CAPTURE PUMPING TECHNOLOGY
The author wishes to give credit for the cover illustration to the artist, friend and colleague: David J Spamer, Pixel Dimensions, 4900 Knightswood Way, Granite Bay, CA 95746.
CAPTURE PUMPING TECHNOLOGY 2nd Fully Revised Edition
Kimo Welch
2001 Elsevier Amsterdam
- London
- New
York
- Oxford
- Paris
- Shannon
- Tokyo
ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 P.O. Box 211, 1000 AE Amsterdam, The Netherlands
© 2001 Elsevier Science B.V. All rights reserved.
This work is protected under copyright by Elsevier Science, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier Science Global Rights Department, PO Box 800, Oxford OX5 1DX, UK; phone: (+44) 1865 843830, fax: (+44) 1865 853333, e-mail:
[email protected]. You may also contact Global Rights directly through Elsevier's home page (http://www.elsevier.nl), by selecting 'Obtaining Permissions'. In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London WI P 0LP, UK; phone: (+44) 207 631 5555; fax: (+44) 207 631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of Elsevier Science is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier Science Global Rights Department, at the mail, fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher 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. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made.
1st edition 2001 Library of Congress Cataloging in Publication Data A catalog record from the Library of Congress has been applied for.
ISBN: ISBN:
0-444-50882-1 Hardbound 0-444-50941-0 Paperback
@ The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in The Netherlands.
This book is dedicated to the memory of Arline H. Welch, my mother. In the early years of her life she was an ordained minister in the Salvation Army. In this work, and throughout her life, she dedicated herself to the caring, educating and loving of all children.
This Page Intentionally Left Blank
TABLE OF CONTENTS
Preface ................................................................................................................. x/i Biographical Note ........................................................................................... xiii Acknowledgements ................................................................................................ x/v 1 BASIC T H E O R Y ....................................................................................................... 1.0 Introduction, Pump Classification .......................................................... 1.1 Pump Capacity .......................................................................................... 1.2 Understanding Pump Behavior ............................................................... 1.3 The Ideal Gas Assumptions .................................................................... 1.4 Definitions of Temperature and Pressure ............................................. The Ideal Gas Law, 3 Manipulating the Equation of State, 5 1.5 Counting Molecules (or Atoms) ............................................................. 1.6 Density, Pressure and Molecular Velocities ......................................... 1.7 Vapor Pressure ......................................................................................... Surface Pumping, 11 Pumping on a Liquid Cryogen, 12 1.8 Mean Free Path .......................................................................................... 1.9 Thermal Conductivity of Gases ............................................................... Gas Flux Incident on a Surface, 16 1.10 Pumping Speed, a Convenient Abstraction ........................................... 1.11 Throughput .................................................................................................. 1.12 Conductance, Another Convenient Abstraction .................................. 1.12.1 Conductance Model Applications, 22 Species and Temperature Dependence, 22 Pressure Dependence of Conductance, 22 System Geometry Dependence for Molecular Flow, 23 Molecular Conductance for Different Gases, 24 1.13 Voltage, Current and Impedance Analogies ......................................... 1.14 Dalton's Law and Linear Superposition ................................................ 1.15 Selective and Variable Pumping Speed ................................................. 1.16 Measuring Pump Speed ........................................................................... 1.16.1 Rate of Pumpdown, 33 1.16.2 Single Gauge Dome Method, 36 1.16.3 Three Gauge Method, 37 1.16.4 Fischer-Mommsen Dome, 39 Three Gauge Versus Fischer-Mommsen Results, 41 Speed Measurement Errors Due to Trace Gases, 41 1.17 System Diagnostics With Any Pump ...................................................... Measuring Throughput by Rate-of-Pressure-Rise, 43 vii
1 1 1 1 2 2
5 6 8
12 15 17 19 20
25 28 32 33
42
TABLE OF CONTENTS 1.18
Electrical Discharges in Gases ................................................................ lonizing Gases, 46 High Pressure Electrical Discharges, 46 Low Pressure Ionization Processes, 50 Ionization Gauge Sensitivities, 53 Partial Pressure Gauges, 53 Ion Collectors Spectra Interpretation, 57 Gauge Calibration Methods, 59 1.19 Vacuum Seals, 62 1.19.1 Elastomer Seals, 62 Outgassing From Elastomers, 63 1.19.2 Metal Seals, 65 1.20 Comments on Helium Leak Detection .................................................. 1.20.1 System Applications, 71 1.20.2 Leak Checking Systems Appended With Capture Pumps, 76 Problem Set ........................................................................................................... References ............................................................................................................
44
71 76 78
2 SPUIq'ER-ION PUMPING .................................................................................... 83 2.0 Introduction .................................................................................................... 83 2.1 The Penning Cell ............................................................................................ 84 2.2 I+/P Characteristics in Penning Cells .......................................................... 86 Qualitative Pump I/P Characteristics, 90 2.3 Pumping Speed Abstraction .......................................................................... 93 2.4 The Making of Sputter-Ion Pumps ............................................................... 94 2.4.1 Pumping Mechanisms and Materials ................................................. 94 Sputtering and Sputter-Yield, 94 Physisorption and Binding Energies 97 Chemisorption, 99 2.4.2 Diode Pumps ......................................................................................... 102 The Need for Clean Pumping, 102 2.4.3 Noble Gas Instabilities ....................................................................... 105 SLAC Pump Instability Problem, 106 Diodes With Slotted Cathodes, 108 The Galaxy~ Diode Pump, 109 2.4.4 The Triode Pump ............................................................................... 110 Magnetic Fields in Diode and Triode Pumps, 113 Field Emission in Triode Pumps, 114 2.4.5 The Differential Sputtering Pump ..................................................... 114 The High Energy Neutral Theory, 115 Beware of Shortcuts, 118 2.5 Hydrogen Pumping .................................................................................. 119 2.5.1 Hydrogen Pumping in Diode Pumps ............................................... 121 Diffusion in the Cathodes, 127 Titanium Cathode Material, a Model, 130 Ti-6A£-4V Cathode Material, 133 viii
TABLE OF CONTENTS 2.5.1 Hydrogen Pumping in Diode Pumps ................................................. 121 Beta-Stabilized Cathode Materials. 133 Diodes Withe Aluminum Cathodes. 134 Hydrogen Burial in Diode Pump Walls. 136 Conclusion on Hydrogen Pumping. 137 138 2.6 Triode Pumping .............................................................................................. 2.7 Transient Speed Effects ................................................................................. 142 2.8 Pumping Gas Mixtures ................................................................................... 143 2.9 High Pressure Operation ............................................................................... 145 Advantages in Low Pressure Operation. 149 149 2.10 The Sputter-Ion Pumping of Helium ..................................................... Regeneration After Pumping Helium. 153 2.11 Pumping Elements Located in Antechambers or Pockets ................... 154 Distributed Sputter-Ion Pumps (DIPS). 156 2.12 Pump Power Supplies .............................................................................. 158 2.13 Magnet Designs ........................................................................................ 160 164 2.13 More on the Nature of Penning Discharges .......................................... 164 2.13.1 Space Charge Distribution in Penning Cells ................................... Discharge Modes. 164 The Stark Effect. 165 Frequency Shift in RF Cavity. 166 Ion Sputtering Patterns. 167 Ion Currents and Retarding Potential Probes. 168 Electron Beam Probes. 169 Smooth.Bore. Magnetron Model. 169 2.13.2 More on Sputtering Patterns on Pump Cathodes ........................... 170 Noise in Sputter-Ion Pumps. 171 172 2.13.3 Striking Characteristics ...................................................................... 173 2.13.4 Use of Very High Magnetic Fields ................................................... 2.14 Other Considerations ............................................................................... 173 Maintenance and Trouble Shooting. 173 2.15 A Summary of Advantages and Disadvantages .................................... 175 Appendix 2A .Electrostatic Getter-Ion Pumps ............................................... 177 179 Problem Set ............................................................................................ 181 References ............................................................................................................. TITANIUM SUBLIMATION PUMPING ............................................................... 3.0 Introduction .................................................................................................... 3.1 Sticking Coefficients ....................................................................................... 3.2 Pump Speed vs. Sticking Coefficient. a ....................................................... Titanium and Conductance Limited Operation. 203 Dependency of a on Gas. Film Thickness and Temp., 204 3.3 Synthesis. Displacement and Dissociation of Gases .................................. 3.4 Sublimation Sources ....................................................................................... Filamentary Sources, 211 Constant Current Operation, 212 Constant Voltage Operation, 214
ix
195 195 1%
200 208 210
TABLE OF CONTENTS 3.4 Sublimation Sources ....................................................................................... 210 Radiantly Heated Sources, 216 E-Gun Sources, 218 3.5 Combination Pumping ................................................................................... 220 Peeling of Titanium Films, 222 3.6 Advantages and Disadvantages of TSP Pumping ....................................... 222 Problem Set ............................................................................................................ 223 References ............................................................................................................. 224 4 N O N E V A P O R A B L E GETTERS (NEG) ............................................................. 4.0 Introduction .................................................................................................... 4.1 Mechanical Features ...................................................................................... 4.2 NEG Pumping Mechanisms .......................................................................... The Pumping of CO, CO ~, N ~, O ~, 233 Hydrogen Pumping, 235 The Pumping of Hydrocarbons, 240 Pumping Speeds for Gases and Gas Mixtures, 241 4.3 Sintered NEG Structures ........................................................................ Some Interesting Applications, 248 4.4 Advantages and Disadvantages of NEGs .................................................... Problem Set ........................................................................................................... References ...........................................................................................................
229 229 229 232
5 C R Y O P U M P I N G ...................................................................................................... 5.0 Introduction .................................................................................................... 5.1 Cryocondensation v s . Cryosorption ............................................................ 5.1.1 Cryocondensation Pumping .............................................................. 5.1.2 The Clausius-Clapeyron Equation ................................................... Condensation at Higher Pressures, 260 5.1.3 Thermal Transpiration ....................................................................... 5.1.4 Adsorption Isotherms ........................................................................ Classifications of Adsorption Isotherms, 267 Helium and Hydrogen Isotherms on 4.2* k Surfaces, 271 5.1.5 Speed and Capacity of Cryopumps .................................................. 5.1.6 Cryotrapping ....................................................................................... 5.1.7 Sieve Materials .................................................................................... Plugging of Sieve Materials, 279 Surface Bonding of Sieve Materials, 283 5.2 Sorption Roughing Pumps ............................................................................. 5.2.1 Staging of Sorption Pumps ................................................................ Effects of Neon When Rough Pumping, 291 5.2.2 Dewars and Bakeout Regeneration Heaters ................................... 5.2.3 Safety Considerations ......................................................................... 5.3 Liquid Helium Cryopumps ............................................................................ 5.3.1 Classification of LHe Cryopumps .................................................... 5.3.2 Chevron Design ..................................................................................
255 255 255 257 258
246 248 249 251
261 264
274 276 277
284 287 292 293 295 295 299
TABLE OF CONTENTS 5.4 Closed-Loop, Gaseous Helium Cryopumps ............................................... 301 Chevron Design, 303 Sticking Coefficients, 306 Thermal Loading of Cryopumps, 307 Sputtering Applications, 309 Cryopump Applications, 312 System Configuration, 313 Regeneration of Cryopumps, 315 Reverse Cycle Regeneration, 318 The Placebo Effect, 319 Sources & Remedies of He Gas Contamination, 320 5.5 Closed-Loop, Gaseous Helium Refrigerators ............................................ 321 Some of the History, 321 The GM-Cycle Refrigerator, 323 The Expander, 323 The Compressor, 326 Refrigerator Capacity, 329 Networking Cryopumps on Complex Systems, 329 5.5.4 Meissner Coils & Traps Using Vapor Refrigerants ...................... 331 Problem Set ............................................................................................................ 331 References ............................................................................................................. 334 SUBJECT INDEX ................................................................................................ 345 AUTHOR INDEX .............................................................................................. 353
xi
PREFACE
This is a practical textbook written for use by engineers, scientists and technicians. It is not intended to be a rigorous scientific treatment of the subject material as this would f'dl several volumes. Rather, it introduces the reader to the fundamentals of the subject material, and provides sufficient references for an in-depth study of the subject by the interested technologist. The author has a lifetime teaching credential in the California Community College System. Also, he has taught technical courses with the American Vacuum Society for about 35 years. Students attending many of these classes have backgrounds varying from high-school graduates to Ph.D.s in technical disciplines. This is an extremely difficult class profile to teach. This book still endeavors to reach this same audience. Basic algebra is required to master most of the material. But, the calculus is used in derivation of some of the equations. The author risks use of the first person/, instead of the author, and you instead of the reader. Both are thought to be in poor taste when writing for publication in the scientific community. However, I am writing this book for you because the subject is exciting, and I enjoy teaching you, perhaps, something new. The book is written more in the vein of a one-on-one discussion with you, rather than the author lecturing to the reader. There are anecdotes, and examples of some failures and successes I have had over the last forty-five years in vacuum-related activities. I'll try not to understate either. Lastly, there are a few equations which if memorized will help you as a vacuum technician. There are less than a dozen equations and half that many rules of thumb to memorize, which will be drawn on time and again in designing, operating and trouble-shooting any vacuum system. These key elements are identified with the symbol '~" in the text. The student will want to master the derivation of these concepts, where the daily user eventually accepts them as the laws of physics, and applies them.
xii
ABOUT THE AUTHOR
Kimo Welch has worked in vacuum-related industries for the last 45 years. This background includes work in the microwave tube industry, at General Electric, Raytheon and Litton; in the vacuum components and equipment industry at Varian; and, work in the high energy physics industry at the Stanford Linear Accelerator Center, and the Alternating Gradient Synchrotron Department and Relativistic Heavy Ion Collider at Brookhaven National Laboratory. He has managed large R&D departments, manufacturing operations and P&L centers for several companies. He has a teaching credential in the community college system of the State of California, having taught courses in electrical engineering and mathematics. As a youngster, Kimo worked in the shipyards of Stockton, where he was a journeyman welder. He later spent time on the board, and also worked as a microwave tube technician. He eventually went on to obtain an undergraduate degree in physics from the University of the Pacific, and a Masters degree in Electrical Engineering at Stanford University. He is a member of the Society of Vacuum Coaters, and on the Education Committee of that Society. He is an Emeritus member and Fellow of the AVS (formerly the American Vacuum Society). It is this practical and theoretical experience, in conjunction with his teaching experience, that makes this book of interest to the reader.
xiii
ACKNOWLEDGEMENTS I am forever indebted to my wife, Junella, for her love, understanding and support in all things, including the writing of this book. Also, I express my sincere appreciation to Dr. J.V. Lebacqz, Dr. Robert Jepsen, Mr. James Lind, and Mr. Bert Ryland for their support and guidance in various stages of my career. I also consider Dr. Derek Lowenstein and Dr. Satoshi Ozaki, of Brookhaven National Laboratory, as mentors who honored me with their trust during my nine years of senior project management at that institution. I have indeed been privileged to work with such accomplished technologists and true gentlemen. I prize the many vigorous and enthusiastic technical discussions which I have had with colleagues and friends including Henry Halama, Peter Hobson, Marsbed Hablanian, Ralph Longsworth, John Helmer, William Wheeler, David Harra, and my friends and colleagues at CERN including Alain Poncet, Cristoforo Benvenuti and Hartmut Wahl, over past decades. These exchanges have shown light on the subject of this book. My sincere thanks to Marion V. Heimerle for typing and editing much of the first edition, and to Joseph E. Tuozzolo for editing the work and working many of the problems at the end of the chapters. Lastly, my thanks to the staff of the Brookhaven Libraries for their help, and to the Associated Universities, Inc., Brookhaven National Laboratory and the Department of Energy for the use of their libraries on weekends and evenings.
xiv
CHAPTER 1
BASIC THEORY 1.0 Introduction, Pump Classification There are three types or classifications of UHV(1 ) (ultrahigh vacuum) pumps: 1) momentum transfer pumps; ( 4 2 ) 2) capture pumps; and, 3) hybrid pumps, the latter comprising combinations of the former two pump classifications. Examples of momentum transfer pumps include diffusion, turbomolecular, and compound turbomolecular pumps. Examples of capture pumps include all forms of cryopumps, sputter-ion and other getter pumps. This book will deal only with capture pumps. Momentum transfer pumps serve as a conduit for compressing gas, and conveying it along their innards, to be exhausted at higher pressures, called forepressures. Because of this, momentum transfer pumps have no inherent capacity limitation; that is, no limitation in the amount of gas they can remove from the vacuum system. Hybrid pumps feature some form of mechanically cooled trap or array at the inlet of the momentum transfer pump. The refrigerated array in addition to affording high water vapor pumping speed, may also serve to trap backstreaming pump fluids from entering the system. Also, when such traps are used with turbomolecular pumps, the combination is called a cryoturo pump.
1.1 Pump Capacity When discussing vacuum pump capacity, there is sometimes confusion, as pump capacity has two meanings. One meaning relates to the total amount of gas that the pump can remove from the vacuum system prior to having to be serviced in some manner. The second meaning relates to the rate at which the pump can remove the gas from the vacuum system without exceeding its design limitation. The former is typically referred to, simply, as pump capacity. The latter is referred to as throughput capacity. Both of these concepts will be discussed at length. All capture pumps share one common feature: they store gases which are pumped. Because of this, all capture pumps have a finite capacity for the amount of gas they can pump. This capacity differs depending on the species of gas being pumped, and the pressure at which it is being pumped. Once this capacity is exceeded, the pump must either be replaced, or refurbished by some process unique to the type of pump. The refurbishment process might require a complete rebuilding of the pump, as in the case of sputter-ion pumps, or merely executing some sort of pump warm-up cycle as in the case of cryopumps.
1.2 Understanding Pump Behavior You have a significant advantage as a user of capture pumps if you have a quantitative feel for the capacity and throughput limitations of your pump. Without this insight, problems of misapplication occur, and there will often be misinterpretation of pump performance. Also, you are at an advantage if you are able to predict when the pump has exceeded its finite storage capacity, and you are able to recognize these symptoms. Better yet, if you can anticipate when this storage capacity will be
2 BASIC T H E O R Y exceeded, you will have control over your equipment rather than the equipment controlling you. This system insight requires the ability to make estimates of the rate at which gas is being pumped. These are simple concepts, which I refer to as counting molecules. They require a basic understanding of the behavior of gases, and the models or devices of man used to describe the behavior of gas in a vacuum system (e.g., pump speed, conductance, etc.) . If you have mastered these concepts, go on to Chapter 2. If not, take the time to study this material, as it will afford you a better understanding and control of your equipment.
1.3 The Ideal Gas Assumptions The kinetic theory of gases predicts the macroscopic behavior of a gas by statistically averaging the behavior of individual particles or molecules of that gas. There is nothing complex or abstract about this theory, with the single exception that we are dealing with microscopic particles which we cannot see. Not being a theoritician, I find comfort in the fact that the major advances in kinetic theory were almost totally empirical; that is, engineers and scientists fussing around in a laboratory or shop, over a period of several centuries, reached certain conclusions about the behavior of gas. For example, Newtonian mechanics - dealing with concepts of force, momentum, inertia, etc. - was, for the most part, founded on empiricism. Charles', Boyle's and Dalton's Laws were founded entirely on empirical observations. Of course, there were notable theoriticians who contributed to the kinetic theory. Maxwell's theoretical derivation of the distribution function was of singular importance. But, I am getting ahead of myself. There are some assumptions, with regard to the behavior of gases, essential to the development of kinetic theory. These include: A. Gas comprises a very large number of molecules or atoms. B. The size of these particles or molecules is small compared to the space which they occupy. C. Collisions of these molecules with their neighbors and with the walls of the container are elastic. D. The behavior of these molecules may be described with Newtonian mechanics. E. External forces on the molecules such as gravitational, electrostatic, etc., are negligible.
1.4 Definitions of Temperature and Pressure For the most part, except at very low temperatures and very high pressures, the above assumptions are realistic. But, temperature and pressure are terms which we have not yet def'med. There are several ways in which temperature may be defined. It is known that when molecules are contained in a vessel, the hotter the vessel, the greater the average velocity of the molecules. Therefore, perhaps we could use this
BASIC T H E O R Y 3 average molecular velocity as a means of defining the temperature of the vessel (and gas). Steam and ice are common states in nature. Another approach in defining temperature might be to use the ice point, T i, and steam point, T s, of water as two temperature states, arbitrarily assigning O* C to the ice point and 100" C to the steam point (i.e., the scale is arbitrarily divided into 100 equal parts, as in the * C scale). Defining pressure as force per unit area (e.g., lbs./in 2 , Nt/m ~ , etc.), experiments conducted with different gases led to the experimental observation with regard to pressure, P, at steam and ice points: Ts
limit Pi'* 0
Ti
Ps
1.366.
Pi
The term "_a" is used by mathematicians to mean is defined as. Using this definition and requiring that (T s - Ti) = 100" K, leads to Ti = 273.16" K and a value T s = 373.16" K. Again, the scale between the ice point and the steam point was conveniently and arbitrarily divided into 100 units.
The Ideal Gas Law Constant pressure thermometry is essentially founded on Charles' Law from which it was determined that, within limits, the volume assumed by a fixed quantity of gas will vary linearly with temperature (Fig. 1.4.1). 10
....
i ....
LIQUEFACTION / TEN I
F i g u r e 1.4.1. C h a r l e s ' Law w h e r e t h e v o l u m e of a g a s a t c o n s t a n t pressure varies linearly with temperature.
i
0
0
100 TEMPERATURE-
200 OK
Boyle determined that for a fixed quantity of gas at a fixed temperature, the volume occupied by that gas varied inversely with pressure (Fig. 1.4.2). i0
'
'
'
'
I
"
'
'
'
P xVis F i g u r e 1.4.2. B o y l e ' s Law w h e r e t h e p r o d u c t of t h e p r e s s u r e a n d v o l u m e of a n i d e a l g a s is e q u a l [o a c o n s t a n t .
I
0
i
i
l
i
,
i
I
PRESSURE -
l
10
5
Torr
4 BASIC T H E O R Y Avogadro, in experiments with various elements, discovered that they combined in chemical reactions in certain proportions. From this he was able to def'me what chemists now call a mole of something. One mole of anything is -6.023 x l0 s a of those things (e.g., atoms, molecules, boxcars, etc.). According to Avogadro, the volume occupied by any gas, at a fixed temperature and pressure, is directly proportional to n, the number of moles of that gas. The symbol N O is frequently used in the literature to denote Avogadro's number. These three important fmdings are mathematically summarized as follows: Charles'Law:
Vcx T
(1.4.1a)
Boyle's Law:
Vcx 1 / P
(1.4.1b)
Avogadro's Law:
Vcx n
(1.4.1c)
6.023 x l0 s a particles ~ 1 mole
(1.4.1d)
6.023 x 10 2 a particles 1 mole
N O•
Combining the empirical results of Charles, Boyle and Avogadro, we have: Vcx
n TIP.
(1.4.2)
The algebraic manipulation of (1.4.2) and assigning of a proportionality constant "~1", called the Universal Gas Constant in chemistry textbooks, yields the following: trl
PV =
n~ T.
(1.4.3)
The value of the term '~" depends on the units selected for P, V and T (of course a mole is as defined by Avogadro). In vacuum-related work, the terms P, T and V may have different units. For centuries, chemists typically have used atmospheres (e.g., 0.5 Atm., 1.5 Atm., etc.) as the unit of pressure, P. Volume, V, was typically given in liters, and temperature in degrees Kelvin. Vacuum technologists have made a jumbled mess of units in the short span of fifty years. I will use Torr as the unit of pressure (P), liter (Z) as the unit of volume (V), and degree Kelvin as the unit of temperature (7). To convert these terms to other units, you are referred to a basic vacuum technology handbook.(2 ,a ) However, just this once: 1 Torr = 133.32 Pascals. The value of J1, for the units I've selected, is then: ]t ~ 62.36 Torr-Z/mole * K
(1.4.4)
Equation (1.4.3) can be rearranged as follows: P = (n/V)~l T, or P = p • T,
(1.4.5)
where p is the density of gas (i.e., moles/unit volume). Physicists refer to (1.4.5) as the equation of state.
BASIC T H E O R Y 5
Manipulating the Equation of State Most of the equations we vacuum technicians use to describe processes going on in a vacuum system have their origins in (1.4.3). For example, if we differentiate (1.4.3) with respect to time, while assuming temperature is constant we obtain: (1.4.6)
P dV/dt + V dP/dt = ~ T dn/dt
The term dV/dt in (1.4.6) is def'med as speed, and has the symbol S. The concept of speed is discussed in Section 1.10. The term to the fight of the equal sign in (1.4.6) is called throughput, and has the symbol Q in the literature. Throughput is discussed in Section 1.11. If the pressure in a vacuum system is not changing in time, then the term VdP/dt is zero. Therefore, (1.4.6) may be rewritten in the familiar form: (1.4.7)
SP = Q
We will return to (1.4.6) and (1.4.7) and variations thereof time and again in the following sections and chapters.
1.5 Counting Molecules (or Atoms) The simple expression given in (1.4.3), and variations of it, will be used throughout the text. Equations (1.4.1d), (1.4.3) and (1.4.4) are worth memorizing, as they are invaluable tools to the vacuum technologist in counting gas molecules pumped by a capture pump. Let's work a simple example using these three equations. Assume you have a sealed vessel where V = 1.5 Z, the pressure in the vessel is P = 750 Torr, and the temperature of the vessel (and gas) is T = 300* K. How many gas molecules are contained in the vessel?
(P)(v3 -"
or,
tl,
(750 Torr)(1.5 Z) (62.36 Torr-Z/mole * K)(300* K)
~
0.06 moles.
(1.5.1)
Using Avogadro's number (which you memorized), we know: 6.023 x 10 2 3 molecules or,
6.023 × 10 2 a molecules 1.0 mole
= =
1.0 mole, "1.0".
(1.5.2)
We know that we can multiply any number by one and not change its value. If we multiply the result of (1.5.1) by (1.5.2) we conclude there are --3.6 x 10 2 2 molecules in the vessel.
6 BASIC THEORY
1.6 Density, Pressure and Molecular Velocity It is somewhat incredible that there were no restrictions placed on the type of gas, when stating the gas laws given in equations (1.4.1a) - (1.4.1d). Incredible, in that these laws apply to molecules and atoms of all sizes. The vessel in the example given in the last section could have contained small and relatively light helium atoms, with the atomic weight o f - 4 amu (i.e., atomic mass units). Or, it could have contained very large and comparatively heavy xenon atoms, which have an atomic weight of ---130 ainu. For the same P, V and T in the above example, there would be the same number of atoms in the vessel, regardless of the atomic weight of the gas. The density of gas atoms, that is the number of atoms per cubic centimeter, would be the same whether He or Xe. What would happen to the number of molecules in the vessel if we increased the temperature of the vessel? Nothing, as it is a sealed vessel, and the number of molecules per unit volume would remain the same. However, according to (1.4.3), the pressure would increase with increasing temperature. Pressure is defined in Section 1.4 as the force, F, exerted per unit area, A, on the walls of a vessel by a gas (i.e., P = F / A ) . In order to determine the pressure, we need only use some sort of mechanical pressure gauge. This, however, tells us nothing about the density of gas in the vessel, unless we also know the temperature. Pressure has no meaning at the very center of the vessel, but density does. Therefore, in order to determine the density of gas molecules in a vessel, we must know P, V and T. This is the distinction between density and pressure. Some vacuum gauges measure the mechanical force exerted by a gas on a wall or membrane comprising the gauge envelope. On the other hand, both total and partial pressure ionization gauges measure the density of the gas. What is the distinction? Assume you have a sealed envelope appended by a "mechanical" vacuum gauge. If you increase the temperature of the envelope (and gas), the pressure measured by the gauge will increase in accordance with (1.4.3). That is, if the initial pressure is Pi and initial temperature Ti, and the final temperature Tf, the final pressure in the chamber, Pf, will be: Pf = Pi T f / T i. In this case the density of the gas has not changed, but the pressure has. A mechanical gauge would detect this pressure change, but an ionization gauge would measure no pressure change. We know that pressure on the walls of the vessel stems from collisions of the gas atoms on the walls. How is it that the much heavier Xe atoms produce the same force (i.e., change in momentum per second) on the walls of the vessel as that imparted by the lighter He atoms? We can only conclude that the number of Xe atom wall collisions per second must be less than the wall collisions of the He atoms, under the same conditions. Knowing that there are the same number of atoms in the vessel, for the same P, 1,I, and T, we can only conclude that the Xe atoms must be moving about the vessel at a much slower velocity than the He atoms. This proves to be the case; that is, there is a relationship between the velocity of the atoms (or molecules) of an ideal gas and the atomic (or molecular) weight of that gas. Also, we know that when we heat the gas in a container, the pressure increases on the walls of the container. Therefore, the velocity of the molecules must increase with temperature.
BASIC T H E O R Y 7
Molecular Velocities In 1859 Clerk Maxwell theoretically derived an expression which made it possible to predict how molecular velocity varied with temperature and molecular weight. He predicted that in a very large collection of molecules, there would be a distribution of molecular velocities for one type of molecule at a given temperature, rather than one distinct velocity. His theoretical derivation was later empirically verified in a number of famous experiments. His findings, referred to as Maxwell's Distribution Law, are summarized in the following equation: m
N v = 4"nN{
my 2
}a/2 v 2 exp{ . . . .
27tkT
2kT
}
dv,
(1.6.1)
where the terms are:
N Nvdv k m
= = = = =
the velocity of the molecules or atoms in meters/sec, the total number of molecules, the number of atoms found in the velocity interval v and v + dv, Boltzmann's constant (-1.38 x 10 -2 a joule/* K), the mass of the molecule in kg.
For example, assume we have a collection of 10 6 molecules of H 2 in some container at a temperature of 273* K, and that we are able to measure the velocity of each of these molecules in the container. If we then made a plot of the number of molecules as a function of their velocity, we would be able to construct a plot similar to that shown in Fig. 1.6.1. Z ~:
600
+
500
pa "~ ~ ~
I
,
~
,
f
,
~
,
f
,
Vp V avg.
400
Vrms
_
-
300 200
~
j_"tail"
100
>
0
Z
0
1
2
3
4
5
VELOCITY- k i l o m e t e r s / s e c . Figure
hydrogen
1.6.1.
Maxwellian Distribution
molecules
at t e m p e r a t u r e s
of v e l o c i t i e s
of 273
of 10 6
°K a n d 473
°K.
This figure shows two population vs. velocity curves. One is for when the container is at 273* K, the second is for when the container is at 473" K. We see that there is a
8 BASIC THEORY
distn'bution of molecular velocities for each temperature. These two curves are examples of Maxwell's Distribution Law. There are three special velocities that are of interest. One is called the rootmean-square velocity, Vrms. The average velocity, Vav... of the whole population is the second. The third is the most probable velocity, (i.e.r'~ a small interval of velocities where we are likely to find the greatest number of molecules), Vp. Using (1.6.1), one can solve for these three velocities.(4 ,s ) The results are as follows: Vrms ~
1.7 { -
kT
} 1/2,
(1.6.2)
m
Vavg ~ 0.98 Vrms, and,
Vp
~ 0.82 Vrms.
The relationship (kT / rn) ~ appears in each of the velocities of interest. This relationship, or a variation thereof, is used extensively in making vacuum calculations. In (1.6.2), m is the mass of the gas molecule, in kilograms. In making vacuum calculations it is more convenient to use M, the atomic or molecular weight of a gas, rather than the mass in kilograms of the gas particle. This is another area of possible confusion to the student. Chemists insist on using units of grams, where physicists use units of kilograms. However, it can be shown that the mass of a gas, in kilograms, is simply: m
= lO'a M / N O (kilograms).
(1.6.3)
We can now express Vrms in terms of the atomic or molecular weight of the gas species: Vrmscx (T / M) t /2.
(1.6.4)
We will frequently use this relationship in making calculations in the following sections. 1.7 V a p o r P r e s s u r e The concept of vapor pressure is frequently misunderstood. It is discussed here as it is an extremely important concept as it relates to cryopumps, an important class of capture pumps. We have thus far defined the meaning of pressure, density and temperature. Assume that we have a sealed box such as shown in Fig. 1.7.1, and that the box contains some element in a bulk (i.e., solid or liquid) state. Assume that the walls of the box are at a uniform temperature Tsl , and that this temperature has been held constant for an extended period of time. Above the bulk material in the box, some of the molecules or atoms of the bulk material will exist in a gaseous phase. These gaseous atoms will behave according to the ideal gas laws. The.~,,gaseous atoms will move about in the em,,pty part of the box with a Vrmscx (T s ~ / M)/~. Both the bulk atoms and the gaseous atoms will be at the same temperature Tsl. The gaseous atoms will collide with the walls of the box,
BASIC T H E O R Y 9 exerting pressure, and also return into the bulk material. At the same time, atoms within the bulk material with enough thermal energy will escape from the bulk material to exist for a while as gas above the bulk material. Ts /-T s
/-Ts
/
VlO I/P
o
o
o
o
o
SATURATION
ov
I~ A O
0
o
0
o
P
O
o
0
o
0
OoOoOoOoOoOoO~//]
~I/
r//~Oo PRESSURE 00~/ o[.,
[o SATURATION o I / I o ° PRESSURE ° o T H
ISOTHERMAL BOX A
o / o L
°
°
,OoOoo%,.OoOoO H i /////J(/.d
5
/////////A
//////////
O
1
V/////////
o
o
O ~O
00
o
o
O ~O
O
O
<]'A
,..,
,-..
,--,
,.-,
,.-,
....
.-.,
..
ISOTHERMAL BOX C
I S O T H E R M A L BOX B
Figure 1.7.1. Three boxes at i s o t h e r m a l t e m p e r a t u r e Ts • Boxes A & B have sufficient bulk m a t e r i a l to support a s a t u r a t i o n vapor pressure above the material, where this is not the case in box C. In time, an equilibrium state will be reached where the rate at which atoms leave the bulk material is the same as the rate at which other atoms impinge onto, and remain behind as part of the bulk material. The rate per cm2 at which particles pass some plane - real or imaginary - is referred to as a net flux of particles. When the net flux of particles at the surface is zero, the bulk material is said to be in equilibrium with the gas. The gas above the bulk is said to be saturated with atoms, because, for every atom that escapes the bulk material to become part of the gas, another atom leaves the gas to become part of the bulk material. Each element has a specific saturation pressure, Ps 1, for a given temperature, T s 1" Change the temperature of the box to Ts 2, and the pressure above the bulk material will eventually change to a predictable saturation pressure, Ps 2, uniquely characteristic of the element. Figure 1.7.2 represents the situations where there is sufficient bulk material, in three cases, but the temperature is changed in each of the three instances. In this case, the density or pressure of gas varies, as a function of temperature, over the bulk material. If we were to plot the pressure of the given element as a function of temperature, we would find a smooth, monotonic curve unique to that element. The saturation pressure of that element is sometimes called the saturation vapor pressure, or the equilibrium vapor pressure, or just the vapor pressure. /-Ts2
,Ts3 /
000;0500000001
)OO
/~o° SATURATION oO
Oq
o°o ' <
o SATURATION o~ ooq oo PRESSURE oo )oo oo( oooo p ~ o o o o o
ISOTHERMAL BOX B
ISOTHERMAL BOX C
"lo
~o°o /p
O
)oo
PRESSURE o O
p<,~o "'-
O
O
4 o?o?o o o?o °,
ISOTHERMAL BOX A
'
)OOOOO " ~ 0 0 0 0 0 ( OOOOOOOOOOOOO )OOOOOOOOOOOOO(
Figure 1.7.2. Three boxes at i s o t h e r m a l t e m p e r a t u r e s Tsl, Ts2 and Ts3, where Tsi < Ts2 < Ts3 and corresponding vapor pressures of the m a t e r i a l are Psi < Ps2 < Ps3.
10 BASIC THEORY Examples of such vapor pressure curves for the common gases are given in Figure 1.7.3.( 6 ) Some smoothing of the reference (6) data was done, where there were obvious errors in the tabulated data. There are similar curves for all of the elements. Note that the curves in these two figures don't seem to alter their smooth monotonic characteristics, when a material changes from a gas to liquid, or liquid to solid state. A summary of more current and precise vapor pressure data, reported by others for hydrogen, helium-3 and helium-4, is given by Haefer and shown in Fig. 1.7.4.(7 ) It is a common misconception that some mysterious process occurs at the point of phase change. This is not the case. We will revisit these two figures in considerable detail in some of the following chapters.
-3 10 -4 I0
o E-
l F-T-I
-5 10 -6
I0
--7
c~
o
I0
-8 I0
> -9 10
I0
-10
-11 10
10
20
40 TEMPERATURE
60 -
80
DEGREES
100
200
KELVIN
F i g u r e 1.7.3. V a p o r p r e s s u r e of s o m e of t h e common g a s e s as r e p o r t e d by H o n i g & H o o k . (6)
300
BASIC THEORY 11 10
10
10
~ 0
10
0
-1
-2
_3 ~
I 10
~-1
O a,
10
10
10
10
10
-4
-5
-6
-7
-8
-9 0.3
0.4
0.6
0.8
1.0 TEMPERATURE
2.0
4.0 -
DECREES
6.0
8.0
10.0
20.0
KELVIN
Figure 1 . 7 . 4 . Vapor p r e s s u r e of 3He, 4He, H2 a n d Ne as given by Haefer.(7) Surface Pumping To achieve the true vapor pressure in a bounded environment, it is required that there be enough bulk material to supply sufficient atoms to support a saturated gas state. If there is not sufficient bulk material, the atoms in the box sort of behave according to the ideal gas law (i.e., PIe" = n~ T). This is not completely correct as, when the atoms impinge on the walls of the box, they tend to stay on these walls for a short period of time. If many billions of atoms are impinging on the walls, even though these atoms may dwell on the walls for say only a few p sec., the net effect is that the pressure will be less than would be predicted by the ideal gas law. The walls of the box have sort of a pumping effect, and at any instant in time, there will be a large population of molecules resident on the walls. Assume that we squirt more of the same gas into Box B of Fig. 1.7.1. The pressure in the box will momentarily be higher. This means that the rate at which gas will impinge on the bulk (i.e., see (1.9.6)) will be greater than the rate at which it is leaving the bulk. Yet, gas will leave the bulk material at a rate sufficient only to support the saturation vapor pressure at that temperature. Therefore, if the rate of arrival of gas onto the bulk surface is greater than the rate of departure of gas from that surface, the surface is pumping gas. This type of pumping is called cryocondensation. If we stop squirting in gas, the pressure will eventually equilibrate at the vapor pressure of the gas.
12 BASIC THEORY
Pumping on a Liquid Cryogen We noted in Fig. 1,6.1 that there was a distribution of velocities of H2 molecules in the container, as defined by MaxweU's Distribution Law. The velocity of the molecule def'mes its kinetic energy. This implies that some of the molecules in the gas have much greater kinetic energies than others. Those out on the tail of the Maxwell distribution curve have much higher kinetic energy than the average population. There are other energy states possessed by the molecules besides their kinetic energy. These include their vibrational and rotational energy. There are similar energy distribution states for molecules and atoms resident in both solids and liquids.(8 ) Molecules or atoms resident out on the tail of the distribution function have the greatest probability of escaping the bulk material and going into the gas phase. Departure of the higher energy molecules from the bulk, in essence has the effect of ever so slightly reducing the temperature of the bulk material. For example, assume that by some process 10% of the molecules on the tail of the distribution function escaped the bulk material and became gas, and that this gas was removed from the container so that it could not return to the bulk. If we were to then construct a new distribution function for the balance of the bulk material, the average energy of the bulk material would be reduced because of the departure of the higher energy particles. A reduction of the average energy essentially implies a reduction in temperature of the bulk material. Liquid cryogens (i.e., very cold liquids) are commonly used in cryooumps. For example, liquid He is frequently used as a cryogen in special cryopumps. ('9) It is also used widely in superconducting applications. Helium liquefies at a temperature of 4.2* K. A process commonly used to reduce the temperature of liquid He below this 4.2* K temperature is to pump on the gaseous He which escapes from the liquid with some sort of mechanical vacuum pump. Pumping away the higher energy He atoms which escape from the liquid He has the effect of reducing the temperature of the liquid He.(1 0 ) The cooling effect of pumping on liquids can at times be troubling. For example, if there is an abundance of some type of material in your vacuum system which is saturated with water vapor (e.g., some sort of multi-layer insulating material), pumping on this material may cause the temperature to sufficiently decrease so as to cause the formation of ice in the material. The outgassing of the ice may be much slower than the outgassing of the room temperature water. This can lead to pump-down problems. This can be ameliorated by cycle-backfilling the system with a warm gas, or limiting the rate of system pumpdown.
1.8 Mean Free Path The mean free path of a particle in a medium is defined as: The average distance a particle will travel bz a medium before it suffers a collision with some otherparticle in that medium.
Ref'med calculations used to predict the behavior of gases in vacuum systems require that one be able to calculate the mean free path of gas molecules in the system. These calculations are used to determine the relative ease with which molecules will
BASIC T H E O R Y 13 be conveyed or conducted from one place to another in a vacuum system. They are therefore called conductance calculations, and will be discussed in Section 1.12. There are numerous other applications where the concept of a mean free path is of importance. For example, the solid state physicist is concerned about the mobility of electrons (and holes) in solids. The conductance (i.e., the inverse of resistance) of an electron in a solid is directly related to the number of collisions the electron suffers as it traverses the material under the influence of an electric field. The electron microscopist is concerned about gas scattering of the electron beam in the microscope. In this case the particle is an electron and the medium is the gas in the microscope. Physicists working with plasmas and particle accelerators also are on the long list of technologists concerned with mean free path concepts. Figure 1.8.1 depicts a chamber which is attached to a vacuum pump through an aperture. The chamber contains a large number of molecules which bounce off the walls, and which also collide with other gas molecules. If molecules bounce off the walls of the chamber many times prior to experiencing a collision with another gas molecule, then collisions with neighboring gas molecules will have little influence on whether or not the molecule eventuallyfinds the aperture and escapes the chamber. ~
I ~ ~_ ~¢~ I ¢~ / ] ~ ~ ~ ~ _~ ~. I I \ _~ ~- - ~ - - ~ r - - - ~ ~ I I/~r~ ' ~ - . ~ ~ - \1
~
VACUUM
F i g u r e 1.8.1. V a c u u m c h a m b e r in which the pressure is s u f f i c i e n t l y low w h e r e one molecule has little influence on the departure of o t h e r s t h r o u g h the aperture.
" ~ APERTURE
I
[~VACUUM
PUMP
With a very long mean free path, the probability of the molecules escaping the chamber will primarily be dependent on v, the surface collision rate, and the area of the aperture. The velocity of a molecule is proportional to (T / M) g . This means a molecule's opportunity to escape through the aperture diminishes with its molecular weight and increases with its temperature. If the mean free path is long compared to the dimensions of the chamber, the likelihood of a molecule escaping from the chamber is not influenced by its neighbors. In vacuum terminology, when the pressure in a vacuum system is such that the behavior of one gas molecule - with regard to escaping through an aperture or moving along a conduit - is essentially independent of the behavior of neighboring gas molecules, this pressure range or interval is called the molecular flow region. This is analogous to being on an elevator in New York City. If the elevator is very crowded, you may have to disembark (momentarily) so that those behind you may get off at their selected floor. On the other hand, if there are few people on the elevator, all may disembark at will without bumping into other occupants. It is reasonable to assume that the mean free path, ),, will be inversely propor-
14 BASIC T H E O R Y tional to the number of molecules per unit volume, N / V, and the cross-sectional area of these molecules, 7td 2 / 4, where d is the diameter of the molecule. The diameter of gas molecules differ from species to species. Therefore, )~ will differ with species and mixtures of different particles. The mean free path of some species in a medium of the same gas is given by the following equation:(a ) (1.8.1)
a = k T / [2 'A 7t P (da) 21,
where the subscripts "aMindicate the given species. In our calculations we will frequently pretend that there is such a thing as an air molecule. This term actually means that on the average the mixture of all gases found in the atmosphere at sea level behaves collectively in some manner. For example, the average molecular weight of air is M ,* 28.7 amu. The mean free path of an air molecule in air, at 20" C, is: 4.94 x 10 -a )" air =
Torr-cm.
P
(1.8.2)
The actual constituents making up air, at sea level, are important to know when using capture pumps. For example, certain noble gases in air are more difficult to sorption pump - a term which will be explained in Chap. 5 - than other gases. Also, the presence of argon in air can cause problems with certain types of sputter-ion pumps. These phenomena will be discussed later. Some of the air gases found at sea level, including average values for urban pollution, are given in Table 1.8.1.(1 1 )
Table
1.8.1.
Partial
pressure
of various
gases
at
ATOMIC MASS UNITS ( a m u ) *
SYMBOL
PARTIAL PRESS. ( T o r r )
Nitrogen
N2
5 . 9 3 x 10 +2
Oxygen
02
1 . 5 9 x 10 +2
32 40
GAS SPECIES
28
Ar
7 . 1 0 x 10 0
Carbon
dioxide
C02
5 . 0 4 x 10 -1
44
Carbon
monoxide
Argon
CO
3 . 8 4 x 10 - 2
28
Neon
Ne
1 . 3 8 x 10 - 2
20
Helium
He
3 . 9 8 x 10 - 3
4
Methane
CH4
1 . 2 2 x 10 - a
16
Kr
8 . 6 6 x 10 - 4
84
S02
7 . 6 0 x 10 . 4
64
He
3 . 8 0 x 10 - 4
2
Ne0
2 . 2 8 x 10 . 4
44
Xe
6 . 6 1 x 10 . 5
131
He0
variable
Krypton Sulfur
dioxide
Hydrogen Nitrous
oxide
Xenon Water
vapor
*Approximate
values
of t h e
most
abundant
sea
18 isotopes.(ll)
level.
BASIC T H E O R Y 15
1.9 Thermal Conductivity of Gases Mean free path considerations enter into an understanding of the thermal conductivity of gases. Also, an understanding of gas ionization processes, at reduced pressures, involve calculations of the mean free path of electrons in that gas. This will be discussed in the section on sputter-ion pumps. The thermal conductivity, K, of a gas or mixture of gases is given by:(1 2 ) K = (1/3) p
(1.9.1)
Vavg ), ¢ c v
= r/ E c v, where p is the gas density, c v the specific heat of the gas at constant volume. We have already defined ), and Vav . The term 17 is defined as the viscosity of the gas. Excluding the E term, (1.9.1) is ]erived from elementary kinetic theory. The term takes into account the rotational and vibrational energy states of various types of gas molecules, where 93"
=
-
5
4
,
(1.9.2)
and for monatomic gases 3' = 5/3, where for polyatomic gases 3, --1. The thermal conductivity of gases is particularly important in understanding the limitations of various forms of cryopumps. For example, most cryopumps have a f'mite refrigeration capacity. When pumping on a system with a cryopump, should the pressure exceed a certain value, it is possible to exceed this refrigeration capacity, and swamp the pump. Also, assuming that c v is about the same for air and H 2, what would be the relative thermal conductivities of the two gases at the same pressure? This has important implications in the starting of cryopumps. At low pressures, the thermal conductivities of all gases are directly proportional to the pressure, over a very wide pressure range (e.g., see Fig. 1.9.1). We note that at pressures less than ~ 10 -4 Torr, the thermal losses for the greater part are dominated by radiation between surfaces. For this reason, this represents the lower practical limit of thermocouple-type (TC) gauges, which depend on heat loss from an element as an inference of pressure. At pressures above a few Torr, the thermal conductivity becomes essentially independent of pressure. 0
"r~EE" IN CONDUCTIVITY CUR
_ Io-'
Figure
1 9.1 •
conductivity
Thermal
~
.
of
air
as
a f u n c t i o n of pressure.
_
-
~ ~~
t-
.~
I
.
_
- T H E ~ M , ~ CONDUCTr~TY 1 0 - z _ I S LINEAR IN PRESSURE
z~
~
-
.._--------__
\
-
10 -a
"N/ / /
_ -
10_ 4 _
.
/
~
-
/ NEAR LOWER LIMIT OF TC ~ -
f
GAUGE
_
_
-
i0-5
/
~.,"~ ii I 1 0 -4 10 -a
t ll l 10 -z
t II t 10 -I
PRESSURE
-
l~
i
10 Torr
o
Ill] ~ ~ll 10 l LO z
16 BASIC T H E O R Y Gas
Flux Incident
on a Surface
Using kinetic theory, it can be shown that the flux, v, on a surface, or through an imaginary plane in some vacuum vessel, is given by.'( 1 a ) v
=
N {
kT
}t/2.
(1.9.3)
27tin The only new symbol here is N, the density of particles per cubic meter. Assume we wish to determine the rate at which molecules impinge on the plane defined by the aperture in Fig. 1.8.1, with a nitrogen pressure, P, in the chamber. The density of molecules in the chamber is simply the total number of molecules in the chamber, divided by the total volume. Using (1.4.1d) and (1.4.3), we can express the density of molecules in the chamber, in terms of molecules per liter, as: N
=
PN o
(1.9.4)
~T
Substituting this into (1.9.3), and noting there are 10 a Z / m a , w e h a v e : PN o v
=
10 3
kT
(_____ } t / 2 .
~T
(1.9.5)
27tm
In (1.9.5) the flux density, v, is given in particles per square meter. With a little algebraic manipulation, and using (1.6.3), this equation can be put in a form expressing v, the flux density per cm 2, as a function of pressure and temperature: v
=
3.51 x 10 2 2 p / (MT) t / 2 .
(1.9.6)
Assuming the chamber is filled with N2, and the gas is at room temperature (i.e., 293" K), the rate at which N 2 molecules strike the wall of the box is simply: v = 3.88 × 10 2 o p molecules/cm2-sec.
(1.9.7)
Equation (1.9.7) is an extremely useful equation for developing a feel for the physics going on in your vacuum system. For example, assume that every molecule which strikes some special surface in your system sticks to that surface. In this case the molecule is said to have unity sticking coefficient on the given surface. Perhaps it is trapped by some chemical process occurring on the surface - this process is called chemisorption. Or, perhaps there is a cold surface in the system, so that when the molecule strikes the surface, sufficient energy is removed from the molecule so that weak attractive surface forces prevent it from escaping back into the vacuum system - this process is calledphysisorption. If one molecular layer on the special surface comprises --101 5 molecules/cm2 ,(x a ) how long would it take, at a pressure of 10-6 Torr, for one molecular layer to build up on the surface by either chemisorption or physisorption? The answer is, less than three seconds. This can become a serious problem in some applications. For example, assume that you were subliming a special semiconductor
BASIC THEORY 17 material, such as As, onto a substrate for the purpose of growing some solid state device - a process called molecular beam epitaxy (MBE).('i4) The material has to be sublimed at a fairly slow rate, say one A/see, so that it will settle in and become a well ordered structure. A molecule has dimensions of the order of 2-3 A. If the material being sublimed is chemically active with the gas in the system, and the pressure is 10-6 Torr, this implies that for every molecule sublimed, a chemically active gas molecule arrives on the surface and has an opportunity to form an undesirable chemical compound. Because of this problem of possible gas contamination of a deposited film, MBE is typically accomplished at pressures ranging from 10" 1 o to 10-9 Torr.
1.10 Pumping Speed, a Convenient Abstraction The concept of pumping speed is merely a mathematical abstraction, created by man as a convenient tool to describe the behavior of gas in a vacuum system. Pump speed is defined as a volumetric displacement rate. Note that this definition implies nothing about pressure, mass flow, gas density or anything that has to do with gas molecules. The symbol "S" is usually used in the literature to represent pump speed. To best understand the concept of speed, refer to Fig. 1.10.1. Assume the following:
1) 2) 3)
4)
5)
The piston shown in Fig. 1.10.1 forms a perfect gas seal with the cylinder within which it is contained. There is no outgassing from the components. The piston is attached to a crankshaft which can be rotated either manually or at some fixed rate. The volume above the piston is one liter when it is fully withdrawn from the cylinder, and zero liters when it is fully extended. There are two valves located in the piston head. One valve, V1, is used to introduce gas into the piston head to some desired pressure. The second valve, V~, automatically opens as the piston approaches the top of the cylinder - venting gas and closing the instant the piston reaches top dead center.
'~ . I -
F i g u r e 1.10.1. Mechanical model of a p e r f e c t v a c u u m p u m p w h e r e s p e e d is m e r e l y A V/At.
EXHAUST VALVE
PERFECT VACUUM SEAL
ROTATING CRANKSHAFT PISTON
18 BASIC T H E O R Y If we manually turn the crankshaft one full cycle, with all the above assumptions, there would be no gas contained above the cylinder head at the end of a cycle (i.e., when the piston is again fully withdrawn). The amount of gas which was vented through V 2 at the top of the cycle would simply be P × V, where P is the pressure above the piston prior to the start of the cycle, and V is one liter, the volume of the cylinder head when the piston is fully withdrawn. The next time we crank the piston a complete cycle, the net amount of gas vented out of V2 will be zero, as none was contained in the one-liter volume at the start of the cycle. We may repeat the process, but, as long as V 1 is closed, no gas is vented out of V 2 in subsequent cycles. Assume now that we turn on a motor which is attached to the crankshaft, and that the shaft rotates at one revolution each second. Clearly, the piston under this condition sweeps out the one liter volume of the cylinder every second. We say that the piston has a volumetric displacement rate of one liter per second. The rate the volume is swept out, S, is 1 Z/sec. This is the speed of the pump. I have described a perfect pump which has a speed of one liter per second. Every second (At = 1 sec.) a volume (A V = 1 liter) is swept out of the cylinder head. Or, S =
AV At
-
one liter
(1.10.1)
one second
The above equation implies nothing about the flow of a gas. It is merely a statement of the rate at which the piston is sweeping out the volume; the volumetric displacement rate. If the rpm of the motor could be varied to say two revolutions per second, the volumetric displacement rate, or speed, would be: lzl S =
dV dt
=
two liters one second
.
(1.10.2)
This mechanical model, defining pump speed, is easy to visualize in applications involving mechanical pumps. These pumps either have pistons or vanes which sweep out volumes at a given rate. However, the model is totally applicable to all forms of vacuum pumps. We must stretch our imaginations a tittle when visualizing a mechanical model in the ease of, for example, cryopumps or sputter-ion pumps. Capture pumps have no moving parts within the vacuum. How then does a volumetric displacement rate model apply? It applies only as a convenient tool for making calculations; it is an abstraction, rather than as a physical reality. However, I have found it convenient to picture capture pumps, in my mind's eye as consuming abstract or elemental volumes at some specified rate. For example, assume the pump in Fig. 1.10.2 has a specified speed of 1000 Z/second. Picture this pump as consuming elemental volumes at a rate of 1000 Z/second, at the input flange of the pump. These imaginary volumes, as in the case of the mechanical pump model, are consumed by the pump at unspecified pressure. The pump may be a capture pump, turbomolecular pump, etc. This same abstraction applies to all forms of vacuum pumps. We will return to the concept of pumping speed after first learning about the concept of throughput.
BASIC T H E O R Y 19 VACUUM CHAMBER ABSTRACT VOLUMES
PUMP INLET VACUUM PUMP F i g u r e 1.10.2. A b s t r a c t v o l u m e s a t u n s p e c i f i e d p r e s s u r e b e i n g " c o n s u m e d " by t h e p u m p .
1.11 Throughput Throughput is defined as a gas flow rate (d(PI0/dt). For example, it might represent: 1) the rate at which gas is leaking into a vacuum system; 2) the rate at which molecules are outgassing from the walls of the system; 4) outgassing from o-rings our other system elastomers; or, 4) the rate at which combinations of several sources of gas travel into a vacuum pump. The symbol "Q" is frequently used in the literature to represent throughput. The units of throughput which we will use are Torr-Z/second (i.e., P x V / t ) . We note that the temperature, T, is not specified, and remember from (1.4.3) that knowing T is also required to be specific about the number of gas molecules involved. When data involving throughput are given in the literature, it is usually assumed that P V is specified at room temperature (i.e., 293* K), unless otherwise noted. Returning to the mechanical pump model of Fig. 1.10.1, assume now that the motor is disconnected, and the piston withdrawn so that the volume above the piston is one liter. Assume that this volume is pressurized to 10-s Torr, by introducing gas through V 1. After a pressure of 10-3 Torr is achieved in the cylinder head, V 1 is closed. We assume in our model that the gas is neither cooled on expansion into the cylinder head, nor heated during the compression cycle (i.e., it remains at 293* K). If, while invoking the five assumptions of the previous section, we manually "cranked" the pump through one complete cycle, 10-3 Torr-Z of gas would be exhausted through V2 at the top of the cycle. This is merely P x V vented through V2 during some unspecified time interval. If we now turn on the motor and repeat the process once a second, the rate at which gas is exhausted out of V2 (and introduced into the system through V1 ) would be P x A V / A t. Or,
Q=Px------r~ Q
AV At
=
P×
S
=
10-3 Torr-Z/sec.
(1.11.1)
20 BASIC T H E O R Y If we increase the pressure to 10-2 Torr between each cycle, the throughput, Q, would increase to 10" 2 Torr-Z/sec., but the pump speed would remain the same, etc. The important thrust of this is that, for a constant pumping speed, throughput varies linearly with pressure. This was the prime motivator in creating the concept of pumping speed as a constant volumetric displacement or consumption rate.
1.12 Conductance, Another Convenient Abstraction Conductance is another mathematical model or tool, devised to be able to describe the behavior of gas within the geometrical confines of a vacuum system. We were introduced to the concept of imaginary volumes entering the flange of a pump. The concept of conductance is similar. That is: Conductance is a measure of the ease with which elemental, abstract volumes, at unspecified pressure, pass from one place to another place in a vacuum system, per unit of time.
Remember, this is the definition of a model, rather than a physical reality. With this model, we pretend that little abstract volumes at zero pressure bounce around inside the constraints of a vacuum system. Mathematical equations comprising the model let us make calculations of the ease with which these volumes travel from place to place in the system. Then, we assume the little volumes are filled with gas at a specified pressure, and make a statement about the transport of the gas in the system. For example, assume that the vacuum manifold in Fig. 1.12.1a is separated by a thin foil with an aperture in the center. Assume that d, the diameter of the aperture, is much less than D, the diameter of the manifolds. For reasons that will later be explained, assume that D >> ),g, the mean free path of gas species g in the vacuum system. For the moment, disregard the fact that there is gas in the system. A conductance model enables us to calculate the rate at which abstract volumes pass from Manifold A, through the aperture of conductance C a, and into Manifold B. It is reasonable to assume that the rate at which these elemental volumes accomplish this passage will be directly proportional to the area of the aperture. That is, in Fig. 1.12.1a, Td 2 C a (A -* B) oc
~/sec. 4
MANIFOLD A
MANIFOLDB
,,i\
/\
AV At
OF CONDUCTANCE Ca
Figure 1.12.1a. The conductance of the foil aperture is the measure of the rate it will pass abstract volumes from Manifold A to Manifold B, at unspecified pressure.
BASIC THEORY 21 Also, in Fig. 1.12.1b, ~d 2
Ca(B-* A) oc
Z/sec.
MANIFOLD A
AV
MANIFOLD B
1
/\
/1\
4g>-~
.z\
Y
Z_ THIN FOIL APERTURE OF CONDUCTANCE Ca F i g u r e 1.12. lb. T h e foil a p e r t u r e h a s t h e i d e n t i c a l c o n d u c t a n c e f o r t h e p a s s a g e of a b s t r a c t v o l u m e s , a t u n s p e c i f i e d p r e s s u r e , f r o m M a n i f o l d B to M a n i f o l d A.
Assume now that Manifold A is at a pressure PA, and Manifold B is at a pressure PB" As shown in Fig. 1.12.1c, we then pretend that the elemental volumes escaping from Manifold A to Manifold B retain the pressure of Manifold A when arriving at Manifold B, and vice versa. Therefore, there is transport of gas in these elemental volumes in both directions. MANIFOLD A AT PRESSURE PA
Av A--t -XP
~
MANIFOLD B AT PRESSURE P B
©
e
//
\/
Av
~ \
~
XPA
O
/
~ T H I N FOIL APERTURE OF CONDUCTANCE C a F i g u r e 1.12.1c. A " m o d e l " f o r t h e n e t p a s s a g e of g a s t h r o u g h a n a p e r t u r e c o n d u e t a n c e s e p a r a t i n g m a n i f o l d s at d i f f e r e n t p r e s s u r e s .
The net flow of gas in the system, Qnet, is: anet cx
[ QA
]
-
[ QB],
71;d 2 ['
4
7Id 2 PA]
-
[
{PA
-
4
PB],
717d2
Qnet cx
[
4
]
PB}"
(1.12.1)
22 BASIC THEORY
1.12.1 Conductance Model Applications Species and Temperature Dependence We have used the above conductance model to predict the flow of gas through an aperture in a vacuum system. However, we have not specified the type of ~as. We know from (1.6.4) that the Vrms of any gas species is proportional to (T / M) 7" of that species. We know, for example, that for a given T, He molecules move about in a vacuum system much faster than do N2 molecules. Common sense tells us that an H2 molecule would more readily escape from Manifold A to Manifold B, than would an N2 molecule. Therefore, to predict the exact amount of gas transported between the two manifolds, we must take into the account the gas species and temperature. This is a simple process, once the exact conductance equations (i.e., models) are known for one gas species.
Pressure Dependence of Conductance In the example given in (1.12.1), we imposed the restriction that ~Xg>>D >> d. That is, the mean free path of one gas species g, in the gas of all other speoes present, was so great compared to the aperture diameter, d, that the presence of other gas species had little influence on the former species g, passing through the aperture. When the mean free path in a gas is large compared to the dimensions of the system, we call this the molecular flow region. In the molecular flow region, a molecule travels from one place to another in the system in a totally random manner. It bangs into the walls of the system far more frequently than it collides with neighboring molecules. Therefore, the configuration of the system is the only influence on the molecule in its sojourn within the system. It is by accident that the molecule finds a pump appending the system and is removed from the system. Providing a larger conductance between the source of the gas and the pump increases the probability that the molecule, in its random sojourn about the system, will find the pump. The pressure difference across the conductance is an effect of gas being captured at the pump, rather than a cause forcing gas into the bowels of the pump. However, if a gas molecule suffers frequent collisions with neighbors - in fact, more frequent than collisions with surfaces in the system - then its neighbors begin to have a significant influence on the molecule's travels within the system. Should the frequency of collisions with neighbors be high, this implies that the mean free path of the molecule in the system is short. There are equations similar to (1 12 1) for conditions where ?~.. is comparable to or much less than the dimensions of the system. In these circumstances, the conductance expressions are no longer independent of pressure. As a consequence, the expression for the throughput becomes nonlinear in pressure. That is: .
Q = C(P,d) [PA -
PBI,
.
(1.12.2)
where the expression C(P¢I) means C is a function of P and d. When the mean free path of gases in the system is comparable to or less than dimensions of the vacuum system, these nonlinear equations in P apply, and the calculations of system Q become messy. The pressure realms where these nonlinear equations apply are
BASIC THEORY 23 coined viscous and turbulent flow regimes. Capture pumps seldom operate at pressures where these special calculations are required. However, if constructing a large cryopumped space simulation chamber, where the initial pumpdown or roughing time is an important consideration, such calculations would be used. Also, for example, if one is manufacturing and selling large quantities of leak detectors, costs of vacuum manifolding might play a significant role in the competitiveness of your company. Formulas for making viscous and turbulent flow calculations are given in references (~), (s)and (4 2 ).
A rule of thumb is that these nonlinear (i.e., in pressure) equations will always yield conductances (i.e., C(P,d)) which are greater than will be derived using the molecular flow equations. Therefore, using the molecular flow calculations will yield conservative results in roughing times.
System Geometry Dependence for Molecular Flow The conductance of an aperture in the molecular flow region, and for air at 293" K, is given by the following equation: n
Ca, air = 11.6A Z/sec,
(1.12.3)
where A = the aperture area in cm2. The conductance of a round manifold in the molecular flow region, and for air at 293" K, is given by the following equation: D 3
n Cm,ai r = 12.1----- Z/sec, L where D and L
(1.12.4)
= the diameter of the tube in cm, = the length of the tube in cm.
Equation (1.12.4) is sometimes called the long tube formula as it applies to tubes where L / D e 10. However, even for a value of L / D = 10, there is a 16% error in the value of (1.12.4) over that calculated by Clausing.( 4 s ) Because of this he created what is now called the Clausing factor, o~, which, when multiplied with the long tube formula, corrected it for all values of L / D to within 1%. However, using tabulated o~ values is inconvenient, and an equation equivalent to (1.12.4), but which compensates for short tube effects, is somewhat awkward to manipulate. A modified conductance formula for tubes of all lengths is:(4 s ) Ca
Cm, air =
(1 + 0.80 x / D)
(1.12.4a)
where C a is the conductance of an aperture of diameter D, and x is the length of the tube. For values 0.05 < L / D < 600, (1.12.4a) is in agreement with Clausing to within 8%. At the risk of displeasing the more rigorous technologists, I offer you some advice to follow when making UHV vacuum calculations. Outgassing rates in vacuum systems may vary, in time, as much as five or six orders in magnitude. Because of this, though you may accurately know the pump speed, and have pre-
24 BASIC THEORY cisely calculated the various conductances of the system leading to the pump, you will do well to predict system pressures to within ×2. In such instances, calculating conductances to three decimal-place accuracy is a waste of time. There are often approximations which afford sufficient accuracy to adequately predict the behavior of gases in your system. For example, the equation for calculating the conductance of a long rectangular tube, for air at room temperature, is as follows: (2,3)
a~ b2 Cn = 30.9 where
a
and, if then
b L a/b K
L(a +b)
K
Z/sec,
(1.12.5)
= rectangular tube width in cm, = rectangular tube height in cm, = length of rectangular tube in cm, = 0.1 0.2 0.4 0.6 0.8 1.0 = 1.44 1.29 1.17 1.13 1.12 1.10, respectively.
Equation (1.12.5) is an exact expression for the conductance of a rectangular tube. However, in one of the problems at the end of this chapter, you will be asked to compare the conductance for air in a round tube (i.e., using (1.12.4)), with diameter D and length L, with that of a rectangular tube of dimensions a, b and the same L, where: a × b =
7t D 2 . . 4
(1.12.6)
From this, you will find that solving for the equivalent D of (1.12.6) and using this D in (1.12.4), which you memorized, will usually provide more than sufficient accuracy in your calculations. M o l e c u l a r C o n d u c t a n c e for D i f f e r e n t G a s e s
In the previous sections it was emphasized that Cai r (x (T / Mair) ~ , and we may use (1.12.3) and (1.12.4), equations applicable to an air conductance, to derive the value of a conductance for other gases. Noting the constant 11.6 in (1.12.3), we must conclude that the expression (T / Mair) ~ went into deriving the constant. That is:
11.6 = k (T/Mair) x/2,
(1.12.7)
where k is a constant. If we wish to derive the conductance of an aperture for some other gas with molecular weight M,,, we need only multiply 11.6 (i.e., (1.12.3)) by the value (Mair) g and divide the sam~e expression by the value (Mg) ~ . Refer to the atomic mass units of Table 1.8.1 for values of Mg for the different gases. Also, we remember from Section 1.8 that the value of Mai r ts --28.7 amu. Let's work out a couple of examples where we modify (1.12.3) to compensate for a different gas at a different temperature. Assume that we have calculated the conductance of an aperture for air, and we want to know the value of this conductance for helium. The atomic weight of helium, MHe, is 4 amu. Therefore,
BASIC T H E O R Y 25 CHe
= {Cai r} x
[.
Mair]
x/ 2
(1.12.8)
MHe = {ll.6A}x
[
28.7 amu 4 amu
11/2
This conductance for He is the value at a temperature of 293* K. We may calculate the conductance of the aperture for He at, say, 400* K as follows: CHe,400* K = {CHe,293" K } x
400*K [ ~ 1 293* K
1/ 2 .
(1.12.9)
1.13 Voltage, Current and Impedance Analogies In the molecular flow region (i.e., S and Q are linear in P), there is a one-to-one and onto correspondence - mathematicians call this an isomorphism - between vacuum conductance, pumping speed and throughput, and the electrical concepts (i.e., linear circuit theory) of electrical conductance, voltage and current, respectively. This is not coincidental, for similar models were created much earlier to describe the flow of charge in electrical circuits. The correspondence (*--*) is as follows, where the subscripts v are for vacuum values, and the subscripts e are electrical values: C v *--* G e = 1/R (electrical conductance of a component), Sv~-. Ge Pv *'" Ve (voltage across a linear component), Qv *"* Ie (current, or charge flow rate).
This correspondence between linear vacuum circuits (i.e., in the molecular flow region) and linear electrical circuits provides an aid in expanding use of the vacuum S, C and Q models using elementary circuit theory.(1 s ) The definition of throughput, with constant T, was: Q
=
d(PIO
dt
dV = P---dt
+ V
dP
dt
.
(1.13.1)
Assuming steady state conditions (i.e., dP / dt = 0), yields:
or,
Q
= PxS,
(1.13.2)
Q
dn = BT~----. dt
(1.13.3)
The electrical equivalence of (1.13.2) is merely: I
-
VeX
o
-
VeX
lm,
(1.13.4)
26 BASIC T H E O R Y
I = dq/dt q.. ] Ve or,
G=I/R
Q=SP I
=GV
T
e
Figure 1.13.1. Electrical correspondence between V & S, G & C and I & Q. where current,/, or charge flow (i.e., dq/dt) is analogous to the flow of molecules, dn/dt, in a system. An electrical circuit which represents (1.13.4) is given in Fig. 1.13.1. Were the circuit to have a second resistor, as in Fig. 1.13.2, we can combine the resistors into the reciprocal of one equivalent resistor with electrical conductance Ge: Re
=
1
Or,
Rt 1
=
Ge
And,
+
R2, 1
+
Gt
G2
G t G2 Ge
=
.
(1.13.5)
G, +G2
The current,/, in the circuit in Fig. 1.13.2 is:
G1 G2 I = Ve
• Gt
~ R 1 'VVV~ I = dq/dt +
R2
Ve
T
+G2
Figure 1.13,2. The correspondence between eleetrical and vacuum conduetances,
(1.13.6)
The corresponding vacuum circuit is shown in Fig. 1.13.3, where the vacuum conductance of a manifold leading to a pump, C m , is combined with the pump speed, Sp, to yield an equivalent pump speed, S c, at the chamber: 1
1
Sc
sp
Sc =
VACUUM CHAMBER AT
1
WHICH
Cm
SpCm
+Cm
~ ~
~ ~
SPEED
IS S c
MANIFOLDOF C0NDTANCE Cm
(1.13.7) P U M P WITH SPEED S p
Figure 1.13.3. Speed produced at vacuum chamber for a given Sp and Cm.
BASIC T H E O R Y 27 Knowing the pressure in the chamber, Pc, we may calculate the throughput (e.g., due to outgassing) from the chamber, Qc, accordingly:
ac = PcSc = Pc
SpCm
Sp + Cm"
(1.13.8)
Ka'rchhoff's first rule in electrical circuits originates from the assumption that charge is neither created nor destroyed in a circuit. There is an analogy in vacuum circuits, where we assume that mass (i.e., PV) is preserved. For example, if we assume that the only source of gas in the system shown in Fig. 1.13.3 originates in the chamber, then the rate at which gas is leaving the chamber, Qc, must be the same rate at which gas is entering the throat of the pump, Qp. Therefore,
Qc
=
1:I Pc Sc =
Qp, Pp Sp.
(1.13.9)
We will later use (1.13.9) to diagnose possible problems in a vacuum system. Also, in the next section, you will learn how to take into account the added effect of an outgassing manifold between a chamber and pump. However, let's first pursue the electrical and vacuum analogies a little further. The similarity between (1.13.6) and (1.13.8) is obvious. Also, we have used simple circuit theory to derive (1.13.8). Equation (1.13.8) is not totally correct as we have not taken into account conductance of the aperture leading into the pipe from the chamber. This aperture conductance, Ca, is combined with the conductance of the manifold, Cm, to yield a total conductance, Ct, of:
Ct "1
= Ca" l + Cm" l .
(1.13.9a)
In (1.13.9a), conductances in series are combined as reciprocals to yield a total conductance, Ct: Ct
=
C. Cm q+Cm
.
(1.13.9b)
Any number of these series vacuum conductances may be combined as reciprocals, as in (1.13.9a), to yield a total equivalent conductance, Ct, between the chamber and pump. Parallel conductances are combined by simple addition of the various component conductances. The total conductance, Ct, for air may be calculated by using (1.12.3) and (1.12.4) to determine Ca and Cm, respectively. Then, the total speed delivered to the chamber, Sc, is given by:
Sc
=
SpCt . Sp + Ct
(1.13.10)
If conductances are required for other gases, or air at some other temperature, then equations equivalent to (1.12.8) and (1.12.9) may be used to modify Ct, the total conductance for air. It is not necessary to calculate each of the conductances for the
28 BASIC THEORY new gas or temperature, and then combine the results. Voltage analogues of vacuum systems can be used to model very complex vacuum systems. For example, such an analogue constructed to model the 30 100-meter sectors of the Stanford 2-mile Linear Accelerator.( 4 6) The model proved most useful in diagnosing system problems, including identifying the outgassing of certain valve o-rings seals due to loss-tangent heating, and locating hard-tof'md leaks in the linearly distributed vacuum system. Similarly, we modeled the 12 480 meter cryostats of Brookhavens Relativistic Heavy Ion Collider.( 4 7 ) Each cryostat contains numerous superconducting magnets. These magnets are used to steer the 3.5 km circular orbits of counter-rotating particle beams. By imposing gas impedances (i.e., contrived conductances) within the cryostats and selectively pumping on the cryostats with a few turbomolecular pumps, we were able create helium pressure gradients stemming from leaks. In fact, we were able to meet the objective of locating helium leaks in the cryostats to within one magnet interconnect.(4 8 ) 1.14 Dalton's Law and Linear Superposition
Dalton's Law states that the partial pressures of gases in a mixture behave individually according to the ideal gas laws. For example, assume a vessel of volume V1 contains partial pressures of argon and helium of pressures PA and PHe, respectively. Assume a second vessel of volume V~ is isolated from the first by means of a valve, and that it contains no gas (this is sometimes referred as being under vacuum). The total initial pressure in Vt is simply Pi,A + Pi,He" Were we to open the valve between the two vessels, the amount of argon would remain the same (i.e., n~ T before and after). The argon in Vt would expand into V~, and the final argon pressure, Pf, A, according to (1.4.1b) would be: Vt Pf, A = Pi,A
V1
Similarly, Pf, He = Pi,He
+ V2
V1 V1 + V2
The final pressure is merely the sum of the individual partial pressures of helium and argon (i.e., their linear superposition). Linear superposition theory applies to both electrical and vacuum circuits. For example, assume there is a time varying voltage source, Ve t (t), in the circuit in Fig. 1.14.1a. The current in the circuit, It (t), will vary as shown in Fig. 1.14.1b. Or, I1 (t) = G x V e x (t).
(1.14.1)
I t ( t ) = dq/dt
Amps z
G=I/R
V TIME
Figure 1.14.1 a.
Figure 1.14. !b.
t
see
BASIC T H E O R Y 29 Similarly, we may solve for current, 12 (t), as a function of a second voltage source, V e 2 (t), in the circuit of Fig. 1.14.2a (i.e., the same conductance)" 12 (t) = G x Ve2 (t). I2(t)
=
(1.14.2)
Amps
dq/dt
I
I
I
1
I
[-, Z
1/
~ --
I
--
Figure
t
see
TIME
1.14.2a.
Figure
1.14.2b.
Were we to install both voltage sources in the same linear circuit, the principle of linear superposition implies that all we must do to find the total resultant current, It(O, is add each of the individual current components given by (1.14.1) and (1.14.2). That is: It(t )
=
Iz =
(t)
+ I2(t)
o x
(t)
(1.14.3) +
o x v
(t).
Or, It(t ) = G x [Vel(t) + Ve2(t)].
(1.14.4)
This total resultant current as a function of time, It(t), is depicted in Fig. 1.14.3b. Note that by the manipulation of (1.14.4), we may express the total voltage imposed on the circuits as the algebraic sum of Ve 1 and Ve 2. It(t)
=
Amps
dq/dt + b-., z
G=I/
L) i.
Figure
.-
t
see
TIME
1.14.3a.
Figure 1.14.3b. The correspondence in a linear vacuum circuit might relate to the following cases. Case #1: There are several sources of different gases, but each gas originates from one location in the vacuum system (e.g., from within a chamber such as shown in Fig. 1.13.3). Case #2: There may be several sources of the same type of gas, but from different locations in the system. Case #3: Combinations of Cases #1 and #2. The first two cases will be discussed. Case #3 is part of the problem set. CASE #1, Different Gases and Sources: Assume that there is an air leak, Q air, in a flange seal appending the vacuum chamber shown in Fig. 1.13.3. Also, assume that some source within the chamber has an H2 outgassing rate, QH 2" What is the total chamber pressure Pc and Pp, the pump pressure? Assuming we know the pump speed for these two gases, what are the partial pressures of each of the gas species at both chamber and pump? Let's first
30 BASIC THEORY solve the problem for the air leak. Then we will solve the problem for the H ~ outgassing. Using (1.12.3), (1.12.4) and (1.13.9b), we can calculate the total conductance for air, C t air, from chamber to pump. Knowing this conductance, we can then use (1.13.10) to 'calculate the equivalent speed for air, S c air, delivered to the chamber. We can then calculate the partial pressure of air in t~e chamber, Pc, air accordingly: Qair
= Pc, air Sc, air
Qair
= Pc, air
Sp'air Ct'air
(1.14.5)
Sp,ai r + Ct,ai r
The speed of all pumps varies with the type of gas being pumped. Therefore, (1.14.5) specifically refers to both the pump speed and total conductance for air. A similar equation may be developed for the hydrogen outgassing in the chamber: Qc, H2
= Pc,H2 Sc,H2
Q c,H 2
= Pc,H 2
Sp,H 2 Ct,H 2
(1.14.6)
Sp,H 2 + Ct,H 2
Because the gas is now hydrogen rather than air, we compensated for both the conductance and pumping speed of H 2 in (1.12.6). We can solve for Ptotal by the algebraic manipulation of (1.14.5) and (1.14.6) so as to add the partial pressures of the two gases (i • e," P_c , ,r a,2 and P c , a l r• ) • We did the same to determine the sum of V~ 1 and Ve2 in (1.14.4). In this case, the total pressure in the chamber (analogous to total voltage) is simply: Pc, total = Pc, air + Pc, H2 =
Qc, air
+
Sc, air
Qc,H~ ~ . Sc,H2
(1.14.7)
Using the conservation of mass law, it is a simple matter to calculate the partial pressures of air and hydrogen at the pump. That is: Qc, air Pc, air x Sc, air Also,
Qc, H 2
= Qp,air = Pp,air x Sp,ai r.
(1.14.8)
= Qp,H = Pp,H2 XSp, H2.
(1.14.9)
The manipulation of (1.14.8) and (1.14.9) enables one to determine the total pressure at the pump. Ppump
=
Pp,air
+
Pp,H2
BASIC THEORY 31
--
Qair
~
QH~
+
Sp,air
Sp,H2
CASE #2, Outgassing Manifold and Chamber: It can be shown, using slightly more advanced linear circuit theory,(1 6 ) that the pressure as a function of length along a long, uniformly outgassing cylindrical manifold, such as shown in Fig. 1.14.4, is given by: P(x)
where
7tq
= Pp -t
2kD 2
Pp
= = = = =
(1.14.10)
pressure at the pump, distance from the pump in cm, diameter of the manifold in cm, length of the manifold in cm, outgassing rate Torr-Z/sec-cm 2 = 12.1 (i.e., f((T / M) g) in (1.12.4)).
X
D q k
and
[2x,~ - x2], the the the the the
p,
x)
10 -
0
8
P(t)
-~
-
~ 4 (0)
LONG 0UTGASSING MANIFOLD
Q z
0
o
1.14.4.
Pressure
1,
t/z
I
,] x
atl4
t
DISTANCE ALONG 0UTGASSING MANIFOLD OF LENGTH 1.
PUMP Figure
1
t/4
profile
in
a long,
uniformly
outgassing
manifold.
Returning to Fig. 1.13.3, assume that the total chamber outgassing is Qc and that the outgassing rate of the manifold between chamber and pump is q. Assume that the gas, in both cases, is air. The total outgassing rate of the manifold, Qm, is merely the inner area of the manifold, A, multiplied by q, or: Om
=
[A]q
Qm = [ T t x D x 2
]q.
Therefore, the pump pressure is: Pp = [ a c + a m ] / S p . The pressure in the chamber due to outgassing of the manifold, Pm(2 ), is found by calculating the pressure at x = 2 in (1.14.10).
32 BASIC T H E O R Y 7tq
Pm(l) = Pp + ..
(1.14.11)
[~1 2kD 2
The pressure in the chamber due to outgassing in the chamber, Qc, is: Pc
SpCt
= Qc
Sp + C t
•
(1.14.12)
The total pressure in the chamber due to both chamber and manifold outgassing, Pt, is just the linear superposition of (1.14.11) and (1.14.12). Or, Pt = Pm('~) + Pc" In the case of particle storage rings, the majority of the vacuum system may comprise long beam tubes. If the pressure in the beam tube is excessive, this can cause beamemittance growth. In such cases, the average pressure (and species) within the beam pipe becomes important. The average pressure in the beam pipe may be found by integrating (1.14.10) with respect to x, and dividing by .~, where in this case .~ is the distance half way between pumps. Or, 1
Pa
= = Pp +
of~P(x) dx 2
3
2
2
[Ttq,~ /2kD ]
(1.14.13)
1.15 Selective and Variable Pumping Speed It is noted in (1.14.5) and (1.14.6) that pump speeds vary depending on the type of gas. This is true for all pumps. Much more will later be said about this as it relates to specific types of capture pumps. At high pressures, the speed of all high vacuum pumps (i.e., capture pumps, momentum transfer pumps, etc.) deteriorates for reasons unique to the type of pump. For example, the oil in the jets of a diffusion pump stack may start to randomize; the rotor on a turbomolecular pump may start to slow down; the thermal capacity of a cryopump may be exceeded, etc. The mechanical model of pump speed stressed that the speed of a pump was treated as being independent of the pump pressure. However, we know that all pumps have both high and low applicable pressure limits. For example, even the best mechanical pump has a base or blank-off pressure of ---10-4 Torr. The base or blank-off pressure of a pump is the minimum pressure which can be achieved when the pump is valved off at the input port. All pumps have a minimum blank-off pressure (irrespective of many of the published data sheets). Also, this blank-off pressure may vary with the amount and type of gas which has been pumped. Because of this, we must modify our mechanical model of pump speed to compensate for these effects. The high, operating pressure limitations of different capture pumps will be treated in their respective sections. The low pressure limitation of any pump is a bit simpler to model.
BASIC T H E O R Y 33 Assume that you have measured the blank-off pressure of some pump. The implication of a blank-off pressure, P o , is that the usable speed at this pressure is zero. Speed of the pump as a function of P, the operating pressure, is given by the following equation: P®
S = S m xa { 1 ......
},
P
(1.15.1)
where Sma x is the maximum inherent speed of the pump. This maximum speed limitation primarily stems from geometrical or mechanical limitations of the pump. However, it will also depend on the condition of the pump. With capture pumps, Po is critically dependent on the amount and type of gas which has been pumped. However, (1.15.1), an example of which is given in Fig. 1.15.1, is valid for all types of vacuum pumps. e) ff~
[.-,
Sm
I ua r.ra
°
=-~
0
I 10 n
p~
/L
) = Sm~
1 -
1 1 10 n+l
I I I] L 10 n+Z
[ 1 1 t, I , ] I I p 10 n+4 10 n+3
P
1
PUMP PRESSURE- TORR Figure 1.15.1. Pump speed as a function of base pressure.
1.16 Measuring Pump Speed Perhaps a half dozen methods exist for determining the speed of a pump. Four of these methods are now common practice in industry, and are reported in the literature. They include 1) Rate of Pumpdown; 2) Single Dome; 3) Three Gauge Dome; and, 4) Fischer-Mommsen Dome methods. Each method has certain advantages and limitations. All of these methods of measuring pump speed require the use of calibrated vacuum gauges for the applicable test gas.
1.16.1 Rate of Pumpdown By monitoring the rate a pump evacuates a vessel (sometimes called the pumpdown rate), we attempt to deduce the speed of the pump in question. Assume a vessel of known volume V is attached to a pump with a valve and manifolding having a total conductance C t. Assume that there is a gauge attached to the vessel, and the indicated pressure of this gauge is P(t). The equivalent speed delivered to the vessel is Seq .. Figure 1.16.1 represents such a system.
34 BASIC T H E O R Y t)
~ I
I
1
I
I
I
I
I
I
VACUUM VESSEL OF VOLUME V
E OPENED
VACUUM
GAUGE
l
PUMPDOWN
d
~
CURVE"
-
P
1
o
1
l
t
PUMPDONN TIME
Figure which
1.16.1. Determining pressure decreases
t
~ M A N I F O L D & VALVE OF CONDUCTANCE Ct
/
A C U U M P U M P OF SPEED Sp << Ct.
pump speed in a vacuum
by measuring the rate at v e s s e l of k n o w n v o l u m e .
This method of making speed measurements is usually used to measure the speed of roughing pumps, rather than UHV pumps. Because of this, calculating Ct, as the pressure "passes" through the various pressure regions, becomes very difficult. That is, C is a function of geometry, pressure and time. It is difficult to compensate for C(P,x~). Therefore, the test apparatus is usually designed so that C t >> S. Then, from (1.13.10), we may make the approximation S --S e . During pumpdown, the rate at which gas (i.e., P(t) x V) is removed from the vess~, Qv, is: Qv =
d(VP)
= P
dt
dV dt
+ V
dP(t) dt
.
(1.16.1)
The volume of the vessel is a constant (i.e., d V / d t = 0). Therefore, Qv =
v
dP(t) dt
.
(1.16.2)
If we neglect outgassing from the manifold, the gas which is being evacuated from the vessel, Qv, is the same gas which is entering the pump (i.e., S x Pp). With C t ~ S, then P --- Pp, and:
Qv -" S P(t).
(1.16.3)
Setting (1.16.2) equal to (1.16.3) and integrating from the time the valve is opened to some time, t, in the test yields: t
f ( S / V ) dt = o P or,
!(t)
dP(t)/P{t),
P(t) = Po e - ( S / IO t,
(1.16.4)
where Po is the pressure in the vessel at t = 0. Equation (1.16.4) is called the system
pumpdown equation. Taking the natural log of both sides, and solving for S, we f'md:
BASIC THEORY 35
t~ S ( t ) =
V
In
t
Po
e(t)
.
(1.16.5)
A plot of pressure, P(t), as a function of time is called a pumpdown curve. We may select any P0 and P(t) along this curve to determine S(P), pump speed, as a function of pressure. One assumption made in solving for (1.16.4) was that pump speed was independent of time (i.e., pump pressure). This is not the case, for it was noted in (1.15.1) of the previous section that speed is a function of pressure, particularly when operating the pump near its blank-off pressure. If speed is a function of pressure, and pressure is a function of time, then speed must be a function of time. Therefore, we may not treat S as a constant, as was done in development of (1.16.4). Using (1.15.1), where Po is the blank-off pressure of the pump, the equivalent of combining (1.16.2) and (1.16.3) becomes: Po
Sma x { 1 -
- - - - } P(t) = V P(t)
dP(t)
dt
When rearranging this equation, integrating, and solving for P(t), we f'md:
P(t) = Po ÷ Po e - ( S / V) t.
(1.16.6)
Another, far more serious, limitation of this pumpdown-rate method of speed measurement is that it neglects the effects of outgassing in the vessel, o-rings and manifolding. Outgassing rates in the vessel, Qo(P4), may stem from several sources which vary both as a function of time and pressure. If we could predict these, we could modify (1.16.2) and (1.16.3) accordingly: Po
Sma x ( 1
P(t)
} P(t) = V
dP(t) dt
k
+ Z Qi(P, t), i= 1
(1.16.7)
where )7Qi(p¢) is the sum of "k" outgassing sources. Outgassing of the chamber will result in very large errors at reduced pressures. Solution of (1.16.7), for some specific sources, is given as a problem at the end of this chapter. The outgassing problem is somewhat remedied by baking the system and using pure N 2 for the test gas. We may also use a simple graphical method for determining pump speed. Assume the chamber shown in Fig. 1.16.1 has a 100 Z volume, has negligible outgassing and has been filled with N2 to a pressure Po = 100 Torr. Assume that we wish to test the speed of a mechanical pump with a blank-off pressure Po of 10-4 Torr. Assume we measure an average vessel pumpdown curve, for the pump under test, similar to that shown in Fig. 1.16.2. Setting (1.16.2) equal to (1.16.3) we obtain: V
S(P,t) =
P(t)
dP(t) x ------.
(1.16.8)
dt
Of course, dP(t)ldt is merely the slope of the pumpdown equation at some chosen P(t). By graphically differentiating this curve at say P = 1.0 Torr, we find that
36 BASIC THEORY
AP(t)/At --0.1 Torr/sec. Using these values of P, AP(t)/At and V in (1.16.8) we determine the pump speed is approximately 10 Z/sec. P(t) 10
-2
2 1
1
I
1
I
I
I
1
I~
1
1
I -
10
o [--,
Log
P(O) -
Log P ( t )
1
i 10
[.x.]
10
a~ ,.~
10
10 -4
_ -
-3
PUMP BLANK-OFF PRESSURE
/
At
mid
10 S =V
Log
(P(0)/P(t)) At
-~.
0
t l I J
[
50
PUMPDOWN TIME -
Figure 1.16.2. Graphical determination the rate at which pressure decreases
l
!
t.
l
150
100 see.
of pump speed using in a known volume.
Note that we used the same gauge to determine both AP and P in (1.16.8). If there is a calibration error in the gauge reading, a gauge correction constant for both AP and P would cancel in (1.16.8), assuming gauge linearity. Therefore, though we know the absolute speed of the pump at the indicated pressure, we do not know the absolute pressure at that reading. In a plot similar to Fig. 1.15.1, there would be some uncertainty in the placement of S(P) along the P-axis. The response times of the gauge and gauge electronics may cause errors in speed measurements. Also, problems of gauge outgassing will obscure the data if tubular ionization gauges are used to determine vessel pressure during a high vacuum pumpdown measurement. An example of this will be given in the chapter on cryopumping. 1.16.2 S i n g l e G a u g e D o m e M e t h o d This method, requiring an apparatus similar to that illustrated in Fig. 1.16.3. has been used for many decades to measure the pumping speed of diffusion pumps.(1 7 ) Gas is introduced into the dome, at a known rate, through the variable leak, V L. On entry into the dome, the gas is directed by a small internal pipe so as to bounce off the top of the dome. This is done to minimize gas beaming effects within the dome. The top of the dome is sloped so that oil which backstreams from the diffusion pump, when condensing on this surface, will run down the sides of the dome rather than drop down onto the pumping stack and cause erratic pressure bursts. The rate at which gas is introduced into the dome, Qd, must be measured by some means. This is often done by measuring AP/At of a known volume, V1, f'dled with gas at a known pressure, P 1 (t). The throughput into the dome is given by:
BASIC THEORY 37 dP1 Qd = Vt
(1.16.9)
dt
NUDE BAYARDALPERT GAUGE
GAS INLET ~ 3D/2
I
PUMP INLET
\\'q
D/a
~
1
V// D~
J
APPARATUS
Figure 1.16.3. Single gauge d o m e , speed m e a s u r i n g
-
apparatus.
It is assumed that the pressure in the dome, Pd, is the pump pressure, Pp. Therefore, the speed of the pump is:
S(P) =
V1 Pd
x
dP1
.
(1.16.10)
dt
Note that both of the gauges must be calibrated for the test gas, and volume V1 must be known, in order to accurately determine the pump speed. A stopwatch is usually used to determine A P 1/At. Use of a strip chart recorder to plot P1 (t) will make possible the graphical differentiation of this value with respect to time. Alternatively, a commercial gas flow meter may be used to establish Qd" Flow meters can presently be purchased from manufacturers which are NIST (National Institute of Standards and Technology) traceable, and accurate to within a few percent. These flow meters must, however, be calibrated for each gas species.
1.16.3 Three Gauge Method The three gauge method of making speed measurements requires the use of an apparatus similar to that shown in Fig. 1.16.4. This technique came into prominence in the early 1960s, at which time it was used to make speed measurements on sputterion pumps.(1 8, t a ) Three vacuum gauges are required. Pump throughput, Qp, is determined by measuring the pressure difference along tube L 1, of calculated conductance C 1. This manifold is a long tube. Therefore, (1.12.4) is used to determine C1. Assuming air as the test gas, throughput is then: Qp
= C1 [P1 - P2 ]
(L)I) 3 = 12.1
L1
[PI-P2].
(1.16.11)
38 BASIC THEORY - - ~ ~_2~=
L1
t
~--D~
L2
Da/Z
!
Cl j Pz
-
-
D,
GAS INLET
NUDE BAYARDALPERT GAUGES
L3
177
=~-~ ~_ZD,---
P
Figure 1.16.4. Three gauge speed m e a s u r i n g dome.
PUMP ~
INLET /
The speed of the pump is found by setting Qp = Sp P_. We can't directly measure the pump pressure. But, we can calculate this value. NPote that manifold L s is also a portion of a long tube. This means that the portion of its conductance from the gauge to the pump flange is given by: (02)
Ca
= 12.1
3
La
.
(1.16.12)
The throughput into the pump is also given by the following: Qp
= Ca [Pa - P p ] .
(1.16.13)
Setting (1.16.13) equal to (1.16.11) and solving for Pp, we find: Pp
= Ps
+
C1
[P1-P2].
(1.16.14)
Of course, Sp Pp = C I [ P x - P2 ].
(1.16.15)
Ca
Solving (1.16.15) for Sp, and substituting (1.16.14) for Pp, yields the following expression for Sp: Sp
=
CaC1 [P1 CaPa -
-P2]
CI[P1 - P2]
.
(1.16.16)
There are numerous problems associated with using this method of speed measurement. It is advisable to bake out the system. This is true for most speed measuring apparatus. However, the long tubes in this configuration add significantly to the error in speed measurement (see problem at end of chapter). Secondly, the three vacuum gauges must be calibrated to precisely plot S.~ as a function of Pp. If all the gauges were normalized with respect to each other, the magnitude of Sp could be precisely determined, though the S o curve would be translated by some error along the P-axis. Gauge normalization ig required as usually the
BASIC T H E O R Y 39 indicated pressure, Pni, of any of the n gauges will not agree, though all are at the same absolute pressure, Pabs" That is: Pabs ¢ P1 i ~ P2 i ~ P3 i.
(1.16.17)
Gauge normalization is accomplished by shutting off the pump and back-filling the test apparatus with the test gas. Normalization constants, ks and ks for example, are then found for two of the three gauges, by requiring that at constant pressure: P1 i = k 2 P 2 i = k3P3 i. Remember, these are the indicated gas pressures. Most commercially available Bayard-Alpert Gauges (BAG) will be accurate to within +-- 20 % for N2 or air. However, note that there are significant differences in ionization gauge sensitivities for the different gases.(2 o, 2 1)
1.16.4 Fischer-Mommsen Dome The Fischer-Mommsen Dome method of speed measurement requires the use of an apparatus similar to that shown in Fig. 1.16.5. It is sometimes called the two gauge method. It was developed at CERNt~ 2 ) for making speed measurements on sputter-ion pumps. For this reason, it is often called the C E R N method. The apparatus has been slightly modified in subsequent AVS and ISO standards.(2 3,2 4 ) I n the late 1970s it also became a popular apparatus for making speed measurements on cryopumps. Data reported in reference (2 s ) were taken using a modified CERN Dome.
D
0.1D ~ 1 ~ - I ILl
i .2 , mJ
~ GAS INLETr~
D
I
I
ALPERT GAUGE
/
L-m
[__ n _.J ........
NUDE BAYARD-
I
PUMP~_P~ q IPNLET
t ~]~
~
APERTURE ~__ALL-M EALL-METAL VALVES
Figure 1.16.6. Modified CERN speed m e a s u r i n g dome. The geometry of the dome was determined through Monte Carlo calculations. This amounts to using a computer to theoretically predict the random sojourn of numerous molecules launched into the system at the input port. These calculations predict the proper location and diameter (i.e., L 2 and Dg 2 ) of the aperture leading to the vacuum gauge indicating pressure P2. TheoreUcally, the indicated value of the pressure reading P2 is equivalent to the pressure Pp at the pump flange. That is, we need not take into account, in speed calculations, tl/at the gauge is a distance equivalent to length L 2 in Fig. 1.16.5. There is some error in this assumption depending on
40 BASIC THEORY the capture coefficient of the pump.( 2 2 ) The pump capture coefficient is merely the probability that a molecule on entering the pump inlet will be captured (i.e., will be pumped). The diameter of the aperture separating the upper and lower chambers of the dome, D a, is selected to maintain the pressure difference between the two chambers to prescribed values. Therefore, some pretest estimate of the probable pump speed is needed. If the size of this aperture is too small, the pressure difference across the aperture will result in P t being much greater than P2. The throughput introduced into the dome is simply the product of the pressure difference between the two chambers and the conductance C a . Or, Qp
=
C a [Pi - P2 l-
(1.16.18)
In that Pp corresponds P2, the pump speed is simply:
Sp
= Ca '
[Pi - P ~ ]
.
(1.16.19)
P2 Note, as in the case of the three gauge method, the two gauges used in this apparatus must be normalized with respect to each other. This becomes a nuisance when making speed measurements over pressures of several orders in magnitude, and with different gases. 0.1D
N U D E BAYARDALPERT GAUGE
i GAS INLET
°
,
D
J
!
_
APERTURE CONDUCTANCE
D/2 ALL- METAL VALVES Figure 1.16.6. Modified CERN speed m e a s u r i n g dome. At Varian, in 1976, I modified the conventional CERN configuration shown in Fig. 1.16.5, and installed two all-metal valves at the locations of the gauge ports (i.e., see Fig. 1.16.6). These two valves were connected to a single gauge which was used, by manipulation of the valves, to determine both of the indicated pressures P 1 and P2. Assuming gauge linearity with pressure, use of an apparatus similar to that shown in Fig. 1.16.6 eliminates the need of determining gauge normalization constants. However, the gauge must still be calibrated for the test gases to accurately locate the speed curve on the S-P coordinate system.
BASIC THEORY 41
Three Gauge versus Fischer-Mommsen Results Over the years, I've found that the three gauge method will yield indicated pump speeds which are 10% - 15% higher than results of the Fischer-Mommsen apparatus, for the same pump. Similar disparities are reported elsewhere.(2 6 ) This is true for all of the gases. Also, for reasons which are discussed in the chapter on cryopumps, the theoretical speed of a cryopump for H 2 0 should be within a few percent of actual measured speeds. However, cryopump speed tests for H2 O, with the Fischer-Mommsen method, will yield results which are approximately 20% lower than the theoretically predicted speed values. I quickly emphasize that, because of apparatus wall pumping effects, making H 2 0 speed measurements is very difficult and subject to errors. In the f'mal analysis, as long as the method of speed measurement is def'med when reporting results, the departure of this measuring equipment from absolute speed is of little consequence. The Modified CERN Dome method of speed measurement is far more convenient to use than any other method of which I am aware. Test results also prove to be very repeatable.
Speed Measurement Errors Due Trace Gas Contamination Speed measurement errors will be exacerbated by the presence of trace contaminant gases. For example, assume that you wish to verify the speed of a pump for hydrogen using a modified CERN dome. The actual hydrogen speed - unknown to you - is 5,000 Z/s. Assume also that the speed of the same pump for argon is 3,000 Z/s. These are reasonable speed numbers for a 250 mm 4, cryopump. Assume that, unknown to us, the hydrogen test gas is contaminated with 1% argon. This trace contamination of argon will have two effects: 1) the ionization gauge has a greater sensitivity for argon than hydrogen (i.e., see Table 1.18.2); and, 2) the aperture between the upper and lower portion of the dome will have a lower conductance for argon than for hydrogen. Let's select the conductance of the dome aperture, C a, such that P1/P2 = 3 for hydrogen, while guessing a hydrogen speed of = 5000 Z/s. Then Call ~ = 2,500 Z/s for hydrogen, whereas the same aperture will have a conductance of 559 Z/s for argon. Let's introduce the test gas into the dome at a total rate of 10-4 Torr-Z/s. Then, the throughputs of argon and hydrogen into the dome will be 10-6 Torr-Z/s and 9.9 x 10-5 Torr-Z/s respectively. Assume the dome vacuum gauge is calibrated for nitrogen. Using the throughput values of argon and hydrogen, actual pump speeds for these two gases, and (1.16.19), we can readily calculate the absolute pressures of these two gases both above and below the dome aperture. If we modify the absolute pressures of argon and hydrogen to the values which will be indicated by the pressure gauge for each gas, and plug the sum of the upper and sum of the lower indicated pressures into (1.16.19), we calculate a pump speed of 5,505 Z/s for hydrogen. That is, the trace contamination of 1% argon in the hydrogen test gas has resulted in overstating the pump speed by more than 10%.
42 BASIC THEORY
1.17 System Diagnostics with Any Pump There is a formal method in the use of logic whereby if: 1) we claim an assumption to be true; 2) we thereafter test the assumption by making measurements or calculations which are known to be exact or true; and then, 3) we reach a conclusion which is a contradiction to the original assumption, then the original assumption is false.(2 T ) For example, assume that we make a claim that the average mass of each of ten apples in a bowl is greater than 0.1 kg (the assumption). Thereafter, we precisely weigh each of the ten apples (the measurement), and find when we sum their masses and divide by ten (the calculation), that the average mass is 0.08 kg. We have reached a contradiction (i.e., 1.0 kg ~ 0.8 kg), and our original assumption was disproven. Similar use of logic may be used to troubleshoot a pump attached to a vacuum system. Assume that a problem exists with a production sputtering system similar to that depicted in Fig. 1.17.1. Say that the pressure in the vacuum chamber is a factor of ten (10) higher today than it was yesterday. We do not keep log books on each piece of equipment in the factory. But, we should! The operators observed that the chamber base pressure today is 2 x 10-6 Torr, where yesterday, they recall it was 2 x 10" 7 Torr.
~-~
VESSEL OF KNOWN - ~ VOLUME ~ ~
-
VACUUM
-GAUGE
T
~
~
BELLJAR TOOLING
CRYOPUMP
Figure
1.17.1.
Coating s y s t e m with p r o b l e m s
HELIUM MPRESSOR
in base p r e s s u r e .
Assume that last night the production engineers put new tooling in the chamber, and as part of a scheduled maintenance program, the maintenance personnel installed a refurbished compressor on the cryopump. The old compressor was due for an adsorber change. Today, because of the pressure problem, the maintenance people leak checked the system and found no leaks! Also, the temperature of the second stage of the cryopump was --10" K (i.e., it was "OK"). The maintenance people claim: "The tooling is dirty or has a virtual leak!" The production people claim: "The tooling is OK, but the maintenance people screwed up the pump!" Rather than become embroiled in these sometimes emotional confrontations, we will resort to a formal logical process in an attempt to locate the problem. We will start with the assumption that: 1) there is nothing wrong with the pump; 2) we will then make measurements and calculations based on this assumption; and, 3) if we reach a contradiction, the original assumption is incorrect. Remember that in UHV-type applications, we should expect errors in measurement accuracy to be typically as much as ± 20% of results predicted by theoretical calculations.
BASIC THEORY 43 The manufacturer's data sheet indicates a pump speed, S.., of 1000 Z/sec for air. We may make calculations for any gas in testing the claim. BVut,for now, we will assume that the gas is air. Assume we calculate the total conductance leading from chamber to pump to be Ct = 1000 Z/sec. Using (1.13.10), we calculate the equivalent speed delivered to the chamber to be: ~ Sc
Sp C t
-
= 500 g/sec.
Sp+C t Then, the total assumed throughput into the chamber, Qt A, and the conclusion drawn by the original assumption of the value of Sp, is given 6y: Qt
= [500Z/sec] × [2× 10-6 Torr], = 10"a Torr-Z/sec.
Measuring Throughput by Rate-of-Pressure-Rise We may use another method for directly measuring the throughput, Qc m, into a chamber of volume Vc. Assume, after pumping for an extended period ot/ time, we close the valve between the pump and chamber. We can conclude something about the rate at which gas is coming into the chamber - either as the result of outgassing or because of leaks - by monitoring the pressure in the chamber, Pc, as a function of time. That is, using (1.16.2), dP c
Qc,m = Vc
•
dt
(1.17.1)
X l 0 -5 10
J
1
J
I
l
1
w
I
1
8
P(
I
F i g u r e 1.17.2. D e t e r m i n i n g the outgassing or leak rate into a vessel by measuring the ra re- of- pressurerise.
~4 2
0
t 0
t 200
400
600
800
TIME AFTER CLOSING VALVE -
1000 see.
This is called the rate-of-pressure-rise method of measuring throughput. Assume that Vc = 1000 Z, and if we plot Pc(t) we obtain a curve such as shown in Fig. 1.17.2. Graphically differentiating the curve, we determine APe~At -10 -5 Torr/100 see; or, 10-7 Torr/sec. Plugging APe/At into (1.17.1) and multiplying this value by Vc, we determine that Qc,m = 10-4 Torr-Z/s. Then,
44 BASIC THEORY Pc Qc,A
Qc~m
Sc,A Cc S c + Cc
(1.17.2)
AP c
- - - At
"
Or,
10" a Torr-Z/sec 10-4 Torr-Z/sec
~
Qc,A ~ Qc,m"
1.0 ± 20%
-.,,.-
n
Note that the same gauge was used to measure both Pc and APe~At. Therefore, an error in gauge calibration in (1.17.2), for some unknown gas, may be excluded. We have made the assumption of a value for the pump speed, have tested the assumption, and therein reached a contradiction. Therefore, the original assumption was incorrect, and the pump is for some reason not delivering the anticipated speeds to the chamber. From this we conclude that either there is some obstruction between the pump and chamber, or the pump is defective. Had we found agreement, to say within ± 20% between Qc A and Qc, m, we could safely assume that the problem did not reside in the pump. i'f a leak could not be found in the system, the disparity between Qc A and Qc m probably is caused by some internal outgassing source. I've found that'even the ~anufacturers of vacuum systems fail to use this simple technique in logic to troubleshoot vacuum systems on the floor. Also, it is best to keep formal log books of the history and performance of complicated vacuum systems. 1.18 E l e c t r i c a l D i s c h a r g e s in G a s e s
It is reasonable for you to ask about the possible relevance of gas ionization processes in a book on capture pumps. Of course, you know that sputter-ion pumps require electrical discharges to function, and some of the material in this section will be drawn on in the chapter dealing with these pumps. However, it is even helpful to know a little about electrical discharge processes when designing and using certain types of cryopumps. For example, closed-loop gaseous helium cryopumps have sealed compressors which provide high pressure, helium gas to the cryopump. This high pressure gas is conveyed through hoses to an expander, where refrigeration is produced. There are electrical motors in both the sealed compressor and in the expander. If there is a leak in the high pressure He system, the pressure may drop below some value at which electrical breakdown will occur. This electrical arcing could cause irreparable damage to the sealed compressor, damage the expander motor, or result in PCB trace shorting within the expander housing. Another problem of electrical breakdown has been encountered by many of us as we tried to leak check in proximity to a sputter-ion pump high voltage feedthrough while the pump was energized. The feedthrough assembly sputtered and arced when we introduced the He, but seemed to function properly in the presence of air. Why is this?
BASIC T H E O R Y 45 Electrical discharges may constitute the momentary sparking or burn-off of metallic whisker in the presence of an electric field - in this case the metal atoms are ionized and form part of the gaseous discharge. With a robust power supply, large currents may be drawn through a sustained discharge. Electrical discharges may be extremely weak and imperceptible to the eye, as in the case of vacuum ionization gauges operating at a low pressures. Or, an intensely glowing discharge - this is the origin of the expression glow discharge - may be visible to the eye, and occupy a large volume, as in the case of certain types of thin film sputtering operations, or in the starting of certain types of sputter-ion pumps. Our visual perception of the discharge stems from light which is given off as a consequence of electrons recombining with ions in the discharge, rather than the process of ionization. Under certain circumstances when starting diode sputter-ion pumps, the intensely glowing discharge may f'dl the entire volume of the vacuum system. This can cause serious damage to sensitive electronic equipment contained within the system, or even promote the polymerization of hydrocarbons on insulators or once optically clean surfaces. Such glow discharges can also cause safety problems. For example, several decades ago I constructed a small sputter-ion pump system to process microwave tubes. This was in the early days of ion pumps, and before I knew anything about the concepts of pump speed, throughput, etc. The tubes being processed had oxide cathodes which gave off copious quantities of CO and CO ~ during what was called cathode conversion. This excess gas caused an ion pump throughput problem (i.e., the pump swamped). We installed an LN2 (liquid nitrogen) trap in the system, with the hope that this would help with CO and CO 2 pumping. (The student will immediately refer to the vapor pressure curves for CO and CO2 at LN2 temperatures; not much help for CO, but beneficial for CO2 .) The LN2 trap was vertically suspended in a well forming part of the vacuum chamber, and its flange rested on a slightly oversized o-ring. On roughing the vacuum system and attempting to start the ion pump, the volume was initially filled with an electrical discharge. Positive ions collected on the reservoir, with a charge build-up equivalent to the pump starting potential. Had I placed the probes of a voltmeter between the reservoir and ground, I would have measured the full potential used in starting the ion pump (i.e., several hundred volts). The LN2 reservoir, insulated by the oversized o-ring, developed the full potential. I inadvertently discovered the problem when refilling the trap. A shocking experience. Lastly, we must use some form of gauge to measure pressure in our vacuum systems. Measuring pressures s 10-e Torr requires the use of ionization gauges. Such gauges are used to ionize gas. The ion current which is produced by the gauge is, over a very large pressure range, directly proportional to the pressure within the system. We will learn, in Chap. 2, that sputter-ion pumps are very similar to a specific type of ionization gauge. It soon becomes evident that when designing or using many forms of capture pumps, having a better understanding of the physical processes involved in the ionization of gases will be very useful.
46 BASIC T H E O R Y Ionizing
Gases
When the outer electrons of atoms or molecules are ripped off the particle by some process, the particle is said to be ionized. If one electron is ripped off the particle, we refer to the ion as being singly ionized; if two electrons are removed, it is said to be doubly ionized, etc. The most common process of ionization in vacuum systems occurs when energetic electrons collide with gaseous molecules or atoms. If there are numerous such collisions of electrons within the gas, and there is some form of sustaining electric field, the volume in the vacuum system may become highly populated with both ions and electrons. Once an electron parts company with the molecule, it is said to be a free electron. A gas which is rich in free electrons and ions, in the presence of an electric field, conducts current and is called an electrical discharge. It is sometimes called a plasma. However, the word plasma is usually used to describe a volume containing the same number of electrons and ions of opposite charge. In this instance, on the average, it is electrically neutral. High
Pressure
Electrical
Discharges
In electrical discharges there is a transient state, where the discharge is initiated by a single electron. An avalanche process occurs, such as illustrated in Fig 1.18.1, where an initial electron, by ionizing collisions, creates additional free electrons, etc. _
_
+
kEY: O"
SECONDARY ELECTRON
,,r~;
IONIZED GAS
!_:+ GAS MOLECULE
i,,'-", <~ ]'~'~0 '" -] '-> ELECTRIC%' n -
..... ~
,-
FIELD
~
% ':
~'-)
."~
"
Q
@
0 i~
"
(
.,-
F i g u r e 1.18.1. T o w n s e n d d i s c h a r g e i n i t i a t e d by a p h o t o e l e c t r o n in t h e p r e s e n c e of an e l e c t r i c field. The avalanche process is called a Townsend Discharge. The first electron may have been launched into the vacuum as a consequence of field emission from some very sharp, metallic, whisker-like point, with extremely high electric fields.( 2 a ) The electron may have been thermally emitted from some type of hot filament or cathode;( ~ 9 ) or, it may have been dislodged from a surface by a photon or some cosmic particle which penetrates the walls of the vacuum envelope. Also, technologists have used radioactive sources, which are beta or alpha emitters (i.e., by radioactive decay, they emit either electrons, positrons or alpha particles), to initiate discharges.(a o ,a 1 )
BASIC T H E O R Y 47 the electric field, and impinge on the opposite plate. An electron dislodged from a surface by a photon is called a photoelectron. If, as it traverses the distance between the plates, the electron comes very close to a gas molecule, and if it has accumulated sufficient energy in the acceleration process, it might ionize the gas molecule. The gas molecule - now a positive ion - would be accelerated toward the negative plate (called the cathode). But, now there are two free electrons being accelerated toward the positive plate (called the anode), the initial photoelectron, and the electron stemming from the ionization of the gas molecule. Each probably have different energies, but both are picking up velocity as they progress toward the anode. The distance the initial electron traveled prior to striking the gas molecule is simply ), eg, where the subscripts denote the mean free path of an electron, "e", in a particular gas, "g". Because of ionizing collisions with gas molecules, the electron current, le, progressively increases closer to the anode. This current may be calculated in the following manner.(Z 2 ) Assume that there are n electrons entering the back of the imaginary 1 cm 2 surface shown in Fig. 1.18.2. PHOTON BEAM P H0 TOELECTRONS
CATHODE.
~
+
ANODE ~ ELECTRIC -0)
FIELD
ION
@--
-
@--- o _ _ - - 0
SECONDARY_.. ELECTRON
,.---I~) ~ / ~ x ,4-//
6---- [
ELECTRONS
F i g u r e 1.18.2. E l e c t r o n a n d ion c u r r e n t s in a n e l e c t r i c a l d i s c h a r g e i n i t i a l e d by s e c o n d a r y a n d p h o L o e l e c t r o n s . Assume that this square is &x cm thick. Def'me c~ as the number of ionizing collisions which occur as n electrons travel the distance Ax. We know that c~ will be inversely proportional to the mean free path of the electron in the gas, and that it will vary as some function of the electric field, E, between the plates. That is: c~
=
~(Z)
.
(1.18.1)
eg Then, the number of free electrons produced, An (dn), for a given distance, &x (dx), is simply:
dn or, n
-- o~nd~ = c e °'x.
48 BASIC T H E O R Y The constant of integration c is found by setting x = 0; that is, it is the number of photoelectrons initially emitted from the cathode, no, each cm 2. If qe is def'med as the charge of an electron, then the product n o x qe is simply Jo, the photoelectron current density drawn from the cathode. That is, the current density anywhere between the two plates is:
J
= Jo ecxx"
(1.18.2)
Equation (1.18.2) assumes that the electric field is uniform between the two plates (i.e., there are no regions of field depression due to space charge). For every free electron created in the space between the two plates, a positively charged ion is created (i.e., we assume gas is singly ionized). These ions are accelerated toward, and bombard the cathode. When charged particles bombard a surface, they will dislodge and create free electrons.(a a ,a 4 ) The electrons which are knocked loose from a surface, as a consequence of the bombardment of primary ions or electrons, are called secondary electrons. The ratio of the number of secondary electrons produced for a given number of primary particles bombarding a surface is called the secondary emission coefficient, and is usually given the symbol -1 in the literature. Defme the following: ns no nc n
= = = =
the number of secondary electrons created / sec-cm2 of cathode, the number of photoelectrons emitted / sec-cm2 of the cathode, the total number of electrons emitted / sec-cm2, the number of electrons reaching the anode / sec-cm 2.
By the above def'mitions, clearly, nc
= n o + n s.
(1.18.3)
The number of free electrons produced between the two plates is simply (n - n c). Also, for every free electron produced in the space between the two plates, there is an ion produced (assuming only singly ionized gas). Therefore, the number of secondary electrons produced by ion bombardment of the cathode is, ns
= ~,(n-
nc).
(1.18.4)
Substituting (1.18.4) into (1.18.3) yields, nc
= n o + -),(n- nc).
(1.18.5)
Substituting (1.18.2) into (1.18.5), and assuming e ctd >> 1, the electron current reaching the anode is found to be: eCtd le
= Jo
1 - -re ct//
(1.18.6)
We see that the electron current becomes very large as the denominator of (1.18.6) approaches zero. In fact, electrical breakdown occurs as the quantity (1 - "t e cta) approaches zero, or,
BASIC T H E O R Y 49 e c~a
=
1 / 7.
(1.18.7)
We have not yet clearly defined the mathematical expression for c~. We merely indicated that it must be inversely proportional to ~'e and be related to some function of the electric field, E. With considerable i n s i s t into experimental evidence, Cobine made the following assumption,(a s ) ot
(1.18.8)
= A P e -BP/E,
where P is the pressure, E the electric field, and A and B are some constants which we can later assign values to based on experimental evidence. If we assume the field between the two plates is uniform, we may express the field in terms of the voltage applied to the plates, Vs, divided by the distance between them (i.e., E = V s / d). Here the subscript "s" denotes the sparking potential, or the voltage at which breakdown occurs. Taking the natural log of (1.18.7), substituting (1.18.8) into this expression, and solving for Vs, yields: Vs
=
BPd
In [{AP d}/{ln (1/7)}1
.
(1.18.9)
Equation (1.18.9) is known as Paschen's Law. We noted at the beginning of this section that electrical breakdown is observed near sputter-ion pump high voltage feedthroughs in the presence of He, but not in the presence of air. This leads us to believe that (1.18.9) may be different for different gases. Of course, this is the case. That is, the constants "A" and "B" vary depending on the gas species. Setting a Vs / a (Pd) = 0, to solve for the minimum breakdown voltage, we find, e (Pd)min =
A
1 In
e Vs,min. =
A
(1.18.10a)
7 1
B In
7
,
(1.18.10b)
where e -- 2.72. Experimental data for several gases are given in Table 1.18.1.(3 6 ) Using these data, we may solve for the values of "A" and "B" for different gases. These values may then be substituted into (1.18.9), and the breakdown voltage determined for different gases as a function of pressure and electrode spacing. This is left as an exercise at the end of this chapter. (Hint! The term "In(i/7)" cancels when substituting general expressions for "A" and "B" in (1.18.9).
50 BASIC T H E O R Y Table 1.18.1. Minimum sparking potentials and pressuredistance product for different gas species. (s 6 ) Gas Air A H2 He CO2 N2 N20 02
Vs,mi n (volts) 327 137 273 156 420 251 418 450
(Pd)min (Torr-cm) 0.567 0.9 1.15 4.0 0.51 0.67 0.5 0.7
The sparking potential will vary significantly depending on the shape of the electrodes. However, assume for the moment that high voltage feedthrough of a sputterion pump may be represented by parallel electrodes. Assume that when we were leak checking around the feedthrough of the sputter-ion pump we introduced approximately one atmosphere of He near the high voltage feedthrough. With these assumptions, and using the values of "A" and "B" for He and air in the equivalent of (1.18.9), we can solve for the ratio of the breakdown voltage for He vs. air. We find that,
VHe
0.09.
(1.18.11)
Vair Equation (1.18.9) is found to be valid for pressures greater than ten atmospheres. Almost all of the high pressure He compressors used in gaseous helium cryopumps, mentioned earlier in this section were originally produced for use in domestic air conditioners. They were intended for use with a type of Freon ®, which proves to have very good electrical breakdown properties - four or five times better than air. It is now obvious why we must take great care not to operate these compressors at reduced He pressures. Also, we see in (1.18.11) that the high voltage feedthroughs of an ion pump will be able to stand off only about 10% of the voltage in the presence of He vs. air.
Low Pressure Ionization Processes In (1.18.1) we noted that the number of ionizing collisions which occurred as a function of the distance an electron traveled in a gas was inversely proportional to the mean free path of the electron in the gas. If we assume that the gas molecules have a certain diameter, say a few A, and are stationary compared with the electrons, and if we further assume that the electrons occupy only a point in space (i.e., their diameter is zero), simple calculations indicate that, )" eg ~'
5.7 ), g.
(1.18.12)
BASIC T H E O R Y 51 Equation (1.18.12) is an approximation for, on the average, the distance an electron will travel in a gas between each collision with a gas molecule. We cannot assume that when an electron collides with a gas molecule it will necessarily result in the molecule being ionized. The probability of an electron ionizing a gas molecule is inversely proportional to ), e-" But, it also depends upon such variables as the energy of the ionizing electron, and~he energy required to tear loose the outer electron from the gas molecule. The probability of a gas molecule being ionized by an electron is termed the molecule's ionization cross section. The symbol o is used in the literature to denote this property. A very comprehensive study of gas ionization cross sections was done by Rapp and Englander-Golden.(a s ) An energetic electron beam is launched into a chamber containing the test gas. It is probable that some gas molecules will lose more than one electron in collisions with primary electrons. Because of this, the gas is said to have a total ionization cross section, o T" This total ionization cross section is given by: I+ oT
=
I+ IN ,f
= = = =
,
I-N~
the the the the
(1.18.13)
ion current produced by gas ionization, ionizing electron beam current, density of the gas, distance the electron beam travels in the gas.
A test apparatus similar to that shown in Fig. 1.18.3 is used to make measurements of the total ionization cross section of different gases.( a 7 ) The test gas is introduced into the ionizing chamber to a pressure of ~ 5 x 10-s Torr.
ELECTRON
-,,
+ t
i
+
,,--~,
A r t
VARIABLE VOLTAGE -/
/~,
- ',r-\
HIGH PRESSURE /TONIZATION C H A M B E R
/
~
,
i /+ ~-~
"
Figure cross
. . .-
ELECTRON BEAM
)
/!
/"4
",,
J HIGH V A C U U M +." CHAMBER
i
/A3
#,
Ji,
1.18.3. Test apparatus sections of various
/,+(;'"
ION COLLECTOR PLATES
,,
\ ELECTRON
COLLECTOR
used to determine the total gases as reported by Tare and
ionization S m i t h . {37)
A collimated electron beam of varying energies is launched into one end of the apparatus. The electron beam is strongly focussed by a magnetic field. The electrons travel a distance ~ within a high pressure ionization chamber to the electron collector. Ions created by the electron beam as it travels the length of the chamber are collected on the negatively biased ion collector plates.
52 BASIC T H E O R Y Figure 1.18.4 is a plot of typical results of such measurements for some of the more common gases.(S s ) In this figure, total ionization cross sections are plotted as a function of the electron beam voltage. It is a common convention to express o T in units of 7tao 2 , where a o is the first Bohr orbit radius of the hydrogen atom (i.e., 5.29 x 10 -9 cm).
z ~J
m ~
10 °
oq oq G
Z 0
N .<
-I
Figure 1.18.4. Total ionization cross sections as measured by Rapp and Englander-Golden.(aS)
L_
10
2;
o
.< [--,
o v,
i0 2 10 i
10 2
10 3
10 4
ELECTRON ENERGY (eV)
Ionization G a u g e Sensitivities There are two forms of ionization gauges which are widely used. The first is the familiar Bayard-Alpert gauge (BAG); the second is the Penning ionization gauge (PIG; sometimes called a cold cathode ionization gauge). An outstanding reference on the physics of these and other types of gauges is given by Redhead, Hobson and Kornelsen.(a 9 ) As mentioned, these gauges are used to ionize gas, and deduce total pressure by the amount of ion current produced, as a function of pressure. If we rearrange (1.18.13) it will take a more familiar form:
I+
= I - (OT £) N.
(1.18.14)
Substituting (1.9.4) into (1.18.14) yields: I+ I+
= I-{(OT
£)
= I - { s } P.
N O
XT
}P
(1.18.15) (1.18.16)
The term inside the brackets of (1.18.15) is defined as the sensitivity, s, of the BayardAlpert gauge, and has the units of Torr" 1. The term I - is called the gauge emission, and is the adjustable electron current emitted from the filament. Years ago, gauge sensitivity was def'med as the product I - x s. However, this led to manufacturers of gauges proclaiming very high gauge sensitivities, when in fact all they did was increase the gauge emission for the same gauge geometry as their competitor. Electrons are launched into the gauge structure of a BAG, where they orbit about the electrodes in such a manner that the equivalent length £ is very large. Therefore, the probability of ionizing gas is greatly increased, for a given pressure. The value of £ will vary depending on the geometry of the Bayard-Alpert gauge.
BASIC THEORY 53 The energy of the electrons accelerated from the cathode of a Bayard-Alpert gauge is usually fixed, and of the order of 200 volts. These gauges are usually calibrated to indicate the pressure, in Torr, for N2. If we normalize Rapp and EnglanderGolden's data to N2, the relative sensitivity of a gauge for the various gases will be given by values in the following table. Data shown in the two far right columns are summaries of relative gauge sensitivities.(4 o, 4 1 ,s 4 ) Work by Rapp and EnglanderGolden are more current. Table 1.18.2. Ionization cross sections of v a r i o u s gases normalized t o t h e c r o s s s e c t i o n of N 2 a n d compared with relative vacuum gauge sensitivities. GAS N2
100
eV
1.00
200
eV
1.00
300' eV
BAG
PIG
1.00
1.00
1.00
He
0.37
0.32
0.30
0.39
0.49
He
0.15
0.15
0.16
0.17
0.25
Ne
0.25
0.34
0.38
0.32
Ar
1.13
1.05
1.03
1.09
02
1.06
1.12
1.14
1.10
CO
1.05
1.04
1.04
1.05
C02
1.39
1.43
1.46
1.36
CH4
1.45
1.33
1.30
-
C2H4
2.29
-
-
Data t a k e n f r o m r e f e r e n c e s
0.95
36, 40, 41 a n d 54.
Penning ionization gauges are called cold cathode gauges because they do not have hot cathodes or filaments which emit ionizing electrons. There is a self-sustaining discharge in the gauge which is replenished by secondary electrons which are emitted as the result of ion bombardment of the cathode. In this case the expression for sensitivity, s, has the units Amps per Torr, and the ion current, I +, is given by the following: I +
= s P.
(1.18.17)
We will discuss (1.18.17) at considerable length in the next chapter.
Partial Pressure Gauges There are three elements common to all partial pressure gauges (PPGs). They are: i) the ion source; ii) the analyzer or mass filter; and, iii) the ion collector. These elements are illustrated in Fig. 1.18.5. Gases are ionized in the ion source, the ions launched into the mass filter, and the ions are collected at the exit to the filter in the collector. The ion sources of all PPGs are just some form of ionization gauge. Because of this, the sensitivities of all PPGs for gases are inherently similar to the sensitivities of ionization gauges for the same gases. Therefore, one can't interpret the ion currents reaching the collector of a PPG as being in direct proportion to their respective partial pressures. For example, if we were to compare the amplitudes of ion currents of equal partial pressures of nitrogen and helium, the nitrogen peak will be --x7 greater than the helium peak (e.g., see Table 1.18.2). This is because nitrogen has a
54 BASIC THEORY much greater ionization cross-section than helium. ION ~ SOURCE
Figure
ANALYZER (m/e filter)
~
ION COLLECTOR
1.18.5. The b a s i c b u i l d i n g b l o c k s of all p a r t i a l p r e s s u r e g a u g e s .
All analyzers or mass filters of PPGs exploit the m/e (mass to charge) ratio of the ion species created in the ion source. Vacuum technicians have been very innovative throughout the years in developing various forms of analyzers. However, all of these analyzers have one thing in common: they differentiate between ion species based on either the ability of the ion to successfully pass through a selective analyzer (i.e., filter), or based on the time it takes the ion species to transit through the analyzer. The former is a spatial filter, whereas the latter is a temporal filter. Filters of the latter form are called time-of-flight analyzers. Ions having the same energy are launched into the "analyzer", simply a drift tube, with a kinetic energy of V,mv 2 = eV, where m is the mass of the species (often converted when displayed to amu or atomic mass units), v its velocity, e the charge of the ion, and V the ion accelerating voltage. If the length of the analyzer is L, then the time, t, it takes an ion species to transit the analyzer is simply: t = L (m/2eV) g. That is, heavier ions will take longer to reach the collector than lighter ions of the same charge. So, if all of the ions are launched down the drift tube at the same instant in time, say by some pulsed voltage, when each ion species reaches the collector depends on its m/e value. A lighter ions of the same charge as a heavier ion will win the race from the starting line (the ionizer) to the finish line (the collector). MAGNETIC FIELD OF MAGNITUDE B--~\ MONOERGIC MULTI-SPECIES ~
///
'\,,
] ~, / ~
~
ION _ \ X .... ~ SOURCE ~ / ~ \ ) / x
m l/e Figure
1.18.6.
90 ° SECTOR ..... MAGNET
|
1~/ 2
< m2/e
90°Magnetic
I
4"( /.-~
/F ma//e
~ // j I0N ")..,~//N~ ~ C O L L E C T O R
< ma/e
Sector
mass
spectrometer.
With spatial type filters we vary or tune analyzer parameters so that for given settings, species having only one value of m/e can make it through the analyzer to the ion collector. Some use magnetic fields to create m/e dependent ion orbits that selectively permit only one species to reach the ion collector (see Fig. 1.18.6). These are called magnetic sector analyzers. A somewhat collimated ion beam, comprising all of the ionized species, is launched from the ionizer into the gap of a permanent
BASIC THEORY 55 magnet. These ions are monoergic (i.e., all have the same kinetic energy). However, only ions having the "right" charge to mass ratio (i.e., m~/e of Fig. 1.18.6) make it too the ion collector. In order for species having some other m/e ratio to pass through the slit entrance of the ion collector, the accelerating voltage, V, must be changed. This is the means by which the analyzer is "tuned". The magnet shown in Fig. 1.18.6 is called a 90* magnetic sector; that is, the beam of properly tuned ions is bent 90* prior to reaching the ion collector. Other magnetic sector analyzers might bend the ion beam 60", 120" or even 180" prior to it entering the ion collector. In order for the beam to have a radius of curvature in the magnetic field there must be two equal and opposite forces acting on the beam particles: 1) mv ~/R, the centrifugal force; and, 2) evB, the Lorentz force; where B is the magnitude of the magnetic field in Tesla, v the velocity of the ion (meters/sec.), and R the radius of curvature of the orbit. Equating the magnitude of these two values, and solving for R, yields: mk,
R -
meters
eR
(1.18.18)
Of course, the velocity of the ion is just v = (2eV/m) V~. As a point of clarification, the term "e" in (1.18.18) constitutes the magnitude of the charge of an electron (i.e., 1.60 × 10" 1 9 coulombs), and m the mass of the ion in kilograms. However, in plotted spectra we often see the term "m/e". In this case "m" is the amu of the ion, and "e" the magnitude of the unit charge (i.e., 1, 2, etc.). If we plug the term for the ion velocity into (1.18.18), we find: Ri =
B
(2 m i V/ei)~
meters
(1.18.19)
Def'me "i" as the attributes of ion 'T' and "i + 1" as those of an ion of one amu unit larger. Solving for Ri/R i + 1 w e find that the orbits of these two ions, for the same e, V, and B, change according to:
Ri / Ri + 1 = (mi / mi + 1) Z This suggests that the larger the mass of neighboring ions, the smaller the difference between the two respective radii. Therefore, unless you have a very large and expensive magnetic sector PPG, the ability of the instrument to resolve between neighboring species diminishes at higher m/e values. Another type of spatial filter exploits the natural cyclotron resonances of orbiting ions. These operate somewhat like cyclotrons. A stable ion beam, accelerated by an RF field and trapped by DC electric and magnetic fields, is stored in circular-like orbits, and impinges at grazing incidence on an ion collector.(S 1 ) These PPGs were given the descriptive name omegatron, and were in common use in the 1950s. They required the use of permanent magnets with very high fields (i.e., the magnets were large and cumbersome), and the gauge tubes, usually constructed of glass, were fragile envelopes. (Yes, and the system was at 600* C when it happened.) The last classification of spatial-type analyzers exclusively uses combinations of DC and AC electric fields to selectively filter out all the in-transit ions but those of
56 BASIC THEORY unique m/e value. The earlier models of such analyzers were called Farvatrons. They worked somewhat like multi-cavity klystrons, in that when tuned only ions of the "right" m/e value could successfully sojourn through the RF "cavities". Other ions of the wrong m/e value, crashed into the electrodes of the analyzer. The most popular present-day variety of these AC/DC analyzers is the QRGA (quadrupole residual gas analyzer). It too works on the basis of selectively filtering out all in-transit ions except those having a certain m/e value. Ions are launched from an ion source down the center (i.e., z-axis) of an array of four parallel, equally spaced conducting rods (e.g., see Fig. 1.18.7). Alternating and DC potentials are applied to rod pairs on opposite sides of the launched ion beam. For given AC to DC rod potentials, all but the select m/e ions crash into the analyzer rods prior to reaching the collector. The trajectories of ions having the right m/e ratio are such that they stagger through the filter rods, and make it to the ion collector.
i //
I
//" ~ \I--Y
Z~
~,~
/ '~\
-~×
I
I Figure 1.18.7 Quadrupole mass-to-charge filter.
Ion Collectors Ion collectors usually comprise some form of metal cup. These cups are referred to as Faraday collectors (e.g., see Fig. 1.18.8). The cup configuration is used to prevent secondary electrons, dislodged by the impinging ion beam, from escaping the collector. Remember, a negative electron leaving a surface is "seen" by electronic circuitry in the same manner as an arriving positive ion. Ion collectors may also include some form of electron multiplier (EM). Electron multipliers are used when the ion current is small. The ion current impinges on a surface at entrance of the EM. These ion dislodge secondary electrons (see Fig. 1.18.9). The secondary electrons are accelerated down the innards of the multiplier, successively impinging on additional surfaces, called dynodes, and dislodging additional secondary electrons. This creates an avalanche condition along the multiplier, until at the exit of the EM, the electron current measured may exceed the entering ion current by a factor of many thousands.
BASIC T H E O R Y 57 ANALYZER EXIT --'"" "'i
POSITIVE ION BEAM
ELECTRON SUPPRESSOR
|
....
~
I
FARADAY CUP
"
f /
v
J 1 ,
Figure
1.18.8.
F a r a d a y c u p ion c o l l e c t o r .
The magnitude of the ion current amplification is called the gain of the EM. The gain of an EM will change with changes in the applied voltage. In order to measure the gain one must be able to measure the ion current entering the EM (i.e., striking the first dynode), and the electron current leaving the EM. However, most low cost QRGAs do not have provisions for measuring the gain of the EM. ANALYZER EXIT --"
SECONDARY ELECTRONS
,--
POSITIVE
~
ION--
ELECTRON COLLECTOR
@
~
2
3 4
5
6
7
8 O I0 II 12 13 14 ;
,
.
t
,
..
~
l
¢
_ ,
,
CL_ ~
I
~
r I
:
, i
I
I
-H.V. Figure
1.18.9.
Fourteen-dynode
blind electron
Venetian
multiplier.
In many applications EMs are not required. Absent the use of an EM, if the ion signal is very small it may either: i) go undetected; or, ii) require very slow m/e sweep times in order to be measured. The disadvantages in the use of EMs include the fact that their gain will vary with: i) the m/e species; ii) time and number of exposures to air; iii) the degree of baking; iv) exposure to certain contaminating gases; and, v) the degree of use.
Spectra Interpretation One major advantage in interpreting QRGA spectra is that the m/e values are linearly displayed along the x-axis as the filter voltages are varied. That is, if one connects the x-axis input of any x-y recorder to a voltage proportional to the DC potential of the f'llter, and uses the y-axis as the input of species current amplitude, the species will be linearly arrayed along the x-axis based on rn/e. You can literally use a ruler to scale and identify the amu of various species along the x-axis. Figure 1.18.10 shows a QRGA spectra. The x-axis linearly displays the m/e of gas species in the vacuum system. You may be surprised to note that the partial pressures of H 2 (m/e = 2), He (m/e = 4), N 2 (m/e = 28) and Ar (m/e = 40) were identical. The
58 BASIC THEORY primary reason for the difference in QRGA current amplitudes for the four gases is due to their differences in ionization cross-sections. -6 ARGON PARTIAL PRESSURE 3.8 x IO TORR BAYARD-ALPERT GAUGE OFF DURING TEST QUADRUPOLE EMISSION 250 /.z.A
nA
350
rr 1.1.1 J
a.
300
~ w
~
25o 200 i
150
~ ~
I00
o
5o
rc, c j
_
2 4
..._ ~ . ~ , ~ / ~ _ _ I I 1 I l
12 14 16 18 20
I
I
28
32
MASS TO CHARGE R A T I O -
40
44
m/e
Figure 1.18.10. Ouadrupole gas analyzer spectrum showing relative species current amplitudes when measuring the same partial pressures of H2, He, N 2 and Ar. A common limitation of all anayzer filters exploiting the m/e ratio of ions is that they are not able to distinguish between singly or multiply ionized species. For example, assume that an ionized beam of argon is launched into an m/e filter. You will observe at the collector a major peak at m/e = 40 (i.e., singly ionized argon; 40/1). You will also observe an m/e peak at amu 20. This is probably doubly ionized argon (i.e., it has an m/e peak of 40/2). How can we discern that the m/e peak at amu 20 is not singly ionized neon (i.e., m/e = 20/1)? The trick in so doing is to establish simultaneous equations involving the peak amplitudes of all of the m/e values displayed, and solving for the probable species present based on the solution of these equations. For example, if neon were in fact present, you would expect a small m/e peak at m/e = 10 (i.e., doubly ionized neon). Lacking an m/e peak at 10 probably implies that the m/e peak at 20 is doubly ionized argon. Modern-day commercial PPGs can be purchased having features to crunch the numbers, and report tabular results of probable species present in the system. This in situ spectrum interpretation becomes extremely helpful when measuring complex gas mixtures. In such cases, not only may there be peaks of singly and multiply ionized species, but often complex molecules will be fragmented into several species in the ion source. These broken up fragments, themselves ionized, result in characteristic spectra which are unique to the gas, and are known as fractionating or cracla'ng patterns of the gas.(4 9 ) Certain m/e analyzers exhibit what is called high mass discrimination. This merely means that the higher m/e species are not as effectively transmitted down the analyzer as are the low mass species. An example of high mass discrimination is
BASIC THEORY 59 shown in Fig. 1.18.11. In this figure the x-axis shows the m/e value, the y-axis the normalized signal amplitude of each species, and the z-axis (out of the page) the relative changes in QRGA resolution setting. All data were normalized to doubly ionized argon toward the middle of the sensitivity-m/e surface. Notice how the sensitivity-m/e surface seems to roll off at higher resolution settings and m/e values. 'X. AM
MEASURED
AT HALF PEAK HEIGHT
co liT)
M/AId -- . . _ l e Z )
)
. . . . . . . . . . .
(Z;
--
_j " r
, ,o-'~"
l0 t
0 i ¢'
~I I
~0 -a
// /
2
•
4
t4
t6
MAST SOCHARG REATO I 18
20
,'
i
,
2B
32
40
" 44
/
/
Figure 1.18.11. Quadrupole system relative sensitivity as a function of resolution. Resolution is merely the ability of the instrument to resolve the difference between adjacent peaks. That is, adjacent peaks don't overlap into each other, but are clearly distinguishable one from the other. Reference 50 provides several quantitative interpretations of the meaning of resolution.
Gauge Calibration Methods Gauges (and gauge controllers) may be obtained from United States manufacturers which have been calibrated in some manner traceable to the NIST (National Institute of Standards and Technology). These include forms of Bayard-Alpert gauges, capacitance manometers, and more recently, forms of spinning rotor gauges.( s s ) Spinning rotor gauges can be calibrated at NIST with a repeatability of --0.25%, and this technology transferred, with training, to an industrial setting with assurances of repeatability in the range of 1-2%.(s 6 ) The spinning rotor gauge may then be used as a secondary standard for the calibration of other high vacuum gauges. You may prefer to have "primary" gauge calibration capabilities at your own facility. A simple system used to calibrate both total and partial pressure gauges is shown in Fig. 1.18.12. When using this system we invoke the equation: Seff P = Q
where
S e fr P
= the effective speed delivered to the dome in Z/s, = the pressure in the dome (to be deduced) in Torr,
(1.18.18)
60 BASIC THEORY and,
Q
= the gas throughput introduced into the dome, in Torr-Z/s. GAS
- ~ i
G~S t~
[
V ~
[~
'
Vo
PUMP T
GAUGE TO BE
"" ....... KNOWN
INLE CONDUCTANCE
ROUGHING--I~IPORT Figure
1.18.12.
Gauge
calibration
system.
In calibration processes we wish to have assurances of both accuracy and repeatability. If we can claim that any two of the terms in (1.18.18) are accurate (perhaps traceable), we can claim accuracy in our calibration method. Of course, in this exercise we wish to calibrate the gauge. Therefore, we must establish both S e fr and Q to the desired level of accuracy, and use error bars when claiming accuracy in P. There are three common methods used to quantify the value of Q: i) flow meters; ii) measuring the pressure difference across a known conductance; and, iii) measuring the rate-of-change of gas pressure in a known volume (i.e., VAP/At). Gas flow meters may be obtained from suppliers which are NIST traceable to within 1-2%. However, most of these flow meters vary in calibration with gas species. Using a known conductance method to measure Q has its own problems, as the gauges used to measure the pressure on each side of the conductance must be calibrated. The use of a calibrated "low-vacuum" gauge (i.e., capable of operating at pressure of 1-1000 Torr), such as a capacitance manometer, to deduce VAP/At is elegantly simple. It merely requires that the low-pressure gauge be certified by the manufacturer as NIST traceable, with error bars provided for the operating range of interest. The volume, V, within which the test gas is stored, must also be known. For example, assume that the gauge to be calibrate is attached to the dome illustrated in Fig. 1.18.12. The gas is introduced into the test dome through a variable leak valve. The throughput, Q, introduced into the dome is just:
Q
= VAP/At
(1.18.19)
Assume that the volume, V, in (1.18.19) comprises the known test gas volume and associated piping between V 1 and V 2 (i.e., valves V 1 and V 2 ). The volume can be either calculated or accurately determined by measuring a weight change when filled with water. The volume of the capacitance manometer and associated piping between V 1 and the variable leak can be found by volumetric expansion methods. Knowing this smaller volume may be important when wanting to introduce small values of throughput, Q, into the dome. Using these methods, one may determine
BASIC T H E O R Y 61 the respective volumes easily to within an accuracy of < 1.0%. The next task is to establish accuracy in the value of S e f. r This is a simple task. Assume that the manufacturer of the pump appending the dome claims a pump speed, So, of 1000 Z/s ---20%, for nitrogen, and we wish to calibrate the gauge for this gas. "If we accurately machine the hole in the dome aperture plate with a diameter of 1.040 cm, this will create an aperture conductance of 10.0 Z/s for nitrogen. What would be the error bars in S e fr assuming the vendor pump speed claim? If we invoke an equation similar to (1.13.10), we find that if the pump speed is actually 800 Z/s, S e fr = 9.877 Z/s. However, if the speed of the pump is actually 1,200 Z/s, S efr = 9.917 Z/s. Taking the average of the calculated values of Serf, we can predict a value of speed delivered to the dome of S e fr = 9.897 Z/s ---0.2%. We can establish the volume in (1.18.19) to within less than one percent. Errors bars were provided by the capacitance manometer manufacturer. We can establish S efr in (1.18.18) to within less than one percent. The only variable unaccounted for, with respect to accuracy, is time, t, in (1.18.19). An inexpensive digital wristwatch will prove more than satisfactory in this regard. On the other hand, young engineers today use graphical programing software (e.g., LabVIEW~), and RS-232 or RS-485 outputs from the vacuum gauges to crunch the numbers, using some form of PC. The f'mal results will be some deduced absolute pressure, Pabs, associated with each indicated pressures, Pind, measured. The gauge calibration constant, k, is found by finding its value from: Pabs
= kPin d
(1.18.20)
Knowing the measure of uncertainty in all of the variables, one can easily establish error bars in Pabs.(5 2 ) There are some considerations to keep in mind when calibrating gauges: It is best to make the complete system bakeable, and to use all-metal valves and flange seals. Make sure the gas, when introduced into the dome, is not beaming directly at the gauge. Reasonable dimensions for a calibration dome can be found in the literature.(5 a ) Select the volume V of VAP/At so that the pressure in the dome undergoes only slight changes in time, but the pressure changes in the test gas volume are significant. If the conductance of the hole in the aperture plate is too small, errors will be encountered in gauge calibration due to pumping effects of the PPG or ionization gauge. The pump speed of a typical ionization gauge for most chemically active air gases ranges between 0.1-0.3 Z/s with a gauge emission of 4 mA. It is best to locate the test apparatus in an air-conditioned room. The capacitance manometer, variable leak and pressure in the test gas volume will varywith temperature, and cause reading errors.
62 BASIC THEORY
1.19 Vacuum Seals Criteria to be considered when selecting vacuum seals for any application include: safety; reliability; bakeability; gas permeation rates; material outgassing rates; mating surface materials; mating surface finishes; ease of installation; required sealing force; system stress/strain; applicable flange sizes; parts cleaning; virtual leaks; initial costs; replacement costs; availability (replacement); radiation "hardness"; fragility; loss-tangent heating; electrical insulation; and, shelf life. This listing of twenty-one items seems unnecessarily lengthy. But, I've had to deal with problems in numerous instances when either a colleague or I disregarded one or more of the criteria in this listing. I'll share some of these instances with you, and recommend further reading on the subject of vacuum seals.
1.19.1 Elastomer Seals Most capture pumps are attached to vacuum systems using metal seals, or are distributed within the system when, for example, using some forms of nonevaporable getters or cryogenically cooled coils. Some forms of cryopumps prove the exception to this rule. Elastomer or o-ring seals have numerous advantages, low cost and ease of installation being major considerations. However, the primary disadvantages in their use are: i) high outgassing rates compared to metal seals; and, ii) the total reversibility of gas absorption and desorption from the bulk material.(6 1,6 6,8 7 ) This gas sorption reversibility presents a major problem when o-rings are used as vacuum seals in large vacuum chambers. Whereas the o-ring outgassing process will occur over a period of hundreds of hours, the readsorption of, for example water vapor in Viton, becomes complete on air exposure in only about 24 hours. An example of the outgassing rates of typical o-rings is shown in Fig. 1.19.1. The elastomer samples were suspended in an all-metal test chamber. The time-dependent vacuum outgassing of all substances can be expressed as follows: qm(t) where
qm (t) ql t
_- q l
x ( t / t l )-c~
(1.19.1) 2
= the material outgassing rate in torr-L/s-cm , = the material outgassing rate at time t 1, = the pumpdown time in seconds, minutes or hours, 1.0 for most metals, whereas, 0.6 for most elastomers.
If the value of t is measured in seconds, minutes or hours, of course t I must have the same units of time. After the air gases are pumped out of an unbaked system, the primary outgassing constituent is water vapor. The sources of water vapor outgassing include metal surfaces within the system, and diffusion of water vapor out of o-ring seals, if any. Intuitively one might expect the value of ct for metals to vary drastically according to the surface condition. This usually isn't the case. For example, Schram reported an
BASIC T H E O R Y 63 a for anodized aluminum - a metal "sponge" - of 0.9, whereas unbaked aluminum had an cz of 1.0.( 6 4 ) The dominant factor is the value of q 1 which for anodized aluminum is - 4 4 times greater than that of fresh or unbaked aluminum. -4 I
O b~
I
Y 1 L_
17 B u n a - N
~9 Q)
~9 op,,~
I I[
1 0-5
O Polyimide
--
10 -6
I
~
10
~em c
7 ~ -
•
t
+ Viton-A
10-8
I
10
1
I lj[
,
_ I ill
2
l 3
~ zll
10 10 Duration Under Vacuum-
l
i l
4
10 10 minutes
F i g u r e 1 . 1 9 . 1 0 u t g a s s i n g Rate of T h r e e P o p u l a r E l a s t o m e r s as a F u n c t i o n of Time in V a c u u m . Assume that a vacuum chamber is being pumped down. The air gases, such as nitrogen, oxygen, and argon, are evacuated very quickly. After this, the surfaces in the chamber outgas water vapor, so that the pressure, P(t), as a function of time will take the form:
P(t)
= P 1 x (t / t 1 )-a
(1.19.2)
where P1 is the pressure in the chamber in say 1 hour (i.e., t 1 ). If we take the log of both sides of (1.19.2) and solve for ct, we find that: [log P(t)- log P x ] -cz
=
(1.19.3) [ l o g t - logt 1 ]
This suggests that if we plot P(t) on log-log paper as a function of time, we can actually measure the slope to the curve. By analyzing the slope of the log-log curve, we will be able to determine if the outgassing curve stems from elastomers or from metal surfaces within the system. If the plot curves early in the pumpdown, rather than being a straight line, this means that P(t) is the superposition of both metal and elastomer outgassing. Of course, if the plot is initially a straight line, it will eventually become a curve as the base pressure of the pump starts to dominate.
64 BASIC THEORY The value of c~ ~ 1.0 for metals means that for the same surface area, a material having an a ~ 0.5 will take much longer to outgas. The total outgassing rate of an o-ring would be A x q(t), where A is the effective surface area of the o-ring exposed to the vacuum. This "effective" exposed area was reported by de Csernatony to be approximately half of the total surface area of the o_ring.(6 s-) I have verified, on a number of occasions, that this is a very good estimate of the equivalent surface area. In one of these instances a manufacturer of semiconductor coating systems believed that they had a chamber cleaning problem. They concluded this because of problems of protracted pumpdown. Their chambers were aluminum, and seemed to be cleaned in a proper manner. The aluminum surface area of their chambers was -7.2 × 10 ~- cm 2 whereas the total equivalent surface area of the o-rings was a shocking 2750 cm 2. The pumpdown curve of their system in time very closely replicated the 5% curve of Fig. 1.19.2, where the relative percentages shown in the figure show the proportion of elastomer surface area to that of clean aluminum. Their protracted chamber pump-down problem stemmed from the use of an excessive amount of o-rings in their system design. 10
-5
I I
I
1"1 I
I
O-Ring ~ 10
"1 I
1
% of Am
I l~__ -
-6
o b~
~ I0
-7
--
'
-8
.
.
.
0 %
.
- Am~72,000
em 2
s~a,aoo i /s 10
-9
_
10
l
1
I 1t
I
10
I II
2
Time
I
10 -
3
I lit
I
10
4
10
minutes
F i g u r e 1.19.2. P u m p d o w n of a B e l l - J a r of S u r f a c e A r e a Am v. R e l a t i v e O - R i n g S u r f a c e A r e a Lastly, o-rings have another disadvantage: gases diffuse through the o_rings.(6 6,6 7,6 8 ) Anyone with experience in leak checking systems recognizes the problems associated with the diffusion of He through o-rings during leak checking. However, it has been shown that other gases will diffuse through o-rings. Water vapor proves to be one of these gases when using Viton o-rings.( 6 a ) Yoshimura showed the water vapor diffusion rate through Viton depended markedly on the humidity within the room. Steady-state water vapor outgassing rates from diffusion through a Viton gasket were shown to be -1.7x 10" 7 Torr-Z/s-cm 2 under extremely humid conditions, and -5.4 x 10-9 Torr-Z/s-cm 2 in fairly dry laboratory conditions. If we compare these findings with the data shown in Fig. 1.19.1, we conclude that the
BASIC THEORY 65 initial water vapor sorbed in the bulk the advantages and by Peacock.( 6 9 ) variety of materials
outgassing from the Viton is predominantly coming from water material. Perhaps the most comprehensive paper published on disadvantages of the use of elastomers in vacuum systems is given A comprehensive summary of the outzassing properties of a is given in Appendix C of O'Hanlon.(S o
1.19.2 Metal Seals Varieties of metal seals include: ConFlat@-type (i.e., knife-edge or stepped flanges with coined gaskets);( s z ) Wheeler @ (special wire seals);( s s ) Foil (AI, In, Cu);(s 9 ) Wire (A1, Au, Ag, Cu, In, etc.);~ s 9 ) C-rings, coated with soft metals such as Pb or In, and which may or may not be reinforced with a spring; C-rings of Inconel, jacketed with a ductal metal such as AI or Cu, and spring reinforced;( 6 o ) C-rin~ with "diamond" edges, and spring reinforced (e.g., the HelicoFlex Delta®);~ 6 1 )6 2 ) and, what are called "diamond" seals.( 6 z ) The ConFlat ~ flange seal, invented by W. Wheeler at Varian Associates in the early 1960s, is by far the most popular, robust and reliable of all of the metal seals.(S z ) A cross-section of this seal configuration is illustrated in Fig. 1.19.3. The seal configuration in the lower section of this figure illustrates a nonrotatable flange configuration, whereas the flange pair in the upper section a rotatable configuration.
GASKET ----..4 ( "\ "" "" " "x, t /' ""../ .... '*" ,~ ROTATABLE / .~"~.,~" k-,7~!.,",'.:i~::fi:~...:;'~.. FLANGE ASSEMBLY ", ",.. "<..... "..... , .....~/. >-. .... I',_'-.-b,. ' ~ .
"!< / / . - y . : , ~
"- : , . ",,". " , \
""~
/
l1 l, l
'
,_
,
,
I
,r
1
FLANGE ASSEMBLE KNIFE-EDGE
Figure
,-'/
\ TUBING
WELDED
1.19.3. ConFlat
FLANGE ~- INSERTS r-SOLID
t
I / ..... .
",,~J ,, V" / /
/rLANGr, S / ~
-,
...~
,-
\IM'~~<'/~':~'70
._./-X,~.~,
",]/,.""
,
i.'"(-/ j"j,/
Rotatable and non-rotatable ® flange configurations.
When the flanges are bolted together, and the seal knife-edges cut into the gasket material, the material on the outer periphery of the knife-edge becomes hydraulically confined when wedged between the cone landings and the outer shoulders within which the seal is captured. Though after torquing the bolts the flange faces may appear to be in face-to-face contact, this is not the case. There is stored energy in
66 BASIC THEORY both the bending moment of the flanges and the strain of the bolts so that even with a severe baking, the gasket material remains highly compressed. The copper gaskets are punched out of rolled sheet. The compressing or "coinhag" of the gasket has the added benefit of sealing "piping" defects in the OFHC (oxygen free, high conductivity) copper gasket material. In the late 1970s I designed a product a product called the HiQ ® sputter-ion pump. It was assembled and sealed together with Cu gaskets and facing knife-edges.l, r ~ ) But, the seal design lacked the outer restraining shoulders. Assemblies rarely leaked on initial assembly, but on bakeout of the devices a large percentage (--20%) of the Cu gaskets developed leaks in the plane of the gasket due to "piping" in the gasket material. Wheeler explained that not only do the outer flange shoulders provide flange-to-flange indexing, but they also assure the coining of the Cu gaskets to inherently seal potential piping leaks therein. After the addition of the flange shoulders to the design, leaks were no longer experienced as a consequence of bakeout. Any metal may be used for the flange material as long as it is harder than the gasket material. Because of its high hot strength, low outgassing properties, good weldability, and reasonable cost, 300-series (e.g., 304, 304L, 316L, 321, etc.) stainless steel is widely used as a flange material. If fully annealed mild steel (i.e., 1020C) is used for the flange material there may be problems of knife-edge roll-over when using "half-hard" OFHC copper gaskets. In fact, even 300-series stainless steel flanges will have knife-edge roll-over problems if they are first annealed by some brazing or firing process. When this may be likely, the knife-edge is cut so that its included angle is increased from the conventional 70* to a 90* included angle. Though this prevents roll-over problems, it does require higher sealing forces. Sealed stainless steel flanges with Cu gaskets may be baked to temperatures of > 500* C, though the gasket material may sinter to the flange material at these temperatures. By silver plating the gasket it is possible to avoid this sintering problem at high bakeout temperatures. The rotatable flange pairs, particularly those of smaller sizes (i.e., the miniConFlat ~ flanges), may develop leaks if the seal is subjected to high bending moments through the flange joint. This is because the portion of the Cu gasket bearing the compressive load of the bending moment, may be radially extruded away from the knife-edge. Because of this, close attention must be paid to dimensional tolerances between the o.d. of the gasket and the i.d. of the gasket retaining shoulder. However, this never appears to be a problem with nonrotatable flange pairs. Some in the literature recently disputed this.(7 1 ) Their findings may be valid assuming: i) reliability during bakeout of the flange pair is not a consideration; and, //) if one or both flanges are rotatable, there is no possibility of a bending moment on the flange joint in the application. ConFlat ® flange sealing forces are comparatively high (i.e., see Table 1.19.1). If the seal joint must be frequently taken apart aluminum gasket material may be substituted for the Cu gasket. But, because of the very poor hot strength of AI, one should assume that a flange pair having an AI gasket is bakeable to s 150" C. Most applications require the use of Cu gaskets. Therefore, because of market demand for the Cu gaskets, AI gaskets of the same size prove more costly. Table 1.19.1 summarizes the required sealing force for aluminum (AA6061T6) and half hard copper gasket materials, and for knife-edge included angles of 70* and 90*.
BASIC THEORY 67 Table 1.19.1. C o n F l a t ® f l a n g e s e a l i n g f o r c e r e q u i r e d as a f u n c t i o n of g a s k e t m a t e r i a l and k n i f e - e d g e included angle. KNIFE-EDGE ANGLE
Nt/cm
70 ° . .
Aluminum
1788
1021
70 ° .,
Copper
4169
2381
90 °
Aluminum
2022
1155
90 °
Copper
4717
2694
,,
•
SEALING FORCE
GASKET MATERIAL
lbs.jinch
There are many applications where it is not feasible to use ConFlat ® flanges. For example, many particle accelerators have high radiation backgrounds in certain regions, and for prolonged periods after operation.(6 3 ) Securing or disassembling ConFlat @ flanges is a time-consuming process, and might result in unnecessarily prolonged exposure of personnel to background radiation (i.e., a safety issue). We would instinctively want to resort to the use of o-ring sealed flanges rather than ConFlat @ flanges in such instances. However, because of the high radiation background during operation, elastomer seals would harden and fail with time.( 8 9 ) Because of this, other metal flange sealing methods have been developed. Also, it is necessary at times to electrically isolate flange pairs from each other. For example, induced eddy currents in beam pipes of pulsed-magnet accelerators must be avoided, as the currents would cause opposing or "bucking" magnetic fields as well as cause beam pipe heating. This problems is solved by electrically "floating" the chambers from ground and each other. An example of such an application is shown in Fig. 1.19.4, which illustrates one of the 24 sectors of the Alternating Gradient Synchrotron (AGS) at the Brookhaven National Laboratory. ~-
._j._.,~¢"
I~
[I"
MAGNET
SECTOR ~--ISOLATION VALVE
-~d,
X._~
ENAMZL
~
COATED - ~
MAGNET
ROUGHING ~ VALVE ~
~
__L_
EhECraODE
MAGNET
~r')~.. _..k~ I \. SEC'FOR PIRANI de ~
~ '~t'-~
PENNING GAUGE
/
MAGNET
"1
HYBRID IMAGE _~ CURRENT 'PASSING CIRCUIT
.._...2..__
Figure 1.19.4. S c h e m a t i c r e p r e s e n t a t i o n
of an AGS v a c u u m sector.
The requirement for flange seals in the AGS included: 1) electrical insulation between flange pairs; 2) sealing and operational reliability; 3) immunity to radiation; 4) quick assembly or disassembly; and, 5) seal tolerance to high humidity environments. The initial solution to this sealing challenge was the use of indium coated Inconel c-rings. This proved most satisfactory when the c-rings were new, and freshly coated with In. However, the In coating tended to corrode in high humidity environments (even when stored in sealed plastic bags; a shelf-life problem). Also, it is believed that the corrosion process in instances caused hydrogen embrittlement of the Inconel base material. The corrosion at times resulted in leaks spontaneously opening in
68 BASIC THEORY nel base material. The corrosion at times resulted in leaks spontaneously opening in the seal, and the embrittlement the spontaneous cracking of the Inconel c-rings through the minor diameter, with catastrophic consequences. Because of this reliability problem, alternate sealing methods were explored and tested. This included testing many of the metal seals listed at the introduction of this section. The minor dimensions of some seals which were tested are illustrated in Fig. 1.19.5. Ineonel X750, ..... 0.1mm X 904 0.4ram Spring ___j,,~,,z "Diamonds" :3 5.0mm *
"".,, ~......
AI or Cu
,, - -7~
A.
"~)" 90~"~", omm
l i~4
600 eonelLining
Helicoflex Delta'.1~
Inconel Spring .~...........
...
j
I-- Jacket ~5.0mm ~21~ + ~'"' ~k~/.~ ~ Ineonel ~ Lining F C. Aluminum-Jaekted Seal.
--~ ~t___ 1 ~ 2.5 mm
' '
B. 1100 Aluminum "Diamond". Inconel Spriig . ... \ ~- 5.0ram ~
~
Lead .... Coating
~ ~
", Copper 1 Lining D. Lead Covered C-Ring.
Figure 1.19.5. Minor dimensions of metal seals tested against stainless steel, aluminum and enamel-coated surfaces. We sought to quantify the required sealing force to achieve vacuum tight seals for various seal configurations and for fiat sealing surfaces of stainless steel, aluminum and enamel-coated stainless steel. This comprised the mix of flange materials use in the AGS. The enamel coating is used on flanges to electrically isolate one flange from another, and the clamp used to secure a flange pair.( r s ) Metal test parts were machined to a ---32 micro-inch finish, and were free of nicks and scratches. In the test sequence, the seal plates were loaded with a hydraulic ram, and the seals leak checked with a leak detector having a sensitivity of - 2 x 10-1 o Torr-Z/s He. Any indication of a leak constituted a seal failure. Results of these tests are shown in Table 1.19.2. After each test, the load was removed from the flange pair. However, flanges were not parted or radial indexing changed between each test, so repeated sealing of the same seal does not indicate the seals are reusable, though the HelicoFlex Delta ® seal proved to be reusable - 8 5 % of the time, in other tests.(6 1 ) Flanges were usually secured in the AGS using Marmon clamps. The upper limit of acceptable seal force was set at 1750 Nt/cm, based on limitations of the seal damps. The lower seal force test range was arbitrarily established. Without exception, the HelicoFlex Delta ® seal sealed the first time with a force per unit length of seal of < 438 Nt/cm (i.e., -250 lb./inch). This sealing force is about 10% of the force required when using Cu-gasketed ConFla~ flanges.
Table 1.19.2. Force required for various vacuum seal configurations to achieve vacuum tight seals between stainless steel, aluminum and enamel coated surfaces.*
*Leak detector sensitivity (MDS): 2 x 10-lo ~ o r r - L / S .
70 BASIC THEORY This suggests that the seals would be most suitable for use in o-ring grooves of diffusion pumps, turbomolecular pumps with aluminum flanges, or cryopumps. The stress/strain data of the various seals would be beneficial in making the selection of the proper seal sizes. These data, for the seal configurations noted, are given in Fig. 1.19.6. 1.0
I
I'
~
I
I
-
I
1
~
1
I
I
Aluminum /
1.0 -
f
~ .s
~'~ [ [
I
t
I
J
f"c 0.15
Cu C-Ring w/ Spring-
Deltai~l
- 0.10 ~
0.5
0.05 _-
0
0
/
~.__~~r~~
0.5
1.0
A1 J a c k t e d
_
1.5
2.0
c ~
0
Sealing Force - kiloNew-t~ons/cm Figure 1.19.6. Seal Deflection as a F u n c t i o n of Sealing Force. (Right Scale for "Diamond"; Left Scale all Other Seals} Sputter-ion pumps (SIPs) used in most applications are supplied with ConFlat @ flanges. Use of such flanges assures their suitability for UHV application. Also, manufacturers typically bake out the pumps at temperatures of 350-400* C prior to shipment to the customer. The disadvantages in the use of these flanges in high radiation regions have been noted. Also, it is difficult to electrically isolate flange pairs. Therefore, some other form of electrical isolation would have to be provided were this feature needed at or near the pump flange joint. A scheme had to be developed that facilitated: i) the manufacturer baking the pump at the usual temperatures prior to shipment (a good quality assurance provision); ii) provision for use of an all-metal seal on the pump; and, iii) a simple (i.e., low cost) means of electrically isolating the pump from the beam chamber. During the construction of SLAC, all of the many miles of vacuum manifolding was cleaned, welded, leak checked and baked at 350* C at the manufacturers facility. The manifolds were terminated with peal-off flanges, to which were brazed Cu pinch-off tubes for sealing the manifolds after bakeout and during shipping. A similar scheme was adopted with the hundred or so SIPs used in the AGS. This is illustrated in Fig. 1.19.7. The pump was provided with a peal-off flange, with pinch-off tube, was removed with a pair of pliers, the flange touched up with a file, and the pump thereafter installed by mating it with an enamel-coated flange. The enamel coated flange material has been shown to have very low outgassing rates, and good dielectric strength, promising wider future applications.(7 o )
BASIC THEORY 71 WELDED FLANGE WITH PINCH-OFF WHEN PUMP RECEIVED FROM VENDOR
~xxxxx×xxx'xx'-,x~
/---/--TIG WELDED PEEL-OFF JOINT Figure
AGS BEAM HELICOFLEX® SEAL CHAMBER DELTA SEAL RETAINER FLANGE \ (COATED) "
..._
..,¢,-
-1..-
~
--
LpSuPUTTEFR~INOGNE/
/ FOUR-SEGMENT MARMON CLAMP
1.19.7. P u m p a n d f l a n g e as p r o v i d e d v e n d o r w i t h p e e l - o f f TIG weld j o i n t .
by
1.20 Comments on Helium Leak Detection A wide variety of helium leak detectors (LDs) are manufactured and sold throughout the world. One need only look on the web sites of the Society of Vacuum Coaters (SVC), the AVS (formerly the American Vacuum Society), or the Association of Vacuum Equipment Manufacturers International (AVEMI) to appreciate the availability of this equipment, and the numerous companies manufacturing leak detectors. These companies will be happy to explain the features of their equipment. Therefore, this section will primarily discuss the theoretical aspects associated with the use of these LDs to leak check vacuum systems, and provide some understanding of their use in this application. All conventional helium leak detectors use mass spectrometer tubes, similar to that illustrated in Fig. 1.18.6.
1.20.1 System Applications Two considerations are important when using a LD to check for leaks in a vacuum system: i) the response time; and, ii) the realized sensitivity. We return to equations (1.4.3) or (1.4.6) as the starting point. That is, at constant temperature: P d V/ dt + VdP I dt = ~ T dn l dt
(1.20.1)
The first term on the left is merely the value P x S, and the term to the far right we call throughput and represent by the symbol Q. Plugging these expressions into (1.20.1) yields the expression: PS + V~'l& = Q.
(1.20.2)
Assume that we have a system such as illustrated in Fig. 1.20.1, and that we have the option of attaching a leak detector at either point "A", "B" or "C" in the system. The figure suggests that the system is pumped with a momentum transfer pump (i.e., a diffusion or turbomolecular pump), but much of the discussion below will also apply
72 BASIC T H E O R Y to systems pumped with capture pumps (e.g., cryopumps or sputter-ion pumps). VACUUM VESSEL MANIFOLD & VALVE V1 OF CONDUCTANCE C t ~\ HIGH VACUUM PUMP WITH SPEED Sp \ FORELINE, VALVE V 2 \ & FOREPUMP \
Figure 1.20.1. Possible leak detector when leak
OF VOLUME V _/PPG ~ ~,. .... ( - ~ ' ~ -
",, " \ ~ - - ~ ~ ~ ~ ~'x ~ " t
. ......... ~
~
VALVE /- V a
4k,-~-~. . . . . . .
" -~i ~. . . . . . . . . . "~
locations eheeking
for attaching a a vacuum system.
....
Let's first disregard the type of high vacuum pump used on the system, and assume that the system has an air leak o f - 3 . 7 x 10-6 Torr-Z/s. Invoking (1.12.8), if He gas was substituted for air at the leak entrance, an air leak this size would correspond to a He leak of Q 0 = 10" s Torr-Z/s. This is a very large leak in a high vacuum system. Note, when we quantify the size of a leak for any gas we assume that on one side of the leak is an atmospheric pressure of that gas, and a vacuum exists on the other side of the leak. Assume that when we leak check the system, we conduct four tests: we throttle V 1 (i.e., valve V 1 ) so as to vary the speed at the chamber provided by the high vacuum pump to the four discrete values of 0, 10, 100 and 2000 Z/s. Assume that the chamber has a volume, V, of 1000 ~. Our leak checker is moved around the factory on wheels. Therefore, when we couple the leak detector up to the system at point "A", we must use a long hose, and the conductance of this hose for He is such that the He speed delivered to the vessel by the leak detector is 2 Z/s. What would the He pressure be in the vessel as a function of time were we to spray He on the leak at time t = 0 seconds, and for the four cases? Remember, because of the superposition law (e.g., see Section 1.14), we can forget about all the other gases in the system, and concentrate on what He is doing in the system. We can find the He pressure as a function of time by solving (1.20.2) for the rise in pressure in the chamber on spraying the leak with He. The solution is:
Qo
where
PHe(t)
= - - - - + c × e -St/V S
PHe(t)
= the He partial pressure in the vessel with time, = the constant He leak rate into the vessel in Torr-Z/s, = the speed at the vessel due to both the system and leak detector pumps, in Z/s, = some constant, = the volume of the vessel (i.e., 1000 Z).
ao
S C
and
V
(1.20.3)
BASIC T H E O R Y 73 Prior to spraying He on the leak, the only other source of He near the system leak is that in found in the atmosphere. Under normal conditions, the partial pressure of He in the atmosphere is - 4 x 10-a Torr (see Table 1.8.1). Even with a system pumping speed of only 100 Z/s, the partial pressure of He in the system due normal atmospheric conditions would equilibraite at --5 x 10" 1 1 Torr. Let's consider this to be a negligible He background. Therefore, if we plug in the initial condition of PHe(t), just before spraying He on the leak (i.e., Pile(0) = 0), we find that the value c = - Q o / S, and (1.20.3) has the solution:
ao PHe(t)
=
S
[ 1 - e -St~V]
(1.20.4)
If we multiply both sides of (1.20.4) by S, we find: Q(t)
= Q o [1- e "St~V]
(1.20.5)
or, by dividing both sides of (1.20.5) with S, we find: e(t)
= e o [1- e "St~V]
(1.20.6)
Prior to continuing, it will help to understand the meaning of some of the terms in the above three equations. The value of Q(t) shown to the left of (1.20.5) is the rate at which He is being consumed by the pump, whereas the value Qo is the constant rate at which He is being introduced into the system. Initially these two do not have the same value, as not all of the He being introduced through the leak is instantaneously consumed by the pump; some of the introduced He appears as free gas in V, the vessel volume. Therefore, the larger the volume, the longer it takes for the values of the two throughputs to converge. Similarly, the value P o in (1.20.6) is the maximum He pressure which will eventually be observed in the vessel. If the vessel is large, it will take longer for PHe(t) to converge on (i.e., become equal with) P0" Also note that small values of S will extend the convergence time. The various leak checking scenarios we have postulated are given in Table 1.20.1 Also, the values of PHe(t) in the vessel as a function of time, and for the four cases, are shown in Fig. 1.20.2. The response time when leak detecting is arbitrarily defined as the time when the He signal reaches --63% of its maximum signal level. This occurs when time t increases such that the value S t / V = 1, in the above equations. When this happens e "St/V -0.37 and the value in the brackets is the stated 0.63. The value V/S is called the leak checking or system time constant, r . For example (1.20.5) could be rewritten: O(t) = Q o [1- e -t/r ] (1.20.7) When t = 7", this is called one time constant; when t = 2 r , this is called two time constants, etc.
74 BASIC T H E O R Y T a b l e 1.20.1. Leak d e t e c t o r (LD) r e s p o n s e t i m e s a n d maximum sensitivity for four s y s t e m c o n f i g u r a t i o n s . * Case Sp S p e e d No. L/s** 1
'0
2 3 4
10 100 2000
LD
Speed
L/s**
LD He Qmax Vessel He Torr-L/s Pmax T o r r
Response Time-sec 1000 83.3 9.8 0.99
1.0
x
5.0 8.3 9.8 5.0
10 -5
1.7 x 10 -6 2.0 x 10 -7 1.0 x 10 -8
x x x x
10 -.6 10 -7 10 -8 10 -9
* H e l i u m l e a k r a t e 1.0 x 1 0 - 5 T o r r - L / s . **Speed d e l i v e r e d to v e s s e l . 10
-5 1
©
I
Case
Q
I
[ II
10
10
~,~$1-
I II
[ l'
2 L/s
)
o oCo
©
o
O x x X xxw: x x
"s2-
-7
12 A
~,
S3 -
102 ' L / s
-8 A
• ,,-,i
|
|
S 4 -- 2 0 0 2
c~ 10
I I'
1
d
No • -"
-6
r/?
10
I
-9
I
[ [1
10 0
t
[ [1
10 1 Time
of H e l i u m
t
I tl
10 2 Leak
L/s I
10 3 -
lit 10 4
seconds
F i g u r e 1.20.2. P a r t i a l p r e s s u r e of h e l i u m in t i m e in a 1 , 0 0 0 l i t e r v e s s e l w h e n i n t r o d u c i n g a helium l e a k of 10 - S T o r r - L / s a t t i m e t - O. Some very important conclusion are reached regarding leak checking when studying the above table and figure. These include: 1. When simultaneously pumping on the vessel with a large (e.g., 2000 ;t/s) pump as well as the leak detector pump, the total He signal reaching the leak detector decreases in direct proportion to the size of the two pumps. For example the maximum He signal (i.e., throughput) to the leak detector eventually equilibrates at Qmax = 1.0 x 10 -s Torr-Z/s when pumping on the vessel with only the leak detector (i.e., Case 1). However, in Case 4, the maximum He throughput to the leak detector is Qmax = 1.0 x 10 -s Torr-Z/s. This means the realized sensitivity for leak
BASIC THEORY 75 checking has been diminished by a factor of 1,000 as a consequence of the simultaneous use of both pumps. 2. The time it takes the He partial pressure in the vessel to maximize decreases with increasing pump speed. In fact, as shown in Table 1.20.1, response time decreases with increasing system pumping speed. 3. Observations i and 2 above graphically illustrate what is called the response-
time~sensitivity trade-off when leak checking anything.
4. Note that the terms within the brackets of (1.20.4) - (1.20.6) contain no information about the system pressure (or throughput), other than how their magnitudes change in time. 5. The amplitudes of the P(t) curves shown in Fig. 1.20.2 are directly proportional to the introduced leak rate. Therefore, if for example, in Case 4 the He leak rate was 10-6 Torr-Z/s rather than 10-5 Torr-Z/s, the He Qmax into the leak detector would be 10-9 Torr-Z/s. Also, the maximum vessel He pressure, Pmax would be 5 x 10"1 o Torr. This tells us something about the sensitivity which would be required of the PPG shown attached to the vessel in Fig. 1.20.1, were this PPG used for leak checking the system. 6. Were the leak detector hooked up to point "B" rather than point "A", the Case 2-4 curves in Fig. 1.20.2 would be very similar, except that their magnitudes would be less due to the pressure drop caused by conductance C t. The Case 1 P(t) curve would be identical to that observed at point "A". 7. In most applications it is best if all of the He introduced through the leak can be pumped by the leak detector. But, we saw in Case 1 that this sacrificed system response time when it was attached to point "A". What would happen if we attached the leak detector to point "C", keeping both V 1 and V 2 wide open in the process. Perhaps because a high total gas throughput into the forepump, we are not able to close V 2 completely. If we throttle open the valve at point "C" so that the LD is sampling only 10% of the He throughput stream leading to the forepump, the He signal at the LD will equilibrate at 1.0 × 10-6 Torr-Z/s,, and the response time of the system will be that of Case 4! In fact, if we can totally close V 2 and let all of the He throughput stream into the leak detector, we will have the best of both worlds: the response time of Case 4, and the He leak realized sensitivity of Case 1. Lastly, if we differentiate (1.20.6) with respect to time we will obtain a summary equation for how the He signal response time will vary as a function of S, 1,I,and Q o" That is:
e(t) dt
=
Po
s V
e'St/V
(1.20.7)
This equation clearly shows that the rate of increase in system pressure (i.e., the He signal) is directly proportion to the speed of the pump, and inversely proportional to the system volume. Note that the product P o × S, is merely Q o, the magnitude of
76 BASIC THEORY the He leak rate. That is, a larger leak means a faster response time.
1.20.2 Leak Checking Systems Appended With Capture Pumps Contrary to common belief, both sputter-ion pumps and cryopumps have very high initial pumping speeds for helium. However, both have the drawback that as helium is pumped the base pressure of the pump will rise in time, rather than remaining at a fixed value. This doesn't happen with momentum transfer pumps with high compression ratios. This background pressure rise in time, called a memory effect, is quantitatively discussed for sputter-ion pumps in Chapter 2, and for cryopumps in Chapter 5. As the base pressure of a specific gas for any pump increases in time, the partial pressure of that gas will also increase in time if the gas throughput is held constant. For example, Fig. 5.4.4 shows the pumping speed of a 200 mm cryopump for He and H 2 . If the initial speed is -4,000 Z/s for H 2 at 10-6 Torr the same pump will have a speed of -2,000 Z/s for He at this pressure. However, note how the speed for He "quickly" decreases as more of this gas is pumped. In this test, the He Q(t) had to be continually decreased in time to maintain a constant He pressure of 10-6 Torr. After pumping only -800 Torr-Z of He, the speed of the cryopump at 10-6 Torr decreased to zero. This means that if more He gas is introduced, the He system pressure will creep up in time. This same effect occurs with sputter-ion pumps, these pumps having the disadvantage that they can't be regenerated, and the He background will remain high for all time. If the system He pressure is creeping up in time as a consequence of spraying He on a leak, the background He pressure will f'LXthe lower limit of leak detection sensitivity. For example, after pumping ---10 Torr-Z of He with this same cryopump the background pressure of He will be --10 -9 Torr; 100 Torr-Z of He will lead to a background pressure of 10"a Torr, etc. Therefore, when leak checking any system pumped with capture pumps it is advisable to "valve out" the pump, and leak check the system at the equivalent of point "A" of Fig. 1.20.1. If under these conditions the throughput of all the system gas is so high that the LD can't handle the gas load, it is advisable to use some sort of auxiliary ("dry") pump in parallel with the leak detector. Remember, if 90% of the gas is pumped by this auxiliary pump you have decreased the realized sensitivity of your leak detector by only 10%, etc. Lastly, if you find no leaks in the system with the capture pump "valved out", you may have the nagging suspicion that a leak in fact exists in the capture pump. If you wish to leak check the capture pump, turn it off and leak check it as a passive volume. There are no "short cuts" in this regard.
Problem Set 1) Construct a table of the room temperature root-mean-square velocities of H2, He, Ne, H2 O, N2, CO and CO 2. 2) What is the vapor pressure of He, H2 and Ne at 10" K? 3) The vapor pressures of all other gases are less than or equal to what value at 20* K?
BASIC T H E O R Y 77 4) Assume that the inner surface of a box, with a height, width and length of 10 cm, is covered with one monolayer of gas, and that the pressure within the box is 10" 8 Torr. What would be the pressure if the temperature of the box was 300 ° K and the surface gas was liberated into the volume of the box? 5) Prove the last sentence starting at the bottom of page 27. (Hint! A legitimate proof might be to first claim that the sentence is incorrect, and then reach a valid contradiction.) 6) What is the ratio of the thermal conductivity of H2 vs. air at room temperature, at a pressure of 10 -2 Torr, assuming equivalent values for ~, and ). for the two gases? 7) Assume that the room temperature partial pressure of H 2 0 in your vacuum system is 10 "e Torr. Assume that there is a unity sticking coefficient for this gas on some surface within the system. How long would it take to build up one monolayer of H 2 0 on the surface? Under the same conditions, how long would it take if the gas was H 2 ? Xe? 8) Assume that each time the piston in Fig. 1.10.1 is fully withdrawn, the volume above the piston is back-filled to a pressure P. The piston sweeps out the volume at the rate of/x V/At. Construct a table of throughput vs. pressure for P = 0, 10-4,10-3, 10-2 Torr, and values of/x V/At = 1, 2, 100 and 10 3 liters per second. 9) Construct a table of the minimum conductances of an aperture, with a diameter of 1.0 cm, for the gases H 2, He, H 2 O, air, N2, CO and CO 2 at room temperature. 10) Construct tables of the molecular conductances of a round manifold with diameter D, and length L, where D = 5, 10 and 20 cm and L = 242 cm, for the gases in the previous problem, but at 400* K. 11) Calculate the molecular flow conductance of a rectangular tube for room temperature air, where a = 10 cm, b = 25 cm and L = 250 cm. What is the conductance of a round tube under these same conditions, where the diameter of the tube is D and 7rD 2/4 = ab? 12) Assume the diameter of the manifold in Fig 1.13.3 is 15 cm and the length of the manifold between the chamber and pump flange is equivalent to 200 cm. Assume that H 2, H 2 O and CO are outgassing from within the chamber because of some process, but not from other surfaces with the vacuum system. If the speed of the pump for H 2 , H2 O and CO2 is 2000, 4000, and 1200 Z/sec, respectively, what are the partial pressures of these gases at the pump flange and in the chamber, if the outgassing rate of each is 10 -6 Torr-Z/sec? 13) The same conditions exist as in the previous problem. However, the manifold between the pump and chamber has an H 2 O outgassing rate of q = 10 -g Torr-Z/sec/cm2. What would the chamber and pump pressures be for the three gases? 14) A long manifold of diameter D and length L has a roughing pump attached to one end and a sputter-ion pump attached to the other end. Most of the bulk gas has been evacuated from the manifold by the roughing pump. We want to eventually turn on the sputter-ion pump, but we must wait until the pressure at sputter-ion pump reaches a value of P(L,t). The manifold has an outgassing rate of q(t)/cm2. The roughing pump \
78 BASIC THEORY has a speed of S r. Express the length of the manifold between the two pumps as a function of Sr, q(t), P(L, t) and D. 15) Assume that ~Qi = Q0 t'l + Q1 t'° .5 in (1.16.7), where Q0 and Q1 are constants. Solve (1.16.7) for the general expression of P(t). 16) You are conducting H 2 speed measurements of a sputter-ion pump using a Fischer-Mommsen Dome. Assume that the speed of the pump for H2 is 150 Z/sec and for Ar it is 10 Z/sec. The diameter of the dome aperture is 1.5 cm. Unknown to you, your bottle of H2 has been contaminated with 1.0% Ar. What is the percent error in the speed measurements as a consequence of the presence of Ar? Remember to take into account the relative gauge sensitivities for H 2 and Ar. 17) You are calibrating a quadrupole gas analyzer. You have a bottle containing a high-pressure mixture of several different gases, each of known partial pressure. A variable leak valve is connected between the bottle and the upper chamber of the modified CERN Dome shown in Fig. 1.16.6. You have replaced the BAG, normally used with the dome, with your gas analyzer. Assume that CA << Sp for all the gases, and that the rate at which the gases pass through the variable leak is proportional to (Mx)'~, where M x is the molecular weight of each gas species. Prove that the ratio of the partial pressures of each of the gases introduced into the dome will be the same as their partial pressures in the bottle. 18) Solve for the values of "A" and "B" for each of the gases listed in Table 1.18.1. 19) Assume that the aperture used when introducing the gas mixture into a test dome such as illustrated in Fig. 1.18.12 was sized so that the speed delivered to the dome for nitrogen was 0.25 Z/s. Assume that the introduced gases were mixed in proportions such that the partial pressure of oxygen should have been the same in the dome as that of nitrogen. What is the oxygen pumping speed of the QRGA used to create the spectrum Fig. 1.18.107
References 1.
2. 3. 4. 5. 6.
Kaminsky, M.S., Lafferty, J.M., Dictionary of Terms for Vacuum Science and Technology, Surface Science, Thin Film Technology, Vacuum Metallurgy, Electronic Materials (American Institute of Physics, New York, 1980). Guthrie, A., Wakerling, R.K., Vacuum Equipment and Techniques (McGraw-Hill Book Company, Inc., New York, 1949). O'Hanlon, J . F . , A User's Guide to Vacuum Technology (John Wiley & Sons, Inc., New York, 1980). Halliday, D., Resnick, R., Physics For Students of Science and Engineer(John Wiley & Sons, Inc., 1963), p. 516 ff. Sears, F.W., Thermodynamics (Addison-Wesley Publishing Company, Inc., London), p. 223 ft. Honig, R.E., Hook, H.O., "Vapor Pressure Data for Some of the More Common Gases", RCA Review, 21, 360 (1960).
BASIC THEORY 79 7.
Haefer, R.A., Cryopumping, Translated by J. Shipwright, and R.G. Scurlock, (Clarendon Press, Oxford, 1989). Simon, F.E., Kurti, N., Low Temperature Physics (Pergamon Press, Ltd., London, 1952), p. 32 ft. 9. Benvenuti, C., "Characteristics, Advantages and Possible Applications of Condensation Cryopumping," J. Vac. Sci. Technol. 11, 591 (1974). 10. Barron, R., Cryogenic Systems, (McGraw-Hill Book Company, New York, 1966), p. 65 ff. 11. Butcher, S.S., Charlson, R.J., An Introduction to Air Chemistry (Academic Press, New York, 1972), p. 4. 12. Dushman, S., Scientific Foundations of Vacuum Technique (John Wiley & Sons, Inc., New York, 1949), p. 44 ft. 13. Dushman, S., o_12.ci_.[t.,p. 17 ft. 14. Pamplin, B.R., Bachrach, R.Z., Editors, Aspects of Epitaxs', Vol. 12 (Pergamon Press, London, 1986), p. 45 ft. 15. Leach, D.P., Basic Electric Circuits (John Wiley & Sons, Inc., New York, 1969). 16. Welch, K.M., "The Pressure Profile in a Long Outgassing Vacuum Tube", Vacuum, 8, 271, 1973. 17. Hablanian, M.H., "Recommended Procedure for Measuring Pumping Speeds", J.Vac. Sci. Technol. A5(4), Jul/Aug, 2552 (1987). 18. Rutherford, S.L., "Pumping Speed Measurements on Sputter-ion Pumps", Vacuum, 16, 643 (1966). 19. Tom, T., Munro, D F., "Speed Measurements of Ion Getter Pumps by the 'Three-Gauge' Method", Trans. of 3rd Int. Vac. Cong., 1965, (Pergamon Press, London, 1966), p. 377. 20. Brombacher, W.G., A Survey of Ionization Vacuum Gauges and Their Characteristics, NBS Technical Note No. 298, 1967. 21. Redhead, P.A., Hobson, J.P., Kornelsen, E.V., The Physical Basis of Ultrahigh Vacuum, (Chapman and Hall, Ltd., London, 1968), p. 336 ft. 22. Fischer, E., Mommsen, H., "Monte Carlo Computations on Molecular Flow in Pumping Speed Test Domes", Vacuum, 17, 309 (1967). 23. International Organization for Standardization, ISO/TC 112/SC 3(Sec-20)44, April, 1970. 24. American Vacuum Society Standards (tentative), AVS 4.1-4.8, 4.10, J. Vac. Sci. Technol., 8, 664 (1971). 25. Welch, K.M., Flegal, C., "Elements of Cryopumping", Industrial Research~evelopment, March 1978. 26. Denison, D.R., McKee, E.S., "A Comparison of Pumping Speed Measurement Methods", J. Vac. Sci. Technol., 11(1), 337(1974). 27. Ball, R.W., Principles of Abstract Algebra (Holt, Rinehart, and Winston, New York, 1963), p. 31. 28. Wells, O.C., et al., Scanning Electron Microscopy (McGraw-Hill Book Company, New York, 1974), p. 94 ff. 29. Spangenberg, K.R., Vacuum Tubes (McGraw-Hill Book Company, Inc., New York, 1948), p. 24 ft. 30. Komiya, S., Sato, H., Hayashi, C., "Triggering Sputter Ion Pump in Extreme High Vacuum", Proc. 13th Nat. AVS Syrup., 1966 (Herbick and Held Printing Company, Pittsburgh, Pennsylvania, 1967), p.19. 0
80 BASIC THEORY
References 31. Downing, J.R., Mellen, G., Rev. Sci. Inst. 17, 218 (1946). 32. Cobine, J.D., Gaseous Conductors (Dover Publications, Inc., New York, 1958), p. 144 cf. 33. Bruining, H., Physics and Applications of Secondary Electron Emission (Pergamon Press, London, 1954). 34. Colligon, J.S., "Ion Bombardment of Metal Surfaces", Vacuum, 11, 272 (1%1). 35. Cobine, J.D. o_0.12,cit., p. 162. 36. Thomson, J.J., Thomson, G.P., Conduction of Electricity Through Gases Vol. II (Dover Publications, Inc., New York, 1969), p. 487. 37. Tate, J.T., Smith, P.T., "The Efficiencies of Ionization and Ionization Potentials of Various Gases Under Electron Impact", Phys. Rev. 39, 270 (1932). 38. Rapp, D., Englander-Golden, P., "Total Cross Sections for Ionization and Attachment in Gases by Electron Impact. I. Positive Ionization", J. Chem. Phys. 43(5), 1464 (1%5). 39. Redhead, P.A., Hobson, J.P., Kornelsen, E.V., o_12.cit, pp. 253 - 366. 40. Hultzman, W.W., "Characteristics and Performance of Several Mass Spectrometer Residual Gas Analyzers", National Aeronautics and Space Administration, Lewis Research Center, Cleveland, Ohio, Report No. NASA TN D-7554, 1974. 41. Conn, G.K.T., Daglish, H. N., "Cold Cathode Ionization Gauges", Vacuum 3(1), 24 (1953). 42. Hablanian, M.H., High Vacuum Technology, A Practical Guide, 2nd Edition (Marcel Dekker, Inc., New York, 1997), pp. 137 - 318. 43. Clausing, P., Physica, 9:65 (1929). 44. Guthrie, A., Wakerling, R.K., Vacuum Equipment and Techniques, (McGrawHill Book Company, Inc., New York, 1949), pp. 35-37. 45. Chambers, A., Fitch, R.K., Halliday, B.S., Basic Vacuum Technology (Adam Hilger, New York, 1989), p. 43. 46. Neal, R.B., "The Stanford Two-Mile Accelerator", J. Vac. Sci. Technol 2(3), 149 (1%5). 47. Ozaki, S, "RHIC Project", IEEE Particle Accelerator Conference, Accelerator Science and Technology 5(5), 2901(1991). 48. Hsueh, H.C., et al, "Measuring and Locating Internal Leaks in the RHIC Insulating Vacuum System", Presented at the 46th International Symposium of the American Vacuum Society, Seattle, Oct. 1999. 49. O'Hanlon, J.F., A Users Guide to Vacuum Technolo_~¢, 2nd Edition (John Wiley & Sons, New York, 1989), pp. 458-463. 50. O'Hanlon, J.F., A Users Guide to Vacuum Technolo_ff~/, 2nd Edition (John Wiley & Sons, New York, 1989), pp. 134-136. 51. Alpert, D., Buritz, R.S., "Ultra-High Vacuum. II. Limiting Factors on the Attainment of Very Low Pressures," J. Appl. Phys. 25(2) 207(1954). 52. Melissinos, A.C., Experiments in Modern Physics (Academic Press, New York, 1966), pp. 467-471. 53. Welch, K.M., et al, J. Vac. Sci. Technol. A17(5), Sep/Oct 1999.
BASIC THEORY 81
References 54. Filippelli, A.C., Vacuum Design of Advanced and Compact Synchrotron Light Sources,/kiP Proceeding No. 171 (ALP, New York, 1988), p. 240. 55. Fremerey, J.K., "The spinning rotor gauge," J. Vac. Sci. Technol. A3(3) May/June 1985, p. 1715. 56. Looney, J.P., NIST; private communications. 57. Wheeler, W.R., Carlson, M., Trans. 8th Nat. AVS, 2nd Int. Vac. Cong., p. 1309, Pergamon Press, Oxford (1962). 58. Wheeler, W.R., Trans. 10th Nat. AVS Symp., 159(1963). 59. Roth, A., Vacuum Sealing Techniques (Pergamon Press, London, 1966) pp. 308 ft. 60. Ishimaru, H.J, J. Vac. Sci. Technol., 15, 1853(1978). 61. Welch, K.M., et al., Vacuum, 41(7-9), 1924-1927(1990). 62. HelicoFlex Delta ® Seal is a registered trademark of Cefilac, S.A., located in Saint-Etienne, France. The product is also manufactured and sold by Garlock Helicoflex, Located in Columbia, South Carolina. 63. The Alternating Gradient Synchrotron (AGS) at BNL (Brookhaven National Laboratory or Protron Synchrotron at CERN (European Organization for Nuclear Research) are typical examples. 64. Schram, A., LeVide No. 103, 58(1963). 65. de Csernatony, L., Crawley, D.J., Vacuum, 17(10), 551(1967). 66. Dayton, B.B., Transactions of 1959 American Vacuum Society Syrnposium (Pergamon Press, London, 1960), pp. 101-119. 67. Perkins, W.G., J. Vac. Sci. Technol. 10(4), 543(1973). 68. Yoshimura, N., J. Vac. Sci. Technol. A 7(1), 110(1989). 69. Peacock, R.N., J. Vac. Sci. Technol., 17(1), 330(1980). 70. Biscardi, C., Hseuh, H-C, Mapes, M., J. Vac. Sci. Technol. A 18(4), 1751(2000). 71. Kurokouchi, S., Morita, S., J. Vac Sci. Technol. A 19(2), 693(2001). 72. Welch, K.M., J. Vac. Sci. Technol., 13, 498(1976). 73. Biscardi, C., Hseuh, H-C, Mapes, M., J. Vac. Sci. Technol., A 18(4), 1751(2000).
CHAPTER 2
SPUTTER-ION PUMPING 2.0 Introduction Two types of sputter-ion, or electronic, pumps came into commercial use in the last few decades. The first of these, introduced on a commercial scale by the GranvillePhillips Company, was called the Electro-Ion pump, and was a simplified version of two varieties of pumps reported on earlier by Herb, et al., and referred to as an Evapor-ion or Orbitron pump.(1,2 ) A version of Herb's Evapor-ion pump was offered for sale for a brief time by Consolidated Vacuum Corooration. Leybold also offered an electrostatic pump for sale for a brief period.(s') These types of pumps were called electrostatic pumps because of the exclusive use of electrostatic fields in their operation rather than both electric and magnetic fields. They are referred to in the past tense, as they were soon overshadowed by a second type of pump of much simpler construction. For completeness, a bibliography of some of the salient publications dealing with electrostatic pumps is given in Appendix 2A. No further discussion is given on these pumps in this chapter. The main emphasis of this chapter will be on the second variety of sputter-ion pumps, first developed on a commercial basis by Varian Associates, and called by that company the Vaclon ® pump.(4 ) It is a sputter-ion pump which makes use of both electric and magnetic fields to achieve the pumping action of a low pressure, electrical discharge. These electrical discharges are referred to as Penning Discharges. The origin of all of these electronic pumps came from studies of various forms of vacuum ionization gauges and the phenomenon of sputtering of metals by gas ions; that is, observations of the somewhat annoying pumping effects of vacuum ionization gauges and gas pumping effects during sputtering started it all. Hundreds of technical papers have been published on low pressure Penning discharges in the last half-century. About 10% of these have attempted to theoreticaUy model the low pressure characteristics of Penning discharges. Redhead noted that none have had complete success in this effort,(s ) though some of these papers have combined empirical and theoretical results and met with reasonable success. Great care must be used in interpreting data of any one or two publications, toward the formulation of a unified low pressure Penning discharge model. This is because slight changes in cell geometry and symmetry, magnetic field and its uniformity, electric field, pressure and gas species can significantly alter the properties of the discharge. I will first qualitatively describe a Penning discharge, discuss its properties as they relate to varieties of ion pumps, and then eventually discuss some of the theoretical considerations of these discharges.
84
SPUTTER-ION PUMPING
2.1 The Penning Cell In 1934 Gaede published a review paper on various types of vacuum gauges with which he had been working.(S ) The emphasis of his work to that point was on the ref'mement of what came to be known as the Knudsen Gauge.(7 ) For completeness, in that paper he reported on experiments with gauges including an ionization manometer, configured as shown in Fig. 2.1.1. He recognized, as reported much earlier by Philips,( "8) that if one applied a magnetic field axial to the electrodes to which a high voltage was applied, at low pressures (i.e., ~; 10" z Torr) a stable electrical discharge would be initiated. He noted that the magnetic field tended to trap the electrons in the void between the electrodes, stating "... the emitted electrons (from the cathode) follow along the lines of the magnetic field (within the anode) and thus oscillate until they strike a gas molecule." Presumably the current "drawn" by the electrical discharge was noted to be linear with pressure. However, Gaede rejected this type of gauge for general applications. He noted that when the gauge was used, "... the gas pressure sank (i.e., decreased) rapidly about 30 parts as though a pump were attached." Gaede's gauge, though showing striking similarity to cold cathode gauges later described in the literature, was actually a hot filament gauge. That is, the filaments forming the cathodes of the gauge were thermal emitters. The two leads connected to the anode facilitated outgassing of this electrode by resistance heating. VACUUM SYSTEM GLASS BULB --~
POSITIVE ANODE WITH Y/~//~/; PROVISIONS FOR --~,~~~11 RESISTIVE HEATING
RESISTIVELY HEATED~ ~ ~ THERMAL CATHODES V / l i l Y
MAGNET SOLENOID
+ HIGH VOLTAGE
Figure 2.1.1. Gaede's m a g n e t i c a l l y confined discharge gauge.
Such hot filament gauges had been reported on decades earlier.( 9 ) Gaede's contribution was in using a magnetic field to intensify and confine the discharge. Had Gaede turned off the filaments, with sufficient magnetic and electric field, the discharge would have been sustained within the electrodes in a large pressure interval. Penning was also studying such discharges - that is, low pressure electrical discharges sustained by the use of electric and magnetic fields. I believe that Penning's initial work, rather than being related to gauging, was directed at using the sputtering action of these discharges to produce thin films. He submitted an application for a United States patent on such sputtering apparatus in 1936,( 1 o ~the same month he published a paper on the subject in Physica,(1 1 ) and he was awarded a German patent on such apparatus in 1935. In a review paper, Guentherschulze noted that this ion sputtering phenomenon had been reported on as early as
SPUTTER-ION PUMPING
85
1858.(x 2 ) However, Penning identified a practical application of this phenomenon. Two of the configurations claimed in Penning's patent are shown in Fig. 2.1.2. (The words claim and claimed are used in a legal sense in this book). In one, a hollow cylindrical tube served as the anode. Cathode plates of the desired sputtering materials - perhaps two different materials - were placed at each end of the cylindrical anode. The substrate, to be coated with the sputtered material, was placed as close to the anode cylinder as possible. At low pressures, ions created by the swirling cloud of electrons trapped within the anode volume bombard and sputter the cathodes. The selected cathode material is spewed onto the substrate located within the anode. CYLINDRICAL TUBE ANODE ~ ~ ~f--~ - " ~ - ~--~~___~ SUBSTRATES /\ f-/-~---~-~ TO BE f ~ "-~--~ ~ ' ~ COATED ~
WIRE ANORING ~z--- 4 c n ~ - ~ ~ ' -
~
M o To ~ HI~GH () OF ~(?EAcLT~T~I)~IALS - VOLTAGE + - VOLTAGE + Figure 2.1.2. Examples of two of the s p u t t e r - c o a t i n g cells noted in Penning's 1936 patent application.
In his patent application, Penning pointed out that as positive ions bombard the cathode surfaces, sputtering of these surfaces occurred. He referred to this sputtering as cathode disintegration. Though it was not included as a claim in the above patent, he noted therein that "Cathode disintegration may also be used for the purpose of reducing the pressure of a gas in a closed chamber. The cathode particles disintegrated combine with gas molecules and thus bring about a reduction in pressure." Penning experimented with numerous other configurations, including what later came to be known as magnetron and inverted magnetron sputtering devices. Later variations of the Penning cell, some radical departures in scale and configuration, were found to produce very satisfactory sputtered films (e.g., see t s, x 4, t s ). The original work by Penning proved to be the advent of the sputtering technology which is presently used to produce functional, decorative and semiconductor related thin films having applications in a wide variety of major industries. Penning did not report on using a hot cathode as a source of electrons in these crossed field discharges. However, he verified Gaede's findings regarding current vs. pressure in magnetically confined discharges. Because of the absence of a hot filament, these electrical discharges became known as cold cathode discharges. With the advent of the magnetron rf oscillator, they were later sometimes described as crossed field discharges. This is because, similar to the magnetron, the electric and magnetic fields applied to the discharge apparatus, or cell, are somewhat orthogonal to each other. The analogy between the Penning cell and a magnetron proved appro-
86
SPU'ITER-ION PUMPING
priate, as you will later see. Qualitatively understanding the properties of a Penning electrical discharge was the starting point of understanding the workings of ion pumps. In early 1937 Penning published an article characterizing several gauge configurations, one of which was a minor variation of the original Gaede ionization manometer.(1 e ) Of course it did not use a hot cathode f'dament. In this article he made mention that the pumping of gas occurs because of the electrical discharge. As a practical gauge and pump, these devices were first reported on by Penning in 1949.(x r ) He commented on the need to thoroughly degas the gauge configuration in order to get consistent (not accurate) results, as he noted that the gauge burnt away gas. The property of the gauge to burn away gas was used in a leak detecting scheme he reported on in this paper. He reported that "..., owing to the discharge, a certain amount of the gas disappears. In the state of equilibrium the amount burnt away per second by the discharge is obviously equal to the sum of that leaking in ...". The property of these electrical discharges to consume gas had been reported on decades earlier.(t a ) Four years later, in April 1953, Gurewitsch and Westendorp applied for patent on various forms of single-cell pumps, and the use of different cathode materials.(t 9 ) Rather than curse the darkness - that is the nuisance of the pumping effects of these gauges - they took the best that both Penning and Gaude had to offer and patented it as a pump! Their patent included various single-cell configurations, some including the use of a hot filament. They described the physics of the devices as " ... compared to that of a cylindrical anode magnetron operating near cut-off voltage." They claimed use of various types of cathode materials (i.e., Ti, C, Mg, AI and stn. stl.) and the use of a heated filament in some of their diode pumps, to aid in the build-up (i.e., ignition) of the space charge in the cell. They subsequently reported results on use of both the Ti and C cathodes, indicating that the use of Ti cathodes aided considerably in the pumping of hydrogen.(2 0 ) They reported speeds of -0.03 Z/sec for H 2. Incidentally, most magnetron oscillators are effective pumps. When working as an engineering associate at Raytheon, in the late 1950's, I conducted an experiment where I deliberately introduced a continuous flow of H 2 into a high power magnetron during high voltage testing. The magnetron continued to operate stably, at an H2 pressure of--10 -4 Torr for several hours. Arcing and missing pulses only occurred after removal of the H ~ leak. In 1955, Gale reported on speed measurements on a single cell gauge/pump used to append a high voltage generator.(2 1 ) He reported measured pump speeds ranging from 0.01 to 0.03 Z/sec for various gases. But, I'm getting ahead of myself
2.2 ]: +/P Characteristics in Penning Cells The linear relationship between pressure and ion current in Penning cells, at pressures ~ 10" 3 Torr, resulted in the first widespread interest in what then came to be known as a Penning gauge. For several decades it was also known as a Philips gauge, perhaps in honor of C.E.S. Philips,(8 ) or because Penning worked for the company N.V. Philips, located in the Netherlands. Over the years, technologists attempted to establish parameters, including the geometry of the cell, magnetic field and applied voltage, which would yield well behaved (i.e., constant and smooth) I + / P or sensitivity characteristics as defined by (1.18.17) of Chap. 1. Conn and Daglish published two classical empirical studies of
SPUqTER-ION PUMPING
87
Penning discharges of that era.(2 2,2 ~ ) In the second paper, published in 1954, they systematically determined the gauge sensitivity, s, as a function of pressure, P, magnetic field, 13, and voltage, V, of a number of Penning cell configurations, including those first described by Penning. Serious discharge instabilities were reported for various configurations (i.e., the current jumped about in certain pressure regions). Penning first reported seeing such instabilities at pressures < 10-4 Torr. He termed these as "small jumps" in indicated pressures, and reported smoothing them out in his published data.(1 7 ) The Penning cell characterization studies of Corm and Daglish were limited to pressures in the range of 10-s to 10-2 Torr. Their description of the visual and electrical characteristics of the discharges at various pressures is a tribute to their scientific technique. Their 1953 paper showed a remarkable insight into the mechanisms and characteristics of Penning discharges.(2 2 ) They noted at that time the similarity between these gauges and magnetrons, and even made measurements of the rf signals produced by the discharges. One very important finding of Conn and Daglish was the discovery of the cell configuration that later became known as the tn'ode. This triode configuration (potentials reversed) is shown in Fig. 2.2.1. It did not evidence the instabilities characteristic of the remaining configurations, termed diode configurations. Of course, no one at that time, including Conn and Daglish, recognized the significance of their triode discovery. Though showing figures and test results of several diode configurations in that paper, they omitted a figure of the triode, for which they gave only the test results. CATHODE PLATES +
ANODE RINGS
H Figure 2.2.1. The first "triode" Penning cell as reported by Conn and Daglish.
Jepsen was the first to point out the relationship of soutter-ion pump speed, S, to the '7 +/P" characteristics of the cells of the pump.(2 4 ) This was dramatically demonstrated in later work published by Rutherford.(2 s, 2 e ) He quantitatively showed the direct correspondence between S and I +/P of sputter-ion pumps of various cell sizes. Rutherford and others at Varian showed how I+/P tended to significantly decrease at lower pressures for a given cell geometry.(2 r ,2 8 ) But, for the moment let's explore how a diode Penning cell works. Figure 2.2.2 represents a single-cell diode pump. The cathode plates of this pump are typically made of a chemically active metal such as Ti, Ta, or combinations thereof.(2 0,2 g ) The anode cylinder and pump housing (not shown in the figure) are usually made of a nonmagnetic stainless steel. The anode is operated at a positive, dc voltage with respect to the cathode plates; it is usually of the order of 3 to 7 kV. The magnetic field, applied axially to the anode, is usually 0.1 - 0.2 Tesla (i.e., 1000 to 2000 gauss). Single cell pumps of this type have speeds of the order of 0.1 2.0 Z/sec. If a free electron is somehow launched into the volume within the anode, it becomes trapped by the Lorentz force of the magnetic field. That is, with the appropriate magnetic field, it is physically impossible for this electron to reach the positive anode. This first electron may have been dislodged from some surface by a cosmic
88
SPUTTER-ION PUMPING
particle, or it may have originated from field emission from the cathodes. In any event, this initial electron, trapped by the magnetic field, swirls about within the anode volume. The mean free path of the first electron for ionizing collisions may be very large, and it may travel great distances in its curved orbits within the anode cylinder (e.g., several hundred kilometers at a pressure of 10-9 Torr). However, it will eventually experience an ionizing collision with a gas molecule. This collision will result in at least one additional free electron orbiting within the anode volume, etc. A Townsend avalanche process ensues (the stn'king of the discharge) until, at pressures 10-4 Torr, a dense and somewhat uniform, steady-state cloud of negative space charge (i.e., electrons) is stored within the anode volume. These stored electrons cause the electrical potential within the volume to become depressed to a value approaching that of the cathodes.(a o ,a 1 ) The potential depression causes there to be a component of electric field in the anode volume which is radial to the axis of the anode cell, similar to that suggested by Fig. 2.2.2. NEGATIVE CATHODE . . . . . . . .fT.. ~/.... ELECTRON CLOUD TRAPPED BY ~ MAGNETIC FIELD
xt=
=m WITH RADIAL .~ ~ [ CO~~MPONENTS ~.~p "........~ " f 1-.~"= =--_ OSITIVE/~" ~ . . . . _ " CYLINDERS ~z~ - [ ~ ¢~m
Figure 2.2.2. The negative s p a c e - c h a r g e cloud of electrons, trapped by the magnetic field of the Penning cell, causes there to be a large c o m p o n e n t of radial electric field.
During the transient buildup of the negative space charge, positive ions bombard the cathodes and dislodge secondary electrons. These secondary electrons aid in the buildup of negative space-charge in the anode volume. At steady-state, the negative space-charge cloud remains constant within the volume, and consistent with the applied voltage.(a 2 ) Additional secondary electrons are prevented by the space-charge cloud from entering the anode volume. In fact, a mechanism must exist for depletion of the trapped electron space charge cloud as additional free electrons are created with creation of each positive ion. That is, at steady-state, for every positive ion which bombards the cathode, at least one electron must reach the anode. The mechanism permitting electrons to eventually reach the anode, while crossing magnetic lines of force, has been coined by plasma physicists as crossed field mobility or migration.( a a ) It means that electrons are given the mobility to cross magnetic lines of force as the result of collisions with gas molecules. They are deflected out of their quasi-circular orbits due to collisions with gas molecules, and some eventually reach the anode. This collision process, and the existence of spurious rf noise generated by the discharge,(a 4,3 s ,3 6,3 7 ,s ) makes it possible for electrons, with potential energies nearer that of the anode, to also reach the cathode. Define the total amount of electron space charge stored in an anode cell as Xq. At a constant pressure, a greater cell sensitivity, s, means that for the same pressure there is a greater rate of gas ionization and, therefore, more current generated by the discharge. It is reasonable to assume that the rate at which gas is ionized within the
S P U T r E R - I O N PUMPING
89
cell will be directly proportional to the amount of space charge stored in the cell. If we were able to double X q in a given cell configuration, we would be able to double I +, (i.e., the ion current created by the discharge). That is, I + = kz XqP,
(2.2.1)
where k l is some constant of proportionality. We recognize k x Xq as being the gauge sensitivity, s. It is also reasonable to assume that there is some linear relationship between the rate of gas ionization and the rate at which gas is pumped (i.e., the throughput, Q) in a Penning cell.(3 8 ) This is to say that for a given gas species, Q
= k2 I +,
(2.2.2)
where k2 is some constant of proportionality. We have defined throughput as Q = SP. Using this relationship, (2.2.1) and (2.2.2) we find:
S = kx k2 Zq
(2.2.3)
That is, the speed of the pump, S, is directly proportional to the total amount of electron space charge, Zq, stored in the Penning cell (cell hereafter). Speed dependence on pressure is not indicated in (2.2.3). This is known not to be the case; that is, the speed (and amount of stored space charge) is pressure dependent. If we plot the slope of the log 1 o of current vs. log x o of pressure for most Penning discharges we empirically observe: logI2 - loglz logP2 - logPx
logi2/I1
=
logP2/P1
=
n.
(2.2.4a)
Substituting/+ for 12 and P for P2, and solving for I + as a function of P, we find: I+
= kl pn
= kx Z q ( P ) e
(2.2.4b)
where X q(P) means that the stored charge in the cell varies as a function of pressure. The value of n varies with all of the parameters listed in Table 2.2.1. Depending on these parameters, there will be distinct and abrupt changes in n with pressure. The slope of (2.2.4a) will vary in value, over the pressure range of 10" x 2 to - 5 x 10-4 , from 1.5 > n > 1.0, respectively. Solving (2.2.4b) for Zq(P) P, we find that the speed of a pump as a function of pressure is: S
=
k z k2 P n- x
(2.2.5)
The constant k2 (see (2.2.2)) has a special significance as pointed out by Jepsen.(2 4 ) It is the efficiency with which gas is pumped by a sputter-ion pump. That is: Q
=
17I +
= s P = 17 = S P / I +
(2.2.6)
The term 17 or the ratio S P / I + appears frequently in the literature. Equation (2.2.6) clarifies its meaning. That is, it is merely the number of gas molecules
90
SPUTTER-ION PUMPING
pumped for each ion generated by the discharge.(S ~ ) The dependence of I + on n (i.e., (2.2.4b)) was most apparent in Penning's original data,(: r ) and has been the subject of over 50 subsequent publications - some relating to sputter-ion pumps (e.g., 2 6,2 6 ) and others dealing with various forms of vacuum gauges, (e.g., 4 0 ) as discussed in a review paper by Redhead.(S ) W e conclude that anything we can do to increase the amount of electron space charge stored in the cellwill go to improve the pumping speed of the cell. Recognition of this space charge dependence has led to some rather unique approaches in designs of sputter-ion pumps.(4 :, 4 2 ) It has also prompted numerous studies of the space charge characteristicsof these cells. This willbe discussed at length in Section 2.13. However, for the time being, the parameters which significantlyalter Zq and cell pumping speed are given in Table 2.2.1. Typical values used in commercial sputter-ion pumps are also given in thistable. Table 2.2.1. Parameters effecting Penning cell sensitivity and sputter-ion pump speed and typical values. Anode Voltage Magnetic Field Cell Diameter Cell Length Anode/Cathode Gap Pressure (P n)
V B d £ a P
3.0- 7.0 kV 0.1 - 0.2 T 1.0 - 3.0 cm 1.0 - 3.2 cm 0.6- 1.0 cm 1.0 s n < 1.5 Torr.
Qualitative Pump I/P Characteristics I first will qualitatively describe how the sensitivity of pump cells varies with the parameters listed in Table 2.2.1, and then discuss construction of various forms of sputter-ion pumps. Figure 2.2.3a shows how IfP varies as a function of magnetic field, B, at a given pressure (Welch, SLAC 1969). I/P ~. o
30
....
-"cRmc~" i - MAGNETIC _
. . . . . . . .
•
FIELD
[
[:] [:3
r
l
1
{
1
l
!
I
I ARGON 6.0
20
PRESSURE
x
10 -6
-
Tort
l
-
]D
;~ ;>
1
kv
A
No 10
a,~
z
i*
a
kV
~..
•
0
1
0
i
i
Dimensions of cell - millimeters
° • Oo~o 0 I
,
0.05
1
1
1
1
I
l
I
i ,
i
o
FIELD
Figure 2.2.3a. Penning argon as a function of
-
•
(
, i
0.15
0.10 MAGNETIC
o
•
o
,
t , i
0.30
•
T
0.35
Tesla
cell sensitivity magnetic field
at and
6.0 x anode
10 -6 Torr potential.
The dimensions of the pump cell are given in the figure. We see that there is a value of B below which there is no discharge in the cell. This is called the cutoff field. On discharge ignition, the cell I/P, and therefore I +, increases linearly with increasing B field until it reaches a maximum value. This maximum value, dependent on anode voltage, is called the critical magnetic field. A further increase in B beyond the
SPU'I'TER-ION PUMPING
91
critical magnetic field will result either in a constant value of I/P, or possibly a gradual fall-off in its value. In this case, the words cutoff and critical magnetic field do not have the same meaning as with magnetrons. The sharpness or kurtosis of the I/P peak (i.e., knee) at the critical magnetic field, increases with pressure. Because of the possible drop-off in I/P beyond the critical field, further increases in B can actually result in less pumping speed for the same voltage and cell configuration. The spacing between the cathodes and the anode also results in variations in I/P. This effect is shown in Fig 2.2.3b, where changes in I/P, are given for anode-tocathode spacings of 3.1 and 5.9 turn, with the same magnetic field and anode potential. In Section 2.10 it will be shown that if the gaps are too narrow between the cathode plates and a multi-celled anode array it will restrict the passage of gas into the cells. This results in an apparent reduction in cell sensitivity. I/P 30
,
,
v ,
f ,
,
,
,
,
i
,
i
i
i
[
l
i
I
i
I
I ARGON
-"h'~ a3.4 mill
_~
-
,,
.,~
"h"--, 27.9
mm
2.2.3b.
ANODE5.0POTENTIAL: kV
0
Penning
function
of
"
-1
i i ~
i
i
i
[
[
0.05
I
i
cell
sensitivity
cathode
to
i
anode
, , ,
i
0.10
t l i ,
0.15
MAGNETIC
a
-
Torr
i0
~'~ Z (/] 0
as
10 -0
_
~
Figure
PRESSURE x
-
I
Dimensions of cell - millimeters
6.0
FIELD
at
6.0
spacing
x
and
-
t
, t
0.20
T
0.25
Tesla
10 -6
Torr
magnetic
argon field.
The increase in I/P with V is linear up to the critical magnetic field. Then, I/P becomes nonlinear in V. This effect is shown in Fig. 2.2.4 for the same cell configuration given in Fig. 2.2.3a. Within limits, an increase in pump voltage usually results in an increase in pump speed. I/P 25 O
_
ARGON
~r/l2 0
~
-
< 15 I
-
;~ 1 0 ~
-
6.0 KEY
PRESSURE
x
10
~
o,oo
A
0.126
[]
0.180 0.230y
P
if/
q
/ /
TESLA
°
o
Torr
2".
L-.
[]
[]
[]
O
0
()
Figure 2.2.4. Penning cell sensitivity as a function of anode voltage, and for various magnetic fields.
[] -~) _
~ 5 Z m
0
I
I
1.0
2.0 ANODE
1
3.0 VOLTAGE
J
4.0 -
I
5.0
kV
6.0
Kilovolts
The correspondence of I/P with speed, S, is graphically demonstrated in Figs. 2.2.5 and 2.2.6 where S and I/P, respectively, for H 2 and N2, are given for a multi-cell element pump as a function of magnetic field.(4 a )
92
SPUTTER-ION PUMPING x/P
~/sec 10
1.0 H 2 PRESSURE: 2 x 16 s Torr Ng PRESSURE: 5x l{)TTorr .l
r-
~?
I
~ ~
Ha PRESSURE: 2x l(T8Torr N 2 PRESSURE: 5 x I O-7Torr
'
Nz SPEED
0.5
5
~
°°°oo
o?.#e#
J
0
i
o Z
1
i
1
.~
0.4 0.8 MAGNETIC FIELD -
Figure 2.2.5. Pump f o r He a n d N2 function
DD3DDDD
,..,1
SPEED 0
t
of
i , i
1
0
T
Tesla
speed as a field. magnetic
1 1 0
.2
1 1 1
0.4 0.8 MAGNETIC FIELD -
Figure 2.2.6. Normalized sensitivity f o r Hz a n d a
function
of
magnetic
t
1 1
1.2
T
Tesla pump
N2 a s field.
For reasons given in Section 2.13, the magnitude of I/P also varies with cell length, ,~. As noted earlier, some of the first Penning cells were constructed with wire anodes. That is, the length of these anode cells was approximately zero. For example, the triode of Fig. 2.2.1 has an I/P for air, at 10" s Torr, of ---1.6 A/Torr. This would correspond to end effects of a cylindrical anode of finite length. Therefore, I/P is f'mite for an anode cylinder of near zero length, and increases linearly with anode length thereafter. There are three caveats regarding the length of the anode cell. 1) Should the cell be too long, then the conductance leading to the volume within the cell (i.e., where the ionizing electrons are stored) becomes restricted. Therefore, though the amount of space charge stored in the cell may increase with cell length, the cell sensitivity model is no longer applicable, as the space charge is less accessible to the gas. 2) For a longer cell, a greater amount of magnetic material will be required to achieve a magnetic field for the optimum I/P. 3) The third consideration has to do with optimizing the surface area within the anode - where most of the chemically active gases, excluding H 2, are pumped. This is discussed below. The value of I/P also varies as a function of the cross section of the gas species. That is, the sensitivity of the cell varies with the type of gas. This does not necessarily mean that because the sensitivity for a given gas is low, that the speed for that gas is correspondingly low. This is often to the contrary. For example, the approximate relative cell sensitivities of Ar:N2 :H2 are 1.2:1.0:0.49 respectively (see Table 1.18.2). However, a conventional diode pump, with Ti elements, will pump these gases with speeds in the ratio approximately 0.02:1.0:1.4.(4 4,4 s ) The implication here is that different gases are pumped more effectively than others for the same ion current. For example, for the same N2 and H2 pump currents (note: this corresponds to different pressures for the two gases), the H 2 throughput will be almost x3 greater. You will have to think about this effect for a moment. The key is in the ratio of the number of gas molecules pumped to the number of ions created in the discharge. This is called thepumping efficiency,( 4 4 ) (i.e., the term "1'/"in (2.2.6)). Lastly, I/P varies as a function of pressure for different cell diameters. This has been discussed at length in the literature.(2 8,4 s ,2 a ) This was dramatically demonstrated in a paper published by Rutherford in 1%3.(2 6 ) In this paper he reported
S P U T r E R - I O N PUMPING
93
measurements of I/P of pump modules having cells with diameters of 1.27, 2.54 and 5.08 cm. In each case, the cell aspect ratio, defined as the cell length, i , to diameter, d, was held constant at £ M = 1.5. Rutherford's results are shown in Fig. 2.2.7. As a generaliTation, for a fixed magnetic field, the larger the cell diameter, the greater the I/P at lower pressures. I:l As a very rough rule of thumb, for N 2" I/P --10 AfForr for a typical sputter-ion pump cell at 10-6 Torr. Therefore, a properly designed pump with ten cells would have an I/P of--100 AfI'orr, at this pressure. On a lighter note, while at Varian, I developed a sputter-ion pump specifically for operation at very low pressures. The diameters of the cells and magnetic field were chosen for this purpose. We submitted this pump, for evaluation purposes, to a potential customer, with the condition that the cell geometries of the design must not be divulged. Keeping technical developments a secret is similar to keeping a secret in a church choir; it is not possible. Several months later, I was asked to referee a paper on sputter-ion pumps. It contained dimensions of the cells and speed data on this very pump and other competitor's pumps. The results proved most favorable to Varian. The paper was published. The authors never knew that I was one of the referees. 200
1 I II I I 1 11 1 1 I1 I I 11 1 1 11 36, 12.7mm? ~ Va = 3.0 kV _ cells, w i t h - - - - - - - - - - - 7 " "~ ,._ ALL TESTS B= 0.20T ~ . _ i Q ~-/ 3 50 8 m m ~ "xt" 150 , 1 /f / - cells, w i t h
~.
_
<, >..,
/
/ ~=010t
/
__ .
.
.
/
/
38. i2.7mm~_
~ c e l l s ,
.
.
.
.
.
.
with
_
.
too
~r~ Z m
//6, / /
/ /
~.~
I
25.4 m m ? • ceils, w i t h ~ B : 0.10
--
• _ . . .
0
I 10 -10
.
.
.
.
i II
111
10 -9
l 10 - a
Rutherford's I/P and for various
,,"
'-"
-
-_
st
36, 12.7 m m ? L__ cells, w i t h B-O 10T
I I1
! 10 -7
PUMP PRESSURE Figure 2.2.7. vs. pressure,
-"
• #'
,,.
1
- S
1 1t
J lo -a
_
I
I I]lO -5
Torr
measurements cell diameters
with a cell and magnetic
t/d
of 1.5, fields.(~6)
2.3 Pumping Speed Abstraction The definition of the speed of a sputter-ion pump is the same as that described in Chap. 1 for a mechanical pump. We must use our imaginations a bit to envision how volumetric displacement rate applies to a Penning cell. Envision elemental abstract volumes, at unspecified pressure, finding access to the discharge volume within the cell. The rate at which these volumes enter the cell is influenced by the conductance of the gap between the cathode plates and anode cylinder. In fact, this is the definition of conductance. The conductance accessing the anode cell and the amount of space charge therein is crucial to achieving a high pumping speed.
94
SPU'VFER-ION PUMPING
The speed of a sputter-ion pump is analogous to a farmer of old cutting grain with a scythe. If the farmer walks along and sweeps his scythe at a constant rate, the rate at which the grain falls before the scythe is dependent only on the density of grain stalks in hispath. In this illustration the swirling scythe, which is being swept at a constant rate, is analogous to the constant and swirling electron space charge cloud. The stalks of grain, perhaps of varying density, are analogous to gas, perhaps at varying pressure. In Section 1.14, I discussed the implications of Dalton's law and how it related to the application of linear superposition in calculating system partial pressures for a given pump speed (e.g., problem 12, Chap. 1). The assumption of linear superposition proves correct for turbomolecular pumps, diffusion pumps, cryopumps and most other types of pumps. It is also true, to the first order, for stably operating sputter-ion pumps. However, because of the complexity of sputter-ion pumping mechanisms, both the sequential and simultaneous pumping of different gases impacts on the speed of each of the species (e.g., see 4 7 - s o ). Also, as will be seen, it is possible for the speed of sputter-ion pumps to be selectively negative for some species (e.g., noble gases) while, at the same time, positive for others.
2.4 The Making of Sputter-Ion Pumps 2.4.1 Pumping Mechanisms and Materials English journal publications reporting on electronic pumping mechanisms, alone, number in the hundreds. Excellent review papers have been published on pumpimz effects in hot-filament ion gauges (i.e., forms of electrostatic capture pumps).(s 1") Approximately 100 additional papers exist which report on the pumping mechanisms of sputter-ion pumping. Were I to reference all of these publications and treat this subject in depth, this alone would require a complete volume. Rather, I will describe these electronic pumping effects with simple models. There are two primary mechanisms responsible for the electronic pumping of gas. One is physisorption the other chemisorption. One need not have command of quantum physics to understand these simple mechanisms (Richard Feynman is quoted as saying "I think I can safely say no one understands quantum mechanics", anyhow). Both mechanisms to one degree or another are enhanced by sputtering. Sputtering, physisorption and chemisorption will be qualitatively described.
Sputtering and Sputter-Yield When an ion bombards a solid (e.g., metal) surface, it can be very disruptive to that surface, causing a great deal of damage on an atomic scale. For example, an argon ion with a few hundred volts energy will cause considerable damage to a metal lattice to a depth of --100 A.(s ~ ) This damage may be in the formation of small dislocations in the lattice structure. These dislocations can serve as conduits for the diffusion of H2 and the inert gases. Damage may also take form in the splashing off of some of the ground zero metal atoms onto adjacent surfaces. This splashing process is called sputtering, and is the primary mechanism leading to sputter-ion pump-
ing.(3 s )
The cathode material being bombarded is often called the target material. The effectiveness of the primary particle in sputtering or splashing off the target material
SPUTTER-ION PUMPING
95
is called the sputter yield. It is the ratio of the average number of target material atoms sputtered per bombarding ion. Sputter yields of ~ 1.0 are common. Yields depend on the momentum, charge and mass of the ion, the angle at which it strikes the surface - this is called the angle of incidence - and properties of the target material. The angle of incidence is the angle formed by the trajectory of the ion with an imaginary line perpendicular to the tar]get surface. Sputter yields have been published for many gas-metal combinations.(5 a, s 4 ) 2.5
I
I
I
z.0 Z
a:
f.~
/
1
/
2.5
Z) °
~. a.0
[..,
1.0
-
7//
0.5
_
-
45 keV IONS WITH
/
o Z
NORMAL
l
I
[
1
1
z
1 .5
N
I l I i 1.05 keY ARGON IONS
INCIDENCE
~
1.5
~
1.0
i
bl M
_
~ o Z
o
0.5 /
0
I
0
i
10
I
20
I
30
I
40
Z
50
ATOmiC NUMBER OF BOMBARDING IONS Figure 2.4.1. Sputter-yield o f Si, normalized t o Si s p u t t e r - y i e l d f o r Ar, v s . Z o f b o m b a r d i n g ions.(55)
o
l 0
I
1
1
30 ANGLE
OF INCIDENCE
1 60
90
- DEGREES
Figure 2.4.2. Sputter-yield with incident angle of ion.(5a)
The following generalizations regarding sputtering may be made as they relate to sputter-ion pumping: 1) The greater the ion momentum (i.e., pump voltage), the greater the sputter yield. 2) The lighter the gas ion, the less the sputter yield for the same energy and target material. For example, He and H~ ions have very little sputter yield compared to N2 and Ar ions accelerated through the same voltage. Figure 2.4.1, illustrates the dependence of sputter yield on ion mass, as reported by Andersen and Bay.(S s ) 3) The sputter yield, for the same gas ion and energy, does not necessarily differ markedly for different target materials. Oechsner outlines the theoretical models reported by others which explain this characteristic.( s a ) This effect relates to the surface binding energy of the target material and the amount of energy per unit volume which can be deposited into the lattice of the target material by the bombarding particle. 4) The sputter-yield gradually increases with the angle of incidence of the bombarding particle to about 70", then decreases quickly with increasing angle (e.g., see Fig. 2.4.2).( s 3 ) Brubaker was the first to point out the significance of this effect as it related to sputter-ion pumping.(S 6 ) Variation in sputter-yield with angle of incidence is analogous to the effect of a flat stone skipping over water. There is a modest plop when we throw the stone on edge directly into the water (i.e., zero angle of incidence). However, if we throw stones of similar shape in the water, at progressively increasing angles of incidence, they will make progressively larger splashes as they enter the water. Finally, there is an angle of incidence which is less than 90", but at which the stone does not break the surface of the water. It makes a very slight splash and skims along the surface. An even greater angle of incidence results in even less perturbation to the water.
96
SPUTTER-ION PUMPING
The flux or intensity of material splashed off the surface of the cathodes, by the bombarding gas ions, is said to follow the cosine law for near zero angles of incidence. This merely means that the flux intensity is greatest perpendicular to the cathode surface, and that it gradually diminishes in proportion to the cosine of the angle q,, shown in Fig. 2.4.3. The flux intensity will also vary as r" 2, the length def'ming a surface at a distance r from the sputter source. These two relationships are given in (2.4.3a). For reasons which will become clear, a large proportion of the chemically active and noble gases are pumped on the anodes. Therefore, the size (i.e., surface area) and shape of the anodes is important.(4 7 ) The cosine-law effect has important implications as it relates to the cell aspect ratios; that is, the ratio of the cell length to diameter (i.e., ~/d). It is known that with magnetic fields ~ 0.15 T, and at low pressures, the density of ion current is greatest at the cathodes at points subtending the axis of the anode structure (e.g., see s 7 ,s s ,s 9,6 o ). Assume, therefore, that most of the material sputtered from the cathodes originates at these points. One can calculate the flux intensity of sputtered material through an imaginary surface (see Fig. 2.4.3) by the following: v vdA
= k4 r" 2 cos q~
(2.4.1a)
= (k4 r'2 c o s ¢ ) ( r 2 sin¢bd¢dfl)
(2.4.1b)
The rate at which sputtered material is deposited on the anodes, R, is then:
R
=
(N/Tt) sin 2 ~.
(2.4.1c)
I have defined the quantity, N, as the total rate at which material is sputtered from the cathode. Such preferential cathode sputtering (centered on the anodes) will be evident to anyone who takes apart and services an extensively used pump. Using (2.4.1), the proportion of material sputtered on the anodes vs. the total amount sputtered is calculated to be, for .~/d, as shown in Fig. 2.4.4.(6 1 ) Some sputtering of the cathode plates occurs off-center of the anodes. Ions are known to strike the cathodes beyond the projected anode radius.( s 7 ) However, at low pressures, the ion current density under the shadow of the anode cells is negligible compared to that on-center with the anodes. Therefore, there is a preferential erosion of cathode material toward the centers of the anodes, and a net build-up of sputtered material in the regions nearest the anodes.(S 9 ) d~
z[ " 1 0 0
7,5 I
I
¢~ 50 z 0
<
Figure 2.4.3. Elemental area t h r o u g h which flux passes. Figure 2.4.4. S p u t t e r - c o a t i n g of anode cell vs. cell l / d .
o
25
z o
0
1 0
0.5
1.0
g=o.2d
I
1.5
l/a 2.0
CELL LENGTH TO DIAMETER RATIO
SPUTTER-ION PUMPING
97
Physisorption and "Binding Energies ~ Physisorption is the catch-all term used to describe all gas capture pumping mechanisms, excluding chemisorption. For example, if a gas molecule or atom is buried and trapped under a coating of sputtered fdrn, this is called physisorption. If a high energy, neutral atom is buried deep within the metal lattice, this too is called physisorption. Lastly, if a gas molecule has no chemical aff'mity for a metal, but finds defects within the metal lattice into which it diffuses (e.g., Ar), or is small enough to diffuse into interstitials of the lattice (e.g., He) - in both cases because of a concentration gradient created by the bombardment of noble gas ions - this too is called physisorption. If that gas is being adsorbed into the bulk of the metal, it could also be called absorption. Noble or inert gases are pumped exclusively by combinations of these three physisorption mechanisms. The physisorption of noble gases, and the formation of mobile bubbles of these gases in metals, became of interest to those studying fission reactor materials.(6 2,6 s ) Of course, the pumping effect due to van der Waals forces is another form of physisorption. These mechanisms will be discussed as they relate to specific types of pumps. Physicists and chemists many times will model the flow or dynamics of physical processes in terms of energy states and changes in these states. This only becomes an intuitive process after years of thinking in these terms. We lay people tend to think in terms of concepts made tangible by our day-to-day experiences. For example, time, mass, velocity, and force are concepts most of us have a feel for. Therefore, when told that two bodies will have a mutual attraction for each other, related in some way to the distance at which they are separated, we can accept this as a condition of nature, and carry on. Such is the case with two atoms or molecules. When they come in close proximity to each other, there are competing forces of attraction and repulsion between the two atoms. Though we have never seen force fields (e.g., gravitational or electrostatic fields), we daily experience their effects. And, we can accept that there must be some sort of force field emanating from each atom into free space. The force of the attractive field between two atoms varies as 1/rT, where the mutually repelling force varies as 1/r 1 s, where r in each case is the separation between the atoms. That is, the force, F(r), between the two atoms varies as:
F(r) -
f -
,
g
r 1 3
r 7
(2.4.2)
where f and g are positive constants which vary in value depending on the atomic or molecular species. In this equation, the repulsive force is given a positive value (i.e., it requires the input of work to overcome) and the attractive component is given a negative value. The work which must be done, zxE, to bring the two atoms close together by a distance zxr, from a great distance, is simply: AE = -
F(r)×/xr
The negative sign implies that the particles are initially attracted to each other and therefore do work on the system (i.e., give up heat), rather than the system having to do work on the pair of atoms. The work required to bring the atoms together from an inf'mite distance is simply:
98
SPUTTER-ION PUMPING /"
¢o
oo
F
f
E(r) =
12 r 1 2
g
,
6 r6
(2.4.3)
Equation (2.4.3) is said to be the energy state of the system (i.e., two atoms or molecules). Both (2.4.2) and (2.4.3) are plotted in Fig. 2.4.5, which shows both the force and energy well for two hydrogen molecules.(6 4 ) Similar, comparatively weak attractive forces exist between all molecules and atoms, the constants f and g varying depending on the gas species. The depth of the energy well, varying from species to species, is called the heat of adsorption. It is merely a measure of the tenacity or strength with which the two particles are bonded. For this reason, it is often called the binding energy of the two atoms. Equation (2.4.2) is the more familiar van der Waals force equation. It expresses the force responsible for the cryosorption, condensation and liquefaction of all gases. 10+7
10+5
A
o
/10+3
POTENTIAL ENERGY
r~ t
/
.~ cJ
MOLECULES
o
r
~:
r~ Z
10-1
~ ,-,'~
c
~l' ~'-" "~ "" " "~ - ~ . ~ . - e F(r)
-,, -n ~ .~ .~,
Z
;>
Z
aa _ l O - a o
N
-10-1
t
~
---- P O T E N T I A L ENERGY "WELL"
E-
-~
]
." - 1 0 +1
J
0
1
2
SEPARATION Figure
2.4.5.
distance
3 BETWEEN
Potential
of s e p a r a t i o n
4 MOLECULES
and force of two h y d r o g e n
energy
r
5 -
0
~
6
7
Angstroms
as
a
of
function
molecules.
{64)
Equation (2.4.3) is the van der Waals energy equation. The fact that the energy equation is negative in sign means that energy or heat is given off in the process of the combining of atoms or molecules. In fact, all physisorption processes are exothermic. When we supply refrigeration for the pumping of gases, the refrigerant or refrigeration system must be able to absorb calories stemming from the heat of sorption; but, this is covered in another chapter. It is sufficient to note that cryocon-
SPUTTER-ION PUMPING
99
densation and cryosorption (e.g., at temp. < 293* K) are included in the pumping mechanism termed physisorption. And, the dominant mechanism responsible for this form of physisorption is van der Waals forces between the molecules. Van der Waals forces also exist between liquid and solid surfaces and gases. An atom or molecule coming near to the surface will be attracted to that surface, fall into the potential well and stick to that surface for a f'mite period of time. To calculate these forces with reasonable accuracy, one need only assume a certain density of atoms per unit volume of the bulk material, and then integrate to fred both the force of attraction and energy terms for a gas molecule some slight distance away from the surface. These calculations were made by Redhead, et al. (i.e., (6 4 ): This book is out of print. But, the student of UHV technology should seek it out in some good technical library.) Energy units most frequently used to describe the strength of the bond between a surface and some gas molecule are electron volts/atom or kcal/mole of gas. One electron volt per atom corresponds to - 2 3 kcal/mole. As a rule of thumb, the magnitude of heats of physisorption are usually s 10 kcal/mole, where heats of chemisorption are > 10 kcal/mole.
Chemisorption The above calculations were made for H2 molecules, rather than atoms of hydrogen. Had we brought two atoms of hydrogen together, the potential well would have been much deeper. It would have represented a chemical bond between the two atoms. For example, N 2 is a chemical bond of one nitrogen atom to another nitrogen atom; CO a chemical bond of a carbon atom to an oxygen atom, etc. In fact, all atoms of gases, except the inert gases, when found in their natural states, are chemically bound to either similar atoms of the same gas (e.g., N2, H 2 , 0 2 ), or other atoms (e.g., CO, CO2, H2 O, CH4 ). Such chemical bonds stem from the sharing of valence electrons in the outer orbital shells of the atoms. These gases also form stable chemical compounds with many of the active metals. Chemisorption is the removal of gas from the system (i.e., pumping) through the formation of chemical bonds between the gas and some chemically active metal (e.g., Ta, or Ti cathode materials). Therefore, the cathode materials must have a chemical affinity for the gas. There are two chemisorption mechanisms which occur in ion pumps. The first involves dissociative pumping of the gas through the formation of stable chemical compounds; the second involves the pumping of a gas by dissociation and diffusion into the cathode plates and, to a lesser degree, the anodes. The first, chemisorption model, depicted in Fig. 2.4.6, occurs as follows: 1) ions bombard the cathode surface and sputter cathode material onto some nearby surface (e.g., the anode); 2) the chemically active film of sputtered atoms, residing on these surfaces, awaits the arrival of neutral gas molecules; 3) gas molecules which impinge onto these active sites are physisorbed for a brief time; 4) gas molecules dissociate on the surface either spontaneously- the fundamental mechanism - or by achieving some slight activation energy; so, 5) the dissociated gas forms a stable chemical compound with the film (e.g., N 2 + 2Ti --, 2TIN). Of course, the gas molecule may already be present on the surface, residing there because of physisorption, at the time the sputtered cathode material arrives. The gas molecule may require an energy nudge before it is dissociated and combines with the sputtered cathode material. In a nutshell, this is pumping through sputtering and chemisorption by a dissociative process.(8 s )
100
SPUTTER-ION PUMPING CATHODE ~T BUILD-UP :ATHODE EROSION
SPUTTERED TITANIUM ATOMIC NITROGEN TITANIUM NITRIDE DIATOMIC NITROGEN IONIZED NITROGEN
Figure 2.4.6. C h e m i s o r p t i o n of n i t r o g e n by s u r f a c e d i s s o c i a t i o n a n d c o m b i n a t i o n with c h e m i c a l l y active, s p u t t e r e d t i t a n i u m films.
With the exception of CO, all diatomic gases are dissociatively adsorbed on metals.(6 4 ) The process is sometimes represented by an energy diagram, taken from Bond,(6 6 ) and shown in Fig. 2.4.7. Curve 1 represents the energy diagram for the physisorption of a diatomic molecule on a clean metal surface. Curve 2 represents the energy, diagram for the chemisorption of each atom of a diatomic molecule Energy E in this diagram represents the energy required to dissociate the molecule into atomic species. We see that the closer the molecule of Curve 2 comes to the surface, the lower the molecular dissociation energy becomes. This phenomenon is the process which facilitates chemisorption on metal surfaces.
A~/
_j
PHYSISORPTION
CaEMISORPrXON
I
ON
Z 0
:a
"~
N
0 ~ "~ gh
L Sp
PHYSISORPTION
HEAT OF I~CHEMISORPTION
*~'
i
] ~
,
F i g u r e 2.4.7. P o t e n t i a l energy diagram for the dissociative chemisorption of a gas m o l e c u l e . (Be)
DIATOMIC GAS
oLrc
L
0 1 2 3 4 5 DISTANCE FROM SURFACE -
6 7 ANGSTROMS
The large positive value of E* does not represent, in this case, a mutual repulsive force between the metal surface and the molecule. Rather, it represents a measure of the work required to dissociate the molecule as a function of its distance from the surface of the metal. In Curve 1, E...!/ represents the physisorption energy, where E c represents the depth of the chenusorption potential well. If the physisorbed molecule could gain just enough energy to escape the shallow Ep well and reach the point of intersection of the two energy curves, it could dissociate and fall into the chemisorption potential
S P U T t E R - I O N PUMPING
101
well. The energy barrier, Ea, is very small in value compared to E* . Energy E a is referred to as the activation energy for the dissociative chemisorption of the gas by the metal surface. The gas molecule might have been energetically disturbed (i.e., become metastable) due to a prior collision with an energetic ion or electron.(e 7) In this encounter, the molecule is given just enough energy by the collision that when it later lands on a surface, on which a chemically active atom resides, it is readily dissociated and chemisorbed. Depending on the surface material and gas molecule, E a could have a value less than zero. In such an event, gas molecules would directly combine with the sputtered films without first requiring the intermediate dissociation process by some energetic particle. For a clean surface, unpopulated by gas, E a will have the smallest value, and it will increase in magnitude with increasing gas coverage of the surface. Some gases have sufficient chemical affinity for materials to spontaneously dissociate when encountering an atom of a metal. For example, O2 spontaneously dissociates and combines with a number of metals (e.g., In, Ti, A2, Cr, etc.). However, once a 50 - 100 A oxide layer develops on the surface of the metal, the oxide becomes a diffusion barrier to the formation of additional oxide. Sputtering away these oxides will replenish active metal sites. /~
JJJ J
jJJ 4
g D ~ I I D g gO i /~ IONIZED @ CATHODE +I~ ~ "- ~ r , ~ BURIEDHIGH t~ ~ ENERGYNEUTRAL ANOD~EI ~ ~ .~ 'REFLECTED" '/C + NEUTRAL SURFACE--~[~ ~ ~ ~ l / H DIFFUSIONIN F CATHODESON
,\
\ ~
HYDROGEN _ DIFFUSION INTO
0 DISTANCE IN CATHODE
Figure 2.4.8. Chemisorption of hydrogen in the pump cathodes by dissociation and diffusion, and physisorption in anode by high-energy neutral burial.
The second chemisorption process, depicted in Fig. 2.4.8, also requires a chemical aff'mity between the cathode material and the gas. In this mechanism, on impact with the metal surface, the gas ions dissociate into an atomic state (e.g., H2 + + e" -. H + H). A concentration gradient develops in the bulk material as a consequence of the dissociated surface gases. The gas then diffuses, in an atomic state, into the cathode material. The cathode plates serve as a bulk reservoir for the pumping of gas. Hydrogen is very prevalent in vacuum systems. Therefore, it is important that we understand the limitations of sputter-ion pumps in pumping this gas. I will return to this subject later in the chapter. The relative activity of metals for chemisorption of different gases, as reported by Bond,( 6 8 ) in an extension of Trapnelrs work,( 6 s ) is given in Table 2.4.1. We see that Ti and Ta would serve as comparatively good sputter-ion pump cathode materials, where it is noted that aluminum would not seem to be a very good cathode material.(6 9 ,z o ,7 1 ) However, this was recently investigated, and there is some evidence, contrary to the indication in Table 2.4.1, that both CO and N2 are chemisorbed on A2.(7 2,7 a ) Some success has been reported in the use of pumps constructed with one A~ and one Zr cathode,(7 4 ) and pumps constructed with cath-
102
SPUTI'ER-ION PUMPING
odes of Ti-Zr-Al, though the alloy composition was not noted.( ~ s ) Also, Lu has recently reported on work with A~ -Y and A l -Ce composite cathodes.(76 ) Table 2.4.1. Metal and semi-met~,s adsorption properties for various gases.t66,6 a ) Group A
Metals
02
Ca, Sr, Ba, Ti,
Zr, Hf, V, Nb,
Ta, Cr, Mo, W, Fe, 1 Re 2
B 1 Ni, C02 B2 Rh, Pd, Pt, Ir2
C D E F
Notes:
A A A A A
AI, Mn, Cu, Aua K Mg, Ag, 1 Zn, Cd, In, Si, Ge, Sn, Pb, As, Sb, Bi A Se, Te NA
C2H2 C2H4 CO
H2
CO2
( = decreasing heats of adsorption = ) A: Gas Adsorbed NA: Gas Not Adsorbed A A A A A A A A A A A A A A NA A A A NA NA A NA NA NA NA NA NA
NA NA
NA NA NA NA
NA NA
N2
A NA NA NA NA NA NA
1) Adsorption of N 2 on Fe and 0 2 on Ag is activated at 0* C Metal probably belongs to this group, but film behavior unknown. Au does not adsorb 0 2.
As noted, all sorption processes are exothermic. That is, heat is given off during the process. The relative absolute values for the heats of chemisorption for the various gases decrease as shown in Table 2.4.1. The heats of chemisorption for various metals, used in soutter-ion pumps at one time or another, have absolute magnitudes accordingly:(64 ) Ti,Ta > Nb > W, Cr > Mo > Fe > Mn > Ni,Co > Rh > Pt,Pd > Cu,Au. An unending variety of materials is discussed in the literature as possible candidates for sputter-ion pump cathodes (e.g., see Holland (T 7 )). Suggestions for the possible use of many materials have far outweighed actual experimental evidence of the use of these materials. I will not make an attempt to summarize much of this speculation for you. However, it should be noted that both triode and diode sputter-ion pumps with aluminum cathodes have been shown to very effectively pump N 2 for extended periods. In fact, steady-state speeds for N2, and with aluminum cathodes, were shown to exceed those of pumps with titanium cathodes.( ~ 2,7 a )
2.4.2 Diode Pumps The Need for "Clean" Pumping Varian Associates played a major role in the commercial development of sputter-ion pumps. The principal product of Varian Associates in the 1950's was microwave tubes. Many of these tubes had thermal emitting, oxide cathodes. These cathodes were fragile devices in that if they were contaminated by diffusion pump oils, cathode emission characteristics would be significantly degraded. Because of this, technologists sought alternate methods of pumping these vacuum tubes during their final stages of high-temperature bakeout.
SPU'ITER-ION PUMPING
103
Some klystron amplifiers were almost two meters in length, operated at very high beam voltages (e.g., 100-250 kV) and generated peak rf power in the MW range. After aging the tubes on diffusion pumped vacuum systems, they had to be pinched off the systems and shipped somewhere to a radar transmitter. The need of sustaining high vacuum in these devices during subsequent operation was further motivation for the creation of clean, portable pumps which permanently appended the microwave tubes. Whereas the magnetron oscillator pumped very effectively on itself, linearbeam microwave tubes were less effective self-pumps. The presence of gas in these linear beam devices caused spurious oscillations and noise in the rf output. This required that better high temperature bakeout techniques be developed and appendage pumps be permanently attached to some tubes. Numerous scientists and engineers at Varian worked on these requirements. The assumption leading to the development of the larger sputter-ion pumps was elegantly simple: If one pump cell had a measurable pumping speed of S, would not an integer, n, of these cells have n x S pumping speed? Reikhrudel and others tested this assumption and reported on the first multi-cell sputter-ion pump in 1956.(6 9 ) Their sputter-ion pump had cells arrayed in series, analogous to the pages of a book, where the odd numbered pages comprised the cell cathodes and the even numbered pages, anode rings. The cathodes of this multi-cell pump were operated at the same negative potential with respect to the anode cylinders, sandwiched between the cathodes. The entire assembly, fitted in a long glass cylinder, was inserted into a solenoid which provided the magnetic field. They conducted tests with Ta, Mo, Ni and A~ cathodes, finding only the latter to be unsatisfactory in the pumping of "air", He and Ne. Varian engineers constructed their first multi-cell pump with a fiat array of parallel anodes, sandwiched between two cathode plates. Hall, Helmer and Jepsen submitted an application for patent of such a multi-cell pump on July 24, 1957.(4 ) It was reported on by Hall a year later.(7 8 ) A diagram of the first commercial pump of this nature is shown in Fig. 2.4.9. The anode, cantilevered on the center rod of the high voltage feedthrough, comprised the plainer array of 36, 1.3 cm square anode cells, 1.9 cm in length. The cathodes were made of Ti and held in place by spacers. The pump housing was made of 304 stn. stl. (i.e., stainless steel) and TIG (i.e., Tungsten Inert Gas) welded. Flanges of the earlier pumps had Alpert step seals, though three years later, pumps of all sizes were sealed with ConFlat ® flanges.(7 9 ) This is a good example of the ratcheting effect of technology. The UHV sputter-ion pump came before the ConFlat ® flange, but made the development of reliable, bakeable, take-apart flanges important. A large Alnico V~ magnet provided the external magnetic field. The lighter ferrite magnets had not yet found wide industrial use. The speed of this pump was of the order of 8 Z/sec for N 2. Variations of this simple design are used as appendage pumps to this day on many of the microwave tubes produced throughout the world. This first Varian pump proved to be an excellent appendage pump for use on large microwave tubes. At the time it was being developed, I was a tube technician at the G.E. Microwave Laboratory in Palo Alto. In 1957, we were developing a large, highly classified, super-power klystron for a radar system. We heard rumors of the magical electronic pumps which could be used to append our tubes. We obtained prototype pumps, and developed a battery-operated, high-voltage power
104
SPUTTER-ION PUMPING
supply so that appendage pumps, attached to klystrons, would pump on the tubes, even while in transit to the transmitter. //
/
._ ~ \\ \
\
\\ \
I\ I~
HIGH VOLTAGE FEEDTHROUGH
ANODEARRAY TiCATHODES
~-J~ / /
3 m m THICK L
L -
I
I
I
L__i
TIG WELDS
~\ \ \ , , \ , , \ \ \ \ ~ . \ \ , \ \ , . \ ~ . ~
_ . . . . . .
~lrnm I ~_LL
~
I
",~
_~
!
~
1
I
U "~. . . . .
F
[~)
STL. PUMP BODY Figure
2.4.9.
The
first
Varian,
multi-cell
sputter-ion
pump.
This appendage pump and even larger pumps under development at Varian were reported on that same year by Hall(S o ) and a year later by Jepsen.(2 4 ) In August of 1958, Lloyd and Huffman made patent application for a pump with modular elements which fit into pockets defined by the envelope of the pump body.( a 2 ) Figure 2.4.10 shows the concept of this modular design. HIGH VOLTAGE FEEDTHROUGH
,~ l~/
ALNICO MAGNET WEDGES (12) / f
~
HIGH VOLTAGE INSULATORS
\
\
ANODE
/ \ \
~~_CATHODE
~
/ ~
~
~
J
PLATES
-1 /
PUMP ELEMENT "POCKETS" (12) Figure 2 . 4 . 1 0 . R e p r e s e n t a t i o n of a v a r i a t i o n of t h e f i r s t " p o c k e t " pumps, patented b y Lloyd a n d H u f f m a n of V a r i a n A s s o c i a t e s . ( a 2 )
Pump elements were fabricated so that the anode arrays were supported off of the cathode assembly frame by shielded ceramic insulators. The insulators had to be
SPUTTER-ION PUMPING
105
optically shielded from the cathodes to avoid electrical breakdown problems caused by sputtered cathode material. The design of these insulator assemblies became more sophisticated with time, as it was also important to minimize insulator contamination problems associated with the starting of pumps (i.e., the unconfined discharge mode). The assembled elements were inserted into separate pockets in the pump body weldment. Thereafter, a variety of unique pump configurations began to appear on the market.(8 3,8 4 ) However, this development by Lloyd and Huffman represented the next major breakthrough in sputter-ion pumping as: 1) it made possible manufacturing economies associated with the modular construction; 2) it suggested that the size of sputter-ion pumps was virtually unlimited; and 3) it incorporated innovative, and equally modular, magnetic circuit designs. All subsequent pump designs were minor variations of this modular design.
2.4.3 Noble Gas Instabilities The diode pump proved to have a fundamental flaw. It was discovered that, after pumping on an air leak for an extended period of time, the pump would from time to time violently belch up gas in some sort of periodic fashion. The gas which was preferentially regurgitated by the pump was argon. These instabilities, called pump memory effects, would result when pumping on a gas mixture containing a partial pressure of an inert gas, or when pumping on a pure inert gas. Such instability problems were recognized early in the development of these pumps.(5 8, s s ,8 6 ) At times they were evidenced by gradual cyclical fluctuations in pressure; on other occasions, as very rapid increases in pressure. Brubaker was the first to report the cyclical, almost sinuosoidal, variations in pressure, sometimes characteristic of these instabilities.(S 6) Malter reported on this type of instability problem a year later. (8 6 ) 1.5 ~"
o
I
/.,,
HIGH
SPUTTER-ION PUMP
f.~.
1.0
~ X
HIGH ARGON
c¢1 c¢?
0.5 Z
~
0
0
0
~360 hr. INT0 "rEST
100
PRESSURE
200 300 TIME - Seconds
0.5 APERTURE CONDUCTANCE 4OO
TO AUXILIARY DIFFUSION PUMP
F i g u r e 2.4.11. E x a m p l e of s i n u s o i d a l - t y p e argon instabilities observed with a sputter-ion pump, when simultaneously pumping on the system with a second stable auxiliary pump.
The sinusoidal-type instabilities are difficult to duplicate. They usually involve use of very controlled parameters and conditions, (e.g., highly regulated power supplies, etc.) and the simultaneous pumping on the system with some form of stable, auxiliary pump. An example of this type of instability is shown in Fig. 2.4.11 (Welch, Stanford Linear Accelerator Center (SLAC), 1969). Extensive tests were conducted at that
106
SPUTTER-ION PUMPING
time on multi-cell pump elements, (32 cells, d = 1.98 cm 4', ,t = 1.59 cm) with Ti cathodes, which were inserted into custom-built pump pockets. In other studies at SLAC, the more abrupt or violent type of instabilities were also duplicated under laboratory conditions. An example of this type of instability is shown in Fig. 2.4.12. In this case, in the absence of a stabilizing auxiliary pump, at the onset of the instability, the pressure gradually rises, over a period of hours. After this, abrupt, periodic peaks or spikes in pressure are observed, after which, the pump will again be stable for a period of several hours or even days. The pump and apparatus were identical to that of Fig. 2.4.11, except in this case the auxiliary pump was not used. 1
I I -4 10
- -
~
Z o
~
/"
-5
113
J
-6
PUMP 0.017. -7
o
t
1
_ji
VOLTAGE: 5kV REGULATION
l
2
t
3
4
l
5
6
7
TIME Figure 2.4.12. with titanium
/I Jl
i
lO
lO
1
se
I
~
i
~2
Example cathodes,
of instabilities after pumping
8
I 9
10
t 11
12
f 13
14
hours
in a system argon at 5
x
with one sputter-ion pump 10 - 7 T o r t for a few hours.
The shape and periodicity of the instability wave functions will vary with any of the parameters of Table 2.2.1, with the volume of the vacuum system, and with the presence of auxiliary pumping on the system. In fact, it was noted in these studies that under highly controlled circumstances, including temperature, the pump elements described above could pump on Ar leaks for extended periods. In one instance, a new pump pumped on an Ar leak, at a pressure o f - 2 x 10" 7 Torr for only 5½ hrs, having pumped only -2.4 x 10-3 Torr-Z Ar prior to the onset of instabilities. On readjusting the anode voltage from 5.0 kV to 4.2 kV, this same pump pumped on an Ar leak, at a pressure of --2 x 10-6 Torr, for -700 hrs, pumping -3.0 Torr-Z Ar, with no evidence of instability. Reducing the Ar pressure for several days, to --2 x 10" 7 Torr, did not prompt the onset of instabilities. These results suggested two things regarding sputter-ion pumps: 1) there are narrow voltage "bands" within which, under highly controlled conditions, a given pump may stably pump Ar; 2) in the absence of the reporting of all of the experimental data and test parameters, discretion should be used in interpreting published claims of instability or stability relating to a given pump configuration.
SLAC Pump Instability Problem High vacuum systems which are frequently cycled back to air are less prone to argon instabilities.(8 s ) However, systems with air leaks, yet operating for extended periods at pressures even < 10" 8 Torr, may evidence argon instabilities in a matter of a year of so. This became a serious problem in the SLAC Two-Mile accelerator, and
SPU'VFER-ION PUMPING
107
prompted the above studies. Anomalous pressure bursts in sectors of the accelerator resulted in the interruption of operation of the machine. The vacuum system is described elsewhere.( s 7 ) For purposes of this discussion, it is divided into thirty, 100 m sectors. A simplified schematic of one of these sectors is given in Fig. 2.4.13. 500 Z / s e e
ION P U M P S j d ~
KLYSTRONS MANIFOLDS WAVEGUID VACUUM VALVE S-BAND VACUUM
130 m m ¢ VACUUM ANIFOLD
WAVEGUID
I----
i00 METER ACCELERATOR TUBE-
Figure 2.4.13. One of t h i r t y SLAC v a c u u m s e c t o r s . (87)
The only vacuum connection between each of the sectors existed at the conductance-limited, disk loaded waveguide (i.e., the accelerator beam pipe) located in the accelerator tunnel. Therefore, from a vacuum standpoint, the sectors were somewhat independent. Each sector was pumped by four, 500 Z/sec diode sputter-ion pumps, equally spaced along a 100 m, 200 mm ~ manifold running the full length of the sector. In early 1967 we observed anomalous pressure bursts in many of these sectors. We suspected argon instabilities. R.S. Callin, of the SLAC Vacuum Group, substantiated this in a series of measurements with a partial pressure analyzer. He reported the following: 1) There were steady-state, pressure gradients in the 100 m manifolds of six sectors evidencing these periodic bursts. The average of pressures at one end of the manifolds was --4 x 10-9 Torr, where the average of pressures at the other end of the troublesome sectors was --3 x 10-8 Torr. We had long suspected there were air leaks in these sectors, but they could not be located. 2) The gas bursts, determined by Callin to be argon, always initiated in the pump at the lowest-pressure end of the 200 mm ~ manifold. During a burst, the pressure would increase to s 5 x 10-s Torr in --15-30 minutes, and then recover to the original pressure in an equivalent or less time. 3) Argon pressure bursts in the pump furthest removed from the air leak would cause one or two of the adjacent pumps to also become unstable. 4) Initially, occurrences of argon instability were random in time. Eventually, they occurred on a regular basis - usually in the mid-morning when the klystron gallery had warmed up from the night before (no heating of the gallery was required in sunny California). It was at first perplexing that the pump furthest removed from the leak was the first to become unstable. We suspected it might have something to do with an argon
108
SPUTFER-ION PUMPING
enrichment phenomenon created by the distributed pumping system. Assume that the pumpin~ speed of the sputter-ion pumps for Ar was --2% of the speed for air.(8 s ,8 8 )"Admittedly, this speed is very difficult to stably quantify in conventional diode pumps.(2 4 ) Assume that the air speed of the pumps is the rated 500 Z/sec; at the lower pressures, the speed is probably -- 100 Z/sec; but, neglect this fall-off in I/P with pressure. Assume there is --1.0 % Ar in the air, and that the inlet leak rate is proportional to the partial pressure of the atmospheric gases. To simplify the problem, assume that, excluding Ar, the remainder of the in-leaking gas is N2. The pressure at the leaking end of the manifold is 3 x 10 "8 Torr. With these assumptions, one can calculate the results of Fig. 2.4.14. That is: 1) There is an Ar enrichment process that occurs at the location of the leak. Where the partial pressure of Ar inthe atmosphere is assumed to be --1%, the partial pressure of Ar near the leak is --20 %. 2) The partial pressure of Ar at the pump furthest removed from the leak is --23 % of the partial pressure of Ar at the pump nearest the leak. But, 3) the Ar pressure at the furthest removed pump is --99.7 % of the total pressure, neglecting system background gases. Each of these pumps had sixteen elements identical to the element described at the beginning of Section 2.4.3. Assuming the pumps operated two years prior to the onset of instabilities, calculations show that --5 x 10 -2 Torr-Z of argon was sufficient to induce instabilities in a single element. However, in this case, varying the pump voltage did not eliminate the instability problem. From this, we concluded that the pumping of chemically active gases mixed with the in-leaking Ar aided in the pumping and retention of Ar. The noble gas stabilization effect of pumping mixed gases has been reported elsewhere.(8 9,4 8,2 4 ) The argon enrichment phenomenon, stemming from a linear system configuration, has not. 10--7
10 3
TOTAL
o
10
~q
10
/I
z o b10 2 o o
>
I
~ PRESSURE
/
--
-9
10 1 Z N a PARTIAL -~ !PRESSURE_
Q9
-10
10
lo o ~2: I
It
-Ii
~
0[...,
RATINEOFuff~s/N 2 10-1
-
-12 ~ 10 0
CLOSEST
/- Znd PUMP
PUMP
~ 3 r d PUMP-~ 33
DISTANCE
PUMP -~-
67
FROM AIR LEAK -
F i g u r e 2.4.14. Argon e n r i c h m e n t a t one e n d of a m a n i f o l d along
FURTHEST~,
o <
m
100 meters
a s s o c i a t e d with an a i r l e a k which pumps are distributed.
Diodes with Slotted Cathodes Several schemes were developed for the stable pumping of noble gases. It was found that if slots were machined into the cathode)plates, stable Ar pumping could be achieved up to pressures of--10 -5 Torr.(8 5 A patent application was made by Jepsen on such a pump,(9 0 ) only two years after publication of results of the first
SPUTTER-ION PUMPING
109
multi-celled pump.( T a ) This slotted cathode pump, as described in Jepsen's patent, is shown in Fig. 2.4.15. mm CATHODE \ \ \
\
TITANIUM
ANODE ",
ATOMS -~
SLOTTED ~ CATHODE
F i g u r e 2.4.15.
APPED
Slotted
titanium
cathodes
for noble
g a s p u m p i n g . (90)
It was assumed that, because most of the ions striking the cathodes do so at large angles of incidence, the sputter-yield was greater. This had the effect of increasing r/, the pumping efficiency. Also, it was believed that noble gas molecules, weakly physisorbed in the lower regions of the troughs formed by the slots, would be physicaUy trapped by subsequently sputtered material. The slotted-cathode pump did not prove to be the perfect panacea for noble gas pumping with sputter-ion pumps. For example, Vaumoron showed that these pumps would also become unstable when pumping on a 20% Ar, 80% N2 mixture, at pressures in the 10-5 Torr range.(9 t ) However, these pumps could pump on air leaks for extended periods without the evidence of noble gas instabilities.(8 s ) This improved stability was later discovered as resulting from the burial (i.e., physisorption) of high-energy, neutral atoms. But, this is discussed below. The slots had to be machined in each of the cathode plates. This was an expensive operation. The slotted diode pump was short-lived, as less expensive alternate solutions were soon discovered.
The Galaxy Diode Pump Over forty years after the invention of l~umps with slotted cathodes, Rutherford invented the Galaxy~ diode pump.(2 s s) The configuration of a single-cell diode pump is illustrated in Fig. 1.4.17A. Rather than using an expensive machining operation to create slots, spiral grooves, passing through the entire material thickness, were machined in the cathodes using water-jet cutting. The cutting action stems from the use of high pressure water jet containing an abrasive slurry. According to Rutherford, the spirals can be cut using computer-aided manufacturing (CAM) from a CAD (computer-aided drawing). CAM, CAD and water-jet cutting did not exist forty years ago. Rutherford's spiral slots serve the same purpose as the troughs in Jepsen's slotted cathodes. However, assemblies comprising multiple anodes, subtended by the galaxy-like grooves in the cathodes can be used as either diode or triode assemblies. Galaxy~ diode pump speed characteristics for all gases, including argon, are reported as very similar to performance observed with Jepsen's diodes with slotted cathodes.
110
SPUTTER-ION PUMPING
Spiral Slot Water-Jet Cut in Cathode /
Walls of Pump Body
................ L .... ~ .... ~ /~r ~t~~H~t~ ~
i
Anode ~Cylinder
~Titanium
Cathode s h Spiral Slots
Figure 2.4.15A. Single-cell Galaxy @ pump invented by Rutherford in 1996.( 255)
2.4.4 The Triode Pump Vanderslice made patent application for the first triode sputter-ion pump in September 1958.(9 2 ) It was a sputter-ion pump with positive anode, negative cathode and slightly more negative collector plate. That is, the anode, cathode and collector, arrayed in this order, were operated at three different voltages. This triode pump effectively pumped noble gases. The cathode was somewhat transparent to bombarding ions and cathode material sputtered by these ions. Some of the ions could pass through the cathode mesh-structure and bombard the collector material, but with negligible sputtering. It was believed that noble gas ions, weakly physisorbed on the collector, were sputtered over with cathode material and trapped within the collector. Similar to the single-cell diode patent of Gurewitsch and Westendorp,(1 9 ) also of the General Electric Company, Vanderslice's patent described a single-cell, triode sputter-ion pump. Clarke, of the General Electric Company, later patented a form of multi-ceU triode pump.(9 3 ) In May of 1959 Brubaker made patent application on the first multi-cell, triode sputter-ion pump.( Q4 ) Later that year he presented a paper on the characteristics of this new triode pump vs. the conventional diode sputter-ion pump.(S 6 ) This was the first comprehensive paper on the noble gas instability problems of conventional diodes. He reported on various manifestations of these instabilities, from violent pressure bursts to, at times, almost sinusoidal variations in pressure. He postulated that Ar, shallowly buried in the cathodes of diode pumps, was liberated by subsequent sputtering of cathode material. He described an experiment where there was a factor of 30 increase in the Ar pumping speed of a single-cell diode as the result of the use of an auxiliary filament to sublime Ti over weakly trapped Ar atoms in a Ti cathode. This over-sputtering of material, as proposed by Herb in 1953,(1 ) was thought to be the main mechanism responsible for the pumping of noble gases. The effect is real as evidenced in the SLAC instability problem (i.e., a mixture rich in N2 will augment the stable pumping of Ar). Brubaker's triode was configured as
SPUTTER-ION PUMPING
111
shown in Fig. 2.4.16. In this pump configuration, the anode is operated at + 3 kV with respect to ground, the cathode at ground potential, and an interposing, somewhat transparent auxiliary cathode grid, at - 3 kV with respect to ground. \
\
ANODE
~
ITs, ' - .
x
_x,
~--CATHODES [ ( F i g u r e 2.4.16. B r u b a k e r ' s m u l t i - v o l t a g e t r i o d e p u m p . The c a t h o d e m a t e r i a l , n o t s p e c i f i e d , was p r o b a b l y t i t a n i u m . ( 5 6 ) ..-~'CELLULAR" AUXILIARY CATHODES
P
Brubaker reported triode speeds for Ar of about 25% that of air. He also reported that the triode had four times the pumping speed for air as that of the diode. Though the former results have proven to be the case, the latter has not for similarly sized pumps. In fact, for reasons to be discussed, triode pumps typically have less pumping speed for air, per unit pump volume, than do diode pumps. However, Brubaker was the first to note improved pumping of triode pumps for Ar. He reported improved pumping as "... ions incident at a grazing angle do more sputtering than those which strike the surfaces at normal incidence." He also pointed out the apparent ease with which argon instabilities could be initiated in a diode pump, with slight variations in voltage, where triode pumps seemed much more immune to this phenomenon. The triode pump was effective in pumping noble gases. But, the correct model for the pumping of these gases remained a mystery. Two years after the publication of Brubaker's paper, Hamilton, of the Consolidated Vacuum Corporation, discovered a strange phenomenon.(4 7 ) He varied the electrode potentials of a triode pump, as shown in Fig. 2.4.17, and measured associated changes in pump speed. CATHODES ~. (CELLULAR Ti) ~
. . . . . . . .
oq) 35
I
I
z5
o
l
I
+6.5 kV--
f
Z
ANoor
1
_~fCATHODE ~" 30 r POTENTIAL I
i I H ! i I
ANODE ~. - ~ POTENTIAL ~,
10 COLLECTOR J (PUMP WALL)
Figure 2.4.17. Hamilton's triode speed results as a function
0
1
2
3
1
t
4
5
COLLECTOR VOLTAGE -
6
kilovolts
pump configuration and argon of e l e c t r o d e potentials.(47)
Hamilton determined that if the electrodes of the triode pump furthest removed from the anode (i.e., Brubaker's cathodes) were operated at the same potential as the anode, pumping speed for the gases, including Ar, was comparable to that observed with Brubaker's potential settings. These results caused quite a stir in the industry. Of course, the industry immediately adopted the much simpler, single potential
112
SPUTTER-ION PUMPING
power supplies for use with triode pumps. But the perplexing aspect of Hamilton's results was that the proposed mechanism for triode pumping of noble gases went out the window with his findings. That is, it was believed that Ar ions were shallowly buried in the equivalent of Brubaker's cathodes (i.e., see Fig. 2.4.16) - the pump walls, in Hamilton's case - and subsequently covered over by material sputtered from the auxiliary cathodes. Hamilton's results indicated ion burial in the cathodes (i.e., pump walls) could not be occurring, as the kinetic energy of any ion reaching the cathodes had to be zero, assuming no significant rf energy was being generated in the discharge. Through use of autoradiography, Andrew and others reported that a significant amount of the noble gases were pumped at the anodes.(S a ,~ s ,~ 6 ,a r ). This too seemed energetically impossible, for how could gas ions reach the anode? Rutherford speculated that anode pumping was due to rf voltages present in the discharge.(S s ) The triode pump effectively pumped noble gases, and diodes to some extent, but an understanding of the mechanism was still not at hand. The Varian StarCell ® triode pump, first reported on by Pierini and Dolcino in 1983,(9 s ) is the most recent development in triode pumps.(9 9,1 o o, 1 o 1 ) The conventional triode (i.e., a pump with the Hamilton potentials, hereafter) was expensive to manufacture because of the high labor content associated with assembling the grid-cathode module. Each Ti cathode strip had to be individually mounted in the stn. stl. cathode frame. Also, after extended use at high pressures or the extended pumping of gases containing H2 (e.g., H2 O), one or more of the grid strips tended to burn through and short to the pump wall or anode. I
,~
"~
i
I ~/)f~-----~~/}~,~--~zt~J~ ~ ~ i1
/
~~(~k.~~~(~ ~
/-:~b~,~'~'~ ~ • \ Ill ~J< ~ ~
mm
~
/
I
~
I
e00~ /
" mm
~
/ I
I
I
I -I-
I
I
\ SIX
"STARS"
MESHINGLAMINATES
i
~CATHOuD~BEN ~LA~ INATE
i
"STARS; /~-..~ BEND
-___" . . . . . . . . . . . . . . . [ ~ I / z5 4 I
\
l\\
I \/BI-LAMINATE
CATHODE (LOWER)
Figure 2.4.18. Pierini and Dolcino's StarCell ® pump element. (98)
SPUIq'ER-ION PUMPING
113
The StarCell ® pump configuration, shown schematically in Fig. 2.4.18, eliminated the cathode strip shorting problems of the conventional triode pump. Also, the conventional triode pumps, after pumping H2 for extended periods, are subject to thermal run-away problems.(t o 2 ) This problem, discussed below, is somewhat ameliorated by the StarCell® cathode configuration (i.e., due to the extended cathode surface area of this pump). However, the primary advantage of this pump configuration is in the reduction of triode manufacturing costs. Special tooling made possible the formation of the stars in the cathode plates. This investment in tooling eliminated the repetitive, high labor costs associated with assembling the grid assemblies of the conventional triode pumps. These labor savings can be passed on to the buyer. However, replacement cathodes must be purchased from the original equipment manufacturer.
Magnetic Fields in Diode and Triode Pumps Subsequent to Hamilton's discovery,(4 z ) all commercial triode pumps were manufactured and sold with the cathodes operated negative with respect to ground, and the anodes and pump bodies at ground potential. Brubaker's cathode was eliminated, as the walls of the pump body served this function thereafter. For reasons which will become evident, eliminating the use of titanium pump walls may have been a fundamental error. Operating the cathodes at a negative potential required that, for electrical isolation, additional spacings, g, had to be provided between the cathodes and pump body (e.g., see Fig. 2.4.19). Therefore, for the same cell dimensions (i.e., £ and d) the magnetic field gap would have to be greater for the triode pump. A larger magnetic field gap, for the same amount of magnetic material, would result in a decrease in magnetic field for the triode. GROUNDED
PUMP
CATHODE/ / - ~
I _ _
"POCKET" GAP
__
_ _ _ . _ ~ / / / /
/
/
/
/
/
/
/
/
TITANIUM GRID CATHODES
~
J
~OL~T[0~~n ................... ..... ........ m m•
t "
mm
-
I l I° ° I I D Ill ° ° I B nllll .
.
.
.
.
.
.
.
...... ANODE
NDER
+
F i g u r e 2.4.19. C o n v e n t i o n a l t r i o d e a n d d i o d e s p u t t e r - i o n pumps c o n f i g u r e d to be i n s t a l l e d in p u m p " p o c k e t s " of t h e s a m e size.
Assume that the magnetic circuitry of the diode pump was designed so that the magnetic field was optimum for the maximum I/P (i.e., near the critical magnetic field of Fig. 2.2.3). An increase in the magnetic field gap of the pump would result in a corresponding decrease in magnetic field and consequential decrease in triode pump speed. We see that because of the need for a larger magnetic field gap, the speed of the triode pump would be less than that of the diode. Some manufacturers sold both diode and triode pump configurations. The above considerations suggested that the size of the diode pump pockets would have
114
SPUq"I'ER-ION PUMPING
to be different from pockets used for the triodes. Also, the fall-off in the triode magnetic field had to be addressed. Therefore, for reasons having to do with both the economics and the physics of sputter-ion pumps, these manufacturers decided to make all pumps have the same size pockets and, therefore, the same magnetic field. Anodes of triode pumps were shortened to compensate for the two additional electrical gaps and to make use of the same pockets and magnetic field circuits used with diode pumps. This resulted in a triode configuration comparable to that shown to the far right of Fig. 2.4.19. The value of I / P is directly proportional to £, the anode length. Therefore, a reduction in anode length in the triode pump resulted in a corresponding reduction in the speed of the pump. It was known that the fall-off in speed of sputter-ion pumps with a decrease in cell diameter becomes significant only at low pressures. This is shown in Fig. 2.2.7. Therefore, in order to increase the high-pressure speed of the triode pumps, the diameter of the cells was decreased so that more cells could be packed into the same volume. Because of this, at high pressures, the pumping speed per unit volume of the triode pump is very comparable to the diode pump. However, at lower pressures and with nominal magnetic fields, this is not the case.
Field Emission in Triode Pumps There appears to be a tendency for triode pumps to have problems with field emission from the cathodes.(t o 3,9 8, t 0 4 ) This is called a problem as in many applications pump current is used as an indication of system pressure. For example, a new 120 Z/sec triode pump will have of the order of ---100/.t A field emission current on initial start-up. This corresponds to current equivalent to a pressure of - 5 x 10-8 Torr. It is believed that whiskers develop on the grids of the cathodes each time a triode pump is started, and that these whiskers cause the field emission. The whiskers may be burned off by a process called hi-potting - the momentary application of a ---20 kV, AC voltage. Unless the pump is hi-potted, field emission current would obscure indicated pressures inferred by pump current from the Penning discharge. The presence of field emission in a pump may be verified by removing the magnets, and while under high vacuum, measuring the pump current as a function of pressure. If field emission is present, a plot of the ln(/V-2) vs. V - 1 , where V is the pump voltage and I the current, will result in a straight line which is called a FowlerNordheim plot.( 1 o s )
2.4.5 The "Differential" Sputtering Pump In 1965, James and Tom, of Perkin-Elmer (Ultek), applied for patent on a new type of diode sputter-ion pump.(2 9 ) They mistakenly assumed that the current model for the pumping of inert gases, as postulated for triodes, was correct. Therefore, they constructed a pump having two types of cathode materials, as shown in Fig. 2.4.20. One cathode material was reported to be more readily sputtered than the other. In a paper which they published a year later, they assumed that noble gas ions, weakly buried in the cathode with the lowest sputter yield, would be preferentially buried by material sp uttered from th e cat h ode with highest sputter-yield. (1 o 6 ) They also suggested, in this early paper, that "... the high rate of deposition of atoms onto the anode surfaces (due to a high sputter-yield cathode) can substantially increase the pumping of fast neutrals that impinge on the anodes."
SPUTTER-ION PUMPING f/ / / / / / / /
METAL ANODE-~
tl
LOW YIELD ------~t SPUTTERED ATOM [q
/ [ \ © o ./ I \ _._.t,~tlf o I ©+ ~ '-'
ARGON ~- ~ ~r Iq i o TITANIUM
~ / / //
o
o ÷
O
/ / /- /
~TRAPPED BY HIGH SPUTTERYIELD TITANIUM
© ~ [ o ~ A
o
tt/
]"
~t
o
115
..
SPUTTERED Ti ATOM ATOMS FROM
PUTT'A.O0
CATHODE
1
Figure 2.4.20. The "differential ion" ("DI") sputtering model proposed by James and Tom to explain the pumping of a r g o n . (29)
In their patent application James and Tom suggested that one cathode might be constructed of Cu and the other of Ti. They also proposed the use of Ta and Ti cathodes in that patent. Evidently, they assumed that both Cu or Ta had lower sputter-yields than that of Ti. In the first article which they published, no mention was made of the presumably proprietary Ta cathode material.( 1 o 6 ) However, competitors and other researchers had no mercy, and soon the cat was let out of the bag. Three years later, in 1969, James and Tom published results of what came to be known as the D l p u m p (i.e., Differential Ion), with Ta and Ti cathodes.( x o 7 ) These DI pumps very effectively pumped argon, and pumped chemically active gases with speeds only a few percent less than pumps with two Ti cathodes. For example, we at SLAC modified all of the pump elements of the two-mile linear accelerator, using Ti/Ta cathode pairs. Argon instabilities were never again observed at SLAC. However, there were some inconsistencies in the "differential sputtering" model of James and Tom. It was known that, for a broad range of ion energies, Ta and Ti have very comparable sputter-yields. Therefore, the differential sputter-yield model couldn't be correct. The correct model for noble gas pumping was still not fully understood. However, as first suggested by Lafferty in 1961,( 9 s ] people began to speculate that the burial of high energy neutrals must play an important role in the pumping of noble gases. Neither Lafferty nor Tom and James speculated on the origin of these neutrals. Lafferty concluded that some charge exchange process must exist to explain the presence of noble gases pumped in the anode assembly.
The High Energy Neutral Theory The atomic structure of a metal comprises an ordered, three-dimensional array of positive ions of that material. These metal ions are awash in a sea of electrons, so that the net electrical charge of a piece of metal is zero. Some electrons are tightly bound to, and shared by, neighboring ions in the lattice. This gives the lattice a structure unique to the element. Other electrons, though belonging to particular metal ions at any moment, are free to meander about the bulk of the lattice structure. These are called conduction electrons. They stagger their way through the lattice in a preferred direction when an electric potential is applied to the metal. On an atomic scale, the metal ionic sites have a positive charge. When a posi-
116
SPUTTER-ION PUMPING
tive gas ion bombards the surface of a cathode plate, it is confronted by these positive point charges, and it is deflected at some angle, due to mutual coulomb repulsion. If the ion bounces back from the cathode surface, while at the same time picking up an electron from the metal, the ion is electrically neutralized. If, when bouncing back, it also retains much of its original kinetic energy, it is then said to be a high energy neutral. The reflected high energy neutral does not suffer coulomb repulsion on its next encounter with a solid surface. Therefore, as a high energy neutral, it will be more deeply buried into some solid surface in line-of-sight with the primary collision. The higher the kinetic energy of the reflected neutral, the deeper the burial and permanent entrapment in the adjacent pump element, be it cathode or anode. Jepsen was the first to demonstrate that the creation of high energy neutrals at the cathodes, and their subsequent burial on reflection, was the primary mechanism by which noble gases were pumped in all sputter-ion pumps.(1 o 8 ) The model was elegantly simple, appears in many freshman physics textbooks,(S 1 ) and has been verified by the work of a number of investigators (e.g., see: t 0 g, t 1 0, 7 0, 7 4, t 1 1, t z 2, 9 z, 9 o ). The model had to explain the noble gas pumping of diodes with slotted cathodes, triodes, and the Ta/Ti DI pumps. Jepsen calculated the possible angles of deflection and energies of reflected neutrals for a given cathode material. He made the assumption that collisions between gas ions and ions in the metal lattice of the cathode were elastic (i.e., kinetic energy was conserved), and he treated the interaction of the two ions as a free-body problem. For example, assume the bombarding gas ion, represented in Fig. 2.4.21, has an initial kinetic energy of ½m 1 U 2, where m 1 is its mass and u its velocity. Assume that a metal cathode ion, of mass m 2 , has no initial velocity. In the collision process the gas ion will be deflected to an angle 0, and the metal ion to an angle 4', both depending on the impact distance, d, and the mass of each ion. Define the velocity vector of the gas ion after the collision as v , and the velocity vector of the metal ion as s. Conservation of kinetic energy and momentum results in the following equations: 0
=
-mlv
sin0 + m 2 S s i n 4 >
mxU
=
mlv
Y,m x U 2
=
~zm 1 V
(2.4.4)
cosO + m 2 s c o s q )
(2.4.5)
~zm 2 S 2
(2.4.6)
2
-k-
The sin q~ and cos 4) terms may be eliminated by squaring and adding (2.4.4) and (2.4.5). Def'ming R = m 2/m z, solution of v 2/U 2, i.e., the ratio of the energy of the reflected neutral vs. its energy as an incoming ion, is: v 2
{ cos 0 4---(R 2 _ sin 2 0 ) ½} 2
U2
(R + 1) 2
(2.4.7)
The table in Fig. 2.4.21 gives results of (2.4.7), as a function of 0, m 2 and m 1 , for four gases scattered off A~, Ti, Mo and Ta. We see that for normal incidence, neutrals of Ar scattered off Ti have < 10% of the energy of the primary ion. Reflected neutrals of Ar on Ta retain 41% - 64% of their original energy. Of course, when m 2 < m 1, then m 2 < m 1 sin0 implies that neutrals can only be scattered in the forward direction, as in the case of Ar on A~. In this case, the maximum pos-
S P U T t E R - I O N PUMPING
117
sible scattering angle, 0, is --41". This does not mean that Ar cannot be pumped by a pump with A~ cathodes. It merely means that the pump must be designed so that most of the Ar ions strike the cathodes at grazing angles of incidence, as in the case of triode pumps.(7 2 ) GAS 0 : 0 9/8 ?/4 Z?/a ?/Z 57/aa?/47?/a
VELOCITY TARGET
METAL
S --~/
+o-r-g---- .w~--- I ....
\
GAS ION OF MASS r n I
'
N,
¢'~\\ v ---j NEUTRAL GAS ATOM VELOCITY
Figure 2.4.21. Kinetic ion, as a function
energy of the
o
He Ne Ar Kr He Ne Ar Kr He Ne Ar Kr He Ne Ar Kr
1.000 0 . 9 7 8 0 . 9 1 7 0 . 8 3 2 0 . 7 4 2 1.000 0 . 8 9 0 0.621 0 . 3 3 3 0 . 1 4 4 1.000 0.781 . . . . 1.000 . . . . . 1.000.0.907 0.952 0.902 0.046 1.000 0 . 9 3 7 0 . 7 7 7 0 . 5 8 0 0 . 4 0 7 1.000 0 . 8 7 8 0.581 0 . 2 7 3 0.091 1.000 0 . 7 3 6 . . . . 1.000 0 . 9 9 7 0 . 9 8 7 0 . 9 7 3 0 . 9 5 7 1.000 0.98310.937 0.871 0 . 7 9 9 :1.000 0 . 9 6 7 0 . 8 7 8 0 . 7 5 9 0 . 6 3 8 1.000 0 . 9 3 1 i 0 . 7 5 6 0 . 5 4 6 0 . 3 6 7 1.000 0 . 9 9 4 0 . 9 7 6 0 . 9 5 0 0 . 9 3 0 1.000 0.968]0.{]83 0 . 7 6 9 0 . 6 5 3 1.000 0 . 9 3 8 0 . 7 7 9 0 . 5 8 4 0 . 4 1 3 1.000 0 . 8 7 2 0 . 5 6 3 0 . 3 4 4 0 . 0 6 8
?
0.661 0 . 6 0 0 0 . 5 6 3 0.550 0 . 0 6 2 0 . 0 3 3 0 . 0 2 3 0.031 . . . . . . 0.793i0.75110.724 0.715 0.286 0.313:0.177 0.166 0.030 0.014 0.009 0.008 . . . 0.941 0.927 0.918 0.915 0.734 0.682 0.650 0.639 0 . 5 3 7 0 . 4 6 4 0.421 0 . 4 0 7 0.247 0.178 0.145 0.135 0.891 0 . 8 6 7 0 . 8 5 3 0 . 8 4 6 0.554 0.483 0.4400.436 0 . 3 9 1 0 . 3 1 8 0.181 0 . 1 7 0 0.019 0.008 0.005 0.005
ratio of reflected neutral atom to bombarding masses of the gas and target metal atoms.
The Ar pumping characteristics of the conventional diode, DI, slotted diode and triode pumps started to make sense to investigators. In the case of the conventional diode pump (i.e., pumps with two Ti cathodes), the reflected neutrals of Ar ions have very low energies and are therefore shallowly buried in the adjacent anode and cathode. A very slight change in one of the pump parameters, such as pressure, magnetic field (e.g., due to temperature changes), or voltage, can cause a slight modification in the space charge distribution. The magnetic field of a typical sputter-ion pump varies from 30 - 40% across the anode array (of identical cells). This suggests that the discharge modes (i.e., space charge distribution) in, for example, anode cells with high magnetic field could have one mode configuration, where the space charge distribution might differ markedly in cells with the lower magnetic field. Rudnitskii and Dallos have shown that such mode shifting occurs even at low pressures.(S 7,1 o s ) There may also be mode mixing, or unstable moding in cells between the two magnetic field extremes. A slight change in the space charge distribution in a few of the cells, due to some perturbation in a pump parameter, could result in ions bombarding previously unscathed portions of the cathodes. From these areas, weakly buried neutrals would be subsequently liberated. This is tantamount to a negative pumping speed for the gas.(1 1 s ) The liberated neutrals in effect increase the local pressure in the cell. This gas in turn is ionized, resulting in an increase in the intensity of ion bombardment of the cathodes, etc. This is in essence a positive feedback situation, which, depending on the amount of weakly buried Ar, can result in pressure runaway. If the system is damped by the presence of a second stabilizing pump, the feedback may be slightly negative, and may result in a sinusoidal instability, or dithering effect in pressure. (I have also seen H ~ induced, sinusoidal pressure variations with a single pump, pumping on a closed system.(4 s ) In this case, the pump self-stabilized due to changes in the discharge characteristics, pumping the H 2 elsewhere on the cathodes.) In the case of diode pumps with Ta/Ti cathodes, neutrals reflected from the Ta cathodes have energies of the same order as the primary ions, and are deeply buried in the anode and adjacent cathode. Therefore, with slight changes in the discharge,
118
SPUTTER-ION PUMPING
there is less probability that previously buried neutrals will be desorbed. In the case of triode pumps, the choice of cathode material is of less importance in the pumping of noble gases. This was demonstrated by Liu's experiments with A~ cathodes,(T 2 ) and in earlier work by Hall, when using this material as part of a triode cathode assembly.(1 1 4 ) Most of the ions impinging on the cathode grids do so at grazing incidence. Therefore, regardless of the cathode material used, a deflected, neutralized ion retains much of its original kinetic energy and becomes buried in the walls of the pump. A similar situation exists with the slotted diode pump. The grazing angle of incidence of ions on the slots of the cathodes makes possible the creation of high energy neutrals which in turn are buried and sputtered over at the bottom of the slot trenches. However, because it is possible in time to completely sputter through the cathode plates of a slotted diode pump, it is possible for Ar to be later liberated and cause instabilities. This is not possible with the triode pump (i.e., a pump using Hamilton's potentials). In this case, the potential of the anode is the same as the walls of the pump body. This means that the kinetic energy of even the most energetic ion created in the discharge will be zero at the walls. Therefore, there is no ion sputtering of the pump walls, and it is impossible for ions to desorb previously pumped gas. The walls of the pump body apparently are not appreciably sputtered by energetic neutrals. In 1970, Vaumoron and De Biasio conducted an extensive series of experiments which put Jepsen's fast neutral theory to the test.(1 1 s ) They tested the theory with cathode materials including Ti, Cu, Mo, Ag and Ta, and for the gases He, Ne, Ar, Kr and Xe, in some 22 combinations of materials and gases. They found excellent correlation with the fast neutral theory, at pressures < 10-4 Torr, providing R z 2.4. Baker and Laurenson conducted noble gas pumping experiments with cathode material xpairs of A£/Ta, Ti/Ti and Ti/Ta and for the gases He, Ne, Ar and Kr.(1 1 6 ) These results also verified Jepsen's fast neutral theory, though intermittent pressure pulses were noted when pumping Ne, Ar and Kr with the A3./Ta cathode pair. Where Vaumoron's tests were conducted at steady-state pressures, Baker's tests involved deducing the speed of different cathode pair materials from the rate of pressure pump-down of the noble gas. The advent of the slotted diode, DI and triode sputter-ion pumps overcame one major limitation of these pumps; that is, the pumping of noble gases.
Beware of "Shortcuts" A note of caution is given regarding noble gas pumping provisions. Because of the fall-off in speed noted for triode pumps at the lower pressures, some would prefer to use diode pumps for their low-pressure applications. However, because of the cost of Ta, it has been suggested that some of the elements of the diode pump might have the conventional Ti/Ti cathode pairs, where the others might be constructed with TafI'i elements. No data has been published which suggests that such mixing of elements in a given pump or system will afford immunity to instabilities in those elements with the Tiffi cathode assemblies. Granted, the system will be more tolerant to Ar leaks, as the preponderance of Ar will be pumped by the noble elements. But, the conventional diode elements (i.e., elements with Tiffi cathodes) may still become unstable after pumping a given amount of Ar. Also, the use of auxiliary pumping, such as a cryopump, DI or triode pump, to pump on a system once evidencing argon instabilities, may not cause the problem to
SPUTTER-ION PUMPING
119
go away. Such was the case when argon instabilities were noted in LINAC pumps at BNL (the Brookhaven National Laboratory). We first observed that argon instabilities existed in a number of conventional diode pumps used to pump accelerator modules (tanks). We sought to determine if instabilities would persist in some of the pumps, if Ti/Ta elements were substituted in other pumps on the same tank. We used a cryo-pump as an auxiliary pump. It was coupled to one of the tanks so as to afford a speed for Ar of---10 z Z/sec. The argon instabilities, verified with a quadrupole residual gas analyzer, persisted thereafter.
2.5 Hydrogen Pumping Hydrogen is prominent in all UHV systems. Because of this, the pumping characteristics of sputter-ion pumps for this gas have been extensively studied. For example, microwave tube components are frequently brazed in hydrogen furnaces. This process results in large amounts of H 2 being liberated during the final bakeout of these tubes. Also, both fusion(1 t 7 ) and fission work gave further impetus to the study of the H 2 pumping capabilities of sputter-ion pumps.(1 1 8,1 1 9 ) A s of 1973 Singletonhaddone the most definitive work on the sputter-ion pumping of H2 to 2,1 2 1 ) date.(1 2 _ As noted in Fig. 2.4.8, H 2 is primarily pumped by diffusion into the cathodes of the pump. But, as you will soon see, the diffusion process is rather complicated. Excluding Li, Ca, and Na, titanium and alloys of this metal have the greatest sorption capacity for H 2 , ( 1 2 3,1 2 4) and to first order, this is completely reversible.(1 2 s,7 1,1 o 2,1 2 3 ) I noted earlier that Gurewitsch and Westendorp, the first to report on use of Ti as a cathode material, used this material for this very reason.(1 9 ) In order for there to be substantive diffusion of gas in metals, three requirements must be met: 1) the metal must have a chemical affinity for the gas; 2) the gas must first dissociate into an atomic state; and 3) there must be a concentration gradient of the atomic species to promote its diffusion in the bulk material.(1 2 6 ) However, as will be shown, gases may be artificially implanted into materials for which they have a low chemical affinity, and diffusion forced by contrived concentration gradients. Only -2.5% of the ions created in a low-pressure H2 Penning discharge are H + ions. Therefore, in order for H2 to be significantly pumped, H2 + ions must first dissociate on or near the cathode surfaces. Also, if the surfaces are clean, H 2 molecules similarly dissociate at the cathodes. This requires that there be surface sites which will promote this dissociation. These are sometimes called energy sites. This merely means that there are locations on the cathode where the value of the activation energy, Ea, of Fig. 2.4.7, is sufficiently depressed so that molecular dissociation and chemisorption is promoted. If the surfaces of the cathodes are contaminated by the presence of, for example, TiN or oxides of Ti, E a can become sufficiently positive where the surface dissociation of the H 2 + ions will not occur.(4 g ) Gas contamination will also inhibit the subsequent surface release of hydrogen in solution with a metal.(1 2 r ) The energy sites are created by the sputter-cleaning of the cathode surface. However, even at high energies, the lighter H2 + ions have poor sputter-yield.(1 2 8,1 2 9 ,s z ) More specifically, at 7.0 keV, the sputter-yield of H 2 + on Ti is only 0.01.(1 3 0 ) Because of this, when introducing pure H 2 into a pump, it may take a long time for the surfaces to be scrubbed clean by the bombarding H 2 + ions. Therefore, the pump speed may initially be very low for H 2 and
120
SPUITER-ION PUMPING
gradually increase with time. Eventually, as the cathode hydrogen concentration gradient equilibrates, a steady-state speed will be achieved which will be --50% of the maximum speed (see Fig. 2.5.3). It is important to note that the high pumping speed for molecular H2 is only temporary with Ti cathodes at high pressures. However, this effect can be long term with cathodes of Ti alloys. Rutherford and Jepsen used noble gas ions as a scrubbing agent to enhance the pumping of H2, reporting this effect in 1961.( 4 9,1 a t ) They used Ar for this purpose, though this has been known to have disadvantageous side effects, as reported by Dean.(1 a 2 ) Dean used this technique to reduce the total system pressure, comprising mostly H2, in a low-energy, sputter-ion pumped electron storage ring. However, he recanted doing so, as the beam decay times of an electron storage ring vary to first order with Z 2 , where Z is the atomic number of the residual gas species. Though the total pressure in the ring was decreased by scrubbing the pump with Ar, because of its higher Z, the beam decay effects from the residual Ar proved to outweigh the beam lifetime benefits of the reduced pressure. Dean thereafter used He as a scrubbing agent. Though it has a much reduced sputter-yield, the net benefits were reported to be positive. Rutherford reported on the sustained H2 pumping of scrubbed Ti cathodes even after removing the high voltage to the pump,(4 g ) observing evidence of pumping almost two hours later. In light of the negligible benefits which might have stemmed from the prior sputtering of the cathodes by H2, this was conclusive evidence of the pumping of H2 by a diffusion process. Singleton noted that this sustained pumping effect did not exist when scrubbing the cathodes with N 2 ions.(1 2 t ) He, and others, believed H 2 pumping probably ceased, on turning off the power, only because the cathodes became contaminated with N2 or other gases.(4 g, 1 0 2 ) Also, he noted that with the simultaneous pumping of CO and H2, the speed for H2 decreased over that observed when pumping pure H 2. Audi reported a similar effect, ha certain pressure intervals, with the simultaneous pumping of N 2 and H 2 with a triode pump.(5 o ) Cummings also noted that the H 2 speed of a distributed ion pump, when pumping H 2 and CO, was significantly reduced.(1 3 3 ) Argon not only has a high sputter-yield, but it creates ruptures or dislocations in the cathode metal lattice. This damage tends to increase the effective surface area of the cathode, and also creates conduits for the diffusion of hydrogen into the bulk of the material. The pumping of pure H 2 will also cause cracks and rifts in Ti cathodes. But, this is for different reasons.(1 3 1 ) In this case, changes in the crystalographic structure occur with differing H2 concentrations in the metal. In other words, there are volume changes in the cathode material, which take place on a macroscopic scale, as the result of variations in the atomic hydrogen concentrations in the Ti. During the sustained pumping of H 2, these surface rifts increase the effective surface area of the cathodes. However, Singleton and Audi have shown that the gains in H 2 pumping will vanish on the subsequent and sometimes simultaneous exposure to other active gases.( 1 o 1,1 o 2 ) Diffusion into the cathodes is a surface area and diffusivity-limited process. In an effort to improve on both, Jepsen constructed cathodes of sintered Ti granules, with particle dimensions < 0.2 mm ~.(1 3 4 ) The speed per unit area of this sintered Ti cathode was x 7 greater than pumps with conventional Ti cathodes, and decayed to a steady-state speed of half this value, 25 minutes after removal of pump power. If the sintered granules formed hemispheres, the maximum effective increase in cathode surface area would be - x 2 . Therefore, diffusion effects, into the bulk of the
SPUTI'ER-ION PUMPING
121
sintered matrix, must have played a significant role in the enhanced H2 pumping. This technology, then primarily having application in fusion-related programs, was not pursued commercially. New UHV applications have encouraged the revisiting of this technology in the form of NEG (nonevaporable getter) pumps.
2.5.1 Hydrogen Pumping in Diode Pumps In the early 1970's, Vissers experimented with various schemes for detecting the presence of hydrogen in molten Na heat exchanger loops of reactors.(1 1 8 ) These Na loops interfaced with secondary water cooling loops. If a water leak developed at the interface - a serious problem in these systems - the concentration of nascent hydrogen would increase in the Na-cooled loop. Vissers developed a water leak monitor to detect changes in the concentration of hydrogen in these liquid Na loops. It comprised a sputter-ion pump attached to a reinforced, Ni diaphragm which in turn was inserted into the molten Na stream. Nascent hydrogen in the molten Na, stemming from a water leak, would diffuse through the Ni diaphragm and into the sputter-ion pump. One could therefore pump the hydrogen, while at the same time use the pump current as an alarm, to indicate a water leak into the Na loop. One problem with this scheme was that the speed of the pump for H ~ varied so widely, that it was difficult to do quantitative measurements. For example, if the speed was initially low, the H 2 pressure indicated by pump current would have the value L If, however, the speed increased in time by x 3, for the same H 2 throughput, the current for the new pressure would be 1/3. Variations in H2 speed of this magnitude are quite possible.(1 2 0 ) This implied that one could not rely on pump current readings as an accurate indication of the magnitude of a water leak into the Na loop. A second problem had to do with increasing the capacity of the pump for H2, so as to prolong the life of the instrument. Though the idea for the original instrument was Vissers', people at Varian later helped in this effort by continuing research on H~ pumping. Hill experimented with and made patent application on the use of various alloys of Ti in sputter-ion pumps, specifically aimed at the waterleak monitor application. In a patent application claim in 1975, he noted that greater hydrogen capacity is to be achieved by the use of one of the Ti alloys, "... stabilized to maintain it in the body-centered cubic (bcc) crystal lattice form" (presumably, rather than the hexagonal, close-packed (hcp) form of pure Ti).(1 z s ) In 1967 Lamont reported on the enhanced sputtering of bcc cathode materials, when he made patent application on what became known as the diode post pump.(t 3 6 ) (He gave credit to Brubaker, who first used posts in a triode configuration.(S 6 )) The posts in Lamont's pump were grounded to the cathodes and protruded part way into the anode volumes. He claimed that the use of bcc materials enhanced cathode (i.e., post) sputtering over that of the pure hcp Ti. I experimented with variations of Vissers' H2 monitor, and developed an instrument which was more immune to variations in pump speed.(1 1 9 ) The monitor was formed by putting two Penning cell assemblies in series, separated by a small aperture. Cells close to the H2 source had A l cathodes, with reduced pumping,(7 3 ,t 2 o ,1 z 7 ) where the cell furthest removed from the source had Ti cathodes. If the current to the two pumps was summed, there would be an amplification effect which nulled out, for the greater part, current changes stemming from changes in pump speed. However, quantitative tests still had to be conducted on a variety of Ti alloys to address the H2 capacity problem of the water-leak monitor. I was then
122
SPUTTER-ION PUMPING
managing the Pump and Instrument R&D Department at Varian. We were all wonking managers, and I took on the cathode materials study program. We had to con-
struct some sort of materials test vehicle with the following features: 1)it had to be bakeable; 2) it had to be a diode configuration (triodes have problems in pumping large quantities of H z ) ; 3) it had to facilitate measuring and controlling cathode temperature; 4) replacement of the cathodes had to be possible without welding or machining; and, 5) we needed the assurance that all boundary conditions would remain the same except the one variable, the choice of cathode material. The above conditions prompted the development of the Varian H ~ Q @ pump as the test vehicle, and for which I submitted a patent application shortly thereafter.(' ) The pump was constructed as shown in Fig. 2.5.1.
SPU'Iq'ER-ION PUMPING
123
The cathodes were sealed to the pump body with ConFlat~-type seals, as shown in the blow-up in the figure. Titanium water-cooling tubes were vacuum brazed on the outside surfaces of the 4.7 mm thick cathode plates (1). The magnetic circuitry was formed by the pump body (2), the cathode clamp rings (3) and the steel end-caps (4), which captured the ferrite magnet disks (5). The total surface area of both cathodes under the projected area of the total anode assembly (i.e., not just the anode cell diameters) was --190 cm 2 . At a pressure of--2 x 10"4 Torr, the total pump sensitivity for N2 was --400 Afforr, dropping to - 7 0 % of this value at --3 x 10-8 Torr. The pump sensitivity was --120 A/Torr for H2 at --2 x 10-4 Torr. The magnetic field varied between 0.09 - 0.13 T across the 19, 21 mm ~ x 28 rnm long anode cells. If you are a stickler for detail, you will note that the pump shown in Fig. 2.5.1 has 19 anode cells. In the final design, one cell was removed to make room for an anode support bracket. The anode structure shown in the figure had a mechanical resonance which caused it to self-destruct when in transport (i.e., while in the trunk of a salesman's automobile). Data apply to the 19-cell configuration. The pump operated at a voltage of 7.5 kV, decreasing by only - 7 % at a N2 pressure of 10-3 Torr (i.e., when used with a robust, matching power supply). We initially experimented with anodes comprising several plates, with aligned holes. Anodes, constructed in this manner, were first suggested by Hall, et al., in 1957.(4 ) Though such anodes have found wide acceptance in distributed ion pumps, used in electron-positron and other storage rings,( ~' s ,2 8,1 3 9,7 0,1 4 0 ) we found them unacceptable for the HiQ ® pump applications, as the anode plates could be melted, due to electron bombardment, during the glow discharge starting of the pump. With this test vehicle, we conducted tests on a number of cathode materials. The first test was with an alloy comprising Ti-10V-11Cr-3A~. It was a variation of the more common alloy, Ti-13V-11Cr-3A~, reported on by Hill.(1 3 s ) It proved to be the most promising of the materials tested in terms of H 2 capacity. We were able to pump ... 16,000 Torr-Z of H 2 with this cathode material, at a steady-state pressure of -2.0 x 10-4 Torr, and with the near-constant speed of 50 Z/sec. To appreciate the significance of this number, tests of others are cited. For example, Singleton reported pumping, with some decay in speed, the equivalent of ---103 monolayers of H2, with a single-cell pump with Ti cathodes.( 1 2 o ) The Ti-10V-11Cr-3A~ tests corresponded to pumping the equivalent of -1.8 x 10 6 monolayers of H2. In 1980, Lan Zeng-Rui and Lu Ming reported on the H2 pumping of pump having one cathode of Ti and the second of an unspecified Ti-Mo alloy.(a4 1 ) They reported pumping -6.7 Torr-Z of H 2/cm 2 prior to discontinuing their tests, while noting that the speed of their pump did not "... decrease." (The Mo serves as a ~-stabilizer in Ti alloys.(1 4 2 )) In tests with the Ti-10V-11Cr-3A~ alloy cathodes, a total o f - 8 4 . 2 Torr-Z of H2/cm 2 was pumped prior to end of life. McCracken and Maple measured trapping efficiencies of 10-30 keV H + ions on 0.25 mm thick Ti, Ta, Zr and Mo "cathodes".(7 1 ) They noted negligible change in trapping efficiencies after accumulating the equivalent of --10 ~ 9 H + ions/cm2 on Ta, Zr and Ti. Tests with the Ti-10V-11Cr-3A~ alloy led to trapping the equivalent o f - 5 . 5 x 10 2 1 H + ions/cm2, prior to end of life. In all tests, the quantity of H2 pumped was determined by measuring the value of VdP/dt of a known volume, pressurized with H 2, and attached to the system through a variable leak. One would think these results with the Ti-10V-11Cr-3A~ material to be a smashing success. Rather, they were a shattering failure, as the restrained cathodes, because of hydrogen embrittlement, tended to spontaneously shatter into a dozen or
124
S P U T r E R - I O N PUMPING
so pieces at the end of life. This presented some difficulties, in light of the HiO ® pump configuration. Also, though we first thought the Ti-10V-11Cr-3A~ material to be a common, readily available alloy, about this same time we learned that it was the left-overs of a mill-run made for an air-frame manufacturer; we thereafter called this material unobtanium. We speculated that the more common Ti-13V-11Cr-3A~ alloy would be no more promising (i.e., less shattering). This was a double disappointment. But, in truth, this is the way we all muddle our way through development programs.
Figure 2.5.1A Ti-10V--llCr-3AI cathode of a HiO ® pump after pumping -- 16,000 Torr-Z of H 2. Being optimists, we started to view the test vehicle (i.e., the pump) as having as much, if not more, market potential than the materials thus far tested. Therefore, we continued to evaluate the high throughput pumping capabilities of the pump, for various gases, and while using cathodes made of pure Ti. In the meantime, we explored the possibility of using other materials. One material looked particularly attractive, from an availability standpoint. This material, Ti-6Ai-4V (6-4 material, hereafter), was likened to 1020C carbon steel used in the farm implement business (i.e., in terms of its availability). It was a desperate stab in the dark, but cathodes of this material were fabricated and installed in several test vehicles. The results proved most promising. The H 2 speed of this 6-4 material, compared with pure Ti, is given in Fig. 2.5.2.
SPUTTER-ION PUMPING SI 40
P) l
I I I
1
1 I I
1
--
1 I II
I
I I [
l
"'l t 1
1
I il ~
l
I Jl
I
//-
HYDROGEN SPEED Ti-6A/-4V l
30
~
125
eo
I
10NITROGEN SPEED T i - 6 A i -4V
i
I II
0 10-a
i LL
10 - 7
1
""
I 11
10 - 6
l
t Ill
10 -5
PUMP
PRESSURE
Figure 2.5.2. Pumping speed cathodes and for Ti-alloy
1
I I[
10 - 4 -
10 - a
10
-2
TORR
for hydrogen hydrogen with
and
nitrogen with Ti c a t h o d e s .
pure
When pumping H2 with the pure Ti cathodes, end of life was manifest as a sinusoidal instability. The onset of these instabilities occurred after pumping only - 1 0 a Torr-Z of H 2 at a pressure o f - 1.0 x 10 -4 Torr. By reducing the pressure to - 7 x 10 -6 Torr, additional H2 could be pumped before the cathodes became saturated. This suggested, as indicated in Fig. 2.5.2, that H 2 pumping was being diffusion limited at the higher pressures. Torr-~
X/see Z
0 CD r.~
10,000
80
r.~
7,500
60 r,~
I
\/-"c" 40
J ~
- ~ l
r~
POINT ,, ~-
Z
~
POINT TOTAL HYDROGEN
J e f f -
/
'
>, 5,000
! POINT
~ ~--~- . . . .
~
o
UV,ROGENPUMPING
0
.
-"NO ""
""
~
.---...
_
20 D ~1.8 x 10 -° e m Z / s e c
i
[ 0
100
I
t
200
300
LIFE-TEST
400
HOURS
Figure 2.5.3. Speed and quantity of hydrogen pumped with Ti-6A1-4V cathode material.
Z
500
2,500
O
126
SPUTFER-ION PUMPING
Note that at H2 pressures ~ 1.7 x 10 "s Torr, the speed of pumps with Ti cathodes starts to decrease with increasing pressure. The test vehicle precluded this reduction in speed as being due to either thermal effects or changes in the discharge characteristics. Therefore, I conclude that this must be the onset of diffusion-limitations of the pure Ti. The maximum H~ capacity observed with Ti cathodes, for operating pressures of 7.0 x 10 -6 to --1.0 x 10 -4 Torr, and cathode temperatures ranging from 70 - 200* C, was - 3 x 10 a Torr-Z. Figure 2.5.3 shows the change in speed of the 6-4 material as a function of the quantity of H ~ pumped. The H 2 capacity of pumps with the 6-4 cathodes, ranged from -1.05 x 10 4 (at a pressure of -2.5 x 10 -4 Torr, and with water cooling) to 1.3 x 10 4 Torr-Z (at a pressure of -4.0 x 10 "s Torr, and without cooling). The onset of the sharp decrease in H~ speed at --10 "a Torr, with the 6-4 material is not due to diffusion-limitation effects. Rather, it is due to a mode change in the discharge characteristics of the Penning cells.(1 4 a ) The onset of this mode transformation was at a pressure of -2.3 x 10 -4 Torr when pumping pure N2. An hysteresis effect was not observed. The H : speed of the pump with the 6-4 material at pressures ~ 3 x 10 -4 Torr was probably not steady-state, as the capacity of the 6-4 cathodes could be increased slightly by reducing the operating pressure to --4 x 10 "s Torr (an arbitrary reduction in operating pressure). A variation in apparent H2 speed, with operating pressure, was not observed when using the unobtanium cathodes, suggesting a much higher diffusivity in this totally, f~-stabilized, bcc material. In pumps with the 6-4 cathodes, end of H~ life was manifest as an abrupt decrease in pump speed, rather than the dramatic implosion with the unobtanium cathodes. A summary of transient and steady-state pumping results with the pure Ti and 6-4 cathodes is given in Table 2.5.1. Table 2.5.1. P u m p s p e e d for m o l e c u l a r h y d r o g e n for c a t h o d e m a t e r i a l s of T i - 6 A / - 4 V a n d p u r e t i t a n i u m . Hz
PRESS. Torr -3
1.4x10
-3
1.0xl0
-4
2.6x10
-4
1.OxlO
-5
1.7xlO 1.OxlO 1.OxlO 'c" 2 . 3 x 10 'd" 2 . 3 x 10 NOTES:
-5 -6 -4 -4
ION DENSITY
I-I~/cm2- sec (note 5.53
x
10
3 . 9 5 x 10 1 . 0 3 x 10 3.95 xl0 6.70 xl0 3 . 9 5 x 10 3 . 9 5 x 10 9 . 0 8 x 10 9.08
"a"" B a s e d "b"-
"a")
Total
x
10 on
15 15 15 14 13 13 12 14 14
H z SPEED Liters/see Ti
"6-4
Ti
lO.O
34.0
2 . 4 2 x 10
8.7
36.6
1.51xlO
5.9
31.5
2 . 6 6 × 10
9.4
28.5
1.63×10
19.4
29.0
5.71 x 10
20.0
29.5
3 . 4 6 x 10
18.8
15 14 14 13 13 12
8.24 xlO 6.34 ×10 1 . 4 2 ×10 4.94 ×10 8.54 xl0 5.11×10 5.40 xl0
67.0
2.67 xl0
-
25.0
9 . 9 6 x 10
area
constant ~190
"c": After
pumping
~0.5
"d":
pumping
~55.3
After
3 . 2 6 x 10
31.2
H~ P±UMPED /H-~ ION
"6-4" 15
-
assumed
cathode
H 2 MOLECULES PUMPED/cm2-sec ("b")
"pump
15 15 15 15 13 13 12 15 14
sensitivity"
of
Ti
"6-4
0.44
1.50
0.38
1.60
0.26
1.38
0.41
1.25
0.85
1.28
0.88
1.29
0.83
1.37
-
2.94
-
0.91
120 A / T o r r .
cm e .
Torr-:g Torr-~
//cm z /cn/2
(Point (end
"B"
o f Fig.
of life,
Fig.
2.5.3.). 2.5.3.).
Data on the N~ speed and life of pumps with 6-4 cathodes was reported elsewhere.(4 s ) Briefly, the end of life occurred after pumping -6.2 x 10 a Torr-Z of N2, at a pressure of -1.3 x 10 -4 Torr. The failure, after pumping N2 at this pres-
SPU'ITER-ION PUMPING
127
sure for -10 a hours, was caused by N 2 + ions drilling small holes through the 4.7 mm thick cathodes. After the N ~ life tests, it was evident, from the comparative cleanliness of the cathode plates, that most of the nitrogen was pumped on the anodes. Also, a crusted build-up of Ti and pumped gas on the anodes was colored the golden brown of TiN.
Diffusion in the Cathodes The cathodes are bombarded by ions ranging in energy from 0 - 7.5 keV. KenKnight and Wehner report that: 1) there is little difference in the H2 + sputter-yield of metals vs. oxides or nitrides of the same metals; 2) in the range of 1.0 - 7.5 keV there is, in fact, an increase in sputter yield with decreasing energy (e.g~, H2 + on Cu); and, 3) the sputter-yield of 7.5 keV H 2 + ions on Ti is 0.0113.1 a 0 ) I will make the assumptions that the energy distribution of ions in the pump is approximately linear; that the sputter-yield of all ions is comparable; and, that the surface contaminants have a thickness s 100 A. For the given sputtering rate, and an operating pressure of -2.3 x 10-4 Torr, the surfaces of the cathodes will be sputtered clean in s 1.5 hours. The range of penetration of the bombarding ions will differ according to energy of the ions and density of the material. In work with ion implanters, the beams are monoergic and, prior to annealing, when bombarding surfaces with high doses (i.e.e 10 x 5 implants/cm2 ), the dopant tends to stratify about an average range.(X 4 4 ) In our case, the energy of the beam varies from 0 - 7.5 keV. Young experimented with the range of electrons, H + and H 2 + ions in A~ .(1 4 5 ) He determined that the range, R, in/.tm, of H~ + ions in A~ is: R = 0.015 E ° .a 3, where E is the energy of the ion in keV. Therefore, because of the difference in densities, we would expect the range in Ti to be --0.01 x E ° .a 3. We then calculate that the maximum range of the H 2 + ions in Ti to be --500 A. This concurs with later findings reported by Andersen and Ziegler.(2 2 5 ) Of course range, R, is distributed about an average value. The macro-fissures and cracking often observed when pumping large quantities of H2 with Ti,(1 3 1 ) and very evident in the 6-4 material at end of life, were not at all evident when pumping H2 with the unobtanium. The mechanisms for cathode rift formation, with the adsorption of H 2, are extremely complex, beyond the scope of this book, and certainly my complete understanding. However, qualitatively, the observed cracking probably occurs for the following reasons. At room temperature the o~-phase of Ti (i.e., the crystalographic structure of pure Ti) is hcp. Pure ~-Ti, at room temperature, has a low H 2 solubility limit of -20 ppm (by weight).(1 4 2 ) As H1 goes into solution with the metal, some of the Ti must convert to a -t-phase, which is fcc (face centered cubic). Going into solution amounts to the H1 atom jumping into an interstitial site, when given the opportunity by lattice vibrations. Once the H 1 atom squeezes in, the lattice is distorted, and seeks the lower energy, fcc configuration (i.e., with the H 1 atom interstitially in place). Such lattice distortions are carried to an extreme with the formation of titanium hydride, and ranging from Till1.5 3 to Till1.9 9 in composition.( 1 4 6) The hydride is also a fcc arrangement, with the H~ atom residing in the tetrahedral interstices. As an extreme example, the density of Till 2 is -86.5% that of pure Ti.(1 2 3 ) Therefore, each dimension of a cube of pure Ti would have to increase by -6.4% in accommodating sufficient H2 to form Till2. These local volume changes cause stresses between nearest-neighbor volumes, i.e., unit cells in the lattice structure, inducing
128
SPUTI'ER-ION PUMPING
rifts and cracks in the cathodes. These induced rifts are called interfacial defects.(X • 7 ) The Ti-6A~-4V material is referred to as an c~-fl alloy. The A~ tends to substitutionally stabilize its neighbors in the c~ (hcp) form, where the addition of V acts as a fl-stabilizer.(t • 8 ) This means that in the absence of H 2, some of the alloy is in an o~-phase (hcp) form, where some of the material is in the bcc form (i.e., the /~-phase). The increase in H 2 capacity of the 6-4 material is probably not due to the presence of A;.. In fact, Paton and Williams, in a review paper, suggest to the contrary.(1 • 6 ) They report that only a metastable, supersaturated H 1 solution can exist in the c~-stabilized material, resulting in high internal stresses. These stresses are f'mally manifest in the formation of titanium hydrides which precipitate along the c~-fl interface, causing fractures or cleavages. The presence of vanadium, a fl-stabilizer, is probably responsible for the higher solubility and diffusivity of the 6-4 material. The H 2 solubility of completely ~'-stabilized materials (e.g., the unobtainium) ranges from x 40 to x 100 that of pure, supersaturated Ti, with the higher concentration resulting in catastrophic embrittlement.( 1 • 8 ) It is apparent that, because of crystolographic changes with H x concentration, the Ti and 6-4 materials become anisotropic, and a simple diffusion model does not apply.(1 • 9 ) Also, the spalling and cracking noted on the surfaces of the Ti and 6-4 cathode materials, suggests that late in life, zones on the surface of the cathodes may have operated at much higher temperatures than would be predicted by simple heat transfer calculations, dealing with an isotropic media. A thermal jump condition would exist at the interfaces of the spallated zones. A diffusion barrier also would be created at these interfaces. With substantial H 2 pumping, the concentration of H 1 in the Ti results in the formation of titanium hydride. At this time, the isotherm of titanium hydride comes into play, which, with excessive extrapolation (noted by Singleton(1 0 2 )), is ---10-s Torr at 170" C. Temperature also plays an important role in the diffusion process. It probably was in some part responsible for McCracken's results,(7 1 ) and the sharp increase in the speed of pumps with the 6-4 material at pressures z 10"4 Torr. From ambient temperatures to --500* C, the solubility of H 1 in Ti is approximately constant.(1 2 s ) However, the diffusivity, D, will vary markedly with temperature. The expression for diffusivity in three dimensions is:(1 • 7 ) D
= (1/6)a 2 v e x p ( A S d / R ) e x p ( - A E d / R T )
= D O exp(-AEd/RT ) where a
R
=
u.
(2.5.1)
the distance between interstitial jump sites encountered by the diffusing atom, the frequency of oscillation of the atom (i.e., the number of attempts per unit time at changing sites), the universal gas constant in calories per mole -* K,
and AE d and AS d are the activation energy and activation entropy for diffusion per mole of diffusing species, respectively, in units of calories/mole. The diffusivity term appears in the equation of Fick's first law of diffusion, in an infinite slab (i.e., a flat sheet of material of finite thickness d in the x-direction, but unbounded in the y and z-directions), as:
S P U T t E R - I O N PUMPING
Q
= - n a C(x,t) / ax
129
(2.5.2)
where Q is the throughput or transport of H 1 within the slab, and C(x,t) is the concentration gradient of H x in the slab. Fick's second law of diffusion is given in the less familiar form:
o c / o t = o / a x (n a c / a x )
(2.5.3)
Because of the hodge-podge gradient of crystalographic species, the diffusivity becomes a function of depth in the material; or, D = f(x). Secondly, where temperature gradients exist (e.g., exacerbated by spalling), then the diffusivity will take the form D = f(Do(x),T(x)). Then, (2.5.3) has the form:
a c / a t = a / o x Do(x)((exp(AEd/RT(x)) O C/ax,
(2.5.4)
where, T(x) means that the temperature, T, is a function of x, the distance within the slab. Assuming we were able to express T(x) exactly, expressing Do(x ) is another matter. At a constant pressure, there is qualitative similarity in the speed-vs.-time characteristics of pumps with cathodes of pure Ti and the 6-4 materials. Let us examine these results more closely. Note that the speed of the pump with the 6-4 material, shown in Fig. 2.5.3, initially appears to be high (i.e., -66 Z/sec). Actually, this is the speed of the pump --2 hours into the test. It is difficult to measure the initial transient speed of any pump at time t = 0, but, based on H 2 + implant considerations, and assuming an initial contaminant layer o f - 1 0 0 A, we calculate a speed, at t = 0, o f - 1 5 Z/sea In fact, there appear to be seven phases involved in the pumping of hydrogen in the 6-4 material - many of which occur simultaneously. These are: 1) A process of the gradual sputter-cleaning of the surfaces by the bombarding ions (i.e., interval "A"-"B" of Fig. 2.5.3). 2) A process of the gradual restructuring of the material, to a finite depth, 6, the implantation laminate (see Fig. 2.5.5), as a consequence of H x implantation and the pumping of ionized and molecular hydrogen (i.e., interval "A"-"B" of Fig. 2.5.3). At point "B", the pumping ratio of H 2/H 2 +, given in Table 2.5.1, suggests that the rate at which molecular hydrogen is being removed from the system is ---x3 that of the impingement rate of the H 2 + ions. At the end of the life-test this ratio was --- 0.91. 3) The creation of a second bulk laminate of thickness ), (t) - 6, where the high H x concentration at x = 6 forces the diffusion into and restructuring of the material in the range 6 < x < )~ (t) (i.e., interval "A"-"C"). 4) A decrease in pumping speed as ), (t) increases (i.e., the bulk laminate grows in depth) to the point of ), (t2), and the incoming flux is in equilibrium with the gradient in the bulk laminate (i.e., interval "B"-"C" of Fig. 2.5.3). 5) A steady-state equilibrium of H ~ flux entering the implantation laminate and the H 1 flux leaving this laminate and entering the region x > 6 (i.e., the point "C" of Fig.
130
SPUq'TER-ION PUMPING
2.5.3, which is --29 hours into the test). 6) The steady-state growth of the bulk laminate by diffusion of implanted hydrogen from the initial implant depth, through the interface x = d; and into the bulk laminate (i.e., interval "C"-"D" of Fig. 2.5.3, and A (ts) of Fig. 2.5.5). 7) The point where the surface concentration needed to cause diffusion from the implantation laminate and into the bulk laminate (further increasing its thickness) exceeds Till2, resulting in a net speed of zero (or the onset of thermal problems initiated by spallation of the material; i.e., point "D"). The above model suggests there are three diffusion processes going on: 1) the diffusion of implanted H 1 in the implantation laminate 0 < x < 5 ; 2) the diffusion of H 1 into the bulk laminate 6 ~ x < A (t); and 3) diffusion into the bulk material (of negligible consequence in Ti, as will be shown). Crank refers to such processes as either "pseudo-Fickian" or "non-Fickian" in nature.(1 s o ) This merely means that the presence of the diffusing material alters the diffusing medium so as to modify the diffusivity of the material in the medium (i.e., D = f(C(x)), as we suspected). I will leave solution of such equations to the mathematicians. What, then, can we conclude from the data in the two figures? A great deal, if we are willing to risk assuming the seven perceived steps apply in the hydrogen pumping model.
Titanium Cathode Material, a Model Let us assume that an implantation laminate is created, of thickness 6, and with a concentration of---Till2, which remains constant after a short time. A 500 ./k laminate would accommodate only 1.6 Torr-Z of H 2 (i.e., 8.5 x 10" a Torr-Z/cm 2) prior to filling to a state of the creation of Till 2. Let us assume the H 1 concentration in the laminate reservoir is (76 -1.14 x 10 2 a and see where this leads. For the time being we will assume that D ~ f(x,t) at x > ZX(t). Assuming the bulk of the cathodes is initially, void of H t , the H t concentration in the cathodes, of thickness d, is then given by:(t s t )
C(x't)=C6[1
-
x __2 ~ _1 sinnStx
d
7t t
n
d
exp(-D(nTt/d) 2 t)l (2.5.5)
This equation stems from the solution of Fick's first and second laws of diffusion, assuming zero initial concentration of H 1 in the material and the above assumptions, including D ~ f(x). Room temperature (RT) diffusivity data were not found for pure Ti. Available data, taken at temperatures >_ 500 o C, suggests a RT Ti diffusivity of D --4.25 x 10-t 2 cm 2/sec.(t s 2 ) Also, the RT diffusivity of H t in Tills: isDtc - ( 2 -t¢)(3.0 x 10-t 2) cm2/sec, 1.96 > ~; > 1.35, as given in this same reference. Two other references in this same work place the RT diffusivity of H 1 in Till 1 .s 5 at 0.48 to 3.4 x 10" 1 2 cm 2/sec. A summary of these data is given in Fig. 2.5.4. Diffusion processes in the cathodes result in concentration gradients, in time, as shown in Fig. 2.5.5. At time t 1, the cathodes are sorbing H 1 and an implant laminate of depth 6 is established. At time t2, the edge of the bulk laminate, x - ),, starts to grow away from the surface such that )~ (t) > ~, etc. Early in this process,
S P U T r E R - I O N PUMPING
131
the diffusivities in the implant laminate, bulk laminate, and bulk cathode material are all similar, so that the distinction between d; and ), is at best fuzzy. This is also true for all times at low operating pressures. l0
-11
I_
I
I
1
I
/BASED
]
I
I
1
1
I
_~ Z
Hiq @
ON
PUMP RESULTS
~_
<~
t-ca ¢.)
i
~ 10 -1~
~
Zca ¢2 o a;
10-13
0
2.0
1.0
HYDROGEN CONCENTRATION Figure 2.5.4. Hydrogen vs. concentration in
IMPLANT LAMINATE ESTABLISHED A FEW INTO THE TEST
(//-HOURS
BULK
X ~
\
LAMINATE O TIME ts
I ~-f- TIME t 1 \
1
~
\ \ ~~
\ ~
~-
BULK CATHODE
MATERIAL (UNMODIFIED)
?
DISTANCE INTO CATHODE
TilH? Figure
diffusivity
titanium.
2.5.5.
hydrogen
Proposed
diffusion
model in
for
eathodes.
Let us first assume two things: 1) D), = 4.25 x 10-t 2 cm 2/sec (i.e., the diffusivity at the edge of the bulk laminate); and, 2) a steady state speed, at 10 -4 Torr, of --9 Z/see, for the pump with pure Ti cathodes. This corresponds to a flux 3.12 × 101 4 H 1 atoms/sec-cm2 passing the plane at x = ),. Evaluating (2.5.2) and (2.5.5) at x = 0 - ),, and solving for C~ vs. time, results in an increasing hydrogen concentration in time as shown in Fig. 2.5.6. We see that pumping --34 hours at this H~ flux, the concentration, at x = )~, necessary to sustain pumping at this flux, exceeds that of Till 2 - a physical impossibility. z
lO ~3
10
- HI FLUX EQUIVALENT 9 Z/see 0 ,,, 10-4 T o r r
-
Z
't.
2;
°4
t-.,
~
.
o
x
o
THEORETICAL 10
~x~.vM
22
Z ~
__-'2~_- .
2:
5 10
21
0
I
-10
I
I
I
~
20
TIME -
1
30
at ¢3 •
/ NEEDED TO SUSTAIN FLUX OF 3 . 1 2 x 1 0 14 H1 a t o m s / e m z - s e e 10
Figure 2.5.6. concentration
- ~ -
-"°---IV_ / C ~ CONCENTRATION
5= --
~ .
~7,
H1
flux
Maximum exceeded
at
possible i n Ti ~10 -4 Tort.
[--, X
1
10
40
hours
Let us try another approach; let us arbitrarily assume that C), stabilizes at 8.8 x 10 2 2 Hz atoms/cm 3 , and calculate how the flux through the plane x = ~, i.e., the speed of the pump, varies in time. This results in a change in hydrogen pumping speed as shown in Fig. 2.5.7. This result more closely corresponds to what we observe in the laboratory. For a constant value of C 6 we see that the speed of the pump has decreased after --560 hours, to --23 % of the initial speed. With this model, we calculate that a total of - 7 2 0 Torr-Z of H2 was pumped into the bulk of the cathode, with 280 Torr-Z unaccounted for (i.e., the experimentally observed 1000 Torr-Z). The values of C), - 8 . 8 x 10 2 2 Hz atoms/cm a , and D), - 5 . 0 x 10 "1 ~ cm2/sec were estab-
132
SPU'Iq'ER-ION PUMPING
lished by iterative computer calculations of the best fit for the H 1 flux in the Ti cathodes at a quasi, steady-state pressure of 1.6 x 10"s Torr. You will recall that this is the pressure above which the speed decreased markedly. z
10 23
10 _
z
-~S-
15
_ H~ FLUX EQUIVALENT__~ TO INITIAL 9 :£/sec
,..] o
z
Figure 2.5.7. V a r i a t i o n i n H1 assuming steady-sLate bulk laminate HI c o n c e n t r a t i o n .
~ ~ to 22
1014
~
Z ~ ,=.,
[///AS[ S U M[ E D
C~ I / C O N C E N T R A T I O N OF 8.8 xl0 zz H1 a t o m s / c m
[--,
z
~
lO
0
150
I
300
I
• ~ t"
a
10 600
450
TIME - h o u r s
Because D = f(x), we may not venture into the material zone x s )~, with (2.5.2) and (2.5.5). However, let us look in the region ~ < x, assuming C)~ 1.14 x 10 2 a H 1/cma, and observe the total flux passing a plane x = £, for various values of £, while approaching the bulk laminate from the right. The purpose of this exercise is to determine how much H 1 has diffused into the bulk of the cathode as a function of time. Invoking (2.5.2), and counting the atoms passing the plane x = £, for time t, amounts to solving the following equation: t
f [-D a / a x (C(x,t), Ix = £ ] d t = number of H1 atoms/cm 2
(2.5.6)
0
Evaluating (2.5.6) at £ --10 -2 cm, we find that after 560 hours, a total of only --12 Torr-Z of H2 has diffused into the cathodes. This conclusively establishes that the mechanism for the pumping of H 2, in Ti, at high throughputs, is almost exclusively through the creation and growth of the modeled bulk laminate. The speed of the pump becomes zero when the concentration of H1 in the implant laminate must exceed Till 2 in order to drive H 1 through the bulk laminate. Because of limits on the diffusivity out of the bulk laminate (i.e., its growth), the hydrogen capacity of the pump depends on the influx rate. Had we operated at lower pressures, or modestly higher temperatures, pump capacity would have increased (as was observed). From the above findings, because of the small diffusivity value of pure Ti, we might have assumed that the cathodes were semi-infinite mediums to the diffusion of hydrogen. This problem has the simple solution:(1 s a )
C(x,t) = C 6 [1 - erf(x/2(Dt) ½)]
(2.5.7)
In using (2.5.7), we still have to assume that D ~ f(x), which we know is incorrect. However, assuming C& = 1.14 x 10 2 a H 1 atoms/cma, what would D have to be at x = 6 --0 to support the observed flux of 3.12 x 101 4 H1 atoms/cm2-sec (i.e., a speed of --9 Z/sec at --10 -4 Torr), at the edge of the implant laminate? Using (2.5.2) and (2.5.7), we calculate that after 34 hours of pumping, D -2.5 x 10" 1 2 cm 2/sec, a reasonable result.
SPU'ITER-ION PUMPING
133
Ti-6A~-4V Cathode Material Let us assume the model which we used in the case of pure Ti applies to the 6-4 material. Note that the speed of the pump, as shown in Fig. 2.5.3, decreases almost linearly in time. I suspect that end of life was due to thermal problems stemming from spaUing of the cathodes. If this is the case, we may express speed of the pump as: S - 4 4 Z/sec - (1.16 x 10-s Z/sec 2 ) x t. From this we calculate that the speed of the pump is zero at t e -- 3.79 x 10 6 seconds. Extrapolating linearly to S = 0, the average speed over the total life would have been ---22Z/sec. From this, we calculate that the cumulative H 1 in the material per unit area, at S --0, is --6.64 × 10 2 1 H atoms/cm2, assuming that no H z leaked out the back of the cathodes. With this accumulated flux, we may establish a lower limit on the probable diffusivity of the 6-4 material, by assuming the maximum possible H ~ concentration is Till 2. (Note that the depth of spalling was of the order of 1.0 - 1.5 mm.) We make these calculations by using the equivalent of (2.5.5), but at x = :~ --- 0, and assuming C), --1.1 x 10 2 3 H 1 atoms/cm 3 . This simple approach suggests that in the absence of H 1 in the 6-4 material, D >_ 1.8 × 10-9 cm2/sec. Results of this calculation are shown as the dashed curve in Fig. 2.5.3. On the other hand, using (2.5.2) and (2.5.7), we calculate that the maximum possible diffusivity of the 6-4 material is D --3 x 10-1 0 cm 2/sec at, for example, time, t = 29 hours. Two questions remain: 1) If speed at the higher pressures is diffusion limited, why does it not increase markedly with decreasing pressures, assuming activation sites are continuously being generated by ion bombardment? 2) Why is there a disparity of almost a factor of two in the early speed data of pumps with the 6-4 vs. Ti materials? You will note from Table 2.5.1, that for all pressures, the rate of pumping by the pure Ti was less than the impingement rate of H2 + ions. This was not the case with the 6-4 material. This can be explained by differences in solubilities of the two materials in conjunction with probable differences o f - - x 100 in diffusivities. Remember that the impingement rate of molecular hydrogen is --×370 that of the H2 + ions, for all pressures. This impinging flux of gas creates the equivalent of a concentration gradient at the surface. If the concentration at the surface is sufficiently low, H 2 will impinge on the surface, be dissociated and diffuse into the bulk. If the solubility limit at the surface of the material is at a maximum, and the diffusivity is insufficient to deplete this H ~ concentration, no molecular H2 will be pumped.
Beta-Stabilized Cathode Materials These materials have very high diffusivities. For example, Zeilinger and Pochman, using neutron radiography, established that the diffusivity of ~-titanium (i.e., Ti-13V-11Cr-3A£ ), is -- 2.8 × 10" T cm 2/sec(1 8 3 ) at RT. With such high diffusivities, a bulk laminate does not have an opportunity to build. Also, the surface implant concentration varies in time, but never achieves the Till2 level (i.e., at the time of catastrophic failure). There are two possible models for the pumping of hydrogen; one assuming the escape of H 2 out the back side of the cathodes, the other assuming negligible H 2 loss through the cathodes. Assuming the latter to be the case, the following equation approximates the concentration as a function of time:(1 5 4 )
134
SPUTFER-ION PUMPING
qd Dt 3x 2 - d s C(x,t) = ~ I_"77" d'6d 2
2°°-1n
-~2~
- cos
~ n
nTtx d
exp(-D(n~t/d)
2 t)],
(2.5.8)
where q is the H 1 implantation rate per cm s, on the face of the cathode. (Note that this equation is not applicable for small values of t.) With a speed of --50 Z/see at 10-4 Torr, 16,000 Torr-Z would be pumped in -3.2 x 10 e sec. The above equation predicts a f'mal H1 concentration atx = 0 of --1.27 × l0 s 2 Hx atoms/cm a . At the back side of the cathode, the concentration would be --1.17 x l0 s 2. Assuming a linear concentration gradient at failure, the total calculated average H 1 concentration is -1.2 x 10 2 2 H ~ atoms/cm3, which, assuming all the gas was retained in the cathodes, leads to a f'mal measured average concentration of -1.17 x 10 2 2 H x atoms/era a . This closely corresponds to the H1 concentration which Paton indicates is sufficient to cause catastrophic failure in B-alloys of Ti.( ~ 4 e )
Diodes With Aluminum Cathodes The pump elements of the LINAC at BNL had to be replaced (see pages 118-119). I was disposed to replace the cathodes with the Ti/Ta combination, as there was some evidence of noble gas instabilities in the existing pumps. However, the cost of Ta was prohibitively high. Someone had noted to Dr. Derek Lowenstein, head of the BNL Collider/Accelerator Department, that aluminum might be an acceptable substitute for SIP cathode material. Hydrogen and water vapor were prominent gases in the LINAC. Neglecting the noble gas instability problems with the LINAC, I was skeptical about the use of aluminum cathodes for hydrogen pumping. This came from "reading between the lines"in the literature, and because of the problems encountered with the TRISTAN machine in Japan, as related to me by Dr. Satoshi Ozaki, formerly director of that project. I told Derek about my reservations, relating to the pumping of hydrogen with aluminum cathodes. He suggested: "prove it." He wasn't being flippant. Rather, he was asking me to technically verify the hydrogen pumping problem, prior to making a major investment in Tifl'a cathode material. Work had been done by Liu, Lin and Lee, with aluminum cathodes in triode pumps.(7 s ) Based on hydrogen solubility data in aluminum, their results seemed inexplicable. To verify their results we modified a diode pump and installed cathodes of pure aluminum. The pump comprised two element assemblies, each having 63 stainless steel (SS) anode cylinders, 20.3 mm in diameter and 25.4 mm long. The projected aluminum cathode area under each anode array totaled 588 cm s. We first conducted speed tests for N 2 and using titanium cathodes. This served as a benchmark for the aluminum cathode test data. To make sure the cathodes were saturated (i.e., if this effect existed) we conducted speed measurements at high pressures, and then reduced the pressure to successively lower values in subsequent measurements. Results of the speed tests with aluminum cathodes and for the gases CO, "air" and N s are shown in Fig. 2.5.8a and 2.5.8b, the later evidencing saturation for CO. There was only modest decay in pump speed with time. Therefore, this demonstrated that aluminum SIP cathodes might be effective in the pumping of N s and CO. When using aluminum cathodes, the measured speed for N s exceeded that measured in the same pump when using titanium cathodes. The speed for "air" using aluminum cathodes decreased more dramatically at lower pressures. We can only speculate on the reason for this.
SPU'ITER-ION PUMPING S(P) 200
1
1 11
1
I I II I I I I "AIR" WITH -~ A1 C A T H O D E S \
N2 WITH
1
135
I I 1 -
150
5q
-
Oq £-
I
100
~ ~/
CO WITH A1 CATHODES
50
0
-9
-a
10
-7
10
1
PUMP PRESSURE
Figure diode
1 II
10
10 -
-6
l
I I1
-5 10
TORR
2.5.8a. Steady-state pumping SIP with either titanium or
speed of the same aluminum cathodes.
200 ]_
~
[...
1.3 x 10 -6 Torr
150
--
~
X
¢x,
5 x 10-6 T o r r -
100
?0
100
200
300
400
500
ELAPSED TIME - Minutes Figure pump
2.5.8b. D e c a y i n t h e CO s p e e d with aluminum cathodes and
of a diode sputter-ion stainless steel anodes.
Hydrogen speed tests, with new aluminum cathodes, were far less promising. Speed data were very unstable, pressure runaway was evidenced after pumping a total of only 2-4 Torr-Z of hydrogen. This corresponded to a hydrogen pumping capacity of only 1.7-3.4 x 10-a Torr-Z of H 2 per cm2. The triode hydrogen results of Liu and his colleagues could only be explained if one assumed that the hydrogen was primarily being pumped as neutrals in the SS walls of the SIP. This led us to ask: "how much of the H 2 gas is pumped in the walls and anodes of diode SIPs?" Andrew and others had shown that significant amounts of inert gas neutrals are buried in the anodes of SIPs.( 8 9,9 s ,9 6,9 7 ) Further, is the burial of neutrals in the SS walls and anodes of SIPs the cause of the gradual decay in low pressure hydrogen pump speed observed with all SIPs? We could answer the ques-
136
SPUTTER-ION PUMPING
tion of hydrogen burial in the anodes by rebuilding the same SIP used for the above hydrogen tests, only this time using new aluminum cathodes and titanium anodes. Under such circumstance, there should be some low, but steady state hydrogen speed after pumping 2-4 Torr-Z of hydrogen. Such a pump was constructed, and the speed v. amount of hydrogen pumped is shown in Fig. 2.5.9. The results of this experiment proved to be one of those rare instances when you have a "hunch", do a test, and you hunch is experimentally proven correct. The initial hydrogen speed of this pump was --.120 Z/s. But this speed rapidly decreased to a steady-state value of -6.6 Z/s. If the hydrogen pressure was increased to ;:5 x 10-6 Torr during speed measurement, the pressure gradually increased in time. We speculate that this may have been due to the thermal desorption of hydrogen from the cathodes (i.e., --15 mW/cm 2 ). However, if the hydrogen pressure was maintained at s 5 x 10-6 Torr, the pump was completely stable. The pump continued to stably pump hydrogen at the lower pressures until --15.28 Torr-Z of hydrogen was pumped, at which time the test was terminated (i.e., over 150 h into the experiment).
z~l 6 10
I
J
I
!
!
m
I
×=
o 0
I
~ I
_~ 6.6 :g/see.
0
10
i
!
I
~ 1
DIODE WITH ALUMINUM CATHODES AND TITANIUM ANODES
,-
I
I
1
1
4.0
!2
!
l
\
i
J
1
8.0
HYDROGEN P U M P E D
-
t
! 12.0
-
I
Torr-
r
t 16.0
;f
F i g u r e 2.5.9. H y d r o g e n p u m p i n g s p e e d of a diode SIP with a l u m i n u m c a t h o d e s a n d t i t a n i u m a n o d e s . Hydrogen
Burial
in Diode
Pump
Walls
In another experiment, new diode pump cathode/anode assemblies were installed in a pump which had previously been used to conduct He pump speed measurements. When baking this refurbished pump into turbomolecular pump, He outgassing from the walls of the pump amounted to -1.3% of the total He which had been pumped by the pump in the earlier experiments.(2 4 7 ) If this amount of He was buried in the walls, would not an equivalent amount or more of hydrogen be so buried when pumping this gas? We set out to substantiate this by constructing a sputter-ion pump with Ti anodes and cathodes, and lining the element pockets with sheets of Ti (the "special" pump). We then conducted H 2 speed tests at low pressures as a function H 2 which had been pumped. We compared these data with that of an identical pump having SS anodes and no Ti lining of the pockets (the "standard" pump). After a 48 h, 250* C bakeout of the standard pump the base, with no other pumping, the H 2 base pressure in this pump was 6.4 × 10° 1 o Torr in 144 h at RT. Hydrogen base pressure in the special pump was 2.5 × 10-1 o Torr in 24 h, after a 24 h bake at 250* C.
SPU'VFER-ION PUMPING
137
Initially the hydrogen speed of both pumps was -225 Z/s in the 10"~ to 10-6 Torr pressure range. However, after pumping -74.2 Torr-Z H 2 the speed of the standard pump was 23.6 Z/s at 1.1 x 10"a Torr, whereas the speed of the special pump at 2.6 × 10"9 Torr was -169 Z/s after pumping 61.9 Torr-Z of H 2 . Calibration of all instruments was traceable to NIST. This series of tests proved to us beyond a doubt that, lacking the provisions of the special pump, the burial of H 2 in the walls pressures.and anodes(2of4 8a)pump results in an ever decreasing pump speed for this gas at low
Conclusions on Hydrogen Pumping The model for the pumping of H 2 in sputter-ion pumps depends on both the solubility and the diffusivity of the gas in the cathode material. Depending on the diffusivity, there appear to be two extreme models in the pumping of hydrogen. The first deals with materials with moderately low diffusivities, such as Ti. In this case, the hydrogen is implanted to a limited depth in the material. Thereafter, a bulk laminate is established, the depth of which depends on two diffusivities - the diffusivity in the parent material, prior to reaching a maximum solubility, and the diffusivity in the bulk laminate. The depth of the bulk laminate grows after the H t concentration at the inner edge exceeds the solubility of the material for a given crystal lattice structure. If the crystal lattice is able to reorder itself to accommodate more H z, the depth of the bulk laminate will grow. Eventually, equilibrium exists between diffusion at x > 3, (i.e., further growth in the laminate) and the concentration at the surface of the cathode, and no additional H2 is pumped, for the given flux. The degree of recovery, in further pumping of H 2, depends on rate of diffusion into the bulk material in the absence of further implantation of H z into the material. This model is very similar to that proposed by Singleton.(t 0 2, t 2 o ) The other extreme of the model is when the diffusivity of the material is so large that for the given flux an equilibrium surface concentration is never established. Therefore, three elements are essential to the extended pumping of H2: 1) a high, inherent solubility for hydrogen (e.g., the ~i'-alloy); 2) facility of the material to reorder its crystal lattice structure to accommodate additional H 1 once the solubility limit is exceeded (e.g., pure Ti, and the 6-4 material); and 3) a high diffusivity (e.g., the 6-4 and/9-alloy materials). Note that a high inherent solubility corresponds to there being a low equilibrium pressure over the metal when the gas is in solution with the metal. Also, increasing the cathode surface area has the obvious advantage of decreasing the H 2 flux. Of course, all of this focuses in the chemistry occurring in the cathodes. But, as has been shown, the of burial of neutrals in the anodes and pump walls is also an extremely important in consideration in how it impacts on pump speed at low pressures and the ultimate pressure which can be achieved in a UHV system.
138
S P U T t E R - I O N PUMPING
2.6 Triode Pumping Two primary forms of triode pumps are in use at this time. One configuration features gridded cathodes, as shown in Fig. 2.6.1. Hereafter, it is called the GCT pump (i.e., gridded cathode triode). The second is the Stareell ® pump, marketed by Varian Associates, and shown to the right in Fig. 2.4.18. Use of the former pump configuration continues to date primarily as distributed ion pumps (DIPs), used in particle storage rings throughout the world (see Section 2.10).
--Z'~ / /3/./ ~
///
v WALL
///
~ /
Lf-///
/ ./
/
-35.6 mm I
!
I ~ n I
I
! AREAOF CATHODE
i I a th-,.J
-tSr~rnr
I
'-............... ./../ ,:::~ ---_ ~ I / / / / ././../../../../............./../../..j../..X.:'., Z# ~ ~X
TITANIUM OR Ta/Ti _~ CATHODE PLATES
Figure 2.6.1. Cathode surface compared with the effective
~/ - -
"A"
..~
[/ I
UNDER ANODE , - --6 7 c m z " / AREA OF GRIDS U N D E R ANODE
/ / .,,.
I
I I
Y I
I I
__ i a I 0.6 ~0 ~"~-Zz~ bm~ - ~
-~i i- ~
25 4 " m,m
{ T / /
l
/
i
..~l
" +
..4"}/" _
GRIDS "B" GRIDS
_
TITANIUM GRID CATHODES (0.63ram x 3.9ram)
area of conventional diode and~triode --13.3 cm 2 area of the StarCell ~ triode
pumps pump.
Solutions of ion trajectories proximate to the GCT pump grids are complex, but the), may be found using a method described by Spangenberg, for gridded tubes.(1 s s ) This approach does not yield exact solutions, and requires iterative computer calculations. In the absence of a significant RF component in the discharge, all of the ions impinge on the cathodes. Also, though grids from time to time will be either melted or sputtered through, there is little evidence of back-side sputtering. Hamilton did note the back-side pumping of Ar in an early triode pump configuration.( t 7 ) However, most of the Ar ions strike the grids the first pass through, and many are reflected off these grids and pumped in the walls of the pump body as energetic neutrals. The surface area of the cathodes plays an important role in the pumping of H 2. An enlargement of the GCT pump element is shown in Fig. 2.6.1, where one cell of the StarCell ® element is shown in Fig. 2.4.18. Assuming that ions impinge on both sides of the A and B grids of Fig. 2.6.1, the equivalent cathode surface area, under a single cell, is --6.9 cm 2. With a similarly sized anode and assuming both sides of the radial vanes accommodate gas - a reasonable assumption with neutral reflections the StarCell ® has an equivalent cathode surface area --13.3 cm 2. A diode cathode, under the same anode cell, would have a surface area of --6.7 cm 2. Therefore, from a cathode surface area standpoint, the StarCell ® pump should have a decided advantage over conventional diode and GCT pumps in the low pressure pumping of H 2 . Assuming the same I/P characteristics of diode vs. triode pumps, and the absence of other variables, this would certainly be the case.
SPUIqqER-ION PUMPING
139
As the H~ pressure and associated throughput is increased, triodes evidence problems of thermal run-away much sooner than diodes. The cathodes of diode pumps are operated at ground potential. Because of this, there is some thermal contact of these cathodes with the walls of the pump. Though a high thermal jump condition may exist between the cathode and the pump wall, there is some exchange of heat energy to the wall. Because the cathodes of triode pumps must be electrically isolated from the walls of the pump, they are also thermally isolated from the walls. This means that the primary mechanism for the dissipation of the ion beam energy is through radiation. The thermal problem is exacerbated by the fact that the construction of all triode cathodes makes the conduction of heat, in the plane of the cathode, negligible. Assuming a cell sensitivity for H 2 of --5 Afforr-cell, and a pump operathag voltage of --5 kV, power dissipated in each cell would be 2.5 x 10 4 W/Torr-cell. At 4 x 10"s Torr, 0.5 Watts would be dissipated on each cathode of a given cell. Depending on the configuration of the cathode, this power would be sufficient to raise the temperature of a thermally isolated cathode to as much --300* C. Snouse reported cathode temperatures as high as 500"C during the starting of triode pumps.(1 s 6 ) Baker reported problems with thermal runaway when pumping D2 with triode pumps.(1 s T) His solution was to reduce the voltage so as to decrease the power dissipated in the triode elements. For the above reasons, triode pumps have very low throughput limitations for H2. For that matter, in the absence of water cooling, this is also true, to a lesser extent, of diode pumps. They too are not immune H~ thermal runaway problems.(1 s a ) In a recent paper, Liu, et al., reported on the pumping soeed of a GCT pump, with A~ cathodes, and for the gases H 2, O 2, N2 and Ar.( "7 ~ ) They modified a Varian 60 Z/sec GCT pump, by removing the Ti grids (see Fig. 2.6.1) and simply replacing them with A~ grids. Speed findings of this paper are important to the understanding of active-gas pumping mechanisms of triode pumps, having the conventional Ti cathodes. They reported pumping speeds vs. pressure, at saturation conditions, for the above four gases. In this instance, I interpret saturation as being comparable to Point "C" of Fig. 2.5.3, for the given gas. Liu observed only moderate differences in the speed of the pump, with A~ vs. Ti cathodes, for three of the four gases, H~ being the exception (i.e., comparing his data with speeds reported by the manufacturer, with Ti cathodes). Denison measured the speed of a new triode pump for Ar and determined that it was -39% of the speed of the pump for N~.(1 s Q) Late in life, the speed reduced to --20% of the rated N2 speed. Therefore, Liu might have expected an initial speed for Ar of--24 Z/sec, and a decrease to --12 Z/sec late in life. Liu reported a maximum speed for Ar of ---18Z/sec after saturation. Liu reported a peak speed of--95 Z/sec for N~ with the A~ cathodes. This corresponds to an -50% increase over that which would be observed with Ti cathodes. These results probably in part stem from the high sputter-yield of A t , compared to Ti. Comparative sputter-yield data for N2 + on A2 and Ti were not found. However, one may draw some conclusions from published data for other gases. Based on the sputter yields of Ne + and Ar + on Ti and A~ ,(s 4 ) the sputter-yield of N2 + on A t is probably .-x 2 that of Ti. The sputter-yield of 1.0 keV Ar + ions, normal to an A t target, is --2. Also, you will recall that in all triode pumps, the ion bombards the cathode with a high angle of incidence. Based on 1.05 keV Ar + sputtering data, we would expect the increase in the sputter-yield of A~ and Ti, with angle of incidence, to be comparable (i.e., maximizing at --x2 that of normal inci-
140
SPU'ITER-ION PUMPING
dence).(s a ) However, we should not give too much significance to sputter-yields of triode configurations vs. those observed in diodes. For example, the nitrogen pumping efficiency (17, as defined in (2.4.6)) for the HiQ ® pump is -25%, where for triodes it is --26%; i.e., very comparable. However, I have observed only a --25% gain in N2 speed in a diode pump with A~ cathodes, suggesting some gains from the sputter-yield of triode configurations.(7 a ) At an incidence of 7t/3, the lower the energy of the bombarding ion, the further the departure from the cosine law (for example, for 7t/3, 5.0 keV Kr + on W, most of the sputtered W atoms are at - 7t/3).(s a ) Lastly, for Ar + ions with energies of - 1 keV, --30% of the sputtered A~ neutrals and --40% of the sputtered Ti neutrals will have energies > 7.4 eV (i.e., the dissociation energy of N 2 ).(s a ) From the above, we can envision a model for the pumping of N2 with A~ cathodes, as follows: 1) For every N 2 ion bombarding the cathode, --4 A~ atoms are sputtered off the cathodes. 2) A deposit of A~ builds up on the walls of the pump. The deposit is always rich in weakly physisorbed N2 and some A~ N. 3) Weakly physisorbed N2, in the A~ deposit, may be dissociated by both bombarding N2 and A~ neutrals, to promote further formation of A~ N within the bulk and on the surface, respectively. Therefore, the increased pumping of N 2 with A~ vs. Ti cathodes is essentially due to the higher sputter-yields of the A~ cathodes, and perhaps the energetic neutral dissociation of N2 in and on the sputtered A~. Liu used depth profiling, Auger analysis to determine the concentrations of gas and metal on various surfaces in the pump after pumping air for 3000 hours. Okano made earlier use of this technique in assessinz the regions of pumping in a diode pump with juxtaposed A~ and Zr cathodes.(T 4-) Analysis of a metal plate, substituted for the wall of Liu's pump, indicated uniform concentrations of N 1 and A~, in equal proportions. Also, a uniform 40% nitrogen concentration was found in the cathodes, off anode center, at depths of -450 A, the limit of the profiling. However, based on earlier work of Audi, we would expect the N 1 in the cathodes to decrease with further depth profiling.(9 9 ) Also, Hamilton observed that N2 is pumped by both physisorption and chemisorption in sputter-ion pumps. He noted that physisorbed N2 (probably implanted in the cathodes) could be later desorbed on the introduction of Ar.(4 7 ) Regarding H2 pumping, Singleton reported that saturation (i.e., in this case meaning zero speed) occurred in A~ cathodes after pumping only -10 -2 Torr-Z H 1/cm 2 of cathode.(1 2 0 ) (In a similar experiment, I found close agreement with Singleton, and negligible capacity for H 2 when using A~ cathodes in a multi-celled, diode pump.(7 a )') This corresponds to filling up a beam-fabricated, implant laminate to a point of saturation. However, an important distinction is that in the case of Ti, the crystal lattice of the implant laminate (or, bulk laminate) will reorder itself to permanently accommodate more H 1 up to the point of formation of Till 2. This is not the case with A~, which, as reported below, has very low solubility and negligible diffusivity for H 1. Therefore, we might expect that on turning off the pump with A~ cathodes, after filling the implant laminate, a great deal of the hydrogen would quickly diffuse back into the vacuum system. Liu reported a steady-state H 2 speed of--32 Z/sec at a pressure of 10" 7 Torr, decreasing to --20 Z/sec at -10 -6 Torr. We note that hydrogen sorption in A~ is endothermic. Work reported by Hatch establishes RT solubility limits of H 2 in A~ from 0.56 to 4.9 x 10 -6 Torr-cm a/cm a A~.( 1 6 0) Further, up to a point, H2 solubility in A~ tends to decrease with increasing amounts of trace elements.(1 6 1 )
SPU'I~I'ER-ION PUMPING
141
Because of this, we would expect the solubility of H 2 in 6061 A~ to be less than that of 1100 A~. Once the H1 concentration exceeds the very low solubility limit, perhaps as the result of quick-quenching or implantation, there is some evidence that H1 in A~ forms small bubbles which diffuse and coalesce into larger bubbles, causing internal blistering,( 1 6 2) behaving similar to inert gases in metals.(1 6 s, 1 e 4 ) The solubility of H 1 in Ti, at one atmosphere and RT, is --1.8 × 10 6 Torr-cm3/cm z Ti; this is × 101 2 higher than in A~, give or take a few orders of magnitude. Therefore, diffusion into the A~ cathodes was of limited significance in the H 2 pumping observed by Liu. We expect the steady-state N 2 and H 2 speeds of any pump with Ti cathodes to be comparable at, for example, 10" 7 Torr. Therefore, the fact that Liu reports only a 50% reduction in H2 speed with A~ cathodes, vs. the speed of the same pump with Ti cathodes, suggests the pumping of H 2 in Liu's experiment stemmed in part from the burial of high-energy neutrals in the stainless steel walls and anodes of the pump. After filling the implant laminate in the A~ cathodes, the H 2 population is sustained in the A~ cathode only as the result of further ion impingement, but the net speed of the A~ cathodes would very quickly become zero. After this the walls of the pump consumed the H 2 . We are also led to this conclusion by results reported by Norton.(1 6 s ) He reports that the permeability, i.e., the product of D × s (s is the solubility) when extrapolated to room temperature, is --× 10T greater in stn. stl. than in A~. Were the walls of the pump constructed of A~, we would expect the speed of the pump to approach zero on the pumping of--2 - 3 × 10"2 Torr-Z H 2 / c m 2 . Of course, the presence of A~ 2 O3 on the walls would cause further decrease in the GCT pumps capacity for H 2 .(7 o ) Calder and Lewin report measured values of H 1 concentration in stn. stl., at RT, of >300 Torr-cm z/cm 3 metal (i.e., ~ 2 × 101 9 H1 atoms per cm 3 ).(2 o 6 ) Eschbach notes that the RT, H1 diffusivity of stn. stl. to be --8.2 × 10 -1 4 cm 2/sec.(Z e e ) This limited diffusivity may in part account for the reduction in apparent H2 pump speed at pressures ~ 10"7 Torr in Liu's experiment. Using data for the solubility of H 1 in stn. stl., we are still unable to account for the steady-state speed at --10" T Torr. For example, the 60 Z/sec triode pump has a stn. stl. cathode surface area (i.e., the stn. stl. area under the anode cylinders) of--500 cm2. Using Calder's number for the solubility, and Eschbach's number for diffusivity, we predict with (2.5.7) the maximum possible flux of H1 o f - - 2 × 10 6 atoms/cm 2 -sec. Liu's H 2 data suggests a flux of --4 × 101 1 H 1 atoms/cm 2-sec. Assuming neutrals were implanted in the anodes with equal facility would only reduce this value by - × 3 . Therefore, an implant laminate, similar to that in the case of Ti, must be constructed by neutrals impinging on the walls of the pump. McCracken, using 18 keV H + ions, bombarded a number of materials, including stn. stl. targets, and measured the reflection coefficient as a function of time for beams of intensities of -0.9 mA/cm 2.(1 e 7 ) He determined reflection coefficients by measuring the H 2 pressure build-up as a function of time. Interpreting his data, a beam reflection function, R, for stn. stl. can be approximated by R --k o sinc~t, 0 ~ t 5 seconds; R = ko, t > 5 sec, where ~ = 7t/10 sec. and k o = 0.9 mA H 1 +/cm2 (i.e., 5.6 × 101 5 H1 + per cm 2 -sec). The gas accommodated is then merely o j "5 ( 1 - R ) d t . This has the value of--101 e H1 + ions/cm 2 . With Young's A~ range data, t 1 4 5 ) one calculates the maximum range for 18 keV H 1 + ions in stn. stl. to be --600 A. Assuming that the H 1 could readily diffuse out from this maximum range, a laminate filling would occur with an H 1 concentration of -1.7 ×
142
SPU'ITER-ION PUMPING
10 2 x Hx atoms/cm s , a concentration x l00 higher than reported by Calder and Lewin. Using these energy-promoted solubility limits, and compensating for range, we would predict that Liu's pump would cease to pump H 2 at 10" ~' Tort after a few hours. It is probable that a steady-state cathode implant laminate had not yet been fabricated at the time of the measurements. Also, neutral burial of H t in the anodes would extend this to --10 hours. However, we should avoid taking any of these calculations too seriously, as the laws of physics always seem to be challenged by the quantities of gas which are consumed by sputter-ion pumps - at least our understanding of these laws. Recent work substantiates the above model - that is, a significant amount of the H2 in Liu's experiments was pumped in the walls of the pump.( 7 3,2 ,t e ) It is possible that the effects of pumping substantive quantities of neutral H 2 in the stn. stl. walls and anodes of pumps may explain why the hydrogen speed of pumps, at low pressures, falls off faster than predicted by I/P considerations. On reducing the H x flux into the stn. stl. walls and anodes, the hydrogen therein, formerly in equilibrium with the higher flux intensities, is liberated into the vacuum system. This has the effect of temporarily reducing the net pumping speed at lower pressures. Of course, cathode contamination by other active gases at low pressures causes further reduction in speed with partial pressures comparable to that of H 2.
2.7 Transient Speed Effects Early publications tended to overstate the relative speed of pumps for H2 vs. the other gases (e.g., N2 ). This was because few had full appreciation for the transient speed characteristics of these pumps. The speed of sputter-ion pumps will vary widely depending on the history of pump use. For example, if you bake a new pump overnight at say 300* C and then conduct N2 speed measurements, starting at 5 x 10-g Torr, you will initially observe speeds x 4 higher than will subsequently be noted at this same pressure after a week of pumping. If the initial measurement is conducted at 10-8 Torr, it will take about 8-10 hours for the speed to decay to a steadystate level. It will take progressively less time to reach steady-state speed values, the higher the initial pressure at which measurements are conducted. However, my experience has been that the time required to achieve a steady-state speed is not a linear function of pressure. The steady-state speed ultimately achieved has been coined the saturation speed. This does not mean that the pump has reached its capacity for the gas in question. Rather, it means that at the given pressure, the speed of the pump does not significantly change with time. If speed data are taken while progressively increasing pump pressure, the measured speed will initially be higher than the saturation speed, and decay according to the above time-table. If speed measurements are taken, starting at higher pressures, and then at progressively lower pressures, the initial speed measured will be low, and increase to the saturation speed level according to the above time-table. This effect was first reported by Jepsen.(2 4 ) Up to this point, I have stressed that the speed of a cell is proportional to I/P. However, because of transient effects, and pump saturation effects (e.g., see Section 2.5), this, as noted by Dallos, proves to be an oversimplification.(1 o 3 ) When making meaningful speed measurements, it is necessary, but not sufficient, to have pumped a prescribed amount of gas prior to taking data. For example, the 1976 edition of the PNEUROP standard requires that prior to taking speed
SPU'ITER-ION PUMPING
143
measurements, the equivalent o f - 4 x 10"2x S Torr-Z of the test gas must be pumped where S is the rated speed of the pump.(a n 8 ) The pump is presumably saturated at this point. However, were one to pump this m o u n t of the test gas, at say a pressure of 10-5 Torr, and by reducing the leak rate, subsequently establish a pressure of 10"a Torr, the observed speed thereafter would be excessively low for several hours. Conversely, were one to achieve an equilibrium speed at 10"a Torr some time after having pumped the required saturation quota, on varying the leak rate so as to increase the pressure to 10"e Torr, the initial measured speeds would be far higher than the subsequent steady-state speed. Reaching a steady-state speed - I prefer this to the term saturated - is primarily a time-dependant function, and extremely pronounced when making speed measurements with sputter-ion pumps. For example, these long-term, transient speed changes do not exist in cryopumps, or the momentum transfer pumps. With these pumps, speed-dome outgassing and gauge pumping effects, due to pressure changes, are more significant than actual changes in pump speed. Because of these effects in sputter-ion pumps, I suggest that after first satisfying some saturation and scrubbing quota, meaningful data can only be obtained by dwelling at a given pressure for a minimum time, tm, where, t m = kx kx k2 ks Pt
x exp(k2 x (logPt/ks)), = 24 hours, = -0.96, = 10-9 Torr, and = the test pressure, in Torr, Pt > 10"g Torr.
(2.7.1)
Equation (2.7.1) is not founded on some complex theoretical analysis. Rather, it is based on decades of experience in the speed testing of sputter-ion pumps. The above expression yields minimum test durations for steady-state speed results at 10-9 Torr of -24 hours, and 30 minutes at a pressure of 10"s Torr. Equation (2.7.1) applies to each reading. For example, should the pressure be altered by x2, it is necessary that (2.7.1) be invoked for the new change in pressure, if meaningful data are to be obtained. The above equation suggests that it will take weeks of elapsed time to make meaningful measurements of pump speed for various gases, and pressures varying orders in magnitude. This proves to be the case. Neglecting set-up time, similar speed tests with a cryopump might take only 2-3 days.
2.8 P u m p i n g G a s M i x t u r e s The effects of pumping mixtures of gases appears straight forward. Singleton reported the changes in the H 2 speed of a single-cell, diode pump as a consequence of increasing the partial pressure of N 2 .(1 0 2 ) At an H 2 partial pressure of --2 x 10"~ Torr, when increasing the partial pressure of N2 from --10 "g to 10"s Torr, the H2 speed of the pump increased by -x2.5. As the partial pressure of N2 was decreased, the H2 speed correspondingly decreased. Singleton concluded that the increase in H2 speed with increasing N2 pressure stemmed from the chemisorption of H 2 on free Ti sputtered by the heavier N 2 ions. On turning off the voltage, there was no sustained H 2 pumping, as in the case of pumping Ar and H 2 mixtures. This indicated that the sorption of N2 on the cathodes saturated activation energy sites, and precluded the dissociation of molecular H 2. Audi has done the most comprehensive work in the pumping of gas mixtures in
144
S P U T r E R - I O N PUMPING
triode pumps. In 1988 he published results of pumping N2 and H 2 in a StarCell ® triode.(S 0 ) That same year he reported results of pumping He and N 2, Xe and N 2, Ar and H2, and N 2 and H2 mixtures in a StarCell ® pump.(1 0 t ) The symbolism "X--Y", used below, is defined as: "measuring the pumping speed of gas X while progressively increasing the partial pressure of gas Y and holding the total pressure constant" ... In interpreting Audi's results, we must keep in mind general observations regarding ion pumps: 1) The I/P of sputter-ion pumps varies as a function of pressure. 2) The I/P of the StarCell ® triode pump for N2 is a maximum at -10 "6 Torr, decreasing as pressure is either increased or decreased. 3) The I/P for different gases varies as a function of the sensitivity of the Penning cells for the gas species. This results in a translation of the I/P curve along the pressure-axis. 4) The I/P for gas species X, in a mixture of gases X and Y, varies as a function of the partial pressure of gas X, and is independent of the partial pressure of gas Y. Conversely, the I/P of gas Y is independent of the partial pressure of gas X. 5) If X--Y, and the sputter-yield of gas Y >> X, even with the steady-state implant of gas X in the cathodes, as the pressure of gas Y is increased, eventually gas X will be sputtered away from the cathodes at the rate exceeding that of cathode implantation. Therefore, at that point, the measured speed of gas X would be solely due to pumping on surfaces other than the cathode. This is somewhat ameliorated by the unique configuration of the StarCell ® cathodes. Some gas may be back-scattered, and implanted as neutrals on the back side of the slanted cathode vanes. Neutrals would also less readily sputter-desorb gas which has been implanted on these back-side surfaces. The I/P (i.e., speed) of air, H 2, and He are shown in Fig. 2.8.1. The I/P data for He were calculated assuming speeds are proportional to I/P, He speed is -x0.25 the air speed, and the speed maximizes at --x 4 the pressure of the air maximum (i.e., another I/P consideration). Ionization cross-section data of Rapp and EnglanderGolden, for 300 eV electrons, were used to define the I/P peaks relative to N 2 (i.e.,
air).(169)
150
6
I/ I -
100 -
,
~
m
__
! HYDROGEN/ _ SPEED ~ - - ~ ~ I E "-
..SPEED~~./
N,TROGEr
r..,,i/P
"PRAY"
_
_
-
[
__
I/PCONSIDERATIONS,
SPEED FOR HELIUM -BASED ON ONE DATUM t¢ ~ / A ~ P ! I A N D
50
10 -9
10-8
10 -7
10 -6
J
I II
10 -5
PUMP PRESSURE - Torr F i g u r e 2.8.1. S h i f t in p u m p s p e e d " p r o f i l e s " of a 120 . ~ / s e c . pump with pressure, due to changes in I/P for different
StarCell ~ gases.
Audi's results for H2-*N2 are given in Fig. 2.8.2. Audi noted that for all pressures when He--N2, the speed for He decreases. Therefore, results for He--N2 are
SPU'YFER-ION PUMPING
145
explained by Observations 3, 4 and 5. The difference between the He speed for a saturated vs. unsaturated pump is probably attributable to saturation effects of neutral He pumping on the anodes and walls of the pump, and the sputtering away of He implanted in the cathodes. The fall-off in He speed of an unsaturated pump, with He-* N 2, is probably due to the decrease in I/P of He with the increasing partial pressure of N 2. ~:
2'01
o ~ tm
I
I
[_ /
l
l
~ Nr/"]
.a <
z
I
"
4 x l O -6 Torr t-(total)
l
[
[-
1.00
¢ f
_
"
F i g u r e 2 . 8 . 2 . V a r i a t i o n i n H2 s p e e d of a S t a r C e l l ® p u m p at constant total pressure, with increasing N2 p r o p o r t i o n . ( l ° l )
1 x l 0 -8 Torr
j¢'/ /
[
/ /
/
[
/
. M
4 x l O -7 Torr (to tal)
50 PERCENT NITROGEN PRESSURE
1 )0
Audi's H 2-* N 2 data were more revealing (i.e., Fig. 2.8.2). There are two competing mechanisms when H 2-+ N 2. One mechanism deals with the increase in H 2 speed as a consequence of an increase in the I/P of N 2, as the N 2 pressure is increased in proportion to the total pressure. The second deals with the H 2 liberated from the cathodes, (i.e., the net flux), as a consequence of cathode sputtering by the N2 + ions. Regarding the first mechanism, when H 2 -, N 2 at a total pressure of 10-6 Tort, and the N2 partial pressure is 10" 7 Torr, the N2 speed of the pump is - 7 0 Z/sec (i.e., Fig. 2.8.1), and the rate at which N2 molecules are being pumped is -2.3 x 101 4 N2 molecules/sec. However, assuming unity sputter-yield, the rate at which free Ti atoms are being created is twice this number (i.e., 17 -0.25). If when H 2-+ N 2 the total pressure is the same, but the partial pressure of N 2 is -6.5 x 10" T Torr, the speed of the pump is 115 Z/sec, N2 is pumped at a rate of --2.5 x 101 s N2 molecules/sec, and the rate of creation of free Ti atoms is twice this number. Therefore, though we have increased the pressure by --x6.5, we have increased the rate of creation of free Ti by - x 11 merely because of liP consideration of the N2. A decrease in H 2 speed at yet higher N 2 pressures would suggest that the rate of H 2 liberation due to sputtering of the cathodes by N 2 is exceeding the net gain in pumping of H 2 due to the sputtering of free Ti. As shown in the H2-*Ar tests conducted by Audi, there will be a precipitous reduction in H 2 speed at certain pressures. The H 2 pressure at which this occurs is a balance between: 1) the concentration of H 1 stored in the implant laminate of the cathode; and 2) the sputtering rate of Ar + ions, both as it relates to being a source of free Ti and erosion of the cathodes to liberate implanted H i . Arguments of a similar nature apply to the pumping of all gases (e.g., the effect of N2-+ He in all sputter-ion pumps(2 4 T )).
2.9 High Pressure Operation The problem of thermal run-away in sputter-ion pumps is well known.(1 s e ) It is particularly pronounced when pumping hydrogen, or when operating at high pressures of other gases, after the pump cathodes were previously loaded with hydrogen (e.g., as a c o n s e q u e n c e of pumping significant amounts of water vapor).(4 a, 1 0 2,1 2 t , 1 s 7,1 7 1 ) In the late 1950's, the primary product of Varian continued to be microwave tubes. Many of these tubes had oxide cathodes; that is,
146
SPUaq'ER-ION PUMPING
nickel-matrix cathodes coated with a mixture of carbonates of Ba, Sr and Ca, suspended in a nitrocellulose binder. The major source of gas from these cathodes came from a process called cathode conversion, which occurred some time during the bakeout of the tubes. That is, cathode heater power was applied and the carbonate mixture decomposed to form oxides of Ba, Sr and Ca; thus, the term oxide cathode. This decomposition process resulted in large throughputs of CO and CO 2. Because of heating effects, sputter-ion pumps could not handle the gas loads during cathode conversion. This in part led Jepsen to develop the first water-cooled sputter-ion pump. He applied for patent on such a pump in 1961,(1 7 2 ) and a modified version of this pump, with the cathodes water-cooled, in 1965.( 1 7 3 ) When Jepsen developed the first water-cooled pump, I was a maverick engineer at Raytheon. I was building equipment to process tubes - some having the dreaded oxide cathodes. They were dreaded by people using ion pumps to process their tubes, as the CO and CO 2 given off during conversion often swamped the pumps. A prototype of the earlier version, water-cooled pump was lent to me by a Varian salesman. The problem then was that I didn't have the vaguest notion what throughput meant. I was building hundreds of thousands of dollars of tube processing equipment, but didn't understand some of the fundamental vacuum concepts. Therefore, this water-cooled, sputter-ion pump languished in a storage cabinet - I was afraid to even use it - until it was retrieved by the salesman. It is ironic that decades later, I would develop the next generation of water-cooled pumps at Varian,( t a 8 ) and what became known as the Hi-Q ® pump.(4 s ) These water-cooled pumps helped somewhat in handling high gas loads. But, even this approach had its limits; it is primarily a question of power consumption. Throughput, Q, as defined by (1.11.1), is simply Q = S x P. Assume that the sensitivity, s, of each cell in a sputter-ion pump is 10 A/Torr and that the speed of each cell is -1.0 Z/sec. Assume the exponent n -1.0 in (2.2.4b). If we were to build a pump with a speed of 500 Z/sec, we would need -500 cells in the pump, each having a sensitivity of 10 A/Torr. Therefore, we would have a total pump sensitivity of 5000 A/Forr. This pump would have a pumping efficiency of I'/ = 0.1 Torr-Z/coul., for N2. (The speed rating for sputter-ion pumps is usually published for N2 or air). If we were to operate the pump at 5 x 10-4 Torr, current drawn by the pump, at this pressure, would be 2.5 A. If the power supply maintained a constant 7.5 kV at this high pressure, the power dissipated in the pump would be -19 kW. In this example, the throughput of the pump would be 0.25 Torr-Z/sec. A 15 cm ,t' turbo-molecular pump, cryopump or diffusion pump would be able to pump greater than x 10 this throughput, with only - 1 0 % of the power consumption of an ion pump. This is obviously a contrived example, used to make the point that sputter-ion pumps are not high throughput devices. To minimize current drawn in Penning gauges, ballast resistors are put in series with the gauge power supply.(2 a ) Similarly, because of changes in the impedance of the pump discharge with pressure, in the early days of pump development, ballast resistors were used to limit the pump current. Some resorted to the use of lightbulbs for these ballast resistors.( ~ 8 ) Of course, the maximum power dissipated in the pump occurs when the impedance of the pump is exactly that of the resistance of the ballast resistors. This implies that only half of the voltage is across the elements of the pump. Some form of power-limiting circuit is required on even the very large sputterion pumps.(1 s 8 ) As noted, large pumps may be constructed with numerous elements inserted into individualpockets. Using this technique, sputter-ion pumps have
SPUqTER-ION PUMPING
147
been constructed having speeds ;~5000 Z/see.(8 s ) In order to operate these pumps at pressures of--5 × 10-5 Torr, the DC power supply must be somewhat robust. But, this can lead to problems. For example, the diode sputter-ion pumps used on the LINAC at BNL have rated speeds of 1500 Z/sec. For years these pumps were plagued with cathode warping problems, under what was thought to be low throughput conditions. The pumps are constructed with 14 pockets, each of which contain two pump elements. The total pump sensitivity is --2 × 10 4 A/Torr (i.e., r/ = 0.075 Torr-Z/cod., for N 2 ). Each of the three power supplies used to power the 45, 1500 Z/see pumps distributed along the LINAC is capable of supplying 12 A at 5.5 kV. Separate ballast resistor banks are used to limit the current to each of the pumps. Each bank once comprised 60, 500 f~, wire-wound resistors, providing a total ballast resistance for each pump of 1875 t'). With this resistance, we calculate that the maximum power which can be dissipated in each pump is -4.0 kW. The 2.5 mm thick cathode plates each measured 10 × 22 cm. We calculate that the maximum power density dissipated in each of the cathodes plates is -0.3 W/cm2. Assume that, because cathode plates are in poor contact with the walls of the pump, they must dissipate this heat through black-body radiation. The maximum temperature they would achieve, under these conditions, is --200* C. Therefore, this didn't seem to be the cause of the cathode warping problem. The flaw in the above calculation is that I have assumed that the power is uniformly dissipated in each of the pump elements. This proved not to be the case. We observed from time to time that, because of excessive heating, cathodes in pumps would buckle and short to the anodes. In some instances, melting of the Ti cathodes occurred in one or two spots having the same diameter as the anode cylinders. Surface tension kept the material in place, but it was evident the material had melted completely through the plate. Near the melting temperature, the vapor pressure of Ti is --3 × 10"s Torr. This suggested that a self-sustaining, localized thermal runaway process was occurring, and that power dissipation in spots was > 100 W/cm 2. We sealed the input flange of one of these pumps with a viewing window, roughed the pump with a turbo-molecular pump to --10 -5 Torr, and started the pump. We were able to observe the discharge within most of the elements. Sure enough, though the discharge in all of the elements was confined, the discharge in one of the elements glowed a bright orange-red, where the others glowed with a quiescent blue. We maintained the power into the pump for several minutes, and on turning it off, noted that the cathode plates of the element which had the bright, orange-red discharge were now glowing orange, from heating. We repeated the process several times. Eventually, the orange-red discharge aged out of the first pump element. However, it was observed to move to a second and then third element. Before the glowing problem went away, we modified the ballast resistor bank, in series with the pump, to have a resistance of 7.5 lff~. Thereafter, with early start-up (i.e., P > 10-5 Torr), we were still able to initiate thermal run-away problems; the orange-red glow still wandered about the pump from element to element. Lamont suggested that thermal run-away in pumps could be caused by hysteresis effects in the discharge.( 2 ~ ) He noted that with near-constant voltage and decreasing pressure, I/P tended to be somewhat constant at high pressures. But, with increasing pressure, because of this hysteresis phenomenon, the I/P of a discharge peaks to a value as much as x5 that observed with the same but decreasing pressure. This effect is shown in Fig. 2.9.1. The very high value of I/P, shown in this figure at low pressures, stems from the very long anode used in this experiment (i.e.,
148
SPUT'rER-ION PUMPING
d -25 mm and ,~ -51 mm). P. 100 o
I
I
INCREASING PRESSURE ~
-
l I
Figure 2.9.1. Hysteresis Lamont, for d ~25.4 13 = 0 . 1 0 T, a n d V a
~ 5o r~ Z
effect mm, l = 3000
reported by --50.8 mm, v o l t s . (27}
_
0
t6.5
DECREASING ~ PRESSURE I
z
1
4
1
1
l
1
6
NITROGEN P R E S S U R E
l o- 4
-
Torr
If outgassing, due to heating, was sufficient it would cause all of the elements to share in the gas load, and uniformly share the power dissipation throughout the pump. However, if, after first pumping down and passing through the anomalous pressure region, the pressure increased locally in one of the elements, the current drawn to that element would increase to × 5 of that previously observed and cause the power dissipated in that element to correspondingly increase. A local increase in pressure could be due to metallic contaminants, local outgassing, or even the vapor pressure of the cathode material. For example, Tom reported on the effects of using high vapor pressure cathode materials in sputter-ion pumps.(t r s ) The orange-red discharge problem in the LINAC pumps stemmed from the materials we used when refurbishing these pumps. The process included disassembling the pump elements and sandblasting the cathodes and anodes with what we thought to be pure SiO2 grit. We noted that more often than not, the orange-red glow discharge was confined to one of the gaps in a pumping element. On disassembling a problem element, we noted there were strange metallic streaks on the outer fringes of the cathode plate which had evidenced the red glow. Technicians commented: "Oh, yeah, we've seen that on lots of cathode plates." An Auger analysis of these streaks indicated the presence of Si and Ca. Of course, it became evident that we had shot ourselves in the foot when we cleaned the cathodes by sandblasting. We discovered that the sandblasting material was in fact soda-lime glass beads, comprising only 73% SiO2, and a remainder of 15% NaO, 7% CaO, 4% Mg, and 1% A~ 2 0 3 . When sandblasting a cathode which had been extensively used, the fine grit was implanted in the microfissures created by the prior pumping of hydrogen (e.g., water vapor). A subsequent washing and 800* C vacuum firing process did not completely remove the grit. On reuse of the cathodes, ions desorbed the imbedded impurities, creating a local high pressure. As the cathodes heated, the vapor pressure of the volatile constituents, listed above, became operative. Local heating caused an increase in local pressure, which caused further heating, etc. A local thermal run-away condition was created in the pump by the presence of the contaminants introduced by sandblasting the pump cathodes. Of course, in retrospect, we know the orange-red glow was probably due to the Ca. The NaO component condensed on the cold wall of the vacuum furnace. We altered our pump refurbishment process thereafter. Three lessons were learned by this: 1) how not to clean sputter-ion pumps; 2) the presence of contaminants on the surface of one element, in a pump containing
SPUTTER-ION PUMPING
149
many elements, can cause overheating in the one element, while the others go unscathed; 3) power delivered to a pump should be such that if it is all diverted to one element, the element will not self-destruct.
Advantages in Low Pressure Operation On the brighter side, what is a disadvantage with sputter-ion pumps at high pressures is a major advantage in their use at low pressures. For example, you will recall that each of the SLAC vacuum sectors was pumped with four, 500 Z/sec pumps, and there were thirty such sectors, making a total of--120 such pumps. Assume that for the gas N2, I'/ = 0.1 Torr-Z/coul. for each pump, and a pump voltage of 5.0 kV. Then, if the average pressure in each sector was 10" 8 Torr, the total power drawn by the 120 pumps would be 30 W, or 0.25 W/pump! Where turbo-molecular, diffusion and cryopumps were more efficient at the higher pressures, of the order of 100 - 200 kW would be required to power --120, of the latter, 500 Z/sec pumps. The low power consumption of sputter-ion pumps at low pressures is one of the primary reasons they have received such wide acceptance in the high-energy physics community.
2.10 The Sputter-Ion Pumping of Helium It was initially proposed that the cold bore of the RHIC be pumped with SIPs. However, there was some concern about possible memory effects which might occur as a consequence of pumping He in this application. Therefore, we set out to determine the following: i) which type pump had the greatest capacity for He;//) which had the greatest long-term speed for this gas (a corollary of ii); iii) what is the effect of pumping He with the various types of SIPs; and, iv) could any of the pumps be regenerated (i.e., be somehow brought back to their original low pressure He speeds) after the pumping of large amounts of He? In all types of sputter-ion pumps the effects of pumping He tends to be much less dramatic than when pumping argon. That is, He pumping merely results in a gradual increase in the pump's base pressure with increasing amounts of He pumped. Much work has been done in the last 40 years in measuring the speed and capacities of SIPs for inert ~ases of > 20 amu. However, little data existed on the He speed and capacity of SIPs. t, 1 0 a, 2 4 9,2 s o, 2 5 1 ) The first quantitative results for the pumping of He with diode and triode pumps was reported in 1994 .(2 4 7 ) By definition, the base pressure of a pump is that pressure at which the speed of the pump is zero. We arbitrarily defined the base pressure of the pumps tested versus the amount of He pumped, as that He partial pressure achieved by the pump 20 h subsequent to the removal of the He source. (This is a very optimistic definition of capacity.) Remember Po in (1.15.1) of Chapter 1, is the base pressure of the pump, and if Po = P, the operating pressure, the speed of the pump is zero. All capture pumps have the characteristics that Pe varies a function of f Qdt, where f Q d t is the quantity of a particular gas pumped at the end of some time t. This effect is illustrated in Fig. 2.10.1, where the speed curve is shown to gradually shift to the right as more gas is pumped. That is f Q1 dt < f Q 2 dt < f Q 3 dt. Therefore, if Pe is a function of f Qdt of a given gas, it is meaningless to make a statement about the speed of the pump for that gas with out first quantifying the amount of gas pumped.
150
SPU'Iq'ER-ION PUMPING %
I
) ) I
I
Sn
-Smax.
(1
I ) I
" I
I I'
Po//P)
100 Smax.
~X
75
a~
50
z
25
0
fQ1 t
0
I0
J /I I
-9
fQ L/
[ t 1
Pl 10 - 8 P2
I/
10
I I 1
- 7 P3
I0
-6
PUMP OPERATING PRESSURE - Torr Figure 2.10.1 Increases in a capture pump's base pressure (i.e., P1, Pz & P3) as a c o n s e q u e n c e of pumping increasing a m o u n t s of
a particular
gas
(i.e.,
f Qldt,
f Q2dt
& f Q3dt
respectively)
A description of the five types of pumps tested is given in Table 2.10.1. Subsequent to the He capacity and speed tests we measured the nitrogen speed of the pumps. This was done to establish their relative sizes rather than classify them based on vendor speed definitions. The projected cathode areas noted in this table includes that of the pseudocells formed by nesting anode cylinders. The 80 Z/s diode and 60 Z/s DI @ diode pumps were identically configured except for the cathode materials. These two pumps and the StarCell triode pump were new at the time of test, whereas the other two pumps were used in previous hydrogen speed and capacity tests. Table 2 . 1 0 . 1
Pump
configurations
for w h i c h h e l i u m
TYPE PUMP
SPEED L/see.
NB
CATHODE MATERIAL
DIODE, STD. 80 L / s e c I DIODE, DI® 60 L / s e c I TRIODE, StarCell ® 120 L / s e e DIODE, STD. i i 0 L / s e c DIODE, STD. 270 L / s e c
39 z 36 3 79 4 1405 ND
Ti/Ti Ta/Ti Ti/Ti
NOTES:
I) Enlarged p u m p
aperture, w /
Al/Al Ti/Ti 203 m m
NO. OF CELLS 64 64 106 126 126
c a p a c i t y was m e a s u r e d . CELL CELL PROJECTED L E N G T H DIAM. CATHODE mm mm AREA-cm 2 14.4 14.4 25.4 25.4 35.6
17.8 17.8 19.8 20.3 30.7
ConFlat ® flange to m a t c h
2} Speed measured @ 3 x 10 -7 Torr after p u m p i n g
1.8 Torr-L N2.
3) Speed measured @ 4 x 10 -7 Torr after p u m p i n g
0.33 Torr-L N2.
4) Speed measured @ 2 x 10 -7 Torr after p u m p i n g
0.19 Torr-L N2.
464 464 835 1176 2186 CERN
dome.
5) Speed measured @ I x 10 -7 Torr after 4 hrs. elapsed time; speed w / Ti/Ti cathodes and under similar conditions, was ~II0 L/see.
We used a test apparatus as illustrated in Fig 2.10.2. A NIST traceable capacitance manometer and volume V 1 (including capillary tube and gauge) were used. The throughput of test gas into the dome could be verified by calculating V 1 dP 1/dt. Also, the total amount of test gas introduced into the pump could be accurately
SPUTTER-ION PUMPING
151
measured by calculating V x AP 1 between each test. Between successive He measurements, the entire gas introduction system was evacuated (i.e., just in ease the variable leak was not entirely closing). A long SS capillary was used to separate the gas introduction system from the test dome. This made it possible to bake the dome and pump under test for 48 h at 250* C without risk to the capacitance manometer. VARIABLE I LEAK/
1
I/
APERTURE CONDUCTANCE
NUDE BAG
QRGA TEST GAS
Figure 2.10.2
D/2
,
[ ~
Modified F i s c h e r - M o m m s e n
ALL-METAL VALVES speed m e a s u r i n g
dome.
A partial pressure gauge is required when making speed measurements with one gas which might cause desorption of a second gas in the SIP. For example, assume that we are presently making pump speed measurement with gas a, after previously pumping a significant quantity of gas b. Assume that gas b is inert so that it is pumped exclusively by physisorption. On pumping gas a, gas b is desorbed to some partial pressure in the test dome. If there is little pumping of gas b by the BAG or QRGA, the desorbed gas b will be at the same partial pressure on both sides of the aperture conductance. Assume Pah is the pressure of the test gas a on the high pressure side of the aperture, and Pal is the pressure of test gas a on the low pressure side of the aperture. Using similar notation for test gas b, and (1.16.19) of Chapter 1, we f'md that the indicated speed, Si, of the pump is:
Si
= Ca
[(Pah + Pbh)- (Pal + Pbl)]
(2.10.1)
(Pal + Pbl) If Pbh = Pbl, these terms cancel in the numerator. However, the term Pbl remains in the denominator, and causes an error in the measurement by so being. After bakeout of the system and pump, the He capacity and speed tests progressed as follows: i) speed measurements were conducted in the pressure region of 10-9 to 10-6 Torr (i.e., see Table 2.10.2 below);//) the variable leak was then shut off, and the gas introduction system evacuated; and, iii) the He background pressure in the pump was measured > 20 h thereafter. Helium capacity results for four of the five pumps are shown in Fig. 2.10.3, whereas speed results are given in Table 2.10.2. The diode pump with aluminum cathodes pumped -3.4 Torr-Z of He prior to
152
S P U T t E R - I O N PUMPING
pressure run-away. Sequential He base pressure measurements were not made with this pump. 10-7
o
1
-
t
I
II1
I
~_ O 60 _ [] 80 × 120 ~ 270
0_ 8
I
II1
1
1 I1
1
L/sec
DI®Diode
L/see L/see L/see
Std. Diode StarCell ® Std. Diode/,)"
1 11
I
1 I,t-
_ _~_
(p
/
_
i
10 -9 51J
5x.] .<
10
/
-10
10 -11
lO-la
l
10
-4
AMOUNT Figure 2.10.3. subsequent
Table pumps
Helium
as
function
TYPE PUMP
DIODE, STANDARD 80 L / s e e
DIODE, D I ® 60 L / s e e
TRIODE, S t a r C e l l ® 120 L / s e e
DIODE, STANDARD 110 L / s e e w / AI CATHODES
DIODE, STANDARD 270 L/see
*One
-3
10 OF
/
1
10 PUMPED
-2
l
10
I tl
-1
HELIUM
1
10 -
I II
0
10
Torr-Liters
Helium base pressure of various pumps ~20 hours to the pumping of controlled amounts of helium.
2.10.2. a
I tl
standard
pumping of
He
speeds pressure
HELIUM SPEED
of
a
and
variety quantity
of
sputter-ion
of
He
fTQodrtr - LHe
pumped.
L/see
o*
HELIUM PRESSURE Torr
12.2 18.3 14.7
0.2 1.8 0.2
1.2 x 10 - 8 2.1 x 10 -7 8.0 x 10 -7
7.6
neg
2.4 x 10 -6
1.7[} x 10 -3 0.0814 0.226 0.514
12.6 14.4 7.0
0.8 1.3 0.4
3.5 x 10 -9 3.0 x 10 -8 1.2 x 10 -7
7 . 0 3 x 10 -4 0.0824 0.661 9 . 2 6 x 10 -4 8 . 3 3 x 10 - 3 0.0889 0.486
13.5
1.0
1.7 x 10 - a
13.5 25.5 28.1 23.4
1.0 neg 0.7 0.7
2.9 1.6 1.2 1.2
x x x x
10 -8 10 -7 10 -6 10 -6
51.0 51.6 13.5 19.9
9.0 1.2 0.7 1.0
2.4 1.2 2.1 2.6
x x x x
10 -8 10 -7 10 -7 10 - 7
6.1 x 10 -7
4.8
ND
57.0
3.0
1.3 x 10 - a
79.0 98.0 10.0
neg neg 1.5
5 . 9 x 10 -7 5.0 x 10 -7 1.0 x 10 -6
deviation.
....
1.000
5 . 0 0 x 10 - a 0.050 1.538 1.749 2.451 1.13 x 10 - a 1.120 6.635 7.700
SPUTI"ER-ION PUMPING
153
Helium Capacity Observations The above measurements suggest the following: 1) There is little if any difference in He pump capacity or speed between the standard 80 Z/s diode pump and the DI ® pump. This is of course not the case when pumping argon where the DI ® pump would be expected to have - x 2 0 times the argon speed as a conventional diode of the same configuration. 2) The projected cathode surface area of the 270 Z/s is only --x4.7 that of the 80 Z/s diode. Yet the He base pressure of the former pump is - x 1000 times lower than that of the 80 Z/s diode on pumping an equivalent amount of helium in the lower pressure regions. The base pressure of a pump at low pressures and subsequent to pumping He depends both on the projected cathode area, and I/P characteristics of the pump. Because of the larger diameter and longer cells of the 270 Z/s pump, the discharge intensity of each cell in the larger pump is much greater. Therefore, the speed at very low pressures is proportionately greater. 3) The triode pump appears to have proportionately lower capacity for He with progressively greater charges of He (i.e., the data for the triode and smaller diodes converge with increase He charge. Note however, that the speed of the triode pump seems more constant with charge than the comparably sized diode pumps. Note also that the triode did not evidence a capacity curve inflection with higher charge. Further He capacity tests of the StarCell ® pump would have been worthwhile.
Regeneration After Pumping Helium Pumps were regenerated by baking, and glow discharge cleaning with augmented pumping. After the regeneration, nitrogen speed measurements were taken. The purpose of this was to determine if the pumping of nitrogen would cause renewed desorption of previously pumped He. Three of the pumps were regenerated by baking: the DI ® diode, 80 Z/s diode and the StarCell ® triode pump. The regeneration of the standard 80 Z/s diode amounted to 48 h, 300* C bakeout on itself. This was tantamount to the pump stewing in its own juices over this period. After this regeneration, the He partial pressure in the pump was 4.0 x 10-1 o Torr (i.e., a - x 100 improvement in base pressure). Well, this was short lived, as after pumping -0.33 Torr-Z of N2, and with an N 2 partial pressure of 4.0 x 10" T Torr, the He partial pressure equilibrated at 3.8 x 10" ~ Torr. Pumping N 2 at a pressure of 1.3 x 10-6 Torr caused the He background in the pump to creep up to 2.5 x 10-6 Torr. That is, the He pressure exceeded the N 2 pressure by ---x2. After pumping 1.0 Torr-Z of N 2 and then removing the N 2 leak, the He partial pressure equilibrated at 8 x 10" 1 o Torr. The standard 80 Z/s diode was regenerated by baking it on itself at 275* C for 16 h, and then augmenting the pumping on the SIP at this temperature with a turbomolecular pump for 2 h. With this augmented pumping and while still at 275* C, N 2 was introduced to the pump at a pressure of--3 x 10"s Torr. The pump was left energized (i.e., -200 W) for 7.5 min., the turned off for 2.5 min., for a total of five such cycles. At the conclusion of this regeneration process - during which time the pump consumed -2.66 Torr-Z of N 2 - the pump base pressure was < 2.6 x 10" 1 4 Torr in 22 h, the minimum discernable signal (MDS) of the QRGA. However, in the process of pumping an additional 1.07 Torr-Z of N 2 at room temperature, the He background pressure crept up to 9.6 x 10-7 Torr with the N 2 pressure set at 2.4 x 10-6 Torr.
154
SPUTTER-ION PUMPING
The StarCell ® pump was regenerated by baking for 28 h at 300* C, with the high voltage off, and while pumped on with a turbomolecular pump. After this, and with the pump still at temperature, the pump was cycled through several glow discharge scrubbings with N 2 at a pressure of 5 x 10"s Torr. After this bakeout and glow discharge cleaning, the base pressure of the pump was below the MDS of the QRGA in 22 h. On the subsequent pumping of N 2 , the He partial pressure always remained --x50 less than the N 2 pressure. However, this may not have been because we had found the perfect regeneration recipe. It also stems from the fact that much of the He is pumped by burial in the wall of the pump, where it can't be desorbed by energetic ions. However, as in the case of the DI ® pump, the He pressure in the StarCell ® gradually increased in time when holding the N 2 pressure constant. This effect is shown in Fig 2.10.4. In these room temperature tests, the throughput of N~ into the two pumps was in addition to that stemming from the regeneration processes. From these observation, one must conclude there is no perfect recipe for the regeneration of any SIP after pumping He. Further, it is evident that there are advantages in the use of triode-type pumps in such applications as there is less induced He desorption caused by the pumping of a second gas. 10
-5 -
-
I
1
DI ® D i o d e
0
I
10
-6
StarCell ® Triode
C~
210
/
-7
Nitrogen Pressure~1.3 x 10-6Torr 10
-8 0
0.2
0.4
NITROGEN PUMPED Figure 2.10.4. regeneration
Helium and as
0.6
0.8
-
Torr-Liters
1.0
partial pressure subsequent a consequence of p u m p i n g
to p u m p nitrogen.
2.11 Pumping Element Located in Antechambers or "Pockets" Assume that a pump element is located in an antechamber or pocket situated adjacent to a much larger chamber, as shown in Fig. 2.11.1. This pocket may represent one of many pockets built into the body of a pump having numerous pump elements. It might also represent a DIP (distributed ion pump) located in the beam pipe of a storage ring. Though rarely cited, Jepsen was the first to publish the solution to this boundary value problem.(1 r 6 ) Referring to Fig. 2.11.1, Jepsen's results and derivations thereof are given below.(1 r r )
SPUTrER-ION PUMPING
/••/•">Xo
//
//
\~-':"~ ~'Z~ ~
--
PUMP "POCKET" OF DEPTH xo AND WIDTH 9
I
CATHODES --//
155
I I
g
N ANODE CYLINDERS --Y OF LENGTH l Figure 2.11.1. P u m p "pocket" d i m e n s i o n s needed to calculate the effective speed at the opening, for a known i n t r i n s i c speed. Assume there are N cells in the given element which is inserted into the pocket, each of which has an intrinsic speed of S O. The intrinsic speed of the total element is simply NS o. The I / P of a given cell is directly proportional to the length of the cell, I . Therefore, assuming that the gap, g, between the two cathodes is fixed, the intrinsic speed of any one cell is given by: So
= I~
= K(g-2a),
(2.11.1)
where the constant K is a function of magnetic field, voltage, cathode material and cell aspect ratio, and a is the anode to cathode spacing. The intrinsic speed of the total element then becomes: S
= NS o = N K ( g - 2a) = ¢0N1 K ( g - 2a),
(2.11.2)
where N 1 is the number of anode cells per unit width of the anode element (i.e., the length of the pocket). The effective speed of the pump, S', depends on the intrinsic speed of the total element, and the conductance C, leading into the antechamber. The approximation is made that this conductance, C, say for N 2, is equivalent to that of a slab projecting into the two gaps between anode and cathodes, in the antechamber where:(1 7 s ) C
= 2 x 30.9a 2 ~)/x o Z/sec-cm 2 A
k x a 2 o)/x o Z / s e c - c m 2
(2.11.3)
The ratio of the effective speed, S', delivered to the chamber, to the intrinsic speed of the element is found by solution of the following two differential equations:
156
SPU'ITER-ION PUMPING dQ(x) = -
P(x) (S/xo) dx,
(2.11.4)
dP(x) = -
Q(x) (Cxo)-1 dx,
(2.11.5)
where x o is the depth of the pocket, and Q(x) is the throughput at any plane x in the pocket. Equation (2.11.5) is less obvious than (2.11.4). Using a transmission line analogy, the expression (Cxo)" 1 can be looked at as the impedance, Z, per unit depth in the pocket, where (Cxo)" 1 = Z/xo" Using (2.11.4) and (2.11.5), and the boundary condition dP(x)/dx = 0, at x = x o, the expression for the pressure, P(x) is:
P(x) = c 1 [exp(kx) + (exp(2kxo))X(exp(-kx))],
(2.11.6)
where c 1 is some constant, and k = (S/xo 2 C). Substituting the value of P(x) in the right hand side of (2.11.4) and integrating from x = 0 to x o yields the total rate at which gas is pumped throughout the pocket. By the same token, the product S'P(x), evaluated at x = 0, represents the same throughput. Or, xo
s'P(o) L f P(x)(S/xo)
(2.11.7)
0
This results in the following expressions for the intrinsic speed, S', and the effective speed per unit width, to, of the pocket:
S'/S = (tanh(S/C) ~ )/(S/C) y'
(2.11.8)
Equations (2.10.2) and (2.10.3) may now be used in (2.11.8) to solve for the optimum pump configuration. Note that kl in (2.11.3) varies depending on the type of gas and temperature of the chamber. The assumption in (2.11.3) is that the chamber is at RT and the gas is N2. Assume that the gaps, a, are held constant, and the total intrinsic speed remains the same. The total intrinsic speed may be expressed by S = x o x to x Sa, where S a is the pumping speed per unit area of the anodes. By making these substitutions, differentiating (2.11.8) with respect to to, and setting this value equal to zero, we find the maximum effective speed, S', delivered to the chamber is given by the following relationship: x o -- 0.125 a (kl/Sa) ½
(2.11.9)
In that the pump comprises integer arrays of anode cells, we should not over interpret the significance of (2.11.9). However, assuming an anode-to-cathode gap o f - 0 . 7 5 cm, and S a -0.25 Z/sec-cm 2 , leads to the value x o -1.7 cm. This suggests what was already intuitively obvious; that is, because of conductance limitations, for a constant value x o x to, making to >>x o will yield the most effective use of cathode material, and the highest S'/S ratio. Of course, in component pumps, magnetic circuit design considerations impose practical limits on the ratio to/xo.
Distributed Sputter-Ion Pumps (DIPs) The development of electron-positron storage rings was ultimately responsible for the development of DIPs. G.K. O'Neill noted in 1963 that, in the Princeton-Stanford storage ring, there was "... a rather violent pressure rise due to synchrotron
SPU'Iq'ER-ION PUMPING
.(
157
light., t r ~ ) This ring was originally pumped with diffusion pumps. He speculated that "... soft gamma rays from the synchrotron light make photoelectrons which are in turn responsible for the cracking of pump oil." This effect is modeled as follows: 1) As the result of being bent in a circular orbit (i.e., accelerated radially), the electron (or positron) beam emits a highly collimated beam of photons (i.e., synchrotron radiation) tangent to the orbit of the electron beam; 2) The photon beam impacts on the walls of the beam chamber, dislodging photo-electrons; 3) The photo-electrons leave the surface, but are bent in a circular orbit by the magnetic field used to bend the primary beam; 4) On impact with the chamber walls, the returning photoelectrons desorb gas. The desorbed gas, generated the full length of the beam chamber, raises the average pressure in the machine, and results in decay in primary beam current due to scattering. Pumping on each end of the beam chamber is only a partial solution, due to conductance limitations of the beam chamber. Fischer and Mack, using synchrotron radiation from the Cambridge Electron Synchrotron Accelerator, did the first quantitative measurements of gas desorption off of Cu surfaces resulting from the synchrotron radiation.(1 8 0 ) Their work was done under UHV conditions and led them to the conclusions that a "separate function system" would be needed to pump the gas desorbed by synchrotron radiation. By this they meant that there would be two pumping systems, one to handle the ambient gas load, and a second "distributed pump" to handle the gas desorbed by synchrotron radiation. Their proposed distributed p u m p comprised TSP (titanium sublimation pumps) filaments, strung along the length of the beam chamber. The filaments were located behind parallel-plate electrodes, located in the top and bottom of the beam chamber. These plates served to separate counter-rotating electron and positron beams, and also served to shield the beam from sublimed Ti, which because of its high Z, would cause beam scattering. Anashin, et al., in 1968, were the first to report the use of distributed sputter-ion pumps in the VEPP-2 electron-positron storage ring, located at the Institute for Nuclear Physics in Novosibirsk, Russia.(t 8 1 ) These diode pump assemblies were distributed along the lower plane of the beam chamber, behind the lower beam separation electrode. They made use of the same magnetic field used to bend the electron-positron beam, to support the Penning discharge. Actually, as early as 1959, Jepsen suggested such a possibility, stating: "In special applications ... elements may be placed directly in the system ..., thus improving conductance ... In some cases incidental or stray magnetic fields associated with the system can be utilized,.(2 4) This scheme, as in the case of the Fischer-Mack TSP approach, required a larger gap in the dipole bending magnets and resultant increase in cost. Two years later, Cummings, et al., described the proposed vacuum chamber configuration of the Stanford (SLAC) Electron-Positron storage ring. As shown in Fig. 2.11.2, it featured an extruded A~ beam chamber, with the DIP housed in an electrostatically shielded antechamber on the inboard side of the chamber radius of curvature.(t a z, 1 s 2 ) This was the first major application of extruded A~ beam chambers in storage rings. The approach was met with some skepticism. Norman Dean, who managed construction of the vacuum system, related to me: "They said we were crazy to use extruded A~ chambers in SPEAR." The SLAC DIP, also making use of the magnetic field of the dipole magnets, was located in the same plane as the electron-positron beam, thus minimizing the gap of the dipole bending magnets. Since the original work at Novosibirsk, and then SLAC, the use of both diode and triode DIPs, of varying configurations, has found •
158
SPUI'rER-ION PUMPING
wide acceptance in storage rings (e.g., 7 o, 1 • 0, i a • - 1 ll •,2 a a ). Also, the use of extruded A£ beam chambers has found wide use in accelerator applications. ~ 1 8 4 ANODE CYLINDERS, 12.5 m m ~) x 18.0 m m LONG, ~ DISTRIBUTED ALONG CHAMBER
~
/[
B
RIBBED WALL TO DECREASE PHOTOELECTRON YIELD
~.
\
l i: ....J" " i ~ \ ~ . ~ \ \ \ . - ~ ~ ,-4-~- ; , ~ - [' .... - . . . . . .
....
HIGH VOLTAGE FEEDTHROUGH
Figure
2.11.2.
..
///
TITANIUM CATHODES Distributed
\\\\\\\\~
\\\
:3.5 m EXTRUDED / ALUMINUM CHAMBER pump
in
SLAC e - e +
1
'~'~
COOLING WATER PASSAGE storage
r i n g . (133)
2.12 Pump Power Supplies Little new transpired in the development of sputter-ion pump power supplies in the last twenty-five years until the advent a dozen or so years ago of switching power supplies. Because of their light weight and power efficiency (e.g., -80%) these types of power supplies are widely used in airborne radar and ECM systems.(1 9 s ) Varian Associates first exhibited a sputter-ion pump switching power supply at the 1989 IUVS Symposium, held in K5 In, Germany. Current
Feedback
Voltage Feedback
vpc Iv O
0
I
0
Signal
IDC
PULSE WIDTH MODULATOR
TRANSFORM UP & RECTIFY
I
Dl LINE POWER
Signal
/
D2 v
! L
I m
c+
VDC
[~
H.V.
L_
C2 =
v
F i g u r e 2.12.1. Block d i a g r a m of V a r i a n SIP s w i t c h i n g p o w e r s u p p l y ( r e p r o d u c e d w i t h p e r m i s s i o n of V a r i a n V a c u u m ) . A block diagram of the Varian switching power supply is shown in Fig. 2.12.1.(2 s 2 ) In this supply the line voltage is rectified to DC. Components O x and 0 2 are switching transistors. If they are both open, the DC voltage charges capacitor C x and C 2 and no voltage appears across the transformer. However, if one of the transistors is closed, the associated capacitor discharges through the transformer,
SPUTTI~R-ION PUMPING
159
thus inducing a voltage in the transforming and rectifying output circuit. The switching transistors are alternately switched at very high frequencies (e.g., 50 to 100 KHz). The wave forms and frequencies of the switching signals may be modified to produce the desired output voltages. The pulse width modulator is actually a controller which may be programed to both control and limit both the output current and voltage of the supply. Further, because they can be easily controlled with low level signals, they lend themselves to being networked in very large systems. HOST COMPUTER Ethernet !
i
1
1
I
i
RS-485 Data Bus PUMP I PS
I
i
Pt MP ***o ~ [
S
UP TO 31 Intelligent Power Supplies Or Other Devices
F i g u r e 2.12.2. S c h e m e for n e t w o r k i n g m u l t i p l e SIP p o w e r s u p p l i e s to a s i n g l e h o s t c o m p u t e r . Figure 2.12.2 illustrates a networking scheme which was used to interrogate and control --1000 intelligent ORGAs, ion gauges and sputter-ion pumps power supplies at the RHIC.( ~ s 3 ~ The PLCs (programmable logical controllers) control and receive the status of the numerous intelligent device controllers (e.g., Pump Power Supplies), and report status back to the host computer. Such networking schemes are also used to control and monitor numerous cryopumps on large cluster tools. In order to deal with thermal problems at high pressures, pump power supplies in the early days had ballast resistors. With a robust power supply, the power dissipated in these ballast resistors can be excessive. This prompted Hall to use conventional household light bulbs for these resistors.(4 ) Rutherford applied for patent on a power supply which pulsed the high voltage to the pump at a pulse repetition frequency (i.e., duty cycle) which was varied in an inverse relationship to the peak power drawn by the pump.(4 9 ) In this manner, it was possible to limit the power drawn by the discharge at high pressures. The industry workhorse presently comprises a high voltage transformer in series with a full-wave, bridge rectifier circuit. The high voltage transformer is loosely coupled, so as to become saturated as the impedance of the pump decreases (i.e., with increasing currents at high pressures). This has the effect of limiting power drawn by the pump. This scheme, along with a voltage-doubler circuit, was patented by Ouinn and Mandoli in 1967.(1 9 6,1 9 T ) Because of the loose coupling of the transformer, it is called a soft transformer. This technology, switching power supplies, along with a log-reading circuit (i.e., a circuit to measure pump current over a broad dynamic range), patented by Mandoli,(1 9 8 ) comprise the only significant developments in pump power supplies in the last 30 years. Some manufacturers are now offering pump power supplies with RS 232, as well as RS 485 interfaces to facilitate remote monitoring and control through a computer interface.
160
SPU'IWER-ION PUMPING
If pump current drawn from the power supply is to be used as an inference of very low pressure, care must be taken to filter out power supply ripple. Capacitance of the pump and high voltage cable will cause AC currents which will obscure low pressure pump readings. For example, the capacitance of a 20 Z/sec diode pump is of the order of --60 pF. The high voltage cable leading to the pump has a capacitance of -90 pF/m. Assume that the power supply provides 5.0 kV to the pump, with 1.0% regulation. The ripple of a full wave, bridge rectifier circuit is at a frequency of 120 Hz. At zero pressure, AC currents in the system would be 12/.tA and be equivalent to an operating pressure of--4 x 10" 8 Torr. The problem is much more pronounced when very long runs of high voltage cable are used, as in accelerator applications. For example, the power supplies attending the Brookhaven AGS are located on the average --200 m from the pumps. The pumps terminating the cables have capacitances of only --200 pF, but the cables have capacitances of -0.018/.tF. Assuming similar regulation and operating voltages, AC currents (i.e., noise in the system) would be equivalent to operating pressures o f - 4 × 10-7 Torr; with 0.1% regulation, this would be --4 x 10-8 Torr. If some sort of computer-aided current sampling system was used to measure pump current, signal noise would appear as flutter on top of the DC discharge currents. Sputter-ion pumps tend to be inhospitable loads. They arc, sputter and frequently short out. As noted in Section 2.9, this requires that some form of current limiting provision exist in any power supply used on these pumps. Also, unlike the simple bridge rectifier supplies, with soft transformers, switching power supplies tend to be inherent noise generators. The high voltage cables connecting power supplies to sputter-ion pumps are essentially TEM waveguides. Therefore, provisions must exist for filtering out the noise from the supply to the pump or this noise might prove troublesome in the user's system. The reverse is also true: the TEM waveguide (i.e., pump high voltage cable) will transmit noise generated in the pump back to the power supply.
2.13 Magnet Designs In the early days of sputter-ion pump development in the United States, various forms A~ NiFeCo alloy magnets were used. These alloys, identified by the tradename Alnico ®, were first developed in the mid 1930s. In 1940 a breakthrough in magnet design occurred with the development of Alnico-V, a domain oriented, permanent magnet alloy comprising the above elements and a pinch of copper.(1 9 9 ) Casimir tells of the accidental discovery of the ceramic-type magnets which are extensively used in today's sputter-ion pumps.(2 o o ) Apparently, some time in 1950, a technician at Philips, when mixing up a chemical recipe for an experiment, made an error in the proportions of the mixture, and the result was a permanent magnet of unique properties. Of course a great deal of further development work followed (and continues) at Philips, but this error resulted in what came to be known as ferroxdure in Europe and magnedure in the United States; a sintered, ceramic compound of BaO. 6Fe2 03 known as barium hexaferrite. An excellent review paper was published on the oroperties of forms of this ferroxdure in 1977 by van den Broek and Stuijts of Philips.~ 2 o 1 ) They point out in this article the financial significance of the discovery, as magnets could now be produced using relatively inexpensive and readily available materials. A comprehensive treatment of ferromagnetic materials,
SPU'I'rER-ION PUMPING
161
including the Alnico alloys and all forms of the hexaferrite materials, is found in a recent work edited by Wohlfarth.(2 o ~ ) Helmer published the first comprehensive theoretical treatment of the use of hexaferrite (ferrite, hereafter) magnets in pump applications, including multiple gap, periodic structures.( 2 o a ) He filed for a patent in 1961 on, what I later reinvented and coined, a confined magnetic field circuit, using ferrite magnets.(1 a 8,2 o t ) Two years after Helmer's publication, Kearns, of the General Electric Company, pubfished an article on some of the more practical considerations dealing with pump magnetic circuit configurations in which either the ferrite or Alnico magnets are used.(2 o s ) Kraus's chapter on ferromagnetic materials serves as an excellent refresher on the design of permanent magnet circuits.(2 o e ) The above references most adequately cover the design of magnetic circuits for sputter-ion pumps. Sputter-ion pumps manufactured throughout the world today are almost exclusively provided with ferrite magnets with steel magnetic return circuits. Minor exceptions to this are the very small appendage pumps (e.g., -0.2 - 2.0 Z/sec); some of which are still offered with cast, Alnico-type magnets. These Alnico magnets are used in this case strictly for reasons of economics. It is more cost-effective to purchase these small, cast-alloy magnets from, for example, Indiana General or the Crucible Steel Company, rather than buy the ferrite magnets and build the required magnetic circuitry. FERRITE MAGNET SLABS
IRON
w-fq--fq-fqn ~r-'t-fqrf-'rE
POLE-~Fll !! !1 Ill III !! Ifl IF]/BENT IRON YOKE
F i g u r e 2.13.1. I r o n y o k e p o l e piece with ferrite magnets.
FERRITE ~ MAGNET ~
IRON
POLE
(*( + ~*I ~., PUMP \~... i ~ J . , ~ ~ "POCKETS"
F i g u r e 2 . 1 3 . 2 . Closed l o o p m a g n e t i c c i r c u i t with iron pole pieces.
In ferrite magnet packages, slabs of the material are located on the inner faces of steel (iron) yoke assemblies such as shown in Fig. 2.13.1. These assemblies are then slipped over the pockets containing the pump elements. An example of a magnetic circuit which might be used in a very large, multi-pocketed, sputter-ion pump is shown in Fig. 2.13.2. Depending on the number of pump pockets, all magnetic circuit designs presently used are variations of these two circuits. For over 15 years pumps similar to that shown in Fig. 2.4.9, and the smaller appendage pumps, were offered with horseshoe-type, cast Alnico-V or Alnico-VIII magnets. These magnets, besides being very heavy, had excessive stray magnetic fields. Because of concern for the effects of these stray fields near microwave tubes, Helmer developed a magnet package with a confined field.(2 o i ) That is, there was negligible stray field even at the surface of the magnetic circuit. It was specifically intended for use with pumps similar to that shown in Fig. 2.4.9, which appended large klystron tubes, having military applications. As a spin-off of work on the HiQ ® pump, in 1975 1 developed a compact "confined field" magnet package for use with this same pump, and to replace the Alnico-VIII, horseshoe magnets offered in the commercial market. This magnet package, shown in Fig. 2.13.3, had the advantage
162
SPU'ITER-ION PUMPING
of weighing only 8.9 lbs., where the horseshoe magnet weighed -18 lbs. It also had the advantage of very low fringing magnetic fields. It is noted herein to emphasize that it is possible, when needed, to construct magnetic circuits for pumps having very low fringing fields.
FERRITE DISKS
,
SPLIT
"\
>,
ROLL-PINS
..... ; -~~-
,,
\/1
--5 m m THICK _ _ ~ IRON END-CAPS
Figure ~.13.3. "Confined field" magnet package with ferrite magnets.
The magnetic circuit shell was made of annealed, 1010C steel. Plots of the field, in the magnet gap and proximate to the package, are given in Figures 2.13.4 and 2.13.5, respectively. There is an inherent efficiency associated with a confined field circuit. The normally bugling or fringing fields are wasted in the conventional horseshoe ferrite or alloy magnets. Because of the boundary conditions imposed by the steel circuitry, the fields for the greater part are confined to the interaction region. For reasons cited by Helmer, you will note that field uniformity within the gap is better with the alloy magnet than with the ferrite package, though in speed measurements this proved of no consequence. Helmer suggested that use of steel sheets on the ferrite faces would flatten out the field in the ferrite package.(2 o a ) In special applications, requiring low fringing fields, such confined field packages could be fabricated for much larger sputter-ion pumps. However, because of the comparatively low energy product of the ferrite materials, magnetic discs of these materials (i.e., see Fig. 2.13.3) would have to be too thick for use in smaller appendage pump applications. Because of this, about this same time, I designed a smaller conf'med field package for use with a smaller appendage pump (e.g., --2 Z/see). I used SmCo magnet discs, stabilized to 300* C, in this application. The design, a scale-down of the package shown in Fig. 2.13.3, had SmCo discs measuring 9.5 mm thick x -38 mm q~. Using a shell wall fabricated of 3.2 mm thick steel, TIG welded to end-caps -4.8 mm thick, a field of -0.13 T (1300 Gauss) was obtained in the 50.8 mm gap. This was slightly greater than the maximum 0.125 T observed with the alloy magnets. The design, though a technical "success" at the time, was not fiscally viable
SPU'ITER-ION P U M P I N G
163
as the cost of SmCo magnets was prohibitive. The relative cost of SmCo magnets has not decreased in the last 25 years, though in some special applications, this design is still in use. 0.150
I
I
f
T
FERRITE MAGNET PACKAGE FIELD PROFILE
0.125
T
1
_
~
T
~
T
I
~
ALNICO-VIII HORSESHOE MAGNET
\x F
(13
[" 0 . 1 0 0 I
I
-ANODE
FERRITE DISKS
4~
0.075~
-"~'--- --1
0.050
//,
!
l
~/ I
1
I
I
B
I
I
1
t
L
f
~ ~ ~ - ~
-
:, 0 -12
-10
1
-8
1
-6
_ _ 3
t
-4
-2
0
2
4
I
8
10
12
Centimeters
DISTANCE AXIAL TO PUMP -
Comparison of both fringing and useable magnetic fields field" package and the conventional horseshoe magnets.
Figure 2.13.4. in "confined
15.0
6
I
I
I
I
I
I
I
I
I
I
I
I
(/]
12.5
I
I
I
'F
--
10.0
7.5
J2-----
.t? f
-
x
~~~_~MAGNETIC
Z
~B e
FIELD
ORIENTATION
5.0 Z
z
2.5
0
5
10
DISTANCE FROM DEFINED SURFACES F i g u r e 2.13.5. F r i n g i n g m a g n e t i c
15
20
Centimeters
field of f e r r i t e m a g n e t
package.
It is best to anneal the steel magnetic circuit after forming and welding, as these processes cause stresses which in turn decrease the effective permeability of the steel. Also, if magnetic fields of the order of 20-30 Gauss are measured near the
164
SPUTTER-ION PUMPING
surface of the steel circuit, the steel is near saturation and the return path not optimized. For example, the field at the center of the broad face of the magnet package, shown in Fig. 2.13.5, could be reduced from -15 Gauss at contact, to a value approaching the magnitude of the earth's magnetic field merely by increasing the thickness of the 5 mm end-caps. The maximum magnetic flux density which must be supported by the end-caps is proportional to d/t, where d is the diameter of the ferrite disk, and t is the thickness of the end-caps. Therefore, to further reduce the fringing fields, we have the option of either using a material with higher permeability (e.g., silicone-bearing irons), or increasing the thickness of the end-caps to support the high flux. Pumps with ferrite magnets may be baked at temperatures of -350* C without loss of field in the gap on returning to room temperature. For example, the magnetic circuit shown in Fig. 2.13.3 was baked at -400* C with a total recovery in the field on returning to room temperature.
2.14 More on the Nature of Penning Discharges 2.14.1 Space Charge Distribution in Penning Cells The development of sputter-ion pumps and various forms of Penning gauges prompted considerable study into the properties of the negative space charge stored in the cells, particularly in the LPPDs (i.e., low pressure Penning discharges at P 10 - 4 Torr). Since that time these discharges have been measured for oscillations and RF content, poked and probed with charged beams and Langrnuir probes, optically studied, and otherwise investigated and modeled with attention rarely given a phenomenon. Just a few of these studies will be cited. In that the speed of sputterion pumps is directly proportional to the cell sensitivity, l/P, and I/P is known to be directly proportional to the stored space charge in the cells, knowledge of the characteristics of this discharge became of paramount importance. First, it is important to note that there appear to be numerous modes, or SCD (i.e., space charge distribution) configurations in a simple cell. The SCD will vary with Va, Bz, d, £, a, and P, the anode potential, axial magnetic field, cell diameter, cell length, anode-to-cathode gap and pressure, respectively. Because of this, findings on the SCD characteristics have at times led authors to what appear to be conflicting conclusions, as all of these parameters were not the same from one study to the next. This does not mean that the findings were not scientifically accurate. Most probably they were accurate, and many of these experiments showed a scientific ingenuity and technical understanding which I to this day hold in awe. However, the f'mdings in most cases were not sufficient to suggest a particular unified SCD theory.
Discharge Modes Hooper qualitatively describes the various discharge modes as a function of Va, P and Bz, as shown in Fig. 2.14.1.(3 7 ) This figure was a modification of a mode diagram first proposed by Shuurman in 1966.(2 0 r ) This figure may not be used as a rigorous interpretation of all existing modes. For example, when making noise measurements with magnetic loops, located within a sputter-ion pump operating in the LPPD region, I observed four distinct modes when traversing Hooper's modes I and II. They were evidenced by abrupt changes
SPUTTER-ION PUMPING in noise, with associated changes in sonable model. HIGH
MAGNETIC
MODE
II
165
I/P. However, Fig. 2.14.1 is qualitatively a rea-
TRANSITION MODE III
FIELD
HIGH PRESSURE
MODE IV
Jgi<< ~e MOST OF POTENTIAL DROP AT SHEATH NEAR ANODE. THEREFORE, ?e NONUNIFORM WITH ANODE RADIUS. CENTER FILLED WITH QUASI-NEUTRAL PLASMA, WITH POTENTIAL NEAR THAT OF CATHODES. NO LOW PRESSURE LIMIT. ;> LOW
MAGNETIC
~<<~,
MODE
I
FIELD
O~ e / 0 r 0 o
[...,
~ 0
THEORY UNDERSTOOD; DISCHARGE STABLE. NO LOW PRESS.
~ ~ ~ ~ ~ ~ ~ .~ ~ ~ ~ ~ -m~
NO DISCHARGE
~7
ze
MODE V
PRESSURE Figure 2.14.1. Hooper's d i a g r a m of d i s c h a r g e modes. (37) The distinctive feature of all modes of LPPD is that negative space charge is trapped within the cell, and depresses the cell potential on axis to a value approaching that of the cathodes. This results in radial electric fields within the discharge (e.g., see Fig. 2.2.2). As the pressure is increased to ~ 10-5 Torr, space charge neutralization effects occur, and a neutral column develops at the cell center, growing as the pressure is further increased.( 1 4 3 )
The Stark Effect Knauer and Lutz(2 0 9 ) made use of the Stark effect(2 1 o ) to measure the degree of depression in a Penning cell with the geometry (1.2, 1.6 cm) at an H 2 pressure 2 x 10-4 Torr. This notation "(1.2, 1.6 cm)" is used hereafter to denote a cell aspect ratio (i.e., £/d) of 1.2 and cell diameter of 1.6 cm. The Stark effect is the splitting of the spectral lines of, for example, ionized hydrogen, on recombination, as a consequence of the presence of an electric field. The magnetic field was 0.15 T and anode voltage 3.0 kV. This classic paper has been described by others and the results reproduced in numerous subsequent publications. They introduced H 2 into the cell, and by measuring the splitting of the Hfl line, they deduced the magnitude of the electric field within the cell. From this, they could calculate both the space charge density consistent with the field and the potential within the anode. The SCD
166
SPU'ITER-ION PUMPING
was configured as a hollow sheath. The inside diameter of the sheath was -0.25 the diameter of the anode, peaking at a density of --1.6 x 10 ~ 0 electrons/cma at ---0.44 the diameter of the anode, and gradually falling to zero at the anode wall. Using the implied space charge density and electric fields, their calculations of electron current reaching the anode were within 30% of that predicted using classical cross-field mobility calculations. At the above pressure, a film exposure time of--½ hour was required. Therefore, using this technique for determining space charge within the cell at significantly lower pressures was impractical. At higher pressures and magnetic fields, excessive broadening of line-structure occurred. This effect was attributed to the presence of fluctuating electric fields, thereby deduced as having amplitudes of the order of 4 kWcm. This departure (i.e., increase) in I/P with increasing pressure is the effect noted by Lamont (2.0, 2.5 cm) and shown in Fig. 2.9.1.(2 7 ~ Lamont suggests that the mode observed in this experiment corresponds to mode II of Fig. 2.14.1.
Frequency Shift in RF Cavity In another classic experiment, Lange took an entirely different approach in an attempt to define and quantify the SCD of an LPPD.( 2 1 1 ) He measured the frequency shift in a cell (1.2, 2.5 cm) also serving as an RF cavity, resonant in the TM0 t 0 mode. From this frequency shift he was able to determine the magnitude of electron space charge stored in the cell. At this resonant mode, it was possible for the cathode plates to be electrically isolated from the anodes. They were inserted within the anode cylinder. Conn and Daglish indicated that cell configurations of this type (i.e., cathodes inserted within the anode cylinder) were very unstable at pressures z 10-5 Torr.(2 3 ) This did not appear to be the case over the majority of the pressure range of Lange's experiments (i.e., 10"1 1 to 10-4 Torr, Ar). Lange calculated expected frequency shifts from 18 possible space charge distributions, with variations both in r, the radius of the anode, and axial distance, z. The SPD possibilities entertained at most, would result in errors < x 10 of the absolute space charge density implied by frequency shift measurements. He then arbitrarily normalized all his data to a particular electron space charge distribution, p e (P e uniform in r, and linearly decreasing with z). It is interesting to evaluate Rutherford's data of Fig. 2.2.7, for the different cell sizes and a magnetic field of 0.1 T. If we assume the space charge is uniform and configured as right-angle cones, the bases of which meet at the center of each cell (i.e., one of the distributions evaluated by Lange), we find very close correspondence between I/P findings and cell sizes in Rutherford's data. This also serves as further verification that I/P cx £, the cell length. Lange's findings pointed out three very important features of LPPD: 1) depending on anode voltage,/9 e is nearly constant over 5-6 orders of magnitude; 2) at low pressures, p e is directly proportional to anode voltage; and, 3) there is close correlation between Lange's observed variation in p e as a function of pressure and Lamont's f'mdings( 2 7 ) of I/P vs. pressure. This last point emphasized the correspondence between I/P and p e which was pointed out in Section 2.2 (i.e., ~q <
SPU'Iq'ER-ION PUMPING
167
the higher pressures. Similar frequency-shift measurements were conducted by Agdur and Ternstr0 m with an rf cavity operating in the TM 0 1 o mode.(2 1 2 ) However, these measurements, as others referenced therein, were conducted at much higher pressures. Lange's work represents the only work done of this sort at very low pressures.
Ion Sputtering Patterns Prior to reporting on the Stark Effect measurement, Knauer reported on an experiment where he used the sputtering pattern appearing on the cathode to deduce the characteristics of the space charge distribution, p e(r), in a Penning cell (1.2, 1.9 cm).(a 6 ) He made a study of the sputtering patterns for several different gases in the pressure interval 10-6 to 10-4 Torr. In this case he used a magnetic field of 0.25 T and varied the anode voltage up to 4.0 kV. He noted that there was a mode shift when reducing the magnetic field to < 0.12 T, but at fields > 0.12 T, there was "... little change in behavior (of the discharge)." He observed that a plateau of cathode material was preserved on the cell center line during sputtering. He noted that the diameter of this plateau varied depending on the magnetic field, anode voltage and species of gas. He made calculations of the probable origin of ions initiated in the discharge based on the cathode sputtering patterns. Such sputtering patterns were explained by Knauer by assuming that all of the sputtering ions originated in a sheath close to the inner anode surface. Using Fig. 2.14.2a, where r c is the ion cyclotron radius and rp the plateau radius, he calculated that: rp = (ra2 + rc2 ) ½ - r c.
(2.14.1)
Given that r c = mivi/qiBz, where mi, vi and qi are the mass, velocity and charge of the ion, respectively, and vi = (2qiVa/m i) ½, then: rp = (ra2 + 2miValqiBz 2 ) ½ - (2rniValqiBz 2 ) ½
(2.14.2)
ELECTRON SHEATH
AXIAL
PLATEAU OF UNSPUTTERED MATERIAL ON CATHODES
'
ANODE ~
/
CYLINDER
,/
@
ION
CYCLOTRON
RADIUS
_,og2
. . . .
. . . .
TRAJECTORIES Figure
2.14.2a.
Figure
P ROBE
2.14.2b.
Figure 2.14.2. Knauer's observations of c a t h o d e s p u t t e r i n g and ion probe measurements leading to an anode "sheath"
patterns m o d e l . (36)
Equation (2.14.2) was found to be in excellent agreement with experimental results, totally explaining the plateau effect. Hirsch also reported on this effect two years later.(2 1 z ) However, this seems to be the only other instance it has been observed. This may be due to the fact that slight anode cylinder asymmetries (e.g., 0.05-0.1 mm)
168
SPUTFER-ION PUMPING
would "wash out" such a plateau. Note that the streaks of Si and Ca observed on the cathodes of the BNL LINAC pump elements (i.e., see Section 2.9) were all oriented in a direction suggesting curvature of these ions by the pump's magnetic field. This same year, Knauer applied for a patent on a sputter-ion pump comprising solid anodes.( 2 t 4,2 2 7 ) The claim of this patent was that a sheath of electrons would be stored on the outer surface of the solid cylin-ders. This configuration was not subsequently exploited, as I suspect that it was probably more applicable at higher pressures and magnetic fields.
Ion Currents and Retarding Potential Probes Using two retarding potential probes, one on axis with the cell and the second radially off center and hidden behind the cathode plate (i.e., see Fig. 2.14.2b), Knauer measured the energy distribution of ions created in the discharge.( s 6 ) From these results and subsequent calculations, he determined that even a 5% potential depression along the axis of the cell would cause the axial velocity component of even the most energetic ions (i.e., in terms of the radial component) to miss the radial probe. From this, Knauer concluded that essentially all of the potential depression occurs near the anode. He then cut an 8 mm $ hole in the center of the cathode plate opposite that having the probes. An 80% reduction of axial ion probe current was then observed, while current to the radial probe remained the same. From this he concluded that most of the ion current initially measured with the axial probe was due to particle reflections from the opposite cathode. Earlier experiments of a similar nature, conducted by Helmer and Jepsen,(3 5 ) led them to different conclusions. A Faraday cup was used to measure the axial ion current from a cell (1.7, 1.2 cm). They determined the following: 1) ion energy, measured at the Faraday cup, is almost mono-energetic, having an energy spread of only 10% of the average; 2) the beam is highly collimated; 3) there is a direct relationship between ion energy and anode voltage, becoming linear above a certain value; 4) there is a decrease in average ion energy with increasing pressure; and, 5) ion current to the collector represented x0.2 of the total current to the respective cathode, whereas the aperture represented only x0.063 of the total cathode area. They concluded from these measurements that the mean ion energy was a measure of the potential depression at the center of the cathode, but that this potential depression was not as significant as that suggested by Knauer's experiments. Note, however, that essentially the same results were obtained by Knauer, and Helmer and Jepsen when their experiments overlapped (i.e., the same B z and pressure). Other investigators have made similar use of ion current measurements to deduce something about SCD in an LPPD. For example, Young used a Faraday cup to measure total ion current passing through a 3 mm $ hole on the axis of a cell (1.0, 1.2 cm), over the pressure range of -10" 1 0 to 10-4 Torr.(S s ) He found that at 10-1 o Torr, 60% of the ion current was collected on the central 6% of the cathodes and at --10 -4 Torr this number reduced to 20% of the total discharge current. That same year, Smirnitskaya and Nguyen-Khiu-Tee made observations similar to Young.(6 0 ) Results of experiments published by Rudnitskii a year earlier showed that all the above experimental results were probably consistent and correct.( s 7 ) He graphically illustrated how cathode current densities shifted between two mode patterns with changes in Va and B z. There also appeared to be some mode mixing. With low magnetic fields and voltages, such as used by Young (0.097 T, 2.0 kV),
SPUTTER-ION PUMPING
169
Rudnitskii's data suggests that most of the current would be confmed to the central part of the cathodes. This also proves to be the case when comparing Rudnitskii's and Smirnitskaya's data for the case of low B z (0.13 T) and high Va (7.0 kV). With higher magnetic fields and lower voltages, as used by Knauer (0.15 - 0.25 T, 3.0 kV), Rudnitskii's data is in very close agreement with Knauer's results. That is, a cylindrical sheath exists as described by Knauer, in the above section discussing use of the Stark Effect. Hirsch used a radial probe, running the full length of the anode cylinder, to intercept azimuthal currents in a Penning cell.( 2 1 3 ) He measured ion current as a function of probe depth in the cell, for various pressures. This and other similar tests give questionable results, as the discharge often will readjust itself to the new boundary conditions created by the presence of the probe, and give misleading results.
Electron Beam Probes Dow used a transverse ribbon, electron beam as a Penning discharge probe.(3 1 ) He hoped to use the measure of azimuthal drift of the beam, on transversing the discharge, to deduce the average radial field therein, and therefore, the SCD. He was not able to make self-consistent, quantitative determinations of these fields. However, he did make qualitative determinations of the radial fields in a small cell (2, 1.2 cm) as a function of pressure, Va and B z. The ribbon beam, on passing through the cell, appeared on a phosphorescent screen. Though forming a straight line when launched into one end of the cell, it was distorted by the electric fields in the cell on passing through the discharge. The squiggles observed indicated distinct variations in discharge modes of the SCD, as noted above. Others have reported use of electron beams to probe Penning cells, at low pressures, including Knauer (without a discharge present),(3 6 ) and Hirsch.(2 1 s )
Jepsen's Smooth-bore Magnetron Model As noted earlier, Jepsen recognized that the space charge potential depression on axis of an LPPD resulted in a crossed-field type configuration similar to that of a magnetron. Though others had commented on the similarity between these cells and magnetrons, Jepsen expanded on this point in modeling the discharge in a 1961 article.(a o ) This, and the subsequent work of Schuurman,~ ~ 1 8 ) served as the basis from which subsequent semi-empirical models of the discharge were formulated, including both normal and inverted (magnetron) discharges.( 2 1 7 ) The discharge mode modeled by Jepsen is represented as mode II in Fig. 2.13.1, predicting behavior of the discharges at and above the transition from mode I to mode II (i.e., decreasing I/P with increasing Bz) , where Schuurman's model dealt with mode I, and as pointed out by Redhead, did not agree with experimental findings in mode 11.(5) Jepsen's model was a smooth-bore, coaxial magnetron, with a filamentary cathode radius. It showed a linear relationship between I/P and £, anode length. Values of I/P predicted by the model were a factor of 2-5 higher than values empirically observed with these discharges.( 1 0 8 ) In the years that followed, many excellent articles appeared dealing with optimizing the speed ~I/P) of various pump geometries. This includes works by Malev and Traktenberg,1,l 8 8 ) Hartwig and Kouptsidis,( 2 1 8 ) and the more theoretical work of Wutz.( ~ 1 9 ) The working models which were subsequently developed
170
SPU'Iq'ER-ION PUMPING
became semi-empirical modifications of Jepsen's and Schuurman's work.
2.14.2 More on Sputtering Patterns on Pump Cathodes When disassembling a used, multi-celled, diode sputter-ion pump, I have invariably observed distinct star-like, sputter-errosion patterns on the cathode plates. These observations have been made on hundreds of sputter-ion pump cathodes. The stars will be observed regardless of the cathode material, or gases pumped. If the pump has been operated exclusively at very low pressures, the width-to-depth ratio of the sputtered stars will be small; perhaps 1:5. This ratio increases with higher operating pressures, and may be of the order of 20:1 in an extensively used pump. Nevertheless, the stars are always present. The stars will take shapes similar to that depicted in Fig. 2.14.3. For a cell size of the order of 20 mm~ and B z of ~ 0.18 T - the maximum field in any commercial pump which I have studied - the stars will have four points or spokes. If the cells are nested in a rectangular array (as opposed to close packed), additional stars will appear in the parasitic cell regions created between the cells. Stars in these pseudo-cells will usually have eight spokes. A-
ANODE
ICYLINDERS
FOUR-POINTsPuTTERED_~__~ +~+me ÷~\/*/+: ÷~÷:Q\
"STARS" EROSION
\ ~+÷
~~+-~;+/~,,,'v 4,~___~ ~
n = 0
~"--~
towARDS n= F i g u r e 2.14.3. S p u t t e r i n g p a t t e r n s i n v a r i a b l y o b s e r v e d on c a t h o d e s ,
2
n = 1
I 1 n = 3
F i g u r e 2.14.4. N o d e s of o s c i l l a t i o n in a s m o o t h - b o r e magnetron.(22°)
Each spoke of a star is always curved, and the curvature of the spokes of all the stars will be oriented in the same clockwise or counter-clockwise direction. Star patterns of facing cathode plates are mirror images. The orientation of star spokes in closepacked cell arrays differs from that observed in rectangular arrays, such as shown in Fig. 2.14.3. Lastly, the spokes of adjacent stars are always oriented in a preferred direction, suggesting a mutual coupling mechanism exists between cells. I have never seen an exception to the above observations in a cathode of a multi-celled, diode pump element. The stars would not be discernible in triode pumps. Though these erosion star patterns are prevalent, sputtering still occurs over the entire face of the cathode, and is greatest on center-line with the cell. However, existence of the stars suggests some nonuniform SCD exist in all multi-celled sputterion pumps. A half-century ago, Brillouin described various possible modes of oscillation in the SCD of a cylindrical, crossed-field diode (i.e., a magnetron).(2 2 o ) These are reproduced in Fig. 2.14.4. The n = 0 mode, shown in the upper left corner of the figure, corresponds to the SCD uniformly expanding and contracting radially within the cell (magnetron). The n = 1 mode is a dipole-type of oscillation. The n = 2 mode corresponds to an oscillation of the space charge about two fixed axes. The SCD at one extreme takes the shape of an ellipse of distorted space charge
SPU'VrER-ION PUMPING
171
along the y-axis, where at the other extreme of the oscillation it is distorted into an ellipse having a major axis along the x-axis. I suggest that the above noted stars are the result of n = 2 space charge oscillations in the sputter-ion pumps. The stars in the parasitic cells, having eight spokes, correspond to the n = 4 mode of oscillations. Helmer and Jepsen first reported on the existence of these stars, noting eight-point stars in cells having square cross section, and four-pointed stars in cells with slightly elliptical, but near-round shape.( s s ) A photograph of the discharge in a - 8 mm square cell, shown in that paper, evidenced eight distinct stars (i.e., n = 4) at a pressure o f - 1 0 -4 Torr. They attributed these stars to space charge oscillations. Short of the modes suggested by Brillouin, I know of no one to date having modeled the dynamics of the SCD which would create these stars. The magnetron analogy persists.
Noise in Sputter-Ion Pumps Conn and Daglish first reported measuring low frequency oscillations in these discharges.(2 2 ) Using a Lecher wire system, they observed oscillations in LPPDs in the frequency range of 10 - 26 MHz. Helmer and Jepsen also reported on noise and oscillations in LPPDs.(S s ) They reported noise in these discharges stemming from plasma oscillations, occurring at -1.0 GHz; electron cyclotron frequency oscillations, fc, occurring at -3.3 GHz (depending on the magnetic field), and, low frequency, rotational oscillations at -150 MHz. Helmer and Jepsen used a split-ring magnetron configuration to measure some of the low frequency oscillations. Redhead has done the most extensive work on these LPPD instabilities and oscillation at very low pressures.(S 4 ) A review paper by Hooper cites much of the early work done in this area.(S 7 ) In a recent review paper, Redhead cites additional important work done in this area, to the present.(5 ) He suggests that the decrease in I/P at very low pressures is probably caused by the loss of energetic electrons from the discharge as a consequence of inherent instabilities in the discharge, including diocotron oscillations. Diocotron oscillations occur at higher magnetic fields, and are proportional to the amount of space charge stored in the discharge. Reporting on work by Malmberg and Driscoll,(2 2 1 ) he noted how at a pressure less than ~ 10-8 Torr, the electron containment time in a discharge became constant, suggesting an anomalous electron transport phenomenon. By improving uniformities in the electric fields within cells, and the magnetic field, containment times could be increased. Oscillations in SCD, suggested by the sputtering patterns in cathodes, probably stem from space charge instabilities caused by slight asymmetries in each anode cell. Where on the one hand the ordered nature of the sputtered patterns from cell to cell may be due to electromagnetic coupling of the discharges between each of the cells, on the other hand, it may be due to cell asymmetries introduced by fabrication techniques used in building the anode structure (e.g., spot welding). In work at SLAC, using very sensitive superheterodyne receivers with bandwidths o f - 2 5 MHz and MDSs of -90 dbm, I made measurements of noise emanating from sputter-ion pumps of various configurations in the frequency range of 1.9 8.0 GHz. The pumps were coupled to, in one case, by inserting the pump element into an S-band waveguide. In the other cases, magnetic loops were inserted into the pump housing. Of course, magnetic loops often have inherent resonances. Therefore, data taken with one loop configuration could never be considered conclusive.
172
SPU'VFER-ION PUMPING
The following observations were made: 1) Noise was manifest in several forms, including broad-band Gaussian-type noise, increasing with pressure, Va and B z. 2) There was a pressure dependent, low frequency amplitude modulation of the Gaussian-type noise. 3) There were impulsive-type, narrow-band bursts of noise within the Gaussian noise. 4) Depending on the pressure, Va and B z, there was at times a striking similarity in the plots of noise amplitude as a function of frequency and that reported by Glass and his colleagues for magnetrons near cut-off.(2 2 2 ) 5) Noise thresholds existed at fc as a function of voltage, where the amplitude of the noise would increase three orders in magnitude with slight changes in Va. 6) No noise was evident at frequencies > fc, to a level of -90 dbm.
2.14.3 Striking Characteristics Penning discharges do not instantaneously ignite with the application of high voltage (e.g., see (2 2 a ,2 s 4 )). The delay in ignition of the discharge depends on pressure, cell geometry, applied voltage and magnetic field. This delay is sometimes called the strila'ng characteristics of the cell. There have been numerous articles published on this topic, and the theory is fairly well understood. Redhead was the first to publish a comprehensive theoretical treatment of the striking characteristics of an inverted magnetron Penning cell vs. pressure,(2 2 4 ) and he and Hobson the I + / P characteristics of such cells at extremely low pressures,(2 1 7 ) and Lassiter the striking characteristics specifically for He at very low pressures.(2 2 6 ) The striking characteristics of an ensemble of Penning cells, of different configurations, was reported on by Laurent.(1 4 0 ) The delay in the striking of the discharge, particularly at very low pressures, was reported on by de Chernatony and Craig.(2 2 z ). This delay-effect prompted the use of hot filament, electron sources in some single cell devices. For example, Westendorp's patent on single-ceU pumps included configurations with hot filaments,(1 9 ) as did a multi-cell pump patented by Mark and Henderson several years later.(2 2 s ) The General Electric "trigger discharge" cold cathode gauge makes use of such a hot electron source to initiate the discharge at low pressures.l,2 2 9,2 a 0 ) The use of hot electron sources, radioactive beta emitters and sharp points (field emitters) to initiate a Townsend (i.e., starting) discharge was suggested by Conn and Daglish in 1953.(2 2 ) Komiya used a 6 a Ni beta source to initiate a discharge in a multi-cell pump at a pressure of 3.5 x 10" z 1 Torr.(1 2 2 ) Short of using a radioactive seed in the cell,(2 5 4 ) there is no quick fix for this ignition problem, other than increasing both B z and V a to practical maximums. The problem plagues microwave tube engineers who, after baking both tube and appendage pump (without magnet) thereafter, when high voltage testing the tubes, have problems initiating discharges in the pumps. The appendage pumps used on small microwave tubes such as TWTs are usually the mini-pumps, and have very small cells, further exacerbating starting problems. I have often wondered if it might be possible to integrate a Tesla coil in a pump power supply, and by intermittent use of this coil, facilitate greater ease in starting of these mini-pumps. In the larger, multicelled pumps, once discharge starts in one cell, it quickly propagates throughout the rest of the cells. I observed this effect in noise measurements of pumps at very low pressures, noting quantum increases in the noise level associated with the sequential ignition of cells.
SPUT/'ER-ION PUMPING
173
2.14.4 Use of Very High Magnetic Fields The use of very high magnetic fields to support LPPDs is often a requirement in DIP (distributed ion pump) applications.(1 a s ,1 a 0 ) These pumps often exploit use of the fringing magnetic field of magnets used to bend particle beams in storage rings and accelerators (e.g., see Section 2.10). This was first proposed by Jepsen in 1959.(s 4 ,s a 1 ) He did much of the early work relating to the use of very high magnetic fields (~e., up. to --1.0 T) in sputter-ion pumps which here-to-fore has not been referenced.(2 a 2 ) He showed, in a 1959 patent application, that for a given cell diameter, dc, there was a critical magnetic field below which the discharge extinguished. For a fixed voltage, this field was def'med by: dc
= kl /Bz,
(2.13.3)
where k l is a constant. This relationship has its origin from the cutoff parabola of magnetrons.(1 s s ) Based on empirical measurements, he noted that for cells of the same length, £, and £/d > 1, to a first approximation the speed of the cell, S O is given by: SO
= ks (d -
dc).
(2.13.4)
The constant ks is a function of Va, the cathode material and the gas being pumped. The number of cells which can be arrayed per unit area is given by: na
= ks / d s ,
(2.13.5)
where the constant k3 depends on the packing geometry of the array. For close packed and rectangular arrays, k3 has values of (4/3) ~ and 1.0 respectwely. The 1
,
product of (2.13.4) and (2.13.5) yields the speed of the array assembly. On differentiating this product with respect to d, and setting this value equal to zero, we find the optimum cell diameter for the maximum speed at a given magnetic field; this is found to be when d = 2d c. Jepsen provided empirical data in the above reference for fields ranging from 0.04 to 1.0 T. He suggested an optimum value of k 1 = dcB z --5.6 x 10-2 T-cm. Jepsen's patent application was made four years prior to Rutherford's paper, showing that the speed of pumps mysteriously decreased as a function of pressure, for the same magnetic fields (e.g., see Fig. 2.2.7).(s 8 ) Nevertheless, this early work by Jepsen is still applicable and was used by others over a decade later in the design of DIPs using high magnetic fields.(1 8 8, s a 4 ) However, at low pressures, the onset of moding or oscillations at high magnetic fields may cause pumping instabilities. As noted, these instabilities have not been adequately modeled. Therefore, empirical work is required when tailoring DIPs for use in high magnetic fields.
2.15 Other Considerations Maintenance and Trouble Shooting These pumps are actually very simple devices to trouble shoot. When experiencing difficulties, we must first make sure that it is truly a pump problem, rather than a system problem (see Section 1.17). Assuming it is a pump problem, there are only a
174
SPU'ITER-ION PUMPING
few items which need be checked. If the pump current is the only measure of system pressure, we must determine if we truly have a pressure problem, or if the indicated pressure is high as a consequence of leakage current. Note that a sputter-ion pump at air behaves the same as a sputter-ion pump at very low pressures. This is often the cause of problems in a system. Short of the existence of some other form of gauging, there is no way to be completely sure that your pump is not at air. In the case of triode pumps, field emission leakage current may be eliminated by hi-potting, while under vacuum. Prior to this, we might want to determine if there is leakage current in the anode insulators or high voltage feedthrough. This may be checked by: 1) turning off the power supply; 2) disconnecting or locking out the primary power source; 3) switching the meter on the high voltage supply to the voltage setting, to make sure the voltage has drained to zero; 4) removing the high voltage cable and measuring the resistance across the pump electrodes. Note that the resistance should be measured using both polarities, as the equivalent of a thinf'dm diode may exist on an insulator, and yield a high impedance in one direction and a much lower impedance with reversed polarity. Because of thin film effects, insulator leakage may also disappear after venting the pump to atmosphere. If a low resistance is noted, this may stem from leakage across an anode insulator or across the high voltage feedthrough. In the case of the latter, if it is due to internal sputtering, there is no simple remedy. A replacement must be purchased from the OEM (original equipment manufacturer). Because of high electric fields, in the presence of high humidities and moderate temperatures, an accelerated corrosion of the exterior parts of the high voltage feedthrough may take place. The exterior fields may be reduced by the careful design of the high voltage connector. When leakage currents exist as the result of sputtered material on the anode insulators, the pump element may be disassembled, and the insulators completely refurbished by sandblasting (preferably with A~ 2 0 a grit), washing in an industrial dishwasher, and then firing them in air at 1000" C. Are the magnets weak or improperly installed? Irreversible changes in the magnetic field will occur if the magnets are exposed to either very high, or, in the case of ferrite magnets, very low temperatures. If the magnets are improperly installed so that the fields are bucking, though a discharge may exist in the cells, it will be comparatively weak. If the magnets were cracked because of some prior thermal or mechanical shock, this too will significantly reduce the field. Lastly, you should make sure that some extraneous magnetic member is not shunting the field in the gap. If high voltage is being applied to the elements (i.e., there is not an internal or external open circuit) and the magnetic field is adequate, then the pump has the same speed it had when new. The problem then is that the effective speed has been diminished because of sources of gas. We saw in Section 1.15, that the speed of a pump will be diminished at low pressures depending on the base pressure of the pump. This is not to be confused with I/P effects. If there is a source of gas leaking into the pump, or originating from within the pump, though the speed of the pump may be the same as when new, the effective speed at low pressures is diminished as given by Remember, certain types of sputter-ion pumps effectively pump He.(2 4 T) Therefore, to get the highest sensitivity when leak checking the pump, the pump should first be turned off. Also, in Section 1.18, we made calculations showing how an atmosphere of He will break down under high voltage much more readily than an atmosphere of air. Because of this, use caution when leak checking around a
Q.15.1).
SPUTTER-ION PUMPING
175
pump which is energized. If there is no leak into the pump, then the source of gas must either be due to contamination inadvertently (or deliberately) introduced, or outgassing from saturated cathodes. Outgassing from saturated cathodes is usually the last probable cause for poor performance of a pump, as with the exception of H 2 bearing gases, most of the air gases form very stable compounds with the cathode materials. End of life in this case happens when the cathodes are sputtered through. For reasons cited in Section 2.5, there is a rate limitation in the pumping of H 2-bearing gases. In this case, the cathodes of diode pumps can be refurbished by sandblasting them with A~ 2 O3 grit, washing them in an industrial dishwasher and then firing them at --800* C, for a few hours, in vacuum. This same process is also used to refurbish the anodes. The element insulators may also be refurbished at the same time, using the previously noted procedures. Lastly, cathodes of diode pumps which have been eroded through are easily replaced with fabricated parts. Alternate methods of cleaning stn. stl. parts, such as the pump body, were reported on by Sasaki.(2 5 e )
2.16 A Summary of Advantages and Disadvantages Operating Efficiency: At low pressures, the operating efficiency of these pumps, in terms of Watts/Z/sec., exceeds all other pumps. However, we have seen that at very high pressures this is not the case. These pumps have very poor throughput handling capability. Capacity: In that these are capture pumps, they have a finite capacity for all gases. End of life may be manifest as a gradual decrease in the base pressure of the pump, and resultant decrease in speed at the lower pressures for a given gas. A more dramatic example of end of life might be when, in the case of diode pumps, holes are sputtered through the walls of the pump body. This does not have catastrophic results on the system, as the drilled holes are usually initially very small and are evidenced as a minor system leak. When either the throughput or total sorptive capacity of the pumps is exceeded, pumps will become difficult to start, or evidence thermal run-away problems. These pumps can also be rendered useless when pumping forms of hydrocarbons.(2 3 s, 2 3 6 ) For reasons noted, their capacity for pumping the noble gases is limited, particularly in the case of conventional diode pumps. Source o f System Contaminants: In 1%3, it was shown by Litchman(2 a ~ ) Riviere,(t ) and Reich(2 3 8 ) that these pumps synthesize methane. Carbon, which is always present in the Ti cathodes, combines with H2 therein to create this gas. This effect was also pointed out by Jepsen in 1963, when he noted it was prevalent when pumping H2 and H2 O.(2 3 9 ) This can be troublesome when using these pumps at very low pressures.(4 6 ) However, components in the vacuum system, other than the sputter-ion pumps, often prove to be a far greater source of contamination.(2 4 o )
Source of Charged Particles & Neutrals: These pumps are also sources of charged particles which can cause damage to instrumentation and cause system noise problems. In fact, when starting diode sputter-ion pumps, the glow discharge originating in the pump can spread throughout the entire system to which it is appended. Because of this, screens are used at the pump inlet flange to shield the system from
176
SPU'Iq'ER-ION PUMPING
these charged particles.(2 4 1,2 4 2 ) The spread of the discharge is much more prevalent in diode pumps, as with triode pumps the walls of the pump are at the anode potential. Note, however, that use of these screens will not shield the system from neutrals and sputtered material generated within the pump.( 2 4 3 ) These neutrals can also cause system contamination problems (e.g., the coating of optical surfaces).
Pump Stam'ng. The starting of these pumps can be more difficult with pump age, or if they are contaminated by hydrocarbons or water vapor from the roughing pumps or some other source. Usually, the older the pump (i.e., the greater its use), the harder it is to start. This is primarily because of thermal desorption effects resulting from ion bombardment of the cathodes. Bell and his colleagues showed that pumps in goocl condition would very effectively pump down systems with large water vapor loads.(2 4 4 ) With diode pumps, on starting, the walls of the pump are bombarded with low energy ions which desorb gas. This compounds the difficulty in starting. This is not the case with triode pumps.(1 s 6,2 3 s ) On the other hand, it has been shown where triode pumps can present starting problems as the result of delayed, thermal run-away. Because of wall and pump element gas desorption considerations, the use of an isolation valve between the pump and system has obvious benefits. Use of pump isolation valves is fiscally impractical in very large accelerators, having numerous pumps. The SLAC two-mile accelerator made use of isolation valves at each of the pumps distributed along the accelerator (e.g., see Fig. 2.4.13). The valves eventually were no longer used. However, dry N 2 was always used when venting the sectors and they were always sorption roughed.( ~ 4 s ) These precautions effectively minimized both the hydrocarbon contamination and water vapor levels in the system.(1 ~ 0,1 7 4 ) Triode pumps must be hi-potted after starting to eliminate field emission problems (i.e., assuming current at low pressures is used to indicate pressure). Thermal run-away problems can also occur when starting all sputter-ion pumps. Because of this, when starting these pumps the high voltage should be cycled on and off according to some recipe which depends on the characteristics of the power supply, size and surface area of the pump, the surface area and aspect ratio of the system, and the roughing provisions. The fact that the roughing pump might be capable of achieving low pressures, even < 10"s Torr, is no assurance that starting problems will be avoided. For the above reasons, the pump starting recipe must be empirically estabfished for the total system. Roughing pumps of various forms can be the source of hydrocarbon contamination. Therefore, sorption pumps are often used for roughing in instances when system hydrocarbon contamination can not be tolerated.
Magnetic Fields and High Voltages" The possible existence of high fringing magnetic fields can cause problems, as noted in Section 2.13. Pump magnets and the requirement of the use of high voltages present operation problems and represent potential safety hazards. Therefore, when designing a vacuum system in which sputter-ion pumps are to be used, the potential hazards of both the pump magnets and high voltage must be given as much consideration as the vacuum performance. Also, in UHV applications, these pumps are often baked, when operating, to temperatures as high as 350* C. Care must be taken that the appropriate type of high voltage cables are used in such applications.
SPUTTER-ION PUMPING
177
Advantages of Sputter-Ion Pumps: They afford clean, dry pumping. Excluding methane, these pumps neither synthesize nor become the origin of hydrocarbon contaminants. They are bakeable, immune to very high radiation fields, can be operated in any orientation and are not a source of system vibration. When used in conjunction with TSP pumps, and baked at temperatures of--250* C, pressures < 10"1 1 Torr may be achieved. This combination of TSP and sputter-ion pumping is the least expensive and most reliable means of achieving very low pressures. Sputter-ion pumps can be constructed to suit the application, including built-in, distributed pumps, and the appendage and large component pumps. At low pressures they are by far the most efficient method of pumping. When designing, building, or purchasing a sputter-ion pump, one key to success is in first defining the pressure at which you need a given speed.
APPENDIX 2A ELECTROSTATIC GE'I~ER-ION PUMPS 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13)
Herb, R.G. and Davis, R.H., "Evapor-Ion Pump", Phys. Rev. 89, 897 (1953). Divatia, A.S. and Davis, R.H., "Construction and Performance of Evapor-Ion Pumps", Proc. 1st Nat. AVS Symp., 1954 (W.M. Welch Manufacturing Company, 1955), p. 40. Alexeff, I. and Peterson, E.C., "Evapor-Ion Pump Performance with Noble Gases", Proc. 2nd Nat. AVS Symp., 1955 (Committee on Vacuum Techniques, Inc., 1956), p. 87. Swartz, J.C., "Evapor-Ion Pump Characteristics", Proc. 2nd Nat. AVS Symp., 1955 (Committee on Vacuum Techniques, Inc., 1956), p. 83. Reich, G. and N6 ller, H.G., "Production of Very Low Pressures with GetterIon Pumps", Proc. 4th Nat. AVS Symp., 1957 (Pergamon, Inc., New York, 1958) p. 97. Herb, R.G., "Evapor-Ion Pump Development at the University of Wisconsin", Proc. 1st Int. Vac. Cong., June, 1958, Vacuum 9, 97 (1959). Holland, L., "Theory and Design of Getter-Ion Pumps", J. Sci. Instrum. 36(3), 105 (1959). Klopfer, A., Ermrich, W., "Properties of a Small Titanium-ion Pump", Proc. 6th Nat. AVS Symp., 1959 (Pergamon Press, Inc., New York, 1960) p. 297. Holland, L., and Laurenson, L., "Pumping Characteristics of a Titanium Droplet Getter-Ion Pump", Brit. J. Appl. Phys. 2(9), 401 (1960). Kumagai, H., et al., "Characteristics of Titanium Evapor-Ion Pump", J. Vac. Sci. Technol. 1, 433 (1960). Warnecke, M. and Moulou, P.C., "On a Miniature Titanium Pump", Le Vide 85, 41 (1960). Adam, H. and Bachler, W., "Operational Procedures and Experiences with a High-Speed Ion Getter Pump", Proc. 2nd Int. Vac. Cong., 1961 (Pergamon Press, Inc., New York, 1962) p. 347. Gould, C.L. and Dryden, R.A., "One Year of Operating Experience with Getter-Ion Pumps", Proc. 2nd Int. Vac. Cong., 1961 (Pergamon Press, Inc., New York, 1962) p. 369.
178
SPUITER-ION PUMPING APPENDIX 2A, CONTINUED ELECTROSTATIC GETTER-ION PUMPS
14) Gould, C.L. and Mandel, P., "A Sublimation Pump", Proc. 9th Nat. AVS Syrup., 1962 (The Macmilian Company, New York, 1963) p. 360. 15) Hooverman, R.H., "Charged Particle Orbits in a Logarithmic Potential", J. Appl. Phys. 34(12), 3505 (1963). 16) Herb, R.G., Pauly, T., Welton, R.D., Fisher, K.J., "Sublimation and Ion Pumphag in Getter-Ion Pumps", Rev. Sci. Inst. 35(5), 573 (1964). 17) Maliakal, J.C., Limon, PA., Arden, E.E., Herb, R.G., "Orbitron Pump of 30-cm Diameter ~, J. Vac. Sci. Technol. 1(2), 54 (1964). 18) Mourad, W.G., Pauly, T., Herb, R.G., "Orbitron Vacuum Gauge", Rcv. Sci. Imtrum. 35(6), 661 (1964). 19) Nazarov, A.S., Ivanovskii, G.F., Men'shikov, M.I., "Getter-lon Pump with Filamentary Titanium and Chromium Evaporators", Instr. & Exper. Tech., No. 6, 934 (1964). 20) Douglas, R.A., Zabritski, J., Herb, R.G., "Orbitron Vacuum Pump", Rev. Sci. Instrum. 36(1), 1 (1965). 21) Andrew, D., "The Performance Assessment of Sputter Ion Pumps", Vacuum 16(12), 653 (1966). 22) Gretz, R.D., "A High Vacuum Titanium Getter Pump", Vacuum 16(10), 537 (1966). 23) Holland, L., Laurenson, U, Fulker, M.I., "The Vacuum Performance of a Combined Radial Electric Field Pump and Penning Pump", Vacuum 15(12), 663 (1966). 24) Bills, D.G., "Electrostatic Getter-Ion Pump Design", J. Vac. Sci. Technol. 4(4), 149 (1967). 25) Denison, D.R., "/'he Effect of Gas Quantity Pumped on the Speed of GetterIon and Sputter-Ion Pumps", Proc. 14th Nat. AVS Symp., 1967 (Herbrick and Held Printing Company, Pittsburgh, 1968), p. 87. 26) Denison, D.R., "Performance of a New Electrostatic Getter-Ion Pump", J. Vac. Sci. Technol. 4(4), 156 (1%7). 27) Maliakal, J.C., "Residual Gas Analysis in an Orbitron Pump System", Proc. 14th Nat. AVS Symp., 1967 (Herbrick and Held Printing Company, Pittsburgh, 1968) p. 89. 28) Della Porta, P. and Ferrario, B., "Magnetless Gauge Appendage Pump Utilizhag Non-Evaporable Getter Material", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969) p. 369. 29) Denison, D.R., "Getter Properties of Yttrium as Used in a Getter-Ion Pump", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969) p. 377. 30) Maliakal, J.C., "Residual-Gas Analysis in an Orb-Ion Pump System", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969) p. 361. 31) Naik, P.K. and Herb, R.G., "Glass Orbitron Pump of 5-cm Diameter", J. Vac. Sci. Technol. _5(2),42 (1968). 32) Kuznetsov, M.V. and Ivanovsky, G.F., "New Developments in Getter-Ion Pumps in the U.S.S.R.", J. Vac. Sci. Technol. 6(1), 34 (1969).
SPU'VFER-ION PUMPING
179
APPENDIX 2A, CONTINUED ELECTROSTATIC GETI'ER-ION PUMPS 33) Bills, D.G., "Electrostatic Getter-Ion Pump Performance", J. Vac. Sci. Technol. 10, 65 (1973). ' 34) Singleton, J.H., "The Performance Characteristics of Modern Vacuum Pumps", J. Play. E6, 685 (1973). 35) Naik, P.K. and Verma, S.L., "Performance of the Modified Orbitron Pump", J. Vac. Sci. Technol. 14, 734 (1977). 36) Feidt, M.L. and Petit, B., "Measurement of Ion Energy Distribution in an Orbitron Device", J. Vac. Sci. Technol. 18(3), 987 (1981). 37) Feidt, M.L. and Paulmier, D.F., "A Model for Optimum Ionizing Characteristics of an Orbitron Device", Vacuum 32(8), 491 (1982).
Problem Set 1) Assume that the speed of a sputter-ion pump is 100 Z/sec at a pressure of 10"s Torr, the magnetic field is 0.10 T, and the anode cells have a diameter of 12.7 mm. What would be the speed of the pump at a pressure of 10-a Torr and for the magnetic field settings of 0.1, 0.15 and 0.2 T? 2) Assuming the sputter-yield of 1.05 keV ions, normal to a target, is 1.94 for A~ and 1.13 for Ti, what would the sputter-yield be for these two materials, at this same argon ion energy, if the ions were incident at an angle of 70" ? 3) Assume that the sensitivity of a cell of "zero" length is proportional to the diameter of the cell. Assume that the diameters of anode rings of Fig. 2.2.1 are each --16 mm ~, and the I/P of the cell is 1.8 A/Torr at a pressure of 10"s Torr. Neglecting, ~, the length of the cells in Fig. 2.2.7, what would be the I/P of each of the anode arrays given, for a magnetic setting of 0.1 T? 4) Assume that the I/P of a cell is directly proportional to the space charge stored in the cell. Assume that the space charge is of uniform density, and the shape of the discharge forms two right-angle cones, the bases of which join at the mid-point of each cell. Based on these assumptions, what would be the total I/P for the three configurations given in Fig. 2.2.7, for a magnetic field of 0.1 T and at a pressure of 10"s Torr? 5) Assume that each cell of the pump shown in Fig. 2.4.9 has an I/P of 5 A/Torr at pressures ~ 10-6 Torr, and that the pump is connected to a robust power supply capable of maintaining a voltage of 5.0 kV even at very high pressures. What would be the total power drawn by this pump at pressures of 10-6 , 10"s and 10-4 Torr? What would the average power density dissipated in the cathodes be at these three pressures? 6) When operating at high pressures, why don't the anodes get as hot as the cathodes? 7) Assume that there are two pumps attached to a vacuum system, with all metal valves. One pump has ten cells, each with an I/P of 10 A/Torr and the second pump has 250 cells, each with an I/P of 10 A/Torr. Assume that gas is let into the system through a controlled leak. Prove that for the same leak rate, the current drawn by each pump, when individually pumping on the same leak, would be the same.
180
SPUTTER-ION PUMPING
Problem Set (continued) 8) Assume that each of the cells in the pump elements shown in Fig. 2.4.10 have the same I/P as in problem 5. What would be the power drawn by the pump for the same voltage and pressures of problem 5? 9) You have built a very large accelerator to which 100 pumps, comparable in size to that shown in Fig. 2.4.10, are appended. Assume that each pump cell has an I/P which decreases according to the data given in Fig. 2.2.7, for the 12.7 mm ~ cells and a field of 0.15 T. What is the total power drawn by the pumps at a pressure of 10-° Torr? 10) Calculate the results of Fig. 2.4.14, using the given assumptions. 11) You suspect your pump has a field emission problem, and with a 6.0 kV setting, the current drawn by the pump is 100/.tA. How can you differentiate between current due to field emission, leakage currents and real I/P effects? 12) What is the maximum possible scattering angle of Kr ions off of Ag and Ti cathodes? 13) Assume that the density of Till2 is -86.5% that of pure Ti. Prove that the dimensions of a cube of Ti pure would have to increase by -6.4% to accommodate sufficient H 2 to form Till 2. 14) Using (2.5.2) and (2.5.6), show that for large times, t, the total amount of gas which has diffused through a semi-infinite slab from time 0 - t is given by f Odt = (DC/d)(t - d 2/6D), and is called the "breakthrough" equation. 15) What is the ratio of the speeds of pumps when the ratio w :xo, of Fig. 2.10.1, is 2:1, vs. 1:1 for the same anode area and cathode-to-anode gap.
SPU'ITER-ION PUMPING
181
References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.
Herb, R.G. and Davis, R., "Evapor-Ion Pump", Phys. Rev. 89, 897 (1953). Herb, R.G., Pauly, T., Welton, R.D., Fisher, K.I., "Sublimation and Ion Pumping in Getter-lon Pumps", Rev. Sci. Instrum., 35(5), 573 (1964). Adam, H. and Bachler, W., "Operational Procedures and Experiences with a High-Speed Ion Getter Pump", Proc. 2nd Int. Vac. Cong., 1961 (Pergamon Press, Inc., New York, 1962), p. 374. Hall, L.D., Helmer, J.C., Jepsen, R.L., U.S. Patent No. 2,993,638, "Electrical Vacuum Pump Apparatus and Method", f'ded 7/24/57, awarded 7/25/61. Redhead, P.A., "Instabilities in Crossed-Field Discharges at Low Pressures", Vacuum 38(8-10), 901 (1988). Gaede, V.W., "Tiefdruckmessungen" (vacuum gauge review paper), Zeitschr. f. Teclm. Physik 12, 664 (1934). Guthrie, A. and Wakerling, R.K., Vacuum Equipment and Techniques (McGraw-HiU Book Co., Inc., New York, 1949), p. 115. Philips, C.E.S., Proc. Roy. Soc. (London), A64, 172 (1898). Cobine, J.D., Proc. Nat. Acad. Sci. U.S., 2, 683 (1916). Penning, F.M., U.S. Patent, "Coating by Cathode Disintegration", filed 11/7/36, awarded 2/7/39. Penning, F.M., "Die Glimmentladung Bei Niedrigem Druck Zwischen Koaxialen Zylindern in Einem Axialen Magnetfeld" (sputtering apparatus paper), Physica 3(9), 873 (1936). Guentherschulze, A., "Cathodic Sputtering, an Analysis of the Physical Processes ", Vacuum 3(4) (1953). Gill, W.D. and Kay, E., "Efficient Low Pressure Sputtering in a Large Inverted Magnetron Suitable for Film Synthesis", Rev. Sci. Instrum. 36(3), 277 (1%5). Wasa, K. and Hayakawa, S., "Low Pressure Sputtering System of the Magnetron Type", Rev. Sci. Instrum. 40(5), 693 (1%9). Welch, K.M., "New Materials and Technology for Suppressing Multipactor in High Power Microwave Windows", Stanford Linear Accelerator Pub. No. SLAC-174, 1974. Penning, F.M., "Ein Neues Manometer fur Niedrige Gasdrucke, Insbesondere Zwischen 10-3 und 10.5 mm" (paper on various configurations of cold cathode gauges), Physica _4(2), 71 (1937). Penning, F.M. and Nienhuis, K., "Construction and Application of a New Design of the Philips Vacuum Gauge", Philips Tech. Rev. 11(4), 116 (1949). Soddy, F., Proc. Roy. Soc. (London), A80, 92 (1908). Westendorp, W.F., and Gurewitsch, A.M., U.S. Patent No. 2,755,014, "Ionic Vacuum Pump Device", filed 4/24/53, awarded 7/17/56. Gurewitsch, A.M. and Westendorp, W.F., "Ionic Pump", Rev. Sci. Instrum. 25(4), 389 (1954). Gale, A.J., "Cold Sealed Getter/Ion Pumped Supervoltage X-ray Tubes", Proc. 2nd Nat. AVS Symp., 1955 (Committee on Vacuum Techniques, 1956), p. 12. Corm, G.K.T. and Daglish, H.N., "Cold Cathode Ionization Gauges for the Measurement of Low Pressure", Vacuum 3(1), 24 (1953). Corm, G.K.T. and Daglish, H.N., "The Influence of Electrode Geometry on Cold-Cathode Vacuum Gauges", Vacuum 3(2), 136 (1954).
182
SPUTtER-ION PUMPING
References 24. Jepsen, R.L., "Important Characteristics of a New Type Getter-Ion Pump", LeVide 80, 80 (1959). 25. Rutherford, S.L., U.S. Patent No. 3,258,194, "Magnetically Conf'med Glow Discharge Apparatus", filed 7/29/63, awarded 6/28/66. 26. Rutherford, S.L., "Sputter-Ion Pump for Low Pressure Operation", Proc. 10th Nat. AVS Symp., 1963 (The Macmillan Company, New York, 1964), p. 185. 27. Lamont, L.T., "Physical Processes in the Penning Discharge from 10-8 to 10-a Torr", Proc. 14th Nat. AVS Symp., 1967 (Herbrick and Held Printing Co., Pittsburgh, 1968), p. 149. 28. Pierini, M., "Use of Discharge Intensity for the Evaluation of Pumping Characteristics of a Sputter Ion Pump", J. Vac. Sci. Technol. A2(2), 195 (1984). 29. Tom, T. and James, B.D., U.S. Patent No. 3,398,879, "Asymmetric Ion Pump Method", flied 10/7/66, awarded 8/27/68. 30. Jepsen, R.L., "Magnetically Confined Cold-Cathode Gas Discharges at Low Pressures", J. Appl. Phys. 32(12), 2619 (1961). 31. Dow, D.G., "Electron-Beam Probing of a Penning Discharge", J. Appl. Phys. 34(8), 2395 (1963). 32. Lange, W.J., "Microwave Measurement of Electron Density in a Penning Discharge", J. Vac. Sci. Technol. 7(1), 228 (1970). 33. Chen, F.F., Introduction to Plasma Physics (Plenum Press, New York, 1974), p. 154. 34. Redhead, P.A., "Oscillations in Magnetically Confined Cold-Cathode Discharges at Very Low Pressures", Canadian J. Phys. 43, 1001 (1965). 35. Helmer, J.C. and Jepsen, R.L., "Electrical Characteristics of a Penning Discharge", Proc. I.R.E. 49, 1920 (1961). 36. Knauer, W., "Mechanism of the Penning Discharge at Low Pressures", J. Appl. Phys. 33(6), 2093 (1962). 37. Hooper, E.G., "A Review of Reflex and Penning Discharges", Advances in Electronics and Electron Physics, 27, 295 (1969). 38. Blodgett, K.B. and Vanderslice, T.A., "Mechanism of Inert Gas Cleanup in a Gaseous Discharge", J. Appl. Phys. 31(6), 1017 (1960). 39. Dallos, A., "The Pressure Dependence of the Pumping Speed of Sputter Ion Pumps", Vacuum 19(2), 79 (1969). 40. Peacock, N.T. and Peacock, R.N., "Some Characteristics of an Inverted Magnetron Cold Cathode Ionization Gauge with Dual Feedthroughs", J. Vac. Sci. Technol. A6(3), 1141 (1988). 41. Ono, S. and Yokoo, K., "The Multipactor Ion Pump for Ultra-High Vacuum", Proc. 8th Int. Vac. Cong., 1980 (Suppl6 ment a la Revue, "Le Vide, les Couches Minces, No. 201", 1981), p. 299. 42. Yokoo, K. and Ono, S., "High-Speed Ion Pumps with a Multipactor Cathode-the Multipactor Ion Pump", Vacuum 32(5), 265 (1982). 43. Welch, K.M., "Diode Sputter Ion Pump Speed for H 2 and N2 as a Function of Axial and Transverse Magnetic Fields up to 1.2 Tesla", Stanford Linear Accelerator Technical Note No. TN-72-10, July 1972. 44. Jepsen, R.L., "The Physics of Sputter-Ion Pumps", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and the Physical Society, London, 1969), p. 317.
SPUTTER-ION PUMPING
183
References 45. Welch, K.M., "New Developments in Sputter-Ion Pump Configurations", J. Vac. Sci. Technol. 13, 498 (1976). 46. Chou, T.S. and McCafferty, D., "Pumping Behavior of Sputter-Ion Pumps", J. Vac. Sci. Technol. 18(3), 1148 (1981). 47. Hamilton, A.R., "Some Experimental Data on Parameter Variation with Triode Getter-lon Pumps", Proc. 2nd Int. Vac. Cong. and 8th Nat. AVS Symp., 1961 (Pergamon Press, Inc., New York, 1962), p. 388. 48. Konjevic, R., Grant, W.A., Carter, G., "Ion Pumping Effects in Inert Gas Mixtures in an Ionization Tube", Vacuum 18(9), 511 (1968). 49. Rutherford, S.L., U.S. Patent No. 3,159,332, "Methods and Apparatus for Enhanced Sputter-Ion Pump Operation", filed 8/14/61, awarded 12/1/64. 50. Audi, M., "Pumping Speed of Sputter Ion Pumps", Vacuum 38(8-10), 669 (1988). 51. Berman, A., "Methods of Pumping Speed and Gas Release Measurement in Ionization Gauge Heads--A Review", Vacuum 32(8), 497 (1982). 52. Bowden, P. and Brandon, D.G., "The Generation of Dislocations in Metals by Low Energy Ion Bombardment", Phil. Mag. 8, 935 (1963). 53. Oechsner, H., "Sputtering--A Review of Some Recent Experimental and Theoretical Aspects", Appl. Phys. 8, 185 (1975). 54. Maissel, L.I., Physics of Thin Films, Vol. 3 (Academic Press, New York, G. Hass and R.E. Thun, Eds.), Chapt. 2, "The Deposition of Thin Films by Cathode Sputtering", p. 61. 55. Andersen, H.H. and Bay, H.L., "Sputter-Yield Studies on Silicon and Silver Target", Radiation Effects 19, 139 (1973). 56. Brubaker, W.M., "A Method for Greatly Enhancing the Pumping Action of a Penning Discharge", Proc. 6th Nat. AVS Symp., 1959 (Pergamon Press, London, 1960), p. 302. 57. Rudnitskii, E.M., "Ion Current Density Distribution at the Cathode of a Penning Discharge", Soviet Phy.-Tech. Phys. 12(5), 666 (1967). 58. Young, J.R., "Pressure Dependence of the Axial Ion Current in a Penning Discharge", J. Vac. Sci. Technol. 5(4), 102 (1968). 59. Komiya,S. and Yagi, N., "Enhancement of Noble Gas Pumping for a SputterIon Pump", J. Vac. Sci. Technol. 6(1), 54 (1969). 60. Smirtnitskaya, G.V. and Tee, N.K., "On the Ion Current Distribution on the Cathode in an Ion Pump Discharge at Low Pressures", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969), p. 356. 61. Ho, W., Wang, R.K., Keng, P.T., "Calculation of Sputtering Ion Pump Speed", J. Vac. Sci. Technol. 20(4), 1010 (1982). 62. Grant, W.A., Carter, G., "Ion Trapping and Gas Release Phenomena", Vacuum 15(10), 477 (1965). 63. Reed, DJ., Harris, F.T., Armour, D.G., Carter, G., "Thermal Evolution Spectrometry of Low Energy Helium Ions Injected into Stainless Steel and Nickel Targets", Vacuum 24(4), 179 (1974). 64. Redhead, P.A., Hobson, J.P., Kornelsen, E.V., _ThePhysical Basis of Ultrahigh Vacuum (Chapman and Hall, Ltd., London, 1968).
184
SPUTtER-ION PUMPING
References 65. Winters, H.F., Horne, D.E., Donaldson, E.E., "Absorption of Gases Activated by Electron Impact", J. Chem. Phy. 41(9), 2766 (1964). 66. Bond, G.C., Catalysis by Metals (Academic Press, Inc., New York, 1962) p. 69. 67. Wagener, S., "Efficiency and Mechanism of Barium Getters at Low Pressure", Brit. J. Appl. Phys. 2, 132 (1951). 68. Trapnell, B.M.W., "The Activities of Evaporated Metal Films in Gas Chemisorption", Proc. Roy. Soc. A218, 566 (1953). 69. Reikhrudel, E.M., Smirnitskaya, G.V., Burnisenico, A.I., "Ion Pump with Cold Electrodes and its Characteristics", Radiotekh. Electron 2, 253 (1956). 70. Momose, T., Kanazawa, K., Hisamatsu, H., Ishimaru, H., "Inversely Operated Aluminum Alloy Distributed Ion Pump", J. Vac. Sci. Technol. A__77(5),3092, (1989). 71. McCracken, G.M. and Maple, J.H.C., "The Trapping of Hydrogen Ions in Molybdenum, Titanium, Tantalum, and Zirconium", Brit. J. Appl. Phys. 18, 919 (1967). 72. Liu, Y.C., Lin, C.C., Lee, S.F., "Pumping Mechanism for N2 Gas in a Triode Ion Pump with All00 Aluminum Cathode", J. Vac. Sci. Technol. A6(1), 139 (1988). 73. Welch, K.M., "Pumping Mechanisms in Sputter-ion Pumps; Low Pressure Operation" IEEE Conference on Particle Accelerators, 91CH3038-7 4_, 2269 (1991). 74. Okano, T., Ohsaki, A., Tuzi, Y., "A Zr-AI Composite-Cathode Sputter-Ion Pump", J. Vac. Sci. Technol. A_.22(2),191 (1984). 75. Ishimaru, H., "All-Aluminum-Alloy Ultra High Vacuum System for a LargeScale Electron-Positron Collider", J. Vac. Sci. Technol. A_.22(2),1170 (1984). 76. Lu, M. and Fang, P.W., "New Cathode for Sputter Ion Pumps: Aluminum Mixed Rare-Earth Alloy", J. Vac. Sci. Technol. A5 (4), 2591 (1987). 77. Holland, L., "Theory and Design of Getter-Ion Pumps", J. Sci. Instrum. 36, 105 (1959). 78. Hall, L.D., "Electronic Ultra-High Vacuum Pump", Rev. Sci. Instrum. 29(5), 367 (1958). 79. Wheeler, W.R. and Carlson, M., "Ultra-High Vacuum Flanges", Proc. 2rid Int. Vac. Cong. and 8th Nat. AVS Symp., 1961 (Pergamon Press, Inc., New York, 1962), p. 1309. 80. Hall, L.D., "Properties and Behavior of Electronic Ultra-High Vacuum Pumps", Proc. 5th Nat. AVS Symp., 1958 (Pergamon Press, Inc., London, 1958), p. 158. 81. Halliday, D., Resnick, R., Physics for Students of Science and Engineering, Part I (John Wiley and Sons, Inc., New York, 1963), p. 191. 82. Lloyd, W.A. and Huffman, G.A., U.S. Patent No. 2,983,433, "Getter Ion Vacuum Pump Apparatus", filed 8/1/58, awarded 5/9/61. 83. Zaphiropoulos, R. and Lloyd, W.A., "Design Considerations for High Speed Getter-Ion Pumps", Proc. 6th Nat. AVS Symp., 1959 (Pergamon Press, Inc., New York, 1960), p. 307. 84. Andrew, D., "The Development of Sputter-Ion Pumps", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969), p. 325.
SPUTI'ER-ION PUMPING
185
References 85. Jepsen, R.L., Francis, A.B., Rutherford, S.L., Kietzmann, B.E., "Stabilized Air Pumping with Diode Type Getter-Ion Pumps", Proc. 7th Nat. AVS Symp., 1960 (Pergamon Press, Inc., New York, 1961), p. 45. 86. Malter, L. and Mandoli, H., "Electron Tube Processing with Getter-Ion Pumps", Vacuum 10, 121 (1960). 87. Neal, R.B., "The Stanford Two-Mile Linear Electron Accelerator", J. Vac. Sci. Technol. 2(3), 149 (1%5). 88. Rutherford, S.L., Mercer, S.L., Jepsen, R.L., "On Pumping Mechanisms in Getter-Ion Pumps Employing Cold-Cathode Gas Discharges", Proc. 7th Nat. AVS Symp., 1960 (Pergamon Press, Inc., New York, 1961), p. 380. 89. Andrew, D., Sethna, D.R., Weston, G.F., "Inert-Gas Pumping in a Magnetron Pump", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and the Physical Society, London, 1969), p. 337. 90. Jepsen, R.L., U.S. Patent No. 3,070,719, "Cathodes for Magnetically-Conf'med Glow Discharge Apparatus", filed 10/11/60, awarded 12/25/62. 91. Vaumoron, J., "The Study of the Influence of Certain Parameters of a Penning Pump on its Stability in Argon", LeVide 25(145), 26 (1970). 92. Vanderslice, T.A., U.S. Patent No. 3,080,104, "Ionic Pump", filed 9/25/58, awarded 3/5/63. 93. Clarke, Pj. and Roman, N., U.S. Patent No. 3,325,086, "Triode Ion Pump", filed 9/20/65, awarded 6/13/67. 94. Brubaker, W.M. and Berry, C.E., U.S. Patent No. 3,535,055, "Cold-Cathode Discharge Ion Pump", filed 5/25/59, awarded 10/20/70. 95. Lafferty, J.M. and Vanderslice, T.A., "The Interplay of Electronics and Vacuum Technology", Proc. I.R.E. 49, 1136 (1961). 96. Kawasaki, K., Sugita, T., Kohno, I., Watanabe, J., "Study of Pumping Action of a Diode Type Getter Ion Pump for Inert Gases by the Tracer Technique", Japan J. Appl. Phys. 2(2), 315 (1963). 97. Willis, R.D., Allman, S.L., Chen, C.H., Alton, G.D., Hurst, G.S., "Pumping Inert Gases by Electron-Impact Ionization Sources and Associated Memory , Effects", J. Vac. Sci. Technol. A2(1), 57 (1984). 98. Pierini, M. and Dolcino, L., "A New Sputter-Ion Pump Element", J. Vac. Sci. Technol. Al(2), 140 (1983). 99. Audi, M. and Pierini, M., "Surface Structure and Composition Profile of Sputter-Ion Pump Cathode and Anode", J. Vac. Sci., Technol. A4(3),. 303 (1986). 100. Audi, M., Dolcino, L., Doni, F., Ferrario, B., "A New Ultrahigh Vacuum Combination Pump", J. Vac. Sci. Technol. A5(4), 2587 (1987). 101. Audi, M. and DeSimon, M., "The Influence of Heavier Gases in Pumping Helium and Hydrogen in an Ion Pump", J. Vac. Sci. Technol. A6(3), 1205 (1988). 102. Singleton, J.H., "Hydrogen Pumping by Sputter-Ion Pumps and Getter Pumps", J. Vac. Sci. Technol. 8(1), 275 (1971). 103. Dallos, A. and Steinrisser, F., "Pumping Speeds of Getter-Ion Pumps at Low Pressures", J. Vac. Sci. Technol. 4(1), 6 (1967). 104. Wear, K.B., "Electrical Characteristics of Sputter-Ion Pumps", J. Appl. Phys. 38(4), 1936 (1967).
186
SPUITER-ION PUMPING
References 105. Bloomer, R.N. and Cox, B.M., "Some Effects of Gases Upon Vacuum Breakdown Initiated by Field Emission of Electrons", Vacuum 18(7), 379 (1968). 106. Tom, T. and James, B.D., "Inert Gas Ion Pumping Using Differential Sputter-Yield Cathodes", Proc. 13th Nat. AVS Symp., 1966 (Herbick and Held Printing Company, Pittsburgh, Pennsylvania, 1967), p. 21. 107. Tom, T. and James, B.D., "Inert Gas Ion Pumping Using Differential Sputter Yield Cathodes", J. Vac. Sci. Technol. 6(2), 304 (1969). 108. Jepsen, R.L., "The Physics of Sputter-Ion Pumps", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969), p. 317. 109. Brodie, I., Lamont, L.T, Jepsen, R.L., "Production of High-Energy Neutral Atoms by Scattering of Ions at Solid Surfaces and Its Relation to Sputtering", Phys. Rev. Lett. 21(17), 1224 (1968). 110. Kornelsen, W., "The Ionic Entrapment and Thermal Desorption of Inert Gases in Tungsten for Kinetic Energies of 40 eV to 5 keV", Can. J. Phys. 42, 364 (1964). 111. Shock, C. and Kistemaker, J., "Fast Ion Scattering Against Metal Surfaces", Advances in Electronics and Electron Physic, Vol. 21 (Academic Press, New York, 1965, L. Marton and C. Marton, Eds.), p. 67. 112. Vaumoron, J.A., "Influence of Certain Parameters of a Penning Pump on its Stability in Argon", LeVide 145, 26 (1970). 113. Harra, D.J., Abstract: "Pumping Mechanisms of Sputter-Ion Pumps", J. Vac. Sci. Technol. 11, 331 (1974). 114. Hall, L.D., "Multicomponent Getter Films for Getter-Ion Pumps", J. Vac. Sci. Technol. 6(1), 44 (1969). 115. Vaumoron, J.A. and DeBiasio, M.P., "Argon and Rare Gas Instability With Heavy Metal Cathode Pumps", Vacuum 20(3), 109 (1970). 116. Baker, P.N. and Laurenson, L., "Pumping Mechanisms for the Inert Gases in Diode Penning Pumps", J. Vac. Sci. Technol. 9(1), 375 (1972). 117. Rutherford, S.L., Jepsen, R.L., Frances, A.B., "Study Program on the Utilization of Vacion® Pump Elements for DCX", Final Report, August, 1960, prepared for Oak Ridge National Laboratory, Varian Engineering Report No. 267-1F. 118. Vissers, D.R., Holmes, J.T., Nelson, P.A., Bartholme, L.G., U.S. Patent No. 3,683,272, "Method and Apparatus for Determining Hydrogen Concentration in Liquid Sodium Utilizing an Ion Pump to Ionize the Hydrogen", filed 11/24/70, awarded 8/8/72. 119. Welch, K.M., U.S. Patent No. 4,047,102, "Dual Chamber, High Sensitivity Gas Sensor and Pump", filed 5/3/76, awarded 9/6/77. 120. Singleton, J.H., "Hydrogen Pumping Speed of Sputter-Ion Pumps", J. Vac. Sci. Technol. 6(2), 316 (1969). 121. Singleton, J.H., "The Performance Characteristics of Modern Vacuum Pumps", J. Phys. E6, 685 (1973). 122. Komiya, S., Sato, S., Hayashi, C., "Triggering Sputter Ion Pump in Extreme High Vacuum", Proc. 13th Nat. AVS Symp., 1966 (Herbick and Held Printing Company, Pittsburgh, Pennsylvania, 1967), p. 19. 123. Dushman,S., Scientific Foundations of Vacuum Technique (John Wiley and Sons, Inc., New York, 1949), p. 581.
SPUqqqER-ION PUMPING
187
References 124. Z~idel, A.N., Spectra-Isotopic Method for Determination of H~ in Metals (Butterworths, London, 1961). 125. Riviere, J.C. and Allison, J.D., "Gas Evolution During Baking of Sputter-Ion Pumps", Supplemento AI Nuovo Cimento, 1(2), 520 (1963); also, Vacuum 14(3), 97 (1964). 126. Fast, J.D., "Foreign Atoms in Metals", Philips Tech. Rev. 16(12), 341 (1955). 127. Klaichko, I.A., Kunin, L.L., Fedorov, S.P., Larionov, I.N., "Studies of Interaction Between Gases and Metals", UCRL Translation #1375, from Russian Akademiia Nauk SSSR Komissiia po Analiticheskoi Khimii, Trudy 10, 49 (1960). 127. Klaichko, I.A., Kunin, L.L., Fedorov. S.P., Larionov, I.N., "Studies of the Interactions Between Gases and Metals", UCRL Translation #1375, from Russian Akademiia Nauk SSSR Komissiia po Analiticheskoi Khmii, Trudy 10, 49 (1960). 128. Carter, G., Nobes, MJ., Armour, D.G., 'The Erosion Energy Efficiency of Sputtering", Vacuum 32(8), 509 (1982). 129. Colligon, J.S., "Ion Bombardment of Metal Surfaces", Vacuum 11, 272 (1961). 130. KenKnight, C.E. and Wehner, G.K., "Sputtering of Metals by Hydrogen Ions", J. Appl. Phys. 35(2), 322 (1964). 130. KenKnight, C.E., Wehner, G.K., "Sputtering of Metals by Hydrogen Ions", J. Appl. Phys. 35(2), 322(1964). 131. Rutherford, S.L. and Jepsen, R.L., "Enhanced H2 Pumping With Sputter-Ion Pumps", Rev. Sci. Instrum. 32, 1144 (1961). 132. Dean, N., "Helium Treatment of Vac-Ion Pumps to Remove Deposited Argon", Rev. Sci. Instrum. 36, 12~ (1965). 133. Cummings, U., Dean, N., Johnson, F., Jurow, J., Voss, J., "Vacuum System for the Stanford Storage Ring, SPEAR", J. Vac. Sci. Technol. 8(1), 348 (1971). 134. Jepsen, R.L., U.S. Patent No. 3,147,910, "Vacuum Pump Apparatus", filed 8/30/61, awarded 9/8/64. 135. Hill, E.F., U.S. Patent No. 4,097,195, "High Vacuum Pump", filed 10/10/75, awarded 6/27/78. 136. Lamont, L.T., U.S. Patent No. 3,460,745, "Magnetically Confined Electrical Discharge Getter Ion Vacuum Pump Having a Cathode Projection Extending into the Anode Cell", filed 8/23/67, awarded 8/12/69. 137. Lamont, L.T., "A Novel Diode Sputter-Ion Pump", J. Vac. Sci. Technol. 6(1), 47 (1969). 138. Welch, K.M., U.S. Patent No. 3,994,625, "Sputter-Ion Pump Having Improved Cooling and Improved Magnet Circuitry", filed 2/18/75, awarded 11/30/76. 139. Brothers, C.F., Tom, T., Munro, D.F., "Design and Performance of a 50,000 ,~/sec Pump Combining Cold Cathode Ion Pumping and Active Film Gettering", Proc. 10th Nat. AVS Symp., 1963 (The Macmillan Company, New York, 1964), p. 202. 140. Laurent, J.M., "Study of Ion Pumps for Operation in the Low Magnetic Bending Field of a Large Electron Storage Ring", Proc. 8th Int. Vac. Cong., 1980 (Suppl6 ment a la Revue, "LeVide, les Couches Minces, No. 201", 1981), p. 164. 141. Zeng-Rui, L. and Ming, L., "Pumping Mechanism for H~ and Cathode Material in a Sputter Ion Pump", op. cit., p. 295.
188
SPUTTER-ION PUMPING
References 142. Donachie, M.J., Ed., Titanium and Titanium Alloys (American Society for Metals, Metals Park, Ohio, 1982). 143. Lamont, L.T., "A Model for the Penning Discharge in the Neighbourhood of the Transition Point", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969), p. 345. 144. Beanland, D.G., Ion Implantation and Beam Processing (Academic Press, New York, 1984, J.S. Williams and J.M. Poate, Eds.), Chapt. 8, "High-Dose Implantation", p. 297. 145. Young, J.R., "Penetration of Electrons and Ions in Aluminum", J. Appl. Phys. 27(1), 1 (1956). 146. Paton, N.E. and Williams, J.C., Titanium and Titanium Alloys (Mj. Donachie, Ed., American Society for Metals, Metals Park, Ohio, 1982), Section IV, "Effect of Hydrogen on Titanium and Its Alloys", p. 185. 147. Barret, C.R., Nix, W.D., Tetelman, A.S., The Principles of Engineering Materials (Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1973), p. 152. 148. Wanhill, R.J.H., "Aqueous Stress Corrosion in Titanium Alloys", Brit. Corrosion J. 10(2), 69 (1975). 149. Philibert, J., "Diffusion in Solids", Sci. & Technol., August, 1968, p. 47. 150. Crank, J., The Mathematics of Diffusion (Oxford University Press, Oxford, 1975), p. 254. 151. Ibid, p. 50. 152. Kubaschewski, O., Titanium: Physico-Chemical Properties of Its Alloys, Special Issue No. 9 (International Atomic Energy Agency, Vienna, 1983, K.L. Komarek, Ed.), Chapt. IV, "Diffusion", p. 441. 153. Crank, J., The Mathematics of Diffusion (Oxford University Press, Oxford, 1975), p. 21. 154. Ibid, p. 61. 155. Spagenberg, K.R., Vacuum Tubes (McGraw-Hill Book Company, Inc., New York, 1948), p. 125. 156. Snouse, T., "Starting Mode Differences in Triode and Diode Sputter-Ion Pumps", J. Vac. Sci. Technol. 8(1), 283 (1971). 157. Baker, L.C. and Bance, R.D., "The Pumping of Deuterium by Triode Ion Pump in the 10"s Torr Region on a Van de Graaff Accelerator", Vacuum 20(6), 251 (1970). 158. Turner, C.M., Marinuzzi, J.G., Bennett, G.W., "Improvement of Hydrogen Pumping of Penning Discharge Getter-Ion Pumps", IEEE Trans. Nucl. Sci., NS-14, 831 (1967). 159. Denison, D.R., "Comparison of Diode and Triode Sputter-Ion Pumps", J. Vac. Sci. Technol. 14(1), 633 (1977). 160. Hatch, J.E., Editor, Aluminum, Properties and Physical Metallurgy (American Society of Metals, Metals Park, Ohio, 1984), p. 18. 161. Dushman, S., .Scientific Foundations of Vacuum Technique (John Wiley and Sons, New York, 1962), p. 570. 162. Celik, M.C. and Bennett, G.H.J., "Effects of Hydrogen Inclusions on Blistering in High-Purity Aluminum Sheet and Foil on a Laboratory Scale", Metals Technol. 6(4), 138 (1979).
SPU'I'TER-ION PUMPING
189
References 163. Barnes, R.S. and Mazey, D.J., "The Migration and Coalescence of Inert Gas Bubbles in Metals", Proc. Roy. Soc. (London), A275, 47 (1963). 164. Kleuh, R.L. and Mullins, W.W., "Some Observations on Hydrogen Embrittlement of Silver", Trans. Metallurgical Soc. AIME, 242, 237 (1968). 165. Norton, F.J., "Gas Permeation Through the Vacuum Envelope", Proc. 8th Nat. AVS Symp. and 2nd Int. Vac. Cong., 1961 (Pergamon Press, New York, 1962, L.E. Preuss, Editor), Vol. I, p. 8. 166. Eschbach, H.L., Gross, F., Schulien, S., "Permeability Measurements with Gaseous Hydrogen for Various Steels", Vacuum ~ 543 (1963). 167. McCracken, G.M. and Maple, J.H.C., "Reflection of Hydrogen Ions From Metal Surfaces in the Energy Range 5-30 keV", Proc. 7th Int. Conf. on Phenomena in Ionized Gases, Beograd, 1965 (Gradevinska Knjiga Publishing House, Beograd, 1966, B. Perovic and D. Tosic, Editors), Vol. I., p. 137. 168. PNEUROP, Editorial Body of Twelve European Countries, Vacuum Technology Sub-committee, "Acceptance Specifications, Part VI, Acceptance Rules for Sputter Ion Pumps" (British Compressed Air Society, PNEUROP Editorial Body, 1976). 169. Rapp, D. and Englander-Golden, P., "Total Cross Sections for Ionization and Attachment in Gases by Electron Impact. I. Positive Ionization", J. Chem. Phys. 43(5), 1464 (1965). 170. Jepsen, R.L., King, L.T., Rogers, R.M., U.S. Patent No. 3,116,764, "High Vacuum Method and Apparatus", filed 3/30/59, awarded 1/7/64. 171. Souchet, R., "Study of the Degassing Spectra of Materials for the Construction of the Vacuum Chamber of an e + e" Colliding Beam at 1.8 GeV", Institut National de Physique Nucleaire et de Physique des Particules, Laboratoire de L'Accelerateur Lineaire, Report No. 4-72, June, 1972. 172. Jepsen, R.L., U.S. Patent No. 3,149,774, "Getter Ion Pump Method and Apparatus", filed 1/27/61, awarded 9/22/64. 173. Jepsen, R.L., U.S. Patent No. 3,331,975, "Cooling Apparatus for Cathode Getter Pumps", filed 2/19/65, awarded 7/18/67. 174. Johnson, J.W., Atkins, W.H., Dowling, D.T., McConnell, J.W., Milner, W.T., Olsen, D.K., "Heavy Ion Storage Ring for Atomic Physics Vacuum Test Stand for Pressures of 10-1 2 Torr", J. Vac. Sci. Technol. A_..~7(3),2430 (1989). 175. Tom, T., "Use of a Metallic Ion Source in Cold Cathode Sputter Ion Pumps", Proc. 5th Int. Vac. Cong., 1971, J. Vac. Sci. Technol. 9(1), 383 (1972). 176. Jepsen, R.L., U.S. Patent No. 3,094,639, "Glow Discharge Method and Apparatus", f'ded 10/6/60, awarded 6/18/63. 177. Helmer, J.C., private communication. 178. Guthrie, A. and Wakerling, R.K., Vacuum Equipment Techniques (McGrawHill Book Company, Inc., New York, 1949), p. 39. 179. O'Neill, G.K, "Super-High-Energy Summer Study No. 8, Storage Ring Work at Stanford", Brookhaven National Laboratory, Internal Report AADD-8 (1963). 180. Fischer, G.E. and Mack, R.A., "Vacuum Design Problems of High Current Electron Storage Rings", Cambridge Electron Accelerator Report CEAL-1013 (1964).
190
SPU'ITER-ION PUMPING
References 181. Anashin, V.V., Auslender, V.L., Bender, E.D., Blinov, G.A., Malev, M.D., Osipov, B.N., Popov, A.T., Trakhtenberg, E.M., "The Ultrahigh-Vacuum Pumping System of the VEPP-2 Storage Ring", Proc. All-Union Conf. on Particle Accelerators, Moscow, 1968, Vol. 2, 560 (1970). 182. Cummings, U., Dean, N., Johnson, F., Jurow, J., Voss, J., "Vacuum System for Stanford Storage Ring, SPEAR", Stanford Linear Accelerator Center, SLACPUB-797 (1970). 183. Zeilinger, A., and Pochman, W.A., "New Methods for the Measurement of Hydrogen Diffusion in Metals", J. Appl. Phys. 47(12), 5478 (1976). 184. Falland, Chr., Hartwig, H., Kouptsidis, J., Kueppershaus, R., Schumann, G., Schwartz, M., Wedekind, H.-P., "First Operational Experience with the 2.3 km Long UHV System of the Electron Storage Ring PETRA", Proc. 8th Int. Vac. Cong., 1980 (Supple- ment ~ la Revue, "LeVide, les Couches Minces, No. 201", 1981), p. 126. 185. Blechschmidt, D., Bojon, J.-P., Chapman, G., Jensen, D., Monnier, B., Nordstrom, H., "Linear Sputter-Ion Pump for Integration in High Magnetic Fields", Proc. 8th Int. Vac. Cong., 1980 (Supple- ment ~ la Revue, "LeVide, les Couches Minces, No. 201", 1981), p. 159. 186. Hartwig, H. and Kouptsidis, J.S., "Design Performance of Integrated SputterIon Pumps for Particle Accelerators", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. R~ denauer, F.P. Viehb8 ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 93. 187. Momose, T., Chen, J.R., Kanazawa, K., Hisamatsu, H., Ishimaru, H., "A New Distributed Ion Pump Made of Aluminum Alloy with an Aluminum or a Titanium Cathode in the Transposable Ring Intersecting Storage Accelerators in Nippon e + e" Colliding Ring", J. Vac. Sci. Technol. A._.~7(5),3098 (1989). 188. Malev, M.D. and Trachtenberg, E.M., "Built-in Getter-Ion Pumps", Vacuum 23(11), 403 (1973). 189. Ishimaru, H., Nakanishi, H., Horikoshi, G., "Distributed Sputter Ion Pump and Titanium Getter Ion Pump Combination", Proc. 8th Int. Vac. Cong., 1980 (Supple- ment ~ la Revue, "Le Vide, les Couches Minces, No. 201", 1981), p. 331. 190. Pingel, H. and Schulz, L., "A High-Field Integrated Sputter Ion Pump for BESSY- The Berlin Electron Storage Ring for Synchrotron Radiation", Proc. 8th Int. Vac. Cong., 1980 (Supple- ment ~ la Revue, "LeVide, les Couches Minces, No. 201", 1981), p. 147. 191. Pingel, H. and Schulz, L., "The Ultra High Vacuum System for BESSY - The Berlin Electron Storage Ring for Synchrotron Radiation", Proc. 8th Int. Vac. Cong., 1980 (Supple. ment ~ la Revue, "LeVide, les Couches Minces, No. 201", 1981), p. 119. 192. Ryabov, V.V. and Saksagansky, G.L., "Influence of Asymmetry of Magnetic and Electric Fields on the Parameters of Sputter-Ion Pumps", Vacuum 22(5), 191 (1972). 193. Trickett, B.A., "Vacuum Systems for the Daresbury Synchrotron Radiation Source", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. Rfidenauer, F.P. Viehb0ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 347.
SPUTrER-ION PUMPING
191
References 194. Tsujikawa, H., Chida, K., Mizobuchi, T., Miyahara, A., "Vacuum System for the Test Accumulation Ring for Numatron Project (TARN)", Proc. 8th Int. Vac. Cong. 1980 (Suppl6 ment ~ la Revue, "Le Vide, les Couches Minces, No. 201", 1981), p. 151. 195. Pressman, A.I., Switching and Linear Power Supply, Power Converter Design (Hayden Book Company, New Jersey, 1977). 196. Quinn, D.L., U.S. Patent No. 3,412,310, "Power Supply for Glow Discharge Type Vacuum Pumps with Voltage-Doubler Bridge-Rectifier and a Soft Transformer", filed 3/6/67, awarded 11/19/68. 197. Mandoli, H.A., U.S. Patent No. 3,118,103, "Voltage Doubling Power Supply", f'lled 6/1/59, issued 1/14/64. 198. Mandoli, H.A. and Rorden, J.R., U.S. Patent No. 3,267,377, "Measuring Circuit for Providing an Output either Linearly or Logarithmically Related to an Input", f'ded 6/1/59, awarded 8/16/66. 199. Underhill, E.M., Ed., Permanent Magnet Design (Crucible Steel Co. of America, Pittsburgh, PA, 1957). 200. Casimir, H.B.G., Haphazard Reality (Harper and Row, New York, 1983), p. 281. 201. van den Brock, C.A.M. and Stuijts, A.L., "Ferroxdure", Philips Tech. Rev. 37(7), 157 (1977). 202. Wohlfarth, E.P., Editor, Ferromagnetic Materials (North Holland Publishing Company, Inc., 1982). 203. Helmer, J.C., "Magnetic Circuits Employing Ceramic Magnets", Proc. I.R.E., 1528 (1961). 204. Helmer, J.C., U.S. Patent No. 3,159,333, "Permanent Magnet Design for Pumps", filed 8/21/61, awarded 12/1/64. 205. Kearns, W.J., "A High Efficiency Magnetic Field Design for Large Ion Pumps", Proc. 10th Nat. AVS Syrup., 1963 (The Macmillan Company, New York, 1964), p. 180. 206. Kraus, J.D., Electromagnetics (McGraw-Hall Book Company, Inc., New York), p. 206. 207. Schuurman, W., "Investigation of a Low Pressure Penning Discharge", Physica 36, 136 (1967). 208. Calder, R. and Lewin, G., "Reduction of Stainless-Steel Outgassing in UltraHigh Vacuum", Brit. J. Appl. Phys. 18, 1459 (1967). 209. Knauer, W. and Lutz, M.A., "Measurement of the Radial Field Distribution in a Penning Discharge by Means of the Stark Effect", Appl. Phys. Lett. 2(6), 109 (1963). 210. Herzberg, G., Atomic Spectra and Atomic Structure (Dover Publications, New York, 1944). 211. Lange, WJ., "Microwave Measurements of Electron Density in a Penning Discharge", J. Vac. Sci. Technol. 7(1), 228 (1970). 212. Agdur, B. and Ternstr6 m, U., "Instabilities in Penning Discharges", Phys. Rev. Lett. 13(1), 5 (1964). 213. Hirsch, E.H., "On the Mechanism of the Penning Discharge", Brit. J. Appl. Phys. 15, 1535 (1964).
192
SPUTI'ER-ION PUMPING
References 214. Knauer, W., U.S. Patent No. 3,216,652, "Ionic Vacuum Pump", filed 9/10/62, awarded 11/9/65. 215. Hirsch, E.H., "Excess Energy Electrons", Brit J. Appl. Phys. 15, 909 (1964). 216. Schuurman, W., "Investigation of a Low-Pressure Penning Discharge", Ph.D. Thesis, Rijksuniversiteit Te Utrecht, March 21, 1966. 217. Hobson, J.P. and Redhead, P.A., "Operation of an Inverted-Magnetron Gauge in the Pressure Range 10" 3 to 10" 1 2 mmHg", Can J. Phys. 36, 271 (1958). 218. Hartwig, H. and Kouptsidis, J.S., "A New Approach to Evaluate Sputter-Ion Pump Characteristics", J. Vac. Sci. Technol. 11(6), 1154 (1974). 219. Wutz, M., "Getter-Ion Pumps of the Magnetron Type and an Attempted Interpretation of the Discharge Mechanism", Vacuum 19(1), 1 (1969). 220. Brillouin, L., "Practical Results from Theoretical Studies of Magnetrons", Proc. I.R.E., 216 (1944). 221. Malmberg, J.H. and Driscoll, C.F., "Long-Time Containment of a Pure Electron Plasma", Phys. Rev. Lett. 44(10), 654 (1980). 222. Glass, R.C., Sims, G.D., Stainsby, A.G., "Noise in Cut-Off Magnetrons", Proc. Inst. Elec. Eng. (London), B102, 81 (1955). 223. de Chernatony, L. and Craig, R.D., "The Discharge Mechanism in a Magnetic Ion Pump", Vacuum 19(9), 393 (1969). 224. Redhead, P.A., "The Townsend Discharge in a Coaxial Diode with Axial Magnetic Field", Can. J. Phys. 36(8), 255 (1958). 225. Andersen, H.H. and Ziegler, J.F., Hydrogen Stopping Powers and Ranges in All Elements, Vol. 3 of Stopping and Ranges of Ions in Matter (Pergamon Press, New York, 1977). 226. Lassiter, W.S., "Extension of Measurements of the Striking Characteristics of the Magnetron Gage into Ultrahigh Vacuum", Proc. 14th Nat. AVS Symp., 1967 (Herbrick and Held Printing Company, Pittsburgh, 1968), p. 45. 227. Knauer, W. and Stack, E.R., "Alternative Ion Pump Configuration - Penning Discharge", Proc. 10th Nat. AVS Symp., 1963 (The Macmillan Company, New York, 1964), p. 180. 228. Henderson,W.G. and Mark, J.T., U.S. Patent No. 3,540,812, "Sputter Ion Pump", filed 4/12/68, awarded 11/17/70. 229. Bryant, P.J., "On the Use of a Penning Cell for Pressure Measurements", Proc. 14th Nat. AVS Symp., 1967 (Herbrick and Held Printing Company, Pittsburgh, 1968), p. 55. 230. Bryant, P.J. and Gosselin, C.M., "Response of the Trigger Discharge Gauge", J. Vac. Sci. Technol. 3(6), 350 (1966). 231. Jepsen, R.L., U.S. Patent No. 3,174,069, "Magnetically Confined Glow Discharge Apparatus", filed 11/29/61, awarded 3/16/65. 232. Jepsen, R.L., U.S. Patcnt 3,028,071, "Glow Discharge Apparatus", filed 3/6/59, awarded 4/3/62. 233. Reid, R.J. and Trickett, B.A., "Optimization of Distributed Ion Pumps for the Daresbury Synchrotron Radiation Source", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. Rfidcnauer, F.P. Viehb6 ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 89. 234. Hartwig, H. and Kouptsidis, J.S., "A New Approach for Computing Diode Sputter-Ion Pump Characteristics", J. Vac. Sci. Technol. 11(6), 1154 (1974).
SPU'Iq'ER-ION PUMPING
193
References 235. Bance, U.R. and Craig, R.D., "Some Characteristics of Triode Ion Pumps", Vacuum 16(12), 647 (1966). 236. Kelly, J.E. and Vanderslice, T.A., "Pumping of Hydrocarbons by Ion Pumps", Vacuum 11(4), 205 (1961). 237. Lichtman, D., "Hydrocarbon Formation in Ion Pumps", J. Vac. Sci. Technol. 1(1), 23 (1964). 238. Reich, G., "Investigation of Titanium Sheets for Sputter-Ion Pumps", Supplememo A1Nuovo Cimento 1(2), 487 (1963). 239. Jepsen, R.L. and Francis, A.B., "Interactions Between Ionizing Discharges and Getter Films", Supplemento AI Nuovo Cimento 1(2), 694 (1963). 240. Sheraton, W.A., "Contamination in Sputter Ion Pumped Systems", Vacuum 15(12), 577 (1965). 241. Lloyd, W.A. and Zaphiropoulos, R., U.S. Patent No. 3,042,824, "Improved Vacuum Pumps", filed 6/22/60, issued 7/3/62. 242. Staph, H.E., "Protection of Electrical Components from Damage by Arcing or Plasma Discharges in Vacuum Equipment", Vacuum 15(11), 545 (1965). 243. Wear, K., "By-Products Emitted from Sputter-Ion Pumps", R&D Magazine, February, 1967. 244. Bell, P.R., Moore, E.C., Wyrick, A.J., "Starting Sputter-Ion Pumps and the Outgassing of Wet Metal Surfaces", Vacuum 17(2), 87 (1967). 245. Neal, R.B., "Adsorption Pumping of the Two-Mile Accelerator, Stanford Linear Accelerator Center, Stanford, California, Report No. SLAC TN-67-77, 1962. 246. Bance, U.R., "Development of a diode/triode (D/T) ion pump", Vacuum 40(5), 457(1990). 247. Welch, K.M., Pate, D., Todd, R.J., "Pumping of Hydrogen and Helium by Sputter-ion Pumps. Part I. Helium Pumping", J. Vac. Sci. Technol. A 12(3), 861(1994). 248. Welch, K.M., Pate, D., Todd, R.J., "Pumping of Hydrogen and Helium by Sputter-Ion Pumps, Part II. Hydrogen Pumping", J. Vac. Sci. Technol. A 11(4), 1607(1993). 249. Brothers, C.F., J. Vac. Sci. Technol. 5, 208(1968). 250. Milleron, N., Reinath, F.S., Transactions of the 9th National American Vacuum Society Sumposium (Macmillan, New York, 1962), p. 356. 251. Rozgonyi, G.A., J. Vac. Sci. Technol. 3, 187(1966). 252. The author is indepted to to Mauro Audi of the Torino Varian Vacuum Division for supplying him with these data. 253. Smart, L., et al., "RHIC Vacuum Instrumentation and Control System", Proceedings of the IEEE 1999 Conference on Particle Accelerators, p. 1348. 254. Smart, L., et al., "Enhanced Ignition of Cold Cathode Gauges Through Use of Radioactive Isotopes", J. Vac. Sci. Technol. 14(3), 1288(1996). 255. Rutherford, S.L., U.S. Patent No. 6,004,104, "Cathode Structure for Sputter-Ion Pump", filed Jan. 14, 1997, awarded Dec. 21, 1999. 256. Sasaki, Y.T., "A Survey of Vacuum Material Cleaning Procedures: A Subcommittee Report of the American Vacuum Society Recommended Practices Committee", J. Vac. Sci. Technol. A 9(3), 2025(1991).
CHAPTER 3
TITANIUM SUBLIMATION PUMPING 3.0 Introduction Titanium sublimation pumping is called in the vernacular TSP pumping, or titanium sublimation pumping pumping (sic). In the same vein, titanium sublimation pumps are referred to as TSP pumps. Oh well, I won't attempt to change tradition. TSP pumping is used extensively in UHV applications. Such pumping is a form of chemical getter pumping, or chemisorption pumping. Titanium is sublimed or evaporated onto a surface - the walls of the pump - where it getters gas. To getter a gas means to remove the gas from a system by chemisorption, or the formation of a stable chemical compound of the gas with some chemically active metal. The terms evaporation and sublimation are used synonymously in the literature, when referring to the operation of these pumps. The gettering of ~gases by various metals has been put to practical applications for over a century.(1 ) A survey of some of the earlier work on the properties of getters, including his own substantial work, is given by Wagener.(2 ) Stout conducted one of the early investigations of the sorption of various gases by heated, nonsubliming, Ti filaments.(1 ) A more recent review of the subject of gettering was done by Giorgi, Ferrario and Storey.(S ) Also, papers have been published on the gettering properties of large-scale, sublimed materials other than titanium. For example, Hayward( 4 ) and Okano( s ) have reported on the pumping of sublimed Ta films, Prevot and Sledziewski( 6 ) on the effectiveness of sublimed TiO films in pumping gases, including Ar(!), Haque( r ) the pumping effectiveness of simultaneously sublimed films of Ni and Ti, and Nazarov's( 8 ) pumping comparison studies between sublimed Cr and Ti films. All evidence to date suggests that, when sublimation pumping in large systems (i.e., vs. gettering in sealed-off devices, such as microwave tubes), the most practical and effective material proves to be Ti. For reasons other than chemisorption properties, alloys of Ti and refractory materials are often used.(9,1 0 ) However, Ti is the important, chemically active component. Therefore, this chapter will deal exclusively with the properties and applications of TSP pumping. There have been numerous clever schemes used to evaporate or sublime Ti onto surfaces. I will describe some of these later. Note that many of the original electro-static ion pumps (e.g., see Appendix 2A) were primarily clever methods of dispensing Ti onto pumping surfaces. Granted, ionic pumping plays an important role in these devices. However, the high speed observed for the chemically active gases stems primarily from the Ti sublimed, or otherwise spewed onto the walls of the pump. Because of this, Herb and his colleagues might be viewed as the founders of TSP pumping.(1 1 ) To understand TSP pumping, we must first have an appreciation for the concept of the sticking coefficient of a gas.
196
TITANIUM SUBLIMATION PUMPING
3.1 Sticking Coefficients The sticking coefficient, o~, of a gas molecule is merely a measure of the probability that the molecule, when landing on some surface, will permanently stick. Therefore, it is the measure of the effectiveness with which the pump (or pumping surface) removes gas from the system. The concept is simple; but, as noted in numerous publications, including the def'mitive work of Hayward and Taylor,(1 2 ) later work of Harra,(S 6 ) and the earlier work of Giorgi and Pisani,(1 3 ) the measurement of sticking coefficient values is difficult, and fraught with potential error. Fortunately, most of us need not make such measurements. However, we do need to understand the concept and to interpret the data at our disposal. In Chap. 1, I noted that the rate at which gas molecules impinge on a surface is given by:
where, and
v
= k 1 P molecules/sec-cm2,
kl P M T
= = = =
(3.1.1)
3.51 x 10 2 2/(MT) ~ , the pressure in Torr, the molecular weight of the gas, the temperature of the gas in * K.
Further, in Chap, 1, I defined the concept of conductance as a convenient mathematical model for describing the behavior of gas in a linear (i.e., molecular flow) system. For example, the conductance of an aperture was discussed in Section 1.12.1. We noted in Fig. 1.12.1b, that the rate at which gas was conducted from Manifold B to Manifold A, QB, was directly proportional to the pressure in Manifold B and the area of the aperture separating the two manifolds. Or,
where, and
QB
= Ca PB
Ca PB
= the conductance of the aperture in Z/sec, = the pressure in Manifold B.
(3.1.2)
If the pressure in Manifold A is maintained at aperfect vacuum, no gas molecules will pass from Manifold A into Manifold B. Therefore, in this case, every molecule which impinges on the area defined by the aperture will be permanently removed from the vacuum system of Manifold B. Now, imagine that the aperture of Fig. 1.12.1b is covered over with a metal plate of the exact same area. Assume that the metal plate is somehow cooled to a temperature near absolute zero. If, for example, the gas in Manifold B is N2, every gas molecule which impinges on the cooled surface will permanently stick to the plate. This corresponds to a unity sticking coefficient for the gas and it is permanently removed from the system. There is no difference in the pressure in Manifold B over the case when there was a perfect vacuum in Manifold A and gas could pass through the aperture. In the former case, the rate at which gas was removed from the system depended on the conductance (i.e., area) of the aperture; in the latter case, it depends on the area of the cooled surface. Where the aperture is said to have a conductance, the cold plate is said to have a surface conductance. If all of the gas which impinges on it sticks, making the plate even colder will have no benefit in terms of the further pumping of gas. Therefore, the speed of the surface pump is limited only by the area of the cold plate. If we wished to increase the speed of the
TITANIUM SUBLIMATION PUMPING
197
surface pump, we would have to increase its area. Therefore, the speed of this surface pump is said to be surface conductance limited. We must keep in mind that the conductance of an aperture or pumping surface differs with the temperature and species of gas (i.e., see (1.12.7)). A similar situation exists when gas is chemisorption pumped on surfaces. We noted in Section 2.4.1, that when a gas molecule, such as N2, lands on a surface on which resides a metal with which it is chemically active, the N ~ molecule will perhaps stick, dissociate and form a stable chemical compound with that metal. It is reasonable to assume that: 1) there is a relationship between the number of active metal energy sites per unit surface area (defined in Section 2.4.1) and the probability that a molecule, when landing on the surface, will stick and form a chemical compound; and, 2) the sticking probability (coefficient) will diminish with time if the sites are not in some way replenished, but the flux of gas impinging on the surface persists. Both assumptions are correct. For example, for one gas-metal system - this term is often used in the literature when noting the reaction of a specific gas with a specific metal - the sticking coefficient, ct, might vary from an initial value of unity and, with additional gas exposure, decay to zero. Therefore, though the material sublimed on the surface may initially have very effectively removed gas from the system, in time the sticking coefficient of the gas on the pumping surface will decay to zero. Understanding of the dynamics of maintaining a balance between the incident flux of gas and the availability of chemisorption sites is what TSP pumping is all about. Two experimental methods were proposed by .Clausing for determining the sticking coefficients of various gases on Ti films.(1 0 ) The approach used exploits the anisotropic state created in a vacuum system when some or all of the surfaces in the system have nonzero sticking coefficients for a gas (e.g., see the work of Giorgi and Pisani(1 a )). An anisotroptic state in a vacuum system merely means that the flux of gas in the system is not equal and uniform in all directions. Clausing's experimental technique, in somewhat refined form, was used by Harra some time later to determine the precise sticking coefficients of N2 on Ti films.(1 4 ) The method, as reported by Clausing and Harra, is given below. VACUUM V E S S
,#~o
/¢/
/7
E
""'~
\
~
F
U
I
I
/
/
"e
3.1.1.
i
Ii/"
~
Figure
GAS SOURCE E R )
S
Apparatus
~
~ ~ e
~ \
for
~.~
~
Gas rLUX~,g
~,
REFLECTED ON PRIOR
-, ~ \.
GAS FLUX WITH /// ONLY RADIAL --~ COMPONENT lee
achieving,
lee , a r a d i a l
flux
~
V~
,.,,..
///
~../J"
of gas.
(14)
Figure 3.1.1 represents a spherical vacuum vessel. Gas is introduced into the vessel through a diffuser located at the center of the chamber. The purpose of the diffuser
198
TITANIUM SUBLIMATION PUMPING
is to ensure that all gas entering the chamber is dispersed in a radially uniform fashion to the walls of the chamber, as though originating from a point source located at the center of the chamber. The rate at which gas enters the chamber is measured with some form of throughput, or Q-meter. The flux density of entering gas molecules impinging on the walls of the p u m p (i.e., chamber), v e, is proportional to Q / Ac, or, v e oc Q / A c Torr-Z/sec-cm 2, (3.1.3)
a Ac
= the gas throughput entering the chamber, Torr-Z/sec, = the area of the chamber in cm2.
Knowing the value of Q, or rate at which gas enters the vessel, we can solve for v e by using (1.13.3), or
where,
ve
= k2 Q molecules/sec-cm 2 ,
k2 NO fl T
= = = =
(3.1.4)
N o / ~ TA c, 6.02 x 10 2 a molecules per mole of gas, 62.36 Torr-Z/mole * K, the temperature in * K.
Assume that we have sublimed a fresh film of Ti on the inner surface of the chamber. On introducing a gas, such as N 2, through the diffuser, some of the molecules of the gas flux, v e, will stick to the chamber on their first encounter with the wall. However, this is not a perfect pump, so some of the gas is reflected back off the walls. This gas is reflected away from the walls according to the cosine law (i.e., see (2.4.3a)). In the second and subsequent encounters this reflected gas has with the walls, the angle of incidence on impact can vary from 0 to 7t/2 radians. Therefore, the total flux density of gas incident on the walls at any time (i.e., the total number of molecules striking the wall per sec-cm 2 ), v t, is:
where
vt
= v e + Vg,
ve
= k 2 Q molecules/sec-cm 2, which impinge .I. to the walls, = the subsequent impingement rate of molecules, not sorbed on the first wall collision, per sec-cm 2.
vg
(3.1.5)
From the definition of sticking coefficient, c~, given at the beginning of Section 3.1, clearly: c~
= v e / v t"
(3.1.6)
The key to determining the value of o~ is to be able to measure any two of the three components of (3.1.5). For example, if we measure v t and v g, we can solve for v e using (3.1.5) and express the sticking coefficient with o~
= (vt-Vg)/V
t = 1 - V g / V t.
(3.1.7)
On the other hand, if we were to able to measure only v e and v g, we could solve for the value of the sticking coefficient by substituting for v t, found with (3.1.5), into (3.1.6), and we find: o~
= ~,e / (v e + Vg)
TITANIUM SUBLIMATION PUMPING = (1 +
Vg/Ve)'l.
199
(3.1.8)
The above analysis involves only simple algebra and the counting of molecules which we learned how to do in Chap. 1. The difficulty arises when attempting to experimentally differentiate or separately identify two of the three flux components. Gas beaming effects can lead to large experimental errors (e.g., see the work of Kornelsen and Domeij(1 s )). Clausing was the first to propose methods for the separation of the components.(1 0 ) Harra and Hayward described an apparatus of similar nature, but with added features.(1 4 ) This latter apparatus is represented in Fig. 3.1.2. The RGA (residual gas analyzer) shown was used to measure the rate at which Ti vapor is deposited on the chamber wall. It is calibrated by measuring ion current (i.e., Ti +, or amu 48), and by subsequently normalizing changes in ion current, with time, to the subsequently weighed Ti source (e.g., a filament). Baffle "A" serves to shield the BAG1 (Bayard-Alpert Gauge #1) from the line-of-sight gas flux from the diffuser (i.e., v e) and Ti vapors from the source. Therefore, BAG1 actually affords an indication of gas stemming from v g, as defined by (3.1.5). Baffle "B" shields BAG2 from Ti vapors from the source, but not from gas emitted from the diffuser. Therefore, measurements made with BAG2 are an indication of both components v e and v g (i.e., v t)" GAS "Q- ETER"
Ii
I BAG 2 QRGA
BAG
Figure 3.1.2. Apparatus used by Harra and Hayward to measure sticking coefficients of nitrogen on titanium films.(14)
VACUUM x~N "A" VESSEL ~ ~ . . ~ . ~
DIFFUSER ~ / ~ ~ TITANIUM FILAMENT
AUXILIARY ~ PUMP
Assume that both gauges are calibrated to read absolute pressure. Also, assume that they have negligible pumping speed for the test gas. This latter assumption implies that every gas molecule passing through the apertures and into the antechambers housing the gauges, later leaves these antechambers. Assume that the apertures leading into the gauge antechambers both have areas A a cm2. With this apparatus, we have two methods at our disposal for determining the values of ct in time. First, we can use the Q-meter to directly establish v e at any instant in time. Then, using the indicated pressure, P 1 of BAG1, we can determine the value of v g with (3.1.1). Then, using (3.1.8) we find: o:
-
(1 + kl P1 A a / k 2 Q)-I.
(3.1.9)
On the other hand, we can measure implied values of v g and v t by the insertion of the reading of BAG1 and BAG2 respectively, into (3.1.1), and determine the value of
200
TITANIUM SUBLIMATION PUMPING
a using (3.1.7). This reduces to: a
= (1 - P 1 / P2 )"
(3.1.10)
Using methods similar to those described above, the sticking coefficients for various gases on room temperature and LN~ (liquid nitrogen) cooled surfaces have been studied by a number of researchers. Harra has published an excellent review paper on this subject.(x e ) In a recent recapping of published data. Grigorov proposes slightly different interpretations of Harra's summary findings.( 1 7 ) However, prior to discussing some of these specific results, I will first discuss some of the general results and how they are applicable to building and using a TSP pump.
3.2 Pump Speed vs. Sticking Coefficient The components of a pump are quite simple and include only three elements, as depicted in Fig 3.2.1: 1) a source from which titanium is sublimed or evaporated through heating; 2) a power supply to heat the titanium; and 3) a surface onto which the Ti is sublimed and is accessible to the arriving, chemically active gas. Liquid nitrogen (LN~) is sometimes used to cool the pumping surfaces. However, for the moment, let us neglect the possible beneficial effects of such cooling. PUMP FLANGE A2"rACHED TO
/7 ,'!
L':,
TITANIUM BAFFLES
"-
'~
/
~ ' ~ ~
FILAMENT / "~-- HOLDER J
',,
i~':"
-~ --\\
{USUALLY HOLDING 3 - 4 FILAMENTS)
~
PUMPING
/ / \ ~'' .,~.~. . .'. , ~/ (~ SURFACE-"-. 1000 cm z ) "-..
WELDED 30 em ~
STN. STL. DOMES 3.2.1.
/
VACUUM GAUGE
FILAMENT 7
~...
Figure
/
~:\\~ f
Example
\'~
of a t i t a n i u m
~:,"
~.:>"
sublimation
SPUTTER-ION PUMP PORT pumping
_~
system.
There are two procedures used to deposit Ti when operating a TSP pump. Procedure #1: Ti is periodically sublimed onto the pumping surface, and the film is allowed to saturate with gas over a period of time. This is called batch sublimation. If the Q into the system is constant, as the film saturates (i.e., a diminishes), the pressure will rise. Therefore, when the pressure approaches a value which we do not wish to exceed, we will have to sublime additional material onto the surface to restore the original sticking coefficient, etc. The time between batch sublimations will depend on the pressure, the thickness and temperature of the film, and, of
TITANIUM SUBLIMATION PUMPING
201
course, the gas species. For example, at a starting pressure of 10" t x Torr, it might take days, or even weeks before a film is saturated to the point where the pressure rises above the desired level. Procedure #2: Ti is continuously sublimed onto the surface, throughout the process, at a rate which for a given system throughput, Q, the pressure remains constant. That is, we replenish the pumping sites at a rate that is in perfect stoichiometric balance with the rate at which they are being occupied. Continuous deposition is usually used in high throughput applications. In either case the speed of the surface, S s, for a gas species "i" is simply:
where,
Ss
= ot i CiA c Z/sec,
ai
= the sticking coefficient for gas species 'T', = the aperture conductance of I cm 2 area for species 'T', = the total pumping area of the chamber, cm2.
ACi~
(3.2.1)
In Procedure #1, the batch deposition method, a i will change in time as the film becomes saturated. Such an effect is shown in the two curves given in Fig. 3.2.2, where the change in sticking coefficient, ai, is given for N2, on a thick deposit of Ti.(1 4 ) Curve "A" gives values of c~ for a batch deposited film of 8.0 x 101 ,L Ti atoms/cm 2 deposited onto a previously exposed base Ti film o f - 1 . 8 x 101 7 atoms/cm 2. o
10
f__ ~ × l _
0 --
10
-1
II
I
X XXX~
I II
x XK X X ~
oo
x o 0
--
% o
-
_
~
%
I z
×~ 10
10
Figure 3.2.2. Harra and Hayward's sticking coefficient, ~, data for Nz o n Ti. C u r v e "A" g i v e s v a l u e s o f cx for a "batch" l a y e r o f 8 . 0 x 10 TM Ti atoms/cm 2 deposited on a previously exposed base of ~1.7 x 1017 atoms/cmZ o f Ti. C u r v e "B" g i v e s v a l u e s o f =: f o r a fresh layer of 8.3 x 1014 atoms/cm z o f Ti d e p o s i t e d on an exposed base z . (14) o f ~ 1 0 1 8 Ti a t o m s / c m
× x
- CURVE "A"
o r,.)
-
×
-2
f,)
Z
I I'
x
o o
x x
-3
x
o o o
L) b-, rd~
< x
10
o o
-4
×
g o o
× ×
o
10
-5
1
I 11 1013
Iql
1014
SORBED NITROGEN
1
[ I 1
1015 -
molecules/cm
1016 z
Curve "B" shows values of a for a batch film of 8.3 x 10 x 4 Ti atoms/cm 2 deposited on a previously exposed base of Ti atoms --101 s Ti atoms/cm2. Rather than show the increase in pressure as a consequence of film saturation, it is customary to show the decrease in sticking coefficient as a function of the amount of gas pumped. We see from this figure that major changes in c~ occur with the amount of gas pumped, as well as the apparent thickness of the Ti substrate onto which the film is deposited.
202
TITANIUM SUBLIMATION PUMPING
For any given amount of gas pumped, the speed of the film for N2 is found by substituting the value of ~ for the amount of gas sorbed into (3.2.1). For example, assume that the area of the pump, Ac, is l0 s cm2, and we have thus far pumped 101 4 N2 molecules per cm 2. From Curve "A" of Fig. 3.2.2, we can determine that the value of cz has decayed to -0.18. If the chamber is at RT (room temperature), CN2 is -11.7 Z/sec-cm2. Therefore, the resultant instantaneous speed of the pump for N2 is -2110 Z/sec. We will later establish some approximate values of ct for the different gases. However, for now the troubling aspect of the result of Fig. 3.2.2 is that in order to predict the behavior of the pump, we must know something about the Ti film exposure up to that point in time. For high throughput, continuous deposition applications, data are usually presented in a form equivalent to that shown in Fig. 3.2.3.( 1 0,1 4, x s ) 10
0
-_A
LZE
t
l l
'1
I
Z 10-1
"D"
0
2; 1
-
i I l
o-2
F i g u r e 3.2.3. H a r r a a n d H a y w a r d ' s Nz s t i c k i n g c o e f f i c i e n t , oc d a t a . C u r v e "C" was t a k e n w i t h a c o n s t a n t s u b l i m a t i o n of Ti, w i t h i n c r e a s i n g p r e s s u r e . C u r v e "D" was t a k e n w i t h c o n s t a n t N2 t h r o u g h p u t a n d v a r y i n g Ti s u b l i m a t i o n r a t e . ( 1 4 )
<
l q,
q 10-a
I 10
I 11
-2
10
-i
1
I 11 10
0
Nz MOLECULES S O R B E D / s e c p e r Ti ATOM D E P O S I T E D / s e c
In this case, the sticking coefficient of N 2 on the Ti film is given as a function of the ratio of the rate at which gas molecules are sorbed to the rate of Ti deposition per cm 2 . This method of presenting data, as in the above batch case, also normalizes out pressure data in the presentation of results. Data of Curve "C" of Fig. 3.2.3 were taken with the sublimation rate held constant, where Curve "D" data were taken with the throughput held constant. Harra noted that with either a constant sublimation rate, R, with varying pressure, or constant pressure with varying sublimation rate, R, the sticking coefficient could be approximated by:
where and
cz
= czm (1 - k2 ka Q /R),
Ct
= = = -
Rm k2 ks
(3.2.2)
the maximum observed sticking coefficient, the Ti sublimation rate in atoms/sec-cm2, N,~/~ A.. T, a ~[ata "fi~ing" constant.
Fitting the data of Fig. 3.2.3, Harra found values of ct m ---0.49, k a ---1.8 for the Curve "C", the constant Ti sublimation rate data, where a m -0.87, k a -1.6 for the Curve "D", the constant leak rate, Q, data. Clausing reported on continuous sublimation data for a number of gases (e.g., H 2, D 2, N 2, CO and CO 2 ) on both RT and LN 2
TITANIUM SUBLIMATION PUMPING
203
surfaces.(Z o ) We will return to these and other results, and eventually establish
engineering values of k3 for the different cases (i.e., see Table 3.2.1). Some investigators have used measurements of the total pumping speed of a Ti film, sublimed onto a surface, to determine the sticking coefficients of gases.(X 9 ) In this case, speed per unit area is usually given as a function of time, and for the given gas. Assuming no measurement errors due to beaming effects, etc.,(1 2 ) then the total amount of gas pumped per unit area, up to time t, is simply fS(t)P(t)dt up to this time. When the speed, or sticking coefficient, decreases to zero, the fdm is said to have reached its capacity. Note that though the speed or sticking coefficient for a particular gas may be very high at the start, the rate at which it decreases with gas sorption is of equal importance. Therefore, in describing the effectiveness of a film in pumping a gas, we must quantify both the capacity of the film and the sticking coefficient or speed as a function of the amount of gas pumped. These are two distinct concepts. That is, we are dealing with both the rate of pumping as a function of the amount of gas pumped and the capacity of the film.
Titanium and Conductance Limited Operation Let us recast (3.2.1) into a more intuitively useable form. First, we know that the throughput into the chamber is Q = SsPc, where S s is given by (3.2.1) and Pc is the pressure in the pump chamber. If we substitute this value for Q in (3.2.2) and the resultant value of c~ into (3.2.1) and solve for S s, we find the general equation for the pumping speed for the gas species "i", as a function of pressure, sticking coefficient, and sublimation rate, is: Ssi
= (C~mi CiAc) / (1 + (k2 k3 C~miCiAcPci/R)),
(3.2.3)
where, the various terms are given in (3.2.1) and (3.2.2). This expression, though having a lot of terms, is actually quite simple and reduces in the limits to two important expressions. In the first case, when 1 ,> (k2 k3 C~mi C i A c Pci / R), or the rate of Ti sublimation is very high compared to the pressure in the pump, then, Ssi
~ (¢~mi Ci Ac).
(3.2.4)
In this case, the speed of the pump is independent of the pressure, and depends solely on the area of the pumping surface and, of course, the maximum sticking coefficient for the gas species "i". This mode of operation is called surface or conductance limited pumping. Remember, this is a surface conductance, having to do with the product C i x A c of (3.2.1), rather than the conductance of some manifold or aperture separating the pump from the chamber. When 1 ,~ (k2 k3 C~miC i A c Pci / R), or when the pump pressure is high compared to the rate of Ti sublimation, then, Ssi
~ R / k 2 k 3 Pci.
(3.2.5)
In this case, the speed of the pump is directly proportional to the sublimation rate and inversely proportional to the pressure. The implication here is that there will be a decrease in pump speed at the higher pressures, in this case limited by the rate of sublimation of Ti on the surface of the pump. This mode of operation is called titanium limited pumping. This difference in conductance vs. titanium limited pumping
204
TITANIUM SUBLIMATION PUMPING
is illustrated in Fig. 3.2.4. Nitrogen speed data, on RT surfaces, are shown for three different Ti sublimation rates. These data are averages of unreported results, using a special pumping configuration reported on by Harra.( ~ o ) l0
4
I
I Ill
~.~ o°
l
I II
I
a~
.........
~
10
-8
I
I
I 11
I
I I ~-
~'~
..... . 1.0 g i n / h r . --". _ _.----w SUBLIMATION /d~ ",~ \ RATE -/ . ~ -~ \ 0.5 g m / h r . "x. =. k SUBLIMATION ~ -,.~ -RATE -~ ., --
1_ . . . . . . . . . . . .
--"':_
SPUTTER-ION PUMP ONLY
~..
to x
I 1I I
~ l:sP PUMP
- _ ~ ' ~ 0.01 g m / h r . _'~ SUBLIMATIONS- ~ RATE i ~ I /*N~ TITANIUM ] LIMITED ~
03
'
I
SPUTTER-ION
I II 10
-7
I
_ /
I II 10
-6
"--,.
/// TITANIUM__~ LIMITED
1
i II 10
-5
i
I 1I
PUMP INLET FLANGE NITROGEN PRESSURE Figure 3.2.4. Nitrogen both titanium and
10
x
~-"'-. ".
-4
1
1 I l 10
-3
Torr
"combination" pumping speed conductance limited conditions.
for
D e p e n d e n c y of a on Gas, Film Thickness and Temperature We see from the findings of just Figures 3.2.2 and 3.2.3 that this business of sticking coefficients is quite involved, even for the gas N ~, on Ti, sublimed on room temperature surfaces. To date, there is much disagreement in the literature regarding absolute values of initial sticking coefficients, whether or not one gas poisons the f'dm in the pumping of another, the preferential displacement of one gas by another, etc. I will discuss some of these findings, and eventually establish some recommended engineering values of sticking coefficients and film capacities. The apparent sticking coefficient of a gas and the capacity of the film are known to markedly vary with the 1 thickness of the film at the time of exposure to the gas, ~ atio of the rate gas is pumped to the rate Ti is sublimed, surface temperature at-the time of film sublimation, i surface temperature at the time of gas sorption, method of film deposition (i.e., batch vs. continuous), gas species, i! Ti deposition rate, independent of the presence of gas, gas desorption and synthesis at the source (e.g., H 2, CH 4 ), partial pressures of gases at the time of sublimation, 1 contamination of the film for one gas by another gas, 11) dissociation and liberation of gases at the film surface, 12/effects of film annealing, and 13 time dependent effects (e.g., surface and bulk diffusion processes). Returning to Harra's data for N2 on Ti, the effects of film thickness, implied in the results of Fig. 3.2.2, are graphically illustrated in Fig. 3.2.5, where a summary of the initial, maximum sticking coefficient, am, is given throughout the sequence of twelve
TITANIUM SUBLIMATION PUMPING
205
experiments reported by Harra.( 1 4 ) Data shown here are for both batch and continuous sublimation runs. 1.0 i [.., z
I
1 I'1' "
CONTINUOUS DEPOSITION
0.8
1
0.6 o
e~ X
BAT(;H, DEPOSITION
Figure 3.2.5. Harra and Hay-ward's data for the variation in ~, the sticking coefficient of nitrogen on titanium, as a function of the cumulative thickness o f t i t a n i u m . (14)
0.4
z 0.2
v
x t
10
16
10
17
t
t
LO
18
CUMULATIVE TITANIUM DEPOSITION- a t o m s / c m 2
The implication of this figure is that the sticking coefficient for N2 appears to increase with film thickness. This effect has been reported elsewhere for N2 and other gases.(X 8,1 g, 2 1 ) This effect is believed to stem from increases in the surface roughness or effective area vs. the projected surface area of the film. Of course, for the gases H2 and D2, capacity of the films would vary with film thickness due to simple diffusion related considerations.(X o, 1 g, 2 x- 2 4 ) Such increases in effective area increase the number of pumping sites per projected surface area and thereby increase the effective value of c~. Results of Elsworth, et al. seem to disagree to some extent with Harra's findings, regarding the effect of RT pumping of N 2 vs. f'drn thickness. They indicate that after the first of many batch films were exposed to N 2, both c~m and film capacity remained the same thereafter.(1 g ) The implication in their discussions was that the first batch exposure gave depressed values of a m because of possible experimental errors stemming from outgassing from their Ti source. Others have observed this effect.(2 1 ) Both source outgassing and gas beaming effects will lead to errors in such measurements. Clausing observed the effects of film texture (i.e., high values of sticking coefficients) of films deposited under batch conditions, and in a high partial pressure of He.(1 0 ) He noted that when the film was sublimed under high partial pressures of He (e.g., 2 x 10-3 Tort, on a surface at 10" C), the films deposited in this manner "... have a velvety black appearance and are very poorly crystalized." Such films had higher sticking coefficients for the gases H2, N2 and CO than those films batch deposited under high vacuum conditions. This same effect - that is, an apparent improvement in sticking coefficients of films deposited in high partial pressures of He - was also reported on by Clausing for films deposited on LN2 cooled surfaces.(X 0 ) This enhancement of sticking coefficients on LN2 surfaces was also reported on the same year (i.e., 1%1) by Sweetman.(4 1 ) This was markedly apparent for the gas H 2, where Clausing reported the difference in a m was x 4 higher for Ti f'dms deposited under a high He pressure. In most applications, sublimation of films under high partial pressures of He is not practical. However, results of this experiment emphasized the importance of surface texture or roughness as it relates to values of a. In this same reference, Clausing reported significant increases in
206
TITANIUM SUBLIMATION PUMPING
sticking coefficients and capacities of Ti f'dms, deposited in high vacuum at near LN2 temperatures, over those deposited on RT surfaces, for all gases tested. This effect was also reported the same year (i.e., 1961) by Sweetman.( 4 1 ) This effect has been widely noted in the literature (e.g., see 9,1 T,2 s,2 e,4 ), though Gupta and Leck report it is significant only for the gases H 2 and N ~.( 2 1 Another outstanding early work on TSP pumping on surfaces at temperatures of-190" C and 20* C was reported by Elsworth, Holland and Laurenson.(1 9 ) They tested the sorptive capacity of Ti films for the gases H2, D2 and N2, at the two temperatures, and for varying film thickness. They expanded this work even further by measuring the sorptive capacities of Ti films at temperature Tm, which were evaporated into surfaces at temperature Te, where Tm was either -190" C or 20* C and Te was either -190" C or 20* C. Lastly, they measured the sorptive capacity of films at -190" C, which were deposited at -190" C and subsequently degassed at a temperature of 20* C, and they then remeasured the pumping speed of the f'dm at -190" C. Using their notation "T - e "/Tm " they determined for N2 1) under 20* C/20" C conditions, virtually irreversible N2 chemisorption occurred, even if the film was later baked to -170" C; 2) under -190" C/-190" C conditions, the equivalent of ---100 monolayers of N~ were sorbed at full capacity (one monolayer of N2 corresponds to - 6 x 101 4 molecules/cm 2); 3) under 20* C/-190" C conditions, the equivalent of--3 monolayers were sorbed at capacity; 4) under -190" C/-190" conditions, and the film was subsequently baked at 20* C, - 3 monolayers of N 2 were sorbed on reaching the capacity of the film. This and other work of Elsworth, et al., verified the findings of Clausing, wherein he noted it appeared the beneficial effect of films deposited at LN2 temperatures "... anneals out on warming to room temperature. ,,(10) Clausing also noted that even slight traces of O ~ or N 2 "... drastically reduced the value (of c~) obtained when hydrogen is sorbed onto a new deposit." Because of this latter effect, he submitted that what appeared to be a film annealing effect might be due to gas poisoning. However, Elsworth, et al., verified Clausing's findings of an actual film annealing effect. Also, Gupta and Leck later verified that film poison ing by other gases was also a real effect which could lead to experimental error when making sticking coefficient measurements.(2 1 ) The conclusions of Elsworth, et al., regarding the pumping of N~ on -190" C surfaces, included: 1) once the capacity of fresh film, deposited on a -190" C surface, is reached, the value of ct m was completely reversible on warming the surface to 20* C and then returning to -190" C; 2) the capacity of films once saturated at -190" C, when baked at 20* C under UHV conditions, and again cooled, was -0.03 of that of fresh f'dms deposited at -190" C. However, this latter capacity was still x300 greater than would be expected for simple physisorption on a cold metal surface. This led investigators to later speculate of the existence of an intermediate mechanism in the sorption on N 2 on cooled Ti surfaces.(1 8, ~ s ) That is, there is some process short of N2 + 2Ti - 2TiN (i.e., complete chemisorption), but stronger than physisorption. The findings of Elsworth, et al., regarding the pumping of H ~, were as follows: 1) for 20* C/20" C conditions, ct m ---0.01 (compared with c~m -0.1 for N2 ); 2) for -190" C/-190" C conditions, a m -0.4; 3) the capacity of a 350A film under 20* C/20" C conditions was ~; under 20* C/-190" C conditions it was --0.11~ ; and, under -190" C/-190" C conditions it was 7.4~', where ~ = 7.8 x 101 s H2 mole-
TITANIUM SUBLIMATION PUMPING
207
cules/cm 2 . Steinberg and Alger noted a very important effect relating to the poisoning of Ti films by one gas, and the subsequent pumping of H 2 and D 2.(2 a ) They noted that the bulk Ti of a thickly deposited film, which has been once contaminated on the surface by some other gas, can be made to effectively pump H2 and D2 if a thin overlay of fresh Ti is deposited on the contaminated layer. We noted in Chapter 2 that H2, on dissociation into atomic hydrogen, diffuses readily in many materials. Therefore, it is reasonable to assume that the fresh overlay of Ti produces energy sites which dissociate the H 2. On dissociation, the atomic hydrogen diffuses through the poisoning f'dm barrier of oxide or nitride and on into the bulk Ti deposit. This is precisely the pumping model proposed by Steinberg and Alger. Gup.ta and Leck considerably expanded on the work of Clausing and Elsworth, et al.(2 1 ) Their work was done under rigorously controlled UHV conditions and after outgassing and characterizing their Ti sources, including the filament and holder. They studied the Ti film sorption of a wide variety of gases on both RT and LN2 (i.e., 77* K) surfaces. They classified the gases under study into three groups: Group #1: the inert gases CH4 and Ar; Group #2: the intermediately chemically active gases H2, D2 and N2; and Group #3: the very chemically active gases 0 2 , C2 H2 (acetylene), CO, CO2 and H 2 0 . Results of sorption of most of these gases, under 300* K/3(D* K conditions, are given in Fig. 3.2.6. In each case, the deposited Ti film had a thickness equivalent to --- 101 6 Ti atoms/cm 2, each batch being deposited in a 100 second time-span. 10 0 -
/__ co,
coe&
Oz
~
/
\
-
I
z 10
]
[
CO J l
I
Figure 3.2.6. Sticking coefficients, ~, f o u n d by Gupta a n d Leek, for t i t a n i u m RT films of ~ 1 0 1 6 a t o m s / c m 2 " e a c h b a t c h was d e p o s i t e d in ~100 sec. (21)
¢0 0
z 10 ¢0 O9
I0
_-
10 13
l
llt
10 14
I
i Jl
Y
10 15
L li
10 16
GAS COVERAGE - m o l e c u l e s / c m 2
Group #1: Argon was not sorbed on Ti films, even at 77* K. For methane (i.e., CH4 ), sorption at RT was barely detectable, with values of a m --4 x 10-4 and c~ --10 -4 with a coverage of---10 -1 3 molecules/cm 2 " Values of the respective a ' s increased by -..x 10 in conditions of 300* K/77" K. The CH4 pumped on 77* K Ti f'llms, even if covered over with additional Ti prior to warming to -150" K, resulted in "copious" desorption of the CH 4 at this temperature. Group #2: Nitrogen results were similar to those reported by others. That is: 1) the values of c~m can vary as much as x 2 depending on film batch deposition rate. Note that results for N 2, in Fig. 3.2.5, seem to be bracketed by Harra's results shown in Fig. 3.2.2. 2) The values of a increased by --x 5 on Ti films on 77* K surfaces, over that of RT surfaces. However, it was not clear if the conditions of these experiments
208
TITANIUM SUBLIMATION PUMPING
were 300* K/77" K or 77* K/77" K. Group #3: 1) The values of a m for all of these gases ranged from 0.9 to 0.99 when batch deposited at RT. 2) Substrate temperature had only a second order effect on values of am, the value for CO varying the greatest, with c~'s of 0.8 and 0.98 on 373" K and 77* K surfaces, respectively. E n g i n e e r i n g V a l u e s for F i l m S p e e d s a n d C a p a c i t i e s Harra, in two review papers, summarized sticking coefficient data for several gases, pumped under both batch and continuous conditions, and on surfaces at 77, and 300* K.(X 6 ,a 6 ) This review paper comprised results and considerations of some 25 references reported in the literature, including his own work. Results of this summary are given in Table 3.2.1. Table 3 . 2 . 1 . " E n g i n e e r i n g " v a l u e s for m a x i m u m sticking coefficients and speed for v a r i o u s g a s e s o n Ti f i l m s a t 77 °C a n d --300 °C as s u m m a r i z e d by H a r r a . ( 1 6 , 3 6 ) TEST GAS
MAX. STICKING MAX. SPEED a COEFFICIENT- ? m L i t e r s / s e c - c m e
MAX. CAPACITY OF VALUES OF FILM - x 1015 m o l e c u l e s / c m 2 b CONSTANT k3 d
300 °K
77 °K
300°K
77°K
H2
0.06
0.4
2.6
17
8-230 c
7-70
1.1
Dz
0.1
0.2
3.1
6.2
6_11 c
H20
0.5
7.3
14.6
30
CO
0.7
0.95
8.2
11
5-23
50-160
1.0
Ne
0.3
0.7
3.5
8.2
0.3-12
3-60
Oe
0.8
1.0
8.7
11
24
C02
0.5
4.7
9.3
4-12
NOTES:
300 °K
77 °K
a) S p e e d c a l c u l a t e d b a s e d o n g a s ~t Wide v a r i a t i o n s d u e in p a r t to Wide v a r i a t i o n s d u e in p a r t to C o n s t a n t u s e d w i t h (3.2.2); i.e.,
283°K
77°K 0.7
1.8
1.9
1.0 (300 °K) 1.0
]
-
b e i n g a t RT. film roughness. bulk diffusion into film, for c o n t i n u o u s deposition.
These are very reasonable working numbers for use in the design of TSP pumped systems. The value of film speed per unit area, S s, given in the above table, was calculated using (3.2.1), and while assuming that the temperature of the gas was 293* K. In counting molecules, there is a limit to the number of gas molecules which can be pumped by a Ti film. Harra reported that for each atom of Ti, one molecule of H2, CO, 0 2 or CO2 can be oumped; for each molecule of either H 2 0 or N2, two atoms of Ti are required.(1 6 ) 3.3 S y n t h e s i s , D i s p l a c e m e n t a n d D i s s o c i a t i o n o f G a s e s In the early days of development of electrostatic ion pumps, and in the use of TSP pumps in combination with sputter-ion pumps, it was believed that the presence of an electrical discharge in some way activated sublimed films so as to create a synergistic effect of the combined pumping modes. That is, the apparent combined speed of the sputter-ion and TSP pumps was greater than that observed for each, when independently operated. In these measurements, total system pressure was used as the indication of speed. For example, assume a f'Lxed N2 throughput, Q x, is introduced into a system and that the sputter-ion pump has an N2 speed of S i and the TSP pump a speed of Ss, yielding pressures of Pi and Ps, respectively, when individu-
TITANIUM SUBLIMATION PUMPING
209
ally operated. If the combined speeds were the sum of the individual speed we calculate that: Psi
= PsPi / (Ps + Pi)
(3.3.1)
where Psi is the pressure observed with both pumps simultaneously operating. The problem is that in every case it was observed that combined speed Ssi > S i + S s. Therefore, it was concluded that some form of "activation" was occurring to the sublimed f'dm as the result of the discharge of the sputter-ion pump. The flaw in (3.3.1) was in assuming that the throughput of gas was constant under all three conditions. Indeed, the throughput of N2 or other experimental gas might have been constant, but the total throughput of gas into the system was not the same when operating the TSP pump. This pump, and other forms of pumping involving the spewing of Ti onto the walls of the pump, creates methane (CH 4 ). The background of CH4 only increases when using the TSP pump. The synthesized CH4 increases the indicated total pressure so as to lead to lower indicated speeds for the test gas. When the sputter-ion pump is operated simultaneously with the TSP pump, the sputter-ion pump pumps the methane generated at the TSP. Therefore, this indeed does prove to be a synergistic effect, but not for the reasons assumed. This was verified in a classic experiment by Francis and Jepsen in 1963.(2 T ) This synthesis of CH 4 was observed by Klopfer and Ermrich early in the development of these pumps.(2 8 ) Wagener, at the same time, reported observing high CH4 backgrounds above several types of getter films deposited from hot sources.(2 ) He also noted that though there was observable sorption of CH4 on Ti films, it was slight, and the film had negligible capacity for this gas. Holland showed that the production of methane was related to the content of carbon in the Ti source.(5 5 ) He noted an enrichment of carbon in a molten droplet, replenished by wire feed, as the material sublimed. Yet, he concluded that the CH4 and CH3 synthesized during the experiments was manufactured at the deposited film. Kuz'min reached this same conclusion in a later work.(9 ) Gupta and Leck were the first to prove that the CH 4 was synthesized on the hot Ti source, rather than on the surface of the film.(2 1 ) They accomplished this by introducing D 2 into the system subsequent to the batch sublimation of the film. They then noted a displacement of CH4 by the sorbed D 2, but no CD 4 (or, presumably CD..H. ^ y. where integers x + y = 3 or 4). Conversely, they noted the production of CD 4, and CDxHy, x + y = 4, when introducing D 2 into the system with the filament turned on. Halama showed that CH4 and H 2 were the limiting components in achieving pressures < 10" t o Torr in proton colliders.(2 9 ) Because of the poor pumping of CH 4 by Ti films, Halama noted that sputter-ion pumps must be used in combination with TSP pumps in UHV applications. Edwards devised a recipe for accelerating the outgassing of CH4 from sublimed films, for applications at pressures < 10" 1 o Torr.(3 o ) It essentially involved a post-sublimation, 100" C bake of the sublimed film. He hastened to advise us, in a paper submitted three months after the first, that the same benefit would not be achieved by subliming the Ti with the pump walls at elevated temperatures (i.e., during bakeout).( s 1 ) Chou and McCaferty expanded on Halama's work in characterizing the pumping of CH4 by sputter-ion pumps at very low pressures (i.e., see ref. 4 6, Chap. 2). Gupta and Leck reported on the displacement of one chemisorbed gas by a second.(2 1 ) By displacement, it is meant that one gas, previously chemisorbed on
210
TITANIUM SUBLIMATION PUMPING
the Ti film, is subsequently replaced by a second and the first is liberated into the vacuum system. These results are somewhat controversial. Harra sites some conflicting publications regarding this displacement phenomenon.(1 6 ) Nevertheless, Gupta and Leck submit that experimental evidence indicates that a displacement process preferentially occurs as shown in Table 3.3.1. In this table, 0 2 will displace all other gases noted; CO will displace all the gases but 0 2 ; H2 all gases but 0 2 and CO, etc. I f'md it difficult to conceive that a displacement process of the nature "O 2 + 2TiN -* 2TiO + N 2 (gas)", or "20 2 + 3TiN -* 2TiO 2 + TiN + N2 (gas)" occurs. However, it is possible that gas molecules either physisorbed or sorbed in some intermediate phase on the Ti surface, are subsequently displaced by another gas. Gupta and Leck coined this gas displacement phenomenon as a pump "memory" effect (similar to that noted with sputter-ion pumps, in Section 2.4.3). They did not report on the liberation of H 2 on the pumping of H 2 0 on RT, Ti f'dms, though they listed this as one of the test gases. PUMPED GAS
DISPLACED GAS CH4
CH4
N2
H2
CO
02
no
no
no
no
no
no
no
no
no
N2
yes
H2
yes
yes
CO
yes
yes
yes
02
yes
yes
yes
o(:~x
no
Table 3.3.1. Gupta and Leck's gas displacement and sticking coefficient d a t a f o r s e v e r a l g a s e s o n RT L i t a n i u m films. F o r e x a m p l e , N2 d i s p l a c e s CH4, H2 d i s p l a c e s N2 a n d CH4 , CO d i s p l a c e s CH4, N2 a n d H2 , e t c . (el)
yes
<1-03 0.3 0 . 0 5 0 . 8 5 0 . 9 5
In the process of chemisorption, most gases are first dissociated. For example, I noted that this is the method by which H 2 and N 2 are pumped. This also proves to be the case with water vapor. In this case, however, H 2 0 + Ti -* TiO + H 2 (gas). Of course this is not observed at LN2 temperatures, as the water vapor has unity sticking coefficient on surfaces at this temperature. However, Harra (and others whom he cited) observed that on RT surfaces the chemisorption of H 2 0 on continuously deposited f'dms results in the liberation of copious quantities of H 2 .(2 0 ) As long as the sublimation rate, R, is sufficient (i.e., the condition (3.2.4) prevails), the partial pressure of H2 remains low. However, when conditions of (3.2.5) begin to prevail, the partial pressure of H 2 starts to rise, as the O in H 2 0 wins in the competition for the limited amount of fresh Ti. The poor pumping of Ti films for H2 O, reported on earlier by Wagener, was probably due to film poisoning effects (i.e., this self-poisoning mechanism).(2 )
3.4 Sublimation Sources There are three primary sources used at this time for subliming Ti onto pumping surfaces: 1)filamentary sources; 2) radiantly heated sources; and 3) E-gun sources, where the Ti is heated as the result of bombardment with high energy electrons. These will be discussed in the order given.
TITANIUM SUBLIMATION PUMPING
211
Filamentary Sources Titanium has poor hot strength at temperatures needed for reasonable sublimation rates. Because of this, wires of this material had to be bundled together with a second refractory material which possessed good hot strength and supported the Ti at or near its melting temperature. Several material combinations have been used to accomplish this goal. Some of these schemes are illustrated in Fig. 3.4.1. Kornelsen used a twisted pair of 0.076 mm ~ W and 0.25 mm #, Ti wires wound over a 0.25 mm W mandrel.t3 3 ) Clausing used a scheme where he tightly overwound a 25 cm long, 4.3 mm $ Ta rod with one layer 0.76 mm $ Nb wire followed by windings of two layers of 0.89 mm ~, Ti wire.(1 0 ) The Ti alloyed with the Nb wire forming an alloy with a higher melting temperature. With this scheme, he was able to achieve Ti sublimation rates as high as -3.0 gm/M for up to four hours and with a power input of--1.0 - 1.5 kW. McCracken and Pashley mention an early Varian product consisting of a tn'filar overwinding of a 0.76 mm ,~, W core with two 0.76 mm $ Ti wires and a third 0.38 mm $ Mo wire.(3 4 ) In a review paper, Herb sites these and some of the other early variations of sublimators, including other filamentary sources.(3 s ) ~100 mm "BUNDLED" ~ FILAMENT{IO)
\ .__ \ T.~Sc~R~'X I ~
~
/
/
~
~,--~~6A_ 3_,. . . . ,~=t~t,~ - ~ OF ~88WI~X~ ...........
___ / ~ ~ UNDERLAYER / /i' ~axxx'zr~rrzrt~ ~ OF 0.75 m m - / - ~ ~ --~t~ Nb WIRE / ~ ..... ~ ...... ~
as
/
Figure 3.4.1. Varieties of reported by Clausing,(lO)
'
r/
I
~ __"
~
150 m m
-
-._-- . -- .- - .
~
I
f
" B u n d l e d " f i l a m e n t with 0.76 m m is W core w i t h a n o v e r w i n d i n g of two 0.76 m m ~ Ti f i l a m e n t s a n d one f i n a l 0.3fl m m l ~ Mo f i l a m e n t . ( 3 4 )
An 85% Ti, 15% Mo f i l a m e n t w i t h ~2.0 m m iS d e v e l o p e d by K u z ' m i n for p r e d i c t a b l e ' o p e r a t i o n a t t e m p e r a t u r e s e x c e e d i n g th~.t of p u r e Ti, or of " b u n d l e d " f i l a m e n t s . ( ) * C o n f i g u r a t i o n p r e s e n t l y in p o p u l a r use. "bundled" and McCracken,(34)
alloy filaments a n d K u z ' m i n . (9)
Reliability problems and inconsistent sublimation rates are pointed out in many papers dealing with these bundled sources. The primary problem was that if the proper input power was exceeded in the wire bundles, blobs of molten Ti would form, fall from the bundle, and hot spots would develop in the assembly. This and the unpredictability of the thermal contact of the Ti with the refractory metal conductor, led to wide variations in sublimation rates for the same input power. The technical breakthrough in these f'llamentary sources was made by Kuz'min, who developed and patented (in 1961) a sublimation filament comprising an alloy of Ti and Mo, with a 15 to 20% Mo content, by weight.(9 ) This T ~ o alloy had superior hot strength, and enabled operation of the filament at temperatures far exceeding the melting temperature of pure Ti. Using a 25 cm long, 2.0 mm ~ wire of this material, he was able to achieve a sublimation rate o f - 0 . 3 gm per hour with a current of--44 A and total power o f - 3 0 0 W. He stated that "For maintaining a constant evaporation rate within certain (unspecified) limits, it is sufficient to maintain a constant current through the evaporator." Publications in the following years took issue with the idea of constant current operation (i.e., vs. constant voltage or constant
212
TITANIUM SUBLIMATION PUMPING
power operation). The objective is to be able to predict the sublimation rate at all times during the life of the filament. This is prompted by a desire to get the best utilization of the available Ti, neither wasting Ti when sublimating, nor starving the pumping surface by operating Ti limited. As the Ti becomes depleted in the T ~ o alloy mixture, changes occur in the physical and electrical characteristics of the wire making correlation of sublimation rate vs. filament power, in time, difficult. Constant Current Operation McCracken and Pashley published the first study of sublimation rate characteristics of the 85% Ti, 15% Mo alloy fdaments with constant current usage.(a 4 ) The fdaments, manufactured by Imperial Metal Industries, Ltd., were 25 cm long, with a diameter of 2.1 mm. With constant current operation, they observed sublimation rates as shown in Fig. 3.4.2 for operation at a constant current of 50, 55 and 57 amperes. 3.0
' ~i ~
)-
~
I McCracken
57.0 A
'
I
I
("tick"
m a r k s are not necessarily reported d a t u m )
Figure
~ 2.0 I E--
a~ Z
--
1~ /
~83.7%Ti 74% Ti 16.3 % Mo 26% Mo McCracken ~ _ / ~ iA\ /_5. 5. ...0. / ~ , ~
E-~ -e 10"
Lawson & Woodward,48.9 A"
~83% Ti McCracken
m
50.0 A
Oq
0
3.4.2.
Sublimation rate differentiating McCracken's filament weight l o s s d a t a , (34) a n d c o m p a r e d with normalized Lawson and Woodward constant current sublimation r a t e data. (24) Where Lawson and Woodward maintained constant current in their filament, McCracken had to cycle his filaments f r o m RT to the "constant" c u r r e n t s e t t i n g s i n o r d e r to measure changes in weight.
data found by
A 0
1.0 TIME -
2.0
3.0
hours
Data of Fig. 3.4.2 were obtained by differentiating their published weight loss data with respect to time. Changes in sublimation rates observed in this figure were comparable, with exceptions noted below, to rates noted by Lawson and Woodward when operating a smaller diameter filament at the equivalent constant currents.(2 4 ) McCracken noted that during extended operation of the filaments, a point was reached where large crystals developed in the wire. Some of these adjoining crystals had slips or dislocations along their grain boundaries, some crossing the full diameter of the wire. However, they noted no failures as the result of this crystal growth, as long as the filament current was kept below a certain value (i.e., 57 A). Lawson and Woodward published the first definitive study of changes in the physical properties of these Ti/Mo filaments in time.(2 4 ) Their filaments measured ---1.6 mm (t' with a length of 21.7 cm. They studied changes in wires initially comprising 85% Ti, 15% Mo by weight. First, they noted that none of the Mo was sublimed from the f'dament. They noted that 40% of the Ti in the wire "can be evaporated", presumably on reaching the end of filament life. Also, when operating the filaments at a constant
TITANIUM SUBLIMATION PUMPING
213
voltage, 24% of the Ti can be evaporated at a constant, predictable rate. They indicated that the large crystal growth - which they termed a "macrocrystalline" structure - occurred on the wire reaching an alloy mixture of 74% Ti and 26% Mo. The onset of this structural transformation was independent of the evaporation temperature between the limits of 1550-1800" K. This result is somewhat confusing as if: 1) no Mo is sublimed from the filament, and, 2) the initial alloy mixture is 15% Mo and 85% Ti; then, 3) on reaching an alloy mixture of 26% Mo and 74% Ti through evaporation of the Ti, an alloy of 26% Mo and 74% Ti suggests that -50% of the Ti has been sublimed at the onset of the macrocrystalline state. However, the total mass of the wire has decreased by-42%. Lawson and Woodward noted increases in the total emissivity of - x 2 in these wires as a consequence of the filament's crystalographic transformation. These changes in emissivity resulted in decreases in filament temperature for the same power. Wires were clamped in a holder, similar to that shown in Fig. 3.2.1, but comprising "fairly heavy copper clamps"; the implication being that these heavy clamps imposed a low-temperature thermal boundary condition at each end of the wire. Therefore, the central part of the wire initially operates at the highest temperature. This portion of the wire loses Ti at the greatest rate and therefore is the first portion of the wire to undergo the crystalographic transformation. It then, because of its higher emissivity, operates cooler than neighboring portions of the wire which have not yet gone through the transformation. The regions of transformation "grow" from the center of the wire towards the clamps. This fortuitous transformation mechanism prevents the central portion of the wire from overheating and premature filament failure. They also noted slight increases in the resistivity of the filament material (i.e., - 9 % ) as a consequence of this metallurgical transformation. However, for constant current operation, changes in the alloy resistivity are of much less importance to evaporation rates than are increases in wire resistance stemming from a decrease in the wire diameter due to sublimation of the Ti. They showed sublimation rate data for filaments operated under both constant current and constant voltage conditions. Under constant current conditions, under early stages of sublimation, the sublimation rate gradually increases, and eventually reaches a value of --x 4 that of the initial value. Similar findings were noted by Chou and Lanni (--x 6) under constant current conditions.(a 2) This increase in sublimation rate stems from an increase in the power dissipated in the wire (i.e., constant current, but increasing wire resistance due to decreases in wire diameter). They too proposed that a peaking in sublimation rate was observed as additional Ti was sublimed, and the wire transformed over to the macrocrystalline state. They indicated that thereafter, the filament temperature decreased due to increases in emissivity, and the sublimation rate correspondingly declined. There is some disagreement in the published data regarding when, in the life of the filament, this maximum sublimation rate occurs. Chou and Lanni note, in their sublimation rate vs. time data, that the peaking in rate occurs after sublimation o f - 1 0 % of the usable Ti. Yet, they report that it occurs when the Mo content of the filament is -26% "by weight". Their data does not seem to indicate this is the case. For example, if one integrates their sublimation rate vs. time data, the expended Ti at filament failure is -1.2 gm. This closely agrees with my findings for such filaments. On the other hand, the peaking in sublimation rate in time, if occurring after sublimation of 10% of the Ti, one would expect that the alloy of the filament would be in the proportions of--16% Mo and 84% Ti, rather than the suggested 26% Mo.
214
TITANIUM SUBLIMATION PUMPING
Integrating similar constant current data reported by Lawson and Woodward, we fred that thepeaking in sublimation rate occurs after expending --50% of the total Ti sublimed, including sublimation results from increases in current late in the experiment. This peak corresponds to a filament alloy mixture corresponding to --26% Mo. Their results are also shown in Fig. 3.4.2. Lawson and Woodward's constant current data were normalized to McCracken's data by taking into account the differences in the lengths, diameters and areas of the respective filaments. Of course, the higher the sublimation rate with constant current operation, the sooner rate peala'ng occurs in time. However, I believe there is a second and equally important rate-effecting mechanism at play in these constant current experiments. That is, crystalographic changes in the wire occur as the consequence of temperature cyclimz. This effect has been reported in the literature with pure Ti sources.(4 2") For example, in order for both Chou and McCracken to make periodic measurements of weight changes in their constant current sublimation rate measurements, they had to cycle their f'daments from RT up to the high temperatures associated with their respective constant current settings. From his data, we speculate that McCracken probably cycled his filaments 5 to 10 times. Chou implied that he cycled his filaments of the order of 70 times throughout his measurements. It is known from Ti/Mo alloy phase diagrams given by Brewer and Lamoreaux( a ~ ) that Ti/Mo alloys go through a crystalographic phase transformation when cycled from RT to temperatures ranging from -600 to 900 C. A temperatures less than this, TFMo alloys have a mix of both bcc and hcp crystal structure. However, above this temperature range, the alloy becomes exclusively bcc. If the temperature changes occur very slowly, the atoms of the alloy will reorder themselves into the crystal structure appropriate to the particular temperature. However, if these temperature changes are made very quickly, the filament will become a hodgepodge mixture of both the bcc and hcp crystals. This has the effect of increasing the surface roughness or area of the filament. As a consequence, filaments which are cycled numerous times through the crystalographic transformation temperature eventually develop a very rough surface texture. This in turn results in an increase in the effective surface area (or emissivity), and a reduction in operating temperature, for the same input power. Lawson and Woodward's data were literally taken at a constant current (i.e., without cycling the filaments). Therefore, I suggest that Chou and Lanni mistakenly interpreted the peaking in their sublimation rate data as being due to Ti depletion, as observed by Lawson and Woodward, rather than stemming from increases in surface roughness as a consequence of temperature cycling.
Constant Voltage Operation Lawson and Woodward report that almost constant filament sublimation rates - in fact, slightly increasing rates - are achieved when operating filaments at constant voltage, until - 2 4 % of the total Ti is sublimed.(2 4 ) They did not cycle the filaments during their tests. I conducted a series of measurements on 2.0 mm ~, 85% Ti / 15% Mo filaments having lengths of ---17.5 cm. The usable filament length was of the order of 15 cm., as each end was clamped in a holder similar to that shown in Fig. 3.2.1. Sublimation rate measurements were made on a sample of six filaments. The filaments and holders were placed in a water cooled chamber, combination pumped with a 20 Z/sec sputter-ion pump. Two gauges were used to measure the indicated pressure.
TITANIUM SUBLIMATION PUMPING
215
One was optically baffled from the Ti vapors, the second was deliberately put in line-of-sight with the Ti vapors. The intensity of the Ti vapors, monitored by the unshielded gauge, was used to determine the relative sublimation rate of each filament, in time. The second gauge was used to make sure that actual chamber pressure was sufficiently low so that errors did not occur in the inferred Ti flux readings taken with the first gauge. Six f'daments, with dimensions as described above, were used in the following tests. All were weighed prior to the experiment and had approximate weights of -2.7 gm. All experiments were conducted with a constant voltage setting, sufficient so that the initial current drawn by the filaments was 50 A. All of the filaments were outgassed prior to sublimation measurements. The first three filaments were each operated for one hour, and then removed and weighed. One of the three was inadvertently operated at a constant current of 50 A for one hour. The other two were operated at an initial current of 50 A, which decreased to 48.5 A at the end of one hour of operation. The weight loss in the first filament corresponded to an initial equivalent sublimation rate, Ro, of 0.071 ± 0.017 gm/hour. 0.20
'
I
'
I
'
24% OF Ti USED °.
0.15
[.~
NORMALIZED CONSTANT VOLTAGE DATA OF LAWSON & W00DWARD (24)
-~-. :
-
.
0.10--
z
I!
o
FILAMENT CYCLED ~ 1 7 0 TIMES FROM
" /
~ - :.
ZERO VOLTAGE TO
~
VOLTAGE INITIALLY
-
~
Figure 3.4.3. Sublimation rate d a t a of L a w s o n a n d W o o d w a r d for a literally constant filament voltage, compared with Welch data where the filament was c y c l e d o n a n d off, b e t w e e n a voltage setting initially yielding 50 A c u r r e n t in the filament.
-
0.05
0
, 0 font:
1
t
10 0.1"TIME
I
20 _ hours
The remaining three f'daments were operated in a manner similar to that of actual use. The current was initially set at 50 A, the voltage thereafter held constant and the current allowed to decrease in time as the resistance of the filaments increased in time (see Fig. 3.4.3). However, the voltage was cycled, 10 minutes on and 20 minutes off, throughout the life of the filaments. The sublimation rate, R(t), throughout the life of the filaments followed the relationship:
R(t) where and
= Roe'at Ro a t
= 0.073 ± 0.017 grn / hour, = 0.0455 / hour, = time of sublimation in hours.
(3.4.1)
216
TITANIUM SUBLIMATION PUMPING
At end-of-life, the average weight loss of each f'dament was 1.16 - 0.01 gm, or --50% of the Ti in the wire had been sublimed. Of course, because of end effects, the high temperature portion of the filament had far greater depletion in Ti than this (i.e., -73%). Constant voltage results are given in Fig. 3.4.3. Lawson and Woodward's data are also shown in this figure, but are normalized to the larger diameter f'damerits of my tests. Note that when normalizing all of the above data, -5.0 cm was subtracted from the reported length of the fdament. This was done because of end effects induced by securing the filament to the holders. We must conclude from Fig. 3.4.3 that, as in the case of constant current operation, the cycling of the f'dament through a crystalographic phase transformation causes changes in the total emissivity of the filament. In my constant voltage experiments, the f'daments underwent -170 temperature cycles, where in the case of Lawson and Woodward the voltage was in fact literally held constant. Where Lawson and Woodward reported no decrease in sublimation rate up to a depletion of 24% of the total Ti, I observed sublimation rate decreases, in time, at the beginning of filament operation. Therefore, from the above constant voltage and constant current data, it is obvious that not only is the proportion of Ti and Mo in the f'dament important to predicting sublimation rates in time, but the number of temperature cycles through the crystalographic transformation temperature of the filament is of equal, if not greater importance. Because of the perplexing problems posed in predicting sublimation rates of these f'damentary sources in time, Kuznetsov(3 a ) used the electron emission of the f'dament as an inference of its temperature. A ref'mement of this technique was later reported on by Strubin.(3 9 ) Electron emission from a hot filament can be predicted by using what is called the Richardson-Dushman equation.(4 0 ) Presumably, by knowing the temperature of the filament, one can thereby predict the sublimation rate. As you have by now observed, knowing the temperature is necessary, but not sufficient to the prediction of filament sublimation rates. That is, the rate of Ti sublimation at a given temperature depends on both the Ti and Mo proportions and the number of temperature cycles the filament has undergone.
Radiantly Heated Sources These sources are heated by the thermal radiation of a secondary W filament. Herb and his colleagues were the first to report on this type of Ti source.( 3 s ) They reported on using a 0.5 mm ~ W coil, inserted into a 1.6 cm ~ tube with a length of 2.5 cm and a wall thickness of 3.3 mm. No sublimation rate data were given. Harra and Snouse reported on a refined radiantly heated source, which came to be known as the Varian TiBaU® source.(4 2 ) A cross section of this source and another similar source, developed by me, is shown in Fig. 3.4.4. In the case of both the TiBall ® and MiniTiBall ® sources, current is passed through a W filament, completely surrounded by the Ti ball, and with a return path through the Ti sphere. Part of the inner surface of the TiBall ® sphere is serrated for purposes of achieving a more uniform radiant heating of the Ti. The ball assembly is mounted on a flanged holder, similar to that used with filament sources. For example, two such TiBall ® sources could be coaxially positioned near the center of the cheveroned array depicted in Fig. 3.5.1, and each provide -35 gm of usable Ti, prior to venting the system for source replacement. The TiBall ® source was capable of providing sublimation rates of up to -0.6
TITANIUM SUBLIMATION PUMPING
217
gin/hour with a maximum power of 750 watts. The smaller source had less uniform sublimation rates, and required the same power needed to operate the standard fdament sources found in common use today (i.e., - 3 8 0 W).
~23.3 mm
T U N G S T E- -?~~ ~ ~ ~~~ , , , ~ I , ~
.E TZR
'
/
"Mini" TiBall ®
M0 XBDEN oU
-,
~~OLDER
STANDARD TiBall ® F i g u r e 3.4.4. Two v a r i e t i e s of r a d i a n t l y h e a t e d s u b l i m a t i o n s o u r c e s " t h e TiBall ®, d e v e l o p e d by H a r r a a n d S n o u s e ( 4 2 ) a n d a s m a l l e r , r a d i a n t l y h e a t e d s o u r c e d e v e l o p e d by Welch.
Sublimation rates as a function of sublimed Ti for these two sources are given in Fig. 3.4.5. End-of-life of these ball or bell sources more often than not occurs when the Ti wall thins down to the point where a window to the outside develops. This window may be only ---1.0 cm 2 in area, but this is sufficient that radiation losses from the W filament, through the window, make further sublimation of the Ti negligible. 0.8
I
Q
~
1
I
I
- ~ ~
0 )
i
10
I~ --15 GRAMS OF USEABLE TITANIUM I
20
1
t
l
STANDARD TIBALL ~SUBLIMATION RATE ~-'~"- S- AT MAX. POWER
0.4
0.2
1
--35 GRAMS 0FTITANIuMUSEABLE- ~
0.6
Z o
I
~
i
30
1 \
l
40
SUBLIMATION RATE O 50 g ("380 Watts)
50
l
60
I
70
I
80
I
90
100
TIME - hours F i g u r e 3.4.5. S u b l i m a t i o n r a t e s , in t i m e , of t h e s t a n d a r d a n d MiniTiBall ® r a d i a n t l y h e a t e d s u b l i m a t i o n sources.(42)
These types of sources have one major disadvantage. If they are not operated with some level of standby power, their life is significantly decreased. The standby power must be sufficient to keep the Ti source at a temperature in excess of --880* C.(4 2 ) This is due to the fact that distortion occurs in the Ti source as a consequence of temperature cycling it through a crystalographic phase transformation which occurs in pure Ti at --880* C. At temperatures ~880" C, Ti exists in an a or hcp phase. At temperatures >_880* C, it changes into a ~/or bcc crystal structure. The temperature
218
TITANIUM SUBLIMATION PUMPING
cycling of these pure Ti sources soon results in a knurled, distorted body of Ti. Martin encountered this distortion when using the TiBallO in a fusion-related program at Princeton.( ~ 6 ) Distortion usually leads to a shorting of the W f'dament to the Ti wall. Also, the very rough texture of the surface, as the result of the temperature cycling, results in a greatly increased surface area (or total effective emissivity). As a consequence of this increase in surface area, radiation losses increase, and sublimation rates diminish to unusable levels. I had the bright idea of making a source out of a B-stabilized material so as to minimize distortion of the bell (i.e., the MiniTiBall ~) on temperature cycling. I had a bell made of such an alloy. It worked magnificently. I was able to cycle the bell over 250 times from RT to operating temperature with only slight distortion of the configuration. Cycling of a pure Ti source a similar number of times led to failure due to shorting of the filament, and after only - 3 0 % Ti utilization. The problem with the new material was that though the bell survived the temperature cycles, the sublimed film wouldn't stick to surfaces including A2 2 0 a, stn. stl., Cu and glass. Film flakes, with a thickness of a only few/.t m, lay in piles on the bottom of the bell jar. This was due to residual stresses in the sublimed film. In the final analysis, no solution to this phase transformation distortion problem has been discovered. Therefore, to date these sources must be operated with standby power. Standby power, for the smaller source shown above, is -95 watts, and for the larger source -200 watts. Though there is negligible sublimation at temperatures stemming from these power levels (theoretically in the millions of years of life) heat from the source may be troublesome to extreme high vacuum applications such as storage rings. Also, in applications such as storage rings, operating at pressures in the 10" t o Torr region and lower, TSP users are less concerned with the availability of large quantities of Ti than they are concerned about the reliability of the source. Filamentary sources usually number 3 to 4 on one holder. This provides redundancy in the system in the event of the failure of any one f'dament. Because of these two considerations, radiantly heated sources have not found wide acceptance in particle accelerator applications.
E-Gun Sources Gould and Mandel, subsequent to a visit with Professor Herb at the University of Wisconsin, launched into an E-gun Ti source development program at Brookhaven National Laboratory.( 4 3 ) The Alternating Gradient Synchrotron at Brookhaven was once pumped with wire-fed electrostatic ion pumps. It was evident, when reading between the lines of publications dealing with these electrostatic pumps, that they had serious reliability problems. Gould and Mandel developed an E-gun Ti source so as to eliminate the wire-fed Ti sources. It comprised a simple Ti plug, mounted on a Ta rod within the center of the pump. The Ti plug was bombarded by -1.0 kV electrons emitted from a hot W filament, as shown in Fig. 3.4.6. Gould and Mandel noted at that time that "It is perfectly feasible to visualize a continuous feed rod (i.e., Ti source) ..., depending only on the life of whatever filament or multiple filaments W • t! ere mstalled. Gretz reported on an appendage pump having a similar configuration to Gould and Mandel's E-gun source.(5 a ) A variation of Gould and Manders proposed rod-fed source was reported on by Burthe and Munro, of the Ultek, Perkin-Elmer Corporation, only four years later.(4 4 ) Ti rods, with diameters of -3.18 cm and of varying lengths, could be
TITANIUM SUBLIMATION PUMPING
219
manually fed, as needed, into beams of electrons emitted from six or more short, heated W f'daments. A later variation of this E-gun source, depicted in Fig. 3.4.6, made use of a toroidally wound, heated W electron source. This later version provided more uniform heating of the Ti rod.
/ //'(~-/ .).3.
/ \\
J
(~) [ I .
I
~ HOT TUNGSTEN ~ CATHODE
SYSTEM FLANGE ~ l BOMBARDING
~ ] -
]-']
j
Figure 3.4.6. A form of "E-gun" t i t a n i u m sublimation source developed and sold by the Ultek, Perkin-Elrner Corporation. The need of these sources arose with the need for higher sublimation rates than could be reliably achieved by filaments. E-gun sources were initially developed for semiconductor and decorative and optical coating applications.( 4 5 ) The primary pumping application of these sources was in space simulation work. When pumping very large chambers with forms of combination pumping (see below), including TSP pumps, they were Ti limited. That is, even with multiple filament sources, they could not sublime enough Ti to take advantage of all of the available surface area. The above source could sublime Ti at a rate of up to 0.5 gm/hour, operated at less than one-third the power of the TiBall ®, had --55 gm of usable Ti, and did not require standby power. The Ti source rod was automatically advanced. This was controlled with a servo loop tied to the pervience of the electron gun (i.e., the filament emission to the Ti anode, for a given voltage). Because of the need for less power than the TiBall ®, Ultek advertised this source as the "cool sublimator". A second form of E-gun source, reported on by Robertson, was developed and sold by Varian Associates.(S 4 ) Rather than having a continuously fed source of Ti, the source comprised layers of Ti sheet, as depicted in Fig. 3.4.7. A heated W f'dament served as the source of electrons. This E-gun could sublime Ti at a rate of up to 7.0 gm/hour and provided --200 gm of usable Ti. It operated at up to 3.0 kV, and required 3.0 kW of beam power at the maximum sublimation rate. It was bakeable to 450* C and did not require water cooling. For reasons noted below, these high sublimation rate, E-gun sources had rather limited use in pumping applications. Their present advantage is that, with the appropriate pumping arrays, they can provide pumping speeds of 50,000 - 200,000 Z/sec in very hostile environments (e.g., high ionizing radiation fields).
220
TITANIUM SUBLIMATION PUMPING
r3m==~=m~mtmm~------~HOT TUNGSTEN "CATHODE FILAMENT I" ~- . # II ".~
----£ 7 ]1'
CERAMIC PEDESTAL
%
,i i ; ~/~1 ' , ~
~ ~
&~
WITHa TITANIUStROD
I I X.~REGION "FLOATING" II
Fz ure 347 A form of E un "sgublimabt~oVasO22eA,ssdoei~op(d41'ant~tasnil~ m
[ ~
/J
FLANGETO SYSTEM (~152 mm ~ )
3.5 Combination Pumping As is now evident, some gases are not chemically pumped by a deposited layer of Ti. Also, some gases are synthesized on the hot filament, or liberated from the Ti film due to molecular dissociation. Because of this, various forms of c o m b i n a t i o n pumping are used to pump these by-products of TSP pumping, and other gases such as He and Ar which are present in many vacuum systems. Three forms of combination pumping are often integrated into one system. These are cryopumping on LN2 cooled surfaces (primarily to pump H 2 O), TSP pumping, and sputter-ion pumping. An example of the configuration of such a system is shown in Fig. 3.5.1. In this system, the TSP pump affords speeds of the order of many thousands of liters per second for chemically active gases, the sputter-ion pump pumps gases including the CH4 and the inert gases, and the LN 2 surfaces pump the H 2 0 . Pumping systems, having configurations similar to that shown above, were developed primarily for thin f'dm deposition work in the semiconductor industry. As chip component densities became greater and line-widths smaller, the effects of contamination from the pumping systems became more significant. Diffusion pumped systems have problems of oil backstreaming. Also, if such systems are inadvertently let to air while the main isolation valve is open to the semiconductor deposition chamber - a process referred to as being d u m p e d - this could result in system down time of weeks, and serious damage of the work in process. Because of this, use of diffusion pumped systems became less popular and systems such as shown in Fig. 3.5.1 came into brief prominence. Of course, even these systems became less popular with the advent of the gaseous helium cryopump. Note that the system shown in Fig. 3.5.1 is rough pumped with sorption pumps. This too was a requirement of many of the semiconductor manufacturers. There was concern that possible oil backstreaming from mechanical roughing pumps might also
TITANIUM SUBLIMATION PUMPING
221
pose a threat of system contamination. Essentially all of the UHV surface science equipment (e.g., Auger Electron Scanning systems) presently marketed use some form of TSP and sputter-ion pumping. A load lock or preparation chamber may be rough pumped with a cryopump of some form, but the UHV work is done in a chamber pumped with TSP (usually on RT surfaces) and sputter-ion pumps. BELL-JAR BASE-PLATE t ~ - ~ WATER-COOLED, FINNED ARRAY, TRANSPARENT TO
GAS BUT A BAFFLE TO SUBLIMED TITANIUM __~
LN
~
--
F---~
~-. I/
P
2 PUMPING SURFACE I
'
~-"7-~ ]
l
.•••0' I
V-~ F
VACUUM GAUGE
-RING SEAL TO PERMIT REMOVALOF
1
CRYOSORPTION ROUGH
PUMPS
/ SPUTTER-ION PUMPS IN WELDED "POCKETS"
Figure 3.5.1. A c o m b i n a t i o n p u m p c o n f i g u r a t i o n , r e p o r t e d on by Harra, and m a k i n g use of TSP, s p u t t e r - i o n and cryopumping.(Zo)
As an interim measure, TSP and sputter-ion pumping was also used in space simulation systems.(6,4 4,4 6,4 r ) This was always augmented with LN 2 pumping. This combination pumping was also prompted by oil back-streaming problems from diffusion pumps, which caused the contamination of work pieces. However, these large space simulation chambers are now almost exclusively pumped with closed loop, gaseous helium cryopumps, augmented with LN 2 shrouds. Combination TSP and sputter-ion pumping is also used extensively in particle accelerators and storage rings in operation throughout the world (e.g., see refs. 2 9,4 a,4 9 ). Many of the accelerators and storage rings noted in Section 2.10 use this form of combination pumping. Pressures in the 10" 1 a to 10" 1 1 Torr range may be reliably achieved through this means.(4 8-s 1 ) TSP pumping surfaces are made comparatively large, so that the pump speed is large, and infrequent use of the sublimation pump is required to maintain the needed operating pressure. In such instances, the Ti films need not be cleaned off the pumping surfaces. This is because the operating life of the pump approaches the useable life of the accelerator and the accumulated thickness of the sublimed films is therefore negligible (see problems 5-11). As a footnote to combination pumping, for reasons having to do with safety, I strongly discourage the use of TSP pumping in conjunction with cryopumps capable of sorbing (i.e., temporarily storing) large quantities of flammable gases. More on this is discussed in the following chapters.
222
TITANIUM SUBLIMATION PUMPING
Peeling of Titanium Films The subliming of Ti onto LN 2 cooled surfaces fell into disfavor in later years. It was determined that the advantages gained in values of sticking coefficients and f'dm capacities with such use were outweighed by disadvantages of premature peeling of deposited films. This effect is not reported in the literature, though peeling effects of films on various surfaces have been reported. Harra reports the peeling of the sublimed f'dms initiates after deposition of -0.03 gm/cm 2 .(2 o ) This was a worla'ng system and the f'dm was used intermittently (i.e., for thousands of cycles) to pump large quantities of air. Assuming, because of the sorption of gases, the film has a density o f - 4 . 0 gm/cm a , this corresponds to a f'dm thickness of -7.5 x 10 "a cm. Martin made measurements of the peeling of films deposited on surfaces of various materials, including sheets of Ti, Ta, Mo, Cu, A£ and stn. stl.( 2 6 ) Half of each of the test sheets were sandblasted. He noted unsandblasted surfaces peeled with thinner films, compared to sandblasted surfaces. This work was of a qualitative nature, though he noted a "maximum" film thickness of 0.023 cm on peeling. Conservatively, a usable thickness of--5 x 10"a cm seems reasonable. The peeling of films causes a deterioration in pump performance. As the film peels back from the surface of the pumping array, the film tends to operate at elevated temperatures, when the sublimation pump is turned on. This is because thermal conduction along the peeling film is poor, yet its surface area is comparatively large. As the result of increases in temperature of the peeling film, H2 may be thermally desorbed, and result in unwanted increases in system pressure. Because of this peeling effect, it is necessary to clean the surfaces of the pump from time to time. Therefore, one must either have access to the deposition surface for cleaning, or provisions must exist for removing the pumping array from the system. The clever configuration shown in Fig. 3.5.1 served this purpose. Titanium was sublimed onto the inner surfaces of a cylinder comprising numerous cheveroned baffles. The baffles were transparent to the transmission of gas into the TSP pump, but completely cut off the sublimed Ti vapor. The cheveroned array could be easily removed from the system for periodic cleaning. A word of caution: peeling films of Ti are pyrophotic. That is, they may spontaneous ignite, or spark when vented to air. Also, when scraping these films off of metal surfaces, if enough work is put into the film, it is possible to ignite the f'dm and have a localized "flash fire". This is not a major safety hazard. But, we can often injure ourselves when responding to a surprise, and this is merely a word of warning. Secondly, film refuse should not be heaped together or discarded in containers having other flammable refuse.
3.6 Advantages and Disadvantages of TSP Pumping There are disadvantages to the use of TSP pumps. First, in being capture pumps, they have a finite capacity. In high sublimation rate applications, the pump surfaces must be periodically cleaned (see section on film peeling). The power radiated from Ti sources, in the absence of water cooling, will raise the temperature of surfaces within the vacuum system. This causes system outgassing. These pumps are low throughput devices (see problem 3). TSP pumps will not pump the inert gases, and gases including CH 4 are synthesized at the hot Ti source. Because CH4 is a comparatively inert gas, it is virtually unpumped by the Ti film,
TITANIUM SUBLIMATION PUMPING
223
and requires the use of combination pumping. Also, some form of rough pumping is required, as in the case of all high vacuum pumps. The sublimed Ti vapor can cause problems in a vacuum system if adequate line-ofsight baffling does not exist between the Ti source and other vacuum equipment. For example, sublimed films can cause breakdown or leakage currents in high voltage insulators, the contamination of working films and optical surfaces, and damage to instnanentation within the system, without appropriate shielding from the Ti flux. The need for the shielding of surfaces from the Ti flux requires that gas impedances be interposed between the system working volume and the pump, thus reducing the available speed of the fdm surface to the system. Combination pumping with TSP and sputter-ion pumps is the most reliable and cost effective method of achieving UHV pressures in bakeable, lumped systems.(S e ) By lumped, I mean that the pumps are lumped elements rather than elements distributed along some great length, as in the application of NEG pumping in, for example, LEP.( s 2 ) TSP pumping is the most economical of all forms of pumping, in terms of operating power, or W/Z/sec at very low pressures. TSP pumps are bakeable, and with the exception of C H , and possible Ti vapors introduce no contaminants into the system. They are safe to operate, though when being serviced peeling Ti films pose potential hazards, as noted in Section 3.5. Systems in which they are used are simple to design. They may be operated in any position, are immune to high radiation fields, and present no source of vibration to the system.
Problem Set 1) Using the data of Fig. 3.2.5 and the pump configuration of Fig. 3.2.1, calculate the intrinsic speed of the pump for the RT gases H2, N2 and O2, assuming a gas surface coverage 101 a 10 z 4 and 101 s molecules per cm 2 2) Show that without a pumping synergistic effect, and the synthesis of CH4, (3.3.1) should be true. 3) Assume a perfect stoichiometry of TiN is achieved in a continuously operated system. What would the throughput capability be, in Torr-Z/sec, of a TSP source having a sublimation rate of 0.5 gm/hour? 4) Use Table 3.2.1 to calculate the speed of a TSP pump for the gases H 2, CO, N2 and 0 2 . Assume the temperature of the pumping surface is 77°K and the surface area is 103 cm2, and the gas is at RT. 5) Assume that TSP combinations pumps are designed as shown in Fig. 3.6.1. Assume that the pump is operated at RT. For a freshly deposited film, what is the intrinsic speed of one pump for the gases H 2, CO, N 2, O 2 and CO 2 ? 6) Assume a system configuration similar to that of Fig. 3.6.1. The H 2 outgassing rate of the long beam pipe attached to this pump is -10" 1 2 Torr-Z/sec-cm2. The distance between each pump is 2x cm. The outgassing of other gases from the beam pipe is negligible. Assume the values of c~ for H 2 vary as shown in Fig. 3.2.5. Assuming an H 2 coverage of 101 a molecules/cm 2, what is the average pressure in the system if it is periodic in sets of the system shown in Fig. 3.6.17 Neglect outgassing from the pump and connecting manifold.
224
TITANIUM SUBLIMATION PUMPING
•
~
[ t
I
~'~
~ ! ~
2x cm -
,~
=-
8.3 cm ~ I.D.
_,.~~ /f- - - ~ -"~ ]~ TSP ~ ~" FILAMENTS [ ~
1_,.,___-------20 L/see - ~
_[
I
i
PARTICLE ~ i < ACCELERATOR / f " "N BEAM PIPE [ [~ / t l l ]
PUMPING
I
I
Figure 3.6.1. Particle accelerator storage ring pumping system similar to t h a t described by Halama for the "ISABELLE" accelerator ring.(50)
7) What is the _average pressure of the system in problem 6, if the beam pipe outgassing rate is 10-z a Torr-Z/sec-cm2 H 2 ? What is the average pressure if the gas is CO and the outgassing rate is the same? 8) Assume that the H2 sticking coefficient initially decreases as shown in Fig. 3.2.6, and extrapolate coverage data until a gas/film stoichiometry of TiM2 is reached. Assume a film thickness of--200 A. Construct an analytical expression for the sticking coefficient as a function of coverage. 9) Using the results of problem 8, express the speed of the film as a function of coverage. 10) Assume that the diffusion of H 1 is not a limiting factor. Express the average pressure in the system as a function of time, assuming the H 2 outgassing rate of problem 7, the pumping speed found in problem 8, and until such time as a stoichiometry of TiM is reached. 11) Assume we operated the pump in problem 5 with a sublimation rate of 0.05 grn per hour, under batch sublimation pumping conditions. Conservatively, how long would we be able to operate the sublimation pump prior to the onset of film peeling?
References 1. 2.
3. 4.
Stout, V.L. and Gibbons, M.D., "Gettering of Gas by Titanium", J. Appl. Phys. 26(12), 1488 (1955). Wagener, J.S., "Properties of Getters in Electron Tubes", Proc. 4th Nat. Conf. on Tube Techniques, 1958 (New York University Press, New York, 1959), p. 1. Giorgi, T.A., Ferraio, B., Storey, B., "An Updated Review of Getters and Gettering", J. Vac. Sci. Technol. A3(2), 417 (1985). Hayward, D.O. and Taylor, N., "The Adsorption and Absorption of Hydrogen by Evaporated Tantalum Films", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969), p. 115.
TITANIUM SUBLIMATION PUMPING 5.
6. 7. 8.
9.
10. 11. 12. 13. 14. 15. 16. 17. 18.
19. 20.
225
Okano, T., Iimura, K., Tominaga, G., "A Tantalum Evaporation Pump", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. Rtidenauer, F.P. Viehb6ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 81. Prevot, F. and Sledziewski, Z., "The Titanium Evaporation Pump, its Application to Nuclear Fusion Experiments and Space Simulation", J. Vac. Sci. Technol. _9.(1),49 (1972). Haque, C.A., "Simple Titanium and Nickel Sublimation Pump (TNSP)", J. Vac. Sci. Technol. 13(5), 1088 (1976). Nazarov, A.S., Ivanovskii, G.F., Men'shikov, M.I., "Getter-Ion Pump With Filamentary Titanium and Chromium Evaporators", translated from Pribory i Tekhnika l~ksperimenta, No. 5, pp. 157-161, 1963, Instrum. & Exper. Tech., 934 (1964). Kuz'min, A.A., "Laboratory Ultrahigh-Vacuum Apparatus With Filamentary Solid-Phase Titanium Evaporators", translated from Pribory i Tekhnika l~ksperimenta, No. 3, pp. 126-130, 1963, Instrum. & Exper. Tech., 497 (1963). Clausing, R.E., "A Large-Scale Getter Pumping Experiment Using Vapor Deposited Titanium Films", Proc. 2nd Int. Vac. Cong., 1961 (Pergamon Press, Inc., New York, 1962), p. 345. Herb, R.G., Davis, R.H., Divatia, A.S., Saxon, D., "Evapor-lon Pump", Phys. Rev. 89, 897 (1953). Hayward, D.O. and Taylor, N., "Some Fundamental Problems in the Measurement of Sticking Probabilities of Gases on Metal Films", J. Sci. Instrum. 44, 327 (1967). Giorgi, T.A. and Pisani, C., "On the Concept of Pumping Speed in Gettered and Cryopumped Systems", Vacuum 16(12), 669 (1966). Harra, D.J. and Hayward, W.H., "Sorption of Nitrogen by Titanium Films", Proc. Int. Symp. on Residual Gases in Electron Tubes, 1967, Supplemento AI Nuovo Cimento Serie I. 5(1), 56 (1967). Kornelsen, E.V. and Domeij, B., "Simple Differential Pumping Stage for Connecting High to Ultrahigh-Vacuum Systems", J. Vac. Sci. Technol. 3(1), 20 (1966). Harra, D.J., "Review of Sticking Coefficients and Sorption Capacities of Gases on Titanium Films", Proc. 22nd Nat. AVS Symp., 1975, J. Vac. Sci. Tcchnol. 13(1), 471 (1976). Grigorov, G.I., "Apparent and Real Values of Common Gas Sticking Coefficients on Titanium Films and Application to Getter Pump Devices With Periodic Active Film Renovation", Vacuum 34(5), 513 (1984). Biryukova, N.E., Vinogradov, M.I., Efimov, M.N., "Sorption of Nitrogen by Continuously Deposited Titanium Films", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969), p. 184. Elsworth, L., Holland, L., Laurcnson, L., "The Sorption of N2, H2, and D 2 on Titanium Films at 20 ° C and -190 ° C", Vacuum 15, 337 (1965). Harra, D.J., "Improved Titanium Sublimation Pumping Techniques for Long-Term, High-Throughput, Clean Pumping", J. Vac. Sci. Tcchnol. 12, 539 (1975).
226
TITANIUM SUBLIMATION PUMPING
21. Gupta, A.K. and Leck, J.H., "An Evaluation of the Titanium Sublimation Pump", Vacuum 25(8), 362 (1975). 20 Reichardt, J.W., "The Kinetics of the Hydrogen-Titanium Reaction", J. Vac. Sci. Technol. 9(1), 548 (1972). 30 Steinberg, R. and Alger, D.L., "Access to Uncombined Titanium Through an Inhibiting Film in Sublimation Pumping of Deuterium", J. Vac. Sci. Technol. 10(1), 246 (1973). 24. Lawson, R.W. and Woodward, J.W., "Properties of TitaniumMolybdenum Alloy Wire as a Source of Titanium for Sublimation Pumps", Vacuum 17, 205 (1967). 25~ Grigorov, G.I. and Tzatzov, K.K., "Theory of Getter Pump Evaluation. Sticking Coefficients of Common Gases on Continuously Deposited Getter Films", Vacuum 33(3), 139 (1983). 60 Martin, G.D., "Pulsed Gas Load Pumping of Hydrogen by Vapor Deposited Films", Princeton Plasma Physics Laboratory, Princeton, New Jersey, PPPL Report #MATF-1193, 1976. 270 Jepsen, R.L. and Francis, A.B., "Interactions Between Ionizing Discharges and Getter Films", Supplemento AI Nuovo Cimento 1(2), 694 (1963). 80 Klopfer, A. and Ermrich, W., "Properties of a Small Titanium-Ion Pump", Proc. 6th Nat. AVS Symp., 1959 (Pergamon Press, Inc., New York, 1960), p. 297. 90 Halama, H.J., "Behavior of Titanium Sublimation and Sputter-Ion Pumps in the 10-1 1 Torr Range", J. Vac. Sci. Technol. 14(1), 524 (1977). 00 Edwards, D., "Methane Outgassing from a Ti Sublimation Pump", J. Vac. Sci. Technol. 17(1), 279 (1980). 31. Edwards, D. and Lanni, C., "Ti Getter Study", J. Vac. Sci. Technol., 17(1), 1373 (1980). 32. Chou, T.S. and Lanni, C., "Lifetime of Titanium Filament at Constant Current", IEEE Trans. Nucl. Sci. NS-28(3), 3323 (1981). 33. Kornelsen, E.V., "A Small Ionic Pump Employing Metal Evaporation", Proc. 7th Nat. AVS Symp. 1960 (Pergamon Press, Inc., New York, 1961), p. 29. 411, McCracken, G.M. and Pashley, N.A.,"Titanium Filaments for Sublimation Pumps", J. Vac. Sci. Technol. 3(3), 96 (1966). 35. Herb, R.G., Pauly, T., Welton, R.D., Fisher, K.J., "Sublimation and Ion Pumping in Getter-Ion Pumps", Rev. Sci. Instrum. 35(5), 573 (1964). 60 Harra, D.J., "Predicting and Evaluating Titanium Sublimation Pumping", Presented at 6th Int. Vac. Cong., Kyoto, Japan, 1974, Varian Associates Report No. VR-88, 1974. 37. Brewer, L. and Lamoreaux, R.H., Molybdenum: Physic0-Chemical Properties of its Compounds and Alloys, Special Issue No. 7 (International Atomic Energy Agency, Vienna, 1980, L. Brewer, Ed.), Chapter II, "Phase Diagrams", p. 333. 80 Kuznetsov, M.V., Nasarov, A.S., Ivanovsky, G.F., "New Developments in Getter-Ion Pumps in the U.S.S.R", J. Vac. Sci. Technol. 6(1), 34 (1969). 39. Strubin, P., "Study of a New Method to Control Precisely the Evaporation Rate of Titanium Sublimation Pumps", J. Vac. Sci. Technol. 17(5), 1216 (1980).
TITANIUM SUBLIMATION PUMPING
227
0Q Spangenburg, K.R., Vacuum Tubes (McGraw-HiU Book Company, Inc., 41.
20 3Q 40
45. 6Q
47.
80
90
50. 51.
52. 53. 54. 55.
~6o
New York, 1948), p. 30. Sweetman, D.R., "The Achievement of Very High Pumping Speeds in the Ultra-High Vacuum Region", Nucl. Instrum. Methods 13, 317 (1961). Harra, D.J. and Snouse, T.W., "A Radiant Heated Titanium Sublimator", Proc. 5th Int. Vac. Cong., 1971, J. Vac. Sci. Technol. 9(1), 552 (1972). Gould, C.L. and Mandel, P., "A Sublimation Pump", Proc. 9th Nat. AVS Syrup. 1962 (The Macmillan Company, New York, 1963), p. 360. Burthe, J.H. and Munro, D.F., "Bulk Sublimation of Titanium", Proc. 3rd Int. Vac. Cong., 1965 (Pergamon Press, London, 1966), p. 37. Burden, M. St. J., Walley, P.A., "The Evaporation of Metals and Elemental Semiconductors Using a Work-Accelerated Electron Beam Source", Vacuum, 19(9), 397(1969). Wolf, R.A., "The Application of Selective Pumping Techniques to Upgrade Vacuum Chambers", J. Environ. Sci., October 1966, p. 3. Brothers, C.F., Tom, T., Munro, D.F., "Design and Performance of a 50,000 ~/sec Pump Combining Cold Cathode Ion Pumping and Active Film Gettering", Proc. 10th Nat. AVS Symp., 1963 (The Macmillian Company, New York, 1964), p. 202. Ishimaru, H., "Ultimate Pressure of the Order of 10"1 a Torr in an Aluminum Alloy Vacuum Chamber", J. Vac. Sci. Technol. A7(3), 2439 (1989). Johnson, J.W., Atkins, W.H., Dowling, D.T., McConnell, J.W., Milner, W.T., Olsen, D.K., "Heavy Ion Storage Ring for Atomic Physics Vacuum Test Stand for Pressure of 10-1 2 Torr", J. Vac. Sci. Technol. A7(3), 2430 (1989). Halama, H.J., "ISABELLE Vacuum Systems", Proc. 8th Int. Vac. Cong., 1980 (Suppl6ment ~ la Revue, "Le Vide, les Couches Minces, No. 201", 1981), p. 115. Halama, H.J., "Design and Construction of Vacuum Systems for Large Colliders Using Superconducting Magnets", Proc. 9th Int. Vac. Cong. and 5th Int. Conf. on Solid Surfaces, 1983 (Asociacion Espafiol del Vacio y sus Aplicaciones, Madrid, 1984), p. 283. Reinhard, H.P., "The Vacuum System of LEP", Proc. 9th Int. Vac. Cong. and 5th Int. Conf. on Solid Surfaces, 1983, (Asociacion Espafiol del Vacio y sus Aplicaciones, Madrid, 1984), p. 273. Gretz, R.D., "A High Vacuum Titanium Getter Pump", Vacuum 16(10), 537 (1966). Robertson, D.D., "Evaluation of a High-Rate Titanium Getter Pump", Proc. 4th Int. Vac. Cong., 1968 (The Institute of Physics and The Physical Society, London, 1969), p. 373. Holland, L., Laurenson, L., Allen, P.G.W., "The Formation of Hydrocarbon Gas by Titanium Getters Containing Carbon and Hydrogen Impurities", Proc. 8th Nat. AVS Symp. and 2nd Int. Vac. Cong., 1961 (The Macmillan Company, New York, 1962), p. 208. Harra, DJ., Methods of Experimental Physics, 14, Chapter 5.2, "Getter Pumping" (Academic Press, Inc., London, 1979, G.L. Weissler, R.W. Carlson, Eds.), p. 193.
CHAPTER 4
NONEVAPORABLE GETTERS 4.0 Introduction There is a class of getters which effectively pump gases merely as a consequence of being activated by heating to a certain temperature, under vacuum. However, the getter activation temperatures, differing with different materials, are low compared to the operating temperatures of TSP sources. Therefore, appreciable sublimation or evaporation from these getter materials does not occur. Thus, they have been named nonevaporable getters, or NEGs. The NEG comprises a sintered-powder matrix of metal alloys which chemisorb gases. These alloys are typically operated at temperatures ranging from RT (room temperature) to --400* C, though some require activation at slightly higher temperatures. A number of different alloys and pure metals possess this gettering capability. Giorgi, Ferrario and Storey review the properties and applications of some of these NEGs.(X ) Kindl and Rabusin also provide a review of early NEG development work.(2 ) Most of the significant NEG development work to date has been done at the company SAES Getters S.p.A., headquartered in Milano, Italy. Of course, studies of the pumping characteristics of these materials have also involved the work of others throughout the world. Two alloy mixtures have been widely reported on. These are the St 101® and St 707~ materials. The St 101® material is an alloy of 84% Zr and 16% A~ by weight, where the St 707~ material comprises an alloy of 70% Zr, 24.6% V and 5.4% Fe, by weight. In the development of these NEGs, some other NEG was used as a benchmark or measure of relative merit. For example, the St 101® material, first reported on by Della Porta, et al., in 1961,(3 ) was compared with NEGs comprising mixtures of Th and A~. The high activation temperatures of the St 101® material (101, hereafter) prompted the development of the second popular NEG material, St 707~ (707, hereafter), which was first reported on by Boffito, et al., in 1981.~4 ) This chapter will primarily deal with the properties and applications of these two materials. Some mention will be made of the new sintered NEG materials St 172~ and St 185~ (referred to as 172 and 185 respectively hereafter), and recent work done at CERN in sputter-coating chambers with NEG materials. However, I will first discuss some of the forms and mechanical configurations of these NEGs as presently used. After this, I will discuss general aspects of the pumping mechanisms of these NEGs, as presently understood. I will conclude by describing test results of a number of investigators using these NEGs in a wide variety of applications.
4.1 Mechanical Features Boffito describes these materials as "sintered porous bodies".( 4 ) However, I have been informed that a sintering operation is not used in the fabrication process of either the 101 or 707 bonded strips or pellets. Initially, the 101 material was available
230
NONEVAPORABLE GETI'ERS
only in bulk form. That is, the ZrA~ alloy was pressed into pellets having dimensions, for example, of 5-10 mm (b and a thickness of 2-3 mm. Subsequent to formation by pressing, the pellets retained their porosity. Kindl notes that this porosity is equivalent to --0.015 Z/sec-cm ~ for CO (the alloy thickness and gas temperature were not given).(2 ) To increase the surface area of the pellets and facilitate ease of attachment, the pellets were also offered in the shape of washers. Pellets of this and other special shapes are now available in both materials. New applications became possible with the discovery of a method of bonding thin layers of powdered 101 material to metal surfaces. This was first reported on by Kindl and Rabusin in 1%7.(2 ) They reported on a "patented" method of bonding the material to metal substrates. Granules of the material are rolled on to the substrate under high pressure. The process does not involve the use of organic binders. One speculates that some type of thin film is somehow deposited on the metal substrates prior to roll-bonding, and that this thin film acts as a "sticking" agent for the NEG granules. The 101 material may be bonded to Ni plated Fe or Constantan ribbon. Kindl reported problems when attempting to sinter-bond the ZrA£ powder to such substrates.(2 ) Zirconium forms a eutectic with Ni at -960* C. Apparently, if the temperature (and perhaps the Ni plating thickness) was not controlled during sintering of the ZrA~ powder to the substrate, the Ni would diffuse not only in the substrate, but throughout the bulk of the NEG material. This had the effect of significantly reducing the sorptive capacity of the NEG. 30 mm
~
UNCOATED EDGES
0.2 m m ~ - - SUBSTRATE THICKNESS I
METERS IN LENGTH
NEG COATING ~~BOTH SIDES 84%Zr, 16%A/ OR l~Zr,26.4%V,5.4%Fe
I
,I
Ni PLATED Fe 70/urn OR CONSTANTAN NOMINAL SUBSTRATE mm COATING THICKNESS F i g u r e 4.1.1. E x a m p l e of s t r i p NEGs u s e d in d i s t r i b u t e d p u m p i n g a p p l i c a t i o n s .
\
In any event, both the 101 and 707 powder materials may be bonded to either Ni plated Fe substrates, or onto constantan (i.e., a nonmagnetic alloy comprising 55% Cu and 45% Ni). The material is bonded onto strips of either of these two metals, as shown in Fig. 4.1.1. The thickness of the bonded material is -70/.tm, where the substrate may be any thickness; 200 and 400 p m thicknesses are typical values. The powdered alloy, prior to application, has particulate dimensions of the order of 50-100/.tm. Therefore, the 70 p m sintered coating corresponds to about a one or two-particle NEG thickness. Apparently the powder, when pressed into the substrate - probably by high pressure rolling - sticks sufficiently well to itself and the substrate that it may be handled thereafter without falling off the substrate.
NONEVAPORABLE GETTERS
231
For reasons to be discussed, knowing the amount of pressed powder material on the substrate (i.e., g/cm2) is important in the prediction of the sorptive capacity of the NEG per unit area. This is also true as it relates to the relative density of the powder granules. Benvenuti reported progressive improvements in the pumping capacity and speed of 101 materials supplied by SAES for use in the CERN LEP accelerator.(S ) Part of these improvements may have stemmed from increases in the thickness of the 101 material (i.e., 8 2 - 106/.t m thick vs. 70 # m) over the years. However, reported increases in both speed and capacity of the 101 materials may also have stemmed from increases in the density of the sintered powder. Apparently wide variations in powder density are possible. For example, using data from SAES published catalogs, we conclude that the density of powder granules on substrates such as shown in Fig. 4.1.1 is -2.65 g/cm 3 for the 707 material and -2.3 g/cm 3 for the 101 material, assuming a 70 # m thickness in both cases. On the other hand, the density of pellets of the 707 material is calculated to be 4.9-5.1 g/cm 3 , where the density of 101 pellets is -3.7 g/cm 3 . Benvenuti reports that the porosity of newer batches of 101 powder, bonded to Constantan substrates by "cold pressing", is -15%,( s ) compared with a porosity o f - 1 2 % in earlier batches of the 101 material. The bonded material retains sufficient ductility and adhesion so that the strip of metal onto which the NEG is bonded may be bent or shaped into different configurations, depending on the application. Variations of such array configurations have been reported on for use in fusion applications.(6 -t o ) These arrays are periodically activated by passing current through the metal substrate. The possible combinations of array configurations are limitless, and left to your imagination. Audi, et al., report on the use of panels of a 707 NEG structure, similar to that shown in Fig. 4.1.2, installed in a StarCell ® triode SIP.( ~ 1 ) Five years later Romer reported results of the use of a similar 707 structure in a conventional triode SIP, with similar remarkable results particularly relating to low pressure performance of the pumping ensemble.(a 9 ) For reasons noted in Section 2.5, this would have obvious advantages in the pumping of H2 at extremely low pressures. Poncet and his colleagues have used variations of these NEG ,panel arrays, in both linear form and toroidally pleated form, in a particle beam pipe.t t 2,a 7 ) . ~,/
~'~'~'~
5-10 £~/sec-cm 2 V- PROJECTED AREA FoR .YDROGEN
PANE PLEATED 30 m m WIDE NEG STRIPS
Figure 4.1.2. NEG panels a r r a y e d in a plane as described by Ferrario and Rosai. (36)
232
NONEVAPORABLE GETTERS
Another such application is in the construction of modular pumps, such as shown in Fig. 4.1.3. In this case the pleated strips are formed into toroidal-shaped bands. Very high surface area densities may be created with this configuration. The modular pump is activated using independent heaters located within the center of the pump. The substrates of the modular pumps may be coated with either 101 or 707 material, each having advantages which will be discussed. Commercially available pump modules or panels constructed of bent strips coated with either the 101 or 707 NEG material are called SO RB-AC ~ pumps by the SAES company. Those constructed of sintered blades of either the 172 or 185 materials are called CapaciTorr ~ pumps. ~
150 mm FLANGE
PLEATED METAL STRIPS WITH .~400g OF NEG
0 ~ / / / ~ m \
\
i
COAXIAL HEATER F i g u r e 4.1.3. E x a m p l e of m o d u l a r , n u d e NEG c a r t r i d g e p u m p m a n u f a c t u r e d by SAES. The comparatively high pumping speeds pe~ unit area of the strip material led to its use in applications of distributed pumping, x a ) The use of NEGs in this capacity was first proposed by Benvenuti and Decroux at CERN.(1 4 ) They suggested the application of NEGs as distributed pumps for both electron and proton storage rings and accelerators. The most daring example of this application was its use in the LEP e + e" collider.( 6 ,t s ,1 6 ) It was daring in that 27 km of accelerator was designed and constructed using distributed NEGs of the 101 material as the principal pumping mechanism. It proved most successful in this application.(1 7 ,x a ) Strips of NEG material were positioned in an antechamber in an extruded A2 beam pipe, in a manner similar to the distributed sputter-ion pump shown in Fig. 2.10.2 of Chap. 2. Hseuh also successfully used a distributed NEG, comprising the 707 material, in a heavy ion beam transport line at Brookhaven.(1 9,2 o ) Results of these and other applications studies will be discussed.
4.2 NEG Pumping Mechanisms I will first discuss NEG pumping mechanisms in a general sense, as they are presently understood. With this background, I will then give specific findings reported in the literature of the characteristics of the 101 and 707 materials, in some of their present applications.
NONEVAPORABLE GEIq'ERS
233
The Pumping of CO, CO 2, N 2 , 0 The sintered powder is very porous to the transport of gases between the voids in the bulk material. Because of this transparency to gas, the effective surface area of both NEG materials is of the order of ×70 to x 100 that of its geometric surface area.(1 s ,1 ~ ) With exposure to air, or during the pumping of some gases, the surface of the granules become oxidized. This oxide layer passivates the surface; that is, it prevents further surface pumping of gases containing C, O and N. The primary ingredient responsible for surface chemisorption is the Zr. This was shown in the early work of Lampert, Rachocki and Lamartine in the study of St 171®, a NEG comprising a mixture of graphite and Zr.(S s ) Using ESCA (i.e., soft x-ray photoelectron spectroscopy) they determined that the surface of the granules was covered with ZrO2 and ZrC. However, when the material was activated by heating to --900* C, the oxygen and carbon disappeared by diffusion into the bulk material. Miller, reporting on much earlier work of others, described this as also being characteristic of pure Zr.(2 ~ ) He notes that the "protective" film of ZrO2 dissolves into the bulk material at -450* C, permitting further adsorption of 0 2 . He indicated that O 1 (i.e., atomic oxygen) diffuses interstitially in a Zr lattice, with up to even 60 atomic percent oxygen. Also, atomic nitrogen, like oxygen, is absorbed in the interstices of the lattice, but has much less mobility. Because of the ability of all gases to enter voids in the porous body of the NEG, simplistic models involving diffusion processes through a semi-infinite plane are probably not appropriate to NEG pumping. On the contrary, gases are able to diffuse into the bulk nearly through a full 47t steradians of the particle surfaces. This should also be applicable to any H ~ pumping model (e.g., see reference 2 2 ). The above carbon and oxygen diffusion mechanism, applicable to the sorption of CO, CO2 and O2 by pure Zr and the St 171@ material, is also applicable to the 101 and 707 materials. This was shown in the more recent work of Ichimura, Matsuyama and Wantanabe.(2 s ) Using ESCA, they studied the measure of surface activation - this amounts to the freeing up of surface Zr - as a function of activation temperature for three materials (i.e., the 101, 707 materials and a ZrNi alloy). Through use of SIMS (i.e., secondary ion microscopy scanning), they also determined the ratio of sputtered Zr + to ZrO + ions from the surface of the NEG to determine the relative amount of free Zr on the surface as a function of temperature cycling. ESCA and SIMS results were taken at RT after heating the NEG to successively higher temperatures under UHV conditions. They also conducted thermal desorption spectroscopy (TDS) measurements on the three materials. Their ESCA results are reproduced in Figures 4.2.1 and 4.2.2. Z
At.%
0
80
C¢3~
o~
60
o~
40
~ C
Zr-~\
o,
ro
~<
20
~
0
0
200 400 600 8 0 0 1 0 0 0 TEMPERATURE - ° C
Figure 4.2.1. C h a n g e s in s u r f a c e c o m p o s i t i o n of t h e St I01® NEG vs. a c t i v a t i o n t e m p e r a t u r e . (23)
234
NONEVAPORABLE GETTERS z
o
At.% 80
/ C
~ o ~ 60 F i g u r e 4.2.2. C h a n g e s in s u r f a c e c o m p o s i t i o n of t h e St 707® NEG vs. a c t i v a t i o n t e m p e r a t u r e . ( 2 3 )
~'
40 20
a.~
~.2..~..-o
"" ~'''" ,._._.. 0
/-V ~
fFe
200 400 600 flO0 1000 TEMPERATURE - ° C
From their findings, including data in the two figures, they concluded the following. Surface oxygen in the 101 material, as-received, and at RT was attributable to the oxygen in A l 2 Os and ZrO2. After heating the material to --800* C, the remaining "surface" oxygen was interstitial. The as-received surface Zr in the 101 material was actually tied up as ZrO2. Their ESCA and SIMS measurements verified that the surface Zr began to appear in the metallic state only after heating the 101 material to temperatures > 400* C. We see from Fig. 4.2.1 that complete activation of the 101 material occurs after the material is raised to a temperature of 800* C. The surface Zr in the as-received 707 material was also present as ZrO2, though, as shown in Fig. 4.2.2, metallic Zr appeared at much lower temperatures, on heating the material. Using results of TDS, they observed only slight desorption of CO and CO2 from the 101 material up to a temperature of 500* C. Above this temperature there was negligible desorption of these species. Yet, C and O disappeared from the surface of the material at higher temperatures. Therefore, they concluded that both the C and O diffused into the bulk of the material. A similar model applies to the 707 material. However, because of the apparent variation in the diffusivities of C and O in the two materials, activation of the 707 material is almost complete at 400* C, where activation of the 101 material asymptotes at --800* C. In summary, the pumping of gases such as CO, CO 2, O 2, N2, and the oxygen component of H 2 O, occurs on the surface of the powder granules. Carbon is tied up as ZrC and oxygen as ZrO2. After a certain amount of gas has been chemically pumped on the surface, the free Zr sites become saturated. In order to activate the getter, or replenish these sites, the getter must be heated so that the C and O will diffuse into the lattice of the material. On the one hand, the diffusion of these elements into the material creates new surface pumping sites. On the other hand, the bulk material has a finite capacity for accommodating impurities diffusing therein. Because of this, NEGs, as all chemisorption pumps, have a finite capacity for gases. In the case of NEGs, we must define two capacities: 1) the capacity of the pump at saturation of the metal Zr surface sites (i.e., surface capacity); 2) the capacity of the bulk to accommodate diffusing impurities (i.e., bulk capacity). Regarding bulk capacity, little independent experimental data are published on measurements clearly defining the total sorptive capacities of either the 101 or 707 materials for the gases CO, COs, 0 2 , H2 O, N2, etc. In describing the impact of CO loading on the H2 equilibrium pressure, Boffito, et al., show H2 equilibrium data for the 707 material which has presumably pumped -15 Torr-Z CO/g, and data for the 101 material which has pumped - 6 Torr-Z CO/g.(2 4 ) Kindl and Rabusin report pumping CO to the point of saturating of 101 material.(2 ) They pumped "... 50 cm a Torr of CO at 800* C in one hour per mm 2 of geometric exposed surface ...
NONEVAPORABLE GETTERS
235
under a CO pressure of 1 Torr." Assuming a film thickness o f - 0 . 5 mm (i.e., Fig. 3 of the reference) and a f'dm density of - 4 g/cm a , this would equate to a total RT capacity of the 101 material of -6.8 Torr-Z CO/g. Data published in SAES product catalogs indicate the total bulk capacity values shown in Table 4.2.1. A note of caution. In analyzing the published data, it is often very difficult to reduce these data to a common base-line for engineering applications. For example, one such base-line might be the speed per unit area of NEG. In reformatting the data for your use, I have had to scale each datum of published figures which I thought to be salient. Excluding human error, the scaled data will be accurate to "-5%. When all the conditions of a reported test were not stated, I have included the assumptions made in all data interpretation. TOTAL "PRACTICAL" BULK CAPACITY T o r r - Z / g NEG
GAS CO
N2
H2
St 101 ®
2.2
2.2
20.8
St 707 ®
9.0
9.0
20.0
Table 4.2.1. Total "practical" irreversible capacities of the "101" and "707" NEGs for N2 and CO, and m a x i m u m reversible H2 capacities prior to the onset of e m b r i t t l e m e n t problems.
Hydrogen Pumping Unlike the gases noted in the prior section, the pumping of H2 is completely reversible in 101 and 707 NEGs. As noted in the previous chapters, H2 is pumped by fh'st dissociating into an atomic state on the surface of the getter and diffusion at RT, thereafter, into the bulk of the material. We noted that sputter-ion pumps could be activated to have high pumping speeds for H 2, if the surfaces were first scrubbed by sputtering with a noble gas. This scrubbing had the effect of sputtering away oxides and nitrides on the cathode surface which occupied hydrogen dissociation energy sites. These oxides also act as a barrier to the diffusion of hydrogen, though this latter effect is of less importance, as was noted in Chapter 3. Therefore, if the surface of the NEG has been activated so as to create fresh metallic Zr, H 2 is readily dissociated and diffuses into the bulk of the material. In fact, NEGs have such high capacities for H 2, if certain H 2 pumping limits are exceeded, the sintered powder material will become brittle and small pieces of the NEG will spontaneously pop off the surface of the substrate. This effect was noted by Malinowski, when conducting tests with the 707 material.(1 o ) In a figure shown by Knize, Cecchi and Dylla, they indicate that this embrittlement threshold occurs after pumping -18.5 Torr-Z H2 per gram of 101 material.( 2 2 ) Similar embrittlement data (i.e., -.20 Torr-Z H2/g) are reported for both the 101 and 707 materials in product literature distributed by SAES and in several publications by SAES staff.( a 8 ) In the pumping of gas mixtures, including CO, CO2, 0 2 , H 2 0 and H2, the former four gases eventually take up the active surface Zr and cause significant decreases in H 2 speed. Ferrario, et al., reported on the plugging effect or passivation of both 101 and 707 NEG materials to H 2, as a consequence of the simultaneous pumping of the individual gases CO, CO2 and N2.(8 ) Results of these tests are shown in Figures 4.2.3 and 4.2.4. In these figures the decrease in H2 speed is given as a function of
236
NONEVAPORABLE GETFERS
the amount of each of the three gases sorbed. Ferrario used modular NEG panels in these tests, similar to that shown in Fig. 4.1.2. 800
1
d
r.
1
1
I
_.~J~--
_ j /
600
.,..4
I
[-I
/
[]
I
1
1
H z S P E E D WITH N2 PLUGGING
-
p
-
[]
[-1 ,~
400
H2 S P E E D WITH COa P L U G G I N G -
o
200 H2 SPEED WITH~ O CO P L U G G I N G "~~..~
0
I
1
l
,
10- 5
)'
0 t,.-.,
(,x,r
i
i
10- 4
10
TORR L I T E R S / e r a 2 OF E I T H E R Figure 4.2.3. as a function
0
St of
N2,
CO OR C02
101 ® NEG plugging for the the amount o f N3, CO a n d
800
1
d
I
I
[]
~
VT
600 .,-4
I
I
H 2 S P E E D WITH N3 PLUGGING
/
/
pumping o f H2 C03 p u m p e d . ( 8 )
I
I
[3
-3
I
-
~
[]
[]E
400
H2 S P E E D WITH C02 PLUGGING H 2 S P E E D WITH CO P L U G G I N G
Z o
200
_
x (~ I
0 10
-5
1
I
1 10
_
i -4
TORR L I T E R S / c m 2 OF E I T H E R Figure 4.2.4. as a function
X x~
lo -a N2,
CO OR C03
S t 7 0 7 ® NEG p l u g g i n g for the pumping of Hz of the amount o f N z , CO a n d COe p u m p e d . ( a )
Speed measurements were conducted with the material at RT and with a system pressure ranging from 10 -6 to 10 -s Torr. Actual gas conductances to all of the NEG material were unknown. Therefore, the total reported speed data are given. However, because of the comparatively low pumping speeds per unit area, reducing these data to units of Z/sec-cm 2 would give fairly reasonable results. The amount of 101 NEG material reported in each panel was --58 g, where the amount of 707 NEG was --70 g. From this, and knowing the reported approximate densities of 101 and 707 materials in strips, we speculate that the area of NEG material on these panels is
NONEVAPORABLE GETTERS
237
-0.37 m2. This proves to be -x2 larger the mechanical description of the panels. Though the NEG speed for H2 can be drastically passivated by pumping the other gases, apparently some minute amount of H~ pumping still occurs, and a perfect H 2 dissociation barrier is never created.(e, T, a, x 0 ) This fact may eventually prove fundamentally troublesome to fusion-related NEG applications. As noted in most of the above references, sorption of N2 by these NEGs has a negligible effect on the reduction in H2 speed. In fact, there is some hint of possible gas displacement effects of N2 by, for example, O 2,(6 ) and perhaps CO 2 displacement by H 2 O,(1 0 ) though I am unaware of specific studies to date in this area. Systematic studies of possible displacement effects might prove enlightening. Miller reports that volume changes will occur in pure Zr as a consequence of H2 sorption.(2 1 ) For example, sorption of 60 Torr-Z H~/g of Zr will result in an 8.2% volume change in the Zr. This volume change results from a change in the crystalographic structure as a consequence of sorbing the H2. Similar to spalling effects in sputter-ion pump cathodes, stresses occur because of localized volume changes in the matrix. These stresses, due to volume changes, result in the NEG flaking or popping off the surface of the substrate. Because of this, limits are set on the permissible quantities of H 2 which may be pumped prior to coating embrittlement. These limits are more empirical than theoretical as the volume changes progress with increasing quantities of H 2 pumped.(2 x ) The equilibrium pressure of H 2 vs. the amount of H2 sorbed by the bulk of both 101 and 707 materials is predictable, and said to follow Sievert's Law,( ~ s ) at least at high concentrations and temperatures. That is, the log 1 o of the equilibrium pressure is: logP where and
= A + logp 2 + B / T
(4.2.1)
P = the equilibrium pressure of. H 2 in Torr, p = the amount of H 2 sorbed m Torr-Z/g of getter, T = the temperature, * K.
The constants A and B are 4.8 and -6116 respectively for the St 707~ material and 4.82 and -7280 respectively for the St 101® material.( 2 4 ) From (4.2.1) we are able to construct a family of curves, called adsorption isotherms, for the two materials. Fig. 4.2.5 shows some theoretical adsorption isotherms for the 707 material for two temperatures. After pumping, at 450" K, a quantity of H2 approaching some desired limit (e.g., pressure), say point #1 of the curve in Fig. 4.2.4, an auxiliary pump may be attached to the system, and the getter heated to the temperature of point #2 of the figure (i.e., 600* K). At this temperature, the equilibrium pressure is higher. Therefore, if one pumps on the NEG, the hydrogen may be removed from the system. The rate of outgassing will depend on the speed of the appended pump. That is, the pump must be able to achieve pressures lower than that of the equilibrium pressure of the NEG at point #2 or gas will not be removed from the NEG. Auxiliary pumping is maintained until equilibrium point #3 of the curve is established. After this, the auxiliary pump is valved out of the system, and the temperature of the NEG reduced to the original operating temperature, yielding the desired H 2 equilibrium pressure of point #4. Because of the very high H x (and perhaps H 2 ) diffusivity of the NEG materials, with sufficient pumping speed, most of the H 2 can be desorbed from the NEG
238
NONEVAPORABLE GETTERS
in a matter o f - 3 0 minutes. This of course depends on the chosen activation temperature. Boffito, citing work of Ferrario, gives the time/temperature equation for H 2 desorption from both the 101 and 707 materials.(2 4 ) Ichimura, et al., report on their own work and that of others relating to the activation energies and heats of absorption for all of the hydrogen isotopes, and NEGs including the 101 and 707 materials.(2 6 ) -3 10
Z i
10
_
111
1 i11[
r
i11~1
II,_
-.
m
o 10 F i g u r e 4.2.5.
Theoretical hydrogen isotherms o f t h e NEG 707® at two temperatures.(24)
adsorption
10
St
Z POINT
POINT
#3-
z 10 r~
/ -
T
-
o
'
-
\
_
.'~
_
~
_
-8
-9
#4 10
I
0.01
~.~
HYDROGEN EMBRITTLEMENT LIMIT -~
ERATI
llJ 0.1
1.0
10.0
TORR-LITERS H Z/g
Regarding pure Zr, Dushman noted that with the interstitial diffusion of oxygen, "... the volume of hydrogen sorbed at saturation is decreased by that of a volume equivalent of that of oxygen present."(2 ~ ) For a given pumping array, this has the effect of the equivalent of reducing the amount of getter material in (4.2.3) in direct proportion to the amount of oxygen consumed per gram of material. However, Boffito and his colleagues showed this increased H 2 equilibrium effect to be of no consequence in NEGs sorbing, for example, CO far above their rated capacities.(2 4 ) However, only high temperature results (i.e., ~ 475* C) were given. Assuming Sievert's Law is applicable, we may use (4.2.1) to make a simpleminded estimation of the relative H2 capacities of the 101 and 707 materials. Methods for calculating relative capacities are part of the problem set. However, theoretical equilibrium pressures for three levels of H 2 loading of the two materials are shown in Table 4.2.2. Hseuh, et al., reported a departure from Sievert's law for H2 sorption in the 707 material.(1 g ) Hseuh's data were taken at temperatures < 5130" C, and for H2 concentrations up to -50% of the embrittlement limit. The implication of his findings was that molecular hydrogen diffusion, rather than atomic hydrogen diffusion,
NONEVAPORABLE GETrERS
239
was occurring in the bulk material. I have total confidence in Dr. Hseuh's data. On the other hand, it has recently been shown by Ghezzi and De Angeli that, for reasons including gauge calibration challenges, it is very difficult to make accurate hydrogen pressure v. temperature measurements for various concentrations .(a a ) Colleagues at SAES also have difficulty with Hseuh's results. Table 4.2.2. Theoretical assuming Sievert's Law
equilibrium valid for
is
NEG
TEMP.
400
600
753
St
HYDROGEN
MATERIAL 293
pressures
the
St
101 ®
St St St
for
SORBED
0.318
hydrogen
101® and
St
707®
pumping materials.
Torr-Z/g
-
3.18
11.6
9.5
x
10 - 2 2
9.5
x
10 - 2 o
1.3
x
10 - 1 8
707 @
8.5
x
10 - 1 8
I].5 x
101 @
4.2
x
10 - 1 5
4.2
x
10 -16
1.1
x
10 - 1 4
10 - 1 3
5.6
x
707 @
3.3
x
10 - 1 2
3.3
10 - 1 2
x
10 - l °
4.4
x
10 - 9
St
101 @
4.9
x
10 - 9
St
707 @
4.1
x
10 - 7
4.9
x
10 - 7
6.5
x
10 - 6
4.1
x
10 - 5
5.4
x
St
101 @
1.4
x
10 - 4
10 - 6
1.4
x
10 - 4
1.9
x
10 - 3
St
707 @
4.8
x
10 - 5
4.8
x
10 - 3
6.4
x
10 - 2
In Fig. 4.2.6, Hseuh's data are given in an unorthodox format, showing isosteres of his data along with the theoretical isosteres of the 707 material. Only data for coverages in excess of one molecular layer are given. -2
0
10
_--
I
I
-3
~
I
1
l
l
l
1
10- 4
( '/
[B
-6
~
O/
-
/
-7 10
/
~/
/
_ _ -
-/
/
/
/
C:]
/
/l
/
of
i
I I
~/
J
I
1
500
Comparison
the
/MEASURED WITH 0 . 3 1 8
TORRt
1
600
TEMPERATURE 4.2.6.
:-
A S I EWVTEHRT':31L~W
@
400
isosteres
/
/
//
/ -8
10
/
/
I J Ji
/
_
,
/
/
10
I-
E]
E] -
l
V1
/~o
-5
10
1
/
I
-
Figure
l
I /g
/
S I E V E R T ' S LAW / WITH 1 1 . 6 f TORR-\Z /g /o
z r~ 0
I
MEASURED
-_
0
~
1
_--TORR- ~f/g
10
I
f
St
707 ®
1
1
l
t
-
~/g
700
1
1
1
1
800
-OK
measured and theoretical H2 NEG for two concentrations.(19)
of
240
NONEVAPORABLE GETTERS
We see that at very high H2 concentrations, the data and Sievert's Law begin to converge. But, as pointed out by Hseuh, at lower concentrations they diverge to the point of suggesting low temperature equilibrium pressures many orders in magnitude higher than predicted by Sievert's Law. Conceivably, the presence of C, N and O in the bulk material might exacerbate this divergence from Sievert's Law at low H2 concentrations. This is suggested by the data of Boffito, et al.(2 4 )
The Pumping of Hydrocarbons It is not evident that CH 4 is synthesized on activation of either the 101 or 707 materials. However, it is evident that in the absence of an ionizing beam or hot filament,(2 8,2 9 ) there is little or no NEG pumping of this gas at RT by either material. This has been verified by numerous studies. Ichimura, et al., did note the presence of several hydrocarbon peaks as a consequence of heating both the 101 and 707 materials.(2 s ) However, I suspect these hydrocarbons stemmed from sample contamination rather than hydrocarbon synthesis by the NEG. Emerson, Knize, Cecchi and Auciello made pumping speed measurements of 101 material for the gases CH4, C 2 H 6, C3 H 8 and C4 H 1 0.(3 0 ) They used a panel coated with 101 material, which they were able to heat by passing current through the substrate material. Therefore, there was no dissociation of the above gases by some filament or heater used to indirectly heat the NEG (i.e., such as is used in the modular pumps). They then developed a theoretical expression for the speed of these gases on the 101 material. Results of their test data are given in the below Table 4.2.3. I have reduced their data to units of Torr-Z/sec-cm2, assuming the active surface area of the reported pump panel to be --3600 cm 2. Speed measurements were made at pressures of the order of 10-4 Torr using VdP/dt, as indicated by a total pressure gauge, as the measure of speed. This suggests that even though there may have been dissociative by-products created in the process of pumping the larger molecules, values of speed obtained, neglecting ionization cross section differences, were net values of speed for all jzases. Similar data were reported by Emerson, et al., in the pumping of cycloalkanes.("3 t ) Because of the negligible pumping speed of all NEGs for CH4 at reduced temperatures, some form of auxiliary (i.e., combination) pumping is usually used in conjunction with a NEG.(t s, 2 o )
Table 4.2.3. Pumping speed of the several hydrocarbons as a function TEMP.
NEG
oC
SPEED
CH4
-
S t 101 ® NEG f o r of temperature.(30)
IAters/sec-cm
C2 H6
z
C3 Ha
C4 H10
600
1.9
x
10 - 3
7.0*x
10 - 3
1.3
x
10 - 2
1.1
x
10 - 2
400
2.0
x
10 - 4
2.5
x
10 - 3
5.6
x
10 - 3
6.1
x
10 - 3
300
3.8"x
10 - 5
1.2
x
10 - 3
3.0
x
10 - 3
3.9
x
10 - 3
10 - 4
1.3
x
10 - 3
2.2"x
200 * Data
4.2"x extrapolated.
10 - 3
NONEVAPORABLE GETTERS
241
Pumping Speeds for Gases and Gas Mixtures It is of value to be able to predict intrinsic NEG speed per unit area vs. the amount of surface sorption at any time. If there is a decay in speed for a given gas on successive regenerations, we must also know the number of regenerations, and how these regenerations relate to the total bulk sorption of gas. With data such as these, one can then design a system using these NEGs. Much of the reported data, though having value in terms of determining surface and bulk capacities, is difficult to relate to intrinsic NEG speed. 150
1
!
1
I
I
1
I
1
1
6 oq r~ .,-t
co p
100
CO J J ~ SPEED
I
f~
/
50
L
0
Vi-,~I
10-5
10-4
10-3
TORR L I T E R S / c m z F i g u r e 4.2.7. S p e e d N2,
CO a n d
COz
for the gases of t h e St 101 ® NEG m a t e r i a l a f u n c t i o n of t h e q u a n t i t i e s p u m p e d . ( 8 )
as
150
• mmmmn
1
1
I
1
6 © v~
C02 /SPEED
.~ .,.~ 1 0 0 CO• ___.__,.------------< SPEED N2 SPEED-~
50
0
i 10-5
L 10-4
-3
10
TORR L I T E R S / c m z F i g u r e 4.2.8. Ne,
S p e e d o f t h e St 707 ® NEG m a t e r i a l for the gases CO a n d COz a s a f u n c t i o n of t h e q u a n t i t i e s p u m p e d . ( a )
Where gases such as CO, CO2 and N2 interfere with the pumping of H2, the converse is not true. That is, H2 does not interfere with the pumping of the other gases. Because of this, the data given by Ferrario for the speeds of the 101 and 707 materials for H ~, taken during the simultaneous pumping of each of the gases CO,
242
NONEVAPORABLE GETTERS
CO2, and N2, (i.e., Figures 4.2.3 and 4.2.4), may be interpreted as being independent of the presence of H 2.(8 ) These data are shown in the above Figures 4.2.7 and 4.2.8. Also, much data exists on the pumping speeds of these materials at elevated temperatures.(9,1 a ) Sciuccati, et al., give speeds of the 707 material at several temperatures, including RT. However, there is some question regarding the amount of NEG per unit area (i.e., noted as 0.2 g/cm 2 ).(a 4 ) Similar 707 data, reported earlier by. Boffito, et al., indicated a surface density of NEG material of -0.3 g/cm2 .( 4 ) The speed data of the 101 material used in the CERN LEP, published by Benvenuti(s, 1 4,1 s ) and Reinhard,(1 6 ) were most helpful in establishing intrinsic RT speed values for this material. These data, shown in Fig. 4.2.9, were taken for individual gases rather than a gas mixture. However, when used in conjunction with data in Fig. 4.2.3, they are useful in predicting mixed gas system behavior. In reducing the CERN data into the common units which I have chosen, I have assumed that the density of the sintered 101 material was -2.3 g/cm a, and that the thickness of the film was -90/jm.( s ) 10
4
-
I
I II
1
I I1
I
I 11
I
I I I _- I
/SPEED -
¢9
~10
3
rn rn~.
(1) -p-4
I
1 H2
_
lO
/
El
N2
-SPEED rj'j
10
1
CO
_
SPEED
2
fq El
J
10 -7
I ii
I
10 -6
I IL
L
10 -5
1
l~l
10 -4
1 II
10 -3
TORR LITERS/em 2 Figure 4.2.9. Speed of the St 101 ® NEG m a t e r i a l for the gases H2, N2 and CO as r e p o r t e d by B e n v e n u t i and Reinhard.(15,16) The use of 101 has produced very successful results at CERN. During initial machine operation, a great deal of gas is desorbed from the walls of the beam pipe as a consequence of synchrotron radiation. However, the amount is predictable, and decreases in some relation to the product of machine beam current and operating hours. Because of these high initial gas loads, people at CERN predicted that it would be necessary to recondition the NEG 4 - 1 times during the first year of operation.(1 6 ) Benvenuti reported that they planned to recondition rather than activate the NEG.(1 s ) This reconditioning amounted to heating the NEG to a temperature of--400* C for a few minutes, where the activation process, as defined by Benvenuti, requires heating the NEG "... at 700"C typically for about half an hour ..." He reported that this reconditioning cycle was sufficient "... to restore pumping speed." We suspect from results of Fig. 4.2.1, that though the lower temperature regeneration might prove of value in H 2 outgassing, it would have limited benefit in restoring CO and CO2 speeds.
NONEVAPORABLE GETTERS
243
Gr8 bner, et al., indicate that during the first year of operation it was necessary to "regenerate" the LEP NEGs three times; two times within the machine operatinz time interval, and once at the conclusion of the first year of LEP operation.( 1 ~, 1 a The temperature of reconditioning, though not given, was implied to be in excess of 450* C. After the first two regeneration cycles, they indicated that "... the pumping speed for CO and CO 2 had decreased from 500 1/s/m to approximately 70 1/s/m." The latter speed was observed after pumping -0.12 Torr-Z CO/m and 0.13 Torr-Z CO 2/m. Using my common units, these two speeds correspond to -0.93 Z/sec-cm 2 and -0.13 Z/sec-cm2, respectively. The former seem -.x 4 lower than the original CO speed data reported by Reinhard and Benvenuti. Normalized CO and CO2 loading corresponds to -0.011 Torr-Z CO/g and -0.012 Torr-Z CO2/g of NEG material. Ferrario indicates that the 101 material reaches a point of surface saturation after pumping -0.03 Torr-Z CO/g and --0.04 Torr-Z CO2/g.(a ) However, in this case, speeds for the two gases approached zero, rather than the f'mite minimum speed needed to afford reasonable beam lifetimes in the LEP collider. Regarding exposure of the 101 material to atmospheric pressures of various gases, Reinhard reports that "After 5 air exposures, the pumping speed of the reactivated NEG starts decreasing, and after 30 exposures to air, it is only about 30% of the initial value."(1 e ) Regarding N 2, Reinhard reports that "... 30 ventings would result in a pumping speed reduction of 10% only." These results make evident the advantages of venting the "101" NEG material to N 2. Results of these air exposures provide some valuable insight into the total capacity of the NEG for gases such as O 2, CO, CO 2, etc. That is, assume that each time the NEG is exposed to air, a layer of ZrO~ is formed on the surface. On subsequent pumpdown, the only way the ZrO2 can be removed is through activation. That is, the getter must be heated and the ZrO2 reduced to metallic Zr through O1 diffusion into the bulk of the getter. Assume the 101 material has an effective surface area of -..x 100 of that of the geometric area.(1 5 ) On the initial exposure to air, the total amount of 0 2 sorbed will be --7 x 10 ~ e molecules/cm 2.(3 2 ) As a rough approximation, assume that the O 2 speed of the NEG, immediately after activation, is directly proportional to the number of available fresh Zr sites, and that the number of sites decays linearly with each exposure. This being the case, if after 30 air exposures the NEG speed is --x 0.3 that of its original speed, then the average amount of 0 2 sorbed as a consequence of each air exposure is -4.5 x 10 z e O2 molecules/cm 2 of effective surface area. Therefore, the total amount of 0 2 sorbed after 30 air exposures is --1.35 x 101 a molecules per cm 2 of effective area. With a density o f - 2 . 3 g/cm a and a 90/.tm thickness of NEG, we conclude that the speed of the NEG was reduced to 30% of its original value as a consequence of pumping -2.0 Torr-Z/g. These calculations are probably conservative. For example, Halama and Guo have shown, using 707 material, that even after cycling the material to air as much as 68 times, and activating between each cycle, they were able to achieve repeated surface coverages of >60% for CO and CO2, and still observe measurable speeds for these two gases (i.e., see Fig. 4.2.10, below).(3 3 ) Therefore, it is possible the 101 NEG may have sorbed as much as -3.0 Torr-Z/g as a consequence of 30 air exposures. These results are in reasonable agreement with 101 data of Table 4.2.1. The most definitive evaluation of the 707 material to date has been done by Halama and Guo.( a 3 ) Their interest in this study was to determine if the 707 material also possibly had applications in electron storage rings and accelerators.
244
NONEVAPORABLE G E T r E R S
Because of this, they conducted speed measurements for this NEG material, using a gas mixture comprising 50% H 2,35% CO and 15% CO 2. Speed measurements for these individual gases were also conducted. The pump comprised a strip of 3.0 cm wide, 2.2 m long NEG (see Fig. 4.1.1) coated on both sides with 707 material. The strip was mounted in an 8.8 cm ~, 2.5 m long stainless steel tube. Before each measurement, the entire system was baked at 200* C for 48 hours, and the NEG activated by heating it to a temperature of 400* C - 700* C. However, they reported the optimum activation temperature for this material as being 400* C - 450* C, for a duration of--30 minutes. Speed and capacity measurements were taken at RT. In normalizing Halama's data, I have assumed a total NEG surface area of 1188 cm 2, a coating thickness of 70 # m, and a density of 707 material of 2.65 g/cmS. The speeds for the individual gases H 2 , CO and COs were comparable to their respective speeds when pumped in a gas mixture. They noted that when pumping H 2 , the speed "... remains high (after pumping) up to several Torr-Z/m." Results of pumping the above individual gases are shown in Fig. 4.2.10. 400
I
i li
1
I tl
I
I li
1
I
1
11
I
il
NEW >,~ ~ ~ . 3 NEG
oi 300 C 0 2
0q
"~ 200
_ NEG
AFTER
68 AIR & lO N2 EXPOSURES
-
NEW NEG
[] <
CO
[] E3_3
[]
[] []
×
2 e,~
H2
100
03
C02 -~
~~-E~
~
× q
I
10 . 8
I I
10 . 7
I 10 . 6
1I1
I
10 . 5
II
[] []
-
I I I I~~?1 10 . 4
10 - 3
TORR L I T E R S / e m 2 Figure 4.2.10. Speed of the St 707®NEG m a t e r i a l for the individual gases H2, CO and C02 as a new NEG and after exposed to b o t h air and N z, and each t i m e activated. (as) Halama and Guo showed a spread in speed data for a freshly activated, unused NEG (i.e., equivalent to error bars). This variation in speed ranged approximately 4-25% about some average. I have reported the average speeds in Fig. 4.2.10. You will note that there was little difference in the initial speeds for the three gases. In a second series of tests, Halama and Guo measured the speed of an NEG for the mixed gas, as a function of number of exposures to air and N 2, for a total of 68 air exposures and ten N 2 exposures. This was an outstanding series of experiments, and the results provided significant insight into the properties of the 707 NEG. After each air or N2 exposure, the NEG was reactivated, the system again baked, and NEG speed measured for the gases H2, CO and C O s . Data, showing the initial
NONEVAPORABLE GETI'ERS
245
speeds at the end of the 68th regeneration cycle, are also given in Fig. 4.2.10. Also, after each regeneration, speed measurements were conducted until a quota equivalent of --9.26 × 10-4 Torr-Z/cm 2 of H2, --6.48 × 10-4 Torr-Z/cm 2 CO and -2.79 × 10-4 Torr-Z/cm 2 CO2 was pumped. I call this the used NEG. Therefore, they were able to establish the effect of both air exposure and the pumping of substantive quantities of the above three gases, on NEG performance. In a third series of tests, they measured only the initial speeds of a new NEG for H2, CO and COs after exposure to air, activation and a rebake of the system. By conducting these measurements, they were able to distinguish between differences which existed in NEG speed when exposed only to O 2, H 2 0 and N 2 (i.e., the new NEG) vs. both air exposures and the pumping of the gases CO and CO2 (i.e., the used NEG). If a major difference existed between results of these two tests, then changes in NEG speed as a consequence of exposure to air would not be sufficient to characterize the irreversible bulk sorption capacity of the NEG. This proved to be the case, as is shown in Fig. 2.4.11. That is, the speed of the NEG which had pumped the first quota of CO and CO2 (used) was irreversibly reduced by --×5.6 over that of a NEG (new) which had simply been exposed to air, regenerated and only briefly pumped H2, CO and CO2. Also, the difference in speeds diverged to --×20, on further cycling. The implication in this is that oxygen diffuses much more readily into the NEG, than carbon. One wonders how much CO and CO2 had to be pumped by the NEG with no gas loading prior to establishing somewhat steady-state speeds. That is, would O2, alone, have had any significant effect in decreasing the NEG speed? 10 q I
__~~.J-
NEO WHICKI~U~PED OASES /
0q
SUFFICIENT ONLY TO MEASURE SPEED
H2 "-SPEED
0q q) .,=~
~
10 2
K----)~ ~____x__~~.__>~_~~__x___ x
CO SPEED
I
Hz 4 - SPEED i0
I
~..__
CO SPEED
NEG WHICH PUMPED H2, CO AND C02 AFTER EACH ACTIVATION O
10
0
0
10
20
30
40
50
60
70
NUMBER OF NEG E X P O S U R E S TO AIR AND ACTIVATIONS Figure 4.2.11. S p e e d of a n S t 7 0 7 @ NEG f o r t h e g a s e s H z a n d C0; (33) o n e NEG m e r e l y as a consequence of exposure t o a i r " Lhe s e c o n d NEG f o r t h e s a m e r e a s o n and as the result of each time having pumped 0 . 5 , 0 . 3 5 a n d 0 . 1 5 T o r r - ~ f / / m H2, CO a n d COz r e s p e c t i v e l y .
The capacity of the 707 NEG material to sorb gases is remarkable. For example, because the NEG evidences speeds at what correspond to >60% surface coverage at the end of each test sequence shown in Fig. 4.2.11, we must conclude that on air exposure, the NEG must sorb something equivalent to a molecular layer of gas. At the end of the 68th test cycle of Fig. 4.2.11, the NEG had consumed the equivalent of
246
NONEVAPORABLE GETI'ERS
-2.4 Torr-Z CO/g and -1.0 Torr-Z CO2/g. Assuming 68 air exposures and a consequential 100% coverage, we calculate that used NEG, at the conclusion of the 68th exposure, had also sorbed the equivalent o f - 2 . 1 Torr-Z/g of 0 2 . Of course, the new NEG pumped the equivalent of -5.5 Torr-Z 0 2 / g at which time the CO speed was - 2 4 % of the initial speed. Halama and Guo noted that exposure of the 707 NEG to N2 had "... only a small negative effect on NEG behavior (speed)." Because of this, it is suggested that systems containing NEGs be vented to dry N2, rather than air. Benvenuti and Reinhard made similar observations regarding the 101 material.( x s, 1 6 ) One must be careful in interpreting initial, low-coverage, speed data in any baked system having a significant surface area. Also, in the case of the "101" material, if the geometric surface area of the system proves to be x5 to x 10 larger than the geometric surface area of the NEG, it is conceivable that A~, which is probably sublimed from the NEG during a high temperature activation (e.g., 800* C), might cover the vacuum system surfaces and lead to erroneously high initial NEG speed data for CO and N 2 after each activation (e.g., see problem 12).
4.3 Sintered NEG Structures Sintered NEG structures first appeared in 1989 (the St 172~ material). This material comprised granules of sintered St 707, sintered-bonded in a matrix using Zr powder as the bonding glue.(4 o ) In late 1993 a second type of sintered structure was offered. This comprised a mixture of 70% Ti and 30% V (by weight), and this NEG was given the name St 185®.(4 1 ) Blades or disks of both the 172 and 185 materials were fabricated to fit in structures having the same envelope as flanged SORB-AC ~ modular pumps. Blades ranging in size from 0.8 mm x 20 mm x 72 mm for the 185 material and 1.6 mm x 19 x 75 mm for the 172 material are noted.(4 o ) Smaller pumps comprising 0.8 mm x 24 mm x 24 mm stacked disks of the 185 material were also reported on.(4 1 ) These new pumps were given the name CapaciTorr TM-B getter pumps when featuring the 172 material, and CapaciTorr~-D pumps when exploiting the 185 materials. An example of such a pump module, mounted on a 150 mm ~/, ConFlat ~ flange, is shown in Fig. 4 3.1. Both the SORB-AC ® and CapaciTorr TM pumps may be mounted on flanges and penetrate into the vacuum system to afford very high pumping speeds. On the other hand, in that these pumps must be regenerated at temperatures of as much as 500* C, the extraneous heating of parts in the vacuum system might prove troublesome. In such instances, the pumping module can be recessed from the system in a water-cooled jacket. The trade-off in such instances, as with any pump, is a reduction in speed delivered to the system. There is a suggestion that there may be particulate problems with the 172 sinterd matrix. This prompted Giannantonio, et al., to switch from use of the 172 material to the newer 185 material in the experimental region of the D A ~ N E main ring.(4 o ) No data are given that particles would "debond" from the 172 materials, but it was rigorously demonstrated not to be a problem with 185 material. The bladed CapaciTorr TM pumps have the distinct advantage of having both greater speeds and higher gas capacities than comparably sized SORB-AC ~ pumps. This was demonstrated by Li, et al.,(4 1 ) when comparing comparable sized pumps with the 707 v. the 185 NEG materials. They reported initial CapaciTorr TM-D CO speeds, after regeneration, -40% higher than measured with a pump using the 707
NONEVAPORABLE GETTERS
247
material. SAES reports that the improvements are even greater, noting speed improvement of---×2.3 for CO and --×2.7 for H~ at RT, and immediately after regeneration. The disparity in speed becomes even greater with increased quantities of gas pumped.
.
.
i l .....
-
¢
.
.
Figure 4.3.1. Nude CapaciTorr TM -B Pump Mounted on a ConFlat ® Flange and With Screens Surrounding the Pumping Blades.( ~ a )
iii?
Figure 4.3.2. Nude CapaciTorr TM -D Pump Mounted on a 2-3/4" ConFlat ® Flange and With Screens Surrounding the Pumping Blades.(4 a )
248
NONEVAPORABLE GE'Iq'ERS
Some Interesting Applications A perusal of the WEB, as well as the literature, reveals a number of interesting NEG applications. The SAES company has a significant presence in serving the argon sputtering gas purification needs of the semiconductor industry. It takes little imagination to understand how these getters work in this application. Another interesting use of NEGs, not yet in full commercial use, is the WaferGetter. In this application a silicon wafer is coated with a NEG material. On introduction into the chamber it is heated by placement on some thermal pedestal. The robotics for inserting and removing the wafer, and wafer heating system, are already integral parts of the system. SAES customers have shown that inserting these portable pumps in systems will aid in chamber conditioning time, and improve rate-of-pressure rise performance subsequent to their removal. Benvenuti, et al., recently reported on the use of NEG materials sputterdeposited directly onto the walls of vacuum chambers.(4 2 ) The films were deposited using coaxial magnetron sputtering, the chamber wall serving as the substrate. The target material comprised twisted wires of Ti and Zr. The Ti/Zr proportions of the deposited film could be tweaked by altering the number of twisted wires serving as the target material (e.g., two Ti, one Zr; one Ti, one Zr; one Ti, two Zr). The of effectiveness of the film was quantified by measuring the relative electron stimulated desorption (ESD) of the surface films. This was difficult to do, because the films themselves sorbed ESD gases. Therefore, Benvenuti and his colleagues used the relative measure of CH 4 desorption as a measure of the merit of the coating. These experiments, and there will be many more of this nature reported in the near future, were probably in part motivated by the work of Halama and Goa.(a 3 ) They showed a significant improvement (i.e., decrease) in the photodesorption gas yield of a NEG strip, over that when a chamber was merely augmented by a NEG strip removed from the photon beam.
4.4. Advantages and Disadvantages of NEGs NEGs are effective room temperature pumps for all the chemically active gases, excluding hydrocarbons such as CH 4. They do not pump the inert gases. Also, they retain and even have improved sorptive capacities for many gases when operated at elevated temperatures. They do not require the use of high voltage, will function in very hostile radiation fields, and have no preferred operating orientation with respect to gravity. NEGs present no vibration problems to a system. They are neither a source of nor perturbation to magnetic fields when operated at RT. The 707 material is not a source of sublimed material to the vacuum system. However, some investigation is needed regarding possible vaporization of A~ from the 101 NEG during activation. On the one hand, they are low throughput devices. On the other hand, when activated, they provide very high speeds per unit area while at extreme high vacuums. Materials are supplied in both sintered tablet form and bonded to metallic surfaces. Metal strips coated with NEG material may be bent and formed into a wide variety of shapes to suit a particular application. Strips of these materials have been successfully used in both modular and distributed pumped forms. Based on published data, it appears that gases containing carbon cause far greater reductions in speed, on regeneration, than do O 2 and N 2. The reversible pumping
NONEVAPORABLE GETTERS
249
of H2 proves to be an asset, where the embrittlement of these materials, when pumping in excess of --20 Torr-Z H~/g of NEG, presents a risk of problems if not heeded. Based on SAES data sheets, there is some indication that thermal cycling of the material at excessive temperatures for numerous cycles may cause NEG peel-off problems. This has not been reported as a problem in the literature. However, the spontaneous popping off of small pieces of NEG from the metallic substrate, due to H 2 embrittlement, has been cited. The need for high temperature activation requires attention to heat transfer design details and differential expansion considerations, when integrating these NEGs into systems. Activation of the 101 material requires temperatures of the order of 700* C - 800* C, where the 707 material activates at temperatures of the order of 400* C - 450* C. Assuming similar and constant (i.e., vs. temperature) spectral emissivities for the two materials, the relative power requirements to activate the two materials differ by - x 4. A 3 cm wide strip of the 101 material, in an extruded A2 chamber, requires -900 W/m power to achieve 700* C.(1 s ) A method for calculating power radiated from a NEG vs. temperature is given in problem 10. Theoretically, it appears that the 101 material has greater H ~ capacity than the 707 material (e.g., see Table 4.2.2). This may be true at very high H 2 concentrations. However, it has been suggested that Sievert's Law may not be applicable low H2 concentrations in the 707 material (i.e., UHV applications). This may also be true for the 101 material. No such data were found. The cost of the 707 material is slightly higher than that of the 101 NEG. However, this difference is minimal, and it appears that the 707 material has decided advantages in terms of irreversible bulk capacity and post-activation speed.
Problem Set 1) Assume you wish to construct a pump with St 101® material. The pump is constructed in a fashion similar to that shown in Fig. 4.1.3. Assume that the density of the 101 material is -2.3 g/cm3, and that the NEG coating is 70/.t m thick. What is the minimum geometric surface area of NEG required to pump 1.2 x 10 4 Torr-Z H 2 while at the same time not exceeding the embrittlement limit of the NEG material? 2) Assume that you have a substrate panel coated on both sides with 101 material, of the density and thickness noted in problem 1, and measuring 2 m x 2 m. Using the data of Fig. 4.2.9, what would the approximate zero loading, panel speeds be for the individual gases CO, N 2, and H 2 ? 3) Using data of Figure 4.2.3, what would the approximate speed be for H2 after pumping -1.6 Torr-Z N2; -8.0 Torr-Z N2; 8 and 25 Torr-Z CO; 8 and 40 Torr-Z CO2 ? 4) Using data of Figure 4.2.7, what would the approximate speed be for the individual gases after pumping -1.6 and 6.4 Torr-Z N2; -13.4 Torr-Z N2; 16 Torr-Z CO; 16 Torr-Z CO2 ?
Problem Set
250
NONEVAPORABLE GETTERS
5) Assume the panel in problem 2 is exposed to air 30 times and activated each time thereafter. If the panel initially has a speed of --1.0 Z/sec-cm2 for CO, what would the total CO speed of the panel be after the 30th activation? 6) Assume a x 100 effective surface area and 100% O 2 coverage each time the N E G in problem 5 is exposed to air. How much O 2 would have been sorbed by the NEG after the 30th activation. Express your answer both in total Torr-Z 0 2 and in Torr-Z/g of NEG. 7) Using data of Halama and Guo, and assuming the use of 707 material with a density of -2.65 g/cm a , calculate the answers to problems 2 - 6 for this material, excluding N2. Assume a x 70 effective surface area for the 707 material. 8) The power radiated between two parallel plates is given by: W where
= o A [(1/E n) + (1/~ w)" 11-1 [(Tn)' - ( T w ) ' 1, W O
A n Ew
and
Tn Tw
= = = = = = =
power in Watts radiated per cm 2, 5.67 x 10" 1 2 Watts/* K 4 -cm 2 , the area of the panel in cm 2, -0.6, the total spectral emissivity of the NEG, --0.1, the total spectral emissivity of the wall, the temperature of the NEG in * K, the temperature of the wall in * K.
What would be the power, radiated to the RT walls of a vacuum envelope, from a NEG 30 mm wide and 1.0 m long for the temperatures 400* C, 700* C and 800* C? 9) Assuming the same amount of 101 and 707 material, that Sievert's Law is applicable, and that P1 o 1 and P7 o 7 are H 2 equilibrium pressures of the 101 and 707 materials respectively, what is the numerical value of the ratio P1 o 1 / P T o ~ at RT, 200* C and 400* C? 10) Using data of Hseuh and Lanni, speculate on the possible RT equilibrium pressure of the 707 material with a loading of --10 Torr-Z/gm. 11) Assume the panel in problem 2 is located at the center of a very large space simulation chamber. Because of the size of the chamber, we may assume ~w --1. How much power would be required to activate the NEG at temperatures of 400* C and 700* C? 12) Find a copy of Dushman's book (i.e., Reference 25). Look up the definition of Raoult's Law. Use this to calculate the vapor pressure of A~ in the St 101® material as a consequence of heating it to 800* C.
NONEVAPORABLE GEq~FERS
251
References 1. 2. 3.
4. 5. 6. 7. 8. 9. 10. 11. 12.
13. 14.
15. 16.
Giorgi, T.A., Ferrario, B., Storey, B., "An Updated Review of Getters and Gettering", J. Vac. Sci. Technol. A3(2), 417(1985). Kindl, B., Rabusin, E., "A Large-Surface Nonevaporating Getter", Supplem. AI Nuovo Cimento V, 36(1967). Della Porta, P., Giorgi, T., Origlio, S., Ricca, F., "Investigations Concerning Bulk Getters from Metals of the IVth A Group and Thorium", Proc. 2nd Int. Vac. Cong. and 8th Nat. AVS Symp., 1961 (Pergamon Press, Inc., New York, 1962), p.229. Boffito, C., Ferrario, B., della Porta, P., Rosai, L., "A Nonevaporable Low Temperature Activatable Getter Material", J. Vac. Sci. Technol. 18(3), 1117(1981). Benvenuti, C., Francia, F., "R oom-iemperature Pumping Characteristics " of a Zr-A~ Nonevaporable Getter for Individual Gases", J. Vac. Sci. Technol. A6(4), 2528(1988). Cecchi, J.L., LaMarche, P.H., Dylla, H.F., Knoze, R.J., "Technique for In Vacuo Passivation of ZrA~ Alloy Bulk Getters", J. Vac. Sci. Technol. A3(3), 487(1985). Emerson, L.C., Mioduszewski, P.K., Simpkins, J.E., "Performance of Zr-A.~ Getter Pumps Under Transient Load Conditions", J. Vac. Sci. Technol. A2(4), 1583(1984). Ferrario, B., Boffito, C., Doni, F., Rosai, L., "Zr Based Gettering Alloys for Hydrogen Isotope Handling", Presented at the 13th SOFT, Varese, Italy, 1984. (This paper may be obtained through the authors at SAES.) Knize, R.J., Cecchi, J.L., Dylla, H.F., "Compatibility of the Zr-AI Alloy with a Tokamak Plasma Environment", J. Nuc. Mat. !0..3& 104, 539(1981). Malinowski, M.E., "Decreases in Deuterium Pumping by St707~ Getter Alloy Caused by Carbon Dioxide Preexposure", J. Vac. Sci. Technol. A3(3), 483(1985). Audi, M., Dolcino, L., Doni, F., Ferrario, B., "A New Ultrahigh Vacuum Combination Pump", J. Vac. Sci. Technol. A5(4), 2587(1987). Brouet, M., Girasini, M., Poncet, A., Wolf, A., H~itten, L., Poth, H., Habfast, C., "An Ultra-High Vacuum System for Coolers", CERN Technical Report No. CERN/PS/85-4 (ML), CERN, Geneva, Switzerland 1984. Moenich, J.S., "Nonevaporable Getter for Ion Beam Fusion", J. Vac. Sci. Technol. 18(3), 1114(1981). Benvenuti, C., Decroux, J-C., "A Linear Pump for Conductance Limited Vacuum Systems", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. R~idenauer, F.P. Vieb6ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 85. Benvenuti, C., "A New Pumping Approach for the Large Electron Positron Collider (LEP)", Nucl. Instrum. Methods 205, 391(1983). Reinhard, H.P., "The Vacuum System of LEP", Proc. 9th Int. Vac. Cong. and 5th Int. Conf. on Solid Surfaces, 1983 (Asociacion Espafiola del Vacio y sus Aplicaciones, J.L. de Segovia, Editor, 1984), p. 273.
252
NONEVAPORABLE GE'Iq'ERS
References 17. Gr~bner, O., Laurent, J-M., Mathewson, A.G., "The Behavior of the Synchrotron Radiation Induced Gas Desorption During the First Running Period of LEP", 2nd European Particle Accelerator Conf., 1990 (CERN Technical Report No. CERN-AT-VA-90-10, CERN, Geneva, Switzerland, 1990). 18. Gr8 bner, O. "The Performance of the Ultra-High Vacuum System of LEP and the Experience Gained During the First Year of Operation", 2nd European Particle Accelerator Conf., 1990 (CERN Technical Report No. CERN-AT-VA-90-09, CERN, Geneva, Switzerland, 1990). 19. Hseuh, H.C., Lanni, C., "Evaluation of Zr-V-Fe Getter Pump for UHV System", J. Vac. Sci. Technol. A1(2), 1283(1983). 20. Hseuh, H.C., Feigenbaum, I., Manni, M., Stattel, P., Skelton, R., "Ultrahigh Vacuum System of the Heavy Ion Transport Line at Brookhaven", IEEE NS-32(5), 391(1985). 21. Miller, G.L., Metallurgy of the Rarer Metals, 2 Zirconium (Academic Press, Inc., New York, 1954). 22. Knize, Rj., Cecchi, J.L., Dylla, H.F., "Measurement of H2, D2 Solubilities in Zr-A£ ", J. Vac. Sci. Technol. 20(4), 1135(1982). 23. Ichimura, K., Matsuyama, M., Wantanabe, K., "Alloying Effect on the Activation Processes of Zr-alloy Getters", J. Vac. Sci. Technol. A__~5(2), 220(1987). 24. Boffito, C., Ferrario, B., Martelli, D., "Equilibrium Pressures of H2 and D2 for Different Getter Materials and the Effect of CO Impurities", J. Vac. Sci. Technol. A__!(2),1279(1983). 25. Dushman, S., Scientific Foundations of Vacuum Technique (John Wiley and Sons, Inc., New York, 1958), p. 554. 26. Ichimura, K., Matsuyama, M., Watanabe, K., Takeuchi, T., "Absorption/Desorption of Hydrogen Isotopes and Isotopic Waters by Zr-Alloy Getters", J. Vac. Sci. Technol. A6(4), 2541(1988). 27. Dushman, S., op. cit., p. 584. 28. Della Porta, P., Ferrario, B., "A Magnetless Gauge Appendage Pump Utilizing Non-Evaporable Getter Material", Proc. 4th Int. Vac. Cong., 1968 (Institute of Physics and Physical Society, London, 1968), Part 1, p. 369. 29. Shen, G.L., "The Pumping of Methane by Ionization Assisted Zr/AI Getter Pump", J. Vac. Sci. Technol. A5(4), 2580(1987). 30. Emerson, L.C., Knize, R.J., Cecchi, J.L., Auciello, O., "Dissociative Pumping of the Alkanes Using Nonevaporable Getters", J. Vac. Sci. Technol. A.__44(3),297(1986). 31. Emerson, L.C., Knize, R.J., Cecchi, J.L., "Pumping of Hydrocarbons Using Nonevaporable Getters", J. Vac. Sci. Technol. A___55(4),2584(1987). 32. Redhead, P.A., Hobson, J.P., Kornelsen, E.V., The Physical Basis of Ultrahigh Vacuum (Chapman and Hall, Ltd., London, 1968), p. 41. 33. Halama, H.J., Guo, Y., "Non-Evaporable Getter NEG Investigation at the National Synchrotron Light Source", Proc. 37th Nat. AVS Symp., Oct. 1990, Toronto, J. Vac. Sci. Technol. A.._.99(3),2070(1991).
INONE VAI"ORABLE LiI~*l'l'l~R5
22)3
References ~4. Sciuccati, F., Ferrario, B., Gasparini, G., Rosai, L., "In Situ Pumping with NEG (Non-Evaporable Getters) During Vacuum Processing", Vacuum 38, 765(1988). ~5. Lampert, W.V., Rachocki, K.D., Lamartine, B.C., Haas, T.W., "Study of the Activation of Nonevaporable Getters", J. Vac. Sci. Technol. 18(3), 1121(1981). ~5. Ferrario, B., Rosai, L., "New Types of Volume Gettering Panels for Vacuum Problems in Plasma Machines", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. Rf.idenauer, F.P. Vieb8 ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 359. ~7. Pace, A., Poncet, A., "Monte Carlo Simulations of Molecular Gas Flow; Some Applications in Accelerator Vacuum Technology Using a Versatile Personal Computer", Vacuum 41(7-9), 1910(1990). ~. Ghezzi, F., De Angeli, M., "Problems in Determining True Equilibrium Pressures of Hydrogen Getters Over a Wide Pressure Range of Getter Temperature and Hydrogen Sorption", J. Vac. Sci. Technol. A 17(6), Nov/Dec 1999, pp. 3452-3462. ~9. Romer, J.G.M., "Combined Sputter-ion and NEG Pump", Vacuum 43(5-7), 551(1992). 10. Giannantonio, R., Mannini, P., Mazza, F., Dominoni, D., Cloza, A., Zanin, L., "Design and Characterization of High Capacity Nonevaporable Getter Pumps Embedded Inside the Interaction Region of DA~ NE", J. Vac. Sci. Technol. A 17(4), 2093(1999). L1. Li, Y., Hess, D., Kersevan, R., Mistry, N., "Design and Pumping Characteristics of a Compact Titanium-Vanadium Non-evaporable Getter Pump", J. Vac. Sci. Technolo. A 16(3), 1139(1998). ~2. Benvenuti, C., Chiggiato, P., Cicoira, F., L'Aminot, Y., "Nonevaporable GGetter Films for Ultrahigh Vacuum Applications", J. Vac. Sci. Technol. A 16(1), 148(1998). L3. My sincere thanks to my colleagues at SAES Getters for their critique of the first edition of this book. I am particularly indebted to my friend Michael Ferris of the SAES Colorado Springs office for his helpful technical discussions, serving as liaison in the editing process, and for providing the photographs of the CapaciTorr ® pumps.
CHAPTER 5
CRYOPUMPING 5.0 Introduction Cryopumping is the most elementary of all of the forms of capture pumping. By now you have learned most of the fundamentals needed to understand how cryopumps work. In Chap. 1, you learned the def'mitions of pressure and temperature; additionally, the concepts of vapor pressure, conductance, speed, throughput, and the counting of molecules were all introduced. In Chap. 3, I introduced the concept of the sticking coefficient, ct, of a gas on a chemisorbing surface. It was therein noted that there was a counterpart in cryopumping. This counterpart is called a capture coefficient, and is sometimes denoted by the symbol c. It has the exact same meaning as c~. Therefore, I will stick with c~. Assuming that you have mastered these definitions and concepts, excluding hardware considerations, there is little new to be learned in achieving an understanding of cryopumping. From a physics standpoint, there are two simple effects which describe the mechanisms of cryopumping. One is defined as cryocondensation pumping and the second as cryosorption pumping. These two effects are respectively related to two physical phenomena: one is the vapor pressure of gases as a function of temperature; the second is called an adsorption isotherm of a gas. With a few minor exceptions, these are the only new physical concepts introduced in this chapter. One exception deals with the concept of thermal transpiration; a second exception deals with the phenomenon of cryotrapping; a third exception deals with the use of molecular sieve materials in cryosorption pumping. I will first discuss the above concepts, and then how they relate to the more practical considerations of cryopumps such as their speed, capacity, and blank-off pressures. After this, I will discuss the various forms of cryopumps, including cryosorption roughing pumps (sorption pumps) which operate at LN2 (i.e., liquid nitrogen) temperatures, the elements of liquid He cryo-pumps (LHe pumps) and then conclude with a discussion of closed-loop, gaseous helium cryopumps (CLGHe pumps). References given in the literature on various aspects of cryopumping number in the thousands. Those references given in the tables and text herein are given as representative work. They were selected on this basis and are few in number. 5.1 C r y o c o n d e n s a t i o n vs. C r y o s o r p t i o n Cryopumping is merely the temporary storage of gases on cooled surfaces. There are two defined extents, or measures, of this pumping. One measure is called cryocondensation; the second is called cryosorption. These are merely words used to distinguish between two extents of pumping on cold surfaces.(1 5 g )
Cryocondensation is the pumphlg of gas due to a phase change of the gas into a solid or a liquid (i.e., an ice or frost of the gas).
256
CRYOPUMPING
Cryosorption is the pumping of gas on a sorbent or surface where the amount of gas which can be pumped is dependent on the surface density of the gas, and the equilibrium pressure of the gas over the sorbent or surface. Both cryocondensation and cryosorption pumping exists in part because of the comparatively weak van der Waals' (i.e., dispersion) forces at and near the pumping surface. At reduced temperatures, the thermal energy retained by a molecule (i.e., translational, rotational, vibrational, etc.) is less. In fact, the energy retained by a gas molecule as a function of temperature is one of the methods used to arbitrarily define temperature (e.g., Section 1.4 of Chap. 1). A molecule becomes thermalized when impinging on a cold surface. That is, it gives up heat energy to the surface and in doing so is itself cooled to reduced temperatures. The amount of heat energy which a molecule or atom gives up with each collision with a cooled surface is called the accommodation coefficient. It will vary from 0.01 to 1.0, depending on the gas and surface and their respective temperatures.(1 ) It is generally believed that for the same van der Waals' force, the likelihood of a molecule being captured on a surface increases both with the decreasing temperature of the gas being sorbed and the surface temperature.(2,3 ) Exceptions have been reported regarding the temperature dependency of the gas being pumped. (4, s ) Intuitively, we would speculate that the heavier gas molecules, when giving up sufficient thermal energy to be cryopumped, because of their very mass would be less readily bumped off the surface by another gas molecule once initially sticking to that surface. Therefore, heavier molecules would tend to be more readily cryosorbed on surfaces, including ice of the same and different gas species. Conversely, the lighter the gas molecule (e.g., He, H2 and Ne), the more difficult it is to cryopump at a given temperature. With some notable exceptions, this proves to be the case. The exceptions to this mass-dependency occur when a gas molecule has a measure of polarity, as in the case of H 2 0 . In this case, other forces come into play, and the polar molecules are more readily pumped at comparatively higher temperatures. Of course, there are degrees in molecular polarity. Being polar means that the gas molecule, because of its average internal electrical charge distribution, distorts the electrical charge distribution of surface molecules (e.g., the ice or metal substrate). These charge distortions cause an attractive force between the molecule and the surface, which is additive with the van der Waals' forces. Smooth, round atoms having filled electron shells are nonpolar. This is the case with all of the inert gases. Therefore, primarily van der Waals' forces cause these gases to stick to surfaces. This qualitatively explains why at the same temperatures H2 is more readily cryopumped than He, CH4 more readily than Ar, and H 2 0 more readily than Ne; the former gases are the lighter, yet more polar, gases. A more quantitative measure of the relative strengths of these forces between molecules of the same and other species is found from the boiling points (and critical temperatures) of the respective species. For example, Dushman, reporting on work of others, gives the near-RT (i.e., room temperature) capacity of a carbon (charcoal) surface for different gases as a function of their respective boiling points.(S ) These data, in modified form, are given in Table 5.1.1. We see from the table that to first order, the lighter the molecule, the lower the possible surface coverage of the gas at a given temperature. More correctly, the lower the boiling temperature (or critical temperature) of the gas, the lower the possible surface coverage of the gas at a given
CRYOPUMPING
257
temperature. Note that the equilibrium pressure of the gases in Table 5.1.1 was 750 Torr, and the cryosorbing surface was maintained at -288* K. Nevertheless, this correlation between possible surface coverage and boiling points of gases persists even at very low pressures and temperatures. Table 5.1.1. The charcoal surface,
extent of coverage vs. the respective Torr-:~ /gram
GAS COC12 SOz CHz C1 NH3 H2S HC1 NzO C2 I-I2 C03 CH4
CO
03 N3 H3
MOLECULES 1 /ern 2 1.3 1.1 7.9 5.2 2.8 2.1 1.5 1.4 1.4 4.6 2.6 2.3 2.3 1.4
330.0 285.0 207.B 135.8 74.3 54.O 40.5 36.B 36.0 12.0 6.8 6.0 6.0 3.B
1) Assumed "-,900
m2/g
x x x x x x x x x x x x x X
1015 1015 10 TM 10 TM 10 TM 10 TM 10 TM 10 TM 10 TM 1013 1013 1013 1013 1013
of gas boiling
COVERAGE ? "m
molecules points of BOIUNG POINT OK
1.140
281.3 263.0 249.3 239.7
0.309 0.176 0.258 0.088 0.032 0.026 0.028 0.009
189.3 183.5 184.5 194.5 111.5 81.0 90.0 77.3 20.1
-
on a 288°K the gas.(6)
211.2
of charcoal.
5.1.1 C r y o c o n d e n s a t i o n P u m p i n g We observe the effects of cryocondensation frequently in our everyday lives. For example, from time to time during the winter I find that overnight a thick layer of ice has accumulated on the windshield of my automobile. This is simply the cryocondensation of H2 O on the window. Water accumulates on the surface of a cocktail glass. This too is the cryocondensation of H2 O. The limiting consideration in the measure of the thickness of the ice on the car window is the surface temperature of the ice on the window and the water vapor content in the atmosphere. As long as there is a source of H2 O in the atmosphere and the ice on the windshield is sufficiently cold, ice will continue to build up on the windshield by cryocondensation. REFRIGERATOR BOX AT 30°K ~ - . ~
VACUUM //~VESSEL
W_//I//////I//////A
BAG "
Hil .
,
.
w
vA
II
u im...LLLLq H
Figure 5.1.1. Vacuum vessel maintained at a temperature of 30°K and covered with N z "ice" with a vapor pressure o f ~ 2 x 10 - S T o r r .
PUMP NET N e GAS FLUX AT SURFACE IS ZERO
_~
Similarly, assume the vessel represented in Fig. 5.1.1 is vacuum baked so as to remove the internal surface gases. After this, assume that it is enclosed in a special refrigerator which is used to vary the temperature of the vessel walls to a uniform
258
CRYOPUMPING
value, and which is set at 30* K. Assume that we valve out the pump and inject sufficient RT N2 gas into the volume to build a thick layer of N 2 ice on the vessel walls. On turning off the gas source, what will the eventual equilibrium N 2 pressure be once it has thermalized to the wall temperature? It is the vapor pressure of N2 at 30*K. Referring to Honig and Hook's data (i.e., Fig. 1.7.3 of Chap. 1),(7 ) we note that the N2 pressure in the volume would be - 2 x 10-s Torr. As long as the temperature of the N 2 ice on the inner surfaces of the volume remains at 30" K, the pressure in the volume will be - 2 x 10 -s Torr. If we now change the temperature of the vessel to 25* K, the N 2 pressure will start to decrease and eventually equilibrate at a value of -10" T Torr, the vapor pressure of N2 at 25* K. Reducing the temperature to 20*K would result in an eventual equilibrium pressure of --10 "1 1 Torr. In this case, to be at equilibrium means that the rate at which gas is departing from and impinging on (and sticking to) the N 2 ice is constant and the same. Therefore, there is a net gas flux of zero at the surfaces. In fact, if we know the pressure, we can calculate this flux using (1.9.6) of Chap. 1. Analytical expressions for the vapor pressure of all gases may be derived from the famous Clausius-Clapeyron equation.
5.1.2 The Clausius-Clapeyron Equation The saturation vapor pressure of a gas may be analytically derived as follows.(S ) The origin of this derivation is the Clausius-Clapeyron equation which describes the slope of the saturation vapor pressure curve as follows: dT
=
dP where
T P Vg vf Sg sf
hfg
and
Vg -
vf
Sg-
sf
= = = = = = =
the the the the the the the
=
T(vg - vf) ,
(5.1.1)
hfg
temperature of the surface, pressure over the liquid, saturated-vapor specific volume, saturated-liquid specific volume, saturated-vapor entropy, saturated-liquid entropy, heat of vaporization.
Note that Sg - sf _~ hfi, / T, and that v~. A ~ T / P . If we assume that v~. >> vfi and ~ as some linear~;t h at hfg may be expressed function of the temperature (i.e., hfg = kl + k2 T, where kl and ks are some constants), then (5.1.1) takes the form: dT
de
~ T/P
(k~ + k2 73
~ T2 / P
(kx + k2 T)
(5.1.2)
On separating the variables and integrating both sides, we derive the simple algebraic equation for the pressure of a gas as a function of temperature: loglo P
= A - BT-1
+ ClogloT
(5.1.3)
CRYOPUMPING where
P T A
= = = = = =
B
C Ca
and
259
the pressure in Torr, the temperature in * K, Ca /2.303, kl /2.303~, ks/2.303~, the combination of the constants of integration.
The values of A, B and C differ for the specific gas and units of measurement. Also, when readin~ the literature be aware that one person'sA may be another person's B, etc.(7,9-1 2") Also, hfg might have initially been represented as an inf'mite series in T, and lead to an expression similar to (5.1.3), but containing many more higher order terms in T. Table 5.1.2 gives values for A, B and C for some of the common gases. Table 5.1.2. Clausius-Clapeyron constants for H2 and Dz from Lee,(12) and others from Honig & Hook, (7) and Redhead, et al. (14) B
C
3.881
42.730
0.50
196
D2 CH4 He0 Ne N2 CO 02
4.234 7.123 10.285 6.889 8.362 8.732 9.082
62.970 482.871 2637.006 109.270 387.428 445.831 480.871
0.50 -
2208.7
289 1955
-
12061.7 499.8 1772.1 2039.2 2199.5
431 1333 1444 1630
Ar C02 Kr Xe
7.703 9.734 7.797 7.775
419.566 1349.809 581.967 801.641
-
1919.1 6174.0 2661.9 3666.7
1558 4041 2158 3021
GAS Hz
A
AH
AHv
The data for H2 and D2 were taken from the work of Lee.(1 2) Data for the remainder of the gases were calculated by "force-fitting" data of Honig and Hook into (5.1.3),(T) while at the same time assuming C-0. The value of AH in this table has different interpretations depending on the temperature at which the data are taken.(1 a ) For example, if the data are taken on the liquid side of the melting point, B would be proportional to AHs, the heat of vaporization, and AH = AH s. If the data were taken on the solid side of the melting point, B would be proportional to AHv, the heat of sublimation, and AH = AH v. In either instance at constant temperature, AH s
=
AH v + AHf,
where
AH v
and
AHf
= the heat of vaporization (i.e., liquid to gas), = the heat of sublimation (i.e., solid to gas), = the heat of fusion or melting (i.e., solid to liquid).
A/4
(5.1.4)
In applications involving cryocondensation pumping, the heat of sublimation is usually applicable. In cryosorption pumping, the heat of vaporization is usually thought to apply. The term heat of sorption is often used synonymously with heat of vaporization. Data for the heat of vaporization of several gases, taken from Redhead, et al., are given to the far right of Table 5.1.2.(1 4 ) There are obvious differences in the values of AH and AH v appearing in this table. These in part stem from
260
CRYOPUMPING
the approximations used in arriving at (5.1.3), as well as the AHf component in (5.1.4).
Condensation at Higher Pressures Practical cryocondensation pumping involves the pumping of gases at pressures greater than their respective equilibrium vapor pressures. In the previous physical example, we might have continuously introduced gas into the cooled volume, but at a rate which did not appreciably alter the temperature of the N2 ice. The introduced gas might have a temperature higher than that of the ice. The continuously introduced gas causes there to be an imbalance in the surface flux. When this flux imbalance exists, there is a net pumping by the N 2 ice. That is, N 2 gas is cryocondensing on the N2 ice. This is illustrated in Fig. 5.1.2. As long as the induced pressure is higher than this vapor pressure, there will be a net flux of gas into the ice. (However, we will shortly see that there is at least one exception to this.) This is simply surface pumping by cryocondensation. Also, as long as the surface temperature of the ice remains constant, the rate at which gas is pumped when cryocondensation exists does not vary with the amount of gas pumped. That is, there is no surface gas population density dependency with cryocondensation pumping, only a temperaturespecies dependency.
.ox F i g u r e 5.1.2. V a c u u m at a temperature of w i t h Na " i c e " w h i c h g a s a t a p r e s s u r e >2
/
vessel maintained 30°K and covered cryoeondenses N~ x 10 - S T o r r .
CRYOCONDENSATION PUMPING ON Nz ICE
r-
GAS N~on)
--------------- . . . . . . . . . . . . . . .
The sticking coefficient, c~, of a gas for cryosorption or cryocondensation is merely the ratio of the average number of molecules which stick, when impinging on a cooled surface, divided by the total number of molecules impinging per unit area. For reasons to be noted, it proves necessary to separate the surface gas fluxes due to the vapor pressure or some other equilibrium pressure, from those fluxes due to induced gases at higher temperatures. Therefore, the sticking coefficient disregards effects of vapor or equilibrium pressure and is expressed by: c~ where and
=
(v i - v v) / v i
ui vv
= the impinging flux rate per cm2, = the departing flux rate per cm2.
(5.1.5)
The speed of a cryocondensation pump is obviously area limited. For example, assume that we continuously introduce gas into the vessel so that the flux impinging on the N2 ice is xl000 greater than the introduced flux of gas departing from the ice. This corresponds to a sticking coefficient for N 2 of c~-0.999; i.e., ,--1.0. That is, on average 99.9% of the introduced gas impinging on the surface sticks. If we were to
CRYOPUMPING
261
reduce the temperature of the N 2 ice so that the impinging flux of gas sticking to the surface was now 99.99%, the new sticking coefficient would be c~-0.9999; i.e., -1.0. For the reduced operating temperature, there is essentially no perceivable change in the speed of the pumping surface (i.e., see (5.1.8)). Therefore, if we wanted to increase the speed of the pump, further reduction in the ice temperature would be of no avail. We could only increase the speed of the pump by increasing the geometric area of the pumping surface. This is called surface conductance limited pumping. If the ice is held at a constant 20* K, the rate at which N2 leaves the ice is equivalent to that of its equilibrium vapor pressure plus a second component of gas flux. This second component is that portion of the continuously introduced gas which when impinging on the surface, rather than sticking, is reflected back from the surface.
Transpiration
5.1.3 T h e r m a l
An interesting phenomenon occurs when two chambers are joined together through a given conductance and the chambers are maintained at different temperatures. The pressure of gas in each chamber will be different even when the net flux of gas passing between the two chambers is zero. This phenomenon is called thermal transpiration and is variously coined in the literature as the thermomolecular effect.(t s-1 r ) As you will see, this has applications in the design and evaluation of cryopumps. We must often indirectly deduce the behavior of gases being pumped on cold surfaces in one chamber, through the direct measurement of throughput, Q, and pressure, P, observed in a second chamber maintained at a different temperature. For example, assume that the two vacuum chambers shown in Fig. 5.1.3 are maintained at different temperatures, and that they are separated by an aperture of area A a. Assume that one chamber is extremely cold and maintained at temperature T2, and the second is at T1 = 293 ° K (i.e., RT). Assume that the gas in the RT chamber comprises air at a pressure of 10 .6 Torr, and that there are no sources or sinks (i.e., pumps) of gas in either chamber. If the pressure in both chambers is constant, what is the pressure in the cold chamber? If the pressure is neither rising nor falling in either chamber and there are no sources or sinks of gas, the net flux of gas passing through the aperture of area A a must be zero. Using (1.9.6) of Chap. 1, we can express the flux of gas leaving the warm chamber as: v 1 where
= Aa P1 M Tx
A a 3.51 x 10 2 2 P1 / ( M T I ) ~ , = = = =
the the the the
(5.1.6)
area of the aperture in cm2, pressure in the warm chamber in Torr, molecular weight of air (i.e., -28.7),(2 2 ) temperature of the warm chamber in ° K.
An identical equation may be constructed for v 2, the flux leaving the cold chamber and entering the warm chamber. In that v t = v 2, we arrive at the relationship:
or
P1 (T1)-~
=
P2 ( T 2 ) ' ~ ,
P2
=
P1 (T2 / T 1 ) ~ .
(5.1.7)
262
CRYOPUMPING
We could have just as easily invoked (1.12.3) and (1.12.7) of Chap. 1, and arrived at the same result. In this case though, we must also take care to note that the conductance separating the two chambers differs depending on the origin of the gas (e.g., see problem 2). BAG
S GAS i
~I~AT
REFRIGERATOR BOX TEMPERATURE
i
~Pa V
( T2,~1/2 = PI~,TI j
VESSEL
PREooATuRE
~
PRESSURE
PUMP (off)
T2
TEMPERATURE
T2
~ - OR "SINKS" IN V A C U U M VESSEL AT T1 = 2930K Figure
~_
EITHER VESSEL
APERTURE OF AREA A a
5.1.3. T h e r m a l
transpiration
effect.
Now assume that there is some sort of mysterious pumping process going on in the cold chamber. We need not understand the process for now. We will only assume that the surfaces in the cold chamber are sticky to the gas being introduced in the warm chamber and passing through the aperture. Using an equation identical to (3.2.1), we can express the pumping speed of the surface in the cold chamber as: Ss where and
= c~ Cs, a Cs
(5.1.8)
= the sticking coefficient of the cold surface, = the conductance at the surface (i.e., dependency on molecular weight and temperature of the gas).
Assume that gas is introduced through a variable leak into the warm chamber at a rate Q. Clearly, Q = S s p2 = a C s P2. From this result, and by invoking conservation of mass, the following is found:
a where
and
C1 P1 - C 2 P 2 C1 C2 Cs k
= czC s P 2 ,
(5.1.9)
= k A a (T1 / M)~, = k A a (T2 / M) y' ,
= k A s(T2 /M) ~, = a constant of proportionality (see section 1.12.1).
From these simple equations we may derive Serf, the effective speed produced in the warm chamber due to the pumping of the cold chamber, assuming we know the value of T1, c~, A s and A a (see problem 3); a , assuming we know Q and P1 (see problem 4); or a , assuming we know P1 and P2 (see problem 5). As a word of caution, one should be careful in indiscriminately invoking (5.1.7) as an identity. It is only an identity when the net flux of gas through the separating conductance is zero. That is, (5.1.7) is true only when there is no net flow of gas. One frequently encounters this error in the literature. From this simple model, one finds the speed delivered
CRYOPUMPING
263
to the warm chamber due to pumping by the cold chamber as:
otCx A
S efr
=
s (A a + O~As)
(5.1.10)
Note that the temperature of the cold chamber does not appear in (5.1.10), though it is implicit in the value of the sticking coefficient, ct, of the gas on the cold surface. Therefore, we see that ct is all important in predicting the behavior of cryopumps. There are numerous excellent publications on the subject of the sticking coefficients of gases on cold surfaces, and some I feel are classics. One of these includes the work of Dawson and Haygood.(1 a ) This work contains a comprehensive listing of sticking coefficients of the gases N 2, CO, O 2, Ar, CO 2 and N 2 O, at temperatures ranging from 77°K to 400°K and on cryofrosts of the same gas at temperatures ranging from 10*K to 77°K, and similar data for 13 other RT gases on 77*K surfaces. It also contains a most lucid treatment of elementary kinetic theory in the definition and interpretation of the sticking coefficient of a cryosurface in one chamber adjoined by a second at RT. In another early classic, Levenson used a quartz crystal microbalance to measure the sticking coefficients of molecular beams of Ar, CO 2, Kr and Xe as a function of gas and substrate temperature.(a ) The work of Brown, Trayer and Busby is also frequently cited in publications dealing with the sticking coefficients of gases on ices of the same species (i.e., cryocondensation pumping).(4 ) Through use of collimated molecular beams, they measured the sticking coefficients of the gases N 2, Ar and COs as a function of gas temperature, cryofrost temperature and angle of incidence of the gas beam with the surface. The results indicated only slight variations in o: for the three gases over gas temperatures ranging from 300 ° K to 1400 ° K. The implication of this finding is that cryopumps may be used to cryopump very hot gases. This is not a surprising result (e.g., see problem 6). The marked dependency of (x on v i, which they noted for 300 ° K CO 2 gas on CO ~ ice of varying temperatures, is shown in Fig. 5.1.4. 1
I
1
~3.2 x 1017 MOLECULES_
1.0 I
~. 0.8 Z L9
0.6
<~3.2 x 1016 MOLECULES
O
z
(~
t
0.4
(..)
per cm a-see
-
0.2
72
74
76
78
80
82
84
86
88
90
92
SUBSTRATE SURFACE TEMPERATURE - °K Figure 5.1.4. The effective capture coefficients of 300°K C02 beams of varying intensities and impinging on C02 "cryofrosts" of varying temperatures, as reported by Brown, Trayer and Busby.(4)
2~
CRYOPUMPING
Results of this work (i.e., Fig. 5.1.4) proved troubling to many readers, as there is a strong temptation to lend significance to the zero intercept of the various flux intensities. That is, we would intuitively assume that the sticking coefficient would vanish at that flux intensity equivalent to the vapor pressure of the cryofrost. Brown, et al., reported that for CO 2, Ar and N2 this was not the case. That is, the sticking coefficient of these gases vanished at surface flux intensities far lower than that equivalent to the vapor pressure. I have modified their data, in Fig. 5.1.4, to reflect beam intensities, as they reported on total flux. The inequality signs for the flux values result from my assumption of no beam divergence at the cryosurface. We can qualitatively state that a cryopumping surface with a sticking coefficient of ~ 0.5, whatever the conditions, would pump a significant quantity of gas. With this criterion, if we assign a pressure equivalency to the beam flux of ~ 1.6 x 101 s CO2 molecules/sec-cm2, and vapor pressure equivalency of the ice at the point of c~ -0.5 shown in Fig. 5.1.4, we determine that, at least for CO2, effective cryocondensation pumping occurs when the p u m p operating pressure is ---x 100 that of the vapor pressure of the ice. Note that as the CO 2 beam intensity was increased to 3.2 x 10 ~ 7 molecules/sec-cm 2, this would require a p u m p operating pressure margin of --.x 2000 that of the vapor pressure. In molecular beam experiments similar to those of Brown, et al., Bentley and Hands reported similar f'mdings for the sticking coefficient of CO2 on cryofrosts of same, but disagreement in results for the gases N2 and Ar.(S, 1 9 ) For the latter gases, they concluded that ct vanishes at or very near that equivalent to the vapor pressure of the cryofrost. They proposed a simple model for the anomalous cryocondensation behavior of CO2 based on the tendency of CO~ to nucleate and to stick to other surface CO 2 molecules more readily than to a clean cryopanel. Bentley and Hands noted that the effect became much less pronounced once several monolayers of gas had accumulated on a cryopanel.(1 9 ) This phenomenon would result in an hysteresis effect in sticking coefficient data of the nature shown in Fig. 5.1.4, as c~ would depend on the thickness of the cryofrost. This strange phenomenon seems unique to CO~, though it may have been observed by Hobson with the gas Xe,( 2 o) and Chubb, et al., with H 2 condensation on 3.7* K surfaces.(S ~ ) It is referred to as a super saturation effect. Bfittner measured the cryocondensation speeds of H 2, D 2 and a 1:5 mixture of H2 :D2 on a surface cooled to 4.2"K (i.e., cryofrosts of the gases). He determined that the speed for H 2 decreased to zero very near the equivalent of the vapor pressure of the H 2 . (2 1 ) Chubb and Pollard reported near-unity capture coefficients for RT H 2 on ice of H 2 at a temperature of 3.6 - 3.9* K, and for sorbed H 2 fluxes of the order of x 15 to x 160 that of the surface efflux due to the vapor pressure of H2.(2 ) They reported that "... there was no dependence of (the) sticking coefficient upon gas flow rate ..." for the indicated fluxes.
5.1.4. Adsorption Isotherms The second form of cryopumping involves cryosorption which in turn relates to an adsorption isotherm. You frequently encounter this term and the terms isostere and isobar in the literature. Mathematical definitions of these terms are as follows: an isotherm is defined as o = f (P, Tc) at constant T, an isostere is defined as P = f (T,o c) at constant o, an isobar is defined as o = f (T, Pc) at constant P,
CRYOPUMPING where
O
and
P T
265
= the density of molecules of gas on a surface per cm 2, = the equilibrium pressure in the system, = the system temperature.
Note that we could have just as well used the volume of gas, V.+.., at standard temat~ perature and pressure, in place of o. This appears in some places in the literature. Regarding adsorption isotherms, assume that we have a vacuum vessel identical to that shown in Fig. 5.1.1, which we bake under vacuum to remove all of the surface gases. Assume we fix a gauge to this system to enable measurement of the pressure within the vessel, while knowing both the vessel and gauge temperatures and thereby compensating for their differences using (5.1.7). The vessel surface area is A v and the volume Vv. In this instance, after cooling the system, we will introduce a very limited, known amount of gas. Some of the gas will land on the surface, stick there for a while, and then come off again as free gas. If all of the gas is continually in a gaseous state, the pressure in the vessel, Pv, will behave according to the ideal gas law (i.e., Pv = n~ T / V v ) . However, if for example, at any instant in time half of the gas is resident on the walls, then the pressure in the vessel will be half of that predicted by the ideal gas law, etc. Therefore, there is a net pumping on the walls. If we know the amount of gas introduced, and the subsequent equilibrium pressure in the vessel, we can calculate the average population density of gas on the surface of the vessel at any time. A measure of the surface population density is usually given the symbol o in the literature, where o represents the extent of surface coverage in molecular or atomic layers. For example, o = 0.1 indicates a 10% coverage; o m is usually reserved to note a surface coverage of one molecular layer or a monolayer, o = 2 indicates coverage of two monolayers, etc. Sometimes the symbol 0 is also used in the literature to represent surface coverages relative to the total number of possible adsorption sites, and sometimes as the ratio of o / o m" As a rule of thumb, a molecular layer comprises --101 s molecules (atoms) per cm 2 . See Redhead, et al., for more precise values of o m" When plotting experimental data, it is usually the convention to display the independent variable (i.e., the cause) along the abscissa or x-axis and the dependent variable (i.e., the effect) along the ordinate or y-axis. There are two ways one might perceive an adsorption isotherm: 1) one might assume that the surface population density causes the equilibrium pressure; or, 2) one might assume that sustaining the equilibrium pressure over the surface causes the surface population density. Each are reasonable scenarios, and I will show data both ways. Assume that we successively introduce additional known quantities of type A gas into the system, and after each introduction we wait for the pressure to equilibrate and then make a pressure measurement. From this sequence of measurements we are able to plot a curve of the equilibrium pressure Pe vs. the average surface coverage, o. Data of Pe vs. o at constant temperature, T1, are called an adsorption isotherm. A n isotherm merely means that pressure vs. coverage data are taken at the same temperature. Such data might take the form represented in Fig. 5.1.5a, where the equilibrium pressure is given as a function of surface coverage at a given temperature. If we were to repeat the measurement all over again, but this time at a different temperature, T2, where T2 > T1, we would observe a distinctly different pressure for a given coverage. We note that as the temperature is increased, the equilibrium pressure is higher for the same gas coverage.
2~
CRYOPUMPING i0
-2
,,I n , ,,,/I,
o [-4
ISOTHERM OF GAS "A" AT
I r~
i0
T2
~ 10-4
(y
--
H.E.RM OF
-
--5
> T:
_ -
i0
Figure 5.1.5a. Variations in equilibrium pressures, for the same coverage, but at different temperatures.
-3
~l 1012
SURFACE TEMP. T I -
IN
t 10 la
SURFACE COVERAGE
I Jl]
I
10 TM
-2
_
AT
/ /
,.
I nl
I0
--
~~'o 1015
- molecules/era 2
1
I [1
1
I 11
/Tz
1
1 11 T1/ ? --
GAS "C"
t ~
-3 10
:/3 ~t3
Figure 5.1.5b. There are unique adsorption isotherms for each gas and for the same surface temperature.
N m -~
lO - 4
:
~r~
10
1 1012
I II
1013
SURFACE COVERAGE -
1 10 TM
lit 10
molecules/cm 2
Were we to repeat this series of measurements for three distinct gases, identified as types A, B and C in Fig. 5.1.5b, each gas would have its own distinct T-dependent adsorption isotherm. That is, at the same surface temperature, each type of gas would have a different equilibrium pressure for an identical surface coverage, as illustrated in this figure. For all gases, the measured pressures of the adsorption isotherms are in each case less than the vapor pressures of the given gas at the particular temperature. However, as surface coverage increases to the point of several atomic layers, the equilibrium pressure of an adsorption isotherm converges to the much higher value of the vapor pressure of the species at that temperature. In the case of adsorption isotherms of gas A, the cold surface has a finite speed for gas A as long as the pressure of the adsorption isotherm is less than the operating pressure of gas A. This effect is graphically illustrated in Fig. 5.1.6. Assume an adsorption isotherm for gas A as shown in Fig. 5.1.6a. With greater and greater surface coverage, the equilibrium pressure over the surface continues to increase. Clearly, after pumping o i amount of gas, the speed of the pump is zero at P i ; i.e., the speed of the pump at P1 has decreased to zero. However, at higher pressures, and for the same o x coverage, the speed increases as we depart from the base pressure of the pump, P z . Eventually, the pump speed asymptotes at some Sma x value, which is dependent on the area of the cooled surface and the sticking coefficients of the gas being pumped. This amounts to surface c o n d u c t a n c e limitation for cryosorption pumping. This effect is shown in Fig. 5.1.6b. After pumping enough gas to increase the surface coverage to o 2, the speed of the pump becomes zero at P2. With increasing surface coverage, the zero speed point of the pump progressively increases. This results in a family of speed curves such as shown in Fig. 5.1.6b. Each speed curve is dependent on the respective surface coverage. We conclude that when discussing the speed of a cryopump used for
CRYOPUMPING
267
cryosorption pumping, we must have know-ledge of the quantity of gas pumped up to that point in time and the pressure. This is similar to requirements in the use of NEG pumps, noted in Chap. 4. Also, when making claims about the capacity of a pump to cryosorb gases (i.e., the amount of gas which it can pump), these claims only have meaning if the pressure at which the speed measurements are taken is also given. For example, if the speed measurement for a gas A was taken at pressures somewhere on the knee or flat portion of the speed curve belonging to the o 2 curve, the speed for that gas would be zero or negative at pressures
I
I II
1
o
//
i
P3 M -4 - - ISOTHERM OF ~ lO GAS "A" AT r~ TEMP. T 1
~
N -5 EIO ~
jf
P2 /
I
I II
Ill
/
Figure 5.1.6a. Variations in equilibrium pressure for surface c o v e r a g e s O 1,O 2 a n d o3 and temperature TI.
"*~/
/
Pl
o"
-6 10
I 1012
O1 1013
l
rl II 02
1014
03
1015
SURFACE COVERAGE - m o l e c u l e s / c m 2
1
I tl ~j.~..~ ~
l s max.
1 11 ~.~
I
I I'
~.
100 Figure 5.1.6b. Variations in pump speed due to changes in base pressure because of gas surface coverage, where ot
~ < :~ ~ o z
75
L)
0
o~--.~ /
5o
%
25
I 10-6
I/ P1 10- 5 P2
-4 P3 10
PUMP OPERATING PRESSURE -
1
-3 10
Tort
The rate at which the zero speed of the pump increases depends on the rate at which gas is pumped and is the true measure of a cryopump's capacity. In practical applications, when the system pressure increases to some predetermined limit as a consequence of the zero speed point of the pump increasing in pressure, we conclude that pump capacity has been exceeded and the cryopump needs a regeneration. I will later return to the subject of cryopump regeneration.
Classifications of Adsorption Isotherms In Fig. 5.1.5b I showed the adsorption isotherms of fictitious gases as being smooth functions and devoid of inflection points. An inflection point is a point along some curve (e.g., the P vs. o curve) where the curve seems to depart from a trend and
2~
CRYOPUMPING
change directions (i.e., d 2 0 / dP 2 ~ 0). All adsorption isotherms have a minimum of one inflection point and some as many as three or four. To further complicate their theoretical modeling,, all evidence hysteresis effects which may or may not be time dependent functions.(2 4 -~ 6 ) In addition, Stern and DiPaolo report on a conditioning effect occurring with some sieve materials.(2 9 ) That is, an isotherm of a gas on a given material may differ for the same temperature after, in an interim period, having been returned to RT. An in-depth treatment of the physical process resulting in these complex P vs. 0 functions is beyond my complete understanding. However, ... Adsorption isotherms often have a unique and repeatable form, which is attributable to the gas, temperature and sorbing surface. Roberts provides an excellent qualitative review, with a comprehensive biblio~aphy, on the classification of about 13 different types of adsorption isotherms.(2 7 )" Dushman discusses several types of adsorption isotherms reported in the early literature. I liberally use this reference
below.(2 8 )
No unified theoretical model presently exists which fits the peculiarities of all of the variously shaped isotherms, though some theoretical models are very accurate for specific gases on specific types of materials. For example, one model might be found to agree with experimental results in a pressure range varying --13 orders in magnitude for certain gases on a specific material,( 2 o ) where the experimental results with a different gas on another material were found to diverge from the model.(2 9 ) I will briefly note some of the historical developments in this area to provide you with a familiarity with some of the terms and their origins. To avoid repetition, I will use the terms o, P and T as defined in the previous section. There are several popular semi-empirical adsorption isotherm models which we vacuum technicians find frequently cited in the literature. One such model is called Henry's Law. It is sometimes applicable only at very low pressures,(2 g, 3 o ) while at other times it is applicable at high pressures.( 3 1") It is a variant of Henry's Law dealing with the solubility of gases in liquids. This law states that the amount of gas which will go into solution with a liquid is directly proportional to the pressure of the gas over the liquid. The vacuum variant of Henry's Law predicts that 0 will vary linearly with pressure. Or, o
= kl P,
(5.1.11)
where k i is some constant of proportionality. In this case, if Henry's Law is applicable, when the P-o data are plotted on log-log paper, the slope of the data will fall on a 45* line. The parabolic or Freundlich isotherm is also frequently cited.(3 2 ) In this case, o
= kl p m ,
(5.1.12)
where kl is a constant and m < 1. Freundlich's model includes Henry's Law. Langrnuir was the first to attempt to model adsorption processes in terms of gas-surface physics relating to potential occupancy sites, variations in the number of sites possibly occupied by different gas species, and how all of this affected the amount of surface gas which could be accommodated. Because of Langmuir's work, his insightful approach started vacuum technicians thinking in these sort of terms. In honor of him the term Langmuir is often used in publications to denote a gas coverage of one monolayer. His model dealt with surface coverages where o < one Lang-
CRYOPUMPING
269
muir. The equation which he developed for an isotherm is sometimes called the hyperbolic isotherm, and takes the form: o
=
o s (kl p)l/m
(5.1.13)
1 + (kl p)l/m
where k l is a constant, o s denotes the total number of available sites per cm 2, and m represents the number of surface sites which is occupied by a particular gas species. Almost a quarter of a century later another important adsorption model was derived by Brunauer, Emmett and Teller.( a s ) They were troubled by published data indicating that cryosorption could include the build-up of surface gases to an extent exceeding a monolayer of gas - perhaps to the extent of several monolayers as in the case of cryocondensation. Their interest was in the use of molecular sieve materials (see Sieve Materials, below) as catalysts. It was noted that the shape of the P vs. o isotherms of gases on these sieve materials was concave downward at low pressures and at higher pressures became convex. They coined this effect an s-curve isotherm. They expanded on Langmuir's theory to include its applicability to gas coverages exceeding a monolayer, with the added assumption that the same forces which came into play in cryocondensation were applicable to cryosorption at higher temperatures. They derived the following equation: o
=
omkl P
, [Pv - P][I + (kx - 1)(P/Pv) ]
(5.1.14)
where k i is some constant much greater than unity, o m is a Langmuir and Pv is the vapor pressure of the test gas. Equation (5.1.14) may be rearranged into the form: P
O(Pv-P)
=
1 Omkl
+
(klomkl
1)
P//Pv"
(5.1.15)
Therefore, assuming the theory was correct, a plot of P / o (Pv - P) as a function of P / P v should produce a straight line, the ordinate intercept being 1 / o m k x and the slope having the value (kx - 1) / o m k x. This enables one to determine the values of both k 1 and o m" They expanded on (5.1.14) to include coverages exceeding one monolayer. However, (5.1.15) proved so successful in characterizing the s-curve adsorption isotherms of materials that it developed into an industrial standard for specifying surface areas of sieve materials used for catalysis and in vacuum applications. In recognition of Brunauer, Emmett and Teller's work, it became known as the B E T method for determining the area of sieve materials.(a 4,a 5 ) Nitrogen is usually used as the standard gas in this application and the sorbing material is cooled to LN2 temperatures. A decade later, Dubinin and Radushkevich published what is now abbreviated in the literature as the DR adsorption isotherm.( a 6,9 8 ) This equation takes the form:(2 9 ) log o
= log o m - k x (log Pv / P) 2 ,
(5.1.16)
270
CRYOPUMPING
where k 1 is some constant and Pv is the vapor pressure of a liquid of the gas. Stern and DiPaolo have found fairly good agreement in this equation when pumping N2 on sieve materials (Linde 5A) cooled to LN2 temperatures.(2 g ) However, the DR theory and their results diverged at pressures ~ 5 × 10 -4 Torr. Troy and Wightman could f'md no general experimental agreement with either the B E T or DR theory. Their experiment involved the pumping of CH 4, N 2, Ar and Kr on stn. stl. surfaces ranging in temperature from 77 to 90* K. Halama observed similar departures in the DR theory with measured results for He on 5A sieve, but at pressures z 10" 1 o Torr.(2 s~ Hobson found close agreement with the DR theory and measured isotherms of Ar, Kr and Xe on porous silver, and over a very broad pressure range. This latter experiment must have been a test of both scientific technique as well as endurance. The results are shown in Fig. 5.1.7. lO
22
f.n O [--, 1 0 < 10
10 O
~
10
~
lo
~ o
10
Z
I0
I
31
XENON ISOTHER~~~
20
~
[] E] E] [l L] [] []
I
19 / 18
~
^,A~ ~
2
17
J
ARGON ISOTHERM ~ _ KRYPTON ISOTHERM l
15
f lo-11 lo-lO lo-9
1o-O lo-7
to- 6 lo -5
1 lo-4
! i
10- 3 lo -2
TOTAL PRESSURE ABOVE ADSORBED LAYER -
1o-~
lo 0
101
10 2
10 3
Torr
F i g u r e 5.1.7. A d s o r p t i o n i s o t h e r m s of Xe, Kr a n d Ar on a p o r o u s silver a d s o r b e n t at a t e m p e r a t u r e of 77.4 °K, as r e p o r t e d by Hobson.(eo)
We see from this figure that in order to conduct this experiment, Hobson had to measure equilibrium pressures ranging 13 orders in magnitude. Detailed attention to U H V technology was a prerequisite to the experiment. Also, achieving equilibrium pressures at times requires elapsed times of an hour or more for each datum. With all this in mind, this has to be one of the classic experimental papers on the measurement of adsorption isotherms. Hobson predicted in an earlier theoretical paper that the isotherms of gases would depart from the DR theory at very low pressures, and begin to converge with Henry's Law.(3 o ) The theoretical model which he presented in ref. (2 0 ) accommodated predictions of the onset of Henry's Law at low pressures. However, he did not observe this effect. Also, it was not observed in similar experiments which he conducted with He on porous silver at LHe temperatures.( 3 7) In this case, the lower pressure limit of the experiment was --5 × 10-1 1 Torr He. Hseuh, et al., noted that the Henry's Law limit was in no manner applicable for He on activated coconut charcoal at temperatures ranging from 4.2 to 18.2" K.( 3 8 )
CRYOPUMPING
271
Defining a universal model to predict an adsorption isotherm becomes very complex, as they are critically dependent on the type of gas (e.g., polar vs. nonpolar) and the sorbing material (e.g., the nonpolar carbon materials vs. the highly polar zeolites).(1 o o ) Existing models are still somewhat flawed. For example, using graphical (or analytical) methods similar to those described relating to the B E T method, Hobson found a disparity between that of the DR method and B E T methods in defining o m (i.e., equivalent to the effective area of the porous silver) for the gases Ar, Kr and Xe on porous silver at LN 2 temperatures. There was reasonable agreement for Xe (i.e., within 30%), but a divergence in data for lighter molecules to the point of a disparity of x 4.3 for the B E T vs. DR methods. The effective surface areas of his porous bed, as predicted by the B E T and DR methods, were 1.1 x 10 5 cm2 and 2.4 x 10 4 cm2, respectively. Secondly, there was disagreement in implied surface areas for the different gases when comparing results of one method for the three gases. These were minor in the case of the B E T method (i.e., 25%), but significant in the case of the D R method (i.e., -.x 2.5).(2 o ) In a paper published three months later, Hobson reported measurement of He adsorption isotherms, presumably on the same porous silver bed, but this time at LHe temperatures.( s 7") He made a direct comparison of these results with that of Danilova and Shal'nikov, who measured the isotherms of He on a smooth Cu plate also maintained at LHe temperature.(S 9 ) From this, Hobson was able to conclude that the effective area of his porous silver bed was - x 1000 that of a fiat Cu surface, or -3.5 x l0 s cm2 for He. Therefore, for the same bed, and different gases, there is a variation of over an order in magnitude in the predicted effective surface area using these three different techniques. Recognizing Hobson's exacting technical skills, one therefore must conclude that much about this theory is still found wanting. Hobson suggests that we classify cryosorption pumping as a process where o 5 o,,~ and cryocondensation pumping when surface coverages are such that o 5o m"(4 o ) This is a reasonable scenario if we envision that van der Waals' forces " extend into free space something greater than the equivalent of three molecular layers (e.g., see Fig. 2.4.7). However, o s 5 is a statement about the probable upper limit of the adsorption process as, for example, He on LHe cooled surfaces has a coverage o - 3 x 10-2 o m at 10-6 Torr.(3 9 ~The effectiveness with which gases are cryosorbed relates to equivalent energy sites created by surface-gas and surface gas-gas forces. The forces between gas molecules and the parent surface tend to be greater than the forces between gas-gas molecules on the surface. Therefore, the heat of sorption will initially be greater than the heat of vaporization of a gas on a liquid of the same species. However, as the surface coverage increases, the heat of sorption decreases and finally converges to the heat of vaporization (or sublimation).(4 t )
Helium
a n d H y d r o g e n I s o t h e r m s on 4.2* K S u r f a c e s
By far the most comprehensive current work on He and H 2 isotherms was done at CERN by Wallen.(X 6 o, 1 6 t ) A discussion of the various forms of adsoption isotherms and their application to his results in these two references. He explored isotherms of these gases on 304LN st. stl., Cu plated stn. stl. and Pyrex over the temperature range of -1.9 - 4.2* K. His work was motivated by work on the Large Hardron Collider presently under construction at CERN. Most of this superconducting collider, measuring 27 Km in circumference, will operate at --1.9 ° K.
272
CRYOPUMPING
Helium, stemming from possible leaks into the beam tube, and hydrogen from beam tube end-effects, were of particular interest. A summary of his adsorption isotherm findings for H2 in the pressure interval of 10 "1 1 to 10 -6 Torr. with the three mentioned surfaces maintained at 4.2* K, is given in Fig. 5.1.8.( 1 6 o 10 - 6 _
[]
/', 0
10 -7
[]
~_~ 0 o°
I 10
Key:
-1
/-,.
10 -9
~ S t n . Stl. S u r f a c e [] Cu P l a t e d S t n . S t l . © Pyrex Surface
3
~9 0
10 - 1 0
10-11
-C~ ~ o
l x l O 15
l J
2 x l O t5 Hydrogen
izJllllll
IJlllillz
I11111111
3 x l O 15 Molecules
Sorbed
4 x l O 15 per
5xlO 15
cm 2
Figure 5.1.8. Adsorption isotherms of hydrogen at 4.2 K on three surfaces (reproduced with permission o f E r i k W a l l e n o f C E R N ) . (160)
Actually, Wallen's hydrogen work was far more comprehensive than just the measurement of isotherms of this gas. He measured isotherms of H 2 on cryofrosts of CO, CH4, and CO 2 at 4.2* K on Cu plated stn. stl., and the concurrent pumping (i.e., cryotrapping) of H2 with the three gases, in different proportions and at 4.2* K. The nature of these experiments is given at the bottom of Table 5.1.3. Wallen's He adsorption isotherm data, at 4.2" K, and on the same three surfaces, are shown in Fig. 5.1.9.(1 6 1 ) Considering the error bars in Hobson's similar data on Pyrex, reported almost four decades earlier,(1 6 2 ) I find remarkable agreement in these two experiments. The results of Danilova and Shal'nikova, on a polished cooper surface, seem to depart significantly from that of either Wallen or Hobson.(a 9 ) The current interest in the isotherms of He on 4.2 ° K surfaces has very practical significance as there are numerous superconducting accelerators, with 4.2"K vacuum envelops planned or in operation throughout the world. Being in charge of the RHIC Vacuum Group, I was concerned about the He partial pressures which might exist in the RHIC beam pipes in the event of He leaks into the U H V cold beam tubes, and the end-effects of H 2 streaming into the 4.3 ° K cold-bores. In 1991 I collaborated with Hobson to model what might occur in the event of leaks of various sizes, and with and without the use of sorption pumps. One asked: What would the effect be on beam lifetimes depending on the size of a He leak and with and without the use of sorption pumps? By that time, because of measurements reported in Chapter 2, I had dismissed the use of sputter-ion pumps as a viable means of pumping He cold-bore leaks.
CRYOPUMPING
273
Measurements were made by Edwards and Limon where they noted that it took almost three minutes for a helium signal from a contrived leak to be detected in a 20 m long, -4.2* K cold-bore tube.(X 6 3 ) We set out to construct a theoretical model of the phenomenon using a variation of the DR adsorption isotherm, and using DR constants derived from previous work of Hobson. Using Hobson's data and making six basic assumptions in the model, we calculated: i) the velocity of either a He or H 2 wave front in the cold-bore tube as a function of the size of a leak; and, ii) the velocity of the wave fronts with the use of sorption pumps scattered along the beam pipe. We published these theoretical results in 1993.(1 6 4 ) 10
-5
,3
o b-, I
10
-6
10
-7
Key: 0 P y r e x
1"71 Cu P l a t e d /~ S t n .
10
.~ 10
Surface t Stl.
4
Surface
~
)
Surface
./
-8
•
-9
/
,-...4
10
-10
10
-11
I ~
1013
l
1
I
Helium
Atoms
Sorbed
I
I i i 1
10 15
1014 per
cm 2
F i g u r e 5.1.9. A d s o r p t i o n i s o t h e r m s of h e l i u m o n three surface materials a t 4.2 K ( r e p r o d u c e d with permision of E r i k W a l l e n of CERN).(161)
Six years later Hseuh and Wallen put our theoretical model to the test.(1 6 s ) When introducing a He leak of known magnitude into a ¢6.9 cm, 400 m, 4.2"K RHIC beam tube, they measured a He wave front velocity in very close agreement with our theoretical model. Wallen reported on tests made a year earlier at CERN, using a ¢4.3 cm, 75 m long beam tube at temperatures of 1.9" K and 4.25* K.(1 6 6 ) The He wave velocity findings at 4.25" K were within 4% off that predicted by our theory, and the He wave front velocity at 1.9*K was within 17% of our theory. To my knowledge, no one has made measurements of the wave front velocities of hydrogen at these temperatures. This in part may be due to the fact that the wave front velocities are predicted to be -10-50 times slower for H 2 than for He. Table 5.1.3 lists representative examples of both cryocondensation and cryosorption pumping experiments reported in the literature. It covers a wide variety of gases and surfaces, including cryofrosts of similar and different gases. It is somewhat chronological.
274
CRYOPUMPING Table 5.1.3. Examples of some cryocondensation, cryosorption
SURFACE TEMP. o K 4.2 4.2 10-77 77 3.5-99 77 & 87 3.8-77 4.2 2.18-3.68
APPROX. PRESS. Torr 10- t 2 _ 10-11_
10-4 10-fi
of
the and
referenced cryotrapping
GASES
REF.
literature pumping. SURFACE
on
1
varied varied not indicated
He Hz, He N2,C0,02,Ar, C02,N20 13(e.g., HeO-CHaCOCH3) He, Ar
162 45 18 18 2
Pyrex CT by CF of CF of CF of
1 0 - 9 _ 10-4 molecular beam 2 x 1 0 -11_ 10-5 7x10 -1° 7 x 1 0 -8
Kr, Xe Ar, C02, Kr, Xe He H2, D 2
107 3 39 87
Mo CF Cu CF
20-92 77-90 2.5-4.2 2.5-4.2
molecular
10-4 10- 1 2 - 7 x 1 0 -7 10 -t2--- 4x 10-7
Ne, Ar, C02 CH4, N2, Ar, Kr H2 H2
4 108 12 12
CF of S. 304 s t n . s t l . Cu a n d CF of S. CF of S & CF of At.
3.7-5.2 2.3 4.2 2.5-4.3
10 - l z - 5 x 1 0 - a 8 x 1 0 -12_ 4 x 1 0 - 1 1 10 -6 _ 10-4 2x 10-14__ 10-6
D2 H2 H2, D2 Hz, HD, Da
12 91 96 92
Cu Ag CF CF
& CF of S. p l a t e d stn. stl. of S & s t n . s t l . of S & N e , A r , N z , K r .
4.2 4.2 3-300 2.23-4.2
g 1 . 3 x 1 0 -4 ~ 1 . 3 x 1 0 -4 molecular beam not indicated
He N2, Ar, C0z He, He, N2
103 103 19 42
CF CF CF CF
of of of of
S & At. D2. S & Cu. S & Nz SCT.
3-300 7.3-15.3 4.1-4.9 4.2
molecular beam 8 x 1 0 -12--- 8 x 1 0 -6 3 x 1 0 -5 _ 10 -4 10 -7 6 x 1 0 -5
N2, Ar, C0z Ha Hz, D2, N2 He, Dz, H z + D e
5 48 94 21
CF CF CF CF
of of of of
S & Cu. C02. S a n d s t n . stl. S & CCT of H2.
4.2 4.2 3.7-4.3 6.0-11
2x 10 -6 - 2 x 1 0 -4 not indicated 4 x 1 0 -6 - 4 x 1 0 -5 5x10 - 8 _ 7x10 -4
Dz, D z + H e H2 He H2
46 61 97 50
CF CF CF CF
of of of of
S & CCT of He. S a n d s t n . stl. S a n d Cu. CO , Kr & Xe.
160 160 160 160
Cu p l a t e d s t n . s t l . Cu p l a t e d s t n . s t l . CF of C02, CO & CH4. CCT, Cu p l a t e d s t n . s t l .
161 161
Cu p l a t e d s t n . s t l . Pyrex & 304 L stn.
beam
10 - 9 _
D 2
..,
Ar. S. S. S.
films & glass. of S; q u a r t z c r y s t a l . plate. of S.
.
77 4.2 4.2 4.2 1.9-4.2
4.2 1 Key:
8x10 -7-
2 x 1 0 -5 8 x 1 0 -7 ~ 4 x 1 0 -12_ 8 x 1 0 - 7 ~4x10 - 8 x 1 0 -7 < 1 0 -11 _
<113_11 -
,., 1 0 - 5
< 1 0 -11_
,...10-5
He w /
Xe H2 H2 COz, CO o r CH4. He He
stl.
CT • C r y o t r a p p i n g . CCT: C o n c u r r e n t Cryotrapping with more readily pumped gas. CF of S: P u m p e d o n c r y o f r o s t s of s a m e g a s o r s u b s t r a t e .
5.1.5 S p e e d a n d C a p a c i t y o f C r y o p u m p s In a previous section, we noted that when the pressure of N2 in the vessel was greater than the equilibrium vapor pressure of the N 2 ice, the net flux of gas was into the N2 ice, and the surface of the ice was a cryocondensation pump. Also, similar mechanisms applied to cryosorption pumping. That is, if the equilibrium pressure related to a given adsorption isotherm was less than the pressure due to introduced gas, the surface would pump gas as the system sought a new equilibrium. In both cases, the system attempts to equilibrate at a state where the net flux of gas impinging on and leaving the surface is zero. In most cases of cryocondensation pumping, the surface ice has zero speed for a gas A at a pressure equivalent to the vapor pressure of the ice A, and it has finite speeds for gas A at pressures greater than the vapor pressure of ice A. An exception to this was noted in the case of CO 2. There-
CRYOPUMPING
275
fore, we must recognize and compensate for the fact that the values c~ and (1 - Pe / Po) may stem from two different effects. The equation for the pumping speed of a cryocondensing or cryosorbing surface is as follows: S
= a C (1 - Pe / Po)
where
C Pe
and
Po
(5.1.17)
= the conductance of the pumping surface, in Z/sec., = the equilibrium vapor pressure of the condensed ice, or the adsorption isotherm equilibrium pressure, = the operating pressure above the ice or surface.
This is simply a variation of (1.15.1) of Chap. 1. In this case the base pressure of the pump is simply the equilibrium pressure of the ice when cryocondensation pumping, and the adsorption isotherm equilibrium pressure in the case of cryosorption pumping. In (5.1.17) we have assumed that the gas impinging on and leaving the cold surface is at the same temperature as the surface. This is implicit in the term C. In order to compensate for possible differences in the free gas temperature, Tg, and the temperature of gas departing the surface, Ts (assuming a unity accommodation coefficient), we must modify (5.1.17). Remember that Pe / Po corresponds to the equilibrium fluxes of gas leaving and entering the surface. In other words, Pe / Po = v v / v i" Making this substitution in (5.1.17) gives us a new equation for the speed of a cryosurface which includes the effects of gas temperature. That is: S
where
=
a Cg [1 - (Pe / Po)(Tg / Ts)V'],
O~
= = = = = =
Pe Po and GAS SOURCE
the the the the the the
BAG
sticking coefficient (5.1.5), equilibrium pressure of the cryosorbed gas, pressure (induced gas) above the cryosurface, surface conductance for gas at temperature Tg, induced gas temperature, cryosurface temperature. APERTURE OF
REFRIGERATION CE AT TEMP. T2
\~i41 \lVI
vrssrL AT
II VA FIRSTSTAGEOF [~fPUMP AT TEMP. T2
11
y
II VA
PREATuRE---~---L~ ROUGH PUMP
/
.
~
VACUUM VESSEL AT TI = 2 9 3 ° K Figure
5.1.10.
The
(5.1.18)
~
~
~'~-PUMP "4/A
~ T I > T2> T3
elements
& WITH AREA A2
AT
SECOND STAGE OF PUMP AT TEMP. T3 & WITH AREA A3
of a t w o - s t a g e
cryopump.
When cryocondensing or cryosorbing gases, (5.1.18) is always applicable. We may now combine (5.1.8) and (5.1.18) to predict the behavior of a cryopump of the more complex configuration such as shown in Fig. 5.1.10. In this case there is a cryopanel of area A 3 located in the colder chamber. The temperature of this additional panel
276
CRYOPUMPING
is Ta, where Ta < T2 < T1. If a cooled chevron was placed over the aperture connecting the two chambers, this would fairly well describe the simplest form of multi-stage cryopumps. More will be discussed later on the purpose of such cryochevrons and cryopump stages.
5.1.6 Cryotrapping Cryotrapping is the concurrent or sequential cryopumping of two or more gases for the purpose of trapping a less readily pumped gas in the cryodeposits of a more readily pumped gas. (1 s g ) For example, def'me D as a less readily pumped gas and E as a more readily pumped gas at some temperature. We might pump the D gas in a mixture rich in proportions of the E gas. The vapor pressure of the E gas might be very low at the given temperature, but that of the D gas very high. If the D gas becomes trapped in the cryofrost matrix of the E gas during its cryocondensation, we say that the D gas has been concurrently cryotrapped. An excellent example of cryotrapping was noted by Wallen where he pumped H 2 simultaneously with varying proportions of either CH4, CO or CO2 at a temperature of 4.2* K.( ~ 6 o ) He showed that the H~ equilibrium pressures could be reduced orders of magnitude, over that of hydrogen alone, depending on the proportions of the "E" gases, CH 4, CO or CO 2 in this case. In another example, assume that we have cryocondensed a thick cryofrost of D gas. However, at the cryopanel temperature the vapor pressure of the D gas is higher than desired. In this case we might introduce sufficient amounts of E gas so as to cover over the D gas with a eryofrost of E gas and thereby depress the observed system partial pressure of D gas. This would be sequential cryotrapping. Overlays of E-type gases have also been used to shield D-type gases from desorption effects due to energetic particles, as Hilleret reported doing with N2 on cryodeposits of H 2. (4 2 ) Again, Wallen conducted experiments where he cryosorbed H2, at 4.2* K, on cryodeposits of either CO, CH 4 or CO 2.(1 6 o ) Cryofrosts of CO or CH 4 showed little benefit as a sorption bed for H 2. On the other hand, this was not the case for CO 2. After depositing a CO 2 cryofrost layer of 2 x 101 8 molecules/cm 2, sorption of 5 x 101 s molecules/cm 2 of H 2 on the CO 2 bed resulted in an H 2 equilibrium pressure o f - 2 . 6 x 10" 1 1 Torr, whereas the same amount of hydrogen on a Cu plated stn. stl. surface yielded a pressure equivalent to the vapor pressure of H 2 at 4.2* K (i.e.,-.- 9 x 10-T Torr). Several requirements become obvious if cryotrapping is to work. First, there must be no surface gas displacement effects. For example, suppose Hobson tried to depress the 2 x 10" a Torr Ar partial pressure, in his experiment of Fig. 5.1.7, by the subsequent introduction of Xe. At this pressure, 0 was < 0 m" Indeed the Xe would have been more readily pumped than the Ar, but it probably would also have displaced Ar molecules off the surface as it successfully competed for physisorption energy sites. Hseuh, et al., observed this gas displacement effect when pumping H 2 and He on activated coconut charcoal.(a s ) We observed a similar effect with TSP pumping. However, in this case gases were displaced in competition for chemisorption sites (i.e., see Table 3.3.1). Therefore, sequential cryotrapping probably has no significant application in cryosorption pumping. Also, this competition for energy sites may explain why cryocondensation pumping of certain gas mixtures leads to
CRYOPUMPING
277
unexpected results. For example, B~ ttner reported that the cryocondensation speed of a mixture of H 2 :D2 (1:5) on supercooled He surfaces was less than their respective speed components when individually pumped at the same temperature.(2 1 ) Some of these anomalous effects, when sieve materials are used, may also be attributable to sieve plugging or unplugging effects (see below). Secondly, if cryotrapping is to work, the D gas must be somewhat insoluble in the E gas. For example, if the D gas is soluble in the E gas, it will readily diffuse in the cryofrost matrix of the E gas. If on reaching the surface of the E gas, the heat of sorption of the D gas on the E gas is less than the D gas heat of vaporization, the D gas will outgas into the vacuum system and eventually equilibrate at a pressure corresponding to its equilibrium isotherm on the E gas. Halama, et al., saw this ~ery effect when attempting to cryotrap He with H 2 on LHe cooled surfaces.(4 a, 4 4 There have been many successful examples of the applications of cryotrapping. For example, in an early paper Hengevoss and Trendelenburg reported on the cryotrapping of both He and H2 with the concurrent pumping of Ar on 4.2"K surfaces.(4 s ) Chou and Halama concurrently cryotrapped both H 2 with D 2 in the proportions of D2 :H2 of --10:1.( 4 4 ) Batzer, et al., used a directed stream of Ar to pump He and isotopes of hydrogen on 4.2* K surfaces in a fusion-related application.(4 e ) The flux of Ar:He had to be maintained in the proportions of-30:1. They resorted to this method of pumping He because the H 2 plugging of sieve materials decreased its capacity for He. They also pointed out the need for UHV conditions when studying the cryopumping of one gas. They indicated that cryotrapping of the study gas by system background gases could lead to errors. Longsworth and Webber conducted similar concurrent cryotrapping experiments for Ar and H2 in a CLGHe cryopump.(2 4 ) Also, a number of sequential cryotrapping experiments have been conducted to determine the effectiveness with which D gases are cryosorbed on cryofrosts of E gases (e.g., see (1 2,4 7-s o ,g o )). These studies were undoubtedly in part aimed at determining the heats of sorption of D gases on the E gases as they relate to the mentioned solubility-diffusion considerations. Also, it has been concluded that D gases are less readily desorbed from metal surfaces if a substrate of E gas is first deposited on the surface.(1 2,9 o ) It is believed that the E gas serves as a phonon barrier between the D gas and the metal substrate. Some of these works and others related to cryotrapping are referenced in Tables 5.1.3 and 5.1.5.
5.1.7 Sieve Materials In cryosorption pumping it is obvious that the equilibrium pressure within the vessel is critically dependent on o, the measure of surface coverage. Further, for a fixed pressure, o depends on the temperature of the pumping surface. As will be discussed, there are certain practical limits to the temperatures which can be readily achieved in various forms of cryopumps. Because of this, cryocondensation pumping has limited application in the UHV pumping of the gases He, H2 and Ne. For example, the practical lower temperature limit of CLGHe cryopumps is -10" K. We see from the data of Haefer( s 2~ (i.e., Fig. 1.7.4 of Chap. 1) that the gases He, H2 and Ne would not be effectively cryocondensed at these temperatures (i.e., at UHV pressures). Therefore, these gases can only be pumped at lower pressures by cryosorption. Sorption roughing pumps are operated at LN2 temperatures (i.e., -77* K).
278
CRYOPUMPING
Therefore, if these pumps are to be used to rough pump substantive quantities of gases, cryosorption pumping must come into play. In fact, the vapor pressure of LHe (liquid He) is --750 Torr at --4.3* K. At this temperature, a LHe cooled surface would be of marginal value in the pumping of H 2 in UHV applications and useless in the cryocondensation pumping of He. However, I have noted examples of the effective cryosorption pumping of both of these gases at temperatures < 20* K. Therefore, many gases, including the difficult gases He, H2 and Ne, must be pumped in most applications by cryosorption pumping. But we have seen that cryosorption pumping has an inherent capacity limit established by the effective area and temperature of the cold surface. As the surface coverage of gas increases, the equilibrium pressure rises. I model this in my mind's eye as being due to mobile surface gases. That is, because some molecules temporarily residing on the surface still have sufficient thermal energy to move about on the surface, from time to time they bump into their neighbors and give these neighbors sufficient energy to escape the surface, as gas. It is reasonable that the consequences of this unneighborly behavior (i.e., higher pressures) would depend on the surface population on the gas. Solid state physicists explain the higher pressures, with higher surface coverage, as stemming from phonons. A phonon is a quantum bundle of heat energy propagating through the crystal lattice. These phonons, residing in the cooled substrate, bump into the surface gas and cause its ejection. If, for a given temperature, the phonon concentration is constant in the substrate, it is reasonable that with increasing surface coverage more gas would be ejected. No matter which model you pick, the consequences are the same, the greater the surface population, the higher the equilibrium pressure. Therefore, the only hope of increasing the cryosorption capacity of a pump for He, H 2 and Ne is to increase the effective surface area of the pump. You will recall that charcoal was used in making the measurements leading to the data in Table 5.1.1. This material has very strange and unique properties. That is, for a comparatively small mass of the material, charcoal evidences very large effective surface areas. Materials with this unique property have come to be known as molecular sieve materials, or just sieve materials. Actually, the first reported vacuum use of charcoal was in a paper read by Professor Sir James Dewar before the Royal Society of Edinburgh in 1874. I discovered its existence when it was referenced in a vacuum application note appearing in the journal Nature a year later.(5 3 ) In this second paper he reported heating the charcoal while pumping on it with what he called "a mercury pump". He then sealed off the charcoal in the glass vile, while it was still hot. The cryogen was the RT air which cooled the vile after bakeout. He reported on various electrical and mechanical experiments conducted in the sealed vile which served as verification of the improved vacuums achieved by the cooled charcoal. Over the last three-quarters of a century, much work has been done in studying the properties of these sieve materials at both RT and cryogenic temperatures. The variety of materials studied has included almost everything imaginable (e.g., cement, coal, rubber, glass, etc.).( 5 4 ) Study of their RT properties was ziven impetus because of the development of gas masks before and during WWI.(5 5"~ Unfortunately, these sieve materials must still be used as adsorbents in this application. An adsorbent is a material on which gas is adsorbed. An adsorbate is the gas which is adsorbed on an adsorbent. These are often abbreviated as sorbents and sorbates respectively. A review of much of the early work on the pumping properties of various sorbents for various sorbates, and over wide temperature ranges, is
CRYOPUMPING
279
given by Dushman.(S 6 ) A second extremely important RT application of these sieve materials is in the area of catalysis.(Z z ,s 7 ) Use of catalysis on an industrial scale requires large surface areas. Because of this, a variety of man-made sieve materials were specifically developed for catalysis applications. Some are called artificial zeolites. Zeolite is an aluminosilicate, crystalline matrix. It occurs naturally in some forms, the man-made material being distinguished as artificial, and usually comprising some form of calcium or sodium aluminosilicate.(S 8 ) Because of their large surface areas, these sieve materials have been investigated for vacuum applications. Some of these materials are listed in Table 5.1.4, an embellishment of a summary published by Stern, et al.,(3 4 ) and Redhead, Hobson and Kornelson.(S 9 ) I find the results of this table to be nothing short of remarkable. For example, a gram of activated coconut charcoal has the equivalent effective surface area o f - 8 9 0 m2 (i.e., using the B E T method). The Linde 5A artificial zeolite has an effective surface area of -600 m 2/g, and the 13X artificial zeolite --515 m 2/g, etc. Table
5.1.4.
Effective
SIEVE MATERIAL
areas
of
some
FORM
sieve
materials.
AREA 1 me/gm
P o r o u s Silver 0.22 2 sintered matrix P o r o u s Silver 0.71 3 sintered matrix Alumina spheres, 3 mm¢ 287 Silica Gel, Type ID granular 311 Linde 13X ( a r t i f i c i a l z e o l i t e ) 514 powder Linde 5A ( a r t i f i c i a l z e o l i t e ) 600 pellet Silica Gel, Type R granular 784 Coconut Charcoal granular 889 S a n a n C h a r c o a l , Type S - 8 5 powder 1170 Coal granular 1050-1150 C o c o n u t C h a r c o a l , Type PBC granular 1150-1250 NOTES: 1. M e a s u r e d u s i n g t h e BET m e t h o d . (33) 2. M e a s u r e d u s i n g Ar r a t h e r t h a n N2. 3. M e a s u r e d u s i n g He r a t h e r t h a n Nz.
(34'59)
POUR SIZE-.~
50 150 9-10 5 22 10-30
REF. 20 37 34 34 34 34 34 34 110 35 35
These results imply that if there was some convenient way to bond these materials onto cryogenically cooled surfaces, it would be possible to cryosorption pump many orders of magnitude more gas than would be possible on similarly cooled metal substrates of the same geometric surface area. For example, Danilova reported the cryosorption pumping of only --6 x 10-7 Torr-Z He/cm 2 on a Cu surface cooled to 4.2"K and before an equilibrium pressure of --10 -7 Torr was observed.( 3 9) Halama, by cooling the surface onto which a sieve material was bonded to 4.2" K, was able to pump --23 Torr-Z He/cm2 of geometric surface area and achieve the same equilibrium pressure.(2 6 ) That is, for the same projected surface area, --4 x 10r more He could be accommodated. Similarly, Sedgley, et al., epoxy-bonded activated coconut charcoal to a cryopanel.(3 s ) When the surface was cooled to 4.2* K, and the equilibrium He pressure reached -~10 -5 Torr, they reported the pumping of--2.0 Torr-Z He/cm 2 of bonded material. This corresponds to --x 1.2 x 10 6 more He than could be accommodated on a Cu panel at equivalent temperatures. The term activated coconut charcoal (i.e., vs. less active) has to do with the
280
CRYOPUMPING
relative sorptive capacity of the charcoal. This relative sorptive capacity varies with the ref'mement of impurities out of the charcoal and vacuum and air roastin~procedures. Use of proper procedures and recipes leads to activated charcoal.(S s-) The mysterious properties of these materials are qualitatively explained by Fig 5.1.11. In Fig 5.1.11a, the cryogenic pumping surface is a fiat, metal plate. On an atomic scale, the surface is comparatively smooth, so that gas can randomly move about the surface, and suffer collisions with neighboring molecules on the surface. Gas on the surface of the plate is pumped gas, where gas which is bumped off contributes to the pressure in the system. Bulk sieve material comprises a complex matrix of caverns or cavities, with interconnecting apertures. For example, Bolton notes that the unit cells of the Linde 13X zeolite matrix comprise "... eight large cavities, each with a free diameter of about 13 A".(s 7 ) These cavities are interconnected by apertures with diameters of -7.4 A. The Linde 5A zeolite has internal cavities with diameters of the order of 11 A, which are interconnected by apertures having diameters o f - 4 . 2 A. Because of these caverns within the bulk of the sieve material, copious quantities of gas can be pumped. "FREE"
EJECTED
GAS -~-~@ ~ ~ - GAS SURFACEx • .[: ' O T| -, [• • GAS~ i ~[~ " , " CRYOPUMPINGy SURFACE Figure
5.1.1 l a .
SURFACE DIFFUSION
Cryosorbing
plate.
GEOMETRIC
INTERCONNECTING
SURFACE\
INTERNAL ~ CAVERNS
k
//~APERTURES
~
~
F i g u r e 5 . 1 . 1 l b . C r o s s s e c t i o n of s m a l l c h u n k of s i e v e m a t e r i a l .
The mechanism for gas pumping, depicted in Fig. 5.1.11b, is as follows. Gas impinges on the geometric surface area of the sieve. Because of a build-up of surface concentration gradients, the gas diffuses on the surface of the sieve and as free gas down into the caverns within the bulk material. If a molecule, when residing on the wall of one of these inner caverns, is bumped off by a neighboring gas molecule, it lands on a nearby cavern wall. This struggle for real estate between molecules occurs within the innards of the sieve and therefore has no effect on the pressure over the geometric surface area of the sieve. I have coined this process as a labyrinth effect. Note that if the interconnecting apertures are large, gas is able to more readily diffuse down into the bulk of the material. However, this accessibility argument works both ways. That is, it will be easier for free gas within the bulk to find its way out of the labyrinth. For this reason, both the sieve surface area and the size of the interconnecting apertures are important. Table 5.1.5 is a chronological summary of some of the publications relating to the use of different sieve materials for the pumping of a variety of gases and in applications involving sorption, LHe and CLGHe cryopumps.
Plugging of Sieve Materials The preponderance of the surface area of a sieve material is located within the body of the sieve. Therefore, it is important that the entrance apertures to the body of the
CRYOPUMPING
281
sieve remain open so that gas has access to the inner surfaces. Because of the relatively small sizes of these entrance apertures, they can become plugged by certain gases. The phenomenon associated with the plugging of sieve materials is called persorption .(a a ) It merely means that the bigger or more polar the sorbate molecule, the greater the likelihood that it will plug the entrance apertures.
Table 5.1.5. on the use SURFACE TEMP. o K
Examples of sieve
APPROX. PRESS. Torr
77 4.2
not indicated 10 -4 - 7 5 0 4 x 1 0 -6 - 4 x 1 0 -4 10-zl _10-6
20.2 20.2 20.2-40.4 .. 4.2
10-8 _ 1 0 - 4 10 -5 - 1 . 0 10 -7 - 1 0 . 0 5 x 1 0 -8 _ 7 x 1 0 -4
20.2 4.3 77.3 10.7
7 x 1 0 -9 _ 1 0 -4 8 x 1 0 -10 _ 4 x 1 0 -7
4.2 9 77.4 77
77.3
of the referenced literature and forms of cryotrapping.
GASES
SIEVE MATERIAL I Charcoal & porous MS-5A, ACC. MS-13X, ACC. CCT w i t h At.
34 34 34 68
Seven sieve m a t e r i a l s . M S - 13X, s i l i c a gel. ACC. MS-5A.
3 x 1 0 -7 - 2 x 1 0 -5
H2 He, Hz He, Ne, N2 He, H2
67 101 29 79
MS-SA. MS-5A. MS-5A, MS-10X, MS-13X. ACC.
5 x 1 0 -11 _ 2 x 1 0 -3 not indicated 10 -11 _ 1 0 2 10 -6 - 7 5 0
He H2 Ar, Kr, Xe Air
37 49 20 99
P o r o u s silver. On Ar c r y o f r o s t . P o r o u s silver. MS-5A.
He*, H2 H2 He He
86 47 25 26
MS-SA. On CO z c r y o f r o s t . MS-5A. MS-SA.
8 x 1 0 -8 5 x l O -8 10-11 10-11
_ f l x l 0 -5 - 3 x 1 0 -4 _10-7 _10-7
He,H2,Ne,N2,02,Ar,CH4 Air, H20 N2, Ar He, H2
REF. 60 80 104 45
10 - 6 - 102
4.2*,20 12.4-21.5 ;4.2 4.2
of some materials
para-Ha,
ortho-H2 He He He
glass,
4.3, 20 4.2 ~20 4.2
3 x 1 0 - a - 4 x 1 0 -3 ~<1.3 10 -4 10 -6 _ 7 x 10-4 not indicated
He, H2, D2 H2 He H2 +He, D2 +He, H2+D2
63 103 106 44
MS-5A. On D2 c r y o f r o s t . ACC. CCT.
~.2 2.23-4.2 7-20 7.5-15.3
1 0 " a - 1 0 - a (H z) not indicated
>/10 -6 10 -~1 _10-5
He, Hz, D2, D Z+He He, He, N2 D2 He
65 42 85 48
MS-5A. SCT w i t h Nz. M S - 5A. On CO2 c r y o f r o s t .
4.2 4.2-18.2 4.2 2.18, 4.2
2 x 1 0 -7 - 4 x 1 0 -4 10 -9 _ 2 x 10-6 10 -7 --6x 10-5 8x10-10 _10-5
He, H2, He+H2 He, He+H2 H2, D2, Hz+D2 He
62 38 21 51
MS-5A. ACC. CCT of H2 w i t h D2. On Ar c r y o f r o s t .
4.2 10-20 4.2 4.2
varied varied 6 x 1 0 -5 - 4 x 1 0 - 4 not indicated
He, Hz, D2, Tz He He, He+D2 He, Hz
43 43 46 61
ACC. ACC. CCT w i t h Ar j e t s t r e a m . ACC.
18 - 2 2 6.0-11.0 9.5 10-18
10-7 _ 1 0 - 4 8 x 1 0 -10 - S x l 0 - a not indicated 10-10 _ 1 0 -9
Hz H2 He, Hz He
105 50 70 84
Z X - 1 5 c h a r c o a l (PRC). On C02 e r y o f r o s t . ACC. ACC.
4.2-10 7.5-8.0 4.3 4.3
not indicated 4 x 1 0 -6 _ f l x l 0 -5 not indicated not indicated
He He He, H2, He+H2 He+Ar
69 35 64 64
ACC. V a r i e t y of c h a r c o a l s . ACC. CCT of He w / j e t of Ar.
5(?)-23 4.3 10.4-17.5 11-25
1 . 2 x 1 0 -5 - 2 . 3 x 1 0 -4 4 x 1 0 -6 _ S x 1 0 -5 10-7 _ 1 0 - 4 2 x 1 0 -6 - 1 0 - 4
He He He Hz
109 64 102 102
ACC; CLGHe c r y o p u m p . Six v a r i e t i e s of c h a r c o a l ACC., CLGHe c r y o p u m p . ACC., CLGHe c r ~ o p u m p .
_
1 Key:
SCT: S e q u e n t i a l c r y o t r a p p i n g of o n e g a s w i t h a n o t h e r . CT : C r y o t r a p p i n g . ACC: A c t i v a t e d c o c o n u t c h a r c o a l . CCT: C o n c u r r e n t C r y o t r a p p i n g w i t h m o r e r e a d i l y p u m p e d gas. CF of S: P u m p e d on c r y o f r o s t s of s a m e g a s or s u b s t r a t e . MS-5A, 10X a n d 13X a r e z e o l i t e p r o d u c t s of t h e U n i o n C a r b i d e
Corporation.
282
CRYOPUMPING
The phenomenon of persorption is used to determine the size of the entrance apertures in sieve materials. This plugging of entrance apertures can cut off all gas access to the inner sieve labyrinth. Gases which prove particularly troublesome in the plugging of sieve materials include oil vapors and water vapor. Also, any gas which might cryocondense on the outer surface of the sieve, so as to bridge over the entrance apertures, will cause plugging of the sieve. For example, Stern, et al., reported on the plugging of Linde 5A sieve by N2, when pumping H2 at --20* K.( s T ) Even H2 and D2 have been noted to plug sieve materials for the pumping of He, when the sieve was operated at --.4.2* K.~.4 6,6 z-6 5 ) This has been observed with both activated coconut charcoal(6 z ) and the artificial zeolites.(6 2 ) There may also be some gas displacement effects as well as plugging effects in these cases. In the early days of CLGHe cryopumping, these pumps were often used to replace diffusion pumps on vacuum systems. Diffusion pumps are notorious for oil backstreaming into traps, valves and systems to which they are appended. If, when replacing the diffusion pump with a CLGHe cryopump, the remainder of the system is not first completely cleaned of traces of the pump oil, eventually this oil will find its way to the sieve material and cause irreversible plugging. Problems due to the sieve plugging (e.g., negligible subsequent H2 capacity) might take several months to become evident. Within time the system might have been shut down and the pump cycled back to RT several times. However, eventually the oil will get to the sieve material by creeping over RT surfaces. Sieve materials are frequently used in devices to trap out possible oil backstreaming from mechanical pumps. These sieve traps may be periodically baked to high temperatures (i.e., --300* C) to get rid of the oil accumulation, and refresh the sieve trap. However, baking to this extent is not possible with many forms of cryopumps. For example, if the sieve material of a CLGHe cryopump becomes plugged with hydrocarbons, the element to which the sieve material is bonded must be discarded, and replaced with a new element. This is for two reasons. First, the sieve material is usually bonded to the surface of a cryopanel using some form of nonbakeable epoxy-resin. Secondly, the CLGHe mechanisms would be damaged by high temperature bakeouts. Sieve materials are also plugged by H ~ O. This gas is present in most systems. Also, if it is initially cryocondensed on some surface other than the sieve, on warming up the pump at some later time, the water vapor will eventually find its way to the sieve material. Because of the tenacious manner in which H 2 O sticks to many of these sieve materials, the only way in which it can be removed is through bakeout. This is particularly true of the artificial zeolites,(2 9,9 9 ) and there is some suggestion that it also applies to porous silver sieve materials.(2 o ) For this and another reason, these zeolites are primarily confined to use in sorption pumps and bakeable LHe cryopumps. The above H2 O sieve plugging problem would seem to present an insoluble obstacle to the use of these materials in CLGHe cryopumps. Fortunately, activated coconut charcoal does not require baking to remove most of the pumped water vapor. In fact, very pure charcoal is said to be almost hydrophobic in nature.( 6 6 ) That is, at the conclusion of some number of pumping cycles, the pump may be brought back to RT. Water vapor consequently finding its way to the charcoal may be subsequently pumped away at RT using some form of trapped roughing pump. This proved most fortuitous, as it made possible the practical use of mechanical
CRYOPUMPING
283
refrigerators, capable of achieving temperatures of 10-20" K, in the cryosorption and cryocondensation pumping of all gases. Longsworth constructed a summary of the sorption properties of activated coconut charcoal in the temperature range of 10-30" K. This is given in Fig. 5.1.12.(1 4 2 ) I will return to this subject when discussing closed loop gaseous helium cryopumps later in this chapter. k~ ©
I
1 0 +4 30
10° 10
-4
~, 10
-8
= 10
-1~
-16
-,
10
~
10 -eo
K - ~ -_..
J ~/
-20 K--~ _ - / - / )("
Or]
,.Q .~
1
//
/
"~
./"
/f
,- / > i
/
/ /
,
/ /
0
/. / /.// / / / / / / ~'-~-15
" / r
I
~-12
K-
K-
/"
:/, 1
---
/ J r / ....
/
/
/
/ ,/7
100
/
/
i-
/
,
/
/
~-10
K
"
200
300
400
500
S t a n d a r d c m 3 of H y d r o g e n p e r g r a m of a c t i v a t e d c h a r c o a l F i g u r e 5.1.12. A d s o r p t i o n I s o t h e r m s of H y d r o g e n on AcLivated C o c o n u t C h a r c o a l v. C h a r c o a l T e m p e r a t u r e . ( 1 4 2 )
Surface Bonding of Sieve Materials Developing methods, free of organic binders, for the bonding of sieve materials to surfaces has been a challenge. The artificial zeolites are ceramic-like and have the form of grains or nuggets. Coconut charcoal may be purchased in small, regularly shaped cylinders of--4 mm ~ x 7 mm, or in irregularly shaped chunks varying in size from < l m m t o > 4mmg~. I surmise that technologists used some form of glue or epoxy to bond sieve materials to metals prior to 1964, as Hemstreet, in a patent application at that time, noted the disadvantages of this practice.(S s ) He noted that the epoxy or glue tends to diffuse into part of the bulk of the sieve and cause plugging. He described in the final patent a number of methods used to successfully bond both charcoal and artificial zeolites to metal surfaces. Some involved the formulation of a slurry comprising inorganic binders, sieve material ( < 1 mm cp) and metal particles ( < 5 mm ~). This slurry was applied to a roughened, pretreated surface and then fired in either N 2 or Ar, so as to sinter the matrix together and to the metal surface. Other methods, also noted in the patent and elaborated on by Stern, et al., involvedputting the bulk slurry into cross-milled grooves machined into the metal surface.( "6 ~ ) Thereafter, the plate was baked to remove the inorganic binders. Presumably, it was still required that the matrix be sintered to the grooves, cut in the plate with a circular end-mill, though Granier and Stern also used this same method with an A~ plate.( 8 s ) This technique was used by the Excalibur Corporation to bond sieve materials to the third stage of a bakeable LHe cryopump (e.g., see below).
284
CRYOPUMPING
Hseuh, et al., report using a 6.3 mm thick bed of 3.5% AgSn solder to bond coconut charcoal.(4 a ,6 1 ) The charcoal was pressed half way into the alloy while it was heated to a molten state under vacuum. He indicated that when subsequently inverting the bed and tapping on the metal plate, - 1 0 % of the charcoal fell off the surface. However, the surface of the solder could not be seen in the vacant sites as some of the broken charcoal remained stuck to the surface. Tobin, et al., report the successful bondinlz of coconut charcoal to Cu plates with the use of a "silver-based braze alloy".(6 9 ~" According to Tobin, this was only applicable to the coating of small, flat surfaces. It was applied to a copper plate sing a P, 80% Cu) and an oxyacetylene torch.(8 8 ~ sheet of Sil-Fos (i.e., 5% Ag, 15% Coupland reported on the use of sintered Ni as a sieve.(T o ) Hobson reported on a technique of flame-spraying an 8.5% CaAg alloy onto a stn. stl. surface.( 2 o ) The Ca was thereafter oxidized with steam, and leached out of the matrix with 20% acetic acid, leaving a 1.25 mm thick bed of porous silver bonded to the surface. The bed could be baked to -500* C. All commercial manufacturers of CLGHe cryopumps use some form of epoxy to bond the activated coconut charcoal to the second stages of the cryopumps. It appears that this practice was rediscovered some time in the early to mid 1970s. I recall, when visiting CTI Incorporated in August of 1975, that they were producing and selling three varieties of cryopumps at that time. All featured the use of activated coconut charcoal on the second stages. Turner and Hogan, when reporting on a modified Taconis-cycle cryopump in 1966, were unaware of the technique.t~ 1 ) Cryopumping papers published as late as 1977 reported the need of the use of turbomolecular pumps to augment CLGHe cryopumps in the pumping of He and H2.( ~' 2 ) This same year Visser, et al., reported on the use of epoxy to bond charcoal to the second stage of a CLGHe cryopump. In May of 1976, Longsworth read a paper (unpublished) on CLGHe cryopumps before a joint meeting of the New York and Philadelphia Chapters of the AVS. The implication of this paper was that he had used epoxy to bond charcoal to the second stages of cryopumps on or before 1975. This technique has also been successfully used in LHe cryopump applications. (3 s )
5.2 Sorption Roughing Pumps Sorption roughing pumps are not, as thought by some, scientific novelties which are confined to use in some laboratory. On the contrary, they are widely used in a number of very practical applications. For example, with few exceptions, all UHV surface science apparatus comes equipped with sorption roughing pumps. All large molecular beam epitaxy systems are sorption roughed. On a much larger scale, Neal instituted the use of sorption pumps to rough the sectors of the two-mile SLAC accelerator.( ~ 3 ) Each sector comprised a volume of --7 x 10 a Z with a surface area of - 3 x 10 6 cm2 of stn. stl. and Cu. Prior to the advent of the use of CLGHe cryopumps in the semiconductor industry, there was a brief period when many coating systems were combination-pumped systems. All these systems were cryosorption roughed. The primary motive in using sorption pumps in all the above cases was to avoid possible catastrophic problems stemming from back-diffusion of oil from mechanical roughing pumps. A definitive work on the practical application of sorption rough pumps was published in 1972 by Frederick Turner of Varian Associates.(7 4 ) I liberally reference his work herein.
CRYOPUMPING
285
If one were to pour a quantity of sieve material into a small volume with flanges on each end, such as shown in Fig. 5.2.1a, the bulk sieve would afford an effective trap for gases including water vapor and oils. These are called sieve traps and are often used in conjunction with mechanical pumps to protect, to some extent, the system being rough pumped from backstreaming pump oils.(r s ) As will be later discussed, these RT sieve traps are also used to filter out impurities in gas streams. For similar reasons, bulk sieve materials have also been used, both at RT and reduced temperatures, as traps over mercury and oil diffusion pumps, as shown in Fig. 5.2.1b. FLANGE
~m~
4'
i
~~.
TO
SIEVE
).
,SCREE
]
[' t
,
.
L_
I
,
~/7/ BAKEOUT
I I I TO MECHANICAL PUMP
\
JACKETS
Figure 5.2.1a. Mechanical pump sieve trap. (75)
~,\'i
~
. t-.\\~.J ' ~J l
i / I r l V////~ ' V/~ , TO DIuFFMUpSION
Figure 5.2.1b. Sieve trap used to limit oil or Hg backstreaming.
The first commercial use of bulk sieve materials for sorption roughing came about as a consequence of the use of sputter-ion pumps. These pumps and the devices to which they were appended (e.g., klystrons) could be damaged by backstreaming oils from mechanical roughing pumps. Jepsen submitted patent application for a commercial sorption roughing pump in 1959.( r 6 ) This pump made use of bulk coconut charcoal. A cross section of the pump is shown in Fig. 5.2.2. To effect pumping, the vessel containing the charcoal was cooled on the exterior by filling a dewar in which it resided with LN 2. PUMP
SCRI~
)Y
F i g u r e 5.2.2. R e p r e s e n t a t i o n of t h e f i r s t c o m m e r c i a l s o r p t i o n r o u g h i n g p u m p as d e s c r i b e d by J e p s e n . ( 7 6 )
286
CRYOPUMPING
This type of sorption roughing pump was widely adopted in many applications requiring the preservation of clean surfaces and UHV conditions. However, the use of coconut charcoal had a couple of disadvantages. First, the charcoal tended to fragment into carbon dust within the pump. This dust could accidentally be carried over into a U H V system if the isolation valve between the pump and system was inadvertently opened while the system was under vacuum and the pump at atmospheric pressure. Secondly, the presence of LN2 suggests that under certain circumstances there may be LOs (i.e., liquid oxygen) present. Carbon dust and LOs comprise a potentially explosive mixture. This was recognized, and the artificial zeolites were subsequently used in sorption pumps of similar configuration. RELIEF C0RK
VENT
', {
~
2:
23
j~ oO6Oo o oOoOoOo
ooOoOoOo °°°°°°
'
~
./~-ALUMINUM FINS
IEVE
OoO o OO 0
o
REEN
,
I
Figure 5.2.3. R e p r e s e n t a t i o n of a s i n g l e - l e n g t h s o r p t i o n p u m p , with --1400 g sieve c h a r g e , offered by t h e Varian V a c u u m Division.
These pumps are extremely simple devices. Bulk sieve material is simply poured into the volume of the pump. A coaxial screen is located down the center of the pump so that gas will have access to the sieve material the full length of the pump. Single sorption pumps are available which accommodate from -1.4 to 4.2 kg of sieve material. The pump, and accompanying LN2 dewar, are simply made longer if greater capacity is needed. It is important that the sieve material be uniformly cooled. Various commercial designs were developed to facilitate this uniform cooling of sieve material. Two such schemes are shown in Figs. 5.2.3 and 5.2.4. The design in Fig. 5.2.3, a sorption roughing pump manufactured and sold by Varian Associates, makes use of extruded A~, finned quadrants. The finned quadrants are longitudinally welded to form a tubing having an inner finned structure. A blank A~ cap is then welded on one end of this cylinder and a flanged cap, with an A3. to stn. stl. transition joint, is welded to the other end. This simple configuration, with a screen and some sieve material, comprises the Varian sorption pump. The fins, radially protruding into the innards of the cylindrical pump, afford a means of
CRYOPUMPING
287
cooling even that sieve material located near the center of the externally cooled pump. ~---,,,70.0 mm ti I--~ [ [
I ~
PUMP FLANGE
PRESSURE REUEF CORK
~
CIRCULATION
o°
°o
o' )
SCREEN
o °o
t
~
,-,114 mm 0
=
Figure 5.2.4. R e p r e s e n t a t i o n of a s i n g l e - l e n g t h s o r p t i o n p u m p , with ~1400 g sieve c h a r g e , offered by P h y s i c a l E l e c t r o n i c s , Inc.
Physical Electronics commercially manufactures and sells a pump of slightly different design. This sorption rough pump is represented in Fig. 5.2.4. In this configuration, cooling of the more centrally located sieve material is accomplished by the use of six stn. stl. tubes, extending the full length of the sorption pump and serving as convective passages for the flow of LN2. The visible top of the sorption pump is in fact not an integral member of the vacuum vessel. Rather, it serves as a splash shield for the LN2 which is vigorously circulating about the outer shell and within the six convective tube passages. Copper fins are attached to the six tubes, radially extended out into the volume of the pump, and thus provide the needed sieve temperature uniformity during cooldown. These two pump designs are merely given as representative examples of commercial sorption pumps (e.g., see (7 T )).
5.2.1 Staging of Sorption Pumps The staging of sorption pumps is defined as the sequential use of more than one pump to rough a volume, but with no more than one pump evacuating the volume at the same time. For example, assume you have three sorption pumps appending your UHV system for the purpose of roughing the system. Such a configuration is schematically represented in Fig. 5.2.5. The reasons for this staging of pumps will be clarified, after first clarifying the meaning of staging. The sorption pumps each have their associated valves leading to a common manifold. These may be elastomer sealed valves. The roughing manifold is in turn attached to the UHV system through one valve, which may perhaps be of all-metal construction. In some applications, the first roughing stage of this threestage roughing is accomplished by some form of bulk roughing; pump such as a carbon-vane, gas-aspirator, or even a metal-bellows pump.t7 a ) By means of the
288
CRYOPUMPING
first-stage roughing, anywhere from 50 to 80% of the bulk gas is removed. After this, the sorption pumps come into play. Though these pumps might be simultaneously cooled with LN 2, roughing is accomplished by the sequential valving in and out of each pump. THERMOCOUPLE GAUGE
~ ,
t,,\N l]I
Figure
,
, ''' I!' ! ~ 1 ~ [il i i i~J LN2 SPLASH
,
I 111
l1
5.2.5.
~ ,,,
Typical
UHV ISOLATION VALVE
~
I
,['~:-~, ~ [~ !1~-~] I;q ti~
~
~~
.F. ~_ - - ~ I
t [l .........]'li I- ~i i ~
~
[~
configuration
of
BAKEOUT
~ ~
Ii ~
SLEEVE
~
a multi-stage
JACKET
POWER CABLE /
Pb[yMsO3~ARBELNEE
sorption
system.
This will lead to a pump-down curve such as given by Turner,(Z 4 ) and shown in Fig. 5.2.6. In this instance, Turner used three sorption pumps to rough pump a 200 Z vessel, initially at one atmosphere of air. No first stage bulk roughing pump was used. a :- l I [ - ~ ~ m t ~
10
t
~
T
I
I
'
[
l
J
I
-
FIRST STAGE
_ ~
ROUGHING
2
o ,o!
[-., I
10 1
....
a:;
-
O~
o
10
10
~"
SECOND STAGE G I "
-1
_
-2
10-3
\
-
1
I
0
1
I
5.2.6.
single-high
The
THIRD STAGE -ROUGHING -
l
1
1
5 TIME
Figure
I
PUMPDOWN OF A 4.72 ~ / s e c . PUMP
staged
sorption
roughing pumps filled
1 10
-
15
Minutes
of a with
200 liter volume using three L i n d e 5A s i e v e m a t e r i a l . ( T M )
CRYOPUMPING
289
In this measurement, each sorption pump was used individually to pump on the vessel, and it was then valved out prior to use of the next sorption pump. This is what is meant by the staging of sorption pumps. Note that though the first knee of this pressure vs. time curve occurs at ~ 3 minutes into the pumpdown, over half of the bulk gas is pumped in this interval of time. For comparison purposes, the theoretical pumpdown curve of a 4.7 Z/sea (i.e., 10 ft a/min) mechanical pump is also shown in this figure. These sorption pumps were charged with -1400 g of artificial zeolite. Several of the artificial zeolite materials were exhaustively characterized by Stern and his colleagues.(S 4 ) Typically, the Linde 5A sieve is used. An adsorption isotherm for the gas N 2 and one of these pumps, as taken by Turner, is given in Fig. 5.2.7.(r 4 ) 150
120
1
I II
1
I II
I
I 11
1
I II
1
I 11
1
Ill
1
1 II
- Na ISOTHERM AFTER
%
TORR-~f/g -~
!
90
) I 6o
~
N
z ISOTHERM
__
o 30
/ 1
10. 4
1 11
[
10-3
I I1
TO 77.3 K BEFORE EACH M E A S U R E M E N T l
l0- e
I 11
10-1
1
t 11
i
10 0
NITROGEN PRESSURE F i g u r e 5.2.7. V a r i a t i o n s sorption pump charged
I Ill
10 1
l
llll
1
10 a
Ill
10
Torr
in N2 a d s o r p t i o n i s o t h e r m s , w i t h t i m e , in a w i t h 1350 g of L i n d e 5A s i e v e m a t e r i a l . (74)
The pump was charged with 1350 g of Linde 5A sieve material and refrigerated with LN2. Because of the thermal conductivity limitations of the sieve material, and the aforementioned diffusion-related processes associated with molecular sieve pumping, achieving the equilibrium pressures of an adsorption isotherm is a long-term process. This is evident in the data given in Fig. 5.2.7. For example, Hobson reported that when pumping both Kr and Xe on porous Ag at LN 2 temperatures, it often took up to an hour for pressures to equilibrate.(2 o ) Similar equilibrium pressure, time-dependent effects were reported by Stern, et al., for the gases H2 and He on a variety of sieve materials at 20.2* K and 77.3* K;( 2 9 ,a 4 ) also see, for example, Halama and Aggus( 2 s ,2 6 ) for He on Linde 5A sieve at >4.2* K; Sedgley, et al., for the pumping of He on six varieties of charcoal;(a s ) and Longsworth for H2 on charcoal and supercarbon.(2 4, ~ Q ) Because of this time dependency, one must use judgement in applying any adsorption isotherm data to the calculation of pumpdown performance of sorption pumps. Contrary to common belief, the gases He, H2 and Ne are sorption rough pumped. Adsorption isotherms for the separate gases He, H~ and Ne, taken by Turner, are given in Fig. 5.2.8.( 7 4 ) These isotherms are given for one sorption pump, with a charge of 1350 g of Linde 5A sieve material at LN 2 temperature (i.e.,
290
CRYOPUMPING
-77.3* K). Of course, the data in this figure were taken after the extended times needed for true equilibrium to be achieved. 10 5
-
I
-
I
111
-
III
-
O9
lO
I
J
I II
I
Ill
1
10
HYDROGEN
3 -
_
ISiTHERM ~ NEON ISOTHERM
2
1
_ ~-
-
- HENRY'Sj/~/f 10
I I ~ J ~ l L
O
I
I
t
10 4
I
I
77.3 OK PUMP TEMPERATURE
LAW
-
_
!
/ , , /
_
~
_
0
10
.< c~
HELIUM -1 10
~OTHERM _ _
-
-2
10
_
,/,,
_
1
-5 10
1 I1~
I
-4
10
-3
10
I
-2
1
11
I
11
-I
10
10
10
EQUILIBRIUM PRESSURE Figure 5.2.8. Adsorption isotherms sorption pump containing ~1359
for
He,
I
Ne
g of Linde
1 11
0
1
10
1
1 II
10
2
Torr and H2 5A s i e v e
pumped with m a t e r i a l . (74)
a
To qualitatively appreciate the importance of staging, let us conduct a simple gedanken experiment (i.e., a mental exercise). Assume that we have a vessel with a volume of 1.0 Z and that it is pressurized with H 2 to 750 Torr. Assume that a valved manifold is constructed which enables us to simultaneously pump on the vessel with 750 sorption pumps, each having 1350 g of Linde 5A, and each chilled to LN2 temperatures. In that we have simultaneously valved in all 750 sorption pumps, it is necessary that each sorption pump, on reaching equilibrium, pump the equivalent of only -1.0 Torr-;~ of H2. Referring to Fig. 5.2.8, we can therefore predict that on achieving equilibrium, the total system H 2 pressure will be --4 x 10 -4 Torr. Now let us attach only two of the same sorption pumps to the system, and repressurize the 1.0 Z volume with 750 Torr of H 2. This time, let us valve in only one of the two sorption pumps. The pressure of the first sorption pump, on achieving equilibrium, is again found from Fig. 5.2.8, and determined to be - 5 x 10 -2 Torr as a consequence of pumping -750 Torr-Z H2. Now, let us valve out the first sorption pump. The H2 pressure in the 1.0 ~¢ volume is still - 5 x 10 -2 Torr. Now let us valve in the second, unused sorption pump. This sorption pump must pump the equivalent o f - 5 x 10" 2 Torr-Z of H 2. Again referring to Fig. 5.2.8, we see that the f'mal equilibrium pressure probably falls somewhere in the mid 10" s Torr range. We have learned two things from the exercise: 1) that sorption rough pumps, even if cooled to only LN 2 temperatures, very effectively pump H 2 ; 2) that we can obtain better ultimate H 2 pressure by the staging of two sorption pumps than could be achieved with the simultaneous pumping of 750 such pumps.
CRYOPUMPING
291
Effects of Neon When Rough Pumping We have seen the obvious benefits of staging pumps when pumping a single, difficult gas such as H 2. Similar benefits are gained when using these pumps to rough-pump vessels filled with air. For example, Fig. 5.2.9 shows pumpdown data taken by Turner for two conditions: the first gives total vessel pressure, as a function of time, when using two sorption pumps to simultaneously pump down a 100 Z vessel at an atmosphere of air; the second gives the total vessel pressure, in time, when using the sorption pumps to sequentially pump down the vessel under identical conditions. 10
1
o
[-,
1 s t PUMP VALVED OUT ~ - 2 n d PUMP VALVED IN
\,
0 10
.
I
-1 10
TWO PUMPS
=
SEQUENTIALLY
=
SIMULTANEOUSLY PUMPING
~r~
10 ["
o [...,
~TWO
-3 10 •
16 -4
" ' ~ - - - ~-~STAGED PUMPS
i
0
1 1 I
1 I
5
1 1
1
10
15
PUMPDOWN TIME -
20
25
Minutes
Figure 5.2.9. Pumpdown o f a 100 l i t e r v o l u m e , i n i t i a l l y a t atmosphere, with two sorption pumps. In o n e c a s e t h e p u m p s are sequentially staged, in the other, used simultaneously. (74)
The results of Fig. 5.2.9 are explained by the presence of Ne in the air. That is, the difference in the final pressures of the two pumpdown curves is due to the partial pressure of Ne in the air. We recall from Table 1.8.1 that the partial pressure of Ne in the atmosphere is --1.4 x 10" 2 Torr. The total pressure of Ne in the vessel which was simultaneously sorption pumped is less than its partial pressure in the atmosphere. This merely indicates that there was some cryotrapping of the Ne in the sorption pumps. The reason one is able to achieve better total pressures with staged pumping is modeled in Fig. 5.2.10. It is suggested that substantive quantities of Ne are pumped in the first of the two staged pumps as a consequence viscous drag effects. That is, the Neon is swept along with the O 2 and N 2 evacuated from the system. If the first pump is valved out of the system in sufficient time, (i.e., at time t, of Fig. 5.2.10) and the second pump valved in, the Ne is retained in the first pump. However, if one dawdles in valving out the first pump (e.g., to time t 2 of Fig 5.2.10), Ne will back diffuse from the first pump, and eventually contaminate the system with a Ne partial pressure equivalent to that found at atmospheric pressure. As a rule of thumb, one should attempt to valve out the first pump on achieving a roughing pressure of--0.1 Torr. Stern and DiPaolo noted the benefits of staging sorption pumps and referred to the above effect as entrainment pumping.( g g ) Three important lessons are learned by these results: 1) gas displacement effects probably exist when sorption roughing;
292
CRYOPUMPING
2) though the component of Ne is only -1.8 x 10 "s of the total atmospheric pressure, only -65% of this small Ne component could be removed from the system by cryotrapping; 3) the staged roughing of systems will also aid in the pumping of the more difficult gases. 10
10 1
10
-
•
o
I
1 ''
I
Ne ° / T
N~
H ~.~ H ~ , o.[A I ~ ~ ~--._~ i~o~. 0 FJoooet-I I t"]o°~ °o H I KI 0 c~t"l I TIME t
\
PUMP -INLET
0
10
~4 .
r~
<
-1
10
1
O--z
.
BASE -PRESSURE _
_
_
NITROGEN
-4
k ~ "--f- PUMPDOwN
\
-5
roughed
.
I
10
Figure
i jl
)
10
10
.
TIME t2
-- NITROGEN __~ \
I
I"
0
5.2.10. system
i
L1 Limitation as
a
I
Minutes
in observed
consequence
_
1
1
Le TIME -
Z
base
of N e o n
p r e s s u r e of s o r p t i o n backstreaming.(74,99)
5.2.2 D e w a r s a n d B a k e o u t R e g e n e r a t i o n H e a t e r s The artificial zeolites must be baked to remove sorbed water vapor. If this is not done, though other air gases are liberated on warming up the pump, H2 O will remain behind and plug the sieve material on subsequent use. This is also required on initially charging a pump with new sieve material. Because of this, some form of sorption pump bakeout provision must exist. This does not mean that each time a sorption pump is used it must be baked. On the contrary, in UHV system applications, they are often used a number of times between regenerations (bakeouts). This is totally dependent on the partial pressure of H 2 O in the gases being pumped. The regeneration process merely involves heating up the valved-off pump for an hour or so. Released gases escape through a pressure relief valve which appends all of these sorption pumps. It is not necessary to pump on the sorption pump during regeneration, though associated bulk roughing pumps are often valved in to the sorption pumps during the last ten minutes or so of bakeout. Figure 5.2.11 shows H 2 0 isotherms on the Linde 5A sieve material.( T 4 ,s o ) Four H 2 0 isotherms are shown in this figure. In this case, a reduction in the capacity of the sieve material for H 2 O at higher temperatures is interpreted as a measure of the thoroughness of the bakeout regeneration. Stern and DiPaolo have shown
CRYOPUMPING
293
that when the amount of adsorbed H 2 0 in the Linde 5A sieve material exceeds 7% by weigl~t, the capacity of the sieve material for N~ is almost completely suppressed.(2 9 ) This corresponds to sorption o f - 7 0 Torr-Z/g of sieve material. From this we conclude that bakeout temperatures in excess of 250* C are required to eliminate sufficient water vapor to regenerate the sieve material. In practice, bakeout at a temperature of--300* C proves adequate.(S o ) Also, we conclude that some form of bulk roughing might be helpful in the regeneration process. However, this does not prove to be essential. Metal bakeout jackets are provided by vendors who sell these pumps. They take the form of split sleeves which, as shown in Fig. 5.2.5, fit over the cylindrical pumps. The resistance of the heater is usually matched to available domestic voltages. For example, one need only plug in the power cable to a 110 VAC, 14, outlet and come back in an hour or so to a regenerated pump. These sleeved heaters may be immersed in LN ~ without damage. Therefore, they may be permanently installed on the sorption pumps. 180
,
I,,
"~I 150
,,,
,
I
l II
I
l II
I
y
l l{
I
/
I II
-
!
120 100 °C -~ ISOTHERM
o < >
,,,
25 °C [SOTHERN-~
O
I
I
90
/
ISOT~IERM 60
o
30
/as0°c ~//
:/
/
/
ISOTHERM \ /
-
1 II
-4
10
1() 3
10-2
10-1
10 0
101
10
10
EQUILIBRIUM PRESSURE - T o r r F i g u r e 5.2.11. A d s o r p t i o n isotherms of w a t e r v a p o r m o l e c u l a r s i e v e a s r e p o r t e d by B a n n o c k , ( 8 ° ) a n d
o n L i n d e 5A T u r n e r . (74)
Two types of dewars are commonly used. One is constructed of an inexpensive plastic foam, or polystyrene material. The second comprises a stn. stl. vacuum therm o s bottle. The plastic dewars will be damaged if not removed during bakeout. However, in low profile vacuum systems, the bottoms of the sorption pumps are often located so close to the floor that it is impossible to remove the dewars. In this case, the more expensive stn. stl. dewars are used and remain in place during pump regeneration.
5.2.3 Safety Considerations Of course, there are certain inherent hazards associated with the use of liquid cryogens such as LN2. Many establishments wisely require the use of protective goggles, etc., when handling and using these cryogens. Above all, pressure relief provisions
294
CRYOPUMPING
must exist when using these pumps. Pressure relief provisions which are an integral part of a purchased pump should not be altered or tampered with in any manner. As shown in Figures 5.2.3 and 5.2.4, pressure relief valves on commercially sold pumps are always an integral part of the pump vacuum envelope. This is by design so as to make it impossible for the pump to be installed on a system without the pressure relief provision. The importance of these pressure relief valves is related in the following potential horror story which fortunately had a happy ending. In days past, these pressure relief valves comprised rubber corks which were inserted into small tubes, in turn welded to the flange or neck of the sorption pump (e.g., see Figures 5.2.3 and 5.2.4). When the pressure built up in the pump as the result of regeneration, the cork would pop out of the small tube and the gas therein, presumably not toxic, harmlessly vent into the laboratory. The corks were loosely attached to the neck of the pump with some sort of light chain or cable. Of course, in time the chain would break or be somehow lost and the user would find himself on his hand and knees, in a dark corner of the lab, looking for the missing cork which had been launched across the room during the last regeneration. Also, in the quiet hours of the night, as you puttered with some other work, a cork popping out of the vent tube could be very startling. Because of this, a Viton ® sleeve vent valve was invented as a substitute for the cork and the cork was eliminated (see Fig. 5.2.3). Many production facilities making use of these sorption pumps have provisions for automatic regeneration and LN2 dewar filling. One such company used three, triple-length sorption pumps (i.e., each charged with --4.2 kg of sieve material), arrayed on a manifold similar in configuration to that shown in Fig. 5.2.5. In scheduled preventive maintenance programs, they typically changed the sieve material in their pumps each year, while at the same time replacing all of the elastomer o-rings in the pump valves. On one occasion, having done this maintenance, they activated the automatic bakeout and post LN 2 filling provisions. Some time later, but prior to the staged roughing of the main chamber, the base pressure of each sorption pump was determined by opening the valve over that pump and measuring the pressure with a thermocouple gauge located in the main roughing manifold. Inexplicably, the base pressure of one of the pumps was > 1 Torr, where the other two evidenced the expected low base pressures. It was therefore concluded that perhaps the troublesome pump had not been given sufficient time to cool to LN ~ temperature. The two well-behaved pumps were used for the staged roughing of the main vacuum chamber and then valved out of the system. The operator then left the scene, with the automatic LN ~ dewar filling cycle still activated so as to further cool the troublesome pump. Several hours later the operator returned and found that the base pressure problem still existed in the third pump. From this he concluded that the pump bakeout cycle had not adequately regenerated the new sieve material in this third pump. He activated the bakeout regeneration cycle. Moments later, the bottom of the troublesome pump exploded. Fortunately, the explosion was contained by the stn. stl. dewar, but a fountain of sieve material and cryogen spewed out from the top of the dewar. Most fortunately, the operator wore safety glasses, and because of this was not seriously injured. The event was later duplicated under laboratory conditions. It was determined that: 1) in the process of the preventive maintenance work, the technician had mistakenly assembled the valve over the troublesome pump without a bonnet seal;
CRYOPUMPING
295
2) this pump had pumped on a leak at near atmospheric pressure for ---8 hours, while the dewar was automatically topped off with LN 2 ; 3) due to this, the pump had f'dled with a mixture of LN2 and LO2; 4) on turning on the bakeout heater, cold gases and condensed liquid from within the pump caused a contraction of the Viton ® vent sleeve when escaping between the sleeve and vent tube; 5) the flow of cryogen was so restricted by the ever-contracting Viton ® sleeve that the pump exploded at a pressure of -700 Nt/cm2. The solution to the problem was simple. It was determined, that when using a Viton ® cork, rather than the sleeve, under identical conditions, the cork would pop out at a pump pressure s 50 Nt/cm 2. The cork was restored, while keeping the sleeve venting provision.
5.3 Liquid Helium Cryopumps The creation and handling of liquid helium is a mix of both art and science. Excellent references exist on these two subjects, including a book by Scott,(a 1 ) and a second by Barron.(a 2 ) These two books are also excellent references on properties of materials for use in cryogenics applications. The only method of practically achieving temperatures s 10*K is through the use of LHe. Some of the early methods used to liquefy this gas will be briefly discussed in another section. For now, it is sufficient to note that all cryo-pumping experiments reported in the literature at temperatures s 10" K were done with LHe. A quick glance at Tables 5.1.3 and 5.1.5 indicates that a good portion of the data therein was obtained with LHe cryopumps. However, many of these pumps were of the nature of experimental apparatus rather than practical cryopumps. 5.3.1 C l a s s i f i c a t i o n of L H e C r y o p u m p s Before first classifying these pumps, I will describe some of their salient features. Figure 5.3.1 shows a cross section of a cryogen-fed cryopump for which Claude made patent application in 1968.(8 3 ) Boiling pool LHe cryopumps, defined below, had been used much earlier. However, this particular cryopump has some important general features. FIRST STAGE (2)
PUMP ~FLANGE
CHEVRON ARRAY - ~
RT PUMP BODY (1)
' c
?////or
LN 2 SSAGES
SECOND STAGE COATED WITH SIEVE MATERIAL
LHe
SECOND STAGE ARRAY (3) AT -~ LHe TEMP. Figure
5.3.1. Early cryogen-fed
RADIATION SHIELD AT LN2 TEMP. (4) eryopump
patented by Claude. (a3)
At first glance, it has a striking resemblance to features of many of the CLGHe cryopumps reported on in the literature (e.g., see Fig. 5.4.1). First, the pump has a
296
CRYOPUMPING
flanged, RT body (1). Afirst stage array (2), cooled with LN2, is used to shield the second stage array (3) which is maintained at LHe temperatures. The first stage array is a chewon-type structure which serves to prevent RT radiation, coming from the system to which the pump is appended, from shining directly on the LHe cooled surfaces. However, this chevron is somewhat transparent to the passage of gas. (It is the convention in cryopumping vernacular to refer to the warmest stage as the first stage and thereafter, in multi-stage cryopumps, to number successive stages in the order of their decreasing temperatures. This precedence stems from multi-stage gas cooling and refrigeration processes.) A third metal, bowl-like member (4) closely follows the contour of the pump body, but is thermally isolated from the body. It too is cooled with LN 2 and serves as a radiation shield between the RT pump body and the second stage. Being at the same temperature as the first stage chevron, it is also classified as part of the first stage of the cryopump. The second stage, suspended by wire cables, is completely surrounded by surfaces which are cooled to LN 2 temperature. Using a parallel plane approximation and unity spectral emissivities for all surfaces, the equation is given in problem 8 of Chap. 4, we calculate that thermal radiation on the second stage as a consequence of the interposition of the first stage is reduced by ---x210. The second stage of Claude's pump was coated with a sintered Ni powder sieve material to facilitate cryosorption pumping of certain gases. The cryogens were force fed through tubes attached to the arrays. In Claude's case, they were LHe and LN 2. They could just as well have been cold gases, or one cryogen a cold gas and the second a liquid, etc.(6 ~ , 8 a )
!ii 20 °K ~ STAGE-2 CHEVRON
=
f
77°K STAGE-1 - N CHEVRON
PUMP -~FLANGE ~
LN z BOILING POOL LHe ~rBOIL-OFF
LHe BOILING POOL 4.2 °K STAGE-3
\
"-4 _-4
LINDE 5A SIEVE COATING
F i g u r e 5.3.2. T h r e e - s t a g e E x c a l i b u r CVR-100 b o i l i n g pool a n d c r y o g e n - f e d c r y o p u m p .
Had the arrays themselves been the containers of pools of liquid cryogen, this pump would have been classified as a LHe boiling pool cryopump. An example of a combined boilingpool and cryogen-fed cryopump is shown in Fig. 5.3.2. Variations of this cryopurnp, once manufactured by the Excalibur Corporation, were ~eported on in numerous cryo-pumping articles.(2 s, 2 6, a a, 4 a, 6 2,8 a, 6 s, a s, a 6
CRYOPUMPING
297
The pump shown in Fig. 5.3.2 is actually a three-stage cryopump. Two stages are boitingpoot stages (i.e., the first (1) and the third (3) stages) and the middle stage (2) a cryogen-fed stage. In this example, an intermediate 20* K stage (2) is interposed between the first LN2 stage and third LHe stage. The gaseous He, which boils off the pool in the third stage, passes through tubing attached to the chewon comprising the second stage, thereby cooling this stage to --20* K. The third stage was coated with sieve material as described in a previous section. This three-stage cryopump evidenced thermal instabilities under certain operating conditions. When operating at very high pressures, gaseous He, boiling off the LHe pool in the third stage, super-cooled the second stage to near that of LHe. Therefore, gases which would normally not condense on this stage at 20*K (e.g., H2 ) did so at temperatures near 4.2* K. However, on reducing the influx of gas being pumped, the boil-off from the third stage would decrease and the second stage increase in temperature. This caused gases condensed thereon to be liberated, which raised the pressure, etc. This phenomenon, reported on by Fisher,(6 a ) was at times evidenced as cyclical pressure instabilities, pressure run-away,(6 s ) and longterm pumpdown subsequent to high pressure operation. You will note in a following section that anomalous long-term pumpdowns under some circumstances are also evidenced by CLGHe cryopumps after high pressure operations. This is sometimes referred to as apressure clamping effect. Improved variations of three-stage LHe cryopumps have been reported on in the literature. For example, Hseuh and Worwetz reported on a three-stage cryopump which had three boiling pool stages.(6 1 ) The second and third stages were LHe stages and the first stage was a LN 2 stage. A schematic representation of this pump is shown in Fig. 5.3.3. REGENERATION PORT LINDE . . _ 7 . ~ ~ "BAYONETS" [
-~ '
LHe 3rd STAGE
'!" .'--',(,'((((((~(,'(((,'(,'~ ': ~' " '
!
LHe 2nd -~ ! STAGE ~ - - - - - - ~ LN2
L
POOL (1st STAGE)
LHe ' ~"
r ' '
COCONUT C t t A R C O A L _ ~ ~ , ~ (EPOXIED) ' Figure
5.3.3.
[ ~~'~
-- . ~:ik(((((((((((((((qC'g(((~((((((((((((((((((((((((([ ~ ~
Three-stage
~400nam~ 1 cryopump
--~ .,~ ' of H s e u h
and
Worwetz. (61)
The intermediate boiling pool stage served three purposes. First, it avoided the pressure instability problems associated with the pump shown in Fig. 5.3.2. Secondly, the third stage was coated with sieve material. This material was needed in order to pump substantive quantities of He. By placing an optically dense second stage array, cooled to LHe temperatures, in front of the third stage, H2 and its isotopes would be cryocondensed on the second stage. This served to prevent plugging of the sieve material by H 2 and reserved the sieve material for the pumping
298
CRYOPUMPING
of He. Thirdly, Chubb, et al., had reported that stray IR would cause desorption of H2 on LHe surfaces.(a z ) The LHe array interposed in front of the final stage precluded all possibility of this occurring on the third stage. The H2 possibly desorbed from the second stage would be pumped by the third stage. As we will see, Benvenuti, et al., showed that under certain conditions such intermediate IR shielding stages were not required in LHe cryocondensation cryopumps (i.e., those not using sieve materials).( a g-9 2 ) A third type of cryopump is the linear-arrayed pump. Pumps of this variety are used primarily in large space simulation chambers and fusion-related apparatus. Hood orovides an excellent review of the design and applications of these pumps.(9 z ) An example of the configuration of these pumps is shown in Fig. 5.3.4. They are usually two-stage, cryogen-fed pumps. Many arrays such as shown in Fig. 5.3.4 comprise extruded A~ panels, with integral cryogen piping. Again, the cryogen-fed LHe cryopanels are optically shielded from the RT walls of the chamber and equipment within the chamber by cryogen-fed LN 2 panels. Such pumps are used in linear module configurations in the pumping of neutral beam fines in fusion applications.(~ 4) Coupland describes the design and performance of some of these pumps.( ~ o, 9 s ) (Also see Duffy and Oddon,(9 6 ) and Thibault, et al.(S t )) RT V A C U U M R T VACU UM - - - - ~ / / / / / / / / / / / l WALL WALL
END VIEW OF "DOUBLE CHEVRON"
Figure 5.3.4.
)~/// //-/-/
/~/ -J--/ f//- ~/,,/I
RT VACUUM ffALL
END VIEW OF "Z-ARRAY"
Two linear eryopanels described by Hood. (93)
In fusion-related applications it is necessa.ry that these pumps be able to pump He and the isotopes of hydrogen.( s t ,6 4,9 7 ) It was noted in section 5.1.7 that sieve materials at LHe temperatures can become plugged by the isotopes of hydrogen and become ineffective in the pumping of He. Because of this plugging problem some have resorted to the use of the concurrent cryotrapping of He, on 4.2* K surfaces, with Ar. Batzer, et al.,(4 6 ) describe such a system. The cryotrapping scheme, as further described by Sedgley, is shown in Fig. 5.3.5.(6 4 ) As noted, the Ar cryotrapping of He and H 2 on 4.2* K surfaces was reported on much earlier.(4 s ) In the above case, jet streams of Ar were directed onto 4.2* K cryopanels. Helium pumping speeds of the order of---1.0 Z/sec-cm 2 were achieved in so doing. The speed of a similar panel coated with coconut charcoal produced speed of the order of ---5.0Z/sec-cm 2 .(6 4 )
CRYOPUMPING
Ar JET SPRAY ~ ~ -
LHe
C R Y O P A N E L Y/////////////////////////////
299
LNe /S-CRYOPANELS
/
Figure 5.3.5. Cryopanel c o n f i g u r a t i o n u s e d by Sedgley, et al., for t h e Ar c r y o t r a p p i n g of He.(TM
~L
RT VACUUM WALL
From the above we are now able to lend some organization to our classification of LHe cryopumps and place them in categories as follows:
1. cryocondensation , 2. cryosorption, 3. cryogen-fed, 4. boiling pool, 5. arrayed and lumped (i.e., large appendage pumps), 6. one-stage, two-stage or three-stage, and 7. combinations of the above. 5.32 Chevron Design It has been noted that in multi-staged cryopumps some form of higher temperature cryopanel is often used to shield the lower temperature stages or cryopanels. This serves in some cases to protect sieve materials from plugging effects. In other cases, the intermediate cryopanel is used primarily to shield the lower temperature panel from thermal radiation. In this latter instance it is merely an issue of economics. Though the gas boil-off from LHe cryo-pumps is always captured and recycled, the very cost of liquefying this gas (i.e., the energy required) is x25 to ×30 greater than that of producing, for example, LN 2. Therefore, LHe cryopumps make use of LN 2 cooled louvers or chevrons as radiation shields for LHe cryopanels. Similar arguments exist for CLGHe cryopumps; that is, it is much more difficult to produce 50 W of refrigeration at 10*K than 50 W at 77* K. Therefore, louvers or chevrons attached to the first stages of these pumps are used to shield the second stages of CLGHe cryopumps. A perfect shielding louver or chevron would be one where no thermal radiation is able to pass through and impinge on the colder stage of the cryopump. However, excluding what are called the highly condensible gases (e.g., H 2 O), most of the gases must be pumped on the colder stages of cryopumps. If no radiation can possibly pass by the shielding louver or chevron, molecules will also be impeded from being pumped on the colder stages. Therefore, design of the shielding louvers or chevrons is a compromise between acceptable radiation loads on the colder stage and the conductance of the shielding stage for the passage of gas. The gas conductance of shielding cryopanels is often called its molecular transmissivity. Similarly, the measure of thermal radiation or light which can pass the shielding cryopanel is called the thermal or optical transmissivity. The molecular or thermal transmissivity of a cryopanel is merely the ratio of the passage of particles or light through the chevron relative to that passing through a comparably sized aperture.
300
CRYOPUMPING
There are three basic types of optically dense cryopanels. These are shown in Fig. 5.3.6. The first is the louvered cryopanel, the second the chev-roned cryopanel and the third the z-chevroned cryopanel. The louvered cryopanel, nestled behind the louvered cryoshield (i.e., in the upper left corner of the figure) was first proposed by Santler.(1 x o ) Because of this, it is sometimes referred to as the Santler chevron. I am not sure of the origin of the remaining configurations. A comprehensive study of the transmissivities of louvers, chevrons and other labyrinth configurations was reported on by Levenson, MiUeron and Davis.(1 1 1 ) They used Monte Carlo type calculations to predict the theoretical passage of gas through the optically dense structures, and then empirically verified their findings with vacuum measurements. Monte Carlo calculations involve the use of a computer to calculate, on average, the random flow of individual molecules from one place to another in a vacuum system. It is assumed in these calculations that the molecules, when momentarily sticking to some surface along the way, come back off the surface according to the cosine law (i.e., see Chap. 3). LHe or LH2 LOUVERS S CHEVRON "SANTLER CHEVRON"
//
LOUVER
Figure 5.3.6.
CHEVRON
Z-CHEVRON
End view of eryopanel louvers and chevrons.
Def'me the spacing between the chevrons or louvers shown in Fig. 5.3.6 as W, and their lengths as L. Levenson, et al., determined that the molecular transmissivity of the louver, with zero overlap, L >> W, and an angle O = 60", converges to -0.5. This finding is directly applicable to cryopumps such as shown in Figs. 5.3.1 and 5.4.1. Also, for the conventional chevrons, with L >> W, and for angles cz = 60", 90* and 120", the molecular transmissivities converged to -0.19, --0.25 and --0.28, respectively. H](drogen desorption by IR (i.e., infrared) radiation, first reported by Chubb, et al.,(a 7) was particularly troublesome when attempting to cryocondense H 2 on LHe cooled surfaces. There was an apparent departure of the hydrogen vapor pressure from the Clausius-Clapeyron equation. This was noted by Lee,(1 2 ) and Benvenuti and Calder.( 9 o ) Evidently, this departure stemmed from IR radiation impinging onto the cryopanel on which the H 2 was pumped. Benvenuti, et al., determined that there was a photon (i.e., fight) energy threshold of --45 p m below which the stray IR passing through a shielding chevron would not desorb cryocondensed H2 .(9 2 ) They postulated that the IR photons created phonons in the metal substrate which in turn bumped off cryocondensed H 2. They arrived at this conclusion by noting that if the H 2 was first cryocondensed on a cryofrost of a more readily cryocondensed gas, the H 2 pressure, on subsequently cryocondensing on the former gas, more closely fit the predicted Clausius-Clapeyron equation. From this they concluded that the predeposited cryofrost established the equivalent of a phonon cushion between the
CRYOPUMPING
301
metal surface and the cryocondensed H2. More importantly, they determined just how sensitive the cryocondensed H2 was to possible desorption by low energy photons. As an aside comment, this effect may have very important implications in both proton and heavy ion coUiders. That is, where synchrotron radiation may not be of sufficient energy to desorb photoelectrons, which in turn desorb gas, phonons or plasmons created by very weak synchrotron radiation may cause desorption of surface gases and pose beam scattering problems. Because of their f'mdings, Benvenuti, et al., made refined calculations and measurements of the molecular and light transmissivities of chevrons of various geometries. Their initial conclusion was that some form of z-chevron would be needed to properly shield the surface on which H 2 was cryocondensed.(9 1 ) These z-chevrons have molecular transmissivities of 50-75% of the conventional chevrons, and because of this are less desirable for high pumping speeds. However, in subsequent studies, they determined that by blackening the fh-st stage conventional chewon with a low reflectivity material, and silver plating the H2 cryocondensing surface, they were able to achieve high molecular transmissivities with very low light transmissivities. Results of their findings of molecular transmissivities are summarized in Fig. 5.2.7.(a 9 ) Depending on the surface photon reflectivity, they were able to achieve photon transmissivities of the order of -1-5 x 10-4 for chevrons with ct = 90* and 120". 0.25 I
I
I
[-.. oq
o~
0.20 0.15
~q
z < CHEVRON TRANSMISSIVITIES vs. D / h , THE ANGLE ? AND WITH P - ZERO. D/h ~
2.5
5.0
10.0
20.0
40.0
¢x:'= 90 °
0.174
0.217
0.231
0za5
0za8
°C=120 °
0.222
0.258
0.269
0.274
0.275
Figure 5.3.7. as reported
/
0.10 0.05
> gr3
0.00
/J D -6.8 h
J o
Chevron molecular transmissivities by Benvenuti, Blechschmidt and
I
40 oc-
P -0.1 W I l l l 80 120
160
DEGREES
vs. geometry Passardi. (ag)
5.4 Closed-loop, Gaseous Helium Cryopumps (CLGHe) I will fhrst discuss the design and operation of these cryopumps, then some of the history behind the development of the refrigerators, and then discuss some of the design and operational features of these CLGHe refrigerators. Closed-loop, gaseous helium cryopumps (cryopumps hereafter) are very simple devices. The key elements are the refrigerator and, of secondary importance, the cryopanels and pump body. This order of importance stems from the relative cost and complexity of the refrigerators, compared to that of the cryopanels and pump body. The cryopanels comprise simple, fabricated metal components. Their design merely requires that a little thought be given to considerations discussed in the previous sections. The cross section of a typical cryopump is shown in Fig. 5.4.1.
302
CRYOPUMPING
PUMP~ FLANGE " ~ l 2 n
H,
~
end STAGE
COOLING STATIONJ 10K to 20K
~
1st STAGE ARRAY & INLET ARRAY (4)
~.+de._ilki2=,q\\,x.~A ~1£/ " "=1 = ' ` "" ~
~
~ ~
/ ~ ~ \
~--~~
II, II
(8)
MOTOR DRIVE ~ POWER INPUT
'
ilPl
~ '
~
~
N %. _'ill '_ ~ 1st S T A G E " ~ " ~ ...... COOLING STATION - - ~ ~ ~ 50K to 80K ~] ' i , I' [d~ III SCREEN ~ PRESSURE ~ RELIEF VALVE--~
'1t
l][~
~
[J~ RT P U M ~ I fj - BODY (7)
k,.__
~
~
COMPRESSOR
( Lt~-c~BE~
]~J~ [ .------li~ 7 ~ ~ ~ HIGH PRESSURE ~ ~ ~ . ~ . . , NEAR RT He _
HELIUM
~ ~
SORBENT
H"- MATERIAL H (6)
/ /
~ HELIUM EXPANDER
\
METAL HOSE (3)
(1)
Figure 5.4.1. Cross section of a typical t w o - s t a g e , c l o s e d - l o o p g a s e o u s h e l i u m c r y o p u m p with Gifford-McMahon r e f r i g e r a t o r . The refrigerator comprises an expander (1), to which the arrays (4,5) are attached, and a high pressure He compressor (2). For the present, assume that the refrigerator is merely a device which provides refrigeration capacity at two stages. The term refdgeration capacity means that for a given, simultaneous input power to the two stages, the refrigerator will sustain a given temperature at the respective stages. In the absence of cryopanels, the refrigeration capacity might be measured by inserting the expander into a vacuum dewar. Inputting power to the two stages could be accomplished by passing current through electrical resistors mounted to the two cooling stations. This thermal loading with resistors is often used as a final refrigeration inspection test by manufacturers. From convention, the capacity of the first (i.e., warmest) stage is measured at 77* K, and the second stage at 20* K. CLGHe refrigerators are typically capable of achieving temperatures as low as --9 to 10" K at the second stage and -30* K at the first stage. As will be shown, achieving too low a first stage temperature can sometimes cause problems, and is usually not desirable. Work has recently been reported on the use of two-stage expanders which, through the use of rare earth materials in the r e g e n e r a t o r beds, can achieve second stage t e m p e r a t u r e s of 3.6-4.0* K.(1 e 7,1 e 8 ) Kuriyama, et al., first reported on a two stage refrigerator in 1993 which, through use of a mix of rare earth materials in the second stage,, could • • produce 1 W of refrigeration at a second stage temperature of 4.2 O K. t 1 ~ o) These refrigerators have a finite capacity. For example, the second stage refrigeration capacity of the typical --20 cm ~ cryopump ranges from 2.0 to 4 W, with the simultaneous loading of 10 to 25 W at the first stage. Similarly, the capacities of --30 cm ~ cryopumps range from 3 to 5 W at the second stage and 40 to 60 W at the first stage. In the vernacular, if a refrigerator is a "4/25" it has the refrigeration capacity to simultaneously absorb 4 W on the second stage and 25 W on first stage while sustaining temperatures of 20* K and 77* K respectively, etc. However, when several cryopump manufacturers and I established the AVS Recommended Practice for measuring the performance and characteristics of cryopumps, we avoided specifically classifying refrigerators based on their refrigeration capacity.(1 s 9 ) This
CRYOPUMPING
303
proved not to be necessary. Chevron
Design
The first and second stage arrays are attached to the respective cooling stations with threaded bolts. Thin sheets of indium foil (-0.25 mm) are sandwiched between the cooling stations and arrays to provide good thermal intimacy between the two surfaces. This indium foil retains its ductility at very low temperatures, thus preventing a thermal jump condition from existing between the arrays and cooling stations. 3000
i IIi
1
H20
2000
I
!1
I
/
I I1
/ ~
f
/
I
1 11
I
1 II
VARIATION IN SPEED WITH PRESSURE DUE TO LOOSE 1st STAGE ARRAY
1000 )4--
Ar
/REDUCED H2 S P E E D ~ DUE TO Ar* PLUGGING 0
1 I II
1 I 111
10 -7
10-6
10-5
10-4
10 -3
10-2
PUMP PRESSURE - Torr Figure 5.4.2. Pump H20 and He speed anomalies s t e m m i n g from a loose first stage chevron and Ar*(3000 Torr-Z) sieve plugging. If the arrays are not securely fastened to their respective cooling stations, anomalous pumping effects will occur. For example, Fig. 5.4.2 shows pumping speed data of a 20 cm (b cryopump taken on a modified CERN dome (see section 1.16.4 of Chap. 1). The speed for H2 O was observed to be inexplicably low. Though measuring the speed of a pump for H 2 O is a very difficult undertaking, we knew that results were steady-state and the effect was real. We subsequently discovered that the first stage inlet array was loosely attached to the first stage array. As a consequence, at pressures <4 x 10-8 Torr, most of the water vapor was pumped on the second stage of the cryopump. Above this pressure, water vapor cryocondensed on the first stage inlet array. Recalling the vapor pressure data of Figures 1.7.3 and 1.7.4, we note that, with the exception of He, H2 and Ne, all of the gases would be readily cryocondensed on the second stage of a cryopump, operating at temperatures of < 20* K. The gases He, H 2 and Ne can only be effectively cryosorption pumped at these temperatures. Because of this, activated coconut charcoal is epoxied to the underneath surface of the second stage array (item (5) of Fig. 5.4.1). The purpose of hiding the charcoal in
304
CRYOPUMPING
this location is to protect it from plugging effects of those gases which pass through the first stage chewon and are cryocondensed on the second stage array. The perfect configuration for hiding the charcoal from potentially plugging gases is one in which no gas has access to the sieve. Of course, this would also preclude the pumping of He, H2 and Ne. Therefore, protecting the sieve material from the plugging gases, and yet making it accessible to gases which must be cryosorption pumped, is a compromise. A good rule of thumb in the design of a cryopump is that on removing the first stage array you should not be able to see the coconut charcoal from any viewing angle. This rule is often broken, and leads to measurement data which can be misleading, to say nothing about limiting the capacity of the cryopumps for He, H 2 and Ne. For example, I once observed a high partial pressure of H 2 in the pumpdown of a system. This high H2 partial pressure was not, as some thought, due to outgassing of H2 from the walls of the system. Rather, it stemmed from the dissociation of H2 O, into O + H 2, in a quadrupole residual gas analyzer. The high H 2 pressures existed because the cryopump could not pump the H2. This, in turn, stemmed from an improper second stage array design resulting in excessive exposure of the sieve material, and its consequential plugging during the pumpdown. If for example, the data sheet of a 250 mm q~ cryopump claims an H 2 speed 5-6,000 Z/s, and yet the sieve material is visible from the pump inlet, the actual H2 speed subsequent to pumping substantive quantities of cryocondensable gases might be in the range of 3-4,000 ~/s. Of course, if one plans to exclusively use a cryopump for the cryosorption pumping of He, H2 and Ne, then the above rule of thumb would not apply. The best measure of the design effectiveness in shielding the sieve is the measure of the change in speed of a cryosorption-pumped gas as a function of the amount of a second gas which has been cryocondensation pumped. For example, Ar is cryocondensed on the second stage of a cryopump, where H 2 must be cryosorption pumped. If we intermittently pump one gas and then the other, we should have a good measure of the effectiveness with which the sieve material is being protected for the sorption pumping of H 2.( 1 s g ) I conducted such a test on a cryopump in the late 1970s. The second stage array is shown in Fig. 5.4.3. It had the advantage of being very directional to the pumping of cryocondensables, but the sieve material was fairly open to the pumping of He and H2 .(1 1 e ) The sieve material was bonded to the underneath side of 36 metal leaves, which were in turn individually attached to a fined brazement. One idea was that if the sieve material was inadvertently plugged by oil, the leaves could be discarded and replaced. Secondly, the layered configuration provided inherent protection from cryocondensing gases. This array afforded an extended surface area (i.e., -1500 cm 2 ) of bonded activated coconut charcoal. We first quantified the speed and capacity of this array for He and H2. These results are shown in Fig. 5.4.4. Speed data and capacity data shown in Fig. 5.4.4 were taken on a modified CERN dome. The He and H2 pressures were held constant at 10"e Torr throughout the tests. We arbitrarily defined the capacity of the pump as that point at which the speed for the given gas had decreased to one e-fold of the initial speed. This was totally arbitrary. That is, we might have held the pressure constant at 10"8 Torr or 10-s Torr, etc., and similarly defined capacity. For very low initial throughputs, the H2 speed of this pump was --4700 Z/sec. However, as has been noted, the cryos-
CRYOPUMPING
305
orption speed will vary markedly with throughput (e.g., see Halama, et al.,(2 s, ~ s ) Hseuh, et al.(3 8 )). After pumping a few Torr-Z of H ~, the speed of the pump for this gas leveled off at --4,000 Z/s. We note that the capacity of the cryopump for He, as herein defined, is --8% of that of H 2. However, the initial He speed of the cryopump at 10 -e Torr was --50% of the H2 speed. Again, I emphasize the importance of specifying the pressure at which the speed and capacity measurements are taken when making claims about these values. ~--1~60
/."
/'//_
"<",~/'0.."~L 3//7
//~;/ / / //
Figure
I COPPER m m ¢I--~ ¢ / f B R A Z E M E N T ' N
[ ] ,
' "
// /
A second
5.4.3.
~._\\ \ ~ ~ -"-%~,\q J'.O'~ "/ ~ \\ "< \\ / ~',,,/'~ stage
5000 4000 ~
array
with
i
~
properly
l
!
.:,
;
,~_HYDROGEN S P AT 10 - 6 T o r r
~-
0
shielded
l
charcoal
(116).
!
, --'--"~~-,,,
..-~ i
0
O F CHARCOAL BONDED UNDERNEATH
2rid STAGE TEMP. ~14°K
3000
1000
36 "LEAVES" W~tH
A TOTAL OF 1500 cmz
1
2
4
QUANTITY
PUMPED
1
6 -
I
8
Torr-Liters/10
3
t
10
F i g u r e 5 . 4 . 4 . R e l a t i v e s p e e d a n d c a p a c i t y of a c l o s e d l o o p , gaseous helium eryopump for the gases He and He, with ~1500 cm 2 epoxy-bonded coconut charcoal on second stage.
We then measured the speed of the same cryopump for H2 as a function of the amount of Ar which had been pumped. These data are shown in Fig. 5.4.5. The H 2 speed was measured at 10 -6 Torr. A total of --8 × l0 s Torr-Z of Ar was pumped before the speed of the cryopump for H 2 had decreased to -50% of its initial value. I believe that the initial, very pronounced, drop in H2 speed, with Ar pumped, stemmed from Ar plugging of the lower-most leaves in the second stage array. These lower leaves account for --17% of the H2 capacity, but because they are so
306
CRYOPUMPING
much more accessible to both Ar and H2, they accounted for -25% of the initial H2 speed. These lower leaves form the equivalent of an inverted cup, the inside of which is coated with activated coconut charcoal. The almost linear nature of the subsequent decrease H2 speed, with the pumping of Ar, suggested that other mechanisms were at play. For example, it is more probable that for the greater part, the decrease in H2 speed stemmed from the physical blockage of access of H2 to the sieve material by thick layers of cryocondensed Ar ice. This method of measuring the degree speed decrease for one gas usa consequence of sieve plugging b~,a second gas was adopted in the AVS Recommended Practice on cryopumps.~. 1 s 9 ) 4000
?
¢9 ~9
HYDROGEN SPEED
Y
3000 Q) .,.~ ,-3
o 2000 L 7
0
2
4
ARGON P U M P E D
6 -
8
10
Torr-Liters/105
Figure 5.4.5. H y d r o g e n s p e e d of CLGHe c r y o p u m p as a function of t h e a m o u n t of a r g o n p u m p e d ; the sieve "plugging" effect.
Sticking Coefficients We now have the basic tools to calculate the performance of a cryopump based on our knowledge of the type of first stage chevron or louver used, the surface area of the second stage and the sticking coefficient of gases on this sta~;e. The sticking coefficients for several gases, as reported by Dawson and Haygood,( 1 8 ) are given in Table 5.4.1. Table 5.4.1. gases as
The sticking coefficients of some of the common a function of gas and surface temperature. (18)
CRYOSURFACE TEMP. °K
_
10 12.5 15 17.5 20 22.5 25 77
77 °K 1.0 0.99 0.96 0.90 0.84 0.80
0.79
GAS AND GAS TEMPERATURE NZ C02 CO 02 Ar 300 OK 77 °K 300 °K 7 7 ° K 3 0 0 ° K 77 °K 3 0 0 ° K i 9 5 ° K 3 0 0 ° K 0.65 1.0 0.75 1.0 0.68 1.0 0.90 0.63 0.98 0.70 1.0 0.68 1.0 0.85 0.62 0.96 0.67 0.90 0.67 1.0 0.85 0.81 0.66 0.61 0.92 0.65 1.0 0.86 1.0 0.85 0.80 0.66 0.60 0.90 0.63 1.0 0.85 0.87 0.63 0.60 0.79 0.66 1.0 0.85 0.85 0.63 0.60 0.79 0.66 1.0 0.85 0.85 0.63
With the exception of CO, they noted that the sticking coefficients of the above gases, at gas temperatures of 400* K, were -0.5 for the above surface temperatures. For CO at 400* K, a -0.73. Also, they reported a sticking coefficient of 300*K H 2 O, on a 77* K surface, to be 0.92. From these data, our knowledge of molecular
CRYOPUMPING
307
transmissivities through louvers, and with the use of (5.1.18) we are able to make fairly reasonable estimates of the speed of a two-stage, CLGHe cryopump for condensable gases. From the data of Fig. 5.4.4, we deduce that the speed per unit area of the activated coconut charcoal, to first order, is --3 Z/sec-cm 2 for H2 and -0.5 Z/sec-cm2 for He. This is the initial speed with zero loading of the sieve, a sieve temperature of 14" K, an H2 and He temperature of ---60* K, and at a pressure of 10"6 Torr.
Thermal Loading of Cryopumps There are three possible sources of refrigeration loading or heat input to the two stages: 1) thermal radiation; 2) gaseous convection; and 3) heat of sorption losses. Regarding radiation loading, the sizing of the cryopanels must be properly matched to the refrigeration capacity. Also, the mass of the first and second stage array will impact on the cool-down time of a cryopump. This is because of the specific heat of the materials used to construct the arrays. In the early days of cryopump development, many of us overlooked this aspect. On the one hand, massive arrays provide a measure of thermal inertia when pumping impulsive gas loads. On the other hand, this very inertia will cause protracted machine down-time (e.g., coaters and implanters) during cool-down subsequent to a regeneration cycle. Another compromise is made. First stage chevrons or louvers of many commercial cryopumps are often either electro-polished or brightly nickel plated. This gives the customer a good feeling, looks good, but is of little functional value. This is because the total spectral emissivity of the chevron approaches unity after pumping a few/.t m of H 2 O.(1 1 2-1 1 4 ) Of course, if your system has no water vapor ... On the other hand, brightly polished surfaces are of benefit on the outer surface of the first stage cryopanel (i.e., the bowl-like structure attached to the first stage cooling station shown in Fig. 5.4.1). Gases which would readily condense on the chevron of the first stage are quickly condensed as they pass between the pump wall and the first stage cryopanel. Because of this, the highly condensibles do not reach the surfaces deep within the gap between the pump body and first stage cryopanel. Brightly polished surfaces will serve to reduce radiation losses in this region. I learned this lesson the hard way when inventing a unique first stage cryopump array. The design featured a three-dimensional, first stage array that filled a solid angle of---37t steradians.(1 1 5 ) The idea was to make the projected surface area of the first stage chevron exceed that of a circle having an area equivalent to the 20 cm #, pump. This array, within which the second stage array resided, was housed in a bulged pump body. The first stage chevron was blackened. This configuration had decided molecular transmissivity advantages, but the first stage thermal radiation load was excessive, and required refrigeration capacity which could be better otherwise used. A little judgment must be used in the proper application of a cryopump. For example, while at Varian Associates, a customer once called me on the phone to discuss a problem he was having with one of our cryopumps. He exclaimed, "It won't pump." I was unable to solve the problem over the phone, so I suggested that he send the pump back to the factory. On disassembling the cryopump, we discovered that the indium foil, sandwiched between the second stage station and associated array, had melted. Indium melts at -156" C. From this we learned that the cus-
308
CRYOPUMPING
tomer had mounted the pump so that it had a direct view of the hot zone of a high temperature furnace. This was an obvious misapplication of cryopumping. Also, quartz lamps are sometimes used to preheat and outgas substrates prior to coating. Light from these lamps should not shine directly on the first stage chevrons. The thermal load due to operation at high pressures is doubly troublesome to cryopumps. First, there are the losses associated with the heat of sorption of the gases being pumped. Secondly, there is the thermal load stemming from gas convection losses. Another rule of thumb is that the thermal load stemming from cryocondensation pumping on the second stage amounts to -0.7 W per Torr-Z/sec. You can construct more precise values using the data of Table 5.1.2. Thermal loading on the first and second stages due to operation at high pressures causes the temperature of these stages to increase. These temperaturerefrigerator loading effects are quantified in a following section. Though many of the gases are readily cryocondensed at 10 - 20* K, slight increases in the temperature of the second stage will cause gases which are cryosorption pumped thereon to be liberated from the sieve material. Therefore, the base pressure of the system for a cryosorbed species may have a low value under conditions of moderate throughput, but increase markedly with only modest increases in the temperature of the second stage. -3 10
*6.2
-*Torr- Z/g
*1.9
at 293 °K
*0.075Z
;A ?2 _:
16 4 I
10
:~
*0.3
-
g8
M -7 10 10
12
11
13
14
15
16
17
ACTIVATED CHARCOAL TEMPERATURE Figure
5.4.6.
-4 lO o E~
I
-
Adsorption
I
isosteres
l/
of helium
l
on
l
PJ
/
/
°K
-
c h a r c o a l ! 102)"
coconut
y
*187.5-~ -5
I0
18
"15
~
°rr
l
I -
at 2 3
_
/
!
"37.5
_
2.5/
168 Z F.~ ~7 o 1 -8 lO 10 Figure
5.4.7.
~ o v
CORRECTED FOR
!
~0-.4"~~ I
I
12 14 16 18 20 ACTIVATED CHARCOAL TEMPERATURE Adsorption
isosteres
of
I
I 22
°K
I 24
(102)
h y d r o g e n on c o c o n u t charcoal.
CRYOPUMPING
309
For example, the adsorption isosteres for He and H2, as reported by Lessard, are given in Figs. 5.4.6 and 5.4.7, respectively.(1 o 2 ) These data have not been adjusted for thermomolecular effects (see problem 14). Assume that you have accumulated --150 Torr-Z of H2, per gram of charcoal, on the second stage, and the operating temperature of this stage is -14" K. From Fig. 5.4.7, we note that the H2 equilibrium pressure would be -10 -7 Torr. Note that the H2 speed at this pressure is zero. Now, assume that you introduce a steady flow of Ar at a rate of 5 Torr-Z/sec. With this throughput of Ar, we calculate, using the above rule of thumb, that the increase in thermal load on the second stage will be -3.5 W. It is reasonable to expect that the temperature of the second stage will increase by several degrees Kelvin as a consequence of the increase in thermal load, due to heat for sorption of Ar, to the second stage. Assume that it increases to -18" K. If this happens, the H 2 pressure will equilibrate at > 10-4 Torr. Therefore, as a consequence of pumping high throughputs of Ar, we have rendered the pump useless in the pumping of H 2 at pressures s 10" 4 Torr. For purposes of illustration, I have used an extreme case. It is possible that a condition of thermal run-away might occur as a consequence of the release of H2 in the above example. That is, the thermal conductivities of gases are proportional to m "~ , where m is the molecular weight of the gas. If we are able to stably operate at a given pressure of Ar, while hovering on the fringes of swamping the pump, thermal convective losses due to operation at the same pressure of H 2 will be -.x 4.5 greater. Many times, this causes problems in the starting of cryopumps subsequent to regeneration. More will be said about this.
Sputtering Applications Because of the problems associated with heat of sorption loading of the second stage at high operating pressures (e.g., sputtering pressures), some form of throttling device is often used to interpose a high impedance between the pump and gas source. Because of the pressure drop across this impedance, the required high sputtering pressure is thereby achieved in the chamber, while the pump operates at a much lower pressure. This is also done when using diffusion pumps at excessively high sputtering pressures. The common practice is to put an aperture plate directly over the diffusion pump flange. However, the LN2 trap is positioned above the aperture so as to afford high pumping speed for H 20. It is important to maintain high water vapor speed during sputtering applications. Similar schemes have been adapted to cryopump sputtering applications, both with and without the use of LN 2 traps.(X 1 7 , 1 1 8 ) One such configuration is shown in Fig. 5.4.8. The variable aperture is mounted directly to the first stage of the cryopump. During the initial phases of pumpdown, the aperture is adjusted to be full-open, thus affording high speeds for all gases during the pumpdown and component preheating phase. During the sputtering phase, the aperture is throttled back to limit the Ar speed delivered to the coating chamber. However, because of the cold surfaces of the variable aperture (i.e., -70* K), the speed for H 2 0 remains very high. The speed for H 2, a byproduct of all sputtering operations, will be -..x 4.5 that of the Ar speed in the sputtering chamber. This variable aperture has been extensively used on hundreds of sputtering machines.
310
CRYOPUMPING
Figure 5.4.8. Variable aperture used in sputtering
The invention shown in Fig. 5.4.8, though clever, proves to be more marketing than good engineering, as a futed aperture plate (i.e., see Fig. 5.4.9) would prove more than satisfactory in most applications. For example, assume that we have a have a "5M5" refrigerator attending a 250 mm 9 cryopump. Assume that the speed of the cryopump for argon is -2,000 21s and -4,000 21s for H2. We have an argon sputtering application where the process must occur at a pressure of 5x Torr. If we pumped directly on this chamber with the cryopump during the sputtering operation, the throughput into the pump would be 2,000 21s x 5 ~ 1 0 Torr - ~ = 10 Torr-21s. Using the rule of thumb for heat of sorption, this throughput would correspond to a -7 W thermal load on the second stage of the cryopump. But, if it's a "5135"refrigerator, the temperature of the second stage cooling station will be >20° K. This suggest we need some sort of throughput restriction between the pump and system. SECURED TO
FIXED OF ARGON (:ON DUCTANCE Figure 5.4.8A. Fixed aperture plate used in sputtering applications.
Assume that the intrinsic speed of the second stage array is -4,000 21s for argon and 5,500 21s for H2. Lets remove the inlet array and secure an aluminum futed aperture plate to the fust stage array of the cryopump. Let's size the holes in the plate so that the speed for argon will be 400 21s. Then, during sputtering under the same conditions, the heat of sorption load to the second stage will -1.4 W.
CRYOPUMPING
311
What about the speed for other gases? Well the speed for H 2 0 will be 8,750 Z/s and that for H 2 --1,350 Z/s. What about initial chamber pumpdown? Assume that the pump has a speed for air of --500 Z/s with the fixed aperture plate installed, that the chamber has a volume of 5,000 Z, and that we cross over from roughing at 50 mTorr. Excluding water vapor, the air pressure in the chamber, at 10, 50 and 100 seconds will be approx. 18 mTorr, 3.4 x 10-4 Torr, and 2.3 x 10"e Torr respectively. The message here is that the air gases are very quickly pumped out. But, as we all know, the water vapor partial pressure will take a long time to recede due to outgassing from the walls and tooling within the chamber. Some very successful manufactures of sputtering equipment handle the argon throughput problem by using throttle valves over the cryopumps. They fix the input rate of argon or other sputtering gas, and then control the sputtering pressure with the throttle valves. This has the disadvantage of throttling back pumping of all the system gases, including H 2 0 and H 2. If it is a baked UHV cluster tool, and excluding gases desorbed from the targets, this may not be a problem. Note that with the use of the fixed aperture plate solution one must control flow with fixed pumping to achieve the sputtering pressure, rather than pressure with throttle valves. Operation of cryopumps at very high pressures can lead to some very interesting effects. One such effect I have coined the pressure clamping effect. In the early days of cryopumping, I designed a cryopump specifically for sputtering applications. It featured the previously noted 37t steradian first stage array, housed in the center of a cylindrical LN2 reservoir. This was a three-stage cryopump. The first stage of the cryopump was an LN2-cooled cylinder, similar to that of Hseuh's pump shown in Fig. 5.3.3. The second stage was the 37t-array which comprised the first stage of the CLGHe cryopump, and the third stage was the 20* K stage of the CLGHe cryopump. A variable aperture was mounted on the LN2 stage.(1 1 7 ) When Ar sputtering, the pressure in the region of the first stage array was of the order of 0.01 Torr. Because the 37t-array was so effectively shielded by the LN2 stage, it operated at a temperature of--40* K. At this temperature and sputtering pressure, Ar cryocondensed onto the 37t-array. On turning off the Ar source, the temperature of the 37t-array equilibrated at -35* K. Therefore, the Ar ice, formed on the 37t-array during the sputtering operation, had an equilibrium vapor pressure of--10 -4 Torr. The pressure clamped near this value until all of the Ar ice had sublimed away from the 37t-array, and was pumped onto the 20*K stage. This took several hours to occur. The lesson learned here was that the first stage of the CLGHe cryopump was operating at too low a temperature. This resulted in an Ar pressure clamping effect. The fix was simple. We merely put a thermal shunt of the first stage of the CLGHe cryopump so that it always remained at temperatures ~60" K. However, only you and I know that I made an error in designing this pump. Actually, there were two problems. At the time, the 37t-array was mounted on a refrigerator of moderate first stage capacity. Because of this, cool-down times after regeneration were excessive (i.e.,--224 hours). Similar effects are noted when operating a cryopump at very high pressures. In some instances, the speed of the pump seems to mysteriously increase at higher pressures. This effect is sometimes erroneously attributed to increases in first stage louver transmissivities as a function of pressure.(1 1 a ) In fact, these effects usually stem from the start of cryocondensation on the first stage of the cryopump. This effect is shown in Fig. 5.4.9, where Ar speed of a 20 cm ¢, cryopump is given as a
312
CRYOPUMPING
function of pressure. The base pressure of the system was ---10-1 o Torr prior to the measurement. The measurements were taken on a modified CERN dome. The pump was appended with a 20 cm 4, ConFlat ® flange. The dome was baked to 250* C and the cryopump warmed to -85* C. The cryopump was turned off and the system pumped on with an auxiliary pump during this bakeout. I
1 11
[
I
I 11[
--
1 1I
-
1
1
_
.f~-~
~-J~
ARGON PUMPING -- ON SECOND STAGE 1200
1
I
1 I1
I
Ill
1
1 11
I 1 ARGON PUMPING ON FIRST STAGE
1600
1400
I II I
)
= q
I
1000
800 L 7
0 ~" I
1
I tJ
10-8
10 -7
1 11
10 -6
1
1 11
1
1 11
10 -5
1
10 -4
ARGON PUMP PRESSURE Figure
5.4.9.
Increase
in
argon
speed
I 11
due
to
10 -3
10 - 2
Torr first
stage
pumping.
The refrigeration capacity of the first stage of the cryopump used in tests shown in Fig. 5.4.9 was somewhat marginal. The clue that it was the pressure clamping effect was that even though the second stage cooled to <20" K, and therefore had high speeds for Ar, there was a protracted pumpdown after the tests. This was due to the gradual sublimation of Ar off the first stage. The reverse of this effect was noted by Lessard when pumping Ne on the second stage of a CLGHe cryopump.(1 o 2 ) That is, he initially observed an inexplicably high pressure when pumping Ne. However, with the gradual heating of the second stage to higher temperatures, there was a precipitous drop in the Ne pressure. This was due to the fact that at the initial lower temperatures, Ne cryocondensed on the second stage. However, when heating the second stage to slightly higher temperatures, the cryocondensed Ne came off the outer surface of the second stage and was cryosorption pumped by the sieve material.
CryopumpApplications Cryopumps are presently manufactured and sold by a number of companies. In the early days of CLGHe cryopumping, some companies offered cryopumps with reverse Stirling-cycle refrigerators.(1 0 6,1 2 0-1 2 2,1 3 9 ) These refrigerators had vibration problems, operated at high rpm, and expander seal wear (i.e., reliability) was impacted. Today, refrigerators of commercial CLGHe cryopumps are machines exclusively usin~ some variation of the Gifford-McMahon refrigeration cycle.(1 2 a ,1 2 4 ) The literature abounds in publications dealing with various applications of these latter cryopumps.(1 2 s- 1 4 2 ) Some believe that because of recent advances in two-stage pulse tube refrigerators, and there inherently low
CRYOPUMPING
313
vibration levels, these refrigerators will eventually replace the conventional GiffordMcMahon cycle machines in cryopump applications.t t 7 x ) The driving impetus for commercialization of the present cryopumps has been the semiconductor industry. It was recognized very early that hydrocarbon contamination caused problems with the properties of thin films.(: 3 ~ ) Some have argued with me that the primary reason for the use of cryopumps in this application was the expense associated with LN ~, required for the trapping of diffusion pumped systems. However, I can recall in the early 1950s, as a microwave tube technician at the General Electric Laboratory in Palo Alto, CA, servicing freon refrigerators used to cool diffusion pump traps. This technique was widely used throughout the General Electric facilities, having first been reported on by Leever in 1954.(x 4 s ) Cryopumps afford a dean, dry, and reliable means of pumping. Cryopumps of the above varieties are used on most functional (i.e., optical and magnetic media), decorative and semiconductor coaters presently offered throughout the world. They are also used on the beam lines and end stations of most ion implanters. Some of the above references (i.e., 1 2 s- 1 4 2 ) discuss the advantages and disadvantages of the use of cryo-pumps over other methods of pumping. Some also discuss the use of cryo-pumping in combination with pumps including sputterion pumps. For reasons to be discussed below, this practice is to be discouraged. The primary limitation to the base pressure achieved by these pumps is the outgassing stemming from the use of elastomer seals (e.g., the pump flange seal). As noted above, one can readily achieve pressures of < 10" 1 o Torr without the presence of these elastomers. This was also demonstrated by deRijke.(t s t ) Kikuchi, et al., used all-metal, Helicoflex Delta seals in lace of elastomers and achieved pressures < 10-t ~ Torr with a cryopump.(1 4 ,t )Phese all-metal seals require very low sealing force, and therefore, may often be used as a direct replacement for elastomer seals.(1 ,t s, 1 4 s ) At times, metal burst diaphragms are substituted for the cryopump pressure relief valves. This is done to avoid the outgassing from the elastomer seal used in these relief valves. If you decide to use a metal burst diaphragm, rather than building your own, I suggest you purchase a commercially available assembly. Also, have a fellow technician check your calculations on the possible consequences of the use of this diaphragm. You should assume that the entire system to which the cryopump is attached will at some time be pressurized to the limit of the burst diaphragm.
System Configuration Several aspects should be considered when designing a cryopumped system. These include functional as well as safety considerations. There are some main guidelines which should be followed when designing or specifying a cryopumped system. The configuration of a typical sputtering system is represented in Fig. 5.4.10. The configuration of your system need not be as complex as this system. It all depends on your application. The valves V 1 - V4 are the conventional pneumatic or manual roughing valves. Valves V s - V s are the smaller, combination vacuum and gas handling valves. Some form of rough pumping is always required when using cryopumps. First, it is required to pump the work chamber down prior to opening the main system valve. Cryopumps are tolerant of large gulps or impulsive gas loads. For example, a 20 cm q~ pump is typically able to accommodate an impulsive gas load of 10-50
314
CRYOPUMPING
Torr-Z with no difficulty. However, should this practice be carried to an extreme, some gases may be viscously swept into the bowels of the pump and conceivably cause partial plugging of the sieve material. Secondly, a rough pump is required during the regeneration of the cryopump (see Regeneration, below).
#2TC~
GAUGE
SYSTEM ARGON ROUGHINGI VENT INPUT [ PUMP I z~Vs ~ 7 /
BAKEOUT JACKET --
ROUGHING_ ~ VALVES ] PUMPING LINE
V2 ~1SIEVE ~ TRAP -~--, v3
lION
GAUGE
MAIN SYSTEM VALVE
FIXED PRESSURE RELIEF ~ l VALVE
#3 TC GAUGE- ~
l
IG A U G E
APERTURE PLATE
CRYOPUMP V4+
HELIUM COMPRESSOR
~7V5 "WARM"NITROGEN
A
PUMP PURGE
Figure 5.4.10. Schematic representation of a cryopumped batch sputtering system with pump regeneration provisions. It is always good vacuum practice to minimize the number of penetrations into the high vacuum system. For this reason, valves V4 and V s are the only penetrations into the cryopump. Also, an additional valve might be placed between the manifold at the point where valve V 7 is attached, and the work chamber. One can then make possible compromises on the quality of these vacuum/gas valves. Use of a Main Valve Over the Cryopump: A cryopump temporarily stores the gas which it has pumped. In the event of a power failure, these gases will be liberated from the first and second stages of the cryopump. This isn't a catastrophic or instantaneous event. Actually, the cryocondensed gases slowly sublime off the surfaces. The cryosorbed gases are more quickly liberated. The rate at which the gases are liberated depends on the thermal inertia of the arrays, and the heat input to the arrays by convection and radiation. The key is to contain these gases in a safe and ecologically responsible fashion. For example, the gases might comprise an explosive mixture (i.e., see problems 16-18). There are possible sources of ignition in most vacuum systems. Therefore, you don't want this mixture disgorged back into the system. This would eventually happen if there was a power failure and no main isolation valve existed between the work chamber and the cryopump. Lastly, the main valve should be a pneumatically actuated valve. You have no assurances that you will be present during a power failure. Also, the electronics of the valve actuator should be latching. If the power fails and the valve closes, on return of power, the valve should not automatically open. No Possible Sources of Ignition Below the Main Valve: I have received some criticism in certain areas as being an alarmist regarding this issue. I reason with these criticisms in this manner. Anyone who has worked in vacuum related industries longer than a few years has walked up to a vacuum system, which he thought was under vacuum, and turned on the ionization gauge. More than once I have mistakenly carried this practice to the extreme of attempting to degas a gauge when a system was at atmospheric pressure. The gauge filled up with a white, smokey
CRYOPUMPING
315
powder and was ruined. I've done this several times over the years. Perhaps I am more fallible than most. But, had there been an explosive mixture in the chamber to which the gauge was affixed, an explosion would have followed. While lecturing on cryopumping at a very large corporation in the late 1970s, I stressed the importance of not having possible sources of ignition below the main valve seal. I noted that one person in the audience was nodding his head in a manner which indicated he knew what I was talking about, to the point of having a personal experience in this area. Seeking a testimonial to this effect, I questioned him. He was dressed in a white smock, and was obviously a vacuum technician. He immediately became guarded in his response. His supervisor, dressed in a dark suit and tie, was seated several rows behind him. On pressing the issue, the technician noted that though they hadn't had an explosion of a cryopump, "... the pump came apart very fast as a consequence of gas combustion". For this reason, I stress that one should never locate an ionization gauge or other possible sources of ignition (including sputter-ion pumps) below the main valve seal of a cryopump. While at Varian Associates, in the late 1970s we conducted an experiment to determine the effect on a cryopump in which such an explosion had occurred. The pressure relief valve in the pump was adjusted to relieve at a pressure o f - 9 0 0 Torr. Using remotely actuated valves, we concocted a stoichiometric mixture of 900 Torr of 0 2 and H2 within a cryopump. The pump was housed within a sandbagged bunker constructed in the Varian parking lot. We remotely ignited the gas mixture. Indeed, the pump came apart very fast. We called this an explosion. The results of this test demonstrated to us that there was no practical method of constructing a cryopump which would be immune to the released energy of such an explosion. Secondly, it made us zealots regarding the avoidance of the use of possible sources of ignition below the main valve of any cryopump. Toxic Gases: In many instances, cryopumps are used to pump toxic gases, such as carrier gases containing dopants for ion implantation. For this reason, consideration should be given to the manner of disposing of these gases on regeneration of the cryopump. A closed, off-gas venting system is needed in such cases. Sustained Untrapped Pumping on a Cryopump: When using a mechanical pump to pump on a cryopump during some phase of the regeneration cycle, care should be taken that oil backstreaming into the cryopump is not occurring. Such oil backstreaming will most certainly cause sieve plugging.
Regeneration of Cryopumps Because of their finite capacity for certain gases, cryopumps must be regenerated from time to time. Regeneration simply involves the warming up of a cryopump so that cryopumped gases may be pumped away by some auxiliary pump. Some studies of regeneration procedures appear in the literature.(1 4 6,1 4 7,1 7 3,1 ~ 4 ) Regeneration is usually the first thought that comes to mind when cryopump pressures seem inordinately high. Except in high-throughput sputtering applications, regeneration is required on a very infrequent basis (see problem 20). More rules of thumb are applicable. Two forms of cryopump regeneration are used: i)full regeneration; and, ii) quick or partial regeneration. The former amounts to warming the cryopump arrays to above room temperature for some period, flushing the cryopump with nitrogen gas, pumping on the cryopump with some form of dry pump, and then chilling down
316
CRYOPUMPING
the arrays to their operating temperatures. A quick regeneration involves similar operations as the full regeneration except the arrays are not warmed to room temperature, and the expander motor remains on during the process. These two regeneration methods are illustrated in Figures 5.4.11 and 5.4.12. 350
I
300 Stage Array
Second I
250 200 150
[-~
I
t
100
•
Firs;rrSat?ge~
A 50 [,3 0
20
40
60
80
REGENERATION TIME -
100
120
Minutes
F i g u r e 5.4.11. "Full r e g e n e r a t i o n " of a 200 m m d i a m . , low p r o f i l e c r y o p u m p a f t e r p u m p i n g ~3.8 x 10 5 T o r r - L of a r g o n ( r e p r o d u c e d with p e r m i s s i o n of Genesis V a c u u m Technology). 160 140
~
First S:age Array
120 i00 E-~
80 60 40
20
S e c o n d Stage Array
/ / E3WARM-uF 0
I
t
COOL-DOWN
1
10 20 30 REGENERATION TIME -
I
~
I -~I
40 Minutes
50
Figure 5.4.12. " P a r t i a l r e g e n e r a t i o n " of a 300 m m d i a m . c r y o p u m p with ~3 x 105 T o r r - L of a r g o n ( r e p r o d u c e d with p e r m i s s i o n of Genesis V a c u u m Technology).
A quick or partial regeneration of a cryopump is only possible under very controlled circumstances, for if you warm the pump sufficiently to liberate some of the cryosorbed gases, and then cool the pump down, some of the cryocondensed gases may be liberated in the process. These gases will find their way to the sieve material and cause plugging problems. A partial regeneration is made possible by the use of heaters, attached to the cryopump arrays, which are used to control the temperatures of these arrays in a prescribed fashion. The true measure of the success of any
CRYOPUMPING
317
regeneration process is the time between the regenerations as dictated by increased pressures of cryosorbed gases. There is some suggestion that if the second stage array is kept warmer than the first stage array during warm-up and rough pumping, the effectiveness of the regeneration is improved. If the opposite proves the case the required time between regenerations becomes progressively shorter, suggesting a progressive plugging of the sieve material. On warming a cryopump during full regeneration, the charcoal on the second stage is the last surface in the cryopump to reach room temperature. Because of this, during regeneration water vapor originating from other surfaces within the cryopump will ultimately end up on the charcoal. During the regeneration process, you must allow sufficient time for this water vapor to be pumped away. Slightly warming the pump arrays will accelerate the process. Flushing the cryopump with a warm purge-gas accelerates the regeneration process. Dry argon or nitrogen are suitable for this purpose. The same valve through which the cryopump is rough pumped during regeneration is often used as the inlet for the purge gas. This has disadvantages, however, as the implication is that the purge gas is flushing out of the pump through the pressure relief valve. This can cause problems as charcoal dust may be carried over, and cause leaks in the pressure relief valve seat. Some form of screen or shield is still needed to make sure that activated coconut charcoal, which is known to fall off second stage arrays, does not cause valve sealing problems. Use of a stand-pipe arrangement will eliminate such problems with the roughing and gas inlet valve. Some form of screen is usually used to protect the pressure relief valve seat. Without a similar stand-pipe arrangement for the pressure relief valve, it is not advised that the purge gas be of sufficient pressure to cause release of this valve. This is because the flushing gas may carry debris to the valve seat and cause subsequent sealing problems. The warm purge-gas serves another purpose. This is to dilute the partial pressures of He or H2 which might be resident in the cryopump during regeneration. A partial pressure of 10"3 Torr of these gases may be sufficient to make it difficult to start the cryopump. Therefore, the purge-gas is helpful in assuring the dilution and viscous conveyance of these gases to the rough pump during regeneration. One may use the purge-gas to dilute He and H2 partial pressures, through the sequential venting and roughing of the cryopump over several cycles. Sales people often extol the fact that their cryopumps can be started at N2 or air pressures as high as 2 - 3 x 10-2 Torr. Indeed, this may be the case. However, when starting a cryopump at this pressure, speculate on where you think this gas is pumped. Yes, it is all pumped on the charcoal, and will partially plug the charcoal for the subsequent pumping of He and H2. For this reason, one should achieve the best possible base pressures prior to starting the cryopump. The roughing pump used during regeneration must be trapped or you will have problems with oil plugging of the sieve material. Sieve traps are often used for this purpose. Note in Fig. 5.4.10 that an isolation valve is located on both sides of the sieve trap. This is done for a very specific reason. Mechanical roughing pumps back-stream oil at operating pressures ~ 0.1 Torr. Assume that the sieve trap shown in Fig. 5.4.10 is regenerated by a high temperature bakeout. Of course, during this process valve V2 is kept open and valve V3, closed. If after this regeneration process, the roughing pump is allowed to pump on the sieve trap for an extended period, and at pressures near its blank-off pressure, the sieve trap will become
318
CRYOPUMPING
loaded with oil. It will then be more of a source than remedy to the back-streaming of oil into the cryopump on its the subsequent regeneration. New developments in dry rough pumps may completely avoid these problems.( 1 7 u ) Because of the above consideration, cryopump component vendors and system manufacturers often provide automatic regeneration control provisions with their equipment. These controls are usually micro-processor based systems that take the cryopump through the regeneration sequence, including the regeneration and subsequent isolation of the sieve traps.
Reverse Cycle Regeneration Twenty years ago Sarcia, of Oerlikin-Burhle U.S.A. made application and was awarded a patent on heating the arrays of a cryopump by reversing the motor direction of a displacer.(1 r s ) By merely reversing the direction of the expander motor, heat may be generated at the expander cooling stations. Nothing special need be done with expander valving. Just reverse the direction of the motor. The problem was that the motors on expanders of many companies were not reversible. This was not the case with Genesis Vacuum Technology's (formerly ETI's) expanders. David Crone and I tested out the scheme just to see if it would work in regenerating cryopumps.(1 6 9 ) It worked very efectively and our results were reported on at the AVS International Symposium in 1999. A typical pump quick regeneration cycle is shown in Fig. 5.4.13. 160
i
140
FIRST STAGE
120 o
100 80 60
(
40 20
SECOND ST GE _ ~ ARRAY
i
i./
E 0
~
10
20
30
40
50
60
REGENERATION TIME - Minutes F i g u r e 5.4.13. R e g e n e r a t i o n of a 200 m m d i a m . c r y o p u m p a f t e r p u m p i n g --1.14 x 10 6 T o r r - L of a r g o n , a n d by r e v e r s i n g t h e d i r e c t i o n a n d c o n t r o l l i n g t h e s p e e d of t h e c r y o p u m p m o t o r . (168)
We cycled the system through over 400 reverse cycle full regeneration cycles to determine if some mysterious expander seal wear was occuring as a results of heating. This proved not to be the case. Our bench mark was the time it took the refrigerator to chill down after each regeneration. This was invariant for the full number of cycles. Note that the second stage array temperature is not higher than that of the first stage arrays. This might suggest a heater is still needed on the second stage.
CRYOPUMPING
319
The Placebo Effect Assume that over several days we observe the pressure in a cryopumped system to rise from 2 x 10"8 to 1 x 10-r Torr. Also, we note that the temperature of the 2nd stage of the cryopump has increased in the same time from say 12*K to 15" K. From this we hypothesize that: i) the temperature is rising because the pressure is rising; and, ii) the pressure is increasing because the cryopump needs regeneration. Therefore, we regenerate the cryopump, and sure enough, after regeneration the temperature of the 2nd stage returns to the proper value of 10 - 12" K, the pressure to 2 x 10" a Torr, and we have reinforced our original hypothesis. An aside: A cross section of the expander of a typical cryopump is given in Fig. 5.5.1. The first stage displacer operates at a temperature gradient from RT to -77* K (i.e., the temperature of the first stage). The second stage displacer operates at temperature gradient of from 77"K to say 10 - 20*K (i.e., the temperature of the second stage). Any gas other than He (and 100-1000 ppm H 2 ) which is in the high pressure He stream, will cryocondense or coat out on surfaces within the regenerator beds (i.e., the lead shot or bronze screens shown in Fig. 5.5.1). This may take hours or even days to happen, and may be a gradual accumulation process. The regenerator bed acts like a dialysis or kidney machine for helium stream. It all depends on the amount and type of contaminant gas. When the contaminant gas is cryocondenced on these beds, it obstructs heat exchange processes which occur between the He gas and regenerator beds. Let's examine our original hypothesis in the context of the observed 2nd stage temperature increases. We know there are three sources of heat input to a cryopump: i) thermal radiation;//) heat of gas convection; and, iii) heat of sorption. This first possible source is straight forward, if we have not changed the system by turning on quartz lamps or other heaters. Gas convection losses can immediately be dismissed (e.g., what is the lowest pressure one can measure with a conventional thermocouple gage?). This leaves heat of sorption as a possible cause of the temperature increase. Let's see, if the pump has a speed of 2,000 Z/s for the system gas, and the pressure is 10-5 Torr, heat of sorption losses would be only 0.02 W. Therefore, this thermal source at 10-7 Torr is insignificant. None of the three sources of heat input to the cryopump caused the temperature increase. Then, with essentially constant heat input, the only other possible reason for the temperature increase is a decrease in refrigeration capacity for a f'Lxed heat load. This decrease in refrigeration capacity can be caused by gas contamination of the regenerator beds, or seal wear. We reject seal wear, because after "regeneration" we properly cooled down. The only possible explanation of the observed decrease in performance is that the refrigeration capacity has diminished for some reason. This results in a gradual increase in the second stage temperature. The effect of the increase in second stage temperature is a gradual increase in system pressure. You observe the pressure increase and erroneously deduce that it is causing the temperature increase. Therefore, you regenerate the pump. Sure enough, after the regeneration of the cryopump, the subsequent system pressure and temperature return to their proper low values. You conclude that the pump needed regeneration. What in fact has happened is that the contaminant gas, which collected in the regenerator-displacer bed, was liberated back into the He inventory on pump regeneration. The process of conta-
320
CRYOPUMPING
minant collection on the regenerator-displacer bed will start all over again. I have also noted cyclical pressure variations, with a periodicity of an hour or so, in cryopumps as a consequence of gas being collected in the regenerator beds, then being liberated as a consequence of second stage temperature increases, then again collecting, etc. Neon contaminant in the He will cause this effect. If a slight second stage temperature increase causes a slight system pressure increase, the vacuum system gas is probably either He or H~. Assuming that the sieve on the second stage of the cryopump has not been plugged by oil, you should be able to make rough calculations of the probable H2 loading which would results in this pressure increase. Is it consistent with the cryopump capacity and operating pressures? Count the molecules before assuming that regeneration is required.
Sources and Remedies of He Gas Contamination If the contamination collected in the regenerator bed is an air gas, the process will be completely reversible. That is, when the cryopump is warmed up, the contamination completely desorbs from the regenerator bed. Air gas contaminants can be introduced into the He inventory through contaminated hoses, the expander assembly or the compressor. One knee-jerk reaction to a malfunctioning refrigerator is to start swapping out components; for example, changing the compressor. This is obviously the worst thing to do. Think about it. If contamination is the problem, you now have one set of hoses, one expander and two compressors which are contaminated. I've seen such cross-contamination contagion shut down a customer's implanter manufacturing line. Air gas contaminants can be purged from the refrigerator system by pressurizing and venting the system approx, five times with five-nines pure helium. The process is as follows: i) let the system warm to room temperature; ii) turn on the refrigerator for a few seconds to flush the collected air gases out of the expander; iii) vent the helium out of the system to -25 psig; and, iv) pressurize the system to the recommended static helium pressure. Repeat steps ii)-iv) four to five times, and if it's air gas contamination, it will remedy the problem. Waming.t Do not operate the refrigerator at helium pressures below the specified static charge pressure. If you do, this will result in arcing in the compressor or expander, and may cause irreparable damage to one or both.
If the regenerator bed contamination comprises oil or water vapor, you have a serious problem. These substances collect and stay in the regenerator bed even when it is warmed to room temperature. In such cases the pump must be removed from the system, the expander assembly cleaned, and the regenerator-displacer assemble replaced. You would do well to thoroughly clean or replace the helium hoses in such circumstances. If it was oil contamination it probably carried over in the helium stream from the compressor. Did you replace the compressor adsorber (see Fig. 5.5.3) at the recommended times? I'd change them at twice the frequency recommended by the manufacturer. The replaceable compressor adsorbers are very inexpensive.
CRYOPUMPING
321
5.5 Closed-I.x~p, Gaseous Helium Refrigerators Some of the History( ~ 4 9 ) Most of us have seen the diagram which shows the chronology of the development of the semiconductor industry. The commercial genesis of this industry was the Fairchild Corporation. On a much smaller scale, there are counterpart origins of many of the technologies in the vacuum industry. We saw in Chap. 4 that SAES Getters, in Italy, was the origin of the significant contributions made to NEG technology. Varian Associates played the major role in the development of sputter-ion pumps. The late Harrold Farrow started the Excalibur Corporation which took liquid helium cryopumping from a laboratory curiosity into commercialization. The work of Linde, a Division of the Union Carbide Corporation, in liquid helium cryopumps and artificial zeolites was similarly notable. Arthur D. Little (ADL), working in concert with the Massachusetts Institute of Technology (MIT), was the corporate origin of the closed-loop gaseous helium refrigerators, variations of which are used in today's cryopumps. MIT owned ADL from 1935 to 1952, as A.D. Little bequeathed controlling interest of the corporation to MIT on his passing. Therefore, there was a close relationship between ADL and MIT. They worked jointly during WWII on many technical projects. The impetus for the refrigerator work at ADL had nothing to do with vacuum. Rather, it was directed at finding some more convenient and safe way of producing liquid helium. Prior to the start of WWII there were two places in the United States where one could conduct experiments at 4.2* K. One was at the University of Chicago, the second was at the Naval Research Laboratory. Researchers had to take their experiments to these facilities to conduct 4.2* K work. The liquid helium at these facilities was produced by high pressure, cascading Joule-Thomson expansion of N2, H2 and He. This process was potentially hazardous because of the necessity of working with high pressure H 2 in the second stage of expansion. JouleThomson (JT) expansion is merely the cooling effect stemming from the ansions of high pressure gas to lower pressures and below the inversion curve.( 1 5 ex~ In 1946, the late Professor Samuel C. Collins and his colleagues at MIT and ADL, successfully developed what became known as the Collins Helium Crvostat. This was a simple, less hazardous machine for the production liquid helium.( x s 1 ) This technology made use of a gaseous helium, mechanical refrigerator to produce refrigeration in two stages, and equivalent to that achieved by the JT-expansion of N2 and H 2 of the former process. The third stage was the JT-expansion of the He resulting in its liquefaction. Industry and laboratories came to MIT thereafter and asked them to produce more of these helium liquefiers. MIT was not in the business of building and selling equipment. Therefore, ADL took a licence under the MIT patent to build and sell these machines. By the mid-1960s the were 250 of these machines in use at laboratories throughout the world. Of course, not all of the machines were produced by ADL. In 1955 Dudley Buck, at MIT, demonstrated a superconducting switch for computer applications.(1 s s ) He called it the cryotron memory. Insurance companies and other large institutions needed reliable and fast methods of making calculations. The cryotron memory device presented this potential. It required the use of LHe, and had market potential for the sale of additional Collins Helium Cryostats.
322
CRYOPUMPING
The Collins Helium Cryostat needed a full-time operator to keep it up and running. It was belt-driven, required considerable care and attention, and would operate about 1000 hours before it had to be rebuilt. Therefore, ADL undertook the development of a continuous duty, unattended, closed-cycle machine which would achieve 4.2* K, and which would operate continuously for one year between maintenance intervals. The third stage of this machine was the JT-expansion of He. The late Professor W.E. Gifford was a mechanical engineer at ADL at the time. Professor Gifford and Dr. Howard O. McMahon made major refinements in these helium mechanical refrigerators.(1 2 z ,1 2 4 ) They developed a closed cycle machine which is presently called the GM-cycle (Gifford-McMahon-cycle) refrigerator. Applications for these refrigerators blossomed. For example, they found use in defense electronics and satellite communications. These applications inherently brought issues of reliability and low maintenance to the forefront. In 1952 ADL purchased back the stock of the company, owned by MIT, as part of the investment portfolio of a retirement income plan for its employees. In 1967, ADL created an 80% ADL subsidiary of its cryogenic activities. The name of this subsidiary was 500, Incorporated. Two years later, in 1969, the 80% ADL ownership was transferred to the ADL employee retirement portfolio, and the name was changed Cryogenic Technology, Inc. (CTI). CTI worked for a number of years with vacuum technologists in an to gain acceptance of these machines in cryopumping applications. They built ten of a variety, on speculation, and specifically for cryopumping applications. The refrigerator was referred to as called the Creacher. The name came from a combination of the words cryogenics and teacher. Presumably, CTI sought with this machine to teach vacuum technologists about the merits of cryo-pumping with mechanical refrigerators. The machine, first reported on in 1963, was a modified Taconis-cycle refrigerator, capable of achieving 10" K at the second stage, and making use of LN2 at the first stage.(1 5 a ) Three years later, Hogan and Turner reported on use of this refrigerator in a vacuum application.(7 1 ) However, because it lacked the use of sieve material on the second stage, it could not pump He, H2 or Ne. The idea of gluing coconut charcoal to a second stage array was literally incomprehensible to a UHV vacuum technologists at that time. In 1967, Boeing, at Kent, Washington, had the task of testing the NASA Ranger and Surveyor, unmanned satellites that went to the moon. These satellites had TV and IR cameras. They were trying to test them in a diffusion pumped, 12.2 m ~ x 24.4 m tall space chamber (i.e., a 12.2 m ~ right angle cylinder subtended by two 12.2 m ,~ hemispheres). The chamber was pumped by two diffusion pumps, each of---1.2 m ~. The diffusion pump backstreaming problem was causing havoc with the optics of the systems. CTI, working with Douglas McKinney of Boeing, built the world's first GM-cycle cryo-pumped space chamber. The cryopumps were augmented with two, 2400 Z/sec sputter-ion pumps and TSP pumps, as vacuum technologists had still not rediscovered activated coconut charcoal. The chamber had an LN2 shrouded wall. According to John Harvell, of CTI, it was James A. O'Neil, of CTI, who made the breakthrough of gluing activated coconut charcoal to the second stage of a GM-cycle refrigerator. He was experimenting with the cryotrapping of He with Ar. He constructed the equivalent of a diffusion pump, where the pumping medium was Ar, rather than oil. He had little success with this cryotrapping apparatus with a second stage temperature of 20* K. Therefore, he resorted to gluing activated
CRYOPUMPING
323
charcoal to the second stage. In 1961, W. E. Gifford left CTI and took a chair of Professor of Mechanical Engineering at Syracuse University. While there, he started a small company called Cryomech. Professor Gifford was advisor to Ralph C. Longsworth who received his Ph.D. at Syracuse University. Longsworth managed Cryomech thereafter. Gifford and Longsworth developed a GM-cycle, pneumatically driven, rotary valve machine while at Syracuse. An agreement was reached between CTI and Gifford to crosslicenee this design. Ralph Longsworth eventually left Cryomech and headed up a cryogenics development group at Air Products and Chemicals, Inc. (APCI), in Allentown, PA. While there, he developed a ref'med version of the pneumatically driven, rotary valve, GM-rnaehine which featured a modified Solvay-cycle. Samuel Collins also consulted for the Advanced Products Division of APCI in the late 1950s and early 1960s. The above scientists and technologists are the people who developed and ref'med the refrigerators which are presently used in all closed-loop gaseous helium cryopumps presently manufactured and sold throughout the world. It all started with development of the Collins Helium Cryostat. In the years that followed, hundreds of articles appeared in cryogenics journals describing various versions of the basic GM-cycle machines. Also, in excess of 200 United States patents were subsequently issued on various improvements to these machines. While at Varian Associates, I was for a time in charge of R&D for the Pumps and Instruments Group. In 1975 my marketing counterpart, John McLaughlin and I visited Cryomech, APCI and CTI to learn what these new cryopumps were all about. We had the pleasure of meeting for the first time Professor Gifford at Cryomech, and Dr. Ralph Longsworth at APCI. We quickly became converts to this new technology. The semiconductor industry was blossoming, and we knew that cryopumping would play a major role in this area.(1 2 7 ) In July of 1975 we put together a five-year business plan for Varian's participation in this market. The key to success was the GM-cycle refrigerator. CTI's pricing strategy was such that it was impossible for us to purchase their machines and integrate them into cryopumps, while at the same time compete with them in the open market. For the next few years, Varian purchased cryopump refrigerators from APCI, and developed a line of cryopumps and cryopumped systems. The first system was a batch coater, but the most successful was the Varian-3180, the first cassette-to-cassette coater of the semiconductor industry. Varian eventually developed their own line of cryopump refrigerators as did many other vacuum companies. But, the origin of all these machines was the GM-cycle refrigerator developed at CTI (ADL), for the Collins Helium Cryostat.
The G M - C y c l e R e f r i g e r a t o r The key to success in the development of the refrigerator resided in two areas. First, one had to develop a method of separating out contaminants from the He gas stream. As noted, any gas or contaminant other than He will cause problems in the refrigerator expander. This includes oil vapors, water, and all of the air gases. The second challenge was to develop a reliable expander seal which would operate at 40 80" K for several thousand hours. An example of a typical expander, with regenerator-displacer bed, is shown in Fig. 5.5.1. There are two regenerator-displacer pistons coupled in series. These
324
CRYOPUMPING
pistons reciprocate within the thin-wall cylinder heads. In case of the CTI design, they are driven by mechanical coupling to an electric motor, through a scotch-yoke assembly. In the Gifford-Longsworth (APD hereafter) approach, the pistons are driven pneumatically by exerting a He pressure difference on a third, slack-piston fixed to the RT end of the first stage expander. From a reliability and functional standpoint, I perceive of no disadvantage to either approach used to drive the regenerator-displacer pistons. Though the reciprocating frequency of the APD machines is - x 2 that of the CTI approach, the stroke of the latter is - x 2 that of the former. Therefore, seal wear will be the same in either case.
E
ND STA( bANSION )LUME
"BED" / I
IUM
LEAD SHOT
WALL '.SS STE, BING
ALLOY
PISTON ~ ~
T STAGI ~ANSION )LUME
BRONZE SCREEN DISKS
HELIUM SEAL REGENERATOR-DISPLACER PISTON A S S E M B L Y
H E L I U M PRESSURIZATION
Figure 5.5.1.
CYCLE
HELIUMEXHAUSTCYCLE
Closed-loop, gaseous h e l i u m expander assembly.
The regenerator bed comprises either a lead-alloy shot (the second stage) or a combination of lead shot and bronze screen material. The specific heat of bronze is higher than that of lead above a temperature o f - 5 5 * K. Therefore, bronze alone or a mix of bronze and lead shot is used as the first stage regenerator bed, and usually only lead shot in the second stage regenerator bed. Mentioned was made that rare earth materials are at times used in the second stage regenerator beds.(1 6 7,1 6 8 ) Materials in these regenerator beds have high surface-to-volume ratios. They alternately serve as a means of storing heat calories, on one phase of the refrigeration cycle and giving up heat calories in a different phase of the cycle. The specific heat of the regenerator bed material is important in this function. The magnitude of the specific heat of the regenerator bed materials, at reduced temperatures, is the primary limitation to the minimum temperatures which can be achieved by these machines. Also, any contaminant which deposits on the surface of the regenerator material, so as to create a thermal barrier between it and the gaseous He, will impact on the refrigeration capability (capacity) of the refrigerator. For this reason, high purity He is required.
CRYOPUMPING
325
High and low pressure He valves, within the expander assembly, open and close in a manner synchronous with the reciprocating displacer pistons. When the pistons are approaching the top of the stroke, the high pressure He valve into the displacer is opened. As the displacer piston is withdrawn from top dead center, high pressure He flUs the void volumes above the 1st and 2nd stage piston. The high pressure valve is then closed, and the low pressure valve opened when the regenerator-displacer piston is almost fully withdrawn. As the regenerator-displacer piston assembly is on the rise toward top dead center, the He expands in the voids above the 1st and 2nd stages, and exhausts through the low pressure He valve. This expansion process causes cooling of the He in the volume created above the respective pistons, and that He filling the void volume in the regererator-displacer beds. Some of the refrigeration created by the expansion is used at the first and second stage cooling stations to cool the arrays attached thereon. Some of the remaining refrigeration produced by the expansion process precools the regenerator beds, in anticipation of the next portion of the cycle. That is, the exhausting He picks up calories from the regenerator bed on passing over the surfaces of the bed material. During the next pressurization cycle, the He is precooled by the regenerator beds. In this case, the regenerator beds pick up calories from the high pressure, RT He. The role of the regenerator bed is now clear. It functions as a thermal ballast in the refrigeration cycle. The amount of refrigeration produced at the two stages depends on the amount and effectiveness of the regenerator beds, the volume before and after expansion, and the pressure of He at the two stages of the cycle. This, in very simple terms which even I can understand, is how these expanders work. If the He seals leak, He will bypass the regenerator beds, and cause a reduction in refrigeration capacity. Also, the shunting He will result in thermal losses. The reliability of these seals is very high in todays GM-cycle machines. Usually the seals require replacement every 30,000 - 40,000 hours, though Harvell notes in one particular instance the seals lasted 65,000 hours, and still had not failed.(1 s 4 ) r
I
~ -
i
1.
SEALED OFF END - - - - _ _ _ ~ ~
OF CAPILLARY TUBE
"~
THIN
[1[ [
BRAZED TO M[ SECOND STAGE ~ ] 1
~-
CAPILLARY ~ TUBE
/
EXPANDER PUMP BODY WELD COLLAR
I " I
BOURDON-TUBE / - PRESSURE-~ ~
'~l
[
.J
/
I
F i g u r e 5.5.2. H y d r o g e n v a p o r p r e s s u r e b u l b u s e d to m e a s u r e the temperature of t h e s e c o n d s t a g e of t h e e x p a n d e r .
326
CRYOPUMPING
Two methods are used to measure the temperature of the second stage of the expander. One method involves the use of what is called a hydrogen vapor bulb. As shown in Fig 5.5.2, a small tubular pipe, sealed of on the end, is wrapped around and brazed to the second stage cooling station. The long, thin capillary tube, posing negligible thermal shutting of the second stage, is then brought out through the bottom of the expander pump body weld collar. It is then attached to a conventional high pressure gauge. By measuring the hydrogen (vapor) pressure one is thereby able to deduce the temperature of the second stage (see problem 22). The second, and far more current method of measuring the temperature of the second stage is through use of silicon diodes. The forward voltage drop in these diodes changes significantly as a function of temperature. It is nearly linear, and -dV/dT is --50 mV/* K in the 1 to 35 * K range. Also, the current source need not be highly regulated.(1 s s )
The Compressor The compressor is needed to supply high purity, high pressure He to the expander. A schematic representation of a typical compressor assembly is shown in Fig. 5.5.3. For the last few decades, the Hc compressor modules of most cryopump refrigerators have been modified air-conditioning compressors. For example, in the late 1970s, the CTI Model-21, APCI (APD) Model-202 and Model-204 refrigerators used piston-type, Tecumseh air-conditioning compressors. The CTI Model-350SC refrigerator used an RCA Whirlpool, rotary vane compressor. Longsworth, who had worked for two years at Carrier Air Conditioning, in Syracuse, proposed the use of these compressor modules in this application, to Gifford. About the same time, the same compressor modules came into use at ADL. Prior to that, oil lubricated He compressors had been used in this application. The technology for removing the oil vapors from the He stream was developed at ADL. BIFIIAR WOUND WATER AND
HIGH
THERMAL EPOXY ~
VA+
SOLENOID--~
PRESSURE
~
t-"~t~O~,] t
(°+l
I
~t
F . . . . . . . . . . .
~
u
2
HIGH
RETURN LINE -~
OIL
¢1
OIL COMPRESSOR L. H
HIGH PRESS. SHUTDOWN~ SWITCH
VALVE
Closed
t "-~ V2
VENTING VALVE
HELIUM RETURN
COMPRESSOR 5.5.3.
loop,
~
HELIUM BYPASS CHECK-VALVE
RELIEF - / " ' "~ VALVE
L O W PRESS. SHUTDOWN -~
Figure
PRESSURE HELIUM
gaseous
helium
/~-
Vt HELIUM SUPPLY
eryopump
compressor.
CRYOPUMPING
327
The total He mass flow required for a typical 20 cm ~ cryopump is --4.7 Atm.-Z/sec (10 SCFM) and that of a 30 cm ~ cryopump, twice this mass flow. The He compressor system is pressurized to a static pressure of --140 Nt/cm2 (--200 psig). When operating, the pressure difference between the low pressure and high pressure ports of the expander is --140 Nt/cm 2. The above static and dynamic pressures are only approximate values and will differ with the size and type of refrigerator. Early-day sealed compressor modules were designed to operate with freon ®. The compressor motor is an integral part of the compressor and is therefore, also housed in the compressor case. The heat of compression of He is much greater than that of freon ®. Therefore, the compressor module is usually modified to include provisions for additional cooling of both the compressor and the high pressure He prior to its conveyance to the expander. This is sometimes accomplished by bifdarwrapping the high pressure He and water cooling lines and bonding them to the compressor casing with thermal epoxy. Infant mortality compressor problems in cryopump refrigerators were rampant in the mid and late 1980s. Problems were encountered with these compressors because new powdered metallurgy techniques used by a particular compressor manufacturer to fabricate many of the compressor components.(1 5 6 ) Cryopump compressor failure problems nearly put some cryopump companies out of business. Also, this venture in powdered metallurgy cost the compressor manufacturer many hundreds of millions of dollars because of home refrigerator problems. After removing the freon ® from the sealed units, the units are flushed clean and charged with a low vapor pressure oil. The coalescer, immediately downstream of the compressor, serves to remove oil droplets or fog from the He stream. This oil returns to the compressor through a bypass, oil return line. This takes the form of a capillary tube or some form of aperture from the high, to low pressure He circuit. It is important that this oil return circuit not become plugged or, in time, oil will carry over to the expander. The adsorber, located downstream of the coalescer, serves to filter out the remaining gas impurities in the He stream. As you might guess, this adsorber is filler with either, and sometimes both activated charcoal and artificial zeolite. It has a f'mite capacity. Because of this, it must be changed every 8,000 - 16,000 hours. Some manufacturers claim adsorber life of as much as 30,000 hours. However, this is not the place to economize. Therefore, I always change the adsorber every 5,000 hours. Most compressor packages have some form of elapsed time clock to serve as warning that preventative maintenance is required. Because the motor is housed within the compressor casing, problems of electrical breakdown or discharge will occur if the He pressure becomes too low. For this reason, a low pressure shutdown switch is located on the He return line. Also, if the compressor is charged with too high a He pressure, the compressor will overheat. For this reason, a high pressure shutdown switch is installed on the He supply line. A He bypass valve is used to shunt the supply and return He lines. This serves a very important function. Oil which might have been carried over into the He lines, as a consequence of shipping or handling, poses a threat to the expander. Therefore, immediately after receipt of the compressor, the compressor unit should be operated for a couple of hours prior to installation of the expander. This age-in of the compressor will afford assurance that oil that might have gotten in the lines, will be flushed back into the compressor by the He stream.
328
CRYOPUMPING
Valves V1 and V~ may be special sealed-off, AeroQuip fittings which retain the high pressure He, rather than valves manipulated by the user. The AeroQuip fitting have the advantage that when mating a pair together, the volume of trapped gas between the female and male fitting is theoretically zero. The prevents the introduction of gas contaminants when attaching hoses to the expander or compressor. The charging and venting valves are self-explanatory. These are also used to top off the He supply in the event leaks in the compressor package. Also, in the event of He contamination, the compressor package is pressurized and then partially vented through the respective valves. Of course, high purity He should be used. As noted, the He expander will not work properly if the regenerator beds become contaminated. A contamination is any gas or vapor other than He. The expander is similar to a dialysis machine. All the system impurities, if not f'dtered out by the compressor adsorber, will eventually collect in the regenerator beds. While at Varian, many of the cryopumps we manufactured were integrated into Varian ion implanters. We once had a contamination problem with a batch of compressor adsorbers, and didn't know it at the time. Some were contaminated by air gases, and some were not. In refrigerators with contaminated adsorbers, after 1020 hours of cryopump operation, sufficient gas contamination accumulated in the expanders to seriously degrade their refrigeration performance. We had measured the refrigeration capacity of each refrigerator by thermal loading them with resistors. But this, and the final test pump procedure then involved only 3-4 hours of cold operation. The problem became evident when dozens of the cryopumps were installed and operating on ion implanters in various stages of manufacturing. If the manufacturing staff observed some anomalous cryopump behavior on an implanter, they did one of two things. They first exchanged compressors. If the problem persisted, they then regenerated the cryopump. Because of the previously described placebo effect and the above dialysis effect, in a matter of a few days every refrigerator system in the ion implanter factory became contaminated. The solution to the problem involved exchanging all the adsorbers, and the multiple flushing of the expanders, hoses and compressors with high purity He. Thereafter, every cryopump refrigerator was operated for 50 hours as part of the cryopump final test. The lesson in this for you is that if you are observing some anomalous behavior in your cryopump, never swap compressors to determine if the problem is due to this unit. If the cryopump expander has cryosorbed the contaminant on the regenerator beds, and you then swap compressors, you now have two contaminated compressors as well as one contaminated expander. Though the compressor shown in Fig. 5.5.3 is of a 1975 vintage, all of the key components shown therein are still applicable to today's modern day compressors. Scroll-type helium compressors have become the current technology. The compressor capsules are specifically designed for helium applications. Cooling is provided in the capsules by circulating the lubricating oil through a separate external heat exchanger. The compressors all still have oil coalescers, perhaps preceded in the helium stream by some form of knock-off assembly which removes the bulk oil from the helium stream. Also, some compressor packages use separate ballast tanks in the helium return line. These serve to smooth out the pressure variations during operation. The scroll-type helium capsules are manufactured in countries including Japan and in the United States.
CRYOPUMPING
329
Refrigerator Capacity The first and second stages of a Gifford-McMahon refrigerator are in series. Helium passes through the first stage regenerator bed prior to reaching the second stage regenerator bed. Therefore, we intuitively know that the thermal loading of the first stage will impact on the thermal capacity and temperature of the second stage. The size and aspect ratios (i.e., length to diameter) of the regenerator beds and the amount of bed material (e.g., lead alloy shot) in the regererator-displacer pistons impacts on the refrigeration capacity of the respective stages. Also, the amount of He flow and its expansion in the two stages impacts on refrigeration at the respective stages. An example of the interdependence of the simultaneous thermal loading of the stages of a Gifford-McMahon refrigerator is shown in Fig. 5.5.4.(1 s 7 ) 25
20
15
10 8 30
40
50
60
70
FIRST STAGE TEMPERATURE -
80
90
OK
F i g u r e 5.5.4. T e m p e r a t u r e v a r i a t i o n s in t h e f i r s t a n d s e c o n d s t a g e s of a G i f f o r d - M c M a h o n r e f r i g e r a t o r w i t h t h e r m a l l o a d i n g of b o t h stages.(157) (Data reproduced with the pemission of Leybold-Heraeus; Refrigerator Mod. # RDG-510.)
These data were provided by Dr. Dieter MOiler of Leybold. Following the convention, the simultaneous capacity is - 6 W at 20* K with -13 W at 77* K. With this capacity, one would have more then ample refrigeration for the first stage of a 20 cm pump, with a very robust second stage capacity. N e t w o r k i n g C r y o p u m p s on C o m p l e x S y s t e m s The batch coating system shown in Fig. 5.4.10 is a very simple configuration compared to the present day technology of cryopumped systems. Cluster tool used in the semiconductor industry, and for magnetic disk coating may have as many as 10 - 12 high vacuum pumps. These might include cryopumps, turbomolecular pumps, or combinations of cryopumps and cryoturbo pumps.
330
CRYOPUMPING
Most PVD (physical vapor deposition) systems can be pumped exclusively with cryopumps. However, many LPCVD (low pressure chemical vapor deposition) applications are pumped with a hybrid pump called a cryoturbo pump. Cryopumps can not be used in many CVD applications because of the corrosive gases either created or used in the processes. However, the internal aluminum parts of turbomolecular pumps can be coated by electroless nickel plating. The nickel coating affords an immunity to the corrosive gases. However, the water vapor pumping speed of turbomolecular pumps is somewhat limited. Therefore, a pump was developed by Toshiba and Ebara where a single stage GM exl~ander was used to cool an array residing on top of the turbomolecular pump.(1 7 6 ) This was coined a cryoturbo pump. The cryopanel attached to the single stage expander is highly transparent to the passage of gas to the turbomolecular pump. This affords both high water vapor speed, and yet only modestly restricts the turbomolecular pump speed for other gases. The main advantages of the use of cryoturbo pumps are that: i) they are immune to many of the corrosive gases used in CVD applications; and, ii) they require very infrequent regeneration. For example, in some PVD application regeneration of a cryopump might be required once or twice a week. Of course, there is an associated argument that the cryopumps are regenerated only when the system is down for target changes. If magnetically levitated turbomolecular pumps are used - the semiconductor industry favors this - the cost of a cryoturbo package will be approximately 2-3 times that of a cryopump of comparable speed. A turbomolecular pump having ceramic bearings, and used in a cryoturbo package, will cost approximately ~ - 2 times that of a conventional cryopump with the same nitrogen speed. However, magnetically levitated turbopumps (i.e., the rotors) have the advantage that there are no bearings assemblies exposed to the corrosive gases of CVD applications. As an alternative to cryopumping with two-stage GM machines, many OEMs (original equipment manufacturers) are using turbomolecular pumps to pump on the process chambers, and augmenting the water vapor speed with copper coils or plates cooled by either vapor refrigerants or single-stage GM expanders respectively. The complexity of cluster tools, implanters and other systems made it necessary to have computer controlled regeneration and control of the pumps. Though Varian offered a microprocessor based batch coater, with automatic regeneration in 1978, Perkin-Elmer, in 1984, was the first to offer networking capability of multiple cryopump controllers. They used a IEEE-488 data bus for multi-drop applications. In 1987 Varian sold the first totally cryopumped cluster tool (the M-2000), featuring fully automated valve controls, and regeneration provision. This system featured a star networking system, rather than a multi-drop network as in the Perkin-Elmer case. The IEEE RS-485 ("BITBUS") control system architecture came into prominence in 1983 (e.g., see Section 2.12). Starting in the early 1990s, many cryopump companies offered cryopump controllers which could be controlled the RS-495 multi-drop network. CTI patented a clever scheme where the mounted the individual pump controllers directly on the cryopumps.(t 7 r ) They called this the ONBoard cryopump. The semiconductor industry is presently promoting use of more modern data bus systems having peer to peer capability.
CRYOPUMPING
331
5.5.4 Meissner Coils and Traps Using Vapor Refrigerants It was noted that Leever reported on Freon cooled diffusion pump coldtraps in 1954.(1 4 s ) He used a two-stage expansion refrigerator, the first stage using Freon 12, and the second Freon 13. With this scheme he was able to achieve a cryotrap temperature of--180 K. The vapor pressure of water vapor at this temperature is -4.3 x 10"s . Therefore, the trap was not a very good water vapor pump. However, its primary purpose was to prevent backstreaming of vapors from oil diffusion pumps. A couple of year later Meissner reported on the use of a liquid nitrogen cooled copper tube, coiled within a bell jar so as to completely surround the work area in an evaporation system.( 1 T a ) The copper coils within a 14-inch bell jar had a total surface are of --3,000 cm 2, and the water vapor speed provided at the work was
-44,000 ~Is.
In 1972 Missimer (not Meissner; the similarity in names confused me for a number of years) f'ded a patent application for a mixed gas refrigeration system which became known as a mixed refrigerant cascade (MRC) cycle, or autorefrigeration cycle (ARC).(1 7 9 ) These refrigerators were originally developed for the liquefaction of natural gas. In 1974 Missimer founded Polycold Systems, where he and his colleagues developed a variety of refrigerators for cooling cold traps, chevrons, and piping distributed within vacuum systems similar to the Meissner coils. Use of these distributed coils in vacuum systems had the advantage of providing high water vapor speeds (i.e., ten of thousands of liters/sec) directly within the work chamber. The MRC cycle actually became know as the Polycold cycle,(1 7 1 ) after the company Polycold Systems, Inc., founded by Missimer. Through use of mixes of halocarbons or fluorocarbons he was able to develop refrigerators which, on expansion of cryogen vapors, produced temperatures of 128 to 143" K, and with capacities of several hundred watts. In 1979 Missimer made application to patent an invention which made possible to heat the coils or traps for purposes of regeneration.(1 8 o ) The Polycold technology is used throughout the world in a wide variety of functional and scientific coating and etching applications.
Problem Set 1) Why do the curves for Ar, Xe and Kr, in Hobson's data, shown in Fig. 5.1.7, abruptly cease their gradual trends with surface coverage, and become vertical lines with continuing coverage? 2) Use (1.12.7) and (1.12.3) to obtain the results of (5.1.3). Assume that the gas on entering the cold chamber is instantaneously thermalized to temperature T2. 3) Using (5.1.4) and (5.1.5), and given the values of Tx ,A a andA s, derive Serf, the effective speed produced in the warm chamber due to the pumping of the cold chamber, in terms of t~, Cx, A s and A a. Assume that the gas on entering the cold chamber is instantaneously thermalized to the unknown temperature T2. 4) Using (5.1.4) and (5.1.5) and given the values of T1, A a and As, derive the sticking coefficient t~, assuming you have measured the values of we know Q and P 1. Assume that the gas on entering the cold chamber is instantaneously thermalized to the unknown temperature T2.
332
CRYOPUMPING
Problem Set 5) Using (5.1.4) and (5.1.5) and given the values of T1, T2 ,A a and A s, derive the sticking coefficient ~, assuming you have measured the values of P 1 and P2. 6) Calculate the change in kinetic energy of 1 Torr-Z of Ar when taken from RT to zero velocity. If this is done once every second, what is the rate at which work must be done bring this RT gas to rest (i.e., watts of power)? What is the ratio of the above power to that of the heat of sorption for Ar? 7) It was noted in Section 5.2.5 that when roughing a system with two sorption pumps, one should valve out the first pump at --0.1 Torr. What is so special about this pressure level? 8) Calculate the base pressure which could be theoretically achieved by the simultaneous use of 100 sorption pumps, each containing --1350 g of Linde 5A sieve materials and cooled to LN2 temperature, and used to pump a one liter volume of Ne at pressure of 10 Torr. 9) Repeat the calculation in problem 8. However, this time use only two sorption pumps, and stage the first at its equilibrium pressure. 10) Referring to Fig. 5.1.8, calculate the effective speed delivered to the RT chamber. Assume that gases are immediately thermalized on entering the respective chambers, and that no pumping occurs on the walls of the chamber at temperature T2. 11) The chamber to the left of Fig. 5.1.8 is at RT and the one to the right is very cold. A test gas introduced in the RT chamber is therein thermalized and subsequently pumped by the cold chamber. The RT gas on being pumped by the cold chamber has a sticking coefficient of c~ 1 • Assume now that a chevron is placed in the aperture separating the two chambers, and that this chevron is at an intermediate temperature so as to prechill the gas prior to it reaching the cold chamber. Assume that c~ 2 is the new sticking coefficient of the prechilled gas in the cold chamber. What is the conductance of the chevron, assuming the ratio of measured speeds, in the RT chamber and with and with out the chevron, is --0.5. 12) Assume the two chambers in Fig. 5.1.3 are initially at RT = To. The volumes of the two chambers are given as V1 and V2. Assume that there is an initial He pressure in both chambers is given as Po" If the temperature of the chamber of volume V2 is cooled to temperature T2, and there is no surface pumping in either chamber, express the final pressure in both chambers in terms of the givens. 13) What was the probable temperature of the first stage louver of the cryopump for which speeds are given in Fig. 5.4.2? 14) Assume that during measurement of the isosteres of Figures 5.4.6 and 5.4.7, the gases were at a temperature of--77* K. However, assume that the pressure gauge used to measure the pressures was located in an adjacent RT chamber. How do you compensate for this temperature difference? 15) Construct adsorption isotherms for He and H~ using the data of Figures 5.4.6. and 5.4.7. Compensate for thermomolecular effects by assuming conditions given in problem 14.
CRYOPUMPING
333
Problem Set 16) Assume that the vessel in Fig. 5.4.10 has a volume of 200 liters. Assume that this is a batch coater and that you are using an e-gun source to deposit A~ films on some substrate. After loading the bell-jar with the substrates which are to be coated, you rough the bell-jar to 0.1 Torr, isolate the roughing pump, and then valve in the cryopump. You limit the roughing pressure of the bell-jar to this value to avoid back-streaming of oil from the roughing pump. Assume that the coating process takes --20 minutes, and that the partial pressures of H 2 0 and H2 during this process are 2 × 10 -7 Torr and 5 x 10-6 Torr, respectively. Assume the H2 pumping speed in the vessel is 10 3 Z/sec. Assume that you complete 100 batch cycles prior to deciding to regenerate the cryopump. On warming the cryopump to room temperature, what is the partial pressure of O 2 and H 2 in the cryopump if the isolated pump has a volume of 50 liters. 17) Hydrogen and air comprise an explosive mixture in proportions ranging from 4.0 to 75 percent air, by volume. Conceive of a scenario in the example given in problem 16, wherein the mixture of O2 and H2 is not potentially explosive. (Hint: in real evaporative coating applications, this is difficult.) 18) Assume that no pressure relief valve was used on the cryopump in problem 16. What would the total pressure be in the cryopump after the 100 coating cycles. If --290 Torr-Z of H 2 and half this much O 2, when burned, corresponds to --1.0 g of TNT in explosive energy,(9 4 ) how much potential equivalent of TNT explosive energy is stored in the cryopump. 19) Derive the He and H 2 pumping speeds per unit area of activated coconut charcoal at 14" K by using the zero loading data of Fig. 5.4.4 and through use of (5.1.18). Assume the second stage is configured as shown in Fig. 5.4.3, but that the slots between the leaves are on average rectangular. Assume the cryopump is a 30 cm ~ pump and that the first stage is shielded by 60* louvers. (Hint: start by setting up an equation similar to (2.10.8).) 20) Assume that a second stage array similar to that described in Fig. 5.4.3 is used in the cryopump in problem 17, and that the partial pressure of H2 is ---10-6 Torr during the coating operation. Assume that we delay regeneration of the cryopump until the speed of the cryopump has decreased to 2000 Z/sec. How many batch coatings will we be able to make prior to the requirement of cryopump regeneration? 21) Work problems 16-18 for the case described in problem 20. 22) What is the pressure indicated by a hydrogen vapor bulb of a cryo-pump if the temperature of the second stage was 20* K?. 23) Develop an equation similar to (5.1.7), but which expresses gas densities P x and p 2 in the two chambers of Fig. 5.1.3, with zero net flux passing through the aperture joining the two chambers. The gas density in the chamber at temperature T1 i s p z . 24) Assume a eryostat containing a large superconducting magnet comprises three very large interconnected volumes Vx, V2 and 113 at temperatures Tx, T2 and T3 respectively. Assume that gaseous He is leaked into the cryostat at a rate Q, and that none of this gas is pumped on the cold surfaces within the cryostat. Assume that at time t, the leak is turned off. What is the partial pressure of He, at equilibrium, in the three chambers?
334
CRYOPUMPING
References 1. 2. 3.
4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.
20.
Redhead, P.A., Hobson, J.P., and Kornelsen, E.V., The Physical Basis of Ultrahigh Va.cuum (Chapman and Hall, Ltd., London, 1968), p. 61. Chubb, J.N. and Pollard, I.E., "Experimental Studies of Hydrogen Condensation on to Liquid Helium Cooled Surfaces", Vacuum 15, 491 (1965). Levenson, L.L., "Condensation Coefficients of Gases Measured with a Quartz Crystal Microbalance, II, Ar, Kr, Xe, and CO2 ", Proc. 14th Nat. AVS Syrup., 1967 (Herbrick and Held Printing Company, Pittsburgh, 1968), p. 95. Brown, R.F., Trayer, D.M., and Busby, M.R., "Condensation of 300-2500 K Gases on Surfaces at Cryogenic Temperatures", J. Vac. Sci. Technol. 7(1), 241 (1970). Bentley, P.D. and Hands, B.A., "The Condensation of Atmospheric Gases on Cold Surfaces", Proc. Roy. Soc. (London), A359, 319 (1978). Dushman, S., Scientific Foundations of Vacuum Technique (John Wiley and Sons, Inc., New York, 1949), p. 389. Honig, R.E. and Hook, H.O., "Vapor Pressure Data for Some Common Gases", RCA Review, 360 (1960). Barron, R., Cryogenic Systems (McGraw-Hill Book Company, New York, 1966), p. 271. Dushman, S., op. cit. (1949), p. 742. Heafer, R.A., Cryopumping- Theory and Practice (Clarendon Press, Oxford, 1989), p. 61. Eisenstadt, M.M., "Condensation of Gases During Cryopumping - The Effect of Surface Temperature on the Critical Energy for Trapping", J. Vac. Sci. Technol. 7(4), 479 (1970). Lee, T.J., "The Condensation of H2 and D2" Astrophysics and Vacuum Technology", J. Vac. Sci. Technol. 9(1), 257 (1972). Honig, R.E., "Vapor Pressure Data for the Solid and Liquid Elements", RCA Review, 567 (1962). Redhead, P.A., et al., op. cit., p. 25. Bennett, M.J. and Tompkins, F.C., "Thermal Transpiration: Application of Liang's Equation", Trans. Faraday Soc. 53, 185 (1957). Podgurski, H.H. and Davis, F.N., "Thermal Transpiration at Low Pressure. The Vapor Pressure of Xenon Below 90* K", J. Phys. Chem. 65(2), 1343 (1961). Edmonds, T. and Hobson, J.P., "A Study of Thermal Transpiration Using Ultrahigh-Vacuum Techniques", J. Vac. Sci. Technol. 2(4), 182 (1%5). Dawson, J.P. and Haygood, J.D., "Cryopumping", Cryogenics 5(2), 57 (1965). Bentley, P.D. and Hands, B.A., "The Initiation of Deposition of Gases on Cryopumps", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. R~idenaur, F.P. Viehb6ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 73. Hobson, J.P., "Physical Adsorption Isotherms Extending from Ultrahigh Vacuum to Vapor Pressure", J. Phy. Chem. 73(8), 2720 (1969).
CRYOPUMPING
335
References 21. B(ittner, P., "Measuring Pumping Speed, Condensate Thickness and Condensate Stability in a Large Cryopump", Proc. 8th Int. Vac. Cong., 1980 (Suppl6ment ~. la Revue, "Le Vide, les Couches Minces, No. 201", 1981), p. 316. 22. Guthrie, A. and Wakerling, R.K., .Vacuum Equipment and Techniques (McGraw-HiU Book Company, Inc., New York, 1949), p. 249. 23. Redhead, P.A., et al., op. cit., p. 40. 24. Longsworth, R.C. and Webber, RJ., "Cryopump Vacuum Recovery After Pumping At" and H 2 ", Submitted to J. Vac. Sci. Technol., February, 1991. 25. Halama, HJ. and Aggus, J.R., "Measurement of Adsorption Isotherms and Pumping Speed of Helium on Molecular Sieve in the 10" 1 1 _ 10-r Torr Range at 4.2* K", J. Vac. Sci. Technol. 11(1), 333 (1974). 26. Halama, H.J. and Aggus, J.R., "Cryosorption Pumping for Intersecting Storage Rings", J. Vac. Sci. Technol. 12(1), 532 (1975). 27. Robens, E., "Physical Adsorption Studies with the Vacuum Mirobalance", J. Vac. Sci. Technol. 17(1), 92 (1980). 28. Dushman, S., Scientific Foundations of Vacuum Technique (John Wiley and Sons, New York, 1962), p. 376. 29. Stern, S.A. and DiPaolo, F.S., "The Adsorption of Atmospheric Gases on Molecular Sieves at Low Pressures and Temperatures. The Effect of Preadsorbed Water", J. Vac. Sci. Technol. 4(6), 347 (1967). 30. Hobson, J.P., "Theoretical Isotherms for Physical Adsorption at Pressures Below 10-1 o Torr", J. Vac. Sci. Technol. 3, 281 (1966). 31. Dushman, S., op. cit. (1962), p. 388. 32. Dushman, S., op. cit. (1962), p. 384. 33. Brunauer, S., Emmett, P.H., and Teller, E., "Adsorption of Gases in Multimolecular Layers", J. Amer. Chem. Soc. 60(1), 309 (1938). 34. Stern, S.A., Mullhaupt, J.T., Hemstsreet, R.A., DiPaolo, F.S., "Cryosorption Pumping of Hydrogen and Helium at 20* K", J. Vac. Sci. Technol. 2, 165 (1965). 35. Sedgley, D.W., Tobin, A.G., Batzer, T.H., and Call, W.R., "Characterization of Charcoals for Helium Cryopumping in Fusion Devices", J. Vac. Sci. Technol. A5(4), 2572 (1987). 36. Dubinin,(1947).5)M'M" (2 Radushkevich, Proc. Acad. Sci. USSR 55, 331 37. Hobson, J.P. and Williams, B.R., "Cryosorption Pumping of Helium on Porous Silver at 4.2* K", J. Vac. Sci. Technol. 6(6), 965 (1969). 38. Hseuh, H.C., Chou, T.S., Worwetz, H.A., and Halama, H.J., "Cryosorption Pumping of He by Charcoal and a Compound Cryopump Design for TSTA", Proc. 8th Symp. on Engineering Problems of Fusion Research, 1979 (IEEE, New York, 1980), p. 1568. 39. Danilova, N.P. and Shal'nikova, A.I., "Adsorption of Helium on a Copper Surface at 4.2* K", Instrum. Experim. Tech. No. 6, 1464 (1967). 0° Hobson, J.P., "Cryopumping", J. Vac. Sci. Technol. 1...0_0(1),73 (1973). 41. Redhead, P.A., et al., op. cit., p. 27.
336
CRYOPUMPING
References 42. HiUeret, N. and Calder, R., "Ion Desorption of Condensed Gases ~, Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. R~idenauer, F.P. ViehbSck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 227. 43. Halama, H.J., Hseuh, H.C., and Chou, T.S., "Cryopumping of Hydrogen and Helium", Advances in Cryogenic Engineering, 27 (Plenum Press, Inc., New York, 1981), p. 1125. 44. Chou, T.S. and Halama, H.J., "Cryopumping of Deuterium and Helium Mixtures on Smooth 4.2"K Surfaces", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. R~idenauer, F.P. Viehb8 ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 65. 45. Hengevoss, J. and Trendelenburg, E.A., "Continuous Cryotrapping of Hydrogen and Helium by Argon at 4.2* K ~, Proc. 10th Nat. AVS Symp., 1963 (The Macmillian Company, New York, 1964), p. 101. 46. Batzer, T.H., Patrick, R.E., Call, W.R., "A TSTA Compound Cryopump", J. Vac. Sci. Technol. 18(3), 1125 (1981). 47. Tempelmeyer, K.E., Dawbarn, R., and Young, R.L., "Sorption Pumping of Hydrogen by Carbon Dioxide Cryodeposits', J. Vac. Sci. Technol. 8(4), 575 (1971). 48. Arakawa, I., Kobayashi, M., and Tuzi, Y., "Effects of Thermal Spikes on the Characteristics of Cryosorption Pumps with Condensed Carbon Dioxide Layers", J. Vac. Sci. Technol. 16(2), 738 (1979). 49. Hengevoss, J., "Influence of the Temperature History of Condensed Argon on its Hydrogen Adsorptivity at Low Temperatures", J. Vac. Sci. Technol. 6(1), 58 (1969). 50. Arakawa, I. and Tuzi, Y., "Temperature Dependence of Pumping Speed for H2 of a Cryosorption Pump with a Condensed Gas Layer Sorbent", J. Vac. Sci. Technol. A4(3), 293 (1986). 51. Thibault, J.J., Disdier, J.M., Arnaud, G., and Boissin, J.C., "Design and Preliminary Tests of the Liquid Helium Cooled MFTF Cryopumping System", J. Vac. Sci. Technol. 18(3), 1140 (1981). 52. Heafer, R.A., Cryopumping - Theory and Practice (Clarendon Press, Oxford, 1989), p. 410. 53. Dewar, Sir James, "Charcoal Vacua", Nature 12(7), 217 (1875). 54. Dushman, S., op. cit. (1949), p. 462. 55. Dushman, S., op. cit. (1949), p. 466. 56. Dushman, S., op. cit. (1949), p. 387. 57. Anderson, R.B. and Dawson, P.T., Editors, Experimental Methods in Catalytic Research (Academic Press, Inc., New York, 1976). 58. Hemstreet, R.A. and Kenmore, N.Y., U.S. Patent No. 3,296,773, "Adsorbent-Coated Thermal Panels", awarded January 10, 1967. 59. Redhead, P.A., et al., op. cit., p. 51. 60. Steele, W.A. and Halsey, G.D., "The Interaction of Gas Molecules with Capillary and Crystal Lattice Surfaces", J. Phy. Chem. 59, 57 (1955). 61. Hseuh, H.C. and Worwetz, H.A., "Performance of BNL-TSTA Compound Cryopump", J. Vac. Sci. Technol. 18(3), 1131 (1981).
CRYOPUMPING
337
References 62. Dillow, C.F. and Palacios, J., "Cryogenic Pumping of Helium, Hydrogen, and a 90% Hydrogen-10% Helium Mixture", J. Vac. Sci. Technol. 16(2), 731 (1979). 63. Fisher, P.W. and Watson, J.S., "Cryosorption Vacuum Pumping of D2, He, and H~ at 4.2"K for CTR Applications", Proc. 24th Conf. on Remote Systems Technol. (ANS, Washington, D.C., 1976), p. 12. 64. Sedgley, D.W., Batzer, T.H., and Call, N.R., "Helium Cryopumping for Fusion Applications", J. Vac. Sci. Technol. A6(3), 1209 (1988). 65. Watson, J.S. and Fisher, P.W., "Cryosorption Vacuum Pumping Under Fusion Reactor Conditions", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. Rtidenauer, F.P. ViehbSck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 363. 66. Dushman, S., op. cit. (1949), p. 484. 67. Stern, S.A., Hemstreet, R.A., and Ruttenbur, D.M., "Cryosorption Pumping of Hydrogen at 20*K II. Development and Performance of Cryosorption Panels", J. Vac. Sci. Technol. 3(3), 99 (1966). 68. Grenier, G.E. and Stern, S.A., "Cryosorption Pumping of Helium at 4.2* K", J. Vac. Sci. Technol. 3(6), 334 (1966). 69. Tobin, A.G., Douglas, W.S., Batzer, T.H., and Sedgley, D.W., "Evaluation of Charcoal Sorbents for Helium Cryopumping in Fusion Reactors", J. Vac. Sci. Technol. A5(1), 101 (1987). 70. Coupland, J.R., Hammond, D.P., B~ chler, W., and Klein, H.H., "Experimental Performance of a Large-Scale Cryosorption Pump", J. Vac. Sci. Technol. A5(4), 2563 (1987). 71. Turner, F.T. and Hogan, W.H., "Small Cryopump with Internal Refrigerator", J. Vac. Sci. Technol. 3(5), 252 (1966). 72. Forth, H.J. and Frank, R., "Installation of Cryopumps in the Estec-Space Simulation Chambers", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. Rtidenauer, F.P. VienbSck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 61. 73. Neal, R.B., "The Stanford Two-Mile Linear Electron Accelerator", J. Vac. Sci. Technol. 2(3), 149 (1965). 74. Turner, F., "Cryosorption Pumping", Varian Tech. Pub., VR-76, 1972, Varian Associates, Inc., Palo Alto, CA. 75. Baker, M.A. and Laurenson, L., "The Application of Foreline Sorption Traps", Vacuum 19(2), (1969). 76. Jepsen, R.L., U.S. Patent No. 3,116,764, "Sorption Pumping Apparatus", f'lled March 1959, awarded January 1964. 77. Gareis, P.J. and Hagenbach, G.F., "Cryosorption", Ind. and Engng. Chem. 57(5), 27 (1965). 78. Welch, K.M., "A New Dry Vacuum Roughing Pump", Research~evelopment, February, 1972, p. 42. 79. Longsworth, R.C., "Performance of a Cryopump Cooled by a Small Closed-Cycle 10-K Refrigerator", Adv. Cryo. Engng. 23, 658 (1978). 80. Bannock, R.R., "The Role of Molecular Sieve Pumping", Vacuum 12, 101 (1962).
338
CRYOPUMPING
References 81. Scott, R.B., Cryogenic Engineering (D. Van Nostrand Company, Inc., Princeton, New Jersey, 1959). 82. Barron, R., Cryogenic Systems (McGraw-HiU Book Company, New York, 1966). 83. Claude, G., British Patent No. 1,225,608, "Improvement in or Relating to Cryogenic Pumping Devices", filed August 1968, awarded March 1971. 84. Longsworth, R.C. and Lahav, Y., "A Cryopumped Leak Detector", J. Vac. Sci. Technol. A5(4), 2646 (1987). 85. Fisher, P.W. and Watson, J.S., "Cryosorption Pumping of Deuterium by MS-5A at Temperatures above 4.2 K for Fusion Applications", J. Vac. Sci. Technol. 15(2), 741 (1978). 86. Powers, R.I. and Chambers, R.M., "A Clean Cryo-Vacuum System with High Pumping Speed for All Gas Species", J. Vac. Sci. Technol. 8(1), 319 (1971). 87. Chubb, J.N., Gowland, L., and Pollard, I.E., "Condensation Pumping of Hydrogen and Deuterium on to Liquid-Helium-Cooled Surfaces", Brit. J. Appl. Phys. 1(2), 361 (1968). 88. Tobin, A.G., private communication, March 29, 1991. 89. Benvenuti, C., Blechschmidt, D., and Passardi, G., "Molecular and Radiation Transmissivities of Chevron Type Baffles for Cryopumping", J. Vac. Sci. Technol. 19(1), 100 (1981). 90. Benvenuti, C. and Calder, R.S., "The Desorption of Condensed Hydrogen from Various Substrates by Infrared Thermal Radiation", Phys. Lett. 35A(4), 291 (1971). 91. Benvenuti, C., "Characteristics, Advantages, and Possible Applications of Condensation Cryopumping", J. Vac. Sci. Technol. 11(3), 591 (1974). 92. Benvenuti, C., Calder, R.S., and Passardi, G., "Influence of Thermal Radiation on the Vapor Pressure of Condensed Hydrogen (and Isotopes) Between 2 and 4.5 K", J. Vac. Sci. Technol. 13(6), 1172 (1976). 93. Hood, C.B., "The Development of Large Cryopumps from Space Chambers to the Fusion Program", J. Vac. Sci. Technol. A3(3), 1684 (1985). 94. Graham, W.G. and Ruby, L., "Cryopumping Measurements Relating to Safety, Pumping Speed, and Radiation Outgassing", J. Vac. Sci. Technol. 16(3), 927 (1979). 95. Coupland, J.R., Hammond, D.P., Obert, W., "Experimental Performance of an Open Structure Cryopump", Vacuum 32 (10/11), 613 (1982). 60 Duffy, T.J. and Oddon, L.D., "Beam-Line Cryopump", UCRL Preprint No. UCRL-77236, 1975. 97. Langhorn, A.R., Kim, J., Tupper, M.L., Williams, J.P, and Fasolo, J., "Performance of Doublet III Neutral Beam Injector Cryopump System", J. Vac. Sci. Technol. A._.22(2),1193 (1984). 98. Redhead, P.A., et al., op. cit., p. 46. 99. Stern, S.A. and DiPaolo, F.S., "Cryosorption Pumping of Air on Molecular Sieves at 77 K - The Ultimate Achievable Vacuum", J. Vac. Sci. Technol. 6(6), 941 (1969).
CRYOPUMPING
339
References 100. Dubinin, M.M., "The Potential Theory of Adsorption of Gases and Vapors for Adsorbents with Energetically Nonuniform Surfaces", Chem. Rev. 60(1), 235 (1960). 101. Gareis, P.J. and Stern, S.A., Liquid Helium Technolom/, "Pumping of Helium and Hydrogen in the High Vacuum Range by Means of Cryosorption Arrays", Proc. Intl. Inst. of Refrigeration Commission, 1966 (Pergamon Press, Oxford, 1966), Vol. 6, p. 429. 102. Lessard, P.A., "Cryogenic Adsorption on Noncondensibles in the High-Vacuum Regime", J. Vac. Sci. Technol. A...!7(3),2373 (1989). 103. Halama, H.J., Lain, C.K., Bamberger, J.A., "Large-Capacity Cryogenic Pumping of D2 and H2 for Fusion", J. Vac. Sci. Technol. 14(5), 1201 (1977). 104. Kuluva, N.M. and Knuth, E.L., "Sorption Pumping at Pressures Less Than 10-5 Torr", Proc. 9th Nat. AVS Vac. Syrup., 1962 (The MacmiUian Company, New York, 1963), p. 237. 105. Liu, B-K., Ren, J-S., Cui, X-H., "On Cryosorption Pumping of Hydrogen with the ZDB-150 Type Cryopump Cooled by a Two-Stage Closed-Cycle Refrigerator", J. Vac. Sci. Technol. 20(4), 1000 (1982). 106. Visser, J., Symersky, B., Geraerts, A.J.M., "A Versatile Cryopump for Industrial Vacuum Systems", Vacuum 27(3), 175 (1977). 107. Endow, N. and Pasternak, R.A., "Physisorption of Xenon and Krypton on Glass and on Molybdenum Films", J. Vac. Sci. Technol. _3(4), 196 (1966). 108. Troy, M. and Wightman, J.P., "Physisorption of Ar, Kr, CH4, and N2 on Stainless Steel at Very Low Pressures", J. Vac. Sci. Technol. 8(6), 743 (1971). 109. Be, S.H., "New Cryopumps for a Resonator of the RIKEN Ring Cyclotron", Vacuum 38(7), 543 (1988). 110. Santeler, DJ., "Pressure Simulation of Outer Space", Proc. Sixth National Symposium on Vacuum Technology Transactions, 1959 (Pergamon Press, New York, 1960), p. 129. 111. Levenson, L.L., Milleron, N., and Davis, D.H., "Optimization of Molecular Flow Conductance", Proc. Seventh National Symposium on Vacuum Technology Transactions, 1960 (Pergamon Press, New York, 1961), p. 372. 112. Caren, R.P., Gilcrest, A.S., Zierman, C.A., "Thermal Absorptances of Cryodeposits for Solar and 290"K Blackbody Sources", Proc. 1963 Cryogenic Engineering Conference, 1963 (Plenum Press, New York, 1964), Advances in Cryogenic Engineering, Vol. 9, p. 457. 113. Cunningham, T.M. and Young, R.L., "The Radiative Properties of Cryodeposits at 77* K", Proc. 1962 Cryogenic Engineering Conference, 1962 (Plenum Press, New York, 1963), Advances in Cryogenic Engineering 8, p. 85. 114. Merriam, R.L. and Viskanta, R., "Radiative Characteristics of Cryodeposits for Room Temperature Black Body Radiation", Proc. 1968 Cryogenic Engineering Conference, 1968 (Plenum Press, New York, 1969), Advances in Cryogenic Engineering, Vol. 14, p. 240.
340
CRYOPUMPING
References 115. Welch, K.M., U.S. Patent No. 4,311,018, "Cryogenic Pump", filed 9/22/80, awarded 1/19/82. 116. Welch, K.M., U.S. Patent No. 4,295,338, "Cryogenic Pumping Apparatus with Replaceable Pumping Surface Elements", filed 10/18/79, awarded 10/20/81. 117. Welch, K.M., U.S. Patent No. 4,285,710, "Cryogenic Device for Restricthag the Pumping Speed of Selected Gases", f'fled 9/18/78, awarded 8/25/81. 118. Klein, H.-H., Heisig, R., and Augustine, C.M., "Use of RefrigeratorCooled Cryopumps in Sputtering Plants", J. Vac. Sci. Technol. A2(2), 187 (1984). 119. Bentley, P.D., "The Modern Cryopump", Vacuum 30(4/5), 145 (1980). 120. Haarhuis, G.J., "Stirling Cryogenerators and Cryopumping", LeVide No. 171-172, 351 (1974). 121. Visser, J. and Scheer, J.J., "Twenty-Kelvin Cryopuming in Magnetron Sputtering Systems", J. Vac. Sci. Technol. 16(2), 734 (1979). 122. Visser, J. and Scheer, J.J., "A Cryosorption Pumped Rapid Cycle UHV System", J. Vac. Sci. Technol. 19(1), 122 (1981). 123. McMahon, H.O. and Gifford, W.E., "A New Low-Temperature Gas Expansion Cycle", Advances in Cryogenic Engineering 5 (Plenum Press, Inc., New York, 1960), p. 354. 124. Gifford, W.E., "The Gifford-McMahon Cycle", Advances in Cryogenic Engineering 11 (Plenum Press, Inc., New York, 1966), p. 152. 125. Becker, G.E., "Operation of a Cryopumped UHV System", J. Vac. Sci. Technol. 14(1), 640 (1977). 126. Benini, M. and Morbidi, P., "Performance of Closed-Cycle Refrigerator Pump", Proc. 8th Int. Vac. Cong., 1980 (Suppl~. ment ~. la Revue, "LeVide, les Couches Minces, No. 201", 1981), p. 275. 127. Bridwell, M.C. and Rodes, J.G., "History of the Modern Cryo-pump", A3(3), 472 (1985). 128. Dennison, R.W. and Gray, G.R., "Cryogenic Versus Turbomolecular Pumping in Sputtering Applications", J. Vac. Sci. Technol. 16(2), 728 (1979). 129. Denision, D.R., "Characteristics of Cryo- and Cryo/Ion Pumped Vacuum Systems", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. R~idenauer, F.P. Viehb6ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 69. 130. de Rijke, J.E., "Performance of a Cryopump-Ion Pump System", J. Vac. Sci. Technol. 15(2), 765 (1978). 131. de Rijke, J.E., "Factors Affecting Cryopump Base Pressure", J. Vac. Sci. Technol. A8(3), 2778 (1990). 132. Hands, B.A., "Recent Developments in Cryopumping" Vacuum 32(10/11), 603 (1982). 133. Longsworth, R.C., "Characteristics of a 550-mm ID Low Profile Refrigerated Cryopump", Advances in Cryogenic Engineering 27, 1980 (Plenum Press, Inc., New York, 1981), p. 1135. 134. Longsworth, R.C.,. "Refrigerators for Cryopumps" for AVS Course "Fundamentals of Cryopumping", 1978.
CRYOPUMPING
341
References 135. Longsworth, R.C., "Characteristics of Cryopumps Cooled by Small Closed-Cycle 10-K Refrigerators", Advances in Cryogenic Engineering 23 (K.D. Timmerhaus, Editor, Plenum Press, New York, 1978), p. 658. 136. Minata, M. and Stolz, J., "System Level Vibration of a Low-Vibration Cryopump for Semiconductor Processing Equipment", J. Vac. Sci. Technol. A7(3), 2361 (1989). 137. Muhlenhaupt, R.C., "A Comparison of Cryopump and Diffusion Pump Performance on MMA 29x45 ft. Vacuum Chamber", J. Vac. Sci. Technol. 20(4), 1005 (1982). 138. Ikegami, K., Nakajima, S., Be, S.H., Ohsako, N., Morimoto, K., Kikuchi, T., and Morishita, S., "Design and Performance Characteristics of Refrigerator-Cooled Cryopumps for the RIKEN Ring Cyclotron", Vacuum 38(2), 99 (1988). 139. Ruthlein, H. and Forth, H.J., "The Application of Cryopumps with Integrated K 20 Refrigerators in Vacuum Deposition Technology", Proc. 7th Int. Vac. Cong. and 3rd Int. Conf. on Solid Surfaces, 1977 (R. Dobrozemsky, F. R~idenauer, F.P. Viehb6ck, A. Breth, Postfach 300, A-1082 Vienna, Austria, 1977), p. 77. 140. Sch/ifer, G. and H~ifner, H.-U., "Cryopumps for Evaluating Space Simulation Chambers", J. Vac. Sci. Technol. A5(4), 2359 (1987). 141. Welch, K.M. and Flegal, C., "Elements of Cryopumping", Industrial Research~evelopment, March, 1978. 142. Welch, K.M. and Longsworth, R.C., An Introduction to the Elements of Crvopumping (K.M. Welch, Editor, Varian Associates, Palo Alto, CA, 1980). 143. Leever, R.C., "A Low Temperature Mechanically Refrigerated Cold Trap", Proc. 1st Nat. AVS Symp., 1954 (W.M. Welch Manufacturing Company, 1955), p. 19. 144. Kikuchi, T., Ohsako, N. and Hayashi, Y., "Capability of Obtaining Extreme High Vacuum by Commercial G-M Refrigerator-Cooled Cryopump", Vacuum 41(7/9), 1941 (1990). 145. Welch, K.M., Mclntyre, G.T., Tuozzolo, J.E., Skelton, R. Pate, D.J., and Gill, S.M., "Metal and Elastomer Seal Tests for Accelerator Applications", Vacuum 41(7/9), 1924 (1990). 146. Longsworth, R.C., and Booney, G.E., "Cryopumping Regeneration Studies", J. Vac. Sci. Technol. 21(4), 1022 (1982). 147. H/i frier, H.-U., Klein, H.-H., and Timm, U., "New Methods and Investigations for Regenerating Refrigerator Cryopumps", Vacuum 41(7/9), 1840 (1990). 148. Lefrancois, M., Montuclard, J., and Rouaud, C., "Low-Load Metal Seals to Replace Elastomer O-rings: the Helicoflex-delta (A) Seals", Vacuum 41(7/9), 1879 (1990). 149. The author is indebted to Mr. John T. Harvell and Mr. John Peterson, of CTI, who were interviewed in April, 1990, to obtain some of this historical data. Also, I am indebted to Dr. Ralph Longsworth and Mr. Richard L. Rerig for a similar interview in May, 1990. All have been my friends for many years, even with recorder in hand.
342
CRYOPUMPING
References 150. Barron, R., op. cit. p. 81. 151. Smith, J.L. and Robinson, G.Y., "A Tribute to Samuel C. Collins: September 28, 1898-June 19, 1984", Advances in Cryogenic Engineering 31 (R.W. Fast, Editor, Plenum Press, New York, 1986), p. 1. 152. Barron, R., op. cit., p. 122. 153. Chellis, F.F. and Hogan, W.H., "A Liquid-Nitrogen-Operated Refrigerator for Temperatures Below 77* K", Advances in Cryogenic Engineering 9_, (K.D. Timmerhause, Editor, Plenum Press, New York, 1964), p. 545. 154. Harvell, J.T., private communications, April, 1990. 155. Buck, D.A., "The Cryotron: A Superconductive Computer Element", Proc. Inst. Radio Engrs. 44(4), 482 (1956). 156. O'Boyle, T.F., "Chilling Tale, GE Refrigerator Woes Illustrate the Hazards in Changing a Product"", The Wall Street Journal, Eastern Edition, May 7, 1990, p. 1. 157. Mr1ller, D., private communications, June, 1990. 158. Sondericker, J., "Production and Use of High Grade Silicon Diode Temperature Sensors", Advances in Cryogenic Engineering 27, 1980 (Plenum Press, Inc., New York, 1981), p. 1163. 159. Welch, K.M., Andeen, B., de Rijke, J.E., Foster, C.A., Hablanian, M.H., Longsworth, R.C., Millikin, W.E., Sasaki, Y.T., Tzemos, C., "Recommended Practice for Measuring the Performance and Characteristics of Closed-Loop Gaseous Helium Cryopumps", J. Vac. Sci. Technol. A 17(5), 3081(1999). 160. Wallen, E., "Adsorption Isotherms of H ~ and mixtures of H2, CH4, CO, and CO2 on copper plated stainless steel at 4.2 K", J. Vac. Sci. Technol. A 14(5), 2916(1996). 161. Wallen, E., Adsorption Isotherms of He and H~ at Liquid He Temperatures", J. Vac. Sci. Technol. A 15(2), 265(1997). 162. Hobson, J.P., "First Adsorbed Layer of Helium at 4.2 ° K", Can. J. Phys. 37 300(1959). 163. Edwards Jr, D., Limon, P., "Pressure Measurements in a Cryogenic Environment", J. Vac. Sci. Technol 15(3), 1186(1978). 164. Hobson, J.P., Welch, K.M., "Time-Dependent Helium and Hydrogen Pressure Profiles in a Long, Cryogenically Cooled Tube, Pumped at Periodic Intervals". J. Vac. Sci. Technol. A 11(4), 1566(1993). 165. Hseuh, H-C., Wallen, E., "Measurement of Helium Propagation at 4.4 K in a 480 m Long Stainless Steel Pipe", J. Vac. Sci. Technol. A 16(3), 1145(1998). 166. Wallen, E. "Experimental Test of the Propagation of a He Pressure Front in a Long, Cryogenically Cooled Tube", J. Vac Sci. Technol. A 15(6), 2949(1997). 167. Iwasa, Y., Ito, S., "A New Type of Cryopump with a Metal Cryopanel Cooled Below 3.6 K by a two-stage GM Refrigerator", Vacuum, 47(6-8), 675(1996). 168. Hashimoto, T. Li, R., Kuriyama, T., Nagao, M., "Recent Progress in Applications of Magnetic Regenerator Material", Cryogenic Engineering, 31, 131(1996). 169. Crone, D.W., Brady, J., Foreman, M.O., Welch, K.M., "A Highly Simplified and Reliable means of Regenerating Closed-Loop Gaseous Helium Cryopumps", Presented at the 46th International Symposium of the AVS, Oct. 1999, unpublished.
CRYOPUMPING
343
References 170. Kuriyama, T., Takahashi, M., Nakagome, H., Hashimoto, T., Nitta, H., Yabuki, M., "Development of 1 Watt Class, 4 K GM Refrigerator with Magnetic Regenerator Materials", Advances in Cryogenic Engineering 39, (Plenum Press, New York, 1994), p. 1335. 171. Radebaugh, R., "Advances in Cryocoolers", ICEC16/ICMC Proceedings (Elsevier Science, Amsterdam, 1997), p. 33. 172. Davis, R.P., Abreu, R.A., Chew, A.D., "Dry Vacuum Pumps: A Method for Evaluation of the Degree of Dry", J. Vac. Sci. Technol. A 18(4), 1782(2000). 173. Mundinger, H-J, H~ifner, H-U, Mattern-Klosson, M., Klein, H.H., Timm, U., "A New Cryopump with a Very Fast Regeneration System", Vacuum 43(5-7), 545(1992). 174. Kiese, G., Mundinger, H-J, Voss, G., "Increasing Uptime of Cryopumps with Fast Intelligent Regeneration System Technology", Vacuum 44(5-7), 693(1993). 175. Sarcia, D.S., U.S. Patent No. 4,339,927, "Gas-Driven Fluid Flow Control Valve and Cryopump Incorporating Same", filed July 6, 1981; issued July 20, 1982. 176. Okumura, K., Kuriyama, F., Murai, Y., Tsujimura, M., U.S. Patent No. 5,062,271, "Evacuation Apparatus and Evacuation Method", filed May 4, 1990; issued Nov. 5, 1991. 177. Gaudet, P.W., Eacobacci, Mj., Harvell, J.T., Lepofsky. R.J., Roche, D.E., Bender. S.A., "Electronically Controlled Cryopump", filed May 20, 1991; issued Oct. 27, 1992. 178. Meissner, R.C., "A High Vacuum Laboratory for the Vapor Deposition of Conductors and Dielectrics", Vacuum Symposium Transactions (Pergamon Press, London, 1956), p. 15. 179. Missimer, D.J., U.S. Patent No. 3,768,273, Self-Balancing Low Temperature Refrigeration System", filed Oct. 19, 1972; issued Oct. 30, 1973. 180. Missimer, Dj., U.S. Patent No. 4,176,526, "Refrigeration System Having Quick Defrost and Re-Cool", filed Sept. 22, 1978; issued Dec. 4, 1979.
SUBJECT INDEX A Accommodation coefficient, 256 Adsorbate, see Sorbate Adsorbent, see Sieve Materials Adsorber, He Compressor, 326-328 Adsorption isosteres, helium on activated charcoal, 308 hydrogen on activated charcoal, 308 Adsorption isotherm, 2.64 Ar, Kr, Ke, 270 BET method, 269, 270 classification of, 267 DR, 269 example of, 266 Freundlich, 2 ~ He, H 2, Ne on Linde 5A sieve, 290 He on 4.2* K surfaces, 271-273 H 2 on 4.2* K surfaces, 272 H 2 0 on Linde 5A sieve material, 293 Henry's Law, 2 ~ hyperbolic, 269 isobar, definition of, 2 ~ isostere, definition of, 264 isotherm, def'mition of, 2 ~ limits of surface coverage, 271 N 2 on Linde 5A sieve, 288 parabolic, 2 ~ surface conductance limited, 2 ~ time dependency, 289, 290 zero speed intercept, 267 Author Index, 353
Bakeout, sorption pumps, 292 Brunauer, Emmett and Teller, see Adsorption Isotherm, BET, 269, 270 C Calibration, Vacuum Gauges, 59-61 Capacity, TSP film, "engineering values", 208 parameters effecting, 204 pealing of, 222 Capture coefficient, 255 CO 2 v s . flux intensity, 263 CO 2 v s . temperature, 263 Cathode materials, SIP aluminum, 101, 102, 119, 121, 134-136, 139 body centered, cubic, 121-131 bubble migration in, 141 chemisorption, 99-102 diffusion into, see Diffusion gas v s . material, 101 hexagonal, closed packed, 121 high energy neutrals, from, 115-117 oxides on, 101 sputter-yield, 94-96 Ta and Ti cathodes, 114, 115 test vehicle, 121-123 Ti-6 A~-4V, 124, 125 "unobtainium", 123, 124 Chemisorption, active sites, 99 affinity, 99 chemical bond, 99 cathode material vs. gas, 102 chemical, 101 dissociative pumping, 99 potential energy diagram, 98
346
SUBJECT INDEX
Chewon design, cryopump, 298-301 louver, 299-301 transmissivity, molecular, 301 transmissivity, optical, 301 Chevron, Santler, 300 Clausius-Clapeyron constants, 259 Clausius-Clapeyron equation, 258-259 Closed-loop, gaseous helium cryopumps, see CLGHe Cryopumps CLGHe cryopumps, applications, 312, 313 argon plugging of sieve, 303, 304, 306, chevron design, 303-306 compressor, helium, 326-328 design elements, 301-302 expander, 323-325 dialysis effect, 328 helium speed, 303-305 history, brief, 321-325 hydrogen speed, 303, 305 placebo effect, 319, 320 pressure clamping effect, 297, 312 protection of sieve material, 303-305 Recommended Practices, AVS, see reference (1 s 9 ) refrigerator, Gifford-McMahon, 323-328 regeneration of, 315-320 safety, 313-315 second stage array, 305 sputtering applications, 309-312 variable aperture, 308 fixed aperture, 310 sticking coefficients, 306 system configuration, 313-315 thermal loading of, 309-310 CLGHe refrigerators, 323-329 Coconut charcoal, activated, see Sieve Materials Combination pumping, TSP, 204, 220, 221
Conductance, abstraction, 20, 21 definition, 20 electrical analog, 28-29 geometrical approximations, 23 species dependency, 24-25 temperature dependency, 24-25 combining in series, 26 Cross field mobility, 88 Cryocondensation, def'mition of, 255 Clausius-Clapeyron equation, 258, 259 sticking coefficient, 2~, 306 surface conductance limited, 261 vapor pressure, 258, 259 various gases, references, 274 Cryopumping, 255-331 speed and capacity, 267, 274-276 Cryopumps, high pressure operation of, 296, 307-312 Cryosorption, amount vs. gas boiling point, 257 definition of, 256 examples of, 274 vs. cryocondensation, 255 Cryotrapping, 276, 277 definition of, 276 examples of use of, 274 He with Ar jet spray, 299 D Dalton's Law, pumping speed, 28 Density, gas molecules/unit volume, 5,6 Diagnostics, system, 42-44 Diffusion, cathode, bubble migration, 141 bulk laminate, 129-132 concentration gradient, 129 hydrogen, into mixed o~-/5' cathodes, 132, 133 hydrogen, into/5'-stabilized cathodes, 133, 134 hydrogen, into Ti cathodes, 130, 131
SUBJECT INDEX Diffusion, cathode, implant laminate, 129 non-Fickian, 130 pseudo-Fickian, 130 Diffusion model, cathodes, 127-134 Diffusivity, 128 neutron radiography, 133 Distributed sputter-ion pump, 156-158 Dubinin and Radushkevich see Adsorption Isothern, DR E Efficiency of pumping, Penning cell, r/, 89, 92, 109, 126 Elastomers, vacuum seals see Vacuum Seals Electrical circuit analogies, linear superposition, 28, 29 pressure, 25 speed and conductance, 25 throughput, 25 Electrostatic pumps, 83 bibliography, 177-179 Equilibrium vapor pressure, see Vapor Pressure Fick's First Law, 128, 129 Fick's Second Law, 128 Field emission, Fowler Nordheimplot, 114 Fischer-Mommsen speed dome, 43 Freundlich isotherm see Adsorption Isotherm Fusion, heat of, 259 G Gas, surface flux or intensity, 16 thermal conductivity, 15 molecular velocities, 7, 8 Gases, electrical discharges in, 48 ionizing, high pressure, 46-50 Townsend discharge in, 46 Gauge, Calibration Methods, 59-61 Gifford-McMahon refrigerators, 323-328
347
H Helium Leak Detection system applications, 71-76 response/sensitivity trade-offs, 94, 95 with capture pumps, 76 Henry's Law, 2 ~ High energy neutral theory, cathode material, 115-117 gas species, 115-117 gas pumping, 114-117 High voltage power supplies, sputter-ion pump, battery operated, 103 capacitance of high voltage cables, 160 soft transformer, 159 switching, 158 Hydrogen pumping speed, sputter-ion pump, 119-142 diode pumps, 121-138 reduction due to oxides and nitrides, 119-120 titanium, 119 Hyperbolic isotherm, see Adsorption Isotherm
Inert gas pumping, see Sputter-Ion Pumps, Noble Gas Pumping Ionization of gases, cross section, 51 high pressure, 46-50 low pressure, 50-53 Townsend discharge, 46 K Kirchoff's First Rule, see Throughput
Langmuir, definition of, 2 ~
348
SUBJECT INDEX
Leak Detection see helium leak detection Liquid helium cryopumps (LHC), 295 - 299 boiling pool, 296 classification of, 299 cryogen-fed, 295, 296 M
Magnetic field, sputter-ion pump, diode vs. triode, 113, 114 Magnets, A~-Ni-Fe-Co alloys, 160 design of, 161 ferrite magnets, 161 high gradient, see reference horseshoe vs. confined field, 161-164 samarium-cobalt, 162 very high field, 91, 92, 173 Mean free path, def'mition of, 14 Memory effects, 105, 106, 152 Metal seals, Conflat~, 65-67, 103 Helicoflex Delta @, 69-71, 313 Molecules, counting, 5
mean free path, 12, 13, 14 mole, 4, 5 velocity, 6, 7, 8 Monte Carlo calculations, chewon, 300 definition of, 300 N Noble gas instabilities, pressure wave functions, 105, 106 sinusoidal pressure instabilities, 101 Nonevaporable getters, activate, power to, see Chap. 4 problems 10 and 11 activation of, 229-231, 234, 235 activation, St. 101#, 233, 234 activation St. 707~, 234
Nonevaporable getters, adsorption isotherm, 237-239 advantages and disadvantages, 248, 249 capacity, bulk, 234, 235 capacity, surface, 234 embrittlement limit, 235 exposure of St. 101® material to air, 243 exposure of St. 707~ material to air, 243-245 gas mixture, pumping, 241-245 hydrocarbon pumping, 240 hydrogen pumping, 235-240 mechanical features, 229, 230 pellet, 231 plugging gases, N 2, CO, CO 2, 235-237 porosity, sintered matrix, 231 pumping mechanisms, 229-246 reconditioning vs. activation, 242, 243 Sievert's Law, 237-240 Sievert's Law, departure from, 238, 239 sintered NEG structures, 246, 247 speed, decay due to gases containing carbon vs. oxygen or nitrogen, 244 speed, decay when pumping N2, CO, CO 2,241 speed, hydrocarbons, 240 speed measurements, possible errors, 245 St. 101®, composition, 221 St. 707~, composition, 221 strip material, 230 surface area, effective, 232, 233 system, venting to dry nitrogen, 246 temperature activation of, 233, 234 O Outgassing, long manifold, 31
SUBJECT INDEX
Parabolic isotherm, see Adsorption Isotherm Paschen's Law, 49, 50 Partial pressure gauges, 53-59 Pascal, conversion to Torr, 4 Penning cell, aspect ratio, 92, 93 cathode material, 87 cold cathode, 85 current vs. pressure, 88, 89 current vs. throughput, 89 end effects, 92 with hot filament, 84 inverted magnetron, 85 magnetic field, 87 magnetron, 86, 169, 170 parameters, 90 pumping effects, 84, 86 pumping efficiency, r/, 89, 92, 109, 126 pumping speed abstraction, 93, 94 sensitivity, 88, 89 (see also Sensitivity) sensitivity, I / B vs. voltage, 91 sensitivity, l/P, 89 sensitivity, I / P vs. anode diameter, 93 sensitivity, I / P vs. cell length 92 sensitivity, I / P vs. magnetic field, 90-93 space charge within, 88, 90, 91 speed, 87 speed vs. sensitivity, 89 sputter-coating, 84, 85 striking, 87, 88, 172 triode, 87 voltage, 87 Penning discharge, 85 Penning discharges, discharge modes, 165, 165 noise from, 171, 172 rf cavity, space charge from frequency shift, 166, 167 smooth-bore magnetron model, 169
349
Penning discharges space charge, deduction of from cathode current density, 167, 168 space charge, deduction of with probes, 168, 169 space charge distribution, 164171 Stark effect, space charge from, 165, 166 striking characteristics, 172, 173 Phonons, definition of, 278 H 2 desorption, due to, 300, 301 Physisorption, 97-99 binding energy, 98, 99 heat of adsorption, 98, 99 nitrogen pumping, 100 potential well, 98 van der Waal's force, 97-99 Pockets or antechambers, sputterion pump within, 154-158 Polarity, molecular, 252, 253 Pressure, definition, 2, 3, 6 profile in long manifold, 31 Torr to Pascal conversion, 4 voltage analog, 25 - 29 Pumpdown equations, 33 - 36 Pumping on liquid cryogen, 12 Pumping, surface, 11 Pumps, capacity, quantity of gas pumped, 1 capacity, throughput, 1 capture, definition, 1 cryopumps, 255 - 343 momentum transfer, definition, 1 nonevaporable getter, 229-253 sputter-ion, 83- 193 titanium sublimation, 195-227 Pumps, electronic electro-ion, 83 electrostatic, 83, 177-179 evapor-ion, 83 orbitron, 83
350
SUBJECT INDEX R
Rate-of-pressure rise, 43
Saturation, sputter-ion pump, see Speed, Sputter-Ion Pump, Transient Sensitivity, hysteresis at high pressures, 147-148 triode I/P, 92 Sensitivity, I/P, 53, 86, 87 Sieve materials, artificial zeolite, 279 bonding to surfaces, 283, 284 coconut charcoal, activated, 278 H~ on, 10-30" K, 283 effective surface area, 279 labyrinth effect, 280 persorption, 280, 281 plugging of, 280-281 porous silver, 270 use of, references, 281 Sorbate, definition of, 278 Sorbent, definition of, 278 Sorption roughing pumps (SRP), 284-295 applications, 284, 285 construction of, 285-287 dewars and bakeout, 288, 293 neon backstreaming, 291, 292 patent of, 285 safety, 286, 293-295 staging, definitions of, 287 staging of, 287-292 Speed, effect of blank-off pressure, 35 electrical analog, 25, 26 measuring, 33-41 pumpdown with outgassing, 35 rate of pumpdown, 33, 34 selective and variable, 32, 33 Speed and capacity of cryopumps, see Cryopumping
Speed dome, Fischer-Mommsen, 39, 40 modified Fischer-Mommsen, 40,151 single gauge, 36, 37 three-gauge, 37-39 three gauge vs. FischerMommsen, 41 Speed, pumping, an abstraction, 17 definition, 17-19 Speed, sputter-ion pump, transient, 142, 143 Sputtering, angle of incidence, 95 Cosine Law, 96, 97 slotted cathodes, sputter-yield, 108, 109 sputter-yield, 95 yield of Ar vs. mass of target material, 95, 114, 115 Sputtering application, see CLGHe Cryopumps Sputter-ion pump, anode, sputtering onto, 95, 96 battery operated high voltage supply, 90, 103, 104 cathode materials, 99-102 cathode, oxide barrier, 101 cathodes, slotted, 108, 109 cathode sputtering, 100, 114, 115 chemisorption, 94, 99-102 clean pumping, 102 first commercial multi-cell pump, 103, 104 high energy neutral theory, 115-118 high pressure operation, 145-149 hydrogen pumping, 101, 119-142 low pressure operation, 136, 137,143, 149 magnetic field, diode vs. triode, 113, 114 mixing Noble and conventional pump elements, 118, 119 multi-cell assumption, 88 nitrogen pumping, 85 Noble gas instabilities, 91
SUBJECT INDEX Noble gas instabilities with distributed pumps, 106-108 physisorption, 94, 97-99 pockets in pumps, 104, 105 pumping mechanisms, 94 sputtering, 94-96 thermal run-away, 139, 145-149 triode, 110-113 unconfmed discharge, 105 see also, Cathode Materials Sputter-ion pump, advantages and disadvantages, capacity, 175 operating efficiency, 175 pump starting, 176 safety considerations, 176 source of system contaminants, 162 Sputter-ion pump, cathode Galaxy~ diode, 109-110 materials, see Cathode Materials Sputter-ion pump, maintenance and troubleshooting, cleaning, 174, 175 how not to clean, 148, 149 maintenance and troubleshooting, 173-175 Sputter-ion pump, Noble gas pumping, cathodes with two materials, 114-119 differential ion pumping, 114 high energy neutral theory, 115-118 slotted cathodes, 108-110 Galaxy# diode, 109-110 Sputter-ion pump, Noble gas, triode, 110-112 Sputter-ion pump, triode, A;. cathodes, 134, 135, 139-142, 152 cellular cathodes, 111 field emission, 114 gridded cathodes, 110-114 magnetic field diode vs. triode, 113, 114 pumping mechanisms, 138-143
351
Sputter-ion pump, triode StarCell ®, 112, 113 Galaxy~, 109, 110 thermal run-away, one vs. two voltage potentials operation, 111, 112 Sticking coefficient, TSP, engineering values, 208 measurement of, 196-200 parameters effecting, 204-208 Sticking coefficient, cryopumps, 260, 306 Sublimation, heat of, 259
Temperature, definition of, 2, 3 ice and steam points, 3 Temperature measurement, hydrogen vapor bulb, 325, 326 silicone diodes, 326 Thermal conductivity, gas, see Gas Thermal run-away, diode, 145-149 triode, 113 Thermal transpiration, 261, 262 Thermomolecular effect, 261 Throughput, cryopumps, 308 electrical current analog, 25, 26 Kirchoff's First Rule, 27 rate-of-pressure-rise, 43, 44 sputter-ion pump, 146 titanium sublimation, 204, 205 Titanium sublimation pumping, advantages and disadvantages, 222, 223 alternate materials, 195 batch sublimation, 200 conductance limited, 203, 204 dissociation, 208-210 E-gun sources, 218-220 filamentary sources, 211-216 constant current, 212 constant voltage, 214-216 titanium-molybdenum alloy, 212
352
SUBJECT INDEX
Titanium sublimation pump twisted bundles, 211, 212 gas displacement, 208-210 gettering, 195 memory effect, 210 radiantly heated sources, 216-218 safety, 221, 222 sources, 210-220 speed of film, 201 sticking coefficient, measurement of, 196-200 synthesis of gases, 208-210 titanium limited, 203, 204 Townsend discharge, 46 Transmissivity, see Chevron Design Traps, sieve, 285 Triode, sputter-ion pump, burial of gas in pump walls, 138 gridded cathode, 113 pumping mechanisms, 112, 138 thermal runaway, 113 Trouble-shooting system, 42-44 V Vacuum gauge, ionization Bayard-Alpert, 53 cold cathode, 53 partial pressure, 53-58 sensitivity, 52-53 Vacuum Seals, 62-71 Vaporization, heat of, 259 Vapor pressure, 8, 9 of common gases, 10, 11 Z Zeolite, artificial, see sieve materials
AUTHOR INDEX A Abreu, 318 Adam, 83, 177 Agdur, 167 Aggus, 2~, 270, 279, 281, 289, 296, 305 Alexeff, 177 Alpert, 55 Alger, 205, 207 Allen, 209 Allison, 119, 175 Allman, 112, 116, 135 Alton, 112, 116, 135 Anashin, 157 Andeen, 255, 302, 304, 306 Andersen, H.H., 95, 127 Anderson, R.B., 279 Andrew, 105, 108, 112, 135, 178 Arakawa, 274, 277, 281 Arden, 178 Armour, 97, 119 Arnaud, 281, 298 Atkins, 176, 221 Auciello, 240 Audi, 94, 112, 120, 140, 144, 145, 158, 231 Augustine, 309 Auslender, 157 B
B~ichler, 83, 177, 281, 284, 298 Bachrach, 17 Baker, L.C., 139, 145 Baker, M.A., 285 Baker, P.N., 118 Ball, 42 Bamberger, 274, 281 Bance, 139, 142, 145, 175, 176 Bannock, 292, 293
Barnes, 141 Barret, 128 Barron, 12, 258, 295, 321 Bartholme, 119, 121 Batzer, 269, 274, 277, 279, 281, 282, 284, 289, 298, 299 Bay, 95 Be, 281, 312, 313 Beanland, 127 Becker, 312, 313 Bell, 176 Bender, 157, 330 Benini, 312, 313 Bennett, G.H.J., 140, 146 Bennett, G.W., 139 Bennett, M.J., 261 Bentley, 256, 2~, 274, 311 Benvenuti, 12, 231, 232, 233, 240, 242, 243, 248, 249, 274, 277, 289, 300, 301 Berman, 94 Berry, 110 Bills, 178, 179 Biryukova, 202, 205, 206 Biscardi, 68, 70 Bleschschmidt, 158, 173, 289, 301 Blinov, 157 Blodgett, 94 Bloomer, 114 Boffito, 229, 231, 234, 235, 236, 237, 238, 240, 241, 242, 243 Boissin, 281, 298 Bojon, 158, 173 Bond, 100, 101, 102 Booney, 315 Bowden, 94 Boyde, 46 Brandon, 94 Brady, 318 Brewer, 214 Bridwell, 312, 313, 323 Brillouin, 170
354
AUTHOR INDEX
Brodie, 116 Brombacher, 39 Brothers, 123, 149, 221 Brouet, 231 Brown, R.F., 256, 263, 274 Brubaker, 95, 105, 110, 121 Bruining, 48 Brunauer, 269, 279 Bryant, 172 Buck, 321 Burden, 219 Buritz, 55 Burnisenico, 101, 103 Burthe, 218, 221 Busby, 256, 263, 274 Butcher, 14 B/i ttner, 2~, 274, 277, 281 C Calder, 141, 274, 276, 277, 281, 298, 300 Callin, 107 Call, 269, 274, 276, 281, 282, 289, 298, 299 Caren, 307 Carlson, 103 Carter, 97, 108, 119 Casimir, 160 Cecchi, 231, 233, 235, 237, 240, 242 Celik, 141 Chambers, 23, 281, 296 Chapman, 158, 173 Charlson, 14 Chellis, 322 Chen, C.H., 112, 116, 135 Chen, F.F., 88 Chen, J.R., 158 Chew, 318 Chida, 158 Chiggiato, 248 Chou, 92, 175, 213, 270, 276, 277, 281, 284, 296, 305 Chubb, 256, 264, 274, 298, 300 Cicoira, 248 Clarke, P.J., 110 Claude, 295 Clausing, 23, 195, 197, 199, 202, 203, 205, 206, 211
Cloza, 246 Cobine, 47, 49, 84 Colligon, 47, 119 Collins, 321 Conn, 53, 87, 146, 166, 171, 172 Coupland, 281, 284, 298 Cox, 114 Craig, 175, 176 Crank, 130, 132 Crawley, 64 Crone, 318 Cui, 281 Cummings, 120, 157, 158 Cunningham, 307 D Daglish, 53, 87, 146, 166, 171, 172 Dallos, 88, 114, 117, 142, 149 Danilova, 271, 272, 274, 279 Davis, D.H., 300 Davis, F.N., 261 Davis, R., 110 Davis, R.H., 83, 177, 195 Dawbarn, 277 Dawson, 263, 274, 279, 281, 306 Dayton, 62, 64 De Angeli, 239 Dean, 120, 157, 158 De Biasio, 118 de Chernatony, 172 de Csernatony, 64 Decroux, 232, 242 Della Porta, 178, 229, 240, 242 de Rijke, 255, 302, 304, 306, 312, 313 De Simon, 112, 120 Denison, D.R., 41, 139, 178 Dennison, R.W., 312, 313 Dewar, 278 Dillow, 281, 282, 296 DiPaolo, 2~, 269, 270, 281, 282, 289, 291, 292, 293 Disdier, 281, 298 Divatia, 177, 195 Dolcino, 112, 114, 231 Domeij, 199 Dominoini, 246 Donachie, 123, 127 Donaldson, 99
AUTHOR INDEX Doni, 112, 231, 235, 236, 241, 242, 243 Douglas, 178, 281, 284 Dowling, 176, 221 Downing, 46 Dow, 88, 169 Driscoll, 171 Dryden, 177 Dubinin, 269, 271 Duffy, 274, 298 Dushman, 15, 16, 119, 127, 128, 140, 237, 238, 256, 257, 259, 2~, 278, 279, 280, 281, 282 Dylla, 231, 233, 235, 237, 242 E Ecobacci, 330 Edmonds, 261 Edwards, D., 209, 273 Efimov, 202, 205, 206 Eisenstadt, 259 Elsworth, 203, 205, 206 Emerson, 231, 237, 240 Emmett, 269, 279 Endow, 274 Englander-Golden, 51, 52, 144 Ermrich, 177, 209 Eschback, 141
Falland, 158 Fang, 102 Fasolo, 274, 298 Fast, 119 Fedorov, 119 Feidt, 179 Feigenbaum, 232, 240 Ferrario, 112, 178, 195, 229, 231, 234, 235, 236, 237, 238, 240, 241, 242, 243 Feynman, 94 Fischer, 39, 40, 157 Fisher, 83, 178, 211, 216, 281, 282, 296, 297 Filippelli, 53 Fitch, 23 Flegal, 39, 312, 313 Foreman, 318
355
Forth, 284, 312, 313 Foster, 255, 302, 304, 306 Francia, 231, 232 Francis, 105, 106, 108, 109, 119, 175, 209 Frank, 284 Fremery, 59 Fulker, 178 G Gaede, 84 Gale, 86 Gareis, 281, 287 Gasparini, 242 Gaudet, 330 Geraerts, 281, 312 Ghezzi, 239 Giannatonio, 246 Gibbons, 195 Gifford, 312, 322 Gilcrest, 307 Gill, S.M., 62, 65, 313 Gill, W.D., 85 Giorgi, 195, 196, 197, 200, 229 Girasini, 231 G lass, 172 Gosselin, 172 Gould, 177, 178, 218 Gowland, 264, 274, 298, 300 Graham, 274, 298 Grant, 97, 108 Gray, 312, 313 Grenier, 281, 283, 296 Gretz, 178, 218 Grigorov, 206 Gross, F., 141 Gr6 bner, 232, 243 Guentherschulze, 85 Guo, 243, 244 Gupta, 205, 206, 207, 209, 210 Gurewitsch, 85, 87, 110, 119, 172 Guthrie, 4, 23 84, 155, 261 H
Haarhuis, 312 Haas, 233 Habfast, 231
356
AUTHOR INDEX
Hablanian, 1, 23, 36, 255, 302, 304, 306 Haefer, 277 H~i frier, 312, 313, 315 Hagenbach, 285 Halama, 209, 221, 243, 244, 2~, 270, 274, 276, 277, 279, 281, 284, 289, 296, 305 Hall, 83, 103, 104, 118, 123, 146, 159 Halliday, 8, 116 Halsey, 281 Hamilton, 94, 96, 111, 113, 138, 140 Hammond, 281, 284, 298 Hands, 256, 2~, 274, 312, 313 Haque, 195 Harra, 117, 196, 197, 199, 200, 201, 202, 203, 204, 205, 208, 210, 214, 216, 217, 221, 222, 223 Harris, 97 Hartwig, 158, 169, 173 Hashimoto, 302, 324 Harvell, 321, 325, 330 Hatch, 140 Hayakawa, 85 Hayashi, 46, 172, 313 Haygood, 263, 274, 306 Hayward, D.O., 195, 196, 197 Hayward, W.H., 197, 199, 200, 201, 202, 205 Heafer, 10, 259 Heisig, 309 Helmer, 83, 88, 103, 123, 154, 159, 161, 162, 168, 171 Hemstreet, 269, 279, 281, 282, 283, 289, 296 Henderson, 172 Hengevoss, 274, 277, 281, 298 Herb, 83, 110, 177, 178, 195, 211, 216 Herzberg, 165 Hess, 246 Hilleret, 274, 276, 281 Hill, 121, 123 Hirsch, 167, 169 Hisamatsu, 101, 116, 123, 141, 158 Hobson, 39, 52, 98, 99, 100, 102, 169, 172, 243, 256, 259, 261, 264, 265, 2~, 269, 270, 271, 273, 274, 279, 281, 282, 284, 289 Hogan, 284, 322
Holland, 102, 177, 178, 203, 205, 206, 209 Holmes, 119, 121 Honig, 10, 258, 259 Hood, 298 Hook, 10, 258, 259 Hooper, 88, 165, 171 Hooverman, 179 Horikoshi, 158 Horne, 99 Ho, 96 Hseuh, 28, 68, 70, 232, 233, 238, 239, 240, 242, 270, 273, 274, 276, 277, 281, 282, 294, 296, 297, 305 Huffman, 104 Hultzman, 53 Hurst, 112 H~i tten, 231
Ichimura, 233, 234, 238, 240 Iimura, 195 Ikegami, 312, 313 Ishimaru, 65, 101, 102, 116, 123, 141, 158, 221 Ito, 302, 324 Ivanovskii, 178 Ivanovsky, 178, 216 Iwasa, 302, 324
James, 87, 88, 114, 115 Jensen, 158, 173 Jepsen, 84, 88, 92, 103, 104, 105, 106, 108, 109, 112, 116, 119, 120, 123, 127, 142, 146, 154, 157, 159, 168, 169, 171, 173, 175, 176, 209, 285 Johnson, F., 120, 157, 158 Johnson, J.W., 176, 221 Jurow, 120, 157, 158 K Kaminsky, 1 Kanazawa, 101, 116, 123, 141, 158 Kawasaki, 112, 135 Kay, 85
AUTHOR INDEX Kearns, 161 Kelly, J.E., 175 Keng, 96 KenKnight, 119, 127 Kenmore, 279, 283 Kersevan, 246 Keise, 315 Kietzmann, 106, 108, 108 Kikuchi, 312, 313 Kim, J., 274, 298 Kindl, 229, 230, 234 King, 176 Kistemaker, 116 Klaichko, 119 Klein, 281, 284, 298, 309, 315 Kleuh, 141 Klopfer, 177, 209 Knauer, 88, 165, 167, 168, 169 Knize, 231, 233, 235, 237, 240, 242 Knoze, 231, 237 Knuth, 281 Kobayashi, 274, 277, 281 Kohno, 112, 135 Komiya, 46, 59, 172 Konjevic, 94, 108 Kornelsen, 39, 52, 98, 99, 100, 102, 116, 199, 211, 243, 256, 259, 265, 269, 271, 279 Kouptsidis, 158, 169, 173 Kraus, 161 Kubaschewski, 130 Kueppershaus, 158 Kuluva, 281 Kumagai, 177 Kunin, 119 Kuriyama, 302, 313, 330 Kurti, 26 Korukouchi, 66 Kuznetsov, 178, 216 Kuz'min, 195, 206, 209, 211
165, 166 Lamoreaux, 214 Lampert, 233 Lam, C.K., 274, 281 Lange, 88, 166 Langhorn, 274, 298 Lanni, 209, 213, 232, 233, 238, 239, 242 Larionov, 119 Lassiter, 172 Laurenson, 118, 203, 205, 206, 209, 285 Laurent, 123, 158, 172, 232, 243 Lawson, 205, 212, 214 Leach, 25 Leck, 205, 206, 207, 209, 210 Lee, R.S., 159 Lee, S.F., 101, 102, 118, 134, 139 Lee, T.J., 259, 274, 277, 300 Leever, 313, 331 Lefrancois, 313 Lepofsky, 330 Lessard, 281, 308, 309, 312 Levenson, 256, 263, 274, 300 Lewin, 141 Li, 246 Lichtman, 175 Lifshin, 46 Limon, 178, 273 Lin, 101, 102, 118, 134, 139 Little, A.D., 321 Liu, 101, 102, 118, 134, 139 Lloyd, 104, 105, 147, 176 Longsworth, 255, 2~, 277, 281, 283, 289, 302, 304, 306, 312, 313, 315, 321 Looney, 59 Lutz, 165 Lu, 102 M
L L'Aminot, 248 La Marche, 231, 237 Lafferty, 1, 112, 115, 135 Lahav, 281 Lamartine, 233 Lamont, 83, 116, 121, 126, 147, 148,
357
Mack, 157 Maissel, 95, 139 Malev, 157, 158, 169, 173 Maliakal, 178 Malinowski, 231, 235, 237 Malmberg, 171 Malter, 105
358
AUTHOR INDEX
Mandel, 179, 218 Mandoli, 105, 159 Manni, 232, 240 Mannini, 246 Mapes, 68, 70 Maple, 101, 119, 123, 128, 141 Marinuzzi, 139, 146 Mark, 172 Martelli, 234, 237, 238, 240 Martin, 218, 222 Mathewson, 232, 243 Matsuyama, 233, 234, 238, 240 Mattern-Klosson, 315 Mazey, 141 Mazza, 246 McCafferty, 92, 175 McConnell, 176, 221 McCracken, 101, 119, 123, 128, 141, 211, 212 Mclntyre, 62, 65, 313 McKee, 41 McKinney, 322 McLaughlin, 323 McMahon, 312, 322 Meissner, 331 Mellen, 46 Melissinos, 61 Men'shikov, 178, 195 Mercer, 108, 112 Merriam, 307 Miller, G.L., 233, 237 Milleron, 149, 300 Millicon, 255, 302, 304, 306 Milner, 176, 221 Minata, 312, 313 Ming, 123 Mioduszewski, 231, 237 Missimer, 331 Mistry, 246 Miyahara, 158 Mizobuchi, 158 Moenich, 232 Mommsen, 39, 40 Momose, 101, 116, 123, 141, 158 Monnier, 158, 173 Montuclard, 313 Moore, 176 Morbidi, 312, 313 Morimoto, 312, 312
Morishita, 312, 313 Morita, 66 Moulou, 177 Mourad, 178 Muhlenhaupt, 312 M(i ller, 329 Mullhaupt, 269, 279, 281, 289, 313 Mullins, 141 Mundinger, 315 Murai, 330 Munro, 37, 123, 218, 221 N Nagao, 302, 324 Naik, 178, 179 Nakagome, 302 Nakajima, 312, 313 Nakanishi, 158 Nazarov, 178, 195, 216 Neal, 28, 107, 176, 284 Nelson, 119, 121 Nienhuis, Nitta, 302 Nix, 128 Nobcs, 119 Nordstrom, 158, 173 Norton, 141 N5 ller, 177 O Obert, 298 O'Boyle, 327 Oddon, 274, 298 Oechsner, 95, 119, 140 O'Hanlon, 4, 14, 23, 58, 65 Ohsaki, 101, 116, 140 Ohsako, 312, 313 Okano, 101, 116, 140, 195 Okumura, 330 Olsen, 176, 221 O'Neil, J.A., O'Neill, G.K., 157 Ono, 90 Origlio, 229 Osipov, 157 Ozaki, 28
AUTHOR INDEX P Pace, 231 Palacios, 281, 282, 296 Pamplin, 17 Pashley, 211, 212 Passardi, 274, 298, 300, 301 Pasternak, 274 Pate, 62, 65, 136, 137, 149, 159, 313 Paton, 127, 128, 134 Patrick, 274, 277, 281, 282, 298 Paulmier, 179 Pauly, 83, 178, 211, 216 Peacock, N.T., 90 Peacock, R.N., 65, 67, 90 Penning, 84, 86, 87 Perkins, 62, 64 Peterson, E.C., 176 Peterson, J.F., 321 Petit, 179 Philibert, 128 Phillips, 84, 86 Pierini, 87, 90, 92, 112, 114, 123 Pingel, 158, 173 Pisanni, 196, 197 PNEUROP, 143 Pochman, 133 Podgurski, 261 Pollard, 256, 2~, 274, 298, 300 Poncet, 231 Popov, 157 Poth, 231 Powers, 281, 296 Pressman, 158 Prevot, 195, 221
Redhead, 39, 52, 83, 88, 90, 98, 99, 100, 102, 164, 169, 171, 172, 243, 256, 259, 265, 269, 271, 279 Reed, 97 Reichardt, 205 Reich, 175, 177 Reid, R.I., 158 Reikhrudel, 101, 103 Reinath, 149 Reinhard, 223, 232, 242, 243, 2 ~ Ren, 281 Rerig, 321 Resnick, 8, 116 Rezanowich, Ricca, 229 Riviere, 119, 175 Robens, 2 ~ Robertson, 219, 220 Robinson, 321 Roche, 330 Rodes, 312, 313, 323 Rogers, 176 Roman, 110 Romer, 231 Rorden, 159 Rosai, 229, 231, 235, 236, 241, 242, 243 Roth, 65 Rouaud, 313 Rozgonyi, 149 Ruby, 274, 298 Rudnitskii, 95, 117, 168 Rutherford, 37, 87, 90, 92, 93, 94, 105, 106, 108, 109, 110, 112,119, 120, 127, 145, 159, 173 Ruthlein, 312, 313 Ruttenbur, 281, 283 Ryabov, 158
O Quinn, 159 R
Rabusin, 229, 230, 234 Rachocki, 233 Radebaugh, 313, 331 Radushkevich, 269 Rapp, 51, 52, 144
359
Saksagansky, 158 Santeler, 300 Sarcia, 318 Sasaki, 175, 255, 302, 304, 306 Sato, H., 46 Sato, S., 172 Saxon, 195 Sch~i fer, 312, 313 Scheer, 312
360
AUTHOR INDEX
Schram, 63 Schulien, 141 Schulz, 158, 173 Schumann, 158 Schuurman, 164, 169 Schwartz, 158 Sciuccati, 242 Scott, 295 Sears, 8 Sedgley, 269, 279, 281, 282, 284, 289, 298> 299 Sethna, 108, 112, 135 Shal'nikova, 271, 272, 274, 279 Shen, 240 Sheraton, 175 Simon, 12 Simpkins, 231, 237 Sims, 172 Singleton, 113, 119, 120, 121, 123, 128, 137, 140, 143, 145, 179 Skelton, 62, 65, 232, 240, 313 Sledziewski, 195, 221 Smart, 159, 172 Smirtnitskaya, 96, 101, 103, 168, 172 Smith, J.L., 321 Smith, P.T., 51 Snoek, 116 Snouse, 139, 145, 176, 214, 216, 217 Soddy, 86 Sondericker, 326 Souchet, 145 Spagenberg, 46, 138, 216 Stack, 168 Stainsby, 172 Staph, 173 Stattel, 232, 240 Steele, 281 Steinberg, 205, 207 Steinrisser, 114, 117, 142, 149 Stern, 268, 269, 270, 279, 281, 282, 283, 289, 291, 292, 293, 296 Stolz, 312, 313 Storey, 195, 229 Stout, 195 Strubin, 216 Stuijts, 159 Sugita, 112, 135 Swartz, 177 Sweetman, 205, 206
Symersky, 281, 312 T Takahashi, 302 Takeuchi, 238 Tate, 51 Taylor, N., 195, 196 Tee, 96, 168 Teller, 269, 279 Tempelmeyer, 277, 281 Ternstr8 m, 168 Tetelman, 128 Thibault, 281, 298 Thomson, G.P., 49, 50 Thomson, J.J., 49, 50 Timm, 315 Tobin, 269, 279, 281, 284, 289 Tominaga, 195 Tompkins, 261 Tom, 37, 87, 114, 115, 123, 148, 221 Todd, 136, 137, 149, 172 Trachtenberg, 157, 169, 173 Trapnell, 101, 102 Trayer, 256, 263, 274 Trendelenburg, 281, 298 Trickett, 158 Troy, 274 Tsujikawa, 158 Tsujimura, 330 Tuozzolo, 62, 65, 313 Tupper, 274, 298 Turner, 139, 146, 284, 288, 289, 290, 292, 293, 322 Tuzi, 101, 116, 140, 274, 277, 281 Tzatzov, 206 Tzemos, 255, 302, 304, 306 U Underhill, 160 V Vanderslice, 94, 110, 112, 115, 135, 175 Vaumoron, 109, 116, 118 Verma, 179 Vinogradov, 202, 205, 206
AUTHOR INDEX Viskanta, 307 Visser, 119, 281, 312 Vissers, 121 Voss, J., 120, 157, 158 Voss, G., 315 van den Broek, 160 W Wagener, 101, 195, 210 Wakerling, 4, 23, 84, 155, 261 WaUen, 271, 272, 273, 274, 276 Walley, 219 Wang, 96 Wanhill, 128 Wantanabe, 233, 234, 238, 240 Warnecke, 177 Wasa, 85 Watanabe, 112, 135 Watson, 281, 282, 296, 297 Wear, 114, 176 Webber, 2~, 277, 289 Wedekind, 158 Wehner, 119, 127 Weiss, 159 Welch, 31, 39, 62, 65, 66, 68, 85, 91, 92, 101, 102, 116, 119, 121, 122, 126, 136, 137, 140, 141, 145, 146, 147, 161, 172, 174, 255, 273, 283, 287, 302, 304, 305, 307, 309, 311, 312, 313, 318 Wells, 46 Welton, 83, 178, 211, 216 Westendorp, 86, 87, 110, 129, 172 Weston, 108, 112, 135 Wheeler, 65, 103 Wightman, 274 Williams, 127, 128, 134, 270, 271, 274, 298 Willis, 112, 116, 135 Winters, 99 Wohlfarth, 161 Wolf, 221, 231 Woodward, 205, 212, 214 Worwetz, 274, 276, 281, 282, 284, 296, 297, 305 Wutz, 169 Wyrick, 176
Y Yabuki, 302 Yagi, 96 Yakoo, 90 Yoshimura, 64 Young, J.R., 96, 105, 127, 168 Young, R.L., 277, 281, 307 Z Zabritski, 178 Zanin, 246 Zaphiropoulos, 105, 147, 176 Z~ idel, 119 Zeilinger, 133 Zeng-Rui, 123 Ziegler, 127 Zierman, 307 Zigrosser, 159
361
This Page Intentionally Left Blank