HANDBOOK OF CONTAMINATION CONTROL IN MICROELECTRONICS: Principles, Applications and TechnologyDonald L. TolliverWilliam Andrew Inc.HANDBOOK OF CONTAMINATION CONTROLIN MICROELECTRONICSMATERIALS SCIENCE AND PROCESS TECHNOLOGY SERIESEditorsRointan F. Bunshah, University of California, Los Angeles (MaterialsScience and Technology)Gary E. McGuire, Microelectronics Center of North Carolina (Elec-tronic Materials and Processing)DEPOSITION TECHNOLOGIES FOR FILMS AND COATINGS; Develop-ments and Applications: by Rointan F. Bunshah et alCHEMICAL VAPOR DEPOSITION FOR MICROELECTRONICS; Principles,Technology, and Applications: by Arthur ShermanSEMICONDUCTOR MATERIALS AND PROCESS TECHNOLOGY HAND-BOOK; For Very Large Scale Integration (VLSI) and Ultra Large Scale Integration(ULSI): edited by Gary E. McGuireSOL-GEL TECHNOLOGY FOR THIN FILMS, FIBERS, PREFORMS, ELEC-TRONICS, AND SPECIALTY SHAPES: edited by Lisa C. KleinHYBRID MICROCIRCUIT TECHNOLOGY HANDBOOK; Materials, Proc-esses, Design, Testing and Production: by James J. Licari and Leonard R.EnlowHANDBOOK OF THIN FILM DEPOSITION PROCESSES AND TECH-NIQUES; Principles, Methods, Equipment and Applications: edited byKlaus K. SchuegrafRelated TitlesADHESIVES TECHNOLOGY HANDBOOK: by Arthur H. LandrockHANDBOOK OF THERMOSET PLASTICS: edited by Sidney H. GoodmanHANDBOOK OF CONTAMINATION CONTROL IN MICROELECTRONICS;Principles, Applications and Technology: edited by Donald L. TolliverHANDBOOK OFCONTAMINATION CONTROLIN MICROELECTRONICSPrinciples, Applications and TechnologyEdited byDonald L. TolliverMotorola Semiconductor SectorMotorola, Inc.Phoenix, ArizonaReprint EditionNOYES PUBLICATIONSWestwood, New Jersey, U.S.A.Copyright 1988 by Noyes PublicationsNo part of this book may be reproduced or utilized inany form or by any means, electronic or mechanical,including photocopying, recording or by any informa-tion storage and retrieval system, without permissionin writing from the Publisher.Library of Congress Catalog Card Number: 87-31533ISBN: 0-8155-1151-5Printed in the United StatesPublished in the United States of America byNoyes PublicationsFairview Avenue, Westwood, New Jersey 0767510 9Library of Congress Cataloging-in-Publication DataHandbook of contamination control in microelectronics.Bibliography: p.Includes index.1. Integrated circuits--Design and construction.2. Contamination (Technology) I. Tolliver, Donald L.TK7836.H34 1988 621.381'73 87-31533ISBN 0-8155-1151-5PrefaceThe semiconductor industry has witnessed the birth of contamination con-trol in the microelectronics industry within the last ten years. Prior to 1980the importance and future impact of this independent technology was limitedto a relatively few individuals scattered across a limited number of companiesin the United States, Japan and Europe. As the industry entered the higher com-plexities of 64K DRAM manufacturing in the early 1980's the importance ofdefect control and particle reduction in the manufacturing process began to re-ceive more specific focus from the supplier side and the semiconductor manu-facturer.Today a whole industry exists and continues to expand in order to serve thegrowing needs of contamination control for the microelectronics industry.Many semiconductor and microelectronic firms to this day do not recognizecontamination control technology as an independent ingredient of the worldwide semiconductor business. It nonetheless exists in one form or another and iscalled by a variety of different terms, but regardless of the name, it still is con-tamination control.Originally, most of the contamination control suppl iers and users saw thetechnology of contamination control as the exclusive territory of the cleanroom and only the clean room. This led to some ambiguity and confusion as towhat specific categories constituted contamination control and which belongedto other better known processing categories. A technology such as liquid filtra-tion, which was by definition exclusive of the concepts of clean rooms, at thattime was focused toward one of the other complex technology segments whichwent into the manufacturing of semiconductors. Often contamination control,but not necessarily referred to by that name, would be designated as a sub-setof the diffusion/oxidation technology, the lithography and etch operations,the deposition technology, ion implantation, the device engineering and yieldenhancement engineering groups, and very often the quality control organiza-tion that supported a particular manufacturing organization. The contamina-vvi Prefacetion control wheel described below summarizes most of the major categorieswhich actually fall under the umbrella of contamination control.Where is contamination control technology today? The answer is still notclear cut, but there is no denying the fact that it has emerged as a separate anddistinct technology within the mainframe of semiconductor manufacturing. Itstill experiences some confusion in definition and may very well be referredto as particle reduction, yield enhancement, defect prevention, and perhapseven quality control. But certainly contamination control is now recognizedon a more independent basis and is receiving independent budgeting and inde-pendent resources many times greater than we could have imagined in 1975and 1976. Liquid filtration is certainly a well accepted part of contaminationcontrol technology along with many other technologies such as laser particlecounting, process gas purification, and high purity chemical distribution.Preface viiMembers of this highly ubiquitous group of multifaceted technologies arecontinuing to format the boundaries and the specific applications which mayfall into the arena of contamination control. It is certainly clear that clean roomtechnology is only one subset of the entire family of requirements in contamina-tion control. Very important today and in the future of our industry are thenonparticle contamination requirements which clearly fit into contaminationcontrol, but are not always found there. Sub ppb metallic ion contaminationin DI water, process chemicals and process gases are now heavily in focus forcontamination control requirements in the next five years.The purpose of this handbook is to attempt to distill into one volume asynopsis of many of the major categories which go into the subject of contami-nation control. I have purposely focused on the microelectronics applicationsrather than those of other industries. I am best prepared to deal with the issuesof the semiconductor industry rather than those topics which more preciselyfall into industries such as pharmaceutical, aerospace, biomedical and lasers.I n many cases we share the same concerns and the same solutions but for dif-ferent reasons. Certainly it is not my goal to exclude excellent and very worth-while subjects and contributions in these sister industries. However, the needis clear to provide a reference for the microelectronics industry to help definethe needs, solutions and requirements to build upon for the future.Included in this handbook is a broad look at most of the major subjectswhich impact and interact with contamination control technology in the micro-electronics industry. This is only a beginning and there will be future editionswith new subjects and updated technological requirements.We have included a fairly basic treatment of the technology of aerosol fil-tration and aerosol particle measurements methods. Within the operations ofclean rooms the very essence of clean room management and clean room gar-ments is given an indepth discussion.Electrostatics in contamination control is a difficult arena to target forcontamination control. More often the reader will find this subject tied to thearea of EOS/ESD and not in contamination control. The editor belives thatthe treatment of this subject is worthwhile as so many of the inherent par-ticle mechanisms which must be dealt with .on a daily basis are inherent to elec-trostatic technology.The subject chapters under the materials categories such as DI water,gases, chemicals, and lithographic processing are now recognized as major con-tributors to device yield loss and are well entrenched inside contamination con-trol requirements. Associated closely with our materials categories are thechapters on surface particle detection and liquid particle monitoring technology.These subjects would not have even surfaced had this book been publ ishedprior to 1980. The problems were not defined and their actual contribution tothe success of semiconductor processing would probably have gone unnoticed.Near and dear to the hearts of active contamination control engineers inthe semiconductor industry, are the subjects relative to contamination controlin processing equipment. We as members of this everchanging industry totallymissed that one also, prior to about 1978. Today it is clear that in spite of allother efforts that must go on in contamination control, we must learn how toanalyze and control the contamination from the processing equipment itselfviii Prefaceor our other efforts are futile. Two chapters are dedicated to the requirementsof equipment analysis via wafer particle detection and automation in processingequipment as it relates to wafer and cassette automation.A glossary, formally titled A Glossary of Terms and Definitions Relatedto Contamination Control has been included in this printing as an additionalreference for the reader. This glossary has been produced and is under the juris-diction of RP-11 Working Group of the Standards and Practices Committeeof the Contamination Control Division, Institute of Environmental Sciences.The editor and publisher gratefully acknowledge the Institute of EnvironmentalSciences for permission to reprint this glossary. Comments and suggestionsabout this glossary should be made directly to the Institute of EnvironmentalSciences, Mt. Prospect, Illinois.It is certainly the belief of the editor at this writing that the role and im-pact of contamination control in the microelectronics industry will be a domi-nant one and possibly the key factor to success in meeting the manufacturingchallenges of the microelectronics industry of the future.Phoenix, ArizonaJanuary, 1988Donald L. TolliverContributorsMauro A. AccomazzoMillipore CorporationBedford, MAAnn Marie DixonCleanroom ManagementAssociates Inc.Tempe, AZRobert DonovanResearch Triangle InstituteResearch Triangle Park, NCDavid EnsorResearch Triangl e Inst ituteResearch Triangle Park, NCTerry L. FaylorT.L. Faylor & AssociatesSanta Cruz, CAGary GanziMillipore CorporationBedford, MAPeter GiseTencor InstrumentsMountain View, CABennie W. GoodwinAngelica Uniform GroupSt. Louis, MOJeffrey J. GorskiOxnard, CARobert KaiserConsultant to MilliporeArgos Associates, Inc.Winchester, MAGerhard KasperLiquid Air CorporationCountryside, I LRobert G. KnollenbergParticle Measuring Systems Inc.Boulder, COBenjamin H.Y. LiuUniversity of MinnesotaMinneapolis, MNMary L. LongUniversity of ArizonaTucson, AZixx ContributorsRollin McCratyStatic Control ServicesPalm Springs, CAMike NaggarK.T.1. Chemicals Inc.Sunnyvale, CAMihir ParikhAsyst Technologies, Inc.Milipitas, CADavid Y.H. PuiUniversity of MinnesotaMinneapolis, MNDonald L. TolliverMotorola Semiconductor SectorMotorola Inc.Phoenix, AZBarclay J. TullisHewlett-Packard LaboratoriesPalo Alto, CAH.Y. WenLiquid Air CorporationCountryside, I LNOTICETo the best of our knowledge the informa-tion in this publication is accurate; howeverthe Publisher does not assume any responsi-bility or liability for the accuracy or com-pleteness of, or consequences arising from,such information. Mention of trade names orcommercial products does not constituteendorsement or recommendation for useby the Publ isher.Final determination of the suitability ofany information or product for use contem-plated by any user, and the manner of thatuse, is the sole responsibility of the user. Werecommend that anyone intending to relyon any recommendation of materials orprocedures mentioned in this publicationshould satisfy himself as to such suitability,and that he can meet all appl icable safetyand health standards. We strongly recom-mend that users seek and ad here to themanufacturer's or suppl ier's current instruc-tions for handling each material they use.Contents1. AEROSOL FILTRATION TECHNOLOGY 1David Ensor and Robert DonovanIntroduction 1Importance of Aerosol Filtration in Microelectronics 1Filter Media Description 4Historical Background 7Chapter Organization 8Filtration Fundamentals 8Fibrous Filter Theory 8Particle Trajectory Modeling 10Pressure Drop 14Particle Collection Efficiency 16Membrane Filters 28Performance Pred iction 32Particle Collection 32Electrostatics 34Passive Electrostatic Systems 38Active Electrostatic Systems 38Filter Testing Fundamentals 40Particle Generation 40Atmospheric Aerosol 40Dispersion Test Aerosol 40Condensation Test Aerosols 41Instrumentation 43Particle Concentration 43Particle Size Distribution Measurement 46Sampling 52Standard Tests 53Developmental Tests 53xixii ContentsFlat Media Testing 53Laser OPC Modification of Standard Tests 56Point-of-Use Filters 56Filter Applications 60Ventilation System 60Point-of-Use Filters 61Sampling Filters 63Summary 63References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632. INSTRUMENTATION FOR AEROSOL MEASUREMENT 68Benjamin Y.H. Liu and David Y.H. PuiIntroduction 68Size Distl'ibutiol1 of Aerosols 68Particle Deposition on Surfaces 76Instrumentation for Aerosol Measurement 80Optical Particle Counter 80Aerodynamic Particle Sizer 88Condensation Nucleus Counter and Diffusion Battery 96Aerosol Sampling Instruments 100Monodisperse Aerosol Generators 100Performance of HEPA and ULPA Filters 106References 108Bibliography 1093. CLEAN ROOM GARMENTS AND FABRICS 110Bennie W. GoodwinIntroduction 110Clean Room Garment Design/Styles 111Head Covers 112Body Covers 112Coveralls 112Frocks 115Sleeves 115Foot Covers 115Hand Covers 119Accessory Items 119Clean Room Garment Construction 119Seams 119Clean Room Fabrics 121Spun Fabrics 121Synthetic Fabrics 122Woven Fabrics 123Nonwoven Fabrics 123Coated (Laminated) Fabrics 126Chemical Resistance 126Electrostatic Charges 126Static Dissipative Apparel 127Contents xiiiClean Room Garment Life 129Guidelines for Clean Room Garments 130Recommended Clean Room Garment Usage 130Clean Room Garment Rules and Regulations 131Recommended Clean Room Garment Changes 131Clean Room Garment Processing 132Clean Room Garment Sterilization 134Autoclaving 134Ethylene Oxide (ETO) 134Gamma Radiation 134Conclusion 1344. GUIDELINES FOR CLEAN ROOM MANAGEMENT ANDDISCIPLINE 136Anne Marie DixonIntroduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136Clean Room Criteria 137Good Clean Room Work Practices 138Prohibited Actions in Clean Room 139Personnel Disciplines 139Training 140Some New Approaches 140Clean Room Garments 141Classification Demands 141Gowning Procedures and Techniques 142Garment Storage and Reuse 143Secondary Usage 143Clean Room Supplies 143Wipers 144Gloves 144Paper 144Furniture 145Housekeeping Maintenance 145Cleaning the Clean Room 145Tools and Equipment ~ 146Cleaning Procedures 146Ceilings 146Walls 146Floors 146Work Surfaces 146Gowning Rooms 147Test Procedures 147Facility Start-Up Procedures 147Gross Cleaning 147Precision Cleaning 148Equipment Installation 148Equipment Installation and Repair-Existing Facility 149Contamination Control Monitoring 149xiv ContentsThe Clean Room Manager 150Set Objectives 151Organization 151Motivation and Communication 151Measurement 151Qual ity and Productivity 151The Future 152References 152Bibliography 1525. ELECTROSTATICS IN CLEAN ROOMS 153Rollin McCratyIntroduction 153Static Electricity 153Forces Between Charges 154Electric Field 155Capacitance 155Induction Charging 156Triboelectric Charging 156Surface Contamination from Static Charges 157Eliminating Static Charge 158Air Ionization 158Ion Generation 160Nuclear Ionizers 160Electrical Ionization 161Space Charge . . . . . . . . . . . . . . . . . . . . . . 161AC Ionization 162Constant DC Ionization 162Pulse DC Princi pies of Operation 163Electrodes 164Total Room Ionization 164Auto Balance 166Desiccator Cabinets 166Blow Off Guns 169Grounding 171Wrist Straps 172Surface Resistivity 173Bench Top Materials 174Conductive Flooring 175Humidity 175Topical Antistats 176Protective Packaging 176Measuring and Testing Methods 177Static Fieldmeters 177Charge Plate Monitor 178Ion Counter 181Surface Resistivity Meter 182Wrist Strap Testers 182Contents xvIon Balance Meter 183Bibliography 1846. ULTRA HIGH PURITY WATER-NEW FRONTIERS 185Terry L. Fay/or and Jeffrey J. GorskiPure Water and the Semiconductor Industry 185The Pure Water Process 185The Pure Water Industry Today 186The High Purity Water Improvement Industry 187Some New Prospects in Pure Water Technology 188Pure Water Technology 188Contamination Problems 189Organics 189Colloids 190Organisms 190Particles 190Removal Technology 190Details of the Purification Process 191Deionization 192Flocculation 192Reverse Osmosis 193Ultrafiltration 193Electrodialysis 193The Pretreatment Dilemma 194The Steps for the Pretreatment Process 195Degasification 196Storage and Final Polishing 197Requirements for Improved Water 197The Double Pass Reverse Osmosis System 199Double Pass RO for Microelectronics 200The Demand for Continued Improvements 201Using RO Up Front 201The Upstream Requirements for Deionization 202A Triple Membrane Five Step Concept 203Ultrafiltration 203Electrodialysis Reversal 204Reverse Osmosis 204Deionization 204Bacteriological Control. 205The New Concept in Pure Water Systems 205Suppliers for High Purity Water 206A liMe Too" Industry 207Summary 2087. DEIONIZED (01) WATER FILTRATION TECHNOLOGY 210Mauro A. Accomazzo, Gary Ganzi, and Robert KaiserIntroduction 210Semiconductor Product Characteristics and Trends 210xvi ContentsEffects of Contaminants on Product Performance andYield 210Process Water in Semiconductor Manufacturing 212Purity Requirements 212Production of High Purity Water 214Deionized Water Filtration Equipment 216Definitions and Requirements 216Microporous Filters 217Definition and General Description 217Material Properties 223Microporous Filter Configurations of IndustrialInterest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223Particulate Removal by Microporous Filters 228Hydraulic Considerations 230Filter Induced Contamination 231Ultrafiltration Membranes 235Introduction 235Material Properties 236Hydraulic Considerations 236Contaminant Removal by Ultrafiltration 244Ultrafilter Induced Contamination 246Reverse Osmosis Membranes 246Systems Integration and Process Considerations . . . . . . . . . . . .. 246Importance of Overall Process Design and Integration 246Filter Housings 246Requirements 246Design Considerations 247Materials of Construction 247Distri bution Loop Design 247Requirements 247Microorganism Control 247Materials of Construction 248Other Process Elements 249Process Conditions 249Methods of Analysis 251Integrity Testing 251Monitoring for Bacteria 252Monitoring for Particles 252Monitoring for Dissolved Species 253Choosing a Membrane Filtration System 253References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2548. MONITORING SYSTEMS FOR SEMICONDUCTOR FLUIDS 257Robert G. KnollenbergHistorical Perspective 2q7Survey of Measured Levels of Microcontamination in Semi-conductor Process Liquids 259Material Problems in Liquid Media 265Contents xviiSampling Methodologies 267Overview of Measurement Methods 271Detailed Description of Light Scattering Response forLiquid Suspended Particles 274Theoretical Considerations 274Defining the Lower Limit of Particle Size Using Laser LightScattering 282Description-of Instruments 284Volumetric Instruments 284"In-Situ" Liquid Instruments 288Application Problem Areas 292Operating Procedures 292Container Cleaning 295Possible Contaminants 295Artifacts 296Data Interpretation 297References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2999. PARTICLES IN ULTRAPURE PROCESS GASES 301Gerhard Kasper and H. Y. WenIntroduction 301The Challenge to High Technology 301Challenge for Particle Technology 302Contents and Structure of This Chapter 302Particle Sources and Formation Mechanisms 303The Primary Mechanisms 304Case I: Outdoor Atmospheric Aerosol 305Case II: A Heated Metal Tube 306Secondary Generation by Particle Resuspension 307Selection Criteria for On-Line Particle Analysis Equipment 311The Particle Size Range of Interest 312The Particle Concentration Range 312The Choices 313Detection Limits and Counting Efficiency 313I nternally Generated Counts 314Light-Scattering Particle Counters 318Operating Principle 319Key Characteristics of Commercial Optical Particle Counters 319Smallest Detectable Particle Size Versus Sampling Flow Rate 321Background Counts Versus Sampling Flow Rate 321Counting Efficiency and Detection Limits 323Special Comments on High-Pressure Counters 325Condensation Nuclei Counters 326The Operating Principle 326Commercially Available CNCs 327Size Sensitivity and Lower Detection Limit 327Upper Detection Limit 327Internally Generated Counts 328xviii ContentsOn-Line Measurement of Particle Sizes in Very Clean Gases-A Synopsis 328Size Classification With LPC-CNC System 329Differential Electrical Mobility Classification 331Diffusion Batteries 332Particle Sampling 333Particle Generation by the Sampling Line 333Pressure Reduction 334Particle Sampling for Off-Line Analysis 338Inertial Impaction 339Removal of Particles from Compressed Gases by Filtration 341Elementary Filtration Kinetics in Gases 342Achievable Low Levels of Particle Concentration in ProcessGases by Filtration 344References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34610. CONTAMINATION CONTROL AND CONCERNS IN VLSILITHOGRAPHY 348Mary L. LongIntroduction 348Photoresist: What Is It? . . . . . . . . . . . . . . . . . . . . . . . . . . 348Chemistry of Negative Photoresist 348Negative Photoresist Developers 351Chemistry of Positive Photoresist 352Positive Photoresist Developers 354Nonoptical Resists 355Primers 356Hexamethyldisilazane (HMDS). 356Photoresist Functional Requirements 357Image Fidelity 357Etch Protection 358Ion I mplant Mask 358Lift-Off Mask 358Multilayer Mask 359Photoresist Contamination Concerns 359Purity 360Packaging 360Processing 361Contamination Control in Photoresist Processing 362Surface Preparation 362Spin 363Bake 364Exposure 365Develop Cycle 365Inspection 366Etch 366Resist Mask 367Resist Removal 367Contents xixSummary and Conclusions 368Filtration 368Equipment Maintenance 368Personnel Training 368Dirt Traps 36911. CONTAMINATION CONTROL IN MICROELECTRONICCHEMICALS 370Mike NaggarDefining Contamination 270Particle Contamination 370Metal Ion Contamination 371Measuring the Contamination 371Particle Detection Technology 371Metal Ion Analysis 371Equipment Choices 372Sizing up the Contamination 372Controlling the Contamination in Chemicals 375Materials Control as Supplied 375Filtration of Chemicals 375Cleaning of Containers 375High Pressure Nitrogen and/or Air Blow 376Hot DI Water and Hot Nitrogen Clean 376Compatible Chemical Clean 376Chemical Control in the Process 378Storage and Shelf Life 379Processing Compatibility of the Equipment 380Spraying and Immersion Processes 380Spray Processes. . . . . . . . . . . . . . . . . . . . . . . . . 380Immersion Processes 380Process Monitoring 381Conclusion 38112. SURFACE PARTICLE DETECTION TECHNOLOGY 383Peter GiseIntroduction 383Effects of Particles on Circuit Yield 383Historical Perspective 384Inspeetion Methods for Particles on Planar Surfaces 385Wafer Surface Inspeetion Systems 385Light Scattering from Surface Defects 386Basic Light Scattering Processes 387Light Scattering by Latex Spheres 387High Intensity Collimated Light 391Scanned Laser Light 392Principles of the Method 392Rotating Polygon Scanners 393Oscillating Mirror Scanners 394xx ContentsMoving Substrate Scanners 398Signal Processing Methods 400Inspection Methods for Particles on Patterned Surfaces 403Inspection for Particles on Wafers 403Inspection of Particles on Reticles 405Laser Light, In-Situ Inspection 405Laser Light Scattering Remote Inspection 406Laser Light Scattering Inspection of Reticles withPellicles 407References 40913. PARTICLE CONTAMINATION BY PROCESS EQUIPMENT 410Barclay J. TullisIntroduction 410Particles per Wafer per Pass (PWP) 411PWP Relationship to Yield 414Variability of Measurement Tool Data 415Use of Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417Particle Counter Repeatability and Contamination byParticle Counters 418Experiment Structures for Measuring EquipmentContamination 421Experiment Structure 1 421Experiment Structure 2 423Design of Experiments 427Calibration 428Calibration of Gain and Offsets 429Calibration of Count Accuracy Versus Size 430Light Scattering by Particles on Surfaces 431Calibrations on Films 433Pixel Saturation 434Measurement Planning 436Measurement Records 437Measurement Execution 437Specification Standard . . . . . . . . . . . . . . . . . . . . . . . . . 438The Future 441References 44114. WAFER AUTOMATION AND TRANSFER SYSTEMS 445Mihir ParikhIntroduction 445The Need for Automation 445Wafer Damage 446Wafer Contamination 446Defect Density Contribution 447Wafer Misprocessing and Equipment Utilization 449Facility Operating Cost 449Individual Wafer Handling 450Contents xxiManual Handling 450Automated Handling of Individual Wafers 450Particle Control Effectiveness 454Interequipment Wafer-Cassette Transport 456Three Kinds of Wafer-Cassette Transport Systems 456Automated Guided Vehicles 457Tunnel-Track Networks 458The SMI F System 459Automated Facilities 468Conclusion 475References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476GLOSSARY 478INDEX 4861Aerosol Filtration TechnologyDavid Ensor and Robert Donovan1. INTRODUCTIONFiltration processes permeate every phase of modern technologyl andespecially impurity-sensitive manufacturing processes such as microelectronics.Filters remove particles from the air and process gases and from process chemi-cals and water. Manufacturing success in microelectronics, as measured by,for instance, chip or magnetic head assembly yield, depends critically uponhigh quality filtration. This chapter considers only aerosol filtration in micro-electronics, although liquid filtration (for example, of acids, solvents, waterand photoresist) is clearly every bit as important to successfu I microelectronicmanufact uri ng.1.1 Importance of Aerosol Filtration in MicroelectronicsPerhaps no major industry places a higher premium on aerosol filtrationthan microelectronics. The total arnbient aerosol particle concentration isprobably lower in a microelectronic manufacturing area than in any other simi-larly sized manufacturing area on earth. As measured by a condensation nucleicounter, the total concentration of particles of size 0.02 J1m and above in acontemporary microelectronics manufacturing area at rest (no process actvitity,no people in the room) is often in the 10- to 100-ft3(3.5 x 10-4to 3.5 x 10-3cm-3 ) range?The need for such high quality air stems from the small dimensions of thedevices built in today's microelectronic chips and the expectation that thesedimensions will continue to shrink. Table 1 lists dimensions typical of the0.5 J1m technology forecast to be used in production of 4 megabit silicon chipmemories in the 1990s. (These dimensions, however, are only 40 to 50 percentsmaller than those of contemporary state-of-the-art chips.)2 Handbook of Contamination Control in MicroelectronicsTable 1: Dimensional Requirements for a Half-Micrometer Technology3Lateral dimensions:Pattern sizePattern toleranceLevel-level registrationVertical dimensions:Gate oxide thicknessField oxide thicknessFilm thicknessesJunction depth0.5 pm0.15 pm0.15 pm10 nm200 nm0.25-0.5 pm0.05-0.15 pmParticles appearing on the ch ip surface during manufacturing can causechip failure by disrupting surface geometry, by creating defects in thin layersor films, and by introducing impurities. What constitutes a killer particle-aparticle 9f sufficient size to cause chip failure when located in a critical regionof the chip-depends on device dimensions. Arbitrarily, the size of a killer par-ticle can be taken to be one quarter to one half the line width of a chip lateraldimensions and one half the thickness of the various vertical dimensions.3Bythis criterion, particles as small as 5 nm are potentially killer particles of thesilicon chips of the 1990s. (The 1985 state-of-the-art in chip manufacturinguses dimensions only two to two-and-a-half times those listed in Table 1 so that10- to 15-nm particles deposited in certain critical areas can be killer particlesto today's production.) Thus, particulate contamination in microelectronicdevice production represents and warrants unprecedented control methods,of which high quality filtration is a necessary (but not of itself a sufficient) part.The high quality air now achieved routinely in microelectronic manu-facturing areas results from continual recirculation of room ambient air throughbanks of HEPA (High Efficiency Particulate Air) filters which sometimes makeup one entire wall of the manufacturing area or, as is more common now, allor most of the ceiling. In the latter configuration, vents for the return air are atthe floor level along at least two sides of the room or, in some designs, in thefloor itself which then consists of a grating placed above the return plenum-the"raised floor" design. This design promotes vertical laminar flow of air throughthe room whereby clean air from the ceiling HEPA filters flows unidirectionallydownward to the return grates in the floor, sweeping any particles emitted byroom activities out of the room and into the return air being fed back to theHEPA filters. The linear velocity of this air wash through the room is typi-cally 90 ft/min. Obstructions, such as apparatus or furniture or movementin the room, create eddies that upset the laminar flow so turbulent pocketsoften exist. However, the major flow pattern remains that of vertical laminarflow.Controll ing aerosol particles in the ambient air of the manufacturing areais just one aspect of controll ing aerosol particles in microelectronics manu-facturing. Equally important is control of aerosol particles in the many processgases used during chip fabrication. These gases create desired ambients sur-Aerosol Filtration Technology 3rounding the wafers during high temperature processing; high pressure gasesare often used to introduce desired impurities for wafer doping; inert gases areused to clean off wafers or "blow dry" wafers following various wet processsteps. Clearly, the quality of these gases to which process wafers are deliberatelyexposed is as important as room ambient.Typically, these process gases are delivered through point-of-use filters con-sisting of high quality membrane filters rather than the fibrous HEPA filters usedto clean the ambient room air. A membrane filter consists of a porous sheetrather than a packed bed of independent capture sites such as a fibrous filter.One can think of a membrane filter as a solid sheet in which many penetratingholes have been punched and, indeed, a certain subclass of membrane filters(irradiated polycarbonate sheets) are constructed in a process that matches thatdescription closely. However, the more common membrane filter is a solventcast layer which, when dried, leaves a porous but continuous one-piece substratebehind. Variables in this manufacturing process include drying rate and tempera-ture in addition to substrate and volatile solvent properties.Membrane filters are generally classified as surface filters, while fibrousfilters are depth filters. This designation reflects the mechanisms thought todominate particle capture by these differently labeled filters-a surface filterremoves particles from the air stream primarily by interception (the particlesare simply too big to fit through the filter pores); a depth filter depends upondiffusion and inertial impaction, as well as interception, to remove particlesfrom the air streams as they wind their tortuous paths through the filter.The size of the particle being captured and the air stream velocity determinewhich mechanism dominates under which circumstances. For the sub-micronaerosol particle regime to be considered in th is chapter and the filter constructionemployed to control them, the distinction between surface filtration and depthfiltration becomes fuzzy. Indeed, one of the major recent insights in filtrationresearch is recognition and demonstration of the similarity of fibrous and mem-brane filter performance in the filtration of sub-micron aerosol particles.4Inthis particle size regime, the same equations that describe fibrous filter per-formance also predict porous membrane filter performance better than thecapillary tube models traditionally used to model membrane filters.Other roles of filtration in microelectronics include: (1) the collection ofaerosol particle samples from the ambient air in order to subsequently analyzethe composition and properties of those aerosol particles, and (2) the protectionof workers from hazardous or toxic aerosols. In th is latter role, the filter elementis typically part of a respirator canister that also contains sorbent to capturevapors. Sampling -Filters are most frequently membrane filters made from ir-radiated polycarbonate sheets, while respirator filters are always fibrous filtersbecause of their low aerodynamic resistance.Neither of these two functions is unique to microelectronics, and micro-electronics requirements are not the dominant force in the marketplace for suchfilters. Therefore, these roles will not be amplified further. The major applica-tion of filters in microelectronics are the fibrous HEPA filters used to create theclean rooms in which manufacturing is carried out and the membrane filters usedto filter many process gases, some reactive, at their point of use. In both theseapplications, the needs of the microelectronics industry do represent a majormarket and, thus, significantly influence product development.4 Handbook of Contamination Control in Microelectronics1.2 Filter Media DescriptionThe two generic types of aerosol filters used in microelectronics are fibrousfilters and membrane filters. Fibrous filters are a collection of randomly orientedfibers of varying diameters and lengths which are packed, compressed, or other-wise held together as mats (sometimes with the aid of fiber binders) in a fixedvolume through which the air to be filtered passes. Fibrous filters vary from lowefficiency furnace-type filters consisting of a loose fiber web held between twosupport grids to high efficiency "papers" made by the Fourdrinier process. Thefurnish used in the latter includes both fiberglass fibers of submicron diameterfor high efficiency submicron particle capture and larger diameter fibers (up to5 JIm) for strength. The void volumes of fibrous filters are in the range of 85 to99 percent (Figure 1).Figure 1: Type GC-90 glass fiber filter-with binder. 1000X magnification. (Cour-tesy, Micro Filtration Systems)Membrane filters, on the other hand, are homogeneous, solid sheets intowhich penetrating holes or pores have been introduced by various methodsincluding irradiation and etching, biaxial stretching, and solvent volatilization.The latter process is the most common. Solutions of appropriate plastics con-taining volatile solvents are cast into sheets for subsequent curing during whichtime the various components evaporate at different rates to leave behind aporous but continuous sheet. Typical void volumes of solvent cast membranefilters are 70 to 85 percent (Figure 2).Irradiated and etched membrane filters made from polycarbonate sheetslook like the classical membrane filter model (Figure 3)-neatly defined circularholes in an otherwise homogeneous flat sheet. While the void volume of this typeAerosol Filtration Technology 5Figure 2: Cellulose nitrate membrane filter-pore size, 0.45IJ.m. 4800X magnifi-cation. (Courtesy, Micro Filtration Systems)Figure 3: Standard Nuclepore filter. (Courtesy, Nuclepore Corporation)6 Handbook of Contamination Control in Microelectronicsof filter is only 10 to 20 percent, its almost purely surface collection featuremakes this type of filter nearly ideal for examining collected particles by micro-scopy; and it is in this role that it is primarily used. It is also a good physicalmodel of a capillary pore filter.The holes in the polycarbonate sheet are defined by exposing the sheet tofission fragments from radioactive isotopes. The damaged region left in the wakeof the irradiation particles reacts readily with an appropriate etch solution. Holesize is controlled somewhat by etch time.The final membrane filter to be shown is that made from biaxial stretch ingof polytetrafluoroethylene (PTFE). While its chemical inertness is a major ad-vantage of this type of filter, its structure is also noteworthy (Figure 4) becauseit illustrates better than any other membrane filter the similarity between mem-brane and fibrous filters. Under biaxial stretch ing, the PTF E membrane remainsa single solid entity but tears into a delicate lattice of submicron fibrils thatmore closely resemble the fibrous filter structure shown in Figure 1 than themembrane filters of Figures 2 and 3. This similarity in appearance makes plausiblethe claim that similar equations can predict the performance of both types offilters.That membrane filters and fibrous filters are generally manufactured bydifferent organizations means competition between the two filter types.Figure 4: 0.5 pm Teflon membrane-PTFE. (Courtesy, Micro Filtration Systems)Aerosol Filtration Technology 7At present, fibrous filters of the HEPA or ULPA type are used exclusivelyin the filtration of room air. Their combination of high efficiency, low pres-sure drop, and good loading characteristics make them well suited for thisappl ication.For the filtration of high pressure process gases, membrane filters are morecommonly used. They also possess high efficiency, and this application cantolerate a high pressure drop. An advantage widely touted by membrane filtermanufacturers is their freedom from shedding in which parts of the filter breakloose and become reentrained in the filtered gas stream. Fibrous filters are morevulnerable to this shortcoming so, while acknowledging the superior loadingproperties of fibrous filters, membrane filter manufacturers recommend theiruse only as prefilters in high pressure gas filtration, a view not totally sharedby fibrous filter vendors.1.3 Historical BackgroundHistorically, filtration can be traced to efforts over the last 2000 years toprotect workers from toxic dusts and smoke. For a detailed summary of thehistory of aerosol filtration, see Reference 5. The 20th century has seen a burstof activity in the development of filters. The two World Wars stimulated sig-nificant advances in filtration in preparation for anticipated chemical attacks.In particular, the development of respirators has stimulated a developmentof good understanding of the physics of filtration and the fabrication of media.Ventilation systems using high efficiency filters were developed duringWorld War II. Before 1950, small quantities of HEPA filters were bu i1t by theUnited States Army Chemical Corps for the containment of radioactive aero-sols. These filters were called super-impingement or super-interception filters.Later they were called absolute filters and in the 1960s became known as HEPAfilters. These first filters were made from esparto grass, and later asbestos fiberswere added. In the 1950s, glass fiber paper was developed and used exclusivelyas the med ia. The asbestos was removed in the late 1970s because of the poten-t i a health risk.In the late 1950s, the first horizontal laminar flow clean bench was de-veloped at the Sand'ia Laboratories. The air cleaned by a HEPA filter was usedto create a clean environment over the top of the work su rface. Clean roomtechnology was developed during the 1960s in the aerospace industry becauseof the need to fabricate precision mach inery. The tech nology was appl iedto a number of industries requiring a highly controlled environment, such asthe manufacture of semiconductors, precision machinery, and pharmaceuticals.The use of membranes as a filter med iu m dates back to the late 1880s.The early membranes were made by gelling and drying colloidal solutions ofcellulose esters. These early membranes were used only for ultrafiltration ofliquids because the material was unstable when dry. During World War II, theGermans made advances in the technology. This advanced technology thenwas adapted to many civilian areas following the war. (The application of mem-branes to air filtration was reported by Goetz in 1953.6) Today membranesare made from a wide range of materials and are widely used both in the manu-facture of point-of-use filters and also for the sampling of particles.8 Handbook of Contamination Control in Microelectronics1.4 Chapter OrganizationThis chapter will cover the filtration of both air and gases. Three generalareas will be covered: the fundamentals of filter behavior (Section 2); filter testmethods (Section 3); and then the applications of filters (Section 4). Under-standing the fundamentals of filtration is important to allow the user to under-stand and interpret the filter operational data. Filter test methods, to a greatextent, should provide an indication of the performance of the filters. The de-mand for clean manufacturing conditions has resulted in higher efficiency filters,thereby prompting the development of a number of new filter test methods.Finally, the applications of filtration in air and gas filtration reflect the widevariety of filter media and an overall idea of the many uses of filters.2. FILTRATION FUNDAMENTALSThe goal of filtration theory is to be able to satisfactorily predict filtercollection efficiency, resistance (pressure drop/face velocity) and lifetime fromthe independent variables describing its structure and operating conditions. Thereview presented here is of both fibrous filter theory and the capillary tubemodel theory traditionally invoked to predict merrlbrane filter performance.Based on these theories, equations for predicting filter collection efficiency andresistance for both fibrous and membrane filters will be presented, from whichparticle size of maximum penetration and filter media figure of merit can thenbe derived. Filter lifetime cannot yet be predicted theoretically for real filters.While the effects of filter loading upon pressure drop and penetration are beingstudied theoretically using model filters7and lifetime prediction in terms ofpressure drop increase may eventually be possible, only empirical results can bepresented now. In addition, all the theoretical expressions to be presented in thefollowing sections strictly apply to only clean, virgin fibers and filters.2.1 Fibrous Filter TheoryThe classical collection membranes for particle capture from a fluid flowingpast a fiber are illustrated in Figure 5. These five mechanisms are the majorparticle removal mechanisms over the entire particle size spectrum at ambientconditions. Following are brief descriptions of each mechanism.(a) Gravity. Large particles (>5 t-tm) will settle out of the gas stream.Th is action occurs independently of the fiber and depends on thegas velocity and the particle mass.(b) Direct Interception. A particle, following the upstream flow stream-line, collides with a fiber and is collected. This mechanism can beimportant over the full particle size spectrum, depending on theratio of the particle diameter to the fiber diameter.(c) Inertial Impaction. Inertial impaction occurs when the inertia ofthe particle causes it to deviate from its initial streamline and collidewith the collector (a fiber of the filter). Inertial impaction dependson the gas's viscosity and velocity and also on the particles's diam-eter and density. Normally, inertial impaction is an importantAerosol Filtration Technology 9collection mechanism for particles greater than about 0.5 J.Lm indiameter.(d) Diffusion. Small particles 0.5 J.Lm) have a low enough mass tohave their trajectories altered by collision with gas molecules(Brownian motion). Smaller particles have higher diffusion rates.On an average, the particles follow their streamlines. However, therandom deviations from streamlines (due to diffusion) lead to afinite probability of collection by the fiber.(e) Electrostatic Mechanisms. Electrical charges on either the particleor the fiber, or both, create attractive electrostatic forces betweenthe fiber and the particle. This results in collection of the particle.PARTICLES..--/ ,I I' ......,/ FLOW STREAMLINES..... -"I \" I..... ATTRACTIONINTERCEPTIONGRAVITVFigure 5: Classical mechanisms: aerosol capture by a fiber.These mechanisms are considered to act independently of one another,but their contributions are not necessarily additive. Most of the filter modelingwork considers three of the mechanisms to be the most important-directinterception, inertial impaction, and diffusion. The effect of gravity is oftenneglected when discussing small particles because of its minimal effect. Electricalforces have often been neglected because they were considered as only second-order effects. Models which do not include the electrical forces have provensuccessful in predicting performance. Therefore, this assumption appears justi-fied for uncharged particles without externally applied fields. However, if eitherthe particles are charged or external electric fields exist (in an effort to modifyfilter performance), or if both conditions exist, then the electrical forces maydominate. Section 2.3 discusses these electrical forces which will be ignored untilthen.10 Handbook of Contamination Control in MicroelectronicsThe relative importance of diffusion, interception, and inertial impac-tion can be judged from Figure 6 (where the single-fiber efficiencies resultingfrom each mechanism are calculated).s These calculations show that: diffusiondominates for those particles with diameters smaller than about 0.2 J.1m; inertialimpaction and interception dominate for particles with diameters larger thanapproximately 1 J.1m; and all the mechanisms may be significant for particleswith diameters between 0.2 and 1 pm. An interaction term was omitted fromthe plot for clarity. Thus, the three mechanisms' efficiencies do not sum to theoverall efficiency. Other models would provide slightly different estimates ofsingle-fiber efficiency, but the overall trends remain the same for all the models.1.0 0.1 0.01,\\~~, ', I'\ Overall efficiency I" I ,~ I'Efficiency due to ~ , I'diffusion " l, AIPaper filter 0.01 cm thick,', ,4packing density ex = 0.1, "IIfiber radius of 1 ~ m , , I Iair velocity of 10 cm/s 'vI I(after Davies, 1973) ~ , JEfficiency due to interception;' 'f,Efficiency due to inertial impaction I....J "I I '"1.00.10.010.001Particle Diameter, IlmFigure 6: Single-fiber efficiencies (due to diffusion, interception, and inertialimpaction).2.1.1 Particle Trajectory Modeling. In order to fully understand how thethree most important mechanisms operate, it is necessary to consider the variousmodel s which describe them. A popular approach to model ing fibrous filtrationhas been particle trajectory analysis. In this approach, the single-fiber collectionefficiency is estimated by calculating the limiting trajectory allowing particle-Aerosol Filtration Technology 11fiber contact. Figure 7 displays schematically the geometry of a trajectory model.The particle trajectories are calculated by solving the equation of motion of theparticle. The limiting particle trajectory is the trajectory which just brings theparticle in contact with the fiber. The limiting trajectory, at a sufficient distanceupstream of the fiber, is parallel to the main flow direction at a distance (Yc)from the centerline of the fiber. The single-fiber efficiency is the ratio of Yc tothe fiber radius. The single-fiber efficiency can have a value greater than 1. Thisapproach is particularly suitable for the gravitational, inertial impaction, andinterception mechanisms. Albrecht8and Sel1 9used this approach in their pio-neering work, and the subject has been thoroughly reviewed by DaviessandPich.tOStreamline Particle starting position ... GasUooFigure 7: Particle geometry and fiber geometry in a trajectory model.Trajectory modeling has usually been approached by modeling one, or per-haps two, of the fundamental mechanisms to obtain an expression predictingsingle-fiber collection efficiency for the mechanism(s). Different researchersarrive at different results, primarily because of the different assumptions con-cerning the flow field near the fiber and also the simplifications required todevelop solutions. The results can be presented in the form of: plots of theresults of numerical solutions; semi-empirical expressions; or analytical solutionsfor the simplest cases. The separate efficiency mechanisms are then combinedto give an overall single-fiber efficiency. This overall efficiency can be related(through filter properties) to the overall filter efficiency.The study of filtration involves applying classical fluid mechanics to theflow around obstacles. In doing so, it is assumed that: the gas stream is a con-tinuum and the Navier-Stokes expressions are applicable. This assumption shouldbe examined to model high-efficiency filters. The fiber Knudsen number is aparameter that relates the fiber size to the distance traveled by gas moleculesbetween collisions (the mean free path). It is given by:12 Handbook of Contamination Control in Microelectronics(1 ) Kn =IIRtwhereKn Knudsen numberI mean free path of gas molecu IeRf fiber radius.The mean free path for air is given by:(2) I =J!:...['1r M]V2e 2 kTwherep viscosity of gasp gas densityM molecular weightk Boltzmann's constantT absolute temperatureI 0.065 11m at normal temperature and pressure in air.Devienne11classified the flow for which the Knudsen nUrTlber is between 0and 0.001 as the region in which continuum fluid mechanics is appropriate, andthe Navier-Stokes equations can be used to describe the flow. The region boundedby 0.001 < Kn < 0.25 is characterized as the slip-flow regime. Flow within thisregime can be modeled theoretically by using the classical hydrodynamics ex-pressions with modified boundary conditions. The region with 0.25 < Kn < 10is known as the transition regime, and flow for the region for wh ich the Knudsennumber is greater than 10 is known as the free molecular gas flow regime. Thelatter two regimes are not easily dealt with theoretically. For conditions typicalin a clean room (70F and atmospheric pressure), the continuum range is forfiber diameters greater than 130 pm; sl ip flow is for fibers between 130 to 0.5211m; and the transition is from 0.52 to 0.013 pm. Typically, the filters of interestinclude many sub-10 11m fibers. Therefore, the flow regime of interest falls inthe sl ip-flow range.The isolated cylinder model, where the filter is treated as an array of non-interacting cylinders, is the least complicated of all the models. Contemporarymodels treat the filter matrix as: parallel cylinders at random spacing; regulararrays of cylinders; or random'y-rotated layers of parallel cylinders. We willlimit discussion to the Kuwabara theory which is based on the cell model ofrandomly spaced cylinders and is discussed next.The cell model has provided the most successful description of fibrousfilters. The filter is treated as a system of parallel, randomly-spaced fibers wh ichare oriented perpendicular to the flow. Each fiber is surrounded by an inde-pendent body of fluid (Figure 8a). The polygonal cells are difficult to treatmathematically, and the treatment is simplified to regard the cells as identicalconcentric cylinders with the dimension shown in Figure 8b. The packingdensity, a, of the cell is required to equal the fiber volume divided by the totalvolume. For this particular case:Aerosoi Filtration Technology 13Uoo = velocity of gas in the filterp = gas density/1 = gas viscosity.a. Schematic representation of parallel cylinders with the sameradius (randomly placed and homogeneously distributed).Total cell volume =filter volumefiber volume/cell volume = R ~ / R ~ = packing density. 0:Fiber\\ I\ I\ I, /, /" ""..... ,,"............ _---..-Uoo-'"..,---- -......... / Cell Boundary-' .....,/I \I \I \I \I \____.....;.-_-L-_ I/b. Simplified cell used for analysis.Figure 8: Cell model of filter, after KuwabaraY14 Handbook of Contamination Control in Microelectronics(3) a =(fiber radius/cell radius)2This model produces a flow field that is valid for a system of cylinders.The stream function of the Kuwabara12continuum hydrodynamics n10delis given by:(4)whereKrRfUUoc =a(Ju00 r sin 8 [ R ~ ex ex r2]l/; = 2 In ( ~ ) - 1 + ex + - (1 - -) - -2 K Rf r22 2 R ~-0.75 - 0.5 In a + a - 0.25a2radial coordinateradius of fiberfilter face velocityfluid velocity within filter = U/(1 - a)packing densityangular coordinate.K is called the Kuwabara hydrodynamics factor. This description of the flowfield has been validated experimentally by Kirsch and Stechkina.13This formu-lation is the basis of the descriptions which follow.2.1.2 Pressure Drop. The relationship between pressure drop and flow is ofgreat practical significance because the amount of energy expended in collectingparticles is often as important as the efficiency with wh ich they are collected.Daviessdefines filter flow resistance by restating Darcy's law to read:(5)wherew resistanceLl p = pressure dropQ volumetric flow rate.w =d P/QDarcy's law which states that the pressure drop is proportional to flow ratehas been found to hold for many good-quality air filters. Thus the ~ p/G termshould be constant and the filter resistance constant. Actually, many lowerefficiency filters display an increase in flow resistance as the flow rate is in-creased. Daviesscites compression of the filter and/or a change in the natureof the flow from Stokes's flow to laminar, inertial flow in the range 0.5 < Re< 20 as possible explanations, where Re is the Reynolds number based on thefiber diameter and is given by the following:2 Rt Q UooRe =p.whereRf fiber radiusAerosol Filtration Technology 15Filter flow resistance has been modeled theoretically on the basis of eitherthe flow around fibers (isolated or in arrays) or as the flow through channels.These theoretical results have been extended to give an estimate of filter pressuredrop by relating the fiber (channel) resistance and length to the packing density,thickness, and other filter properties. There are also a number of empirical orsemi-empirical expressions for pressure drop as a function of filter parameters.The following section presents an expression for fiber drag derived from modelsof the flow patterns in filters.2.1.2.1 Fiber Drag Expressions. The drag force that a flowing fluid exertson a body immersed in the fluid is often expressed as:(6)Fd = (a) (KE) (Cd)whereFd drag forcea characteristic areaKE kinetic energy termCd drag coefficient.For a cylinder transverse to the direction of the fluid flow, the area termis taken as the projected area. This is 2(Rf)(lf), where Rf is the fiber radius andIf is the fiber length. The kinetic energy term is taken as 1/2 p Uoo, where p isthe fluid density and Uoo is the approach velocity of the fluid to the fiber withinthe filter. Combin ing these expressions with Equation 6, the following ex-pression for the drag force is obtained:(7)A dimensionless drag force per unit length of fiber can be defined as:(8)FdF* =- II UI r 00fwhereF* dimensionless drag force andJ1 fluid viscosity.The dimensionless fiber drag is used because it is convenient mathematically.It can be calcu lated by integrating the tangentia I stress on the fiber over the sur-face of the fiber, using the flow field presented above to provide an expressionfor the tangential stress. The relationsh ip between F* and Cd can be determinedby combining Equations 7 and 8 to obtain:(9) F* = V2 Re Cd2.1.2.2 Filter Pressure Drop. The fiber drag can be related to filter pressuredrop in filters if certain parameters are known. In the development that follows,the filter is assumed to be homogeneous. The total fiber length, tlf, for a filter of16 Handbook of Contamination Control in Microelectronicsuniform thickness or depth L and with a unit area perpendicular to the flow is:(10)wheretit = total fiber length per filter area.The pressure drop across a filter equals the product of the drag force per unitlength of fiber and the total length of fiber per unit area. The drag force perlength of fiber can be obtained from F*, the dimensionless drag per unit lengthof fiber. Combining Equations 8 and 10 as suggested leads to an expression forpressure drop:(11 )where~ p = filter pressure dropIf the Kuwabara12flow field is used for the dimensionless drag, the pressuredrop is given by:(12) JtUoo L 4a~ p = ----R1 K2.1.3 Particle Collection Efficiency. The efficiency of a filter is related topenetration by:(13)E =1 - NINo =1 - PtwhereN is the concentration leaving the filterNo is the concentration entering the filterPt is penetration (== NINo)'Penetration as a function of the properties of the filter is given by:(14) Pt =exp [ - 211 a LJ1r RtwhereRt mean fiber radius~ packing density of the filterL thickness of the filter17 single fiber efficiency.Aerosol Filtration Technology 17Part of a fundamental evaluation of fibrous filtration consists of quantifyingthe contribution of various mechanisms to the single fiber efficiency. Of particularinterest is the single fiber efficiency as a function of particle diameter. By de-termining either the maximum penetrating particle diameter or the minimumefficiency, one can develop a conservative estimate of the filter's protection.This section examines each primary collection mechanism and its contribution tofilter efficiency and then shows the effects of combining different mechanisms.2.1.3.1 Particle Collection by Inertial Impaction. Inertial impaction is animportant collection mechanism only for the largest particles of concern, typi-cally those larger than 111m. The true collection which results solely frominertial impaction (without an interception component) requires that particles(to have their proper mass) be treated as if collision did not occur until thecenter of the particle reached the fiber. This is not physically true. Therefore,pure inertial impaction is simply a theoretical construct.Numerical methods are normally used to solve the equation of motion toestimate efficiency. Thus, the "results are often presented in the form of graphsor tables. As seen in Figure 9, the Stokes number Stk is the key parameter inparticle collection by inertial impaction. Stk is definedsas:(15)wherePp particle densitydp particle diameterUoo = velocity of fluid undisturbed by fiberC 1 + 2.468 I/dp + 0.826 I/dp exp(-0.452 dp/I)I gas mean free path.The Stokes number is the ratio of the distance a particle would travel againstfl.uid10drag ( h a v i n ~ an initial velocity U) to the fiber radius. Other authors (i.e.,Plch ) have defined the Stokes number with respect to the fiber diameter.For potential flow, the single-fiber collection efficiency resulting from im-paction is a function of the Stokes number only. In a viscous flow field, thiscollection is a function of both the Reynolds number and the Stokes number foran isolated cylinder (single fiber) in a system of cylinders. The packing densityis also important.The single fiber efficiency curves in Figure 9 (17st = single fiber efficiencydue to inertial impaction) were originally developed for the interpretation ofmeasurement of droplet sizes in supercooled clouds. A dimensionless group wasused as a parameter to include particle diameter effects over a wide range offlow rates:wherep gas density00:cQ):::J0-CJoo7\o......,(jo:::Jr+Q)3:::JQ).-+o:::J(jo:::Jr+Q.:::J

("')(3coco("')r+(3:::J("')en000to-t:::: ..L..4 ........

- kr-' H)-....::::::

t:::=l-""'t-r-'---

.-'...-

..-"-'

Po ./---"...

0 "..0 100 ./ ,/ 100'"6 1.000>./ i/V I--"'"V"""o 10.000 ./., 50,0001I /' .,VV III'V 1I/ IVV v // VI /1 / ) / V VI LII/ V V '" l/ V/ V VIfV / V/V/ /VI 'IlIl' /.,1 VVV V

VV

- - - -20 100 2Xl I2486OJ1.0...UI

Stk = Qp UooC18J.t RfFigure 9: Collection efficiency for cylinders in an inviscid flow with point particles. (Courtesy, John Wiley and Sons)Aerosol Fi Itration Technol ogy 19Uoo = gas velocitydp particle diameterJ.1 gas viscosityPp density of particlesRep = Reynolds number of the particleStk = Stokes number.2.1.3.2 Particle Collection by Interception. When applied to aerosol captureby an isolated single fiber, interception describes the collection of those aerosolsthat follow the fluid streamlines and make contact with the fiber. Strictlyspeaking, pure interception occurs only when the particles have size but no mass.Thus, the idea of collection resulting solely from interception is conceptuallyuseful but physically unattainable.An expression for single fiber efficiency based on particles interception inKuwabara flow was reported by Stechkina et al.14and is given by:(16) _ (1+R)-1 - (1+R) + 2(1-R) In(1+R)1JR - 2KwhereR the interception parameter, dp/2 RfK Kuwabara hydrodynamic factor defined earl ier.The interception parameter, R, is the ratio of the particle diameter to thefiber diameter. Potential flow collection by interception is a function of R only;for viscous flow, the Reynolds number becomes important. Again, in a systemof cylinders (fibers), the packing density is also important. For a fixed fibersize, the interception mechanism becomes more significant as the particle sizeincreases.2.1.3.3 Particle Collection by Diffusion. Diffusion is significant only forsmall 0.5 J.1m in diameter) particles. For these particles, the fluid streamlinesrepresent only the average path of the particle. Their motion actually consists ofmany random, zigzag deviations from the path of the streaml ines (both towardand away from the fiber). These large deviations from the streaml ines mean thatcontact with the collection surface can occur even though the streamline passingthrough the upstream position of the particle passes the fiber surface at a distancegreater than the particle radius. As the particle size decreases, the importance ofBrownian motion relative to the fluid flow increases until the distinction be-tween aerosol particle and gas molecule becomes insignificant.Diffusion modeling is based on the distance the particle can diffuse in agiven time. The distance equals the square root of the product of the particlediffusion coefficient and time. For collection on fibers, the residence time in thefilter is important. The Peclet number, Pe, relates flu id motion to the diffusioncoefficient and hence is important to modeling this form of collection. ThePeclet number is defined as:(17) 2 Rt UooPe=---o20 Handbook of Contamination Control in MicroelectronicswhereD particle diffusion coefficient = k T C/ 3 1T J.1 dpk Boltzmann's constantT absolute temperaturedp particle diameter.An example of the collection efficiency due to diffusion is given by Lee andLiu:15(18)(K -0.333170 = 2.6 --) Pe-0.6671-aThis expression shows that diffusive collection is a function of the Peclet numberand a hydrodynamics factor. In terms of filter and aerosol properties, collectionefficiency which results from diffusion increases with decreasing particle diam-eter, gas velocity, and fiber size.2. 1.3.4 Particle Collection by Diffusion and Interception. The efficiencyexpression for the combined mechanisms of diffusion and interception15is:(19) 217 = Pe-0.667+ O,R 1-a K 1+RThis combination of mechanisms should be adequate to describe the collectionof particles of diameters less than 0.5 J.1m. Lee and Liu16and RUbow4haveshown that a model based only on these two mechanisms can successfully pre-dict the particle size of maximum penetration and adequately predict filterefficiencies.2.1.3.5 Particle Collection by Diffusion, Interception, and Inertia. The fanmodel of Stechkina et al.14has been shown to fit both particle collection andpressure drop experimental data fairly well in the range of the most penetratingparticle size. The single fiber collection efficiency for the fan model is given by:(20)Diffusion:17 = 170 + lJO,R + lJR + 17St,RlJo = 2.9 K-0.333 Pe -0.667 + 0.624 Pe-1Diffusion-Interception:lJO,R = 1.24 K-0.5 Pe -0.5 RO.667Interception:Impaction-I nterception:lJSt,R =Computed numericallyAerosol Filtration Technology 21These expressions developed by Stechkina et al.14are similar but not identicalto those presented earlier. Stechkina et al. use them to compute the efficiencyof each layer in the model filters. The fan model assumes that the 'fibers haverandom orientation in layers. Each layer can be rotated, creating an illusion ofthe spreading of a lady's fan.These filter models are often applied by including an inhomogeneity factor,. This correction is included in the overall efficiency equations as follows:(21 )The inhomogeneity factor is an estimation of the nonuniformity of a real filterand is defined as the ratio of the drag forces acting per unit fiber length or pres-sure drop and model filter or:(22) . 'YlE =- or E ='Ylstd Ptheoretical~ Pexperimental\tVhere:77 theoretical single particle efficiency77st actual single particle efficiency.To summarize, the theoretical approach to predicting filter performanceconsists of combining fundamental expressions and empirical corrections. Thecomplexity of the filter material makes an exact prediction based on a prioriassumptions and measu res of the filter properties unl ikely. However, under-standing the mechanisms of filter behavior is very useful in gauging the limita-tions of various media.2.1.3.6 Correlation of Actual Performance with Models. Many experimentalinvestigations of fibrous filter efficiency have been reported. Only those modelsthat make use of flow fields based on systems of cyl inders using the Kuwabaraflow equations will be considered here.Figure 10 compares the theories of Stechkina et al.14and Yeh and Liu 17with experimental data collected by Lee and Liu.16The proper constants ac-counting for lack of uniformityI random fiber orientation, and similar nonidealconditions were incorporated into the models. The inhomogeneity factor isabout 1.67 for these polyester filters. Overall, both theories fit the data well.The theory of Stechkina et al. is seen to agree closely at high velocities whiledeviating at low velocities. Stechkina et al.14have stated that some of theassumptions which they made while developing the impaction model (describedin their earlier paper) were inadequate. On the other hand, the theory of Yehand Liu is seen to agree best with the data at low filtration velocities.A semi-empirical method based on earlier work by Friedlander18for gener-alizing filter efficiency data was reported by Lee and Liu.16Starting with thesingle fiber efficiency equation 20 for combined diffusion and interception andusing empirical coefficients22 Handbook of Contamination Control in MicroelectronicsFilter FaceVelocity, cm/s, 3 10Q 30,\,,,,,0.1~G).c~G)~ 0.01Ci3StechkinaYeh and Liu1 0.01 0.1Particle Diameter, JJmFigure 10: Comparison of data of Lee and Liu16with theories of Stechkinaet al.14and Yeh and Liu .17 Fiber diameter of 11 11m; packing density of 0.151.if both sides are multiplied byPe AI (1 +R).V2then the above equation can be reduced to(23)Therefore, the coefficients can be determined by plotting'YJ Pe A/v'1 +AAerosol Filtration Technology 23andwhere71 the single fiber efficiencyR interception parametera the fiber volume fraction or the filter sol id ityK the hydrodynamic factorPe Peclet numberThis correlation (Figure 11) was found to apply over a wide range of experi-mental conditions. Filter performance data can be plotted on a log-log graphwith two dimensionless parameters over a wide range of filter parameters andparticle diameters. The importance of this approach is that a semi-empiricalcorrelation can be used in applications of high-efficiency collection of sub-micron-sized particles. Thus, the correlation can serve as a guide in the designof filter med ia.2.1.3.7 Most Penetrating Particle Fiber. As shown in the description ofcollection mechanisms, fibrous filters function such that there is a particlediameter wh ich has the greatest penetration and thus the lowest efficiency.The most penetrating particle diameter depends on the filter fiber diameter,volume fraction, and the gas velocity. Lee and Liu19differentiated Equation20 for combined diffusion and interception mechanisms with appropriate cor-rections for slip and obtained:(24)wheredpmindpKIkJ1DfUaK IV2k T D ~ 2/9dp min =0.885 (1- oJ (-{{-)(u) for 0.075 < IIdp< 1.3most penetrating particle diameterparticle diameterhydrodynamic factormean free pathBoltzmann's constantviscosity of gasfiber diameter, 2 Rfgas velocityfiber volume fraction (packing density).This equation is shown in Figure 12. Increasing the velocity through thefilter and reducing the fiber diameter will reduce the value of the most pene-trating particle diameter. Emi and Kanaoka20compared the predictions ofEquation 24 with data obtained with HEPA media and found that the trendswere correctly pred icted.The particle size dependent penetrations have been measured for filtermedia typically used in both HEPA and ULPA filters. All of the mechanismsexpected from the theoretical development are evident in the results. The pene-tration data reported by Liu et al.21are shown in Figures 13 and 14. The mostpenetrating particle diameter and its dependence on flow velocity are apparent.24 Handbook of Contamination Control in Microelectronics10 ......---.-.......-0.01100J0/-0Uo em/secaI0 IA 300086'0 0 101-

0 30ct I.1"'- A 3a: (I'0 0.0434A'eu4) 30a. I 30.1513 10 30oy OJ:1:. I 2o l-a'3 -'3 I-a R2o. 77:: 1.6 (t Q)0; c:.... .Y- 1. Obviously, for Nr> 1, all particles are captured bythe filter which acts as a sieve. Spurny et al.27reported an equation for theefficiency of a filter from interception.(29)The impaction of particles around a pore on a filter surface was investigatedby Kanaoka et al.28The equations of motion were solved numerically with anorifice flow field. The results have the fam'iliar Sigmoid shape when the efficiencyis plotted as a function of the Stokes number.The collection of particles by inertial impaction is reported by Pich10as:(30)whereE1=2 Stk v73 + 2 Stk2,6 exp (- _1_ ) - 2 Slk2\ Stk${3= ~1 - ~ 1 - aParticle removal by diffusion in the pores is described by equations de-veloped from diffusion batteries. The entrance effects are neglected, and theproblem is reduced to diffusion of particles to a wall in laminar "flow at verysmall Reynolds numbers. The mechanism is described by a dimensionless param-eter:(31 ) N _ 4LDd - d2 UPwhere D is the diffusion coefficient of the particles. The efficiency of collectionof particles in a tube is described by Gormley and Kennedy29 as:(32)This equation is valid for Nd < 0.03. The equation derived by Twomey30 canbe applied for Nd -> 0.03. It is:34 Handbook of Contamination Control in Microelectronics(33)170 =1 - 0.81904 exp( -3.6568 Nd)- 0.09752 exp( -22.3045 Nd)- 0.03248 exp( -56.95 Nd)- 0.0157 exp( -107.6 Nd)The overall filter efficiency can be obtained by combining the expressionsfor the various particle collection mechanisms. It is assumed that each mech-anism is independent and the pores are not interactive. The total collectionefficiency is given by an expression developed by Spurny et al.:27(34) Er =11Stk + 110 + 0.15 11R - 17Stk 110 - 0.15 11Stk 11Rwhere 0.15 was empirically determined for Nuclepore filters. This equationtends to underestimate interception effects because of the construction of flownear the pore.Membrane filters demonstrate the same general particle size dependentshape of the penetration curve as do fibrous filters (Figure 16). Filters with veryhigh efficiencies or low penetrations are difficult to measure and result in in-complete curves.Nuclepore "filters have a similar kind of behavior except that the "window"between diffusional and inertial interception collection is broader (Figure 17).As the pore size of the filter decreases, the particle size correspond ing to theminimum in collection efficiency becomes smaller. The curves reported bySpurny et al.27exhibited good agreement between theory and the experimentaldata.2.3 ElectrostaticsElectrical forces can play an important role in improving both -fibrous andmembrane filter performance by increasing filter collection efficiency and, insome cases, decreasing pressure drop. In addition to being an important con-sideration in laboratory investigations, electrostatic enhancement of filters hasbeen suggested as a way of obtaining cleaner working spaces.There are three major electrical forces between fibers and particles:1. image force (charged particles only)2. dielectrophoretic force (charged collector only)3. coulomb force (charged particles and collector)The image force is the attractive force created between a charged body and acharge induced in an adjacent, electrically neutral surface (i.e., a fiber or amembrane collector). The charged particle polarizes the fiber by causing acharge redistribution within the fiber. The resulting force can be calculatedby treating the induced fiber charge as a fictional charge located at a pointwithin the fiber. Th is computational approach is the classical method of imageanalysis. The image force is always attractive.The dielectrophoretic force also results from polarization. This force resultswhen charge separation in an electrically neutral body takes place in a nonuni-form electric field. The charge of the sign displaced toward the region of higherAerosol Filtration Technology 35electric field creates a larger coulomb force on the body than the equal chargeof the opposite sign displaced toward the region of lower electric field. P o l a r i z a ~tion in a nonuniform electric field always creates a net electrical force in thedirection of increasing field. Since the electric field surrounding electricallycharged fibers always decreases in magnitude with radial distance from thefiber, a particle will always experience an attractive dielectrophoretic forcepulling it towards the fiber.0.1247950.01~0149z0to-cra::t-wZLUa.0.001 38MILLIPORE TYPE AATEST AEROSOLNoel--- DOP0.00010.01 0.1PARTICLE DIAMETER. fLm1.0Figure 16: Comparison of the experimentally determined aerosol penetrationdata for the AA filter observed in this study and that obtained by Rubow.4(Courtesy, Institute of Environmental Sciences)36 Handbook of Contamination Control in Microelectronicsii' ITTTr '!\+1.00.8E 0.60.4,0.2\ f! 00.001 0.01 0.10 1.0 10.0r(lm)Figure 17: Comparison of the collection efficiency as theoretically predictedand experimentally determined for a 5 J1m pore diameter Nuclepore filter at aface velocity of 5 em/sec. (Courtesy, ASTM)The coulomb force is the famB iar repulsion or attraction between charges.A special case of the coulorTlb interaction exists under concentrated particleconditions from the net charge of the particle cloud or space charge. When theparticle cloud is uniformly charged, mutually repulsive coulomb forces drive theparticle apart and toward any grounded or uncharged surfaces (Le., a fiber).The space charge precipitates or deposits charged particles on surroundingsurfaces until the. space charge electric field decreases to insignificant values.Only then can the charge on an individual particle dominate particle behaviorthrough the image force. Several approaches have been developed over the yearsto uti! ize these forces in various combinations. The particle filter under variousconditions of charge and fields has been reviewed by Pich.32The mathematical expressions to describe filter electrical effects were re-ported by Shapiro et al.33and are summarized in Table 5. The coulomb