Microelectronics Research and Development

March 1986

NTIS order #PB86-205200

Recommended Citation:U.S. Congress, Office of Technology Assessment, Microelectronics Research andDevelopment–A Background Paper, OTA-B P-C IT-40 (Washington, DC: U.S. Gov-ernment Printing Office, March 1986).

Library of Congress Catalog Card Number 86-600509

For sale by the Superintendent of DocumentsU.S. Government Printing Office, Washington, DC 20402

Foreword

Microelectronics, the fundamental! building block of today’s pervasive infor-mation technologies, has progressed at a tremendous rate over the last few de-cades and has become a vital part of U.S. commerce and defense. This fast-changingfield depends critically on aggressive research and development (R&D) programs.Federal participation in microelectronics R&D is twofold: the government direct-ly funds a significant fraction of the activities, and several Federal policies in-directly affect private support of R&D. Hence, Congress has a major role to playin microelectronics R&D—a role that can be illuminated by a better understand-ing of today’s activities in this area.

In November 1984, the House Committee on Science and Technology askedOTA to prepare a background paper on microelectronics R&D as a follow-on tothe OTA assessment of Information Technology R&D: Critical Trends and Is-sues, released in February 1985. This background paper describes the current stateof research and development in microelectronics by examining the range of R&Defforts and the sources of Federal and private support for R&D. It also presentspotential policy concerns that stem from existing arrangements for direct Feder-al support and from changes underway in microelectronics R&D.

Many members of the diverse microelectronics community, spanning the spec-trum from basic research to applied development, contributed their knowledge,insight, and viewpoints to this background paper. These included scientists, en-gineers, managers, and observers in Federal agencies and laboratories, industry,universities, and other organizations. OTA is pleased to thank all participantsfor their assistance. However, OTA assumes full responsibility for the contentsof this study.

Ii;

OTA Microelectronics Research and Development Staff

John Andelin, Assistant Director, OTAScience, Information, and Natural Resources Division

Fred W. Weingarten, Communication and Information Technologies Program Manager

Project Staff

Arati Prabhakar, OTA Fellow and Analyst

Contractor

Susan P. Walton, Editor

Administrative Staff

Elizabeth A. Emanuel Shirley Gayheart*

Patricia M. Keville Audrey Newman

*Deceased, Dec. 11, 1985.

iv

Reviewers

John A. ArmstrongVice President, Logic and MemoryIBM Watson Research Center

Donald S. BeilmanPresidentMicroelectronics Center of North Carolina

Arpad A. BerghDivision Manager, Device Science and

Technology ResearchBell Communications Research

Eliot D. CohenNavy Director, Very High Speed

Integrated Circuits ProgramSpace and Naval Warfare Systems

Command

Larry R. CooperProgram Manager, Electronics DivisionOffice of Naval Research

Kenneth L. DavisHead, Electronics DivisionOffice of Naval Research

Charles B. DukeSenior Research FellowXerox Webster Research Center

James M. EarlyScientific Advisor to the DirectorFairchild Research Center

Jeffrey FreyProfessor, Electrical EngineeringCornell University

Paolo GarginiCoordinator, One Micron ProgramIntel Corp.

George H. HeilmeierSenior Vice President and

Chief Technical OfficerTexas Instruments, Inc.

Evelyn HuProfessor, Electrical and Computer

EngineeringUniversity of California at Santa Barbara

Robert W. KeyesResearch Staff Member, Semiconductor

Science and TechnologyIBM Watson Research Center

Jack S. KilbyDallas, Texas

Conilee G. KirkpatrickDirector, Advanced Technology

ImplementationRockwell Microelectronics R&D Center

Fred W. Kittler, Jr.Vice PresidentJ.P. Morgan Investment

Paul LoslebenAssistant Director, VLSIDefense Advanced Research Projects

Agency

Noel MacDonaldProfessor, Electrical EngineeringCornell University

Kevin J. MalloyProgram ManagerAir Force Office of Scientific Research

Thomas C. McGillProfessor, Applied PhysicsCalifornia Institute of Technology

Carver A. MeadProfessor, Computer ScienceCalifornia Institute of Technology

Robert E. NahoryDistrict Research Manager, Optical

Properties of Semiconductors ResearchBell Communications Research

Martin A. PollackHead, Photonic Devices Research

DepartmentAT&T Bell Laboratories

Richard A. ReynoldsDirector, Defense Sciences OfficeDefense Advanced Research Projects

Agency

Anthony A. Immorlica, Jr.Director, Microelectronics CenterMcDonnell Douglas Corp.

Sven A. RoosildAssistant Director, Electronic Sciences

DivisionDefense Advanced Research Projects

Agency

Joseph A. SaloomSenior Vice PresidentM/A-COM

George M. ScaliseSenior Vice President and Chief

Administrative OfficerAdvanced Micro Devices

Donald J. SilversmithDirector, Solid State and Microstructures

Engineering ProgramNational Science Foundation

Michael A. StroscioProgram Manager, Electronics DivisionArmy Research Office

Other Contributors

Robert S. BauerManager, Integrated OptoelectronicsXerox Palo Alto Research Center

Jules A. BellisioDivision Manager, Digital Signal

Processing ResearchBell Communications Research

Robert D. BurnhamResearch FellowXerox Palo Alto Research Center

Praveen ChaudhariVice President, ScienceIBM Watson Research Center

Melvin I. CohenDirector, Interconnection Technology

LaboratoryAT&T Bell Laboratories

Robert DennardIBM FellowIBM Watson Research Center

James V. DiLorenzoVice President, Research and DevelopmentMicrowave Semiconductor Corp.

Jimmie R. SuttleDirector, Electronics DivisionArmy Research Office

Paul TurnerManager, Technology AssessmentXerox Palo Alto Research Center

Horst R. WittmannDirector, Electronic and Material SciencesAir Force Office of Scientific Research

Edward D. WolfProfessor and Director, National Research

and Resource Facility for SubmicronStructures

Cornell University

Dean EastmanIBM FellowIBM Watson Research Center

Lester F. EastmanProfessor, Electrical EngineeringCornell University

Barry K. GilbertSpecial Purpose Processor Development

GroupMayo Foundation

Kenneth L. GuernseyGeneral PartnerHambrecht & Quist

Richard HendersonHughes Research Laboratories

M.P. LepselterDirector, Advanced LSI Development

LaboratoryAT&T Bell Laboratories

Michael A. LittlejohnProfessor, School of EngineeringNorth Carolina State University

vi

James W. MayerProfessor, Materials Science and

EngineeringCornell University

James H. McAlearPresidentGentronix Laboratories, Inc.

James MeindlProfessorStanford University

C. Joseph MogabMicrowave Semiconductor Corp.

Gordon E. MooreChairman and Chief Executive OfficerIntel Corp.

Stephen NelsonDirector of ResearchCray Research, Inc.

Gregory H. OlsenPresidentEpitaxx, Inc.

Phil ReidTRW

John B. SalazarVice President, Corporate BankingBank of America

George E. SmithHead, MOS Device DepartmentAT&T Bell Laboratories

Louis TothDirector, Ceramics and Materials ProgramNational Science Foundation

P. K. VasudevSenior Staff EngineerHughes Research Laboratories

William D. WartersAssistant Vice President, Communications

Systems and Network TechnologyBell Communications Research

Hugh A. WatsonHead, VLSI Processing Technology and

Fabrication DepartmentAT&T Bell Laboratories

Benjamin A. WilcoxAssistant Director, Materials Science

DivisionDefense Advanced Research Projects

Agency

Gerald L. WittProgram ManagerAir Force Office of Scientific Research

Irving Wladawsky-BergerVice President, SystemsIBM Watson Research Center

Rainer ZuleegMicroelectronics CenterMcDonnell Douglas Corp.

OTA Reviewers

Lauren W. Ackerman Jim Dray Eugene FrankelResearch Analyst Research Analyst Senior Analyst

John A1ic Greg Eyring Chuck WilkProject Director Project Director Project Director

Earl DowdyResearch Analyst

vii

Contents

Chapter Page1. ISSUES IN MICROELECTRONICS

RESEARCH AND DEVELOPMENT,,. 1Main Findings of the Study: Trends

in Microelectronics R&D . . . . 1Findings About Institutional Support . 1Findings About Technological Trends 2

Trends in Fabrication Technology .,. 2Trends in Design . . . . . . . . . . . 2

Potential Policy Concerns . . . . . . . . . . 3Federal Response to Changes in

Microelectronics R& D . . . . . . . . . . . . 3Federal Response to Changes

in the Industry . . . . . . . . . . . . 3Federal Response to Technological

Trends. . . . . . . . . . . . . . . . . . . . . . 5Direct Federal Support for Micro-

electronics R&D: Policy Questions 5Multiple Sources of Federal Funding:

Pros and Cons, . . . . . . . . . . . . . 5Questions Raised by Department

of Defense Activities . . . . . . . . . 6

2. INTRODUCTION: MICROELECTRONICSTECHNOLOGY AND R&D . . . . . . . . 9The Impact of Microelectronics and

Evolution of the Industry . . . . . . . 9The Nature of Research and Development

in Microelectronics . . . . . . 10

3.CURRENT RESEARCH ANDDEVELOPMENT ACTIVITY . . . . . . 11Advanced Silicon Integrated Circuits, . . . 11

Circuits and Packaging, . . . . . . . . 11Design and Fabrication Processes . . . . 12Circuits for Specific Markets . . . . . . . 12

Microelectronics Based on GalliumArsenide and Other CompoundSemiconductors ..,..., . . . . . 12

GaAs Digital Integrated Circuits ..., 13Optoelectronics. . . . . . . . . . . . . . . . . 14

Discrete Optoelectronic Devices . . . 14I n t e g r a t e d O p t o e l e c t r o n i c s 1 5Superlattices andOtherQuantum-

Effect Structures. . . . . . . . . . . . . . . . 15Microwave Devices ..., . . . . ..., . 16

Nonsemiconductor Integrated CircuitTechnologies . . . . . . . . . ., ..., 16

Josephson Junctions . . . . . . . . . . . . . . . . 16Biomolecular Electronics. . . . . . 16

Chapter Page4.INSTITUTIONAL SUPPORT FOR

MICROELECTRONICS R&D . . . . . . . . . . 19Federal Support, . . . . . . . . . . . . . . 19

Department of Defense ., . . . . . 19Research for the Navy, Air Force,

and Army. . . . . . . 20Defense Advanced Research

Projects Agency . . . . . . . . . . . 20DOD Special Programs . . . 21

National Science Foundation ., . . 22Private Sector R&D . . . . . . . . . . . . . . . . . . 23

Captive Manufacturers . . . . ., . . ., . 23Merchant Companies. ..., . . . ., 24

Cooperative R&D . . . . . . . . . . . . . . . . . . . . 25Universities and R&D, . . . . . . . . . . . . . . . . 25Foreign Activities . . . . . 26

Appendix PageA. CURRENT MICROELECTRONICS

TECHNOLOGY . . . . . . . . . . . . . . . 27Electronics Technology From Vacuum

Tubes to Integrated Circui ts 27Microelectronic Devices . 28

Types of Integrated Circuits. ., ., . 28Other Semiconductor Devices . . . 31

Technologies To Produce SemiconductorMicroelectronics, . . . . . . . 31

Circuit Design,..,.. . . . . . . 31Chip Fabrication ., . . . . 32

B. GLOSSARY OF TERMS . . . . . . . . 35

Table

Table No. PageA-1. Levels of Integration . . . . 29

Figures

Figure No. Page1. Periodic Table of the Elements . 13A-l. Generations of Electronics

Technology. . . . . . ..., . . . . . . . 27A-2. Types of Microelectronic Products . . . 28A-3. Microscopic Sizes .. . . . . . . . . . . . . . 30A-4. First of Four Mask Processes

To Make an N-Channel MOSFET...... 32

Vlll

Chapter 1

Issues in MicroelectronicsResearch and Development

Microelectronics is the cornerstone of the in-formation technologies that pervade virtuallyevery aspect of contemporary life. These com-puter and communications technologies arethe basis for changes such as automation,energy conservation, and pollution control inoffices, factories, automobiles, and homes; su-percomputers for applications from weatherprediction to computational research; new ca-pabilities in financial services; new means ofstoring and playing back audio and videorecordings; advanced telephone and televisionsystems; and complex weapons systems fornational defense. Each of these areas is criti-cally dependent on microelectronics technol-ogy. Furthermore, the microelectronics indus-try—and the industries that depend on it—are vital to the U.S. economy.

Research and development (R&D) effortshave fueled progress in microelectronics tech-nology at an extraordinary rate. Since the in-vention of the integrated circuit (IC) 27 yearsago, the capabilities of these devices havemore than doubled every 2 years. Currently,an IC with several hundred thousand compo-nents can be purchased for a few dollars-lessthan the 1950’s price for a single component.

The Federal Government has historicallyplayed vital roles, some direct and some in-direct, in microelectronics research and devel-

opment. These multiple involvements make itimportant to understand both the structureof institutional support and the implicationsof current technological trends in consideringthe many Federal policies that affect micro-electronics R&D.

Today, many factors, including shifts in in-dustry structure and limitations posed bytechnological trends, raise questions concern-ing the types and levels of Federal support formicroelectronics R&D. To address these is-sues, this OTA background paper, requestedby the House Committee on Science and Tech-nology, describes the current state of micro-electronics research and development by ex-amining the technologies emerging from R&Defforts and the range of institutional supportfor R&D. Although other relevant Federal pol-icies are discussed to some extent, the primaryfocus of the paper is the role of direct Federalsupport for microelectronics R&D.

This chapter: 1) summarizes the OTA find-ings, and 2) discusses potential Federal pol-icy implications that they suggest.1

1 In this paper, the term “microelectronics’ is used to describeminiature electronic devices in general, Readers whose back-ground in the technology is limited may wish to review app.A before proceeding.

MAIN FINDINGS OF THE STUDY:TRENDS IN MICROELECTRONICS R&D

Findings About Institutional Support universities all contribute in different ways toA broad range of organizations supports progress in the field. Among international

microelectronics R&D. In the United States, activities, Japanese R&D efforts predominate;Federal agencies and laboratories, private several European nations also support micro-films, cooperative research organizations, and electronics research and development.

2

Within the Federal Government, the De-partment of Defense (DOD) sponsors thelargest share of microelectronics R&D. Fed-eral funds from various sources support workin Federal laboratories, industry, universities,and cooperative organizations.

Private sector R&D in the United States en-compasses a spectrum of activities in both ver-tically integrated companies and merchantfirms.2 Vertically integrated companies gen-erally support a full range of activities frombasic research to applied development, al-though the types and levels of R&D varywidely from company to company. Merchantfirms tend to limit their R&D to the last stagesof development. Some companies in each cat-egory support and participate in R&D in co-operative organizations and in universities inaddition to their onsite efforts.

OTA identified four basic changes occurringin institutional support for microelectronicsR&D in the United States:

1.

2.

In the last few years, many factors haveconverged to alter the structure of themicroelectronics industry. Chief amongthese is the Japanese challenge to U.S.competitiveness, which has led to in-creased shares of U.S. and internationalmarkets for Japanese companies and, con-sequently, reduced profits for U.S. com-panies. This may be leading to decreasedR&D efforts by the industry.The exceptional capability of Japanesecompanies to transform research conceptsinto products is well established in micro-electronics. Now, there is also growingevidence that Japanese basic research ef-forts are outpacing U.S. efforts in someareas of microelectronics, e.g., optoelec-tronics.

*“Vertically integrated” in this context refers to a companythat makes microelectronics to use in the products that it sells,e.g., a computer company that makes integrated circuits forits computers. The microelectronics division of such a companyis often termed a captive manufacturer. Some vertically in-tegrated companies also sell their microelectronic products toothers in addition to using them internally. A merchant firm,in contrast, makes microelectronics primarily to sell to otherend users.

3. Because of international competition andlimited resources for equipment and per-sonnel, cooperative research efforts aregaining popularity as a means to bolsterR&D. These activities, which typically in-volve cooperation among different indus-trial, academic, and Federal organiza-tions, represent a relatively new approachto R&D in the United States.

4. The deregulation of the telecommunica-tions industry is affecting R&D in micro-electronics-related areas. AT&T Bell Lab-oratories’ efforts are becoming moreclosely tied to products than were re-search efforts at the predivestiture BellLaboratories. The role of Bell Communi-cations Research is still not completelyknown.

Findings About Technological Trends

Microelectronics can be separated into tworelated parts: 1) fabrication technology, includ-ing materials, devices, and circuits; and 2) chiparchitecture or design. Advances are takingplace in both aspects of the technology.

Trends in Fabrication Technology

Over the next two decades, the primarytechnological trend in the physical structureof microelectronics is likely to be continuedminiaturization of silicon integrated circuits.Because this scaling down of the dimensionsof ICs has been the key to better and lessexpensive chips, it has been the basis ofprogress in microelectronics over the last 25years. OTA found that, according to experts,this trend will continue for the next 5 to 10years and then will begin to level off in approx-imately 15 years, when minimum dimensionsare between one-tenth and one-fifth of amicron-about one-tenth of the minimum sizecurrently in production. (A micron is one-millionth of a meter. See figure A-3 in app. A.)

OTA identified several other technologicaltrends related to this central finding. Devel-opment activities expected to influence the in-dustry soon center on advanced manufactur-ing techniques required to fabricate circuits

with smaller and smaller dimensions. Theseadvances represent incremental changes incurrent silicon IC technology.

Mid- to long-term R&D efforts, which focuson significantly different technologies withpromise for the next generation of microelec-tronics, are centered on:

● digital and analog (microwave) integratedcircuits made from gallium arsenide(GaAs),

• optoelectronics, and● quantum-effect structures.

OTA found that few microelectronics expertsexpect GaAs integrated circuits to replace sili-con digital ICs. Rather, they believe that thetwo types of ICs will meet complementaryneeds.

The outlook for technologies based on ma-terials other than semiconductors appearslimited for the next few decades. R&D activi-ties in Josephson junction technology, once amajor contender for the next generation of ICsfor computers, are currently receiving limitedattention. Efforts in bimolecular electronicsare only exploratory today, and their promisespeculative. Most experts agree that they willnot come to fruition in the next few decades,if ever.

3

Trends in Design

While circuits continue to shrink and beginto reach limitations, the power of design tools,and hence the flexibility of IC design, will con-tinue to grow rapidly, OTA found. Users havebeen limited to building systems out of stand-ard ICs and other components in the past.Now, complex new design systems coupledwith advanced manufacturing capabilities al-low users to configure chips to perform spe-cialized tasks. Progress in this area will hingeon R&D activities in design software.

Merchant manufacturers are currently pur-suing the expanding markets for applica-tion-specific ICs (ASICs), which include cus-tom and semicustom chips. As the capabilitiesof design systems and the networks that linkthem to manufacturing facilities expand, anengineer will probably be able to design an ICfor a specific need from a workstation, trans-mit the design to a silicon foundry, and receivethe completed special chip within a short timeand for low cost. This level of flexibility couldopen the door to a whole new genre of elec-tronic capabilities.

POTENTIAL POLICY CONCERNS

The state of research and development inmicroelectronics raises several potential pol-icy concerns about the Federal role. One setof issues arises from the changes occurring inmicroelectronics R&D; a second set centers onongoing direct Federal support for microelec-tronics R&D.

Federal Response to Changes in the Industry

In the early days of the merchant IC indus-try, that part of the microelectronics commu-nity generally tried to minimize its interactionwith the Federal Government. In recent years,however, changes in the industry have shiftedthe relationship between the government andmerchant firms, making it necessary for eachto attend more closely to the other: This sig-

Federal Response to Changes in nifies wider recognition of the impact of Fed-Microelectronics R&D eral policy on the industry and a resulting gem

Governmental policies need to recognize the eral trend toward increased reliance on Federal

changes underway in microelectronics, both in policy by the industry.

the industry’s support of and need for R&D, Because Japanese competition is the great-and in the emerging technological trends. est challenge-that U.S. merchant microelec-

4

tronics companies face today, it is the focalpoint of interactions between the industry andthe Federal Government. The U.S. industryhas, in the last few months, asked the govern-ment for help in changing the trade imbalancein three separate cases.3 Another pressing con-cern is the high cost of capital in the UnitedStates relative to Japan.

The current problems that the U.S. compa-nies face pose a paradox. Although R&D, along-term investment, cannot solve industry’simmediate problems, it is a crucial ingredientin industrial competitiveness. Without con-tinued strength in R&D, solutions to the near-term problems will only delay the decline ofthe U.S. companies. Yet microelectronics firmsthat are struggling to survive are likely to ne-glect R&D activity in the face of more imme-diate and pressing problems. For example,they may find it difficult to justify fundingR&D while cutting jobs at an unprecedentedrate. This may lead to a deterioration of theindustrial R&D base.

Federal policies that affect the amount ofR&D available to private companies can becategorized as follows:

●

●

policies that generally strengthen thecompanies by making them more com-petitive, and thus assuring sufficient prof-its to support R&D (e. g., internationaltrade policies and mechanisms to lowercapital costs);policies to ease the financial burden orlower the risks of private R&D (e.g., thetax treatment of R&D4 and intellectualproperty protections); and

‘Semiconductor Industry Association, “Japanese Market Bar-riers in Microelectronics: Memorandum in Support of a Peti-tion Pursuant to Section 301 of the Trade Act of 1974 AsAmended, ” June 14, 1985; “Japanese Accused on Chips, ” NewYork Times, June 27, 1985, p. Dl; and “EpROM Makers FileClaim Against Japanese ‘Dumping’, ” Electronic News, Oct. 7,1985, p. 1.

‘Since 1981, firms have been allowed tax credits on increasedexpenses for research and development, enhancing the R&Dspending power of these companies. This R&D tax credit ex-pired at the end of 1985.

‘Pressure from the microelectronics industry spurred Con-gress to legislate a new form of intellectual property protec-tion in the Semiconductor Chip Protection Act. The act, signedinto law in November 1984, protects masks (used to fabricateICs) from unauthorized copying, and so offers manufacturersan additional incentive to come out with new products by pro-tecting returns on R&D investments.

● direct Federal funding for R&D, the re-sults of which are available to privatecompanies.

The first two types of Federal involvement areindirect ways to make R&D investment eas-ier for companies; the third approach fundsR&D directly.

The lion’s share of direct Federal support formicroelectronics R&D comes from the Depart-ment of Defense, and is therefore driven bymilitary requirements. In general, DOD-spon-sored basic research serves both military andcommercial goals. Development activities formilitary microelectronics, in contrast, do notoverlap completely with activities for commer-cial needs. For example, low-cost, high-volumeproduction capabilities are a high priority forthe commercial manufacture of integrated cir-cuits, but the major DOD program to advanceIC technology, the Very High Speed Inte-grated Circuit (VHSIC) program, focuses ondesign and fabrication of a small volume ofhighly specialized ICs for use in militarysystems.

The DOD style for funding microelectronicsR&D, characterized by long-term investmentin R&D with a clear connection to end uses,appears to have been highly successful forachieving military goals. Some members of themicroelectronics community would, therefore,like to see the Federal Government aim a sim-ilar level of support at commercial needs—apoint of view that has gained momentum inthe face of the current pressures on the indus-try. They cite the influential Japanese Min-istry of International Trade and Industry(MITI) as a useful model. But opinions in themicroelectronics community diverge sharplyon this topic. Opponents of further direct Fed-eral involvement in microelectronics R&D forthe industry argue that the results of R&D willmeet commercial needs and will be availableto industry only if the commercial sectordirects and carries out the work itself. Theyview the MITI model as unacceptable in theAmerican context, given the vast differencesin industry-government relationships betweenJapan and the United States.

One example of a plan to start to bridge thisdifference of opinion comes from the Semicon-

—

5

ductor Research Corp. (SRC), a cooperative This has already occurred in the case ofR&D organization directed by microelectron- GaAs digital integrated circuits research ledics industry leaders. SRC currently supports by the Defense Advanced Research ProjectsR&D at universities with funds from its mem- Agency (DARPA). The microelectronics com-ber companies. In the next few years, it may munity had conducted relatively little researchask the Federal Government to match this lev- on integrated circuits made from galliumel of support. The doubled budget would be arsenide before 1982 when DARPA decided,administered solely by SRC, which would soothe based on the results of 6 years of GaAs basicat least the concerns of SRC’s member com- research that they had sponsored, to fund apanics about the selection of research topics series of pilot production lines to demonstrateand the availability of results.6 the feasibility of using GaAs for ICs. This an-

nouncement kindled the interest of a varietyFederal Response to Technological Trends

.of organizations, and several defense elec-

Limitations to growth that stem from tech-nological trends are less immediate than theeconomic problems that the industry faces,but they too pose questions about the ap-propriate Federal role. The shrinking of cir-cuitry on silicon chips, on which progress hashinged thus far, has required enormous inno-vation, chiefly in manufacturing technologyand engineering exploitation of the conceptsfor transistors and circuit integration. Butprogress in fabrication technology beyond thelimits of silicon scaling will demand a widerrange of more basic R&D activities.

This technological factor may drive ex-panded Federal participation in R&D for po-tential alternative microelectronics technol-ogies. This could take many forms, includingpolicies to encourage basic research in indus-try or greater direct Federal R&D funding. Al-ternatively, Federal agencies may select spe-cific areas in which to focus support withoutsignificantly increasing their total fundinglevels. Because the Federal Government fundsa wide range of microelectronics research ef-forts and interacts with a variety of R&D orga-nizations, it can exert considerable leveragein key areas. In these areas, some Federalagencies might be able to lead the way for ef-forts by participants in the commercial sector,universities, and other Federal agencies by tar-geting R&D monies.

tronics firms mobilized to build a base of activ-ities in GaAs so that they could be involvedwith the pilot lines. Three new commercial ven-tures have already spun off from this work.Perhaps even more significantly, there is someevidence that DARPA’s interest in GaAs ICsmay have helped to convince research organi-zations such as AT&T Bell Laboratories thatthe field deserves an intensive research effort.At about the same time, IBM turned its fo-cus for alternative chip technology fromJosephson junctions to GaAs. In part becauseof DARPA’s initiative, the activities in GaAsICs grew in a few years from a handful of iso-lated efforts in individual laboratories to largeprograms sponsored by Federal agencies, in-dustry, and universities.

Direct Federal Support forMicroelectronics R&D: Policy Questions

Ongoing direct funding of microelectronicsR&D from Federal agencies continues to raisepolicy concerns. The system of multisourcesupport, with several different Federal agen-cies funding microelectronics R&D activities,poses potential policy questions. And since thelargest amount of support comes from DOD,many of the concerns center on the implica-tions of defense R&D spending.

Multiple Sources of Federal Funding:Pros and Cons

Microelectronics, like other science and engi-‘Interview with George M. Scalise, Advanced Micro Devices neering fields, receives R&D funding from sev-

and Semiconductor Research Corp., June 1985. The doubledbudget could total as much as $100 million per year, although eral different Federal agencies. Because a De-SRC’s current annual budget is only approximately$15 million. partment of Science or some other scheme for

6

centralized R&D funding is proposed and con-sidered from time to time, it is important toexamine the pros and cons of the currentmultisource arrangement. In microelectronics-related areas, the system generally avoids po-tential pitfalls and offers advantages over acentralized system.

The potential drawbacks of the multisourcesystem do not pose problems at present. Onepossible difficulty is wasteful duplication ofeffort and competition for resources if the vari-ous agencies do not communicate their plansto one another. The present system, too, couldconfuse or inconvenience researchers seekingfunding, particularly newcomers. However,the researchers within an area typically com-municate with each other and are familiar withthe full range of activities in their area. Theseinformal infrastructures prevent most un-necessary duplication of effort and alert re-searchers to the relevant funding sources inthe area. In addition, the agencies coordinateR&D funding through both formal and infor-mal channels.

The advantages of distributed fundingacross agencies more than compensate for thepotential problems. The arrangement providesresearchers with multiple channels for Federalsupport for promising new ideas; they can turnto a second agency if a proposal is refused bythe first. The present system also permits eachagency to fund R&D to meet its own goals orthose of its parent department. The existingsituation, with several loosely coordinated in-dependent agencies funding different aspectsof microelectronics R&D, appears to serve itspurpose well.

Questions Raised by Department of DefenseActivities

Beyond the questions of the balance be-tween R&D for military needs and R&D forcommercial needs, DOD activities raise twosets of potential Federal policy issues. Theseare:

1. DOD control of research and develop-ment, particularly university researchactivities; and

2. the impact of the Strategic Defense Ini-tiative (SDI) on the structure of DOD mi-croelectronics R&D.

Keeping information about new defensetechnologies within the United States is a crit-ical concern for military security; free and wideexchange of ideas and results is an equallycrucial ingredient in scientific research. Thisdilemma is the basis of an ongoing discussionamong players both within and outside DOD,all of whom are trying to determine the appro-priate type and level of control of defense re-search results. Controls on universities, wheremany foreign students are involved in scien-tific research on campus, are of particular con-cern. Some leading universities have bannedclassified research on campus altogether asa partial solution. Several years of debate re-cently resulted in a policy from the WhiteHouse (the National Security Decision Direc-tive) which makes classification the only mech-anism for control of fundamental research (i.e.,unclassified research may not be restricted).8

The policy does not solve the problems com-pletely. It will, however, greatly simplify theprocess of determining control by reducing thenumber of gray areas.

R&D funding under SDI raises two typesof concerns about DOD research: the impacton the structure of DOD funding in micro-electronics R&D, and further questions aboutthe treatment of university research.

The initiative’s activities in this area mayinvolve a major restructuring of the funds formicroelectronics R&D from the various DODagencies, whether or not it increases the over-all level of DOD support. The transfer of theGaAs IC pilot lines from DARPA to the SD IOrganization (SD IO) is early evidence to sup-port this possibility. Given the wide percep-tion that the current arrangement for DOD-sponsored research (with several differentagencies operating independently but commu-

71nstitute of Electrical and Electronics Engineers, “DOD’sPerle Questions Value of Open Research on Campus,” The In-stitute, July 1985, p. 10.

8“White House Issues Secrecy Guideline, ” Science, vol. 230,Oct. 11, 1985, p. 152.

7-.

nicating with each other) works well, central-ized funding of microelectronics R&D throughSDIO could decrease DOD’s effectiveness inthe field.

The fact that SD I-funded activities are des-ignated as “advanced technology develop-ment” rather than “research” has exacerbatedthe concerns about DOD controls on univer-sity research. There has been concern thatDOD would censor dissemination of all SDIwork, including university activities. From theperspective of a group of university scientistsboycotting SDI, “the likelihood that SDIfunding will restrict academic freedom . . . isgreater than for other sources of funding. ” (A

‘From the ho~cott petition, as quoted in “Star Wars 130y -cott Gains Strength, Science, vol. 230, Oct. 11, 1985, p. 152.

recent policy stating that SD I research atuniversities will be considered ‘‘fundamentalresearch, ” which is to be treated in accordancewith the new National Security DecisionDirective, may have alleviated some of theseconcerns. ‘0) On the other hand, the fact thatSDI’s Innovative Science and Technology Of-fice received approximately 2,700 preliminaryproposals from university researchers over aperiod of just 3 months]’ is strong evidenceof interest from that community.

10] nstitute of E ]ectric~ and Electronics 1? ngineers, ‘‘ SI~ IMemo Bars Controls on Most ‘Star Wars’ Research in Univer-sities, ” The Institute, October 1985, p. 1,

‘lDwight Duston, Innovative Sciehce and Technology office,SDI organization, personal communication, Janu~ 1986. Thisnumber includes preliminary’ proposals in areas other thanmicroelectronics.

Chapter 2

Introduction: MicroelectronicsTechnology and R&D

THE IMPACT OF MICROELECTRONICS AND EVOLUTIONOF THE INDUSTRY

Microelectronics technology has dramati-cally improved the capabilities of computersand communications systems, while also fuel-ing the growth of completely new applications,such as personal computers. Growing ex-tremely fast and providing increasingly power-ful and inexpensive tools to manipulate elec-tronic signals, microelectronics has become thecornerstone of information technologies. It iscentral to such areas as:

●

●

●

●

●

computers, from powerful supercom-puters, through business computers, toinexpensive personal computers;communications systems, includingswitching stations and satellite communi-cations;consumer products, such as electronicwatches, video games, and pocket calcu-lators;control systems for industrial applica-tions, automobiles, and home appliances;andmilitary systems for national defense.

Microelectronics has become a vital part ofU.S. commerce and defense. In both sectors,maintaining a technological edge over the restof the world is the only way to ensure secu-rity—whether military or economic. U.S. com-panies claimed approximately $14 billion ofthe $26-billion world market for microelec-tronics in 1984. I The development of increas-ingly sophisticated weapons systems meansthat virtually every aspect of current militarytechnology depends on microelectronics.

1“The ,Japan(~sc, Semiconductor hlarket and S1 ~1 :10 1 I’uti-tion, ” presentat ion 1)~’ (;eorge Scalist’ for th~’ Sen]i~f)nduct{)r1 ndustr} Ass{)t’lation, J ul~ !4, 19X5.

The dramatic growth of the technology hasprompted observers to describe it as the mi-croelectronics revolution. To date, the minia-turization of circuitry, which leads to productsthat perform faster and better, has been chief-ly responsible for this revolution. Shrinkingthe electronic devices has yielded lower cost,expanded performance, and higher reliability.By any measure—cost for a given function,complexity of circuits, performance—inte-grated circuit (IC) technology has progressedat a tremendous rate since its inception. In1964, Gordon E. Moore predicted that thenumber of components on a chip would con-tinue to double annually, as it had since thebeginning of that decade. Twenty years later,technology has not departed significantlyfrom Moore’s law. Experts predict a slowingof the trend over the next 10 to 20 years.

At the same time, IC design will begin toalter the way that system engineers use ICs.New capabilities provided by design software,silicon foundries, and the networks that linkthem allow the end user much greater flexi-bility in designing chips for a specific appli-cation. This trend, coupled with changes infabrication technology, will affect the industry.

Many other factors, more immediate thanthe scientific and engineering changes, affectthe structure of the American microelectronicsindustry, especially merchant firms. These in-clude international competition, capital re-quirements, capital cost, and shifting mar-kets. U.S. microelectronics companies current-ly face debilitating competition from Japanesesources, and other Asian countries (e.g., Ko-rea) are also preparing to enter the market. Atthe same time, the industry is growing stead-

9

10

ily more capital-intensive, since every techno- Together, these technological, international,logical advance demands more complex equip- and economic trends and forces are alteringment and production facilities. The high cost the microelectronics industry. The changesof capital exacerbates this problem. Finally, mark the beginning of the maturing of the in-although many of the markets for uses of dustry and may signal the beginning of themicroelectronic products are growing, some of end of an era of apparently limitless growth.them grow explosively and then plummet asrapidly, as did the video games market. Theseshifts cause instability in the industry.

THE NATURE OF RESEARCH AND DEVELOPMENTIN MICROELECTRONICS

Since microelectronics has expanded swiftlyand will continue to do so, research and devel-opment (R&D) activities play an extremelycrucial role in its progress. As the technologyand the industry mature, however, the reasonsfor supporting R&D may shift. For example,the approaching “post-shrink” era (beyond thelimits of miniaturization for silicon integratedcircuits) may demand renewed vigor in basicresearch to find a successor to silicon ICs.

To suit the needs of those who use the prod-ucts based on microelectronic devices, R&Defforts in microelectronics are aimed at mak-ing circuits that:

● cost less,• operate at higher speeds (higher fre-

quencies),• require less power and generate less heat,● are more reliable and last longer, and● carry out specific functions.

These goals are prioritized in different waysdepending on the end use. For example, lowcost is typically the chief concern for chips tobe sold in high volume, but high speed maybe the dominant requirement in making com-ponents for supercomputers, and concernsabout power consumption may dominate forICs intended for use in satellites.

For the microelectronics manufacturer, inthe short term, these requirements translateinto:2

●

●

●

●

●

●

making ICs with smaller component de-vices (i.e., smaller transistors, resistors,capacitors, interconnections) and packingthese devices closer together on the chip;using larger wafers so that more chips canbe made on a single wafer;using new processes and new equipmentin chip fabrication;packaging chips in increasingly sophisti-cated ways;designing increasingly complex circuits;anddesigning circuits that are tailored to spe-cific applications.

Scientists and engineers involved in longerterm R&D are trying to anticipate technolog-ical needs beyond the current generation ofproducts. Their concerns center on topics such

new materials for microelectronics (e.g.,gallium arsenide);new devices to replace or augment tran-sistors (e.g., quantum-effect devices);new equipment to fabricate these mate-rials and devices;integrating optical and electronic cir-cuitry; andadvanced design tools.

2App. A: Current Microelectronics Technology provides somegeneral background information on integrated circuit technol-ogy, and App. B: Glossary of Terms gives definitions for tech-nical terms.

Chapter 3

Current Research and Development Activity

Microelectronics research and development(R&D) activities can be separated into threecategories:

1.

2.

3.

activities to improve silicon integratedcircuits (ICs),efforts for compound semiconductor mi-croelectronics (primarily based on galliumarsenide (GaAs)), andinvestigations for integrated circuitsbased on materials other than semicon-ductors.

Design activities span all three categories.Most of the work described here is aimed atmaking better digital integrated circuits.Other semiconductor activities, such as opto-electronics and microwave devices, are also in-cluded here because they are merging to somedegree with IC technology and because allsemiconductor R&D shares a common base ofphysical understanding and process tech-nology.

ADVANCED SILICON INTEGRATED CIRCUITSEfforts now underway to improve silicon-

based microelectronics will be the first typeof R&D to have practical applications. Theseefforts can be grouped in three categories ofsimultaneous activities:

1. the improvement of the physical circuitsand packaging of integrated circuits,

2. the facilitation of the design and fabrica-tion processes, and

3. the design of new types of ICs for specificmarkets.

Circuits and Packaging

The process of reducing the size of devicesin ICs and increasing their packing densitieshas several parts. Scientists and engineers aredeveloping devices–transistors, resistors, ca-pacitors–that have feature sizes of less than1 micron. Despite their small size, these de-vices must be designed to operate correctlyand to control the required amount of power.The interconnections required to hook the de-vices together are also becoming increasinglyharder to make. Each connection must shrinkin width but still conduct electrical currentwith virtually no resistance. The interconnec-tions must lie closer together but still be com-pletely isolated from each other. Designers

must lay out both the devices and the inter-connections in more and more complex pat-terns. Finally, the package for the completedchip must allow signals to enter and leave thechip at high speed, so that the package itselfdoes not obliterate the speed advantage of thenew circuitry.

These steps have been used to scale downsilicon ICs over the past 25 years. Every newreduction in feature size has been significantly

more difficult to achieve than the last, andprogress has been possible only through theintroduction of increasingly complex manufac-turing technologies and device and circuit de-signs. Today’s R&D workers face the great-est challenges yet.

Few trends, however, can continue forever.The remarkable feature of Moore’s law (the an-nual doubling of the number of componentson a chip) is the range over which it extendsbefore meeting unavoidable limits. Two tech-nological factors limit growth. The sizes of theindividual devices and the separations be-tween them eventually become so small thatthe devices cannot function as desired. Insome instances, these dimensions are a fewdozen atom layers. The problems involved ininterconnecting the devices on a chip alsobecome virtually insurmountable. Together,

11

12

these make a relaxation in the rapid growthof microelectronics technology inevitable.Microelectronics experts do not agree on thedetails and consequences of this slowdown,but they generally do agree that these limitswill be reached during the next 10 to 20 years.

Design and Fabrication Processes

Activities to facilitate the IC design proc-ess are focused on design tools, which simplifycircuit layout for IC designers. Particularly asthe design process has grown more complexto accommodate the millions of devices, com-puter-aided design (CAD) systems have be-come virtually indispensable. Currently, thereis no single standardized CAD system; rather,there are several different systems built bydifferent groups of designers. As these sys-tems evolve, they will simplify the design proc-ess and thereby give a wider range of usersgreat flexibility in creating new chips.

Fabrication technology includes the proc-esses for depositing very thin layers of differ-ent metals, insulators, and semiconductor ma-terials on the silicon substrate; changing theimpurity content in the semiconductor; etch-ing the layers; and defining small features inthe layers through lithography. Current R&Dactivities are exploring better techniques tocarry out each of these tasks. For example,densely packed circuits may require new ma-terials with special properties—high electricalconductivity, chemical stability, particularcrystal structure —for interconnections. Also,x-ray, electron-beam, or other lithographictechniques may be needed to replace current

photolithography for better definition ofultrasmall features.

Manufacturing technology is a crucial partof these advances because progress in siliconscaling is based on the introduction and im-provement of highly sophisticated equipment.Some examples are chemical-vapor-deposition(CVD) and evaporation systems to grow thinfilms of semiconductor crystals and metals; li-thography equipment and plasma etchers todefine the ultrasmall features of the IC; andfurnaces and ion implanters to introduce theproper impurities to the wafer.

Circuits for Specific Markets

Currently, most ICs fall into a few stand-ard categories: logic chips, memory chips, andmicroprocessors. However, as the design andmanufacturing capabilities of the IC industrygrow and become more flexible, a range of spe-cialized integrated circuits will play a morecentral role. Application-specific ICs (ASICs)are projected to grow from their current 12 to15 percent of the total IC market to 25 to 30percent of the 1990 market. ’ This category ofintegrated circuits includes custom chips,which are designed from scratch for the par-ticular application, and chips that can beadapted by the user for the specific need. Fur-ther enhancements of the design process willexpand users’ ability to design their own ICs.

‘“A Chip Business That Is Still Growing: Innovation SpursMarket for Application-Specific Integrated Circuits, ” Elec-tronics, July 22, 1985, p. 40.

MICROELECTRONICS BASED ON GALLIUM ARSENIDE ANDOTHER COMPOUND SEMICONDUCTORS

From the vantage point of the chemist or umns (II-VI compounds such as cadmium tel-physicist, there is a logical progression of semi- luride). (See figure 1. In this terminology, theconductor materials in the periodic table from Roman numerals refer to the columns on thesilicon (a column IV material), to compound right side of the periodic table; e.g., “III-V”semiconductors made from the columns refers to columns 111A and IVA. ) This pro-adjacent to silicon (II I-V compounds such as gression is also useful for classifying the rangeGaAs), to compounds made from the next col- of R&D activities in semiconductor microelec-

13

Figure 1 .—Periodic Table of the Elements

N O T E Elements n boldface are those commonly used 10 make semiconductor m a t e r i a l s

SOURCE Off Ice of Technology Assessment

tronics, since the most immediate develop-ment efforts focus on silicon, and longer termwork usually focuses on compound materials.

GaAs and other 111-Vs are now the basis fora variety of discrete microelectronic devices,as described in appendix A. Current R&D ef-forts involving these materials focus on.: 1)GaAs digital integrated circuits, 2) advancedoptoelectronic devices, and 3) monolithic mi-crowave devices. R&D on design tools, whichis critical in silicon technology, is also very im-portant for these alternative technologies.

GaAs Digital Integrated Circuits

Digital integrated circuits based on mate-rials other than silicon continue to receiveattention in the research community. GaAs-based integrated circuits are currently theleading contender for next-generation technol-ogy. Even so, it is important to note that vir-tually no experts believe that GaAs ICs willusurp the position of silicon for most appli-cations.

Gallium arsenide has several intrinsic phys-ical properties that distinguish it from silicon.

The typical devices made from GaAs operatefaster and consume less power than silicon de-vices. They are also less likely to malfunctionin the presence of radiation. However, sincea compound semiconductor is intrinsicallymore complicated than a single-element semi-conductor (silicon), GaAs is a much more dif-ficult material to grow, to handle, and to useto fabricate reliable devices.

At present, many barriers impede prospectsfor making GaAs ICs at production capacitycompared with silicon ICs, which are readilyproduced in quantity. Silicon ICs are currentlyfabricated on wafers with 5- or 6-inch diam-eters, while 3-inch wafers are the largest cur-rent size for GaAs. Standard silicon wafershave far fewer defects and are less brittle thanthe best GaAs wafers. Some processing stepsfor silicon IC technology can be adapteddirectly for GaAs ICs–portions of the litho-graphic procedure, wafer handling, clean-roomrequirements. But the steps involving othermaterials, such as oxides and other insulators,metals, and polycrystalline semiconductormaterial, must be developed specifically forGaAs. This requires a more complete under-standing of the chemistry and physics of the

interfaces between these materials. In additionto all of these fundamental difficulties, GaAsIC technology suffers because experience withit is very limited, relative to silicon. Many ofthe problems, however, will probably be solvedas experience accrues.

The list of organizations supporting (and notsupporting) R&D in this area reveals quiteclearly the microelectronics community’sviews on the applicability of GaAs ICs. DODis the leading Federal supporter of research inthis field, because military applications, par-ticularly in space, require the properties GaAsoffers: high speed for large-scale signal proc-essing, low power to minimize bulk and energyconsumption, and radiation hardness for relia-bility in the presence of radiation. Major com-puter and communications companies-e. g.,AT&T, IBM, and several Japanese companies—are also investigating GaAs ICs, primarilyfor use in the parts of their systems that re-quire the highest speed, e.g., computer frontends. Supercomputer companies, most nota-bly Cray, are attempting to make super-computers based on GaAs ICS. However, thestandard merchant chip makers (e.g., Intel,Fairchild Semiconductor, Advanced Micro Devices), which tend to concentrate almost ex-clusively on short-term development activi-ties, have no onsite efforts in materials otherthan silicon. Some of these companies supportlonger term projects, including GaAs work, atuniversities and through cooperative researchorganizations. This balance of support indi-cates two things:

1. GaAs digital ICs are beginning to findniche applications in a variety of areas,and

2. they will probably not make a significantdent in the standard IC components mar-ket for several years.

Optoelectronics

As described in appendix A, compoundsemiconductors and their alloys are currentlyused to make devices that convert electricalsignals to light signals and vice versa. The de-vices are used for optical communications and

sensor applications. R&D activities in opto-electronics fall into three categories:

1. advanced discrete light sources and de-tectors,

2. integrated optoelectronics, and3. superlattices and other quantum-effect

structures.

The II I-V compound materials used for op-toelectronics include GaAs, iridium phosphide(InP), gallium phosphide (GaP), aluminum ar-senide (AlAs), iridium antimonide (InSb), andalloys of these materials, such as aluminumgallium arsenide (AIGaAs), iridium gallium ar-senide (InGaAs) and iridium gallium arsenidephosphide (InGaAsP). Similarly, important II-VI materials include cadmium telluride (CdTe),mercury telluride (HgTe), and their alloy, mer-cury cadmium telluride (HgCdTe). These ma-terials are designated as binary, ternary, orquaternary depending on the number of ele-ments found in them. The alloys actually rep-resent a range of materials; for example, halfthe atoms in HgCdTe must be tellurium, butthe other half may be any combination of mer-cury and cadmium atoms. The properties ofthe alloy generally lie between the propertiesof the binary materials that compose it. Theparticular composition of an alloy is typicallychosen to have a certain desired wavelengthresponse. The standard approach to fabricat-ing optoelectronic devices is to grow thinlayers of ternary or quaternary alloys on a sub-strate of a binary material.

Discrete Optoelectronic Devices

The first optoelectronic devices for fiberoptic communications were made of GaAsand AlGaAs. The best current devices, how-ever, are based on different compositions ofInGaAsP grown on substrates of InP, struc-tures that generate and detect light over arange of wavelengths that includes those oflowest loss (1.55 microns) and lowest disper-sion (1.3 microns) in optical fibers. Since thesedevices are relatively new, the materials andprocessing problems have not been completelyresolved. R&D efforts in this area focus onmaking devices more reliable and more toler-

15

ant of extreme environments, achieving moreprecise control of the generated light, and de-veloping production processes (with highthroughput and yield) for the devices.

The II-VI compounds are used to fabricatelong-wavelength infrared sensors becausethese materials (especially HgCdTe and CdTe)are sensitive to the wavelengths of inter-est—from around 1 micron to the 10- to 12-micron range. Currently, research on these de-vices centers on materials properties, whichare much more difficult to control in the II-VI compounds than in other semiconductors.

Integrated Optoelectronics

Individual optoelectronic devices can be in-tegrated and fabricated on a single substratemuch as standard electronic devices are in-tegrated on a chip to make an IC. An inte-grated optoelectronic device, typically built ona substrate of InP or GaAs, may be composedof lasers, devices to amplify and modulate thelight signals, and light detectors. In addition,the same chip may have purely electronic de-vices that process electrical signals. Such anintegrated chip has all the advantages of a con-ventional integrated circuit—miniaturization,high speed, low power, fabrication reliabil-ity—and also solves the alignment and vibra-tional-stability requirements for the opticalelements it comprises. In addition, it bringsoptical and electronic devices so close togetherthat signal delays between them are minimized,allowing high speeds to be achieved.

The basic concepts for integrated optoelec-tronics may be extrapolated even further inthe future. Highly advanced crystal growthand processing techniques, currently in theirinfancy, could also open the door to the pos-sibility of structures that would combinesilicon, III-V, and II-VI devices on a singlesubstrate. Such a scheme would allow the flex-ibility to use the optimal material for each por-tion of the complete circuit. For example, a cir-cuit on a single chip could be composed ofInGaAsP lasers, HgCdTe light detectors, andGaAs and silicon digital logic and memory cir-cuits. Concepts of this sort are still in the spec-ulative stage today.

Superlattices and OtherQuantum-Effect Structures

By using sophisticated techniques to de-posit materials very precisely, crystal growerscan make layers as thin as a few layers ofatoms on a semiconductor substrate. Such alayer is approximately one-thousandth of a mi-cron thick-one-billionth of a meter. z Struc-tures called superlattices are formed by grow-ing alternating ultrathin layers of twodifferent materials, e.g., GaAs and AlGaAs.Special methods such as molecular-beamepitaxy (MBE) and metal-organic chemical va-por deposition (MOCVD) are necessary toachieve this extreme level of control duringcrystal growth. These advanced techniquesopen a completely new set of options forsemiconductor materials. Varying the thick-ness of the layers and their composition canyield a superlattice material with differentelectronic and optical properties. Quantummechanical effects dominate the behavior ofthe electrons in superlattices because thestructures have such small dimensions. Thus,the electron transport processes can be dra-matically different from the processes in nor-mal material.

Currently, the range of research efforts inthis area is very wide, spanning the spectrumfrom work in designing better systems forgrowing these precise layers to making devicesbased on superlattices. The devices includeIII-V and II-VI photodetectors, lasers, andtransistors. The greatest overall contributionof this new breed of materials will probablystem from the fact that they can be tailor-made for a particular application. Already,they arevantageareas of

heralded by observers from diversepoints as one ofresearch today.3

the most exciting

‘F’igure A-3 in app. A shows how small a micron is.‘For example, National .4cademy of Sciences, Committee on

Science, Engineering, and Public Policy, The Outlook for Scienceand Technolo~’ 198,5; and CJeorge H. Heilmeier, “Microelec-tronics: End of the Beginning or Beginning of the End?” In-ternational Electron L)e\’ices Meeting, December 198L1.

16

Microwave Devices

Analog microwave devices, operating at fre-quencies from approximately 1 to 60 gigahertz(a gigahertz is 1 billion cycles per second), arecommonly made from GaAs to take advantageof the high speed that the material offers. Cur-rently, monolithic microwave integrated cir-cuits (MMICs) are being developed. These cir-cuits combine various microwave devices ona single substrate, typically GaAs. The devicesare not packed as densely as silicon deviceson a conventional digital IC, but the smallestdimensions are about the same as or smallerthan those for silicon ICs—below 1 micron.MMICs will fill the demand for more compact

and reliable microwave circuitry for applica-tions in radar, transmission of television andtelephone signals, and spectroscopy. A coupleof companies have MMICs on the market al-ready, and others plan to market them soon.4

Most recently, the Department of Defense(DOD) announced that it will launch a majornew initiative for MMICs for defense systems.The new program will be analogous to theVery High Speed Integrated Circuit (VHSIC)program, which addresses digital IC technol-ogy for DOD.

4“MMICS Save Space, Increase Reliability, and Improve Per-formance, ” Electronics Week, May 20, 1985, p. 52.

NONSEMICONDUCTOR INTEGRATED CIRCUITTECHNOLOGIES

Integrated circuits may also be based on ma-terials other than semiconductors. Circuitsmade of superconducting Josephson junctionshave been heavily investigated but are not atpresent expected to find any large-scale appli-cations in digital microelectronics. Bimolecul-ar electronics is currently only a highly spec-ulative field.

Josephson Junctions

A Josephson junction is an electronic devicemade by sandwiching a very thin insulator be-tween two superconductors—materials withzero electrical resistance at very low temper-atures. Like electronic devices made fromsemiconductors, Josephson junctions canswitch or store electronic signals. Despitedrawbacks such as extremely low operatingtemperatures, they offer several advantagesover semiconductor microelectronics. Joseph-son junctions can switch signals at unparal-leled speeds, require very little power, and canbe scaled down to extremely small dimensions.

Computer systems based on Josephson-junction technology were intensively re-searched and developed for over two decades,

most notably at IBM. However, IBM haltedall but its basic research effort in this area sev-eral years ago because packaging, manu-facturing, and reliability difficulties meantthat the systems could not operate as well ashad been originally predicted. Furthermore,while the superconducting technology wasstruggling to get on its feet, silicon and gal-lium arsenide technologies kept progressingat a tremendous rate. IBM’s cutbacks in thisarea symbolized a significant change in themicroelectronics community’s view of IC tech-nology beyond silicon. Currently, althoughsome work on Josephson junctions continuesin the United States and in Japan (especiallyefforts aimed at making high-speed analog cir-cuits), these devices are considered unlikelycandidates to pickup where silicon leaves offin digital ICs.

Bimolecular Electronics

In contrast to semiconductor and Joseph-son junction technologies, the concept of usingbiological systems to process electrical signalsis still in its infancy. Some researchers areturning their attention to biological systems

17

as the extrapolation of the trend to smaller in the fabrication of extremely small struc-and smaller devices leads to molecular-scale tures on semiconductors, and in electronic sen-structures. sors for medical applications. But the poten-

At present, large-scale bimolecular elec-tial for computers based on bimolecular

tronics is a largely speculative area. Some ex-circuitry has not yet been demonstrated.

perts envision-a role for biological materials

.

Chapter 4

Institutional Support forMicroelectronics R&D

Because microelectronics is a key commer- private, Federal, academic, and cooperativecial and military technology, support for re- groups are active in the area. Other nationssearch and development comes from many also support microelectronics R&D; Japan insources. In the United States, a multitude of particular is a growing presence in the field.

FEDERAL SUPPORTThe largest block of Federal funding for

microelectronics R&D comes from the Depart-ment of Defense (DOD). The National ScienceFoundation (NSF) funds basic research inareas related to microelectronics. Several otherFederal agencies and laboratories have pro-grams to meet their specific needs, amongthem the National Aeronautics and SpaceAdministration, the National Institutes ofHealth, the National Bureau of Standards,Lawrence Livermore National Laboratory,and Los Alamos National Laboratory.

Department of Defense

DOD support was crucial to early successesin microelectronics. For example, military re-quirements for smaller circuits were the pri-mary force behind the development of the firstintegrated circuits (ICs). While DOD has con-tinued to fund microelectronics R&D exten-sively and remains the largest source of Fed-eral support for microelectronics R&D,support from private companies has assumeda much larger role, and includes an entire in-dustry that has sprung up in the last few dec-ades. Today, military applications constituteapproximately 10 percent of the total sales ofmicroelectronic products. Microelectronicstechnology is crucial to DOD, and its impor-tance will grow further in the foreseeable fu-ture. According to two experts,

. . . [i]t is safe to say that there is not a sin-gle western military system that is not criti-

cally dependent for its operation on semicon-ductor integrated circuits. ’

Other microelectronic devices, such as sensors,are also becoming increasingly important inmilitary technology.

The defense community, like other usersof microelectronics, requires high-speed in-tegrated circuits that consume little power.Since U.S. defense policy has long been basedon technological superiority, DOD needs thebest possible signal processing and sensor ca-pabilities. Military end users have specificneeds as well, including products like ICs thatare immune to damage from radiation.

Each of the three services—the Navy, AirForce, and Army—supports activities inmicroelectronics R&D in DOD laboratories,such as the Naval Research Laboratory, theAir Force Avionics Laboratory at Wright-Patterson, and the Army Electronic Deviceand Technology Laboratory at Fort Mon-mouth, and through research agencies. Apartfrom the services, DOD’s Defense AdvancedResearch Projects Agency (DARPA) is heav-ily involved in this area. Certain aspects ofmicroelectronics R&D are also covered undera few DOD special programs, such as the VeryHigh Speed Integrated Circuits (VHSIC) pro-

IAlton L. Gilbert and Bruce D. McCombe, “Joint ServicesElectronics Program: An Historical Perspective, prepared forthe U.S. Army Research Office, Electronics Division, April1985.

19

20—

gram and the Strategic Defense Initiative(SDI).

Research for the Navy, Air Force, and Army

The Office of Naval Research (ONR), the AirForce Office of Scientific Research (AFOSR),and the Army Research Office (ARO) handlescientific research for their respectivebranches of the military. All three fund sub-stantial amounts of basic research in micro-electronics, with large shares of these fundsgoing to universities. Their areas of interestcenter on materials and devices. Because thethree agencies are organized in different ways,a completely accurate comparison of their lev-els of support is not possible. However, ap-proximate annual budget figures for microelec-tronics-related work sponsored by AFOSR,ONR, and ARO are $24 million, $13 million,and $9 million, respectively.2

Part of the funding from the three services’research offices goes to support the Joint Serv-ices Electronics Program (J SE P). They spon-sor the program jointly, with each office con-tributing approximately one-third of thefunding. JSEP is a 38-year-old DOD programthat funds electronics research, includingmicroelectronics, at a group of universities. Itis designed to provide its 12 member universi-ties with stable, long-term funding for basicresearch. Total JSEP annual funding has in-creased since the program’s inception to $9.6million in 1984; approximately 75 percent ofthis is spent on research on integrated circuitsand other microelectronics-related areas.3

Defense Advanced Research Projects Agency

The Defense Advanced Research ProjectsAgency (DARPA), which is separate from theservices, supports long-term research for gen-eral military applications, including microelec-tronics R&D. DARPA funds efforts at univer-sities, industrial laboratories, and not-for-profitorganizations. One component of DARPA, the

‘~Horst R. Wittmann, Gerald I.. Witt, and Kevin J. Malloy,AFOSR; Kenneth L. Davis and Larry R. Cooper, ONR; and Jim-mie R. Suttle and Michael A. Stroscio, ARO; interviews anddiscussion, April 1985 to January 1986.

‘Gilbert and McCombe, op. cit.

Defense Sciences Office (DSO) supports mid-to long-term work on materials, processes, de-vices, and circuits. Another branch, the Infor-mation Processing Techniques Office (IPTO)sponsors activities in very large-scale integra-tion (VLSI) design and architecture and someproduction automation work.

In fiscal year 1985, DSO had a budget of ap-proximately $34 million for microelectronics,of which $28 million supported basic researchand the remainder funded exploratory devel-opment. About 40 percent of the total budgetwent to universities for basic research. DSOsponsors investigations of gallium arsenide(GaAs) and other II I-V compounds and theiralloys, and work on II-VI compounds, such asmercury telluride and cadmium telluride, andtheir alloys. Many of these materials have in-teresting combinations of properties that canbe exploited for new applications. DARPAalso supports research on the integration ofbiological materials and semiconductor elec-tronics for ion sensors and other devices.

As of fiscal year 1985, DSO assumed respon-sibility for the administration of the GaAs ICpilot lines and related activities for SDI. Theseefforts were transferred to SD I from DARPAat the end of fiscal year 1984. SDI will spend$23 million in this area in 1985, and fundingmay rise to $40 to $60 million over the nextfew years, depending on the overall level ofsupport for SDI. Most of these funds are usedto support the GaAs IC pilot lines that wereoriginally established and funded by DARPA;additionally, $2 to $3 million from this sourcegoes to universities for basic researchactivities.4

The R&D activities supported by IPTOcover the circuit-design and manufacturingportions of microelectronics technology. Ap-proximately $12 million of the funds thatIPTO puts into microelectronics goes to basicresearch activities in VLSI design and ar-chitecture, most conducted at universities. Inmany cases, this research has later been fun-

4Richard A. Reynolds and Sven A. Roosild, DARPA, inter-views and discussion, October 1984 to January 1986.

neled into commercial enterprises. For exam-ple, IPTO originally sponsored the “CosmicCube” parallel architecture work at the Cali-fornia Institute of Technology; Intel addedfunding to the Federal money and is nowbuilding and marketing a minisupercomputerbased on this architecture. In addition, IPTOspends about $6 million per year on explora-tory development in automation and fast-turn-around efforts for ICs. These funds cover workon the Metal-Oxide-Semiconductor Implemen-tation System (MOSIS), which gives a largeand geographically diffuse community of ICdesigners, particularly at universities, accessto a silicon fabrication facility. With MOSIS,individuals design circuits at their home fa-cilities using a computer-aided design (CAD)system. The design commands are transmittedover a communication network to a manufac-turing site, where the IC is fabricated. At pres-ent, designers use MOSIS to create new sili-con ICs. Activities in progress will also makepossible the creation of GaAs-based ICs in asimilar systems

Since DARPA is charged with longer termR&D responsibilities, the agency is shiftingits focus in microelectronics away from siliconefforts and toward GaAs ICs, while also be-ginning some activities in more esoteric fields.In some cases, work that DARPA initiatedand supported is being picked up by groupsmore interested in near-term efforts. For ex-ample, DOD’s VHSIC program and the pri-vate Semiconductor Research Corp. (SRC) aretaking over the funding of some silicon VLSIprograms at Stanford University that were ini-tially sponsored by DARPA.6

DOD Special Programs

In addition to the four established agencies,DOD has a variety of special programs thatsupport microelectronics R&D. Examples in-clude the Very High Speed Integrated Circuitsprogram and the new Strategic Defense Ini-tiative. VHSIC is wholly dedicated to silicon

‘>Paul I,osleben, DAR PA, interview, August 1985; and pres-entation on ‘ L Silicon as a Medium for New Ideas, I E ~; E Sys-tems, Man, and Cybernetics Society, Sept. 16, 1985.

‘Paul I,osleben, DARPA, interview, August 1985.

21

IC technology; SDI has microelectronics R&Dcomponents as part of a more general mission.DOD also recently announced the start of anew VHSIC-like program to advance the tech-nology of monolithic microwave ICs (MMICs)for defense applications.

The 10-year VHSIC program was estab-lished in 1979 to address specific militaryneeds in microelectronics. The three servicesparticipate in VHSIC, but the Office of the Un-dersecretary for Defense Research and En-gineering provides overall administration.Honeywell, TRW, IBM, Hughes, Texas In-struments, and Westinghouse are the primecontractors for VHSIC, which is scheduled tofinish in 1989.

VHSIC has several goals. The primary tech-nical objective is to establish processes to de-sign and fabricate chips with characteristicsnecessary for defense needs. The program alsointends to ease the adoption of these ICs inmilitary systems. In addition, VHSIC ad-ministrators view their program as a mech-anism to encourage the commercial micro-electronics sector to develop productioncapabilities suited for the military market. Theprogram is intended to function as a bridgebetween designers of military systems and theintegrated circuit community.

The technological goals of VHSIC aredivided into phases. The first phase of the pro-gram, now nearing completion, developed pi-lot lines for the production of ICs with 1.25micron minimum feature sizes and providednew chips for use in military systems. The sec-ond phase will attempt to establish pilot linesto fabricate ICs with 0.5 micron features. Atthe same time, commercial R&D activitieshave been and will be developing ICs with sim-ilar feature sizes, but the VHSIC chips are de-signed for specific DOD applications. Al-though the first phase of activities took longerto complete than originally anticipated, someof the projects have now been carried out suc-cessfully.

Because VHSIC’s other objectives arelonger term and less concrete, their success isharder to assess. Several signs point to

22-.

progress; VHSIC chips are used in several mil-itary systems in each of the services. Andwhile VHSIC has gotten a mixed reactionfrom industry over the years, the program hasdrawn the attention of at least some of thecommercial IC vendors to military appli-cations.

As conceived, VHSIC had a budget of ap-proximately $200 million over 10 years. Sub-sequently, its budget has been expanded to ap-proximately $1 billion to support a range ofadditional activities. These include efforts toencourage incorporating VHSIC technologyin military systems in the three services, thedevelopment of a design automation system,and work on yield enhancement.7

SDI, started in 1983, is designed to studyand perhaps deploy a space-based missile de-fense. Since microelectronics technology wouldbe central to any such system, one programgoal is the development of advanced circuitryfor space-based military operations. In addi-tion to the DARPA GaAs work that is nowfunded through the SD I program element forsensors, SDI Innovative Science and Tech-nology (1ST) office is preparing to supportmultiple activities in microelectronics re-search. At present, the programs are being es-tablished, and funding levels are being de-bated. Although final amounts have not beenset, it is possible that 1ST will spend severalmillion dollars in this area over the next fewyears. The funds may comprise dollars fromother DOD research pockets (e.g., DARPA, asin the case of the GaAs IC work, or the serv-ices’ research offices), so they may or may notconstitute a net increase in the overall DODresearch funding for microelectronics. Cur-rently, contract monitors from other DOD re-search agencies (primarily ONR) are adminis-tering these funds.

National Science Foundation

The National Science Foundation (NSF) ischarged with supporting research in a broadrange of science and engineering fields. Micro-

7Eliot D. Cohen, Navy VHSIC Program Director, interview,July 1985, and additional comments, November 1985.

electronics R&D at NSF is funded primarilythrough the Directorate for Engineering. NSFspends approximately $23 million in areas thatinclude solid-state and microstructure engi-neering, quantum electronics, electronic ma-terials, electrical and optical communications,and VLSI. These funds support individuals orsmall groups of researchers.

NSF established the National Research andResource Facility for Submicron Structures atCornell University in 1977 and has providedit with about $2 million annually for the lastfew years. This facility is also supporteddirectly by industrial sponsors and indirectlyby other Federal agencies. In addition, NSFhas recently established an Engineering Re-search Center at the University of Californiaat Santa Barbara for Robotic Systems inMicroelectronics. NSF plans to give this cen-ter up to $14 million over the next 5 years, asum that the university will probably augmentwith support solicited from industry. The cen-ter will focus on automated systems for ICmanufacturing. 8 To some extent, these centersare evidence of a trend at NSF to concentrateits limited resources in a small number of largefacilities rather than granting small bundlesof money to a large number of investigators.

NSF’s style of supporting research differsa great deal from the DOD approach. Defenseagencies tend to seek out channels-at univer-sities, in industry, or at DOD laboratories—to accomplish their goals, with a focus on enduses. NSF, by contrast, is not a mission-oriented agency, so it responds to proposalsfrom the research community. NSF’s fundingfor microelectronics and related areas is quitesmall compared to the total DOD support, butthese monies provide some counterbalance tothe defense dollars. And while NSF has notplayed the major role in advancing the tech-nology, it has helped to build a base of qual-ified scientists and engineers, e.g., by helpingnew university professors get started withsmall grants. This gives NSF an importantrole as a broad basic-research agency.

‘Evelyn Hu, Associate Director, Center for Robotic Systems,interview, October 1985; and NSF literature on EngineeringResearch Centers.

.

23—

PRIVATE SECTOR R&DMany kinds of private sector organizations

are engaged in microelectronics research anddevelopment. They may be grouped into twobroad categories:

1.

2.

captive manufacturers—the parts oflarge, vertically integrated companiesthat make microelectronic components fortheir own products and services (typicallycomputers or telecommunications) or fortheir defense systems applications(termed “captive” because the primarymarkets for their microelectronic prod-ucts are internal); andmerchant firms-companies that make in-tegrated circuits and other microelectron-ic products to sell to the full range of endusers.

These categories are not mutually exclusive.Several companies (e.g., Texas Instruments)make microelectronic components for internaluse as well as outside sale. However, the divi-sion helps to illuminate the nature of R&D inmany companies, since the size and goals ofthe organization are key determinants of itsapproach to R&D.

Today, the most prominent force changingmicroelectronics companies (and thus theirR&D efforts) is Japanese competition. Mer-chant vendors are especially vulnerable. Thishas major implications for industrial R&D.Companies may cut back on R&D investmentas part of a general belt-tightening effort.Paradoxically, R&D is increasingly importantin the competitive environment. Thus, thecompanies are beginning to look to the Fed-eral Government for R&D support, eitherthrough direct funding for R&D, or throughFederal policies that ease the way for privatesupport, such as the tax treatment of R&D ex-penditures and intellectual property pro-tections.

Captive Manufacturers

Although microelectronics technology is be-ing applied in more and more ways, its pri-mary uses are still concentrated in computers,

telecommunications, and military systems.Because these uses dominate the field, mostcompanies with captive microelectronics oper-ations specialize in them.

Since these firms are generally very largeand relatively stable, they tend to support avery broad range of activities, from basic re-search to product development. For example,companies such as IBM and AT&T have thou-sands of scientists and engineers involved indifferent aspects of microelectronics R&D.

Telecommunications and computer technol-ogies are beginning to merge because of de-velopments in information technology and alsoas a result of the recent deregulation of thetelecommunications industry. The latter de-velopment allowed new entrants into the fieldand permitted AT&T, formerly excluded, toparticipate in the computer marketplace. Thedivestiture of AT&T also split the well-knownBell Telephone Laboratories into two researchorganizations: AT&T Bell Laboratories andBell Communications Research (Bellcore),which serves the seven regional Bell operat-ing companies jointly. Although the divesti-ture of AT&T officially occurred at the begin-ning of 1984, its long-term effects on researchare not yet completely clear. However, thereare preliminary indications of trends at thetwo companies.

Some changes are clearly underway atAT&T Bell Laboratories. R&D activities are,overall, becoming more closely linked withproducts as a result of the new competitiveenvironment that AT&T faces. Even so, thereis ample evidence that at least in areas relatedto microelectronics the organization will con-tinue to pursue a broad spectrum of basic re-search as well as development. The new envi-ronment has both negative and positiveimplications for R&D. Scientists and engineersin the research community are concerned that,whether or not AT&T Bell Labs shifts frombasic research, it will be less likely to sharethe fruits of its research with others. On theother hand, the pressure of competition willprobably drive AT&T Bell Labs to move re-

search into new products and services fasterthan it did in predivestiture years.

The prognosis for Bellcore’s role as an R&Dorganization is, if anything, even less certain.Bellcore is a unique laboratory in the UnitedStates because it serves seven separate, highlyregulated companies. It is not directly linkedto any particular manufacturing facility, andso will probably not experience the same shiftsin R&D as AT&T Bell Labs. In fact, Bellcoreto date exhibits signs of pursuing a vigorousbasic research program in microelectronics-and optoelectronics-related areas. But since itis a completely new organization, Bellcore willneed several years to establish an identity inR&D.

Other manufacturers of products for indus-trial and commercial use, such as Xerox andHewlett-Packard, also contribute heavily tomicroelectronics R&D. These companies, al-though significantly smaller than, for exam-ple, IBM and AT&T, are still large and diverseenough to support good-sized research efforts.Their specific markets tend to shift the direc-tion of R&D activities, so that each such com-pany pursues a somewhat different researchagenda. For example, Xerox’s interest in print-ing technology helped the company to achieveprominence in optoelectronics research.

Of the several captive microelectronics oper-ations that serve the military markets formicroelectronics, many carry out DOD-fundedand internal R&D. Much of the internal R&Dis funded by Independent Research and De-velopment (IR&D) funds which are derivedfrom overhead on defense R&D contracts. Thelargest players in this category are such com-panies as Hughes Aircraft, Honeywell, Rock-well, TRW, and McDonnell Douglas. Manyother companies that are well known for theircommercial efforts are also defense contrac-tors. Typical examples include IBM, AT&T,and Texas Instruments.

Merchant Companies

Merchant companies sell microelectronicproducts to users who incorporate them in avariety of systems—computers, communica-

tions systems, consumer products, controlequipment, and defense systems. California’sSilicon Valley firms (e.g., Intel and AdvancedMicro Devices) are the archetypes of this cat-egory. Generally, merchant firms have concen-trated on producing standard chips (micro-processors, logic chips, and memories), whichhave been used in the larger electronic sys-tems. Custom integrated circuits, which aredesigned for a user’s particular needs, haveheld part of the chip market for many years.As IC design and manufacturing become moreflexible, a wider range of application-specificintegrated circuits (ASICs), including customand semicustom chips, is drawing a largershare of the market. During the 1985 slumpin IC sales, ASICs constituted the one healthysegment of the market.9

Merchant companies, unlike their largercounterparts, tend to limit their R&D to thelast stages of development; their central con-cern is getting the latest design and fabrica-tion technologies into production. Two majorfactors converge to make longer term R&D ef-forts improbable in these firms. First, the com-panies depend on the sale of only one type ofproduct—semiconductor devices—for theirsurvival. They tend to be focused, lean opera-tions, with few discretionary dollars for basicresearch in an area where even a simple exper-imental facility costs several million dollars.Second, two hallmarks of the Silicon Valleyculture are the ease with which workers movefrom one company to another, and the fre-quency with which new companies spring up.Managers in merchant chip firms seldom findthat a heavy investment in long-term researchpays off in this fluid environment.

However, in the last several years, this com-munity grew concerned that its needs were notbeing met by universities and other basic re-search organizations. Manufacturing research,for example, is increasingly crucial to the con-tinued growth of the industry, but it had scantsupport among basic research organizations.

‘“A Chip Business That Is Still Growing: Innovation SpursMarket for Application-Specific Integrated Circuits, ” Electron-ics, July 22, 1985, p. 40,

To correct this, many merchant companiestoday help to support external R&D activitiesthrough different channels. For example, sev-eral merchant semiconductor firms fund uni-versity research through the SemiconductorResearch Corp. These companies also cooper-

2 5

ate in other R&D ventures. Several merchantcompanies also independently support re-search projects at universities. All of theseactivities center almost exclusively on aspectsof silicon technology.

COOPERATIVE R&DCooperative research and development

activities take different forms: organizationsthat are jointly funded from several differentsources or research facilities that are sharedby different groups of workers. These joint ef-forts represent a relatively new approach toR&D in the United States. In microelectronics,several factors are driving the growing trendtoward centralized funding for R&D:

research in microelectronics requires in-creasingly expensive facilities, which fewparticipants can afford alone;advances in microelectronics depend in-creasingly on multiple technical disci-plines, requiring a number of personstrained in different areas; andcooperative research can link academiaand-industry, thereby bringing necessaryfunding to universities and facilitatingthe process of transferring technologyfrom research to development to pro-duction.

Examples of cooperative microelectronicsresearch organizations include the Semicon-ductor Research Corp. (SRC), the Microelec-tronics Center of North Carolina (MCNC), andthe Microelectronics and Computer Technol-

ogy Corp. (MCC).10 Each of these channelsfunds from a variety of commercial firms touniversities and other basic-research organi-zations. MCNC and MCC also carry out in-house R&D activities.

In addition, a plethora of initiatives forshared facilities are emerging from Federalfunding agencies, ranging from NSF, which re-cently reorganized to focus resources on agroup of Engineering Research Centers, to theInnovative Science and Technology part of theStrategic Defense Initiative Organization,which is actively promoting the establishmentof research consortia. While many researcherssupport this trend, others point out that co-operative research is hardly a panacea formicroelectronics R&D. They argue that cen-tralization of resources threatens innovation,and that research done by a large number ofindividual investigators, who come togetherin small groups to communicate and col-laborate, is the most productive approach.

10The structure and Operation of these and other cooperati~reresearch organizations are described fully in U.S. Congress, Of-fice of Technology Assessment, Information Technology R&D:Critical ‘Ii-ends and Issues, ch. 6, OTA-C IT-268 (Washington,DC: U.S. Government Printing Office, Februar~ 1985).

UNIVERSITIES AND R&D

Universities serve two main functions in Support for research at universities comesR&D: they perform basic research suited to from many sources, including military andan academic environment; and they educate civilian agencies of the Federal Government,and train students who subsequently perform industrial organizations, and combinations ofresearch and development in industrial, gov- these. Microelectronics research takes manyernmental, and academic organizations. forms in this setting. Universities across the

26

Nation have individual research programs,typically in such departments as electricalengineering, physics, and chemistry. A hand-ful of universities have large research centers.These include the National Research and Re-source Facility for Submicron Structures atCornell University, Stanford University’s Cen-ter for Integrated Systems, the Microelec-tronic and Information Sciences Center at theUniversity of Minnesota, and the Center forRobotic Systems in Microelectronics at theUniversity of California at Santa Barbara.

The university role in preparing students forthe R&D community has many facets, some

of which have promptedfield of microelectronics.

disagreement in theAlthough many ob-

servers view the current trend toward largecampus engineering facilities as a useful wayto train students for the activities they willundertake in industry, some experts are con-cerned that this trend undermines the well-rounded education that universities ought toprovide for their students. Thus, there is anongoing debate about the best way for univer-sities to fulfill this part of their mission.

FOREIGN ACTIVITIESForeign activities in microelectronics R&D

are so diverse that a full treatment of the topicis beyond the scope of this paper.11 However,several prominent features of the Japanese ef-forts have important implications- for U.S.R&D in microelectronics.

Japan is the largest foreign supporter ofmicroelectronics R&D. Although observershave long viewed Japanese development andmanufacturing activities as competitive withor superior to ‘U.S. efforts in microelectronics,they ‘had generally believed that the UnitedStates excelled at innovation in basic research.Now, however, Japanese basic research effortsare drawing world-tide attention. In the wordsof one panelist of the Department of Com-merce’s Japanese Technology Evaluation(JTECH) Program:

It is often said that the U.S. invents andJapan copies. . . . [S]uch generalizations aregrossly inaccurate and certainly do not favora genuine understanding of our best competi-tor. 12

“For an extensive discussion of foreign R&D efforts in in-formation technology, see U.S. Congress, Office of TechnologyAssessment, Information Technology R&D: Critical Trends andIssues, ch. 7, OTA-CIT-268 (Washington, DC: U.S. GovernmentPrinting Office, February 1985).

l~Federico Capasso, AT&T Bell Laboratories, quoted in “Ja-pan Reaches Beyond Silicon, ” ZllIIE Spectrum, October 1985,p. 52.

Japanese companies continue to transfer re-search concepts to production with greatspeed. In this activity, they draw extensivelyfrom U.S. as well as Japanese basic researchresults. Another JTECH panelist states,

They do not seem to have difficulties withthe “not-invented-here” syndrome that slowstechnology advances into the marketplace inthe U.S.13

The United States has a harder time takingthe same advantage of Japanese work. One ofthe biggest barriers to access to Japanese re-search by U.S. workers is the language dif-ference.

The structure of the electronics industry inJapan strongly affects R&D. In contrast tothe United States, Japan has almost no mer-chant microelectronics firms. Its large, stable,vertically integrated companies can and do in-vest more heavily in long-term R&D than U.S.merchant firms. This suggests that the chal-lenge they pose to U.S. competitiveness willremain and quite possibly increase.

“Robert S. Bauer, Xerox Palo Alto Research Center, quotedin “Japan Reaches Beyond Silicon, IEEE Spectrum, october1985, p. 51.

.

Appendix A

Current Microelectronics Technology

This appendix provides general background in-formation about microelectronics technology, theroots of which extend back to the early part of thiscentury. Today, a vast assortment of integratedcircuits (ICs) and other miniature electronic de-vices are on the market. All of these componentspass through complex design and fabrication proc-esses before reaching consumers.

Electronics Technology From VacuumTubes to Integrated Circuits

Electronics technology over the last century hasadvanced in a series of dramatic breakthroughs:the vacuum tube, the transistor, and the inte-grated circuit. These are often used to designatedifferent generations of technology, as shown infigure A-1.

The history of electronic devices began in theearly 1900s with the invention of vacuum tubes,the first devices to reach major use in manipulat-ing and amplifying electrical currents. This tech-nology is limited by several related features.Vacuum tubes require high voltages and, bymicroelectronics standards, a great deal of power.

The invention of the transistor in 1947 markedthe beginning of a new generation in electronics.The inventors fabricated the device from a crys-tal of semiconductor material to which they addedcontrolled quantities of different impurities. Thetransistor had three points of electrical contact,or terminals. Like a vacuum tube, it could amplifyan electrical signal, but at room temperature, andit required only a small fraction of the power, volt-age, and space that tubes demand. The transistorrapidly became the central component in circuitsfor a variety of applications.

By the late 1950s, transistor-based circuits dom-inated early computers and military systems. Butthe drive for increasingly complex circuitry en-countered some difficult barriers. Millions of con-nections were required to turn the hundreds ofthousands of components into a functioningcircuit—a labor-intensive process with unreliableresults. Further complicating things, the total cir-cuit size was growing unmanageable, although theindividual components were quite small. Theequipment in which electronic devices were used

Figure A-1 .—Generations of Electronics Technology

Beyondvery large-scale

integration

t

VeryEarly 1980s large-scale

integrated circuits

1958 Integratedcircuits

b

t

1947 Transistors

t

Early 1900s Vacuumtubes

Fifth generation

Four th generat ion

Third generat ion

S e c o n d g e n e r a t i o n

Fi rs t generat ion

SOURCE Off Ice of Technology Assessment

demanded smaller, faster circuits that would con-sume less power and could be fabricated reliably.

The next major breakthrough in electronics wasthe invention of the integrated circuit in 1958. Un-til this time, all electrical circuits consisted of sep-arate components-transistors, diodes, resistors,capacitors—connected by wires. The inventors ofthe IC recognized that the wires and other com-

27

28

ponents could also be fabricated from or on thesame semiconductor material that was used tomake transistors.1 In this way, they made an en-tire circuit on a single piece, or chip, of the mate-rial. Using this technique, engineers were able tomake connections and components by etching thesemiconductor and depositing metals and insula-tors in patterns on the chip. The new process elim-inated many of the problems associated with pro-ducing highly complex circuits. Because the hugenumber of connections no longer had to be madeby soldering wires together, the IC technique wasmore reliable and less labor-intensive. The individ-ual components could now be smaller since theywould not need to be handled by people. Thisshrinkage yielded circuits that operated at higherfrequencies and consumed less power.

Progress in microelectronics over the last 25years has been based on further shrinkage of thecircuitry etched onto chips. The number of com-ponents per chip has, on average, almost doubledevery year, Today, up to 1 million transistors andother components are fabricated on a single chipthat may have an area of less than 1 square centi-meter.

Microelectronic Devices

In this background paper, the term “microelec-tronics” is used to refer to miniature electronic de-

‘For an account of the invention of the integrated circuit, see T. Il.Reid, ‘The Chip,” Science 85, February 1985, p. 32, excerpted from T.R.Reid, The Chip: 7’he Microelectronics Revolution and the Men W’hoMade It (New York: Simon & Schuster, 1985).

vices in general. Figure A-2 shows the types ofmicroelectronic devices on the market today. Themost numerous and widely used of these are stand-ard ICs made of silicon. Other semiconductordevices, such as optoelectronic and microwaveproducts and discrete devices, are also part ofmicroelectronics technology. Additionally, a fewexamples exist of miniature devices made frommaterials other than semiconductors (e.g., mag-netic bubble storage devices), but these fall out-side the scope of this paper.

Types of Integrated Circuits

Integrated circuits are commonly classified ina variety of ways and referred to with an impres-sive number of acronyms. They may be catego-rized by function, by the level of integration (num-ber of components), by the type of transistors usedin the circuitry, or by the underlying material (orsubstrate) on which they are fabricated.

Types of Functions.–Integrated circuits may beeither digital or analog (also called linear). In gen-eral, a digital circuit switches or stores voltagesthat represent discrete values. For example, Ovolts and 5 volts may represent the values O and1, respectively. An analog circuit, in contrast, am-plifies or otherwise modifies voltages of any valuewithin a given range (e.g., any voltage between Oand 5 volts). Some integrated circuits are designedto convert signals from analog form to digital form(A-to-D converters) or vice versa (D-to-A con-verters). Custom chips can combine the variousfunctions. Currently, the majority of ICs are digi-

Figure A-2.—Types of Microelectronic Products

Microelectronic products.

1

1 I IIntegrated Discrete Optoelectronic Microwavecircuits transistors devices - devices

Iand diodes . b

t 1? * 1

IDigital Analog Custom and other Lasers Monolithic

Ics (linear) ICs ICs Discreteand light-4 4 < Detectors devices microwaveemitting ICs

t diodes (LEDs)-1 f

Microprocessors a Memories Logic- ICs4

ARead-only Random-accessmemories memories

(ROMs) (RAMs)

SOURCE Office of Technology Assessment

2 9

tal; computers, the major use of ICs, are based ondigital integrated circuits.

The major product categories in digital inte-grated circuits are logic circuits, memories, andmicroprocessors. All three handle information inthe form of electrical signals: logic circuits proc-ess the signals, memories store them, and micro-processors combine the two functions.

Logic circuits carry out the operations requiredto manipulate data in binary digital form. Theyperform mathematical functions. Logic ICs aretypically classified according to the type of tran-sistors used to make them.

Semiconductor memories are microelectronic cir-cuits designed to store information. They fall intotwo groups: read/write memories and read-onlymemories (ROMs). Although both groups consistprimarily of memories that can be accessed ran-domly (i.e., any storage location can be accessedin the same amount of time), the term “random-access memory” (RAM) usually refers only toread/write memories. A computer can store a da-tum in a location within a read/write memory andlater retrieve it or store a new datum there. How-ever, the computer can only retrieve informationthat is already stored (by some external means) ina ROM. Thus, a ROM may be used to store un-changing information or instructions that the com-puter needs regularly, while read/write memoriesare used to store data as they come into the com-puter and as they are processed.

Read/write memories may be labeled “static” or“dynamic” depending on the design of the circuitsthat make up the RAM. Dynamic RAMs (DRAMs)store data in the form of charge on capacitors,and so require that the capacitors be rechargedregularly-every few milliseconds. Static RAMs(SRAMs), on the other hand, store data by chang-ing the state of transistors in the storage element,a technique that requires a constant flow of powerrather than intermittent recharging to maintainaccurate storage. Both static and dynamic read/write memories need a constant supply of powerto the circuit as a whole to operate.

Designers can program information into read-only memories in different ways. They may put theinformation in the physical chip design; this typeis known as a mask-programmable ROM. Alter-natively, chip manufacturers fabricate ROMs thatusers can program themselves (programmableROMs, or PROMs)–a much more versatile prod-uct. Some varieties of these PROMS can be erasedand reprogrammed using light or electrical signals(erasable PROMS, or EPROMs). EPROMs differfrom read/write memories in two ways: they typi-cally cannot be reprogrammed by the computer it-self, but they need no power source to retain in-formation.

The capacity of a semiconductor memory is de-termined by the number of bits that it can store.The maximum capacity available in the early1970s was 1,024 bits, commonly called a kilobit orkbit. Today, 262,144-bit (256 -kbit) RAMs are onthe market, and l,048,676-bit (l-megabit) RAMsare just around the corner. Prices for memorieshave dropped as dramatically as their capacity hasrisen: memory costs approximately one-thousandthof a cent per bit today.

Because a microprocessor is a complete com-puter processing unit on a single IC, it can proc-ess data expressed in a series of binary digits orbits (ones or zeroes). The number of bits used inthe series determines the precision of the datum’sdigital representation; precision increases with thenumber of bits. The original microprocessor, madein 1970 at Intel Corp., was designed to handle fourbits. The industry introduced 8- and 16-bit micro-processors by the end of that decade and 32-bitmicroprocessors within the last few years.

Levels of Integration.—The level of integrationof an integrated circuit refers to the number ofcomponents packed on the chip. The progress inintegration since the invention of the IC is shownin table A-1. Figure A-3 shows how a transistoron an IC compares in size with some other micro-scopic items.

Types of Transistors. -An integrated circuit mayalso be classified according to the kind of transis-

Table A-1 .—Levels of Integration— ———

Design rule Number of ApproximateLevel of integration (microns) transistors years

Scale-scale integration (SSI) . . . . ......... . . . ....30-20 2 to 64 1960-65Medium-scale integrat ion (MSI) . , . . . . . .20-10 64 to 2,000 1965-70Large-scale integration (LSI) . . . . . . . . . 10-3½ 2,000 to 64,000 1970-78Very large-scale integration (VLSI) . . ... 3½ -1 1/4 64,000 to 2 million 1978-86Ultra large-scale integration (ULSI) ... ... ... <1 l/4 >2 mil l ion after 1986N O T E The n u m b e r s (n Ihls table a;e approximate NO standard de flnlt;on extsts for the dlfferenl I ntegratton levels

SOURCE Adapted from James M Early and Bruce E Deal Falrchlld Research Center and C Gordon Bell Encore Computer Co.[)

30

Figure A-3. —Microscopic Sizes

I&l ++ 1 m i c r o n

Transistor on anintegrated circui t

approx. 1-2 microns

/ \

Bacteriumapprox. 3 microns

long

mCross section ofhuman hair

approx. 50 microns

1 Smallpart icle15 microns

o 5 10 15 20 25 30 35 40 45 50 55 60 65 M icrons

SOURCE Off Ice of Technology Assessment Adapted fr)m .] flqure by Jlmmle R Suttle Army Research Of f!ce

tor used in its design. For silicon ICs, two typesof transistors are used: bipolar transistors andmetal-oxide-semiconductor field-effect transistors(MOSFETs). These are further subdivided to de-scribe the specific structure of the device and thedesign of the surrounding circuit. The most com-mon bipolar designations are ECL (emitter-coupled logic) and TTL (transistor-transistorlogic); the most common types of MOSFETS areNMOS (n-channel MOS) and CMOS (complemen-tary MOS). (See the discussion of chip fabricationin this appendix for some additional details.)

The primary differences between bipolar andMOS technologies are speed (bipolar transistorsare faster); power (MOS circuits require lesspower); and ease of manufacture (MOS circuitshave higher yield). Because the yield determines,in large part, the cost of the chip, bipolar chips aretypically more expensive. However, the advantageof higher speed is worth the extra cost and powerrequirements for bipolar ICs in some applications.For example, most makers of large computers usebipolar ICs for the central components in their ma-

chines because they are the only chips capable ofthe high speeds that are required. On the otherhand, personal computer vendors tend to basetheir products on MOS chips, because cost is gen-erally a bigger concern than speed.

Types of Materials.–Virtually all integrated cir-cuits on the market today are made on substratesof the same material that was the base for the firstIC: silicon. The longevity of silicon is due to itsphysical properties as well as to practical economicconsiderations. Silicon IC technology is more eas-ily executed than is the technology of compoundsemiconductors, which have more than one ele-ment, because the chemistry of a material madeof a single element is innately simpler. Further-more, once silicon technology was established, eco-nomic factors dictated that any replacement be farsuperior to silicon to be worth the large costs ofconverting to a new system. And since silicon tech-nology has never stopped progressing-or evenslowed significantly-attempts to replace it havebeen shooting at a moving target. The recent de-cline of Josephson junction technology (which is

31

based on superconductors) for computers haslargely been attributed to this problem.

Despite its great usefulness, however, siliconcannot satisfy all demands of microelectronics.Since transistors made from gallium arsenide(GaAs) can be faster and more impervious to ra-diation damage than silicon transistors, severalcompanies are trying to introduce GaAs inte-grated circuits. These ICs are particularly attrac-tive for some military applications (e.g., space-based circuitry), and they may also find standardcommercial applications. However, they cannotcurrently compete with silicon technology forstandard functions. For example, a fully opera-tional 1,024-bit GaAs RAM is still in the labora-tory phase, while silicon RAMs with over 250times the capacity have been on the market forsome time already, (See ch, 3 for a more detaileddescription of current R&D activities in this area.)

Other Semiconductor Devices

Semiconductor materials are also used to makea host of individual microelectronic devices, asdistinguished from standard silicon integratedcircuits. These include discrete transistors anddiodes, optoelectronic devices, and microwavedevices.

Discrete Transistors and Diodes.—Electronic in-struments, including radios, televisions, and con-trol instruments, use not only ICs but also indi-vidual transistors and diodes. (Diodes are two-terminal devices that allow current flow in only onedirection.) These components are typically madefrom silicon, The transistors may be bipolar orfield-effect transistors.

Optoelectronic Devices. -Semiconductors such asgallium arsenide (GaAs), iridium phosphide (InP),cadmium telluride (CdTe), and mercury telluride(HgTe) are the bases for a range of other micro-electronic devices that require properties that sili-con lacks—the ability to interact efficiently withlight and the ability to operate at extremely highspeeds or frequencies.

Optoelectronic devices depend on the first ofthese properties. Optoelectronics includes light-emitting diodes (LEDs) and lasers, which convertelectrical signals to light signals, and photodetec-tors, which convert light signals into electrical sig-nals. The two major applications for these devicesare fiber optic communications and long-wave-length infrared light detection.

Fiber optic or lightwave communications sys-tems use glass fibers to transmit light signals fromone point to another; the fibers can replace metal

wires that carry electrical signals. Semiconductorlasers or LEDs generate the light signals, and semi-conductor photodetectors convert the receivedlight signals back to electrical signals. These de-vices are most commonly made from layers ofGaAs and aluminum gallium arsenide (AIGaAs) ona substrate of GaAs, or, more recently, from layersof iridium gallium arsenide phosphide (InGaAsP)and iridium gallium arsenide (InGaAs) on a sub-strate of InP. The quaternary materials-differentcompositions of InGaAsP-can generate and de-tect the wavelengths of light that travel along theglass fibers with the least distortion (wavelengthof 1.3 microns) and fading (wavelength of 1.55microns).

The primary purpose of long-wavelength in-frared (IR) detection is to “see” living or hot ob-jects in the dark. The Department of Defense hasspearheaded work on IR sensors, since their po-tential military uses are many. Typically, the de-vices are made from alloys of HgTe and CdTe, be-cause these materials can be adjusted to detect theappropriate wavelengths of light (3 to 5 micronsand 8 to 14 microns). Special types of silicon di-odes can also act as IR sensors in some of thesewavelength ranges.

Microwave Devices.-Microwave (high-frequen-cy) devices take advantage of the intrinsic highspeed of gallium arsenide. They generate, amplify,switch, and receive microwave signals with fre-quencies from approximately 1 to 60 gigahertz (1gigahertz is 1 billion cycles per second). Radar sys-tems and transmission systems for telephone, tele-vision, and telegraph signals, which operate atthese frequencies, as well as military systems, de-pend on these microwave devices. This technologyis at present progressing to monolithic microwaveintegrated circuits (MM ICs), in which microwavecircuits are fabricated on a single substrate, typi-cally gallium arsenide.

Technologies To ProduceSemiconductor Microelectronics

Numerous steps are involved in making anymicroelectronic device. Integrated circuits are themost complex in this respect. The process can beseparated into two parts: circuit design and chipfabrication.

Circuit Design

Once the application for a new integrated circuitis established, the first step in making the chip isthe design of the circuit. The use of a wide variety

32

of computer-aided design (CAD) tools has facili-tated all parts of the design process for large ICstremendously. For a digital IC, the overall circuitwill typically be a combination of logic functionsand storage functions. The designer determinesthe layout of the subcircuits that carry out eachof these functions and the connections between thesubcircuits. Also in this initial stage, the chipdesigner must establish mechanisms for bringingthe external signals and power to the various sub-circuits. Each subcircuit may itself be composedof smaller, interconnected cells of circuitry. Thefundamental units for the designer are the indi-vidual components, such as transistors, diodes,resistors, and capacitors, and the connections be-tween them.

Because the process of designing circuits hasgrown so complex over the years, now involvingover 100,000 components on a chip, design soft-ware is a major research area which is as impor-tant to progress in microelectronics technology asadvancements in chip fabrication. As the capabil-ities of design tools grow, they provide increas-ingly greater flexibility for different chip architec-tures, and therefore open the door to using ICtechnology in completely new ways to fabricatespecialized circuits.

Chip Fabrication

Chip fabrication includes making the circuitryand packaging the completed chip.

The first step of the process is transferring thecircuit designers’ layout onto the semiconductorchip. The substrate onto which the circuits aretransferred is a wafer, or thin slice, of a single crys-tal of the semiconductor. Currently, a typical wa-fer is several inches in diameter, so many chips canbe made on a single wafer and separated after cir-cuit fabrication. The circuit is created by deposit-ing thin layers of metals, insulators, and perhapsadditional semiconductor materials on the wafer;adding n-type impurity atoms (negative chargecontributors) or p-type impurity atoms (positivecharge contributors) to the semiconductor; andetching away precisely defined portions of the vari-ous layers with chemicals or ions.

Illustrating a simplified example of one part ofthis process, figure A-4 shows the steps in the firstphase of the fabrication of a single transistor-ann-channel MOSFET—on a silicon substrate. Theobjective of this first phase is to add p-type impu-rity atoms in two small selected regions near thesurface of the chip, thereby forming “p-wells.” Thefirst six steps of this phase (which constitute li-

Figure A-4.– First of Four Mask Processes To Makean N-channel MOSFET

n-type I I n-typeA. B

siIicon silicon ILight

EI

n-type

I

F

I

n-typesilicon silicon I

p - , t y p e i m p u r i t i e s

I I I I

Metal contacts

SOURCE Office of Technology Assessment

thography) leave a layer of silicon dioxide (SiO2),with small holes etched in it, on top of the sub-strate. This layer serves as a stencil for the p-typeimpurities to which the chip is then exposed: onlythe portions that are not covered with SiO2 becomep-type.

This process is accomplished as follows. Thestarting material is a wafer of silicon to which n-type impurities were added during growth. Thefirst step is to grow a thin layer of silicon dioxideon the surface of this substrate, either by expos-ing the surface to the proper chemicals, or by heat-ing the wafer in a furnace with water vapor (stepA in figure A-4). Next, a layer of light-sensitive ma-terial called photoresist coats the oxide (step B).The portions of the photoresist that are exposed

—

when the surface is irradiated with light througha glass mask (step C) undergo a chemical change.These parts of the photoresist do not wash awaywhen the chip is rinsed in the chemical developer,but the unexposed portions do (step D). Now, thechip is placed in a bath of chemicals that etch awaysilicon dioxide, but leave the photoresist and sili-con intact (step E). Since the portions of the ox-ide that are below the photoresist are not etched,this procedure opens windows in the oxide layer.A different chemical bath removes the remainingphotoresist without damaging the oxide regions(step F). The chip is now ready for exposure to thep-type impurities, which are driven into the chipeither from a piece of source material in a furnaceor by accelerating impurity ions into the material(step G). When the number of p-type impurities ex-ceeds the number of n-type impurities that werethere originally, the wells will exhibit p-type be-havior. Finally, when the p-wells have been formed,the remaining SiO2 can be chemically removed(step H).

After three more mask cycles, the completed de-vice is composed of regions of semiconductor withp- and n-type impurities; silicon dioxide, which actsas an electrical insulator; and aluminum electricalcontacts. In present-day chips, the physical sepa-ration between the regions with p-type impuritiesmay be less than 1 micron, and the entire transis-tor is invisible to the naked eye. Furthermore, thenumber of impurities must be controlled to withina few parts per billion in some regions of thedevice.

Fabricating an entire integrated circuit is sig-nificantly more complicated because a larger num-ber of much more complex masks are necessary.Chips now also require extremely sophisticatedtechniques for selective etching of materials, pho-toresist exposure, and other parts of the process.All of these processing steps must be carried outin an environment completely free of particulate—clean-room facilities. Thus, it is not surprisingthat the trend in chip fabrication has consistentlybeen towards greater automation of the processes,which reduces the chances of contamination or hu-man error. Today, most machines used in chipfabrication can handle cassettes containing manylarge wafers of silicon, minimizing the need for hu-mans to handle the delicate circuitry.

When the final lithographic procedures are com-plete, the large silicon wafer is tested and dicedto separate the individual chips. The good chipsare then assembled in packages appropriate fortheir particular applications. Packaging technol-ogy is a crucial part of the production of ICs.Proper packaging techniques can shield the chipfrom physical damage and some forms of radiationdamage, For commodity ICs that are sold to a va-riety of end users, standard packaging systems arenecessary to ensure that chips are interchangeable.On the other hand, chips designed and used forhigh-speed computation require special custompackages that maximize the speed with which sig-nals enter and leave the chip, since standard pack-aging schemes could obliterate the special speedadvantage of the chip by itself.

Appendix B

Glossary of Terms

Chip: A small piece (typically less than 1 squarecentimeter) of a semiconductor wafer. Also usedto refer to a packaged integrated circuit.

Compound semiconductor: “A semiconductor madeof a compound of two or more elements insteadof a single element like silicon, II I-V semicon-ductors are made from elements in group 111 ofthe periodic table (such as aluminum, gallium,and iridium) and group V of the periodic table(such as nitrogen, phosphorus, arsenic, and an-timony). Binary compounds are made with twoelements, ternaries with three elements, andquaternaries with four elements. ”

Custom and semicustom integrated circuits: ICs thatcan be designed to varying degrees by the enduser for a specific application. Full custom ICsare designed and fabricated from scratch: semi-custom chips allow the user to modify a chip forthe application.

Design rule: Minimum feature size. Current chipstypically employ design rules of 1 micron orgreater. See table A-1 in app. A.

Diode: A two-terminal electronic device that allowscurrent to flow freely in one direction only.

Gallium arsenide (GaAs): A compound semiconduc-tor with properties necessary for very high-fre-quency microwave (analog) devices and opto-electronics. There are also several efforts tomake high-speed digital ICs based on galliumarsenide.

Integrated circuit (IC): Electronic circuits, includ-ing transistors, resistors, capacitors, and theirinterconnections, fabricated on a single smallpiece of semiconductor material (chip). Catego-ries of ICs such as LSI and VLSI refer to thelevel of integration, which denotes the numberof transistors on a chip. ICs may be digital (logicchips, memory chips, or microprocessors) oranalog. The transistors that ICs are made of

p

may be bipolar transistors, or one of a varietyof metal-oxide-semiconductor (M OS) tran-sistors.

Logic chips: ICs that manipulate digital data. Seeapp. A.

LSI: Large-scale integration. See table A-1 in app.A.

Memory chips: Devices for storing information inthe form of electronic signals. See app. A.

Microprocessor: A computer central processingunit on a single chip. See app. A.

Micron: One-millionth of a meter. See fig. A-3 inapp. A.

Microwave circuit: An analog circuit designed toprocess high-frequency signals. See app. A andch. 3.

MMIC: Monolithic microwave integrated circuit.See app. A and ch. 3.

Optoelectronic devices: Devices that convert lightsignals to electronic signals and vice versa. Seeapp. A and ch. 3.

RAM: Random-access memory. See app. A.ROM: Read-only memory. See app. A.Semiconductor: A crystalline material whose elec-

trical conductivity falls between that of a metaland that of an insulator. Semiconductor mate-rials are used to fabricate virtually all micro-electronic devices. Silicon is the most common;others include germanium, gallium arsenide(GaAs), indium phosphide (InP), mercury tellu-ride (HgTe), cadmium telluride (CdTe), and al-loys of these compound semiconductors. See fig.1 in ch. 3 for a periodic table of the elementsshowing common elements used in semicon-ductors.

Substrate: A piece of material, typically a semicon-ductor, on which layers of materials are depos-ited and etched to fabricate a device or a circuit.

Transistor: A three-terminal electronic device thatcan be used to switch or amplify electronicsignals.

VLSI: Very large-scale integration. See table A-1in app. A.

35