bowl Not at Solid-Body Rotation 2.3.3 Regimes of Sedimentation 2.3.4 Stokes’ Law 2.3.5 Settling With Concentrated Solids 2.4 Process Functions 2.5 Summary References Problems

27 27 27

3

Batch and Semibatch Centrifuges 3.1 Spintube 3.2 Centrifugal Filter 3.3 Ultracentrifuges 3.3.1 Analytical Ultracentrifuge 3.3.2 Preparative Ultracentrifuge 3.3.3 Centrifugal Elutriation 3.4 Tubular Centrifuge 3.4.1 General Tubular Bowl Geometry 3.4.2 Ribs and Solids Scraper 3.4.3 Automatic Plunger Cake Discharge 3.5 Summary References Problems

49 49 53 54 55 56 58 60 60 63 67 69 70 70

4

Disk Centrifuge 4.1 Lamella/Inclined Plate Settler 4.1.1 Inclined Plate Settler Principle 4.1.2 Complications in Inclined Plate Settler

73 73 73 74

27 28 32 35 35 38 40 42 43 44 46 46 47

CONTENTS

vii

4.2

75 75 77 77 79 80 80 82 87 87 95 100 100 101 101 102

5

Disk-Stack Centrifuge 4.2.1 General Disk Geometry 4.2.2 Disk Angle 4.2.3 Disk Spacing 4.2.4 Process Functions of Disk Centrifuge 4.2.5 Feed Solids 4.2.6 Manual Disk Centrifuge 4.2.7 Intermittent Discharge 4.2.8 Chamber Bowl 4.2.9 Continuous Concentrate Discharge 4.2.10 Liquid Discharge 4.2.11 Solution to Adverse Heating Effect 4.3 Feed Inlet and Accelerator 4.3.1 Introduction to Low Shear 4.3.2 Hydro-Hermetic Feed Design 4.3.3 Power Loss 4.3.4 Feed Acceleration Visual and Quantitative Testing 4.3.5 Improved Feed Accelerator 4.4 Other Considerations 4.4.1 Materials of Construction 4.4.2 Clean-in-Place 4.4.3 Sterilization-in-Place 4.4.4 Containment 4.4.5 Surface Finish 4.4.6 Temperature Control 4.4.7 Water Requirements 4.4.8 Noise Level 4.4.9 Explosion Proof Design 4.5 Examples of Commercial Disk-Stack Centrifuge 4.6 Summary References Problems

104 108 111 111 112 113 114 114 115 115 115 115 115 119 119 120

Decanter Centrifuge 5.1 Solid Bowl or Decanter Centrifuge 5.2 Feed Rate 5.3 Pool Depth 5.4 Rotation Speed and G-Force

121 121 122 122 123

viii

CONTENTS 5.5 5.6 5.7 5.8

Differential Speed Sedimentation Enhancement Using Chemical Three-Phase Separation Cake Conveyance 5.8.1 Dry Beach 5.8.2 Hydraulic Assist 5.9 Summary References Problems 6

Commercial Applications of Centrifugation in Biotechnology 6.1 Generic Flow Sheet of Biopharmaceutical 6.2 Mammalian Cell 6.3 Yeast Processing 6.4 Hormones Processing 6.5 Insulin Production 6.6 Biotech Separation of Inclusion Bodies 6.7 Vaccines Processing 6.7.1 Concentrated Cell-Based Product 6.7.2 Serum Product 6.8 Enzymes Processing 6.8.1 Extracellular Enzymes 6.8.2 Intracellular Enzymes 6.9 Probiotic Processing 6.10 Aquaculture 6.11 Alternative Meat 6.12 Baker Yeast Processing 6.13 Omega-3 From Microalgae 6.14 Ethanol Production 6.15 Other Biotech Processing 6.15.1 Recovery of Coagulation Factors From Blood Plasma 6.15.2 Tissue From Animal Cells 6.15.3 Laboratory Concentration and Buffer Exchange Using Centrifugal Filter 6.16 Summary References Problems

124 127 127 129 129 130 132 132 133

135 136 138 141 144 145 145 147 147 148 148 148 149 150 151 152 153 154 155 157 157 157 158 159 159 160

CONTENTS 7

8

Concentrating Solids by Centrifugation 7.1 Introduction 7.2 Concentrating Underflow 7.3 Compaction 7.4 Expression or Percolation 7.5 Compaction Testing 7.6 Compaction Pressure 7.6.1 Test-Tube Compaction 7.6.2 Decanter Compaction 7.6.3 Considerations of Cake Compaction 7.7 Recommendations for Increasing Solid Concentration in Underflow 7.8 Summary References Problems Laboratory and Pilot Testing 8.1 Process Objectives 8.2 Solid, Liquid, and Suspension Properties 8.2.1 Solids Properties 8.2.2 Mother Liquid Properties 8.2.3 Feed Slurry Properties 8.3 Bench-Scale Testing 8.3.1 Separability 8.3.2 Flocculant and Coagulant in Bench Tests 8.3.3 Test Variables 8.3.4 Material Balance 8.3.5 Acceleration and Deceleration Time Duration 8.3.6 Settling Velocity 8.4 Centrifugal Filter Testing 8.4.1 Steady-State Membrane Centrifugal Filtration to Determine Protein Diffusivity and Solubility 8.4.2 Transient Membrane Centrifugal Filtration to Determine Protein Osmotic Pressure and Membrane Resistance 8.5 Pilot Testing 8.5.1 Material Balance Consideration for Pilot/ Production Scale 8.5.2 Product (Protein) Yield 8.5.3 Pilot Test Factors 8.6 Summary References Problems

ix 161 161 161 163 164 166 167 168 168 170 170 171 171 172 173 173 174 174 175 175 175 175 176 177 177 180 180 185 185

186 187 188 190 192 199 199 200

x 9

CONTENTS Selection and Sizing of Centrifuges 9.1 Selection 9.1.1 Introduction 9.1.2 Tubular Centrifuge Selection 9.1.3 Disk Centrifuge Selection 9.1.4 Centrifuge Comparison 9.2 Centrifuge Sizing 9.2.1 Sizes and Rates 9.2.2 Dimensionless Le Number 9.2.3 Spintube (Bottle) Centrifuge 9.2.4 Sizing for Disk Centrifuge 9.2.5 Sizing for Tubular, Chamber, and Decanter Centrifuge 9.3 Feed Particle Size Distribution 9.4 Performance of Tubular Centrifuge 9.5 Summary References Further Reading Problems

203 203 203 204 204 205 207 207 208 210 213

10

Troubleshoot and Optimization 10.1 Troubleshooting 10.1.1 Timescale of Occurrence 10.1.2 Mechanical or Process Problem 10.1.3 Process Problems 10.1.4 Mechanical Problem 10.2 Optimization 10.2.1 Separation Metrics 10.2.2 Monitored Variables 10.2.3 Controlled Variables 10.2.4 Simple Optimization Scheme 10.3 Summary Problems

229 229 229 230 230 233 235 235 236 237 237 240 240

11

Visualization and Modeling of Flow and Separation in Tubular Centrifuge 11.1 Flow Visualization 11.2 Improved Moving Layer Flow Model 11.3 Effect of Velocity Profile 11.4 Effect of Friction Within the Flow Layer

243 243 248 251 252

219 222 223 224 224 225 225

CONTENTS

xi

Dimensionless Le Parameter Quantitative Prediction 11.6.1 Total Solids Recovery in Cake 11.6.2 Total Solids Recovery in the Centrate 11.6.3 Particle Size Distribution of Supernatant/ Overflow 11.6.4 Cumulative Size Recovery 11.7 Sedimentation Tests 11.7.1 Experiments on Sedimentation in Rotating Bowl Centrifuge 11.8 Summary References Problems

252 253 253 253

12

Disk-Stack Modeling 12.1 Disk Model 12.1.1 Continuum Phase 12.1.2 Dispersed Phase 12.2 Model Validation 12.3 Complications 12.4 Summary References Problems

261 261 264 264 268 270 271 271 272

13

Performance Projection of Centrifuges in Bioseparation 13.1 Disk Centrifuge 13.1.1 Baseline Case (400-mm Disk) 13.1.2 Effect of Fine Size Distribution (400-mm Disk) 13.1.3 Effect of G-Force (580-mm disk) 13.1.4 Efficiency η in Le Number (580-mm Disk) 13.1.5 Disk Centrifuge for Yeast Processing (500-mm Disk) 13.1.6 Disk Centrifuge for Inclusion Body Separation (260-mm Disk) 13.1.7 Enzymes (580-mm Disk) 13.2 Tubular Centrifuge 13.2.1 High-G Tubular (150- and 300-mm Tubular) 13.2.2 Low-G Tubular (150- and 300-mm Tubular) 13.3 Decanter

273 273 275

11.5 11.6

254 254 255 255 258 259 259

276 277 280 282 284 286 288 289 290 292

xii

14

15

CONTENTS 13.4 Spintube 13.5 Further Discussion on Numerical Simulations 13.6 Summary References Problems

294 296 297 297 297

Rotating Membrane in Bioseparation 14.1 Membrane 14.1.1 Osmotic Pressure Resistance 14.1.2 Gel Resistance 14.1.3 Membrane Fouling and Cake Formation 14.1.4 Two Scenarios on Rotation 14.2 Rotating Disk Membrane With Surface Parallel to the G-Force 14.2.1 Dimensionless Numbers 14.2.2 Governing Equations and Solutions 14.2.3 Gel Concentration 14.2.4 Determining Diffusivity 14.2.5 Parametric Effects 14.3 Rotating Membrane With Membrane Perpendicular to the G-Force 14.3.1 Spintube Equipped With Membrane Module—Centrifugal Filter 14.3.2 Model on Swinging Bucket Equipped With Ultrafiltration Membrane 14.3.3 Comparing Test Results With Predictions 14.4 Summary References Problems

299 299 300 301 302 302

Flocculation With Decanter Centrifuges 15.1 Introduction 15.1.1 Coagulation and Flocculation 15.1.2 Decanter Centrifuge 15.1.3 Problems 15.2 Monotonic Size Distribution Model 15.2.1 Moving Layer 15.2.2 Floc Model 15.2.3 Exponential Floc Size Distribution 15.2.4 Leung Number Calculation 15.2.5 Model Solution

331 331 331 332 332 333 334 336 336 338 339

303 304 306 310 311 313 315 316 320 323 328 329 329

CONTENTS 15.3

Field Test 15.3.1 Two Decanter Tests at Wastewater Treatment 15.3.2 Determining In Situ Floc Size 15.4 Prediction 15.5 Scale-Up 15.5.1 Le scale-Up 15.5.2 Sigma Scale-Up 15.5.3 G/g-Volume Scale-Up 15.5.4 Surface Area Scale-Up 15.6 Summary References Further Reading Problems 16

17

Case Studies of Monotonic and Unimodal Size Distribution Models 16.1 Introduction 16.2 Monotonic Model Equivalent for Disk-Stack and Tubular Centrifuges 16.3 Disk-Stack Centrifuge for Processing Protein From Mammalian Cell Culture 16.3.1 Low Cell Density Cell Culture 16.3.2 High Cell Density Cell Culture 16.3.3 Centrate Solids and Turbidity 16.4 Tubular Centrifuge for Separating E. coli Lysate 16.5 Tubular Centrifuge for Separating S. pneumoniae Flocculate 16.6 Unimodal Size Distribution Model 16.6.1 Unflocculated Suspension 16.6.2 Flocculation in Disk-Stack Centrifuge 16.7 Comparing the Solids Recovery Between Monotonic and Unimodal Size Distributions 16.8 Summary References Problems

xiii 340 340 341 346 346 346 348 348 348 350 350 351 351

353 353 354 356 358 361 363 364 366 367 369 370 375 377 378 378

Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification 381 17.1 Introduction 382 17.2 Mammalian Cells With Cell Debris 383

xiv

CONTENTS 17.3 17.4 17.5

Processing Hybridoma Cell Broth Bimodal Size Model Application of Bimodal Model on Hybridoma Cell Separation 17.6 Variation in Fine Fractions From Debris 17.7 Size of Whole Cells 17.7.1 Whole Cells Average Size 17.7.2 Size Range on Whole Cells 17.8 Effect of Smaller Size (The Debris) 17.9 Size Recoveries 17.9.1 Size Recovery of Small (S) Particles in Centrate 17.9.2 Size Recovery of Large (L) Particles in Centrate 17.9.3 Example on Classification 17.9.4 Smaller Size Fraction Further Apart From Larger Size Fraction in Bimodal Feed 17.9.5 Smaller Size Fraction Closer to the Larger Size Fraction in Bimodal Feed 17.10 Classification of Inclusion Bodies 17.10.1 Conventional Inclusion Bodies Processing 17.10.2 New Inclusion Bodies Processing 17.11 Centrifuges for Inclusion Bodies processing 17.11.1 Disk-Stack Centrifuge 17.11.2 Tubular Centrifuge 17.11.3 Spintube Centrifuge 17.12 Separation by Size and Density Difference 17.13 Summary References Problems 18

Integration of Unified Modeling With Practice in Centrifugal Separations 18.1 Introduction 18.2 Unified Modeling to Centrifugal Separation 18.3 Applications of the Unified Separation Models 18.3.1 Analysis of Test Data 18.3.2 Prediction/Forecast 18.3.3 Guiding Testing 18.3.4 Optimization

384 387 388 390 391 392 392 394 398 399 400 401 405 407 408 409 411 413 413 415 417 418 422 422 423

425 425 425 428 428 429 429 429

CONTENTS

xv 18.3.5 Troubleshooting 18.3.6 Scale-Up/Scale-Down Integration of Unified Separation Models With Practice Summary

429 430

Appendix A: Nomenclature Appendix B: Buckingham-π Analysis for Decanter and Tubular, Disk-Stack, and Spintube Centrifuges Appendix C: Centrate or Concentrate Discharge Through Rotating Impeller Appendix D: Answers to Problems in Chapters 2 17

435

445 449

Index

457

18.4 18.5

431 433

439

In God, I trust

xvii

Preface to Second Edition (2019) Since the Centrifugal Separation in Biotechnology (CSB) has been published in 2007, it has well served the community of biotechnology separation. I have been contacted by practitioners, researchers, engineers, students, users, and equipment manufacturers in the biotechnology and biopharmaceutical community, who told me that they have recommended the text whole-heartedly to others as they find the contents in it are informative and helpful. They said CSB is the most comprehensive treatise in dealing with centrifugal separation in biotechnology. There is a good balance of practice and fundamentals on the subject matter in the text. The text is well illustrated with tables, figures, sketches, and photographs, and is written in layman’s language that is easy to understand. The readers also found the questions at the end of each chapter both challenging and stimulating, which help them to think and digest the contents that have been discussed. Over a decade since the first publication of CSB, there has been an exponential growth in biotech activities. This is especially for the world of biopharma. As an example, there are much more therapeutic monoclonal antibodies that have been under various stages of testing than before and they have been approved, in record numbers, by the European Medicines Agencies and the United States Federal Drug Administration. There is a popular demand from the biotechnology and biopharmaceutical community for a revised edition of the text. After a year and a half of preparation we finally come up with the second edition of the book. We have kept the same approach as our first edition, again with good “balance” between practice and fundamentals, together with the author’s 44 years of experience in separation and filtration from both industry and academia injecting into this delicate balance. We have updated the text with new technologies and new designs for both the disk-stack and tubular centrifuges, which are the workhorse of the industry. A decade ago, flocculation was unheard of for disk-stack centrifuge. Given the bottom feed design which is available today, flocculation can be implemented for disk-stack centrifuge, when process permits. In some applications processing flowable concentrate containing biological cells or bacteria, after separation the concentrate is discharged through external nozzles. The impact of the concentrate discharge stream at high speed from the nozzle disk-stack centrifuge results in destroying the cells or bacteria and losing their probiotic function. Today this probiotic

xix

xx

PREFACE

TO

SECOND EDITION (2019)

flowable concentrate stream can be routed to a small diameter near the axis of the centrifuge, where the concentrate stream can be gently discharged under pressure without being exposed to air causing oxidation and being destroyed from high-speed impact. These are only a few of the new technologies that are presented in the second edition. We have expanded the discussion on clean-in-place (CIP) and sterilization-in-place (SIP), which are very important practice for biotechnology and biopharmaceutical to avoid cross-contamination of their products. Despite there is no satisfactory standard to execute CIP and SIP, the text has discussed good practice that can be used as a general guideline for practitioners. Chapter 6, Commercial Applications of Centrifugation in Biotechnology, discusses several typical flow sheets for biotech and biopharma processing, and this has received very favorable response from the readers. We have included a few more typical flow sheets in biotech and biopharma processings, and they are by no means exhaustive but serve as good examples. Some of the flow sheets are related to separation of recombinant protein, while others are for different applications. Not only have recombinant protein being used for diagnostic, therapeutics, and disease prevention, they have been expanded into human food (beef, pork, and chicken made from recombinant protein) and aquaculture. As medical scientists discover new antigens that are specific to different types of cancer cells, more monoclonal antibodies are also discovered for identifying the antigens associated with these cancer cells so that human immune system can launch selfdefense against them. Despite each type of protein being secreted by the common hosts (mammalian cells, microbial, yeast, virus infected insect cells, etc.) are different, there are certain commonalities in the separation and purification steps. CSB continues to use case studies throughout to illustrate these commonalities so that biotech practitioner not only can use them in generic form but tailor-make for their own process. In the first edition, the author has developed basic models of separation for tubular/decanter and disk-stack centrifuges (Chapter 11: Visualization and Modeling of Flow and Separation in Tubular Centrifuge, and Chapter 12: Disk-Stack Modeling). These models have been used to complement with case studies to discuss separation (Chapter 13: Performance Projection of Centrifuges in Bioseparation). One shortcoming is that the particle size distributions, which were used in Chapter 13, Performance Projection of Centrifuges in Bioseparation, were often lacking. To that end, some of the models being developed may be handicapped in short of that measurement. In the second edition, we have come up with three analytical forms of particle size distribution for the feed to the separating centrifuge, which depend on two to five

PREFACE TO SECOND EDITION (2019) parameters. These analytical forms include monotonic size distribution, unimodal size distribution, and bimodal size distribution. By modifying the parameters, we can easily change the particle size of the feed and investigate the sensitivity of the separation outcome from the centrifuge. We have used these size distributions together with a unified approach on separation using spintube, disk-stack, tubular, and decanter centrifuges to study a lot more cases. This includes a very important application on inferring the flocculated size in centrifuge (Chapter 15: Flocculation With Decanter Centrifuges), which has been a culprit on scale-up/scale-down problems. We have also used the monotonic size distribution to interpret tests being conducted on disk-stack and tubular centrifuges (Chapter 16: Case Studies of Monotonic and Unimodal Size Distribution Models) with both dispersed and flocculated feeds, respectively, and to demonstrate the benefits of flocculation with disk-stack centrifuge (Chapter 16: Case Studies of Monotonic and Unimodal Size Distribution Models). Next, we have adopted the bimodal model to classify biologics with both a smaller and a larger size fraction (Chapter 17: Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification) in the feed to the centrifuge. The size recovery of the smaller fraction, as well as the larger fraction, in centrate and concentrate, respectively, have been developed. One can fine-tune the classification in accordance to the S-shaped size recovery curves to maximize recovery of smaller/larger size fraction with minimum cross contamination. This can speed-up the development or optimization process improving the purity of the products. Finally, in Chapter 18, Integration of Unified Modeling With Practice in Centrifugal Separation, we propose an integrated approach, irrespective of the type of centrifuges, to separation by sedimentation in using known/measured, or estimated analytical representation of, particle size distribution to tackle separation problem. The author has suggested integrating the models into practices at all stages from laboratory testing, pilot testing, and clinical manufacturing to full-scale production. This reassures the reliability of the separation process despite frequently we have only limited test results as a basis to make important decisionon the process. Overall, the second edition not only provides an update to various centrifugal technologies for biotechnology, but fills in a number of missing gaps. I believe this new edition of CSB will provide a more powerful resource for readers to tackle their centrifugal separation problem. I sincerely hope that the text can help to push the biotech and biopharma forward with good practice and advanced technologies in centrifugal separation. Consequently, drugs substances and intermediates, diagnostic and therapeutic protein-based drugs, and health supplements, based on

xxii

PREFACE

TO

SECOND EDITION (2019)

recombinant protein and other biotech processes, becomes available in shorter development time and become more affordable to the general public. Finally, I give thanks to all those who have provided valuable inputs into this new edition. I also thank my wife, Stella, in giving me the extra-time and space to devote in preparation of the new edition of CSB that has taken much longer time than I originally estimated. The support from the rest of my family, my mom, Jessica, Daniel, Jeffrey, and Sandar has been tremendous. I am grateful to God for His keeping and strengthening of me, without which I cannot complete this revision. In Him, I trust. Wallace Woon-Fong Leung

July 2019

Preface to First Edition (2007) In processing biological materials to produce high value-added intermediates or finished products, no matter whether it is liquid or solid, often involves a separation step, especially after the fermenter or bioreactor. In one instance, the high-value solid products in a dilute concentration have to be separated from the waste liquid with spent cells and debris; therefore it is essential to prevent loss of valuable solids in the liquid stream. A further requirement is that contaminants have to be washed from the solids. Alternatively, the high-value liquid containing dissolved protein needs to be separated from the biomass and that the liquid product should be free of solid particles to avoid downstream separation and contamination of purification equipment, such as chromatography column. The difficulty in carrying out separation step is that biosolids do not filter well and often foul and blind the filter, such as microfiltration and ultrafiltration membranes. Instead of filtration, separation by sedimentation utilizing the density difference between the solid and the suspending liquid can be employed. However, the density difference between biosolids and liquid (typically water based) is very small, rendering the separation very slow and ineffective, especially under the Earth’s gravity. In addition, if RNA and other protein materials are dissolved in the liquid, the liquid phase can be very viscous, which further slows down sedimentation. Another difficulty is that the solids concentration in suspension for processing is relatively dilute and requires equipment that has large volumetric capacity for handling the flow and process. Centrifugation has proven to be a rather robust process for enhancing settling by using thousands to almost millions of times the Earth’s gravitational acceleration. In biopharmaceutical processing for producing a recombinant therapeutic protein for antibiotics and drug substances from yeast, microbial, and mammalian cells, such as the Chinese Hamster Ovary cells, centrifuges have been widely used to perform separation, classification of cell debris, concentration of suspension, and separation and washing of solids, such as inclusion body or crystalline protein. No doubt, given the escalating research activities in biotechnology, many new sources of therapeutic proteins and other valuable biological materials will be discovered and developed, and more stringent requirements are demanded from separation/recovery and purification. There will be more

xxiii

xxiv

PREFACE TO FIRST EDITION (2007)

growing needs of centrifugation in combination with other separations and filtrations to perform the often overlooked, yet important, duty. While there are many texts and reference books on bioseparation, there is very little coverage on centrifugation. This book is the first reference book of its kind devoted to centrifugal separation in biotechnology. It is an outgrowth of a series of seminars, short courses and presentations that the author has delivered to biopharmaceutical companies all over the world. This new and challenging topic has received excellent global reception, which is quite comforting and rewarding. The contents of this book are also based on the author’s research and extensive experiences, respectively, in practice, mentoring and lecturing on the subject for over 20 years. The book starts out with an introduction on the topic (Chapter 1) followed by Chapter 2 on sedimentation, which is the key step of separation. Subsequently, various batch (spintube centrifuge, ultracentrifuge) and semi-batch (tubular centrifuge) centrifuges are discussed in Chapter 3. The workhorse of the industrial separation process, disk-stack centrifuge, is presented and discussed in detail in Chapter 4. Also, decanter centrifuge, which is more applicable to high-solids feed under relatively lower centrifugal acceleration, has also been included in Chapter 5. However, the discussion will be brief in favor of giving room to various other topics. Commercial applications of centrifuges in biotechnology are discussed in Chapter 6. This is perhaps one of the most interesting topics for practitioners who are more concerned about where proven processes are and how their new process may build on what is already known and practiced. Despite there being lot of applications discussed in this chapter, unfortunately there might be applications that have been inadvertently omitted, given that the biotech applications are very diverse. Subsequently, we discuss in Chapter 7 the importance and practice of increasing, or at least maintaining, high-solids concentration in the underflow stream of the centrifuge. Laboratory and pilot testing and selection and sizing are essential functions for establishing and implementing the biotech process and they are discussed in Chapters 8 and 9, respectively. A new unified approach in scale-up and prediction with use of a dimensionless Leung (Le) number is introduced. The Le number works for all types of centrifuges, including spintube, tubular, chamber, disk-stack, and decanter centrifuges. This provides a solid foundation for practitioners to scale-up equipment and analyze test results. Troubleshooting and optimization are two important topics of general interest, especially for installed machines, and these are discussed in Chapter 10. Subsequently, modeling of tubular and disk-stack centrifuges are covered in Chapters 11 and 12, respectively, for researchers who are

PREFACE TO FIRST EDITION (2007) interested. Readers who are not interested in modeling can jump directly to Chapter 13. The Le number provides a basis for the scale-up and performance prediction covered in Chapter 13. Here, numerous examples are used to demonstrate the versatility of the numerical simulator built on the Le-approach to forecast performance in parallel with concurrent testing, which is often limited for various reasons. Numerical simulation can also be used to analyze laboratory, pilot, and production test results to validate machine and process performance. Therefore numerical simulation can be used for laboratory screening, pilot testing, clinical manufacturing testing, full-scale production testing, and even for smallscale testing in the laboratory to investigate alternatives and improvements of the existing process under production. Membrane processes, such as microfiltration, ultrafiltration, and diafiltration, are frequently used in bioseparation. Lastly, Chapter 14 is devoted to combining two separation processes: centrifugation and membrane separation. Two examples, respectively, on centrifugal filter in spintube and large rotating membrane systems are discussed. The general approach can be extended to other rotating membrane geometry. Centrifugation has been treated as a black box in the past, as the subject is quite complex and nonintuitive. The subject involves multiple disciplines, such as fluid dynamics, mechanics and vibration, design, material science, rheology, chemical and process engineering, chemistry, biology, and physics. I hope this text will fulfill the quest of knowledge rendering centrifuge a lot more “transparent” to biologists, biotechnologists, chemists, physicists, scientists, researchers, and practicing engineers. The more they know the better they can deploy, comfortably and without reservation, centrifuge as handy process equipment. Problems are listed at the end of each chapter in the text, and they complement and supplement the contents in the chapter. They are also meant to reinforce the concept for the readers through practices, challenging their thoughts and understanding on the topic. Apart from practitioners and researchers, this book is written primarily for 4th year university student in the 4-year undergraduate study and research graduates in their MS or PhD research program in university taking bioseparation, bioprocessing, advanced unit-operation/process engineering, or similar courses. I am grateful to Stella, Jessica, Jeffrey, my mother, and my late father for putting up with me while I was devoted to preparing this book. Both my late father and my dear friend and mentor, the late Professor Ascher H. Shapiro, demonstrated dedication and perseverance in their lives, which inspired me all along, especially during the trying times when I

xxvi

PREFACE TO FIRST EDITION (2007)

was working on the manuscript among other responsibilities that also demanded my undivided attention. I also thank Alice Tang for skillfully helping out with the manuscript work and meeting the publisher’s deadline. Wallace Woon-Fong Leung

2007

1 Introduction 1.1

Introduction

Biotechnology has revolutionized our life in the 20th century [1]. Its impact is only now being felt from engineering food [2 4], engineering and delivering drugs [5,6], to engineering consumable products. Without doubt it will continue to influence our daily lives for years to come. One of the many successful examples is that drugs, such as monoclonal antibodies (mAbs) and some basic drug substances (i.e., the building block for various drugs), can now be manufactured and formulated from bioreaction. One of the commonly used methods in biotechnology is the recombinant DNA technique [7 9]. A desired gene is isolated from one organism, and this is inserted into a small piece of carrier DNA called a vector. It is highly desirable that the recombined DNA (vector plus gene) can propagate in a similar or unrelated host/recipient cell. The mammalian cell, such as the Chinese Hamster Ovary cell, is a popular host cell. Fig. 1.1 shows a schematic of an animal cell which is very similar to that of a mammalian cell. A characteristic size of the mammalian cell is about 10 20 µm. Unlike a plant cell, there is no cell wall for animal and mammalian cells, so they rely on a plasma membrane to keep the intracellular contents intact. High shear stress acting on the cell can rupture the fragile membrane releasing the intracellular material. Yeast (see schematic in Fig. 1.2), in eukaryotic single-celled microorganisms classified as a member of fungus kingdom, has been commonly used as a host cell in the recombinant DNA process, the knowledge and experience of which we have gained from the brewery industry. Unlike a mammalian cell, the yeast cell has a strong cell wall. Yeast cells are smaller than mammalian cells and are typically between 7 and 10 µm. Some common yeast hosts include, Saccharomyces cerevisiae (referred commonly as baker yeast) and Pichia pastoris. Bacteria, such as Escherichia coli (hereafter abbreviated as E. coli) and Bacillus subtilis, have been used as host cells for the recombinant

Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00001-5 © 2020 Elsevier Ltd. All rights reserved.

1

Figure 1.1 Animal cell schematic showing plasma membrane.

Figure 1.2 Yeast cell schematic showing both cell wall and membrane.

Introduction

3

Figure 1.3 Escherichia coli cell schematic.

DNA technique. A schematic of E. coli bacteria cell is shown in Fig. 1.3. Again, E. coli has a sturdy cell wall with both an outer membrane and an inner membrane. E. coli is typically elongated with a dimension of 3 µm long by 1 µm width. Therapeutic protein can be “expressed” by these host cells or organisms with the recombinant DNA. The protein of interest may remain in the cell (intracellular) or be secreted to the exterior of the cell (extracellular). The aforementioned biosynthesis provides more engineering flexibility, specificity, versatility, reliability, and cost-effectiveness. Therapeutic proteins are quite diverse in the application treatments, such as human insulin for diabetes, erythropoietin for anemia and chronic renal failure, interferon-beta and gamma for cancer, DNase for pulmonary treatment, vaccines for hepatitis B, interleukin-2 for AIDS, prourokinase for heart attacks, and tissue plasminogen activator (enzyme) for strokes. Therapeutic proteins are present in many different kinds of mAbs. mAbs are antibodies that are identical, because they are produced by one type of immune cells, and they are all copies or clones of a single parent cell. mAbs are first produced by Kohler and Milstein in 1975 [10], for which they were awarded the Nobel Prize in Physiology or Medicine in 1984. By virtue of the mAb being identical copies produced by one type of immune cells, they have a high specificity for their targets. mAb has been used in diagnosis. There are over 100 different diagnostic products available in the world that are mAb [11]. mAb is also used for

4

Centrifugal Separations in Biotechnology

therapeutics. For example, mAb has been a popular antibody made in the laboratory used for cancer treatment where the antibody is designed to attach as a label to their counterpart protein (antigen) on a specific cancer cell so that immune cells can spot and attack the cancer cells. As an example, the mAb known commercially as alemtuzumab drug (note all mAb drugs have the last three alphabets labeled as “mab,” which distinguishes them being mAb-based drugs) can target at the antigen CD52 found on the cancer cells that causes chronic lymphocytic leukemia [12]. As antigens are discovered to be linked to more specific cancers, more mAbs have also been developed for cancer treatment. Some mAbs work better on certain cancers than others. mAb can also attach to the antigen on breast cancer cells blocking the growth of breast cancer cells. Cancer cells can “turn off the switch” of immune cells to avoid being attacked by the immune system in our bodies. Inhibitors, or commonly known as checkpoints, are mAb produced by the recombinant protein process, that inhibit the protein secreted from the cancer cells in “fooling” the immune cells, thereby allowing the immune cells to carry out their normal functions. As an example, PD-1 is a protein on the immune T-cells. The immune T-cells are normally in a switch-off condition because PD-1 has been attached by their counterpart PD-L1, another protein that both normal cells and cancer cells have. In other words, they have been switched off. Some cancer cells have an abundant PDL1 that is used to attach to the PD-1 of the immune T-cells thereby evading being attacked. On the other hand, mAb can target at either PD1 or PD-L1 and inhibit their binding, thereby allowing the immune cells to attack the cancer cells. For example, pembrolizumab is a PD-1 inhibitor that can treat skin melanoma, non-small-cell lung cancer, kidney cancer, bladder cancer, head and neck cancer, and Hodhkin lymphoma [12]. As another example, atezolizumab is a PD-L1 inhibitor that can treat bladder cancer, non-small-cell lung cancer, and Merkel cell carcinoma [12]. New inhibitors are being developed rapidly over time as more knowledge is being gained on the specifics of different cancers and their behavior. The examples mentioned in the forgoing are just a few under the broad umbrella of immunotherapy for which mAb plays an important role. The main objective of immunotherapy is to enable the immune system of patients to recognize or target specific cancer cells and destroy them [13]. In 2005 the total mAb therapeutics entering first-in-human studies per year is 35, 16 out of which are for cancer treatment. In 2017 the total mAb therapeutics rose to 105, and nearly 80 were for cancer treatment. Indeed, the antibody therapeutics entering clinical study and being approved are in record numbers [14].

Introduction

5

Extracellular proteins secreted from yeasts are produced for making insulin, human serum albumin, and hepatitis vaccines. Insulin drug has reached over USD 24 billion market in 2018 according to a market study in 2019. The fast-growing biopharmaceutical business in producing therapeutic proteins is getting so popular that all major drug manufacturers also carry a parallel line of this business. Unfortunately, the protein expressed from the bioprocess is in very small amounts in a large volume of suspension, that is, low concentration. The two key hurdles in recombinant DNA techniques to produce therapeutic protein [15] are (a) to recover this small concentration of protein after fermentation by separation and (2) to provide high purity of the protein product through separation and purification. It is prudent that both separation and purification processes should be robust and cost-effective for the biopharmaceutical technology to be viable and competitive. Although this text is focused on separation, one should bear in mind that given these two steps are sequential, poor separation can adversely affect purification downstream. Therefore it is prudent to have an integrated approach for downstream processing. To say the least, if there is an upset from the fermenter upstream producing, say, off-spec finer feed, the centrifuge should take on the upset feed and try to produce a consistent output downstream to the filter, membrane and chromatography column downstream in the interim, while the upset condition is being fixed. Otherwise, the entire chain of downstream processes can be seriously affected. Other biotechnology involves synthesis and/or modification of intermediates or final products. Frequently, this is in a suspension form so that mechanical mixing, separation, spray or thermal drying, and other allied processes are required. Given separation is an important task [16 20] in biotechnology in lieu of the above, it can be a very difficult task due to the low concentration of the protein present and the large volume of liquid to handle, the fragility of the cells, the presence of cell debris, fine particulates and colloids, and the high viscosity due to dissolution of intracellular substances, such as RNA. Typically, separation can be achieved by filtration and sedimentation. There are some specific problems relating to each as discussed in the following. Filtering a suspension containing biomass is quite tricky as the material can foul the filter surface, reducing permeate or filtrate flow regardless whether the media is a microfiltration or an ultrafiltration membrane. It is equally challenging to settle biomass as the density of the biomass material is just slightly greater than that of the liquid phase, which often is aqueous based. Given that settling rate is proportional to

6

Centrifugal Separations in Biotechnology

the difference in the two densities, it takes a very long time to separate, translating in simple terms—to an impractically low-capacity, high-cost operation. On the other hand, separation by sedimentation can be much enhanced under centrifugal acceleration. This is possible by introducing the suspension with biomass in a centrifuge rotating at high speed where centrifugal acceleration can be hundreds to millions of times that of the Earth’s gravitational acceleration. 1.1.1

Common Host Cells for Secreting Recombinant Protein

Fig. 1.4A shows the use of centrifugation in biopharmaceutical production of therapeutic protein using the commonly employed mammalian cells, bacteria and other microbial hosts [21], and yeasts. Centrifugation (A)

Fermentation and bioreaction

Mammalian cell Extracellular

Bacteria Intracellular

Centrifuge separation

Extracellular Centrifuge clarification

Lysing Solid product

Depth filter polishing

Yeast

Liquid product

Centrifuge IBClassification

Centrifuge lysate clarification

IB washing

Centrifuge/filter polishing

Centrifuge polishing

Downstream processing

Drug substance and monoclonal antibodies

Figure 1.4A Drug substances produced from fermentation and downstream processes where centrifugation has been widely employed for various duties.

Introduction

7

can be used for separation, clarification/polishing, thickening, classification, and washing-and-separation. With reference to the left bioprocess in Fig. 1.4A, after harvesting from the bioreactor, the cell culture suspension containing mammalian cells is sent to a centrifuge wherein the cells are separated from the liquid product which contains the extracellular protein secreted from the mammalian cells. The separated liquid is then sent to a depth filter for further polishing, removing any solid particulates before sending it downstream for processing. Some mammalian cells also secrete soluble protein intracellularly, despite majority secrete liquid protein extracellularly. With reference to the middle bioprocess in Fig. 1.4A, the protein is expressed intracellular in the bacteria. After harvesting and homogenizing, the protein is released from the lysed bacteria in the inclusion bodies (IBs) that need to be isolated before additional downstream processing. Centrifugation is used to sediment the IBs while the cellular contents, cell debris, and finer materials leave with the liquid phase to wasting. The IB is further washed and separated several times until it reaches the desired purity for downstream processing. Alternatively, the protein from the bacteria may be expressed in the intracellular liquid, and upon homogenizing, this protein is released in the liquid. The task is to remove all solid materials and to recover the liquid bearing the soluble protein. The biomass may have to be washed to ensure all the protein that is adhered to the biomass has been recovered, otherwise this represents a loss or lower yield for the process. With reference to the right bioprocess in Fig. 1.4A, after harvesting from the fermenter, the yeast suspension is sent to centrifugation where the liquid containing extracellular protein is separated from the yeast solids. The liquid leaving the centrifuge may have to be centrifuged again (i.e., clarification or polishing) to remove any particulates and turbidity before downstream processing. Some yeast cells may also secrete protein intracellularly as liquid protein. The above three paths are central to biopharmaceutical production of therapeutic proteins using host cells. These will be discussed in much greater detail throughout the text. Besides mammalian cells, microbials, and yeast, infected baculovirus insect cells can also secrete protein. For example, baculovirus-infected insect cells can express mAb, such as antibody C017-1A that recognizes antigen GA733 on colorectal cancer cell. Once the mAb C017-1A is docked onto the antigen GA733, immune cells can attack the colorectal cancer cells [22]. They are being used for multiple gene delivery

8

Centrifugal Separations in Biotechnology

platforms with subsequent cellular expression of protein for labeling cells that can serve as sensors for monitoring cellular architecture or metabolism, and other functions [23]. 1.1.2

Platform for Protein Expression

Despite we have used as an introduction in Section 1.1.1 the three most common host cells to discuss the downstream separation and recovery of protein, it is more prudent to examine how protein is being secreted irrespective of the host cells, from which the downstream protein separation and recovery steps will typically follow. In general, there are three platforms for recombinant protein expressions from cells and microorganisms. The protein can be secreted extracellularly into the broth of the cell culture in the bioreactor or fermenter. It can also be secreted inside the cells in a bioreactor or fermenter. There are two types of intracellular expressions. In one type, the protein is secreted inside the cell as liquid form. In another type, the protein is secreted as a dense mass, commonly referred as IB, which contains misfolded and biological inactive protein. This is illustrated in Fig. 1.4B. It is possible that some cells or microorganisms secrete mostly intracellular protein but also a small amount of extracellular protein and vice versa. Mammalian cells have been thought of secreting only extracellular protein, but some new processes and host cells have shown that they can also secrete intracellular protein. They can release the protein in the (B)

Extracellular protein (liquid)

Recombinant protein expression

Intracellular protein (liquid)

Protein

Intracellular protein (inclusion bodies) IB

Protein Cell

Cell

Cell

Figure 1.4B Three types of secreted recombinant protein by cells and microorganisms.

Introduction

9

broth during maturity without the need of cell lysis. With the advent of new research and technology, it is expected that more new recombinant processes and new host cells will be discovered over time. Regardless, we can look at these protein expression processes being the “model” for downstream processes, including separation, which is a key step upstream of purification, whether it be extracellular or intracellular protein or even a hybrid of both. Without achieving the objective in separation, the steps downstream of separation will only become more difficult leading to poor protein recovery and low yield. Below, we will briefly describe each of these expression processes and the downstream flow sheet in general terms. 1.1.3

Extracellular Protein

Protein is secreted by the mammalian cells in cell culture or yeast in fermenter extracellularly. When protein mixed with the broth is harvested from the bioreactor or fermenter, the objective is to remove any cells or debris from the broth as the liquid contains the liquid protein. If centrifugation is used to carry out the separation, the liquid product should be clarified (free of suspended cells and cell debris) while all the solids should settle in the concentrate stream. The centrate product will be subsequently sent to the depth filter or microfiltration for polishing, that is, to remove any residual fine solids not being removed by centrifugation. Downstream, the liquid product is washed (reslurry with buffer solution) by diafiltration to dissolve any dissolved impurities, including salts, and subsequently sent to ultrafiltration for concentration. Finally, the protein is purified in the chromatography column. The process flow sheet is shown in Fig. 1.4C. 1.1.4

Intracellular Protein—Liquid

Protein can be secreted inside the cells using mammalian cells, yeast, and bacteria in the bioreactor or fermenter. For yeast and bacteria, at harvest, the cells are lysed breaking the cell walls releasing the protein liquid that will be mixed with the broth. The objective of the separation by centrifugation subsequent to lysis is to remove the cell debris and some unlysed cells from the protein liquid. The liquid lysate is clarified using centrifuge in this step. Subsequently, the liquid protein will go through fine filtration, washing, and concentration, the steps of which are analogous to the case of extracellular protein. Finally, the protein is purified by the chromatography. It is perceived that given the size of the cell debris is very small on the order of submicrons, centrifugation at

10

Centrifugal Separations in Biotechnology (C)

Cell culture/ fermentation

Clarification and cell separation

Cell

(centrifugation)

Protein For example mAb, enzymes, antibiotics, amino acids

Fine filtration (depth filtration/ microfiltration)

Washing and concentration (diafiltration, ultrafiltration)

Purification (chromatography)

Figure 1.4C Extracellular protein secreted into broth.

high speed is necessary to generate large centrifugal force to carry out separation. It is indeed a very important step in the process. Otherwise, any unseparated cell debris will be carried downstream overloading the depth filter/microfilter and even to the diafilter and ultrafilter and possibly fouling the chromatography column. The process flow sheet is shown in Fig. 1.4D. For mammalian cells, there are exceptions for which the intracellular protein secreted can be released into the broth. Therefore periodically the broth is removed from the bioreactor, and separation is made for which the separated whole cells are returned back to the bioreactor. The product centrate with the cell debris will undergo another separation, where debris are removed by high-speed centrifugation. The centrate liquid containing protein is sent to the depth filter for further polishing. 1.1.5

Intracellular Protein—Inclusion Body

Extracellular protein can be secreted quite prolifically from the microbials, commonly bacterial cells, in the production of recombinant protein. Unfortunately, the protein is misfolded and inactive and requires

Introduction

11

(D) Fermentation/ cell culture

Cell harvest

Protein Cell

Cell lysis

Lysate clarification For example enzymes, growth, factors, vaccines

(centrifugation)

Fine separation, washing, concentration

Purification (chromatography)

Figure 1.4D Intracellular protein secreted inside the cells.

subsequent processing to restore its bioactivity before use. At harvest, the cells are lysed releasing the protein in the dense IBs. Subsequently, the IB cells are homogenized releasing the IB cells. The IB cells have to be separated from the cell debris and in order to have effective separation, the properties for which separation is based should be distinctly different. Typically, the IB cells are in the size range of 0.2 1.4 µm [24]. The density of the solids is about 1.3 g/cm3, but the IB cells have large void fraction filled with liquid; therefore the effective bulk density may still be about 1.1 g/cm3 or smaller depending on the void fraction. On the other hand, the cell debris in the lysate is also quite small, 0.5 µm and below, and has bulk density in the same range as the cell debris. In the separation step, the IB cells are separated from the cell debris. High-speed centrifugation is used to settle the IB cells from the liquid leaving the fine cell debris to leave with the liquid centrate from the centrifuge. This fractionation or classification step is not perfect. Given the IB cell size and the cell debris are not too far from each other, there is always some cell debris mixed with the IB cells in the sediment. This interesting case will be taken up in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification, when we discuss classification.

12

Centrifugal Separations in Biotechnology (E) Fermentation IB Cell harvest Cell Cell lysis

Classification Inclusion bodies and cell debris (centrifugation)

Solubilization

Refolding

Solubilization, refolding

Purification (chromatography)

Figure 1.4E Intracellular protein secreted into the inclusion bodies.

The IB cells are washed and centrifuged a few times before sending to the next process step. There the IB cells are solubilized moderately by alkaline-based chemicals to dissolve the protein IB. Subsequently the misfolded protein is refolded to restore the bioactivity of the protein. The protein in liquid is subsequently purified in chromatography. If classification of the IB cells and cell debris is not carried out properly by centrifugation, cell debris will be carried downstream causing contamination and fouling of the chromatography column. On the other hand, IB cells which have not been separated by centrifugation and leave the centrate also represent a loss of yield for the entire process. The process flow sheet is summarized in Fig. 1.4E.

1.2

Centrifugal Separation and Filtration

Industrial centrifugal separation [25,26] can be divided generally into two classes: sedimenting and filtering. It should be noted that straight centrifugation can only separate suspended solids and not dissolved solids with the exception of a centrifugal filter, to be discussed later, which can separate soluble solids. Heavier solids settle to the solid wall of a sedimenting centrifuge under centrifugal acceleration that is much

Introduction

13

greater than the Earth’s gravity. A density difference between the solid and the liquid phase is required to affect separation. A schematic of a sedimenting solid-wall centrifuge is shown in Fig. 1.5. Similarly, a lighter dispersed solid phase, like fat or solids with attached air bubbles, can also float (instead of sink) in a continuous liquid phase, and separation by flotation can be enhanced in a centrifugal field. On the other hand, density difference between the two phases is not required to separate the solid from liquid phase in a filtering centrifuge. Both phases are driven under the centrifugal body force to the perforated wall lined with a filter medium. Liquid permeates (see Fig. 1.6) through the filter medium while solids, comparable or larger than the openings of the filter medium, are retained. Sometimes even smaller solids can be retained as they “bridge across” or “jam” the medium openings precluding them from filtering through. As such, openings in the filter medium can be selected normally two to three times larger than the particle size to be retained. Once a “cake” layer of particles forms on the medium despite some smaller particles that may still percolate through during initial filtration, the cake layer further acts as a frontline filter medium in series with the original medium to retain the incoming particles. Slurry pool

Axis

Solid wall

Cake

Figure 1.5 Solid-wall sedimenting centrifuge. Cake

Axis

Filtrate

Slurry pool

Filter medium

Figure 1.6 Filtering perforate-wall centrifuge.

14

Centrifugal Separations in Biotechnology

It can be seen that filtering centrifuges can separate particles and liquid regardless of their density difference; this is very much different from a sedimenting centrifuge that relies on the density difference of the two phases to drive separation. 1.2.1

Sedimenting Centrifuge

Sedimenting centrifuge can be divided into, respectively, batch and continuous sediment discharge as represented in Fig. 1.7. For batch discharge, this can be further divided to batch and continuous feed. Spintube, ultracentrifuge, zonal centrifuge are all classified as batch feed centrifuges despite some zonal centrifuges can have continuous feed and continuous removal features. In batch feed, a fixed amount of feed slurry is introduced to the centrifuge. Upon separation, heavier solids settle to the bowl wall or tube bottom at a large radius. Here they accumulate temporarily until separation stops. Tubular, manual disk, chamber, multibowl, and solid basket are considered as semicontinuous as they take continuous feed of suspension. Heavier and larger solids from suspension get settled under centrifugal Sedimenting centrifuge

Batch discharge

Batch feed

Continuous discharge

Continuous feed

Disk

Decanter

Spintube

Tubular

Droppingbottom disk

Conventional decanter

Ultracentrifuge

Manual disk

Nozzle disk

Dual-cone decanter

Zonal centrifuge

Chamber and multibowl

Compoundbeach decanter

Solid basket

Nozzle decanter

Figure 1.7 Batch and continuous centrifuge classification.

Introduction

15

acceleration and the sediment is stored temporarily in the bowl until the quality of the separated liquid gets affected by the growing sediment in the bowl. At that point, feeding stops, the centrifuge is allowed to coast down, the liquid pool is drained, and the sediment is removed. The centrifuge needs to be cleaned before the next cycle. For continuous or semicontinuous discharge of sediment, both disk and decanter centrifuges fall in this category. Under disk centrifuge, there are two types depending on how the concentrated solids are discharged—dropping bottom with intermittent solid discharge, and the nozzle disk with continuous solid discharge. On the other hand, there are four types under decanter, a conventional decanter, dual-cone decanter for classifying two solids and a liquid phase, compound beach decanter for dewatering fine solids producing a paste-like cake [27], and nozzle decanter for classifying kaolin suspension with 1 2 µm particles. 1.2.2

Filtering Centrifuges

Filtering centrifuges are also divided into two types: batch and continuous feed (see Fig. 1.8). Under batch discharge, there are two categories. First, a small-batch feed under which perforated spintube and basket and centrifugal filter both belong. Second, a large-batch feed under which conventional basket, peelers, siphon, and inverting bag centrifuges all belong. Regardless of the large- or small-batch feed, they take a batch of feed suspension or “charge” and perform various operations to process the suspension, including filtering, washing, deliquoring/dewatering with or without drying, unloading and decelerating (for peeler) or decelerating and unloading (for regular baskets), and cleaning. Pusher, conical screen, and screenbowl are continuous centrifuges. Feed is continuously introduced in these centrifuges, and cake retained by the filter media is continuously being removed. Filtrate liquid, with minimal suspended solids, is also removed separately. Suspension containing biological solids filter very slowly and they often clog up the filter media, forming an impermeable cake. As a consequence, filtration is often affected by a thin cake filtration in a controlled batch mode under moderate centrifugal gravity so that the cake does not compact to an impermeable “skin” layer adjacent to the filter medium. In addition, solids should be at least greater than 10 µm, otherwise filtration is slow and impractical. If the valuable product is the liquid, it is possible to use a filter aid, such as diatomaceous earth, to enhance the filtration rate provided the filter aid does not interact with the product. In essence, this increases the permeability of the filter cake.

16

Centrifugal Separations in Biotechnology

Filtering centrifuge

Batch discharge

Small-batch feed

Continuous discharge

Large-batch feed

Pusher

Screenbowl

Conical Screen

Perforated spimtube

Conventional basket

Single-stage pusher

Conventional screenbowl

Scroll screen

Centrifugal filter

Peeler and siphon basket

Multistage pusher

Insitu cake washing

Vibrating screen

Invertingbag basket

Tumbling screen

Wide-angle screen

Figure 1.8 Batch and continuous filtering centrifuge classification.

When the product protein is in the solids, such as in the biological cells, filter aid should not be used as it is almost impossible to separate the filter aid from the valuable biological material.

1.3

Pros and Cons of Filtration Versus Centrifugation

Centrifugation followed by depth filtration, two-stage depth filtration, or microfiltration diafiltration can all be used alone for solid liquid separation when the valuable protein is in liquid phase. With reference to Fig. 1.4A, centrifugation is frequently used for lysed cells involving release of protein solids to be separated from the cell debris when bacteria are used as host cells. Table 1.1 compares the pros and cons of centrifugation and microfiltration. It is quite interesting that centrifugation

Introduction

17

Table 1.1 Comparing centrifugation and microfiltration (MF). Pros

Cons

Less shear compared with that generated from tangential flow filtration or crossflow filtration Cell debris does not foul or clog up pores leading to blinding; advantageous for higher feed solids (future trend) Remove solids down to 0.2 µm at high G Compact Less downtime (from fouling) Single pass or multiple passes Robust Slightly higher capital costs

Comparable operating costs (power and maintenance) as MF, which requires periodic replacement of membranes from fouling.

actually generates less shear stress than tangential flow filtration contrary to conventional wisdom, provided a good feed acceleration is adopted in the centrifuge, see Chapter 4, Disk Centrifuge. Also, centrifugation does not suffer from the fouling of membrane that leads to costly membrane replacement and downtime. Also, it is a very robust and reliable system. The solids in suspension for the process of interest are very small, in the domain of 10 µm and below, and sometimes even in the 1 2 2 µm range. While the small density difference affects sedimentation and not as much for filtration, as discussed, the fine biosolids affect depth filtration, clogging flow path of the filter, leading to rapid large pressure drop across the depth filter. For microfiltration, the membrane needs to be stayed “unclogged” or “unfouled” by the use of a crossflow using a high shear rate to scour the membrane surface (see Chapter 14: Rotating Membrane in Bioseparation) and to use cleaning agent to wash the membrane during downtime. Also, diafiltration helps to maintain a lower solid concentration, to prevent fouling. Nevertheless, fouling leading to the replacement of membrane and an associated downtime are the key drawbacks of microfiltration in this application. In addition, a much larger volume of liquid product (two to four times the original volume) results from washing or diafiltration to recover the protein. This implies a higher cost for the process as it involves more concentration by ultrafiltration and other processes downstream for treating the spent liquid. A fourth possibility is to combine centrifugation with depth filtration as an integrated approach to separate mammalian cells. Centrifugation takes up the solids loading from the feed stream leaving

18

Centrifugal Separations in Biotechnology

the fermenter/bioreactor, while the depth filter works best in removing the submicron particles (separated liquid from the centrifuge) from a low-solid stream leaving the centrifuge. This will be discussed in more detail in Chapter 6, Commercial Applications of Centrifugation in Biotechnology.

1.4

Generic Flow Sheet for Biopharmaceutical Process

The recombinant protein process has become very popular for engineering a protein to have specific configurations and functions [2 9]. Various recombinant proteins are commonly expressed through an engineering culture such as yeast, bacteria (e.g., E. coli), and living cells (e.g., mammalian and plant cells). The extracellular (exterior of cells) protein expressed by yeast and mammalian cells are in the liquid phase, while protein expressed by the engineered bacteria is inside the bacteria (i.e., intracellular), which subsequently needs to be lysed to release the protein. The process condition of a cell culture is carried out under specific range of temperature, pressure, and agitation/mixing, with fermentation usually subject to shorter time and more intensive condition (higher process temperature, pressure, and with more rigorous mixing), while bioreaction takes place under longer reaction time and milder conditions (lower process temperature, pressure, and only moderate mixing). Fig. 1.9 shows a generic flow sheet for processing recombinant protein in which protein is expressed extracellularly wherein liquid is the product. The immediate first step—the separation step—is also referred to as primary recovery of protein. After solid liquid separation, by any of the aforementioned four different possible separation methods, the protein buffer liquid may be replaced or diluted with a more appropriate buffer with a different pH and ionic strength followed by a concentration using an ultrafiltration and diafiltration combination. The end product is a concentrated protein solution in a suitable buffer liquid. At this stage, the protein can be purified using ion-exchange or affinity Bioreactor/ cell culture Centrifugation Fermenter

Coarse and fine filtration

Concentration, buffer exchange (UF/DF)

Purification (chromatography)

Sterile filtration

Drug substance

Figure 1.9 Generic flow sheet of biopharmaceutical drug substance.

Introduction

19

chromatography to remove any impurities and contaminants. A final sterile filtration involves the use of a 0.2-µm-sized microfilter to remove bacteria that are incurred during processing. The final product is typically a drug substance or an antibiotic.

1.5

Other Centrifugal Separations

Other than for primary recovery in the downstream process of recombinant protein, centrifugal separation is also used in many biotechnology solid liquid separations in manufacturing of drugs/hormones such as insulin and many others. In the process of manufacturing, drugs (in solid form) frequently contain salt and other impurities. In such cases, the drugs need to be washed by reslurrying followed by centrifugal separation. Also, crystallization and precipitation in the purification step require solid liquid separation by centrifugation. The objective in these processes is to fully capture or recover the valuable suspended solids, unlike the recovery of soluble protein expressed extracellularly by yeast and mammalian cells wherein the product is the liquid. Another equally important objective is to reduce the impurity level of the crystals to an acceptable level for downstream formulation, such as washing followed by centrifugal separation.

1.6

Inputs and Outputs of Centrifuge

Centrifuge has often been considered as a black box, as the mechanics and fluid dynamics are quite complex. Here we will discuss the scientific basis and understanding of centrifugation and operation of various types of available centrifuges as commonly used in separation in biotechnology. In the simplest terms, for a centrifuge processing, a wet feed suspension after centrifuging, a centrate, supernatant, or overflow containing a small amount of suspended solids leave the centrifuge together with a moist concentrate, wet cake, or underflow. This is depicted in Fig. 1.10. There could be also another input such as chemicals (coagulants and flocculants) added to flocculate the feed suspension (not shown in Fig. 1.10). Based on the previous discussion, the valuable protein Feed

Centrifuge

Centrate/effluent/ overflow liquid

Concentrate/cake/ underflow solids

Figure 1.10 Typical input and output streams of centrifugation.

20

Centrifugal Separations in Biotechnology

product can be in the fine suspended solids, such as the IBs, crystals or precipitants containing protein, or in the centrate liquid phase, as in the extracellular protein expression (yeast and mammalian cells). The centrifuge needs to be tuned to separate the product from the rest (waste or recycle stream). Depending on the specific process, as discussed below, some metrics or measures are commonly used to assess the centrifugal separation.

1.7

Separation Metrics

Several measures of centrifuge performance are common—protein yield, suspended solids in the centrate or solid recovery, throughput or capacity, and cell viability. 1.7.1

Protein Yield

For soluble protein expressed from extracellular process, one important measure of the separation performance of the centrifuge is the protein yield Y. Yield is defined as the ratio of the amount (e.g., kg/min or g/min) of protein recovered in the liquid product to the amount (kg/min or g/min) of protein in the feed to the centrifuge. A complete recovery of protein without loss is 100%. Usually, the yield should be very high before the separation process can be considered viable. A 90% or higher in yield is typical. The specific yield depends on how difficult is the separation. An example on protein yield is given, respectively, in Chapter 7, Concentrating Solids by Centrifugation, and Chapter 8, Laboratory and Pilot Testing. For continuous feed centrifuge, the volumetric rate (L/min) and protein concentration of both feed and centrate need to be measured, respectively, for the calculation of yield. For batch feed centrifuge, the volume and protein concentration of both feed and supernatant (i.e., centrate) should be measured, respectively, for the yield calculation. It is evident that liquid loss in the concentrate or cake affects the yield as the protein is dissolved in liquid; therefore the amount of liquid in the concentrate should be minimized (see Chapter 7: Concentrating Solids by Centrifugation). 1.7.2

Centrate Suspended Solids

The centrate suspended solids should be minimized unless this is for classification, wherein finer-sized solids in the centrate are separated from larger solids in the concentrate as found for separating cell debris

Introduction

21

from IBs. A measure of clarity of the liquid centrate is the amount of suspended solids by weight or by bulk volume after the centrate is spun in a spintube centrifuge for a prescribed time. For cell culture, only fine solids in the submicron range escape, with the centrate or supernatant to be ultimately captured by the downstream filter. An indirect method on assessing the centrate suspended solid is to measure the optical opacity (or turbidity) of the centrate liquid. The turbidity measurement should be calibrated against a standard on a frequent basis. A consequence of good clarification is that the solids recovered by sedimentation (kg/h, dry basis) compared with the feed solids (kg/h, dry basis) should be very high. The ratio is referred to the solid recovery Rs. When Rs is at 100%, this implies perfect separation, and there is no solid in the product centrate or supernatant. For cell culture, we may achieve, say, 99.9% recovery of cells by centrifugation, leaving minimal cells escaped in the centrate or supernatant. 1.7.3

Throughput Rate

The volumetric rate or capacity of centrifuge (in L/min) is an important measure of the volumetric liquid throughput capacity that a centrifuge can attain. Once the total capacity (size and number) of fermenter or bioreactor is fixed, the rate of the centrifuge(s) is determined, and this in turn bears out the total capacity (size and number) of centrifuges. Highrate larger centrifuges require fewer centrifuges when compared with low-rate smaller centrifuges. On the other hand, a spare centrifuge needs to be furnished to cover the operating centrifuges when one of these centrifuges is rotated out for maintenance. Again, the operation planning requires the information on centrifuge throughput rate. 1.7.4

Cell Viability

Mammalian cells are gaining popularity in expressing protein as there is more flexibility in pursuing this route. However, unlike plant cells or yeast, mammalian cells are very shear sensitive as they do not have a cell wall. They are highly susceptible to shear, such as during acceleration of the feed stream. Cells can be destroyed in the process, releasing an intracellular protein substance that can be harmful for downstream purification (cross-contamination of product protein) and finer debris which renders the separation problem more difficult, resulting in increased suspended solids loading up the depth filter. As a minimum, cell viability of mammalian cell lines should be maintained at a high

22

Centrifugal Separations in Biotechnology

level during separation using production centrifuges. This will be discussed in detail in later chapters. In subsequent chapters, the principle of centrifugation, types of centrifuges, application of centrifuges, selection, sizing, modeling, and scaleup will be discussed.

1.8

Text Organization

The present text is organized into six areas as illustrated in Fig. 1.11. The basics of rotating flow phenomena and solid liquid separation by sedimentation are discussed in Chapter 2, Principles of Centrifugal Sedimentation. Next, different types of batch, semibatch, and continuous centrifuges are presented in Chapters 3 5. Following this, some typical generic application flow sheets for biotechnology and biopharmaceutical processes are presented in Chapter 6, Commercial Applications of Centrifugation in Biotechnology. Next, the practical end of the subject is given in Chapters 8 10, respectively, on testing, selection and sizing, and troubleshooting and optimization. The modeling on tubular and decanter centrifuges is given in Chapter 11, Visualization and Modeling of Flow and Separation in Tubular Centrifuge, and modeling on diskstack centrifuge in Chapter 12, Disk Stack Modeling. These models require known feed particle size distribution (PSD) in order to predict the performance as detailed by numerous examples in Chapter 13, Performance Projection of Centrifuges in Bioseparation. Unfortunately, 18. Integrated approach

Sedimentation, rotation flow

2. Basicson Rotation flow, Sedimentation

Concentration, expression, filtration

7. Concentration, Dewatering

Centrifuge types

3. Batch, UltraCentrifuge, Spintube, Tubular

Applications

Practice

Modeling and case studies

6. Flow Sheets

8. Testing

11. Modelling Tubular, Decanter

9. Selection,Scale-up 12. Modelling Disk

14. Filtration 4. Disk 5. Decanter

10. Troubleshoot, Optimization

13. Numerical Simulation 15. Floc, Monotonic PSD, Applications 16. MonotonicPSD case studies, Unimodal PSD, Applications 17. Bimodal PSD, Classification

Figure 1.11 Text organization (Note: Numbers in the boxes refer to chapter numbers in the text.)

Introduction

23

the feed PSD is frequently unavailable. In Chapters 15 17, the models are further simplified by adopting a limited number of parameters to characterize the feed PSD from which analytical solutions can be directly obtained on solids recovery and size recovery. This facilitates calibration of the model by determining these parameters using limited sets of test data available. After the models have been calibrated, they should be able to predict centrifuge performance on conditions that have yet been tested. Other than separation by sedimentation, centrifugal filtration is presented in Chapter 14, Rotating Membrane in Bioseparation, and centrifugal compaction and expression to obtain dry cake is presented in Chapter 7, Concentrating Solids by Centrifugation. Along the interest of the readers, they can select accordingly combinations in these six areas in the text to meet their needs.

1.9

Summary

In this chapter, some important applications in biotechnology, such as manufacturing of drug substances purely from biologically derived products, have been presented. These processes are discussed generically so that they can be applicable for various situations. Also, it is important to understand that centrifugation should be partnered closely with other process equipment, both upstream and downstream, to make the entire process work using an integrated approach. Various separation metrics for centrifugation are discussed, including protein yield, centrate suspended solids, throughput, and cell viability. These subjects will be taken up in greater detail in the text. The organization layout of this book has been outlined for the need of the reader.

References [1] C. Ratledge, B. Kristiansen, Basic Biotechnology, Cambridge University Press, 2001. [2] M. Vestergaard, S.H. Chan, P.R. Jensen, Can microbes compete with cows for sustainable protein production—a feasibility study on high quality protein, Sci. Rep. 6 (2016). Article number: 36421. [3] S. Kapoor, A. Rafiq, S. Sharma, Protein engineering and its applications in food industry, Crit. Rev. Food Sci. Nutr. 57 (11) (2017) 2321 2329.

24

Centrifugal Separations in Biotechnology

[4] E. Waltz, Appetite grows for biotech foods with health benefits, Nat. Biotechnol. 37 (2019) 573 575. [5] D. Jao, Y. Xue, J. Medina, X. Hu, Protein-based drug-delivery materials, Mater. (Basel) 10 (5) (2017) 517. [6] R. Langer, Perspectives: drug delivery - drugs on target, Science 293 (2001) 58 59. [7] S. Khan, et al., Role of recombinant DNA technology to improve life, Int. J Genomics. 2016 (2016). Available from: https://doi.org/10.1155/ 2016/2405954. 2405954. [8] M. Singh, L.D. Silva, T.A. Holak, DNA-binding properties of the recombinant high-mobility-group-like AT-hook-containing region from human BRG1 protein, Biol. Chem. 387 (10 11) (2006) 1469. [9] C. Budde, M.J. Schoenfish, M.E. Linder, R.J. Deschenes, Purification and characterization of recombinant protein acyltransferases, Methods: Companion Methods Enzymol. 40 (2) (2006) 143 150. [10] G. Ko¨hler, C. Milstein, Continuous cultures of fused cells secreting antibody of predefined specificity, Nature. 256 (5517) (1975) 495 497. [11] K. Chandrakant, et al., Textbook of Pharmaceutical Biotechnology, Elsevier, 2016. [12] https://www.cancer.org/cancer/chronic-lymphocytic-leukemia.html [13] I. Kimiz-Gebologlu, S. Gulce-Iz, C. Biray-Avci, Monoclonal antibodies in cancer immunotherapy, Mol. Biol. Rep. 45 (2018) 2935. [14] H. Kaplon, J.M. Reichert, Antibodies to watch in 2019, MAbs 11 (2) (2019) 219 238. [15] M. Desai (Ed.), Downstream Processing of Proteins, Humana Press, Totowa, NJ, 2000. [16] M.S. Verrall, M.J. Hudson (Eds.), Separations for Biotechnology, Ellis Horwood Ltd. Publishers, Chichester, and Society of Chemical Industry, London, 1987. [17] P.A. Belter, E.L. Cussler, W.-S. Hu, Bioseparations—Downstream Processing for Biotechnology, John Wiley and Sons, New York, 1988. [18] G. Subramanian (Ed.), Bioseparation and Bioprocessing, Wiley-VCH, Verlag GmbH, 1998. [19] R.G. Harrison, P. Todd, S.R. Rudge, D. Petrides, Bioseparations Science and Engineering, Oxford University Press, New York, 2003. [20] M.R. Ladisch, Bioseparations Engineering—Principles, Practice, and Economics, John Wiley and Sons, New York, 2001. [21] G.L. Rosano, E.A. Ceccarelli, Recombinant protein expression in microbial systems, Front. Microbiol. 5 (2014) 341. [22] J. Zaloudikl, et al., Expression of an antigen homologous to the human C01 7-1A/GA733 colon cancer antigen in animal tissues, Br. J. Cancer 76 (7) (1997) 909 916.

Introduction

25

[23] M. Mansouri, et al., Highly efficient baculovirus-mediated multigene delivery in primary cells, Nat. Commun. 7 (2016) 11529. [24] G. Margreitera, P. Messnerb, K.D. Caldwell, K. Bayer, Size characterization of inclusion bodies by sedimentation field-flow fractionation, J. Biotechnol. 138 (2008) 67 73. [25] W.W.F. Leung, Industrial Centrifugation Technology, McGraw-Hill, New York, 1998. [26] W. Stahl, Industrie-Zentrifugen, Maschinen-Verfahrenstechnik, Dr.M Press, Maennedorf, Germany, 2004. [27] W.W.F. Leung, A.H. Shapiro, Dewatering of fine-particle slurries, Miner. & Metall. Process. 19 (1) (2002) 1 8.

Problems (1.1) Can you engineer protein to be expressed reliably in intracellular contents of the cells in the form of soluble substance instead of in inclusion bodies (solid) using E. coli or Bacillus subtilis? (1.2) Assuming you can successfully express protein in intracellular form as in Problem (1.1), what are the downstream processes to extract, separate, and purify the protein at the time of harvest? (1.3) If the protein from Problem (1.1) is in the solid form, such as inclusion bodies which are 0.5 µm in equivalent diameter and specific gravity of 1.3, after homogenizing the bacteria cell how would you separate the inclusion bodies from other insolubles in the cells which are, say 0.3 µm and smaller, together with other unlysed cells 1 3 3 µm2, assuming the insolubles and the unlysed bacteria have a specific gravity of 1.2? (1.4) Name a few more difficult bioseparation examples? What are the difficulties, and how would you propose to overcome such difficulties? (1.5) In the process flow sheet shown in Fig. 1.9, which is the most difficult step and why? How would you address this most difficult step? (1.6) What are the pros and cons in using high centrifugal force to filter biological solid, such as running equivalent to 10,000 times the Earth’s gravity, through an ultrafiltration membrane? (1.7) Why can’t you always resort to using high centrifugal acceleration, such as half a million times the Earth’s gravity to affect sedimentation as this apparently will overcome small density difference, viscous liquid, and small cell sizes of about 1 µm?

2 Principles of Centrifugal Sedimentation 2.1

Introduction

Centrifugation makes use of high-speed rotation to generate centrifugal acceleration acting on phases with different densities. Under centrifugal acceleration, heavier phases migrate to a location at a larger radius (toward the bowl periphery) while lighter phases are displaced to a location at smaller radius (toward the axis of the bowl). For continuous feed to the centrifuge, dynamics is very important. The fluid dynamics in a rotating frame, such as that observed in a rotating bowl, is highly nonintuitive as compared to the same occurring in a nonrotating reference frame. In this chapter, we will first present some nonintuitive phenomena, such as Coriolis effect [1] and nonlinear pressure behavior. This is followed by a discussion of more intuitive key mechanisms, such as centrifugal force and Stokes’ law on sedimentation.

2.2 2.2.1

Nonintuitive Phenomena Pressure Gradient in a Fluid Under Centrifugal Acceleration

Consider a circular arc segment at radius R with unit depth (into the paper) and subject to a hydrostatic pressure p. At a radial distance R 1 dR, the pressure is p 1 dp (see Fig. 2.1A). The arc segment has an average planar area dA between the two radii. For a fluid under solidbody rotation, the difference in pressure exerted by the fluid on these two respective planes is balanced by the centrifugal force ρdA(dR)Ω2R acting on the fluid mass ρdAdR bound by these two planes. Assuming no radial acceleration, the net difference of hydrostatic forces between the two planes exactly balances the radial outward centrifugal body

Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00002-7 © 2020 Elsevier Ltd. All rights reserved.

27

28

Centrifugal Separations in Biotechnology p + dp

Ω2R p R

dR

Ω

Figure 2.1A Schematic representation of an circular arc element of fluid subject to pressure and centrifugal forces.

force acting on the mass ρdAdR thus
 dp dR dA 5 pdA 1 ðdRÞ ðdAÞ ρΩ2 R p1 dR Simplifying, dp 5 ρΩ2 R dR

(2.1a)

Integrating between radius R at a given radial location and the reference radius Ro, with pressure at p and po, respectively, p 2 1=2 ρΩ2 R2 5 po 2 1=2 ρΩ2 R2o

(2.1b)

Therefore if the fluid pressure at the free surface of a rotating fluid with radius Ro is po, then the pressure at a larger radius R ( . Ro) becomes p 5 po 1 1=2 ρΩ2 ðR2 2 R2o Þ Suppose ρ 5 1000 kg/m3, Ro 5 5 cm (0.05 m), R 5 7 cm (0.07 m), Ω 5 418.9 rad/s (4000 rpm), po 5 1 atm (101,325 Pa), then p 5 1 atm 1 1=2

ð1000Þ ð418:9Þ2 ð0:072 2 0:052 Þ 5 1 1 2:1 5 3:1 atm 101; 325

Under high-speed rotation of 4000 rpm, it is interesting that at a water depth 2 cm below the free surface, a differential hydrostatic pressure of 2.1 atm can be generated. This is much greater than the hydrostatic head due to the Earth’s gravity. 2.2.2

Combined Centrifugal and Gravitational Accelerations

Consider a liquid under rotation with centrifugal force acting radially outward and gravity force acting downward as shown schematically

Principles of Centrifugal Sedimentation

29

in Fig. 2.1B. A free surface is formed between the two radii R and R 1 dR on the fluid. Consider points A and B in Fig. 2.1B. When the pressure at point A is po, then the pressure at point B is given by pB 5 pA 1 ρgdh 5 po 1 ρgdh

(2.2a)

According to Eq. (2.1b), there is no difference in pressure from the effect of the centrifugal field as both points are at the same radii. Invoking Eq. (2.1b) between points B and C, we get pB 5 po 1 1=2ρΩ2 ½ðR1dRÞ2 2 R2   po 1 ρΩ2 RdR

(2.2b)

In Eq. (2.2b), we have set pc 5 po due to point C being on the free surface (see Fig. 2.1B). Equating Eqs. (2.2a) and (2.2b), we obtain dh 5

Ω2 RdR g

(2.2c)

Integrating Eq. (2.2c) between the axis at R 5 0 with h 5 0 (taken as reference height) to a given radius R with a liquid level rise h, we obtain h5

Ω2 R 2g

(2.2d)

Two important observations can be observed from Eq. (2.2d). First, Eq. (2.2d) indicates that there is an increase in free surface height with increasing radius due to centrifugal force in driving the fluid radially outward that balances the gravitational force in “levelling” the surface. In consequence, a parabolic shaped surface is formed. Second, in Eq. (2.2d) the effect of density cancels out as both centrifugal acceleration and gravitational acceleration are body acceleration and they are both proportional to the fluid density. Therefore Eq. (2.2d) applies to fluids with any density.

A g dh

R C

B dR

h

Figure 2.1B Schematic representation of interface as a balance of centrifugal and gravitation forces.

30

Centrifugal Separations in Biotechnology

As an example, suppose the rotation speed Ω 5 104.7 rad/s (1000 rpm), R 5 0.5 cm (0.005 m), g 5 9.81 m/s2, h can be calculated h5

ð104:7Þ2 ð0:005Þ2 5 0:014m 5 1:4 cm 2ð9:81Þ

h (cm)

As will be seen later, when the feed is introduced into a disk centrifuge. If the feed is not fully accelerated in the proximity of the axis, the actual rotation speed of the fluid is only a fraction of the solid-body rotation speed of the centrifuge. The significance of Eq. (2.2d) is that the differential level of the fluid, h, between the inner wall of the rotating distributor and the rotation axis is a good indicator of the actual rotation speed of the fluid, as hBR2. Higher is the water height h, higher is the rotating speed of the fluid; vice versa, the lower is the water rise with increasing radius, the lower is the effective speed or centrifugal acceleration of the feed. This is the basis of the hematic design discussed in Chapter 4, Disk Centrifuge, that seals air from being in contact with the feed to the disk centrifuge causing oxidation. Fig. 2.1C shows an example of the rotating feed distributor tube with diameter 2 cm of a disk stack centrifuge (to be discussed in Chapter 4: Disk Centrifuge). Water is introduced to the feed distributor along the rotating axis of the machine. The three curves in Fig. 2.1C represent the free surface of a rotating fluid column at respective three different rotation speeds under the influence of both centrifugal acceleration and the Earth’s gravity. Accelerating an initial nonrotating feed fluid promptly up to full solid-body rotation is not trivial. This is especially at high rotation speed of the 10 9 8 7 6 5 4 3 2 1 0

Ω2R

g

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

R (cm) Figure 2.1C Example on a rotating column with effective angular rotation speed of fluid corresponding to 1000, 2000, and 3000 rpm, respectively. The curve represents the free surface of the rotating water column under influence of both centrifugal and gravitational acceleration.

Principles of Centrifugal Sedimentation

31

centrifuge, which may be rotating at thousands of rpm, depending on the size of the centrifuge. The smaller is the size, the higher is the rotation speed. Assuming the effective angular rotation speed of the fluid has attained only 1000, 2000, and 3000 rpm, respectively, we can determine the free surface shape of the rotating water column from Eq. (2.2d). At effective rotation of fluid at 1000 rpm, the water has climbed up along the distributor wall at a radial location from the axis, R 5 1 cm, with a liquid height h of 5.6 cm higher than the liquid height at the axis (R 5 0). (Note that the distributor wall and the centrifuge may be rotating at a much higher angular speed, say, 4000 rpm but the fluid may attain a rotation speed only a fraction of that in the process of being accelerated.) If the effective rotational speed of the fluid is at 2000 or 3000 rpm, this climb in water level could be much higher as shown in Fig. 2.1C. Therefore the level for which the fluid level has risen above the minimum level (at rotation axis) is a good indicator of the effective rotation of the feed fluid entering into the centrifuge, which is in process of being accelerated to solid-body rotation. This aspect will be discussed in Section 2.3.2. Also, note that all the free surfaces of the fluid shown in Fig. 2.1C are parabolic shaped, as stipulated by Eq. (2.2d). The effect due to both centrifugal and gravitational accelerations on pressure distribution in the fluid with a free surface can be generalized. This would be applicable in centrifuge designs where the centrifugal acceleration effect is not large, yet the effect of gravity is not negligible either. This applies to the region close to the rotation axis when the feed has not fully accelerated to solid-body rotation. With reference to the two locations with coordinates (R, h) and (Ro, ho), see configuration in Fig. 2.1D, the effect due to combined centrifugal and gravity accelerations can be generalized as p 2 po 5 1=2ρΩ2 ðR2 2 R2o Þ 1 ρgðh 2 ho Þ

(2.2e)

When g effect is negligible, Eq. (2.2e) defaults back to Eq. (2.1b).

h Ω2R

ho

Ro

po

g

R p Ω Figure 2.1D Pressure, p, at (R, h) due to centrifugal and gravity accelerations as compared with reference pressure, po, at (Ro, ho).

32

Centrifugal Separations in Biotechnology

2.2.3

Coriolis Effect

Consider a ball rolling over a turntable rotating anticlockwise as shown in Fig. 2.2. Once the ball enters the rotating turntable, it is subject to an additional Coriolis acceleration that orients perpendicular to the velocity and specifically directs 90-degree clockwise away from the velocity vector. This Coriolis acceleration acts to skew the trajectory of the ball toward a direction opposite to that of rotation. The latter is referred to as retrograde motion [2]. As soon as the ball leaves the turntable, it follows again a straight path. The Coriolis acceleration is given by ~ v~r a~C 5 2 2ΩX

(2.3)

~ is the angular rotation vector (it is directed out of the paper when Ω the sense of rotation is counterclockwise using the right-hand rule) and v~r is the relative velocity vector in the rotating reference frame of the turntable. In Fig. 2.2, take the relative velocity vr 5 v, vc is the average Coriolis velocity as a result of the Coriolis acceleration ac for an infinitesimal time Δt. The path that the ball takes is a curved trajectory toward the clockwise direction (retrograde motion). This is opposite to the rotation direction (counterclockwise) as a consequence of Coriolis velocity and acceleration, directing the ball moving continuously rightward deviating from its initial velocity direction before the ball enters the turntable (see Fig. 2.2). It is noted that given vc 5 1/2acΔt, vc is small as Δt is infinitesimally small. Therefore this does not change the Original trajector Table rotation Ω ac vc 4

v 3 2

Retrograde motion of ball

v

Curved ball Linear ball trajectory trajectory

1

Turntable

v Linear ball trajectory

Figure 2.2 Ball trajectory changing in a rotating turntable rotating anticlockwise.

Principles of Centrifugal Sedimentation

33

magnitude, but only continuously change the direction of the velocity vector v until the ball leaves the turntable. A corn starch suspension was rotated by stirring using an external mixer (not shown), as depicted by the schematic representation in Fig. 2.3A. At t 5 0, stirring stops and the suspension is allowed to settle. Fig. 2.3BD show a sequence of pictures taken at different time intervals after the stirring had stopped. After some time, the suspended corn starch particles settled on the bottom and collected at the center of the container. Fig. 2.3B shows that the suspended particles were still rotating with the flow at the periphery of the circular pan after a short time interval when stirring had ceased. Fig. 2.3C shows that the corn starch

Wall stationary

Fluid rotating from external stirring

Figure 2.3A Fluid rotating clockwise from external stirring in a stationary container.

Wall stationary

Figure 2.3B Liquid with suspended particulates forced to rotate by stirring; stirring stopped at t 5 0, picture taken after stirring stopped at t 5 5 s, liquid still continue to rotate.

34

Centrifugal Separations in Biotechnology

Figure 2.3C Picture taken at t 5 25 s after stirring stopped, liquid coming to stand-still, particulates accumulate near center.

Figure 2.3D Picture taken at t 5 40 s, heavier solids (corn starch) settling and accumulating at the center of the container.

particles settled as they were brought to the container center by secondary flow at a slightly longer time. Fig. 2.3D shows that after 40 seconds, the secondary flow had subsided and particles concentrated and settled at the center of the container. The diffusion coefficient D for corn starch as reported in the literature [3] is 6.59.8 3 1026 cm2/s for temperature between 7 C and 60 C. Given the radius of the container is 20 cm, the time for a corn starch

Principles of Centrifugal Sedimentation

Secondary flow

35

Heavier particles transported to center of container by secondary flow and settle there.

Figure 2.3E Cross section of the diametric plane showing solids distribution being affected by both secondary flow from previous stirring and sedimentation of heavier solids.

particle to diffuse from the periphery (large radius) to the center would have taken much long time, as shown by the following calculation: t5

L2 202 1 5 1:29 years 5 26 D 9:8ð10 Þ ð3600Þð24Þð365Þ

Some recent work [4] shows that the diffusion coefficient is even an order of magnitude smaller than the values reported [3]. This translates to even longer time (at least 10 times) if diffusion is the mechanism behind the transport of particles. Irrespective of the actual value of diffusion coefficient for corn starch, slow mass diffusion could have never transported particles to the center of the container in 40 seconds if not for the fact that the secondary circulatory flow is doing the actual transport of the corn starch. Fig. 2.3E sketches the secondary flow pattern responsible for transport during spindown of the liquid. It is equivalent to the same familiar experience of how tea leaves settle toward the center of the cup after stirring and flow have stopped.

2.3

Intuitive Phenomena

Besides the nonintuitive phenomena presented in the foregoing, there are other phenomena that occur in a centrifuge which are more intuitive, and they will be discussed in the following. 2.3.1

Centrifugal Acceleration

The magnitude of the centrifugal acceleration G is related to the tangential velocity v and the radius R from the axis of rotation via the

36

Centrifugal Separations in Biotechnology

kinematic relationship, G5

v2 R

(2.4)

For the special case whereby the entire body rotates as a solid-body, the tangential velocity v is linearly proportional to the radius R (see Fig. 2.4) thus v 5 ΩR

(2.5a)

The proportional constant of the linear relationship is the angular speed, Ω. Using Eqs. (2.4) and (2.5a), the centrifugal acceleration G for a solid-body rotation becomes, G 5 Gsb 5

v2 5 Ω2 R R

(2.5b)

where sb in Eq. 2.5b denotes under solid-body rotation. The centrifugal acceleration is often expressed in terms of the Earth’s gravitational acceleration g (59.81 m/s2). The ratio between G and g is referred to as relative centrifugal force (RCF) 8 2 v > > > > < gR FG mG G 5 5 5 RCF 5 (2.6a, b) > Ω2 R mg g > Fg > > : g Eqs. (2.6a) and (2.6b) are for general case (not attaining solid-body motion) and for solid-body rotation, respectively. With commonly used

v = ΩR V

R Ω

Figure 2.4 Tangential velocity increases linearly with increasing radius from the axis of rotation under solid-body rotation.

Principles of Centrifugal Sedimentation

37

engineering units wherein v is expressed in m/s, Ω in rpm (or rev/min), diameter D (52R) in mm thus 8 v2 > 203:87 ; v 6¼ ΩR G < D 5 (2.6c, d) > g : 2 27 5:5893 3 10 ðRPMÞ D; v 5 ΩR For the case of a bowl with diameter of 500 mm rotating at 6000 rpm, the G/g ratio becomes 10,061 or G 5 10,061 g. Fig. 2.5 compiles a list of centrifuges with various speeds and diameters under solidbody rotation using Eq. (2.6d). Small diameter centrifuges operate at lower capacity but at higher G, whereas large diameter centrifuges operate at higher capacity but at lower G. The upper limit for each type of centrifuge is only nominal. There are always exceptions depending on the specific design and material of construction as offered by the manufacturer. The angular velocity, rpm, of the centrifuge may not be known if a strobe or a speed tachometer is not available for making measurement. However, if the centrifuge is belt-driven from a motor of known speed (from the name plate), then one can use an approximate kinematic relation to back out the centrifuge speed assuming little-to-negligible belt slip as follows (see Fig. 2.6):

G/g

rpmcentrifuge 5 rpmmotor

Dmotor Dcentrifuge

(2.7a)

Hi speed tubular limit

20,000 18,000 16,000 14,000 12,000 10,000 8000 6000 4000 2000 0

400 440 500 580

Disk 320 240

750

180

Tubular Decanter

Dia. = 150 mm 0

5000

10,000

15,000

rpm

Figure 2.5 Relative centrifugal acceleration versus rpm for various bowl sizes and nominal max G limits (dash curves) for various types of centrifuges.

38

Centrifugal Separations in Biotechnology rpm motor

Dmotor

rpm centrifuge

Dcentrifuge

Figure 2.6 Pulley ratio of driver and driven on rotation speed adjustment.

or rpmcentrifuge 5 rpmmotor

N motor N centrifuge

(2.7b)

D is the diameter of the sheave or pulley, and N is the number of teeth on the gear. It is understood that the motor speed information can be spec out from the motor. The linear velocity of the pulley/sheave and gear are assumed to be the same or nearly the same in both Eqs. (2.7a) and (2.7b). 2.3.2

Fluid in a Centrifuge Bowl Not at Solid-Body Rotation

When a fluid is introduced to a rotating bowl, it takes time for the fluid in the bowl to accelerate to a solid-body rotation. In the interim, the tangential velocity of the fluid is less than that established for solid-body rotation (Eq. 2.5b). It is apparent that the higher is the feed rate the less chance that the fluid is fully accelerated to a solid-body rotation. This will be discussed further in Chapter 4, Disk Centrifuge. Likewise, when fluid is suddenly changed from a given radius to a new radius significantly different from the initial, the fluid slip, and skid as it is not up to the local speed. Meanwhile, there is continuous momentum transfer from the neighboring fluid, and the slip and skid stop when the fluid reaches the local speed. This scenario is commonly found for centrifuges with continuous feeding, especially in the feed zone. The skid and slip of the feed stream can generate shear and turbulence that break up delicate living cells, microorganisms, and flocculated solids. Therefore it is desirable to accelerate the feed to speed so that it can be laid gently onto the separation zone attaining full G for making separation. The acceleration or deceleration of a fluid is by means of viscous diffusion or by secondary flow. Viscous diffusion can transfer momentum from the solid wall of a rotating bowl to the adjacent fluid layer, which is rather slow and ineffective as has been demonstrated earlier. A more effective way of accelerating and decelerating a fluid is by generating secondary flow.

Principles of Centrifugal Sedimentation

39

One can define the acceleration efficiency ηa [5] to quantify the velocity of a fluid v in a container compared with that at solid-body rotation v 5 ΩR. ηa 5

v ΩR

(2.8)

In fact, ηa can change in as much as v can change in space and time. The tangential velocity and centrifugal acceleration for a fluid can be expressed, respectively, as v 5 ηa ΩR G5

(2.9)

v2 5 η2a Ω2 R R

(2.10)

Comparing Eq. (2.5b) for a solid-body rotation to the case of a fluid which may not be under solid-body rotation Eq. (2.9), it follows that the ratio of the two, which is defined as G efficiency ηG, is related to the acceleration efficiency ηa by ηG 5

G v2 =R 5 2 5 η2a Gsb ðΩ RÞ

(2.11)

In summary, the following conditions are applicable to Eqs. (2.8)(2.11) wherein ηa , 1 is for under-accelerated fluid, ηa 5 1 is for fluid establishing solid-body motion, and ηa . 1 is for over-accelerated fluid [5,6]. For example, given a fluid assuming three different possible scenarios ηa 5 70%, 100%, and 140% in a rotating bowl with R 5 0.2 m and Ω 5 500/s. Based on Eqs. (2.4)(2.6c and d), v, G, G/g, and ηG can be calculated and the results are compiled in Table 2.1. Table 2.1 Speed, G, and efficiencies for a fluid under various conditions in a rotating bowl with R 5 0.2 m, Ω 5 500/s

v (m/s) G (m/s2) G/g Acceleration efficiency ηa (%) G efficiency ηG (%)

Underspeeding Solid-body rotation

Overspeeding

70 24,500 2497 70

100 50,000 5097 100

140 98,000 9990 140

49

100

196

40

Centrifugal Separations in Biotechnology

When a fluid at radius R1 and with tangential velocity v1 5 ηa1ΩR1 is taken to a different radius R2, the new tangential velocity can be determined by conservation of angular momentum as follows: v1 R1 5 v2 R2 v2 5

(2.12)

v1 R1 R2

(2.13)


2 v2 ηa1 ΩR21 =R2 R1 ηa2 5 5 5 ηa1 ΩR2 ΩR2 R2 ηG2 5


4
4 v22 1 R1 2 R1 5 η 5 η a1 G1 R2 Ω2 R2 R2 R2

(2.14)

(2.15)

The above implies that when the fluid is brought abruptly from a small radius to a new radius that doubles the original radius, the acceleration efficiency is reduced to 25% of the original efficiency. Likewise, when the new radius is half of that of the original, this overspeeds the fluid to 400% of the original efficiency. Table 2.2 shows an example calculation of speed, G, acceleration efficiency as well as the G efficiency for both increased and reduced radii, respectively. 2.3.3

Regimes of Sedimentation

The sedimentation behavior of a suspension may be classified into four categories in accordance with the solids concentration in suspension and the degree of aggregation of solids. This is illustrated in Fig. 2.7, which Table 2.2 Speed, G, and efficiencies for a fluid moved instantly to a new radius either larger (R2) or smaller (R3) compared with initial radius R1 5 0.2 m and with Ω 5 500/s

R (m) V (m/s) Vsb (m/s) G/g Gsb/g Acceleration efficiency (%) G efficiency (%)

R1

R2

R3

0.2 100 100 5097 5097 100 100

0.3 67 150 1510 7645 45 20

0.15 133 75 12,081 3823 178 316

Principles of Centrifugal Sedimentation

41

Aggregate Zone settling Settling Degree of aggregation

Discrete particles

compaction Cake Compaction expression and Expression

Flocculation Discrete particle Particle settling Settling

Dilute

added Dispersant Added Solids concentration

Concentrated

Figure 2.7 Different regimes of sedimentation.

in essence is a modified Fitch diagram. For dilute concentration and low degree of solids aggregation, solid particles settle independent of each other and they follow the Stokes’ law of sedimentation, which was developed for spherical particles settling under the Earth’s gravity, 1 g (9.8 m/s2). As solids concentration increases, the sedimentation rate of particles is affected hydrodynamically by neighboring particles, despite there is no physical contact between them. Under this condition, the settling rate may be less, or even higher, than the Stokes’ settling velocity. For a given solids concentration, as the particles tend to agglomerate due to weak or negligible electrical repulsion they form an aggregate and settle as a large floc, which can be modeled by fractal analysis. This allows both small and large particles to settle at the same speed, also known as zone settling, without discriminating the size of individual particles. The addition of coagulant and flocculant (polymer) may further promote the formation of agglomerates and flocs leading to zone settling; while introduction of dispersant extends the discrete particle settling condition well into the concentrated solids region in which hindered and zone settling normally prevail (see Fig. 2.7). The former finds applications, such as clarification of waste slurries, while the latter finds applications, such as classification of valuable fine-particle slurries for the coating and pigment market. Dense thick slurry forms networking as particle concentration and the degree of aggregation both increase in a suspension. Under gravitational body force, the solid network or matrix compresses downward (compaction) while liquid expresses counter-currently upward (expression). This is delineated as the region to the upper right in Fig. 2.7.

42

Centrifugal Separations in Biotechnology

2.3.4

Stokes’ Law

Consider the separation velocity vso of a spherical solid with diameter d and with density ρs, settling under gravitational acceleration g in a lighter fluid phase (liquid) with density ρL and fluid viscosity μ. The settling velocity is determined under steady state by balancing all the relevant forces acting on the spherical solid. They are the viscous drag force, buoyancy force from the suspending fluid, and the body weight of the solid. The settling velocity becomes   1 ρs 2 ρL gd2 (2.16) vso 5 18 μ The subscript o refers to settling of an individual particle with no influence from its neighboring particle in an ideal dilute suspension. Under high-speed rotation, the gravitational acceleration g is replaced by the centrifugal acceleration G, given by Eq. (2.6a, b) depending on whether the fluid has attained a solid-body rotation. Thus   1 ρs 2 ρL Gd 2 (2.17) vso 5 18 μ

Viscosity (g-s/cm)

Based on Eq. (2.17), it is evident that high G, larger sized particle, viscosity reduction due to elevated temperature, all enhance separation or settling velocity. Fig. 2.8 shows the viscosity of water with increasing temperature. Between 10 C and 60 C, the viscosity drops rapidly with increasing temperature. As such, it is advantageous to operate at the highest possible operating temperature for improving separation yet without affecting or destroying the product protein and functional biologics. 0.020 0.018 0.016 0.014 0.012 0.010 0.008 0.006 0.004 0.002 0.000 0

20

40

60

80

100

Temperature (ºC)

Figure 2.8 Water viscosity as a function of process temperature.

Principles of Centrifugal Sedimentation

43

The cell density for biological cells in aqueous suspension is nominally about 1.01.1 g/mL, and the liquid density of aqueous suspension is about 1 g/mL. Thus the density difference between the two is up to 0.1 g/mL, which is indeed very small. The viscosity of liquid can be anywhere from 1 cP (water at room temperature) to over 1000 cP depending on the protein and various biologics (such as RNA) dissolved in solution. Separation can be a very difficult task with cell sizes trending from 50 μm down to 1 μm and below as separation rate varies as the quadratic power of particle size. Table 2.3 shows some common biological cell sizes. Some common cells used in protein expression are large mammalian cells, smaller yeast cells and much smaller bacteria cells. On the other hand, cell debris is much smaller and they are typically in the submicron range. They are removed by classification in the overflow after cell lysing. 2.3.5

Settling With Concentrated Solids

As solids concentration increases, the settling rate vs of a particle is reduced as it feels the effect from its neighbors; thus vs should be below that of the Stokes’ free settling velocity vso. The well-known Richardson and Zaki [7] correlation is used frequently to quantify the hindered settling effect.  n vs 5 12φf vso

(2.18)

φf is the actual solid volume concentration of solids in suspension and n is an exponent; vso is the Stokes’ settling velocity of a single particle regardless of whether it is due to gravitational acceleration via Eq. (2.16) or centrifugal acceleration via Eq. (2.17). The hindered settling behavior of Richardson and Zaki (Eq. 2.18) is plotted in Fig. 2.9 for n 5 4.65 and 3.3. As can be seen for the case of n 5 4.65, the separation velocity is reduced Table 2.3 Typical biological cell size. Cells

Size (μm)

Plant cells Mammalian cells Chinese Hamster ovaries Yeast Bacteria Cell debris

60100 1040 10 710 12 0.20.5

44

Centrifugal Separations in Biotechnology 1

vs /vso

0.8 0.6

n = 3.3

0.4 n = 4.65

0.2 0

0

10

20

30

40

Feed solids, percent by actual volume

Figure 2.9 Effect of hindered settling due to suspension solid concentration based on correlation of Richardson and Zaki for n 5 4.65 and 3.3, respectively.

by 40% as the solids volume increases to 10%. The 40% solids on an actual volume basis could have corresponded very well to 60%90% bulk volume (including both solids and liquid volume) depending on packing density. Solids concentration by bulk volume is a more readily available index as it is measured readily in practice using a rotating test tube centrifuge (spintube) spinning for an arbitrary prefixed time duration. For the case with n 5 3.3, the reduction in settling velocity due to hindered settling as compared with n 5 4.65 is less for a given solids concentration. It should be noted that, depending on the solids and free ions in the aqueous phase, the solids may not act as individual discrete particles but may act as a blanket of solids settling all at one rate and speed independent of particle size once solids concentration increase beyond a certain level in which the particles start to form a network (i.e., top right in Fig. 2.7).

2.4

Process Functions

There are several process functions using centrifuges in biotech separation. These are listed below. 1. Separation (solid/liquid, solid/liquid/liquid and solid/solid/liquid separation) Centrifuge can be used for solidliquid separation provided that the solids are heavier than the liquid. Centrifuge can also be used to separate a heavy phase, and two lighter liquid phases, with one of the lighter phases being lighter than the other. As discussed, solids can be lighter than liquid and separation is by flotation of the dispersed solid phase.

Principles of Centrifugal Sedimentation

45

2. Clarification—minimal solids in liquid product Centrifuge can be used to clarify the discharged separated lighter liquid phase. The objective is to minimize the discrete suspended solids in the light continuous phase. Usually, only fine submicron biosolids are left uncaptured by centrifugation and they escape with the light phase. 3. Classification—sort by size and density Centrifuge is used to classify solids of different sizes. One of the several possible applications is to classify crystals of different size range, with finer submicron sizes leaving with the light phase and retaining only the larger sizes in the separated heavy phase. Either of the separated solids can be product. For example, the larger crystals can be the product crystals while the finer crystals are returned to the crystallizer for which the undersized crystals can further grow to larger crystals. Another similar application is to classify smaller size cell debris in the light liquid phase from the heavier products after homogenizing cells. Classification is analyzed in detail in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification. 4. Degritting—remove oversized and foreign particles Degritting is similar to classification where unwanted particles, larger or denser, are rejected in the sediment, with product (smaller or less dense) overflowing in the lighter liquid phase. Another situation is where smaller unwanted particles are rejected in the light liquid phase, and valuable heavier solids are settled with the heavier phase. 5. Thickening or concentration—remove liquid and concentrate solids Centrifuge is frequently used to concentrate the solid phase by sedimentation and compaction, removing the excess liquid phase in the overflow or centrate. This reduces the volume of the product in downstream processing. This is discussed in Chapter 7, Concentrating Solids by Centrifugation. 6. Separation and repulping—remove impurities by washing or diluting With a concentrated suspension containing contaminants, such as salts and ions, it is diluted and washed so that the contaminants are dissolved in the wash liquid. Subsequently, the suspension is sent for centrifugation to remove the spent wash liquid with dissolved contaminants or finely suspended solids. Subsequently, the product can be further concentrated by centrifugation. The aforementioned processes can be combined to achieve several objectives concurrently or in series.

46

2.5

Centrifugal Separations in Biotechnology

Summary

In this chapter, several nonintuitive phenomena, such as nonlinear increase in pressure with radial distance and Coriolis acceleration and associated force, have been discussed. These phenomena dictate the rotating flow that affects separation in a centrifuge. Centrifuge has been designed accordingly to reduce the adverse effect from Coriolis affecting sedimentation and to take advantage of the flow filling the feed to the rotation axis, thereby reducing airfeed contact and undesirable oxidation. Other intuitive phenomena, such as discrete particle settling versus aggregate of particles settling, are discussed. These cover the entire spectrum of real-life possibilities. In particular, this chapter presents Stokes’ free settling of single particles and hindered settling of a cluster of discrete particles at high-solids concentration. Finally, various process functions of centrifugation, covering separation, clarification, classification, degritting, thickening, and impurities removal or valuable recovery by separation and repulping are discussed.

References [1] H. Greenspan, The Theory of Rotating Fluids, Cambridge University Press, London, 1968. [2] H. Greenspan, Centrifugal separation of a mixture, J. Fluid Mech. 127 (1983) 91101. [3] R.B. Leslie, P.J. Carillo, T.Y. Chung, S.G. Gilbert, K. Hayakawa, S. Marousis, et al., Water diffusivity in S-based systems, in: H. Levin, L. Slade (Eds.), Water Relationships in Food, Plenum Press, New York, 1991, pp. 365390. [4] A. Calzetta Resio, R.J. Aguerre, C. Suarez, The drying of amaranth grain: mathematical modeling and simulation, Braz. J. Chem. Eng. 22 (2) (2005) 303309. [5] W.W.F. Leung, Industrial CentrifugationTechnology, McGraw-Hill, New York, 1998. [6] W.W.F. Leung, A. Shapiro, An accelerating vane apparatus for improved clarification and classification in decanter centrifuges, Trans. Filtration Soc. 1 (3) (2001). [7] J.F. Richardson, W.N. Zaki, The sedimentation of a suspension of uniform spheres under conditions of viscous flow, Chem. Eng. Sci. 3 (1954) 6573.

Principles of Centrifugal Sedimentation

47

Problems (2.1) For a bioseparation process by sedimenting under the Earth’s gravity of 9.81 m/s2, a suspension has biological cells with density of 1100 kg/m3 in a liquid with density of 1000 kg/m3 and viscosity of 0.001 Pa s, assuming that particles have equivalent spherical diameter of 10 μm. How long does it take for the particle to settle in a graduated cylinder of height 10 cm? (2.2) If a small dose of coagulant can neutralize the charges so that the particles in Problem (2.1) agglomerate in size to 33 μm with all other conditions the same, how long does the agglomerated particle settle for 10 cm? How would you compare this result with that of Problem (2.1)? (2.3) Repeat the same problem as in (2.1) but with a particle of 1 μm simulating a red blood cell, how long does the particle take to settle to the bottom of a 10-cm high graduated cylinder? (2.4) For a bioseparation process by sedimenting under centrifugal gravity 1000 times that of the Earth’s gravity, a suspension has biological cells with density of 1100 kg/m3 in a liquid with density of 1000 kg/m3 and viscosity of 0.001 Pa s, assuming particles have equivalent spherical diameter of respectively (1) 0.33 μm and (2) 1 μm. How long does it take for each of the particles to settle in a long spintube with a height of 10 cm? (2.5) For the same conditions as in Problem (2.4) with the suspension viscosity increased to 1 Pa s due to presence of soluble matter and contaminants, how long does (1) 0.33 μm particle and (2) 1 μm particle take to settle in a 10-cm distance under RCF (G/g) of 10,000? Can centrifuge still make separation? Why or why not? (2.6) What can one conclude from Problems (2.1)(2.5) on effect of particle size, G/g, and viscosity? (2.7) A feed stream has accelerated to tangential speed of 30 m/s at a radius of 0.1 m. If the feed is allowed to abruptly move to a larger radius of 0.12 m, what is the tangential speed of that feed stream at the new radial location? (2.8) For Problem (2.7), if a pool is already at solid-body rotation, what would be the tangential speed of the pool at 0.12 m radius if the angular rotation of the centrifuge is 300/s? Is there any difference between the feed speed and that of the pool which is already at solid-body rotation?

48

Centrifugal Separations in Biotechnology

(2.9) What is the centrifugal acceleration of the feed expressed as a ratio of g? What is the supposedly G/g (i.e., RCF) for the solidbody pool at 0.12 m radius rotating at angular speed 300/s? (2.10) Express the (1) acceleration efficiency of the feed and (2) the G-efficiency of the feed mathematically in terms of appropriate variables or parameters. What is the relationship between these two efficiencies? (2.11) A water column has an effective rotation speed of only 1000 rpm in a rotating distributor column in a disk stack centrifuge rotating at 7000 rpm. The column has an inner diameter of 1.5 cm. (1) What is the rise in water at the inner wall of the distributor compared to the water level at the rotation axis? (2) Everything being the same, the water is replaced with a molasses suspension with density 5000 kg/m3, what is the rise of the molasses solution at the inner wall of the distributor compared with the molasses level at the rotation axis? (2.12) If the distributor has a wall (along the rotation axis) only 5 cm tall compared with the water height at the axis of rotation, what is the rotation speed in rpm at which the water barely spilled over the wall of the rotating distributor?

3 Batch and Semibatch Centrifuges In this chapter, spintube, centrifugal filter, ultracentrifuge, and tubular centrifuge are discussed. Spintube, ultracentrifuge, and centrifugal filter are bench centrifuges wherein they are batch fed and involve manual removal of sediment. For all three types of centrifuges, there are specific steps/cycles during operation for feeding, rotor acceleration (or acceleration followed by feeding for ultracentrifuge), separation (for spintube and ultracentrifuge) or filtration (for centrifugal filter), rotor deceleration, clarified liquid decanting, sediment removal, and cleaning. Centrifugal filter has various insert modules that facilitate various functions to be performed, including sedimentation, filtration, and purification. Tubular centrifuge operates on a semibatch basis with continuous feed until sediment fills the bowl and the discharged clarified liquid turns turbid, at which point sediment has to be removed. Subsequently, it goes through a series of cycles on pool drainage, deceleration, sediment discharge, and cleaning. Some tubular centrifuge designs require manual removal of sediment, while other designs have built-in automatic sediment removal.

3.1

Spintube

Measurements can be made using spintube under different G’s and t’s to determine the separation characteristics of a given sample. If there are sufficient solids in the feed that forms a decent size of sediment or pellet, a physical method can be used to determine qualitatively the integrity from which the handleability/flowability of the sediment can be inferred. For example, a glass rod can be inserted in the tube [1] at the end of the test to determine the amount of penetration that the rod makes in the sediment. It is a measurement of the yield stress of the sediment. Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00003-9 © 2020 Elsevier Ltd. All rights reserved.

49

50

Centrifugal Separations in Biotechnology

Such a physical evaluation method is largely subjective. Also, chemicals can be added to neutralize the charges and agglomerate the fine particulates in suspension. Spintube centrifuges are typically divided into two different configurations. The first type is a swinging bucket, or horizontal head, wherein the tube support rests at the small diameter on a bracket (or an adaptor) that allows the tube to swing out with the tube axis being perpendicular to the axis of rotation of the centrifuge in the full-swinging position, as illustrated in Fig. 3.1A (right). Numerous combinations of tubes are possible for this configuration. The centrifuge can accommodate a minimum of a pair of spintubes or multiple tubes secured in a cup or a holder. Table 3.1 presents two of the many configurations that are available commercially. One case shows two 50-mL tubes diametrically opposite, and the other case shows as many as four cups spaced 90

Supernatent

θ

Angle-head

Sediment

Horizontal head

Figure 3.1 A Angle-head (left) and horizontal-head (right) test tube centrifuge. Table 3.1 Some examples of horizontal and angled head test tube centrifuges Type

Number

Sample volume

Horizontal head

2 16 Up to 296 microtubes 2 6 60 4 48

10 50 mL 0.25 0.4 mL tube 600 mL bag 250 mL bag 15 mL bag 50 mL tube 0.5 mL tube

Angle head (θ 5 40 2 48 degrees)

Batch and Semibatch Centrifuges

51

degrees apart with each cup holding four 10-mL tubes. For achieving flexibility in testing, the cups and the support for multiple tubes can be interchanged with other bracket geometry that holds a small number of large-volume tubes or large number of small-volume tubes. As an example of the latter, one can run 60 3 5/7-mL, 30 3 15-mL, and 16 3 50-mL cell culture tubes or microplates without taking up additional space. Much smaller (0.25 0.4 mL) and large numbers (up to 296) of microtubes are also available. This can facilitate high-throughput screening. As a demonstration on different possible combinations, Fig. 3.1B (left) shows an adaptor for four swinging buckets, with a one pair of diametrically opposing buckets each carrying four large tubes and the other pair of diametrically opposing buckets each carrying 19 smaller tubes. The adaptor can also be used with microplates and microwells in a plate carrier (see Fig. 3.1B, left). Microplates are small plastic reaction vessels. By design, they are trays or cassettes that are covered with wells or dimples arranged in orderly rows. These wells are used to conduct separate chemical reactions. The large number of wells, which typically amounts to 96 or 384,

Microwells/ microplates

Tube

Bucket

Plate carrier

Adaptor

Figure 3.1B Left diagram: Adaptor with swinging buckets with two diametrically opposing pairs with one pair each having 4 spintubes and the other pair each having 19 spintubes, respectively. Right diagram: Plate carrier carrying microwells and microplates. The covers for the buckets and plate carrier are not shown.

52

Centrifugal Separations in Biotechnology

depending on the size of the microplate, allow for many different reactions to take place simultaneously. This can be useful if the goal is to determine a statistical basis for research results, to test for aberrations in an expected result, or to run large number of unrelated reactions at the same time. Microplates are ideal for high-throughput screening and research. They allow miniaturization of assays and are suitable for many applications, including drug testing, genetic study, and combinatorial chemistry. So far, we have discussed that centrifugation is being used to separate product from reactants based on density difference after chemical reaction, and in some cases, it is used to enhance chemical reaction, process intensification, such as the use of microplates and microwells. Angled head geometry or fixed-angle rotor can take higher centrifugal gravity G as the tubes are in a “solid bowl” rotor with cutouts that serve as tube holders or cups. The inclination between the tube center axes and the vertical is typically between 40 and 48 degrees. This is depicted in Fig. 3.1A (left). Note that particles always settle radially along the direction of centrifugal gravity G, which is perpendicular to the rotation axis. The trajectory of these particles follows the G-field (90 degrees from the vertical axis) and intercepts the wall of the angled tube (40 2 48 degrees from the vertical axis), similar to an inclined plate settler [2] or a disk-stack centrifuge that is discussed in Chapter 4. This reduces the sedimentation path and settling time of particles and enhances particle capture. Various different sizes and number of tubes are available for the angle-head centrifuges, some of which are listed in Table 3.1. For example, four 50-mL tubes can be used in one rotor, or as many as 48 0.5-mL tubes can be used in another rotor. Most high-end bench centrifuge manufacturers provide precise microprocessor-based controls that ensure reproducible test runs in conformity with good laboratory practice standards. To adjust protocols for spinning fragile samples, several acceleration and deceleration profiles are available with some designs. Some designs also provide refrigeration and cooling, maintaining a steady temperature from 210 C to 40 C, even at the maximum speed. This is especially for protein solution that can be denatured if it is being heated up by more than, say, 20 C during the course of high-speed centrifugation. For safety assurance, some designs have a double-lid locking system, armored chamber, and other features to protect the operator. In case of excessive vibration, the rotor stops in seconds. Audible and visual messages alert operators to any anomalies. Some commercial vendors on spintube centrifuges are Beckman and Coulter, Eppendorf, Hermle, Hettich, Kendro, Thermo/Forma/Scientific, and IEC.

Batch and Semibatch Centrifuges

3.2

53

Centrifugal Filter

A centrifugal filter with a membrane filter or a chromatography column can be inserted in a spintube for carrying out a variety of separation and purification functions on biosolids. The centrifugal filter can be a microspintube 15 20 mL with an insertion of a membrane module for separation or a chromatography column (also referred to as membrane absorber) for purification, as shown in Fig. 3.2. The centrifugal filter is inserted in a 50-mL spintube. Typically, the combo configuration can be used where centrifuges can attain 12,000g. Smaller size microspintubes and containing tubes for addressing smaller samples are also available. The following is an example of both separation and purification of RNA using a centrifugal filter. The objective is to separate RNA from a suspension containing a mixture of RNA, DNA, salts, suspended particles, and other impurities. The centrifugal filter is first inserted with a microfiltration membrane, after which the mixture is added. Under centrifugation, micron-sized suspended contaminants and particles are removed by the microfiltration membrane. The microfilter is removed from the spintube, and the filtrate is decanted to a separate container. After cleaning, another microfilter with a chromatography column is inserted in the spintube. The liquid sample, now free from micron-size suspended solids, is first chemically conditioned before pouring in the chromatography column. The liquid mixture runs through the column under G-force that further enhances the rate of the drainage process. Concurrent with liquid draining through the chromatography column, Cap Microcentrifuge tube, 15–20 mL feed, buffer liquid/water, or wash liquid Container Filter or silica membrane

Filtrate, spent wash liquid, product elutate

Figure 3.2 Microcentrifuge tube of 15 20 mL, housed in a larger containing tube, equipped with filter or membrane column.

54

Centrifugal Separations in Biotechnology

RNA is preferentially adhered to the column. Next, the column is washed under centrifugation with desalting liquid to remove salts and contaminants. Furthermore, chemical/buffer liquid is added to the column subject to centrifugation to facilitate the removal of DNA. Upon completion of removing DNA, purified RNA is released under centrifugal field from the column by elution using a conditioning agent. This example illustrates that separation and purification processes can be readily tailor made for the specific process. Each step gets benefits from the enhanced centrifugal acceleration despite the liquid is considered viscous and normally it takes a long time to flow through a column were this drainage process being carried out under the Earth’s gravitational acceleration. In general, sedimentation can be the first step followed by filtration and purification using the chromatography column. The centrifugal filter provides a comprehensive (all-in-one) package with interchangeable modules for sedimentation, filtration, and purification under the enhanced G-field.

3.3

Ultracentrifuges

Ultracentrifuge is also setup in form of an angled spintube, with a titanium rotor that provides mechanical integrity, that is, for high shear and yield strengths. It can reach 500,000 1,000,000g for separating very small particles, particles and liquid with a small density difference, and/or separation in a viscous liquid phase. A schematic of an ultracentrifuge is shown in Fig. 3.3. There are two kinds of ultracentrifuges: the preparative and the analytical. Both have important uses in molecular biology, biochemistry, and polymer science. Theodor Svedberg invented the analytical ultracentrifuge in 1923 and won the Nobel Prize in Chemistry in 1926 for his research on colloids and proteins using the ultracentrifuge.

Supernatant G Sediment

Titanium rotor

Figure 3.3 Ultracentrifuge schematic.

Batch and Semibatch Centrifuges

3.3.1

55

Analytical Ultracentrifuge

In an analytical ultracentrifuge, a sample being spun can be monitored in real time through an optical detection system, using ultraviolet light absorption and/or interference optical refractive index sensitive system. This allows the operator to observe the evolution of the sample concentration versus the axis of rotation profile as a result of the applied centrifugal field. With modern instrumentation, these observations are electronically digitized and stored for further mathematical analysis. Two kinds of experiments are commonly performed in these instruments: sedimentation velocity experiments and sedimentation equilibrium experiments. Sedimentation velocity experiments aim to interpret the entire transient sedimentation process and report on the shape and molar mass of the dissolved macromolecules, as well as their particle size distribution. The size resolution of this method scales approximately with the square of the particle radii, and by adjusting the rotor speed of the experiment, size ranging from 100 Da to 10 GDa can be covered. Sedimentation velocity experiments can also be used to study reversible chemical equilibrium between macromolecular species, by monitoring the number and molar mass of macromolecular complexes, by gaining information about the complex composition from multisignal analysis exploiting differences in each component’s spectroscopic signal, or by following the composition dependence of the sedimentation rates of the macromolecular system. Sedimentation equilibrium experiments are concerned only with the final steady state of the experiment, where sedimentation is balanced by diffusion opposing the concentration gradients, resulting in a timeindependent concentration profile. Sedimentation equilibrium distributions in the centrifugal field are characterized by Boltzmann distributions. This experiment is insensitive to the shape of the macromolecule and directly reports on the molar mass of the macromolecules and, for chemically reacting mixtures, on chemical equilibrium constants. Different kinds of information that can be obtained from an analytical ultracentrifuge include the gross shape of macromolecules, the conformational changes in macromolecules, and size distributions of macromolecular samples. For macromolecules, such as proteins that exist in chemical equilibrium with different noncovalent complexes, the number and subunit stoichiometry of the complexes and equilibrium constant constants can be studied.

56

Centrifugal Separations in Biotechnology

With interference and schlieren optics, the change in concentration can be measured as a dependence on time by the following equation:   1 2 ρ =ρ L p S 5 M (3.1) D RT Eq. (3.1) is known as the Svedberg equation. S is the sedimentation coefficient.   2 ρ 2 ρ p L d vso 5 (3.2) S5 g 18μ S is a ratio of the Stokes’ settling velocity vso to the Earth’s gravitational acceleration g, which is represented in seconds. R is the universal gas constant; T is the temperature in Kelvin; ρp and ρL are the density of the particle and liquid, respectively; and D is the diffusion coefficient. Molecular measurements typically give S in units of 10213 s and is in honor of Svedberg. For reference, 1 Svedberg is defined as 10213 s. From measurements, molecular weight M, sedimentation coefficient S, diffusion coefficient D, and partial specific volume, 1/ρp can be derived. 3.3.2

Preparative Ultracentrifuge

Preparative ultracentrifuges are available with a wide variety of rotors suitable for a large range of experiments. Most rotors are designed to hold tubes that contain samples. Swinging bucket rotors or horizontal heads allow the tubes to hang onto hinges, so the tubes reorient to the horizontal position as the rotor initially accelerates. Fixed-angle rotors are made of a single block of metal and hold the tubes in cavities bored at a predetermined angle. Zonal rotors are designed to contain a large volume of samples in a single central cavity rather than in tubes. Some zonal rotors are capable of dynamic loading and unloading of samples while the rotor is spinning at high speed. Fig. 3.4A shows a feed suspension is being introduced to the zonal centrifuge operating at, say, 2000 rpm. This is after a sucrose solution with 10%, 20%, 30%, and 40% with respective densities 1.0381, 1.0810, 1.127, and 1.1764 g/cm3 is first introduced to the zonal centrifuge. Fig. 3.4B shows that after separation, unloading can be effect by introducing water at the center to displace the product to the periphery of the centrifuge. Fig. 3.4C shows just the opposite, where water is introduced at the periphery displacing product toward the center of the centrifuge.

Feed

Figure 3.4A Feed is introduced to the zonal centrifuge after loading gradient solution, such as sucrose or cesium chloride solution.

Water Product

Figure 3.4B Product is unloaded by injecting water/buffer liquid at the center, while product is removed at the periphery.

Product Water

Figure 3.4C Product is unloaded by injecting water/buffer liquid at the periphery, while product is removed at the center.

58

Centrifugal Separations in Biotechnology

Preparative rotors are used in biology for pelleting of fine particulate fractions, such as cellular organelles (mitochondria, microsomes, and ribosomes) and viruses. They can also be used for gradient separations, in which the tubes are filled from top to bottom with an increasing concentration of a dense substance in solution. Sucrose gradients are typically used for separation of cellular organelles. Gradients of cesium salts with much wider density ranges are used for separation of nucleic acids. After the sample has spun at high speed for sufficient time to produce the separation, the rotor is allowed to coast to a smooth stop and the gradient is gently pumped out of each tube to isolate the separated components. This is also referred to as isopycnic separation. The sample is handled gently and carefully so as to avoid disturbing the layers during intake and outtake. 3.3.3

Centrifugal Elutriation

The zonal centrifuge can be operated so that it can classify cells based on the sedimentation velocity. Assuming the cells are of the same density and shape, then classification is by the size only. A special zonal rotor is operated such that fluid is introduced from the periphery of the rotor so that the fluid velocity just balances the particle settling velocity in the separation zone, as illustrated in Fig. 3.4D showing a sectional view in r-z plane with fixed angle in a cylindrical coordinate system. Here, r is the radial distance from the axis, z is the distance along the rotation axis, and angle is along the angular rotation. If there are R, G

v > vs vs = v v < vs Eluent fluid

v=

A

Smaller particles carried out of separation zone by countercurrent flow

Q A

Larger particles settle out of separation zone Medium sized particles trapped in separation zone

Figure 3.4D Centrifugal elutriation with undersized particles removed at small radius and oversized particles at large radius retaining only the desired size particles.

Batch and Semibatch Centrifuges

59

particles of smaller size, they are carried by the fluid leaving the separation zone to exit at the small radius, while particles larger than the ones in the separation zone settle to the large radius and get collected at the exit. This classification process is discussed in detail in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification. For a mixture of cells, separation by classification is possible if there are specific target cells in which modification can be made to their size, such as with the addition of a reagent [3], and these target cells swell in size beyond that of normal cells. The mixture with the swollen target cells is introduced to the rotor, and the eluent fluid is fed at the periphery at a rate so that the radial inflow velocity in the separation zone just balances the settling velocity of regular cells, and they are stationary in position in the centrifuge. Because the swollen target cells are larger in size, they settle out of the separation zone, while cells of finer sizes are carried out of the separation zone by the eluent fluid to exit at the small radius. This is shown in Fig. 3.4D. Alternatively, reagents can be added to get the unwanted cells to swell and settle out of the separation zone, while the target cells remain in the separation zone as shown in Fig. 3.4D. After removal of cell debris and the unwanted cells, the target cells can be washed with other buffer liquid to elute any contaminants attached to the target cells. A third possibility is illustrated in Fig. 3.4E. The rate of eluent can be increased so that swollen cells can be held stationary in the separation zone, while cell debris and normal cells are eluted. After the cell debris R, G

v > vs vs = v

v=

A

All smaller particles (regular and debris) carried out of separation zone by countercurrent flow

Eluent and wash fluid

Q A

Targeted larger particles trapped in separation zone

Figure 3.4E Increasing eluent flow rate to remove undersized particles. Wash liquid can be added to clean the targeted cells.

60

Centrifugal Separations in Biotechnology

and normal cells are removed, wash liquid can be added to further wash the swollen cells (products). A fourth possibility is to increase the G-force so that larger target particles (products) settle quickly toward the large radius, leaving the separation zone that may have the normal cells, while the finer submicron debris leave with the carrier fluid exiting at the small diameter. This is somewhat similar to the second alternative (Fig. 3.4D). In summary, there are various possibilities in the cell classification process: (1) the use of a reagent to modify the cell size (target cells or unwanted cells), (2) different values of G-force to control the particle sedimentation velocity and thus the particle size to be retained in the separation zone, and (3) the eluent flow rate/velocity for use in elutriation.

3.4 3.4.1

Tubular Centrifuge General Tubular Bowl Geometry

Tubular bowl centrifuge or tubular centrifuge is typically vertically oriented. It has a rotating tubular bowl with a length L and a bowl diameter D. The aspect ratio L/D is large anywhere from 5 to 7.3 with the traditional tubular design, a listing of which is presented in Table 3.2. Feed suspension is brought in the centrifuge either from the top (top feed) or from the bottom (bottom feed). A schematic diagram of the bottom-feed tubular is shown in Fig. 3.5A with a more detailed layout shown in Fig. 3.5B. A thin annular pool is maintained between the bowl wall and the inner air core when the centrifuge is operating at full speed. The tubular centrifuge is top mounted (or top suspended) and top driven. When the rotor bowl is at operating speed, the center of gravity of the unit is below the mounting support to maintain dynamic stability with minimal whirling. Table 3.2 Large L/D tubular centrifuge D (mm) Length (mm)

L/D RPM

44 105 127

5 50,000 62,400 7.25 15,000 13,200 6 15,000 15,900

229 762 762

G/g (RCF)

NA, Not Applicable.  Turbine drive based on steam or compressed air.  Motor drive.

Volume Motor rate (L/m) (kW) 0.2 0.4 0.4 12 0.8 25

NA 1.5 2.2

D Overflow

Dp Dh Annular pool

L Particle trajectories

Hub

Feed

Figure 3.5A Schematic of a tubular centrifuge.

Heavy phase out

Light phase out

Feed in

Figure 3.5B Layout of a traditional tubular centrifuge typically with a high aspect ratio (L/D .. 1).

62

Centrifugal Separations in Biotechnology

Industrial-scale tubular centrifuges have bowls of 102 127 mm in diameter and 762 mm long. It is capable of delivering 18,000 20,000g. The smallest tubular, 44 mm diameter by 229 mm long, is a laboratory model capable of developing up to 65,000g (see Table 3.2). It is also used for separating difficult biological solids, cells, and viruses. The bowl is suspended from an upper bearing and drive (electric or turbine motor) assembly through a flexible-drive spindle with a loose guide in a controlled damping assembly at the bottom. The unit finds its axis of rotation if it becomes slightly unbalanced due to process load. In some designs, feed is accelerated to a solid-body rotation by a set of vanes or channels that start at the axis of rotation and terminate at a larger radius below the pool surface. This assures that the feed suspension is accelerated [3] to angular speed before introducing to the annular pool for separation. In other designs, the feed is sent to a set of parallel plates (in radial and azimuthal plane) where feed is accelerated from a small to a large radius inside the radial annular geometry formed between the plates. This type of acceleration is less effective due to viscosity of liquid not being an effective mechanism for momentum transfer, unless the viscosity is very large. Exceptions to the above are (1) when the viscosity of the suspension is very high [4] and the plates are closely spaced on the order of millimeters and (2) where the plates are equipped with accelerating vanes and a smoothener at the large diameter to smoothen the discrete feed streams to a continuous sheet of accelerated feed liquid laying onto the pool [5]. Feed suspension containing solids is introduced from one end of the bowl and clarified liquid exits at the opposite end of the bowl. After feed suspension enters the annular pool, solids of various sizes, shapes, and densities settle while they are also transported along the bowl by the main flow. Recent studies [6] have shown that the flow takes place in a very thin boundary layer, or moving layer, along the surface of the liquid pool regardless of the geometry of the bowl. Particles that have not settled in the moving layer are carried out in the overflow. The effluent discharge takes the form of flow over a weir at fixed discharge diameter under atmospheric pressure. Effluent can also be discharged under pressure through paring discs or centripetal pumps (to be discussed in Chapter 4: Disk Centrifuge). The pressured liquid subsequently flows downstream for processing without further need of pumping. This arrangement also reduces foaming by destroying the kinetic energy of the high velocity liquid centrate impinging onto the stationary housing when centrate liquid is typically discharged via overflow weirs. Not only the kinetic energy of the fluid stream is wasted, it creates additional

Batch and Semibatch Centrifuges

63

liquid air interface in the form of foam, especially when the liquid contains protein. The suspended solids in the centrate of the centrifuge are closely monitored by sampling periodically, or continuously, through turbidity measurement. When the growing sediment in the bowl gets close to the surface of the annular pool, it interferes with the fast-moving liquid. Sediment can be entrained by the liquid leading to resuspension and high turbidity if the resuspended sediment does not settle back in due course. The closer the sediment to the surface of the pool, the more likely the entrainment is and the less time for the resuspended solids to resettle leading to high turbidity. Once a preset limit on turbidity has been reached, the solids need to be removed from the bowl. The rotor goes through coast down, the pool liquid is drained, and the sediment is removed by plow, plunger, and other automatic mechanisms or manually when the bowl comes to a complete stop. Subsequently, the bowl is flushed during the clean-in-place (CIP) cycle with an appropriate cleaning liquid to remove all residual solids. The bowl may be momentarily rotated and stopped to allow wash liquid sloshing inside the bowl to provide an effective rinsing. Some centrifuges also include sterilization in place (SIP) to further kill any biological materials to avoid cross contamination between batches of products. Some special tubular centrifuges have been developed that are hybrid versions between the traditional tubular with the high L/D ratio and solid-basket centrifuge with small L/DBO (1). These hybrid designs (still referred herein as tubular centrifuge) have the L/D ratio in the range of 2.5 . L/D . 1. They take advantage of the solid-bowl basket design that has more bowl volume facilitating temporary storage of cake solids. There are two possible designs on cake discharge. 3.4.2

Ribs and Solids Scraper

One tubular design is the rib design, as shown in Fig. 3.6A. A photograph of the tubular centrifuge with ribs and solids scraper is shown in Fig. 3.6B. This machine can automatically discharge solids when the bowl is filled, eliminating the laborious manual cake removal process as in the case of the conventional tubular centrifuge. The bowl is equipped with a set of annular rings spaced axially and uniformly along the bowl with the ring outer diameter reaching the bowl diameter and the ring inner diameter slightly below the surface of the annular pool. The functions of the set of rings are twofold: to maintain rigidity of the bowl at a rotational speed equivalent to 20,000g and to reduce, if not eliminate,

64

Centrifugal Separations in Biotechnology

Figure 3.6A High-speed centrifuge up to 20,000g with fully automatic cycle, L/D , 1.5. Reproduced with permission from Celeros Separations, LLC.

traveling waves along the axial direction. The liquid suspension communicates between compartments through cutouts, or orifices, adjacent to the inner diameter of these rings. These cutouts are staggered (not in alignment between adjacent rings) to stop traveling waves propagating axially along the bowl. The bowl is first filled by feed suspension after the feed has been accelerated by contacting with a rotating accelerating cone, which directs the feed to the larger bowl diameter, see Fig. 3.7A. Once the

Batch and Semibatch Centrifuges

65

Figure 3.6B A 457-mm diameter high-speed 20,000g centrifuge with suspension SG , 1.5 with fully automatic cycle. Reproduced with permission from Celeros Separations, LLC.

feed suspension fills to the cutouts, it overflows at the cutouts and is uniformly spread into other compartments formed between the annular rings until a constant suspension pool depth is reached among all compartments. Subsequently, the liquid further fills up the bowl until it spills at a small radius end of the bowl opposite to that of the feed accelerator cone. During continuous feeding, the heavier solids settle toward the bowl forming a sediment layer, which increases over time. For a 457mm bowl with a maximum feed rate of 28 L/min, the bowl volume is 36 L, while the solids holding space is 32 L, which is 89% of the bowl volume. There is quite a large percentage of bowl volume to inventory

66

Centrifugal Separations in Biotechnology

Figure 3.7 (A) Feeding, centrate liquid diverted to centrate discharge. (B) Feed stops and pool liquid drains to a lower compartment. (C) Solids discharge to cake chute of high-speed tubular centrifuge. Reproduced by permission of Celeros Separations, LLC.

solids as liquid flows only in a thin moving layer near the pool surface and does not require a lot of space until the sediment surface gets close to the pool surface, at which solid entrainment can take place. This contradicts with disk-stack centrifuge in which, as much as, 50% of the bowl volume is allocated for solid storage (as will be discussed in Chapter 4: Disk Centrifuge). Feeding stops when the centrate turbidity increases as a result of entrainment of the growing sediment by the moving layer. The centrate compartment gate is closed, and the annular pool is allowed to drain to the lower compartment of the machine, see Fig. 3.7B. Upon completing draining of the liquid pool, an eccentrically located scraper or plow, in the form of a comb shape, rotates into position and plows the sediment deposited in the compartments of the bowl, diverting the sediment to a cake chute located at the lower part of the machine, as illustrated in Fig. 3.7C. The residual sediments are removed during the cleaning cycle and CIP. The equipment is equipped with CIP spray balls and CIP nozzles directing water wash, heated caustic and acid wash, water rinse, and water for injection final rinse to all product contact areas. Several sizes are available including bowl diameter of 150, 300, and 457 mm. A specification for two sizes of this type of centrifuge is given in Table 3.3. The bowl diameter of the latter two is larger than the

Batch and Semibatch Centrifuges

67

Table 3.3 Hybrid tubular centrifuge with 1 , (L/D) , 1.5 Hybrid tubular

D (mm) RPM

G/g Volume Bowl (RCF) rate (L/m) volume (L)

Knife discharge

150 457 150 300

20,000 20,000 10,000 5000

Cake reslurry, submerged hub design

15,320 8846 10,832 5417

0.1 1a 0.1 28 0.1 4 1 40

1.1 (1) 36 (32a) 1.3 16.7

(#), solids holding space. a 3-min discharge typical.

traditional large L/D tubular centrifuges with the maximum diameter of 127 mm (see Table 3.2). The centrifugal force is adjustable between 500 and 20,000g with feed rates ranging from 0.1 to 28 L/min and solids capacity from 1 to 36 kg/cycle. 3.4.3

Automatic Plunger Cake Discharge

Another automatic piston-discharge tubular centrifuge allows automatic cake discharge with lower residual cake left in the bowl. The maximum G-force is 20,000g. Depending on the centrifuge size, it can serve 10 10,000 L reactors. The centrifuge operates in a cycle with five consecutive sequences as depicted in Fig. 3.8A E. In the first sequence, when the machine is rotating up to full speed, feed suspension is introduced into the bowl interior pushing the initial piston position at the bottom of the machine upward until the piston is at the uppermost position (see Fig. 3.8A). Subsequently, the piston is secured, and the centrate ports are opened. In the second sequence, the centrifuge is continued to be fed through a center feed pipe. The feed liquid is distributed and accelerated to a larger radius into the annular space of the bowl at the bottom of the machine (see Fig. 3.8B). The feed travels upward along the annulus bound by the inner diameter of the bowl and the outer diameter of the center hub. There is no air liquid interface as the feed fills the entire annular space between the bowl and the hub eliminating oxidation and foaming. (The hub is submerged.) As the feed travels along the bowl, heavier solids settle to the bowl wall accumulating as sediment, while clarified liquid flows adjacent to the hub wall in a thin moving layer. As the centrate liquid reaches the end of the bowl, it is

68

Centrifugal Separations in Biotechnology

Figure 3.8 (A) Feed filling up bowl. (B) Separation with continuous feed introduced into bowl and centrate being removed at the top of the centrifuge and cake accumulating in the bowl. (C) Feed suspension drained from bowl. (D) Plunger moving downward scraping the entire bowl volume removing solids. (E) Plunger at most downward position removing all solids in the bowl. Reproduced by permission of Celeros Separations, LLC.

directed toward the center exiting in an annular passage at the top part of the bowl assembly (see Fig. 3.8B). The third sequence starts when the sediment layer becomes too thick at which the centrate liquid starts to entrain the sediment. At this point, feeding is stopped and draining of the pool starts. In draining, the bowl comes to an abrupt stop. The liquid in the pool drains in reverse to the feed direction (see Fig. 3.8C). In the fourth sequence, upon the liquid pool being completely drained, an air-activated plunger pushes the sediment toward the bowl bottom through the opened solids discharge valve to a solid collector equipped with a liner (see Fig. 3.8D). In the fifth sequence, upon the piston reaching the bottom of the bowl, the sediment should have been fully discharged from the bowl (see Fig. 3.8E). CIP and SIP can be initiated to clean the centrifuge getting ready for the next cycle. Given almost all solid sediments are completely removed by the plunger, the CIP and SIP can be very effective to clean the bowl interior, which is wide open, unlike the baffle design in which some residual solid can be left on the baffles and the bowl wall. For the automatic plunger discharge tubular centrifuge, the small bowl size can start with 1- and 5-L bowl volume for laboratory to pilot scale with a nominal feed rate of over 1 and 7 L/min, respectively. The medium pilot is 10-L bowl with a nominal feed rate of 15 L/min. The full scale can go to 50 L serving a 10,000-L bioreactor with a nominal

Batch and Semibatch Centrifuges

69

Figure 3.9 An automatic plunger discharge (APDII) tubular centrifuge with clean-in-place and sanitary-in-place capabilities.

feed rate of 70 L/min. A versatile arrangement is that the frame for suspending the centrifuge is the same for the 10-L centrifuge and smaller. The bowl can be changed depending on the process volume of feed that is required for separation. This provides a good advantage for a small plant to purchase only the bowl and not the frame or rest of the machine when the process feed rate increases to 15 L/min. Fig. 3.9 shows the installation of an automatic plunger discharge tubular installed for processing multiple products, including Escherichia coli, yeast, and other fungi. The equipment is equipped with control panel and various accessories for CIP and SIP, both of which will be discussed in Chapter 4, Disk Centrifuge, in details. The fermenter is visible on the left side in the photograph.

3.5

Summary

In this chapter, the batch spintube and ultracentrifuge have been presented. The ultracentrifuge, capable of a million times gravity, has

70

Centrifugal Separations in Biotechnology

different operating modes that facilitate the measurement of sedimentation velocity and molecular weight and the separation of small cells and cellular organelles. Also, the versatile centrifugal filter that combines sedimentation, filtration, and chromatography carried out with different functional modules has been presented. Each process can be enhanced by the centrifugal field, including the chromatography process. In addition, a semibatch tubular centrifuge, capable of 20,000g, is discussed with either manual or automatic cake discharge. The traditional tubular uses a high length-to-diameter ratio, which requires manual cake removal, whereas the more modern tubular, which takes advantage of the solid-basket centrifuge design and has a length-to-diameter ratio of order up to 3, is designed with automatic cake discharge and more space for solid storage, and consequently, it can take on feed with higher solids concentration.

References [1] W.W.F. Leung, Industrial Centrifugation Technology, McGraw-Hill, New York, 1998. [2] W.W.F. Leung, R. Probstein, Lamella and Tube Settlers - Part 1 Model and Operation, I&EC process Des. and Dev., 1983. [3] G.V. Childs, J.M. Lloyd, G. Unabia, D. Rougeau, Enrichment of corticotropes by counterflow centrifugation, Endocrinology 123 (1988) 2885 2895. [4] W.W.F. Leung, A. Shapiro, Improved design of conical accelerators for decanter and pusher centrifuges, Filtration Sep. (1996) 735 738. [5] W.W.F. Leung, A. Shapiro, Efficient double-disk accelerator for continuous-feed centrifuge, Filtration Sep. (1996) 819 823. [6] W.W.F. Leung, Experimental and theoretical study of flow and sedimentation of tubular centrifuge for bioseparation, in: AICHE Annual Conference, Cincinnati, OH, November 2005.

Problems (3.1) A suspension of relatively monodispersed particles settles in a vertical test tube under the Earth’s gravity, and it produces a sharp liquid-suspension interface that moves toward the bottom of the tube over time, while a layer of sediment builds up from the bottom of the tube to meet this liquid-suspension interface. If the

Batch and Semibatch Centrifuges

71

Boycott effect Clarified liquid g Traveling interface

Figure 3.10 Boycott effect.

tube is inclined with respect to the vertical (see Fig. 3.10), the liquid-suspension interface travels at a much faster rate toward the bottom of the tube. Explain this phenomenon, which was first discovered by the physician Dr. Boycott in the 1920s. How would this compare to the swing-out spintube and the fixed-angle spintube pair? (3.2) A small tubular centrifuge can attain 60,000 g when operating at an angular speed of 50,000 rpm. What is the diameter of the tubular centrifuge? It can take a maximum feed rate of 1 L/min and allows 100 300 g of sediment deposit in the bowl. A big tubular centrifuge with 457-mm diameter bowl can attain 20,000 g. What is the angular speed at which it should operate? (3.3) A group of cells of comparable size about 3 µm together with cell debris in the submicron range in a mixture need to be classified. A special reagent is added in solution to target some specific cells, which absorb the reagent and swell in size to about 6 8 µm. The mixture of cell debris, regular cells, and the swollen cells (product) are fed into a special elutriating centrifuge with the precision control feed rate. The density of the cells is 1050 kg/m3 and that of the suspension is 1000 kg/m3. The viscosity of the suspension is 5 cP. The mixture enters the separation zone in the centrifuge with an equivalent centrifugal acceleration at 20,000 g and a cross-sectional area of 0.5 cm2. In this rotor, liquid enters from the outer radius and is distributed circumferentially uniform at the separation zone with a counterflow velocity (i.e., directing toward the small radius) opposing the radial directed settling velocity. What is the eluent flow rate, expressed in mL/min, to keep the 3µm cells in the separation zone of the rotor, while the smaller submicron debris will be carried out and exit at the small radius outlet with the liquid, and concurrently, the swollen particles will settle out of the separation zone and exit at the large radius?

72

Centrifugal Separations in Biotechnology

(3.4) Referring to Problem 3.3, what is the eluent flow rate to the elutriating centrifuge if the 6-µm swollen cells are to be kept stationary in the separation zone, while the 3-µm regular cell and submicron cell debris are carried by the flow to exit at the small radius? After removal of the minus 6-µm particles, buffer and wash liquid may be added to wash the 6-µm cells (product) to remove any adhered contaminants.

4 Disk Centrifuge In this chapter, the inclined plate settler concept is first introduced. Subsequently this concept of settling in short distance is extended to disk-stack centrifuges. The various types, features, and functions of diskstack centrifuges for use in biotechnology separation will be discussed.

4.1

Lamella/Inclined Plate Settler

4.1.1

Inclined Plate Settler Principle

The distance for particle settling in a tank of height H is greatly reduced if a set of parallel plates spaced with a distance s apart is inserted in the tank reducing the settling distance from H to s. Instead, plates with a tilt angle θ with respect to horizontal is installed (see Fig. 4.1). The reason for the tilt angle is simply that settled solids left on the plates can slide down by a component of g so that they do not accumulate and clog the channel formed between adjacent plates. n plates

v g s/sin θ H=Lsin θ s

θ

L

s

vs

s/cos θ

Δρgsin θ θ Lcos θ

Δρg Δ ρgcos θ H'=nLcos θ

Figure 4.1 Lamella plate sedimentation. Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00004-0 © 2020 Elsevier Ltd. All rights reserved.

73

74

Centrifugal Separations in Biotechnology

For a particle settling at a velocity vs in a tank filled with liquid height H, the maximum time for the particle located initially at liquid surface to settle at the tank bottom is simply H/vs. If “n” number of inclined plates are installed with plates spaced at an equal distance s in the same tank with liquid height H (see Fig. 4.1), the maximum distance for which the particle has to settle to reach the plate surface is reduced to s/cos θ (see insert in Fig. 4.1). Thus, the maximum time for particle separation is s/cos θ/vs. The ratio of maximum time for separation for the inclined plates to that of the holding tank without the plates becomes (s/H)sec θ. Suppose s 5 3 cm, H 5 3 m, and θ 5 60 degrees, it follows that (s/H) sec θ 5 1/50. This is equivalent to stating that the flow sedimentation capacity of the inclined settler is Hcos θ/s times greater than that of the holding tank, assuming other complications are not being considered. This enhancement ratio is 50 times for the present example. Another related argument to consider is the projected horizontal sedimentation area, which is an indicator of the settling capacity. As shown in Fig. 4.1, for the holding tank, the projected horizontal sedimentation area is approximately (n 2 1)s/sin θ 1 Lcos θ, whereas for the lamella plate, it is nLcos θ. Given L 5 H/sin θ, the ratio of the lamella plate to that of the holding tank becomes nHcos θ/[(n 2 1)s 1 Hcos θ] 5 (H/s) cos θ/[(n 2 1)/n 1 (H/s)cos θ/n]. If n 5 300, this ratio is 43, which is below 50 as determined previously. As n increases to 500, this ratio increases to 45.5, and as n increases further, this ratio approaches Hcos θ/s, which is the same as that obtained previously. Indeed, both short settling distance and equivalent settling area arguments yield similar results. 4.1.2

Complications in Inclined Plate Settler

The foregoing arguments do not account for complications that arise in real situations. 1. These complications include particles that can be entrained by the moving fluid in the channel formed by adjacent plates even if they have already settled on the upper surface of the plates inclined. 2. Particle can accumulate to the inclined plate and may not slide down due to friction and self-adhesion between particles. As to the latter, the angle adopted for the inclined plate settler is typically at a steep angle of 50
60 degrees with respect to horizontal to ensure that sediment slides down the plates and does not accumulate and block the channels between adjacent plates.

Disk Centrifuge

75

3. Flow may not distribute uniformly into each channel. This is by far the most serious problem, and this point will be taken up later. 4. Fluid does not flow in one direction (primary flow) from the bottom to the top. Actually secondary flow exists at the entrance and exit as well as inside the channel [1]. 5. Finally, the interface between the countercurrent flow streams (i.e., primary and secondary flow) may become unstable. The instability can lead to flow turbulence, fluid-particle mixing, particle resuspension, and entrainment of particles by the main flow [2]. In any event, these complications might reduce the capacity of the inclined channels. Instead of having a capacity ratio of inclined plate to conventional tank of 50:1, this ratio might drop to 20:1 after the appropriate complications have taken into consideration. Regardless, the inclined plate settler provides at least an order of magnitude advantage over a conventional settling tank.

4.2

Disk-Stack centrifuge

4.2.1

General Disk Geometry

In this section, the inclined plate settler concept is extended to a centrifuge. Suppose the plate stack of Fig. 4.1 is rotated 90 degrees clockwise, the base H0 of the tank becomes the height of the centrifuge bowl as shown in Fig. 4.2, and the inclined plates shown in Fig. 4.1 become the axisymmetric conical disks with an outer radius R2 and an inner radius R1 as depicted in Fig. 4.2. When the disks rotate about the axis at high angular speed, centrifugal acceleration G is established. Heavier particles settle in the direction toward the large radius, while lighter fluid is displaced toward smaller radius, very much analogous to sedimentation under gravitational acceleration as shown in Fig. 4.1. In a disk-stack centrifuge, particles settle on the underside of the conical disk surfaces from which a G-component, Gsin θ, drives the particles to the peripheral of the disk stack. After leaving the disk stack, the full G-force further drives the particles to a temporary solid-holding space at the large diameter of the bowl. As with the inclined plate settlers shown in Fig. 4.1, the maximum time for separation between the rotating bowl, with disk stack versus the rotating “chamber” bowl alone (no disk-stack centrifuge) with the annular radial extent between R2 and R1 and height H0 , is significantly reduced by a factor of s/H/cos θ. If s 5 1 mm, L 5 100 mm, θ 5 40 degrees, and

76

Centrifugal Separations in Biotechnology R2 H=Lsin θ

θ

Lcos θ

H' = nLcos θ =(R 2 –R 1)ncot θ

R1

L

v

vs

n disks

Δ ρGcos θ Δ ρG Axis

Δ ρGsin θ s s/sin θ

s/cos θ G

Figure 4.2 Schematic of inclined plates (conical disks) sedimentation adopted for a rotating centrifuge.

H 5 64.28 mm, this ratio becomes 0.0203 which is almost 1/50 identical with the value obtained for the inclined plate settler under gravity (g). The important point is that the separation time is much reduced irrespective of the inclined plate settler under 1g or disk-stack centrifuge under thousands of g. Conversely, the capacity is much increased by having the disk-stack centrifuge making separation at a high G and large surface area. Similar complications occur with the disk stack as with its counterpart—the inclined plate settler; however, there is less instability associated with the countercurrent flow for the disk stack due to the strong rotating field that tends to stabilize the flow field in a two-dimensional sense (i.e., Taylor Proudman as discussed in Chapter 2: Principles of Centrifugal Sedimentation). Despite this, due to relative liquid motion in the rotating frame, the flow experiences further complicated Coriolis acceleration that allows the circulatory flow pattern that is represented schematically in Fig. 4.3 [3]. The circulatory flow promotes mixing and reduces the settling efficiency of the centrifuge. In Fig. 4.3A
C, as the liquid flows toward the small diameter (upward of Fig. 4.3A), the Coriolis acceleration directs the flow out of the paper (see Section A-A in Fig. 4.3B). However, when the flow is redirected to the larger diameter (downward of Fig. 4.3A), the Coriolis acceleration directs the flow into the paper (see Section B-B in Fig. 4.3C). The combined throughflow in the channel together with the

Disk Centrifuge

77

Rotation

A-A

A

B

B-B Ω

Longitudinal rib

Longitudinal rib

Ω v

v

v

a = −Ω X v

c Acceleration out of the paper

v

A

ac = − Ω X v Acceleration into the paper

B

(A)

(B)

(C)

Figure 4.3 Schematic of a segment of the angular channel showing secondary flow arises from the Coriolis effect.

secondary flow from Coriolis gives rise to the flow paths as shown in Fig. 4.3A. In most designs, the undesirable Coriolis force is counteracted by the longitudinal ribs (acting as spacers between adjacent disks in some designs). These ribs, e.g. in quantities of 6
8, are uniformly spaced around the circumference of the conical disk. The ribs reduce the extent of circulatory flow and oppose the undesirable Coriolis force (see Fig. 4.3A). The spacing between adjacent disks can be very small, nominally under 1 mm; however, it can be reduced to 0.3 mm or less for low feed-slurry viscosity or as large as 1.5
2 mm for processing viscous feed with high solids content. Fig. 4.4A shows a photograph of multiple disk-stack centrifuges installed in a biopharma plant. 4.2.2

Disk Angle

The inclined angle (with respect to the vertical axis) of the disks stack in a disk centrifuge is typically 35
50 degrees. The number of disks “n” is anywhere between 50 and 300 disks. This quantity varies between different sizes and designs of disk centrifuges. The centrifugal acceleration G ranges between 5000g and 15,000g. 4.2.3

Disk Spacing

The typical spacing between adjacent disks ranges from 0.32 to 1 mm. On the other hand, processing yeast with 30% by volume of feed solids needs more open spacing of 1 mm, whereas tighter spacing such as

78

Centrifugal Separations in Biotechnology

Figure 4.4A Picture of multiple disk-stack centrifuges in operation. Reproduced with permission of Westfalia Separator.

Figure 4.4B Photographed of UniDisc from Alfa Laval. The feed and effluent/centrate holes for the disk are also shown. Reproduced with permission of Alfa Laval.

0.5 mm is most suitable for processing Escherichia coli and lysate with lower feed solids. Disk spacing of 0.32 mm, or even as low as 0.25 mm, can be used for processing mammalian cell broth at low feed solids concentration of 3%
4%. Fig. 4.4B shows a disk design with precision machining on numerous protrusions, or legs, distributed uniformly on the disk surface.

Disk Centrifuge

79

Table 4.1 Summary of various disk centrifuge and their feed and discharge

Solids discharge % by vol. feed solids Concentrate

Manual discharge

Dropping bottom discharge

Nozzle discharge

Manual

Intermittent

Continuous

0
1

0
10

5
20

Minimal

Nonflowable

Paste, flowable

The protrusions provide accurate, uniform spacing between adjacent disks. Given their small size, the protrusions provide minimum interference to the flow. The disk spacing is below 0.3 mm; therefore, nearly 50% more disks (i.e., increasing n) can be stacked up for the disk-stack centrifuge to increase the settling capacity. There are also curved ribs on the disk surface, and the height of the ribs are below that of the protrusions so that the protrusions truly provide the spacing between adjacent disks, while the ribs are used to stop the Coriolis flow. Table 4.1 summarizes different types of centrifuges in terms of feed solids and concentrate discharge. 4.2.4

Process Functions of Disk Centrifuge

There are several important functions in solid
liquid separation of a disk centrifuge combing the high G together with large disk areas (as discussed earlier). 1. Separate suspended solids from liquid phase: The solids are removed from the liquid phase. 2. Clarification/purification: The intent is to reduce the suspended solids in the liquid centrate or effluent phase to a minimum. 3. Thickening or concentration: The feed slurry is concentrated to a suspension with much higher solid concentration. The loss of liquid protein in the concentrate is reduced. Furthermore, the solid loss in centrate should be also minimized. 4. Separation and washing: The solids in suspension may contain contaminants. The suspension is centrifuged first, and the centrifuged concentrate is reslurried with wash liquid to dilute and

80

Centrifugal Separations in Biotechnology

dissolve the contaminants. The resulting slurry is separated again by centrifugation to remove the spent wash liquid containing dissolved and suspended contaminants. The separation and washing process can be repeated as necessary until the dissolved and suspended contaminant level of suspension reaches an acceptable level. Obviously, this should also be balanced by the economics of the process. 4.2.5

Feed Solids

1. Suspended solids: The typical feed solids in suspension, especially for biopharmaceutical applications, are 2%
4% v/v (by bulk volume) for mammalian cells but can increase to 30% or more for yeast. The range of variation is quite large. In the future, the feed solids of mammalian, cell broth might increase to 4%
6% v/v and perhaps even higher due to increase in solids capacity from the upstream bioreactors. 2. Dissolved solids: The dissolved solids in the feed consist of valuable protein product, unfortunately mixed with other contaminants, also in soluble form, that need to be removed in downstream purification. There are three types of centrifuges in accordance to the mode of discharging concentrate solids. Table 4.1 summarizes these three types according to the mode and nature of concentrate discharge, as well as to the percent by volume of feed solids that these machines typically handle.

4.2.6

Manual Disk Centrifuge

In the manual disk centrifuge as shown in Fig. 4.5A, a space or volume in the bowl at a larger diameter is used to temporarily store the sediment from centrifugation. When this space becomes nearly full, the centrate turns turbid due to entrainment of sediment by the incoming feed stream, at which the feed has to be shut off, machine coast down, the pool liquid in the centrifuge drained, and the accumulated solids removed manually. Therefore, for practical purposes, it is important that the feed solid concentration to the centrifuge should be low, otherwise there will too much downtime due to solids discharge and cleaning. This has seriously impact on the operation.

Disk Centrifuge

Clarified liquid Weir Disk stack

81

Feed suspension

Clarified liquid Bowl Feed Sediment Feed acceleration

Figure 4.5A Schematic of a manual disk-stack centrifuge. Lighter liquid Heavier liquid Disk stack

Bowl

Feed Lighter liquid

Heavier liquid Sediment

Figure 4.5B Schematic of a liquid
liquid
solid disk centrifuge showing the rising feed hole near the periphery of the disk stack.

4.2.6.1

Clarification of the Light Phase

A typical application of this design would be for clarifying a liquid stream with a small amount of solids. The light phase in this case is water. Another application is to separate water droplets from continuous oil phase in which the light phase is oil. If there is an interface between oil and water, the centrifuge should be operated such that the interface with radius R should be located outside the disk stack with R . R2 as depicted in Fig. 4.5A, or in the proximity of R2 as depicted in Fig. 4.5B, to maximize polishing of the light phase (i.e., oil) using conical disk area and residence time for settling out the dispersed heavy phase (i.e., water droplets).

82

Centrifugal Separations in Biotechnology

4.2.6.2

Separation

A common application for disk centrifuge is for liquid
liquid
solid separation, such as shown in Fig. 4.5B. This is to separate a lighter liquid (such as oil), a heavier liquid (such as water), and a very small amount of suspended solids (heaviest among all three phases). In the case where oil is lightly present in a continuous water phase, the disk centrifuge can be used to remove the oil droplets from the processed water with the provision that the disk be equipped with a set of equal-radius (equiradius) holes proximate to the small radius R1. When the disks are stacked and holes of all the disks aligned, a continuous channel is formed through which the oil
water feed should be introduced. From this radius to the disk outer radius R2, the small amount of dispersed oil droplets can be separated out from the continuous water phase. The dispersed oil droplets coalesce to form larger droplets, and because of buoyancy, they move to the smaller radius and get discharged at a smaller radius. The clean water phase, which is heavier, moves toward the large radius where it get trapped and discharged though another liquid exit at the larger radius. This maximizes the separation of the undesired discrete lighter phase (e.g., oil droplets) from the continuous heavier phase (e.g., water). This is similar to shown in Fig. 4.5B, except with the rising holes close to the small radius R1 of the disk stack. 4.2.7

Intermittent Discharge

For this configuration as illustrated in Fig. 4.6A, accumulated solids discharge through peripheral ports, which are opened on a time-based mechanism triggered by fixed time intervals or by the increased pressure Clarified overflow R2

D

R1

Disks Concentrate

θ

Discharge

Feed suspension Clarified liquid

Dropping bottom Feed Feed acceleration

Figure 4.6A Intermittent discharge or dropping-bottom disk centrifuge.

Disk Centrifuge

83

due to solids loading in the bowl. During solids discharge, the bowl bottom (a separate cover piece) drops exposing an annular opening for discharging accumulated solids. Alternatively, the bowl bottom can drop when the “pressure” due to accumulated solids exceeds a threshold. The dropping bottom can be controlled hydraulically using water or pneumatically using inert gas with counteracting mechanical springs restoring the closing position of the bowl. 4.2.7.1

Two Intermittent Discharge Designs

There are two types of intermittent discharge. One type uses peripheral discharge where solids are ejected radially outward, and the other uses peripheral discharge where solids are directed axial at the periphery of the disk centrifuge. The radial discharge disk operates at a slightly lower G-force compared with the axial discharge, which has improved mechanical integrity wherein ports in the bowl and lock ring are better positioned. As an example, the radial discharge is capable of 13,000g, whereas the axial discharge can operate at higher speed and Gs of 15,000g. The radial discharge can process all types of solids except solid with shear-thickening behavior. On the other hand, the axial discharge can process only flowable solid. This is because solids leaving the solidholding space need to travel a short axial distance toward the bowl bottom before being ejected. Some designs have built in a pressurized air source to facilitate solid discharge for the axial-discharge disk centrifuge. 4.2.7.2

Angle of Cone for Discharge

The angle of discharge solid-holding space in a disk centrifuge depends on the repose angle of the concentrate. The initial solid surface after the solid-holding space in a disk centrifuge is filled, and the final solid surface after the solids are being ejected from the discharge ports, which are shown in Fig. 4.6B. The angle of repose under G-force of the solid dictates the final solid surface. The angle of repose of granular solids measures how solid grains are locked to each other through surface contact so that they can build-up a pile or structure against gravity force or G-force. (This is similar to the slump test for assessing the added water content in preparing fresh wet concrete.) Two scenarios are shown in Fig. 4.6B: the first scenario is where the repose angle of the solid concentrate φ is greater than the cone angle γ, and the second is where φ is smaller than γ. The latter is more favorable for concentrate discharge because solids should not accumulate with their low angle of repose,

84

Centrifugal Separations in Biotechnology

ρGcos γ Solid

ρG ρGsin γ Final solid surface

Initial solid surface

G

(1) φ, φ>γ (2) φ, φCo Co

Figure 4.6C Feed rate (top diagram), rotation speed (middle diagram), and concentration of centrate suspended solids (bottom diagram) during feed with intermittent discharge.

Disk Centrifuge

Clarified liquid Weir

87

Feed suspension

Clarified liquid Bowl Feed sediment

Figure 4.6D Chamber bowl centrifuge (no disk stack).

between 10 and 15 seconds. This is especially for the bowl rotation speed, depending on gear or belt drive. Later, we will show how continuous concentrate discharge can eliminate this interruption caused by intermittent discharge if this issue becomes unacceptable, especially when the frequency of intermittent discharge is high. The other possibility is to have the feed temporarily replaced by the buffer liquid as discussed in Section 4.2.7.4. 4.2.8

Chamber Bowl

As shown in Fig. 4.6D, the chamber bowl with intermittent discharge has no disk stack. The chamber bowl is similar to a tubular centrifuge but with small aspect ratio (L/D ratio), less than 1. It is suitable for running viscous feed or feed with more concentrated solids. As stated, the clarification capacity is an order of magnitude less than that with the disk-stack centrifuge due to reduced settling area in the absence of disk stack. As such, the feed rate usually is smaller for a chamber bowl when compared with that of disk-stack centrifuge for the same bowl size. However, chamber bowl can take higher feed solids compared with disk-stack centrifuge. 4.2.9 4.2.9.1

Continuous Concentrate Discharge External Nozzle discharge

As depicted in Fig. 4.7A, nozzle disk can be used for continuous concentrate discharge. This type of machine has been used in brewery in the past. Typically, a set of nozzles are located at the periphery of the

88

Centrifugal Separations in Biotechnology Clarified overflow R2

D

R1

Disks Concentrate θ

Discharge

Feed suspension Clarified liquid

Nozzle Feed Feed acceleration

Figure 4.7A Nozzle discharge (continuous solids discharge) disk centrifuge.

bowl or at a smaller radius (for power saving purpose) to discharge flowable solids. The latter feature is not shown in Fig. 4.7A. The total nozzle area is selected to balance feed solids with concentrate discharge solids. Along power savings, the nozzle discharge is directed approximately along a direction opposite to rotation, which recovers part of the energy of the discharge stream (see Fig. 4.8A). The nozzles range in size from 0.5-mm diameter openings for use on the smaller centrifuges to 3.2 mm for the larger centrifuges. The designed number of nozzles per centrifuge, depending on its size, ranges typically from 12 to 24. For satisfactory operation, the minimum allowable nozzle size is at least twice the diameter of the largest particle to be discharged. Large particles must be removed by pretreatment such as screening. The number of nozzles is controlled by the angle of repose of the sedimenting solids and must be selected so that the accumulation of solids between adjacent nozzles do not build into the disk stack and interfere with its clarification effectiveness. Solid concentrate that is not flowable, like yeast or yogurt, can clog the nozzles. This can be discharged using the internal nozzle design that is discussed in Section 4.2.9.3. 4.2.9.2

Small-Diameter Concentrate Discharge

In a special design of disk-stack centrifuge as shown in Fig. 4.7B, the top surface of the disk stack and the inner surface of the conical bowl form a 360-degree annular passage for which the concentrate collected at the bowl diameter can flow radially inward to a collection annular

Disk Centrifuge

89

Centrate Concentrate Rb , pb Ra , pa

Centrate Concentrate/ feed interface

Ri , pi

Figure 4.7B Discharge of concentrate in the annular area between the upper surface of disk stack and the inner surface of upper bowl.

chamber located near the machine axis. A stationary skimmer pipe is used to skim the concentrate liquid off the chamber (see Fig. 4.7B). Suppose the centrate liquid is discharged continuously to a pickup point with radius Ra at a pressure pa, and the flowable concentrate with density ρc flows through the annular channel to the pickup point with a larger radius Rb at a pressure pb. Applying Eq. (2.1b) to the feedcentrate interface “i” with Ri and centrate pickup “a” with radius Ra,R pi 2 pa 5

1 R Ω2 ðR2i 2 R2a Þ 2 f

(4.1)

Also, applying Eq. (2.1b) to the feed-centrate interface “i” with radius Ri and concentrate pickup “b” with radius Rb, pi 2 pb 5

1 R Ω2 ðR2i 2 R2b Þ 2 c

(4.2a)

Subtracting Eq. (4.2a) from Eq. (4.1), we eliminate the interface pressure and get directly the pressure pb in terms of other variables, thus 
 1 
pb 5 pa 1 Ω2 Rc R2b 2 Rf R2a 2 R2i Rc 2 Rf 2

(4.2b)

Furthermore, a stationary skimming pipe is used to skim off the fluid concentrate at radius Rb. With the concentrate having a linear velocity that equates to the circumferential tangential velocity at solid-body rotation ΩRb, and using the Bernoulli theorem for nonrotating fluid, we have 1 pb 1 Rc Ω2 R2b 5 pv 1 1=2Rc v2c 2

(4.2c)

90

Centrifugal Separations in Biotechnology

where pv is the back pressure of the throttle valve and vv is the velocity at the valve. Rearranging, sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2ðpb 2 pv Þ vc 5 1 ðΩRb Þ2 (4.3a) Rc Since pb is fixed by Eq. (4.2b), which depends on Ra, Rb, Ri, pa, and Ω; therefore, by controlling the throttle valve pressure pv, the velocity vc can be controlled, and hence the flow rate Qc of the concentrate, Qc 5 Avc

(4.3b)

where A is the flow-through area of the throttle valve, which can vary, thereby controlling rather easily the concentrate flow rate. The change in concentrate flow may also affect the solids content of the concentrate (as too high rate reduces concentrate solids and vice versa) and the concentrate-feed interface position Ri inside the centrifuge; see Fig. 4.7B. The small-diameter concentrate discharge is an interesting design where the concentrate is brought from the bowl radius with high tangential speed ΩRbowl to a concentrate pickup chamber near the rotation axis at much smaller radius Rb with much lower tangential speed ΩRb. This obviously reduces the concentrate discharge kinetic energy (KE) as KE is proportional to the quadratic power of the linear speed. Otherwise, this energy that has been saved would have been dissipated by discharging the concentrate through external nozzles located at the bowl radius. Not only the external nozzle discharge centrifuge wastes energy, biological cells contained in the concentrate would be killed as the high-speed concentrate, discharging in discrete jets, is abruptly stopped by the stationary casing of the centrifuge. The biological cells in the concentrate would become unviable and useless products. Unlike the case of the internal nozzle design, which is discussed in Section 4.2.9.3, there is no internal nozzle involved, and the control of the concentrate flow is more effective by the throttle valve outside the centrifuge. A small-diameter concentrate discharge design for a commercial disk-stack centrifuge and the corresponding cross-sectional schematic are shown in Figs. 4.7C and 4.7,D, respectively. The 400-mm diameter machine is capable of 8600g handling high-density microbial cell fermentation with the maximum feed rate of 167 L/min (10 m3/h). The dropping bottom and bowl nozzles are only used during CIP flushing of residual solids. Both the centrate and the concentrate are discharged via rotating impeller pumps to stationary chambers, and subsequently the

Figure 4.7C Bactofuge. Reproduced with permission of Alfa Laval.

4a

4b

Figure 4.7D Bactofuge schematic. Feed (1) introduced from below through drive hollow spindle (2). Separation takes place in the disk stack (3). Centrate (4a) and concentrate (4b) are discharged at the top through rotating centripetal pumps, respectively. Solid-holding space (5) is the storage of concentrate and the concentrate flows toward the small radius to the impeller pump at 4b. During CIP, the bowl bottom drops, and CIP liquid push out the solids or residual through external nozzles (7). Reproduced with permission of Alfa Laval.

92

Centrifugal Separations in Biotechnology

flow is directed out of the machine. The velocity of the rotating fluid is converted largely to static pressure. A simplified analysis of the rotating impeller pump for the flowable fluid is given in Appendix C. 4.2.9.3

Internal Vortex Nozzle Discharge

Another continuous concentrate discharge centrifuge is via internal nozzle arrangement; see Fig. 4.7E. Here, the concentrate collected at the large diameter of the bowl is directed radially inward through discrete number of passages and distributed uniformly around the circumference of the bowl. The concentrate from each passage ends up being injected into a small chamber with radius r2 at a tangent to the chamber periphery (similar to feeding a cyclone) creating a free vortex. The concentrate leaves the chamber via a nozzle, with radius r1, located at the center of the chamber (similar to the vortex finder in cyclone for the overflow). The nozzle is used to speed up the flow by a ratio of r2/r1 based on free vortex. After gaining additional speed from the free vortex, the flow is discharged to a liquid pool in the rotating collection chamber. There, the concentrate is picked up by a stationary skimmer pipe. The momentum of the fluid-like concentrate is converted to higher static pressure. The total pressure, which is a combination of the static pressure and the Feed Centrate Stationary pairingdiskStationary concentrate skimmer

Centrate

Nozzle radius, r1 Disks

Concentrate Concentrate collection chamber

r2 Vortex chamber and nozzle Rn

Figure 4.7E Discharge of concentrate via internal passage, vortex chamber, vortex nozzle, and concentrate collection chamber in the rotating bowl and finally being picked up by stationary skimmer. Exploded view of vortex chamber and nozzle.

Disk Centrifuge

93

momentum from the rotation, is used to drive the concentrate flowing along the skimmer pipe. An approximate analysis of the aforementioned process can be carried out. Assuming the concentrate fluid is brought slowly to the nozzle chamber with a nozzle axis radius Rn (see Fig. 4.7E) so that the concentrate fluid is in equilibrium (in solid-body rotation) with the local radius. Therefore, it has a velocity of ΩRn, and this is velocity of the flow being fed tangentially into the vortex chamber with radius r2 (similar to that of the cyclone). The flow leaves the vortex nozzle with radius r1; therefore it should have a higher tangential velocity ΩRn(r2/r1) assuming free vortex. At exit of the nozzle, the flow is directed to a liquid pool in the concentrate collection chamber at a static pressure pb, which may not be too large. Consequently, the total pressure ignoring the magnitude of static pressure is approximately  2 r2 ΩRn 1=2Rc r1 which is used to drive the flow through the skimmer pipe. Given the nozzle radius Rn is fixed by the design, the only parameter that can change the concentrate flow is the nozzle radius r1 of the nozzle chamber. There are different nozzle diameters that are available depending on the process and the desired concentrate flow rate. When too large, a nozzle radius r1 results in the low flow rate, while when too small, a nozzle radius may clog up the nozzle. The nozzle radius can be changed only when the machine is fully dissembled, which is somewhat inconvenient. Therefore, the small-diameter concentrate discharge design offers more flexibility in adjusting the concentrate discharge rate “on the fly” (i.e., while the machine is running) by adjusting the back pressure of the throttle valve. However depending on the process when the angle of repose of the concentrate is too large the concentrate may plug up the annular passage of the small-diameter concentrate discharge design, and the internal discharge nozzle design offers a better solution. 4.2.9.4

Applications of Different Concentrate Discharge Designs

For the intermittent discharge or the external nozzle designs, the linear momentum mΩRbowl of the concentrate stream with mass m discharged at the bowl radius is destroyed once the discharged concentrate hits the wall of the stationary casing. The shear and impact of the concentrate stream on the stationary centrifuge casing would kill the living cells

94

Centrifugal Separations in Biotechnology

and microbials in the concentrate if they are subsequently being recycled back to the bioreactor/fermenter. This can be avoided with both internal nozzle discharge and the small-diameter concentrate discharge, which allow concentrate to be discharged at low velocity and reduced shear avoiding breaking of cells. Indeed, the internal nozzle and the small-diameter concentrate discharge are most suitable application. As an example, for one process, mammalian cells are used to secret protein intracellularly, and at some stages, these cells have reached maturity in which the cells burst by themselves releasing the protein in the liquid broth. During separation, both mammalian cells and the cell debris as well as the suspending liquid broth are sent to centrifuge for separation. The cell debris, which are finer, leaves with the protein liquid to be separated downstream by a depth filter or a microfilter. The whole cells after collected in the bowl wall as concentrate are sent to the small-diameter concentrate discharge or inward nozzle discharge where the concentrate containing delicate cells are gently discharged at lower speed, reducing shear that can lyse the cells. This is especially for mammalian cells that do not have a cell wall. Even if the cells have cell wall such as yeast and bacteria, the high impact can destroy cells. The mammalian cells are recycled back to the bioreactor for further processing producing more intracellular protein. As another example, the small-diameter concentrate discharge design is a good candidate for separating bacteria from liquid by diskstack centrifugation. In one application, the liquid centrate is the product, such as raw milk, and the solid concentrate is the bacteria spores that can spoil the milk, which need to be removed. In another application, the concentrate is the probiotic bacteria product that should be recovered by centrifugation. With regular intermittent discharge or nozzle (external) discharge, the bacteria will be sheared and lysed as the concentrate impinges onto the walls of the stationary casing at high velocity. However, using the small-diameter concentrate discharge, it avoids the impact of the concentrate that can destroy the bacterial cells, and this arrangement also provides continuous concentrate discharge without interrupting the separation process as discussed in Section 4.2.7.5. The small-diameter concentrate discharge is recommended for liquid concentrate that has an angle of repose less than 25 degrees so that the concentrate can flow more readily, while the case of the internal nozzle discharge is recommended for liquid concentrate having a larger angle of repose.

Disk Centrifuge

4.2.9.5

95

External Nozzle Designs

Figs. 4.8A
4.8D show a series of photographs on a nozzle disk centrifuge being disassembled in the shop for inspection. Fig. 4.8A shows the rotor assembly of a nozzle disk centrifuge. The nozzle openings on the bowl are directed at an angle counter to the bowl rotation direction to save power. Fig. 4.8B shows the cutout at the periphery of the disk forming a continuous rising channel for feeding the disk stack. More than 100 disks are stacked as shown in Fig. 4.8C. Finally, Fig. 4.8D shows the lower conical portion of the nozzle disk bowl. 4.2.10

Liquid Discharge

Centrate or effluent liquid can be discharged by centripetal pump (also known as paring disk). The advantages of using centripetal pump are as follows: 1. reduction of energy of discharge stream, 2. reduction of foaming especially when liquid has dissolved protein, and 3. reducing contact with air. 4.2.10.1 Centripetal Pump for Liquid Discharge The working mechanism of the centripetal pump is shown in Fig. 4.9A. For illustration, only a pair of stationary curved channels is used to skim

Figure 4.8A Rotor assembly of a nozze disk centrifuge. The direction of rotation of is clockwise. The nozzles are directed in the anticlockwise direction to save power.

96

Centrifugal Separations in Biotechnology

Figure 4.8B The rising channel, formed by the cutout at the periphery of the disk, forming a continuous channel for suspension feeding in the disk stack.

Figure 4.8C There are over 100 disks. The openings in the bowl with the nozzles (not shown) can be seen.

Disk Centrifuge

97

Figure 4.8D The lower conical portion of the nozzle disk bowl.

Flowable area A, flow rate Q Rotating pool

Rp

Figure 4.9A Stationary centrifugal pump dipping into a rotating pool directed counterclockwise.

the rotating pool. In practical design, multiple pairs of curved channels are used depending on the flow rate. In all cases, the openings of the curved channels dip into the rotating pool to skim the rotating liquid. The openings of the pump face opposite to the rotating direction of the flow are shown in Fig. 4.9A. The rotating liquid enters the openings of the channels and gets decelerated as the channels are curved with the increasing flow area and the increasing flow path to gently decelerate the fluid entering the channel. After collecting at the center of the pump, the fluid is redirected to flow axially in the pump to a connected pipe, which leads to the back-pressure valve. By controlling the back pressure, the kinetic energy of the rotating fluid is converted to pressure. The maximum back pressure that can be established depends on the size of

98

Centrifugal Separations in Biotechnology

the pump and the rotation speed of the centrifuge. It can be as low as 1 bar to as high as 7 bars. In any case, the recovered pressure is of the order of magnitude as the dynamic pressure due to the rotating pool 1 /2ρ(ΩRp)2, with pool surface radius Rp. Using Bernoulli’s equation with energy loss for a nonrotating system assuming negligible change in elevation, p1 1 1=2ρv21 5 p2 1 1=2ρv22 ð1 1 Closs Þ

(4.4)

In Eq. (4.4), location “1” refers to the intake slightly below the pool surface, p1  pa, and v1  ΩR1, and “2” refers to a location in the collection duct. p2 is the pressure build-up at the collection duct to be determined, and v2 5 Q/A, where Q is the total centrate discharge rate and A is the cross-sectional area of the collection duct (see Fig. 4.9A). Closs is the head loss coefficient in converting kinetic energy to pressure for the pump. The foregoing analysis is an oversimplification of the flow loss. Rearranging the equations,  2 1=2ρ QA ð1 1 Closs Þ p2 2 p1 512 1=2ρðΩRp Þ2 1=2ρðΩRp Þ2

(4.5)

If Closs is relatively small compared with unity and p1 is nearly at atmospheric pressure, then the gauge pressure p2 2 p1 at the centripetal pump becomes   1 2 2 1 Q 2 p2 2 p1  ρΩ Rp 2 R 2 2 A

(4.6)

This is the maximum pressure that can be recovered assuming the energy loss or dissipation in the centripetal pump is negligible, that is, Closs  0. Note that part of the pressure recovered is converted to the kinetic energy of the discharge liquid. In practice, a well-designed properly sized centripetal pump can indeed recover a large fraction of the dynamic pressure of the rotating pool minus the kinetic energy of the discharge liquid. The design and size of the pump is tailored to a given flow rate and rotation speed. In contrast, without the centripetal pump, the kinetic energy of the discharge liquid jet (originated from the kinetic energy of the liquid pool) would have been dissipated to heat and foam as the jet hits the wall of a stationary collector. The soluble protein product would further degrade upon impact of the liquid jet on the stationary wall. Note that the maximum pressure regained as given by Eq. (4.6)

Disk Centrifuge

99

Figure 4.9B Stationary paring disk design (cover removed, not shown). Reproduced with permission of Alfa Laval.

can also be expressed as the kinetic energy per unit volume of the pool liquid minus the kinetic energy of the discharged liquid.  2 Q 2 2 (4.7) KE=V 5 1=2ρΩ Rp 2 1=2R A Furthermore, if A is sufficiently large, the kinetic energy of the discharged liquid in Eqs. (4.6) and (4.7) can be neglected. In other words, Q/A ,, ΩRp. A paring disk with 10 inlets distributed around the circumference is shown in Fig. 4.9B. These inlets are positioned to dip slightly into the liquid pool. Note the direction of the rotating pool is clockwise. The fluid with high tangential speed ΩRp is skimmed and directed through the 10 curved diverging channels to the center near the axis. The kinetic energy per unit volume of liquid is converted to pressure head after some minor losses. 4.2.10.2 Hermetic Seal Design at Liquid discharge On a similar principle, a hermetic seal design can also be used in liquid discharge. The centrifuge bowl is completely filled with liquid, which eliminates liquid
air interface. When air is completely sealed off from contacting the feed and discharged liquid, this prevents oxidation and contact with airborne virus and bacteria of the protein product in the broth leading to product denaturing and contamination.

100

Centrifugal Separations in Biotechnology

4.2.11

Solution to Adverse Heating Effect

For separation applications in which protein is secreted extracellularly or intracellularly in the liquid with the cells bursting in the bioreactor releasing the protein, the protein in the centrate can be heated up by more than 20 C from the rotating bowl due to windage, which is air resistance to the rotating bowl from the air mass trapped between the stationary centrifuge casing and the rotating bowl. The air mass adjacent to the rotating bowl is forced to rotate, while the air mass adjacent to the stationary casing is held stationary. The air flow pattern generates vortices, turbulence, and energy dissipation in the form of heat. The higher is the rotation speed to generate high G-force for difficult separation, the larger is the heat dissipation. The temperature rise of 20 C in the product liquid might denature the protein product in the centrate or kill the probiotic bacteria in the concentrate in the smalldiameter concentrate discharge as the concentrate are located adjacent to the inner bowl wall getting the most direct heating from windage at the bowl exterior. Furthermore, the large surface of the bowl promotes efficient heat exchange from the hot air outside the bowl to the contents inside the bowl. There are two ways that can address this problem. 1. The feed can be chilled down approximately 10 C or more before introducing to the centrifuge. 2. A partial vacuum can be applied between the rotating bowl and the stationary casing. This reduces the air mass being trapped between the rotating bowl and the stationary casing, thereby reducing friction and windage and the rise in air temperature from energy dissipation. To accomplish partial vacuum in the air layer trapped between the rotating bowl and the stationary casing, two mechanical seals are required to be installed respectively, at the top and the bottom of the spindle of the centrifuge to seal the air mass trapped between the rotating bowl and the stationary casing from the outside air.

4.3

Feed Inlet and Accelerator

All disk centrifuges are designed with highest surface area and maximum G for use in a given application.

Disk Centrifuge

4.3.1

101

Introduction to Low Shear

Low shear stress is required from centrifuge inlet to centrifuge exit when processing suspension with mammalian and other delicate cells. At the inlet, feed suspension fills to the axis of the machine the flow accelerate gently from the axis, R 5 0 with zero circumferential velocity v 5 0. Also, it is important to have a good feed accelerator design to impart a solid-body rotation to the feed from the axis R 5 0, or close to the rotation axis, to the periphery of the disk stack at a larger radius R2. 4.3.2

Hydro-Hermetic Feed Design

Feed is delivered to the bowl from a stationary feed pipe. The surface of the liquid pool is at a radius Rp as shown in Fig. 4.10A. Under solid-body rotation, the pool liquid should have a tangential velocity v 5 ΩRp in the direction of rotation. This required tangential velocity could not have been delivered by a stationary feed pipe despite the pipe opening can be oriented at an angle to the pool, nor could it be readily delivered by a rotating feed pipe as it is not effective. The design of a rotating pipe is complicated, and most importantly, it does not help as the pipe diameter is small; therefore the imparted tangential speed is also limited. Without the tangential velocity, when the feed is introduced into the pool in a solid-body rotation, there is a mismatch in tangential speed between the feed and the pool liquid, which immediately leads to slip, turbulence, and mixing. With the latter, unfortunately, this promotes oxidation of the cells in the liquid pool, especially for those near the pool surface and in stagnant areas that have contact. This is solved with the hermetic feed-seal mechanism wherein liquid pool can fill to the axis of the machine (see Fig. 4.10B). G Feed pipe

Uniform pool depth

Rp Axis

Figure 4.10A Pool level distribution.

102

Centrifugal Separations in Biotechnology G Rotating flow, higher density, pool shallower

Feed pipe

Baffle

Nonrotating flow, lighter density, pool deeper

Axis

Figure 4.10B Pool level elevated at the pool side with less acceleration. One design uses additional accelerating vanes to ensure accelerating feed stream to solid-body rotation before feeding the disk stack.

A circular baffle is installed perpendicular to the axis at the end of the pipe. The baffle protrudes into the pool separating the pool into upstream and downstream sections. Downstream of the baffle, the pool liquid is either accelerated, or somewhat accelerated, by contact with other rotating surfaces. (As an example, one design employs a set of closely spaced rotating disks acting as a disk pump to impart circumferential momentum to the local pool liquid.) On the other hand, the pool liquid upstream of the baffle is not accelerated, and it appears the liquid there is lighter in density when compared with the accelerated “appeared heavier” liquid pool downstream. As a result, the upstream lighter pool liquid needs a taller liquid column to balance the downstream heavier liquid with a shorter liquid column. This induces a differential liquid head across the baffle, as shown in Fig. 4.10B (see also Section 2.2.2). The liquid upstream can even fill up to the axis of rotation, sealing-off any air contacting the pool, thus eliminating air and oxidation altogether. 4.3.3

Power Loss

Let us consider the energy loss during the feed acceleration as this provides some insight into minimizing shear, especially on processing shear-sensitive feed materials. Let the inlet be at pool surface radius Rp (i.e., entrance to the feed accelerator) and the outlet at the exit of the feed accelerator Rex. The power acquired by the exit liquid stream after acceleration by the feed accelerator rotating at speed Ω is 1 1 Pout 5 RðVÞ2 5 ρΩ2 R2ex 2 2

(4.8)

Disk Centrifuge

103

The power input to the centrifuge from the motor is given by the product of torque T and rotating speed Ω. Based on Newtonian mechanics, which states that the change of moment of momentum equals to torque, the input mechanical power by the centrifuge drive can be determined: Pin 5 TΩ 5 ½ρQðΩRex ÞRex Ω 5 ρQðΩRex Þ2

(4.9)

Subtracting the output power given by Eq. (4.8) from the input power given by Eq. (4.9), the loss power equals 1 1 Ploss 5 Pin 2 Pout 5 ρQðΩRex Þ2 2 ρQðΩRex Þ2 5 ρQðΩRex Þ2 2 2

(4.10)

This is precisely half of the input power [3] which is lost due to shear that ultimately dissipated in viscous losses in feed acceleration. Consider the feed is being accelerated quasi-steadily from the smaller radius Rp to a larger radius Rex, the hydrostatic pressure Δp between these two radii is simply 1 Δp 5 ρΩ2 ðRex 2 2 Rp 2 Þ 2

(4.11)

Multiplying both sides of Eq. (4.11) by Q, and combining with Eq. (4.10), we obtain another form on the power loss as follows: 1 Ploss 5 QΔp 1 ρQðΩRp Þ2 2

(4.12)

The first term of the power loss in Eq. (4.12) represents frictional losses as feed with volumetric rate Q is accelerated from Rp to Rex in the feed accelerator. In this process, there will be shear and pressure forces acting on the fluid from the rotating surfaces carrying out various mechanical work and dissipation. Fortunately, these forces are not that large as they are being spreaded out over a large surface area. As a result, the overall power loss QΔp is similar to that of flow through a “resistor” where there is a pressure drop Δp across the flow resistor. The second term in the power loss in Eq. (4.12) can be minimized when Rp is rendered zero or feed starts accelerating at the axis, Rp 5 0. This is precisely being done in the foregoing discussion with the hydro-hermetic feed design when liquid is filled to the axis. Sealing the center core of the disk centrifuge with liquid also prevents air and oxygen coming in contact and spoiling the liquid product with protein. These are two key advantages for the simple hydro-hermetic design.

104

4.3.4

Centrifugal Separations in Biotechnology

Feed Acceleration Visual and Quantitative Testing

An issue that is often ignored when dealing with centrifugation is whether the feed is fully accelerated before introduced to the separation zone. This is a critical question especially pertaining to high G-centrifuge where additional speed and power are required to generate better separation. Some extensive research has been conducted on feed acceleration of a continuous liquid stream [4
12]. As shown in Fig. 4.11, a simple test rig has been setup to visualize the effect of feed acceleration in a centrifuge. The bowl is cantilevered from support at one end and rotates at 1000 rpm. A ring weir attached to the centrifuge bowl holds a rotating liquid pool in place to prevent it overflowing. The feed is introduced to the feed accelerator via a stationary pipe at the support end. After the bowl has accelerated to full rotation speed, the feed stream is introduced to the pool through an accelerator. Two bowls are setup side by side wherein one configuration has a conventional feed accelerator for bench marking, and the other has an improved feed accelerator. To help visualizing the pool, a pool meter supported on free bearing is driven by two diametrically opposite paddle that extends 3 mm into the pool. The rotating liquid pool at the introduction of the feed stream drives the paddle of the pool meter. The rotation speed of the pool at the point of feeding can be determined using a stroboscope (abbreviated as strobe) by which the strobe frequency is tuned such that the supposedly rotating paddle becomes stationary. This corresponds to the rotational velocity of the introduced feed in the pool. In addition, the pool surface where the feed is introduced can be clearly seen through the front of the bowl with strobing as depicted in Fig. 4.11. Fig. 4.12A shows a photograph of the improved feed accelerator. The bowl is rotating at 1000 rpm. Feed is introduced at 5 m3/h. The pool meter appears stationary when the strobe is tuned with a strobing frequency of 1000 flashes per minute, which is at the same rotation

Bowl

Rotation meter (free wheeling)

Strobe

Feed Axis

Accelerator

Pool Paddle

Figure 4.11 Schematic of the test rig.

Disk Centrifuge

105

Velocity meter recording feed fully accelerated

Calm (mirror finish) pool with color reflection

Figure 4.12A Improved accelerator showing quiescent pool and reflection of color from excellent feed acceleration.

Pool meter Feed Acc.

Pool Turbulence and waves in pool at feed location

Overflow weir

Figure 4.12B Conventional accelerator showing turbulence in pool from poor feed acceleration.

speed of the bowl. This means that the introduced feed stream has also attained an angular speed of 1000 rpm. The pool surface appears calm with reflected color of the pool meter. Were the pool surface being “rough” or “textured” with surface waves, color reflection could not have been possible. Both information (i.e., strobe frequency corresponding to the bowl rotation speed and the calm pool surface) confirm that the feed is indeed rotating at solid-body rotation with the bowl. Conversely, Fig. 4.12B shows a strong contrast under the same operating condition (i.e., same bowl speed and feed rate) as before, but with the pool meter appearing to be rotating backward compared with the bowl rotation. By slowing down the strobing frequency, the pool meter

106

Centrifugal Separations in Biotechnology

Velocity meter recording feed under-accelerated Turbulence and waves in pool at feed location

Figure 4.12C Close-up of conventional accelerator showing turbulence and mixing in pool from poor feed acceleration.

can be made stationary again. However, the angular velocity of the feed stream was determined to be only a small fraction of the full bowl speed of 1000 rpm. This means that the feed is not at solid-body rotation with lower rotation speed compared with the rotating bowl. This situation further deteriorated with the higher feed rate. Most strikingly, at the point where the feed was introduced, waves and turbulence showed up on the pool surface. Fig. 4.12C shows a close-up of this situation, revealing intense mixing due to mismatch of pool velocity with that of the feed stream. Also, there is no color reflection. The pool surface appeared like a rough sea with waves and turbulence. The feed acceleration efficiency is defined as the ratio of the feed stream tangential speed to the tangential speed under solid-body rotation vθ 5 ΩR. ηa 5

vθ ΩR

(4.13)

In this case, R in Eq. (4.13) refers to the pool radius Rp. Another important measure is that the G-efficiency, which is defined as the Gacceleration of the feed at the pool to that of G-acceleration at the same location assuming solid-body rotation, thus
2 

v =R vθ 2
2 5 ηa (4.14) 5 ηG 5 θ 2 ΩR ΩR It is clear from Eq. (4.11) that the G-efficiency ηG can be generated from the acceleration efficiency ηa. The significance is that bad acceleration efficiency implies very poor G-efficiency as ηG varies as the

Disk Centrifuge

107

100%

Improved feed

Improved AcceleratorAcc efficiencies η a = η G = 100 %

G-efficiency

Acceleration and

80%

Conventional acceleration efficiency

60% 40%

ηa =

Conventional G-efficiency

20%

vθ ΩR

ηG =

0% 0

1

2

3

4

5

6

2

7

8

9

vθ ΩR

10

11

Flow rate (m3/h)

Figure 4.13 Conventional versus improved accelerators on feed acceleration.

quadratic power of ηa. Fig. 4.13 plots the results from the experiment described in the foregoing. It reveals that the acceleration efficiency drops sharply with the increasing feed rate with the conventional feed accelerator taking feed rate in the range of 0
10 m3/h. On the other hand, with the improved feed accelerator design, the acceleration efficiency stays constant at 100%, independent of the feed rate. This finding contrast with the well-known centrifugal pump performance behavior wherein the delivered pressure head typically drops off with the increasing flow rate, and vice versa. Unlike centrifugal pump, the improved feed accelerator does not exhibit a trade-off of efficiency with feed rate! The G-efficiency shown in Fig. 4.13 can be generated in lieu of Eq. (4.11) by taking the quadratic power of ηa. As can be expected, ηG would be worse as it is the square of ηa especially when ηa is less than unity. At a feed rate of 6 m3/h, if ηa 5 55%, then ηG 5 (55%)2 5 30%. This implies that the feed velocity is 55% of the solid-body rotation, whereas the G-force is only 30% of that of the solid body. This is considerably disappointing given the G-force is responsible for sedimentation and separation. In contrast, the improved feed accelerator maintains at 100% for both efficiencies ηa and ηG, independent of the magnitude of feed rate. For a disk-stack centrifuge, a set of radial or curved vanes can be used to accelerate the feed stream from the small radius all the way to the large radius. This enforces a solid-body rotation at all radii from R 5 0 to the disk stack outer radius R2. Two improved designs and other features of the disk centrifuge are considered briefly in Section 4.3.5.

108

Centrifugal Separations in Biotechnology

4.3.5 4.3.5.1

Improved Feed Accelerator Improved Accelerator Without Smoothing Disk Section

It is important to accelerate the feed stream to speed in the most efficient manner as discussed earlier. This also reduces the temperature increase (due to intense shear leading to heat dissipation) during acceleration, as well as entrainment of air and oxygen that can adversely affect the process. To get feed accelerated from inlet radius R1 to an exit radius R2, a specially designed feed accelerator can be used, as shown in Fig. 4.14A. The notion on improved accelerator vane was discussed earlier [3,5]. When feed is introduced to the axis of the machine, initially it has no tangential velocity component. Suppose the direction of rotation is clockwise, and the feed is distributed into four inlet channels each with a radial velocity Vr (see Fig. 4.14A). In the reference frame of the rotating accelerator rotating at speed Ω, the feed is moving with a radial velocity Vr superimposed with a backward tangential velocity ΩR1. The resultant velocity relative to the rotating frame is then given by V1r 5 (Vr2 1 [ΩR1]2)1/2 The accelerator channel shown in Fig. 4.14A has opening aligned with the orientation to the incoming feed stream receiving V1r. Even if there is a departure in direction between V1r and that of the channel opening, at least the “angle of attack” is small and the energy loss during mismatch in entrance orientation (resulting in flow separation)

Ω R1

Channel

ΩR 1 vr R2

v1r v1r

ΩR 2

v2

Figure 4.14A This shows four inlet acceleration channels (gray) with the channel first curving backward at R1 and subsequently coming out radially at R2. The rotation is in clockwise direction.

Disk Centrifuge

109

Figure 4.14B Inlet feed accelerator design with rotating pool directed clockwise. Reproduced with permission of Alfa Laval.

is minimized. The flow stream is accelerated with the increasing tangential speed along the channel by the “pressure face” of the channel to exit at R 5 R2. At the exit, the channel is oriented radially, and hence the flow relative to the rotating channel is also directed radially outward. However, because the disk is rotating as a solid body, the feed stream in contact should also acquire this solid-body speed ΩR2. The net resultant velocity in the laboratory frame (i.e., inertial reference frame) should be a superposition of the two orthogonal velocity components, relative channel velocity V1r (assuming channel cross-sectional area at exit, the same as that at entrance), and solid-body tangential velocity ΩR2, as depicted in Fig. 4.14A. The absolute velocity in the laboratory frame is thus V2 5 (V1r2 1 [ΩR2]2)1/2. An accelerator along the idea of the foregoing discussion is shown in Fig. 4.14B. Here the direction of rotation for the accelerator is clockwise. It is clear from the figure that the accelerator is formed from a stack of circular disks with eight cut-outs or “vanes” per disk. When the disks are stacked together and aligned, the cut-outs form accelerating channels accelerating the feed to solid-body rotation to be discharged at a large radius R2. Fig. 4.14B shows that the outlet channels are radially oriented, as does the schematic diagram shown in Fig. 4.14A. 4.3.5.2

Improved Accelerator With Smoothing Disk Section

An improved arrangement is to have the vanes or channels at the exit radius R3 curved forward in the direction of rotation, as illustrated in Fig. 4.15. Note that the exit radius of the vanes/channels

110

Centrifugal Separations in Biotechnology ΩR 2

Typical tangential velocity around entire circumference

R2 Curved channel Ω R1 ΩR1 vr v1r

R3 ΩR 3

Smoothener disk no vane, no channel v3

v1r

Figure 4.15 This shows four curved inlet acceleration channels (gray) with the channel first curving backward at R1 and subsequently curving forward R2 in rotation. The rotation is in clockwise direction.

R3 falls short of the disk radius R2. So between R3 $ R $ R1, there are discrete vanes/channels, whereas R2 $ R $ R3, this section plays the role as a smoothener without vanes/channel to allow discrete jets leaving the accelerator vanes to redistribute uniformly around the circumference before discharging to the separation zone (Fig. 4.16). The tangential speed at exit of the accelerator at R3 would have a component in the tangential direction from the throughflow velocity V1r, plus the solid-body rotation tangential speed ΩR3. The net resultant velocity is the vector sum of the tangential component and the radial velocity (see Fig. 4.15). The disk region without vanes and channels for R2 $ R $ R3 can be used as a smoothener or a smoothening section to reduce the tangential velocity smoothing out the discrete streams (four discrete streams as shown in Fig. 4.15) into a multiple streams or continuous sheet of fluid at radius R 5 R2. Thus, uniformity and solid-body rotation can be both attained [3,5]. An additional benefit is that the radial velocity component of the feed at R2 is reduced [5]. This design minimizes the flow directing radially outward, which can disturb and plunge into the sediment in the solid-holding space of the bowl.

Disk Centrifuge

111 (B)

(A) CIP liquid concentrate

(C)

(D) CIP liquid concentrate

CIP liquid concentrate

(F)

(E) CIP liquid concentrate

CIP liquid concentrate

Vortex nozzle

Figure 4.16 (A) Bowl just after emptying of feed and starting of CIP. (B) Bowl rotating, flush water is introduced into the bowl. (C) Flush water filling up the bowl with water discharged both in the centrate and concentrate outlets. (D) Nozzles are opened and flush water displaces the concentrate trapped in small passages. (E) Bowl emptying pushing out concentrate trapped in the discharge passages of the bowl. (F) Bowl cleaned several cycles already with only clean wash liquid discharging through external nozzles, with exploded view showing the vortex chamber and nozzle. Reproduced with permission of Alfa Laval.

4.4 4.4.1

Other Considerations Materials of Construction

A major challenge is that the disk is used for high-speed separation as well as processing highly-corrosive suspension. Very few materials can meet the stringent requirement. High-strength stainless steel, such as

112

Centrifugal Separations in Biotechnology

duplex and higher-grade duplex, is used for the bowl and accessories construction. The liquid may contain fine particulates that are highly abrasive. Areas that are prone to wear-and-tear are protected by tungsten carbide and nickel-based alloys. Special ceramic sealing is also used when in contact with the liquid containing these abrasive fines. 4.4.2

Clean-in-Place

The objective of the clean-in-place (CIP) is to ensure that the washing from the CIP procedure provides cleaning of all areas of the centrifuge and ancillary equipment that can be reached by the processed product. Therefore, the protocol of the CIP test is to cover all possible areas in the centrifuge plus ancillary equipment (e.g., connecting piping, cyclone) with an indicator to mimic the product in real process. After CIP cleaning, the indicators should be reduced to practically an undetectable level. This ensures that the machine can be cleaned to avoid accumulation of products from the separation process avoiding cross-contamination of products. The CIP test can also identify areas that are difficult to clean, and the CIP procedures (such as repositioning or adding of wash nozzles external to the centrifuge, changing CIP rinse cycles, or rinse liquid composition and temperature) can be adjusted to address these areas. The equipment involved in CIP include dedicated CIP tank, CIP pump, flow transmitter, pressure transducers, and conductive transmitter. There is no standard procedures to running a CIP test. A tracer, for example, riboflavin that is used as a human health supplement for vitamin B2, can be used. A trace amount of riboflavin appears yellowish under fluorescence room lighting, which may be difficult to detect with naked eyes; however, it reveals as fluorescence green color under ultraviolet (UV) lighting or 473-nm laser beam. The fluorescence green color completely disappears when the concentration is reduced below a certain residual concentration, say 1 ppb. Typically, a water-based solution with 10 ppm of riboflavin is introduced into the test centrifuge at rotation speed for a certain period of time. A certain number of discharges (on concentrate outlet) should be performed. Subsequently, the test machine is stopped. The coverage by riboflavin on the machine is inspected. In some cases, additional Riboflavin solution is also sprayed onto the bowl exterior as well, especially in areas where product may or potentially contact. Subsequently, the riboflavin solution in the bowl is emptied, and the bowl and ancillary equipment are cleaned both internally and externally with wash liquid during the CIP. After CIP cleaning, the machine is dissembled; and all parts of the bowl and areas possibly wetted with the product

Disk Centrifuge

113

are being checked with the UV lamp to ensure that no trace of green color (indication of riboflavin residue) is being detected. If 10 ppm is used as the initial indicator solution, reducing down to 1 ppb represents a reduction of 1/10,000 from the initial concentration. In some practice, the initial concentration may actually be higher, say 100 ppm, and the final trace of color still needs to meet a level of 1 ppb, and this represents a more stringent requirement with 1/100,000 reduction. Other practice may elect to reduce down to 10 ppb, which goes back to the 1/10,000 reduction. Two issues are of concern in the CIP testing. First, the UV lamp indicator on green color is only a relative indication and is not an absolute measurement. Therefore a proper calibration should be conducted, which can be as follows. Samples with different concentrations of riboflavin solution, say 1 ppb, 10 ppb, 100 ppb, 1 ppm, 10 ppm, and 100 ppm are prepared, and these samples are detected with the UV lamp, resulting in different intensities of green color, which represent these different riboflavin concentrations, respectively. This “calibration curve” can then be used in the CIP testing. The second issue worth noting is that riboflavin may precipitate, or recrystallize, after drying out if there is a prolonged period between emptying of the bowl and the beginning of CIP rinse. The subsequent wash may not be able to flush out, or redissolve, the precipitated crystals. In practice, reducing the time delay between emptying of the bowl and the CIP cycle together with a warm wash liquid may help. Indicators other than riboflavin may also be used, such as total organic carbon, but the concerns discussed in the foregoing still apply. In Figs. 4.16(A)
(F), a sequence of schematics showing the CIP procedures being conducted on the internal nozzle discharge design disk centrifuge. The sequence of diagrams during CIP wash is self-explanatory. 4.4.3

Sterilization-in-Place

In the case of sterilization-in-place (SIP), superheated steam is used in sterilization. A biological indicator (bioindicator), that is, Bacillus subtilis, is introduced to the centrifuge for a certain time period. In some practice, tablets of bioindicator are placed at critical locations of the disk stack, that is, the discharge nozzle of the concentrate, the bottom of the bowl, and the entrance to the stationary skimming tube of the concentrate for the internal nozzle design. Subsequently, superheated steam with temperature 119 C
121 C is injected into the bowl, as well as from outside of the machine, for 30 minutes. Following the SIP, the machine is dissembled and test samples at different areas of machine are collected, and these samples are cell cultured in the laboratory to

114

Centrifugal Separations in Biotechnology

confirm no bioindicators (bacteria) are found. During SIP, if the temperature recorded by the temperature meters placed at specific locations record temperature below 119 C, the SIP procedure has to be repeated. However, depending on the machine design, it is difficult to ensure that all areas in the machine reachable by the products attained the steam temperature of 119 C
121 C for disinfection as the temperature in these areas may not be measured. The bioindicator tablets tagged to certain locations in the machine at least ensure that the critical locations and vicinity are being sterilized as a minimum requirement. Depending on the application and the biopharmaceutical company, the SIP procedure is carried out once every month, or once daily, or every time when a new process is being run on the centrifuge. Given superheated steam under pressure is used, the bowl should be designed that satisfies the pressure vessel code as stipulated by the certified engineering authority, such as American Society of Mechanical Engineers, or equivalent. More sophisticated control system on steam injection is required external to the machine. Therefore, other than the bowl should be built according to the pressure vessel code, the basic equipment involved in SIP include pure steam generator, pressure transducers, temperature gauges, steam relief valve, steam pressure regulator, and steam trap. The ancillary equipment used in a SIP centrifuge is much more involved compared with the centrifuge only performing CIP. CIP is practiced quite often to ensure no cross-contamination in between runs of different products, or as often as deemed necessary for good general manufacturing practice (GMP). CIP is carried out more than SIP. In some centrifuges, SIP is unavailable unless the machine is designed and equipped with such system. 4.4.4

Containment

The centrifuge should be fully contained for product integrity and operator safety. There should be double-axial seals on the bowl spindle to prevent leakage. In some designs, the seals are cooled and lubricated with circulated water, especially to withstand high sterilization temperature and pressure. 4.4.5

Surface Finish

The bowl should be in high-polish stainless steel surface to ensure effective CIP. There should be no “kink” in path of flow to avoid unnecessary shear on shear-sensitive cells (e.g., mammalian cells).

Disk Centrifuge

4.4.6

115

Temperature Control

Centrifuge bowl hood is jacketed or circulated with cooling water to ensure temperature control during solid
liquid separation. 4.4.7

Water Requirements

Water is used for cooling jacket in bowl hood. In addition, hermetic seal designs (if present), axial seals, a cyclone with cooling jacket in certain designs, and a hydraulic actuated dropping bottom for other designs all require water. 4.4.8

Noise Level

Special designed and cooling liquid jacket are necessary to ensure noise emission is under 80 db to meet industrial standards. 4.4.9

Explosion Proof Design

Explosion proof design is available with inert gas instead of air. This requires inert gas regulating unit with valves and flow switches for flow control. The explosion proof control panel is an important piece of equipment for the system. Also, a solids collector should be equipped with pneumatic pump to transport discharged solids.

4.5

Examples of Commercial Disk-Stack Centrifuge

Some commercially available disk centrifuges are described in this section. Figs. 4.17A and 4.17B show a 500-mm diameter disk stack centrifuge with maximum G of 7400g. Fig. 4.17A shows a complete module with both CIP and SIP capabilities. It can be seen in Fig. 4.17A that there is more ancillary equipment that are required for SIP than just CIP. In addition, the bowl is designed to withstand the maximum pressure up to 3 bars (300 kPa) to withstand pressurized steam in SIP. Fig. 4.17B shows the cutaway of the centrifuge. The feed is introduced at the bottom of the centrifuge through the hollow drive spindle. The feed enters the centrifuge equipped with hematic design such that the feed fills up to the rotation axis preventing air entering in the feed reducing oxidation and foaming. The feed is accelerated from the axis to the periphery of the disk stack where they are distributed to the disk stack. Separation takes place in the disk stack. The clarified liquid is

116

Centrifugal Separations in Biotechnology

Figure 4.17A Culturefuge 400 disk-stack centrifuge. This machine is equipped with both CIP and SIP. Reproduced with permission of Alfa Laval.

discharged by the rotating impeller to a stationary collector where it is pumped out of the machine by the higher pressure of the liquid, see Appendix C. The centrifuge is designed for processing cultured cells and is capable of handling large 20,000 L bioreactors or fermenters. The maximum liquid handling capacity is 333 L/min (20 m3/h). The capacity may be reduced to 70%
80% of the maximum capacity when handling delicate mammalian cell broth. Another disk-stack centrifuge capable of 15,000g is offered by another manufacturer (see Fig. 4.18). The centrifuge services 5000 to 20,000 L fermenter. The maximum solids holding capacity in the disk centrifuge is 13 L. Finally, a photograph of a steam disk-stack centrifuge is shown in Fig. 4.19. Fig. 4.20 shows a picture of the centrifuge with the control panels for use in SIP. The past several examples are used for illustration purposes and by no means are they exhaustive.

Figure 4.17B Culturefuge 400 disk-stack centrifuge cutaway view. Reproduced with permission of Alfa Laval.

Figure 4.18 Model CRA 160-576. Reproduced with permission of Westfalia Separator.

Figure 4.19 Cutaway of model CSE disk centrifuge. Reproduced with permission of Westfalia Separator.

Figure 4.20 Model CSE-170 steam disk centrifuge. Reproduced with permission of Westfalia Separator.

Disk Centrifuge

4.6

119

Summary

The principle of inclined plate sedimentation is reviewed. The same principle is extended to disk-stack centrifuge. The adverse effect due to Coriolis acceleration can be minimized with the use of spacing bars between disks, which oppose the Coriolis-driven flow, and in some designs, they serve as spacing elements between adjacent disks. Also, manual, dropping bottom, external nozzle, internal nozzle, and smalldiameter discharge disk centrifuges are described. Gentle acceleration of the feed stream and the means of discharging the separated liquid streams converting the kinetic energy to pressure head have been discussed. Semigranular and flowable concentrate can be processed. Feed acceleration tests have demonstrated significant difference qualitatively and quantitatively between good and poor feed accelerators for a centrifuge. Factoring into consideration of various issues, a proper design centrifuge can achieve good performance despite separating difficult biological cell suspensions, as it genuinely takes advantage of high centrifugal acceleration to effect separation.

References [1] W.W.F. Leung, R.F. Probstein, Lamella and Tube Settlers - Part 1 Model and Operation, I&EC Process. Des. Dev., 1983. [2] W.W.F. Leung, Lamella and Tube Settlers - Part 2 Flow Stability, I&EC Process. Des. Dev., 1983. [3] W.W.F. Leung, Industrial Centrifugation Technology, McGraw-Hill, New York, 1998. [4] W.W.F. Leung, A. Shapiro, An accelerating vane apparatus for improved clarification and classification in decanter centrifuges, Trans. Filtrat. Soc. 1 (2001). [5] W.W.F. Leung, A. Shapiro, Efficient double-disk accelerator for continuous-feed centrifuge, Filtration Sep. (1996) 819
823. [6] W.W.F. Leung, A. Shapiro, Improved design of conical accelerators for decanter and pusher centrifuges, Filtration Sep. (1996) 735
738. [7] W.W.F. Leung, Feed Accelerator System Including Feed Slurry Accelerating Nozzle Apparatus, US Patent 5,423,734, June 13, 1995, US Patent 5,651,756, July 13, 1997, US Patent 5,658,232, June 13, 1997, US Patent 5,683,343, November 4, 1997. [8] W.W.F. Leung, A Method for Accelerating a Liquid in a Centrifuge, US Patent 5,527,474, June 18, 1996.

120

Centrifugal Separations in Biotechnology

[9] W.W.F. Leung, A. Shapiro, Feed Accelerator System Including Accelerator Vane Apparatus, US Patent 5,520,605, May 28, 1996, US Patent 5,551,943, September 3, 1996, US Patent 5,632,714, May 27, 1997, US Patent 5,769,776, June 28, 1998, US Patent 5,840,006, November 24, 1998, US Patent 6,077,210, June 20, 2000. [10] W.W.F. Leung, A. Shapiro, Feed Accelerator System Including Accelerator Disc, US Patent 5,401,423, March 28, 1995. [11] W.W.F. Leung, A. Shapiro, Feed Accelerator System Including Accelerator Cone, US Patent 5,380,266, January 10, 1995, US Patent 5,527,258, June 18, 1996. [12] W.W.F. Leung, Accelerator System in a Centrifuge, US Patent 5,403,486, April 4, 1995.

Problems (4.1) For a bioseparation process by sedimenting under 10,000g, a suspension has biological cells with density of 1100 kg/m3 in a liquid with density of 1000 kg/m3 and viscosity of 0.002 Pa-s, assuming particles have equivalent spherical diameter of 10 μm, how long does it take for the particle to settle in the spacing between adjacent disks (in a disk-stack centrifuge) of 1 mm? (4.2) Repeat Problem 4.1 but for particles of sizes 3.33, 1, and 0.33 μm? (4.3) For a centrifuge with bowl volume of 8 L and 50% of which is taken up for solid storage for intermittent discharge. Using a simple approach of retention time being the remaining separation volume divided by the feed rate of 40 L/m, what is the retention time? Based on the retention time comparison with the required separation time, which of the following particle sizes (0.33, 1, 3.33, and 10 μm) can be captured by the disk-stack centrifuge? (4.4) Suppose the viscosity of the fluid is increased to 0.005 Pa-s, what is the maximum feed rate that can be maintained to capture the 1 μm particle? (4.5) A centripetal pump is used to skim the liquid pool surface at a radius of 2 cm in a disk-stack centrifuge rotating at 1000/s. The centrate liquid discharge rate is 20 L/m. The centripetal pump loss coefficient Closs equals to 0.4, and the cross-sectional flow area is 1 cm2. Determine the pressure that can be recovered above and beyond the ambient (1) without loss and (2) with loss?

5 Decanter Centrifuge For higher feed solids, decanter or solid bowl centrifuge is a better choice for handling feed with high solids concentration in excess of 5% 10% v/v. However, the maximum G-force that can be attained in decanter is much lower. As shown in Fig. 2.5, large decanter can accommodate high volumetric and solid rates, but the G-force is lower, whereas a smaller decanter takes a lower throughput, yet the G-force is higher. For example, a 750-mm diameter decanter with the length-tobowl ratio of 4:1 can attain only 3000g, whereas a 150-mm diameter bowl can go up to 6000g and higher. Some small decanter designs can actually get up to 10,000g with special support and floating bearing system. However, the G-force range is well below the maximum that can be attained by the disk-stack centrifuge. Despite this, there are some high solids feed bioprocessing applications such as separation in ethanol processing, separation of flocculated biosolids in enzyme processing, separation and reslurrying to remove either contaminants or extracting soluble products, and dewatering of biosolids to yield dry cake, in which the decanter centrifuge is preferred.

5.1

Solid Bowl or Decanter Centrifuge

Fig. 5.1 shows the countercurrent flow decanter or solid bowl centrifuge. After accelerating in the rotating feed compartment or accelerator, feed slurry is introduced to the annular pool for separation. Under high centrifugal force, the heavier solids migrate radially outward toward the bowl wall, displacing the lighter liquid to the pool surface at a smaller radius. Solids are compacted against the bowl wall to form a cake by the centrifugal force. The cake is subsequently conveyed to the smalldiameter solid-discharge end of the conical beach by the screw conveyor rotating at differential speed relative to the bowl. The cake is conveyed above the annular pool in the “dry beach,” and the liquid from the cake further drains back to the pool under G-force, resulting in a drier cake being discharged from the machine. Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00005-2 © 2020 Elsevier Ltd. All rights reserved.

121

122

Centrifugal Separations in Biotechnology

Figure 5.1 Decanter centrifuge.

A gear unit and/or conveyor drive controls the differential speed between the bowl and the conveyor. The solids retention time in the machine can be changed as necessary. The clarified liquid overflows the weirs located at the opposite end (large diameter end) of the machine. The pool depth is controlled by the discharge diameter of the weirs. The performance of the centrifuge depends on various operating variables, such as the feed rate, the pool depth, the rotation speed or G-force, and the differential speed, and they should be optimized for a given separation process. Also, stationary centripetal pumps can be installed to skim the clarified liquid, converting the kinetic energy of the liquid to pressure. This eliminates foaming when discharging liquid especially with soluble protein.

5.2

Feed Rate

The residence time of the slurry in the bowl affects centrate clarity. Decreasing feed rate increases liquid residence time and allows more efficient settling of suspended solids. With dilute suspensions wherein solids concentration is less than 1%, gravity or cyclonic thickening upstream of the centrifuge is recommended to concentrate and reduce the total volume of feed slurry or liquid to be processed. Hydraulic loading affects the main drive motor requirement from the point of accelerating the feed stream [1 3], while solids loading affects the conveyor torque load.

5.3

Pool Depth

The proper pool depth depends on the settling characteristics of the solids in the feed slurry. By reducing the pool depth, a drier cake is

Decanter Centrifuge

123

normally obtained because a longer dry beach is available for cake drainage before discharge. Pool level should not be lowered to a point where centrate clarity suffers or solids conveyability is hindered. On the other hand, when the pool depth is increased, the length of the drying beach is reduced. This generally results in higher cake moisture for granular solids where dewatering is by drainage and desaturation of the cake. For biosolids where the cake is always saturated with liquid, dewatering is by compaction and expression, thus a deep pool allows thicker cake and hence higher compaction pressure is more favorable to express liquid out of the cake. This subject will be taken up later and discussed in Chapter 7, Concentrating Solids by Centrifugation. A deeper pool improves centrate clarity, since liquid retention time is increased providing lighter and smaller particles more time to settle. Increasing the pool depth also eases transport of the cake due to liquid buoyancy, resulting in improved cake conveyability, otherwise high conveying torque results and the sediment may only be discharging intermittently. This aspect will be taken up later in the chapter.

5.4

Rotation Speed and G-Force

Higher rotation speed produces higher centrifugal force for improving settling of suspended solids in the liquid pool and subsequent deliquoring of the cake formed by the sediment. The consequence is lower cake moisture and/or a clearer centrate. However, this does not necessarily always hold. Some solids, especially the finer size fraction, have density very close to that of the liquid (i.e., nearly neutrally buoyant) due to adhesion of contaminants or bubbles to the solid surfaces. They do not settle regardless of the magnitude of the centrifugal force. Some cake drains more readily under lower centrifugal force in which larger voids exist with higher cake permeability or lower specific cake resistance. While solids that form compactible cake tend to pack tightly under high centrifugal force. Increasing G does not always warrant increasing cake dryness for the compactible cake due to equal increase in cake deliquoring resistance. For optimal operation with maximum centrate clarity, cake dryness, and least power consumption, the centrifuge should be operated at the “lowest possible” speed compatible with the process material characteristics and performance requirements. It is a good practice during the initial start-up period to compare cake dryness and centrate clarity at different rotation speeds and Gs using various driver and driven sheave combinations to drive the rotor, or better still, with the

124

Centrifugal Separations in Biotechnology

machine speed controlled by variable frequency/speed drive to provide the capability of speed/G change on the fly without limited by the fixed gear ratio of the gearbox, a subject that will be discussed next in Section 5.5. This allows selection of the optimum centrifugal forces for the specific application.

5.5

Differential Speed

By lowering the differential speed between the conveyor and the bowl, the solids residence time is increased [4]. This usually causes an increase in the cake depth piling against the bowl wall with increasing compacting stress and consequential higher cake dryness. This is often accompanied by increasing conveyance torque as well [5]. Lower conveyor differential provides less turbulence and less resuspension of solids. However, low conveyor differential may also have the opposite effect; the incoming feed solids rate is higher than the solid transport rate offered by the conveyor in which non-transported solids build up in the bowl, get entrained by the high-velocity clarified liquid, and eventually overflow with the centrate liquid. This imbalance in solids feed-to-transport rate often leads to conveyance torque gradually escalating over time. Based on the forgoing discussion, there should be a balance between the solids input and solids removal rate to prevent loss of the centrate clarity and, more seriously, solids buildup causing plugging of the centrifuge. Differential speed ΔΩ may be changed by changing the gear ratio r if available for the gearbox, or changing to a different gearbox (with different r) altogether. The differential speed is related to the bowl speed Ωb and the speed of the pinion shaft Ωp (i.e., shaft protruding from the input end of the gearbox opposite to the conveyor) by the following relation: ΔΩ 5

Ωb 2 Ωp r

(5.1)

The kinematic relationship between ΔΩ and Ωp is shown in Fig. 5.2 for a bowl (with gearbox housing attached) rotating at 3000 rpm for different gear ratios r 5 20, 40, 60, 80, and 100, respectively. A gearbox with larger gear ratio (r . 100) is rated at a higher maximum rotation bowl speed because the differential speed is much lower, generating less friction and heat and less dependent on lubrication to dissipate the excess heat generated compared with a gearbox with lower ratio and higher differential speed. Consider Ωb 5 3000 rpm, and the pinion shaft

Decanter Centrifuge

ΔΩ, RPM

100 90 80 70 60 50 40 30 20 10 0

–3000

–2000

–1000

Forward-drive

125

Ωb=3,000 RPM ratio, r 100 80 60 40 20

0

1000

Backdrive

2000

3000

4000

Ωp, RPM

Figure 5.2 Differential speed versus pinion speed for various gear ratios.

is locked stationary, Ωp 5 0, and with a gear ratio r 5 80:1, it gives a differential speed of 37.5 rpm in lieu of Eq. (5.1). An alternative is to provide an electric back-drive where the pinion is driven by a DC motor, or an AC motor, which can be controlled by a variable frequency drive. A hydraulic motor-and-pump system is also used as back-drive for centrifuge. The hydraulic motor is mounted to the bowl, while the hydraulic pressure actuates the conveyor scroll to rotate relative to the bowl at a lower speed and high torque. By controlling the hydraulic flow rate of the oil, or tuning the frequency of the AC motor, the pinion speed can be adjusted, thus changing the differential speed ΔΩ while the centrifuge is running. For example, using Eq. (5.1) when the pinion rotates in the same direction as the bowl (positive pinion speed in Fig. 5.2), respectively, 1000 and 2840 rpm, with the bowl speed Ωb 5 3000 rpm, the differential speed ΔΩ becomes 25 and 2 rpm. Both values are smaller than the ΔΩ 5 37.5 rpm under the condition that the pinion is locked stationary Ωp 5 0. The small differential speed allows longer retention time, which facilitates deliquored or dewatered cake to higher dryness. This can be easily accomplished by the conveyor back-drive (hydraulic or electric), which acts as a brake, slowing the pinion down without which the pinion would have rotated at the same speed as the bowl. The AC back-drive has a unique advantage in that it further regenerates power back to the main drive as needed. On the other hand, when there are more solids loading, the differential speed needs to be accelerated to transport the cake at a faster differential rate in lieu of the foregoing discussion. Electric drive and motor can drive the pinion opposite to the rotation of the bowl, in other words the pinion speed is negative (i.e. Ωp , 0). Obviously power is input into

126

Centrifugal Separations in Biotechnology

the system. For example, with r 5 80 and the pinion rotating at 21000 rpm, ΔΩ 5 (3000 2 (21000))/80 5 50 rpm. This increases the ΔΩ above and beyond the nominal value of 37.5 rpm when the pinion is locked. This forward conveyor drive finds important and interesting application for processing dewatering of sticky solids, which causes stick-and-slip (due to alternating high-and-low friction condition) among the conveyor, cake, and bowl. There is a maximum differential speed for a given gearbox design (not shown in Fig. 5.2) due to increasing heat dissipation at a high differential speed. Fig. 5.3 shows the case of a fixed gear ratio r 5 40, wherein different bowl speeds are considered. Eq. (5.1) can be used to calculate the differential speed for a wide range of pinion speed for a fixed bowl speed. ΔΩ versus Ωp plot is simply a straight line with a negative slope. The positive pinion speed refers to back-drive with a lower differential speed compared with locked pinion, while the negative pinion speed to forward-drive with higher differential speed compared to locked pinion. For a fixed pinion speed, higher bowl speed results in higher differential speed. The centrifuge can be overtorque due to plugging with unconveyed solids accumulating in the bowl. If temporary reducing or stopping of feed to the machine while maintaining the differential speed between the conveyor and bowl does not clear the jammed machine, the rotation speed and thus the centrifugal force need to be reduced to facilitate cake conveyance. Unfortunately, with fixed gear ratio and locked pinion, the differential speed also reduces based on Eq. (5.1). On the other hand, a centrifuge equipped with an electric or hydraulic back-drive has an advantage in that it allows the machine to adjust to the maximum differential

ΔΩ, RPM ΔΩ –3000

r =40 Bowl RPM

100 90 80 70 60 50 40 30 20 10 0 –2000

–1000

1000 2000 3000

0

Forward-drive

1000

Back-drive

2000

3000

4000

Ωp, RPM

Figure 5.3 Differential speed versus pinion speed for various bowl speeds.

Decanter Centrifuge

127

speed to get cake conveyed out of the machine despite reducing bowl speed or even when the bowl stops rotating.

5.6

Sedimentation Enhancement Using Chemical

Flocculant and/or coagulant is frequently added to the feed slurry to agglomerate fine particles improving centrate clarity. This is frequently adopted for waste treatment in which polymer dissolved in liquid stream is of less concern. It is further discussed in Chapter 8, Laboratory and Pilot Testing. As shown in Chapter 6, Commercial Applications of Centrifugation in Biotechnology, flocculant is used in extracellular enzyme production. Flocculant can help both clarification of the centrate and dewatering of biosolids. This is especially when the solids in suspension are very fine consisting of micron to submicron particles and are not the product.

5.7

Three-Phase Separation

The decanter centrifuge can also be used for three-phase separation such as two liquid phases (e.g., oil and water) and one solid phase, the biosolids. A three-phase decanter is illustrated in Fig. 5.4. Two liquid phases with different densities are discharged simultaneously. Under dynamic equilibrium, liquids and solids assume their radial location in a centrifugal field with respect to the magnitude of their densities. The liquid phases are skimmed off at different radii with the lighter of the two liquids taken off at a smaller radius. Centripetal pump or discharged weirs can be used. The centripetal pump is preferred as it converts the kinetic energy of the lighter liquid stream to pressure, avoiding the lighter phase to be emulsified by the high kinetic energy at discharge, as the

Figure 5.4 Three-phase decanter.

128

Centrifugal Separations in Biotechnology

light phase is predominantly oil phase containing small water fraction. This also applies if the heavy liquid phase is predominantly water containing small oil fraction. Also the pressurized discharge can be used to transport the separated liquid without additional pumping requirement. Another advantage of the centripetal pump is that the centripetal pumps for the two liquid phases can be adjusted on the fly to control to certain extent the quality of the discharged liquid phases. Frequently, oil and water form an emulsion regardless whether it is oil-in-water emulsion or water-in-oil emulsion depending on the percentage of each in the mixture after mixing and agitation as a result of transport and handling; chemical treatment (emulsion-breaking chemicals) and heat treatment (installing steam for preheating to break emulsion) are commonly used to break the emulsion prior to separation. A flow sheet of a three-phase separation with both employment of decanter and disk is shown in Fig. 5.5. The objective is to clarify waste so that the processed streams can be safely discharged or the liquid can be reused. The feed contains 50% oil, 20% water, 20% water-in-oil emulsion, and 10% biosolids, all of which are by weight basis. The mixture after addition of appropriate demulsifying chemicals is sent to a heat exchanger where steam is used in the exchanger to heat up the oil to 80 C 90 C. A three-phase decanter is used to carry out the separation. Upon separation by the decanter, the lighter liquid phase, being skimmed off at a smaller pool radius, contains oil and emulsion with Oil, 50% Emulsion, 20% Water, 20% Biosolids, 10%

Chemical addition

80–90°C Steam

Three-phase decanter

Oil/emulsion Wastewater, suitable for discharge, < 1% w/w oil/solids

Solids, nonfree stackable 30–50% w/w dry solids

Steam

Three-phase disk

Oil (0.1–0.5% v/v BS&W)

Solids pumpable, 10–30% dry solids

Figure 5.5 Three-phase decanter-disk combination for clarifying oil and water streams for bioprocessing.

Decanter Centrifuge

129

minimal water and solids, while the discharged heavier liquid phase (water) contains less than 1% oil and solids, and the discharged cake phase contains 30% 50% by weight of solids that does not bleed and forms a stackable pile. The light phase is further processed downstream. The oil emulsion is reheated back to the process temperature to enhance emulsion reduction before downstream separation with a three-phase disk-stack centrifuge. Subsequent to separation, the oil phase (used as product) contains less than 0.1% 0.5% w/w of residual solids and water, the water phase contains less than 1% w/w oil and solids that can be combined with the discharged water from the three-phase decanter in the upstream separation, and the cake contains 10% 30% w/w solids, which is in a pumpable form. The foregoing example demonstrates an important point about the three-phase separation with separation of two liquid phases. Between the two liquid phases, one can select to have a purified phase and the other a lesser pure phase. The lesser pure phase undergoes further treatment downstream, whereas the purified liquid product from the polishing centrifuge can be discharged or reuse. Note that it is not possible to have two purified products being produced simultaneously.

5.8 5.8.1

Cake Conveyance Dry Beach

When the cake is being conveyed up the beach, the path is neither along the steeper beach angle β nor along a helical path with helix angle α, but along a climb angle γ, which is a combination of both angles. This is illustrated in Fig. 5.6. A simple analogy is that a hiker does not walk up the mountain along the steepest slope but walk up the mountain along a zigzag, meandering path with much shallower gradient. Indeed, the climb angle can be found from geometry as shown in Fig. 5.6: γ  αβ

(5.2)

All angles in Eq. (5.2) are expressed in radians. The differential rotation between the screw and the bowl provides the conveyance, while the resistance is offered by the component of the centrifugal gravity along the climb angle, ρeffG sin γ with ρeff being the effective density of the cake. When the cake is submerged in the pool, ρeff equals to the difference between the dry cake density ρcake and the density of the liquid pool ρL, that is, ρcake 2 ρL. When the cake is above the pool, ρeff equals to the dry cake density ρcake. Therefore at the point when the cake emerges out of the

130

Centrifugal Separations in Biotechnology Helix

Pool emergence

Dry beach

G

β

P

γ

Conical beach of bowl

α Pool surface

πD

Figure 5.6 Helix, beach, and climb angle geometry in the “unwrapped” conical beach. P is the pitch of the helix and D is the bowl diameter. The schematic is somewhat distorted as D is actually reducing up the conical beach while P stays constant. The dashed line represents pool surface. Resistance to cake conveyance increases significantly at the pool emergence.

pool, the resistance force is the highest. ρeff jumps up spontaneously by a factor of ρcake/(ρcake 2 ρL). Consider the case of biosolids wherein ρcake 5 1.01 g/cm3 and ρL 5 1 g/cm3, the ratio ρcake/(ρcake 2 ρL) becomes 101. This factor is very large and is typically the situation encountered for processing biosolids where ρcake and ρL are very close to each other. Consequently, there is a significant resistance to cake emerging out of the liquid pool. Typically, the pool is adjusted so that it is close to the spillover (i.e., pool near the conical discharge) to facilitate cake discharge. However, some small dry beach area should be set aside for drainage of liquid from the cake back to the pool. 5.8.2

Hydraulic Assist

Suppose the cake does not need to be dewatered in the dry beach (out of the pool), it can be dewatered by compaction and expression, even in the liquid pool. In this situation, a more innovative solution has been found. To facilitate biosolids cake discharge in lieu of the difficulty of cake transport at pool emergence as discussed in previous section, a cake baffle or restrictive element can be used to separate the pool level across the baffle as shown in Fig. 5.7. Cake has to pass through the opening between the baffle tip and the bowl wall. The pool level downstream of the baffle is maintained by the spillover of the beach, while that of the upstream is controlled by the flow resistance of the baffle, which depends on the opening and the cake flow rate, which in turn depends on the feed solids rate. Thus the driving force for the flow to overcome the resistance offered by the cake baffle is ρeffGΔh. This is a self-regulating

Decanter Centrifuge

131

G Δh

Figure 5.7 Hydraulic assist by cake baffle. Decanter A Decanter B Poly. (decanter A)

Cake solids (wt.%)

32 31 30 29 28 27 50

100

150

200

250

300

Feed rate (L/m)

Figure 5.8 Cake dewatering comparison between two decanters, 457- and 445-mm diameter, dewatering biosolids.

mechanism as higher feed rate implies higher cake rate and consequently higher driving liquid head Δh to drive the cake through the resistance. Given the cake flow resistance depends on the opening of the baffle, the baffle opening (see Fig. 5.7) can be made adjustable depending on the process condition; this has demonstrated very useful for dewatering biosolids [6] and other applications that are shear thickening [7]. Also, an important side benefit that comes along with this arrangement is that given the cake solids is highest near the bowl wall with the highest compaction stress; as demonstrated in Chapter 7, Concentrating Solids by Centrifugation, only the driest cake is admitted downstream of the baffle. Given the cake height is higher upstream of the baffle in the cylindrical bowl section of the decanter, this maximizes the cake solids concentration before the cake is transported to the conical section and eventually out of the machine. This is because cake cannot be furthered dewatered downstream of the baffle in the conical section anyway as (1) the percolated water has nowhere to go given it cannot flow back upstream because of blockage from the baffle and (2) in the conical section the cake height is thinner, therefore there is less dewatering pressure. Fig. 5.8 compares biosolids dewatering between two decanters operating in a biowaste treatment plant. Decanter A is 457 mm diameter

132

Centrifugal Separations in Biotechnology

(length-to-diameter ratio, 3:1), whereas decanter B is 445 mm (3.5:1). Decanter A is equipped with the adjustable cake baffle technology [6] that produces high cake solids at cake discharge when compared with decanter B. At a feed rate of 170 L/m, the difference between A and B is as much as 2% by weight. This is quite significant for downstream processing. It is also noted that cake solids decrease with the increasing feed rate. This is because liquid expressing out of the thicker cake countercurrent to the compaction pressure requires time for the water to percolate through the thicker cake. Higher feed rate does not provide the needed dewatering time as higher differential speed is used to transport the cake out of the machine.

5.9

Summary

The decanter, which can accommodate high feed solid concentration, is applicable for separation, concentration, clarification, classification, and dewatering of biosolids. The machine offers unique advantage for highmass throughput separation and dewatering application with continuous feed and continuous discharge of centrate and concentrate/cake. Nominal operating G-force is between 2500 and 4000g depending on the process and centrifuge size. Coagulation and flocculation are used in processing enzyme and biological waste stream. Feed rate and bowl speed are the key variables to operate. The pool depth adjustment by centripetal pump or by weir change and the differential speed adjustment are two additional variables that affect the centrifuge performance. By monitoring the torque, which indicates cake dryness, and by making proper compensation on the differential speed by the back/forward-drive and pool level, high cake solids can be obtained. Various process issues on separation and dewatering can be overcome by operation and design of decanter. These have been discussed in the chapter.

References [1] W.W.F. Leung, A. Shapiro, An accelerating vane apparatus for improved clarification and classification in decanter centrifuges, Trans. Filtration Soc. 1 (3) (2001) 61 67. [2] W.W.F. Leung, A. Shapiro, Improved design of conical accelerators for decanter and pusher centrifuges, Filtration Sep. 33 (8) (1996) 735 738.

Decanter Centrifuge

133

[3] W.W.F. Leung, A. Shapiro, Efficient double-disk accelerator for continuous-feed centrifuge, Filtration Sep. 33 (9) (1998) 819 823. [4] W.W.F. Leung, R. Havrin, High-solids decanter, Fluid Part. Sep. J. 5 (1) (1991). [5] W.W.F. Leung, Torque requirement for high-solids centrifugal sludge dewatering, Filtration Sep. 35 (9) (1998) 883 887. [6] W.W.F. Leung, Dewatering biosolids sludge with VariGate decanter centrifuge, Trans. Filtration Soc. 1 (2001) 38 44. [7] W. Leung, A. Shapiro, R. Yarnell, Classification of fine-particle slurries using a decanter with adjustable cake-flow control and improved feed accelerator, Filtration Sep. 36 (1999) 32 37.

Problems (5.1) A 457-mm diameter decanter centrifuge is dewatering solid waste from soy bean processing, and the machine is experiencing periodic high torque. Cake is also discharging only intermittently from the conical end of the machine. The G-force at the bowl wall is at 3000g, and the pool level is set 7.62 mm below the maximum pool level (i.e., at a larger radius compared with the lip for discharging the cake), and there is a dry beach in the conical section for which the cake can be dewatered with liquid draining back to the pool. What can the operator do to reduce the periodic high torque and intermittent cake discharge? (5.2) A decanter is running with a bowl speed of 2800 rpm, with a gear box having a ratio r 5 80, what is the differential speed of the conveyor with respect to the bowl? (5.3) The differential speed is found to be too high for Problem 5.2 and it leads to wet cake, and the gear ratio can be changed using another gearbox with either r 5 120 or r 5 40, which ratio should be selected and what would be the differential speed with this new ratio? (5.4) It is desired to reduce the differential speed to 5 rpm to get a drier cake, and a AC back-drive is used to do the job. What would be the pinion speed? What direction should the pinion rotate? (5.5) The conical beach angle is 10 degrees, the helix angle can be approximated by L/(πD) with L being the lead/pitch, and D is the local bowl diameter. The maximum pool as set by the conical beach is 76.2 mm for a bowl with 457-mm diameter. What would be the climb angle at the cone cylinder junction and at the conical discharge diameter? Is the climb angle constant? Why and why not?

6 Commercial Applications of Centrifugation in Biotechnology Several commercial applications of centrifugation are commonly used in biotechnology. In almost all applications, products may be the biomass itself, a soluble extracellular component in liquid, or an intracellular component, either solid or liquid, residing in the cell. Since both fermentation and bioreactor are usually very complicated, recovery and purification of the product in the dilute form present a big challenge. Recovery and purification can take up as much as 50% of the expenses of the entire process. The more dilute the fermentation broth is, the greater is the expense on recovery and purification [1]. This chapter discusses a few process flow sheets. The flow sheets are quite generic, and they can be modified to suit a specific application. Of interest is that centrifugation plays a key separation function in the process flow sheet. Also, centrifugation may play more than one function in the flow sheet. These functions, as has been learnt from the previous chapters, include separation solid from liquid, clarification of liquid to remove fines and particulates, classification of solids by sizes and densities, and washing and separation to remove impurities. In addition, there can be multiple centrifugation stages even for the same function in the flow sheet. Finally, for illustration, an example is given with the specific feed rate and concentration, respectively, of feed and effluent of a centrifuge for use in a specific application. Many examples discussed in this chapter refer to disk centrifuge. Tubular centrifuges under high G can also be used provided the capacity of the tubular is within its limit and that semicontinuous operation of the tubular (requiring solids removal and cleaning) is factored into consideration.

Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00006-4 © 2020 Elsevier Ltd. All rights reserved.

135

136

6.1

Centrifugal Separations in Biotechnology

Generic Flow Sheet of Biopharmaceutical

In biopharmaceutical processing, such as shown in Figs. 6.1 and 1.1, there are three popular platforms of recombinant protein from fermentation of yeast or bacteria and from cell culture such as mammalian cells. In fermentation of yeast, or cell culture of mammalian cells, the protein is extracellularly expressed or secreted in the liquid. During harvest, the suspension is sent to a centrifuge for separating yeast, followed by another centrifuge for clarification of the liquid product. In the case of mammalian cells, centrifugation is typically followed by a depth filter to clarify the liquid product, see Fig. 1.1. The product can be subsequently concentrated using ultrafiltration or by evaporation, precipitation, and other means for removing excess aqueous phase. The advantage of ultrafiltration is that the appropriate buffer liquid to perform buffer exchange can be added to this intermediate product. The resultant broth is purified using column chromatography. The product goes through another round of formulation, such as crystallization, freeze or spray drying, and final sterile filtration. The latter is to remove viruses sized 0.2 µm and larger before the release of the final product. Although not shown in Fig. 6.1, it is noted that centrifuges are also commonly used to recover crystals downstream of crystallization, even in the formulation stage. Upstream process including bioreactor, fermenter

Downstream process

Extracellular product

Intracellular product

Fermentation of yeast, cell culture of mammalian cells

Cell disruption

Solid–liquid separation Concentration Purification Formulation

Fermentation of bacteria such as E. coli

{ Centrifugation extraction, filtration ultrafiltration { Evaporation, adsorption, precipitation Chromatography freeze drying, { Crystallisation, spray drying, sterile filtration

Final product

Figure 6.1 Generic flow sheet for biopharmaceutical process.

Commercial Applications of Centrifugation in Biotechnology

137

Antibiotics and monoclonal antibodies are commonly processed in this manner. World sales of therapeutic antibiotics in 1991 were estimated at 15 billion dollars, which was approximately 10% of the total world pharmaceutical market. About 80% of these sales came from five most popular antibiotics—cephalosporins, penicillins, macrolides, aminoglycosides, and tetracyclines. Sales of these five antibiotics have doubled with the exception of macrolides being tripled in 1999 [2,3]. Antibiotics have been used to treat infectious diseases caused by bacteria (Gram-positive/negative and aerobic/anaerobic), which cannot be treated otherwise by any other methods. Without doubt, antibiotics become an indispensable item in our life. The sales of antibiotics have been tripled reaching 45 billion dollars in 2019. A microbial cell, such as bacteria, is commonly used to engineer expressed intracellular protein via fermentation. Subsequently, the bacteria are homogenized to release the inclusion bodies that contain protein. An additional separation step is required to remove the protein-bearing inclusion bodies from the lysed cell debris and liquid, which may contain undesirable released intracellular materials from the lysate. This will be discussed later in the chapter. Fig. 6.2 shows three processing circuits somewhat linked together with the top diagram being the first-stage centrifugation immediately after the fermenter (for the removal of extracellular product), followed by either the removal of intracellular product in the soluble form in liquid (middle diagram of Fig. 6.2) or intracellular product in the suspended solid form such as inclusion bodies (bottom diagram of Fig. 6.2).

Extracellular products Substrate Microorganism Air

Extracellular liquid product

Extracellular liquid with soluble protein

Fermenter

Disk

Disk

Biomass waste

Wash/buffer liquid

Intracellular Products

Homogenizer

Disk

Polishing disk

Tank

Biomass

Cellular liquid with cell debris Solids waste

Inclusion bodies products

Wasted cellular liquid with cell debris

Homogenizer

Disk Solids

Liquid waste

Disk Solids

Disk Inclusion bodies

Wash/buffer liquid

Figure 6.2 Generic flow sheet with combined processing of extracellular and intracellular proteins as well as inclusion bodies.

138

Centrifugal Separations in Biotechnology

Referring to the top diagram of Fig. 6.2, the first and foremost is the circuit for processing a extracellular product in which substrate, microorganism, and air are fed to a fermenter under prescribed mixing intensity, process temperature, pressure, and time duration. During harvest, the resulting broth is sent to a disk centrifuge where the extracelluar-liquid product is removed, and the biomass is further repulped to recover liquid product that is adhered to the biomass solids. The biomass suspension is subsequently sent to a second disk centrifuge to remove the liquid (containing residual extracellular protein washed from the biomass solids). The centrate product (which has a lower concentration of soluble product) is combined with the liquid product containing enriched extracellular protein from the first-stage centrifugation. The combined liquid mixture is sent to a disk centrifuge for clarification of the liquid product, removing suspending fines and left-over biomass solids. The underflow or concentrate from the second centrifuge (top diagram in Fig. 6.2) are relurried with wash/buffer liquid. The biomass solid may contain some intracellular protein products (liquid or solid) that need to be recovered. The suspension is sent to a homogenizer in which the biological cells are lysed. Referring to the middle diagram of Fig. 6.2, assuming the intracellular protein product is soluble in liquid after lysing the cells of the biomass, the mixture (i.e., lysate) is fed to a disk, and the liquid product is removed in the centrate, while the solid concentrate is disposed as waste. In the bottom diagram of Fig. 6.2, the protein product is assumed to be in a solid form, such as inclusion bodies. After lysing the cells, the solid inclusion bodies (product) together with the cell debris and other suspended contaminants are released in the liquid for separation. The solid product is separated first by a disk centrifuge followed by washing and separation twice, using mixing tank and disk centrifuge to remove any cell debris and contaminates in the liquid stream (waste). The solid product is finally discharged in the concentrate stream of the last-stage disk centrifuge.

6.2

Mammalian Cell

In mammalian host cells such as Chinese hamster ovary (CHO) or baby hamster kidney cells, high-level expression from 10
100 pg per cell per day of the recombinant protein can be delivered [4,5]. These are used as the system of choice for large-scale production of therapeutic proteins, as they represent quite a stable cell line, expressing also high level of product proteins.

Commercial Applications of Centrifugation in Biotechnology

Liquid protein

Protein solution for Downstream processing

Disk

Depth filter

(A) Substrate Microorganism Air

Bioreactor

139

Cells waste

Figure 6.3A Mammalian cell processing.

With reference to Fig. 6.3A, microorganism, substrate, and air are introduced to a bioreactor. The cultivation cycle is typically relatively long. A bioreactor has lower mixing intensity, shear rate, and process temperature compared with a fermenter. During harvest, the broth is subsequently sent to a disk centrifuge for separation. The feed solid is typically 2%
4% by bulk volume (from spindown in a test tube). Given the mammalian cells, such as the CHO cells, have a thin membrane wall, gentle acceleration of the feed slurry is important so as not to rupture the cells, otherwise this releases the undesirable intracellular contents that contaminate along with the secreted protein product in the liquid. The separated centrifuge centrate liquid (product) may still have unsettled submicron particles, which can be largely removed by a downstream depth filter. The clarified filtrate of the depth filter with minimal particulates is sent to downstream for purification. The concentrate or underflow of the disk centrifuge with remaining spent cells is subsequently being disposed. It is best to look at the centrifuge depth filter as an “integrated system” to treat mammalian cells, rather than centrifuge and filter each carrying out their own operations. Straight batch cultures can be processed through just primary recovery by centrifugation, as the feed contains whole cells and clear fluid. As such, depth filter, shown in Fig. 6.3A, may not be required. However, fed-batch and perfusion culture clarification, which is the current common practice, takes two steps of separation: a primary recovery step by the centrifuge followed by a clarification step by the depth filter (see Fig. 6.3A). Increased cell concentration and longer culture times, which are the current trend, typically generate more cell debris, reduce cell viability, and have more organic constituents in the liquid [6]. It is best for the centrifuge to remove the whole cells and cell debris leaving the submicron fines and colloids to the depth filter, as each equipment has their niche that works best. Fig. 6.3B(i) shows the particle size distribution in the feed with the whole cells ranging between 10 and 20 µm. The cell debris ranges between 2 and 10 µm, and the submicron colloids and fines are typically in the submicron range. The disk centrifuge should be able to remove all

140

Centrifugal Separations in Biotechnology (B)

(i) Feed to centrifuge Number of particles

Submicron colloids and fines Whole cells

Number of particles

0

Cell debris 10 15 20 5 Particle diameter, µm

(ii) Effluent leaving centrifuge — feed to depth filter Submicron colloids and fines

0

10 15 20 5 Particle diameter, µm

Number of particles

(iii) Filtrate leaving depth filter

Submicron colloids and fines

0

10 15 20 5 Particle diameter, µm

Figure 6.3B Particle size distribution of suspension fed to centrifuge for primary recovery followed by depth filter for secondary recovery: (i) feed to centrifuge, (ii) effluent leaving centrifuge and feed to depth filter, and (iii) filtrate leaving depth filter.

the whole cells and cell debris making a cut size about 2 µm; see Fig. 6.3B(ii). The submicron colloids and fines can then be reported to a depth filter. Modern depth filters are equipped with two layers in series, with the first layer of depth filter having an open-pore permeable structure and the second layer having a tighter pore structure. In addition, some designs are further equipped with a 0.1-µm microporous membrane backed up by a porous support downstream of the second depth filter. Not only does the microporous membrane provide a last-defense filtration it also provides back pressure so that the incoming feed distributes uniformly over the two-layer depth filter without short-circuiting the depth filter. Fig. 6.3B(iii) shows the depth filter removing a large

Commercial Applications of Centrifugation in Biotechnology

141

100%

90.6%

99.85%

9.4% 98.4%×9.4%=9.25%

1.56%×9.4%=0.15%

Figure 6.4 Dual-stage centrifugation, with separation in first stage followed by clarification.

part of the submicron particulates leaving a much lower number of submicron particles escaping the centrifuge-filter system. While the aforementioned makes use of size exclusion for trapping slightly larger particles, smaller particles can also be trapped with the filter, as discussed in the following. The cellulose fibers of the depth filter are electrically charged. Filter aids, such as diatomaceous earth, are adsorbed to the pores by the electrical charges of the cellulose fibers. The filter aid improves the permeability of the filter. The fines and colloids, especially RNA, DNA, mammalian cell (such as CHO) proteins, and lipids, can be attracted by electrical forces, or simply by Van der Waals’ attractive force to the pores. This certainly enhances the capture capability of submicron fines and colloids.

6.3

Yeast Processing

The experience of processing yeast as a biotech source of protein is quite extensive [7] as it lends itself to the long-established practices of the brewery industry. Fig. 6.4 shows a schematic of a yeast processing flow sheet. During harvest, the suspension contains fine yeast cells of 0.5
1 µm size, and liquid with the extracellularly expressed protein is sent to a centrifuge for separation. An example is given below with some specific values on the processed streams.

Example 6.1 A centrifuge is used to separate yeast cells from the valuable liquid containing extracellular protein. Suppose the feed contains 30% by bulk volume and the centrifuge is processing at a feed rate of 30 L/m, the

142

Centrifugal Separations in Biotechnology

centrate turbidity is determined to be 300 NTU, which corresponds to 4.5% v/v solids. The centrate flow is 18.8 L/m. The material balance on the solids and volumetric balance of the three respective streams require Qf cf 5 Qe ce 1 Qs cs Qf 5 Qe 1 Qs

(6.1a, b)

Qf, Qe, and Qs are the suspension volumetric rate, respectively, of feed, centrate, and concentrate. cf, ce, and cs are the suspended solid concentration (i.e., yeast for the present example), respectively, in feed, centrate, and concentrate. Solids recovery refers to the amount of feed solids that are recovered in the concentrate stream, that is, Rs 5

Qs cs 1 2 ðce =cf Þ 5 1 2 ðce =cs Þ Qf cf

(6.2)

Eq. (6.2) is derived from Eq. (6.1a, b). If Qe, Qf, ce, and cf are known while the concentrate solids cs is not measured, it is best to approach using these variables as illustrated in the following.    Qs cs Qf cf 2 Qe ce Qe ce Rs 5 5 512 Qf cf Qf cf Qf cf Using the present example,    18:8 4:5 Rs 5 1 2 5 90:60% 30 30 This means that 90.6% of the feed solids are captured by centrifugation in the concentrate stream, while the centrate contains 9.4% of the unsettled solids. The centrate broth is sent to a holding tank for further processing. A slip stream at a rate of 15 L/m from the tank is fed to another disk centrifuge for clarification. The feed is at a solids concentration equivalent to a turbidity value of 300 NTU (4.5% solids), and the centrate leaving the clarifying centrifuge drops down to 10 NTU, which is about 0.15% v/v. The centrate flow rate is at 7 L/m. The solids recovery is thus    7 10 Rs 5 1 2 5 98:44% 15 300 This means that 98.44% of solids are recovered by centrifugation, and 1.56% of solids leave with the product centrate.

Commercial Applications of Centrifugation in Biotechnology

143

Upstream processes

Fermenter

Downstream processes

Solid–liquid separation

{Centrifugation — separation Holding tank

Clarification Purification Concentration Solid–liquid separation

{Centrifugation — clean centrate {Chromatography {Crystallization or precipitation {Centrifugation — recovery of solids

Freezing Reslurrying Final product

Figure 6.5 Yeast processing flow process.

The separation and clarification can be combined in a dual-stage centrifugation process, as shown in Fig. 6.4. The first stage recovers 90.6% of solid yeast and 9.4% solids overflow to the second stage. The second stage recovers a further 9.25% of the total solids feeding into the system (Fig. 6.4). In other words, the combined dual-stage centrifuges remove 99.85% of the feed yeast solid. A small amount of 0.15% of solid yeast leaves with the centrate product containing protein. Typically, separation has a slightly poorer recovery with high volumetric rate and higher solids loading, while clarification deals with lower feed solids and higher solid recovery with less solids escaping in the centrate. For this example, the solid recovery in the first-stage centrifugation, carrying out separation and with higher feed solid concentration, is 90.6%, while that of the second-stage centrifugation, carrying out clarification and with lower feed solid concentration, is increased to 98.4%. Fig. 6.5 shows a complete circuit for yeast processing. Downstream of the dual-stage centrifugation, the centrate may be sent to a depth filter depending on the amount of suspended solids before being routed to a chromatography column. After the protein solution is purified, it is concentrated by crystallization or precipitation. The protein crystals are

144

Centrifugal Separations in Biotechnology

washed and separated by centrifugation. The solid recovery of this step is very important as the solids now contain valuable protein, and any loss of solids in the centrate represents losing the valuable product. The concentrate is frozen for further processing. When the frozen product is taken out at a later time, it is reslurried with appropriate buffer liquid, and the process of purification, concentration, and separation is repeated until the product meets the requirement for downstream processing.

6.4

Hormones Processing

Hormones are chemicals in the bodies that regulate and control the metabolism and organ functions. Working with the nervous system, they coordinate all essential functions of the body. A common hormone, such as insulin, regulates the blood glucose of the body. Separation by centrifugation has been widely used to process insulin for use by diabetic patients. Fig. 6.6 shows a generic process of separating and related processing of hormones. Microorganism, substrate, and air are common ingredients to the fermenter. After fermentation, the broth is fed to a nozzle disk where the liquid phase is removed as waste stream, and the separated biomass concentrate discharges continuously through the nozzles. Subsequently, the biomass after being resuspended in wash/buffer liquid is sent to a homogenizer for cell disruption, releasing the solid inclusion bodies. The cellular liquid, with cell debris typically in the submicron sizes, is classified by another nozzle disk in the centrate with moderate G-force. As shown in Fig. 6.6, the concentrate with solid inclusion bodies containing contaminants is reslurried and separated by centrifugation sequentially in two stages. The inclusion bodies free from contaminants are sent downstream for further solubilizing and refolding.

Cellular liquid with cell debris (waste)

Liquid waste Substrate Microorganism Air

Cell disruption

Fermenter

Nozzle disk

Homogenizer

Biomass

Nozzle disk Solids

Wash/buffer liquid

Liquid waste

Nozzle disk

Nozzle disk

Solids Wash/buffer liquid

Figure 6.6 Hormones processing flow sheet.

Inclusion bodies product for downstream processing

Commercial Applications of Centrifugation in Biotechnology

Innoculation with genetically modified E. coli

Finish

Fermentation

Harvest — lyse cells and recover protein by centrifuge and filtration

Deep freezing

Centrifugation

Reslurrying

145

Refolding — treatment of preinsulin w/ buffer solution to active tertiary form

Enzyme clevage with trypsin

Crystallization

Chromatography

Washing and separation

Figure 6.7 Generic insulin flow sheet using centrifugation.

6.5

Insulin Production

Insulin is an important hormone controlling the glucose level of blood in human. A generic insulin production sheet is shown in Fig. 6.7. First, the bacteria, such as Escherichia coli, are genetically modified to express a specific protein under fermentation. The bacteria go through a fermentation process where temperature, agitation rate, and physiological conditions are closely monitored to ensure the process is well controlled and optimal protein is released in solution. During harvest, the bacteria cells are lysed, and the protein in solids is recovered by centrifugation followed by depth filtration. In the refolding step, the preinsulin is treated with buffer liquid to activate to tertiary form. An enzyme trypsin is used for cleavage. The product solution is purified with chromatography column. The protein is further concentrated and purified by crystallization, washing, and centrifugation. This process is repeated several times until the desired concentration and purity are reached. The resulting product is frozen to maintain freshness before final processing.

6.6

Biotech Separation of Inclusion Bodies

A suspension with E. coli cells, after being engineered and fermented (usually relatively short cycle), is sent to be homogenized so that the cells are lysed to release the inclusion bodies. This is shown in Fig. 6.8. The concentration of the feed stream is generally about 10
20% v/v (spun solids). E. coli is typically 3.5 µm by 1 µm, and the inclusion bodies are about 0.8 µm by 0.8 µm. The specific gravity (SG) of the inclusion bodies is about 1.2, and the SG of cell debris after lysing is about 1.05. The cell debris is typically less than 0.4
0.5 µm. Here the inclusion bodies of 0.8
1.2 µm (SG 5 1.2) is separated from the cell debris of 0.4
0.5 µm (SG 5 1.05) by classification. The lighter smaller cell

146

Centrifugal Separations in Biotechnology

1 μm

E. coli Concentrate

Centrifuge Homogenizer Reslurrying

Separation

Cell debris

Centrifuge Reslurrying

Separation

"Single wash"

Centrifuge Reslurrying

Separation

3–5 μm

Excess liquid

Cell debris

"Double wash"

Centrifuge Inclusion bodies

Cell debris Protein

Figure 6.8 Escherichia coli and inclusion bodies.

debris reports to the centrate, whereas the heavier larger inclusion bodies report to the concentrate. Conversely, if the cell debris also has the same SG as 1.2, then the effective size is more like 0.2 µm (50.4 [(1.05 2 1)/(1.2 2 1)]1/2) to 0.25 µm (50.5 [(1.05 2 1)/(1.2 2 1)]1/2) after making density correction. One can simplify this problem as classification only by size, rather than classification by both size and density difference. The larger inclusion bodies of 0.8
1.2 µm can be separated more readily from the cell debris of the size of 0.2
0.25 µm compared with that separated from the cell debris with the size of 0.4
0.5 µm. The overflow rate of the centrifuge is tuned such that there is minimal cell debris settle in the concentrate, and almost all cell debris leave with the centrate. In Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification, a model has been developed to quantify this interesting classification problem. The dual process of reslurrying centrifugation is repeated until little debris left behind in the centrate. Frequently two to three rounds of the dual process are required for removing the contaminants. Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of

Commercial Applications of Centrifugation in Biotechnology

147

Inclusion Body Classification also shows how the separation process can be improved by modifying the upstream homogenizer operation.

6.7

Vaccines Processing

There are two types of vaccines: a concentrated cell-based solid product and a serum liquid product, both of which can be processed by centrifugation. 6.7.1

Concentrated Cell-Based Product

The starter culture and air are introduced to a prefermenter as depicted in Fig. 6.9. The virus culture from the prefermenter is sent to the main fermenter where additional air and nutrient solution are added. Afterward, the virus culture is taken to a downstream disk centrifuge for separation. The excess liquid in the overflow is steam sterilized before discharge. The raw vaccine (concentrate from centrifugation) is sent to a downstream mixing tank where additives are added. These additives include the following: 1. Suspending fluid (e.g., sterile water, saline, or protein solution) 2. Preservatives and stabilizers (e.g., albumin, phenols, and glycine) 3. Adjuvants or enhancers that improve the function of vaccine. After adequate mixing, the mixture is sent to a freeze dryer wherein excess water is removed in the overflow centrate, leaving the concentrated vaccine product in the underflow.

Nutrient solution Air Starter culture

Liquid to steam sterilizer

Prefermenter Virus culture

Fermenter

Disk

Virus culture Water to waste

Raw vaccine Additives

Mixing tank

Freeze dryer Concentrated vaccine product

Figure 6.9 Vaccine concentrated product flow sheet.

148

Centrifugal Separations in Biotechnology Nutrient solution Air

Starter culture

Serum product for downstream processing

Prefermenter

Fermenter

Disk

Virus culture

Virus culture Biomass

Killing tank

Biomass to waste

Figure 6.10 Vaccine serum flow sheet.

6.7.2

Serum Product

As shown in Fig. 6.10, a common production of vaccine in the form of serum liquid follows similar flow sheet as illustrated in Fig. 6.9 although the fermentation and process condition may be different. After two fermentation stages, the virus culture is sent to a disk centrifuge. The serum product is in the centrate liquid, which requires further processing downstream, and the biomass captured in the concentrate is starved and deprived of food and air in a killing tank, with the dead biomass subsequently being disposed.

6.8

Enzymes Processing

There are two types of enzymes: extracellular enzymes and intracellular enzymes [8,9]. The processing of both processes is described below, especially where centrifugation plays an important role. 6.8.1

Extracellular Enzymes

Biological enzymes play a key role in expediting a given biological
chemical process. Enzymes can be produced from raw materials— substrate, microorganism, and air in the fermenter at suitable temperature and physicochemical condition. With reference to Fig. 6.11, during harvest, the fermentation broth, after the addition of appropriate flocculant to agglomerate all the fine biological solids, is separated by a disk or decanter centrifuge to a centrate liquid containing extracellular liquid product and a concentrate with the biomass. The separated biomass is reslurried in buffer or wash liquid to recover additional protein product (that could have been lost with the solids), and the resulting suspension is separated under centrifugation (disk or decanter) one more time. The overflow of the centrifuge contains diluted

Commercial Applications of Centrifugation in Biotechnology Flocculant Substrate Microorganism Air

Fermenter

149

Extracellular liquid

Disk/decanter Biomass

Wash/buffer liquid Extracellular liquid

Tank

Disk/decanter Biomass waste Extracellular liquid product

Polishing disk Biomass waste

Figure 6.11 Extracellular enzyme processing flow sheet.

liquid product from washing, and the underflow biomass is disposed of. Both centrate streams coming out of the two-stage centrifuges (the firststage centrifuge centrate containing enriched extracellular protein and the second-stage centrifuge centrate containing diluted recovered protein) are combined and sent to storage. A slip stream of the liquid from the storage tank is sent to a polishing disk centrifuge to remove suspended submicron particles. The clarified extracellular enzymes liquid product is discharged in the centrate of the disk centrifuge for downstream processing. The removed solids in the concentrate are wasted. Flocculant is required for use with both disk and decanter for the firststage separation as the biological solids are quite fine and can easily escaped with the centrate product. However, flocculant is not used for the second-stage separation as the solid concentrate discharged from the first-stage centrifuge still contain flocculated solids that can provide flocculation in the second-stage separation. Typically, decanter works best for high feed solids, while disk centrifuge works best for lower feed solids. The disk centrifuge should be used for the third-stage centrifugation, which is for clarification of the liquid product. 6.8.2

Intracellular Enzymes

As shown in Fig. 6.12, a mixture of microorganism and substrate is introduced with air to the fermenter. The broth is thickened or concentrated by centrifugation, with liquid overflowing to waste and the concentrate biomass is further reslurried with wash liquid. The resulting suspension of biomass is sent to a homogenizer for cell lysing to release the intracellular liquid product. The liquid stream containing cell debris and liquid (with intracellular protein) is separated by a high-G centrifuge

150

Centrifugal Separations in Biotechnology Intracellular liquid product

Liquid waste Substrate Microorganism Air

Cell disruption

Fermenter

Centrifuge

Homogenizer

Biomass

Centrifuge Cell debris waste

Wash/buffer liquid

Figure 6.12 Intracellular enzyme processing flow sheet.

Nutrient solution Air Freeze dried starter cultures

Microorganisms in nutrient solution

Seed fermenter

Disk

Prefermenter

Freeze dryer

Main fermenter

Packaging (O2 absent)

Wastewater Live culture product (e.g., probiotics — Lactobacilli or Bifidobacteria)

Figure 6.13 Probiotic processing.

to isolate the protein product in the centrate liquid, while the solids and cell debris are removed in the concentrate for wastage.

6.9

Probiotic Processing

Probiotics supplement consists of good bacteria for healthy living. As an example, lactobacillus is a Gram-positive, anaerobic (no air/oxygen) bacteria. These nonpathogenic bacteria are located in our bodies in the digestive, urinary, and genital systems and play important functions in regulating our body functions, including protecting against our bodies from pathogens invasion. Therefore, they have been used to treat skin disorders, diarrhea in human, and in particular vaginal infections in women. In producing probiotics, freeze-dried starter cultures are sent sequentially to a series of fermenters: seeding fermenter, prefermenter, and main fermenter, as shown in Fig. 6.13. Each fermenter operates at a slightly different condition (e.g., temperature, pH, different nutrients, and doses) to continue nurturing the bacteria during different stages of their growth. All fermenters are fed with air and nutrient solution. During harvest, the suspension containing the bacteria and the nutrient solution is sent to the disk stack where the bacteria are separated from the nutrient liquid. It is important that the bacteria are alive after separation, otherwise probiotics without viable bacteria becomes simply protein

Commercial Applications of Centrifugation in Biotechnology

151

without the benefits as prescribed. Therefore, the feed to the disk centrifuge as well as the concentrate discharge (containing the bacteria) should be gentle and not to incur high shear that can kill the bacteria. For concentrate discharge from the external nozzle disk, as the concentrate with high rotational speed is abruptly stopped by the stationary walls of the collector, the impact and shear upon the centrifuge concentrate hitting the collector wall can kill the bacteria. Therefore, internal nozzle discharge and the small-diameter top discharge (see Section 4.2.9.2), which are both gentle reduction of the concentrate, would be a more judicious choice to protect the discharge bacteria from the high shear. After separation, the liquid from the disk centrifuge containing the nutrient is sent to wastewater treatment. The concentrate from the centrifuge is sent to freeze drying where additional water is removed. The solids containing live bacteria from the freeze dryer is sent to downstream packaging. The bacteria cultures are packaged under oxygenexcluded environment and further stored under low temperature to keep the bacteria alive. This can retain the bacterial activity for several months in yogurt and other health supplements.

6.10

Aquaculture

Biotechnology has been used to produce animal and fish feed. The crude protein contents required by feeds for terrestrial livestock feeds varies from chickens requiring 18%
22%, swine 13%
26% to cattle 12%
16%. On the other hand, aquacultures require much higher protein 35%
60%. Bacteria culture can provide high protein over 80%, meeting the needs of aquacultures. A modified corn-to-ethanol process has been proposed for making aqua-food [10]. The proposed flow sheet is shown in Fig. 6.14. After saccharification, the intermediates are sent to a centrifuge for clarification. The liquid centrate containing starch is sent to an anaerobic fermenter, where two bacteria acidogenic Clostridia and Acetogens are being used to boost the yield of adenosine triphosphate (ATP) and the cell mass, which typically is low for anaerobic fermenter. In boosting ATP, the byproducts are the butyric and acetic acids. Downstream of the fermenter, the centrifugal separation produces the product protein that serves as the aquaculture feed after drying and a co-product, butyric acid. Butyric acid is a common animal feed additive as it helps to improve the gut health of animals. The advantages of anaerobic fermentation is that no large agitator, oxygen sparging equipment, pressure-rated bioreactors for steam sterilization are needed. The wet cake from the first-stage centrifuge is further

152

Centrifugal Separations in Biotechnology Corn milling Mashing Saccharification Clarification Clarified starch

Anaerobic fermentation

Wet cake

Centrifuge Modified DDG

Corn oil

Storage

Drying Butyrate salt

Centrifuge Protein

Evaporation

Drying

Storage

Storage

Storage

Figure 6.14 Aquaculture feed derived from protein produced by bacteria anaerobic fermentation using corn feed [10].

dewatered with another centrifuge to produce corn oil and dry distiller grain (DDG) for processing and storage.

6.11

Alternative Meat

An interesting application of biotechnology is to develop alternative meat derived from genetically modified soy plant [11]. Alternative beef, chicken, and fish are made by soy protein—leghemoglobin that carries heme, which is a molecule containing iron. Heme occurs in both plants and animals. Heme provides the red color and also texture that has appearance (red color from heme), taste, and flavor like beef after it has been cooked. Further advantages of alternative beef include reduction of saturated fats, sugar, and salt. By recombinant process using yeast as a host, the yeast cell can secret leghemoglobin that can be processed with the generic yeast-based flow sheet as shown in Fig. 6.5. The foregoing shows different flow sheets depicting various combinations of processing extracellular and intracellular protein; nevertheless these bioseparation processes encompass one or a combination of the following: 1. Separation of the mixture to liquid phase and concentrated solid phase.

Commercial Applications of Centrifugation in Biotechnology

153

2. Washing of solid phase to remove contaminants or recovery of liquid product attached to the solid phase. 3. Classification to remove cell debris in the overflow from the product concentrate. 4. Clarification or polishing to remove any fine solids in the product liquid stream.

6.12

Baker Yeast Processing

The baker yeast processing is an important example as some of the biotech processes also follow similar flow sheet which is depicted in Fig. 6.15. Substrate is needed for the fermentation of yeast. Preparation of substrate for the fermenter requires readily available hydrocarbon obtained from fructose derived from molasses. Preparation of molasses requires dilution and preheating followed by clarification of the molasses solution removing the suspended solids. The saturated liquid, after sterilized and cooled, is sent to the fermenter. In the fermenter, it is mixed with excess air, yeast (for seeding), and chemicals (acids for pH adjustment). After 15
16 days in the fermenter, the yeast is ready to be harvested. It is first separated by a disk centrifuge (Disk 1). The feed to the centrifuge has 5%
8% dry solids by weight. Fine debris is removed with the centrate liquid, which is sent to wastewater treatment. The concentrate discharged from the first disk centrifuge is reslurried with the Molasses, chemicals, water

Molasses dilution preheating

Molasses clarification

Sterilization, cooling

Drain/ wastewater treatment

Two-stage washing Fresh wash liquid

5%–8%w/w Fermentation, cooling, t∼30°C

Disk 1

Disk 3

Disk 2

Liquid effluent recycled as wash liquid Air, yeast, chemicals Dewatering

= Washing

Dry pressed yeast (baker yeast)

Drying Extrusion

Cream yeast ∼ 20%w/w

Yeast blocks (refrigerated) Packaging

Figure 6.15 Baker yeast processing flow sheet.

154

Centrifugal Separations in Biotechnology

wash liquid, which is the centrate from the downstream disk centrifuge (Disk 3) containing very low suspended solids. The mixture is sent to the second disk centrifuge (Disk 2) for separation. On the other hand, the concentrate discharged from Disk 2 is reslurried with fresh wash liquid, and the resultant suspension is sent to the third and last disk centrifuge (Disk 3), while the centrate discharged from Disk 2 is sent to wastewater treatment. The centrate from Disk 3 is recycled to wash the concentrate of Disk 1 as described, while the concentrate largely free from impurities and suspended solids has a creamy texture with concentration about 20% w/w. Before packaging, this creamy yeast is further dewatered by a vacuum filter, and the cake from the filter is dried pressed yeast that can be used as baker yeast without further need for refrigeration. On the other hand, crumbled yeast from the concentrate of Disk 3 centrifuge is extruded to yeast block and need to be refrigerated for freshness for future use. Given yeast is flowable, and it be best to be processed by the internal vortex nozzle disk centrifuge as discussed in Section 4.2.9.3. The vortex nozzle controls both the velocity and the volumetric flow rate given the fixed nozzle area. However, the yeast stream with higher density has the higher mass flow rate, vice versa the yeast stream with lower density has the lower mass flow rate. It becomes self-regulating. With the higher amount of suspended molasses in the stream, it is best to discharge through the external nozzle disk design as discussed in Section 4.2.9.4 to avoid clogging of the vortex nozzles. If there are too much suspended solids from the molasses leaving the harvester, the harvested stream should first go through a strainer to remove large particles before sending to the separating disk (Disk 1) to avoid clogging. It is evident that the three disk stack centrifuges, shown in Fig. 6.15, play an important role, respectively, in separation (Disk1) after the harvester and subsequent repeated stages of reslurrying and separation (Disk 2 and Disk 3).

6.13

Omega-3 From Microalgae

Eicosapentaenoic acid (EPA; C20:5,n-3) and docosahexaenoic acid (DHA; C22:6, n-3) are two common omega-3 fatty acids. Instead of extracting omega-3 fatty acids from cold-water fish oils, they can also be extracted directly from microalgae, which is food for the cold-water fish. Referring to the flow sheet in Fig. 6.16, after cultivation in open pond, race track, or photoreactor, upon maturity, the algae is removed for harvest. First, the algae suspension is centrifuged to remove water,

Commercial Applications of Centrifugation in Biotechnology

155

Isolation and purification of microalgae

Cultivation (race track/photoreactor/open pond)

Centrifugation (separation and dewatering)

Lysing (mechanical or chemical)

Centrifugation (clarification)

Cell debris

Lipid extraction

Molecular distillation/ molecular sieving

Purification

Omega-3 (DHA, EPA)

Figure 6.16 Omega-3 from microalgae.

and the dewatered algae is collected and lysed mechanically or chemically. The lysate containing the valuable product is clarified by centrifugation one more time. Lipid is extracted from the clarified liquid. Subsequently, the extract goes through molecular sieving or distillation before being purified to yield EPA and DHA. By the way, the metabolically engineered yeast, Yarrowia lipolytica, can produce tailored made omega-3 (EPA, DHA) and omega-6 (arachidonic acid (ARA), Gamma-linolenic acid (GLA)) fatty acid mixtures [12].

6.14

Ethanol Production

In the next example, the more traditional bioprocessing [13,14] whereby ethanol is produced for biofuel by dry mill corn process is discussed. Ethanol is ethyl alcohol and is produced and stored on site. It is a fuel component made primarily from corn and various other grains. It can be used for (1) an octane enhancer in fuels, (2) a nonpetroleum-based

156

Centrifugal Separations in Biotechnology Hammer mill Whole corn

200 proof

Jet cooker

Enzymes

Slurry tank 5% gasoline

CO2 Liquefaction

Water sources

Mash cooking Yeast

denatured

Molecular sieves

Fermentation

ethanol final product

200 proof ethanol

190 proof ethanol Beer

DDGS final product

Syrup Condensate

Drum dryer Wet grain

Evaporators

Distilling system

Decanter centrifuge Thin stillage

Whole stillage

Figure 6.17 Dry corn mill process.

gasoline extender, and (3) an oxygenated fuel additive that can reduce carbon monoxide vehicle emissions. Typically, ethanol is employed in its primary form for blending with unleaded gasoline and other fuel products. The ethanol production process is illustrated by the schematic in Fig. 6.17. Corn is metered to the hammermill by a computer-controlled weigh belt feeder, then ground, and pneumatically conveyed to the tank in the form of slurry for further enzymatic processing. The addition of heat, water, and enzymes further breaks the ground corn into fine slurry. The slurry is heated for sterilization and is pumped to a liquefaction tank where other enzymes are added to convert starches into glucose. The processed corn is transferred to the fermenter, into which yeast is added for a 50-h fermentation process. A vacuum distillation system can be used to separate the mash from the alcohol derived from fermentation. The alcohol is routed to the dehydration equipment, wherein 145-proof alcohol is produced from the distillation stripper. This intermediate product is passed to the rectifier from which 190-proof alcohol is produced. It is further dried to 200proof alcohol by the molecular sieve. A 5% gasoline is added to provide a mixture of 200-proof denatured ethanol product. The mash streams from the distillation stripper are sent to decanter centrifuges for dewatering. The aqueous phase (thin stillage) separated from the decanter is pumped to a steam driven evaporator. A thick syrup remains after evaporation. The wet cake from the centrifuge is transported to a rotary dryer to dry the moisture of the wet cake, producing golden

Commercial Applications of Centrifugation in Biotechnology

157

dried distillers grains with solubles. One pass through the rotary dryer produces a 50% moisture product, while repeated passes produces a 10% moisture product-Dried Distiller Grain with Solubles (DDGS). The drier product, which has a long shelf life, can be pneumatically transported to storage to cool and ready for shipment, while the heavier wetter product is shipped by trucks.

6.15

Other Biotech Processing

6.15.1

Recovery of Coagulation Factors From Blood Plasma

Coagulation factors can be purified from the outdated plasma containing other highly abundant proteins, such as albumin, immunoglobin, and fribrinogen. Plasma, after centrifugation to remove its cellular components, including erythrocytes, leukocytes, and platelets, is typically in a buffered solution based in tri-sodium citrate. Barium chloride solution is added to the citrated plasma, which subsequently induced the precipitation of barium citrate. Dissolved coagulation factors (including factors II, V, VII, XI, and X) will be adsorbed onto the barium citrate precipitate. The resultant mixture is sent to a centrifuge to separate the barium citrate precipitate on to which the coagulation factors are adsorbed. Relatively low centrifugal gravity, such as 1000 g, is needed for the separation for a period of 10
20 min at 4 C. This avoids overcompacted solids formed under high G, which would be difficult for subsequent dissolution. After removing the supernatant or centrate in continuous operation, the centrifuged solid or pellet is redissolved in buffer (usually the equilibration buffer for the ion-exchange column downstream) before buffer exchange and subsequent ion-exchange chromatography downstream for selective partial purification of the coagulation factors. At the right set of ion-exchange resins and conditions, such as pH and ionic strength, coagulation factors will remain bound to the column while highly abundant contaminants, such as albumin, immunoglobin, and fibrinogen, will flow through the chromatographic column. Serial elution of the coagulation factors from the ion-exchange column is usually done with a buffer containing high salt. 6.15.2

Tissue From Animal Cells

Proteins of interest in cells from animal tissue can be isolated and studied in the laboratory. One technique involving centrifugation is to homogenize the tissue before centrifuging the lysate. Insoluble cell

158

Centrifugal Separations in Biotechnology

membrane and debris will be collected at the bottom of the spintube, while the intracellular contents (i.e., proteins in the cytosolic fraction) left in the supernatant can be processed to recover the valuable soluble protein. (This is identical to the process as delineated in the middle diagram of Fig. 6.2.) The cell membrane left as pellet in the centrifuge spintube can be reslurried and washed several times, or as many as practical, until the contaminants are removed. At the end, detergent such as Tween 20 or NP-40 can be introduced to dissolve the membrane containing phospholipid. The resultant solution can be analyzed using standard laboratory techniques. 6.15.3

Laboratory Concentration and Buffer Exchange Using Centrifugal Filter

The centrifugal filter (see Fig. 6.18), as described in Chapter 1, Introduction, and modeled in Chapter 14, Rotating Membrane in Bioseparation, is used commonly in the laboratory for concentration and buffer exchange, which replaces the traditional concentration and dialysis steps. For example, a protein in buffer A solution is fed to the centrifugal filter equipped with the appropriate molecular weight cutoff (MWCO) membrane. (The centrifugal filter, as shown in Fig. 6.18, has a low MWCO of only 5000.) Protein with molecular weight larger than the MWCO is retained and concentrated by the centrifugal filter

Spintube

Cassette with UF membrane and retentate

Permeate

Figure 6.18 Centrifugal filter in filtering a protein buffer solution.

Commercial Applications of Centrifugation in Biotechnology

159

(membrane) with the removal of buffer A in the filtrate. Buffer B is added to the centrifuged proteins left on the membrane to displace any residual buffer A liquid, and the mixture is centrifuged again. This washing and centrifugation process is repeated several times until the protein is practically in buffer B as the suspending medium.

6.16

Summary

Centrifuges have been used for various separation functions and processes from the traditional ethanol production to therapeutic protein recovery. These include hormones, insulin, vaccines, and enzymes. It is also used in blood fractionation and various biotech separations in small and large scales. Centrifuges are also used for separation in food (e.g., baker yeast, alternative meat), health supplements (e.g., omega-3, probiotics), and aquaculture. The applications discussed in this chapter are by no means exhaustive. Typical flow sheets for separation of extracellular- and intracellularexpressed protein are discussed. Other biotech processes, such as recovery of concentration factors in blood, recovery of certain red blood cells, and/ or white blood cells, are also of great value. Other applications, such as cell tissue characterization and concentration and buffer exchange, are also carried out routinely, and they too are of great interest.

References [1] M. Shuler, F. Kargi, Bioprocess Engineering, Basic Concepts, second ed, Prentice Hall, Upper Saddle River, NJ, 2002. [2] W.R. Strohl, Industrial Antibiotics: today and the future, Biotechnology of Antibiotics, second ed, Marcel Dekker, New York, 1997, pp. 1
47. [3] D. Lowe, Antibiotics, in: C. Ratledge, B. Kristiansen (Eds.), Basic Biotechnology, second ed, Cambridge University Press, 2002. [4] M. Butler (Ed.), Mammalian Cell Biotechnology. A Practical Approach, Oxford University Press, New York, 1991. [5] R.E. Spier (Ed.), The Encyclopedia of Cell Technology., John Wiley, New York, 2000. [6] M. Pailhes, C. Lambalot, R. Barloga, Integration of centrifuges with depth filtration for optimized cell culture fluid clarification processes, Bioprocess. J 3 (3) (2004) 55
58. [7] K. Wolf, Non-conventional Yeasts in Biotechnology., Springer-Verlag, Berlin, 1996.

160

Centrifugal Separations in Biotechnology

[8] T. Godfrey, S. West (Eds.), Industrial Enzymology, second ed, Macmillan Press, London, 1996. [9] R.N. Patel (Ed.), Stereoselective Biocatalysis, Marcel Dekker, New York, 1999. [10] A. Karpol, S.W. Jones, B.P. Tracy, Single-cell proteins as aquaculture feed, Chem. Eng. Prog. (May 2019) 48
51. [11] H. Schmitz, F. D’Cruz, Challenges and opportunities in a legacy sector, Chem. Eng. Prog. (May 2019) 33
37. [12] D.M. Xie, E.N. Jackson, Q. Zhu, Sustainable source of omega-3 eicosapentaenoic acid from metabolically engineered Yarrowia lipolytica: from fundamental research to commercial production, Appl. Microbiol. Biotechnol. 99 (2015) 1599
1610. [13] M.C. Flickinger, S.W. Drew (Eds.), Encyclopedia of Bioprocess Technology: Fermentation, Biocatalysis and Bioseparation, John Wiley, New York, 1999. [14] A.L. Demain, Small bugs, big business: the economical power of the microbile, Biotechnol. Adv. 18 (2000) 499
514.

Problems (6.1) A yeast broth containing 20% suspended solids (under upset condition as the norm is more 25%
30%) is sent to a disk centrifuge for separation at a rate of 25 L/m. The centrate liquid leaves the disk centrifuge at a rate of 13 L/m with 3% suspended solids. (1) Determine the solids recovery of this first-stage centrifugation. The centrate is sent to a holding tank, and a slip stream of 12 L/m is sent to a disk centrifuge for clarification. The centrate leaving the clarifier centrifuge is at a flow rate of 5 L/m with turbidity of 1 NTU. Assuming 1% suspended solids is equivalent to 66.7 NTU as an approximate calibration, (2) determine the solids recovery of the second-stage centrifugation as well as (3) the solids recovery or capture for the entire process. (6.2) Design a flow sheet where protein is expressed in a bioreactor intracellularly in the cell liquid/plasma. The processed protein needs to be in a buffer liquid A. (6.3) Fig. 6.3 refers to mammalian cells with larger cell sizes, design a flow process with all the appropriate separation steps if yeast (6
8 µm), cell debris (2
5 µm), and submicron colloidal particles are all present in the broth and the whole yeast cells need to be separated from the debris and colloids. (6.4) Design a separation flow sheet for a biofuel other than ethanol.

7 Concentrating Solids by Centrifugation 7.1

Introduction

With a dropping-bottom disk centrifuge as shown in Fig. 7.1, the solids accumulate in the solids-holding space during the period in between solid discharge. The accumulated solids become more concentrated and compacted under high centrifugal force at the bowl diameter, and concurrently expressed liquid percolates through the accumulated solids/cake opposite to the centrifugal force toward the small diameter. In Chapter 4, Disk Centrifuge, we have discussed how the conical angle in the solid-holding space of a disk centrifuge may affect the discharge as well as compaction of solid. There are several benefits associated with discharging a concentrated shot. When the valuable protein product is in the liquid, it should be collected at the centrate and not in the concentrate or underflow. When the concentrate becomes too wet, it loses the valuable product in the rejected concentrate. In addition, handling of a very wet concentrate may be problematic as it sticks to walls and is difficult to convey unless the concentrate is being reslurried.

7.2

Concentrating Underflow

In the foregoing, the importance of concentrating the underflow is discussed. One of the key benefits of separation with extracellular protein product in the liquid centrate is that concentrating the underflow reduces product loss in the liquid with the concentrate, thereby increasing the protein yield. The following example illustrates this point.

Example 7.1 Consider a fermentation broth with 5000 L of which after spin-down the solid represent 3% v/v (bulk volume). This suspension after all contains

Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00007-6 © 2020 Elsevier Ltd. All rights reserved.

161

162

Centrifugal Separations in Biotechnology

Clarified liquid

Disk stack Solids compaction

Feed

Countercurrent liquid expression Separated solids

Feed

Bowl

Figure 7.1 Concentrate piling against solids-holding space in a dropping-bottom disk-stack centrifuge. Table 7.1 Example on compaction affecting product yield. Concentrate discharge (L)

% (v/v) solid

Bulk solid Liquid (L) carryover (L)

Yield (%)

214 214 214 214 214

70 75 80 85 90

150 161 171 182 193

98.7 98.9 99.1 99.3 99.6

64 54 43 32 21

4850 L of liquid with protein product and 150 L of solid (bulk volume). Assuming a total discharge volume of 214 L from the concentrate of the machine with a 70% v/v, this implies indeed 150 L of solids v/v and an additional 64 L of liquid carryover in the concentrate. The carryover liquid contains the valuable protein product. The product yield is the ratio of the liquid leaving the centrate 4786 L (54850
64 L) to the original 4850 L in the feed; thus, the yield is 98.7%. Table 7.1 shows additional calculation of what the expected outcome with improved compaction from 70% successively to 90% in increments of 5%, translating to product yield increase from 98.7% to 99.6%. Also, if the concentrate is too watery, it does not discharge properly and tends to hang or stick onto the walls of the bowl, and possibly the walls of the centrifuge housing outside the rotor. This may also present problems when handling the wet concentrate downstream of the centrifuge.

Concentrating Solids by Centrifugation

163

There are two mechanisms that are responsible for concentrating solid in the hold-up space of a disk centrifuge. They are compaction of granular solids (hereafter referred as grains) and expression or percolation of liquid in the porous bed, both taking place simultaneously, as depicted in Fig. 7.1 for a disk-stack centrifuge.

7.3

Compaction

Compaction can be understood with a simple analogy. Imagine five acrobats standing on top of each other building up a “human ladder,” as shown by Fig. 7.2 (right diagram). The top acrobat has no weight from above imposed upon him, the second acrobat bears the body weight of the top acrobat, the third bears the weight of the two above, the fourth supports the three above, and finally, the fifth supports the weight of all four acrobats above. Assuming that the weight of each acrobat is the same, the loading would then be evenly increasing from top to bottom in the direction of the gravitational acceleration. Likewise, it is expected that the equivalent hydrostatic pressure for fluid, or more precisely the compaction pressure for solids, increases nonlinearly with increasing thickness of the concentrated layer in the holding space in the disk-stack centrifuge. This increase in pressure translates to an increase in solid concentration through compaction and re-arrangement of solid grains in the bed. The solids concentration at the bowl wall increases first rapidly with increasing pressure but subsequently increases gradually at diminishing rate until it finally levels off to an asymptote (see left diagram of Fig. 7.2).

0

Solids concentration

A

Pressure

Concentrate thickness

Solids concentration at bowl wall

A

Wall B

B

Pressure

Figure 7.2 Schematic representation of a human pile, one standing on top of another.

164

Centrifugal Separations in Biotechnology

7.4

Expression or Percolation

Liquid flows through a porous medium under the influence of a pressure gradient. According to Darcy’s law, the superficial fluid velocity in a porous medium is proportional to the driving pressure gradient and is inversely proportional to the viscosity of the fluid. (Note the superficial velocity is basically the flow rate divided by the cross-sectional area of bed, ignoring the fact that fluid flows in the tortuous porous path between grains of the bed.) Thus, v5

Q K dp 52 A μ dz

(7.1)

The fluid can be either gas or liquid. The proportionality constant K of Eq. (7.1) is the permeability of the porous medium and has a unit of m2. The permeability of porous medium can span quite a wide range from 10210 m2 for unconsolidated porous media to 10216 m2 or less for low-permeability consolidated medium. Referring to Fig. 7.3, the pressure drop has to be smaller in the direction of flow or increasing z (i.e., toward the screen). Let us examine a slice of the concentrate medium in the holding space as represented in Fig. 7.4, where the centrifugal acceleration G is directing solid compaction toward the bowl wall (similar to Earth’s gravitational acceleration g compacting the “acrobat ladder” toward the ground). The solids will be more packed toward the wall in the direction of the G-force and likewise less packed in the direction opposite to the G-force. The hydrostatic liquid pressure is therefore highest near the wall, approximately ρGh 1 pa, where pa is the overburden pressure acting on the surface of the porous medium or cake and h is the concentrate thickness. Applying Darcy’s law Eq. (7.1) to determine the liquid flux v, A

Porous cake/media

z –dp/dz

Liquid permeate Screen Q

Figure 7.3 Darcy’s law on percolation.

Concentrating Solids by Centrifugation

165

h(t = 0)

Space free-up from cake consolidation

h(t)

Liquid percolation

G

Solid bowl wall

Figure 7.4 Darcy’s law on percolation and concurrent compaction.

v5

Q K ðpa 2 ½pa 1 ρGhÞ K 2 5 ρG A μ h μ

(7.2)

The liquid flux v is independent of the overburden pressure pa as it is present at both the surface and at the bottom of the media. The pressure gradient being the difference of the pressure at these two locations becomes independent of pa. Let t be the time for liquid to percolate across the concentrate, treated herein as a packed porous bed with thickness h. Thus, the velocity of percolation is of the order, v

h t

(7.3)

Combining Eqs. (7.2) and (7.3), the time of percolation or expression of liquid can be estimated as, t

μh ρKG

(7.4)

It can be inferred from Eq. (7.4) that higher bed permeability K, higher centrifugal gravity G, lower liquid viscosity μ, and smaller bed thickness h all lead to shorter time for liquid percolation t. Example 7.2 illustrates the use of Eq. (7.4).

Example 7.2 The concentrate is semi-permeable with the following properties: Kinematic viscosity μ/ρ (cm2/s) h (cm) G/g g (cm/s2) K (cm2) t (min) Note: 1 D (Darcy) 5 1028 cm2.

0.05 (5 3 water at room temperature) 5 10,000 981 10210 4.25

166

Centrifugal Separations in Biotechnology

Thus, it takes over 4 minutes for the viscous liquid to percolate through the bed. If the discharge frequency is much faster than this time, the liquid has yet percolated through to the surface of the porous bed, and the concentrate would be wetter as it carries this additional moisture. This also implies that the protein yield (in liquid) in the centrate is reduced. The time for compaction, as measured by reduction in h(t) over time t, for biological solids is usually faster compared to the time of liquid percolation. This can be determined in the laboratory using the procedure as described below.

7.5

Compaction Testing

As shown in Fig. 7.5, a specially designed bucket centrifuge [1] has been used to determine the compaction and expression of a concentrated biological suspension. The bucket centrifuge diameter is 30 mm. The bucket can be spun at different times, G-level, and feed solid consistency to establish various concentrate thicknesses. The large-diameter end of the bucket centrifuge is equipped with a removal cap. After the end cap is removed, the concentrated cake sample can be pushed out of the bucket centrifuge using a plunger having a diameter slightly smaller compared to the inner diameter of the bucket centrifuge. The sample should maintain intact during removal, and it can be dissected in segments along the radial direction (i.e., along the bucket centrifuge axis) to determine the solid concentration of each segment as a function of radius from which solid pressure/stress under centrifugal loading can be inferred. Fig. 7.6 shows a general behavior of solids concentration Ws of the biosolids that can be concentrated by centrifugation under increasing compaction stress ps and increasing time. For a given time t, the solids Ws increase with increasing ps, first linearly and subsequently with diminishing return. Likewise, for a given level of ps, increasing time increases both consolidation and liquid expression.

Compaction

G

End cap removable

Dissected cake Liquid expressed sample after test

Axis

Figure 7.5 Bucket test in determining compaction and percolation.

Concentrating Solids by Centrifugation Ws

Equilibrium

167

t∞ t3 t2 t1

Increasing time t ps

ps1

0

Figure 7.6 Solids concentration Ws under compaction stress and increasing time. .4 NORTH ATTLEBORO SEWAGE 50/50 PS/WAS WITH PERCOL 757 , 7.9#/T

THIN CAKE (≤25 mm) RPM TEST CAKE A 27.9 mm 3,000 13.0 mm 3,000 B 17.0 mm 2,000 C D 17.3 mm 1,000

Ws (w/w)

.3

A

.2 B

THICK CAKE (66 - 94 mm) rpm LUB. 3,000 rpm UNLUB. 3,000 rpm LUB. 2,000 rpm UNLUB. 2,000

DC

.1

0

0

5

10

15 ps (kPa)

20

25

30

Figure 7.7 Test results on biological materials in a bucket centrifuge.

Fig. 7.7 shows some results from testing a biological sample from a wastewater treatment plant under large compaction time (i.e., under equilibrium where the kinetics have died out, i.e., t-tN as shown in Fig. 7.6). Both thin cake samples with thickness 13
28 mm and thick cake samples with thickness 66
94 mm have been tested. The tests were conducted with and without lubricating (grease lubricant) the bucket wall. It was found that the effect of lubrication was negligible.

7.6

Compaction Pressure

As the concentrate media is compacted, force is transmitted from solid grain to solid grain, as depicted in Fig. 7.8. A loose structure with wide

168

Centrifugal Separations in Biotechnology

ps Grain

Figure 7.8 solid pressure or stress transmitted across grain-to-grain contact.

opened pores can be compacted under body force to a much tighter structure with less porosity and lower flow permeability. The solids stress or pressure ps can be expressed as an integral, ð (7.5) ps ðRÞ 5 ðρs 2 ρL Þ Ω2 Rφs ðRÞdR where φs(R) is the solids volume fraction in cake at radius R, Ω is the rotational speed, ρs is the solids density, and ρL is the liquid density The integral Eq. (7.5) can also be approximated by a finite sum X ps ðRÞ  ðρs 2 ρL ÞΩ2 Ri φsi ΔR  ðρs 2 ρL Þφsa Gh (7.6) i

Where φsa is the average solids volume fraction in cake column, G is the Ω2Ra, h is the cake height, and Ra is the average radius. 7.6.1

Test-Tube Compaction

The solids pressure has been determined for the data of test-tube centrifuge displayed in Fig. 7.7 via Eq. (7.6), and the data are found to be well correlated by a power law as depicted in Fig. 7.9. Thus, Ws φ 5 s 5 ps n Wso φso

(7.7)

where n 5 0.23 for the data set shown in Fig. 7.9. Note that this correlation does not depend explicitly on the cake thickness and G or rotation speed Ω. 7.6.2

Decanter Compaction

Solids concentration by weight as a function of compaction pressure has been obtained for some cake samples from a continuous-feed decanter processing biosolids [2]. The results for two different test runs using, respectively, 2000 and 2500 g, are charted in Fig. 7.10. As can be seen,

Concentrating Solids by Centrifugation NORTH ATTLEBORO SEWAGE 50/50 PS/WAS PERCOL 757 , 7.9LB/TON Gb = 2085 g

–0.2

log10Ws

–0.4

RPM LUB/UNLUB

3000 3000 3000 3000 2000 2000 2000 1000

L L L UL L L UL L

h(IN) 1.11 0.52 2.60 2.70 0.67 3.00 3.70 0.68

169

TEST

A B C D E F G H

30% W/W

Ws

–0.6

=

Wso

φs φso

= ps0.23 FA GC DC B C F D C FD G F GF G G E GD

–0.8 H

–.1 –1

G

FG

0

C D

1

2

log10 ps (kPa)

Figure 7.9 Biosolids by weight correlated with compaction pressure. 40

2000 g 2500 g Power (2000 g) Power (2500 g)

2000 g y = 9.1095x0.2548

Ws (%)

30

2500 g y = 3.8935x0.4661

20 10

100

1000

ps (kPa)

Figure 7.10 Cake solids by weight versus compressive solid pressure or stress from cake sample obtained from a continuous fed 460-mm diameter decanter on biological sludge.

the cake solids for the 2000 g testing indeed follow a power law relationship where the power index is 0.25. This is similar in result to that obtained from the test-tube centrifuge. However, at 2500 g, the power index increases to 0.46, which shows more response of cake compaction with higher G for a continuous-fed centrifuge. This is because the cake is confined by the side walls of the test tube during compaction regardless of what the G-force is, while the cake is not confined by physical side walls in a continuous-fed centrifuge during cake compaction [1].

170

7.6.3

Centrifugal Separations in Biotechnology

Considerations of Cake Compaction

Despite the difference in results between the test tube and continuous centrifuge, the test-tube experiment is still useful providing a conservative estimate of cake compaction. The samples used are much smaller, and testing can be controlled much more readily. It can be used as a first estimate of what to be expected; however, the actual centrifuge test would provide a more accurate measurement of how dry the cake can achieve. Bear in mind, a larger size centrifuge would have higher cake solids than a smaller centrifuge because of thicker cake pile in the solid-holding area. One final note is that irrespective of test-tube or continuous-fed centrifuge, the cake solids achieved by compaction and expression from biosolids cake increases with increasing compaction pressure, first linearly and subsequently with a diminishing return. It can be approximated by a power law. This pretty much confirms of what we have discussed in the two concurrent mechanisms of dewatering of biosolids cake: cake compaction and expression.

7.7

Recommendations for Increasing Solid Concentration in Underflow

There is a good basis for reducing solids concentration in the concentrate: one reason is to reduce loss of valuable liquid product and the other reason is to reduce liquid content, thus reducing the bulk volume for downstream processing, especially if the concentrate is the product. Even if the concentrate is a waste, it is desirable to reduce the water content, which also reduces both mass and bulk volume for transport, handling, storage, and downstream processing. The following are recommendations and concerns for increasing the concentrate solids: • Increase solids residence time for thicker drier cake to effect compaction and liquid expression. • Concentrate depth should not be too thick, such that it interferes with separation zone (disk stack or pool surface for tubular centrifuge). • Concentrate cannot be too concentrated, such that it presents discharge problem and plugs the openings of the discharge port of the centrifuge. • Concentrate should not be distributed unevenly around the circumference leading to unbalance and mechanical vibration, as well as uneven dewatering of the concentrate/cake layer. The tubular centrifuge allows the sediment or concentrate to be compacted as more settled solids from the incoming feed are added until the sediment becomes sufficiently thick. The solids adjacent to the bowl

Concentrating Solids by Centrifugation

171

wall are getting concentrated due to (1) liquid expressing out of the layer and moving away toward the small radius and (2) increasing compaction pressure from thicker concentrate/sediment. Feeding stops when the concentrate layer becomes too thick which interferes with the effluent or centrate, resulting in entrainment of solids in the centrate. The liquid pool is subsequently drained off and the concentrate is removed by mechanical plough or an automatic plunger as described in Chapter 3, Batch and Semi-Batch Centrifuges. The drainage of pool liquid prevents the latter from mixing and further wetting the drier concentrate. A special disk centrifuge without disk stack also operates in a similar way as a tubular centrifuge. After cake has built up to a certain thickness, feeding stops. The centrifuge is slowed down and the pool liquid is subsequently drained. Nitrogen can be used to blow-dry the concentrate to provide sterilization. Subsequently, the centrifuge ramps up to speed again to establish the G-force and the bowl bottom opens to discharge the concentrate. The cycle time is longer than a regular disk centrifuge without draining the pool. The concentrate discharge is drier without any pool liquid and with special provision as described.

7.8

Summary

To achieve a high protein yield during separation, it is a must to minimize the loss of liquid in the concentrate/underflow/cake discharge. The underlying physics on cake compaction and percolation of liquid through the cake has been discussed. Cake compaction depends on the stress developed, which in turn depends on the G-force and the concentrate thickness. There is general increase in solids concentration initially with increasing solids stress to be followed by diminishing return. The kinetics of dewatering dictates whether there is adequate residence time for the cake to drain the trapped liquid; this is especially for the case of intermittent discharge disk that operates on a time cycle. Furthermore, we have shown some experiments that have been carried out on cake compaction and percolation to determine the compaction and kinetics of dewatering, and how these relate to continuous feed centrifuge. The dewatering testing can be carried out similarly for any bioprocess.

References [1] W.W.F. Leung, Industrial Centrifugation Technology, McGraw-Hill Inc, New York, 1998. [2] W.W.F. Leung, Dewatering biosolids sludge with varigate decanter centrifuge, Trans. Filtration Soc. 1 (2001) 38
44.

172

Centrifugal Separations in Biotechnology

Problems (7.1) Sediment after being separated in a disk stack operating at 5000 g gets compacted in the solid-holding space of a dropping-bottom disk-stack centrifuge. The concentrate thickness is 3 cm. The viscosity is 0.03 P. The density of the biosolids is 1 g/cm3. The permeability of the sediment is 8(10211) cm2. What is the time for liquid percolation across the concentrate? If the time between adjacent discharges is 4 min, would this be sufficient? Why and why not? (7.2) The concentrate thickness remains 3 cm as with Problem (7.1). The liquid viscosity is increased to 0.1 P due to increase in the soluble matters. Density of the biosolids is 1 g/cm3. The permeability of the sediment is 8(10211) cm2. What is the rotational speed in rpm to operate a 500-mm diameter disk centrifuge, such that the time in between shots is equal to the percolation time across the concentrate of 5 minutes? (7.3) A compaction testing under an angular speed 300/s has been conducted using a bucket centrifuge in the laboratory. This is similar to the one described in Section 7.6. The density difference between the biosolid and liquid is 0.1 g/cm3. The concentrate has been dissected along the radius in 1 cm increments. Table 7A shows the test results. Table 7A Compaction result from a bucket centrifuge test. Sample

R (cm)

φs (
)

1 2 3 4 5

5 6 7 8 9

0.05 0.07 0.08 0.085 0/088

Δps

ps

(1) Determine the relationship of the solids volume fraction versus the compaction pressure. Plot in log
log scale the solids volume fraction φs versus the compaction pressure ps. (2) How does this result compare with that shown in Fig. 7.9? (3) What is the exponent if the correlation is based on a power law between φs and ps?

8 Laboratory and Pilot Testing In this chapter, the objectives of centrifugal separation are reviewed. This is followed by discussion of characterization of solids, liquid and suspension. Bench scale and pilot tests will be presented for biotechnology separation.

8.1

Process Objectives

There are six process objectives in biotechnology separation. 1. Separation—Suspended solids are separated from the liquid phase. The solid phase can be the valuable product being recovered in the concentrate and the liquid a waste, or the solid phase can be the waste while the liquid is the product recovered in the centrate or overflow. 2. Clarification—The liquid phase or broth is the product and as such the amount of suspended solids should be minimized. This application may take place wherein the centrate from the first-stage centrifugation is sent to the second-stage centrifugation to further remove fine dispersed suspended solids. For example, a suspension with solids concentration 300
1000 ppm can be centrifuged (in the second-stage centrifugation) with centrate reduced to 10
20 ppm. 3. Dewatering/deliquoring of concentrate—This is to remove additional liquid in the concentrate. Some processes require ultra-dry concentrate for special downstream processing, while others demand high protein yield with minimized loss of liquid (with dissolved protein) in the concentrate discharge. 4. Washing or repulping concentrate—This can be used to strip off valuable protein adhered to the solids (cells or crystals) in an effort to increase the protein yield of the separation process, but this Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00008-8 © 2020 Elsevier Ltd. All rights reserved.

173

174

Centrifugal Separations in Biotechnology

increases the liquid amount for downstream processing as shown in certain flow sheets in Chapter 6, Commercial Applications of Centrifugation in Biotechnology. 5. Classification of debris—This is to remove cell debris from the released inclusion bodies after bacteria (such as Escherichia coli) is lysed. The inclusion bodies need to be recovered in the concentrate while the finer cell debris should be removed in the centrate/overflow. There are other applications where separation of particles can be both accomplished by difference in particle sizes and/or difference in densities. 6. Filterability and properties of protein—The objective is to determine whether the protein solution is filterable, the filtration rate, and associated resistance from osmotic pressure under low transmembrane pressure and gel resistance under high transmembrane pressure, respectively. The gel concentration and diffusivity of protein can also be determined.

8.2

Solid, Liquid, and Suspension Properties

It is important to characterize the properties of solids, liquid, and the suspension as a whole. 8.2.1

Solids Properties

1. The density of solid density is not important. Instead, what is more important is the difference in density between the suspended solids and liquid. The greater is the density difference is, the greater impact it has on separation. The specific density of biosolids does not differ greatly from that of water, hence the driving force is very small. Centrifugation with high G can certainly compensate to a large extent on such small density difference. 2. The shape of the solid also plays an important role in settling and consequently separation. Solid is typically amorphorous, yet thin elongated-shaped solids have some interesting and peculiar settling properties, with the longer axis settling much faster than that of the shorter axis [1]. 3. When solids are suspended in liquid, charges can be induced on the solid surface from ions presence in the liquid. The zeta potential provides a measure of the electrical charges on the solid surface [2]. 4. The particle size distribution (PSD) is extremely important. The sedimentation rate of particles in a stationary liquid varies as the second

Laboratory and Pilot Testing

175

power of the particle size. While larger particles have no problem to separate, small particles have difficulty to separate. Hence, it is important to know the particles size and their distribution in the suspension to be separated by the centrifuge. 8.2.2

Mother Liquid Properties

1. The density of mother liquid affects separation. Typically, the liquid in biotech separation is largely aqueous or water based. 2. The viscosity of liquid, as shown in Chapter 2, Principles of Centrifugal Sedimentation, is a function of temperature. Increasing temperature often reduces the viscosity and enhances the separation rate. Viscosity of the liquid can be increased with increase in soluble intracellular materials from cell lysate, such as RNA and proteins. 8.2.3

Feed Slurry Properties

1. The viscosity of suspension depends on the viscosity of the liquid and the concentration of suspended solids. Increasing the concentration of suspended solids and/or higher liquid viscosity leads to higher suspension viscosity. 2. In addition, the pH or hydrogen ion concentration of the suspension affects the acidity (or alkalinity) of suspension. 3. The ionic strength of suspension affects the charges (zeta potential) of particles and consequently particle
particle interaction in the concentrate.

8.3 8.3.1

Bench-Scale Testing Separability

The separability of solids in a suspension is of great interest. Does the sample separate? If so, how much time and under what G-force would the solid phase be separated from the liquid phase? Furthermore, is the supernatant (i.e., centrate or overflow) cleared of suspended solids? How much suspended solids are recovered by centrifugation? In addition to separability, what is the solids concentration in the concentrated underflow and how does it compare to that of the feed? What is the factor of concentration? As discussed in Chapter 3, Batch and Semi-Batch Centrifuges, spintubes are common centrifuges used for laboratory testing. The time duration that it takes to fill up the solids holding space for intermittent discharged centrifuge can be easily estimated

176

Centrifugal Separations in Biotechnology

from spintube testing. This determines the discharge frequency for the intermittent discharge disk centrifuge. (Note the dewatering time should also be factored into consideration on discharge frequency, see Chapter 7: Concentrating Solids by Centrifugation.) 8.3.2

Flocculant and Coagulant in Bench Tests

In some bioprocess, especially when the protein product is in the liquid, flocculant can be used to assist agglomerating finely dispersed biological solids. Flocculants are polyelectrolytes with long chain polymers with charges in the chains serving as an agglomeration agent or “mop” from which fine particulates, which are difficult to separate, can be attached. The flocculants can be anionic (negatively charged), cationic (positively charged), or even nonionic. The polymer is selected that is compatible with the liquid solvent without adverse effect on the dissolved protein, enzyme, and downstream processing. On the other hand, coagulants are simple electrolytes—acids, bases, and salts. Upon dissolution, the inorganic ions from these electrolytes, such as sodium ion, magnesium ion, or aluminum ion, can be used to neutralize outstanding charges (typically negative that stay on particle surfaces) and allow particles to agglomerate due to the attractive Van der Waals’ force. Most particles and colloids carry negative charges, consequently positive ions from these electrolytes are very effective to neutralize the charges from these particles. In essence, upon charge neutralization the “charged double layer” surrounding and keeping the particles in repulsion collapses. The effectiveness of these ions is in the order as stated because ion with large positive charges, such as Al13 ion, can neutralize more outstanding charges than a smaller Mg12 ion, which in turn is better than an Na1 ion, which carries less positive charge. At times, coagulants followed by flocculants are both used. Flocculants have been used in separating extracellular enzymes (see Chapter 6: Commercial Applications of Centrifugation in Biotechnology) in which fine solids are collected by these long chain flocculants to report to the underflow, leaving the product enzyme free from suspended solids. In the laboratory it is of interest to determine what type of flocculant and coagulant are appropriate for the process and at what polymer dosage (measured by kilograms of flocculant per tonne of solids) to be treated? A systematic series of spintube testing can address the above issues. Selection of coagulant and flocculant depends on the type of solid, pH, ionic strength of liquid, and process temperature. Through

Laboratory and Pilot Testing

177

systematic experimentation, optimal coagulant or flocculant dosage can be derived to obtain reasonably sized flocculated solids (or commonly referred as flocs) that are strong enough to withstand shear forces in a centrifuge. In this process, proper mixing is required to provide the needed energy for flocculation. Also, proper dilution of flocculant is important especially if they are in solid or condensed emulsion form. The polymer is folded up and needs to be properly diluted with liquid (typically water) so that the polymer chain and functional groups are fully extended out to effectively capture the biosolids and fines. It is best to follow the manufacturer’s specifications, with some laboratory dilution trials for fine tuning. In carrying out spintube tests, different Gs and ts and chemical dosages on coagulant or flocculant, should be tested. Some extreme values of G, t, and chemical dosage should also be tested to determine operation and performance beyond the normal range. 8.3.3

Test Variables

The variables in spintube testing comprise the following: • • • •

centrifugal gravity G, time duration t, feed solids concentration cf, and flocculant type (anionic, cationic, nonionic) and dosage D and coagulant type and dosage D.

G and t are the most readily tested variables, as all modern bench centrifuges can allow readily programmable manual to adjust these two variables. Feed solids can be varied by either dilution with the appropriate buffer liquid or concentration by filtration. Both coagulant and flocculant types and dosage require trial-and-error using a standard mixing device, with several bench mixing devices running concurrently for making comparison. 8.3.4

Material Balance

The separation efficiency or solid capture efficiency Rs depends on G, t, D, and cf. The solid recovery is a function of these four variables. Furthermore, the solid recovery under steady state can be expressed as function of the feed solid concentration cf, centrate solid concentration ce, and concentrate solid concentration cs. The cs are the concentration of solids by

178

Centrifugal Separations in Biotechnology

bulk volume. The commonly used method to determine the bulk volume is by spindown of the sample (feed, centrate/supernatant, or concentrate) under 10,000 g for 3
5 minutes. Note that the separation test referred to below has much shorter time of centrifugation as it is supposedly to determine if the sample, which has much shorter residence time than 3
5 minutes in the centrifuge, can be separated despite the Gs, are comparable in both cases. Alternatively, cs can also represent number of biological cells per unit volume in which a totally different measurement needs to be carried out and not by spindown of the sample. 8.3.4.1

Material Balance Consideration for Bench Scale

With reference to Fig. 8.1, before spinning the contents of the spintube should be homogeneous with suspended solids concentration cf occupying volume Vf. After spinning, it is separated into two phases, a lighter phase (supernatant) with volume Ve with solids concentration ce, and a heavier phase (pellet) adjacent to the bottom of the tube with volume Vs and with solids concentration cs. Balancing total volume and solids respectively before and after spinning Vf 5 Ve 1 Vs

(8.1a)

Vf cf 5 Ve ce 1 Vs cs

(8.1b)

Rearranging Eqs. (8.1a) and (8.1b), the volumetric recovery (i.e., volume recovered in concentrate versus original feed volume Vs/Vf) Vs cf 2 ce 5 Vf cs 2 ce

(8.2)

Ve

Vf

Vs t=0 Feed

t Centrate

Sediment

Volume, mL Vf

Ve

Vs

cf

ce

cs

# cells/mL

Figure 8.1 Material balance of a spintube before and after spinning.

Laboratory and Pilot Testing

179

The solids recovery Rs becomes Rs 5

cs Vs cs cf 2 ce 1 2 ce =cf 5 5 cf Vf cf cs 2 ce 1 2 ce =cs

(8.3a)

Also, when the centrate concentration is very much smaller than that of the concentrate Rs  1 2

ce ðce {cs Þ cf

(8.3b)

The concentration factor CF becomes, cs V f  cf Vs Vf cs 2 ce cs as 5  ðce {cs and ce {cf Þ Vs cf 2 ce cf

CF 5

(8.3c)

Typically, when the sample is spun at 10,000 g for t 5 3
5 minutes, one can determine Vf/Vs, which also equals to the concentration factor CF. For a given G, the solids recovery increases with increasing centrifugation time. For a fixed time, the recovery increases with increasing G. This is illustrated in Fig. 8.2A. Vice versa, when Rs is plotted against G with t as a parameter, as shown in Fig. 8.2B, similar behavior results. In fact, Rs can be well correlated with the multiplication product Gt [3], as depicted in Fig. 8.2C. Flocculant may be used in applications in

(A) %R

(C)

%R

100

100

G

Gt

t %R

(B) %R 100

(D)

100

t

Inc. Dosage or dec. cf G

Figure 8.2 (A
D) Sketch of centrifugation behavior.

Gt

180

Centrifugal Separations in Biotechnology

which the product is the liquid phase and solid the waste. The solid
recovery curve is plotted versus Gt for a fixed flocculant dosage. A set of curves can be generated readily for different flocculant dosages. Increasing flocculant dosage generally increases the solid recovery, as illustrated in Fig. 8.2D. 8.3.5

Acceleration and Deceleration Time Duration

Despite using a powerful motor drive, it takes time to get the centrifuge rotor to speed because of inertia of the rotor. Some specially designed bench centrifuge can get up to 10,000 g in 2 seconds, however, most other spintube centrifuges can accelerate to, say, 3000 g in a matter of 15 seconds, and it may take another 20 seconds, or longer, to coast down with pneumatic/hydraulic brakes and other electrical retardation mechanisms. Suppose the operating speed Ωo 5 3500 rpm and it takes an elapsed time ta 5 10 seconds to get to the operating speed Ωo. Assuming a linear increase in angular speed over time ta the integral of G
t is simply, ð ta  2 ð ta Ωo t 1 Gdt 5 Rdt 5 Gta (8.4a) t 3 a 0 0 Likewise, if td is the time elapse for deceleration, the G
t integral for deceleration is  ð td  ð td Ωo ½td 2t 2 1 Gdt 5 Rdt 5 Gtd (8.4b) 3 t d 0 0 The total effective time for the test run including both acceleration and deceleration assuming t is the operating time duration at steady rotation speed Ωo should then be 1 ðGtÞtotal 5 Gt 1 G½ta 1 td  3

(8.4c)

An example is used in the problems at the end of this chapter to illustrate the importance of making this correction, otherwise the test data would be incorrectly interpreted. 8.3.6

Settling Velocity

The settling velocity of a particle depends on size, shape, solid concentration, and other variables. Thus far, the Stokes law has been used to

Laboratory and Pilot Testing

181

calculate the settling velocity with correction of hindered sedimentation from concentrated suspension by the hindered settling factor λ (5vs/vso). 8.3.6.1

Apparatus for Visualizing Sedimentation Behavior

A special spintube arrangement can be setup whereby both vso and vs (or λ) can be measured. Suppose the spintube is rotated to the position as shown in Fig. 8.3, it is illuminated by a set of focused light sources spaced closely apart and aligned in a radial direction (top of Fig. 8.3). The light source can be either near infrared or ultraviolet light depending on the suspension. The transmitted light is measured by the Charged Couple Device (CCD) sensor located on the opposite side of the spintube (bottom of Fig. 8.3). From the transmission profile measured, both along the radius and over time, the location of the interfaces (suspension
liquid interface and the cake
suspension interface) and the solid concentration at a given location may be inferred. Eight to twelve spintubes can be used at any time during a test run and measurement on the transient transmission profile on each spintube is stored for subsequent data retrieval and analysis. 8.3.6.2

Sedimentation Behavior for Mondispersed Suspension

Fig. 8.4 shows a sketch of the transmission profile for a suspension with monodispersed particles. The liquid
suspension interface is very

e

Figure 8.3 Special spintube setup equipped with light transmission and CCD sensor to detect the transmission of light through the suspension. Reproduced with permission from LUM Gmbh.

182

Centrifugal Separations in Biotechnology

t=0 t1

t3

Air–water interface

Transmission%

100%

t2

t1

t2

Sediment buildup

t3 Increasing time, water– slurry interface increasing in radius toward bowl dia.

Increasing t

0 Envelop from profile generated at t1

Radius

Figure 8.4 Typical spintube air
water interface position (top) and transmission profile over time for a monodispersed suspension with distinct air
water interface moving from small to large radius; also, sediment is building up counter-currently from large to small radius.

distinct as all particles of the same size settle at the same velocity. The liquid
suspension interface increases with time at a relatively constant rate, but this rate can actually increase as it moves to a larger radius due to increasing G, as G is proportional to radius. After a certain time, the interface reaches the bottom, intercepting the sediment
suspension interface, which moves inwards from a large to a smaller radius. For monodispersed particles, the cake forms a regular structure and the cake
suspension interface is also very distinct. 8.3.6.3

Sedimentation Behavior for Polydispersed Suspension

In contrast, there is no distinct liquid
suspension interface when the PSD in suspension is polydispersed, as depicted in the sketch of Fig. 8.5. Larger particles settle fastest, followed by smaller particles, and the colloidal size in the submicron range. At any instant, the larger particles may be located close to the large radius of the tube, while the smaller ones are still at the small radius close to the water
air interface.

Laboratory and Pilot Testing

100%

183

Transmission%

Last profile

Increasing time

Air–water interface

Sediment buildup

t3 t2

t1 First profile 0

Radius

Figure 8.5 Typical spintube air
water interface position (top) and transmission profile over time for a polydispersed suspension; no distinct air
water interface moving from small to large radius; also, sediment is building up counter-currently from large to small radius.

These slow settling fine particles reflect, scatter, and absorb the light reducing light transmission. Consequently, there is a gradual increase over time in light transmission from the small radius to the large radius as these fines settle (see Fig. 8.5). This is a slow process and the behavior is very different from that of the monodispersed suspension sedimentation shown in Fig. 8.4. 8.3.6.4

Sedimentation Behavior for Monodispersed Suspension With Hindered Settling

For sedimentation with monodispersed suspension, there is a distinct liquid
suspension interface that can be traced over time. A plot of such for a real suspension is shown in Fig. 8.6 for feed solids 5.5%
19.55% v/v (bulk volume basis). They all exhibit a linear increase first, followed by a horizontal line (where sediment builds up to meet the suspension
liquid interface). Note that for feed solids concentration 5.5%
10.2%, they all show a slight concave upward behavior instead of a linear increase, due to increasing settling velocity at a large radius, as discussed. The slope (or average slope of the curve) represents the average settling rate over the period, and it decreases

Radial position of the interface (mm)

184

Centrifugal Separations in Biotechnology 115

110 Free settling

105

100

5.5% 8.3% 10.2% 14.55% 19.55% 39.2%

Hindered settling

95

90 0

500

1000

1500

2000

2500

3000

Time (s)

Figure 8.6 Radius of air
water interface for monodispersed suspension with different feed solids concentration showing free settling and hindered settling behavior. Reproduced with permission from LUM Gmbh.

with increasing feed solids. Particles do not contact each other; however, they feel the hydrodynamic effect of settling of neighboring particles. (Imagine a driver driving a small car on the motorway who feels the effect of a bigger car passing by despite there is no physical contact.) At a bulk solid concentration of 39.2% in the feed, the particles form a network and there could be physical contact during sedimentation. As can be seen in Fig. 8.6, the curve exhibits a convex upward (due to increasing G with R) followed by a concave downward shape until the suspension
liquid interface meets the growing sediment. The specially designed bench centrifuge as shown in Fig. 8.3 can provide useful measurement on the settling rate of a given suspension for particles with various sizes, shapes, and densities. Also, the concentration effect can be quantified, that is, the hindered settling factor can be measured instead of using correlation. In addition, it can measure the sedimentation not only for dispersed particles but also for flocculated particles. More sophisticated tool and analysis can be made using the technique in tracking the sedimentation behavior of a suspension [4].

Laboratory and Pilot Testing

8.4

185

Centrifugal Filter Testing

There are two scenarios that need to be considered. One has to do with the rotation effect generating secondary flow that reduces concentration buildup of protein on the membrane surface, while the other uses centrifugal pressure as discussed in Chapter 2, Principles of Centrifugal Sedimentation, to generate the transmembrane pressure to effect membrane filtration. 8.4.1

Steady-State Membrane Centrifugal Filtration to Determine Protein Diffusivity and Solubility

We are interested to determine the filterability of protein under the effect of rotation, and infer the properties, such as diffusivity at a given protein concentration as well as the solubility at a given temperature and pH. The first effect is to pressurize the feed protein solution so that the pressure upstream of the membrane is higher than downstream, and this transmembrane pressure drives the protein and associated liquid toward the membrane, see Fig. 8.7. Due to the selectivity of the membrane, the large molecular weight protein is retained while only the liquid can permeate through the membrane. The fluid system is in rotation such that secondary flow (Ekman flow) keeps a thin viscous boundary layer and thus also a much thinner mass boundary layer of protein on the membrane surface (due to large Schmidt number, which is the measure of viscosity effect to diffusion of protein solution). The entire experiment is run under steady state. Pressurized solution

Permeate channel

Membrane

Stationary housing

r2

Permeate Axis

Ω

Porous support

Permeate

Figure 8.7 Rotating membrane to control protein buildup on membrane surface with steady-state permeate/filtrate.

186

Centrifugal Separations in Biotechnology

There is a balance between advection of protein toward the membrane and the back diffusion from the membrane surface with higher concentration toward the bulk solution. The diffusion rate depends on the diffusivity at a given concentration at the membrane. By measuring the permeate rate, we can back out the diffusivity corresponding to a given concentrated protein at the membrane surface. The diffusivity of protein can be determined at any concentration until it reaches the solubility or gel limit, which can also be inferred. In Chapter 14, Rotating Membrane in Bioseparation, a model has been developed and the model is used to analyzed the test work reported earlier [5]. 8.4.2

Transient Membrane Centrifugal Filtration to Determine Protein Osmotic Pressure and Membrane Resistance

Rotation can setup a high static pressure that can be used as transmembrane pressure to effect membrane filtration of protein solution. This is shown in Fig. 8.8, and a photograph of the spintube centrifugal filter is shown in Fig. 6.18. Due to the batch feed, as the solution level drops the transmembrane pressure is reduced. The major resistance to permeate flux is due to the osmotic pressure at low concentration polarization [6] and not due to the cake buildup as there is no cake. Cake resistance is due to a growing cake layer with approximately constant properties, but the resistance increases over time due to growing cake thickness. However, this is not the case here. Unfortunately, the flux curve in membrane filtration of protein solution looks similar to that of cake filtration, but the correct physics is that there is no cake and this problem Support Membrane Rotating tube

Ω Feed

Filtrate

Axis Retained protein

Figure 8.8 Rotating membrane where high-speed rotation sets up transmembrane pressure driving filtrate/permeate through membrane in dead-end filtration.

Laboratory and Pilot Testing

187

should not be confused with cake filtration as has been analyzed in such a way [7], which is misleading. The correct analysis is given in detail in Chapter 14, Rotating Membrane in Bioseparation (published in the first edition of the text in 2007) and the model developed is used to analyze the transmembrane permeate flux. The flux curve is typically linear first and subsequently exhibits a diminishing return at large time due to lower driving transmembrane pressure as the solution in the tube has depleted and concurrently the osmotic pressure resistance (concentration polarization) has further increased. The test work can be facilitated much more accurately, of course, using the photo-sensor setup for filtration [7], which monitors on-line the decrease in liquid level over time. Otherwise, we have to either measure the filtrate volume collected in the spintube, which requires a sensitive load cell rotating with the tube, or monitor the decrease in liquid level in the rotating tube using a stroboscope. The osmotic pressure as function of concentration can be inferred (see Chapter 14: Rotating Membrane in Bioseparation). In both centrifugal membrane filtration cases considered, given the details involved, it is best to take up the subject in a separate chapter (Chapter 14: Rotating Membrane in Bioseparation) for readers who are interested.

8.5

Pilot Testing

Pilot testing is absolutely essential as it takes a bench viable technology to a demonstration phase where “biotech materials” can be produced on a semicontinuous small-scale basis. It is one of the most critical phases. Often problems arise which are not encountered in bench-scale testing. This could be due to many factors, among which are the absence of flow dynamics with inflow and outflow which is not present in bench testing, and larger amount of feed materials and discharge materials (centrate or concentrate) for handling compared to bench scale, which is handled manually. This addresses whether the scale-up is appropriately carried out from lab bench-scale to pilot scale on the bioprocess in question. Even sample (i.e., feed, centrate, and concentrate) handling, such as transport, becomes quite different between a continuous pilot test operation and a bench test. Prior to discussing pilot testing, a discussion of basic concepts on material balance with continuous flow is in order.

188

Centrifugal Separations in Biotechnology

8.5.1 8.5.1.1

Material Balance Consideration for Pilot/Production Scale Material Balance by Volume Fraction

Our discussion on material balance will not be just limited to pilot-scale tests, it can also apply to production trials or full production. In pilot or production tests, continuous feed and discharge streams have to be dealt with. Balancing the volumetric rate Qs and the solids concentration individually for feed, concentrate, and centrate, material balance equations can be obtained similar to Eqs. (8.1a) and (8.1b) and as such similar results can be obtained. Qf 5 Qe 1 Qs

(8.5a)

Qf cf 5 Qe ce 1 Qs cs

(8.5b)

Qs cf 2 ce 5 Qf cs 2 ce

(8.6)

from which

The solids recovery Rs becomes Rs 5

cs Qs cs cf 2 ce 1 2 ce =cf 5 5 cf Qf cf cs 2 ce 1 2 ce =cs

(8.7a)

Also, when the centrate concentration is very much smaller than that of the concentrate, Rs  1 2

ce ðce {cs Þ cf

(8.7b)

One immediate application of the result is to determine the solids recovery in the continuous pilot and production tests. Another use is to back-out the time for the intermittent discharge of the concentrate. Given Vs is the solid holding volume of the disk centrifuge, the time to fill the space ts can be determined from the definition with the use of Eq. (8.7a)     1 2 ce =cs Vs Vs   5 ts 5 (8.8a) cs Qs Qf cf 1 2 ce =cf In the test, Vs is given by the geometry of the test centrifuge, Qf is fixed by the operating feed rate, and the cs are measured for the test conditions on all three streams.

Laboratory and Pilot Testing

189

Frequently, operators ignore the concentration of centrate solids, that is, assuming ce{cf; thus, this also implies ce{cs. Taking ce/cf  0 and ce/cs  0, ts 

Vs ηd cf Qf

(8.8b)

The discharge time is a rough estimate anyway and is used as an initial point for adjusting discharge. cf can be measured from the bench test and given Vs and Qf are both readily obtained ts can be estimated. Once the machine is set to discharge at this initial guess, subsequently, the discharge time can be better tuned (either longer or shorter) once the turbidity is monitored in-line with the discharged centrate liquid. It is to be noted that concentrate discharge efficiency ηd is added to Eq. (8.8b) to account for possibility that solids may adhere to the space in between openings of the intermittent discharge disk centrifuge. This effect essentially reduces the hold-up volume and thus also the time in between intermittent discharge, ts. 8.5.1.2

Material Balance by Mass Fraction

Instead of examining the concentration as expressed by number of biological cells per unit volume or by bulk volume concentration, one can also examine the weight fraction Wi of a given species i and the mass flow rate Mi of that species. The species i can be the feed f, centrate e, or solid s. Balancing the mass rate of liquid and solid in the feed, centrate, and solid, as well as balancing the biological material (excluding liquid) of the feed, liquid and solid, Mf 5 Me 1 Ms

(8.9a)

Mf Wf 5 Me We 1 Ms Ws

(8.9b)

Again Eqs. (8.9a) and (8.9b) follow very much like Eqs. (8.1a), (8.1b), (8.5a), and (8.5b). The results should be similar but with appropriate change in variables. Solids recovery by weight is thus Rs 5

Ws Ms Ws Wf 2 We 1 2 We =Wf 5 5 Wf Mf Wf Ws 2 We 1 2 We =Ws

(8.10a)

Ignoring the centrate concentration in relation to that of the concentrate Rs  1 2

We ðWe {Ws Þ Wf

(8.10b)

190

8.5.2

Centrifugal Separations in Biotechnology

Product (Protein) Yield

In what follows, all quantities and relationships that are required to determine protein yield are examined (see Fig. 8.9). A key assumption to simplify the mathematics is that we ignore the suspended solids in the centrate. Let φf be the “actual” (not bulk) volume fraction of solid in feed, and φs the volume fraction of solid in cake. Material balance on volumetric rate and liquid rate are given respectively in the following equations: Qf 5 Qe 1 Qs   Qf ð1 2 φf Þ 5 Qe 1 Qs 1 2 φs

(8.11a) (8.11b)

Subtracting Eq. (8.11a) from Eq. (8.11b), and after arranging Qf Vf φ 5 5 s Qs Vs φf

(8.11c)

Note in Eqs. (8.11a
c ), Qs refer to the rate of feed slurry and concentrate flow in a continuous centrifuge, whereas Vs refer to the volume respectively of feed slurry and concentrate in a spintube, as shown in Fig. 8.9. From Eqs. (8.11a) and (8.11b) φf Qe 512 Qf φs

(8.11d)

Ve = Qet

Vf = Q f t

Vs t=0 Feed Volume, mL Vf Solids vol. fraction: φf Liquid fraction: 1– φf Protein conc. Cf

t Cent rate Ve 0 1 Ce

Sediment Vs φs 1– φs Cs

Figure 8.9 Schematic representation of spintube before and after centrifugation, assuming no suspended solids in centrate. This is the basis for calculating yield of protein in product centrate.

Laboratory and Pilot Testing

191

Further Cf, Ce, and Cs are the concentration of dissolved protein in liquid for feed, centrate, and cake solids, respectively, balancing the protein in the three streams, so that   (8.11e) Qf ð1 2 φf ÞCf 5 Qe Ce 1 Qs 1 2 φs Cs Protein yield Y1 can be defined as the ratio of protein in the centrate to that of the feed, Y1 5

Ce Qe C 1 2 φf =φs Ce 1 2 Vs =Vf Ce 1 2 1=CF  5 e 5 5 Cf 1 2 φ f Cf 1 2 φf Cf 1 2 φf Cf Qf 1 2 φf (8.12)

The yield of protein can be determined from Eq. (8.12). It requires the solids volume fraction of feed φf, concentration factor CF 5 Vf /Vs, and concentration ratio of protein Ce/Cf. The concentration factor can be obtained from spintube testing. When φf  0, CFc1, then protein yield depends primarily on the concentration of protein in the centrate and feed, respectively, Y1  Ce/Cf. It is clear from Eq. (8.11e) and Eq. (8.12) that if the concentrate has little liquid content, that is, φs  1, then the protein yield approaches 100% as all the protein in the feed goes to the centrate and the concentrate has no liquid, that is, no soluble protein. This is impossible to completely remove the liquid in the concentrate. In practice, the concentrate leaving the centrifuge is repulped and goes through a second-stage centrifugation to remove as much protein as possible in the centrate of the second stage, or subsequent stages for that matter, by dilution followed by separation. Suppose the concentration ratio CF2 5 (Qf /Qs)2 5 (Vf /Vs)2 and the protein concentration in the centrate Ce2 is the same as the feed Cf2. Furthermore, the solids concentration of the feed is assumed to be negligible, that is, φf2  0. Then, the yield Y2 from the second-stage centrifugation (based on the unrecovered protein from the first stage) using Eq. (8.12) becomes !    Ce Qe 1 1 12 5 ð1 2 Y1 Þ 1 2 (8.13) Y2 5 1 2 CF2 CF2 Cf Qf ð1 2 φf Þ Total yield from the two-stage centrifugation becomes:   1 Ytotal 5 Y1 1 ð1 2 Y1 Þ 1 2 CF2

(8.14)

192

Centrifugal Separations in Biotechnology

Note that both Eqs. (8.13) and (8.14) have taken into approximation that in the second stage the feed solids is negligible φf  0 and the Ce  Cf.

Example 8.1 φf 5 0.05 CF1 5 CF2 5 7 (Ce)1 5 (Cf)1 Y1 5

Ce 1 2 1=CF 1 2 1=7 5 0:9022 5 1 2 0:05 Cf 1 2 φf

 1 Y2 5 ð1 2 0:9022Þ 1 2 5 0:0838 7 Ytotal 5 Y1 1 Y2 5 0:9022 1 0:0838 5 0:986

The protein yield from the first stage is 90.2% and an additional 8.4% protein yield has been obtained in the second stage. Therefore, the two stages in combination have a total of 98.6% protein yield. 8.5.3

Pilot Test Factors

Pilot test centrifuges can be used for confirming production of product protein. Pilot disk centrifuge of nominal diameter 180 and 240 mm are being used for testing and feasibility study. Also, the pilot tubular centrifuge of nominal diameter 150 mm is used for testing. A photograph of the 180-mm pilot disk stack centrifuge is shown in Fig. 8.10. Also, Fig. 8.11 shows the back side of the pilot disk stack centrifuge with the turbidity meter for monitoring centrate clarity together with other ancillary instrumentations. 8.5.3.1

Monitored Variables in Pilot Tests

The monitored variables in the pilot test include the following: • rpm/G of centrifuge—The rotational speed or centrifugal gravity of the centrifuge should be monitored with tachometer or comparable speed-measurement device. The set rotational speed should be adjusted to adapt to the process need. Given that centrifugal acceleration varies to the second power of the rotation speed and solid
liquid separation is typically a strong function of G, the speed is one of the

Laboratory and Pilot Testing

193

Figure 8.10 Skid mounted pilot disk centrifuge showing the front view and the control panel. Reproduced from Rohr and Teebe’s paper (2002) by permission from the American Filtration and Separations Society.

key controlling parameters. The test range should be made at a wide range of Gs with a nominal mean value selected for constant operation. Also, higher and lower G-force should be tested to check cost
benefit effects. • Flow rate of feed and centrate—Another readily adjustable parameter is the feed rate. The feed and centrate rates should be closely monitored. Other than the nominal feed rate have to be tested, higher and lower feed rates should also be tested for use in scale-up. • Centrate solids and turbidity—The centrate solids should be monitored, for example, using turbidity as a monitoring index on a continuous basis as it is the key gauge variable measuring performance. This is especially when the centrifuge is performing clarification or polishing. Also, individual or “grab” samples should be taken periodically to measure the percentage by weight or percentage by bulk volume (after centrifuged for 2
3 minutes under 10,000 g). With low solid concentration, it may be difficult or inaccurate to measure based on percentage by weight as the tare weight is large compared to the sample weight.

194

Centrifugal Separations in Biotechnology

Figure 8.11 Skid mounted pilot disk centrifuge showing the back view with back-pressure control, turbidity monitoring, centrate, and feed ports. Reproduced from Rohr and Teebe’s paper (2002) by permission from the American Filtration and Separations Society.

• Solids concentration in feed and underflow—Both feed and underflow concentration should also be monitored. The former measures the solid loading and the latter measures the concentrate concentration and handling. Also, the latter measures the yield of the protein product—loss of liquid protein in reject concentrate. The solid concentration is best measured using percentage by weight or percentage by bulk volume (after centrifugation in a spintube). Occasionally, if the feed contains low amount of solids, as with clarification application, turbidity monitoring should be conducted on the feed stream. • Cell concentration—The number of cells concentration cells per milliliters in feed, centrate, and underflow (continuous discharge) are monitored for say mammalian cells (such as the commonly used Chinese Hamster Ovary). Also, the lysed cells are closely monitored using say lactate dehydrogenase (LDH) technique especially when lysed cells can generate undesirable intracellular substances released in the liquid that contaminates the expressed protein product.

Laboratory and Pilot Testing

195

• Flocculant Dose—Flocculant is used for extracellular enzyme processing. Most other bioprocesses do not use flocculant due to (1) solid being the product and (2) liquid being the product; however, the polymer dissolved in solution might affect the liquid product. For the extracellular enzyme application, a suitable flocculant is selected such that it is compatible with the liquid product. • PSD of feed and centrate—The PSD of feed and centrate should be monitored. Given sedimentation rate varies as the second power of particle size, PSD with a large fine fraction in submicron range may experience difficulty of separation. This is an important measurement that has often been ignored. Despite particle counter is not an direct measurement of particle size, as it depends a certain physical phenomena (e.g., optical scattering), which the counter is based on. Despite this, it is a relative measurement that monitors an important property of the feed for which the separation is based upon. It would be very helpful if the same instrument is being used throughout the tests to provide a bench-mark for comparison and even making prediction, and also between centrifuge scale-up/scale-down (see Chapters 15
17). • Protein concentration in feed, centrate, and underflow—Protein concentration in all three streams should be monitored to accurately determine the yield. • Process temperature plays a key role in the viscosity of the liquid and hence that of the suspension. Liquid viscosity is reduced at higher temperature but there is a limit on the process temperature due to (1) protein may denature at an elevated temperature, (2) increases corrosion, and (3) economics. High-speed centrifugation may generate an increase in temperature by as much as 20 C between the incoming feed and the discharged centrate. This is one important concern that needs to be addressed if the protein can be spoiled by an increase in the temperature rise. There are ways to address this problem as discussed in Chapter 4, Disk Centrifuge. • Viscosity may be monitored using commercially available viscometer that are available in different geometry, such as coaxial, capillary, and cone-and-plate. Occasionally, in-line viscometers can be used. These viscometers provide only relative measurement on viscosity for the suspension/liquid, especially for liquids with high soluble protein and suspended solids. • pH should be monitored to keep protein in solution and avoiding precipitation or dissolution of solids due to high solution acidity (low pH) or alkalinity (high pH).

196

Centrifugal Separations in Biotechnology

• Mechanical and electrical conditions should all be monitored to ensure that the machine is running under normal conditions. These encompass: 3 mechanical vibration to ensure the masses in the rotor are evenly distributed without unbalance, 3 torque to ensure it is within limit preventing torque overload, 3 electrical current to ensure again the loading and associated power drawn is within limit, and 3 bearing temperature is within limit and the lubrication system of the rotating parts is working properly. • Other pertinent process parameters for the process should also be monitored and controlled. 8.5.3.2

Metrics of Pilot Tests

Several parameters below are used to gauge the performance. Some are more important than others. • Centrate clarity (solids % by weight or turbidity in Nephelometric Turbidity Unit (NTU))—The centrate clarity or suspended solids are most commonly used to gauge separation. Too high centrate solids lead to overload of downstream depth filter in the case of processing mammalian cells. High turbidity is another way of measuring poor separation as turbidity and suspended solids are closely correlated for low solid concentration. Also, turbidity/centrate suspended solids usually increase gradually with increasing volumetric feed rate. At very high feed rate, turbidity/centrate solids can increase sharply with small increment of rate, as depicted in Fig. 8.12. This is due to concentrate accumulating in the disk stack solid holding space, which has increased between concentrate discharges or shots. Some solids get Turbidity

Steep rise due to centrate interfering with sediment

Gradual increase 0

Inc. underflow discharge frequency

Qf

Figure 8.12 Turbidity as a function of volumetric feed rate.

Laboratory and Pilot Testing

•

•

•

•

197

entrained by the feed stream and start migrating to the disk stack and carried by the fast-moving centrate flowing up the disk channels. Solids recovery Rs in concentrate is another measure of centrate clarity. As with the centrate suspended solids depend on the feed solid concentration, the solid consistency in the concentrate also depends on the feed solids. The solids can get entrained in the disk channel can cause sudden precipitously drop in solids recovery. Yield Y (soluble product in centrate vs feed) is usually in the high 90%. It is an important measure of not losing protein during separation of solids from the suspension. Cell viability should be maintained at a maximum. Moreover, cell viability should not be significantly decreased between feed and the centrate, otherwise the centrifuge would have acted unintentionally as a homogenizer. It can be measured by conventional laboratory tests such as LDH release assay or multiplex cytotoxicity assay. The solids throughput/capacity should be maintained. This means the volumetric feed rate should be maintained constant if feed solids concentration is relatively constant.

The frequency distribution of feed particle size affects separation as bigger particles settle much faster than smaller ones, by virtue of Stokes’ law as discussed in Chapter 2, Principles of Centrifugal Sedimentation. Fig. 8.13 shows two different commonly encountered size distribution expressed as cumulative size distribution. These two size distributions are monodispered and polydispersed PSD. In the frequency distribution (not shown), the curve for the monodispersed distribution appears as a spike in the narrow size range where most particles reside. Outside this range, the frequency of occurrence is practically very small to nil. In the cumulative distribution as illustrated in Fig. 8.13, this appears as a discrete step where %Cumulative Undersize

100%

Polydispersed PSD

Monodispersed PSD x (μm)

Figure 8.13 Particle size effect.

198

Centrifugal Separations in Biotechnology Rs 100% Polydispersed PSD

0

Monodispersed PSD

Qf

Figure 8.14 Solids recovery curve.

the particle size resides. Below this size, the cumulative distribution is virtually 0%, and above this size range, the cumulative distribution approaches 100%. On the other hand, the curve for the polydispersed distribution appears as a broad spectrum that spans a wide range of particle sizes. In the cumulative plot the polydispersed distribution as illustrated in Fig. 8.13, appears as a gradual increase from the small size at 0% all the way up to the larger sizes until it reaches 100%. As shown in Fig. 8.14, the solids recovery Rs depends importantly on the feed PSD. The recovery curve is almost a mirror image of the PSD of Fig. 8.13. For monodispersed size distribution, the solid recovery stays close to 100% at small feed rates, and after a critical feed rate has been reached, the solids recovery drops off precipitously with a small marginal increase to a very low value, see Fig. 8.14. For polydispersed size distribution, the solids recovery drops gradually with increasing feed rate, and increases gradually with reduction in feed rate, as illustrated in Fig. 8.14. 8.5.3.3

Flocculant and Coagulant in Pilot Tests

Screening tests should be conducted initially to determine the appropriate coagulant and flocculant that work best for the specific biological solution under investigation. In pilot testing of coagulant and flocculant, usually higher dosage is needed above and beyond that of the laboratory dosage due to flow dynamics (primarily believed to be shear) that are absent in spintube testing. Also, certain adjustment need to be made in pilot testing, such as mixing time and mixing speed, so that there is sufficient time and energy from the mixing to form sizable flocs. This holds especially for the flocs that have slow kinetics. In pilot testing, coagulants and flocculants are added tens of meters away from the centrifuge, usually by piping the feed and chemical stream to meet at an angle, or by T-junction, to deliberately generate mixing and turbulence when the two streams

Laboratory and Pilot Testing

199

meet. The combined stream is sent to the centrifuge 5
20 m away. The turbidity of the centrate is monitored and the polymer dose is adjusted accordingly to minimize the turbidity at a reasonable dosage of chemicals. In Chapter 15, Flocculation With Decanter Centrifuges, the approach of inferring the in situ floc size in test centrifuge is presented. This is another approach that can be adopted in pilot testing and for scale-up to full-scale machine.

8.6

Summary

Bench-scale laboratory and pilot-scale testing have been discussed in this chapter, especially in relation to the process objectives and feed properties. Material balance is used to back-calculate the variables that are difficult to measure accurately but are required as part of the measuring metrics. Laboratory spintube testing can be made to determine the separability and the Gs and ts to make separation. It can also be used to determine whether chemical additives, coagulants and flocculants, are required to enhance separation. A special designed spintube centrifuge system, equipped with optical transmission measurement, provides measurement on the sedimentation behavior of the suspension (monodispersed vs polydispersed particles, and unflocculated vs flocculated suspension) over time and space. Centrifugal filtration is briefly discussed for evaluating protein properties and filterability. Pilot testing should be the next step once bench-scale laboratory test demonstrates the viability of the separation process. Various parameters in controlling and monitoring pilot centrifuge testing are discussed.

References [1] G.I. Taylor, Low Reynolds Number Flow, National Committee for Fluid Mechanics Films (NCFMF), A.H. Shapiro (Ed.), 1961. [2] R.F. Probstein, Physicochemical Hydrodynamics, Wiley & Sons, 1994. [3] W.W.F. Leung, Separation of dispersed suspension in rotating test tube, Sep. Purif. Technol. 38 (2) (2004) 99
119. [4] D. Lerche, Comprehensive characterization of nano- and microparticles by in-situ visualization of particle movement using advanced sedimentation techniques, KONA Powder Particle J. 36 (2019) 156
186. [5] W.F. Leung, R. Probstein, Low polarization in laminar ultrafiltration of macromolecular solution, I&EC Fund. 18 (1979) 274.

200

Centrifugal Separations in Biotechnology

[6] A.A. Kozinski, E.N. Lightfoot, A general example of boundary layer filtration, AICHE J. 18 (5) (1972) 1030
1038. [7] M. Loginov, F. Samper, G. Ge´san-Guiziou, T. Sobisch, D. Lerche, E. Vorobiev, Characterization of membrane fouling via single centrifugal ultrafiltration, J. Taiwan Inst. Chem. Eng. 94 (2019) 18
23.

Problems (8.1) A spintube is spun to a maximum speed of 3500 rpm. The acceleration time ta is 10 seconds, and the deceleration time td takes 20 seconds from 3500 rpm to a stop despite pneumatic brake is being used. The geometric mean radius of the suspension is at 6 cm. (Use this geometric radius instead of the bowl radius at the large-diameter end of the tube as it is more accurate [3]. Note the geometric mean radius is the geometric mean of the bowl/tube radius and the radius of the suspension
water interface.) The following are the test data collected from an experiment on sedimentation characteristics of a biological suspension: t, s

% solids recovery

0 5 10 20 30 40 50 60

Nil 33 40 55 66 73 79 82

The data in the table are the centrifugation time duration at 3500 rpm without accounting for either acceleration or deceleration times. Determine the performance of this test set by calculating (Gt): (1) without acceleration and deceleration effect, and (2) with accounting for acceleration and deceleration using Eq. (8.4c)? What is the difference between the two results?

(8.2) In a spintube test, the feed solid is determined to be 4% v/v, centrate solids at 0.5% v/v, and sediment at 15% v/v, what is the solids recovered (1) in the sediment and (2) in the centrate? (8.3) A small pilot disk centrifuge has a bowl volume of 3 L, 50% of which is used to store sediment in between shots. For a yeast feed

Laboratory and Pilot Testing

201

at 20% v/v and with a feed rate of 4 L/min and discharge efficiency of 80%, what is a reasonable initial guess on the time duration in between shots? (8.4) Design a pilot test using a tubular centrifuge to run broth from a 200-L fermenter. What are the parameters that should be controlled and monitored?

9 Selection and Sizing of Centrifuges This chapter discusses the selection and sizing of centrifuges, which are very important and pertinent in practice, especially for applications in biotechnology requiring hands-on work. We discuss the selection of centrifuges based on the process requirements. A unified approach on the scale-up of all centrifuges based on the dimensionless Le number is presented.

9.1 9.1.1

Selection Introduction

Both disk and tubular centrifuges have been used for the separation of microbial cells or bacteria, yeast, and mammalian cells, especially in biopharmaceutical production. In the other extreme, spintubes are commonly used to carry out the separation in feasibility study of a process and ultracentrifuges are used to determine analytically the properties of a sample in the laboratory as well as other analytical separation, which requires accuracy and high G. Centrifugal separation of microbial cells uses higher G for separation, especially with smaller particle sizes. On the other hand, centrifugal separation of mammalian cells with slighter larger particles, over 10 μm, uses slightly lower G for separation. Mammalian cells are more susceptible to shear and damage due to the lack of cell wall. This is especially important in accelerating feed stream ineffectively to high speed. When under-accelerated feed is in contact with the pool liquid, which is already at solid-body rotation (i.e., v 5 ΩR), there is a significant difference in velocity mismatch that leads to shear, turbulence, and mixing of the feed and the liquid in the centrifuge (see Chapter 4: Disk Centrifuge). This may lead to undesirable lysing of cells inadvertently Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00009-X © 2020 Elsevier Ltd. All rights reserved. 203

204

Centrifugal Separations in Biotechnology

releasing fine cell debris that are difficult to separate and undesirable intracellular contents that contaminate the products. 9.1.2

Tubular Centrifuge Selection

Tubular centrifuge is advantageous to process lower rate applications using a single unit. The cake discharged is drier with lower moisture content when compared with that of a disk-stack centrifuge. Tubular centrifuge operates in a cycle with loading, separation, liquid drainage, concentrate discharge, cleaning, and rinsing (see Chapter 3: Batch and Semibatch Centrifuges). Concentrate in tubular centrifuge has solid consistency that can withstand shear by unloading knife or a plunger during concentrate discharge. When sizing the tubular centrifuge, the entire cycle needs to be factored in the capacity calculation. This is unlike a nozzle or intermittent discharge disk centrifuge in which concentrate discharge is continuous, and there is no downtime required for the operation. 9.1.3

Disk Centrifuge Selection

When sterilization-in-place is not required, centrifuge manufacturers can offer disk models for yeast separation from brewery applications with clean-in-place features. The costs of these brewery centrifuges are lower (from the disk centrifuge to the control panel), and there are more choices for serving the fermenter size of interest. Also, it would be useful to plan out any possible expansion in the next 5 years, or longer if possible, so that existing equipment does not run out of capacity in a hurry. When sizing the manual disk, as with the tubular centrifuge, the entire cycle needs to be factored into the capacity consideration. If the downtime becomes unacceptable and the feed solid is high, then the nozzle or intermittent discharge disk centrifuge should be considered. The nozzle or the intermittent discharge disk is selected based on the nature of the feed solids concentration and the concentrate to be discharged. If the feed solids are high, the nozzle is preferred; otherwise, the intermittent discharge needs to discharge frequently, which interferes with the sedimentation in the centrifuge. The small-diameter top discharge, internal vortex nozzle, and external nozzle disk centrifuges are for flowable concentrate. In particular, the small-diameter top discharge disk is for the case when the concentrate angle of repose is less than 25 degrees, whereas the internal vortex nozzle and external nozzle disk are for concentrate with an angle of repose greater than 25 degrees. On the other hand, the intermittent discharge, in the form of dropping-bottom design, is preferred for nonflowable concentrate. When small quantities of feed are processed and cake dryness

Selection and Sizing of Centrifuges

205

is desired, the manual disk would be preferred, but the feed solid concentration needs to be low, otherwise solid can fill up the bowl relatively fast. 9.1.4

Centrifuge Comparison

Table 9.1 summarizes various types of disk and decanter centrifuges with the typical range of feedsolids and sediment handling. It is noted that manual discharge handles low-solid feed in the range of 0%
1% by bulk volume of solids. The intermittent axial channel (concentrate path at bowl diameter as discussed in Chapter 4: Disk Centrifuge) disk can go up to 0.01%
10% v/v feed solids. The intermittent discharge (radial slot) takes up higher feed solids of 0.2%
20% v/v feed solids. The nozzle disk can process 1%
30% v/v feed solids and provide that the concentrate is in a flowable form (concentrate not stackable with a large angle of repose). All large centrifuges can process in upward of 400 L/min feed rate. On the other hand, decanter can handle much higher feed solids 5%
80% v/v and can concentrate and dewater cake to high dryness. Table 9.2 compares the advantages and disadvantages of the centrifuges. The tubular bowl has high G typically 5000
20,000 g with the exception of one small bench unit that can attain 62,000 g (Table 3.2). It has reasonably good dewatering characteristics, simple to disassemble, and easy to clean. Table 9.1 Basic types of centrifugal separators

Automatic tubular Manual disk Intermittent discharge (axial channel) Intermittent discharge (radial slot) Nozzle disk

Decanter

Sediment

Solids content Maximum flow in feed (% v/v) rate in largest machine (L/min)

Remains in bowl

0
4

30

Remains in bowl Discharge through axial channels

0
1 0.01
10

500 600

Discharge through radial slots

0.2
20

500

Continuous 1
30 discharge through nozzles Screw/conveyor 5
80 discharge continuously

600

600

206

Centrifugal Separations in Biotechnology

Table 9.2 Advantages and disadvantages of various types of centrifuges Centrifuge type Advantages

Disadvantages

Tubular bowl

1. High G 2. Dewatering, drier cake 3. Simple dismantling of rotating assembly 4. Easy to clean 5. Plunger and plough designs, automatic cake removal

1. Limited solids storage 2. Recovery of solids manually and difficult (loss of solid product) 3. Foaming unless centripetal pump used 4. Downtime during cake removal cleaning

Chamber bowl

1. Clarification efficiency stays constant until solid holding space gets filled. 2. Large solids holding capacity 3. Good dewatering 4. Bowl cooling possible

1. No solid discharge 2. Cleaning more difficult than tubular bowl 3. Recovery of solids manually and difficult (loss of solid product) 4. Downtime during cake removal and cleaning

Disk centrifuge

1. Solids discharge possible 2. Liquid discharge under pressure eliminates foaming

1. Poor dewatering 2. Difficult to clean

Decanter

3. Bowl cooling possible 1. Continuous solids discharge (dry cake) 2. High feed solids concentration

1. Lower G 2. Less surface area for clarification compared with disk stack

Traditional tubulars with a large L/D ratio have limited storage, with the exception of the small L/D ratio tubular with a large diameter bowl. Furthermore, the recovery of solids is difficult with the exception of new modern tubular with low L/D ratio, as discussed in Chapter 3, Batch and Semibatch Centrifuges, where solids are discharged by a plunger or with an unloading knife during the solid-unloading cycle. The key disadvantage is the downtime required for either manual or automatic concentrate discharge. The chamber bowl has good clarification and dewatering characteristics and large temporary space to hold solids. As with the tubular, it

Selection and Sizing of Centrifuges

207

requires downtime for solids unloading translating to lower overall capacity. The concentrated solid unloading is manual. Depending on the nature of the solids, it might be of concern for some process materials as it requires a human interface with the process materials. Also, cleaning of the bowl is even more difficult than with the tubular. For disk centrifuge, there are various forms of solid discharge as discussed in the foregoing and there is quite a lot of flexibility for selection, depending on process requirements, budget, availability, and delivery. Liquid is discharged using a centripetal pump to reduce foaming. Also, various forms of hermetic seals are available for both feed and liquid discharge to avoid liquid products having contact with air, which can cause oxidation and spoilage. The key disadvantages are poor dewatering and difficult to clean, especially the disk stack. Decanter centrifuge provides a continuous cake discharge and continuous feed. It has good dewatering capability. The key advantage of all centrifuges is that it can take much higher feed solids concentration. The drawbacks are that it has lower G as compared with the tubular and disk centrifuges and less surface area when compared with the disk centrifuges, rendering the clarification and separation of biosolids less desirable, unless flocculant is used to increase the feed solids sizes, as with processing extracellular enzyme (see Chapter 6: Commercial Applications of Centrifugation in Biotechnology).

9.2 9.2.1

Centrifuge Sizing Sizes and Rates

Fig. 9.1 shows a schematic representation of a disk centrifuge. It is expected that the larger the fermenter, the larger is the centrifuge required. Fig. 9.2A shows that 1-L bowl volume can handle the 50-L bioreactor, a 3-L bowl disk centrifuge serves the 100
1000 1 fermenter, a 20-L bowl serves the 2500
10,000 L fermenter, and a 25
30 L bowl serves the 5000
20,000 L fermenter. This information is also tabulated in Table 9.3 for ease of reference. As can be seen, the centrifuge bowl size increases linearly with the fermenter size. Figs. 9.2B and 9.2C show two different commonly used fermenter sizes, respectively, a 50- and 2000-L bioreactor or fermenter. The maximum and minimum feed rates of respective tubular and disk centrifuges are shown in Fig. 9.3. The feed rate varies approximately as the third power of the bowl diameter from the hydraulic capacity consideration. However, this is much lower from the process viewpoint and the specific feed rate depends on the requirements of a given process.

208

Centrifugal Separations in Biotechnology

Clarified liquid

Disk stack Compacted solids

Feed

Separated Feed solids

Bowl

Figure 9.1 Schematic of a disk. (A) Fermenter volume (L)

100,000 Min size (L) Max size (L)

10,000 1,000 100 10 1 0.1

1

10

100

Disk bowl volume (L)

Figure 9.2A Fermenter size versus disk bowl volume. Table 9.3 Fermenter volume versus disk centrifuge size

9.2.2

Fermenter size (L)

Disk bowl volume (L)

100
1000 1 2500
10,000 5000
20,000

3 20 25
30

Dimensionless Le Number

A new dimensionless Leung number (hereafter abbreviated as “Le”) has been developed for spintube, disk, tubular, and decanter centrifuges as used for separation and clarification sizing. Recently, it has also been

Selection and Sizing of Centrifuges

209

(B)

Figure 9.2B 50-L bioreactor/fermenter. (C)

Figure 9.2C 2000-L bioreactor/fermenter.

used in inferring of flocs in the decanter for clarification (see Chapter 15: Flocculation With Decanter Centrifuges), clarification and separation duties in disk and tubular centrifuges (see Chapter 16: Case Studies of Monotonic and Unimodal Size Distribution Models), and classification of biosolids of two different sizes and possibly two different densities for

210

Centrifugal Separations in Biotechnology Disk

Rate (L/min)

1000

100 Max rate Min rate

10

1 100

Tubular Diameter (mm)

1000

Figure 9.3 Disk and tubular size versus rate, with maximum rate limited by the hydraulic capacity.

all types of centrifuges (see Chapter 17: Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification). Le is a measure of the centrifuge feed rate to the sedimentation capacity of the centrifuge. Le is related to the feed rate to the sedimentation capacity of the centrifuge and should be of the order one for normal operation. Lec1 implies overfeeding while Le{1 underfeeding the centrifuge. The Le number has been used successfully in many applications for fine particle sedimentation (0.1
45 μm) [1
3]. Le is directly correlated with the cut size, that is, the maximum size in the supernatant or the minimum size in the sediment. Despite this, one might argue that there is no cut size that provides a sharp demarcation of the minimum size of solids in the cake/concentrate sample and the maximum solid size in the centrate due to various complexities (e.g., entrainment of sediment in the centrate), yet this provides a powerful concept whereby one can quantitatively size or scale-up centrifuges. Most importantly, the approach presented herein provides a unified approach for sizing different types of centrifuges. Le depends on centrifuge design geometry and properties of suspended particles (size distribution and density) and liquid (density and viscosity). The Le number has been developed for batch sedimentation (spintube) and continuous-feed centrifuges (disk, chamber bowl, tubular, and decanter centrifuges). Indeed, Appendix B shows that the Le number can be derived independently from the Buckingham-π dimensionless analysis for all four types of centrifuges: decanter, tubular, disk, and spintube. 9.2.3

Spintube (Bottle) Centrifuge

As discussed in Chapter 3, Batch and Semibatch Centrifuges, spintube centrifuges have been widely used in the laboratory, a schematic representation of which is shown in Fig. 9.4. A new analytical procedure of

Selection and Sizing of Centrifuges

Rotating tube

211

Pellet or settled heavy phase

Supernatant Axis

H

Figure 9.4 Spintube schematic.

separating a given suspension of the well-defined particle size distribution (PSD) to the desired particle capture has been developed [1]. The dimensionless Le number for the spintube incorporates several key geometric and operation parameters: suspension height H, centrifugal acceleration G, time duration t, suspension viscosity μ, density difference between solid and suspension Δρ, hindered settling factor λ (,1), which depends on the concentration of suspended solids φ, efficiency of spin-up η, and the characteristics particle size xo. The Le number for spintube is defined as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2πμH 2πμ0 H 5 (9.1) Le  ΔρλðφÞηGtx2o ΔρGtx2o μ0 5

μ λðφÞη

  xc 3 5 pffiffiffi Le 5 1:693 Le π xo

(9.2) (9.3)

In the above equations, μ0 is the effective viscosity defined as a ratio of liquid viscosity divided by the product of λ and η. This definition also accounts for hindered settling in the spintube as well as the efficiency of spin-up and nonideal radial geometry.

Example 9.1 Le Calculation—Yeast Cell (7 3 10 µm) H 5 4 cm μ0 /ρ 5 0.01 cm2/s Δρ/ρ 5 0.05 g 5 981 cm/s2 G/g 5 3000 t 5 240 seconds (4 minutes) (G/g) 2 t 5 720,000 seconds xo 5 1 μm Le 5 0.844 xc 5 1.69 Le xo 5 1.43 μm

212

Centrifugal Separations in Biotechnology

Note the cells to be separated are between 7 and 10 μm. If the spintube is operated such that the largest particle in the centrate is below 1.43 μm, then particles in the 7
10 μm range would certainly settle out in the sediment. In other words, the spintube is operated with a cut size equal or smaller than the cell size that is intended to be removed in suspension. For a given G, H, and other parameters, increasing time duration t certainly reduces Le and cut size. This also compensates for the lack of high G-force for some centrifuge designs, or in some processes, G needs to be lowered to avoid over-compaction of the pellet/sediment leading to difficulty for downstream processing (e.g., dissolution of inclusion bodies). Let us consider various biological cells to be separated by centrifugation, as listed in Table 9.4. In the table, it shows the commonly practiced G and t of centrifugation for separating various cell sizes. As can be seen, the cut size should always be smaller than the cell size in order to remove the cells by sedimentation from the suspension. Fig. 9.5 delineates a log-log graph of cell size plotted against cut size for the data in Table 9.4. A 45 degrees line can be drawn through the graph. As depicted in Fig. 9.5, given that the cut size should be smaller than the cells to be separated, all data lie above the 45-degree line as they should. Table 9.4 Biological cells Cells

G/g

t (min)

Cut size (μ)

Cell size (μ)

Cell debris Bacteria cells Yeast cells Mammalian Plant cells

40,000 10,000 3000 200 200

30 10 4 2 1

0.14 0.49 1.43 7.82 11.1

0.2 3 0.2 132 7 3 10 10 3 40 100 3 100

Plant cells

Cell size (µm)

100

Mammalian Yeast

10 Bacteria

1

Cell debris

Lower size range Max size range

0.1 0.01 0.01

0.10

1.00

10.00

100.00

Cut size (µm)

Figure 9.5 Biological cells - observed size versus theoretical cut size.

Selection and Sizing of Centrifuges

213

Table 9.5 Smaller-size organelle Organelle

Diameter (μ)

Nuclei Mitochondria Ribosomes Lysosomes

5
10 1
2 0.02 1
2

Density (g/cc) 1.4 1.1 1.6 1.1

Table 9.6 G, t, Le, and xc Organelle

G/g

t (min)

Le

xc (μ)

Nuclei Mitochondria Ribosomes Lysosomes

1000 10,000 100,000 10,000

1 20 100 20

1.03 0.146 0.008 0.146

1.75 0.25 0.014 0.25

The same exercise can be carried out on small organelle. A tabulation of nuclei, mitochondria, ribosomes, and lysosomes is given in Table 9.5. The organelles are indeed very small biological organisms with diameters ranging between 0.02 and 10 μm. The density of the small organelle is also listed in Table 9.5. Table 9.6 shows the typical time and G adopted to make separation based on laboratory experiences on these applications. Again, from separation considerations, these biological particles should get settled when the cut size is smaller than the particle size. Analogous to Example 9.1, calculations on the cut size can be conducted, and the results are shown in Table 9.6. Fig. 9.6 compares the cut size drawn from Table 9.6 with the cell size drawn from Table 9.5 for the various processes. As expected, all data lie above the 45 degrees line, indicating that the cut size is smaller than the cell size that is required to be separated by centrifugation. 9.2.4

Sizing for Disk Centrifuge

With reference to Fig. 9.7, the dimensionless Le number for a disk-stack centrifuge incorporates the feed rate Q, the angle of the disk stack with respect to the vertical θ, the inner disk radius R1, the outer disk radius R2, the slant length of disk L, the projected area parallel to the axis of the machine Lp, the number of disks n, the liquid viscosity μ, the density difference Δρ, and the characteristics particle size xo.

214

Centrifugal Separations in Biotechnology

Cell size (µm)

10

1

Reticulum microsomes

Microbodies Mitochondria lysosome

nuclei

Cell membranes 0.1

Endoplasmic ribosome

0.01 0.01

Lower size range Max size range

0.10

1.00

10.00

Cut size (µm)

Soluable cell cytoplasm

Figure 9.6 Smaller microbiological organisms - observed size versus theoretical cut size. Axis

Lcosθ

θ

R1

L L

Total projected area Lp = n Lcosθ

R2 G

Figure 9.7 Projected area of disk centrifuge.

The Le number for the disk centrifuge is given by sffiffiffiffiffiffiffiffiffiffiffiffi Q μ Lp Δρ Le 5 ΩR0 xo η R0 5



R22 1R2 R1 1R21 3 Lp 5 nLcosθ

(9.4)

1=2 (9.5) (9.6)

In the above, the effective radius R0 is related to R1 and R2 as expected. Eq. (9.4) is very similar to that of the decanter centrifuge except that R0 replaces the pool radius Rp for the case of a decanter. The clarification length Lp is given by Eq. (9.6). It can also be seen in

Selection and Sizing of Centrifuges

215

Chapter 12, Disk Stack Modeling, that Le can also be expressed in a more compact form as follows: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! u  u 3Q μ tanθ t   3

n ρs 2 ρL R2 2 R31 (9.7) Le 5 Ωxo η It is to be noted that both Eqs. (9.4) and (9.7) yield identical results in the calculation of the Le number. The results from the investigation of settling in a lamella or an inclined plate under the Earth’s gravity show that there are two modes of operation when feed suspension is introduced from the top of the lamella channel. With reference to Fig. 9.8, the first is a subcritical mode whereby the feed layer fills more than half of the channel widthwise and the feed expands to occupy the entire channel, with the exception of a thin clear layer rapidly flowing upwards adjacent to the underside of the channel. The second mode is a supercritical mode whereby the feed layer occupies less than half of the channel and contracts as it is introduced into the channel. Most of the channel is occupied by the clear layer. In contrast, when the feed is introduced at the bottom of the channel, only the subcritical mode is realized [3]. The aforementioned description applies only to a single channel. On the other hand, if there are multiple channels all lining up in parallel, the feed introduced at the bottom might not distribute uniformly into each channel, as demonstrated by the numerical simulation results in Fig. 9.9. It is seen in the figure that not all channels get the same Clarified product

Clarified product

Feed Feed

Sludge

Sludge

25°

Figure 9.8 Subcritical mode (left picture) and supercritical mode (right picture).

216

Centrifugal Separations in Biotechnology

Figure 9.9 Nonuniform distribution (computational fluid dynamics).

amount of feed suspension. The channel adjacent to the main feed stream gets most, whereas the channel to the farthest side from the main feed gets the least. In between, there is recirculation and other complicated secondary flow patterns. The flow pattern for disk stack is similar to that of a lamella settler with nonuniform feed distribution from channel to channel. This shortcoming of nonuniformity is accounted for by using an efficiency index η in Eq. (9.1). 9.2.4.1

Efficiency η in Le Number

There are actually several deficiencies that are all lumped into an overall efficiency factor. These are listed as follows: • Nonuniform distribution of feed into each channel as discussed • Feed not fully accelerated • Entrainment of sediment by a high-velocity liquid stream in the disk stack Given these complexities, it would still be desirable to maintain a high-efficiency factor. A few qualitative implications can be summarized from Eqs. (9.3)
(9.5). Good separation requires small cut size xc or small Le, low-feed rate Q, high rotation speed Ω, large disk stack area A, and low suspension viscosity μ, and vice versa for poor separation. Below is an example to illustrate the point.

Selection and Sizing of Centrifuges

217

Example 9.2 Disk Centrifuge Centrifuge bowl diameter D 5 400 mm Disk outer diameter 2R2 5 300 mm Disk inner diameter 2R1 5 177 mm Disk angle θ 5 40 degrees from vertical Density difference/density Δρ/ρ 5 0.1 Viscosity μ 5 5 cP G 5 12,000 g (7270 rpm); G 5 6000 (5140 rpm) n 5 100 disks Overall efficiency 5 70% Fig. 9.10 shows the linear relationship of Eq. (9.3) where the ratio of the cut size to the reference size is linearly proportional to the Le number. The reference size xo is arbitrary set as 1 μm in here, but it would be more appropriately set as a typical size for the feed, e.g. median size of the feed. The cut size is the largest particle that remains in the liquid centrate or the smallest particle in the sediment under a given operating condition. It is therefore also the separated cell size by centrifugation in the sediment. For the above set of parameters in Example 9.2, we can relate these parameters using Eqs. (9.3)
(9.6). R1 5 8.85 cm R2 5 15 cm From Eq. (9.5), 

152 1ð15Þð8:85Þ18:852 R 5 3 0

12

5 12:06 cm

R0 5 12.06 cm (from Eq. 9.5) Lp 5 (R2 2 R1)/tan θ 5 7.33 cm nLp 5 733 cm

Cut size (μm)

8 6 4 2

Disk centrifuge

0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

Le

Figure 9.10 Cut size is linearly proportional to Le.

3.5

4.0

218

Centrifugal Separations in Biotechnology

From Eq. (9.4), rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Q 1000 0:05 733 60 0:1  Le 5  5140π ð12:06Þð1 3 1024 Þð0:7Þ 30

(9.8)

Q in the above equation is in liters per minute (L/min). From Eq. (9.9), xc 3 5 1=2 Le 5 1:693 Le xo π

(9.9)

Note xo is set equal to 1 μm. Combining Eqs. (9.9) and (9.8), we obtain Eq. (9.10a) for the G-level of 6000 g. Also, for 12,000 g case, 7270 rpm can be used in place of 5140 rpm in Eq. (9.8), this leads to Eq. (9.10b). ! !2 L xc (9.10a) 5 6:34 G 5 6000 g Q m xo ! !2 L xc 5 12:68 Q m xo

G 5 12; 000 g

(9.10b)

Given Q ~ rpm2 (xc/xo)2 and G ~ rpm2, i.e., Q is linearly proportional to G. As such, the coefficient in Eqs. (9.10a) and (9.10b) doubles when the G-level is doubled. This is listed in Table 9.7. These results are plotted in Fig. 9.11. Eqs. (9.10a,b) are linear in a log-log plot. As the cell size to be separated decreases, the feed rate drops off as the second power of the separated cell size, which is quite sensitive. Therefore, it pays off to increase the particle size through the agglomeration of fine biological materials, such as the addition of flocculants to the feed suspension when separating intracellular enzymes. Table 9.7 Processed rate for two different G’s Separate cell size

Processed feed rate Q (L/min)

xc (μ) 0.5 1 2 3

6000 g 1.6 6.3 25.3 57.0

12,000 g 3.2 12.7 50.7 114.1

Selection and Sizing of Centrifuges

219

Feed rate (L/min)

100 Δρ/ρ=0.1 μ=5 cP n =100 disks 10 6000g 12,000g

1 0.1

1.0

10.0

Separated cell size (µm)

Figure 9.11 Performance of a disk for a feed suspension with 5 cP.

Feed rate (L/min)

100 Δρ/ρ=0.1 μ=5 P n =100 disks 10 6000g 12,000g

1 1.0

10.0

100.0

Separated cell size (µm)

Figure 9.12 High feed viscosity (5 P) and solids concentration.

When the viscosity is increased by 100 times to 5 P, the particle size needs to be increased by 10 times, such as by flocculation (see Chapter 15: Flocculation With Decanter Centrifuges, and Chapter 16: Case Studies of Monotonic and Unimodal Size Distribution Models), in order to keep the feed rate the same as before. This is illustrated in Fig. 9.12. A 10-μm particle can be removed in a suspension of 5 P at the same feed rate processed (about 6
7 L/min) as with 1-μm particle at 5 cP. Without considering the 10 times increase in the particle size, the feed rate would have to be reduced by a factor of 1/100. Consequently, highly viscous suspension, containing a high concentration of dissolved protein, can be very difficult to separate with low capacity and require high centrifugal acceleration. 9.2.5

Sizing for Tubular, Chamber, and Decanter Centrifuge

The sizing of tubular and chamber centrifuge is also given by the Le number. Fig. 9.13 shows a schematic of a tubular centrifuge with a bottom feed.

220

Centrifugal Separations in Biotechnology

Overflow

L/D>>1 Rp

L

Particle trajectories

Feed

D

Figure 9.13 Tubular centrifuge.

The Le number can be defined for a tubular chamber and decanter centrifuge as rffiffiffiffiffiffiffiffiffiffiffi Q μ L Δρ (9.11) Le 5 ΩRp xo η where Q is the volumetric feed rate, L is the clarifier length, μ is the suspension viscosity, Δρ is the density difference, Ω is the angular speed, Rp is the inner pool radius, and η is the feed acceleration efficiency. As evident, the Le number for the tubular centrifuge is very much similar to that of the decanter [2]. As discussed, this also applies to chamber bowl centrifuge geometry but without a disk stack. This is illustrated by the example below.

Example 9.3 Tubular Centrifuge Centrifuge diameter D 5 300 mm Length L 5 457 cm Rb 5 300/2 5 150 mm hp 5 35.7 mm Rp 5 Rb 2 hp 5 114.3 mm Density difference/density Δρ/ρ 5 0.1 Viscosity μ 5 5 cP 5 0.05 P G 5 20,000 g (10,920 rpm) or G 5 5000 (5460 rpm) Overall efficiency 5 80%

Selection and Sizing of Centrifuges

221

Using Eqs. (9.9) and (9.11) for this numerical example, it can be easily shown that !2 ! L xc Q 5 0:51 xo m ! !2 L xc 5 2:04 Q m xo

G 5 5000 g

(9.12a)

G 5 20; 000 g

(9.12b)

Eqs. (9.12a,b) are plotted in Fig. 9.14 for G/g 5 5000 and 20,000, respectively. For removing 1 μm particle, the centrifuge at 5000 g needs to operate at a feed rate no more than 0.5 L/min, whereas at 20,000 g, it can be operated at 2 L/min. Likewise, when the viscosity increases by 100 times to 5 P, these feed rates can be maintained if the particle is 10 times bigger, that is, 10 μm instead of 1 μm. This is illustrated in Fig. 9.15.

Feed rate (L/min)

100.0 5000 g 20,000 g 10.0 300-mm dia. Tubular 1.0 5 cP 0.1 0.1

1.0 Separated size (µm)

10.0

Figure 9.14 Tubular centrifuge for a feed with a viscosity of 5 cP.

Feed rate (L/min)

100.0 5000 g 20,000 g 10.0 300-mm dia. Tubular 1.0

5P

0.1 1

10

100

Separated size (µm)

Figure 9.15 Tubular centrifuge for a feed with high-viscosity 5 P and high feed solids.

222

9.3

Centrifugal Separations in Biotechnology

Feed Particle Size Distribution

The performance that has been discussed considers the separation of a single biological cell or particle in the feed (thick solid line in Fig. 9.16). A near-ideal monodispersed PSD is shown as a thinner curve in Fig. 9.16 as there is no monodispersed size in reality. There is always a dispersed distribution around the most populated size. A broader range of particle sizes, that is, polydispersed PSD, is frequently encountered (see Fig. 9.16). Regardless of whichever scenario, the most interesting aspect is that separation performance depends on the feed PSD. Fig. 9.17 sketches the behavior of various forms of solids recovery or solids capture by centrifugation corresponding to their counterpart feed size distribution in Fig. 9.16. For monodispersed PSD, the solids recovery is usually very high for small Le number and it can drop off precipitously at large Le. On the other hand, for a polydispersed PSD, the solids recovery spreads out across a wider range of particle sizes and there is no critical threshold Le as with monodispersed particles, for which there is a sudden change %Cumulative undersize 100%

Ideal mondispersed PSD “Real-World Near” mondispersed PSD Polydispersed PSD (fragmented cells, debris, etc.) Particle size

0

Figure 9.16 Size distribution of feed. %Solids recovery Operation 100%

Ideal mondispersed PSD “Real-World Near” mondispersed PSD

Polydispersed PSD (fragmented cells, debris,

0

Le

Figure 9.17 Separation behavior in accordance to the different particle size distribution.

Selection and Sizing of Centrifuges

223

in solids recovery. It is understood that after feed rate exceeds a critical rate all the particles cannot settle for a monodispersed PSD, this leads to a rapid degradation in performance. Chapters 15
17 consider various analytical forms of PSD and their resultant solids recovery, also expressed in analytical form. These analytical expressions are very useful as they allow the users to explore the effect of feed size distribution on the solids recovery and size recovery as presented in Chapter 11, Visualization and Modeling of Flow and Separation in Tubular Centrifuge, and Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification.

9.4

Performance of Tubular Centrifuge

Finally, Fig. 9.18 shows the separation test results from tubular centrifuges of different designs and sizes [4], all plotted with a parameter that is proportional to the governing Le number, and is expressed herein as c Le. Despite the abscissa has a variable c(Le) with c being a constant, this does not change the shape of the curve, because log(cLe) 5 log(c) 1 log(Le). Therefore, the curve is at most shifted horizontally by a constant log(c) in a log-log plot. Fig. 9.18 bears an important message that all the test data on tubular centrifuges irrespective of the geometries, designs, dimensions, feed rates, and G’s are all correlated by the 100%

0%

P12 Powerfuge pilot V12 V12 (SP)

10%

90%

ViaFuge pilot ViaFuge pilot (SP)

1%

0.1% 0.01

99%

0.1

1

cLe

10

Total recovery in concentrate

Total recovery in centrate

Powerfuge pilot (M)

99.9% 100

*Note c=constant

Figure 9.18 Scale-up of tubular centrifuge tests with Le number. Adapted from T. Simpson, Semi-continuous centrifuges, characterization, modeling, and scale-up, in: Proceedings to the American Filtration and Separations Society Annual Conference, Galveston, April 19
22, 2002, reproduced with permission from Am. Filt. & Sep. Soc.

224

Centrifugal Separations in Biotechnology

Le number. The left coordinate refers to the recovery of solids in the centrate Re (see Eq. (9.11)), while the right coordinate (with axis reversed) refers to recovery in the sediment/concentrate Rs. ws M e we wf 5 Re 5 ws 5 1 2 R s Mf wf 12 we 12

(9.13)

Re and Rs are based on measurements of solids concentration, respectively, in the three streams: feed Wf, centrate We and concentrate Ws under steady-state. Eq. (9.13) is derived based on material balance in Chapter 8, Laboratory and Pilot Testing, assuming all test data are taken at steady-state. The test results in Fig. 9.18 demonstrate clearly that operating at large Le number results in high centrate recovery or poor solids recovery in the concentrate/cake and vice versa. This is independent of the specific tubular centrifuge design so long as the properties of the feed to the centrifuge are the same, and this implies that the feed PSD is the same for all the tests shown in Fig. 9.18.

9.5

Summary

Selection and sizing of centrifuges for biotech applications have been presented and discussed based on the underlying principles. The sizing approach will be apparent when models are being developed in later chapters. Spintube, disk stack, tubular, chamber bowl, and decanter centrifuges can be all scale-up by the appropriate dimensionless Le number. Le depends on the specific set of geometric and operating parameters for each type of centrifuge. Based on the size of cells to be separated, we can determine the cut size, which is linearly related to Le, which in turn depends on the geometry and operation of the centrifuge. The cut size needs to be smaller than the cell sizes to effect separation. If there are any changes in viscosity or solids concentration, these affect Le from which the feed rate and G-force need to be modified. Sizing has been made on centrifuges for use in separating small organelles (0.1
1 μm) and biological cells (1
20 μm and bigger).

References [1] W.W.F. Leung, Separation of dispersed suspension in rotating test tube, Sep. Purif. Technol. 38 (2) (2004) 99
119.

Selection and Sizing of Centrifuges

225

[2] W.W.F. Leung, Industrial Centrifugation Technology, McGraw-Hill, New York, 1998. [3] W.W.F. Leung, R. Probstein, Lamella and tube settlers—Part 1 Model and operation, I&EC Process. Des. Dev (1983). [4] T. Simpson, Semi-continuous centrifuges, characterization, modeling, and scale-up, in: Proceedings to the American Filtration and Separations Society Annual Conference, Galveston, April 19
22, 2002.

Further Reading Leung, 2002 W.W.F. Leung, Plant Design Handbook for Mineral Processing, Soc.Min. Eng., 2002, pp. 1262
1288.

Problems (9.1) An application in-hand is a fermentation broth with a slightly higher feed concentration of 35% v/v solids. Based on the size of the fermenter, the feed rate to the separation and recovery phase immediately downstream is 50 L/min. The objective is to remove as much yeast as possible with no more than 1.0% of the feed yeast escaped with the centrate that contains the valuable intracellular protein. What equipment would be recommended for this application? (9.2) A slurry has 10% solids and the valuable in liquid need to be recovered to have a high yield of at least 90%. Some of the valuables may be adhered to the solids that require washing. What are the flow sheet and equipment required to do the job if the feed rate is 60 L/min? (9.3) A feed suspension of yeast has a mean particle size of 8 μm. The spintube is loaded with suspension with a liquid depth of 0.04 m. The viscosity of the suspension is 6 cP and the density of the liquid is 1000 kg/m3, while that of the solids is 1010 kg/m3. The spintube is run at 3000 g for 4 minutes. What is the Le number for this process separation? What is the cut size in microns? How would this compare with the yeast that needs to be separated by centrifugation? (9.4) The same yeast suspension is tested in another spintube with a 0.045 m depth suspension. The maximum centrifugal acceleration

226

Centrifugal Separations in Biotechnology

(9.5)

(9.6)

(9.7)

(9.8)

(9.9)

for this centrifuge is only 2500 g. What should be the centrifugation time so as to remove the yeast with the same recovery or capture of the cells as in Problem (9.3)? Another application is to fully recover mammalian cells in a broth from harvesting a bioreactor. The minimum particle size in the broth is 9.6 μm and the maximum size is 20.5 μm. A highspeed spintube capable of operating at 5000 g with a sample depth of 0.03 m is used. The broth viscosity is 50 cP and liquid density 1000 kg/m3. The solids density is 1050 kg/m3. What is the minimum centrifugation time in order to fully recover the mammalian cells in the pellet of the spintube? The small-distance settling in a disk stack, where adjacent disks are spaced apart by 1 mm or less, can be mimic by the small-distance settling in a spintube centrifuge in the laboratory. Using a spintube, the smallest possible liquid depth that can be tested with meaningful results is 0.01 m and the highspeed spintube centrifuge in the laboratory is capable of accelerating up to 12,000 g. All the process properties are identical to Problem (9.5), what is the minimum centrifugal acceleration one can operate in order to make a separation in, say, 20 seconds? A 400-mm diameter disk with a disk stack geometry with 100 disks having an inner disk radius 8.85 cm and an outer disk radius 15 cm, capable of operating at 10,000 g, is used in the pilot demonstration. The conical disk makes an angle with the vertical of 40 degrees with the vertical and the overall efficiency of the disk-stack centrifuge is 80%. The unit is used to run the mammalian cell broth of Problem (9.5) with 100% recovery of all the cells. What is the maximum feed rate to the disk-stack centrifuge? The very same centrifuge of Problem (9.7) is used to process yeast suspension upon harvest. The density of yeast is 1080 kg/m3, while that of liquid is 1000 kg/m3. The viscosity of the liquid is 20 cP. The minimum yeast particle in suspension is at 6 μm. Determine the maximum feed rate to the centrifuge? A tubular bowl has a diameter of 300 mm and length 457 mm. The pool depth is measured at 3.57 mm from the inner bowl radius. The feed solid has a density of 1050 kg/m3, while that of liquid is 1000 kg/m3. The viscosity of the suspension is 50 cP. The operating G is at 20,000 g. At a feed rate of 5 L/min, (1) what is the dimensionless Le number for the

Selection and Sizing of Centrifuges

227

tubular centrifuge? (2) What is the smallest cell size to be captured by centrifugation? (9.10) The same tubular centrifuge as described in Problem (9.9) is used to process mammalian cell suspension, the tubular needs to capture or recover even the smallest mammalian cell within the population with size ranging between 9.6 and 20.5 μm. What is the maximum feed rate to the tubular centrifuge that can still capture the finest cell to achieve a high solids recovery?

10 Troubleshoot and Optimization Troubleshooting and optimization of production centrifuges are often dealt hand in hand. A centrifuge that has troubleshoot for various mechanical and process problems may often require process optimization. On the other hand, at any time optimization can be made on a centrifuge that has been performing well to provide process fine-tuning. This is especially upon a new centrifuge being just installed and commissioned to run for the process.

10.1

Troubleshooting

10.1.1

Timescale of Occurrence

It is important to identify a problem as to the timescale for which it occurs. In most cases it would be sufficient to know if the problem is due to long-term degradation (i.e., condition degrading over a long period of time, such as several months) or it happens quickly in the past few days or few hours. Knowing the timescale often allows one to diagnose the root cause of the problem and come up with the solution. For example, the centrifuge may operate fine one day but fail to perform the next day. The centrifuge after all is a mechanical device; it takes what is being fed and reacts to the feed accordingly! So it might very well be the feed slurry that causes the centrifuge to produce offspec outputs, centrate or concentrate, and this has nothing to do with the machine. So the characteristic of the feed suspension to the centrifuge needs to be monitored constantly at all times, whether the machine is performing well or not. This enables an operator to rule out a change due to the process or a mechanical fault. On the other hand, if both the centrifuge and the process results are both deteriorating over time, one needs to investigate changes in the Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00010-6 © 2020 Elsevier Ltd. All rights reserved.

229

230

Centrifugal Separations in Biotechnology

process (feed changing slowly) or mechanical condition of the centrifuge (such as vibration noise increasing over time), and determine whether these problems are coupled or isolated. 10.1.2

Mechanical or Process Problem

As mentioned briefly in the foregoing, problems can be divided into two types: mechanical or process related. If a problem can be clearly identified as one or the other, a solution can be developed promptly. On the other hand, these two problems can be coupled in a complicated way; a problem that appears to be process related is actually caused by a mechanical component failing, or a mechanical-related problem is actually due to process problems. This is unlike an independent mechanical problem, such as a missing or loose bolt causing mechanical vibration, or an independent process problem, such as a concentrated feed leading to buildup of solids rapidly at the solids-holding area for an intermittent disk centrifuge that further triggers frequent concentrate solid discharge/shots. In this situation it would still be constructive to determine which of the two is the root cause of the problem, and to come up with appropriate remedial plan to solve it. 10.1.3

Process Problems

10.1.3.1 High Centrate Turbidity 10.1.3.1.1

High Feed Solids Throughput Causing High Centrate Turbidity This is measured by increasing dry weight of solids per unit time (kg/h DS) (DS represents dry solids). This may be due to higher concentration of suspended solids from a bioreactor or fermenter. The operator should reduce feed rate (if possible) or increase G-force. Another possibility of higher feed solids throughput may be due to higher volumetric feed rate. The speed of the centrifuge or G-force should be increased (if possible) for this case. Obviously, the electrical power drawn will increase as P ~ ρΩ2Q, where ρ is the density of suspension, Ω is the angular speed, and Q is the volumetric feed rate; and all three quantities are increasing. When either the feed rate increases and/or the feed suspended solids concentration increases, these lead to higher viscosity and more hindered settling, therefore the dimensionless Le number increases (see Eqs. (9.4) and (9.7) for a disk stack centrifuge). The operating point (Le, Rs) will shift along the Rs versus Le curve toward the increasing Le direction from point Le1 to Le2. The solids recovery decreases from Rs1 to Rs2. This is illustrated in Fig. 10.1. If the speed or G-force increases, and/or

Troubleshoot and Optimization %Solids recovery

231

Nominal operating condition, Q, RPM

100%

Higher Q or higher solids conc. (Le2>Le1, R s2>1 L

Particle trajectories 1. Baffles

Feed D

Figure 11.1 Schematic of the tubular centrifuge showing baffle arrangement on the right side and no baffles arrangement the on left side. Rotation axis Feed

Rp

h

Ri

H

Effluent

Moving layer

U

Quiescent pool Bowl head

Pool depth

weir

Rb Centrifuge bowl

Bowl head

L

Figure 11.2 Schematic of the tubular centrifuge showing the moving layer flowing over an otherwise stagnant pool.

Fig. 11.4 shows one possible setup wherein water was first introduced into a rotating bowl at 150 g with a pool depth of 45 mm measured radially from the bowl wall (i.e., bowl inner radius). At start of the experiment t 5 0, the red dye introduced immediately stretched longitudinally like a needle showing the Taylor-Proudman column [2], or more generically in the rotating flow as exhibiting a two-dimensional flow pattern atypical of geostrophic flow (pressure gradient balancing the Coriolis effect). This needle-like pattern aligned with the axis of the rotating bowl and lasted for about 60 seconds in the experiment until diffusion took over and started smearing the flow pattern.

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

245

Collection funnel

Dye or dilute acid (0.015 mol/L) trace

V1

Pipe 3

V2

Pipe 1

V3

Pipe 2

Feed

Feed V4 Centrifuge

180-mm and Pipe 2 270-mm diameter bypass bowls

Figure 11.3 Arrangement of dye injection before and after filling the bowl. (A)

(B)

t=0s Bowl ID painted, dye injected at t = 0 (C)

t=5s “Axial needle” (D)

t = 15 s

t = 60 s

“Axial needle”

“Axial needle”

Figure 11.4 Bowl inner diameter painted white for visualization. Red dye introduced to a fully accelerated pool at 150 g and 45-mm pool depth: (A) t 5 0, (B) t 5 5 s, (C) t 5 15 s, and (D) t 5 60 s.

Another flow visualization experiment is described in the following with results shown in Fig. 11.5A and B. The pool was initially filled with black ink, and after filling and accelerating to speed, the feed was changed to water feed. A white pillar attached to the rotating bowl wall showed that a portion of the pillar close to the pool surface (covered by a thin transparent water layer) is exposed without being darkened by the initial “black” pool.

246

Centrifugal Separations in Biotechnology

Figure 11.5 Water introduced to an otherwise “black” pool of liquid. Moving layer is observed by white pillar co-rotating with bowl serving as background: (A) the ruler placed at 5:30 position on the front stationary cover and (B) the ruler placed at the 8:40 position. (A)

(B)

Moving layer visible with florescence dye Stagnant black pool

Figure 11.6 Florescent dye illumination: (A) far view showing pool (black) depth and thin moving layer highlighted in green by florescent dye under UV illumination and (B) close-up view of moving layer.

This layer thickness for a given G and feed rate can be measured by placing a ruler on the Plexiglas weir and sighting, without parallax, the exposed section of the pillar using a strobe to stop the rotating motion. Alternatively, florescent dye was introduced instead of water as a continuous feed. Using an ultraviolet light, an annular ring of bright color reflected from the florescent light was visible, as depicted in Fig. 11.6A and B. Strobe light was not used for visualization, and the moving layer appears as a florescent circular ring with infinite images

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

247

overlaying on each other, wrapping around 360 degrees. This ring thickness “averaged out” over many revolutions of the bowl can be measured and is shown in Fig. 11.6. Finally, one more experimental configuration has been set up that facilitates the quantitative measurement. Instead of using dye, an acid pool was introduced into a rotating bowl. After filling, the acid feed was immediately replaced by water at t 5 0. The conductivity of the effluent liquid leaving the centrifuge was monitored starting at t 5 0. One of the typical results is shown in Fig. 11.7. The conductivity or acid strength in the effluent liquid decreases over time as the pool of acidic liquid was being flushed out from the rotating bowl. After two pool volumes, the conductivity of the effluent drops to the base level and stays there, even for as much as 30 times the pool volumes. The feeding of the bowl with water stopped and the bowl was allowed to decelerate until the pool collapsed (when the centrifugal acceleration is unable to maintain an annular pool with gravitational effect coming into play) with a lot of sloshing and mixing of the liquid inside the bowl. The conductivity level of the effluent liquid immediately jumped up to 0.6 units despite it having been practically at zero after feeding two pool volumes of water. The results of this experiment strongly suggests that the flow pattern of the feed to the centrifuge is in the form of a thin moving layer with little interaction between the moving layer (water without acid) and the rest of the rotating pool 2.00

Normalized conductivity. (baseline removed)

1.80

Initial value

1.60

267-mm ID bowl 140g @ 221 mL/s

1.40 1.20 1.00

Collapse conductivity = 0.6

0.80 0.60 0.40 0.20 0.00 0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

2.25

30.00

Number of pool volumes fed @ 221 mL/s

Figure 11.7 Acid or conductivity versus the total number of pool volumes feeding the centrifuge bowl.

Centrifugal Separations in Biotechnology Moving layer thickness (mm)

248

35 30

Conductivity tests

25

Florescent dye

20 15 10 5 0 0

100

200

300

400

500

Q (mL/s)

Figure 11.8 Comparing moving layer by conductivity test and florescent dye.

(acidic liquid), as represented in Fig. 11.2. Assuming a moving layer model as depicted in Fig. 11.2, the equivalent moving layer thickness can be deduced, based on the initial and final conductivity of the effluent, provided there was thorough mixing between the moving layer and the rest of the “stagnant” pool during pool collapse. Fig. 11.8 compares test results on the florescent dye technique and the conductivity technique for several feed rates of a given test bowl geometry under a fixed G-force. The moving layer thickness, about 5
30 mm, increases with increasing feed rate, 2.4
2.7 L/min. There is a reasonable agreement between the two techniques. Based on this, a moving layer model has been adopted similar to the one originally [1], but with further enhancement in that the feed particles are initially distributed uniformly across the entire moving layer and not concentrated at one radial position, that is, the surface, of the moving layer.

11.2

Improved Moving Layer Flow Model

In the improved model [3], instead of lumping all suspended solids at the pool surface, a more reasonable and likely assumption is that solids are distributed uniformly across the thin moving layer. This thin moving layer, which dictates separation, is at the pool surface with radius Rp. Particles or solids settle across the moving layer to a relatively quiescence zone and form sediment adjacent to the bowl wall, while feed in the moving layer continues to flow toward the overflow weir. To simplify the analysis, consider initially the case where the moving flow layer has over its entire length L (see Fig. 11.9 with the exaggerated vertical scale), a uniform thickness h, and a uniform velocity

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

249

Figure 11.9 Schematic of moving layer without underlying pool with trajectory of particles (dotted line shows limiting trajectory) in the moving layer.

profile, u(y) 5 U. Later, it will be shown that the uniform velocity profile shown on the left of Fig. 11.9, gives the same results as the more general velocity profile u(y) as shown on the right of Fig. 11.9. Furthermore, it will be demonstrated that the longitudinal changes in thickness h and the mean speed U, caused by viscous friction, have very little effect upon the results. The spatial variable along the axis of the bowl is designated as “S.” As mentioned earlier, the concentration of solids is taken to be uniform at the entrance, where S 5 0. The volume concentration of solids in suspension is further assumed to be small, such that solid particle settles independently of the others. Therefore, one can consider each particle size separately, using Stokes’ law of drag for the settling speed in the centrifugal field. In Fig. 11.9, the three slanted lines represent the trajectories of particles of a particular size x, starting at different levels in the flow layer. Since both the forward speed and the settling speed remain constant, the trajectories are straight lines. The lowermost trajectory is for a particle that is captured, and the uppermost for one that escapes with the centrate. The dashed line which starts at the level yc at the inlet of the moving layer that is just barely captured after it moves across the moving layer corresponds to the critical trajectory of particle with size x. Since all particles with size x starting below yc are captured by the moving layer and the velocity and particle concentration are both uniform, the fraction settling out of the moving layer (eventually to the cake) is represented as follows: Zs 5

yc h

(11.1)

Again, since both the speed u 5 U and the settling speed vs are uniform, simple geometry requires that yc vs 5 L U

(11.2)

250

Centrifugal Separations in Biotechnology

Now Stokes’ law for a spherical particle is as follows: vs 5

1 GðΔρ=ρÞx2 18 ν

(11.3)

where ν 5 μ/ρ is the kinematic viscosity of the fluid, and the total volume rate of flow may be expressed as follows: Q 5 2πRp hU

(11.4)

After combining Eqs. (11.1)
(11.4) and noting that the pool area A 5 2πRpL, thus Zs 5

vs A π Rp LGðΔρ=ρÞx2 5 Q 9 νQ

(11.5)

Of course, the value of Zs cannot be greater than 1, so Eq. (11.5) must be truncated at the value of Q, when Zs exceeds unity. Then, denoting Q100 as the maximum flow, at which 100% of the particles of size x are captured by the moving layer, Eq. (11.5) may be written as follows: Q100 5

π Rp LGðΔρ=ρÞx2 9 ν

(11.6)

Take further note that when Q # Q100 Zs 5 1

(11.7)

Eq. (11.6) may alternatively be interpreted as follows: Q5

π Rp LGðΔρ=ρÞxc 2 9 ν

(11.8)

Here, xc is the minimum size for which there is 100% recovery for a given size x to the moving layer when the feed rate is Q. Based on Eqs. (11.6)
(11.8), one may express the fractions to the cake and the effluent with the further proviso that Zs cannot exceed 1.0 as follows:
2 Q100 ðxÞ 5 xx ; x , xc Zs 5 1 2 Ze 5 c (11.9a,b) Q Zs 5 1 2 Ze 5 1; x $ xc sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3 νQ xc 5 pffiffiffi Δρ  2 π ΩRp L ρ

(11.9c)

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

251

1

Zs, Ze

0.8

Ze

0.6

Zs

0.4 0.2 0 0

0.5

1

1.5

2

x/xc

Figure 11.10 Capture fraction for different particle size x.

Given a particular flow rate, Fig. 11.10 illustrates how the fractions to the cake and the centrate vary with the particle size.

11.3

Effect of Velocity Profile

The dotted curved trajectory shown in Fig. 11.9 (right side of schematic) represents a particle that starts at the location s 5 0, y 5 yc and ends at s 5 L and y 5 0. It represents the particle size x that just barely get captured in the moving layer in the critical trajectory. Assuming as before that the particles are uniformly distributed in the incoming stream of the moving layer, the fraction recovered to the cake is thus given by ð yc udy (11.10) Zs 5 ð0h udy 0

The equation for the particle trajectory, dy 2 vs 5 ds uðyÞ

(11.11)

may be integrated between the starting and end points of the trajectory to give ð yc

udy 5 jvs jL

(11.12a)

0

Note that vs in Eq. (11.11) ,0 as defined in Fig. 11.9. The total volume rate of the flow within the layer may be expressed as follows:

252

Centrifugal Separations in Biotechnology

Q 5 2πRp

ðh

udy

(11.12b)

0

Combining Eqs. (11.10)
(11.12b) together with Eq. (11.3), one gets exactly the result expressed in Eq. (11.5). Consequently, the results for the improved model do not depend on the particular velocity distribution within the moving layer.

11.4

Effect of Friction Within the Flow Layer

Viscous stresses acting in the liquid layer as it flows over the stationary fluid in the pool cause an increase of speed and a decrease of the flow layer thickness as the liquid proceeds toward the centrate overflow. While this will affect considerably the flow pattern and the particle trajectories, the effects on the results worked out earlier are believed to be actually quite small for the following two reasons: Remarkably, the values of Zs and Ze derived earlier are determined only by the total flow rate, Q, and do not depend on the particular values of thickness. The thinning of the flow layer is accompanied, of course, by a downward velocity of the liquid in the layer. But this also carries the particles down with the liquid, and it is only the Stokes’ law settling speed relative to the liquid that determines the relative position of the particle within the layer at any longitudinal location.

11.5

Dimensionless Le Parameter

With the introduction of the dimensionless Le number, which is defined as pffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffi ðQ=LÞ Δρ=μ (11.13a) Le 5 ΩRp xo ηa from Eq. (11.9c), the cut size can be written as follows: xc 3 5 pffiffiffi Le π xo

(11.13b)

The significance of the Le number and the cut size in mineral separation and processing has been discussed [4
7]. Regardless of the

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

253

application, the quadratic power of Le number is simply the ratio of the feeding rate to the centrifuge to the sedimentation capacity of the centrifuge. Therefore, it can be seen that Le should be about 1 for normal operation, and Le , 1 infers that the feeding rate is less than the sedimentation capacity, while Le . 1 shows that the feeding rate is greater than the sedimentation capacity. A dimensionless analysis using the Buckingham-π theorem has been carried out to determine all the possible eight dimensionless groups that affect a separation problem in a tubular (likewise applicable to decanter) centrifuge. The results are shown in Appendix B1. Not all eight dimensionless groups affect the separation problem to equal extent. After combining the important ones, the Le number can be derived independently. The independent dimensionless analysis further complements the foregoing derivation on the Le number.

11.6

Quantitative Prediction

11.6.1

Total Solids Recovery in Cake

The total solids capture in the cake is determined using the cumulative undersize fraction of the feed Ff (x) from which the frequency of occurrence ff (x) of size x can be determined. The following relationships are derived from the size distribution and Eqs. (11.9a) and (11.9b): Rs ðLeÞ 5 Ff ðxÞ 5

ðN

ðx

0

ff ðxÞZs ðxÞdx 5

1 x2c

ð xc

ff ðxÞx2 dx 1

0

ðN xc

ff ðxÞdx 5 1 2 Ff ðxc Þ 1 Iðxc Þ

ff ðxÞdx 0 ðx ð 1 1 x IðxÞ 5 2 x2 ff ðxÞdx 5 2 x2 dFf xc 0 xc 0 Iðxn Þ 

n 1 X ðxk21 1xk Þ2 ðFðxk Þ 2 Fðxk21 ÞÞ 4x2c k51

(11.14a -d)

11.6.2

Total Solids Recovery in the Centrate

The total solids recovery in the centrate is simply the complement of that of the cake, from Eq. (11.14a)

254

Centrifugal Separations in Biotechnology

Re ðLeÞ 5 1 2 Rs 5 Ff ðxc Þ 2 Iðxc Þ

11.6.3

(11.15)

Particle Size Distribution of Supernatant/Overflow

The particle size xk in the product centrate (i.e., overflow of centrifuge) can be determined,
2 ! x dx ff ðxÞ 1 2 xc 0 F ðx Þ 2 I ðxk Þ Ff ðxk Þ 2 I ðxk Þ ! 5 f k 5 Fe ðxk ; LeÞ 5 ð
 2 xc Ff ðxc Þ 2 I ðxc Þ Re ðxc Þ x dx ff ðxÞ 1 2 xc 0 ð xk

(11.16) The centrate cumulative size distribution Fe is a function of the size xk and the cut size xc, which is a function of Le. Note that the particle size distribution (PSD) of the feed is required to evaluate the above through I(xk) and Re(xc), where Re(xc) is given by Eq. (11.15). Particle size xk can be considered as a tracer wherein the fraction of particles with a size less than xk in the effluent is being tracked as a function of the separation condition, more precisely Le. 11.6.4

Cumulative Size Recovery

In the size range between 0 and xk, the size recovery (SR) in the product centrate as a fraction of that in the feed suspension can be determined, "
2 # ð xk x dx ff ðxÞ 1 2 x c 0 I ðxk Þ ð xk (11.17) SRðxk ; LeÞ 5 512 Ff ðxk Þ ff ðxÞdx 0

Note that SR is a function of xk and Le through Ff (xk) and I(xk). In the classification of fine-particle suspension, it is clear that the feed PSD plays a vital role in size distribution, SR, and total solids recovery in the product centrate. The size recoveries would be fully utilized when we deal with the classification of two size range of biosolids in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification. In fact, Eq. (11.17) would be used to derive more useful formulae based on the bimodal size distributions

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

255

model to be presented in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification. The results are being used to define the cut size in centrifugal classification, say, lysate with inclusion bodies and cell debris. This is after homogenizing the bacteria cells, which have secreted intracellular protein in the form of the inclusion bodies (0.4
1.4 μm), with size not too much larger than the cell debris (,0.5 μm) generated from homogenizing.

11.7

Sedimentation Tests

11.7.1

Experiments on Sedimentation in Rotating Bowl Centrifuge

In this section, we will illustrate the use of the model in analyzing sedimentation testing with a centrifuge bowl. For these tests, a 1% by weight suspension of 1
10 μm silica (x50 5 3
4 μm) was fed to the centrifuge under different feed rates and rotational speeds, and the recovery was quantified by solids analysis on the three streams, the solids by weight of the feed Wf, effluent We, and the sediment Ws. Based on material balance on the centrifuge under the steady state, the solids recovery Rs can be deduced from Eq. (8.10a) from the respective three weight concentrations: We Wf Rs 5 We 12 Ws 12

(11.18)

The cutoff was determined by comparing the PSD of the feed and the effluent for each sample. Tests were carried out on the 178-mm diameter tubular bowl using 13- and 25-mm ring overflow weir (measured radial inwardly from the bowl wall). The rotational speed was 2947 rpm corresponding to 863 g measured at the bowl inner diameter. Five different feed rates of 32, 95, 189, 379, and 536 mL/s feeding the centrifuge were tested. The viscosity of the slurry is about 1 cP, and the density difference Δρ/ρ 5 1.6. The results are recast in Leb based on the bowl radius Rb instead of the pool Rp as in Eq. (11.13a), so that the dimensionless Le is independent of the pool depth. Furthermore, the effect of the pool radius parameter Rp/Rb can be examined separately. In fact, Leb/Le 5 (Rp/Rb). Fig. 11.11 shows both the range of the feed PSD measured (coarse and fine) using the laser diffraction particle analyzer.

Centrifugal Separations in Biotechnology

%Undersize

256

100% 80% 60% 40% 20% 0% 0

2

4

6 x, µ

8

10

12

Figure 11.11 Fine (lower curve) and coarser (upper curve) feed PSD of silica suspension. 100%

Improved model

90%

Rs

Earlier model

80%

25-mm pool 70%

0

1

2

Le b

Figure 11.12 Comparing the earlier model (lower curve) with the present improved model (upper curve).

Before comparing the newly improved model with the test results, the earlier model prediction (where all feed particles are placed at the surface of the moving layer) is compared with the improved model prediction in Fig. 11.12 using the same feed PSD as shown in Fig. 11.11 (for the finer PSD feed). As seen, the solids recovery is higher for the improved model (curve with bold line) compared with the earlier model (curve with thin line). The earlier model [1] assumes that all feed solids are located at the surface of the pool R 5 Rp, whereas the improved model assumes a spread out of solids across the radial zone R 5 Rp to R 5 Rp 2 h. Thus, the settling distance for the earlier model is the same for all particle sizes, which equals the thickness of the moving layer. On the other hand, in the improved model, solids of all sizes are assumed to be uniformly distributed across the moving layer. Solids closer to the bottom of the moving layer R 5 Rp 2 h settle immediately, whereas solids at the surface of the layer R 5 Rp take a longer time. As such, the total solids recovery should be higher for the improved model, and this difference can be seen for the two curves depicted in Fig. 11.12.

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

257

100%

13-mm Pool 90%

Rs

Tests - 13-mm pool 13-mm pool Repeat Model

80%

70%

0

1 1 cp viscosity

Le b

2

Figure 11.13 Comparing analytical prediction with experiments for the 13-mm pool.

The test results corresponding to the 13-mm pool run are compared with the corresponding model prediction in Fig. 11.13. Note the tests were repeated under identical conditions to confirm the reproducibility of the data, which seems to be very good. Comparing prediction with test results, the model overpredicts the solids recovery at the low rate and low rate Leb , 1 by as much as 5%. (There is a possibility that the test silica suspension contains some fine lighter-density colloids, which are difficult to settle.) However, the agreement is relatively good for Leb . 1. The results for the 25-mm deep pool are depicted in Fig. 11.14. Three sets of tests were carried out to confirm reproducibility. The PSD in the feed has been changed in the third test to a slightly coarser PSD as shown in Fig. 11.11. Thus, prediction is made for the nominal finer PSD and the coarser PSD. Both predictions (coarse and fine feeds) are compared with all three tests in Fig. 11.14. The coarser PSD leads to higher solids recovery as expected. At lower Leb, the prediction is higher than the measurement yet at much larger Leb, the prediction is below the measured. Possible explanation for the discrepancy between the theory and the test data, as shown in Figs. 11.13 and 11.14, is due to other factors not being considered in the improved model: 1. Prediction is based on the PSD as measured subjectively using the PSD analyzer, which uses laser diffraction that is an indirect measurement and the measurements are recast in equivalence to spherical particles. 2. Entrainment and resuspension of the sediment by the complicated Ekman flow on the exit ring weir may present the effect, which may be more pronounced at the high sedimentation rate or low Leb. This leads to a solid capture shy away from 100%.

258

Centrifugal Separations in Biotechnology Test 1 Test 2

100%

Test 3 Model prediction (coarse feed) Model prediction (fine feed)

Rs

90%

80%

25-mm Pool 70%

0

1 cp viscosity

1

Le b

2

Figure 11.14 Comparing analytical prediction with experiments for the 25-mm pool.

3. The very fine submicron silica particles may stay in suspension due to the double-layer electrical forces being present and also would not settle under the relatively lower G (860 g) in the tests. This behavior has been discussed in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification. The test suspension may contain difficult-to-separate colloids. 4. Other secondary flows, which negatively affect separation, are unaccounted in this model. In any event, the model predicts within 5% the performance of a rotating tubular bowl centrifuge. Useful insights are obtained from a very extensive flow visualization study of a rotating bowl with continuous feed in which the moving layer is demonstrated and quantified for the first time. To a first approximation, the thin moving layer thickness does not enter into the equation of the model, which is quite remarkable. Finally, the improved model compares reasonably well the sedimentation tests.

11.8

Summary

In this chapter, a comprehensive flow visualization program is presented for the first time on a rotating bowl centrifuge. Both dye and tracer revealed a thin moving layer present at the pool surface above a stagnant dead pool for a decanter and tubular centrifuge. The moving layer is a function of the feed rate and G. Due to the strong rotational effects, the flow exhibits two-dimensional behavior atypical of the rotating flow. The sedimentation in a centrifugal field is dictated by a dimensionless Leung number, Le, which is related to the feeding to separation capacity

Visualization and Modeling of Flow and Separation in Tubular Centrifuge

259

of the centrifuge, and this has been independently derived by the Buckingham-π analysis. A model, assuming feed particles of all sizes are uniformly distributed across the moving layer at the beginning, is presented to quantify separation. Experiments on sedimentation of fine silica slurry using a test centrifuge are presented, and test results agree well with the model predictions.

References [1] W.W.F. Leung, Industrial Centrifugation Technology, McGraw-Hill Inc, New York, 1998. [2] H. Greenspan, The Theory of Rotating Fluids, Cambridge University Press, London, 1968. [3] W.W.F. Leung, Experimental and theoretical study of flow and sedimentation of tubular centrifuge for bioseparation, in: AICHE 2005 Annual Conference, Cincinnati, OH. [4] W.W.F. Leung, Centrifugal sedimentation and filtration for mineral processing, Plant Design Handbook for Mineral Processing, Soc. Of Min. Eng., 2002, pp. 1262
1288. [5] W.W.F. Leung, Scale-up of sedimenting centrifuges, in: Richard Wakeman (ed.), Filtration Processes
Equipment Scale-Up, Elsevier, 2005, pp. 375
441. [6] W.W.F. Leung, Centrifugal classification of fine-particle suspension, in: Plenary Lecture at Filtration Society Japan Association-INCHEM 2003 Conference, Tokyo Big Sight, Tokyo, November 4
5, 2003. [7] W.W.F. Leung, Centrifugal classification of drill mud, in: Presented at 9th World Filtration Congress, New Orleans, LA, April 18
24, 2004.

Problems 11.1 It has been revealed and backed up by experimental findings that the feed is distributed in a tubular centrifuge in a thin moving layer flowing from the feed end to the centrate discharge end at the opposite side. This contrasts with the conventional wisdom of a plug flow where feed is uniformly distributed in the angular space from the bowl wall to the pool surface and sweeps through the bowl, while heavier solids settle out. What is the phenomenon behind this moving layer? What is the implication from a process standpoint of the moving layer instead of the plug flow and is the machine doing better or worse having the moving layer?

260

Centrifugal Separations in Biotechnology

11.2 Which is the most sensitive parameter among the several parameters listed in Eq. (11.13a) for the Le number? Sensitive refers to small increase in the parameter results in large change in Le, whereas insensitive refers to a large increase in the parameter, which results in small change in Le. Which are the parameters that are difficult to determine in the Le number? What is the basis of introducing a dimensionless number? 11.3 What are other effects, if any, that have not been taken into account in the tubular centrifuge model?

12 Disk-Stack Modeling In this chapter, the separation performance of a disk-stack centrifuge is modeled. A schematic of the dropping-bottom intermittent discharge disk centrifuge is shown in Fig. 12.1. The model discussed in this chapter is also applicable to the nozzle disk and the manual discharge disk. This model is validated with results from a pilot test using a disk-stack centrifuge. As shown in Chapter 13, Performance Projection of Centrifuges in Bioseparation, the disk model can be used to simulate performance of disk-stack centrifuge of different sizes operating at various feed rates and rotational speeds applicable to various bioprocess and biotech applications. A suspension of particles is continuously fed to the disk centrifuge. Under centrifugal acceleration G, heavier solids settle to the upper surface (Fig. 12.3) of the annular disk channel formed between two disk surfaces, displacing lighter liquid phase toward the small radius. The collected sediment slides down the upper surface of the disk channel under the longitudinal component of the G-force until it leaves the disk stack and is collected at the solids holding annular space of the bowl.

12.1

Disk Model

In the simple model [1] presented herein, the following assumptions are made: 1. The solid concentration is relatively dilute, such that the presence of solid does not affect the flow field. 2. The aspect ratio of the channel (longitudinal length to the channel height or spacer height between adjacent disks) is very large and the details of the entrance and exit geometry respectively at both ends can be ignored. For modeling purpose, the triangular region at the inlet of the disk channel can be eliminated; similarly the triangular region at the exit of the disk channel can be added (Fig. 12.2), so that the channel has inlet area and exit area exactly perpendicular to the longitudinal direction of the channel. By virtue of the large aspect ratio and the strong rotational field at high rotation speed, the Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00012-X © 2020 Elsevier Ltd. All rights reserved. 261

262

Centrifugal Separations in Biotechnology

Ri Ro

D n

θ

Figure 12.1 Schematic of a disk-stack centrifuge.

Flow

Figure 12.2 Entrance and exit channel region (annulus) being ignored.

flow develops quickly (i.e., for an entrance distance equivalent to several channel spacing) to a steady velocity profile. 3. The flow is uniformly distributed across all the ‘n’ channels. This assumption is very weak, as shown in Fig. 9.9. In any case, only a single channel is considered in the model. 4. The flow field in each channel (Fig. 12.3) is independent of particle sedimentation; however, the reverse holds in that the sedimentation of a particle depends on the velocity profile in the channel. This statement holds for dilute suspension and where particles are relatively small compared to the channel spacing. 5. The velocity profile in the suspension is assumed to be relatively constant across the entire channel within a given channel, despite there are Ekman layers and Ekman flow adjacent to the upper or lower surfaces

Disk-Stack Modeling h

263

Ri

Upper surface

θ

Particle

G

Ω

1 2 3

L Lower surface

y yc

z Ro

Figure 12.3 Velocity profiles and trajectory of particles in the disk channel.

of the disk surfaces and also a thin clear liquid flowing on the upper surface of the disk channel analogous to the lamella settler [2]. The flow field in the channel is assumed to be dominated by viscous effect. For a rotation speed of, say, 5000 rpm or 524 rad/s, kinematic viscosity 10 times that of water at 0.1 cm2/s, the combined viscous Ekman layers, respectively, at the upper and lower disk surfaces is of order of 2 (ν/Ω)1/2 5 0.28 mm. (The angle effect has been ignored.) Typical spacing between adjacent disks can be as small as 0.25 mm and as large as 1 mm 1 . For high performance, disk spacing is small, therefore the assumption of viscous flow in the channel is not unwarranted. Consequently, we do not consider the geostrophic flow (no viscous effect) in the channel unless the disk spacing is large, over 1 mm. Referring to Fig. 12.4, a particle in the channel is subject to convection from the main flow along z-direction, as well as relative motion due to sedimentation under G. Uniform distribution in Fig. 12.3 implies that the flow rate to each channel is identical and only a single channel needs to be modeled. We separately consider the continuum phase (which is the suspending liquid) and the particles (the dispersed phase).

264

Centrifugal Separations in Biotechnology

U Particle

vs sinθ

θ vs cosθ vs

Figure 12.4 Particle velocity in the channel.

12.1.1

Continuum Phase

The continuity and momentum balances (without considering the settling particles) imply the following. 1. Continuity: The flow rate per unit channel is equal to the overall feed rate Q divided evenly among n channels in n 1 1 disks, thus Q 5 2πnRUh

(12.1)

where R is the radius, U(z) is the local longitudinal velocity in the z-direction, and y is the transverse coordinate measured from the upper surface of the disk channel, and h is the channel height, that is, distance between adjacent disks (Fig. 12.3). This equation holds from the feed entrance R 5 Ro along increasing z to the disk channel exit at the small radius Ri 2. z-momentum: Despite that we have assumed U being constant, independent of the transverse coordinate y. To satisfy continuity or mass conservation, Eq. (12.1), the longitudinal velocity has to increase with reducing radius.
 Q=n 1 (12.2) UðzÞ 5 2πh R 12.1.2

Dispersed Phase

The dispersed phase is sufficiently dilute such that the presence of particles does not affect the flow field, on contrary the rate of deposition and removal of them depend on the main flow. The relative velocity of settling particle in relation to the liquid (in relation to the surrounding liquid under steady-state) with negligible acceleration and deceleration is given by Stokes’ law as discussed in Chapter 2, Principles of Centrifugal Sedimentation.

Disk-Stack Modeling

265

Given x being the particle size (equivalent spherical diameter), Stokes’ law in a centrifugal field states   ρs 2 ρL Gx2 (12.3) vs 5 18μ The above holds under 0 # x ,N, 0 # y # h, 0 # z # L. The frequency distribution f(x) of particle residing in a given size range is the derivative of the cumulative undersize distribution F(x) f ðxÞ 5 FðxÞ 5

dFðxÞ dx

ðx

f ðxÞdx

(12.4a) (12.4b)

0

The frequency size distribution f(x) or cumulative size distribution F(x) is assumed to be known from measurement. The trajectory can be determined for a given particle located at transverse location y at the inlet (i.e., z 5 0) in a flow field, as illustrated in Fig. 12.3. This approach is to determine whether the particle trajectory intercepts the upper collecting surface of the channel prior to traveling through the channel length ending at z 5 L. Consider a particle, with size x starting at z 5 0 and taking on trajectory 1, which gets settled as it travels along the disk channel. If the particle had been positioned differently at the inlet and had taken trajectory 3, it would have escaped from being settled. With reference to Fig. 12.3, trajectory 2, with particle at channel inlet z 5 0 and y 5 yc, is the critical trajectory wherein particle with size x barely gets captured as it travels across the disk channel (Fig. 12.4). Therefore the fraction captured in the sediment Zs for particle size x is Zs ðxÞ 5

h 2 yc h

(12.5)

where h is the channel height as set by the spacer between adjacent disks. Assuming particles of all sizes are initially distributed uniformly across the inlet of the channel from the limiting trajectory, the solids recovery or size capture can be determined. Another important point is to determine the cut size, that is the particle positioned at the inlet lower corner of the channel z 5 0 and y 5 0 (Fig. 12.3) that undertakes the critical trajectory such that it ends up at y 5 h at z 5 L. Particles larger than the cut size would be fully settled in the channel; vice versa, only a fraction would be settled for particles smaller than the cut size. With the above information, the fraction captured for particles of all sizes can be schematically represented in Fig. 12.5.

266

Centrifugal Separations in Biotechnology 1

Zs

Zs , Z e

0.8

Ze

0.6 0.4 0.2 0 0

0.5

1

1.5

2

x/xc

Figure 12.5 Capture fraction for different particle size x.

As stated, the particle settling velocity is Stokes’ velocity relative to the moving liquid under the influence of the G-acceleration (Fig. 12.4). There are two velocity components that should be considered: one component is perpendicular to the underside of the disk surface in which a particle is settling toward with a velocity vscosθ; another is the longitudinal component, vssinθ, which is in the negative direction of the streamwise direction. This second component should be superimposed with the streamwise flow with velocity U(R), that is U 2 vssinθ. Let us consider a small incremental path (dz, dy) along the particle trajectory. Thus it follows that dt 5

dy dz 5 vs cosθ U 2 vs sinθ

or dz U 2 vs sinθ 5 dy vs cosθ

(12.6)

From Fig. 12.3 by virtue of the geometry R 5 Ro 2 zsinθ

(12.7)

It follows that dR 5 2sinθdz and substituting this in Eq. (12.6), 2

1 dR 5 sinθ dy

b R

2 aRsinθ aRcosθ

(12.8)

where
 1 Δρ ðΩxηÞ2 a5 18 μ

(12.9a)

Disk-Stack Modeling

b5

Q=n 2πh

267

(12.9b)

In Eq. (12.9a), η is added to account for the acceleration efficiency. Note that the coefficient a is derived from Stokes’ law, whereas coefficient b is from the continuity Eq. (12.2). Integrating Eq. (12.5) along the critical trajectory where particle initially at y 5 yc just barely gets captured. The two limits have to be R 5 Ro to R 5 Ri, and y 5 yc and y 5 h, respectively. Also for all practical purposes the component of the Stokes’ velocity is assumed to be much smaller than the streamwise velocity component, thus aR2sinθ ,, b/R. ðh ð Ri aR2 cosθ dR 5 2 dy (12.9c) 2 Ro b 2 aR sinθ sinθ yc Assuming vs/U ,, 1, with substitution of a and b from Eqs. (12.9a) and (12.9b), respectively, after rearranging we get
   a cotθ  3 π ΔR ðΩηÞ2 hcotθ  3   Ro 2 R3i 5 h 2 yc 5 Ro 2 R3i x2 (12.10) b 3 27 μ Q=n Assuming uniform particle concentration across the entire channel, the fractional capture of particle with size x from Eq. (12.5) becomes

2  h 2 yc π ΔR ðΩηÞ2 cotθ  3 x   Ro 2 R3i x2 5 5 Zs 5 (12.11) 27 μ xc h Q=n where xc is defined as ffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
  Q=n 27 μ 1   xc 5 π ΔR η2 Ω2 cotθ R3o 2 R3i

(12.12)

From Eq. (12.11) the dimensionless Leung number, abbreviated as Le number, that governs the settling behavior of the disk stack can be defined as vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! u
 u 3Q μ tanθ t     3 n ρs 2 ρL Ro 2 R3i (12.13) Le 5 Ωxo η where n is the number of disks, Ro is the radius of the outer disk and Ri is that of the inner disk, θ is the angle subtended between the disk

268

Centrifugal Separations in Biotechnology

surface and the vertical axis, μ is the suspension viscosity, ρs is the solid density, ρL is the liquid density, Ω is the angular velocity, and η is the acceleration efficiency. The cut size is thus given by xc 3 5 pffiffiffi Le π xo

(12.14)

In Eq. 12.13, n is the number of disk channels. Given the number of disks should be n +1, but as n is usually very large, we refer n also as the number of disks, which is a good approximation. The rest of the governing equations are identical to the ones (i.e., solid recovery etc.) that have already been presented in Chapter 11, Visualization and Modeling of Flow and Separation in Tubular Centrifuge, given that the Zs function is identical with the tubular model.

12.2

Model Validation

It would be useful to validate the model with some experimental results [3]. A suspension of cell culture contains viable mammalian cells 1020 μm, nonviable cells 510 μm, and cell debris less than 5 μm. (In other applications after passing through the homogenizer, cell debris is usually much smaller, less than 0.5 μm.) The size distribution in terms of number of particles is shown in Fig. 12.6. The suspension was separated in a pilot disk-stack centrifuge with a bowl diameter 190 mm. The cell debris and nonviable cells are removed in the centrate and the viable cells are captured in the concentrate. Fig. 12.7 shows the centrate solids by volume leaving the 190-mm disk centrifuge plotted against the feed rate to the centrifuge. (The solids %v/v of the centrifuge centrate was determined from spindown at 14,000 g for

Number of particles

1800 1500

Cell debris

1200

Non viable cells

Viable cells

Feed

900 600 300 0 0

5

10

15

20

25

x (μm)

Figure 12.6 Particle size distribution of feed suspension [3]. Reproduced with permission from the American Filtration and Separation Society.

%Centrate solids, v/v*

Disk-Stack Modeling

0.10 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00

190-mm bowl dia. 10,000g

μ=1.9 cp, Δ ρ/ρ =0.01

Prediction

Qcrit =10.5 L/m

Tests

x c =11.4 μm

0

269

5

10

15

Q (L/m) *Samples spun at 14,000g @ 10 min

Figure 12.7 Centrate solids from a 190-mm disk centrifuge. Solid curve is prediction from the present model and test data are from Reference [3].

%Solids recovery

μ=1.9 cp, Δρ/ρ =0.01

190-mm bowl dia. 10,000g

100

Prediction

99

Tests

98

>99% recovery of solids, @ Q99.5% @Q > > >  > F1 > F ðxc Þ < 3x2c ðx2 2 x1 Þ = SReðSÞ 5 12 ðxc 2 x1 Þ > F1 > > > > F1 > : ; ðx2 2 x1 Þ Simplifying, x2 $ xc $ x1



 ðxc 2 x1 Þ 1  2 2 1 2 2 xc 1 xc x1 1x1 SReðSÞ 5 3xc ðx2 2 x1 Þ

(17.6)

When xc is at xc3 in Fig. 17.9, only the range between x1 and x2 needs to be evaluated as f 5 0 for x , x1 and x . x2 (see Fig. 17.9)  2 ð x2  2 ð x2  x F1 x 1  f dx 5 dx 5 2 x22 1 x2 x1 1 x21 F1 x x 3x x 2 x c 1 c x1 x1 2 c ð x2 ð x2 F1 fdx 5 dx 5 F1 x x1 x1 2 2 x 1 Based on Eq. (11.17), and using the above two equations, 1  xc $ x2 SRe ðSÞ 5 1 2 2 x22 1x2 x1 1x21 3xc

(17.7)

Eqs. (17.5
17.7) represent, respectively, the size recovery of the smaller size fraction in different regimes. 17.9.2

Size Recovery of Large (L) Particles in Centrate

When xc is at location xc3 in Fig. 17.9, by similar argument as with the smaller size fraction xc , xo

SRe ðLÞ 5 0

(17.8)

Classifying Bimodal PSD and Case Study of IB Classification

401

When xc is at xc4 as shown in Fig. 17.9, similar argument applies as with the smaller size fraction.

  xc 2 xo 1  (17.9) 1 2 2 x2c 1xc xo 1x2o xm $ xc $ xo SRe ðLÞ 5 3xc xm 2 xo One can also perceive that the results for the larger size are similar to those for the small size except that xo replaces x1, and xm replaces x2. Note that the fraction F1 for small size and the fraction 1
F1 for the large size both do not enter into the results as they cancel out each other in the derivation. When xc is at location xc5 as shown in Fig. 17.9, we obtain the result, xc $ xm

SRe ðLÞ 5 1 2

1  2 2 x 1x x 1x m o m o 3x2c

(17.10)

Eqs. (17.8
17.10) represent, respectively, the size recovery for the larger size fraction in different regimes. Therefore, Eqs. (17.5
17.7) and (17.8
17.10) represent the basis of size recoveries for the small and large fractions, respectively. In the following, we will use a number of different examples to illustrate their use in optimizing centrifugal classification. 17.9.3

Example on Classification

Suppose the particles in the bioreactor (e.g., mammalian cells bioreactor flow sheet in Fig. 6.3A), or crystallizer (e.g., insulin flow sheet in Fig. 6.7) produces a bimodal size of biologic solids or crystals with cumulative size distribution F shown in Fig. 17.10A. This corresponds to a bimodal feed distribution shown in Fig. 17.8A for which the solids recovery has already been computed in Fig. 17.8B. The solids recovery versus Le for a wider range of Le is shown in Fig. 17.10B. Unfortunately, it is often perceived that when the centrifuge is tuned such that the cut size is about 10 μm, the smaller size fraction of 2
3 μm should all be in the centrate, while the large fraction sizes are all collected in the sediment of the centrifuge. This is only partially correct. As can be seen in Fig. 17.10B, at Le 5 0.591 with cut size xc 5 10 μm, the total solids recovery in the sediment of the centrifuge Rs 5 0.738, and not theoretically 0.72 (51 2 F1 5 1 2 0.28). This indicates that the sediment also contains some finer solid of 2
3 μm being settled in the sediment. Precisely, the difference of the two recoveries, ΔRs 5 0.018 (50.738
0.72), is a measure of the loss of small solids settled together with the large size solid to the bowl wall. It will take

402

Centrifugal Separations in Biotechnology

(A)

Small biologics 2–3 μm, Large biologics 10–11 μm

1 0.9 0.8 0.7 F(x)

0.6 0.5 0.4

F1 = 0.28

0.3 0.2 0.1

Xc = 10μm

0 0

5

10

15

20

25

x, μm

(B)

1 0.018

0.9 0.8

0.738

0.7

0.72

Rs

0.6 0.5 0.4 Le = 0.591

0.3 0.2 0.1 0 0

1

2

3

4

Le

Figure 17.10 (A) Bimodal size distribution of solids from bioreactor or crystallizer. (B) Total solids recovery in the sediment or cake in the classification centrifuge.

theoretically infinite large Le for all the smaller size fraction to leave with the centrate in order to avoid sedimentation. However, under large Le, the larger size fraction would have got settled as well! The solids recovery curve Rs in Fig. 17.10B cannot answer the details on the classification by the centrifuge. Indeed, this is the basis of the size recoveries developed in the foregoing that can address this important issue. We can use Eqs. (17.5
17.7) to calculate the size recovery SRe(S) of the small fraction in the centrate, and these are shown in the semi-log plot Fig. 17.11. The SRe curves have a characteristic S-shape in the semi-log plot, which further amplifies the sensitivity of the abscissa

Classifying Bimodal PSD and Case Study of IB Classification

403

scale at small Le. (Note the corresponding Cartesian plot does not have the S-shape curve but with a steep rise in the curve at small Le.) Similarly, Eqs. (17.8
17.10) are used to calculate the size recovery SRe(L) of the larger size fraction in the centrate stream (see the S-shape dash curve). This represent a presence of the ‘minority’ of larger size fraction in the centrate stream contaminating predominantly the finer size fraction in the centrate. Alternatively, this scenario can represent a loss of valuable large size fraction in the unwanted/waste centrate as well. Note the SRe(L) starts at Le 5 0.591 corresponding to the cut size with xc 5 10 μm. This is because of Eq. (17.4) that states xc 3 5 pffiffiffi Le π xo

(17.4)

Note xo is also taken as 10 μm (Le is based on xo). Therefore, when xc 5 xo 5 10 μm, Le 5 π1/2/3 5 0.591. We need to examine closely the condition of xc 5 10 μm, and this is given in Fig. 17.12. In the latter, when Le 5 0.591 with xc 5 10 μm, SRe(L) 5 0 and SRe(S) 5 0.937, that is, 6.3% of small size fraction in the feed is lost in the sediment by settling. When Le 5 0.6 with xc 5 10.15 μm, SRe(L) 5 0.23% and SRe(S) 5 0.939, that is, 6.1% of small size fraction in the feed is lost in the sediment. The increase in size recovery of the small fraction is 0.2% (593.9%
93.7%), which is quite small as compared to the base level of Rs 5 0.937. On the other 1

93.7%

0.8 0.7 0.6

Le = 0.591

SRe(S), SRe(L), SRs(S), SRs(L)

0.9

0.28x 0.063 = 0.018

0.5 0.4

SRe(S) SRe(L) SRs(S) SRs(L)

0.3 0.2 0.1

6.3%

0 0.1

3.0

1.0

Le

Figure 17.11 Size recoveries in centrate and sediment for, respectively, the small- and large-size fraction.

Centrifugal Separations in Biotechnology 100% Centrate product 99%

1.0% Le = 0.6

SRe(S)

98% SRe(L)

97%

0.8%

SRe(S)

96%

0.6%

95% 94%

93.9% Small size Le = 0.591

92% 91% 90% 0.50

0.4%

93.7%

93%

0.52

0.54

0.56

0.58

SRe(L)

404

0.23% Large size

0% Large size 0.60

0.62

0.64

0.2%

0.0% 0.66

0.68

0.70

Le

Figure 17.12 Zoom-in of SRe(S) and SRe(L) near Le 5 0.6.

hand, the size recovery in the centrate of the larger size creeps up significantly from the initial level of 0% to 0.23%, which is considerably large. This result is attributed to the slow increase of the SRe(S) curve in contrast with the fast increase of the SRe(L) curve near xc 5 xo. Therefore, it is preferable to operate at Le 5 0.591. How accurate can the centrifuge be controlled to operate at Le 5 0.591 versus Le 5 0.6 is another matter? But the lesson learnt is that the increase in SRe(S) is very marginal as compared to much greater increase in the unsettled large solids that are introduced into the centrate product when the cut size is at the minimal size of the large size fraction. Therefore, it is preferable to operate the centrifuge perhaps falling short of xc 5 xo 5 10 μm with Le , 0.591 where possible. In other words, Le can be operated at 0.58 and xc can be at 9.8 μm and SRe(S) 5 0.934 which has a built-in safety margin preventing larger particle size contaminating the centrate. This recommendation could not have been made by analyzing just the total solids recovery given by Fig. 17.10B. This brings out the importance of determining the size recovery SR as is done here. At Le 5 0.591, the loss of fine fraction in the sediment is 6.3% but the fine fraction of the feed is only 28%; therefore, the loss of undesired fine in the sediment (contamination) is 0.063 3 0.28 5 0.018 or 1.8% of the total feed solids. This 0.018 is precisely the discrepancy that has been noted earlier by the total solids recovery in Fig. 17.10B. We can also examine closely the sediment assuming the larger fraction is indeed the valuable product and one wants to minimize the entrainment

Classifying Bimodal PSD and Case Study of IB Classification 100.0%

405

10%

100%

9% 8% 7% 6%

99.0%

5%

Le = 0.591

SRs(L)

6.33% Small size 6.13%

SRs(L)

3%

SRs(S)

Le = 0.6

98.5%

4%

SRs(S)

99.5%

99.75% Large size

2% 1%

98.0%

0%

0.58

0.59

0.60

0.61

0.62

0.63

0.64

0.65

0.66

Le

Figure 17.13 Zoom-in of SRs(S) and SRs(L) near Le 5 0.6.

of the smaller size in the sediment product. This can be seen by calculating the complement of SRe which is SRs (51 2 SRe). This is shown in Fig. 17.13 for SRs(S) and SRs(L) near Le 5 0.6. At Le 5 0.6, SRs(L) 5 99.75% with SRs(S) 5 6.13%. On the other hand, at Le 5 0.591, SRs(L) 5 100% with SRs(S) 5 6.33%. The increase in small size in the sediment is insignificant only 0.2% (56.33%
6.13%), while the solids recovery of the large size can reach 100%, that is, no loss of the large size product in the unwanted centrate. Therefore, operating at Le 5 0.591 is preferred. We can again see the large change of SRs(L) and insensitive change in SRs(S) similar to the case of the centrate. Also, due to difficulty of controlling precisely at a given prescribed condition, it is preferred to operate with Le , 0.591, such as Le 5 0.58 for which SRs(S) 5 6.595% and size recovery for the large size remains at 100%. 17.9.4

Smaller Size Fraction Further Apart From Larger Size Fraction in Bimodal Feed

When the two size fractions, small and large sizes, are further apart, the size cut can be much improved. As an example, the feed size distribution of Fig. 17.4A is used, where the debris (1
1.1 μm) should be referred in this context as the small size biologics and the whole cell should be referred in this context as the large size biologics. The total solids recovery has already been shown in Fig. 17.4B. As can be seen,

406

Centrifugal Separations in Biotechnology

there is a much wider condition of operation for Le where the Rs curve is practically constant. Classification can be carried out by operating at Le 5 0.591 with the centrifuge centrate primarily containing the small size product and the centrifuge sediment containing the large size product, see the size recovery curves in Fig. 17.14. At Le 5 0.591, the centrate has size recovery SRe(S) of 98.9% (which is much higher compared to 93.66% with previous example, see Fig. 17.11) with no large solids in the centrate, SRe(L) 5 0. The sediment product has size recovery of small solids SRs(S) 5 1.1% (which is much smaller compared to 6.33% with previous example, see Fig. 17.13), and the sediment recovers 100% of the large-sized solids. This improved classification or classification is all attributed to the two solid size fractions being much further apart. Here, we have again seen the benefit of using the theoretical model to optimize the centrifuge to perform the classification of biosuspension that contains a bimodal feed solid. The smaller size can be the product that can be separated in the centrate, or alternatively, the large size can be the product that can be collected in the sediment. Regardless, there is always a cross contamination/loss of solids in the centrate and sediment stream. With known feed size distribution, one can carry out a process design to use a suitable type and properly sized centrifuge, or multiple centrifuges, to perform the classification. The operation of the centrifuge can be tuned when the upstream bioreactor has upsets or operate off1

98.9%

0.8 0.7

SRe(S) SRe(L) SRs(S) SRs(L)

0.6 0.5 0.4

Le = 0.591

SRe(S), SRe(L), SRs(S), SRs(L)

0.9

0.3 0.2 0.1 1.1%

0 0.0

0.5

Le

1.0

Figure 17.14 Size recoveries in centrate and sediment for, respectively, the small- and large-size fraction for the feed shown in Fig. 17.4A.

Classifying Bimodal PSD and Case Study of IB Classification

407

spec feed to the centrifuge as has been seen in the foregoing. The bimodal simulation provides the centrifuge to have wider applications in process flow sheets when processing biologics with potential variability in the upstream bioreactors and fermenters. The model can predict the ill effects of off-spec feed and be able to recommend temporary compensation as suggested by the results of the model simulations. This first part will be seen in the next example. 17.9.5

Smaller Size Fraction Closer to the Larger Size Fraction in Bimodal Feed

When the two size fractions, small and large sizes, are approaching each other, upon centrifugation the separated small size solids in the centrate inherently contain the large solids while the separated largesize solids in the sediment inherently contain the small solids. Despite this, the classification model discussed in this chapter can still provide guidance for the best possible operation for this difficult classification as there is no ‘clean cut’ to obtain pure size fraction in the centrate and concentrate streams. . As an example, the feed size distribution of Fig. 17.8A is used, where the small-size biologics is between 5 and 6 μm and the large-size biologics remains at 10
11 μm. The size recovery curves are given in Fig. 17.15. At the cut size of 10 μm with Le 5 0.591, the size recovery of the small sizes in the centrate is only 69.7% because 30.3% is lost to the sediment. It takes much larger Le (c0.591) in order to reduce the loss in the sediment of the small size; however, this entrains the larger size into the centrate stream as well. If Le approaches 1, the size recovery of the small size in the centrate increases to 90% with 10% loss in the sediment, yet the large size particles in the centrate can be as high as 62.7% (see S-shape dash curve for SRe(L) in Fig. 17.15) which is unacceptable! Therefore, the best classification is still at Le 5 0.591 with 69.7% of the recovered smaller size solids in the feed yet without large solids (10
11 μm). Table 17.2 summarizes the results corresponding to the three different scenarios with small sizes, respectively, 1
1.1, 2
3, and 5
6 μm and in all three cases with the same large size biologics 10
11 μm. It is clear that the size recovery of the small size in the centrate is much greater when the small size range is spread out further away from the large size range. Table 17.2 shows, corresponding to the three small size groups, the size recovery of the small size in the centrate are respectively, 98.9%, 93.7% and 69.7% with the centrate free of the large-size solids. In

408

Centrifugal Separations in Biotechnology 1

0.8

69.7%

0.7

SRe(S)

0.6

SRe(L)

0.5

SRs(S) 0.4

SRs(L)

30.3%

0.3

Le = 0.591

SRe(S), SRe(L), SRs(S), SRs(L)

0.9

0.2 0.1 0

0.1

1.0

3.0 Le

Figure 17.15 Size recoveries in centrate and sediment for, respectively, the small size fraction (5
6 μm) and large-size fraction (10
11 μm) for the feed shown in Fig. 17.8A. Table 17.2 Size recovery for three different ranges of smaller biologics but with same large size biologics, 10
11 μm, SRs(L) 5 100%. Small size (μm)

SRe(S)

SRs(S)

1 2 1.1 223 5
6

98.9% 93.7% 69.7%

1.1% 6.3% 30.3%

all cases, the centrifuge is operated with Le 5 0.591 (based on xo) such that xc 5 xo. Any departure from this ideal condition results only in poorer classification. Next, we will apply the bimodal model simulation to study the IB cell classification, which is a very important step for intracellular protein production primarily by E. coli and other microorganisms.

17.10

Classification of Inclusion Bodies

Inclusion bodies (IB) secreted by bacteria, such as E. coli, have many different applications, including producing functional and therapeutic polypeptides, which often aggregate as refractile and insoluble clusters [2]. These include a variety of protein hormones secreted from human

Classifying Bimodal PSD and Case Study of IB Classification

409

amyloidal secretary granules [3], growth factors [2], enzymes, and cytokines [2]. The protein stored in the inclusion bodies are more active than was thought in the past [4]. Because of its mechanical stability and protein functionality, it is being used as enzymes, immobilized biocatalysts, protein delivery into mammalian cells in proteinreplacement therapies, and as nanopills [5]. Further, given the protein is expressed intracellularly, despite the protein might be toxic, or has toxin, it does not harm the host cell given it is localized and contained in the IB. One serious drawback is the low yield due to lysing of the bacterial cells to release the IB. The IB and the cell debris from lysing by homogenizing at elevated pressure are very fine. IB cells are typically under 1 μm and below but it can span a range between 0.5 and 1.3 μm [6
8]. A conventional production is to lyse the bacteria cells to release the IB. Buffer solution is added to the lysed solution, and the cell debris is separated from the IB cells through a series of washing and separation by centrifugation as depicted in Fig. 6.8. Subsequently, the IB cells are dissolved in proper solvent and being refolded to the proper configuration [4,9]. The entire sequence of homogenization, separation, including adding buffer solution for washing and the solubilization, becomes the bottleneck for the production of IB cells [2,4]. If the cell debris is not too much smaller than those of the IB cells, classification becomes problematic as has been discussed. The problem is especially aggravated when the IB cell density is not too far off from that of the cell debris. The following example illustrates how the conventional process may run into a low yield situation. Subsequently, a more novel approach will be presented in that the cell debris can be made even finer, say, less than 0.1 μm by intense homogenization followed by centrifugation. With proper centrifugation by classification, cell debris can be reduced to minimal far beyond that can be attained with conventional approach. The minimized debris can reduce the cross contamination in the solubilization step (less debris also dissolved with the functional IB) as well as downstream purification by chromatography. This is a significant improvement all courtesy of the novel integration approach to homogenization-and-centrifugation as will be seen next after we have first addressed the conventional process. 17.10.1 Conventional Inclusion Bodies Processing After homogenization, Fig. 17.16A shows that the IB cells have a size range between 0.6 and 1 μm, and cell debris having sizes between 0.05 and 0.4 μm taking up 20% of the feed.

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Cell debris

F(x)

(A)

IB 0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

x, μm (B)

1 IB: 0.6–1.0 μm, debris less than 0.4 μm

0.95 0.9

Rs

0.85

83.37%

0.8

80%

3.37%

Le = 0.591

0.75 0.7 0.65 0.6

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Le

1

SRe(S), SRe(L), SRs(S), SRs(L)

0.9

83.1%

0.8 0.7 0.6 0.5 0.4

SRe(S)

Le = 0.591

(C)

16.9% x 20% = 3.38% in sediment

SRe(L) SRs(S) SRs(L)

0.3 16.9%

0.2 0.1 0 0.01

0.10

1.00

10.00 Le

Figure 17.16 (A) Size distribution of cell debris and IB cells after homogenizing the bacterial cells. (B) Solids recovery curve corresponding to PSD of Fig. 17.16A. (C) Size recovery curve corresponding to PSD of Fig. 17.16A.

Classifying Bimodal PSD and Case Study of IB Classification

411

The solids recovery as determined by the model is shown in Fig. 17.16B. As shown, the centrifuge should operate with Le 5 0.591 where Le is based on the minimum size of the IB cells, xo, which is 0.6 μm for this example. The solids recovery is only 83.37% with Le 5 0.591. Given the cell debris take up 20% of all the particle sizes, the ideal situation is to have the solids recovery at 80% sending all the cell debris with a size of 0.05
0.4 μm to the centrate leaving the IB cells 0.6
1 μm settled in the sediment of the centrifuge. But this is not the case. In fact, there is 3.37% of cell debris which settle in the sediment together with the IB cells when operating with Le 5 0.591. The size recovery curve as shown in Fig. 17.16C reveals some interesting clue. At Le 5 0.591, obviously 100% of the IB cells are recovered in the sediment; however, the size recovery for the cell debris in the sediment SRs(S) is 16.9% of all the small size (i.e., cell debris) being fed to the centrifuge. Given the cell debris is 20% of the total feed, the cell debris entrained in the sediment is 20% 3 16.9%, which is 3.38% of all the feed solids. This is precisely the same amount as obtained from the total solids recovery curve as shown in Fig. 17.16B. This relatively large amount of cell debris is removed by subsequent washing and centrifugation as shown in Fig. 6.8. This is one of the key reasons for the low yield and high production cost on the IB cells due to the repeated washing and centrifugal separation as adopted in traditional practice. 17.10.2 New Inclusion Bodies Processing An innovative approach is to homogenize repeatedly the bacterial cells at high pressure, say about 800 bars, such that the cell debris is much finer. The following illustrates this new approach. Suppose the cell debris is homogenized with size no larger than 0.1 μm as depicted in Fig. 17.17A, which is smaller than the 0.4 μm considered in the previous example. The solids recovery with this finer debris is shown in Fig. 17.17B. It reveals several interesting aspects. First, the solids recovery has a relatively slowly decline leading to an almost flat regime between Le of 0.4 and 0.591. Second, the solids recovery in this flat regime is at 80.32%, which is close to the 80% theoretically obtained if all cell debris reports to the centrate. The difference is only 0.32%. The size recovery curve for this finer cell debris is shown in Fig. 17.17C. Operating the centrifuge at Le 5 0.591 shows that 98.4% of all cell debris report to the centrate. In other words, only 1.6% of the cell debris in the feed settles to the centrifuge bowl wall together with the valuable IB product. Given the cell debris is comprised of 20% of the feed, this

Centrifugal Separations in Biotechnology

F(x)

(A)

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Cell debris

412

IB 0

0.2

0.4

0.6

0.8

1 1.2 x, μm

1.4

1.6

1.8

2

1

(B)

0.95

IB: 0.6–1 μm, debris less than 0.1

μm

0.9 0.85

80.32%

Rs

0.8

Le = 0.591

0.75

80.32%–80% = 0.32% in sediment

0.7 0.65 0.6

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Le 1

98.4%

0.9 0.8 0.7 0.6 0.5 0.4

1.6% x 20% = 0.32% in sediment

SRe(S) SRe(L) SRs(S) SRs(L)

Le = 0.591

SRe(S), SRe(L), SRs(S), SRs(L)

(C)

0.3 0.2 0.1 0

0.01

1.6%

0.10

1.00

10.00

Le

Figure 17.17 (A)Size distribution after intensely homogenizing under high pressure the bacterial cells. (B) Solids recovery curve corresponding to PSD of Fig. 17.17A. (C) Size recovery curve corresponding to PSD of Fig. 17.17A.

Classifying Bimodal PSD and Case Study of IB Classification

413

means 0.32% (51.6% 3 20%) of the unwanted solids in the feed goes to the IB product in the sediment. This is precisely the 0.32% shown by the total solids recovery curve in Fig. 17.17B. An important lesson learnt from these two examples is that by homogenizing repeatedly at high pressure of, say, 800 bars on the bacterial cells until the maximum cell debris (,0.1 μm) is much smaller compared to the minimum IB cells (0.6 μm), one can minimize the contaminant in the product IB that settle to the bowl wall by a factor of 10. This minimizes repeated washing and centrifugation as shown in the schematic of Fig. 6.8. This new approach will certainly increase the yield of IB cells and reduce the cost of producing IB cells. This is a good illustration of how the model can help to optimize the process for the production of recombinant protein in IB.

17.11

Centrifuges for Inclusion Bodies processing

The disk stack, tubular, and spintube centrifuges are used to illustrate making the size cut to estimate the performance as predicted by the classification of the new IB process described in Section 17.10.2 with intense homogenization to reduce the fine fraction no larger than 0.1 μm in size. 17.11.1 Disk-Stack Centrifuge A disk-stack centrifuge is used to classify the IB cells after lysing at high pressure with the feed PSD to the centrifuge given by Fig. 17.17A. The density of the IB cells is 1.3 g/mL but the void fraction is large [8]; therefore, the effective density is much reduced. Assuming the void fraction is 67%, then the effective density of IB cells in buffer, largely water (ρ 5 1 g/mL), is (1.3)(0.33) 1 (1)(0.67) 5 1.1 g/mL. Thus, Δρ/ρ 5 (1.12 2 1)/1 5 0.1. This value is also roughly that of the cell debris and will be used in the example below. The following list the geometry and condition of the 400-mm bowl disk stack pilot centrifuge used for the classification:

Example 17.1 Disk Stack (400-mm Bowl) • • • •

Centrifuge bowl diameter D 5 400 mm; Disk outer diameter D2 5 2R2 5 304 mm; Disk inner diameter D1 5 2R1 5 178 mm; Disk angle θ 5 40 degrees from vertical;

414

Centrifugal Separations in Biotechnology

• Density difference/density Δρ/ρ 5 0.1; • Viscosity μ 5 1 cP; • G/g 5 6000 (5179 rpm), 9000 (6344 rpm), 12,000 (7325 rpm), 15,000 (8190 rpm); • n 5 100 disks; • Accelerator efficiency η 5 90%; • This efficiency also includes possible non-uniform distribution of feed to the disk stack; • xo 5 0.6(1026) m; • Q 5 10 2 50 L/min;

Le 5

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi      μ ρ tanθ   3Q=n ρ Δρ R3o 2 R3i

(17.11) Ωxo η ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi v    u 0:001   1 tanð400 Þ u 26 ( Q=100 1 3 10 u 3   3 ) 3 u 60 0:1 304 178 t 2 2000 2000 Le 5    Ωπ=30 0:6 3 1026 ð0:9Þ pffiffiffiffi pffiffiffi π Q 5 683:7 5 5 0:591 Ω 3 In the above, Q is expressed in L/min and centrifuge rotation speed Ω in rpm. For rotation speeds of 5179 (6000 g), 6344 (9000 g), 7325 (12,000 g), 8190 (15,000 g), we can determine the feed rate Q in L/min. Q is plotted in Fig. 17.18 as a function of G/g for the 400-mm diameter disk stack. The trend is shown under the ‘lighter IB’ (as opposed to a “denser IB” that will be investigated subsequently). The feed rate for the lighter IB varies linearly between 20 and 50 L/min with increasing G between 6000 and 15,000 g. As can be seen, this is a rather difficult separation due to the fine IB that needs to be recovered with minimum size xo 5 0.6 μm. However, the good part of this is that the contaminant in the sediment where IB reports to is only 0.32% of the feed, or 1.6% of the total debris that enters the feed. This novel process of homogenizing the cell debris to finer size prior to centrifugal classification reduces the downstream washing and separation. This is much more advantageous when compared with the conventional approach with larger debris (up to 0.4 μm) as discussed. Whether the maximum size of the debris should be

Classifying Bimodal PSD and Case Study of IB Classification

415

120 100

Lighter IB Denser IB

Q, LPM

80 60 40 20 0

0 400-mm disk stack

5000

10,000

15,000 G/g

Figure 17.18 Projected feed rate to a 400-mm disk stack centrifuge after high-pressure homogenizing of the bacterial cells to release finer cell debris in centrate and to recover IB products in sediment.

0.1 μm, or other even finer, depends on the minimum size of the IB for the process being considered. In our example, the minimum IB cell is 0.6 μm, which is 6 times the maximum size of the cell debris, 0.1 μm. The classification model can be used to make simulation before-hand to project performance and make necessary adjustment accordingly for the process. This example demonstrates a very powerful approach of improving the yield of IB products by optimizing combined homogenizing and centrifugal separation, which has a great deal of promise for future therapeutic protein production. 17.11.2 Tubular Centrifuge

Example 17.2 Tubular (190-mm bowl) Centrifuge bowl diameter, D 5 190 mm Pool and hub diameter, Dp 5 2Rp 5 120 mm Clarifier length, L 5 650 mm Density difference/density Δρ/ρ 5 0.1 Kinematic viscosity, μ/ρ 5 1026 m2/s G/g 5 6000 (7516 rpm), 10,000 (9703 rpm), 15,000 (11883 rpm), 20,000 (13722 rpm) • Accelerator efficiency η 5 90% • xo 5 0.6(1026) m • Q 5 1 2 5 L/min • • • • • •

416

Centrifugal Separations in Biotechnology

Le 5

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi     Q μ ρ L ρ Δρ

(17.12)

ΩRp xo η

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   0:001    1 26 60 Q 1 3 10 pffiffiffiffi pffiffiffi 0:1 0:65 π Q 5 4719 5 Le 5   5 0:591   Ω 3 Ωπ 26 ð0:06Þ 0:6 3 10 ð0:9Þ 30 In the above, Q is expressed in L/min and centrifuge rotation speed Ω in rpm. For G/g from 6000 to 20,000, we can determine the feed rate Q in L/min. Q is plotted in Fig. 17.19 as a function of G/g for the 190mm diameter tubular under the curve labeled as “Lighter IB.” The feed rate varies linearly from 1 to 3 L/min with increasing G/g from 6000 to 20,000. For reference, the nominal flow rate for the 190-mm tubular centrifuge is about 10 L/min. As can be seen, this is a rather difficult separation due to the fine IB cells that needs to be recovered with minimum size xo 5 0.6 μm. High G and low feed rate are necessary to achieve good result. In spite of this example, the IB cells has high solids density but the void fraction is also high at 0.67; therefore, the bulk density of the IB cells is only 1.1 g/mL, which is only 0.1 g/mL above that of the aqueous-based buffer.

6

Lighter IB 5

Denser IB

Q, LPM

4 3 2 1 0

0

5000

190-mm diameter pilot tubular

10,000

15,000

20,000

G/g

Figure 17.19 Projected feed rate to a 190-mm diameter tubular centrifuge after high-pressure homogenizing of the bacterial cells to release finer cell debris in centrate and recovering IB products in sediment.

Classifying Bimodal PSD and Case Study of IB Classification

417

17.11.3 Spintube Centrifuge

Example 17.3 In the last example, a spintube is used in the laboratory to carry out classification of the same process. The centrifuge will be operated at Le 5 0.591 to make the cut. Below are the operating conditions of the spintube centrifuge: • • • • • •

Tube length, H 5 4 cm Density difference/density Δρ/ρ 5 0.1 Kinematic viscosity, μ/ρ 5 1026 m2/s G/g 5 1000
12,000 Accelerator efficiency η 5 90% xo 5 0.6(1026) m sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi    ð2πH Þ μ ρ 1   Le 5 η G=g gt ρ Δρ x2o

(17.13)

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ffi u  26  1 1 2πð0:04Þ 114:8 u   10 5 rffiffiffiffiffiffi Le 5 26 u G 0:1 G 0:6 3 10 t ð9:81Þð60Þt ð0:9Þ t g g pffiffiffi π 5 5 0:591 3 where t is in minutes and G/g is dimensionless. The right part of the above equation is needed when the cut size xc 5 xo. Solving, 37; 754  t5  G=g In the above, t(min) is inversely related to G/g, and this is plotted in Fig. 17.20 in a double logarithm scale that yields a straight line with a negative slope labeled “Lighter IB.” When G is at 1000 g, the time for separation t is 37.75 minutes, and when G is increased to 12,000 g, t reduces to 3.15 minutes. The slope is proportional to the square of the quantity μ/Δρ. Therefore, it is possible to run some spintube tests to obtain results similar to Fig. 17.20, and μ/Δρ can be deduced from the slope of the line. This quantity can be used subsequently for projecting the disk stack or tubular centrifuge performance. In any case, the past three examples with results of Fig. 17.18 for the disk stack, Fig. 17.19 for the tubular, and Fig. 17.20 for the spintube all

418

Centrifugal Separations in Biotechnology 100 Lighter IB

t, min

Denser IB

10

1 1000 4-cm spin tube

5000

10,000

G/g

20,000

12,000

Figure 17.20 Time duration for classification of IB cells from cell debris.

refer to the combined optimization of the homogenization and centrifugal classification that result in high product purity IB cells with little amount of cell debris that reduces further separation and purification. They demonstrate in each case how the Le number is used so that either the feed rate to the continuous/semi-continuous centrifuge or the centrifugation time can be deduced.

17.12

Separation by Size and Density Difference

The IB cells are just densely packed protein cluster. Most of the IB are lightly packed; however, there are cases in which they can be densely packed. The IB solid density ρs is 1.3 g/mL. Assuming the liquid buffer is water with density ρL being 1.0 g/mL, then the “bulk” density of IB is ρb 5 ρs ð1 2 εÞ 1 ρL ε Therefore, the density difference between the bulk IB cell including the encapsulated fluid in its void is   ρb 2 ρL 5 ρs 2 ρL ð1 2 εÞ or   ρs 2 ρL ρb 2 ρL 5 ð1 2 εÞ ρL ρL

(17.14)

Eq. 17.14 is plotted in Fig. 17.21 for ρs being 1.3 g/mL and ρL 1.0 g/mL. For the case that has been considered, we assume that the IB is somewhat

Classifying Bimodal PSD and Case Study of IB Classification

419

0.35

(ρb-ρL)/ρL

0.30 0.25 0.20 0.15 0.10 0.05 0.00 0

0.1

IB: ρ s =1.3 g/mL

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ε

Figure 17.21 Density difference between bulk density of IB cells and suspending liquid with IB solid density 1.3 g/mL.

unfolded with a lot of void fraction, such that ε 5 0.67, in which case ρb 2 ρL 5 0.1 g/mL. There are situations in which the IB are more densely packed with less unfolding for which the void is much reduced. Suppose, say, ε 5 0.33 in which (ρb 2 ρL)/ρL 5 0.2 as seen in Fig. 17.21. Thus, we have an interesting situation in which the debris is in the size range 0.05
0.1 μm with Δρ 5 ρb 2 ρL 5 0.1 g/mL, whereas the IB cells are in size range of 0.6
1 μm with Δρ 5 ρb 2 ρL 5 0.2 g/mL. In this case, separation of debris and IB cells is by both size and density difference! Given, Stokes’ settling velocity is given by the following equation: vs BΔρ2 x2o

"sffiffiffiffiffiffiffiffiffi #2  0 2 Δρ2 xo 5 Δρ1 xo 5 Δρ1 Δρ1

We can define a new minimum size for the large fraction, sffiffiffiffiffiffiffiffiffi Δρ2 0 xo xo 5 Δρ1

(17.15a)

(17.15b)

Consider the high-density case in which ε 5 1/3, Δρ2 5 0.2 g/mL. Previously, at ε 5 2/3, Δρ1 5 0.1 g/mL; thus, we can define a new minimum size for the large fraction, sffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffi Δρ2 0:2 0 ð0:6Þ 5 0:8485μm xo 5 xo 5 0:1 Δρ1

Centrifugal Separations in Biotechnology

F(x)

(A)

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Cell debris

420

IB 0

(B)

0.2

0.4

0.6

0.8

1

1.2 x, μm

1.4

1.6

1.8

2

1 0.95

IB: 0.8485-1.4142 μm, debris less than 0.1 μm

0.9 0.85 80.16%

0.75

80.16% -80% = 0.16% fine fraction in sediment

0.7

Le' = 0.59 1

Rs

0.8

0.65 0.6 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Le'

(C)

1

99.2%

0.8 0.7

SRe(S) SRe(L) SRs(S) SRs(L)

0.6 0.5 0.4 0.3 0.2

0.8% x 20% = 0.16% fines in sediment

Le' = 0.591

SRe(S), SRe(L), SRs(S), SRs(L)

0.9

0.1 0

0.01

0.8% 0.10

1.00

10.00 Le'(Δρ1,xo')

Figure 17.22 (A) Size fraction distribution based on xo’ 5 0.8485 μm and xm’ 5 1.4142 μm. (B) Solids recovery versus Le’. (C) Size recoveries versus Le’.

Classifying Bimodal PSD and Case Study of IB Classification

421

Similarly, we can define a new maximum size for the large fraction: sffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffi Δρ2 0:2 0 ð1Þ 5 1:4142μm xm 5 xm 5 0:1 Δρ1 With the new size range for the larger fraction (xo0 , xm0 ) as shown in Fig. 17.22A, we have determined the solids recovery in Fig. 17.22B and the size recoveries for the two sizes in, respectively, the two streams in Fig. 17.22C. As shown in Fig. 17.22B, the solids recovery is plotted against the operating Le0 which is defined based on Δρ1 and xo0 . When Le0 5 0.591, there is only 0.16% solids (primarily the fine fraction) that entrains into the IB product in the sediment. (This 0.16% of total feed solids is further reconfirmed by the size recovery as depicted in Fig. 17.22C.) This is only half that as compared to 0.32% of total feed as shown for the lower density IB in Figs. 17.17B and 17.17C. The improved IB product is all attributed to the heavier and denser IB cells. In the first case, for the lighter IB cells with void fraction of 0.67, the bulk density of IB cells is the same as cell debris 1.1 g/mL, and density difference with the liquid (1 g/L) Δρ1 is only 0.1 g/mL, separation is merely by size difference between the cell debris and the IB cells. However, for the denser IB cells with void fraction of 0.33, the bulk density is 1.2 g/mL, the density difference with the liquid Δρ2 is doubled at 0.2 g/mL, and separation is by combined size and density difference between the cell debris and the IB cells. Given the latter, the separation quality should be better which is quite evident. Further, the feed rate should be even higher as the separation is improved. This can be seen when the cut size xc is set to xo0 , Le0 5 π1/2/3. Given,        0 2 0:2 2 Δρ2 2 5 2xo 2 Δρ1 5 2Q1 Δρ1 5 xo Δρ1 Q2 B xo Δρ1 5 xo 0:1 Δρ1 In general,   Δρ2 Q1 Q2 5 Δρ1

(17.15c)

where Q1 is based on Δρ1 and Q2 is based on Δρ2. The feed rate for a given G for the denser IB is twice that of the lighter IB cells in Fig. 17.18 for the disk stack, and Fig. 17.19 for the tubular centrifuge (see dotted lines in both figures). For the spintube centrifuge, the time duration is 1/2 shorter for the denser IB cells as compared to the lighter IB cells as shown in Fig. 17.20 (see dot-dash line). This is quite interesting. If the IB cells

422

Centrifugal Separations in Biotechnology

can be engineered to be more compact, separation should be much easier with higher throughput for continuous centrifuge and shorter centrifugation time for spintube centrifuge. However, downstream solubilization and refolding of IB needs to be factored into account as well.

17.13

Summary

A model for classifying bimodal feed size distribution has been developed. The bimodal model works on all types of centrifuges. We have demonstrated the use of the model by analyzing a disk-stack centrifuge processing the CHO cells with cell debris for which the monotonic size distribution model does not work. We have further examined the sensitivity of the bimodal model with respect to fine size fraction, average whole cells, range of whole cell size, and the extent of the finer fraction from the coarser fraction. The bimodal model is quite sensitive to each of these parameters, rendering the matching of the model parameters to test data quite unique. The size recoveries for the small and large size fraction, respectively, in the centrate and sediment have been developed. It is used to demonstrate a novel means of combined optimization of both homogenizer and centrifuge to increase the yield of IB cells, for classifying IB cells from submicron cell debris.

References [1] R. Lander, C. Daniels, F. Meacle, Efficient, scalable clarification of diverse bioprocess streams using a novel pilot-scale tubular bowl centrifuge, Bioprocess Int. (2005) 32
40. [2] A. Villaverde, Bacterial inclusion bodies: an emerging platform for drug delivery & cell therapy, Nanomedicine 7 (9) (2012) 1277
1279. [3] M.V. Ce´spedes, et al., Bacterial mimetics of endocrine secretory granules as immobilized in vivo depots for functional protein drugs, Sci. Rep. 6 (2016) 35765. Available from: https://doi.org/10.1038/ srep35765. [4] C. Slouka, J. Kopp, O. Spadiut, C. Herwig, Perspectives of inclusion bodies for bio-based products: curse or blessing? Appl. Microbiol. Biotechnol. 103 (2019) 1143
1153. [5] A. Villaverde, et al., Packaging protein drugs as bacterial inclusion bodies for therapeutic applications, Microb. Cell Factories 11 (2012) 76.

Classifying Bimodal PSD and Case Study of IB Classification

423

[6] S.-L. Choi, S.J. Lee, S.-J. Yeom, H.J. Kim, Y.H. Rhee, Controlled localization of functionally active proteins to inclusion bodies using leucine zippers, PLoS One 9 (6) (2014) e97093. Available from: https://doi.org/10.1371/journal.pone.0097093. [7] G. Taylor, M. Hoare, D.R. Gray, FAO Marston. Size and density of protein inclusion bodies, Nat. Biotechnol. 4 (1986) 553
557. [8] G. Margreitera, P. Messnerb, K.D. Caldwell, K. Bayer, Size characterization of inclusion bodies by sedimentation field-flow classification, J. Biotechnol. 138 (2008) 67
73. [9] Singh, et al., Protein recovery from inclusion bodies of Escherichia coli using mild solubilization process, Microb. Cell Factories 14 (2015) 41.

Problems (17.1) Determine using the bimodal model the solids recovery versus Le (based on xo 5 20 μm) curve for smaller size fraction 5
6 μm and large size fraction 20
24 μm assuming the smaller size fraction being 60%, that is, F1 5 0.6. Plot also the cumulative size under versus particle size. (17.2) Suppose all small size fraction reports to the centrate stream as unsettled solids, what is the solids recovery? (17.3) What is the solids recovery at Le 5 0.591 or xc 5 xo (20 μm)? What is the percent of small size fraction that has not reported to the centrate but settled in the cake? (17.4) A series of tests have been carried out with the following results Le (based on 20 μm)

Solids recovery

0.17 0.25 0.40 0.50 0.70 0.8

0.94 0.748 0.66 0.65 0.55 0.4

Plot these test results on the same solids recovery curve as in Problem 17.1. Does the data match? If not, vary the fraction of smaller size fraction 5
6 μm until the solids recovery prediction matches with the test data? Determine what size fraction F1 matches with the test data?

424

Centrifugal Separations in Biotechnology

(17.5) Determined the size recovery curves of small fraction and the large fraction in the centrate stream for F1 5 0.4 assuming the small fraction has size of 5
6 μm and large fraction 20
24 μm. Also show the corresponding size recovery of small and large fraction in the cake. (17.6) For Problem 17.5, what is the size recovery of small fraction in the centrate when the cut size is at 20 μm corresponding to Le 5 0.591. Likewise, what is the size recovery of small fraction in the sediment? (17.7) For Problem 17.5, what is the size recovery of large fraction in the centrate if the centrifuge is operated such that the size recovery of the small fraction is increased to 94% and 95%, respectively? What conclusion can be drawn? (17.8) For Problem 17.5, given the sediment is the product with large particles, and the size recovery of fine solids in the sediment should be no more than 5%, what is the size recovery of the large sized particles in the sediment?

18 Integration of Unified Modeling With Practice in Centrifugal Separations 18.1

Introduction

In this chapter, we discuss the unified modeling that we have established on centrifugal separation with the objective of clarification of centrate, separation of solids from liquid, or classifying two size fractions of solids in the feed. Subsequently, this unified modeling is integrated with practice in separation at each stage of drug development, from laboratory spintube, pilot demonstration, and clinical manufacturing to production, accomplishing guiding testing, analyzing test results, predicting performance, scaling up/down, troubleshooting, and optimizing performance. This provides a comprehensive integration of all the knowledge and technologies on centrifugal separation and available tools (analytical and numerical models) that have been developed and discussed in various chapters of the book.

18.2

Unified Modeling to Centrifugal Separation

We have developed models on solid liquid separation by sedimentation for spintube centrifuges in Chapter 9, Selection and Sizing of Centrifuges, tubular and decanter centrifuges in Chapter 11, Visualization and Modeling of Flow and Separation in Tubular Centrifuge, disk-stack centrifuges in Chapter 12, Disk Stack Modeling, and decanter centrifuge processing flocculated solids in Chapter 15, Flocculation With Decanter Centrifuges. The design geometry for the centrifuge, operating condition for the centrifuge, and the process properties are captured in the Le number, which is calculated according to the type of centrifuge. In Appendix B, it has been shown that the Le number, aside from being derived in the Centrifugal Separations in Biotechnology. DOI: https://doi.org/10.1016/B978-0-08-102634-2.00018-0 © 2020 Elsevier Ltd. All rights reserved.

425

426

Centrifugal Separations in Biotechnology

aforementioned chapters, is a combination of dimensionless groups that govern the separation process. The dimensionless groups can be obtained independently by the Buckingham-π approach. The process objectives can be clarification removing solids in the liquid containing the expressed protein, separation in which the collected whole cells containing the intracellular protein, and classification in which undesired cell debris are separated from the whole cells/inclusion bodies, or undersized crystals not meeting specification being separated from the larger sized crystals. Once the Le number is determined, the cut size, xc is also determined. The amount of the biosolid fed to the centrifuge that got separated depends on the size distribution of the biosolid, F(x), and their density difference with the suspending liquid. Particle size x in the feed with cumulative size distribution F(x) larger than xc will be settled completely to the centrifuge bowl wall. When x , xc, a portion of which would settle to the centrifuge bowl wall, while the rest escape with the centrate. For clarification of centrate liquid and separation of solid from liquid, the monotonic size distribution, especially with the application of flocculant to produce larger flocculated solids (flocs), has been discussed in Chapter 15, Flocculation With Decanter Centrifuges, for a decanter and Chapter 16, Case Studies of Monotonic and Unimodal Size Distribution Models, for disk stack and tubular. Flocculation in a disk (applicable to other types of centrifuge as well) using a unimodal size distribution, where all particle sizes within the size range of interest have constant occurrence/frequency, has been presented in Chapter 16, Case Studies of Monotonic and Unimodal Size Distribution Models. The size range in the unimodal model can be large as applied to flocculation (maximum size to minimum size ratio xm/xoc1) or it can be quite narrow (1 , xm/xo , 2) to reflect a genuine unimodal size distribution. The fractionation of two different size solid fractions has been detailed in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification. (This can be extended to fractionate three different size solid fractions with two-step classification as well as N different size fractions with N 2 1 step consecutive fractionation, where N is an integer greater than unity.) Examples of all types of centrifuges from spintube, disk stack to tubular have been used to demonstrate the size classification process. Based on all these examples and illustrations, the methodology to the separation process, by sedimentation, is the same irrespective of the type of centrifuge, whether the centrifuge is a spintube, disk stack, tubular, or decanter. This can be considered as an unified modeling approach to centrifugal separation. This is depicted in Fig. 18.1. On the left top of

Integration of Unified Modeling With Practice in Centrifugal Separations

427

100% F(x)

Tabulation/ graph

Total solids recovery in sediment Size recoveries in centrate & sediment (fractionation)

Analytical 0

Size, x

Dimensionless Leung (Le) number Le =

π 3

xc xo

Cut size, xc Le =

Q

Le =

ρ 1 Δρ x 2o

Spin-tube

Tubular/ decanter

ρ Δρ

μ ρ

L

Ω Rp xo η

3Q

Le =

μ ρ

(2πH) η(G/g)gt

n

μ ρ

ρ Δρ

tan θ R 3o R 3i

Disk stack

Ω xo η

Figure 18.1 Schematic representation of unified methodology to centrifugal separation.

the figure, the cumulative size under F(x) can be a set of measured data in tabulation, in the form of a graph, or in the analytical form with pertinent parameters (two for the monotonic and unimodal size distribution and five for the bimodal model) being estimated. Based on F(x), the solids recovery Rs as a function of xc can be determined. But xc is linearly related to Le. If there are distinct multiple sizes in the feed suspension that need to be separated from each other, size recoveries for each size fraction in centrate and sediment can also be determined. Thus, both solids recovery Rs and size recoveries (of large and small sized fractions in centrate and concentrate, respectively) SR as a function of Le can be established. For a given set of process properties (ρs, ρL, μ), the design and operating parameters of a given centrifuge (spintube/disk/tubular/decanter) can be tuned to a prescribed Le to meet the size cut xc requirement in order to achieve the solids recovery or size recoveries objectives, see Fig. 18.1. The methodology or approach in modeling is identical for all types of centrifuge of different designs and sizes except the equation to determine Le is specific for a given type of centrifuge. Fig. 18.1 (bottom right) also shows the relevant equation to determine Le for a specific centrifuge type. This provides indeed a unified modeling to separation. After all, the separation model for each type of centrifuge: spintube, disk stack, tubular, and decanter is based on limiting trajectory analysis using Stokes’ law. This results in the separation efficiency Z(x) (of a

428

Centrifugal Separations in Biotechnology

given particle size x) being identical, namely, Z(x) 5 (x/xc)2. The only difference is the cut size. However, the cut size xc is related to Le via xc/xo 5 3/(π)1/2Le. Hence, it is not too surprising that all these loose pieces can be fitted into a puzzle. The difference among different types of centrifuges is only pertaining to the Le number, embodying the details of centrifuge design, operating condition, and process properties.

18.3

Applications of the Unified Separation Models

There are several applications for the unified separation models: test analysis, prediction/forecast, guide testing, optimization, and troubleshooting. During fractionation of a bimodal feed, we are interested in size recoveries SRe(S), SRe(L), SRs(S), and SRs(L) as discussed in Chapter 17, Classifying Bimodal Particle Size Distribution and Case Study of Inclusion Body Classification, we will only use “SR” as a general representation of any combinations of these four size recoveries for simplicity without further elaboration. 18.3.1

Analysis of Test Data

For a given application with centrifuge feed size distribution F(x), Rs and SR versus Le can be determined. Based on test centrifuge design parameters, operating conditions (speed, feed rate, etc.), process properties, and the corresponding Le can be calculated. Also, Rs and SR can be determined for the tests. Subsequently, the tests results and prediction expressed in the graphical forms: Rs Le and SR Le, can be compared. When the feed size distribution F(x) is determined by particle size measurement, Rs and SR can be determined by numerical integration. When F(x) is not measured, depending on the process clarification/separation or fractionation, the parameters (two parameters per either monotonic or unimodal size distribution and five parameters per bimodal size distribution) have to be estimated. Some test data of the entire data set can be used for this purpose to back-out the missing parameters by matching prediction to test results. This may require some iterations. Once these parameters that pertain to F(x) have been back-out from the test data, Rs and SR can be determined based on a given Le and can be used to compare with the rest of the test data set. This is the procedure adopted in Chapter 15, Flocculation With Decanter Centrifuges, to infer the size of the flocs in the decanter centrifuge.

Integration of Unified Modeling With Practice in Centrifugal Separations

18.3.2

429

Prediction/Forecast

In prediction/forecast, once F(x) is known by measurement or by inferring of parameters from the three models (monotonic, unimodal, and bimodal size distributions), Rs and SR can be determined as a function of Le. Based on the design, operating condition, and process properties, Le can be determined from which we can make prediction on Rs and SR. 18.3.3

Guiding Testing

When F(x) is known or being estimated, it can generate Rs Le or SR Le curves, a set of tests should be planned out so that these curves can be independently verified. If the test results come out differently, fine tuning on F(x) or process properties are required. 18.3.4

Optimization

In optimization, again the appropriate performance curves Rs Le and SR Le for the process need to be determined. Based on benefit-to-cost, one can determine what is the best condition (can be a range of Le, say 0.5 , Le , 0.7) to operate, and this condition can be adjusted depending on changes in upstream and downstream processes as well as the process requirements on separation. Therefore, the process objectives need to be defined clearly first before optimization can be made; otherwise, it is a waste of effort. 18.3.5

Troubleshooting

The unified model can be used to predict the performance similar to discussion on prediction. At times, there could be upset from upstream bioreactors/fermenters in which there are changes on F(x) and process properties (change in viscosity, density of solids, and density of liquid) that results in changes in performance Rs and SR for a given Le. Then, F(x) and process properties in the model can be temporarily adjusted to reflect the upset condition. On the other hand, Rs Le and SR Le have been recorded during the normal process, they can be used as a referee. By comparing the process performance Rs Le and SR Le between the normal and upset conditions, one can compensate temporary upsets external to the centrifuge (say upstream bioreactor or fermenter) by readjusting the centrifuge operating parameters (perhaps lower feed rate or higher speed) to provide consistent

430

Centrifugal Separations in Biotechnology

output. If the problem is in the centrifuge, then one can analyze whether certain parts of the centrifuge affecting process separation has been deteriorated due to wear-and-tear as discussed in troubleshooting in Chapter 10, Troubleshoot and Optimization, or the process operating condition (speed, feed rate, etc.) has been drifted from the initial nominal setting. Other possible scenarios can be dealt with in a similar manner. Often, centrifuge is just a machine and changes often are due to what is being fed to the centrifuge assuming the operating condition is held constant. It follows that the first attempt in process troubleshooting should address the feed properties (size and properties of solids and liquid) before anything else. Again, these considerations have already been discussed in detail in Chapter 10, Troubleshoot and Optimization. 18.3.6

Scale-Up/Scale-Down

In scale-up/scale-down, prediction has to be made using known Rs Le theoretical curve after verified by testing. For a required process specification on solids recovery Rs or bimodal fractionation SR [i.e., SRe(S), SRe(L), SRs(S), and SRs(L)], the required Le to meet the process requirement can be determined. Based on the new smaller size (scale-down) or new larger size (scale-up), the process feed rate Q can be determined, as has been discussed in Chapter 9, Selection and Sizing of Centrifuges. If the calculated Q does not meet the requirement (either too large or too small), then a new size will be redetermined and the process will be repeated until the desired feed rate is met. A small number of iterations is required. When changing to a different size machine, the operating conditions, such as rotation speed, number of disks, disk geometry for disk stack, etc., for a new larger/smaller disk stack centrifuge can be modified. This also applies for making modification in scale-up/scale-down for tubular and decanter centrifuges. In summary, the unified modeling on centrifugal separation is a very powerful approach to address the requirements in the biotechnology separation. The methodology is the same across all different centrifugation which are all based on Stokes law on settling of individual/ discrete particles based on size and density difference. For dense feed with high concentration of solids, we may have to modify the stokes’ law incorporating hindered settling factor as discussed in Section 2.3.5 in Chapter 2, Principles of Centrifugal Sedimentation, and Section 9.2.3 in Chapter 9, Selection and Sizing of Centrifuges.

Integration of Unified Modeling With Practice in Centrifugal Separations

18.4

431

Integration of Unified Separation Models With Practice

Drug development starts with laboratory testing. Separation is of no exception. In laboratory test on separation, spintube testing is being used first to determine whether the whole cells or the finer cell debris can be separated by sedimentation from the liquid. The suspension with solids in a spintube is tested at different G’s and different t’s depending on the ease of separation and objectives—clarification, separation, and fractionation. Analysis on test data can be made on the process in terms of Rs and SR as a function of Le. Concurrently, if F(x) is measured by particle counter or being estimated using the monotonic, unimodal, or bimodal size distributions, Rs and SR can be predicted as a function of Le. Thus, experimental results can be compared with prediction. Adjustment can be made for any discrepancy on process properties, efficiency of spintube acceleration, and inaccurate estimation of parameters for use in the analytical form of F(x). Additional tests can be carried out as necessary and compared with predictions based on better estimate of these parameters. Presumably comparison should be improved. Further testing can be carried out as needed to assure that the parameters adopted are most appropriate conforming to all test results. In a scale-up process, these parameters can be used to project performance of small disk stack, tubular, or decanter centrifuge in pilot testing. Upon establishment of pilot scale centrifuge, protocol guiding testing can be designed to cover a wide range of operating conditions, especially on feed rate and centrifuge speed. Based on the protocol, testing on the pilot centrifuge, test results are analyzed similar to that of the spintube and of course appropriate parameters corresponding to the pilot centrifuge are used. Concurrently, process predictions are made similar to the spintube case and the predictions are compared with the test results all cast in the form of Rs Le and SR Le plots. As before, if discrepancy exists, adjustment and fine-tuning can be made on the process properties, feed size distribution parameters (such as x50 and xo), acceleration efficiency, etc. in the model. It is perceived that additional but perhaps only a limited number of tests are need to be carried out to get better assurance and confidence that the model can provide accurate predictions of process performance. Optimization of the pilot centrifuge can be carried out along specific process objectives (centrate clarity, such as turbidity or centrate solids by weight percent, solids recovery, and size recoveries). The model can be used to predict the optimum, and the results can be compared side-by-side with the pilot centrifuge test results. During pilot demonstration, there could be instances in

432

Centrifugal Separations in Biotechnology

which the centrifuge may get off-spec feed from the bioreactors/fermenters and as such perform disappointingly. The process performance of the centrifuge under normal feed can be used as a comparison so that the problem can be immediately diagnosed, for example, upset from the upstream bioreactor producing finer size solids feeding to the centrifuge, higher viscosity of the liquid due to lower temperature of the feed stream or more concentrated solids in the feed, lighter cells and hence smaller density difference between suspended solids and liquid producing lower driving force for sedimentation. This is process troubleshooting. Once the problem is identified, the centrifuge can be temporarily run in a compensating mode to offset the process upset producing the same product feeding the downstream, for example, depth filter or microfiltration, but at a slower rate. Once the pilot tests have been satisfied, clinical manufacturing of the drugs can be carried out. Again, guiding test, analysis of test results, prediction, optimization, troubleshooting, and scale-up can all be carried out as with the pilot scale. These processes also are repeated at the clinical-manufacturing stage and subsequent large-scale production stage. As can be seen, the model can guide practicing stage-by-stage from lab, pilot, and clinical manufacturing, to production, and from spintube sample testing less than 50 mL of laboratory test sample to large-scale centrifuge separating harvested product from 20,000 L fermenter and large-sized bioreactor. Each stage is being assisted with the unified model that provides guiding testing, result analysis, prediction, optimization, troubleshooting, and scale-up. This certainly provides higher confidence for the plant operators, engineers, scientists, project managers, and decision makers to speed-up the lengthy drug development process at various development stages minimizing testing and risks involved. This reduces, if not eliminates, the perception that separation is the choke point in the drug development. This new approach is schematically represented in Fig. 18.2. The schematic in Fig. 18.2 shows a good interweaving and integration between theory and practice at every stage of drug development, which is of great benefit. Heretofore, only the left “leg” is practiced with trial-and-error leading to numerous problems with poor test protocols without clear objectives, incorrect analysis, and misinterpretation of test results, predictions falling short, lengthy troubleshooting, poor optimization, and incorrect scale-up. These shortcomings can be amended with the integration of the unified modeling as shown in Fig. 18.2 to speed-up the drug development process, in particular on the separation at each stage that are more efficient, reliable, and accommodating.

Integration of Unified Modeling With Practice in Centrifugal Separations Phases Lab development – Spintube tests

Analyze data

433

Simulation Spintube simulator

Guide testing

up

Scale-

Analyze data

Pilot plant

Pilot simulation Guide testing

up

Scale-

Clinical manufacturing

Analyze data Optimize process, troubleshoot

Clinical-manufacturing simulation

up

Scale-

Analyze data

Production Optimize process, troubleshoot

Production-size simulation

Figure 18.2 Integration of unified centrifugal separation modeling with practice.

18.5

Summary

The significance in adopting a unified approach of modeling different types of centrifuges for separation, by sedimentation, is emphasized. The same approach methodology, independent of centrifuge type, is used for planning and guiding testing, analysis of test results, prediction, troubleshooting, optimization, and scale-up/scale-down, etc. This unified modeling when running in-parallel with practice is of great benefit. It provides an extra assurance of process performance, despite test results are often limited and at times difficult-to-interpret. This will speedup the testing at various stages from testing in laboratory, piloting, and clinical manufacturing to production. This will further accelerate drug development while saving extensive testing at each stage.

Appendix A: Nomenclature A a ac b c C Closs CF D d E F f G Gr g H h I IB K KE L Le Lef M N Nu n P Pr p Q

area, m2 coefficient Coriolis acceleration, m/s2 coefficient coefficient, insoluble solid concentration, v/v, number of cells/mL or coefficient solute concentration, g/L loss coefficient [-] concentration factor [-] bowl diameter, m or diffusivity, m2/s particle diameter or disk membrane diameter, m Ekman number [-], Efficiency (Eq. 16.9) cumulative undersize distribution [-] frequency distribution or function [-] centrifugal acceleration, m2/s Grashof number [-] Earth gravity (9.81 m2/s) suspension height, m height, depth or thickness, m integral [-] inclusion body cake permeability, m2, coefficient kinetic energy, J length, m Leung number in centrifugal sedimentation [-] Le number in centrifugal filtration [-] mass flow rate (dry basis), kg/s number Nusselt number [-] number of disks or exponent pitch, m or power, W Prandtl number [-] pressure, Pa volumetric flow rate, m3/s

435

436

Appendix A: Nomenclature

R Ra Re Ro RCF rpm R0 Re Rex Rs r rm S Sc SR SRe(L) SRe(S) SRs(L) SRs(S) s T TCD t ts U u V VCD Vs v vc vr vs vso

radius, m Raleigh number [-] Reynolds number [-] Rossby number [-] relative centrifugal force [-] revolution per minute effective radius, m solids recovery in centrate, % exit radius of feed accelerator, m solids recovery in cake/concentrate, % gear ratio [-] or radial coordinate, m membrane resistance, kg/m2 s sedimentation coefficient, s Schmidt number [-] size recovery, % size recovery of a give particle size in the larger size fraction in centrate, % size recovery of a give particle size in the smaller size fraction in centrate, % size recovery of a give particle size in the larger size fraction in sediment, % size recovery of a give particle size in the smaller size fraction in sediment, % streamflow coordinate, m torque, N-m, turbidity, NTU, temperature, oC total cell density, number of cells/mL time, s time duration between intermittent discharge from dropping bottom disk centrifuge, s constant throughflow velocity, m/s throughflow velocity, m/s volume, m3 viable cell density, number of cells/mL Solids hold up volume in disk centrifuge or sediment volume in spintube, m3 velocity, m/s average Coriolis velocity, m/s relative velocity, m/s settling velocity of a concentrate suspension, m/s Stokes’ free settling velocity, m/s

Appendix A: Nomenclature

v0 w v0 wa vθ W x Y y Z z

normalized membrane wall flux [-] normalized averaged membrane wall flux [-] angular velocity, m/s solid concentration by weight fraction, w/w, [-] particle size, micron protein yield [-] transverse coordinate, m size capture [-] linear spatial coordinate, m; removal efficiency for given size [-]

Subscripts a b c cake d Ek e ex f G g i k L m o p R s w 50 100

acceleration, accelerator, average, or atmospheric bowl or bulk region further from membrane concentrate/cake, cut size, critical trajectory, or clarifier cake solid dimensionless or concentrate discharge Ekman layer centrate exit of accelerator feed centrifugal acceleration, m2/s gel index, inner radius, or species index liquid or load membrane, maximum reference and outer radius, minimum size pool, pinion, projected length, or liquid permeate retention sediment, concentrate, cake, solid, or settling wall 50%, median 100%

Symbols α β Δ

437

helix angle, radian/degree beach angle, radian/degree change

438

δ δ0 η μ μ0 ν φ λ ρ Δρ θ ΔΩ Ω Σ Δp R π l

Appendix A: Nomenclature

boundary layer thickness, m normalized boundary layer [-] efficiency [-] viscosity of liquid, kg/m s effective viscosity, kg/m s kinematic viscosity, m2/s solid volume fraction [-] hindered settling factor [-] density, kg/m3 density difference, kg/m3 angle between disk and vertical, deg/rad differential speed, rpm rotational speed, rpm sigma factor, m2 pressure drop, Pa membrane rejectivity [-] osmotic pressure, Pa, dimensionless group [-] membrane or filter thickness, m

Appendix B: Buckingham-π Analysis for Decanter and Tubular, Disk-Stack, and Spintube Centrifuges It is of interest to examine the general dimensionless π groups that are important for separation in the flow using the Buckingham-π theorem and to compare the results with the dimensionless variables obtained from the analytical model developed in the text. The Buckingham-π analysis can be found in any fluid mechanics study; therefore it is not repeated here except the results of the analysis.

B1

Decanter and Tubular Centrifuge (Separation/Clarification)

For a given clarifier design there are a total of 11 variables that affect the settling of flocs in the moving layer. They are the feed rate Q, clarifier length Lc, pool radius Rp, bowl radius Rb, minimum floc size xo, median floc size x50%, bowl rotation speed Ω, differential speed between conveyor and bowl ΔΩ, viscosity of suspension μ, density difference between floc and suspension Δρ, and suspension density ρ. Given these 11 variables are composed of three basic variables such as mass M, time T, and length L; therefore there should be only eight (511 2 3) dimensionless π groups according to the Buckingham-π theorem. Three variables Δρ, Ω, and Rp are chosen to form dimensionless groups with each of the eight variables. After making such analysis, the dimensionless π groups are: 1. Rossby number π1 5

Q=R2p Q u 5 5 5 Ro ΩR3p ΩRp ΩRp

(B.1)

439

440

Appendix B: Buckingham-π Analysis for Decanter and Tubular

Rossby number measures relative velocity of the feed velocity to the absolute tangential speed of the pool surface. An estimate based on typical values reveals that Ro , 0.1. 2. Ratio of clarifier length-to-pool radius π2 5

Lc Rp

(B.2)

The clarifier length-to-pool radius ratio is typically of orders of 10 based on the configuration and operating of decanter and tubular centrifuges. 3. Ratio of clarifier bowl-to-pool radii π3 5

Rb Rp

(B.3)

The clarifier bowl-to-pool radii ratio is typically about 2. Given flow occurs at the moving layer, this ratio may not have much influence unless Rb approaches Rp or Rb/Rp approaches unity for which the sediment on the bowl wall may affect the moving layer. 4. Ratio of minimum floc size-to-pool radius π4 5

xo Rp

(B.4)

The minimum floc size-to-pool radius is very small. 5. Ratio of floc size-to-pool radius π5 5

x50% Rp

(B.5)

The median floc size-to-pool radius is very small. 6. Ratio of differential-to-bowl speed π6 5

ΔΩ Ω

(B.6)

This ratio is important for cake conveyance. It is also important for sedimentation in the vicinity near the blade surface; however given the distance between adjacent blades is usually much greater than the moving layer thickness, this effect should be negligible. 7. Modified Reynolds number
 ΔρΩR2p Δρ=ρ ρΩR2p π7 5 5 (B.7) μ μ

Appendix B: Buckingham-π Analysis for Decanter and Tubular

441

The modified Reynolds number is based on pool speed and density difference. This ratio is actually a product of a basic Reynolds 2 number based
on density ρΩRp =μ and the ratio of density difference to density Δρ=ρ . It measures the inertia due to settling versus the viscous drag. 8. Density difference ratio

π8 5

Δρ ρ

(B.8)

This ratio is very small given the density difference between biological solids and the suspending liquid is small. As can be seen, this ratio has already been incorporated in the modified Reynolds number, so it becomes redundant. However, Eq. (B7) divide by Eq. (B8) return back the Reynolds number without density difference. Indeed other than the ratios π3 5 Rb =Rp which have been considered not important unless the pool depth is comparable to that of the moving layer thickness, and one other π group (π8 5 Δρ=ρ) becomes redundant, the remaining six π groups are all important. When compared to the settling in surface moving layer, four π groups (π1 ; π2 ; π5 ; and π7 ) out of the six can be combined into a more important π9 group, which is essentially the Leung number. This is seen as follows.   1 π1 1=2 π9 5 π5 π7 π2 !1=2  2 !1=2 Rp Rp π 1 1 Ro 1 π9 5 π7 π2 2 5 Re L x50% c π5 pffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffi Q=L μ=Δρ 5 ΩRp x50%

(B.9)

By squaring Eq. B9, 2πQ  ðπ9 Þ2 5  2 Δρx50% ðΩ2 Rp Þ
μ

2πLRp


 Q 5 2π=18 BLe2  vs Ap

(B.10)

Except for a constant 2π/18, the Leung number is essentially the square root of the ratio of the feed rate Q to the clarification rate by centrifugal sedimentation vs Ap. The Leung number is a combination of four dimensionless π groups (Ro, Re, Lc/Rp, x50/Rp). This is somewhat

442

Appendix B: Buckingham-π Analysis for Decanter and Tubular

analogous to the Raleigh number used in natural convection, wherein Ra 5 GrPr 5 ðΔρ=ρÞRe2 Pr. Thus Ra is also a combination of three dimensionless π groups, such as the density difference ratio, Reynolds number, and the Prandtl number, Pr. Returning to our analysis, similarly, π4 5 xo =Rp and π5 5 x50% =Rp can be combined to form a new dimensionless group, π10 5

xo =Rp 5 xo x50% =Rp

(B11)

Now, π9 and π10 are two important dimensionless variables (after recombination) for the floc settling in the moving layer. These two new π groups that are the direct consequence of the analytical model involve the previous six π groups from the Buckingham-π analysis. As stated earlier the latter does not suggest which π groups are important for the problem and how they should be combined. In summary the results of the Buckingham-π analysis can also be regrouped to derive Le and xo . These two dimensionless numbers involve all the basic dimensionless π groups with the exception of only one π group, Rb/Rp. In our model, there is also an accelerator efficiency ηa to account for the actual pool speed Vt 5 ΩRpηa. The feed acceleration efficiency does not bear out from the Buckingham-π analysis given it is already dimensionless. However, the accelerator efficiency ηa can be incorporated into the denominator of Eq. B9 to come up with identical form as Eq. 15.9 in Chapter 15 on the definition of Le number. The above result is applicable to both decanter and tubular centrifuges. In the absence of flocculation, the only governing parameter is just the Leung number, Le.

B2

Disk-Stack Centrifuge (Separation/Clarification)

For sedimentation in disk-stack centrifuge, there are six variables. These are feed rate Q, liquid viscosity μ, density difference between solids and suspending liquid Δρ, radius R, rotation speed Ω, and reference particle size xo. Using Δρ, R, and Ω which has mass [M], length [L], and time [T], respectively, we can formulate three dimensionless π groups. On using Buckingham-π formulation, we obtain Rossby number similar to Eq. B1, which measures convection-to-rotation velocity, Reynolds number that measures inertia-to-viscous effect, and dimension ratio. π1 5

Q ΩR3

(B12)

Appendix B: Buckingham-π Analysis for Decanter and Tubular

π2 5

443

ΔρΩR2 μ

(B13)

xo R

(B14)

π3 5 Combining, 

π4 π4 5 π4 π23

1=2

  1 Qμ 1=2 5 Ωxo ΔRR3

This is precisely the Le number for the disk-stack, Eq. (B16): rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ρ tanθ 3Q=n μρ Δρ ðR3o 2 R3i Þ Le 5 Ωxo η

(B15)

(B16)

Obviously, n, η, and θ (in radians) are not involved as they by themselves are already dimensionless and cannot be brought out by the Buckingham-π formulation. Similarly in the presence of flocculation, xo from Eq. B11 needs to be included in addition to the Le given by Eq. B16.

B3

Spintube Centrifuge (Separation/Clarification)

For sedimentation in spintube centrifuge, there are six variables. These are tube height H, time for centrifugation t, liquid viscosity μ, density difference between solids and suspending liquid Δρ, centrifugal acceleration G, and reference particle size xo. Using Δρ, H, and t which has respectively mass [M], length [L], and time [T], according to Buckingham-π theorem, we can formulate three dimensionless π-groups. On using Buckingham-π formulation, we obtain the ratio of distance to reach terminal velocity to the spintube height, Reynolds number, and a dimension ratio of height-to-particle-size, respectively, Gt2 H

(B17)

ΔρH 2 μt

(B18)

π1 5 π2 5

444

Appendix B: Buckingham-π Analysis for Decanter and Tubular

π3 5

H xo

(B19)

Combining, 

π23 π4 5 π1 π2

1=2

  1 Hμ 1=2 5 xo GtΔR

This is precisely the Le number for the spintube centrifuge: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ffi 1 ð2πH Þ μ Le 5 xo ηðGtÞ Δρ

(B20)

(B21)

Obviously, 2π is a constant and η is not involved as by itself is already dimensionless and cannot be brought out by the Buckingham-π formulation. Similarly in the presence of flocculation, xo from Eq. B11 needs to be included in addition to the Le given by Eq. B21.

Appendix C: Centrate or Concentrate Discharge Through Rotating Impeller

In some design on centrate liquid discharge or flowable concentrate discharge, a rotating impeller is used. A rather simple analysis on this type of design is made to highlight the underlying principle of this device. The rotating impeller can be compared with the pairing disk or centripetal pump, which are stationary dipping into a rotating liquid pool. Here, this is the reverse with a rotating impeller discharging to a stationary pool of liquid. This is shown in Fig. C.1. The discharge velocity is a summation vectorially of the tangential speed ΩRD due to rotation with Ω and discharge radius RD and the

vB Backpressure valve

pB

Ω

vD

pD vP =Q/AP

v

vD Stationary chamber

ΩRD

Rotating impeller

Figure C.1 A rotating impeller discharging into a pool of liquid in a stationary collector, which is subsequently directed to be discharged outside the machine through a back-pressure valve.

445

446

Appendix C: Centrate or Concentrate Discharge Through Rotating Impeller

velocity from the total flow, Q/AP. AP is the total area of the nozzles in impeller (only one nozzle is shown in Fig. C.1).  2 Q 2 2 (C.1) vD 5 ðΩRD Þ 1 AP This velocity at discharge of the nozzle is reduced as the jet is landed onto the pool in the stationary collector. A loss factor CD , 1 will be included subsequently to account for such loss. The liquid in the stationary chamber with static pressure pD is directed to flow outside the machine through a back-pressure valve, where pressure builds up to pB and with velocity vB. The area of the valve is AB. Bernoulli equation can be written between the two locations discharged at the pool “D” and the back-pressure valve “B”, assuming that the entire flow is in a streamline for which the Bernoulli equation is applicable. 1 1 pD 1 ρv2D 5 pB 1 ρv2B 2 2

(C.2)

Q 5 vP AP 5 vB AB

(C.3)

By continuity,

Combining Eqs. (C.1C.3),   1 1 2 1 1 2 2 2 pB 2 pD 5 ρðΩRD Þ 2 ρQ 2 2 A2B AP

(C.4a)

The flow rate through the impeller pump which is the same rate through the back-pressure valve is identical Q ( . 0). Note that the pressure recovered by the back-pressure valve above pD is primarily the dynamic pressure due to rotation and at discharge radius RD of the rotating impeller. The discharge radius is already near the axis and therefore the dynamic pressure is already minimized. We can also incorporate a loss coefficient, such as Closs, to the dynamic pressure to account for the velocity being less than ΩRD as the rotating fluid decelerated upon impinging onto a stationary pool in the chamber (Fig. C.1). Thus   1 1 2 1 1 2 2 2 (C.4b) pB 2 pD 5 ρðΩRD Þ Closs 2 ρQ 2 2 A2B AP The second and third terms refer to the kinetic energy at the backpressure valve and at the impeller due to flow, respectively. Most likely

Appendix C: Centrate or Concentrate Discharge Through Rotating Impeller 447

AB has smaller area than that of the total area of the impeller nozzles AP. The difference of these two terms amounts to reaccelerating the fluid to a higher velocity and this takes away the static pressure gain pBpD. It is desirable to reduce the additional increase in velocity, so as not to sacrifice for the static pressure gain. This can be achieved by setting the back-pressure valve area AB similar in magnitude as AP so that the recovered pressure is maximum. Eqs. (C.4a,b) are similar to the counterpart, where a stationary pairing disk/centripetal pump is being dipped into a rotating pool and recover the kinetic energy of the rotating fluid to a static pressure (see Eq. 4.6).

Appendix D: Answers to Problems in Chapters 2 17

Chapter 2 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

5.1 hours 0.57 hours 21 days (1) 0.57 hours and (2) 0.051 hours (1) 566 hours and (2) 51 hours Settling rate varies as particle size to the square power varies linearly with G and inversely with viscosity. 25 m/s 36 and 11 m/s 531 and 1101 69%, 48.2%, quadratic power 3.14, 3.14 cm 1261.1 rpm

Chapter 3 (3.2) 429 mm and 8846 rpm (3.3) 2.94 mL/min (3.4) 11.77 mL/min

Chapter 4 (4.1) (4.2) (4.3) (4.4) (4.5)

0.037 seconds 0.41, 3.67, and 40.77 seconds 6 seconds, 1, 3.33, 10 µm 26 L/min 200,000 and 192,222 Pa

449

450

Appendix D: Answers to Problems in Chapters 2 17

Chapter 5 (5.1) raise the pool (i.e., reduce pool radius) (5.2) 35 rpm (5.3) 120, 23.3 rpm (5.4) 2400 rpm, same as the bowl rotation (5.5) 0.53 degree at cone-cylinder intersection, 0.80 degree at conical discharge diameter

Chapter 6 (1) 92.2% (2) 79.2% (3) 98.4%

Chapter 7 (7.1) 3.8 m, yes (7.2) 6741 rpm (7.3) See below: (1)

1 2 3 4 5

R (cm)

φs

5 6 7 8 9

0.05 0.07 0.08 0.085 0.088

Δps (Pa)

ps (Pa)

0.45 0.756 1.008 1.224 1.4256

0.45 1.206 2.214 3.438 4.8636

(2) Similar. (3) 0.2392 1 y = 0.0634x 0.2392

φs

R2 = 0.9553 0.1

0.01 0.1

1

Ps, Pa

10

Appendix D: Answers to Problems in Chapters 2 17

451

Chapter 8 (8.1)

0 5 10 20 30 40 50 60

Gt_at Omega

Gt acc 1 dec

G_total

%Rec

0 4108 8216 16,433 24,649 32,865 41,081 49,298

6847 6847 6847 6847 6847 6847 6847 6847

6847 10,955 15,063 23,279 31,496 39,712 47,928 56,145

33 40 55 66 73 79 82

%Solids recovery

100% 80% 60%

correcting acceleration and decleration no correction

40% 20% 0% 0

10000 20000 30000 40000 50000 60000

(Gt) total (8.2) 90.5% (8.3) 2.34 minutes

Chapter 9 (9.3) 4.62, 7.82, cut size smaller than the mean yeast size (9.4) 310 seconds (9.5) 120 seconds (9.6) 10,000g (9.7) 63 L/min (9.8) 99.2 L/min (9.9) (1) 4.08 (2) 6.9 µm (9.10) 9.6 L/min

452

Appendix D: Answers to Problems in Chapters 2 17

Chapter 10 (10.7) 26%

Chapter 13 (13.2) 2.8

Chapter 14 (14.1) (1) 52,360 (2) 1.91(105) (3) 1(106) (14.2) (1) 0.44 (2) 0.0044 mm (14.3) (1) 13.8 (2) 0.138 mm (14.4) 28.4 mL (14.5) 0.9 mL (14.6) 6 seconds

Chapter 15 Solutions (15.3) 5.792/mm, 0.130 mm (15.4) 1.415/mm, 0.297 mm (15.5) 0.8, 0.473, 100% (15.6) 7, 4.136, 75.74% (15.7) 1

1

0.95

Rs

0.9

x50 =0.297 mm, xm=0.5 mm

x50 =0.08 mm, xm=0.13 mm

0.85 0.8

0.757 0.75

0.472

4.136

0.7

0 xo=0.01 mm

(15.8) 3.708 (15.9) 79.2% (15.10) 3.162, 83.6%

1

2

Le

3

4

5

Appendix D: Answers to Problems in Chapters 2 17

453

Chapter 16 Solutions

Solids recovery (Rs)

(16.1) 5179.8 rpm, 7325.4 rpm, 1.153, 0.8148 (16.2) 26.29%, 52.6% (16.3) 1

0.98

0.9

0.94

10 8 6

0.8 0.7 0.6

0.53

Flocculated

0.26

2

0.5

4

0.4 0.3 0.2

X m /Xo =1

0.1 0 0

0.2

0.4

0.6

0.8

1

0.8148

1.2

1.4

1.6

1.8

2

Le

1.153

(16.4) 94.46%, 98.65%. (16.5) 100%

Solids recovery (%)

Flocculated

90%

90%

Xm /Xo = 8

80% 70% 60% 50% 40% 30%

Xm /Xo = 1

20% 10%

320

57

0%

0

100

200

500 mm-dia., 7500g, X o =2μm

(16.6) 57, 320 L/min (16.7) 117 L/min (or LPM), see plot below:

300

Feed rate (LPM)

400

454

Appendix D: Answers to Problems in Chapters 2 17 100%

90%

Solids recovery (%)

90% 80% 70% 60% 50%

Xm /Xo = 1

40% 30% 20% 10%

117

0% 0

50

100

150

200

250

300

350

Feed rate (LPM)

500 mm-dia., 15,000g, Xo =2 μm

(16.8) 1.32 (16.9) 0.95 (16.10) 0.95

Chapter 17 Solutions

F(x)

(17.1) 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0

5

10

15

x, μm

20

25

30

Appendix D: Answers to Problems in Chapters 2 17 1 0.9 0.8 0.7

Rs

0.6 0.5 0.4 0.3

Bimodal model 5-6 and 20-24μm

0.2 0.1 0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.7

0.8

0.9

1.0

Le (xo=20μm) (17.2) 40% (17.3) 44.55%, 4.55% (17.4) F1 5 0.4 matches the test data.

1 0.9 0.8 0.7

Rs

0.6 0.5

Test data

0.4

F1=0.4

0.3

F1=0.6

0.2 0.1 0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Le (xo=20μm) (17.5)

455

456

Appendix D: Answers to Problems in Chapters 2 17

92.4%

0.9 0.8 0.7 0.6

SRe(S) SRe(L) SRs(S) SRs(L)

Le=0.591

SRe(S), SRe(L), SRs(S), SRs(L)

1

0.5 0.4 0.3 0.2

7.6%

0.1 0 0.1

Le

1.0

3.0

(17.6) 92.4%, 7.6% (17.7) 8.2%, 21%, very sensitive to entrainment of large size fraction. (17.8) 81.5%

Index

Note: Page numbers followed by “f” and “t” refer to figures and tables, respectively.

A Accelerating, 30 31, 38, 103, 122, 245 246 channels, 109 cone, 40 feed stream, 203 204 vanes, 39, 102f Acceleration. See also Centrifugal acceleration; Gravitational acceleration efficiency, 39 40, 106 107, 267 268, 274, 281, 334 of fluid, 38 time duration, 180 Acidity, 175, 195 Acids, 176 Adenosine triphosphate (ATP), 151 152 Adverse heating effect, 100 Aggregation, 40 41 Alkalinity, 175, 195 Aluminum ion, 176 Analytical ultracentrifuge, 55 56 Angle of attack, 108 109 Angle-head test tube centrifuge, 50f Angular momentum, 40 Angular velocity, rpm, 37 38 Animal cell, 1, 2f Anionic flocculant, 177 Annular pool, 60, 62 64, 121 Antibiotics, 137 APDII. See Automatic plunger cake discharge (APDII) “Appeared heavier” liquid pool downstream, 102

Approximate analytical solution, 307 310 mass boundary layer, 308 309 membrane flux, 310 simple rotating membrane setup, 308f “Approximate” test, 332 333 Aquaculture, 151 152, 152f Arc segment, 27 28 Atezolizumab, 4 ATP. See Adenosine triphosphate (ATP) Automatic plunger cake discharge (APDII), 67 69, 68f, 69f B B1Decanter and Tubular Centrifuge (separation/clarification), 439 441 B2Disk-Stack Centrifuge (separation/clarification), 442 443 B3Spintube Centrifuge (separation/ clarification), 443 444 Baby hamster kidney cells, 138 Bacillus subtilis, 1 3, 113 114 Back-drive, 124 126 Back-pressure, 95 98, 139 141, 320 321 Bacteria, 1 3, 7, 18, 150 151, 203, 284 cultures, 151 152 lysing of, 408 409 separating, from liquid by diskstack centrifugation., 94

457

458 Baculovirus-infected insect cells, 7 8 Baffle, 102, 130 131 Baker yeast processing, 153 154, 153f Barium chloride solution, 157 Baseline, 275 cumulative distribution, 275f Bases, 176 Batch centrifuges, 49 centrifugal filter, 53 54 spintube, 49 52 ultracentrifuge, 54 60 Batch filtering centrifuges, 15 Beach, 122 123, 129, 130f Bench-scale testing, 175 184 acceleration and deceleration time duration, 180 flocculant and coagulant in bench tests, 176 177 material balance, 177 180 separability, 175 176 settling velocity, 180 184 test variables, 177 Bernoulli’s equation, 98, 446 Bimodal feed, 405 408, 406f, 408t Bimodal particle size distribution, 395f, 402f bimodal distribution, 389f bimodal size model, 387 388, 387f hybridoma cell separation, 388 390 mammalian cells with cell debris, 383 384 processing hybridoma cell broth, 384 387 size recoveries, 398 408, 403f, 404f effect of smaller size, 394 398 smaller size fraction, 405 408, 408f, 408t variation in fine fractions from debris, 390, 391f

INDEX whole cells, 391 394 Biochemistry, 54 Biological/biologics, 43 cell size, 43, 43t, 90 materials in disk centrifuge, 218 solids, 331 332 Biomass, 5 7, 138 Biopharmaceutical processing, 136 138, 136f, 236f generic flow sheet for, 18 19 Bioreactor, 8 9, 276, 429 430 Bioseparation, 299 membrane, 299 302 perpendicular to G-force, 303f, 315 328 rotating disk membrane, 303 315 Biosolids, 53, 127 132, 166, 168 169, 176 177 Biotechnology, 1 centrifugal separations, 19 and filtration, 12 16 extracellular protein, 9, 10f generic flow sheet for biopharmaceutical process, 18 19 host cells for secreting recombinant protein, 6 8 inputs and outputs of centrifuge, 19 20 intracellular protein inclusion body, 10 12, 12f liquid, 9 10, 11f platform for protein expression, 8 9 processing, 157 159 pros and cons of filtration vs. centrifugation, 16 18 separation metrics, 20 22 text organization, 22 23 “Black” pool, 245 246 Blood plasma, 157 Bottle, 210 213 Boundary conditions, 306

INDEX Boundary layer, 62 63 flow, 299 Bovine serum albumin (BSA), 308 Bowl nozzles, 90 92 Boycott effect, 71f BSA. See Bovine serum albumin (BSA) Buckingham-π analysis, 439 444 Buckingham-π theorem, 253, 338, 426 427 Buffer, 18 19, 157 159, 408 409 Butyric acid, 151 152 C C017-1A antibody, 7 8 Cake, 121, 129 130 compaction, 170 dewatering comparison, 131f dry beach, 129 130 formation, 302 helix, beach, and climb angle geometry, 130f hydraulic assist, 130 132, 131f Cake baffle, 130 131 Calibration curve, 113 “Calm” environment, 332 333 Cancer cells, 4 Cationic flocculant, 177 CD52 antigen, 3 4 Cell culture, 136, 356 debris, 145 146, 360 361 viability, 21 22 Cellulose, 318 Centrate discharge through rotating impeller, 445 448 large particles in, 399f, 400 401 small particles in, 399 400, 399f Centrate solids, 279 feed rate for four different Gs, 294f flocculated feed to centrifuge, 288f

459 function of feed rate for four efficiencies, 278f Le for disk separating mammalian cells, 364f pilot tubular centrifuge, 290f, 291f suspended solids, 20 21 and turbidity, 363 unflocculated feed to centrifuge, 287f Centrifugal acceleration, 5 6, 35 38, 77 combined with gravitational acceleration, 28 31, 29f, 31f pressure gradient in fluid under, 27 28 Centrifugal elutriation, 58 60 Centrifugal filter, 49, 53 54, 158 159, 158f, 316 320, 317f laboratory concentration and buffer exchange, 158 159 microcentrifuge tube, 53f Centrifugal filter testing, 185 187. See also Pilot testing determining protein diffusivity and solubility, 185 186 rotating membrane to control protein buildup, 185f where high-speed rotation, 186f transient membrane centrifugal filtration, 186 187 Centrifugal force, 27 28, 28f Centrifugal gravity, 237, 277 278 adjustable parameters in centrifugation, 277 278 Centrifugal sedimentation principles intuitive phenomena, 35 44 nonintuitive phenomena, 27 35 process functions, 44 45 Centrifugal separation and filtration, 12 16 filtering centrifuges, 15 16 sedimenting centrifuge, 14 15

460 Centrifugal separations, 19, 425 applications of unified separation models, 428 430 unified modeling to, 425 428 unified separation models with practice, 431 432 Centrifugation, 6 7, 27, 35 36, 135 alternative meat, 152 153 aquaculture, 151 152, 152f baker yeast processing, 153 154, 153f biotech processing, 157 159 biotech separation of inclusion bodies, 145 147 dry corn mill process, 156f enzymes processing, 148 150 ethanol production, 155 157 generic flow sheet of biopharmaceutical, 136 138, 136f hormones processing, 144, 144f insulin production, 145 mammalian cell, 138 141, 139f omega-3 from microalgae, 154 155, 155f probiotic processing, 150 151, 150f pros and cons, 16 18 vaccines processing, 147 148, 147f yeast processing, 141 144, 143f Centrifuge, 141 142, 397, 429 bimodal size feed to, 399f in bioseparation, 273 decanter, 292 294, 293f, 294t disk, 273 288 flue gas desulfurization solids, 295f spintube, 294 295, 295f, 296f tubular, 288 292 types, 273 bowl, 99 centrifugal separators, 205t

INDEX comparison, 205 207 feed particle size distribution, 222 223 inputs and outputs of, 19 20 sizing, 203 centrifuge comparison, 205 207 dimensionless Le number, 208 210 disk and tubular size vs. rate, 210f for disk centrifuge, 213 219 disk centrifuge selection, 204 205 schematic of disk, 208f selection, 203 207 sizes and rates, 207 spintube centrifuge, 210 213 tubular centrifuge selection, 204 tubular centrifuge, 223 224 types, 206t volumetric rate or capacity of, 21 Centripetal pump, 95 99, 127 128 Ceramic sealing, 111 112 Chamber bowl, 75 76, 87, 87f, 206 207 centrifuge, 238 239 Checkpoints, 4 Chinese hamster ovary cell (CHO cell), 1, 138, 358 359, 384, 388 390, 394 395 CHO cell. See Chinese hamster ovary cell (CHO cell) Chromatography, 9 10, 12, 53, 157, 409 CIP. See Clean-in-place (CIP) Circular baffle, 102 Circulatory flow, 75 76 Clarification, 45, 79, 81, 173 Clarifier bowl-to-pool radii ratio, 440 Clarifier length-to-pool radius ratio, 440

INDEX Classification, 14f, 16f, 45, 174, 381, 401 413, 402f, 426 427 Clean-in-place (CIP), 63, 85, 112 Climb angle, 129, 130f Clinical manufacturing, 425, 431 432 Coagulant, 127 in bench tests, 176 177 in pilot tests, 198 199 Coagulation, 331 332 factors, 157 from blood plasma, 157 tissue from animal cells, 157 158 Compaction, 161 163 pressure, 167 170 cake compaction, 170 decanter compaction, 168 169 solid pressure or stress transmitted across grain-tograin contact, 168f test-tube compaction, 168 testing, 166 167 biological materials, 167f bucket test, 166f solids concentration, 167f Complications, 270 271, 270f Concentrate discharge, 87 95. See also Intermittent discharge; Liquid discharge; Manual disk centrifuge applications, 93 94 bactofuge, 91f external nozzle designs, 95 external nozzle discharge, 87 88, 88f internal vortex nozzle discharge, 92 93, 92f through rotating impeller, 445 448 rotor assembly of nozze disk centrifuge, 95f small-diameter concentrate discharge, 88 92, 89f

461 Concentrate media, 167 168 Concentration polarization, 301f Conductivity, 247 248 Conical screen, 15 Construction, materials of, 111 112 Contaminants, 18 19, 45, 53 54, 79 80, 138 Continuous feed filtering centrifuges, 15 Continuum phase, 264 Controlled variables, 237 Conveyance, 124, 129 130 Conveyor back-drive, 124 125 Coriolis acceleration, 32, 76 77 Coriolis effect, 27, 32 35, 77f, 243 244 Corn starch suspension, 33 34 Critical flux, 301 302 Cross-sectional area, 164 Crystallization, 19, 136 Cultivation cycle, 139 Cumulative size recovery, 254 255 Cumulative size under F(x), 336 337, 387 388, 426 427 Cut size, 210, 213 theoretical and observed, 214f Cytochrome c, 327 328 D Darcy’s law, 163. See also Stokes’ law on percolation, 164f on percolation and concurrent compaction, 165f Dead cells, 360 361 “Dead-end” filtration mode, 302 Decanter, 238 239 compaction, 168 169, 169f Decanter centrifuge, 121 122, 122f, 127 128, 205, 207 210, 292 294, 331 332. See also Disk centrifuge

462 Decanter centrifuge (Continued) frequency and cumulative undersize distribution, 293f performance prediction, 294t sizing for, 219 221 Deceleration, 38 of fluid, 38 time duration, 180 Degree of aggregation of solids, 40 41 Degritting, 45 Deliquoring, 173 Density difference ratio, 441 Depth filter, 139 141 Desalting, 318 Dewatering, 170, 173 DHA. See Docosahexaenoic acid (DHA) Diafiltration, 9, 16 17 Diatomaceous earth, 15 16, 141 Differential speed, 124 Differential-to-bowl speed ratio, 440 Diffusion, 35, 55 Diffusion coefficient, 34 35, 303 304 Diffusivity, 311 312 BSA diffusivity, 312f Dimensionless Leung (Le) number, 208 210, 252 253 Dimensionless numbers, 304 306 Dirac delta functions, 387 390 Discharge frequency, 84 85 Discharge velocity, 445 446 Discharge time, 85 Disk angle, 77 Disk centrifuge, 273 288. See also Decanter centrifuge baseline case, 275 CIP, 112 113 commercial, 115 118, 116f, 117f cutaway of model CSE disk centrifuge, 118f model CRA 160 576, 117f model CSE-170 steam disk centrifuge, 118f

INDEX containment, 114 disk-stack centrifuge, 75 100 efficiency η in Le number, 280 282 enzymes, 286 288 unflocculated and flocculated feed in, 287f explosion proof design, 115 feed inlet and accelerator, 100 110 fine size distribution, 276 277 G-force, 277 279 inclusion body separation, 284 286, 284f bacteria lysate feeding centrifuge, 285f G-force on classification of, 286f intermittent discharge/droppingbottom, 82f lamella/inclined plate settler, 73 75 complications in inclined plate settler, 74 75 inclined plate settler principle, 73 74 lamella plate sedimentation, 73f materials, 111 112 model, 274f, 282f noise level, 115 operation, 285f performance prediction, 275t selection, 204 205 SIP, 113 114 sizing for, 213 219 efficiency η in Le number, 216 219, 217b, 218t projected area of, 214f subcritical mode, 215f for tubular, chamber, and decanter centrifuge, 219 221 tubular centrifuge, 220b, 220f, 221f

INDEX surface finish, 114 temperature control, 115 water requirements, 115 yeast processing, 282 283, 283f Disk spacing, 78 79 Disk-stack and tubular centrifuges, 354 356 centrate solids and turbidity, 363 high cell density cell culture, 361 362 low cell density cell culture, 358 361 Disk-stack centrifuge, 75 100, 78f, 204, 213, 354, 413 415, 415f adverse heating effect, 100 chamber bowl, 87, 87f concentrate discharge, 87 95 disk angle, 77 disk spacing, 77 79, 78f feed solids, 80 general disk geometry, 75 77 intermittent discharge, 82 87, 82f liquid discharge, 95 99 manual disk centrifuge, 80 82, 81f process functions of disk centrifuge, 79 80, 79t Disk-stack modeling, 353 complications, 270 271 disk model, 261 268 continuum phase, 264 dispersed phase, 264 268 entrance and exit channel region, 262f particle velocity in the channel, 264f velocity profiles and trajectory, 263f model validation, 268 270 Dispersed phase, 264 268 DNA, 53 54, 318 Docosahexaenoic acid (DHA), 154 Dropping-bottom design, 90 92, 204 205

463 Dropping-bottom disk, 233 234 Drug development, 425 Dry beach, 121, 129 130 Dry solids (DS), 230 E Earth gravitational acceleration, 36, 56 gravity, 215, 270 Efficiency, 216 219, 280 282 Effluent, 62 63, 247 248 Eicosapentaenoic acid (EPA), 154 Ekman flow, 185, 257 Ekman layer, 304, 307 Ekman number, 304 Electrolytes, 176 Eluent, 59, 59f Elutriation, 58 60 Emulsion, 128 129 Entrainment, 216, 257, 280 Enzymes processing, 148 150. See also Probiotic processing; Vaccines processing extracellular, 148 149, 149f intracellular, 149 150, 150f EPA. See Eicosapentaenoic acid (EPA) Escherichia coli, 1 3, 3f, 68 69, 77 78, 145 146, 146f, 284, 382, 408 409 bacteria, 353 Ethanol, 155 156 production, 155 157 Explosion proof design, 115 Expression, 164 166 External nozzle designs, 95 discharge, 87 88, 88f Extracellular proteins, 5, 9, 10f F Feed, 101, 104 Feed acceleration, 16 17, 331

464 Feed acceleration (Continued) conventional vs. improved accelerators, 107f efficiency, 106, 107f energy loss during, 102 103 visual testing, 104 107 Feed inlet and accelerator, 100 110 feed acceleration visual and quantitative testing, 104 107 hydro-hermetic feed design, 101 102 improved feed accelerator, 108 110 low shear, 101 power loss, 102 103 Feed particle size distribution, 222 223 size distribution of feed, 222f Feed rate, 122, 230, 277 278 adjustable parameters in centrifugation, 277 278 Feed size distribution, 382 383 Feed slurry properties, 175 Feed solids, 80, 139, 277, 279 comparing, 279f concentration, 277 Feed suspended solids, 237 Fermentation, 3f, 135 136, 145, 148 Fermenter, 8 9, 429 430 Fick’s law, 303 Field test, 340 345 decanter tests at wastewater treatment, 340 341, 341t situ floc size, 341 345 Filterability, 174 Filtering, 12 13 perforate-wall centrifuge, 13 14, 13f Filtration, 12 18 Floc size-to-pool radius ratio, 440 Flocculants, 127, 148 149, 287 288

INDEX in bench tests, 176 177 model, 336 in pilot tests, 198 199 particle size effect, 197f solids recovery curve, 198f Flocculated solids, 287 288, 331 Flocculation, 331 332, 353 354 coagulation and, 331 332 decanter centrifuge, 332 in disk-stack centrifuge, 370 375, 370f, 373f, 374f field test, 340 345 monotonic size distribution model, 333 340 prediction, 346 problems, 332 333, 332f scale-up, 346 349 Flow sheets, 135 with combined processing, 137f generic insulin, 145f Flow stream, 108 109 Flow visualization, 243 248, 246f, 247f, 248f, 251f arrangement of dye injection, 245f florescent dye illumination, 246f inner diameter painted white for visualization, 245f underlying pool with trajectory of particles, 249f Flue gas desulfurization, 294 295, 295f Fluid, 38 dynamics simulation, 270 mass, 27 28 Foregoing analysis, 98 Forward-drive, 126 Fouling, 17, 302 Free surface, 28 29 Frequency, defined, 375 376 Frequency distribution, 197 198, 265 Friction effect within flow layer, 252

INDEX G G-acceleration, 106, 266 G-force, 164 165, 277 279 effect of, 277 278, 278f feed rate versus, 291f G/g-volume scale-up, 348. See also Surface area scale-up; Sigma scale-up GA733 antigen, 7 8 Gearbox, 124 125 Gear ratio, 124 126 Gel BSA diffusivity, 312f concentration, 310 311 flux vs. applied pressure, 301 302 resistance, 300 301 Gelling, 300 Gene, 1 General manufacturing practice (GMP), 114 Generic insulin production, 145 Gentle acceleration, 139 GMP. See General manufacturing practice (GMP) Governing equations and solutions, 306 310 approximate analytical solution, 307 310 model, 306 307 Gravitational acceleration, 31, 31f combined with centrifugal acceleration, 28 31, 29f, 31f Earth’s, 36, 56 Guiding testing, 428 429, 431 432 H Helix, 130f Hermetic seal design at liquid discharge, 99 High cell density cell culture, 361 362, 361f high-density feed, 362f High centrate turbidity, 230 233

465 concentrate not discharging causing, 232 233 finer feed solids causing, 231 232, 232f high feed solids throughput causing, 230 231, 231f lower process temperature and higher viscosity causing, 233 High moisture, 233 High solids recovery, 277, 281 High vibration, 233 235 High-G Tubular, 289 High-speed centrifuge, 64f Higher acceleration efficiency, 281 Higher feed rate, 281 Higher feed solids, 121 Higher solids capture. See High solids recovery Hindered settling effect, 43, 44f Horizontal head configurations, 50 51, 50f, 50t Horizontal-head test tube centrifuge, 50f Hormones processing, 144, 144f Host cells for secreting recombinant protein, 6 8 Hub, 67 68 “Human ladder”, 163 Hybridoma cell broth, 384 387, 384t cumulative size, 386f solids recovery, 385f, 386f separation, 388 390 Hydraulic assist, 130 132 Hydraulic motor, 124 125 Hydro-hermetic feed design, 101 102 Hydrostatic pressure, 103 I IBs. See Inclusion bodies (IBs) Immune cells, 4, 7 8 Immunotherapy, 4 Improved feed accelerator

466 Improved feed accelerator (Continued) bowl just after emptying, 111f curved inlet acceleration channels, 110f inlet acceleration channels, 108f inlet feed accelerator design, 109f without smoothing disk section, 108 109 with smoothing disk section, 109 110 Improved moving layer flow model, 248 251 Impurities, 18 19, 135, 363 Inclined plate, 73 75, 76f In situ floc size, 331 Inclusion bodies (IBs), 7 9, 408 413, 410f, 418, 418f bulk density, 421 centrifuges for, 413 418 disk-stack centrifuge, 413 415 spintube centrifuge, 417 418 tubular centrifuge, 415 416 conventional, 409 411 intracellular protein, 10 12 new, 411 413, 412f separation by size and density difference, 418 422, 419f size fraction distribution, 420f solids recovery vs. Le, 419f Industrial centrifugal separation, 12 13 Inlet, 99, 101 103 Inorganic ions, 176 Insulin, 145 drug, 5 production, 145 Integrated system, 139 Intermittent discharge, 82 87, 82f. See also Concentrate discharge; Liquid discharge; Manual disk centrifuge angle of cone for discharge, 83 84, 84f

INDEX concentrate solids discharge, 85 86 discharge frequency, 84 85 feed rate, 86f intermittent discharge designs, 83 interruption, 86 87 Internal vortex nozzle discharge, 92 93, 92f Intracellular expressions, 8 9 Intracellular protein inclusion body, 10 12, 12f liquid, 9 10, 11f Intuitive phenomena, 35 44. See also Nonintuitive phenomena centrifugal acceleration, 35 38 efficiencies for fluid moved instantly, 40t fluid in centrifuge bowl not at solid-body rotation, 38 40, 39t regimes of sedimentation, 40 41, 41f relative centrifugal acceleration vs. rpm, 37f settling with concentrated solids, 43 44 Stokes’ law, 42 43 tangential velocity increases linearly, 36f Ionic strength, 18 19, 157, 175 177, 302 Isopycnic separation, 58 K Kinematic relationship, 124 125 Kinematic viscosity, 313 Kinetic energy (KE), 90 L Laboratory and pilot testing bench-scale testing, 175 184 centrifugal filter testing, 185 187 feed slurry properties, 175

INDEX liquid properties, 174 175 pilot testing, 187 199 process objectives, 173 174 solid properties, 174 175 suspension properties, 174 175 Laboratory spintube, 425 Lamella plate, 73 75 Large-scale production stage, 431 432 Leung number (Le number), 203, 333, 338 339, 355, 441 443 Leung (Le) scale-up, 346 347, 347f Levelling, 29 Light phase, 128 129 Limiting flux, 301 302 Linear velocity, 38 Liquid, 173 discharge, 95 99 flux, 164 165 intracellular protein, 9 10 liquid-pool radius, 321 322 permeates, 13 14, 300 phases, 127 128 Liquid discharge, 95 99. See also Concentrate discharge; Intermittent discharge; Manual disk centrifuge centripetal pump, 95 99 hermetic seal design at liquid discharge, 99 stationary centrifugal pump dipping, 97f stationary paring disk design, 99f Loss coefficient, 446 Low cell density cell culture, 358 361 feed PSD for low cell density feed, 359f solids recovery vs. Le for low feed concentration, 360f Low shear stress, 101 Low-G Tubular, 290 292 Lower solids recovery, 281

467 Lysing, 408 409 M mAbs. See Monoclonal antibodies (mAbs) Macromolecular species, 55 Magnesium ion, 176 Mammalian cells, 1, 8 9, 21 22, 94, 138 141, 139f, 203, 359, 383 culture, 356 363 dual-stage centrifugation, 141f particle size distribution, 140f Manual disk centrifuge, 80 82, 81f. See also Concentrate discharge; Intermittent discharge; Liquid discharge clarification of light phase, 81 liquid liquid solid disk centrifuge, 81f separation, 82 Mass boundary layer, 308 309 equation, 307 normalized local flux, 310f Material balance, 177 180 consideration for bench scale, 178 180 material balance, 178f sketch of centrifugation behavior, 179f consideration for pilot/production scale, 188 189 by mass fraction, 189 by volume fraction, 188 189 equations, 188 Mechanical problems, 229 230, 233 235 high vibration, 233 235 Mechanical vibration, 233 234 Membrane, 299 302 filtration, 299 300 flux, 310 fouling, 302 gel resistance, 300 301

468 Membrane (Continued) geometry, 317 osmotic pressure resistance, 300 301 resistance, 186 187, 321 scenarios on rotation, 302 Membrane perpendicular to G-force, 303f, 315 328 comparing test results with predictions, 323 328, 326f centrifugal filter test geometries on BSA, 325f filtrate vs. centrifugation time, 323f, 324f, 328f fraction of filtrate, 328f membrane module, 316 320, 317f mixture of enzyme, DNA, salts, 319f swinging bucket model equipped with ultrafiltration membrane, 320 323, 320f Metrics of pilot tests, 196 198 skid mounted pilot disk centrifuge, 193f, 194f turbidity as function, 196f MF. See Microfiltration (MF) Michaels model, 306 Microbial cells, 137, 203 Microfiltration (MF), 299 300 Microplates, 51 52 Microtubes, 50 51 Minimum floc size-to-pool radius ratio, 440 Misalignment, 235 Model prediction, matching data to, 341 345 Model validation, 268 270 Modified Reynolds number, 440 441 Molecular biology, 54 Molecular weight cut off membrane (MWCO membrane), 158 159, 300, 306 307

INDEX Momentum-boundary layer, 307 Monitored variables, 236 237 in pilot tests, 192 196 Monoclonal antibodies (mAbs), 1, 3 4, 7 8, 137, 353 Monodispersed PSD, 197 198 Monotonic size distribution models, 333 340, 353, 375 377, 376f equivalent for disk-stack and tubular centrifuges, 354 356 exponential floc size distribution, 336 338 floc model, 336 Leung number calculation, 338 339 model solution, 339 340 moving layer, 334 335 processing protein from mammalian cell culture, 356 363 tubular centrifuge for separating E. coli lysate, 364 365 for separating S. pneumoniae flocculate, 366 367 Motor drive, 180 Moving layer, 62 63, 333 335, 335f model, 247 248 MWCO membrane. See Molecular weight cut off membrane (MWCO membrane) N N-turbidity unit, (NTU), 282 283 Nanofiltration (NF), 299 300 Net resultant velocity, 110 NF. See Nanofiltration (NF) Noise, 115, 233 234 Nonintuitive phenomena. See also Intuitive phenomena ball trajectory changing, 32f

INDEX combined centrifugal and gravitational accelerations, 28 31 Coriolis effect, 32 35 cross section of diametric plane, 35f liquid with suspended particulates forced, 33f, 34f pressure gradient in fluid under centrifugal acceleration, 27 28 rotating column with effective angular rotation speed, 30f Nonionic flocculant, 177 Nonuniform distribution, 216 Nonvibration problems, 235 Nozzle, 92 93 Decanter, 15 discharge, 87 88 NTU. See N-turbidity unit, (NTU) Nucleic acid, 58 Nusselt number, 313 314 O Oil emulsion, 128 129 Oil-in-water emulsion, 128 Omega-3 from microalgae, 154 155, 155f Operating point, 230 231 Optimizing/optimization, 229, 235 240, 428 429, 431 432 centrifuge, 239 240 controlled variables, 237 disk, 238 239 monitored variables, 236 237 performance, 425 separation metrics, 235 236 simple optimization scheme, 237 240 Organelle, 58, 213t Osmotic pressure resistance, 300 301

469 liquid permeation through membrane, 300f Over-acceleration, 339 Overflow, 254 Oxidation, 101, 207 P Particle size distribution (PSD), 22 23, 174 175, 210 211, 231, 254, 273 274, 276, 284 285, 289, 333, 358, 385 387, 390 size recovery curve corresponding to, 410f solids recovery curve corresponding to, 410f Particle size upstream of centrifuge, increasing, 237 Peeler, 15 Pellet, 49 50, 157 Pembrolizumab, 4 Percolation, 164 166 Performance prediction, 261, 263 Peripheral discharge, 83 Permeability, 164, 300 301 Permeate flux, 186 187, 301, 313, 321 Pichia pastoris, 1 Pilot demonstration, 425 Pilot test factors, 192 199 flocculant and coagulant in pilot tests, 198 199 metrics of pilot tests, 196 198 monitored variables in pilot tests, 192 196 spintube before and after centrifugation, assuming, 190f Pilot testing, 187 199. See also Centrifugal filter testing material balance consideration for pilot/production scale, 188 189 pilot test factors, 192 199

470 Pilot testing (Continued) product yield, 190 192 Pinion, 124 126 Plant cell, 1 Plates, 62 Plexiglas weir, 246 247 Plough, 397 Plunger, 68f, 397 Polishing, 7 Polydispersed suspension, 182 183, 197 198 Polymer science, 54 Pool depth, 122 123 Pool emergence, 130 131, 130f Power-law, 325 Precipitation, 19, 136, 157 Prediction, 333 334, 346, 425, 428, 431 432 Preparative ultracentrifuge, 56 58, 57f, 59f “Pressure face”, 108 109 Primary recovery of protein, 18 19 Probiotic processing, 150 151, 150f. See also Enzymes processing; Vaccines processing Process functions, 44 45 of disk centrifuge, 79 80, 79t Process objectives, 173 174 Process problems, 229 230 high centrate turbidity, 230 233 wet/high-moisture concentrate, 233 Product yield, 190 192 Projection, 273 Project performance, 431 432 Protein, 301, 306 307 crystals, 143 144 platform for protein expression, 8 9 solution, 316 yield, 20 Protocol guiding testing, 431 432

INDEX PSD. See Particle size distribution (PSD) Pulley, 38 Pump system, 124 125 Purification, 5, 53 54, 79, 135, 299, 319 320 Pusher, 15 Q Quantitative prediction, 253 255 cumulative size recovery, 254 255 particle size distribution of supernatant/overflow, 254 total solids recovery in cake, 253 in centrate, 253 254 Quantitative testing, 104 107 R Radial discharge, 83 Raw vaccine, 147 RCF. See Relative centrifugal force (RCF) Recombinant DNA technique, 1, 5 Recombinant protein process, 18 Regimes of sedimentation, 40 41, 41f Rejectivity, 306, 322 Relative centrifugal force (RCF), 36 Repose angle, 83 84, 84f Repulping, 173 174, 284, 284f Retentate, 323 Retrograde motion, 32 33, 320 321 Reverse osmosis (RO), 299 300 Reynolds number, 305, 313 modified, 440 441 Ribs, 76 77 scraper, 63 67, 65f, 66f Richardson and Zaki correlation, 43 44 RNA, 53 54, 318 RO. See Reverse osmosis (RO)

INDEX Rossby number, 439 440 Rotating disk membrane, 303 315, 303f determining diffusivity, 311 312 dimensionless numbers, 304 306 gel concentration, 306 311 governing equations and solutions, 306 310 parametric effects, 313 315 bulk/feed concentration, effect of, 315, 315f dimensionless flux vs. Schmidt number, 314f Reynolds number effect, 313 Schmidt number effect, 313 315 wall permeate flux vs. Reynolds number, 313f wall permeate flux vs. rotation speed, 314f Rotating effect, 299 Rotating flow, 243 244 Rotating frame, 27 Rotation speed and G-force, 123 124 S Saccharomyces cerevisiae, 1 Salts, 53 54, 176 “Scale-down” testing, 296 Scale-up, 203, 346 349, 431 432 Le scale-up, 346 347 of tubular centrifuge, 223f Scale-up/scale-down, 346 347, 425, 430 Schmidt number, 305, 313 Screenbowl, 15 Secondary flow, 33 34, 77f, 185, 258, 320 321 Second-stage centrifugation, 397 Sedimentation, 174 175, 180 181 behavior for monodispersed suspension, 181 182

471 with hindered settling, 183 184 behavior for polydispersed suspension, 182 183 enhancement chemical, 127 equilibrium, 55 in rotating bowl centrifuge, 255 258, 257f, 258f feed PSD of silica suspension, 256f present improved model, 256f tests, 255 258 velocity, 55 Sedimenting, 12 13 centrifuge, 14 15 Selection, 203 Selectivity, 300, 306 Semibatch centrifuge, 49 tubular centrifuge, 60 69 Separability, 175 176 Separation, 5, 8 9, 44, 79 80, 173, 248 metrics, 20 22, 235 236 cell viability, 21 22 centrate suspended solids, 20 21 protein yield, 20 throughput rate, 21 and repulping, 45 Serum, 148 Settling velocity, 42, 180 184 apparatus for visualizing sedimentation behavior, 181 light transmission and CCD sensor, 181f sedimentation behavior for monodispersed suspension, 181 182 with hindered settling, 183 184 sedimentation behavior for polydispersed suspension, 182 183

472 Settling with concentrated solids, 43 44 SG. See Specific gravity (SG) Shear sensitive cells, 21 22 Sheave, 38 Sigma scale-up, 348, 348f Simple optimization scheme, 237 240 optimizing centrifuge, 239 240 optimizing disk, tubular, decanter, and chamber bowl centrifuge, 238 239 solids recovery vs. speed/g and feed rate, 239f Simulations, 296 SIP. See Sterilization-in-place (SIP) Situ floc size, 341 345, 342f matching data to model prediction, 341 345 matching of solids recovery, 343f, 344f pool depth, efficiency, polymer dose, and median floc size in situ, 344t solids recovery prediction with test data, 346f solids recovery vs. feed rate, 342f Size recovery (SR), 254 Sizing, 203 Slurry, 156 Small-diameter concentrate discharge, 88 92, 89f Smaller size fraction, 405 408, 408f, 408t Sodium (Na1) ion, 176 Solid bowl centrifuge, 121 122 cake conveyance, 129 132 differential speed, 124 127, 125f, 126f feed rate, 122 pool depth, 122 123 rotation speed and G-force, 123 124

INDEX sedimentation enhancement using chemical, 127 three-phase separation, 127 129, 127f, 128f Solid(s), 121, 173 concentration, 166 curve, 268 269 properties, 174 175 recovery, 21, 177 178 scraper, 63 67, 65f, 66f high-speed centrifuge, 64f solid-body acceleration, 271 rotation, 39, 106 tangential velocity, 108 109 solid-wall sedimenting centrifuge, 12 13, 13f stress, 168, 168f, 171 Solids by centrifugation compaction, 161 163 affecting product yield, 162t pressure, 167 170 testing, 166 167 concentrating underflow, 161 163 expression or percolation, 164 166 Darcy’s law on percolation, 164f, 165f human pile, one standing on top of another, 163f increasing solid concentration in underflow, 170 171 Solids recovery, 230 231, 276, 276f, 388 390, 389f, 402f, 411, 421 centrate solids versus, 280f comparing, 278f disk centrifuge efficiency, 281f feed rate for four different Gs, 293f effect of fine feed, 277f effect of G-level, 286f

INDEX pilot tubular centrifuge, 290f, 292f Solid throughput, 236 Spacing between disks, 77 79 Specific gravity (SG), 145 146 Speed, 237 Spintube, 294 295, 295f equipped with membrane module, 316 320 normalized solids recovery vs. Le number, 296f testing, 176 177 Spintube acceleration efficiency, 431 432 Spintube centrifuges, 49 52, 210 213, 213t, 417 418 biological cells, 212f, 212t horizontal head configurations, 50f, 50t smaller microbiological organisms, 214f smaller-size organelle, 213t swinging bucket configurations, 51f Spin-up, 211 SR. See Size recovery (SR) “Stagnant” pool, 247 248 Stationary membrane, 299 Stationary pairing disk/centripetal pump, 447 Stationary skimmer pipe, 88 90 Sterile filtration, 18 19, 136 Sterilization-in-place (SIP), 63, 113 114 Stokes’ law, 42 43, 179, 249 250, 252, 264 265, 427 428 Stokes’ settling velocity, 419 Stokes’ velocity, 266 267 Streptococcus pneumoniae bacteria, 353 Subcritical mode, 215, 215f Submerged hub, 290 292 Sucrose gradients, 58 Supercritical mode, 215, 215f

473 Supernatant/overflow, 254 Surface area scale-up, 348 349, 349f Surface finish, 114 Suspended particles, 53 54 Suspension, 15 16, 53 54, 173 monodispersed, 181 182 polydispersed, 182 183 properties, 174 175 unflocculated, 369 Svedberg equation, 56 Swinging bucket, 50 51, 51f, 56, 320 323 T Tangential velocity, 40 Target cell, 59 Taylor Proudman phenomenon, 304, 310 TCD. See Total cell density (TCD) Test analysis, 428 Test variables in spintube testing, 177 Test-tube compaction, 168 biosolids by weight correlated, 169f Testing, 430 Therapeutic protein, 1 3, 5 Thickening, 45, 79 Three-phase decanter, 127 129 Three-phase separation, 127 129, 127f, 128f Throughput rate, 21 Timescale of occurrence, 229 230 Total cell density (TCD), 356 357 Total solids recovery in cake, 253 in centrate, 253 254 Trajectory, 249, 265, 335 Transient sedimentation process, 55 Transmembrane flux, 328 329 pressure, 185 187, 301 302, 321

474 Transport equation, 306 Troubleshooting, 229 235, 425, 428 432 mechanical or process problem, 230 mechanical problem, 233 235 optimization, 235 240 timescale of occurrence, 229 230 Tubular centrifuge, 49, 60 69, 60t, 61f, 67t, 135, 222f, 223 224, 288 292, 353, 415 416, 416f automatic plunger cake discharge, 67 69 feed rate, 292f general tubular bowl geometry, 60 63 high-G, 289 low-G, 290 292 model, 243, 244f dimensionless Le parameter, 252 253 flow visualization, 243 248 effect of friction within flow layer, 252 improved moving layer flow model, 248 251 quantitative prediction, 253 255 sedimentation tests, 255 258 effect of velocity profile, 251 252 ribs and solids scraper, 63 67 scale-up of, 223f selection, 204 for separating E. coli lysate, 364 365, 366f PSD of feed for E. coli cells and S. pneumonia, 365f for separating S. pneumoniae flocculate, 366 367 solids recovery, 289f Tubular model, 238 239, 268 “Tuning of cut size”, 398

INDEX Turbidity, 21, 193, 196f, 357, 363 Turbulence, 332 333, 345 Turntable, 32 33 Two-dimensional flow pattern, 243 244 U UF. See Ultrafiltration (UF) Ultracentrifuge, 49, 54 60, 54f analytical, 55 56 centrifugal elutriation, 58 60 preparative, 56 58 Ultrafiltration (UF), 299 300 Ultraviolet (UV), 112 113 Under-acceleration, 339 Underflow stream/concentrate, 233 Unflocculated feed, 287 Unflocculated suspension, 369 Unified modeling to centrifugal separations, 425 428 Unified separation models applications, 428 430 analysis of test data, 428 guiding testing, 429 optimization, 429 prediction/forecast, 429 scale-up/scale-down, 430 troubleshooting, 429 430 centrifugal separation modelling, 433f integration of, with practice, 431 432 Unimodal size distribution models, 353 354, 367 377, 376f flocculation in disk-stack centrifuge, 370 375 unflocculated suspension, 369 UV. See Ultraviolet (UV) V Vaccines processing, 147 148, 147f. See also Enzymes processing; Probiotic processing

INDEX concentrated cell-based product, 147 serum product, 148 vaccine serum flow sheet, 148f Vacuum distillation system, 156 Validation model, 268 270 centrate solids, 269f solid recovery vs. feed rate, 269f Van der Waals’ force, 176, 331 332 Vanes, 109 110 VCD. See Viable cell density (VCD) Vector, 1 Velocity profile effect, 251 252 Viable and nonviable cells, 356 Viable cell density (VCD), 356 357 Vibration, 52, 233 235 Viscosity, 42 water viscosity as function, 42f Viscous diffusion, 38 Viscous stresses, 252 Volume fraction, material balance by, 188 189 W Washing/repulping concentrate, 79 80, 173 174

475 Wastewater treatment plant, 334 decanter tests at, 340 341 Water flux, 301 Water-in-oil emulsion. See Oil-inwater emulsion Weir, 62 63, 104, 257 Plexiglas, 246 247 Wet/high-moisture concentrate, 233 Whole cells, 391 394 average size, 392, 394f size range, 392 394 Y Yeast, 1, 203 cells, 1, 2f processing, 141 144, 143f, 282 283, 283f Yield, 161 163 product, 190 192 protein, 20 Z Zeta potential, 174 175 Zonal centrifuge, 14, 56, 57f, 58 59 Zonal rotors, 56