annual progress report for through … progress report for july 1, 1980 through june 30. 1981,(u)...
TRANSCRIPT
AD-AI04 727 ILLINOIS UNIV AT URBANA COORDINATED SCIENCE LAB F/6 9/1ANNUAL PROGRESS REPORT FOR JULY 1, 1980 THROUGH JUNE 30. 1981,(U)AUG Al R T CHIEN, H V POOR NOOOl0-479--042
UNCLASSIFIED NL*2IIflIfllfllfllfllfEIIEEIIIIIIIEEIIIIEEIIEIIEEEllullulumllllluEIIEEIIIIIIIEEllllllllEElhE-I//EEEEEEEII
AUGUST, 1981 LEVEL1- FCOORDINATED SLCIfNL' IABORAIdRY
ANNUAL PROGRESS REPORT 1980-81
I
I
I
SI
v UNIVERSITY OF ILLINOIS Ar URBANA-CHIAMPAIGN
1 81 9 28 255
The research reported in this document was made possible through support extendedthe Coordinated Science Laboratory, University of Illinois, by the Joint ServicesElectronics Program (U. S. Army Research Office, Office of Naval Research and theAir Force Office of Scientific Research) under Contract Number N00014-79-C-0424.
Portions of this work were also supported by:
U. S. Air Force - Office of Scientific Research
Grant AFOSR 78-3633 Grant AFOSR 79-0029 Grant AFOSR 80-0084
U. S. Air Force
Contract AF F19628-78-C-0040 AF F19628-81-K-0031
U. S. Army Research Institute
Grant DAHC-19-78-G-O011 Contract MDA 903-79-C-0421
U. S. Army Research Office
Grant DAAG-29-78-G-0114 Contract DAAG-29-80-C-011.Contract DAAG-29-80-K-0069 Contract DAAG-29-81-K-006 2Contract DAAG-29-81-K-0064
U. S. Department of Agriculture
Grant SEA AR 58-519-B-0-884
U. S. Department of Transportation
Contract DOT FA 79-WA 4360
U. S. Navy -- Office of Naval Research
N00014-75-C-0612 N00014-76-C-0806 N00014-79-C-0768N00014-80-C-0858 N00039-80-C-0556 N00014-80-C-0740N00014-80-C-0802 NO0014-81-K-0431 N00014-8 l-K-0014
National Aeronautics and Space Administration
NASA NCC-1-52 NASA NSG-2119 NASA NAG-1-90National Science Foundation
NSF EKG 75-20864 NSF DMR 78-25015 NSF ECS 80-07075NSF DMR 77-23723 NSF ENG 79-08778 NSF IST 80-12240NSF INT 77-20969 NSF ENG 79-01686 NSF DMR 80-24061NSF ERG 77-20436 NSF IST 79-16471 NSF MCS 80-08854NSF DSI 77-26886 NSF SED 79-18501 NSF 1ST 80-12242NSF ENG 78-17815 NSF ECS 79-19396 NSF INT 80-18622NSF ENG 78-06630 NSF ECS 79-16453 NSF ECS 81-06008NSF ENG 78-11758 NSF IST 79-21018 NSF MCS 81-05552NSF MCS 78-13642 NSF MCS 80-07049 NSF SUB/Purdue Univ.
Naval Air Engineering Center
N68335-80-C-0506
University of Illinois (as acknowledged in footnotes in the text)
Physical Electronics Affiliates Program and Communications, Antennas and PropagationAffiliates Program (College of Engineering, University of Illinois)
.4
(the list continues on back page)
.......1
UNCLASSIFIED
SECURITY CLASSIFICATION Or TNIS PAGE ("on Defe Entered)
READ INSTRUCTIONSREPORT DOCUMENTATION PAGE BEFORE COMPLETING FORMI. REPORT NUMBER |2. GOVT ACCESSION NO. 3. RECIPIENT'S CATALOG NUMBER
4. TITLE (and Subtitle) 5. TYPE OF REPORT & PERIOD COvERED
I ANNUAL .ROGRESS REPORT for July 1, 1980 throughJune 3, 1 9 8 1A 6. PERFORMING ORG. REPORT NUMBER
17._ AV 8. _QNRkl;T OR GIRANT NUM6ER(s)
l '. /- / , / N00014-79-C-0424 /
9. PERFORMING OIGANIZATION NAME AND AOORESS 10. PROGRAM EL.EMENT, PROJECT. TASKAREA a WORK UNIT NUMBERS
Coordinated Science Laboratory
University of Illinois at Urbana-ChampaignUrbana, Illinois 61801 ___ _____/-_'
II. CONTROLLING OFFICE NAME ANO AODRESS 2- REPORT W.i3
Electronics Program through the Office of Naval \ AughlilP 81Research, Arlington, VA 22217 ,3. NUMBERor PAGES
15914. MONITORING AGENCY NAME & AOORESS(if different from Controlling Office) IS. SECURITY CLASS. 'o( this repot)
Unclassif ied
IS. DICLASSIFICAT1ON, OOWNGRAOINGSCMEDULE
I6. DISTRIBUTION STATEMENT (of this Report)
IApproved for public release. Distribution unlimited.
17. OISTRIBUTION STATEMENT (of the abstract entered in Slock 20, It ilferent from Report)
1S. SUPPLEMENTARY NOTES
19. KEY WOROS (Continue on reverse side it necesary and identify by block number)
20. ABSTRACT (Continue on r.i'lrs. side It necessary and identify by block number)
This report summarizes all sponsored research activities that occurred in the
Coordinated Science Laboratory during the period July 1, 1980 to June 30, 1981.The summaries are categorized into nineteen technical areas. A comprehensivelist of faculty, graduate students, publications, and supporting agencies during
i l this period of time is included. ti"( .... I v .
D ORM,3 1473 UNCLASSIFIED / 173SECURITY CLASSIFICATION OF 4IS &AGE 'When Oat Entered)
- I -~eL ~I
SECURITY CLASSIPICATION OF TI41S PAGEf*%mn DataaEto")
SECURITY CL.ASSIFICATION OF THIS PAOGIhon Doe Etle.,.)
II
I COORDINATED SCIENCE LABORATORY
3 UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN
I
I ~ANNUAL PROGRESS REPORT 1980-81
1 August, 1981
Accession ForNTIS sGRA&I
I DTIC TABUaannounced
Submitted by: R. T. Chien Justificatio n E_
Edited by: H. V. Poor By bD istribut ion/Availability Codes
Dizt I Special
Approved for public release. Distribution unlimited.
Iv
ityCoeCoe
I!
TABLE OF CONTENTS
Page
P PERSONNEL ............................................................... vii
PUBLICATIONS AND REPORTS ................................................... x
I. PHYSICAL ELECTRONICS
I 1. MOLECULAR BEAM EPITAXY
1.1 MBE Growth and Characterization ................................... 11.2 Transport Properties of Heterojunction Layers ..................... 2
1.3 Theory of Surfaces and Heterojunction Interfaces .................. 3
2. SEMICONDUCTOR MATERIALS AND DEVICES
2.1 Introduction ...................................................... 42.2 Hydride Vapor Phase Growth of High Purity InP ..................... 52.3 LPE Growth of InGaAsP ............................................ 52.4 Fourier Transform Spectroscopy of Shallow Donor Levels ............ 6
3. -QUANTUM ELECTRONICS,
3.1 Excitation and Study of Highly-Excited VibrationalStates of Molecules ............................................... 8
3.2 Energy Transfer Processes in Electronically Excited
Atoms and Molecules .............................................. 93.3 Excited State Chemistry in Gases ................................. 10
4. SEMICONDUCTOR PHYSICS,
4.1 Introduction .................................................... 124.2 Laser and Electron Beam Annealing of Semiconductors .............. 134.3 Studies of Transport Properties in Semiconductors ................ 144.4 Materials Studies of InP ......................................... 15
5. THIN FILM PHYSICS
5.1 Semiconductor Crystal Growth from the Vapor Phase:Ion-Surface, Plasma, and Laser Stimulated Reactions .............. 17
5.2 Growth of Single Crystal Metastable Semiconductors ............... 195.3 Reactive Ion Etching of GaAs ..................................... 205.4 Laser-Surface Interactions ....................................... 21
6. MICROWAVE ACOUSTICS
6.1 Introduction ..................................................... 236.2 Electron Transport in Piezoelectric Semiconductors ............... 246.3 A New Method to Measure the Mobility of Amorphous
Semiconductor Films .............................................. 256.4 Wideband LAW Transducer ......................................... 26
6.5 A Theoretical Determination of Surface Acoustic Wave Velocityand Impedance Differences between Metal Strips and FreeSurface Regions of Metallic Gratings ............................. 26
p - -.--|---., ...-
iv CONTENTS
7. SURFACE STUDIES
7.1 Diffusion and Reactivity on Metals ............................... 287.2 Diffusion and Reactivity on Silicon .............................. 327.3 Atomic Exploration of Crystal Layers ............................. 36
8.' ELECTROMAGNETIC COMMUNICATION, RADIATION AND SCATTERING)
8.1 Millimeter-Wave Integrated Circuits .............................. 418.2 Electromagnetic Radiation and Scattering ......................... 43
9. . PLASMA PHYSICS
9.1 Statistical Properties of Plasma Turbulence ...................... 44
10. RAREFIED GAS DYNAMICS AND COMPUTATIONAL GAS DYNAMICS
10.1 Rarefied Gas Dynamics ........................................... 4510.2 Naval Hydrodynamics ............................................. 46
II. COMPUTER AND INFORMATION SYSTEMS
11. COMPUTER SYSTEMS
11.1 Performance of Instruction Pipelines............................ 5511.2 Test Generation for Complex Digital Systems ..................... 5611.3 Concurrent Error Detection in Arithmetic and Logic Units ........ 5711.4 High Performance and Reliability in Systolic Structures ......... 5711.5 Memory Addressing Architecture .................................. 5711.6 Shared Cache Organizations for Multiple-Stream Computer Systems.5811.7 Multiprocessors with Private Cache Memories ..................... 5811.8 Memory System Architecture for Parallel Processing .............. 5911.9 Computer Architecture for High Level Languages .................. 5911.10 Logic Validation ................................................ 5911.11 Simulation Packages for Complex System Design ................... 60
12. APPLIED COMPUTATION THEORY
12.1 Introduction ................................................... 6112.2 Parallel Computation in VLSI...................................6212.3 Computational Geometry .......................................... 6212.4 Approximation Algorithms ........................................ 6212.5 Storage and Access Costs for Information Retrieval .............. 63
13. ADVANCED AUTOMATION
13.1 Natural Language Understanding .................................. 6413.2 A Knowledge-based System for Intelligent Monitoring
and Diagnosis .................................................. 6613.3 Automated Air Traffic Control ................................... 6613.4 Theoretical Computer Vision ..................................... 6713.5 Pilot Interaction with Automated Airborne Decision
Making Systems .................................................. 6813.6 Human Problem Solving Performance in Fault Diagnosis Tasks ...... 6813.7 Modeling of Human Behavior in Seeking and Generating
Information ..................................................... 68
V
3 CONTENTS
14. INFORMATION RETRIEVAL RESEARCH
14.1 Introduction .................................................... 7014.2 Analysis of Database Data ....................................... 7014.3 An Integrated Man/Machine Interface to Facilitate
Network Resource Utilization .................................... 7114.4 Directory of Computer-Readable Bibliographic Databases .......... 7314.5 University of Illinois Online Search Service .................... 7514.6 Database Analyses ............................................... 7714.7 Automatic Detection of Database Overlap ......................... 77
3 III. COMMUNICATIONS AND CONTROL SYSTEMS
15. COMMUNICATIONS
1 15.1 Multiple-Terminal Digital Communications ........................ 8015.2 Signal Detection, Estimation, and Filtering ..................... 8715.3 Data Compression Theory and Techniques .......................... 91
16. ANALOG AND DIGITAL CIRCUITS
16.1 Computer Algorithms for Integrated Circuit Design .............. 9416.2 A New Structured Logic Array ................................... 9616.3 Hierarchical Design Techniques for VLSI Systems ................ 97
16.4 The Use of Multiple Criterion Optimization for theDesign of Digital Filters ...................................... 97
16.5 Analytical and Computer-Aided Design of New Circuit
Structures for Switched Capacitor Filters ...................... 98
16.6 Automatic Tuning Algorithms for Analog Filters ................. 99
16.7 Fault Isolation in Analog Circuits ............................. 99
17. DECISION AND CONTROL
17.1 Control and Decision Strategies for Systems UnderImperfect Information ......................................... 101
17.2 Implementation Constrained Decomposition and
Hierarchical Control ........................................ 104
17.3 Control Strategies for Complex Systems for Use inAerospace Avionics ............................................ 105
17.4 Large Scale Systems ........................................... 106
17.5 Power System Normal and Security State Assessment ............. 107
17.6 Control of Stochastic Systems Containing ParameterUncertainty ................................................... 108
17.7 Aggregated Hierarchical Control ............................... 109
18. DIGITAL SIGNAL AND IMAGE PROCESSING
18.1 An Investigation of Residue Number Architectures forDigital Filters with Fault Tolerance .......................... 110
18.2 Roundoff Noise and Limit Cycles in Digital Filtering .......... 112
18.3 New Techniques in Digital Signal Processing for
Synthetic Aperture Radar ...................................... 113II
vi CONTENTS
18.4 New Directions in Multidimensional Signal Processing .......... 11418.5 Image Sequence Processing and Dynamic Scene Analysis .......... 11418.6 Underwater Image Processing ................................... 115
19. 'POPULATION/FOOD/CLIMATE STUDIES,t-
19.1 Summary of Progress ........................................... 117
J i
vii5 COORDINATED SCIENCE LABORATORY PERSONNEL
Administration
I Professor R. T. Chien, DirectorProfessor G. G. Judge, Associate Di.rectorL. F. Selander, Assistant to the Director
I Faculty
I Abraham, J. A. Hajek, B. E. Perkins, W. R.Allen, R. Hajj, I. N. Poor, H. V.Adler, R. Handler, F. Preparata, F. P.
m Ahuja, N. Hess, K. Pursley, M. B3 Basar, T. Huang, T. S. Raether, M.Bensoussan, A. Hunsinger, B. J. Rau, B. R.I Bitzer, D. L, Jackson, E. A. Ray, S. R.Brown, D. J. Jenkins, W. K. Rouse, W. B.Brown, R. M. Johannsen, G. Sarwate, D. V.Chien, R. T. Johnson, W. B. Sauer, P. W.I Cho, A. Y. Judge, G. G. Schlansker, M. S.Coleman, P. D. Kieffer, J. C. Slottow, H. G.Cooper, D. H. Klein, M. V. Smith, L. C.
* Cruz, J. B., Jr. Kokotovic, P. V. Steinbach, A.
5 Datta, S. L~ightner, M. R. Stillman, G. E.Davidson, E. S. Looze, D. P. Streetman, B. G.' DeJong, G. Loui, M. C. Swamy, S.DeTemple, T. A. Maran, L. R. Tantaratana, S.Divilbiss, J. L. Mayeda, W. Trick, T. N.Dow, J. D. McEliece, R. J. Tucker, J. R.' Eden, J. G. Michie, D. Van Valkenburg, M. E.Ehrlich, G. Mittra, R. Verdeyen, J. T.El-Masry, E. 1. Morkoq, H. Waltz, D. L.I Flower, R. A. Muller, D. E. Wax, N.Greene, J. E. Munson, D. C. Williams, M. E.Haddad, A. H. Patel, J. H. Yen, S. M.
Academic Staff
l Avramovic, B. R. Krone, H. V., Robins, C. G.Basar, T. U. Lannom, L. W. Rouse, S. H.I Chambers, R. S. Merlin, R. Salman, M.Culton, J. W. Mills, C. Selander, L.Fink, H. W. Mon, K. K. Wada, N.I Hammer, J. M. Peterson, L. J. Welborn, M.Kane, J. Preece, S. E. Wilkus, S.Kirkwood, B. D. Ravlin, H.
viii
Graduate Students
Aazhang, B. Freudenberg, J. Lentine, A.Abrams, S. Fu, T. Lindner, D.Adrian, F. Fuchs, K. Liou, B.Alston, C. Fung, L. Lo, T.Amir, R. Gahutu, D. Low, T.Anderson, K. Garber, F. Lubben, D.Andreatta, R. Garza, L. Maguire, J.Attaie, N. Geraniotis, E. Mahon, S.Ault, S. Gollakota, M. Mak, G.Bannerjee, S. Goodman, B. Marshall, J.Barnett, S. Graf, M. Martin, E.Benhabib, J. Greenberg, K. Martin, P.Bhattacharyya, A. Grieger, J. McCaffrey, C.Bhoosan, S. Grohoski, G. McEwen, R.Bilardi, G. Hadden, G. McNiven, G.Blumer, A. Hall, D. Mendoza, J.Bose, P. Halperin, D. Miller, J.Bovik, A. Hanes, L. Miller, R.Brass, S. Hanlon, M. Morehead, D.Brennan, K. Hayner, D. Morris, N.Brew, W. Henneman, R. Nguyen, D.Bridwell, N. Higgins, W. Norton, R.Brooks, G. Ho, W. Oberstar, J.Burroughs, S. Hocevar, D. O'Brien, J.Byerly, K. Hong, F. Ogier, R.Cadien, K. Hoskins, M. Osmundson, J.Chan, S. Hsu, P. Pahlmeyer, C.Chan, Y. Huang, K. Paleta, R.Chang, C. Hunt, R. Pan, Y.Chen, D. lanno, N. Parent, P.Cheng, W. loannou, P. Paul, D.Chia, W. Janes, D. Peponides, G.Chillarege, R. Johnson, R. Phillips, R.Chin, B. Jones, B. Pleszkun, A.Chou, T. Kafka, H. Pollack, J.Cook, L. Kaplan, M. Pollard, H.Cortelazzo, G. Kastner, R. Posluszny, J.Coulman, D. Keck, K. Pracchi, M.Cross, S. Keever, M. Pun, K.Cruz, R. Kessler, A. Rao, V.Darragh, J. Kielblock, K. Ray, M.Davis, R. Kinsel, J. Ray, S.DeGroat, J. Klinger, R. Reidy, M.DeJule, R. Kollaritsch, P. Renjen, S.Desai, M. Koob, G. Resman, J.Drummond, T. Kopp W. Rivani, Y.Emma, P. Kramer, B. Rivaud, L.Enge, P. Lampe, R. Rockett, A.Fath, J. Leap, J. Rogers, W.Fledderman, C. Lee, C. Romano, L.Fletcher, R. Lee, C. F. Roth, T.Fliegel, F. Lee, S. Ruby, D.Franke, S. Lehnert, J. Saksena, V.
Iix
i Graduate Students
I Salhi, H. Tashima, M. Wei, Y.
Shichijo, H. Top, F. Wen, J.
Skromme, B. Trempel, C. Windhorn, T.
3 Soda, K. Trinh, T. Witkowski, L.
Specht, L. Tsai, R. Wrigley, J., Jr.
Spoor, D. Ufferheide, C. Yeh, P.
I Stadler, C. Van Loon, T. Yen, R.
Stark, W. Vastola, K. Yu, A.
Stothoff, J. Verdu, S. Yuan, C.
Strickland, J. Walker, T. Ziegler, C.
I Tang, Y. Wallace, M. Zielinski, J.
Tao, L. Wang, S. Zinkiewicz, L.
Webster, C.
I Nonacademic Staff
Bales, R. B. Gladin, R. T. MacFarlane, R. F.
I Bandy, L. E. Harris, Rose Murphy, Dixie
Beaulin, W. R. Jewell, Christine Perrero, Barbara
Bouck, G. A. Kesler, Evelyn Roberts, G. L.
Brennecke, Susan Kimball, A. Schrock, Debra
Champagne, Barbara Lawrence, W. I. Thrasher, W. X.
Coad, Geraldine Little, Gertrude Vassos, N.
Decker, Rosemary Lofton, C. M. Wegeng, Rosemary
Ebeling, R. E. Madden Sally Williams, S.
Foster, Mary McLellan, Mary Young, Phyllis
Gardner, 0. E. McNeal, Wilma
II
I
1III
xI3 PUBLICATYi S
1 1. MOLECULAR BEAM EPITAXY
I Journal Articles
1.1 H. Y.orkoQ. L. C. Witkowski, T. J. Drummond, C. A. Stanchak, A. Y. Choand B. I. Streetman, 1"rowth Eonditions to Achieve Mobility Enhancenent9in Al -a, Heterojinctions by ABE,' Electron. Lett. 16, 753-4
(1980.
1.2 L. ". Witkowski, T. J. Drummond, C. M. Stanchak and H. Aorko¢, "HighElectron Mobilities in Modulation Doped AI~aAs/aAs HeterojunctionsPreparel by ABE," Appl. Phys. Lett. 37, 1033-1:35 (1990).
1.3 C. X. Stanchak, H. M4orkoQ. L. C. Witkowski and T. J. Drummond,"Automatic Shutter Controller for Molecular Beam Zpitaxy," Rev. Sci.lnstrum. 52, 105-1)9 (1931).
1.4 T. J. Drummond, H. Morkoq, and A. Y. Cho, "Dependence of ElectronMobility on the Separation of Electrons and Donors in Al 3 1. s/3aAsHeterostructurs J. App. Phys. 52, 1330 (1981).
1.5 1. Keever, H. 3hi:hi"o, 'K. Hess, S. Banerce, L. C. Witkowski, H.Morko;, and 3. 3. Streetman, "Measurements of Hot Electron Conductionand Real Snace Transfer in GaAs-Al Ga1 As Heterojunction Layers," Appl.Phys. Lett. 33, 36-38 (1981).
1.6 A. Keever, T. J. Drummond, K. Hess, H. Morkoq and B. G. 3treetman, "TineDependence of the Current at High Electric Fields in AlI3a As!GaAsHeterojunction Layers," Electron. Lett. 17, 93-94 (1931 ).
1.7 L. C. Witkowski, T. J. Drummond, S. A. Barnett, H. T4orko , A. Y. Cho andJ. 3. Greene, "High Mobility GaAs/AlI As Singie Priod Modulation
Doped Heterojunctions," Electron. LettO7, 66-127 71981).
I 1.9 T. J. Drummond, H. Morko,, K. Hess and A. Y. Cho, "Exnerimental andTheoretical Electron 1obility of Modulation Doped Al 'a, As/naiAsHetarostructures 3rown by Y1olercular Beam "Epitaxy," .J. Apppl.- iys. 52I (Se~tember 1981, to appear).
1.9 D. S. Day, J. D. Oberstar, T. J. Drummond, H. Aorkoq, A. Y. Cho, and3. 13. Streetman, "Electron Traps CIreated by High TemperatuIrs Annealing i-n
BE n-GaAs," J. Electronic Materials 19, 445-453 (M4ay 191).
1.10 T. J. Drummond, ;J. Kopp, HI. .4orko-,, K. Hess, A. Y. Cho and 3. .Streetman, "Effect of Background Doping on the Mobility of (Al,Gas)s'13AsHeterojunctions," J. ADpl. Phys. 52 (Sept. 1991, to apper).
!
xi
P1J3IATItISI
1.11 T. J. Drummrond, 'A. l4orkog and A. Y. Cho. "'43E ;rowth of (l s~~He-te-rostruoetur:es," J. Crystal 3,rowt+h (,to aDvesr).
1.12 T. J. Drummoni, Wg. Kopp and H. '4orko(;, "Thrse ?griod 'Al,3,q~k/'aAsHleterostructure3s *,rith Extremiely High Mobilities," Electron. :Lett. 1',442-444 (1 931 )
1.13 A. Rockett, T. Jf. Drumimond. J. E. Greene an-i H. 4o-rkoq, "Sn Segregationin 'SE 1rown ?za~s," J. Xppl. Phys. (to appear).
1.14 T. IT. Drummnond, I4. Keever, W. Koop. H. "4orko;, K. Hess, A. Y. Tho and3. L 3treetmnan, "Fiall DepDendence of Iobility in A!G qa 3As aA3Haterojunctions at Very Low Fialds," Electron. Lett. (to ap -qr,-
1.15 J'- D. Dow, R. - . Allen, ). F. 3ankey, J. P. Buisson, andi H. P.'Ajalnarson, "Effects of the Einvirontnen-t on Poiit-Defect Energy Levels inSe'aizonductors," J. Vac. Sci. Techn. (to aDPvear).
1.153 . 7- kllen and J. D. Dow, "Theory of Frenkal 'lore Exccitons at SurfaCes."Phys. Rev. (to auoear) .
1.1-' R. E. Ailn and J. D. D)o-w, "Unifiel Theory of Point-Defect Electron-.',States, "ore Ex-2itons, and Intrinsic "Electronic States at Senizondu1ct3rSurfazes," J. Vac. Sci. Techn. (to appear).
,onference ?saver.
1 .13 T. J. Drummond, 2. '4. Stanchak, L. C. -iitkowski and H. :4orkoq, "Modula-
tio DoedGa~/Al 0a A3 Hetaro.jxnction Structures by '4BE for HighSoeei FETs," Pro . of GaAs 1" Sy-mosiun, 1?8O.
2. 33XI2O0' cJCT'R MATERIALS AND DE'lTCES
Journal krticles
2.1 L. 1. 'look, :4. A. Tashima. and . Stillmnan, "Variation of the:.1ickness and Commosition of LPE :nGaksP, tn~,and In? Layers 1ho-wnfrom a Finite Melt by the Step-looling Technique," J3. of Tlectroni:! Mat.
2.? *.. hic hijo, X. lass, and 3. T. Stillman, "SRaulation of High ?ieliTransport in lxa~s using a :4onte Carlo MAethod and ?seudopotntia.. 3-1-1Structures," Appi. 'Phys. Le t t 3 3, 39- 1 (J an. 15 i1 81
Ixii
3 PUBLICATIONS
3 Conference Papers
I 2.3 L. A. Zinkiewicz, T. R. Lepkowski, T. J. Roth, and G. . Stillman, "TheVapor Phase Epitaxial Growth of InP and InGaAs by the Hylride
3 3 93~74(-A -p3 Cl- 2) Technique," Proc. 1980 NATO InP Workshop, pp.I I 349-374 (11,30).
2.4 A. M. Tashima, L. W. Cook, and G. E. Stillman, "The Application of X-rayDiffraction Measurements in the Growth of LPE InGaAsP/InP," Proc. 19803I NATO InP Workshop, pp. 477-493 (1980).
2.5 L. A. Zinkiewicz, T. J. Roth, B. J. Skromme, and G. E. Stillman, "TheVapor Phase - Growth of Inp and Inby the Hyre(In- a-As-Asi-Cl-Hj) Technique," Proc. Int. Symp. of GaAs and Related
Compounds, Vi-nna (1980, to appear)
2.6 L. W. Cook, N. Tabatabaie, 4. A. Tashima, T. H. Windhorn, G. E. Bul-nan,and G. - . Stillman, "Low Leakage LPT-;.3r o wn InGAs/ThP kvalanchePhotodiodes," Proc. Int. Symp. on GaAs and Related Compounds, Vienna(1980), Inst. Phys. Conf. Ser. No. 56: Chap. 6 pp. 361-370 (1981).
3. QUANTUM ELECTRONICS1j B ooks
3.1 Joseph T. Verieyen, Laser Electronics, Englewood Cliffs, J: Prentice
Hall, Inc. (1981).
I Journal Articles
3.2 T. A. DeTemple, S. J. Petuchowski, and H. K. Chun-, "Semiclassical0).Treatment of :ultiphoton Interactions: AC Stark Shifts," Phys. Rev. A
2,2637(18)
3.3 T. A. DeTemple and q. K. Clhunrg, "Semiclassical Treatment of AultiphotonInteractions: Signal Flow 'Zraph Approach," Phys. Rev. A 22, 2647 (1930).
3.4 T. A. DeTemple, L. A. Bahler, and J. Osmundsen, "Semiclassical Theory ofResonant Three-Wave Parametric Interactions: Second 4s oni1Generation," Phys. Rev. A (Aug. 1981, to appear).
!3
xiiiPU3LI2ATIONS
3.5 1. L. Dlabal, S. 3. 4utchison, J. G. "Len and J. T. Verdeyen, "Multiline(430-496 nm) Discharge-Pumped Iodine "onofluoride Laser," Appl. Phys.
Lett. 37, 373 (1980).
3.6 'A. L. Dlabal and J. G. Eden, "'ain and Transient Absorption Profiles for
the Iodine Monofluoriie 490 mm and Iodine Monochlorile 430 nm Bands underDischarge Excitation," Appl. Phys. Lett. 33(7), 433 (1991).
3.7 *4. L. Dlabal, S. B. qutchison, J. G. Eden and J. T. Verdeyen, "Iodine
'4onofluorile 140-kW Laser: Small Signal Gain and Operating Parameters,"
3ot. Lett. 6, 70 (1981).
3.3 J. 3. Eden, A. L. DIabal and S. B. lutchison, "The Interhalogens IF and
I' as Visible Oscillators or Amplifiers," Invited Paper, IEEE J. 'uan.
Elc. Z-7, 1335 (1981).
3. 1 . J. lanno, K. t. 1reenberg and J. T. Verdeyen, "Comparison of the
Etihing and Plasma 2haracteristics of Discharges in CF. and NP-," J.Electrochemiial Soc. (to appear).
3. ,1 . J. lanno, J. T. Verdeyen, S. S. Than and B. I. Streetman, "Plasma
Annealing of Ion in'±lanted Semiconductors," Appl. Phys. Lett. (to
appear) .
3.11 P. A. Ailler, J. A. Halbleib, J. V. Poukey and J. T. Verdeyen, "Inverse
Ion Diode Experiment," J. Appl. Phys. 52(2), 593 (19?1).
3.12 L. T. Specht, "Signal Averager Interface between a 3ioma-ion 5533
Transient Recorder and an LS-11 Microcomputer," Rev. Sci. Inst. 51 17D-
(19-80).
3.!3 . '. Rosenberger and T. A. DeTample, "Far Infrared Superraliance in
Aethyl Fluorile," Phys. Rev. A (Aug. 1931, to appear).
Conference Paners
3.14 R. W. andreatta and J. G. Eden, "Visible Emission Spectrum of )pti-ally
Excited Tetragonal eO, " Presented at the Optical Society of kmeri'aAnn. Meeting, Chicago, tL 'oct. 1933).
3.15 A. L. Olabal, 3. B. "ut-hison and J. 3. Eden, "11 and IF E StateRaliatiie Lifetimes and uenching Rate Constants," 3rd Ann. aseous
Electronics Conf., Norman, )K (oct. 1980).
3.16 . L. blabal, S. B. utchison, J. 3. Eden and J. T. Verdeyen, "!:
Blue-'reen Lasers," 33rd Ann. 2aseous Electroni-s Conf., Norman, )K ():t.1930).
xiv
PUBLICATIONSxi
D 3.17 A. L. Dlabal, S. B. 'utchison, J. G. Eden and I. T. Verdeyen, "70-kW I?Laser," Optical Soc. of America Ann. Meeting, Chicago, IL (Oct. 1930).
3.13 A. 4. McCown and J. G. Eden, "'an at 602 nm on the B - X 3and of 7nI,"CLEO -31, Washington, D. C. (June 1981).
3.19 J. T. Verdeyen, J. 3. Eden, K. Killeen, J. Epler and D. Johnson, "LowTemperature Operation of an S3 Laser using a Radio-Frequency Simmer
Discharge," CLEO 31, Washington, D. C. (June 1981).I4. S-ICONDUCTOR PHYSICS
I Books
4.1 D. X. Ferry, K. Hess and P. Vogl, "Physics and Modeling of SubmicronInslated-3ate Field-Effect Transistors," II in VLSI ElectronicsMicrostructure Science, Vol. 2, edited by N. Einspruch, Academic ?ress(I1913 ).
I 4.2 K. J. Soda, R. Y. DeJule, and B. 3. Street-nan, "C-V and CapacitanceTransient Analysis of Self-Implanted Amorphous Si Layers Regrown by SIeptLine Electron Beam Annealing," pp. 353-360 in Laser and Electron BeamSolid Interactions and Materials Processing, J. F. Gibbons, L. D. Hess,and T. I. Sigman, eds., New York: Elsevier North Holland (1931).
1.3 A. Bhattacharyya, B. '. Streetman, and L. A. Reinisch, "Pulsed LaserI Annealing of BF; Implanted Amorphized Silicon," pp. 526-534 inSemiconductor 3ilicon 1991, R. R. Huff, R. J. Kriegler, and Y. Takeishi,eds. Electrochem. Soc. 191 ).
Journal Articles
4.4 'K. K. Aon and K. Hess, "Resonance Impact Ionization in Sunerlattices,"Solid State Electronics (to appear).
4.5 74. S. Burroughs, K. -ess, N. Holonyak, Jr., W. D. Lailig and -. '.Stillman, "Alloy Clustering and its Effect on Iiipact ionization inTernary III-V Compounds," Solii State Electronics (to appear).
4.6 q. 3hichijo, K. Hess and 3. 2. Stillman, "Simulation of igh-FiellTransport in 3aAs using a Monte Carlo Method and PseUdopotential BaniStructures," Appl. Phys. Lett. 33, 39-91 (1931).
I . 'A. Shichijo and K. 'Hess, Band-Structure-Dependent Transport and ImpactIonization in 'aAs," Phys. Rev. B 23, 4197-4207 (1931).
xv
?PJBLIJATI0NS
4.3 K. ess, "Ballistic Electron Transport in Semiconductors," IEEE Trans on
Electron Devices, (Aug. 1931, to appear).
4.9 H. 3hichijo, K. Hess, and B. 3. Streetman, "Real-Space Electron Transferby Thermioni : Emission in GaAs-Al 3a. -As Heterostructures - AnalyticalModel for Large Layer 4iths," 3oli- State Electronizs 23, 317-32(August, 1980).
4.1) T. Yu, K. J. Soda, and B. G. Streetman, "Annealing of Nitrogen-ImplantedfJaAs( P by a Swept Line Electron Beam," J. Appl. Phys. 51, 4399-4404(Au 790)
4.11 T. J. 3lisson, J. R. Hauser, A. A. Little john, K. '-ess, 3. 3. 3treetman,and H. 3hichijo, "Monte Carlo Simulations of Real-3Dace Electron Transfer
in 3aAs-klaAs Heterostructures," J. Appl. Phys. 51 . 5445-5449 (%ctober,1 9 3).
4.2 ). . Sankey, H. J. ijalmarson, J. D. Dow, ). J. Wolford, and 3. 3.Streetman, "Substitutional Defect Pairs in CaAs1 p,, ?hys. Rev. Lett.15, 1655-1559 (Nov. 1980).
1.13 A. Bhattacharyya and B. 3. Street-man, "Theoretical ConsiierationsRegarding Pulsed CO Laser Annealing of Silicon," Solii State
Communications 36, 671-:675 (ov. 1930).
1.1.4 D. J. Wolforl, B. 3. Streetman, and J. Thompson, "Zero-Thonon Structureof N Tran States in GaAs 1- P ys, Prceedng o t KInternational Confe rence on the hysics of Semiconductors, Kyoto 1980, J.Phys. Soc. Japan t9, Suppl. A, 223-22' (1990).
4.15 3. 3. han, 3. T. Aarcyk, and B. G. 3treetman, "Electrical Properties andPhotoluminescence Studies of ]e-lmplanted GaAs," J. Electronic Materials10, 213-233 (Jan. 1981)
4.16 B. 3. Streetman, "Ion Implantation in Compound Semiconduztor Research,'IEEE Transactions on Nuclear Science :1-28, 1712-1746 "April 1?91).
1.17 D. 3. Day, J. D. )berstar, T. J. Drummond, H. lorkoq A. Y. Tho, and3. '3. 3treetman, "Electron Traps Created by High Temperature Annealing inMBE n-haAs," J. Electronic Aaterials 1). 445-453 (May 1991).
4.13 A. Bhattacharyya, B. 3. 3treet-nan. and K. less, "Theoretical andExperimental Investigation of the Dynamics of Pilsed Laser Annealing ofSilicon," J. Anpl. Phys. (May 1991).
1.19 J. D. )berstar, 3. 3. 3treetman, J. T. Baker, and P. Williams. "iron anviThromiu'u Redistribution in Semi-Inaulating InP," 3. Electrochem.123, 1 3t14-t917 (Aug. 1331 ).
4.20 J. D. )berstar, 3. 3. Streetnan, and 3. A. Sammann, "Cracking of 3i 'Layers on Annealed InP," Thin Solil Films 31, 347-156 '1 I ).
I'5 PUBLICATIONS xvi
3 4.21 A. Bhattacharyya and B. G. Streetman, "Dynamics of Pulsed CO LaserAnnealing of Silicon," J. Phys. D: Applied Physics 14, L67-L72 (191).
4.22 A. Bhattacharyya, V. Iyer, B. G. Streetman, J. E. Baker, and P. Williams,"Electrical Activation and Impurity Redistribution during Pulsed LaserAnnealing of BF"' Implanted Amorphized Silicon," IEEE Trans. onComponents, Hybriqs, and Manuf acturing Technology (December 1981, toappear).
4.23 J. D. Oberstar, B. G. Streetman, J. E. Baker, P. Williams, R. L. Henry,and E. 4. Swiggard, "Boron Contamination and Precipitation during theGrowth of InP," J. Crystal Growth (to appear).
4.24 J. D. Oberstar and B. G. Streetman, "Photoluminescence Studies of SurfaceDamage States in InP," Surface Science (to appear).
I Conference Paver
4.25 K. Hess, "Lateral Transport in Superlattices," Proc. of the 3rd Intl.Conf. on Hot Electrons in Semiconductors (1981, to appear).
I1 Technical Reports
I 4.26 H. 3hichijo, "Theoretical Studies of High Field Transport in III-VSemiconductors," CSL Report R-893 (Aug. 1980).
1 5. THIN FIL14 PIYSIICS
I I Books
5.1 J. 9. Greene, "Thin Film Crystal Growth by Sputtering," in Handbook ofSemiconductors, Vol. III, ed. by S. Keller, North Holland Publishing Co.,Amsterdam (1980), p. 499.
Journal Articles
5.2 S. A. Barnett, '. Bajor and J. E. reene, "Growth of High ualityEpitaxial GaAs Films by Sputter Deposition," Appl. Phys. Letters 37, 735
. K
xviiPUJBLIC ATI JNS
5.5 K. C. 'adien, J. L. Z1ilko, A. H. Eltoukhy and J. E. Greene, "growdth ofSingle Crystal A etastable In~b x i n (ab 1x;
-~1-xi an (IaSb> G,19 , ilms,' J. Vac.3oi. Tech 17 , 4 (15 )
5.4 A. !I. 'iltoukhy, 3. 1. Natarajan, J. E. Greene -and T. L. Barr, "Mechanismsof the Reactive Sputtering of In - Part 3: A General PhenomenologicalModel for Reactive Sputtering," Thin Solid Films 69, 229 (19of).
5.5 A. H. Eltoukhy and J. E. Greene, "Diffusion Enhanced by Low Energy IonBombardment during Sputter Etching and Deposition," J. Appl. Phys. 51444 (I1930).
5.6 J. 3. Greene and A. H. Eltoukhy, "Semiconductor Crystal Growth by SputterDeposition: A Review," Surface and Interface Science 31., (1931).
5.7 J. E. Greene, Crystal Growth by Soutter Deposition: Application toSemiconductors," Critical Reviews in Solid State and Material Sciences
(to appear).
5.3 J. E. Greene, K. C. Jalien, D. Lubben, 3. A. Hawkins, G. R. Erickson, andJ. R. Clarke., "Eoitaxial Ge/GaAs Heterostructures by Scanned 'W LaserAnnealing of a-Ge Layers on GaAs," Appl. Phys. Letters (to appear).
5.9 R. 2. Klin-er and J. E. Greene, "Reactive Ion Etching of GaAs in CCL, 9 ,Appl. Phys. Letters 33, 620 (1931 ).
5.12 B. R. Natarajan, A. H. Eltoukhy, J. E. Greene and T. L. Barr, "Mechanismsof the Reactive Sputtering of In - Part 1: Growth of inN in :Hixed Ar +N2 Discharges," Thin Solid Films 59, 201 (1980).
5.11 B. R. latarajan, A. H. Eltoukhy, J. . 'reene and T. L. Barr, "M4echanismsof the Reactive Sputtering of In - Part 2: Growth of Indium Oxynitrile
in Mixed N2 + J Discharges," Thin Solid Films '9, 229 (1930).
5.12 L. C. Witkowski, T. J. Drummond, S. A. Barnett, H. '4orko;, A. Y. Tho,and J. 3. Greene, "Properties of High Aobility GaAs/Ga Al As SinglePeriod Modulation Doped Heterojunctions," Electron. Lett. _ 1I7 (1931).
Conference Papers
5.13 R. W. Andreatta, J. 3. Eden, D. Lubben, and 3. ". Greene, "Deposition ofThin Amorphous Si and Ge Films by the Ultraviolit Photodissociation of
Silane or Germane," Workshop on the Interaction of Laser Radiation °-ithSurfaces for Application to Electronics, AIT, Boston (May 1991 ).
3.11 S. A. Barnett and J. E. Greene, "Growth of High Quality Enitaxial GaAsFilms by Sputter Deposition," 27th Ann. Amer. Vacuum Soc. Mtg., Detroit,
N 1 (ov. 1930).
xviii
PUBLICATIONS
3 5.15 A. H. Eltoukhy, B. R. Natarajan, T. L. Barr, and J. E. Greene, "A leneralPhenomenological Model for Reactive Sputtering," Intl. Vacuum Soc.Congress, Cannes, France (Sept. 1980).
1 5.16 J. E. Greene, Invited Paper: "A Review of Ion Bombardment Effects Duringthe Growth of Thin Film Semiconductors by Sputter Deposition," AIAA 13th5 Fluid and Plasma Dynamics Conf., Snomass, CO (July 1980).
5.17 X. C. Grender, P. N. U. al, L. C. Burton, L. Rivaud, and J. E. greene,Invited Paper: "Composition Analysis of Heterojunctions by Means ofElectron and Optical Spectroscopies," SPIE 24th Ann. Symp., San Diego, CA(Aug. 1950).
IU 6. MICROWAVE ACOUSTICS
I Journal Articles
6.1 R. Adler, D. Janes, B. J. Hunsinger, and S. Datta, "AcoustoelectricMeasurement of Low Carrier Mobilities in Highly Resistive Films," Appl.Phys. Lett 33(2), 102-103 (Jan. 1981).
I 6.2 S. Datta and B. J. Munsinger, "An Analytical Theory for the Scattering ofSurface Acoustic Waves by a Single Electrode in a Periodic Array on a
i Piezoelectric Substrate," J. Appl. Phys. 51(9), 4317-4923 (Sept. 1990).
6.3 S. Datta, A. Hoskins, R. Adler, and B. Hunsinger, "SAW-LAW Mu'tistrioMode Converter," IEEE Trans. on Sonics and Ultrasonics (to appear).
6.4 S. Datta and 3. J. Hunsinger, "An Analysis of Energy Storage Effects onSAW Propagation in Periodic Arrays," IEEE Trans. on Sonics andI Ultrasonics SU-27(6), 333-341 (Nov. 1980).
6.5 A. L. Lentine, S. Datta, and B. J. Hunsinger, "Analysis of NonperiodicSAW Transducers using a Circuit Model," IEEE Trans. on Sonics andUltrasonics (to appear).
6.6 C. M. Panasik and B. J. Hunsinger, "Scattering Matrix Analysis of SurfaceAcoustic Wave Reflectors and Transducers," IEEE Trans. on Sonics andUltrasonics (to appear).
!I
'I
xixPUBLICATIONS
7. SURFACE STTDIES
Books
7.1 3. Ehrlich and K. Stolt, "Surface Diffusion," Ann. Rev. Phys. Chem. 31,603-637 (1930).
Journal Articles
7.2 B. H. Chin and 3. Ehrlich, "Formation of Silicon Nitriie Structures byDirect Electron Beam Writing," Appl. Phys. Lett. 38, 253-255 (1981).
7.3 '. Ehrlich, "Wandering Surface Atoms and the Field Ion :Aicroscope,"Physics Today 3 , 44-53 (June 1981 ).
7.4 D. A. Reel and 3. Ehrlich, "Surface Diffusion, Atomic Jump Rates andThermodynamis," Surface Sci. 102, 533-609 (1981).
7.5 D. A. Reed and 1. Ehrlich, "Surface Diffusivity and the Time Correlationof oncentration Fluctuations," Surface Sci. 105, 603-623 (1991).
Conference Paners
7.6 r Ehrlich and "K. Stolt, " uantitative Studies of Tndivilual Atoms and
'lusters on Solids," 35th Ann. Proc. Microscopy Soc. Amer. (3. 1. Bailey,ed.), pp. 2-5 (1981).
7.7 H.-1. Fink and 3. Ehrlich, "An Atomic View of Layer Growth on 4(111),"
Abstract, Forty-first Ann. Conf. on Physical Electronics, Xontana St. U.(June 1981 ).
7.8 H.-W. Fink and G. Ehrlich, "Direct Obser-iations of Overlayer growth onW(110)," Abstract, 28th Intl. Field Emission Symp., Lewis and ClarkCoilege, Portland, Oregon (July 1931).
x-x
5 PUBLICATIONS xx
1 3. ELECTROMAGNETIO COMMUNICATION, RADIATION AND SCATTERING
I Journal Articles
3.1 S. Bhooshan and R. Mittra, "On the Design of Transitions between a Metaland Inverted StriD Dielectric Waveguide for Millimeter Wave," IEEE Trans.on Microwave Theory and Tech. MTT-29(3), 263-265 (March 1981).
3.2 J. A. G. Malherbe, T. 3. Trinh and R. Mittra, "Transition from Metal toDielectric 'Waveguile," Microwave J. 23(11), 71-74 (Nov. 1980).
I . S. Safavi-laini and R. Mittra, "High-Frequency Radiation from anElectromagnetic Source Lccated on a Finite Cylinder - A Spectral DomainApproach," IEEE Trans. on Antennas and Propagation AP-2'(6), 304-810
(Nov. 19_0).
Conference Paoers
I 3.4 A. Ciarkoweki and R. Mittra, "Plane-4ave Diffraction by a 'edge - ASpectral Domain Approach," Intl. IEEE/AP-S and URSI Symposium, LosAngeles (June 1991).
I 3.5 R. Kastner and R. Mittra, "A Spectral-Iteration Technique for AnalyzingScattering from Arbitrary Bodies," Intl. IEEE/AP-S and URSI Symp., LosAngeles (June 1981 ).
Technical Reports
I 3.6 R. Mittra, C. H. Tsao and R. Kastner, "Spectral Domain Methods forAnalyzing Frequency Selective Surfaces (FSS)," Interim Tech. Rept. forNSF and ONR, IM Lab Report 31-7 (July 1981).
3.7 R. Mittra and W. L. Ko, "Introduction to the 2eometrical Theory ofDiffraction and the Spectral Domain Approach to High-FrequencyDiffraction," Interim Tech. Rept. for ONR, H4 Lab Report 31-8 (July
I
xxi
PUBLICATINS
9. PLAS14A PHYSICS
Journal Article
9.1 3. Ault, A. Raether, and S. Sobhanian, "Response of the Electric Fiell ina Positive Column in Helium to a Current Perturbation," Journal of Appl.Phys. (Sept. 1931 , to appear).
ID. RAREFIED GAS DYNAMICS A.ND COMPUTATIONAL IAS DYNAMICS
Journal Articles
1).1 S. A. Yen and F. I. Tcheremissine, "Monte Carlo Solution of theJonlinear 3oltzmann Equation," AAA Progress in Astronauti-2s endAeronautiss (to appear).
ID.2 S. H. Yen, "Numerical Solution of the oltzmann and Krook Equations fora Condensation Problem," AIXA Progress in Astronautics and Aeronautics(to appear)
1C.3 S. A. Yen and T. Ytrehus, "Treatment of the Nonequilibrium Vapor "otionNear an Evaporating Interphase Boundary," Chemicl EngineeringCommunications (to appear).
Conference Paner
1).4 S. :4. Yen and D. 3. Hall, "Implementation of Open Boundary Conditionsfor Nonlinear Free Surface *ave Problems," Proc. Third Intl. Conf. onNumerical Ship Hydrodynamics, Paris, France, p. 111-2-i (June IO3I1).
IxxiiPUBLICATIONS
3 11. FAULT-TOLERANT SYSTEMS AND COMPUTER ARCHITECTURE
I Journal Articles
I 11.1 J. A. Abraham and D. D. lajski, "Design of Testable Structures Definedby Simple Loops," IEEE Trans. on Computers (to appear).
I 11.2 R. L. Budzinski and E. S. Davidson, "A Comparison of Dynamic and StaticVirtual Memory Allocation Algorithms," IEEE Trans. on SoftwareEngineering SE-7, 122-131 (Jan. 1981).
11.3 R. L. Budzinski, E. S. Davidson, H. S. Stone, and W. Mayeda, "OMIN: AnCOtimal Algorithm for Dynamic Allocation in a Virtual Memory Computer,"IEEE Trans. on Software Engineering SE-7, 113-121 (Jan. 1981).
11.4 D. 'J. Hammerstrom and E. S. Davidson, "The Use of Second Order Memoriesto Reduce Memory Addressing Overhead," IEEE Trans. on Computers (to
appear).
11.5 B. Kumar and E. S. Davidson, "Computer System Design Using aHierarchical Approach to Performance Evaluation," Comm. of the ACM 23,511-521 (Sept. 1980).
11.6 J. H. Patel, "Performance of Processor-Memory Interconnections forMultiprocessors," IEEE Trans. on Computers 3-30 (Oct. 1981, to appear).
I Conference Papers
11.7 J. A. Abraham and D. D. 'ajski, "Easily Testable, High Speed Realizationof Register-Transfer-Level Operations," Proc. Tenth Int'l Symp. onFault-Tolerant Computing, Kyoto, Japan, pp. 339-344 (Oct. 1980).
11.3 J. P.. Abraham and K. P. Parker, "Practical Microprocessor Testing: Openand Closed Loop Approaches," (Invited Paper) COMPCON Spring 1981 Digest,San Francisco, pp. 308-311 (Feb. 1981).
11.9 J. H. Patel, "A Performance Model for Multiprocessors with Private CacheMemories," Proc. 1981 Int'l Zonf. on Parallel Processing (Aug. 1981, toI appear).
11.10 A. Pleszkun, E. S. Davidson, and 3. R. Rau, "An Address Prediction£ Mechanism for Reducing Processor-Memory Bandwiith," Proc. 1981 !EEEComputer Soc. Workshop on Computer Architecture for Pattern Analysis andimage Database Management (Nov. 1991, to appear).
xx iii
T e c h n i c a l I 7 e n o r t s " P r o m n eP U B L I A T I ) I S
11.1D -. Ya,"-frac Models for X-utiorocessor 'Dncucer Systems,"_"I Tech. iepo rt R-394 j)c. 1?30.
11.12 '1-3. Y eh, "Sharel iacha 9rganizations f1or Auli -- 3tream nou tarSystems," CSL Tech. Report R-904 (Jan. 1931).K
12. APPLED MPUTAID THEORY
Journal Articles
12.1 B. 3. 3aker, D. J7. 3rown, and H. PD. )(atseff, "A 5'4 Al-gorithm forTwo-Dimensional Bin Paclcing," J. Alg-oritlhms 'to appear).
1 2 J. :,. Bentley and D. J. Brown, "A 73eneral Class of Resourie Traleoffs."J. of 3omputear and System Scecs(oaopear)K
12.3_ J. L. Bentley, 'I. ? aust, ani ?. ?. ?rsoarita, "AenroxirnationAklgorithms for Convex Hulls,"' omm. of the AC (tIo annear).
12.1 D. J. Brown. "An 7jprove d BL Lower Boun!." :nfmation ProcessingLetters, II i', 37-39 (Aug. 1990).
12.5 R. 3. Johnson. Jr., "The Jomplexity of a 7',S! Adder." :nf or-mation?rorc-essing Letters 1 1 '2' :1-9 ')ct. 1930').
12.5 'J. Lioski, Jr. and F.?. Preparata, "Fiding- the Contour of a UJnion of1Iso-Oriented Rectangl:?s," J. Algorithms 1,2-5-24519')
12.7 ',. Tipski, Jr. and F. P. Preparata, "Segments, Rectangles, Jontours,"J. Algorithms 2M1', 63-76 (MXa rch 1 931 )
2 3 f. Lioski , Jr. and F. P. Preparata, "Efficienit Algorithms for Fin~injAaximnum Mat:chings in :onvexc Bipartite 3raphs ani Related Problems," AzctaInformatic a (to appear).
12.2 A. 'I. Louti, "A Note on the ?e-bbl ;,ame," :nformation Processing Lett ers11I(I), 24-20- 'Aug. 19390) .
12.19 .J. Nievergelt and F. P. Preparqti, "?lane-SWe Algorithms :orIntersecting 3eovaqtriz! Figure s," Comm. of 'he AM(to appear').-
12.11 F. . Prenarsata, "A New Annroa:ch to ?lanar Point Location." Siam ),,oIcomnputing .to appear) .
xxiv
3 PUBLICATIONS
I 12.12 F. P. Preparata and K. J. Supowit, "Testing a Simple Polygon forMonotonicity," Information Processing Letters (to appear).
I 12.13 F. P. Preparata and J. Vuillemin, "Area-Time Optimal VLSI Networks forMultiplying Matrices," Information Processing Letters 11 (2), 77-73 (Oct.1980).
I 12.14 F. P. Preparata and J. Vuillemin, "The Cube-Connected Cycles: AVersatile Network for Parallel Computation," Comm. of the ACM 2<5),300-309 (May 1981).
Conference PapersI12.15 J. L. Bentley and D. J. Brown, "A leneral Class of Resource Tradeoffs,"
* 1 Proc. of the 21st Annual IEEE Symp. on Foundations of Computer Science,
Syracuse, .Y, pp. 217-229 (Oct. 1980).
12.16 D. T. Lee and F. P. Preparata, "Euclidean Shortest Path in the Presence
i of Parallel Barriers," Int. Conf. on ADplications of 3raph Theory toComputer Science, Linz, Austria (June 1981).
12.17 A. C . Loui, "Simulations among ultidimensional Turing Machines,"Proc. of the 22nd Annual IEEE Symp. on Foundations of Computer Science(Oct. 1981, to appear).
I 12.13 D. B. Muller and P. E. Schupp, "Pushdown Automata, 3raphs, Ends,Second-order Logic and Reachability Problems," Proc. of the 13th Annual
ACM Symp. on Theory of Computing, Milwaukee, WI, pp. 46-54 (May 1981).
12.19 F. P. Preparata and J. Vuillemin, "Optimal Integrated-CircuitImplementation of Triangular Matrix Inversion," 1930 Int. Conf. onParallel Processing, Boyne, 41, pp. 211-216 (Aug. 1980).
12.20 F. P. Preparata and J. Vuillemin, "Area-Time Optimal VLSI Networks forComputing Integer Multiplication and Discrete Fourier Transform,"I.O.A.L.P., Haifa, Israel (July 1981, to be presented).
I Technical Reports
I 12.21 D. J. Brown, B. S. Baker, and H. P. Katseff, "Lower Bounds for On-LineTwo-Dimensional Packing Algorithms," Technical Report ACT-25,Coordinated Science Laboratory, University of Illinois, Urbana, IL (July3 1980).
12.22 A. C. Loui, "Minimizing Access Pointers into Trees and Arrays,"Technical Report ACT-27, Coordinated Science Laboratory, University ofI Illinois, Urbana, IL (June 1981).
I -: : ! -I I I I
PJBLIZATIDNS
13. ADVANCED AUTOMATI)N
3ooks
13.1 N. Ahuja ani B. Schachter, Pattern Models, Wiley (early 1332, toappear) .
13.2 N. Ahuja, "Mosaic Models for Textures," in A. Rosenfeld (ed. Ia
Models, Acalemic Press (1931).
13.3 N. Ahuja and A. Rosenfeld, "Image Models," in P. R. Krishnaiah (ed.,Handbook of Statistics, North Holland (1981).
15.1 N. Ahuja and A. Rosenfell, "Fitting Mosaic Models to Textures," in R. A.Haralick (ad.), Image Texture Analysis, Plenum Press (1 3 1).
15.5 3. DeJong, "An 3ver-tiew of the FRTJXP System," in W. Lehnert and M.Ringle (eds.), Strategies for Natural Language Understanding, Hillsdale,NJ: Lawrence Earlbaum Associates (Dec. 1931, to appear).
15.6 D. L. Waltz, "The State-of-the-Art in N3tural Language Understanding,"in M. Ringle and W. Lehnert (eas.), Strategies for Natural LanguageProcessing, Hillsdale, NJ: Erlbaum Associates (Dec. 13S1 , to appear ,.
13.7 D. L. Waltz, "'enerating and Understanding Scene Descriptions," inJoshi, Sag, and Webber (eds.), Slements of Discourse Understanding,Cambridge Univ. Press, 266-282 (1991).
Journal Articles
13.3 N. khu.ja, R. M. Haralick and A. Rosenfeld, "Neighbor ray Levels asFeatures in Pixel Classification," Pattern Recognition 12, 291- 260
13.3 N. Ahuja, T. Dubitzki and A. Rosenfeld, "Some Experiments with losiLMolels for Images," IEEE Trans. 3yst., Aan, Cybern. 1), 4-74 '1 130).
13.1D N. Ahuja and A. Rosenfeld, "Mosaic Models for Textures," 1E Trans.Pattern Analysis and Machine Intelligence, 1-11 (Jan. 131).
15.11 N. Ahuja, "Mosaic Models for Images, 1: .eometric Properties ofComponents in Cell Structure Mosaics," Information Sciences 2 . 6 - l4
(March 13 31).
13.12 N. Ahuja, "Mosaic Models for Images, 2: leometric Properties of.omonents in 'overage Mosaics," Information Sciences 23, 159-200 (Apri1 ?31.
xxvi
3 PUBLICATIONS
13.13 N. Ahuja, "Mosaic Models for Images, 3: Spatial Correlation in
Mosaics," Information Sciences 24, 43-70 (June 1981).
13.14 N, Ahuja and B. Schachter, "Image Models," ACM Computing Surveys (toIappear).13.15 R. Chellappa, R. L. Kashyap and N. Ahuja, "Decision Rules for Choice of
Neighbors in Random Field Models of Images," Computer Graphics and ImageProcessing 15, 301-313 (April 1981).
* 13.16 R. T. Chien and C. Jacobus, "Intermediate-Level Vision -- 3uiliing- Vertex-String-Surfaces (V-S-S) Graphs," Computer Graphics and Image
Processing 15, 339-363 (1981 ).
13.17 C. J. Jacobus and R. T. Chien, "T-o New Edge Detectors," IEEE Trans.Pattern Analysis and Machine Intelligence (Sept. 1981, to appear).
3 13.13 3. H. Rouse and W. 3. Rouse, "Computer-Based Manuals for ProceduralInformation," IEEE Trans. on Systems, Man, and Cybernetics SMC-10 $3),(Aug. 1980).
13.19 W. B. Rouse, "Human-Computer Interaction in the Control of Dynamic
Systems," Computing Surveys 13( ), 71-99 (March 1981 ).
1 13.20 B. Schachter and N. Ahuja, "A History of Visual Flight Simulation,"Computer Graphics World 3. 16-31 (1980).
0onference Paners
13.21 N. Ahuja, R. r. Chien, R. Yan and 3. Briiwell, "Interference Detectionand Collision Avoidance Among Three Dimensional Objects," Proc. 1stNatl. Conf. on Artificial Intelligence, Stanford, pp. 44-43 (Aug. 1981 ).
13.22 N. Ahuj a, "Dot Pattern Processing using Voronoi Polygons asNeighborhoods," Proc. 5th Intl. Joint Conf. on Pattern Recognition.Miami Beach, pp. 1122-1127 'Dec. 1980).
13.23 N. Ahuja, "Approaches to Recursive Image Decomposition," Proc. PatternRecognition and Image Processing Conf., Dallas (Aug. 1981, to appear).
13.24 3. DeJong, "Generalizations Based on Explanation," Proc. of the 7thiSntl. Joint Conf. on Artificial Intelligence, Vancouver BC (Aug. 1991,to appear).
13.25 G. Johannsen and W. B. Rouse, "A Study of the Planning Process ofAircraft Pilots in Emergency and Abnormal Situations," Proc. 1980 Intl.Conf. on Cybernetics and Society, Boston (9ctober 1990).I
I
xxviiPUBLIJATIDNS
1 5.26 3. Johannsen and W. 3. .ousa, "Problem Solving Behavior of AircraftPilots in Abnormal and Emergency Situations," Proc. 1st Ann. EurooeanConf. on Decision Making and Manual Control, Delft (May 1981).
13.27 3. H. Rouse, W. B. Rouse, and J. M. Hammer, "Design and an Approach toEvaluation of an Dnboard Computer-3ased Information System forAircraft," Proc. 1930 Intl. Conf. on Cybernetics and Society, Boston(ct. 1980).
13.29 3. H-. Rouse and '. 3. Rouse, "Evaluation of an Onboard Computer-BasedInfir-ation System for Aircraft," Proc. 1981 Intl. Conf. on Cyberneticsand Society, Atlanta (3ct. 1981).
Ii.29 D. L. 'altz, "Toward a Detailed Model of Processing for LanguageDescribing the Physical World," ?roc. 7th Intl. Joint Conf. onArtificial Intelligence, Vancouver, BC (Aug. 1981, to appear).
Ienorts and Theses
13.30 3. Brooks, "JQL - A Relational Data Base Query Language," M.S. Thesis,Dept. of "orputer Science, Univ. of Illinois (Feb. 1981).
13.31 9. D. Dankal, "Browsing in Data Bases," CSL Tech. Rnt. T-192, Ph.D.Thesis (ov. 1980).
13.35 J. X. Hammer, "The Human as a Constrained Optimal Editor," Ph.D. Thesis(May 1931 ).
13.34 R. L. Henneman, "Aeasures of Human Performance in Fault DiagnosisTasks," M.S.I.E. Thesis (July 1931).
13.35 ). R. Morehead, "Models of Human Behavior in Online Searching ofBibliographic Data Bases," M.3.I."E. Thesis (lily 1981 ).
13.36 H. R. Tennant, "Evaluation of Natural Language Processors," CSL Tech.Rpt. T-193 (Dec. 1980).
el.
Ixxviii
PUBLICATIONS
I 14. INFORMATION RETRIEVAL RESEARCH
U Books
I 14.1 S. 3. Larson and -. E. dilliams, "Computer Assisted Legal Research," inAnnual Review of Information Science and Technology, Vol. 15, A. E.Williams, ed. White Plains, NY: Knowledge Industry Publications, pp.3 251-286 (1980).
14.2 M. E. Williams, "Criteria for Evaluation and Selection of Data Bases andData Base Services," Chapter 65 in E. B. Jackson, Special
Librarianshi : A New Reader, Metuchen, NJ: Scarecrow Press, pp.673-684 (1 9807.
14.3 X. E. Williams, "Databases, Computer Readable," in The ALA Yearbook: AReview of Library Events 1979, Vol. 5, Robert Wedgeworth, ed. "hicago:American Library Association, pp. 125-131 (1980).
I 14.4 A. E. Williams, ed. Annual Review of Information Science and Technology,
vol. 15. White Plains, NY: Knowledge Industry Publications, Inc., 413
1 pp. (1980).
14.5 A. E. Williams, "Databases, Computer Readable," in The ALA Yearbook: AReview of Library Events 1989. Chicago: American Library Assn., pp.117-1 2271 981).
14.6 A. I. Williams, ed. Annual Review of Information Science and Technology,vol. 16. White Plains: Knowledge Industry Pablications, Inc. (Sept.
1981 , to appear).
1 Journal Articles
14.7 A. E. Williams, Databases and Online Statistics for 1979," Bul. Amer.Soc. for Information Science 7(2), 27-29 (Dec. 1930).
1 .3 M. 3. Williams and L. Lannom, "Lack of Standardization of the Journal
Titla Data Element in Databases," J. Amer. Soc. for Information Science
32(3), 229-233 (May 1981).
14.9 A. S. Williams, "Relative Impact of Print and Database Products onDatabase Producer Expenses and Income - Trends for Database Produc-irOrganizations Based on a 13-Year Financial Analysis," Information
Processing and Management 17(3) (Oct. 1981, to appear).
1I .
. . . . . . . .. . ... . . . . . . . . ... .. .. ... . . . ... . -. J I'
xxix
PUBLICATIONS
Conference Papers
14.10 S. E. Preece and M. E. Williams, "Software for the Searcher'sWorkbench," Communicating Information: Proc. of the 13rd ASS Ann.At.., vol. 17. White Plains, NY: Knowledge Industry Publications,Inc., pp. 403-405 (Oct. 1980).
14.11 A. 5. Williams and S. E. Preece, "Elements of a Distributed TransparentInformation Retrieval System," Communicating Information: Proc. 43rdASIS Ann. :tg., vol. 17, pp. 401-402 (Oct. 1980).
14.12 4. E. Williams, "Thture Directions for Machine-Readable Data Bases andtheir Use," The Role of the Library in a Electronic Society - Proc. 16thAnn. Clini3 on Library Applications of Data Processing, Urbana, IL:University of Illinois, Graduate of Library and Information Science, pp.32-9?3 (1980.
14.13 A. E. Williams and T. 1. Hogan, compilers, Proc. 2nd Nat. )nline Mt3.Medford, NJ: Learned Information, Inc., 554 pp. (March 1981).
11.14 X. ". Williams, "Highlights of the Online Database Field," Proc. 2ndNatl. Online Mt,., Medford, NJ: Learned Information, Inc., pp. 1-5(March 1981 ).
! .15 A. 5. Williams and S. E. Preece, "A Mini-Transparent System Using anAlpha Airoprocessor," Proc. 2nd Natl. Online Atg. Medford, NJ:Learned Information, Inc., pp. 499-502 (March 1931 .
14.17 A. E. iilliams, "Research Toward a Transparent Information System,"Proc. of Symp.: Perspectives on Scientific and Technical Information,Oak Riige, TN: Oak Ridge National Laboratory, pp. 51-60 (Sept. 1981).
15. COMMUNICATIONS
:hapters in Books
15.1 R. J. 'AcEliece, "Some Combinatorial Xspects of Spread-SpectrumCommunication Systems," in New Concepts in Multi-User Thmmunications,J. X. Skwirzynski (ed.), NATO Advanced Study Institute--Series 5,3ijthoff and Noordhoff International Puablishers, Aiphen aan len Rijn.Netherlands, (in press).
15.2 A. 3. ?lursley, "Spread-Spectrum Multiple-Access 'ommuni ations," in
Multi-User Communication aystems, I. Longo (ed.), Springer-Ver!ag,Vienna and New York, pp. 159-1 99 \1 981 ).
xxx
5 PUBLICATIONS
15.3 A. B. Pursley, "Effects of Specular ultipath Fading on Spread-Spectrum
Communications," in New Concepts in Multi-User Communications, J. K.Si-cirzynski (editor), NATO Advanced Study Institute--Series E, Sijthoff
and Noordhoff International Publishers, Alphen aan den iijn,Netherlands, (in press).
5 Journal Xrticles
15.4 D. Alexandrou and H. V. Poor, "The Analysis and Design of DataQuantization Schemes for Stochastic-Signal Detection Systems," IEEE
Transactions on Communications COM-28, 983-991 (July 1980).
15.5 S. P. Au, A. q. Haddad, and f. V. Poor, "A State Estimation Algorithm
I for Linear Systems Driven Simultaneously by Wiener and PoissonProcesses," IEEE Transactions on Automatic Control AC-27 (June 1982, to
1 appear).
15.6 L. Baumert, R. J. McEliece, 3. Rodemich, and H. Rumsey, "A Probabilistic
Version of Sperner's Theorem," Ars Combinatoria 9, 91-100 (1990).
1 .7 T. Cover, R. J. XcEliece, and E. Posner, "Multiple Access ChannelCapacity is Attainable Without Synchronization," IEEE Transactions onj information Theory IT-27 (July 1981, to appear).
15.3 R. Cruz and B. Hajek, "A New Upper Bound to the Throughput of aMulti-Access Broadcast Channel," TEBE Transactions on information TheoryIT-23 (May 1982, to appear).
15.9 L. D. Davisson, R. J. AcEliece, M. B. Pursley, and A. S. Wallace,
"Efficient Universal Noiseless Source Codes," IEEE Transactions on
Information Theory IT-27, 269-279 (May 1981 ).
I 15.13 F. D. Garber and M. B. Pursley, "Optimal Phases of aximal-LengthSequences for Asynchronous Spread-Spectrum Multiplexing," Electroni:!sLetters 16, 756-757 (Sept. 1980).
15.11 F. D. Garber and M. B. Parsley, "Performance of 9ffeet uadriphaseSpread-Spectrum Multiple-Access Communications," IEEE Transactions onCommunicstions COM-29, 305-314 (March 1981).
1 15.12 3. Jajek, "Information Singularity and Recoverability of Ranion
Processes," IEEE Transactions on Information Theory IT-2? (May 1982, to
I appear).
15.13 3. Hajek, "Stochastic Equations of Hyperbolic Type and a No-Parameter3 Stratonovich Integral," Annals of Probability (to appear).
15.1t B. "ajek, "Birth-and-Degath Processes on the Integers with Phases aniGeneral Boundaries," Journal of Appliel Probability 1? (Sept. 1,982, to
3 appear)
I
xxxi Ixxxi PUBLICATIONS 3
15.15 B. Hajek and T. van Loon, "Dynamic Decentralized Estimation and Controlin a Multiaccess Broadcast Channel," IEEE Transactions on Automatic IControl AC-27 (June 1982, to appear).
15.16 B. Hajek and E. Wong, "Representation and Transformation of 3Two-Parameters Martingales under a Change of Measure," Zeitshrift furWahrscheinlichkeitstheorie 54, 313-330 (1980).
15.17 J. C. Kieffer, "Some Universal Noiseless Multiterminal Source CodingTheorems," Information and Control 46, 93-107 (1930).
15.13 R. J. XcEliece, "Practical Codes for Photon Communication," IEEETransactions on Information Theory IT-27 (July 1931, to appear).
15.19 R. J. :cEliece, B. Reznick, and J. Shearer, "A Turan Inequality Arising Iin Information Theory," SIAM Journal of Mathematical Analysis (toappear). I
15.20 R. j. AcEliece and D. V. Sarwate, "On Sharing Secrets and Reed-SolomonCodes," Communications of the ACM (to appear).
15.21 H. V. Poor, "Minimax Linear Smoothing for Capacities," Annals ofProbability (to appear).
15.22 H. V. Poor, "Ontimum Memoryless Tests Based on Dependent Data," Journal Iof Combinatorics, information, and System Sciences 6 (1931, to appear).
15.23 H. V. Poor, "Robust Decision Design using a Distance Criterion," IEEE ITransactions on Information Theory IT-26, 575-597 (Sept. 1930).
15.24 q. V. Poor, "The Rate-Distortion Function on Classes of SourcesDetermined by Spectral Capacities," IEEE Transactions on InformationTheory IT-23 (Jan. 1382, to appear).
15.25 1. V. Poor, "Signal Detection in the Presence of Weakly Dependent Noise I- Part I: Optimum Detection," IEEE Transactions on Information TheoryIT-213 (1982, to appear). 3
15.26 H. V. Poor, "Signal Detection in the Presence of Weakly Dependent Noise- Part II: Robust Detection," IEEE Transactions on Information TheoryIT-23 (1932, to appear).
15.27 H. V. Poor, "Robust Matched Filters," IEEE Transactions on InformationTheory IT-23 (1982, to appear). 3
15.29 H. V. Poor and D. P. Looze, "Minimax State Estimation for LinearStochastic Systems with Noise Uncertainty," IEEE Transactions onAutomatic Control AC-26, 902-306 (Aug. 1331).
15.29 H. V. Poor and Y. Rivani, "Input mplituie Compression in DigitalSignal-Detection Systems," IEEE Transactions on Communications J' -29,I 07-713 (May 1931). *
xxxii
I PUBLICATIONS
15.30 H. V. Poor and J. B. Thomas, "Memoryless Quantizer-Detectors forConstant Signals in m-dependent Noise," IEEE Transactions on InformationTheory IT-26, 423-432 (July 1980).
I 15.31 D. V. Sarwate, "Review of 'An Algorithm to Solve the mxm AssignmentProblem in Time O(mnlogn)'," Computing Reviews 21, 382 (Aug. 1980).
I 15.32 W. E. Stark and D. V. Sarwate, "Kronecker Sequence for Spread-SpectrumCommunications," IEEE Proceedings, Part F: Communications, Radar, andSignal Processing 128, 104-107 (April 1981).
Conference Papers
15.33 A. C. Blumer, R. J. McEliece, M. B. Pursley, and X. S. Wallace,
"Universal Noiseless Source Coding Techniques for Markov Sources,"Abstracts of Papers for the 1981 IEEE International Symposium on3 Information Theory pp. 71-72 (Feb. 1981).
15.34 C.-I Chang and H. V. Poor, "On the Performance of Memoryless DetectionSystems Relative to Systems with Memory," Proceedings of the 1981Conference on Information Sciences and Systems, Johns HopkinsUniversity, pp. 140-1 .6 (March 1981).
3 15.35 F. D. 'arber and M. B. Pursley, "Effects of Selective Fading onSlow-Frequency-.fopped DPSK Spread-Spectrum Communications," 1981 IEEENational Telecommunications Conference, Conference Record (Nov. 1981, to
I appear).
15.36 E. A. 'eraniotis, 'Error Probability for Coherent Hybril Slow-Frequency-Hopped Direct-Sequence Spread-Spectrum Multiple-Access Communications,"1981 IEEE National Telecommunications Conference, Conference Record(Nov. 1981, to appear).
15.37 3. A. Geraniotis and H. V. Poor, "Minimax Filtering Problems forIbserved Poisson Processes with Uncertain Rate Functions," Proceedingsof the 20th IEEE Conference on Decision and Control (Dec. 1981 , to
I appear).
15.38 E. A. Geraniotis and M. B. Pursley, "Performance Analysis for FastFrequency-Hopped Spread-Spectrum Multiple-Access Commuications," IEEENational Telecommunications Conference, Conference Record 1_, pp.17.4.1-5 (Dec. 1980).
15.39 E. A. Geraniotis and M. B. Pursley, "Error Probability for Binary PSKSpread-Spectrum Multiple-Access Communications," Proceedings of the 1991Conference on Information Sciences and Systems, Johns Hopkins3 University, pp. 238-244 (March 1981).
15.40 3. A. Geraniotis and M. B. Pursley, "Error Probability Bounds f:rSlow-Frequency-Hopped Spread-Spectrum Yultiple-Access Communi-ationsover Fading Channels," IEEE International Conference on Commrunications,Conference Record 4, pp. 76.3.1-7 (June 1981).
I0
Axiii
PUBLICATIONS
15.41 3. Hajek, "Markov Properties of Stochastic Differential Equations,"Proceedings of the 1380 Joint Automatic Control Conference, paper TP7-D,(Aug. 1980).
15.42 B. Hajek, "Dynamic Decentralized Estimation and Control in a"ulti-Access Broadcast Thannel," Proceedings of the 19th IEEE Conferenceon Decision and Control, pp. 613-623 (Dec. 1980).
15.43 B. Hajek, "Information of Partitions with Applications to Random AccessCommunications," Abstracts of Papers for the 1981 IEEE InternationalSymposium on Information Theory, pp. 106-107 (Feb. 1981).
15.44 3. Hajek, "Locking Schemes for Pseudo Time Division Multiple Access,"Proc. of the 1981 Conf. on Information Sciences and Systems, JohnsHopkins University, pp. 226-232 (March 1981).
15.15 3. Hajek and E. Wong, "Set-Parametered Martingales and M4ultipleStochastic Integration," Stochastic Integrals Proc., LAS Durham Symp.,Lecture Notes in Math. vol. 851, Springer-Verlag, New York (1931).
15.45 D. R. Halverson, 3. L. Wise, and H. V. Poor, "On an Aspect of LinearCorrection Terms for Nonlinear Detection Systems," Proc. 24th MidwestSymposium on Circuits and Systems (June 1981, to appear).
15.47 D. P. Looze, H. V. Poor, K. S. Vastola, and J. C. Darragh, "MinimaxLinear- uadratic-Gaussian Control of Systems -gith Uncertain Statistics,"Proceedings of the 10th IFI? Conference on System Modeling andOptimization (Aug. 1981 , to appear).
13.43 . J. AcEliece, "The Asymptotic Redundancy of Huffman Coding," Abstractsof Papers for the 1981 IEEE International Symposium on informationTheory, p. 49 (Feb. 1981).
15.49 R. J. McEliece and W. 3. Stark, "An Information-Theoretiz Study ofCommunication in the Presence of Jamming," 1931 IEEE InternationalConference on Communications, Conference Record _, pp. 45.3.1-5 (Jine1981 ).
15.50 'A. V. Poor, "The Rate-Distortion Function on Classes of SourcesDetermined by Spectral Capacities," Abstracts of Papers for the 1931:EEE International Symposium on Information Theory, pp. 52-53 (Feb.1981 ).
13.51 H. V. Poor and D. P. Looze, "Minimax State Estimation for LinearStochastic Systems with Noise Uncertainty," Proceedings of the 19th :EEEConference on Decision and Control, pp. 1)20-1)25 (Dec. 1980).
15.52 A. B. iursley, "Effects of Partial Correlation on the ,utiole-XccessCapability of Direct-Sequence Spread 3pectrum," 1931 IEEE lationalTelecommunications Conference, Conference Record "Nov. 1981, to appear).
Ixxxiv
5 PUBLICATIONS
I 15.53 W. 3. Stark and R. J. McEliece, "Communication in the Presence of Fadingand Partial-Band Jamming," 1981 IEEE National TelecommunicationsConference, Conference Record (Nov. 1981, to appear).
n 15.54 S. Tantaratana, "A Truncated Sequential Test: Performance and Designwith Prescribed Error Probabilities," Proceedings of the 1981 Conferenceon Information Sciences and Systems, Johns Hopkins University,
pp. 333-333 (March 1-981 ).
15.55 K. S. Vastola and H. V. Poor, "On Choquet Capacities and theirDerivatives with Respect to a-finite Measures," Abstracts of Papers for
the 1981 IEEE International Symposium on Information Theory, p. 91 (Feb.198. ).
1 15.56 K. S. Vastola and H. V. Poor, "Performance Analysis of Robust FilteringSchemes," Proceedings of the 23rd Midwest Symposium on Circuits andSystems, pp. 259-263 (Aug. 1980).
15.57 K. S. Vastola and H. V. Poor, "Robust Linear Estimation of StationaryDiscrete-Time Signals," Proceedings of the 1931 Conference onInformation Sciences and Systems, Johns Hopkins University, pp. 512-516(March 1931).
15.58 S. Verda and H. V. Poor, "Least Favorable Signals and Noise for
Discrete-Time Robust Matched Filtering," 1981 IEEE NationalTelecommunizations Conference, Conference Record (Nov./Dec. 19SI , toappear).
Reports
15.59 R. J. McEliece, "A Note on the Ro-parameter for Discrete MerorylessChannels," Jet Propulsion Laboratory TDA Progress Report 42-60, pp.
96-97 (Sept.-Oct. 1930).
I 15.60 R. J. McEliece, "Interleaved Block Codes for the Photon Channel," JetPropulsion Laboratory TDX Progress Report 42-60, Sept. pp. 98-1021 (Oct. 1980).
15.61 R. J. McEliece, "RFI Channels," Jet Propulsion Laboratory TDA Progress
Report 42-60, pp. 103-136 (Sept.-Oct. 1930).
15.62 R. J. McEliece, "RFI Channels, !I," Jet Propulsion Laboratory TDAProgress Report 42-63 (1981, in press).
I 15.63 R. J. McEliece and 3. Rodemich, "The R 0 Parameter for the laussianChannel," Jet Propulsion Laboratory TDA Progress Report 42-64 1931, inpress).
15.64 J. S. Lehnert, "Chip *aveform Selection in Offset- uaternary Direct-Sequence Spread-Spectrum Aultiple-Access Communications," CoordinatedScience Laboratory Report No. 907, University of Illinois (Jan. 1981).
I, ..
xxxv
PUBLICATINS
16. ANALOG AND DIITAL CIRCUITS
3ooks
16.1 .W. :ayeda, "Dividing a System into Almost Unidirectional Blocks," ingraph Theory and Algorithms, Springer-Verlag, New York (1931).
16.2 .. X.yeda, Fundamentals of Digital Signal Processing, Thm Publ. Co.,Japan (1980).
Journal Articles
16.3 "E. I. El-:Aasry and M. . artens, "Determination of Phase-Locked Loop
Statistics with the Aid of RLC Ladder Networks," IEEE Trans. onCommunications nJM-2_91 ), 73-79 (Jan. 1981).
16.4 I. N. Hajj, "Algorithms for Solution Updating Due to Large C-hanges inSystem Parameters," Int. J. on Circuit Theory and Applications 9, 1-11(Jan. 1931 ).
16.5 1. N. 3iajj and S. Skelboe, "Steady-State Analysis of Piecewise-LinearDynamic Systems," IEEE Trans. on Circuits and Systems CAS-23(3), 234-242(Aarch 1931 ).
16.6 1. N. +ajj, P. Yang and T. N. Trick, "Avoiding Zero Pivots in the
Aodified Nodal Approach," EEE Trans. on Circuits and Systems AS-23 (4),271-279 (April 1981).
16.7 . F. Lee and W. K. Jenkins, "Jomputer-Aiied Analysis of SwitchedCapacitor Filters," IEEE Trans. on Circuits and Systems CAS-23(7)681 -692 (July 1931 ).
16.3 A4. R. Lightner and S. I. Director, "Multiple Criterion Optimization forthe Design of Electronic Circuits," IEEE Trans. on Circuits and SystemsCAS-23(3), 169-179 (March 1981).
16.9 T. N. Trick, "A Note on Parameter Value Determination from Node VoltageHeasurements," IEEE Trans. on Circuits and Systems 'Ao-2, 1269-1270(Dec. 1930).
Ixxxvi
5 PUBLIC ATI ONS
5 Conference Papers
16.10 N. Attaie and E. 1. EI-Masry, "Synthesis of the Switched-CaapacitorFilters in the 'laltiple-Input Follow-the-Leader Feedback Topology," 1991IEEE Intl. Symp. on Circuits and Systems, Chicago, IL, pp. 175-179(April 1981).
16.11 N. Xttaie and E. I. El-Xasry, "Optimum Realizations of the Switched-Capacitor Filters," Proc. of the 23rd Midwest Syrup. on Circuits and5 I Systems, Toledo, OH, pp. 433-439 (Aug. 1980).
16.12 ". Cortelazzo, M. R. Liglitner, and W. K. Jenkins, "Frequency Domain* I Design of Multiband Finite Impulse Response Digital Filters Based on theI Minimax Criterion," Proc. 1981 IEEE ISCAS, Chicago, IL, pp. 532-535
(April 1981).
16.13 R. D. Davis, "An Efficient LU Factorization Scheme for the Analysis ofSwitched-Capacitor Filters," Proc. 24th Aidwest Symp. on Circuits andSystems, Albuquerque, NM (June 1981).
I 16.14 R. D. Davis, "A Derivation of the Switched-Capacitor Adjoint Network
Based on a Modified Tellegen's Theorem," Proc. 1981 Int. Symp. onICircuits and Systems, Chicago, IL, pp. 372-875 (April 1981).
16.15 1. D. Davis, "Distortion Analysis of Switched Capacitor Filters," ?roc.
1981 Intl. Syup. on Circuits and Systems, Chicago, IL, pp. 876-879
(April 1981).
16.16 Z. 1. El-Aasry, "State-Space Switched-Capacitor Filter Structures,"Proc. 14th Asilomar Conf. on Circuits, Systems and Computers, PacificGrove CA, pp. 5-9 (Nov. 1980).
16.17 S. I. l-" asry, "Design of Switched-Capacitor Filters in the Biquadratiz:State Space Form," 1931 IEEE Intl. Symp. on Circuits and Systems,Chicago IL, pp. 179-132 (April 191).
16.13 3. 1. El-Aasry and G. 0. Martens, "Transient Statistics of the
First-Order Phase-Locked Loop," Proc. 23rd Midwest Symp. on Circuits andSystems, Toledo, OH, pp. 119-124 (Aug. 1980).
16.19 F. B. 'irosz, Jr. and T. N. Trick, "Some Modifications to Newton's Methodfor the Determination of the Steady State Response of Nonlinear
I Circuits," Proc. 24th Midwest Symp. on Circuits and Systems, AlbuquerqueT (June 1981 ).
16.20 1. N. Hajj and S. Skelboe, "Steady-State Analysis of Piecewise-Linear
Circuits," 1980 European Conf. on Circuit Theory and Design, Warsaw,Poland (Sept. 1980).
16.21 I. 1. Hajj, "An Updating Algorithm for the Shortest Path Problem" I ?81Intl. Symp. on Circuits and Systems, Chicago IL, pp. 302-305 (April
a. 19 -
xxxvii
PUB LIC-A MS
16.22 1. N. Hajj, "Large-Change Sensitivity of Large-'Pzale 'letiforks," ?rOc.211h Mid'west Symp. on Circuits and Systems, Albuqu.erque NT4 (.J,,ne 1931)'.
16.2- D. :oc-3var and T. N1. Trick, "Comparison of Some Automatic TuningAlg-orithms," ?roc. 24th Miiwest Symposium on Circuits and Systems,Albuquerque T4 (June 1931 ).
15.24 D. -3. Hocevar and 'd. K. Jenkins, "Sensitivity Analysis Of SwitchedCapacitor .Circuits," Proc. 23rd Midwest Symp. on Circuits and SYstems,Toledo 0H, pp. 439-443 (Aug. 1930).
13.25 C. F. Lee and Wd. K. Jenkins, "Switchned Capacitor Circuit Analysis Basedon .Modified Nodal Analysis," Proc. 23rd :iiwest Symp. on Circuits andSystems, Toledo OH1, pp. 449-453 (Aug. 1980W.
15-. 25 6. F. L ee and W. K. Jenkins, "Sensitivity Analysis of Switched CapacitorCircuits Based on Modified Nodal Analysis," ?roc. 1930 intl. Conf. onCircuits and Clomputers, Port Chester NTY, pp. 855-358 (Oct. 13,9O).
16.27 .F.Lee and W. K. Jenkins, "A Design Approach for Stray insensitivePractical Switched Canacitor Circuits," Proc. 14th Asilomar Conf. onCircuits, Systems, and ComDuters, Pacific 'rove 'A, pp. 415-1.19 "4ov.1930).
15.23 C. ?. Lee and W. K. Jenkins, "A Comparison of Swiltched Caiacitor Ci--rc ui4tAnalysis Techniques Based on Block Partitioned :acromodels and ModifiedNodal Analysis," Proc. 24th Midwest Sy-np- on Circuits and Systems,Albuquerque NA (June 1 931 ) .
15.29 A4. R. Lightner, P. Yang, and C. L. Pahlrneyer, "PiecewiseNoieaAnalysis of Electronic Circuits," Proc. TEZE ICCC 30, Port Chester NY,pp. 304-807 (Oct. 1,930).
15.30 H. R. Lightner and C. L. Pahlmneyer, "Algorithms and Models o Sneed DCConyvergence in Bipolar Circuits," Proc. 24th Mid-jest Symp. on Circuitsand Systems, Albuquerque T.1 (June 1981).
16.31 P. Yang, 1. N. qajj and T. N. Trick, "SLATE: A Circuit SimulatorProgram with Latency :Exploitation and Node Tearing," Int. Con'. onCircuits and Computers, Port Chrester NY, pp. 353-355 (Oct. 1930).
Technical iReports and Theses
15.32 3. Cortalazzo, "Frequency Domain Design of, Multiband Finite impulseResponse Digital Fil ters 3a sed uoon the' Ainimax Criterion," ]S LTechnical Report R-906 (1930).
15.-33 'u. larza, "An Tn-Depth View of Designing Swiitched Capqcitor Sampledl-Data?il ters, " MA.S. Thesis, Department of Electrical Engineeri ng, Universityof illinois3 at Urbana-Chiamnaivn (May 1931).
Ixxxviii5 PUBLIC ATI ONS
3 16.34 P. Kollaritsch, "REPLICA - A Random, Electronically Reprogrammable,Logic Integrated Custom Array," University of Illinois atUrbana-Champaign, M.S. Thesis (August 1981 ).
1 16.35 C. F. Lee, "An Investigation of Computer-Aided Analysis for SwitchedCapacitor Sampled-Data Filters," Ph.D. Dissertation, Department ofElectrical Engineering, University of Illinois at Urbana-Thampaign (May
I 1981 ).
16.36 ". L. Pahlmeyer, "Algorithms and Models to Speed DC Convergence inBipolar Circuits," M.S. Thesis, University of Illinois at Urbana-Champaign (Aug. 1981 ).
16.37 J. B. Resman, "Impact of Structured Design Methodologies on VLSISystems," M.S. Thesis, University of Illinois at Urbana-Champaign (Aug.199I ).
16.38 P. Yang, "An Investigation of Ordering, Tearing and Latency Algorithmsfor the Time Domain Simulation of Large Circuits," CSL Technical ReportI R-891 (Aug. 1980).
1 17. DECISION AND CONTROL
Journal Articles
17.1 J. Ackermann, "Parameter Space Design of Robust Control Systems," IEEETrans. on Automatic Control AC-25(6), 1058-1072 (Dec. 1980).
1 17.2 F. Albrecht, K. A. Grasse and 3. Wax, "Reproducibility of Linear andNonlinear Input-Output Systems," J. Mathematical Analysis and
I73Applications 79(1), 132-202 (Jan. 1991).
17.3 J. J. Allemong and P. V. Kokotovic, "Eigensensitivities in Reduced OrderModeling," IEEE Trans. on Automatic Control AC-25(4), 321-822 (1990).
17.4 B. Avramovic, J. Chow, P. Kokotovic, and J. Winkelman, "Area
Decomposition for Electromechanical Models of Power Systems," Automatica5 1 , 620-630 (1930).
17.5 A. Bagchi and T. Basar, "Stackelberg Strategies in Linear- QuairaticStochastic Differential Games," J. of Optimization Theory andApplications 35(3).
17.6 T. Basar, "Stochastic Muti-Criteria Decision Problems with Multi-Levelsof Hierarchy," IEEE Trans. on Automatic Control AC-26(2), 549-553(1981).
, !
xxxix
PUBLICATIDINS
1 .7 T. Basar, "Equilibrium Strategies in Dynamic Games with Hulti-Levels ofHierarchy," Automatica 16(4) (1981).
17.3 T. Basar, "On the Saddle-Point Solutions of a Class of StochasticDifferential Games," J. Optimization Theory and Applications W3$I)(1 931).
17.9 J. H. Chow and ?. V. Kokotovic, "A Two Stage Lyapunov-Bellman FeedbackDesign of a Class of Nonlinear Systems," IEEE Trans. on AutomaticControl AC-26, 656-663 (1981).
17.1D D. Cobb, "Feedback and Pole Placement in Descriptor Variable Systems,"Int. J. Control 33, 1-12 (1981).
17.11 J. B. Cruz, Jr., J. S. Freudenberg, and D. P. Looze, "A Relationshipbetween Sensitivity and Stability of 'ultivarable Feedback Systems,"IEEE Trans. on Automatic Control AC-26, 66-71 (1981).
17.12 J. B. Cruz, Jr., D. ?. Looze and W. R. Perkins, "Sensitivity Analysis ofNonlinear Feedback Systems," J. Franklin institute (to appear).
17.13 W. E. Hopkins, J. Xedanic, and 1. R. Perkins, "Output-Feedback PolePlacement in the Design of Suboptimal Linear uadratic Regulators," int.J. Control (1 931 , to appear).
17.11 H. K. inralil, "M4ultimodel Design of a Nash Strategy," J. of OptimizationTheory and Applications 31 (4), 553-564 'Aug. 1980).
17.15 H. K. Khalil and P. V. Kokotovic, "Decentralized Stabilization of
Systems with Slow and Fast Aodes," Large Scale Systems 1(3>, 11-liT(1 980).
17.15 P. V. Kokotovic, "Subsystems, Time Scale, and Multimodeling," Automatica(to appear).
17.17 P. V. Kokotovic, J. J. Allemong, J. R. dinkelman, and J. H. "how,"Singular Perturbation and Iterative Separation of Time Scales,"Automatica 16, 23-33 (1980).
17.13 P. V. Kokotovic, B. Avramovic, J. H. Chow, and J. R. Winkelman,"Coherency Based Decomposition and Aggregation," Automatica ('lov. 1981to appear).
17.19 D. P. Looze and N. R. Sandell, Jr., "Analysis of DecompositionAlgorithms via Nonlinear Splitting Func-tions," J. Optimization Theoryand Applications (July 1981, to appear).
17.20 ". S. Padilla, J. B. Cruz, Jr., and R. A. Paiilla, "A Simple Algorithmfor SAFER Control," Int. J. of Control 32(6), 1111-1113 (Dec. 1?90).
5 PUBLICATIONS
I 17.21 3. Papavassilopoulos, "Algorithms for a Class of NondifferentiableProblems," J. Optimization Theory and Applications 34 (1), 41-82 (May1981).
I 17.22 3. Papavassilopoulos and J. B. Cruz, Jr., "Sufficient Conditions forStackelberg and Nash Strategies with Memory," J. Optimization Theory andADplications 31(2), 233-260 (June 1980).
17.23 R. G. Phillips, "A Two-Stage Design of Linear Feedback Controls," IEEETrans. on Automatic Control AC-25(6), 1220-1222 (Dec. 1980).
I 17.24 R. G. Phillips, "Reduced Order Modeling and Control of Two-Time-ScaleDiscrete Systems," Int. J. of Control 31(4), 765-780 (1980).
5 17.25 H. V. Poor and D. P. Looze, "Minimax State Estimation for LinearStochastic Systems with Noise Uncertainty," IEEE Trans. on AutomaticControl AC-26, 902-906 (Aug. 1981).
17.26 V. Saksena and J. B. Cruz, Jr., "Stabilization of Singularly PerturbedLinear Time-Invariant Systems Using Low-Order Observers," IEEE Trans. on
I Automatic Control AC-26, 510-513 (April 1981 ).
17.27 V. Saksena and J. B. Cruz, Jr., "Nash Strategies in DecentralizedControl of Multi-Parameter Singularly Perturbed Large Scale Systems,"Large Scale Systems (1981, to appear).
17.23 A. A. Salman and J. B. Cruz, Jr., "An Incentive Model of Duopoly with3overnment Coordination," Automatica 17 (Nov. 1981, to appear).
17.29 P. W. Sauer, "On the Formulation of Power Distribution Factors for
Linear Load Flow Analysis," IEEE Trans. on Power Apparatus and SystemsPAS-100(2), 764-770 (Feb. 1981).
17.30 P. v. Sauer, "Explicit Load Flow Series and Functions," IEEE Trans. on
Power Apparatus and Systems (to appear).
17.31 J. Sharp, F. Shupp, and W. R. Perkins, "Energy Conservation and InducedInflation," J. Econ. Dynamics and Control 2(3), 213-231 (Aug. 1980).
17.32 J. R. Winkelman, J. H. Chow, J. J. Allemong, and P. V. Kokotovic,"Multi-Time-Scale Analysis of a Power System," Automatica 16, 35-13(1980).
I 17.33 J. R. Winkelman, J. H. Chow, B. C. Bowler, B. Avr-amovic:, and P. V.Kokotovic, "An Analysis of Interarea Dynamics of Multi-Machine Systems,"IEEE Trans. on Power Apparatus and Systems PAS-100 2), 754-7653 (19 1).
II
xli
PUBLICATIONS
Conference Payers
17.34 1. M. Alunkal, P. W. Sauer, "Load Flow Methods Near Maximum PowerTransfer," Proc. Twelfth Ann. Pittsburgh Conf. on Modeling andSimulation, Univ. of Pittsburgh, Pittsburgh, PA, (April 1981).
17.35 T. Basar, "A New Method for the Stackelberg Solution of Differential'ames with Sampled-Data State Information," IFAC/31 Congress, Kyoto,Japan, (Aug. 1981).
17.36 J. H. Chow and P. V. Kokotovic, "A Two Stage Lyapunov-Bellman Feedback
Design of a Class of Nonlinear Systems," Proc. 19th ISEE Conf. onDecision and Control, Albuquerque, NM (Dec. 1980).
17.37 J. B. Cruz, Jr., "Surviey of Leader-Follower Concepts in HierarchicalDecision-Aaking," Fourth Int. Conf. on Analysis and Optimization ofSystems, Le Chesnay, France, pp. 384-396 (Dec. 1980).
17.33 J. B. Cruz, Jr., J. S. Freudenberg, and Douglas P. Looze, "ARelationship between Sensitivity and Stability of :Aultivariable FeedbackSystems," Proc. JACC, San Francisco, CA, pp. WP8-C-1 to '4P8-C-9 (Aug.1980).
17.39 J. B. Jruz, Jr. and M. A. Salman, "Nash Games with Decision-DependentInformation Structures," Proc. 19th IEE Conf. on Decision and Control,"Albuquerque, T-A, pp. 1130-1186 (Dec. 1980).
17.40 P. A. Ioannou and P. V. Kokotovic, "Error Bounds for Model-Plant
Mismatch in Identifiers and Adaptive Observers," Proc. JACC,
Charlottesville, VA (June 1991).
17.41 P. V. Kokotovic, B. Avramovic, J. H. Chow, and J. R. Winkelman,"Coherency Based Decomposition and Aggregation of a Class of Large ScaleSystems," 3th World Congress of IFAC, Kyoto, Japan (Aug. 1931 ).
17.42 P. V. Kokotovic, R. G. Phillips, and S. H. Javil, "3ingular Perturbation.4odeling of Markov Processes," 4th Intl. Conf. on Analysis andOptimization of Systems, Paris, France (Dec. 1980).
17.43 D. Lindner, W. R. Perkins, and J. V. Medanic, "Chained Aggregation: A3eometric Analysis," ith World Congress of IFAC, Kyoto, Japan (Aug.19331 ).
17.44 D. P. Looze, "A Dual Optimization Procedure for Linear Quadratic, RobustControl Problems, 2nd IFAC Workshop on Control ApDlizations of.onlinear Programming and Optimization, )berpfaffenhofen, W. "ermany(Sept. 1980).
II
xlii
5 PUBLICATIONS
I 17.45 D. P. Looze, H. V. Poor, K. S. Vastola, and J. C. Darraugh, "On Linear-uadratic-Gaussian Control of Systems with Uncertain Noise Statistics,"Proc. 10th Intl. Fed. for Information Processing Conf. on System
Modeling and Optimization (Sept. 1981, to appear).
17.46 D. P. Looze and H. Salhi, "Decomposition and Aggregation in Hierarchical3 Systems," 3th World Congress of IFAC, Kyoto, Japan (Aug. 1981 ).
17.47 ". S. Padilla, J. B. Cruz, Jr., and R. A. Padilla, "A Simple Algorithmfor SAFER Control," Proc. JACC, San Francisco, CA, pp. !P2-1-1 toI WP2-3-5 (Aug. 1980).
17.43 1. A. Padilla, J. B. Cruz, Jr., C. S. Padilla, and R. Milito, "OnAdaptive Nash Games," 8th World Congress of IFAC, Kyoto, Japan (Aug.1981).
17.49 J. V. Pietras and D. P. Looze, "Decomposition of Decentralized Gain
Computations for Interconnected Systems," Proc. JACC, San Francisco, CA,
pp. WA1-C-1 to WA1-C-6 (Aug. 1980).
17.50 H. V. Poor and D. P. Looze, "Minimax State Estimation for Linear
Stochastic Systems with Noise Uncertainty," Proc. 19th IEEE Conf. on
Decision and Control, Albuquerque, NM, pp. 1020-1025 (Dec. 1980).
17.51 V. R. Saksena and J. B. Cruz, Jr. "A Multi Model Approach to StochasticNash Games," 1931 IEEE Intl. Symp. on Circuits and Systems, Chicago, IL,pp. 566-569 (April 1981).
17.52 H. Salhi and D. P. Looze, "Aggregation in Hierarchical Systems:Approximate Decompositions," Proc. 19th IEEE Conf. on Decision and
Control, Albuquerque, NM (Dec. 1980).
17.53 H. Salhi and D. P. Looze, "An Algorithm for the Decomposition of Second
Order Elliptic Operators," Proc. 10th IFIP Conf. on System Modeling andOptimization (Sept. 1981, to appear).
17.54 1. A. Salman and J. B. Cruz, Jr., "Information Structure, OptimalCoordination, and a Game Model of Duopoly," 9th World Congress of IFAC,Kyoto, Japan (Aug. 1981).
17.55 P. W. Sauer, "On the Formulation of Power Distribution Factors forLinear Load Flow Analysis," Paper 305m 614-8, 1980 IEE PowerEngineering Soc. Summer Meeting, Minneapolis, 'AN (July 1980).
1 17.56 P. W. Sauer, Explicit Load Flow Series and Functions," paper 31im034-8, 1981 IEE Power Engineering Society Winter Meeting, Atlanta. 3A(Feb. 1981).
17.57 J. R. Winkelman, J. H. Chow, B. C. Bowler, B. Avramovic, and P. V.Kokotovic, "An Analysis of Interarea Dynamics of Multi-Machine Systems,"
IEEE PES Summer Meeting, Minneapolis, MN, paper 30 SM 533-0 (July 1330).
xliii
PUBLICATIONS
17.53 J. R. Winkelman, J. -H. Chow, P. V. Kokotovic, and 3. kvramovic,"Singular Perturbation Analysis of Power System Dynamizs," SIAM Intl.Conf. on Electric Power Problems: The Aathematical Challenge (March1991).
Technical Retorts
17.59 J. Ackermann, S. N. Franklin, C. 3. Chato, D. P. Looze, "Parameter SpaceTechniques for Robust Control System Design," CSL Report R-890 (DC-39)(Aug. 1980).
17.60 3. R. Avramovic, "Time Scales, Coherency, and Weak Coupling," CSL ReportR-395 (DC-40)()ct. 1980).
17.61 k. Bensoussan, J. H. Chow, and P. V. Kokotovic, "Stabilization andStochastic Control of a Class of Nonlinear Systems," CSL Report R-900(DC-42)()ct. 1980).
17.62 P. X. !oannou and P. V. Kokotovic, "Error Bounds for Xodel-PlantMismatch in Identifiers and Adaptive Observers," CSL Report R-397(DC-41)(Oct. 1980).
17.63 P. V. Kokotovic, "Singular Perturbations and Time Scales in Aodeling andControl of Dynamic Systems," CSL Report R-901 (DC-43)(:Iov. 1930).
17.64 3. 3. Phillips, "Decomposition of Time Scales in Linear Systems andMarkovian Decision Processes," CSL Report R-902 (DC-44)(Nov. 1930).
17.65 A. A. Salman, "Information Structures in Nash and Leader-FollowerStrategies," CSL Report R-903 (DC-45)(Jan. 1991 ).
13. D13ITAL SIGNAL AND I'A.3E PROCESSINI
Books
13.1 T. S. Huang (ed.), Two-Dimensional Digital Signal Processing 1: LinearFilters, 3prinlger-Verlag, lew York (1991).
13.2 T. S. 'Huang, T.hapter I, "Introduction," ibid.
13.3 T. S. Huang (ed), Two-0imensional Digital Signal ProcessinTransforms and Median Filters, Springer-Verlag, New York 191).
I!xli v
PUBLICATIONS
18.4 T. S. Huang, Chapter 1, "Introduction",ibid.
18.5 T. S. Huang (ed.), I Sequence Analysis, Springer-Verlag, New York(Sept. 1981, to appear).
13.6 T. S. Huang and R. Y. Tsai, Chapter 1, "Image Sequence Analysis: NotionEstimation," ibid.
18.7 T. S. Huang and Y. P. 'Hsu, Chapter 4, "Image Sequence Enhancement,"ibid.
13.8 T. S. Huaag, "Image Sequence Analysis: Relationship Between Temporaland Spatial Dimensions," in Transformations in Optical SignalProcessing, W. Rhodes, J. Fienup, and B. Saleh (eds. Society ofPhoto-Optical Instrumentation Engineers, Redondo Beach, CA (1981 , toappear).
13.9 T. S. Huang, "Dimensional Mappings in Fast DFT and Circular ConvolutionAlgorithms," ibid.
19.13 B. T. O'Connor and T. S. Huang, "Stability of 3eneral Two-DimensionalRecursive Digital Filters," Chapter 4 in Two-Dimensional Digital SignalPrssin I: Linear Filters, T. S. Huang (ed.), Springer-Verlag, NewY 191).
Journal Articles
13.11 J. K. Cheng and T. S. Huang, "A Subgraph Isomorphism Algorithm UsingResolution," Pattern Recognition (to appear).
13.12 M. H. Etzel and W. K. Jenkins, "Redundant Residue Number Systems forError Detection and Correction in Digital Filters," IEEE Trans. onAcoustics, Speech, and Signal Processing ASSP-23, 538-545 (Oct. 1990).
19.13 W. K. Jenkins, "Complex Residue Number Arithmetic for High Speed SignalProcessing," Electronics Letters 16(17), 660-661 (Aug. 14, 1980).
18.14 W. K. Jenkins and M. H. Etzel, "Special Properties of Complement Codesfor Redundant Residue Number Systems," Proc. IEEE 69, 132-133 (Jan.1981 ).
13.15 0. R. Mitchell, E. Delp, and T. S. Huang, "Image Coding forPhotoanalysis," Proc. of the Society for Information Display 21 3)279-292 (1980).
L I
19.16 D. C. Munson, Jr., "On Finite Wordlength FIR Filter Design," (note) IEEETrans. on Acoustics, Speech, and Signal Processing kSSP-29, 329 (April1981).
xlvPUBLICATICNS
13.17 D. C. Munson, Jr., "Accessibility of Zero-Input Limit Cycles," IEEETrans. on Acoustics, Speech, and Signal Processing (to appear).
13.13 D. C. Munson, Jr. and B. Liu, "Narrow-Band Recursive Filters with ErrorSpectrum Shaping," IEEE Trans. on Circuits and Systems CkS-23, 160-163(Feb. 1981).
13.19 D. C. Munson, Jr. and B. Liu, "Floating Point Roundoff Error in thePrime Factor FFT," IEEE Trans. on Acoustics, Speech, and Signal
SProcessing (Aug. 1951, to appear).
13.20 3. T. O'Connor and T. S. Huang, "Application of Phase Unwrapping toImage Restoration," Computer Graphics and Image Processing 15, 25-14(Jan. 1981).
13.21 1. Y. Tang and T. S. Huang, "Using the Creation Machine to LocateAirplanes in Aerial Photos," Pattern Recognition 12(6), 431-442 (1980).
13.22 R. Y. Tsai and T. S. Huang, "Estimating the Three-Dimensional MotionParameters of a Rigil Planar Patch," IEEE Trans. on Acoustics, Speech,and Signal Processing (to appear).
13.23 G. Tang and T. S. Huang, "The Effect of Median Filtering on ElgeLocation Estimation," Computer Graphics and Image Processing 15, 224-245(March 1981 ).
13.24 Y. Yasuda, A. Dubois, and T. S. Huang, "Data Compression for TheckProcessing Machines," Proc. IEEE -3, 374-385 (July 1980).
Conference Papers
13.25 1. Cortelazzo, M. R. Lightner, and W. K. Jenkins, "Frequency Domain
Design of Multiband Digital Filters Based on the Ainimax Criterion,"Proc. 1981 IEEE Intl. Symp. on Circuits and Systems. "hicago, IL, pp.532-535 (April 1931).
18.26 A. I. Etzel and W. K. Jenkins, "Hardware Implementation of ?ailureResistant Residue Number Digital Filters," Proc. 1931 IEEE Int. Symp. :nCircuits and Systems, Chicago, IL, pp. 38-91 (April 1931).
13.27 T. 3. Huang, "Noise Filtering in Moving Images," ?roc. 1 90 2EE int.Conf. on Pattern Recognition, Miami Beach, FL, pp. 1161 -1 164 'De.1980).
13.23 T. S. Huang and R. Y. Tsai, "Three Dimensional Motion Estimation," ?roc.First European Signal Processing Conf., Lausanne, Switzerlani, pp.2653-26 ept. 1980).
Ixlvi
PUBLICATIONS
13.29 T. S. Huang and R. Y. Tsai, "Three Dimensional Motion Estimation fromImage Space Shifts," Proc. 1981 IEEE Int. Conf. on Acoustics, Speech,and Signal Processing, Atlanta, GA, pp. 1136-1139 (March 1981).
18.30 W. K. Jenkins, "Redundant Complex Residue Number Systems," Proc. VeryHigh Speed Computing Symp., Georgia Inst. of Tech., Atlanta GA, pp.IV.3-IV.20 (Sept. 1980).
18.31 '. K. Jenkins, "The Effect of Algorithm Selection on the Performance ofMonolithic Signal Processors," Proc. 1980 Nat. Electronics Conf.,Chicago, IL, pp. 425-430 (Oct. 1980).
* 18.32 -. K. Jenkins and C. F. Lee, "Complex Residue Number Arithmetic forDigital Signal Processing," Proc. 14th Asilomar Conf. on Circuits,Systems, and Computers, Pacific Grove, CA, pp. 4.0-493 (Nov. 1980).
18.33 D. C. Munson, Jr., "Determining Exact Maximum Amplitude Limit ?ycles inDigital Filters," Proc. 1981 IEEE Int. Conf. on Acoustics, Speech, andSignal Processing, Atlanta, GA, pp. 1231-1234 (March 1981).
13.34 D. C. Munson, Jr., "Accessibility of Zero-Input Limit Cycles," Proc.1981 IEEE Int. Symp. on Circuits and Systems, Chicago, IL, pp. 321-324(April 1981).
13.33 D. C. Munson, Jr. and W. 3. Higgins, "Error Spectrum Shaping VersusBlock Optimal Structures for Digital Filtering," Proc. 24th MilwestSymp. on Circuits and Systems, Albuquerque, TA (June 1981).
Technical Report
13.36 G. Cortelazzo, "Frequency Domain Design of Multiband Finite ImpulseResponse Digital Filters Based on the Minimax Criterion," TechnicalReport R-906, Coordinated Science Laboratory, University of Illinois atUrbana-Zhampaign, (December 1 930).
19. POPULATION/FOOD/CLIMATE STUDIES
19.1 Paul Handler, "Crop-Crop "orrelations as an Indicator of ClimateTeleconnections in the Region of the Sozthern Oscillation," ClimateChange (to appear).
[
II
1 . MOLECULAR BEAM EPITAXY
3 Faculty and Senior Staff
R. Allen J. E. Greene . . Stillman
A. Y. Cho M. V. Klein B. G. Streetman
J. D. Dow K. Mon N. Wada
H. Xorkoq
I Graduate Students
IK. Cadien W. Kopp A. Rockett
T. Drummond L. 1itkowski
I 1 .1 BE Growth and Characterization*
I During this period, a considerable effort was placed on the preparation
and characterization of single and multiple period modulation Ioped
(Al ,Ga)As/haAs heterostructures. A theoretical study of these
heterostructures was also undertaken and a good agreement with experiments was
i obtained. Segregation effects of Sn in GaAs grown by NBE was also
investigated in detail.
3 Single and multiple period modulation doped structares with mobilities
several times those reported by other laboratories were prepared. Transport
properties of these structures were studied at low and medium electric fielis
and at room and cryogenic temperatures. The maximum room temperature
mobility, 3,400 cm2 /Vs, was obtained when the AlAs mole fraction was about
33I. In addition, the single period structures so far appeared to exhibit
This work vas suported b the Joint Services Electronics Program(U.S Army, U.S. svy, .S. Air ;orce) under contract 1-OR-79-C-04. anT bythe Air Frce Office of Scientific Research under contract AFOSR 0-9014.
!1.. . . .a .1. ' - -
2
1. '0L~TAR3EA'4 EPIrXY
large~r roomt teTD-3-qtUre mobilities than the multiple periol ones. kt 7" ( qn'i
~)~,however, iomparable results were obtainei. 77 '< mobilitites of' 100,YJO1
anl 1) '( mobilities of about 200,000 - ? D,000o &/,vs were obtaiqel
both with single ani m~ultiple period structures. If the AlAs mole fraction is
mie larger than about 0 the jirect-indirect transition, the ele!-Ctron
mobilities steliily 1decline9 uD to the klAs mole fraction of l100'-.
Se--condary ion mnass spectroscopy (STAS) has been used to investigate the
accumulation of Sn at the arow-ing film surface ani the corresponding Sni
lenoletion near the film- substrate interface luring ME lenosition of Sn-loped
,iAs. Sn surface c:oncentrations -9Sn more than two oriers of magnitul-i larger
than the steady state Sn loping level "~)hv enosre. S in f-iMts
g.rw at anproximately constant As/,a surface 2oncentrations was founi to
inceae, hie Sn remainedl constant with increasing g7rowth temperatures for
T ' 53-0O". Both ;3 and dnecreased -at higher T5 values. These results,
as well as those reported by previous investigators, havre been explained by Sn
surf'.ace segregation during deposition in which the segregation is iriven by
the relief of st-rain energy lue to the incor-Doration of over-sized 3n Citoms.
'alculatei Sn Drofileas were shown to nroviie a gool fit to experimental data.
!2'ransnort Oronoerties of *-eterojunc ti4on Layers*
'ransloort properties of -to'dulation doped structures un to ain ee
fieli strength of 2 'c/n were inviestigated. The differentilleto
mobility iroos iuite rapil'j 'io to 200 VI.- ani then stays about c:onstanlt '1D
to 2 'JcVm:. '-figher electric fields cause negative lifferential resistance
eiffacts which f can result in nonun 4form eleC tric fields. Dnce the field is
nonunifort, no reliable wf\") dati c an be obtained. The electri:! fields in 1 Li
jat e norally-off' ?M-s are e--stimtae to be about 2 1 kV/Cm; therefore , our
results shoull be applicable to predicting device performance. Structures
with an extremely high low-field1 mobility showed substantial decrease with
eleactrict fie- LI., partic:ularly 'At 1low tempe rqtures . 1-nerallY the ro ot
t:?Tpersturs mobility. 3,4)0 =2,"1s. WaS Taintained 'AD to 2 <V~cm. *4iiever, at
K . mobility at 2 0/m was about 2 - Oda of its zero-fiel. value. The K~
This work -i-s sunnoortd. by the kir Force Offic e of 3Scientifi:: Reseairch anderonrt .X?0SR 31-003 1, the Offti of lavtal Resear-ch uindIer contrac-t
q)0 -~~D% nd b y th krmy Research Iffice unde-r c.ontrac-t 0)k~k
3
1 1. MOLECULAR BEAM EPITAXY
mobility at 2 kV/cm was about 15-201 of its zero-field value. These results
indicate that a differential mobility of about 25,000 cm2 /Vs at 2 kV/cm with
an electron concentration of about i01c2M - 2 can be obtained at 77 K. This
represents an intrinsic speed improvement of 5 and a speed-power product
improvement of 25 if normally-off FETs were fabricated and operated at '"7 K.
Room temperature operation can result in an improvement of 501 and about 200o
in speed and speed-power product, respectively. By "improvement" we mean to3 compare it to a conventional normally-off GaAs FET using a bulk channel layer.
We have done considerable work during this period on the real-space
electron transfer effect, in which electrons in GaAs-.lGaAs heterostructures
transfer at high fields from the GaAs to the AIGaAs layers. The resulting
negative resistance as been simulated by Monte Carlo techniques and has beenobserved experiment-ally. This effect, which has great potential for ievice
I applications, is discussed further in Unit 4.
1.3 Theory of Surfaces and :eterojunction Interfaces*
The leep levels associated with surface defects in compound
semiconductors have been studied theoretically; the results indicate that
Fermi-level pinning and Schottky barrier formation is produced by surface
antisits defects in most cases, rather than by surface vacancies or subsurface
idefects. We also have investigated the possibility of designing Schottky
barrier heights, and barrier heights at heterojunctions and in superlattices,! via the Ieposition of selected impurities. Predictions of intrinsic surface
states and surface core excitons in III-V and II-VI semiconductors were found
to be in excellent agreement with the available experimental data. Finally,
it was found that impurities that are shallow ionors or acceptors in the bulkJ can become deep traps at semiconductor interfaces.
1I
* This work was supported by the Joint Services Electronics Program* (1.S. Xrmy, U.S. Navy, U.S. Air Force) under contract NOOO14-'9-1'O3 2.
4
2. S34IONDUCTOR IATSRIALS AND DEVMJSS
Faculty and 3enior Staff
a. !. Stillman
Draduate 3tudents
1. 1. Took T.J. Roth "A. XE. Tashima
T. S. Low D. S. Ruby T. H. Windhorn
B. J. 3kromme
2.1 ntroluction
The primary goal of this research is to evaluate ' e potential of
cigh-purity mn? and :nh AsP for optical and microwave ievice applications.
3esiles investigating the vtarious growth parameters and techniques that can
influence the purity of the grown layers, an important part of this work is
the characterization of the purity and crystalline quality of the samples
grown. The total impurity concentration is studied through Hall coefficient
measurements at 300°K and at 770K. The origin and influence of growth
conditions on acceptor concentrations are studied through low temnperature
photoluminescence measurements. Deep levels, and the influence of different
growth parameters on the concentrations and energy levels of these centers,
will be studied through DLTS measurements. The crystalline quality and degree
of lattice match are studied through x-ray diffraction measurements. The
residual shallow donor impurities are studied using far infrared Fourier
transform measurements of pho to thermal ionization spectroscopy. This
technique is being used to study the residual donor impurities in high-purity
SiAs and InP. Far infrared photoconiuctiity measurements on LnIaAsP and
in-aAs have been used to letermine the effective mass in these materials, and
this technique may be useful for studying alloy disorder in these materials. 9
.. . .. .. . . . . . ., , .. . . °*
I5
2.' SEMICONDUCTOR MATERIALS AND DEVICES
2.2 HRydride Vapor Phase 3rowth of High Purity InP*
The vapor phase system, constructed with funds from ONR, is now fully
automated and under computer control for added reproducibility. Previously
high purity GaAs (1177 K 9 38,000 cn2 /V-sec, n7 7 K = 1.7 x lO1 cm-3} had been
I grown in this system but comparable InP had not. Difficulties encountered in
growing InGaAs lattice matched to InP led to a more detailed and systematic
p analysis of the system's growth parameters for InP.
Prior to growth, the substrates are etched in situ with a 3.1B HCI in H9
mixture to provide a clean surface upon which to initiate growth. In addition
the Arsine/Phosphine line has been extended further down into the gas stream.
I The growth rate has increased to the point where it is possible to
substantially decrease the total gas flow (by nearly 90%). These lower flows
result in much less downtime for cleaning purposes and a substantial saving in
the amount of source material used. The InP grown in this system is now
comparable to that grown by any technique, with the best material to late
haviag a liquid nitrogen mobility P77 K = 70,000 cm2/V-sec and carrier
concentration n77 K = 5 x 1014 cm-3 .
2.3 LPE Irowth of InGaAsP**
I High purity InP and InGaAsP epitaxial layers are essential for wide
depletion width and low capacitance PIN and avalanche photodiodes and for
materials characterization in general. Also, for small band gap InGaAsP
alloys, high purity material is required to avoid Zener breakdown rather than
avalanche breakdown in p-n junctions. Methods for obtaining high purity InP
and InGaAsP epitaxial layers on (100) InP substrates using liquid phase
epitaxy have been developed and this high purity material has been
characterized through Hall-effect measurements and far infrared photothermal
ionization measurements. Samples with net carrier concentrations as low as 1
x 1015 cm-3 for InP and 2 x 1014 cm-3 for In.53Ga.47 As, with corresponding
I liquid nitrogen temperature electron mobilities of 67,500 and 67,900 cm2 /V-s,
* This work was supported by the Joint Services 3lectronics Program(U.S. Army, U.S. Navy, and U.S. Air Force) under Contracts DAAG-?9-79-C-1016and NO001 4-79-C-04.24 and by DARPA Contract NO0001 4-77-C-00096.
This work was supported by the Naval Research Laboratory under ContractsN00173-79- -019 and N00173-79-C-01 94, and the Joint Services ElectronicsProgram (U.S. Army, U.S. Navy and U.S. Air Force) under ContractsDAAG-29-73-C-001 6 and NO001 t-79-C-0424.
6
2. 3341DNDUCTOR '4ATERIALS AND DEVIT'ES
respectively have been grown routinely.
Both the growth solution baking technique and the growth temperature have
an influence on the epitaxial layer purity. Baking times of 24 to 13 hours
are required to reduce the carrier concentration to these levels.
ioorporation of smnall amounts of water vapor (< .1 ppm) in the hydrogen
atmosphere in an attempt to reduce silicon transport from the quartz tube to
the grow-th solution during baking has not improved the purity, and
incorporation of larger amounts (> I pm) increases the carrier concentration
and reduces the mobility. An increase in the growth temperature has improved
the purity, especially for In~aAs, presumably due to a decrease in the silicon
distribution coefficient as the growth temperature is increased.
2.4 Fourier Transform Spectroscopy of 3hallow Donor Levels*
Shallow donor energy levels have been measured using pho to thermal
ionization spectroscopy in high purity LPE, .Hydriie VPE, and bulk TnP, grown
in our laboratory and elsewhere. This echnique, which involves the
measurement of photoconductive response versus inciient photon energy at low
tamperature -C) and in a high magnetic field C_< 5.5 T) has also been applied
to high purity LPE, inaAs and InGaAsP samples lattice matched to InP. The
minicomputer based Fourier transform spectrometer facility was set up with the
support of the Naval Research Laboratory.
In the LPE and Hydride 70S InP grown in our laboratory and the LE In?
grown by L. F. Eastman at Cornell, the sample purity was sufficient that peaks
corresponding to lifferent donor species could be resolved in the photothermal
ionization spectra. Though it is not yet possible to identify these ionor
species, the two lonors present in the Hydride VPE material were both common
to the Illinois LP3 material, and one of these was also the lominant donor in
the T'ornell 123 material. A third ionor species may also be present in the
Illinois L2 material at low concentration. It is reasonable to speculate
that a donor common to both LPE and 7Z In? may be silicon, incorporated from
the hot quartz -rowth reactors, especially in view of the high iistribution
coefficient for silicon in InP.
"'his work was supported by the National Science Foundation Under ContractN3F D.R--7-231?9, by the Of fice of Naval Research under 'ontract.XY04-77-C-)53 and by the Joint Services Electronics Program (J.3. Xrmy,U.S. Navy, and U.3. .ir Force) under Contract NOOOIA- 79--D 42
7
1 2. SEMICONDUCTOR MATERIALS AND DEVIgES
The magnetic field dependence of the peak positions in these spectra
allows accurate determination of bottom of the band effective mass (M*),
donor ionization energy R*, and low temperature static dielectric constant Eo-Values of these important materials parameters have now been obtained for InP,
g t. 53 Ga. 47 As and In g 1Ga17Ga.4oP. 60 . For these materials, the respective
values determined for these parameters are:
0(m!) - .00796 + .005, .0437 + .001, and .0635 + .001
R* = 7.4 + .17 meV, 4.30 + .17 meV, and 6.28 + .17 meV
0 = 12.06 + .13, 13.5 + .5, and 13.4 + .5.
The peaks in the spectra of the ternary and quaternary alloy samples are
much broader than would be seen in GaAs or InP of similar liquid nitrogen
temperature carrier concentration and Hall mobility (n 7 7 = 3 x 1014 cm-3 , V.-7
Z 68,000 = 2 /V-s). A mechanism, not present in the binary alloys, where
compositional alloy disorder results in a different local alloy environment
for different donors, could account for the excess broadening seen in the
ternary and quaternary spectra. The degree of this broadening might then be
1 interpreted as a measure of this compositional disorder.
IIIII
8
?acuity ani Ienior 3taff
?7D. Joleman J. 'r. Tsn J.R~. T~
k. . )eTMpIe T. -2. Je-3 -y.
jraduate S3tuierits
* ?ejj~~ ~T. Teap J. )ammuisen
~. ~~abeT,. 3Decht
3.' -,xcitation ani "3tuiy of 'Aighly-",citeed Vibrational States of '4Ioleculas
4 -irectinn inli Significance
Thatr~ trnis iai ita '00)(t3 in t'he mi'2 owa'rc s0.3'troscony of o
?,1-3:,-y Le3 r&Ls of T,013cu1Ir -.ISes. )Tti~ajjy ~p- f~ :3r 7 -S
empioy~i t 1 ?e L- f'i Iam e n t-a vri"rational- ro tational -enn -- y 1 ev 13 ) f i
M. t uI S. Io t ban! s ani ex' i t t i sa te toi- f 7t olue 1 ica o to t ',I,,171i -2, it iauum" hive been little i s - - inq I-iaitut -3 1 -rnn i -s anL tao
es -3troscopy ani st-ia inis are relativel7y -nexplorel.
L3 t~l thr',13t Of bi oktO 9t',11 ':!V 0isis~ ~ OPulatil-13
ai nolv-iZa.tivns o-)f mole1 a' Iar gtates upn to(- the "J.I- '23 L -:on1t in for'
no tential L'a~ -n reiv:ncy t-iiable laser,3, nlnerooti .e ail ~ity
A-l "ia'71113ti- e h iis t'-i j i a.' 311-ge of t',I_? ~
_7 ' _ .1 an by 2-oritrby tb IThit3rYie toie l
9
I3. QUMITUM ELSCTROI'I -S
3.1.2 Summary of ?rogress
A pulsed multimode '02 laser has been used to excite a hot band
transition in room temperature *0S gas to a state .fith three luanta of
vibrational excitation. A transient microwave gain *as observed and is a
measure of the excited state population lifference between Z loublet states.
ssThis inherently small signal has been observed, without signal averaging, by
using a microwave reflection bridge which is transiently unbalanced by -n
infrared pulse exciting the gas sample in part of the bridge. An infrared
pulse truncation shutter and lower noise microwave detection system are being
aided for direct obseration of the excitation relaxation.
3.? -nergy Transfer Orocesses in 'Hlectronically -xcited Atoms and Molecules
3.2.1 Low-?iell Ionization Coefficients of Rare Sases
The electron ionization and excitation coefficients for Ar, Kr, and Ke
have been measured in the low E/p region (1-12 V/cm torr) using a rift-tube
! apparatus. Discrepancies between previous measurements have been resolved and
shown to be due to photoelectric emi3sion attributed to the rare gas
metastables. Usiig this information a revised set of self-consistent near
threshold inelastic cross sections for each rare -as studied was determined.
1 3.2.2 Kinetics of Interhalogen Diatomics
The spectra of the T - k311 states of 1? and 1l were obtained ani forthe former showed clearly resolvable rotational structure. -he lifetimes of
the T states of 2F and ICl were also estimated from pulsed discharge
experiments. The vibrational self quenching rate of the 1"l A 7 state "was
estimated and also the argon quenching rate of this same state -as measured.
3.2.3 Multiphoton interactions
In an unrelated study, ;e have been eXplorinj some effets associated
#Lth the multi-photon nature of near resonant light-matter interazti,"ns
I involving more than one field and a set of coupled energy levels. 'i- have
developei both algebraic algorithms and a graph-algebraic .pproa-ih for solvingthis 2lass of problems in the aliabatic limit. These techniques yore then
Ipplied to the specifi case of resonant second ha.-oni generation in
three-level systems, an example of which o'curs in magnetic resonant :zses.
10
Thits stuily Sheois t"hI.t it is possih - to hiave a very hi--h coonversion, ,-:rt~r
thain 50) , Oumt ene-r-Y intr the seconi na-roaij uniler a el.tei, b'r not too
st ~iv.~nitios The re 4suIts otis terti4 Cr iil cari
-I fothtn. zli~ation ie are io-# applytn- the ihnsofti3ti
STaCifZi3-r syt eT. ini oriir to ietarmi fa :asibil ity of future p'c~ imeqts.
33 'xie 3t-ite 7heristrY in 'iaSeS
ThIe reer rojezt on, Th-Citel Stat-- 7he'istry in 'lases has vJ l'rn
the oast year frou the initial iis' Iqrte -.-ork on plasma e.tching toin'.l
plasmta anneaiing aoal gro-orth 'by elornimpazt) and -tUit4photon IJ'ssoi ti on
of ~ i 2t-n Iuc to : b :? a r 4 ase so13S ?3 as to --ro -., thin fil T3s o0 0i.o or
The zre vious renort Tantioni thie -.viien:!a for the of naieinsin the -2tzhii- of _-i or 3i)- wit' lishaie in-a ahl on h~h
e't2h rait. of thesea materia'is -i-i jre3atly enhancei ~e the .3ampl!e %rs biuasel
to a fe)d t?-ns of voits positivea ii th :esei' t laIn po ential. - -4 hae
3iflce :orre IateI thi behiio-,r t o Lvn-muir prob~e chiAra!,t~ris3tII 'Sf 1 "
inlicate ~ ~ ~ ~ rI a:!sr e~tvl har-ai zarir. Tis behav-ior i:3 only 17 ie
r I isohir~s in ~'~; ischaraes I thle to znrninlt~tie
i: o, io nt - c h ib it this be- h-a v ior This is :orsistent ri th thie fie t thiat
17?- his a Tu-ch lar-yer attichtent cross seCtion) thanl the other g7,ASes -'Z
Thease rqs'11lt3 hiave been ac tlfo- public-ation 3.' notuaev h--
ilentity of the ion is still unknown ani aTiaits the inastallation of the. T.as-5
sce~trotetar on the etchincg systn. This iJs-3 lle for -,ompletioi in the
rthctangyear.
~ ?iata \nealin
LTa this' chase1 of Ios .'r, Ie hae shor0-n thaIt one Ian Ise I --
abnOra'. Iisoh'a'rc;e -;I A 70-f prI3ssure 47a 'tP i217 he ILuM 'It ht3 to
to obtaina -Ilarge Irea e .eutron beim -WiIo, in, turn, iaiic i-!ei to ilnesi
lrge ar-eaS Of ion3 iMniantel SezT-4oni-12tOr naItrl a R . l hs e
r.111'eoc <ot,' the tpes4 anneaei Li thIis 'tner for7 tiT-?s leIs t11-
5C~~flflI3 nave 30t; .!sjst 4it.e 2Ofli~t5ntL7 1715 ""In "hoge anea.ciina
ovn )r 3) tinuteS. -t'.th3 cresentl tLrni il are3 att~mpt;_In t') infe r
I11
I 'I. U .NTU4 ELECTRO31CS
temperature of the sample by measuring the change in optical reflectivity as
I the sample is irraiiated by the electrons. This is a variation on the
technique iescribei by Olson, Kikorowski, Roth and Hess r .
3.3.3 Laser Thin Film 'rowth
The objective of this research program is to investigate the chemistry
and optical properties of excited atoms and molecular radicals proiuced by
ultraviolet photoiissociation and the interaction of these excited species
with a surface to generate a thin film. 3pecifically, an ultraviolet excimer
laser is being used to photoissociate inorganic molecules that contain a
I semiconductor atom (such as SiH4, GeH 4 or SiCl). The spectroscopy of these
laser-jeneratel plasmas as well as the electrical and chemical properties of
I the resulting Ge or Si films are under investigation.
To late amorphous polycrystalline Si and le thin films, arouni I cm? in
area, have been grown by photolytic laser chemical vapor leposition 'LtVO) at
room temperature. Dilute SiH4, (CH3 )43i, or GeH, mixtures in an ine-t carrier
I as wiere photodissociated using krF (1 93 nm) or KrF (241 nm) lasers. Filt.
thicknesses of up to ).5 Um were obtained on St., substrates at leposition
rates ranging from t to 10 A/sec for active gas partial pressures between 5I/an 310 torr and inciient laser intensities up to 25 '4W/cm-. Absorption
I coefficients of as-deposited Ise and Si layers were 2.3 x 10) cm- ani ;.I x
1I cm- , -espectively, ani sheet r9esistivities ', 1O6 2 /0 have been measure I
Tncubation times ani initial growth rites have been leterminel in situ as a
function of V photon flux from 4e-Ne laser absorption measurements. Xl
concentrations of up to several atomic percent have been incorporatel into "e
films by the simultaneous photodissociation of 3eq4 and (Hf3)3Al.
?inally, the chemical characteristics of these films are being stuiiel by
I infrared absorption Auger spectroscopy, ani x-ray scatterin3.
I eferences
31R! T. . 3. Thynne, J. ?hys. Chem. 73, 1536 (1 169).
I '! . L. Uson, 3. A. Kokorowski, J. A. Roth, and L. 0. +ess, "Investigation
of Laser In'iced 3olii-3hAse Crystal 'roth by Time-Isolvei 3ptii!
Reflectivity," ?aper ThC-3, CLZ30 '31, Washington, D.7., June 1 931).
12
?aculty and 3enior 3taff
3. J. Streetrnnn (.ess
3rauate 3tudents
S. 3ainnarjee 11. DeJoule ?*qlrtin
A3ha4.tachariryq r1- 0. reer T )berstir
3. 3orr.ougahs B. Jones 'I. 3h ic hij 0
K. 3rennan H. Ka f ka K. Soi'a
4.1 1-itrod1uction
PL',i3 rqsaar,!h 1n0ts3t'l.y Of blsi: UrOpOertL-? Of ~~'
-)'Loi pfiv~ roc-3sn'i, ani new 3~ie~nn:,ept3. 34oth the 'mitti-. arnl
exn-rirmentalm-3'nthoI3 r3 ,r D 'no ysI in eacoh of these tmoi~
-3-'crininj am -triety of liot eleaztron Dhenomer'm 'mn-. thi:?i- -ff-3cts on pr:?sent iii
l'tm' evii'm perforuancq, 'mSnPeci'1ly "I onne-m:tiori -ithq Tol'.iatton ioping.
WIe 'Ire StUiy-LIJ ion inlant'mtiOn ani 'mnne'iing- of :3i. :ml Ir oonumis,
jncluiing laser 'mni 3lectron beIMm DroCessiftg,. Te excplriTentlnit. Vi 4:S
2I~li -m ! e'Tai-n'tOi 1 t of Ie-l1e3v1? iM pUr it nt i -I mt 3 :'1. r,* 3 i
impolanltation ani arinealin-. ISq!. 3tS If "I i 3 4 rc ' 10'19 1
:nlt'mboratioi -.fithi other u.nits, partiilmr.V th'm3 staiis 'if nmeil ri
I,13
4. SEMICONDUCTOR PHYSICS
4.2 Laser and Tlectron Beam Annealing of Semiconductors*
We have performed both theoretical and experimental work on pulsed laser
annealing of amorphized Si and swept line electron beam annealing of Si and
3aAs. In collaboration with J. T. Verdeyen, oe have also employed gas
discharge annealing to process implanted Si.
In studying pulsed laser melt annealing of Si, we have examined heating
I by band-to-band excitation (1d:glass, 1.06 prm) and free carrier absorption
(207, 10. prn). The technique of finite difference equations was used to
solve the one-dimensional heat conduction equations for the solid and liquid
phases, taking into account the temperature dependence of all the parameters
* involved, especially the substrate parameters, to give a realistic model. 1n
I the 30_ case, we have calculated the temperature variation of free carrier
absorption in n-Si. We have taken into account acoustic leformation potential
Sscattering, optical deformation potential scattering, and ionized impurity
scattering. The deformation Do tentials were adjusted to fit the
-experimentally obserrved values at 3000K. ;ie have investigated the "thermal
runaway" behavior and calculated the threshold power density to cause thermal
runaway and surface melting of Si luring COo laser annealing. For the pulsed
Nd:glass laser annealiag, the melting model is supported by time-resolved
reflectivity measurements using a He-le and an Ar laser. We used 3
2hanneling patterns to study the juality ainl orientation of crystalline
r93rowth. To investigate the electrical activation we 1etermined the Thange
in sheet resistance of the implanted material before and after a laser pulse.
in each case we found excellent reordering of the amorphized Si.
Successful recrystallization of amorphous 3F) + ion-implanted Si has also
been demonstratei by the swept line electron beam (SLEB) technique. High
I percentages of electrically active boron were observed (ith a minimum of
spatial reii3tribution. Active boron was also found in the amorphous-
crystalline transition region, an effect not observed in low temperature
firnace ann aled material. "4e are currently employing the SLEB method in
I recrystallization of deposited 3i layers over Si),.
This work was supported by the Joint Services Electronics Program 1J.S.Army, U.3. Navy, and U.S. Air Force) under Contract NOO14-'7- -3 , by theArmy 7-3search 3ffi~e under Contract DAA2 2q-30-'-0011 , and by the Offi~e of' Naval Research under Conract 'OO0-'o-C-O306.
"m1
14I. SEMICONDUCTOJR PlIT31713
.3 Stulies of Trairspor, "roeertias in 39mi, nIuctors*
-3 have investi~yatal carrier transport in se'niionductor eter o acto
layers ani ievelopel th-3 theory of' mobility ti TIoIUlatiOn-4ope'l struc:tures.
7Th3I resul1ts of this theory zecre 2omparai -.ith ecperimalnts ani -),an! to v
e'c-I lant a ae em. eIt in erai ttnrerqture n~ -1 i n i i-l9 gsav-c-ontinueal to .iork on the ::oncept Of' real SpaCe trans3fer a-Id hae hon
experimentally the valiiity of this concept 'the spill out of electrons f rc,,n
Dotantial -walls) . We belie-ve that this ef fect has haigh ievic3e npote-ntil ani
al so -tany n t -3re9s t in - b as a c2:onsequ-3nces fo-)r the9 t ran spo rt ina lye're I
struc-tures aid suparlattices. The -3ffect can also be --ineralizei1 as i real
soa::e analo-, to a k-space ffc (the 'jinn affect) and this n eralizationY
eqAls to several other Dhvnortena Such -is "resonant" impact ionizatli becauise.
ofthe bani I'~ iscontirwity in heterojunction laye:rs.
*,ic h-ave -etpaniei our :4onte 'arlo nrogramn Iincluiing the? banil structure -is
calcu.latel 'rnm the .3miricAl oseulopote ntiai Tiathol) to incluie t-asnort in.
341.lin ani IaAs. f-e -ers able to alult ionizati on rates i theseq
matferials ;ithout too mnany asSUtnntions. These ialculations inl oe-% the-3t ii~)tes uato f~r hn one 2onluc tion bani. 'Ichae ho
-'at -h"3 nIi~er 2oniucltioni band Ls cruc:ial in siliconi at 'h-igh ltiz'-'s
is theiy are c2urrently r-iachei in small ievicis.
Je nam-re c-ontinuel to 1ievelop numeric al rroaruns to Tmolel -ian snor-t
ahinom-ena in * a r ious s em i con iuc to r lics ani arei cu r ren t Iy L ilu Iin
transien 't Oheno-nena $overshoot/'unlershoz of velocity) in the oneration * f
char'?e cousici leviceqs. ?art of' our theoretical Fork -was ievotei to th
possibilitie3s of ballistic! transport and to the f-ieli ienn iena -i
mneteroltysrs at lo;,, electric fialls.
"Ii fOrIk Y13 31Tn:Ortei by the Mhint lervi es ~etoisnor'').arc .%I.- 7Thn)mi "trt 'MYYI I.3-~' vvy
-1-)1 !-)--JV3 ani byt Amy 'hsearc!h )fIf i In Icer ot rac X
II15S-4. SICOnDUCTOR ?MIYS
4.4 .Materials Studies of InP*
As a part of our continuing studies of ion implantation in compound
semiconductors, we have studied InP carefully during this period. The
redistribution of the compensating opants, iron or chromium, in semi-
insulating indium phosphide has been studied using secondary ion mass
spectrometry (SIMS). Annealing with silicon nitride encapsulation results in
impurity accumulation within the first 1000 ' of the surface followed by
lepletion extending to a depth of - I rm. Profiles resulting from the
implantation of "neutral" elements (1e, B) exhibit accumulation at the surface
1 and also accumulation at the projected range peak. The profiles an be
explained in terms of gettering of the compensating dopant to defect-rich
I regions.
We have examined polycrystalline and single crystal, liuil encansulateiI ,ochralski '.SC) grown TnP for evidence of boron contamination using 311S and
photoluminescence. ?recipitates of boron or a boron compou-ni have been founi
j in In? grown by the LEC method with boric oxile (32_3) encapsulatioi aad
pyrolytic boron nitride (pBN) crucibles. The tensity of precipitates appears
to be solely responsible for the observed phenomenon. At present, the Source
of boron contamination most consistent with our observations is the result of
interaction between 3903 and pBN. A diffusion coefficient of' D(759°0) < I %
13 " cm2 - for boron in InP has been detarminel in this work.
We have observ-ed cracking in chemical vapor-deposited Si0) encapsulatilg
layers on In? wihen these samples were annealed at temperatures above -50"?.
Using optical microscopy, scanning electron microscopy ani A,.ier electron
spectroscopy we have studied the ietaile4 nature of this forn of film f~ilure.
Thermally induced stress between the film and the n? does not fully ac-count
for the phenomena observed. The cracking apparently results from an
interaction between mechanical stress, chemical effects ani possibly defects
I in the 3i), layers.
II Thi3 work was suoport-at by the Joint Services Slectroniis Prozram "J.3.Army 1.. 3 favy, and '.S. err ?orce) under *ontra-t 'Y10V! -- >-)4?14, a-n by
the 5ffie of laval Research under Contra3t 01.-76-7-030-5.
!-
161.3EMI'.ONDyJCTR ?TqyST ,S
?or ionz L-Tplantatio.13 Of In? that are 'hi-hly io'ninatei by elec:troni-c
S toopAac-, the starniarl hussian approximation proviles a oocr fit to the
nbservi experimeqti profile. UJsing four ToTaents basel on tiaoreticil
-onasijerations, -4- have 2onstructe1 profiles ising a tertiaated 3,1,e-wortlh and
?-:)arson I iistribution. It was found that neither the !'Igeworth nor Pearson2 I
provi-ies the accuracy that -.-ouli be -3-cectad f-:on 9L four moment listribution.
The -tomtents ofl the experimental profiles have been empirically1 letarminel.
7:iese moment3 tozpether :iith the Pearson IV listribution qrere sho~m to fit
exparimental profiles over 3t least two Ordlers of magnitule.
.ith the backyrounl I esc rib al abov e, wte ar-urnl1suyn h
effe:cts of 3e, '4, -ini Ri implants in in?. Theise studies ilclu-ie iMpurit7
prc filin-, elcrclactivation, -and interaction with lefects arvlI bac-.kround
vnu ri t ie3S.
I17I
5. THIN FILM PHYSICS
a Faculty and Senior Staff
I J. E. Greene
Graduate Students: I
S. A. Barnett D. Lubben L. Rivaud
K. L. Cadien M. A. Ray A. Rockett
R. E. Klinger L. Romano
I 5.1 Semiconductor Crystal Growth from the Vapor Phase:
Ion-Surface, Plasma, and Laser Stimulated Reactions
We are studying energetic particle-surface interactions and stimulated
* gas phase reactions which control the nucleation and growth kinetics,
chemistry, and physical properties of compound and alloy semiconductors grown
3 from the vanor phase by TUHV ion beam sputtering, plasma-assisted chemical
vapor leposition, and laser-induced chemical vapor deposition. The common
feature of these techniques is that crystal growth proceeds under
non-equilibrium thermodynamic conditions through the production of highly
reactive gas phase species: excited atoms, metastables, radicals, and ions.3 Such species transfer energy to the growth surface upon condensation thereby
altering the surface reactivity as well as adsorption and adatom liffusion
kinetics allowing film growth at lower temperatures, a wiler range in
controlling doping concentrations and tailoring film properties, and the
growth of unique metastable materials. This work is being pursued from both
an analytical and an experimental point of view to establish a ietailed
i understanding of fundamental film growth mechanisms. We have recently
published two invited review papers r5. 1 , 5.61 in this area. Results from
this research have a wide range of applications in adlition to crystal growthI
18
~ncli'n the ac tiv e :esearch fialls of reactive i on et hng n on bean1it*-o-,ranhy, *irochemical analysis , pl a sma ch e".ist an 3aa~
ilteractions3.
5. ', ro-wth of *4ig7h Duality S-ingle 7rystal. la3 Js
by 3Soutter Deoosi tion
"g7o a 4al 'aAs fil-as have been -,rown onl '100) 'r-iopel 3aAs substrates by
3,outte r lenosition "5 . 2 1 from an xqion~ed h.As tar-et. 7n in tential 'Iv I o Ze
samnles were high resistivity 105 to 1ID0 Q2-cm n-type -d-th room temoe-rqtur?
remorte3d -aobilities in hi-,h resistivity IaAs. iesilual oxygen and carbon
contminaionwas Minimized through the use of a liquid nitrogen c2ooled shroud
3urrouniing the discharge during deposition. Tccess arsenic -was provilei to
the growinga film from an effusion cell. Film doping -*as accomnlishediin
ai cher an evaporated source (An) or the addition of a gas phase impurity "1)"
during spoutter lenoosicion. IIn- loped o-tyne films were grown with room
teDeratiAre carriear concentrations between 1.7- x 17)17 And 2. 7 x I1 3 m and
co rresnjo 1ndin holea mobilities of 21,0 to q CM_/'ISec. Thee al. geevr
well W-ith resu-lts for both LP7, and 'ABE films doped to similar concentrations.
"Ihe S dopqi films were ri-tye f'see section 5.1.2). in all cases, the jopant
:noornoration probability was controllable by varying the negative bias on the
2.'.?on.3ombardment gffects on "Elemental Sticking Probabilities
.e havi shown nreviously that low energy < 70 3Q _27 ion bombarldment Of a
g r o wi ng f;im c-a n pl ay a large rol 1- n controlling ele-3m en ta 1s3t ic king
probabilities, allo-wing an extension in the girowth temperature? rangae over
nric stoiChiomjetric 1174v COanounds and alloys can be fornel as we'' as,
allowing the growth of non-eiquilibri-x- phases K see section 3. ?). T"his ocr
through a varietly of mechanisms including self-ion tranning, en'_hanced surfac e
reactivrity, snutter removal of alatomis, ind iac-imltdiissociati-!-?
chemisorotion. For a iiscussion of the3se -mechanisms anirltlec~in e
the review- article by Ireene r,.-71,
19
5. THIN FILM PqYSICS
Sulfur concentrations of up to 1.2 x 1021 cm-3, as measured by secondary
ion mass spectroscopy, were obtained in sputter deposited single crystal 2aAs
layers using an H2S glow discharge source. The incorporated S concentration
increased linearly with increasing H12S partial pressure but was not affected
by varying the Ass/'a flux ratio between 12 and 10. The S incorporation
I probability varied from 0.02 for essentially thermal species to nearly unity
for sulfur containing ions accelerated by a negative substrate potential of
325 V. The enhancement in S incorporation probability was not due only to
implantation processes, but was also related to the strength of the 3a-3 bond.
1Iery sharp doping profiles with no indication of surface segregation were
I obtained for programmed doping steps (for comparison, see results for Sn
segregation in XBE GaAs, from a study carried out in collaboration with
I Prof. Morkog, summarized in Unit I).
5.2 3rowth of Single Crystal Metastable Semiconductors*
As reported last year R1, R21, we have carried out the first detailed
study of the growth of new single crystal metastable semiconductors. The key
feature in stabilizing the growth of these materials was the controlled use of
low energy ion bombardment during deposition to modify elemental sticking
probabilities and to enhance adatom diffusivities. During the past year, we
have investigated the crystal growth, thermal and temporal stability,
electrical properties, and the thermodynamics of a new subclass of metastable
j I semiconductors, (IIl-V) 1 x(rV)1 .Evitaxial metastable ('aAs) x y ompositions between x
).10 and 0.75 have been grown on (I00) 3aAs substrates by ion beam sputtering
in an ultrahigh vacuum system. Electron channeling, double crystal X-ray
liffractometry, and X-ray topography analyses indic:e that the films are of
very high crystalline perfection. Layers with x-values up to at least x =
).47 : ere in the zinc blendp structure. Either n-type or p-type conduction
with n and p varying over several orders of magnitude could be obtained by
varying the film composition, the growth temperature, and the As overpressure
I luring deposition. The equilibrium 3AAs-3e pseuiobinary phase diagram has
This research was supported by the ioint Services Electronics Program(U.5. Army, U.S. lavy, and U.S. Air Force) under Contract NOOO14--'-D424,and by the Department of Energy.
I
* ---
20
. THI. FL,4 HY3JZ
been determined.
Recent thermodynamic -calculations combined -with differential scanning
calorimetry measurements have provided a more detailed understanding of both
the growth and "he thermal stability of these unique materials. EXAFS
measurements, c!arried out in collaboration -4ith Prof. Then of the Aetallurry
enartment at TJgTJ'. =ombined with STE4 and thermal annealing results have
sho'wn conclusively that collisional mixing due to low energy ion bombardment
can be usei to retard phase transformations. "inally, we have b e a un
collaborations writh ?rof. Klein of the Physics Department at "JT1c ani
Prof. Racoah of the Physics Department at Tj7I7 to investigate optica!
oronerties of metastable (,As 1x e. and "'Taob-q xle.. alloys using Raman
s3octroscopy, -electro-reflectance, and Dhotoluminescence.
5.7 Reactive 7on etchin- of 'aAs*
The first Jetailed study of tne reactive ion etching of IaAs r5., w~as
carried out by our group during the past year. )f primary i s the
effect of ion-surface interactions on the mechanisms and Iinetics of surface
reactions. E:xoeriments Were carried out in pure -al.2 and OFj -ischarges, as
w,11 as in dilute '~ClF0-), and -F discharges containing ?D mo \r.
Anisotropic, etching with removal rates, R, of up to D.9 irn/min have been
obtained in reactive discharges operated at 40 retorr and -3 <V, w'hereas the
physical sputtering rate in pure Ar discharges operated under the sane
conditions was only 0. )1 m/min. k oombination of opti:al eM4ission and
absorotion soectroscooies have been used to show that in both pure and lii-ute
-hlorine-containing reactive iischarges, physical souttering of atomic la ind
As is not the primary etching mechanisM for aAs although ion bombardment does
play an important role in the overall orocess. Rather, etching occurs by the
formation and 1esor-otion of volatile a and As chlorides. -n reactive
fluorine-containing discharges, the rate limiting step is the removal of
non-volatile DaF,.
his research was supported by the Jo ,t 3ervices ',I :tronics Program(.3. Army. U.3. lavy and U.3. Air ?orce) under Contrar .;i -II-"--4?4,and by the 3, Research Laboratory, East ?ishkill, ew York.
I21
5. THIN FILM PYSICS
5.4 Laser-Surface Interactions
I In collaboration with Prof. Eden of the Electrical Engineering Department
at JIUC we have initiated a program to investigate film growth by laser
induced chemical vapor deposition (LCVD). In initial experiments we have
studied the growth of amorphous hydrogenated Si and Ge thin films, - 1 cM2 inE area, by photolysis. Dilute SiH4 , (CH3 )4 Si, or GeH4 mixtures in an inert
carrier gas were photodissociated using ArF (193 nm) or KrF (?48 nm) lasers.
I Film thicknesses of up to 0.5 .im were obtained on 3i0 2 substrates at
deposition rates ranging from 4 to 10 A/sec for active gas partial pressuresE between 5 and 30 Torr and incident laser intensities up to 25 W/cm2 .
Absorption coefficients of as-deposited Ge:H and Si:H layers were 2.3 x 105
cm- 1 and 5.9 x 104 cm- ', respectively, and sheet resistivities > 1O6 £/c have
been measured. Incubation times and initial growth rates have been determined
in-situ as a function of UV photon flux from He-le laser absorption
measurements. Al concentrations of up to several atomic percent have been
incorporated into Ge films by the simultaneous photodissociation of 'eH 4 and
(H_5) 3A1.
o A parallel effort, carried out in collaboration with the Physics Division
of Tastman Kodak Research Laboratories, Rochester, New York, has been
established to investigate liquid phase regrowth of semiconductors by OW laser
annealing. We have recently demonstrated the growth of epitaxial Ge/.aAs
heterostructures by scanned CW argon laser annealing of 440 nm thick amorphous
Ie films on (100) semi-insulating GaAs substrates .. Depending on the
incident laser power and scan rate, two modes of film regrowth were obserred.
At low powers (between nu 1.6 and 4.0 W for a beam diameter of n 40 Un) and
scan rates between I and 400 cm/s, polycrystalline Ge with a (100) preferred
orientation was formed by an "explosive" crystallization mechanism. At higher
powers, and over a scan rate range of 20 to 400 cam/s, single-crystal films
I containing some dissolved GaAs in solution were obtained by liquii-phase
regrowth. Typical film resistivities p were as follows: as-deposited, 130
Q-cm; polycrystalline films, 3 x i0 -2 -cm; single-crystal films, 9 x 10- 4
IaN-cm.
Ii
22
5. THTI ?T.A F-YSIC3
Refa3rences
R1 J. L. 7Zilko an-I J. E. Greene, "'Irovth an Phase Stability of Tpitaxial
'Ietastable Thibj -xii Films in GIaAs - ?'r 1: Crystal Growth,"
J. .ppl.- Phys.- 5 1. 1549 (19,90).
R12 j. r. Zilko andI J. E. 3reene, "',rowith and Phase 3tability of 3pitaxialAetastable 1nSbj ,.cBi, im ng~s- at2 Phase Stability,"
J. Appl. PDhys . 51, 1560 (1990) .
23I6. MICROWAVE ACOUSTICS
Faculty and Senior Staff
R. Adler S. Datta B. J. qunsinger
!I Graduate Students
I K. Anderson D. Janes S. Mahon
F. Fliegel A. Lentine R. Miller
M. Hoskins S. Wilkus
6.1 Introduction
The objective of the microwave acoustic research is five fold:
I. to use the interactions between acoustic waves and free carriers in
GaAs to make charge transport ievices,
2. to use traveling wave electric fielis similar to those produced by
acoustic waves to evaluate semiconductor electronic properties,
3. to integrate semiconductor active devices and microwave acoustic
devices,
4. to analyze surface acoustic wave transducers and reflectors from
basic material parameters, and
5. to investigate mode conversion techniques of line acoustic wave
1 generation.
Progress toward this objective is described in the following paragraphs.1I
24
5. MICROWAVE ACOUSTI"S
6.2 Electron Transoort in Piezoelectric Semiconductors*
Acoustic surface and line waves propagating in piezoelectric
semiconductors irag the carriers and produc- an acoustoelectric current. When
the electric fields produced by the waves are greater than the souni wave
velocity divided by the mobility, the carriers are dragged at the sound
velocity. 5'urther increases in acoustic Dower do not increase velocity or the
acoustoelectric current. The carriers are then said to propagate
synchronously and the acoustoelectric current is saturated.
The charge transport devices envisioned in this research depend
exclusively on synchronous carrier transport. Extensive study has been
carried out to deteruine the nature of charge transport in the presence of
very large traveling wave fields. The effect of traps on synchronous carrier
transport is of particular concern and Cr doped gaAs has been studied
extensively because of its high lensity of traps.
The saturated acoustoelectric current, caused by the nonlinear
interaction of photo-generated electrons with large amplitude SAW induced
electric fields in Cr loped gaAs, is found to be too small to be explained on
the premise that all electrons move synchronously at the SAW velocity. This
finding is based on the results of large signal acoustoelectric measurements,
for 3A.J both internal and external to the gaAs substrate. via a separated
media device geometry. The expected linear dependence of saturated current
with SAW velocity is not observed when SAW vith different velocities are used
to induce the saturation. It is concluded that some mechanism other than
complete carrier bunching is responsible for acoustoelectric .urrent
saturation in Cr:GaAs il ].
During the last year, we have iemonstrated that electrons -n be
entrained by an acoustic wave propagating along the surface of a Cr-dopel JAs
crystal. This result was surprising in view of the weak piezoelectricit'y of
3aAs; it was made possible by the unusualproperty of high-frequency acousti-
surface waves of proviling a concentrated fieli over a narrow ::ross section.
This work was supporte d by the Joint Se.vices Electroni s Program uAnler,.onract NO01 400011)_.
25
6. MICROWAVE ACOUSTICS
One can visualize a new class of high-performance charge transfer levices
based on acoustically entrained carriers. Such devices would have no gates
and no complex multi-layer patterns to interconnect the gates; the region
along Ahich the electrons travel would be free for pickup electrodes. Along
with the simpler structure, these devices would have negligible transfer loss,3 making it possible to obtain a time-bandwidth product orders of magnitude
higher than that of present-day CCD's.
I 6.3 X New Method to Measure the Mobility of kmorphous Semiconductor ?ilms*
The measurement of mobility in high resistivity low mobility amorphous
semiconductor films can be difficult. We have used the acoustoelectric
I current described in the last section as a measure of material mobility.
The acoustoelectric method of determining carrier mobility ii in
I semiconductor films has been modified to permit its application to
high- resistance, low-mobility films. The conventional method requires
knowledge of the acoustic loss caused by the mobile carriers; in
high-resistance films this loss becomes too small to be measured. We show
that the required information may be derived from knowledge of the acoustic
power and the device geometry. Our samples were amorphous hydrogenated Si and
Si. 6 Ge. 4 films on nonpiezoelectric substrates, separated by a convenient air
gap (12.5 um) from a LiNbO3 slab carrying surface acoustic waves. One sample
had 0 Q /0 and = ).08 cm2 /V see; another sample 1 0 10 Q2/0 and U = 0.5 cm2 /V
l sec .
In amorphous materials the drift mobility of carriers is usually measured
by observing the time-of-flight of a charge packet through a sample in an
external d.c. electric field. The results, however, cannot be interpreted in1terms of a carrier mobility in the conventional sense lue to the spreading of
the packet by multiple trapping and release, and the concept of time-dependent
mobility has been used to describe such dispersive transport. Recently the
mobility was also deduced from transient photoconductivity and photo-induced
absorption experiments. Using the acoustic-electric method to measure
mobility provides new information. A simple coherent view showing the
connection among all these different measurement techniques has been written
This work was supportel by the Joint Services Electronics Program undercontract NO001 4-79->-)t24.
,eo
26
6. MICROWAVZ ACOUSTICS
C1 2 1.
6.4 Wideband LAW Transducer*
The fundamental advantages of LAW devices are well documented but there
has been no efficient wiiebandwilth transducer with which to i-plement them.
We have recently devised a mode converter that converts surface acoustic wave
SA' power to LAW power. This device consists of a number of unconnected
electroles which extend across the SAW and LAW propagation paths. The
electrodes are bent and spaced so that the electrode spacing, measured in
ohase at center frequency, is the same for the LAV and the SAW beams. The
first generation device hai a 5- bandwidth and a loss of 5 iB r6.3!.
The second ieneration has a bandwilth increased to 101 and a loss reduced
to I IB. With this bandwidth and efficiency, LAW devices may become practical
because efficient SAW transducers can be used to generate the SAW beam which
in turn is converted to the LAW beam by the Multistrip Mode Converter (M14C).
The node converter is reciprocal and LAW waves can be detected by conversion
to SAW, and then a SAW transducer is used to detect them. Efficient moderate
bandwidth LAI delay lines ani convolvers are now possible.
6.5 X Theoretical Determination of 3urface kcoustic Wave Velocity
and :mpedance Differences between Metal 3trips and Free Surface
legions of Metallic 3ratings**
The surface acoustic wIAe is a single-moied, non- disnersive
one-dimensional wave that is frequently modeled using a transmission line
equivalent circuit. Thin electrodes and shallow grooves located in the
surface acoustic wave propagation path produce reflections and shift the
propagation velocity of the wave. The effect of the electrodes nd v grooves
are modeled as an offset in characteristic impedance aniLTT3Z' and a shift In
the propagation velocity QDELTERV). This impedance offset and the veloity
shift, have been theoreti:ally ietarminel from the material icnstants and the
* This work was sunDortei by the Rome Air Development >enter, Deputy fcrSlectronic T-ichnologj, under Air Force contract F-I ?102-31--DJ?.
!his work was sunoortei by the Rome Air Development fntrr Deputy frSlectrsnic Technolog-,' unier Air Force contract F-I -v3-7-ThiD.
27
1 6. MICROWAVE ACOUSTICS
geometrical dimensions. Since the materials are piezoelectric, both the
elastic and electrical loading of the surface have been considered. The
detailed theoretical derivations for the impedance offsets and velocity shifts3 have been published in a series of papers during the course of the development
F6.2,6.4 -6]. A report has been written to tie these studies together and toE provide a simplified analysis of the surface load:ng produced by an electrode
that will be useful to the device designer. The report has been made concise,
with all derivations taken from references or appendices in order to make it a
better design tool.
IReferencesI
j .ril] M. J. Hoskins, S. Datta, R. Adler, and B. J. Hunsinger, "Unusual
Acoustoelectric Current Saturation in Semi-Insulating GaAs," J. Applied
J Physics (submitted for publication).
fR2] S. Datta, "Time-Dependent Drift Mobility in Dispersive Charge
Transport," J. Noncrystalline Solids (submitted for publication).
II
III i
. - -I 2
28
7. SUqFACE STT JDIE-3
?acu2.ty a~nd 3enior Steaf
. '~h~i~h3. 3. Thambers
~ r'AU~t 1S tdpents
S. 3rass IJ. Coulman T. Wrigley, Tr.
q. -4i. Fink
~1 Dif fusion arii o ativity oni Met-ils*
'2'1 1eeompsition of polyltominb 7ses ofl 801113 -3 ~f ineetIn both,halca ani a",-:troiC t-3chnolo- y. 4-3 'i-e been -xmirninj the3 ifl--rao tiO13
o, ne-thane? math tils i3 -i si~Tple but manageable -3xarnpl: of 31uCh prO' -eSSeS.
rhi3 effort has been --oricntrate1l upon linierstaiia' the mechinisrns by hh
1e'~tpatioi o~cus; f plrtiulat interest is *,ehy there-: is -.1 a~trt
- -n3roU to 1-- :limpoitio-I at th-3 surfi2:e -nii hoT- tthi3 is overloote.
. A~3'lr3L-3-1t3 rent)etl last 7-ear .311 iqli"tei that tlhe; i.:t4-ttion lry
_M4e-1aie I n theI t_3nperituire of both the surfi::e ami the 7ases Le~an~-3 onsi-ir-ibly asnalr than the *r'.efiisome time aje 12, i n ima
-:)cp?:eri t -'1 t, n 4&~ ofIliy he liC a 13L3 3-C -its I. rf5 'irl 40 o tt
*P,733ibili tt-: to virit for thise:
1'- 3 WOr 113 eu pp-rrt b the Titional 3 tien!i 71n1in ne,~rn V3-25)1 5. 13~11a 4. tlon To~e 11- ia 7;')'l, I at "T". ty
* '1. 3. lavjr, '13. klir Foree4) Inie - oa3ya: 1 -2
I|29I
7. SURFACE 3TUDIES
1. The beam experiments were done on single zrystal tips, rather than on
evaporated films, wnich served as substrate in the more recent work.3 DThere may be a significant lependence of decomposition rate upon
crystallography, which couli account for the iiffereices.
K 2. if the reaction proceeds through a precursor, that is through a gas
molecularly bound to the surface prior to lissociation, lifferent
3 activation energies are expected in the two measurements.
I 3oth types of effects are likely, and we have carried out several stulies
to lefine more closely the important events at the surface. ",1easurements it
j low. temperatures indicate a heat of physical adsorption of 6 kca/mole. Under
the conditions of the iecomposition experiments, a reservoir of molecularly
hell methane therefore exists on the surface and lissociation shouli take
place via these molecules. This suggests that in a molecular beam experiment
in ghich the gas alone is heated, the activation energy shouli be 3 kcal'mole
higher than in stardard measurements, and should amount to roughly 11
'cal/mole on rhodium film. To test this hypothesis a new apparatus, shown in
?i.. 7.1, has been built. !a this equipment, the gas czan be heatel while
keeping the film, .hich is in contact with a large thermal mass, at a constant
I temperature. The results obtainei in this way are plottel in 7ig. 7.2.
kfter correcting for contributions from gas molecules that thermally
I accommodate to the surface, an activation energy of 10.7 kcal/mole is founi
for methane decomposiag on rhodium films.
I This is in excellent agreement with expectations; more than that, it i3
consistent with other phenomena observed in the interactions of methane githl rhodium. It has previously been found F3,R41 that exciting the v, stretching
moie, rhich has an energy of 3.6 k,al/mole, is ineffective in accellerating
l lecomposition. This is now quite understanable. An energy of 11 kzal/T.ole,
much more than is available in the v. vibrations, must be supplied to methane
in order for it to dissociate. Most likely it is the 5v; bendiai moie that
must be excite a i orier to enhance the reaction on rholium. 1>tailet
predictions of the effect of isotopi: substit7tion hnve been male, basel on
the assumption that vibrational eaxitation is important; thsse preli-tions
are in good accord with 3xperiments.
-- =-• ........ ~i ...... . ........... .......... ,.-...............................................-.....,,......-...........il...
30
r_ L2
~- .-4
-- -2
31
ITemperature (°C)
100 80 60 40
so = 1.97x10 7I -16
1 IO-7
I-10.7 kcal/mole
Gas Temperature-16. Assuming a= 1
. 5x10-8 -J 0 0
_-17-
-17.5- 2.5 x iO9
-18
1.3 1.4 1.5 1.61/(RTg) AP.1213
* Fig. 7.2 Sticking coefficient of methane at different temperatures on
rhodium films maintained at OC.
32
Tq a -work~ 2arriel oat here ha s clIa r if ie-l the role of'he -as in the1eop~iioireac-tion. 3till to be explorel now are th-e -()ntrib,,tior13 of the
solil, ani th-3 e3xte3nt to -htih these are specifically affectei by the ato'ni:
irrinie-tant a" the surface.
7.2 Diffusion ardlaicti-rity on 3ili,-on*
Although tne importance of surface effects in 3sol1ii state i ev i. e s is
:-3narally rezognized, relativrely little is '.nown about thne properties of
surface layers on semiconluztors. 1ae have been : orcentratin,- our e3fforts on
,-btainia-y in formationi on the spatial stability Of sU:ch layers. As Part o f
t'hese inavestigations we havre levelopeil a tachnique frlirect ile-ctron beam
'rtn o iion -.1 T hs nthl ole-:cular nitrozen is psi,-3-,Allv
-iso rbe I on siliina, maintainel at low temperatures at 'iiha monolayer Or
more is fo rtei. Aoleculazr nitroven -on thne surface is then iisao0ciate1l ")y
eec t ron beat irncazt. Dn -.anming the sample, andissoriatei molecular nitroogen
iva-noratas, leaving behiil a nitrogen layer --flerever the el2ectron bat has1
3truck, as 3ho;,fn in i. ..
lirrent efforts have been f)_cusel upon chrceiig tae ::onl iti
.nI'er wvhi-h these nitroiarn layers -Are fo -r el. ),,I stuije;-s i-11icate th.at
?lec tron imnpac_ t I s qui ta e ff J iient in prc ro tiig i sso Ci;a tio n o.f ni tro el.
les, tha 3o x1 elcton n2 suffj,!e to create? a 1,t11 inayr
further bombariment lo-s not increase the arro-zit of nitrogen Cetic1lly b oun I
to tha silicon.
Thei--se situiration la:yers are actually nitriles, io t just chamiso rb e
layers oil nitrogen. It is clear from Fig. 7.4 that the 31iWZV) k-ager 3pnectr-a
of the layers prolucel by -3lecron bombariment are in gaol accori wfith theI
sneCtraI :re7or1e1 for sili-on nitrile prolucci.l by :!hemical1 *Iapcr leposittio
"i'e 'have been able to, sho-w, qowlever, that luring, the early stag-es of th- e
process, *hat appears to be a 2hemi3orbai layer Of nitrog;en logs form. 711.:3
is 1i11.stratel in ?i';. '.5, thare the inteansiti-es, of iarious Aug-er lines -i!re
asaown after ilcreasinag xcpoasures of nitrogen an! cniomitant botbarlment 4ith
eleitrcns. 7,he- 31-37 line chroeit~ of the nitrile ioes not app-:ear at
"his mrk ia s .3uiporte I bZ thje t-) It 3e ui 2es 3 1 1, tr-onI Ics nmram.%. Ary. 1.3. '11vy, J.3. 3kir ~Ore) 11nier otat1D -l D.?.
I
I3
I
I
Fig. 7.3 Secondary electron images of nitride structures created by electron
I beam writing on molecular nitrogen layers held on Si(lO0).
I
34
Si(LVV)
69 54
61
52
83 83
(a) (b)
- *.- Comparison of LVV Auger spectra for nitride layer produced by
electron bombardment (a); and by chemical vapor deposition (b).
IX 10-7 T0RR1000L
900
800 _ N
700
U D 600
>- 500
I z'Z 400
S300
1 200
1 380
1 10083
105 10 15 20 25 30jEXPOSURE (min)
Fig. 7.5 Auger intensities observed after increasing nitrogen exposures
and electron irradiation. The 823-eV nitride line only develops
after an induction period, during which a chemisorbed layer farms.
36
-7. SRIFACE 3TrJDU7,3
tne very start. However, the 330-e3V line attributal to nitrogen is3 apparenit
e ven at iery !,.i e',cposures, ializating that there is a niitrojen laye-r present,
thouyh it is not a nitrila. The nitride layers -10 m -3 after extensive
extyosure are more than one layer thick. The mechanismn of formation ofths
njitril is n 'ot ClearL, but it is quite complex and may involve a c ompetition
between el1ectron impact lesorption ani lissoc-iation.
31ectron heam inducel formation of silicon nitrile- is a partic ilarly
a pe al in-,, proc ess . It gives us the capability of foignitrile structures
,inlar ntroliql ani very clean conditions. A Iliestion immeiately arises-
;13 this techniqule: applicabla to the formition of other films as #eli? 'We hav e
r-3cently exa-minel the feasibility of creaiting oxii-- layers on silicon in this
-,ta y. O)n exposing silicon to molecular ox~ygen, a cnemisorb-i layer fOrms on
the surface. At room temperature ani belo,-w, *cxilation is slow; hwvr
electron bombariment enhances the oxflation rate. As is apparen-,t from 7ig.
7.. it is therefore possible to form oxile structures by ele-Ctron ba4m
.7aliation of an oxymen coverel surfaces. The most alviantageous1 conhitions'1
fir -'his nroce-ss are, now being7 explorei1.
7.3 ktomic ",xplorqtioni of Jrystal Layers*
O)ver the years consilerqble information has been aciumulatei Li this
laboratory about the behivior of iniiviiual metal atoms on metal suirfaces
-7. 1 This has been possible through the use of the fiell ion mic roscope ai
its ability to resolve single atoms. .4'nil -ork on single a :toms has
continuel, w~e have also been ex-imining, the belhavior of overlaye rs, with the
intent of contributing, to a bett-ir unlerstalirlg of crystal anli fIm. Jrowth
phenomena; in the past these oheAno gena a hare not b-?en a 3 ei bl1e to
obser'ration on the atomlic level.
Initially, -.fe have concentratei on a survery of overlay~r formation on
I 1 1) . M4tals fro-a the platinum family form t-o-limensional lay)ers, provile~l
the alatom concentration on the substrate? is hril-h eno ugh. A3 is ap-parelt !I
the fieli ion micrographas in Fig. 7.7, iniividuaal atoms in the la-yers are
ilearly resolvai. This is lute liff-'rent f-om -'nat i s fo0uni 'ne n the
71hi3 work was sunportei b4Y thei Titional Science3 ?ouia tiLon uie-r -rant 7IR3~-.~D1 ml 1 tha Joint .5ervic es "El- tronics Prga 13. Army, '3. 3. Navy,
1..Xr ?o r ey unhe r co n t r ac z m -32)
Ii
I37
,I
IIT
Fig. 7.6 Secondary electron image of oxide structures formed by electron
f beam writing on Si(lO0).
I
38
e l
Fig. 7.7 Platinum metal layers on W(11O). Shown in order are palladium,
iridium, and platinum monolayers, as well as a schematic of a
P(2 x 2) structure.
Fig. 7.8 Field ion image of silicon layer on (110) plane of tungsten
(left), and proposed structure at right.
I39
I'. SURFACE 3TUDIES
U substrate itself is imaged. On a W( I)) plane of the same size, inlivilual
atoms 2annot be liscernel. The spacing in the layers of platinum, pallalium,
U ani irilium must therefore be a multiple of the substrate spacing. This is
not necessarily true for other metals; for rhenium, individual adatots are
5 not resolvei in overlayers on "W(11), and we therefore infer a much lenser
packin.u
A A particularly interesting, behavior has been observed for irilium
clusters. Their arrangement and stability is strongly dependent upon size.
At 460, K for example, a :cluster of 16 or more iridium atoms on W(110) is
stable in a rhombohelraL configuration. A cluster of only 15 alatoms,
however, occurs in a square arrangement. This is not an artifact; the
2luster maintains its shape under repeate - heating intervals. ; hen a single
atom is removed from this layer, the arrangement becomes unstable - the
cluster breaks up into iniividual irilium chains.
This sort of behavior is not limited to irilium. As an important part )f
our studies we have been looking at silicon layers on W(I I D. 7o'derat-e is3
of silicon equilibrated at 300 K yield a very open overlayer, shown La
7.9, made uD of silicon chains oriented along f11. The spafin- be-ween
chains is close to 9.5 . "loser packed layers can be fomed, however, by
ecuilibrating heavier loses at higher temperatures. By field evaporation the
-oncentration of silicon alatoms :aq be lowerel at ill. Repeated fi li
e raporation followed by equilibration converts these lense lqyers into ccein
arrays, analogous to those in ?ig. 7.3. On further relucin the alatom
population, these overlayers isorder slm.ilarly to irilium layers.
The formation of these loosely structured layers on 4(1I) has not been
previously observed by other techniques 361. A survey has been male of
silicon layers on other planes. So far nothing unusual has been observed on
I 31' ), 510) or (I Y). It is already c lear, however, that field ion
microscopic examination is zapable ,f yielding interesting info ' ati)n about
I overlayirs, and these studies as ell as other measureT.ents on ini.-iilual
toTs Ini sall clusters are continuing,
II
. .... 1'
40
7. 3U-RFkCE STUDITS
References
-RI] 2SL Annual Report of Progress 1979-I 930, 7.1.2.
[2] -. -1. Stewart and 2. Thrlioh, J. "hem. Pays. 52, 4672 (975).
FR3I S. Is. Brass, D. k. Reed, and 2. 3hrlich, J. TheM. Phys. 79, 5244 (I 979).
,1.4 J. T. Yates, Jr., J. J. Zinck, S. Sheard, ani W. H. "einbera, .. 'Them.
Phys. 70), 2266 (1979).
K51 1. :ezel ani 71. Lieske, J. Appl. Phys. 51, 2566 (1980).
R6] J. P. Biberian ani 2. A. Somorjai, J. Vac. Sci. Technol. 15, ?9-3
(1 979).
II
41
3. ELECTROXAGNETI- COM4MUNICATION, RADIATION AND SCATTERINGU
Faculty
R. YMittra
I,graduate Students
M. Desai. R. Lampe 3. Ray
R. T<astner J. :aguire T. Trinh
I R. Paleta
3.1 Millimeter-Wave Integrated Circuits
In the past twelve-mcnth period, we have made significant prcgress in the
development of devices and structures for use in millimeter--ave integrated
circuits. A amber of analytical techniques, appli~able to the dielctric
-aveguiies used in these circuits, have also been ievelcped.
k narticular area of interest has been that of dielectric antennas. We
have investigated a type of leaky-wave antenna w.hich ccnsists of a dielectric1 waveguide on which are placed Deriolic metallic perturbations. This antenna
is frequency scannable and is readily integrated with a receiver-transmitter
system. Prior to cur investigation, it was believel that the perturbations
had to be unifo-r. Our experiments have shown that ncnuniform perturbationsresult in significantly better antennas. Antennas "wrth 26 1B sunpresscn Cf
silelobe and endfire radiation levels have been built. Thes? antennas -rn
have beamwidths in the plane parallel to the antenna axis as lcw as 7.1c.
. his3 work InS3 UD rupnortI b; the Jo in t 3e r-vi ces ct rc n'is ?:-cgram"J. 3. Any, 1.3. 'a11y, J.3. Air Fcrce) ,uiler contract N10'
42
3. ELECTROXAMIETIC COMMUNICATION, RADIXTIOT AID SCATTERIIn
For some applications, the radiation pattern of this antenna in the plane
transverse to the guile axis may be too broad. To reduce this beamwilth, a
new structure was developed in which the dielectric guile with periodic
oerturbations on top is embedded in a metal trough with a metal flare attached
on each side. With this arrangement, the antenna behaves like a linear array
in the longitudinal plane, while the radiation pattern in the transverse plane
resembles that of a horn. With this structure we can achieve 70 beamwiiths in
the transverse plane.
Many passive components in a dielectric-based millimeter-wrave integrated
circuit, such as ring resonators and couplers, are comprised of curved
waveguile sections. Therefore, it is necessary to have a good understanding
of the behavior of the fiell in the neighborhood of the bend in order to
prelict the coupling and radiation characteristics. Seneral analytical
approaches to this problem have been very involved to date. We have found a
simple, approximate technique for calculating the field n a cur-ted dielectril
waveqguide of rectangular cross-section.
Dur analysis is based on a conformal mapping which is used to transform a
homogeneous curved waveguide into a straight waveguide wiith an inhomogenecus
refractive index. Solution of the resulting equations yields airy functions
which accurately describe the field both insile and outside the wvteuiie.
in the area of active device design ;we have worked on the nroblem of
TAPXTT oscillators which are directly integrable with dielectric "aveauiles.
Usually this results in excessive mismatch and radiation loss. Through the
use of partial shielding structures and better matching networks we hate
reduced these losses significantly and realized oscillators which wo-k at 34
z.
A new methcd of solving dielectric waveguile discontinuity problems -s
also being developed. Unlike a closed -a.teuile, dielectric #aveg-uiis
support radiation moles with a continuous eigenvalue spectrum. This makes the-
application of mcde-matching and related technicues very difficult. D'r
meth0 is basel on the spectral domain method and utilizes the -77T alcrithm
rather than matrix inversicn. At present small steps cn lieiectric wa'egules
have been haniled successfully. "We are currently inccrporating a variational
i.oressicn to enhance ccn'rverence for larger iisccntinuities.
I43
3. TLECTROMAGNETIC COXNUIZATION, RADtXTION AND SCATTERING
i 3.2 Tlectromagnetic Radiation and 3cattering
This area of our research is primarily concerned with the efficient and
3 accurate calculation of electromagnetic radiation and scattering problems. We
have concentrated our efforts in the high frequency region where moment method
solutions become impractical and where, either by body size or complexity,
asymptotic solutions are not applicable. In the past a spectral domain
aproach has been formulated and applied to thin plates. Recently ws have
been able to accurately extend this technique to two-dimensional bodies of
significant size, which may be either perfectly conducting or iielectric in
I nature. As this technique is iterative and uses the FFT algorithm, it
contains a built-in boundary condition check and is very efficient.
Ie have also Aone experimental and theoretical studies of twist
polarizers. These polarizers can be used in radar systems where traditional
I gimbaled systems are too slow, as the twist polarizer is very light weight and
can be moved rapilly. The theoretical analysis uses the spectral domain
I technique and gives good agreement with experimental results.
IIIi,I!
I:.
9. 7DLASM4A P?YSI"*
Paculty ani Sen.rior 3taff
A. Thaether
uiraluate Stulent
3. K. Aul t
JI 3tatisticql -?ropet'ties of ?lasma Turbulence3
Thei, objective of" this rgenrIIh * t') 3tulY thq statistiznal prope rtie S of
plasma turbulen::e. The syste3% anier ixiastijatiorn is the positive -olcma of i
1:!-iischaroe in helium -dhi~h i3 Ulnstable -,ith resipect to the ST)Oqteneo 1.i
r-rcitatLon of ion acousti3 -,avres . Eeti fiH luttIos fth
turbulent iavre +f.-i are? pi-,k:? up by a Lai-muir probe ani their pr-,b.-b-i'Jty
113tribution is measurel by a fast samnplii- methol. Thei measurcileritin
fL-nt IViSsian beha',ior c-aa be fittel to a Thaqm-Thariir e3xpansion of t~i-e
Probability iistrfibution iricluiing terms up to fourth criler. Thle 101-3,1ss1iA-
mharacter of the- probability isatribution allowe onclusions to be; lrdna13.
th? .neture of th:e ensemble of the faves thlat :composa the turbulenyt fL11ie.
This orkC -413 31pPOrteil by the *ivrtyof Illiiois.
'II45I
U 10. RAREFIED GAS DYNAMICS AND COMPUTATIONAL GAS DYNAMICS
I Faculty and Senior Staff
IS. M. Yen
IGraduate Students
I D. R. Hall S. H. Lee
10.1 Rarefied Gas Dynamics*
Rarefied gas dynamics deals with non-equilibrium gas flow problems in
which microscopic treatment according to kinetic theory is necessary to
determine the effect of intermolecular collisions and gas surface interactions
on both microscopic and macroscopic gas flow properties. Such rarefied gas
flow problems occur not only in aerodynamics, but also in electronics,
aeronomy, environmental fluid dynamics, and other related fields.
7he aim of this research program is to develop numerical methods to solve
a wide range of problems under conditions far from and near thermal
equilibrium. A Monte Carlo method has been developed at the Coordinated
Science Laboratory LRI1 to solve directly the Boltzmann equation and has been
used by the Boltzmann group to solve the Boltzmann equaticn for several
rarefied gas flow problems under a wile range of nonequilibrium and boundary
conditions FR2-R7]. The solutions we have obtained yielded detailed
microscopic and macroscopic non-equilibrium properties, nost of wh-:h have
never been treated and studied before. We have also studied numerical
solutions of other kinetic equations and other numerical methols to 3olv-
I rarefied gas flow problems, including the direct simulation technique.
This work was supported by NATO Research 3rant 1'075 and by the University ofS illinois.
46
10. RAREFIED GAS DYNAMICS AND COMPUTATIONAL JAS DYNAMICS
Nonlinear evaporation-condensation Droblems are encountered in such
diversified areas as upper atmosphere meteorology, the cooling of nuclear
reactors, design of space experiments, oetrochemical engineering, vacuum
technology, and the intaraction of high power laser radiation with metal
surfaces. The treatment of these problems requires first the consideration of
vapor kinetics problems characterized by the nonequilbrium "vapor motion in a
Knudsen layer at the interphase boundary. Under a joint research effort, ye
have successfully studied the evaporation problem [RB1. Our Boltzmann
solutions establish the validity of the Krook solution and Ytrehus" kinetic
theory approach FR9]. This approach can be used to calculate the jump
conditions across the Knudsen layer and the net mass, momentum and heat fluxes
Rilo] and, together with the conventional continuum method, to calculate the
flow parameters at the outer edge of the Knulsen layer for problems in which
evaporation occurs at the interohase boundary.
Studies have also been made for the condensation problem r10.21. The
non-equilibrium vapor near a condensing surface differs from that near an
evaporating surface because of the difference in edge condition at 'he
internase boundary and its behavior is more complex. Our Boltzmann solutions
yield distinct non-equilibrium behavior in the Knudsen layer of condensing
vanor. For example, the temperature gradient becomes negative near the
interphase under certain conditions. The condensation rate as a function of
pressure ratio was found to depend strongly on a substance parameter which i9
proportional to the ratio of the latent heat and its liquid surface
temperature. Some of our results were found to be in agreement with those
obtained by using the simple apnroaches of 0guchi rR11! and Ytreus R12].
We clan to study further the evaporation and the condensation problems by
considering more complex flow and boindary conditions.
10.2 Naval lyrodynamics*
Free Surface wave problems encountered in Naval hyirodynamics are
characterized by complexities in flow geometry, flow features and boundary
conditions. The flow has an unknown free surface and it is propagative aniThis work was Supoortel by the Office of Naval Research inier 7ontract
N00,514-3O-C-0740 and the National Science Founiation under 3rant 'I3 EN77-20436.
1 47
i 10. RAREFIED 1AS DYNAMICS AND 'OMPUTATIONAL GAS DYNAMICS
transient. The boundary condition at the free surface is of a mixed,
parabolic type and it contains highly nonlinear terms. In the steady state,
there also exists a radiation boundary condition, since the waves, once
generated, propagate downstream. These complexities have led to several
computational difficulties: accurately accommodating the free surface
I geometry, satisfying the boundary condition uniformly over the free surface,
treating the radiation boundary condition and implementing the boundary
condition at the contact line of the free surface with the solid surface of a
partially submerged body.
IIn the initial phase of our study, we have developed two time-dependent
numerical schemes with Sulerian grid systems for solving steady and unsteedy
J potential flows for nonlinear free surface problems. In one method
[R13,R14,R16], we use the finite element method to deal with the geometrical
complexity and the free surface boundary condition of the nonlinear free
surface problems. In this scheme, the finite element method is used to make
the field calculation and the finite difference method is used for the time
evolution. We have used this method to solve two problems: a pressure
distribution moving with a constant speed and a moving submerged elliptical
cyliner or a symmetrical hydrofoil.
We have also developed an explicitly time-dependent finite difference
scheme LR17i. Explicit schemes may be more favorable for solving large scale
problems on "supercomputers" which have more stringent implementation
requirements. We have used this method to solve two problems: a pressure
distribution moving with a constant speed and an elliptic cylindrical,
surface-piercing strut accelerating from rest. ur solutions serve to
demonstrate the feasibility of using the method to solve two-dimensional as
well as three-dimensional problems.
The results of the initial phase of our study were presented at the
Second International Conference on Numerical Ship Hydrodiyamis I R15.
I In both numerical schemes mentioned above, the computational domain ;s
xpanded downstream periodically during the computation as the disturbance on
the free surface is propagated close to the downstream boundary. Theref:ore,
the undisturbed condition is applied on thp zut-off downstream bounlary.
AD-AI04 727 ILLINOIS UNIV AT URBANA COORDINATED SCIENCE LAB PIG 9/2ANNUAL PROGRESS REPORT POR JULY 1, 1980 THROUGH JUNE 30. 1981,(U)AUG 81 R T CHIEN. H V POOR N0014t79-C-0424
UNCLASSIFIED NL2,2IIIIIIIIIII
EIEEEEIIEEEEEEEEEEIIEEEEEEEEIIEEEEEEIIIEEEEEmhohEEEEEEIEIIIIEEIIIIIIIIIII ,' mmmomomosm oA,
48
10. RAREFIED GAS DYNAMICS AND COMPUTATIONAL GAS DYNAMICS
In order to obtain the longer time and steady state solutions, we have to
use a fixed downstream boundary set close to the disturbance so that we -an
increase the computation efficiency by dealing with a small computational
lomain. The present phase of our study has been focused on the application of
our numerical schemes to the nonlinear free surface problems with a fixed,
small computational lomain. There are three major tasks involved in this
study: (1) investigate the implementation of the open boundary condition at
the outflow boundary to allow the waves to pass through it, (2) improve our
numerical schemes so that accurate numerical solutions can be obtained using
the open boundary condition at the outflow boundary, and (3) apply the methods jdeveloped to selected nonlinear free surface problems.
We have studied in detail the method to implement the open boundary
condition used by Chan R131. It is based on Orlanski-s scheme LR19] in which
the Summerfell radiation condition is applied at the outflow boundary and the
phase velocity is calculated numerically. 'e have conducted numerical
experiments to study systematically the errors of Than's numerical scheme to
implement the advection equation and to study ways to minimize these errors.
We have studied two approaches for solving notential flows for nonlinear
free surface wave problems using a fixed, small computational domain. :n each
approach, we have used Orlanski's method to implement the open boundary
condition at the outflow boundary. The two approaches differ in the treatment
of high frequency errors that appear in the solution and that inhibit accurate
treatment of the open boundary condition. In the first approach, we used
filtering to minimize high frequency errors. however, it was found to be
effective only at an early time; therefore, it is suitable only for obtaining
transient solutions. In the second approach, we modified our time marching
schemes so that the schemes are characterized by damping. This damping is
more effective in controlling high frequency errors. Transient and steady
state solutions have been obtained for the pressure distribution and the
accelerating strut problems.
Fig. 10.1 shows the pressure distribution problem with the outflow
boundary set close to the iisturbance. The appearance of a two-gril interval
wave in the solution of this problem near the outflow boundary is shown in
Fig. 10.2. The eliminations of high frequency waves in the solution are
shown respectively in Figs. 10.3 and 10.4 for two methods, asing filtering 3
49
I2 ) Y 1 ( -tIy
x -. 5 x 0.5- Outflow
j~ =I~-. Boundary
I - _ _a
4 u Xd
Fig. 10.1 Computational domain with fixed outflow boundary (x )
IB
0.006 1
0.004- r 1I2r =
S0.002-
I0
Hybrid Scheme
0.6 0.8 1.0 1.2 1.4Distance, x PM
Fig. 10.2 Appearance of two-grid interval (2Ax) wave in the solution of
I the pressure distribution problem for Fr - 1/27
500.03 I
00 - Fr 11ll-, t=l
0.01-
F- 0.01 .0
It -0.01-
-002 .. Unfiltered
-0.03 1-1 0 1 2
Distance, x ,-W3
Fig. 10.3 Elimination of 2Ax waves in the solution of the pressure
distribution problem (Fr = 1/27-) at t = 1 by using filtering.
0.03
0.02 -Damped Scheme
.- 0.01 .----Undamped Scheme
a-
-0.02 - 0 - =0.01_V
-0.013-1.0 0 1.0 2.0 3.0Distance, x NP-532
Fig. 10.4 Solution of the pressure distribution problem by using damping
in the time-integration scheme.
- . . . . . . . -. .
51
I 10. RAREFIED GAS DYNAYICS AND COMPUTATIONAL 3AS DYNAMICS
I and damping. Fig. 10.5 shows the steady state solution obtained by using
damping for a ficed computational domain. Fig. 10.6 shows a transient
solution of wave height for the accelerating strut problem.
The results of the second phase of our work were presented at the Third
International Conference on Numerical Ship Hydrodynamics r10.41.
Rqferences
Rq] A. Nordsieck and 3. L. Hick, "Monte Carlo Evaluation of the Boltzmann
Collision Integral," in Rarefied 3as Dynamics, p. 675, Academic Press
(1967).
[R2j 3. L. Hicks, S. A. Yen and 3. J. Reilly, "The Internal Structure of
Shock Wkaves," J. Fluid Mech. 53(l), 35 (1972).
FR13 3. X. Yen, "Monte Carlo Solutions of the Nonlinear Boltzmann Equation
for Problems of Heat Transfer in Rarefied Gases," Int. J. Heat Mass
Transfer 14, 165 (1971).
3R4 S. x. Yen, "Numerical Solutions of Non-Linear Kinetic Equations for a
One-Dimensional Evaporation-Condensation Problem," Int. J. Computers and
Fluids 1, 16' (1973).
[1 S. 4. Yen and W. '1g, "Shock 'eave Structure and Intermolecular Collision
Laws," J. Fluid Mech. 65(1), 127 (1974).
-361]S. A. Y n, "3olutions of Kinetic Equations for the Nonequilibrium ',s
Flow 3etween Lnitting and Absorbing Surfaces," in Rarefied 1as Dynamics
Vol. I D. A.15-1 (DFVLR-Press, 1375).I
1I | !
52 0.03 I
0.02- Fr 1/4' '7, t 1O0- Present, Non linear
- .. Haussling, Linear
-~0.01 \
-0.02- Open Boundary
-0.03 Ii i
1.0 0 1.0 2.0Distance, x NP-53
Fig. 10.5 Steady state wave height solution of the pressure distribution
problem (Fr = 1//27) obtained by using a fixed, small computational
domain. Comparison with a linear solution.
Fig. 10.6 Free surface elevation around an accelerating strut at t-0.16.2(Fr) - 0.025.
U53
10. RAREFIED 3AS DYAXICS AND OPUTAT NA ' DYNA ICS
FR7J 3. M. Yen, "Solution of Boltzmann and Krook Equations for Heat Transfer
Problems with Maxwell and Evaporative Boundary Condi. _ons," Proe. h
Tnt. Symp. on Rarefied Gas Dynamics, Vol. Ii, p. 53 (1976).
ERS] S. M. Yen and T. J. Akai, "Numerioal Solutions of the Boltzmann and
JKrook Equations for a One-Dimensional Effusion-Evaporation Problem," in
Rarefied 'as Dynamics Progress in Astronautics and Aeronautics Vol. 51,
I p. 1175 (1977).
[R?] T. Ytrehus, "Theory and Experiments on 'as Kinetics in Evaporation," in
Rarefied 3as Dynamics, Progress in Astronautics and Aeronautics Vol. 51,J p. 1195 (19*77).
-RI D7 3. 1. Yen and T. Ytrehus, "Numerical Solutions of the Noneauilibrim
Vanor 'lotion Nlear an Evaporating Interphase Boundary," ASME Paper
3O-?IT-82, (\1930).
rill! :4. Hatakeyama and H. )guchi, "Kinetic Theory Approach to Nonlinear
Condensation of Flowing Vapor," Proe. 11th Int. Symposiulm on Rarefied
'as Dynamics, Vol. 2, o. 1293 (19"9.
rR12l T. Ytrehus and J. Alestad, "A Mott-Smith Solution for Nonlinear
Condensation," AIAA Progress in Aeronautics and Astronautics, (toi appear).
RI R31 S. X. Yen and K. D. Lee, "Applic:ation of Finite Element Method to
Potential Flow Problems," Proc. Second International Symposium on Finite
Element Methods in Flow Problems, p. 539 (t9-6).
RI Il S. X. Yen and K. D. Lee, "Design Criteria and leneration of ptimum
Finite Element Meshes," Sixth Int. Conf. on .Numerical Methods in Flui
3 Dynamics, Tblisi, JSSR (June '973).
U,I . .. ... . " -- * " "" . . . ' " " ;' " - 9"-''": i
54
10. RAREFIED GAS DYNAMICS AND COMPUTATIONAL GAS DYNAMICS
[R151 S. M. Yen, K. D. Lee and T. J. Akai, "Finite Element and Finite
Difference Solutions of Nonlinear Free Surface Wave Problems," Proc.
Second Int. Conf. on Numerical Ship Hydrodnamics, p. 305 (197S).
[R161 K. D. Lee, "Application of the Finite Element Method to Potential Flow
Problems," Report T-32, Coordinated Science Laboratory, Univ. of
Illinois (1976).
[R17j T. J. Akai, "Numerical Solutions of Nonlinear Potential Flows with Free
Surfaces," Report T-30, Coodinated Science Laboratory, Univ. of Illinois
(1976).
[R18] Robert K. C. Chan, "Finite Difference Simulation of the Planar Motors of
a Ship," Proc. Second Int. Conf. on Numerical Ship Hydrodynamics, p. 39
(i 978).
EI-191 L. Orlanski, "A Simple Boundary Condition for Unbounded Hyperbolic
Flows," J. Comp. Physics 21, 251 (1976).
1 55
?aculty an-i 3enicr 3taff'
1J. k. Abrahamu . .Da'riison 3. R?. a
R. 1. 3ro wn J. H. patal1 A. 3. 3chlinsker
I iraduate ________
'.Airial D. Halnerin 4,2 1 ~iv en
?. 3cs e L . 'is n -s R. '1 r ton
T. Tho':1P. H-s'uT Tu
3?. "IJ.'1arege C. H-u a n A. ?I 3SZkun-j e"r tI a nH o1 r?. 3?rNm a A. '(qsiler -f. 3ogers1 (. ?i-~hs T. Kinse. '. ren
?.Fr~1 'Koob Utp Uffrh e i e
I4. 3raf X-?3. Mak .vIl
rc ~rhcski X.Y",
I3-isearch ii this area is ccncernei -Afth conputar system.s vhizn a'r botahnigh nerfo-~aince and higjh reliability, ani which 2an bi levelcjeI usilg Ve-ry
H igh 3peei integrateI 'ir-ruit IAqS?) ltechnolo-y.
1'1 erfcmance- cf 'nstruction ?Loi-elines
An inatructici unit Dinelinie iiff-ers fzrcm -4 finctionlal 'i. it ci.ie
-1 ist it is 3ubj!ect to lelays frcm syste'n D~iencmena eicternqT to it s e
e. m.'ernry ant f-incticnal units) . 71huas a sernent in an int'atc nit
Th is fcr*.c -was 3unnecrtei brthe To-.it ?eics -c so-IT,' '.5A ri 1,, 'J.3. Uav .S.. Air ;orse) ln i ar :nrc t n'Y t-- r~t 1. ty th-4I 1.3.Navy nlier cntrict qt-)>9~ nI by th i t-~si ay
56
11 . COMPUTER SYSTEAS
pipeline may take more than one time unit in the event of cache miss, memory
collision, data dependency, etc.
An exact solution for pipeline performance was obtained by considering
the execution time of any given segment to be a discrete ranlom variable which
is governed by a multinomial distribution. The expected wait time of an
instruction in a given segment can be recursively defined in terms of the
expected wait time of the instruction immediately preceding it. If the
difference of these wait times is negative, the actual wait time will be zero.
'nfortunately, this solution is not in closed form.
The effective bandwilth of the pipe can be given as a function of the
expected time gap between the ith and (i+1 )th instructions. The stationary
value of this expectation requires that i be infinite, but it was found that
for values of i as small as 3. estimators can be obtained for the oarameter
ranges of interest with a high degree of confidence.
A cIcsed form solution was obtained heuristically for a pipe consisting
cf N + 2 segments. The two eni segments are variable-time segments, and the
miidle N segments are all constant-time segments. The purpose of this study
was to determine the degree to whi-h a finite-length buffer could decouple the
bai behavior of two variable-time segments. This solution was found by
matching boundary conditions (cn N) against in exponential function of N, and
iomparing against a wide range of simulated values (779 points in all1. This
solution had an error range < Vl. w-ith an RS error of < ?o over the simulated
data used in the comparison.
11.2 Test Teneraticn for 'cmplex Digital Systems
A new and general fault model for microprocessors has been develcpe, and
test techniques have been derived to detect all the faults in the fault Model
>11.3]. A methodology was found for generating combinational structures from
high-level descriptions (using assignment statements, "if" statements, ind
single-nested loops) cf register-transfer level cperators r1 1.1 1. T h e
generated structures are cellular, and are interconnected in a tree structure.
A general algorithm has been developed to test cellular tree structures with a
test length which grows only linearly with the size of the tree. :t can be
proved that this test length is optimal to within a icnstant factor. "Pi ys of
57
k4~cn tI e "I re -- Oen 'Ieasobe fou1ni
1 3 cncurrent ",rrcr *>:tecticn in krithrnetic- ani LoicJnits
I A ne-i tethol cf 2cncurrent i-rrcr Ieatection in onnmute r Artmtcall
cg0ic Unis - a3 -11 een levisel. This 'nethoi excnlcits the inharent yi3 red'niancy of components in the time icmain. This3 tehnijue 4-3 applicable toall coeraticns of a ty-pical ALLY. In this sense, our method is supe'ricr to3 Arithmetic Zcies, -;%hich are apnliicable to only the arithmetic coperaticns of an
ALLY. ?uNrthermore, cur methoi also wforks -with many types or carrI lck' 'AheaiI ci-rcuits. -n terms of the qrrcr coverage canaqbility, our me-hol ccmnares4-avorably with a iunlicatel ALLY methci of error ietection. :I ter-ns of thne
aariiamre cost, our methol is superior to both the Arithmretic c~es anI thie
luplicate1 ALJ.
I ~ I H-iigh -ierf-crmance and, 2eiability in Systolic Structures
S-rs tc arrays hiaver been Tooos-il fo:r coniveni4ent eI eficint
im7)1eentaticn on a TuSI chip. HOW9VSr* systolic! systUem 17CrithIMS -croosei
Ic not use all the inherent psrallel. ismn in the harlwqare. Ichae un!
t3chni4ues Piicncreasa by a factor of two or three the3 Derformance of.systolic syste ms by re arranging, the seque3nce of thie lata flow~ ani maki4,
31si'ght Moifications to the original systolic- system-. :n ii-iion, ahg
level error letecticn technique his been founi fo-r use with systo--ic! systems
1 pe rfo m ing- -atrix multinlication. Th1i s methol Iices not use much axtrahan-dare, "out will nrcvile Drotecticn ag-ainst both transient ani nermanentI errors.
1.5 Yemc:r Aidessing, krchitecture
7-,)r -,he red-?uctioni of processor-memory, allress banlilrh, :%r ' :a -re
-ievlne n ailiress ire ic tion r.1ecaanism. An aliress prediction stack hoilis
the aliresses of instructio4.ns ani Iata vhiich 'nave a high p3robabili-*ty of en3refe3renced soon. The3 reluction of 411rqss baini-filth is ac-hieved by e-ncoliag
Dposit;ions in the stack structure ani se-ndin-- an e ncold cosiric-n to te3 mnemory
3u-bsyst-?m ais anit aliss . Thie remoryf system eie:C-s t"11' stack ositiocn aniI appli!s it to an--Ien'tic-*! stack 3tructure maintainel -fithin the mmoy. .s -- -
the nforati1 in i ts stac, < the Tor syte c o L a Inlt les
58
11 . COMPUTER SYSTMAS
A full address is transmitted on a stack miss.
An T3!U replacement policy and several line ipdate schemes were evaluated
by simulation. T'he simulation results proviie an information theoretic lower
bound on the average number of address bits needed per memory reference. This
bound requires frequency encoding of the stack positions and miss indicator.
Zonventional fixed length binary encoding, which is more easily implementable,
requires only one or two extra bits per address. ?or the frequency enccing,
as the stack depth increased and the stack width remained relatively small,
the average number of bits needed per memory reference continued to decrease.
This phenomenon did not occur in the case of the conventional binary enccling
because after a certain noint a Drice, in terms of extra bits neeled for
encoling, :Yas being paid for the greater number of stack locations. Stacks of
molest size using conventional encoding can reduce the average number Cf
address bits needed per memory reference to below 8 bits, even Ihen the total
address space is very large.
11.6 Shared Cache Organizations for Multiple-Stream cmputer 3ystems
)rganizations of parallel-pipelined multiple instructicn stream
crocessors with shared two-level memory hierarchies have been studied. The
cache memory interference and the shared cache hit ratios in such systems have
been investigated. ?erformance analysis is carried out by using discrete
"4arkov Chain and probability based theorems. ?erformance is evaluated as a
function of hit ratio, the number of processors, the number of pipeline
segments and the cache organization characterized by the number cf lines. the
number cf cache modules per line and the cache cycle time. Some design
tradeoffs are discussed and examples are given to illustrate a variety of
design options ril.121.
11 .7 Multiprccessors with ?rivate cache Memories
We have developed an approximate analytical mcdel for the performance cf
multiprocessors with private cache memories. All Drivate cache memories are
connected to a single shared main memory via an interccnnection netwiork. " e
have x. tended an earlier analytical model to include complex cachIe
organizations like write-through, load-through and buffered -ri te-ba..
%onsilerinj the complexity of such organizations, our approximate analysis is L.
59
3 ac',e tis 3 ratio, bIc 2! t rans fer ti4-me- an, Soz 4 nI ZZ ~ of1 t-
3 -,c" -,, .,7r,. 1
.3 ATerc ry Syster. krc hiteoture fcr ?aaIl rce9sng
U seeac isn Dr c-3re s 3 Wri h ex c lres the1_ :es4~ oe a T merv
,arihnaect-ure fo r the support of naralleil urcessing. 4e have iefin-e'i aI~ ~~ ~~ prlmnr eo y sytem archit:ctlure -,ii:h :=s3tS of1 a Set fmmr
acce2,ss requeas ts3, responses , ani the 3semanti acto asoiae -4t C,4acn W I'
tranactin. Tis arc:hitecture -ill se-,,e to i-sfine a ~om~ai nctcol
i-n an be apclial to a %rariat7 of' Parillel MeMCrv structures a-1 71rlll'?
i.nterconneonaticn teonnologias. Th narallel a h'itect -ure allois the sharing of
a, laa i n a ! e ry intlimate fa shi4o n as _ - 3taiinl3l-!Te -
ocr temcrary Ise to exzloit iata reference lcaility. k. paper 1--scribina, thlis
sha-ell me-rcrr rhtetr has been submitted.
1 '1 ' ,cmnute-r Xr7!hitectu-ire for tivh eel L ang-lsaReS
1 Rsearch hqas been in nrc~ress in - th e area of Ie s i ni ii-ct. v
~nerlrt1~ Y-gugs YDr'~ for h~ia-h le langua,,e su)oort . 'echnques
for the fo-mal leri-ration of Dlls iesi-ned oefcenl urr hligh lfne
Results inlicata -hat it mnay in -,ne futzure be cosibLe to c seaia~
lesi57n efficient ')'L's -^rm a fOrmal sn-cificatiocn cf hehiih le'reL lan-'aa,
tc 'be surportel. '4e are :!urrently exnlorin7 the nossibility, of systematic!ally
1e:sijning the '):, intirprete ,r from a 'formal specific aticn of the sY-.tax a-1
3 semanti_ s of the D'-!. The oblective is to a.cclorg tec7hn--iqus for a :complete
1 systemnatic iesign -,f a haigh le vel language 3suI'Ocrt pr-o esor. Thus, th
-:!sign oroccs3 -.oruli not ony-11ell e3ffic:ient 2anlilate cL-s bt as
harlware s-tructures for the effic-ient Lintirpretation --f aseciD.
:, 3 -iogi 7 al iisati Cn
A 1- utm a ti I e s i -1n v er-ifco f acilityC is 4eInZ IeVilnei fo~r cur1-
k D 3,1stem. f-e hate1 ma ic Orogre 4 in iefinin1 the- crocblem c s'
'-,.Iatcn The orble I su3b- i 4-ides i n c fo ur s rcie c tonics:
I60
11. '10PUTER SYSTEMS
(I ) determining appropriate design language restrictions for the high-level
functional design language, (2) translating the high-level procedural Ilescription cf the system function to an appropriate low-level representation
whic2h is suitable for comparison, (3) translating the hierarchical description I
of the system structure to the low-level representation, and (1) performing
the ccmparison between the low-level iescriptions of the system functicn and
the system structure. Work on the fourth sub-topic may be ailei by recent
,ork in the literature on an analytic verification algorithm which uses binary
decision diagrams to represent iigital functions. We hope to be able tc
incorporate such a verification method as part of our hierarchical lesign
verification system, since this approach could potentially eliminate the need
for simulation in the design verification process.
11.11 Simulaticn ?ackages for Tomplex System Oesign
In any design work, simulation becomes an important part of the design
cycle. 3inca today's ic.iz simulators are based on TTL gate-level structures,
many MO circuits cannot be faithfully mcdeled using them. lircuit level
simulators require a high computational and memory space penalty and cannot be
used for large circuits. A logic level simulator called FETSI4 for "OS
integrated circuits has been developed. The simulator takes into account
slowdo'n due to capacitive loading and therefore gives accurate tuning
estimates in addition to logic values.
A preprocessor based simulation system has also been developed that
allows the specification and system level simulation of digital systems. The
simulator allows hierarchical specification and logical connection of modules
and comprises about 5000 lines of code (including lirect support code). it
has been used to examine multiprocessor bus protocols, memory oganizaticn in
multiprocessor systems, and failure modes for a multiprocessor system. Since
the system can deal with a multilevel specification, it is fairly efficient
and a real time to simulation time ratio of 50,000 for systems with a minorclock cycle of 10 ns has been observed. Theoretical mcdels for such systems
are under development, and will be calibrated using the simulator.
61
1.kT APPLED CY4PrJTATIM~ !HHORY*
I ?aculty ini 3Seriir _____
D. J1. Brow.n D. E. Aillr ?. ?. ?-reparsta3A U4. 3.~. N'amny
,ralUate 3t'lients
.7.3larii; 1. 3. JOhns~on 4.?racw.h-*
J ?.1 ___________
Tie primary obj-3tive of the resear-:h recortel in this section is the
i leeopment of efficien rt zomputational tachniqus n th anliso th
3apabilities Of Mrru Oi9ls Of coiMputatit)-. We are c-oncearned wi th t he
~r- ceSOU- - sUc2h AS tiM-, ejUin~ent, MetMOry, J-.terCOn~neoti~rn - either asci or
neailin h aljorith.'nL 3ilution of given problems. This am ic I isc ip 1 *ne
- cncreta ::omputational -cm-plexity - Is not only -cmntributing to our basic
er3~n1~~of *oputing, bit it is extremely relevant to actul I racti.j both in hari-ware ani in 3ofttars. The relevance is greater, tecloser the
alot 2 omputation moiels arg tc urrent or projected ,omputing systeams.
ThesreforeP, i t is n natural that thta gre-at te,21nological innovation1 rvrsenelby ' -a~-caene t 11,1 V 3T licuitry h1as areat-iy 'hAl a
.3ib 3tan ti4il impact on the iscipLine, o pen -. i ie,# i n i Pnt p~3in.?
-,~hallang img pr-blims. Th-iiel a substantial pairt .7f )ur researih, *-gaaize1
beo )-% in fouir subsections '-?arqllel ,-rnpijtat4ion j.1 *T4 " , "- pttrra
,Thometr_ " Apprxccitatio-n Algoritms" , inl "3toraga a 1 3cs st3 In
:-if~rtation eiva" iraws its motivation frr-t this important tzlhnoli-4al
I ' i or.<ci ;a suppcor-te by t'. 'ia Inat 3erice 1 3M .?-Zt ronqis ?"'gr3'1.7. Art-, 1.3. T'hvy7 'J. 3. Air ?orce) xiler cc ,traczt 'V) -3-12. ni ibthe 'Titionil 3z,-;ienoe x'iiai l ner~at iS'- ~2 C 1-~~? C
30-)3-t1., :3T 31-1 221.0, ini '-T0-1 22 4 -nt 4S'- 62 IS 032
62
1 '. kPPLIED C0:4PUTATI0l THEORY
revolution.
1 .2 Oarallel Tomputation in VT,31
?arallel computation provides a funiamental answer to the neel of
inzreasing computer processing power. Technological a ivances haie male it
possible to realize parallel computing systems. Our stuiies in this area are
zoncernel with the levelopment of realistic computation moiels, the conception
of powerful interconnections of modules (architectures) realizable in the
chosen molees, -Al the analysis and lesign of algorithms to be executei on
those -architectures. 1ithia the commonly accepted "synchronous model" of 7L31
computation, -ie have stulied the realization of algorithms and have levelope4
are-time optimal schemes for matrix operations F 1 2_. 3 , 1 2 .19], for the FFT,
ani *or the multiplication of integers r 12.201. We have also levelopel a
zritirue ani an appraisal of current VL31 models - basel on a physical
analysis of propagation lelays - ani have begun the e3xploration of
three-dimensional TSI technijues 7papers in preparationi.
!?.3 ?omputational Ieometry
Computational geometry is relevant to a number of applicational areas,
such as operations research, statistical analysis, and lesign automation. °ie
have 2ontinued our lines of investigation, and have jevelopeD, time-optimal
algorithms for a %rariety of problems concerning two families of orthogonal
segments r12.6,12.71; this problem is very relevant to VLST lesign. 4e have
also begun the exploration of a class of problems of planar realizability in
the presence of obstacles, and have levelopel a time-space optimal algorithm
for the case where the obstacles are modelled as parallal segments r12.151.
12.4 pproximation Xlgorithms
'otivatel by the intractability of many important problems, ie are
oncqerned with the levelopment of efficient approximation algorithms. 3ur
work with two-limensional bin packing fl_.1,1_.4,12.2 ] moIe _ certain
sc-heluling applications ani also relates to Tr'3t placement problems. f have
reently 3t-liai 'r:31 3treet routing problems ani have obtainet both a lower
bount on the number of tracks requirel ani have levelopel an algorithm whiih
is 'it most ' -5. -orse than optimal papers in preparationl.
mik __._____________!!!_I-
I 63'. A?PT:ED J>y4?rJTAT?:) _""
2,.5 3torqAg -e1 ialmss co,'Sta for :no~tin eiieval
i an y tiforttiorn-r:tria-val syst-In, 3ffLziant storaave ani azeissia-,o
anir: ssentiil. 4-3 1r- st'ilying partizul-ar lats s.ru: tures? al-i to9
sr'e-cestr e1-off aih-3ernt in1 pqrorming assoc-iit-3 ope3rations. 7nr
1.31 ome~n an lvD 1t of 'an a ysis tichniques vhich can be a-ppiel-_ to
'rirtrieval pro blems, lasing zost me'asures ipropriat-i to the- p-4rtJil.Ar
applicationis.
I~ ~ ' 1~ vv loaui' 2.2,12.151 a general cla3ss of risourie traiaoffs
Viat ar4~3e in such proble?,s as parallel sortingalorithms, linesAr re'cursion
sche~.at', 71,3 Layouts, ani sear-chin5, al-orithms. Tset-a'leoffs so-in a
vari.ety of resources, Allt-ing us to tr'aie -.i-e againt space, areal, Oro-ra
I sZ3, iii nixtber of proce ssors.
A lata structuire :fith locil accqss pe-aits an access point-er iato tAhe
Istruictura to charige its *rau rois location only to in aIjaCent '.ocationi
:n tl 'Iutue :uti-neaa 7u:rin- machnines moil stora-e a re ri: r il
P- -1p .1'3 in thsfae"tutrs
multiii~tansionial ,cI !-s for arrys. da 11Vae isel these3 ntachin3s o veJ n-. nehi o nixizin7 acce Pss poiaters into tree-s an ara. -iar
milti-eAl i tre-3 macotne of tima complexity t'n) can be si~nuliteioin-line- boy a
tree Tachi-ne ;i.ith only tw~o aces ias in t im 3 t~ ' q log 'I) 'loi !(oa t.' -I)'
E" r-yMU nutihea 1-3 eini11na. :nachine ani be sinulatai on- Iinie 'by aI--iin-3ns Lonal 7taichine .3 1ta on ly two access ',l 3.i3 1n tterI ~ t~n'~*(1 i- I /de) log t',n)) -l.~1.2. The, simulatio n frtesi3
op ti4T.a!
64
13. .DV3NCED WUTOXATT01
Faculty and 3enior Staff
N. huja J. X. '4ammer 3. q. Rouse
. T. Thien . ohannsen W. 3. Rouse
) Telong 4. B. Johnson D. L. Ialtz
L. R. Aaran
iraduate Students
4. 3rew B. Goolman N. Aorris
N. Briiwell 3. 'ialien Y. Pan
T. Brooks J. Hammer T. Pollack
D. "hen R. Henneman ". Ieily
3. Tross W. qo D. Spoor
J. Fath R. Hunt ]. Websteri. Fletcher .1. Aarshall . Ziegler
D. Ioreheai
I-.1 'Tatural Language 'Jnderstanding*
Research has continued in both theoretical and practical areqs relatel to
coiputer understanding of natural language and the obtaining of iifor ation
from data bases. Attention was focused on e.rplorin ani formalizing the
r-lationship between language and perception. "'vent simulation", the process
of ronstructing a sequence of sets of assertions to represent the meaning of
sentences, ".as proposed and investigated as a method for judging the relative
plausibility of various real -orli interpretations of sentences leszribing
"his -ork "is supported by the )ffice of Naval Research inler ntrant 'To.yvyjrn %-m5-z-9 ..
65
X5 DVXN',!) VUTTATI)N
in 3aDS in tex-t in the absenc:e of "sCrin:ts". ?itveevent simulation termsa rs being iesignei to orovil . a so un i b a si3 fo r -no lls o f laniguage
q a ~i tion , c-oncept Iaarn ia , language 3gene ra t io(n, ani "cormmon sense'- 3.2?91. -Tqe primitives ire basal ii oart on )ur staly of Southeast Asian and
'.ativer Akmerizan languages, ail the methola they use to e3,cre55 spatial and
eent-relatal information.
IThe problem of *?,valuating natural language sy3t9M3 '413 invIestigated, andanumber of novel evaluation technijue3 ani limensions for evaluation verg
1developed. "Evaluation methods ;wers desig;ned to hielp in the gr ininiig
provement of the tested systems; methods wjere tested on ?LANE9 ail other
natuaral language systems, and in extensive analysis of the test results was
ariiout and documented F13 .371.
3 J}L, the relational data base qu.e4ry language system used in ?XS-ss
3,MDletaly rewrittan to allO-d "coooerstive responses", probability. Ald theI lanninig of opti-nal data base seireob strateies 713.301.
'f:dcias completai on 3ROISER, a system whizh automatizally attemo Tts to
find int-3resting patterrs and facts in a lita base 1 131 1. 73RO'iS- iLs Irivfenbyj a T.olel of -;ihat ;-a inte-,resting, in terms of sti-tistic al natterns, measures
of signjifiance, and models of the real world actions ani objects ahc reI reflecte. 'in 1h3 latin base.
I1 'Jehave also been working on a compute r system that can learn. from L,:
ex oeriancs -.ith the wforld. 'fhile designed primarily for a natural language
processing system, the research is innlizable to other AIL lomains as -uell.
hie technique is called 'Explanatory Ichama Azquisition and is a 'inI of one
trial or insight learning; wfith this tachniu-, a natural language pricessing
syst m All be able to learn bcth the voc:abularies and concipts o fne
.co-1ledge domains irectly f-rom its inuut text.- 7or example,, a niatur alI language 3,ysteM .hit '.ows nothing about, say, bolacail -,rll be able toi learnthat conlent frma iaput story describing a blaicaail entini someT al.
IT-itoretizal -,ork 'has been cairriel tout on the use of novel "message
passing" algor:ithms on a hex-agrnal array for computing h-ijh lv'vsa
I features "eg. srnmetr-i and shapq of Objects, anll fo.r lfnn prcitj neighb 7roods fo r obects In sc31enres 3 sults hre3 Ia broa
66
1 3. ADVNCED AUTOMATION
significance, -.3g. for VLSI design of high level vision processing hardware.
13.2 k. (nowledge-based 3ystem for Intelligent Monitoring and Diagnosis*
The goal of this research is to develop a knowledge-based expert system
capable of performing intelligent diagnosis on engineer-designed systems. As
traditional expert systems have been criticized for lack of understanding
about their working domains, our effort is to join the power of expert
heuristics with the completeness of general domain knowledge.
In our approach, an expert system acts as the front-eni for a disgnostic
process. it efficiently postulates plausible failures to explain an observed
abnormality based on the powerful, yet not perfect, diagnostic heuristics
acquirel from human experts. In the background, postulated fault hypotheses
are scrutinizei by a rationalization process. Using a constraint model, the
global effects of each hypothesis are thoroughly deduced and then comparel to
obser"red sensory data for consistency.
By emphasizing coordination between proposal and verification processes,
a basic expert system can be enhanced by the leep-thinking intelligence. The
philosophy of our intelligent expert approach is useful not only for
diagnosis, but for other expert problem-solving systems as -well.
13.3 kutomated kir Traffic Control"*
Air traffic control can be sublivided into a set of tasks each having
only one primary concern. Such tasks include the avoidance of collisions,
dangerous weather and restricted airspaces, ani the control of flow rate to
airports. "However, a set of techniques to perform each of these tasks
inlependently does not produce a system which performs like a human
controller. What is requirei is an architecture that supports a constructive
dialogue among the experts comprising the system. 3uch a systim has behavior
that lepends on the power of the expert interaction language.
This work was suprorted by the National Aeronautics and 3pace Xdmini3trltionunder contract "3S Art I-9.
0 This work was supoorted by the U.S. Department of Transportation inderContract D)T FA7e-iA- 3BO.
Ik AL
67
1 3. XDVV5L'D .V1J'ATfl'I
D'Ar s93ar-ch has proiuce-I te-chnijues li:{e those usei by the e'parts, As
I iell a3 a planning pro:cess involving the u.se of critic isn ani suggestions from
a boly of' experts. ixperts constantly mnonitor traffic for probleM3 in
Darallel. 3hould an -axpert 1etect a problem, it plans a solution ani submits
it to the other experts for analysis. This proc-ss 'nay cycle before a fiial
solution is Sio~tei. The experts share a common conceptuail language
les3Cribing air traffic and each knows its own relative importance in1 th'e
control task. Thus the experts can ' arg-ue" and iecile priorities. T"his3
architecture amphasizzas molularity while maintaining the 1ohesivle natuare Ofhuman controller performance.
1 Still the system will not improve w.ith time uanless it has a learning
capability. hur extensive revie-w of' literature On learning yeiei19 no g-eneral
t '1he3or y, althoug-h some simnie learning has been Iemorastratel. 1Jo rk ~s
coninung to ievelop learning technijue:s to improve expert perfortance
through experience9.
!-).I~ Theoretical ThIc-put-er 'Tision*
Ak book on the general subject of modeling snati-al patterns was comple-tell
r 17.i11. Most of the book is levotel to a iscussion Of the propert4is,
synthesis, extensions ani perspective9 iisnla1y of11 mosaics, c overage pattern--s
1-1 10nd- logrestel 4atle patternis, whqich ar'e e'cnlore1 as mnolels to p"lAnar
t-?-cturqs. *,I,-I,- bascI taxtura measures are also liscussel.
I X~pproacnes to planar iecomnosition for neacia ~i~roe~tta-.,ar '. uI i r,, Square ani triangular juad t-rees were found to be theiI only faible methols, having the same c-omputation-al complexity. For jr*Il
imajes, the type of gril ( triangular or saquare') ieter-mines the- appropDri 1 te
1-.al trse t-pe 'triangular or aquarg).
1 new4 classiflication Of low level1 image T.01-3l. -413 DrOposel 1rIir !t-Ie
I ategori-es of pi-cel based an-i region0' base-1 models 713.141. T-raii-ionl! ojl
were sur-reye-i within this framework.
I , T-his iork was 3upportel by IM,' inier --rant ~-1~'~
68
I . ADVANCED XUTOAATIOY
1.5 ?ilot Interaction with Automated Airborne Decision Making Systems*
The purpose of this program of research is to explore the issues
surrounding problems associated with human-computer interaction in flight
management. During the last year, three projects have been pursued:
multi-task decision making, measures and models of planning behavior, and
onboard computer-based information systems. The multi-task decision making
project has been involved with applying the modeling and function allocation
notions developed earlier to realistic full-mission simulations r1 3.19 1 . On
the basis of a previous review paper ri21, the Diannina project has evolved to
a series of experiments to study the planning process of aircraft pilots in
emergency and abnormal situations -13.25,13.26 1. The computer-basel
information system project startel with an initial fairly simple experiment
1r!3.1 ] and resulted in the design of a system for fuill-mission evaluations
1-1.27,13.291.
13.6 quman ?roblem 3olving ?erformance in Fault Diagnosis Tasks*
The purpose of this project is to develop an understanding of human
performance in fault diagnosis tasks as -ell as ails for use in training
humans in such tasks. In the last year, two large experimental studies of
mantenance trainees were performed. The first experiment -as the third in a
series of studies of human performance using computer simulations of aircraft
power plants. The goal of these three experiments wias to study how suboptimal
performance is influenced by a variety of factors r-il,1 3.551. The second
experiment performed this year involved a second study of transfer of training
from computer simulations to real equipment.
13.7 Modeling of luman 3ehavior in 3eeking and lenerating Information**
This project is 3tuiying online use of computers by scientirii aersonnel
in two areas. The first involves the use of online editors for preparation of
* This work was suported by the lational Aeronautics and 3pae Administration
under Grant No. N1S-2119.
Tais work was supported by the U. 3. Army under -ontract 'DA 13-'--31)4?'.
. This work was supported by the lational Sci nce Foundation under grant13T-791 7 1.IdMOfo- _-. 1.-.tIMU0n
69
1 3. XDVAN"CED kU.kLTl)'4J~
pro grams and locuxments r1 3.3 1. The second involves the onlinle rhigof
f till-t-?xt bibliographic l~atq 'oases r*13.351. T1he goals of both of thIese
subprojects include ieveloping math-3,nati~al models ofl human behiavior that ar,
3Uitable for use as lesiin aiis.
References
'11 1. AI. 13unt, "The Hluman as a Pattern Recognizer and Thformation 3Sekacr i.n3 Fault; Diagnosis Taskcs," ?h.D. Thesis in progress.
1 21 1. Johannsea and W. 3. 'louse, "Mathiematical Conzespts f.o-r :oleling Humani39havior in 3omplex Aan-Maihiie 3ystems," !{uman Factors C5, -7
I (Dcc 1also
70
1. DIFOPMATITO7 RETR!-VXL IESEARCH
Faculty and Senior Stiff
I. "E. 1illiamls L. 4- Lannomn 7. 1. iobins
J. L. Divilbiss 3. 6". Preece J. . Smith
3raiuate 3tuient-s
I. Cheng K . Thuin L. PreeceI
3. Aenlo za 3-.Nn
I I:ntroluct-.on*
During the I 930-1 ?31 time perioli ah I nfo rna t ion ietrieval Resear->na
abr!AtOry R:RL) con-iucted a number of research and levelopnnent -arojects ani
!Jirec:te1 tla operation of the TUni-erSity of illinOi3' online seairch ser-vi:!e.
Aajor activities includei analysis oil database latin: a tes t of t .Ie
fe3aslbili-ty of auto-natica!ly determining the overlap b-?t we -1n b ib Ii o rap hi
latabases; development of a computer-resiabla3 iatabase lire-ctory: mln
mnagem'ent and lireition of the University of Illinois' online searc2h serie
1.2 Analysis of Database Data*
The 2-RIL maintains a body of information about commerially available
databases :callei the Database of D)atabtases. 41e have :cOntiqu-il tto 3t-ly
various statistical characteri3tizs of this population basei on thec, ma t 1r i 1
ii n ur latabase. 3aih 3tatiatlics ani anl S- are isfi. both toreeachr
,the area1 and to the Ase-r3 Ind oro'luCe3 Of ltbss
Thes nartiiular information levelopel in the Stuliies performei In the last
year 3onsilerel the age, size, tyoe ( scientific , melical, e3t::) ,aisu
govermment, private , eti .)~ of the latabasi-s zoversi. xt prisent * 2
4 This oork iss supportil by the University of illinois.
I71
3 o. I MFORNATI3 R;ETRIeV'. L RET3AR!t
latabases are zoverel by our latabase of databases, but any given study Tight
I inclule only i 3-pecifiel subset of those iepenling on the needs of the study.
The most recent analyses are for the year 1390.
i The software used in this research consistel of a program framework
containina slots into which logic couli be inserted to perform specifi lata5 selaction, cleanup, and analysis. The data "as first extractel from the
database by our data access program USER and written into a file of (tag,
m vale) pairs. The analysis programs read that file and selected the tags of
interest. an effort was made to overlap multiple analyses by re-using data
.3 selectel and cleaned for prior use.
I-oss tabulations were male basei on iefined groupings or actual values,
sepending on the iata involved. Thus, cross tabulations veri lone involving
both ra-w year of origin versus groupe-I size and groupel year of origin versus
? roupe 3ize. 3tatistical summaries included means. ranges, and standard
deviations of raw data and correlation coefficients between raw lata ani
3,rounei lata. Detection of bad data included iienlifiiation of records
contaiaing illegal or missina values.
I The %ork involved in this project also developed a framework for further
studies of the iatabases available to researchers today. 71e are currently9 uoldatin the database and improving the coverage and consistency of lata
fields that are likely to be important in future studies.
I 1 1.3 kn Integrated X._an/"achine Interface to Facilitate Net work ResourceUti! iz-ation*
I For the first time in history computerized information reri iva! is
.iiiespread and economically viable. Evidence of this is shown in the volume
of machine-readable records, databases, online so ftwa re ackaes, renlor3,
search services, and searches. The major portion of the currently ublisheli
scientific and technical literature can be ilentified throua:h -orput er
searches, because the reiferences are in -omnuter-readab> f,;.-. The iajority3 of the iorli's currently publishel abstractiag and inlexin literature L.3 1
' This work 41s suoortel by the "National 3cience 7oxi c tn il er Irantl 1
72
14. TNFORMATITI RETRIPEVL RESEARCH
computer-readable form. There are more than 3D0 publicly available databases
containing approximately more than 1T5 million records (citations). lore than
-5 million of these records are now available through the various online
system vendors in the U.S., Canala, and Europe. Some of the major vrendors
include: the 'ational Library of "4edicine ("'L,4) , Lockheed information Sercice
7'3), System Development Corporation SDC), Bibliographic Retrieval Service
,BRS), the Canada institute for Scientific Information (CISTV, and the
,Uropean Space Agency (-SA). These online service organizations use a variety
of sophisticated online software packages such -as LIILL, DIkLOG, OR3 7.
3TA:RS, kI-0 , T 5R, and RECON. Services from the latabases, through the
online vendors, are provided by dozens of centers, libraries and brokers and
they conduct nearly 6 million retrospective searches per year.
Unfortunately, there are no standards governing the foruat of databases,
-he elements and vocabularies included in them, or the online systems for
searching them. The lack of standards forces searchers to become familiar
iith each existing database's coverage and vocabulary and with each system's
access protocol, system features, and command language ani to keep up to late
afith chances made in all of them.
Iany of the differences and variations ian be male less arpqornt by
developing translators or transducers for convertin, the procedures,
conventions, and terminology of one system into equivalent proedures,
ionventions, and teininology of other systems. These converters "gould rake
systems appear alike to the searcher and make databases function as if they
used the same vocabulary. Such ails would make the differences transparent to
the users; thus they can be called Transparency Aids. kn integrated
collection of transparency ails can be said to comprise a Transparent System.
Tle aids then are the Transparency Subsystems.
.fith the reduction of lifferences the databases and systems become easier
to use; the searching environment becomes more user-oriented. 7n a Tre
iser-oriented environment the requirement "or intemeliary searchers decreases
and : e begin to approach the day when Most searching will be done by end users
.iose that nose the search questions).
73
'h~ J jsarZ I program. stuliei methods to Lnzrease -he trsn3-0r y o
r 1 rer iea a stfens by .nakiij the lisoretenaSS Off 3-:?te ta 7Sk3 less obvious aoliby macini the Variety Of system and latabase iiffere nces le ss Obvious. '4i1t hin
:f13 pro jec t wde ')) esignel a veneralizal model for a 7-anspa-enit Sys3tem,
? examinel alt-rnatire factors affectino, toe ITranssarency interfaca,
built ::r-r o n en -. of a iser-oriented terminal, and ( ) inte~grati example
ransparency Subsystems.
The Transparent 3ystem is oompnisel of: user interface Dro~ra.M3;
2 )omandI lan~guages used by the Transparent 3Vysteam; tr:anspareftcy ails
3 P-3rformi-g vIarious functions, such a.s :conversion, o f co-mani lan au a~ --a3or
aselac ti on ofI latabases; data files zcntaiiing lescriotie ani statis3ticalI
data about latibases; 'he coannections to iatabases and 1 nt -o r'K reSOuIrces .
~h -a -ncial o f thea Transoa rent System lays out the relationships a-amo these
lemnts Th ateqtivreS stuly -,cplo)rgs advantacles an 1 14 alvantaoes f
vart"ous Mol :!onfig-urations. 7De'relocment of a nrton syst am
Iemonstrat3s the alvantags of a 7ransnarent System by s3hoiiin, howi subsystems
can inte ract in Der-fo-ming their functions.
The 9 ener-ali1ze m ,odela includes a co-nr-ihensivre 11s' 3 Of -3,c i Stig
I propoise-, and Dotenti-il transparency ails along *Aiith "aestimates of their_ntercoTmTuniCat4on recquirements . The alternatVes reie tIncnid h
f bandl 'i-ith of c:ommumication paths required for various -oigYurations andl the1-
e3ffects on resnonse time and performance? of -han-es L -n f ongurit ion . 7.-he-
?rototyne for the Serhe Vorkbench ':19)implements a struc:tu.red lialoguex
bet;yeen the user ani a mcl seaarch system. The dialogue is then translatedI to the fo rm required by re al --o rli systems to be used. 7-he- t 7u Zt Ur
idalOgue integras aICss to t 'o -3 iS t i 10 transparenIcy ids1, A database
selection adviser and a vocabulary transformer.
I.1 D)ireactory of ,omutar-Raable 3ibliograchiz Databa~ses*
IThe 7omnouter-1eadable 3ibliojranhic_ Ditabases o \ ieto r i anl Da:-taI our:!ebook, cot il -d anl edited by Prfssr:artha "7! viiia 'and ara
Rusea, -la s pubisd i1131-n I~ by (no ;,,l-e a ndu ,iis ry ub 1 2.jt 0n,'1 n1!c1. joa t, 3
the Directo)ry -..rre isued in ' 1ri 17 and ktri4 I I aw re d - r,
I ~~ This iork is3 partiall.7 suaoortil by Kno-,ilelge :nistry PbiainTc
74
74
I. :-NFsiT4A"2DN ETRITIAL -33VA1RCf
including data that was relevant as of December 1973, was published in autumn,
1171. 7qe 3oMDuter generation of the printed proIuct is sunportea by the
-American Society for Information Science. Uhlpating of the inderlying latabase
.n preparation for a new elition to appear late in 1931 is now unierway.
Data 'ollection
1. Database processors in Turope and North Imeri:a were contactel by
mail for information on the latabases they proviled access to.
2. hei 13 Dage questionnaire lesiznei for the first lirectory was
revisel to achieve a more consistent for at anl expandei to inclule
non-U.S. data. A turnarouni questionnaire qq being
*.onutr-oeneratad, containing the information in cur files for the
databases 2overed by the first lirectory. 3xisting data entries need
only to be verifisl or changel and Missing info-,mation supplied by
the database producer.
3. The questionnaires were mailed to the database prolucers ani U.. ani
7anaIian prolucers were subsequently contactel by telephone fr
vrerification of the data and assistance with questions they night
nave.
. :nformation on the availability of new databases was obtainel froT
tne publishel literature ani contacts with latabase
processors/producers. T he producers of new latabases were
interiewel first by phone. The copletei juestionnaire irs then
mailed to them for verification and if necessary, followel up by
t?!ephone. )nly online latabases that rere publicly available were
treatid fully. Databases available for internal use only or "which
are only accessible offline will be listel separately.
5. qeturnei questionnaires ere checkel ani lata entries stanlarlizci.
Data 3ntry
1. Data will be enterel online into the latabase of databases file using
the D3IN program. '-1itries can then be printel out, verifiel, and
corrected ani eiitel as necessary Using the DBEDIT program.
75I . tri ORM'A2-:)P RETRIVIL RES--AR-'H
)ranization
IThe lirsectory is or-ganizei alphabetic2ally by database name and in !u ie
an1 in t -o Iu2 t io I ani fo ur in'iexas: Su.b j-c t Index, Database
3ame~croym/Smonyn Tiea, 'Oroiucer lilax and ?ro7:9ssor :-iex.
-ifonmation oni each latabla f0ollo4s the -9neral format:
(i Basi.- Information
3 2) ro*Iucer/istributor!/',eanerato, 7n:form-atijon
- 3) -railability and Thar-es for Database Tapes3
(V)' uj ect 4atte and S-ope of Data on Tape
,3) Subject -Lalysis/:adeJia. Data35) Data 219ement3 ?resent on T-ane
()Database 3ervices Offerel
1 3) -ser Ails Available
? rn iuo tio n
The biblio.-rachy will be 2omputar tyroaset u ain~ MRL 3o It-war,: al; a
.3ne i al printer on r1IL's microc omputer syste-m. kil- inlexces 4i'Ll then- be
.-e-nratel aUtotfat42a'Ly f,:011 the latabase andI type set similarly.
11. 'Jniversity of T1lnjois Online Se3arc!h Sri
During the past year, the University of ill.inois Online Seariha r'ie
-inl -ar the gene:rAl su1per-iision ani coordination of :RL. expalleli i1ts seri:es
to the stuients, faculty and staff of the university. !a- 3ooperatlon .qit', the
Jai'reqrsity Library, searching is no-i available at 11 jiferqeqt 1 atirs
I arouixi the ,,ampu,3, w.ith f-irther gro-ith pgniin-,. Loc:atiOnS el-rchiftg is
offarel *to unves t at'nt.3, faculty ar.1. Staff inclu1de 77t" eferneRoof the '4ain Library, 'Themistry Library, 3iology Library, Labor and :-ni as trial
Relations Library, Education and Social 3Scienceis Library, Agricuiltuire Library,N 'atural Iistory Survegy Library, ?hysic!s Library. Vetrnr eiie Lbay
H-ealth SCi4enceqs Library, kjric ultural 3coio-aiis R3ealinj Room, T.ecooY Lib)r a r -.l vil the? 3eolojic ai Surveay Library.
aTis fork -iaa 3upportel by the 3sta Of :llii.
76
14. INFORMATIJ RETRIEVAL RE3EARCq
)nline systems available through the 3earch 3erviie include !OCkheed's
DIAL31, 3ystem Development corporation's ORBIT, the Bibliographic Retrieval
3er-i ie system, and the National Library of Medicine AEDLARS system.
-niivilual latabases available from these systems total 176.
Use of all systems luring FY 31 is given in the following table. The
osts given represent the total amount pail by the University to the online
vendors :oncernei. Approximately 90r of this "vas recoverel from the users of
the 3earih 3ervice through the charging of fees.
USE OF ALL 3YST4MS ?Y 91
Tonnec tHours costs
31S 364.D? ?9,596. ?1"Rh 3 :,.36 11 Ii.3ql
SDC T1 .73 1 ,310.00NLA 5.5, 13:3. o
70TAL I 07.16 ,I ,4D. 35
2he total connect hours 'isel represent approKimately 2110 inli-7iiual
s3arc hes, a search being lefined as a single session at the terminal,
regardless of the number of latabases or systems accessed. Of these, -will
)1er iere 2onlucted and paii for, at least in part, by University !iculty,
stilents or staff. The remainier iere coniuctel by 'TIL or library stP'if
either ac Dractice 13 they learned to ise the system or for jeneral rfernce
or !erifi'ition iork.
3eyoni providing general direction and coordination, iRRL performs the
folo0-in specift! tasks relatll to the Online Search Service: negotiAtion of
-ontracts with online vendors: training of new searchers: consultation on
system problems or lifficult searches; keeping statistics on all aspects of
the SearscS 3ervi-e: accounting -ork associatel "with ollecting fees "-ni
paying bills.
J177
S14.. 1rF.ORMATrJM ITRTEVL RTEARCH
1 .6 Oatabase Analys, A
I TRL is coniuctinag ongoing snalyses of the '4ARg database. hi -iork
involves processing the annual files listributel by the Library of 7ongress
and generating a ietailei statistica l analysis of the rate of occurrsn e ani
size listribution of each fiqll of the recordIs. The resulting t bles, 'hi h3 have appearel in the Journal of Library kutomation, show for each fieli 1
the percentage of records in "which it appears, (_) the rata of occurrenceI iithia recoris where it appears, ani ( ) the -.ngth of the fisH in
:haracters. For (a)) ani (3) we report average, minimum, ani msxinum values
3 and staniarl ieviations.
The listribution of records in Dewey ani L, classes is a13o reortel
j along ith average record lengths for each class. -Jummary stati3tics are also
listel separately. Finally, wie report the changes between reporting periols
(show-in g, for instance, -which fiells are usel more now than at the last
renort)
I -e have also begun a secoil oroje ct using the same :ct-puter programs t,-
..alyze the "ONSR latabase for the Jouncil on Library Resources.
These lata allow users of the iatabases to optimize their anpications
for the actual haracteristics of the databases.
1 1.7 lutomatic Detection of Database OverlaD**
This research program addresses the problem of overlap coverage of
journal articles among multiple databases. Trerlap loals to lupli-ative
I processing of material by database producers; luplicative orocessing ani
storage of material by online and batch venlors of database se-rices; ani
J retrieval of luplicative material by users of databases. The net result is
reflectel in costs associated with the databases -nI an unnecessary
1 ependiture of time on the part of users.
.1his fork was supportsl by the council on Library iasources.
.his work was supportel by the 'htionl)' 3ienl c ouietin 1aler >ant- )l 3.
78
14. INFORMATTIN R2TRTEVXL RF,3EARCH
The true -xtent of the problem is unhmown. 3everal studies have
ailressel the Juestion of journal coverage overlaps, some have addressed the
question of article overlaps in narrow subject areas, and one has attacked the
problem of identifying maximum possible overlaps among specific databases.
lone has studied the problem of determining the actual overlaps among database
services.
4e are developing a methodology for automatically determining overlap
among databases and testing and implementing this methodology on seven
latabases: IGIJOL4, BIDSIS Previews, CACon, ,xcerpta "elica, :edline,
3cience 2itation Index (3'.), and Social Science "itation llex (3SSM). The
methodology involves analysis of the databases to determtine the data elements
r-esent in anl common to multiple databases in order to determine effective
keys for matching records in multiple files. The matching process will be
lone in at least two passes. The first pass or broad screen will se a match
key with high discrimination Dower to identify zandidate duplicate records. \
set of finer screens will be used for verifying pairs of r9ecords -s
dunlizates. The second pass will enploy 'farrison keys (which are bit stricg
renresentations of titles) and Hamming iistance techniques 'for overcoming
slight variations in representation of titles) together -ith matches on other
elements as needed.
Results will be tested: (I ) by using manual verification of a
statistically representative sample in order to letermine mismatches, I'li
by using a mo-wn set of duplicate records to see if they are ilentifLel by the
screens. The evaluation in terms of mismatches (wrong matc:hes) and missed
matches will provide a measure of the effectiveness (-f the techniuies.
Further evaluation in terms of gross costs ill be carried out in order to
determine the efficiency of the techniques.
The study will produce overlap stltistics among seven atabases, lata
element freqiency statistics, and a methodology for luplicate etection that
might be used by database producers or atabase vendors. The resultant
methoiology couli be usei as i Dreprocessin g step to reduce storage anI
processing costs Or as a post retrieval step to eliminate the dulicative 1
output deli7ered to users.
4 .. -
1 791 ~,. ~OR4A T)I W~AL R3SMAR"H
Yii thia first year of th-3 project tapes ?iavi been icqu~irgi fro' the
iaib-se proIa::-'s zoveri-.g th-3 1'173 - IT?? perioi. 3vimola subsets h.Ivvz er
reai ai pr-ilminary planniag for ths file stvactur,?s ini Dote:ssiig flow h-s5 be-en lon . Th -e contents Of the SaMpleS hsRVe been Stldiiel' 3t~ti3ti-SIy to
ientify iti seements %xseful in the matching~ proceSS a,1 to allow es3timation3of storage rsqiiirem ants.
80
15. COMMUNICATIONS
Faculty and Senior Staff
T. Basar B. E. Hajek M. B. Pursley
T. U. Basar J. C. Kieffer D. V. Sarwate
A. H. Haddai R. J. McEliece S. Tantaratana
H. V. Poor
raduate Students
B. Aazhang F. D. Garber C. X. Stadler
C. J. Alston 3. A. eraniotis W. E. Stark
A. C. Blumer J. S. Lehnert T. G. Van Loon
C. 1. Chang R. '. Ogier K. S. Vastola
R. Cruz Y. Rivani S. Verdi
?. Bn1e M. S. Wallace
15.1 Xultiple-Terminal Digital Communications*
This is a major research area which includes various problems in
spread-spectrum communications and random-access communications networks. The
topics that were investigated during the past year include the performance of
spreal-spectrum multiple-access communications systems, the effects of fading
on spread-spectrum communications, communications in the presence of jamming,
conflict resolution in random-access communications systems, packet
communications systems, and signal detection in multiple-access Thannels.
Progress in each of these areas is described in the subsections that follow.
'his research 4as supported by the Army Research )ffice unier grantDAG29- 7 3--)114 an cont~a!ts DAA329-31 -K-762 and DAA123-i-K-)06d, by theJoint Seruices Electronics Program under contract NO00! -7- -.0421, by theNaval Research Laboratory under contract NOOO1-3O-C-3O02, and by the National3cience Foundation under grant ENG 79-06630.
81
I I31. 3preql-3pectrum IulItip Ie-k~cess 'cT=trunicqtions 3ystetts
')iaring the Dist 7ear iie have obtliine'i Tany new- results on the problemt of3 evaluating the average probability of error for asyn:!hronous1 lirect-sequ-ence
spra-spatr.in mul tipl3- access (3A)cotmxnications syst,3ms. Because of the
i aturs of the multiple-access interference, standarl numerical integration
techniques require in excessive amount of *omfputation in orier to proviie
aD~roximations that meet a prespeci4fiedl egree of accuracy. We have 1evelopei
'unner and lower bounds whi::h ::-n be evaluate-i with a moierate, amount of
c-~nutation f,-)r i i rect -se que nc - 331A systemns wti th 'ip to fo uir o r f ive
3im~ult-aneous transmitters. "-he bound-ms are Dresenate1 i-n r9 1~ alon- w.ith
nimeric3al resu. t3 woich jistrite thei4r applic ation.
f 'i hve a7'so levelorei accurate3 acnroximations to the average probabil-ity
-t rror for TTI-s~inc SA syst3ens. These approxiTnations are basel on
I the inteljration of th c:hiractaristi- finc tion of the Tiul tin.l:-accqss
Lntarfirence-. 7h3 primary iivantage of these approximations i3 that: the
amrount of corTput ation requiredI for their evaluation growfs only linectrly as th
numTber of s imul taneo us transmitters increases, ihereas the -amount of
I ortputationi reiuirel by 3tandard numerical integrationi 'and Tmost othier %movm
-C 4ima I on techniquesq) in-Creases icPOnajntially7. :n~1.~ -erestI rsuls on t.1. approxMation of the average error probability for binary ?3"
ii~e~t-seuenc 33'A systeMS. 3,Xt:sI'On3 to quat-:rnary s-ystems ith sib
o -ffse-t i-r'hrase and qualature ompoa-2nts ir? given in r1 1. -,, "Il we lso
allow ciD wavqforms of arbitrary shapes s3o that the, mr.1e includes as specialcases the1 forms of iirect-sequence modulation kno-wn as quair-phase-shift
:keyiig / ?3X), offset )PS(, and minimum-shift kgying ('4SK). 7he approx-ination
me9thol thait ie present gives sufficient accuracy for nearly all ipplications.I tevr, if necessary, aven grear accuracy can be achnievei by acn ti
-Atzro'cimation to obtain an expansion point for a Taylor series rceetto
ofth atul robablity of error. By employing thi 4 co2mbi.nation oftha:-racte.ri3t1' fiac tiorn methoi foll.owed by a 3--ries-ixppansion Tethol, -fe -In
3 ob t ai n a n y p r s e i ~ ' e r e o f a c u c y
71arthar r9eSU t 3 ha1-,e Also b-3en ob tine i on thie ce r ") ant aq
1Jirect-3-seii 3T1A sys3~ts ith respect to signal-to-noise ratio) ani el
arametirs. '_ I I~ the signal-to-noise ratio) is iivestig7ate ?I for aiuI t-fmes of)7 ~3 K ani ffset T33K systems. 'ncriclr the effe tsa f
82
15. COOMUNIATI )N3
double-frequency terms are consilere4. n r 1 .2 1 the signal-to-noise ratio is
analyzed for general quaternary D3/SS4A systems. Numerical results are given
for P3(K, offset Q?-3K, and XSK systems. In r.15.10 optimal phases (with
respect to signal-to-noise ratio) are determinel for m-sequences of periods 31
and 53. Using the phases ietermined in [15.101 and the system analysis of
_15.21 we have shown that the signal-to-noise ratio can be significantly
increased by careful selection of the phases of a set of signature sequences.
Our subsequent work on the probability of error 1e.g., 15.391, 311., and
1S1) has shown that similar improvements in the bit error rate are also
obtained. T-he effects of the shape of the chip waveform on the performance ofVirect-sequence *S'4A systems has been investigated. Results on the sine pulse
Irs. the rectangular pulse are given in 15.21 and Fll], and a more extensive
stuiy covering various pulse shapes is presentel in r>5.641.
NTearly all of the analytical results on the performance of
Ii-aet-sequence 3S1A communications systems are restricted to systems for
:wnich the number of chips per data bit is an integer-multiple of the period i
of each signature sequence. That is, there are an integral number of periods
of the signature sequence in each data bit. ?or certain applications,
however, the necessity to discriminate against multipath and to reduce the
probability of interception iictates the use of very long signature sequences.
Ty{pically this will in turn necessitate the use of direct-sequence modulation
with multiple data bits per period of the signature sequence, in rhich case 'I
may be much smaller than p. In a DS/SSA system with multiple iata bits ':.r
period, the performance depends on the partial correlation properties of the
signature sequences 'rather than the aperiodic correlation properties a- ii
the special case p 1 1. In r13.521 we analyze the performance of a D3/3'4A
system for which p I and N and p are relatively prime. This analysis takes
into account the effects of partial correlation. in particular we :ompare the
multiple-access capability of two -lasses of D3/S3'4kX systems with I chins per
b:t when the pe rfo rm ance measure is the mean-square value of the
cro3orrel a ton. 1-3 fini that the multiple-access capability is less for the
21a33 in Wiich p and I are relatively prime than for the class ith one hit
per p-rioi. This is primarily because of improvements that result from the
optitization of the phases of the signature sequences in the case p = 1.
83
315. "OY4M T 1113)N
in ailition to the -io rk Ie sc ribe I above on a IIre t -se quance 33'A2omm~unications sys t ams, several inv as t i 5a t 4ons o ' freqency-hopei 371 k
*ctmxiizti I ystas 11 e also been coiuletal. 3oth fs-rqec-onn£ 1 5.331 and1 slow-fr-?quancy-hopping hiave been :.onsiierei. ?or slow- frequancy-
haoppiviz we have irnrastigatei DP-34 F1 3.351 ani 73-4 "153. 1,01 lata rmolulation, ani
resul ts have been ob ta in-3l fo r hybril iiet- sequxence frequency-ho ped
srreal-snectrum onsem 15.30' ~ ct'i 1h.uiain
15.1.2 _______ o F'adin,3 ____n___ _____________lomuicaic-
3ever-Al of our current research projects are -oncerne1 with the e3ffec:tsI of fiing on the perfor-nance of spreal-spectrum zotmnuicqtLonsrI syste~ns. Th e
faling may be either selective or nonselec t-re -and. either Rician. or Rayleiigh
3otlh frequency-hoppiag and lirect-sequeice sprea.1-spectr'im syetems aive been
2onsiierqi.
:n 13. 40] bounds ani approximations -are leritei TOr- ta e average
probability of erro r I n a n asynchronous 73K sLo-fraquiency-hoppel7,1 k3~
comnuniications system. Three ty-oes of channels are 2onsiiarei: the nonfalian
aiiire ii-e aususi-ti noise zhannel, the nonselecti-re Rician faiing ch-annel.
ani "he ,obyslci r.ii -sense- stationary uncorrelateal-scattaring 'Tz73J3'
fa Iin-? :hannel. 7ie t-vo types of ho pping pqtte-rns Tonsi 1ere 1 3re
lietertnistic frequency-hoppi-ij patterns basal on Reel-3oloton ioles an I
rainiomn frequency-hoppinga patternis. iesults for nonselecti-ve isyleigh fin
are obtainel as a special case of the results for ionselective Ricia-i fiiling.
-n .15 .351 we investigate te perfk . ance of slo-dfeunyhne 31
1sini 73K( iata -noi'lation. The channels 2onsiierei are Rayleigh 13STJ faiing
channels. ?recquancy-selective as wall as time- selec ti-re falinag -aarnels Are
2-,nns. erel fo r syatems -.ith various I ata puilse -davef-)-ms a n d h I-Ann ..
orrelat~loa f~iac-tions. Analr-ical results for ta-e avirage probability o
error are preseqntel, and nutmeric al ivamtples of tlie most -izportvnt caIs.q 'Are
"a e perfortance of lire:ct-seq'Ience sipreqi-spec ,tr-rn 2ttunLzAtioris -a
snela-tiltnathf-aiI 3a nvironnent is invqstigat-eIi h .'. Yi bmn
met,.ois o)f RS. : -tco ere Ieveloptal for zlil-ces:?f"le r
inz i i se ilirtuvl ti4path i-iterfersnce in 1. . :tis tanomf thait )ir
an~r~:taiofto th3 :robsbilhJt7 of error tha-t i.3 'asqi on thes. na-o Vos
84
1 '3. 30O(JNIO ATJ9 S
ratio ' 1 . 2 ] also gives very gool results inen applied to the
specular-multipath problem.
Inhe improvements that cin be obtained by employing error-correcting coles
on a slow-frequency-hoppel SSTA system have been investigated. )ne conclusion
of our results (e.g., 15.401) is that in a Rayleigh faling channel, bit error
rat3s on the order of 10-3 to 10-5 cannot be attained in most
sow-frequency-hoppel 3S'IA communications systems .ithout the use of
error-correcting coles. -owever, the application of certain error-correcting
-oies -will bring the bit error rate lown to this range.
.1.3 Tommunications in the ?resence of Jamming
Last year ye began an investigation of the problem of communicatiqg in
the resence of hostile jamming. 4e have formulated the problem in
game- theoreti : term s, using Thannon'3 mutual information as the payoff
function. le 3howed that under certain restrictions the players in this game
('hi coder and the jammer) have simultaneously optimal (saillepoint)
strategies. These strategies are generally memoryless, ani so in an
information-theoretic sense the commonly-used interleaving strategies are
3urerfluous. Je have also letermined that in the presence of binary zoling,
iniform Iaussian jamming is in general inferior to pulsel jamming. +ofev r,
.ie tfon. that Droviled a molest amount of bani-rL.th a'pansion is tolerable,
suitabl coling can neutralize a pulsel jammer. These results are present~l
in 15. 19 for systems which utilize phase-shift-keying modulation.
?or the case of frequency shift keying (FSK) modulation ",e investigated
the capacity in the presence of partial-band .jamming and nonselective faling.
fe found that there is an optimal positive code rate that minimizes the energy
per bit required to achieve reliable communication. Also, -e showed that for
low iole rates ',' .53 suffices) uniform jamming is optimal for the jammer.
?inally we analyzel the performance of a specifi-, :oIig 3 -heme f)r
nartial-bani jamming. 7hese results are present!?I in 13.531.
19.1.1 onflict Resolution in Iandom-kccess oTmunication "etworks
.onflizt resolution consi3ts of strategies ni-h allow remote 3tations to
share a communication channel of limited zapacity in a lecentralizel f.shion.
fnMich strategy is ippropriate lepends on the iemands of the stations ani on
85
the feiback information available to the stations. le have ierirei innDer
bounds on random- access through-cut, and -we have ilevqlopqi raniom-access
strateg-ies for- situations -mren fieebaclk is severely limiteai and under several
type-s Of statioq demands.
\bounlin- tachflique inspirei1 by the stata-space approach of stochasti.c
control is iitrolucei in F I.~ t is used to establish -an upper bound of.1 25 Dacksts/slot for the throughput of a time-slotted multi-access boroadcast
c*hannel subject to an iifi-nite population of user stations (Tolelel1 by a
?oisson proczess) usinag feedback information consistiing of a '), 1,* or e to3 lenote a slot -fith zero, one, or at least two packets, respectively.
'n '-1 3.19] and 1541a class of simple recursive re-:transmission -ontrol
s st-r-ateg:317ie-s i s i-vrei for the raniom-a.cces!3 broalzast c-harnel, again subject to
-in infinitq Dopiilation of users. They provide stable throug7hput _ip to -
nac~kets/slot undier a variety of feeiback models, and they are ins1ensitive to
fluctuations Ln traffic intensity. Analytical ani numerical pe rfo rman::e
-3%al-,3tions 3re given.
A 2las3 of random- access strategies, t -3 m.ee ackcnowleigement b Rse i
I re transmisston control ".ORC) strategies, are analyscil -i ri-i for both finite-and infinite -nonulations of users. 7hq only e-bced' oipeeta
A31" strategy is that acknowledgement of a successfully transmitte1l packe3t
Tust 'oe given to the transmitter. )n the basis of an equilibriuim analysis and
I the local ?oisson approximation introluel in 3. 5.1 it is siown that the A33-
strategies provide satisfactory stable throughput if- there is an appropriately
3hosen upper bound on the number of retransmission attempts allo.-ed to any one
user. "Thxprssioas do elay, throughput, andi probability of rejectio)n are
founi and suggestions for the c hoice of parameters are given.
knother tyre of user demand is -addressed in 1 15.44. H-ere t.- v users sac.-'a
a have a3 steady stream. of packets to trarismit ani thus a TDA1k ac cess Tole is
arnropriate. jover, if there is limited commwrnication among the use-rs Int
moreover * if thie population of active9 Isers is time-var1yinj gas, for ieample,
in a mobile comunization netwiork) then some strategy for lo'Ckinz' into 771.k
t Jh l.i itet inform ati on must be sedl ? Pre ,iia ry 1rrin 3fs c3 3trateieis and rerformance evaluations are given in 1 . Ul.
86
15. 00M 1" ATI)1S
13.1.5 'eneral 'lumericsl 4ethods and 3ounding Techniques
for ?acket 'ommunication 3ystems
)ne numerical technique we have investigated is the method of phase-ty-e
',arkov processes. Phase-type processes are :1arkov orocesses N t >-vnere the zomponent (t) is itself a Aarkov process, calleq the phase
process. For example, 'it may represent the number of packets queuei for
transmission at a communication station, and t may represent the phase of a
T74A protocol, or t may model fluctuations in incoming traffic intensity, or
both. lumerical results of this application are reported in rZ 61 .
New basic results in the computationai and conceptuql aszects of
chase-type "Aar'-ov processes are presented in i.i- . n parti;ular, a
systematic method for treating general boundary states is found.
The method of phase-type MHarkov processes is also applied in 16 to find
the average jueueing delay in a link of a communication network -hen "he
incoming traffi:- is obtained from a ?oisson stream by a generalized round
robin YTRR) routing rule. Under 1R routing, successive packets coming into a
node are routed out of the node in a pre-determined sequence according to the
order of arrival. The Ielay for 2R7 strategies is often significant-y smallerthan for the standard independent 3ernouli splittiag routing strategy.
A -eneral bouniing technique based on drift analysis is given in j'.
3imple conditions are given to ensure that drift implies convergence. Dne
anlization in _12J is a proof of channel stability under the retransmission
policies introduced in -15.151. A stability result and bounds on wi3ting
times in 11/3/1 jueues are also presente in .'21.
Ln Th±4 it is shown that among all arrival processes /not necessril-y
stationary or renewal tyne , for an exponential server jueue with sr-ciriei
arrival ani sezri-e rates, that the arrival process hiTih minimizes the
average ieliy and related quantities is the process wfith constint inter-rri'al
tites. The proof is biase on a newly discovered convexitv prnrertv of
exponential 3et-ter queues hich is of independent interest. !Lnce the traffic
need not be stationary or renewal tyme, the theorem provides lower bounds
87
1 15. COMMUNI"ATIONS
(which are readily computable by existing methods) to the average delay in a
network of queues under rather general routing disciplines. Some related
bounds and comnarison with some numerical results are presented in rR61.
15.1.6 Signal Detection in Multiple-Access Noise
k project in multi-user communications that has been initiated recently
is the study of optimum signal detection procedures for application in the
multi-user environment. Such procedures are inherently nonlinear because of
the non-*aussian nature of the multiple-access interference in the channel.
Thus the initial goal of this study is to determine the degree of performance
improvement exhibited by optimum receivers over the more conventional linear
(matched filter) receivers, and to identify those situations in which
performance is sufficiently enhanced to justify the use of the more complex
nonlinear systems. Early progress in this study includes the derivation of an
optimum (minimum-error-probability) receiver structure for a two-user channel
j and the levelopment of simplifying approximations to this optimum receiver
that will facilitate the desired performance analysis.
I 15.2 3ignal Detection, Estimation, and Filtering*
Progress in this general area has involved two prinary continuing areas
of study as well as the initiation of a new area. The two continuing topics
of study are: (i ) robustness in detection, estimation, and filtering, ani
(2) the design and analysis of digital detection systems. Progress in these3 two areas is described in the following subsections. The new area of study is
signal detection in multiple-access channels, and early progress in this study
is reported briefly in Section 15.1.6.
1 15.2.1 Robustness in Detection, Estimation, and Filtering
ti general, the area of robustness deals with the design of systems and
3 procedures which are relatively insensitive (in terms of performance) to smal,
deviations from an assumed model. Robust technivles are of interest in a side
* Uhii research was supported by the Ar-ay Research Dffice under contractDAA$29-31 -K-,062, the 'oint 3erices Electronics Program inler contractI 1000)14-7?-,]-,124, th e office of Naval Research under contract110001-31 -,-00!i, and the lational 3cience Foundation under grant'E -S7 9 -1 6 4 53.
88
15. 'OMMUNICATIerS
variety of applications including radar, sonar, and seismology, inasmuch as
inaccurate models are the norm for such situations; and robustness has been
studied widely in the contexts of communications, control, and inferential
statistics. The basic assumption for the analytical study of robustness in
detection and filtering is that the statistics of signals and/or noise are not
ciown exactly, but rather are known to be writhin some lusually nonparametric)
classes representing uncertainty in the underlying model. The primary
technique for lesigning a robust system is to seek a system achieving the best
worst-case performance over the relevant uncertainty classes; i.e., the
primary lesign philosophy is minimax. "dithin this context several detection
and estimation problems have been considered under this project, and these are
described in the follow-ing paragraphs.
The problem of robust linear smoothing of a stationary random signal with
uncertain spectrum observed in additive noise with uncertain spectrum is
consilered in r 1.211. Here, a general solution to this problem is given for
spectral uxncertainty classes of a general type basel on Tnoquet capacities.
This type of model includes stanlard .ncertainty models such as contaminated
mi'ctures as well as several topological models of uncertainty. The usefulness
of these results is extended in r15 .55] by ieneralizing the class of models
that can be treated in the framework of 7i 5.21 ]. xoreover, an extensive
numerical analysis r15.561 indicates that the worst-case performance of the
proposed techniques is generally much better than that resulting from lesigns
which simply ignore the presence of lnicertainty. More general problems of
robust estimation of stationary signals "including smoothing, filtering, and
prediction) also have been considered for the case of discrete time :15.7"].
Tere, a general minimax result is given from which robust solutions to a
variety of problems follow straightforwardly.
Also considered in this general context is the problem of state
estimation in linear stochastic systems with uncertainties. Two approaches
have been consilered for this problem. )ne of them is to consider -inimax
designs for systems with uncertain state and observation noise statistics
[15.23,15.511. Here, under the assumption that the noise is white with
uncertain componentwise correlation, it is shown that the minimax-
mean-square-error state estimator is the *Kalman filter for a least-favorable
I89
I I15. '1414"I kT TI 3
mol1, ani several examples are consiiered. 1elated results for 'inimacinear-qualratic :ontrol are found in .- 1 5.4' ani are Ii3cusseI in the
D-e3!rciusion ani Control unit of this r t other approach to stateDeciion ni Cntro uni of'report.
estimation with ancertainties concerns systems .rith ranlom .jumps 4hich are
moelad by consiiering the staniarl linear system model A-ith aliitional markei
I ?oisson input noise,. suboptimal sequential smoothing algorithm for ;his
moiel is levelopel a ri analyzed in '".52. This scheme involves letecting the
inciience times and estimating the marks of the Poisson input process an! then
reconstructing the state "i±th a (conditional) Kalman smoother. 'he
perfor-ance of this scheme is seen to be superior to that of optimum causal
estimators and linear noncausal estimators, and the imrplementation of the
scheme is superior to the optimum (nonlinear) smoother, :fhich is lescribel by
an infinite sequen,-e -f filtering equations. k relatel study is found in
whi-h consilers tihe problems of' minimax (or robust) filtering anj
Itate estimation for obserrel nonhomogeneous Poisson processes iith uncertain
rate functions. in this situation, analogies with the case of 2ontinuous
observations can be exploitel to ierive minimax designs.
I .3everal tooics in robust signal leteation have also been consilerel.
.iree basic problems have be-n treatei in this context. 7irst, the role of a
generalize measure of signal-to-noise ratio Iknown as the leflection) in
lesignina robust Ietec tion systems is considered in 1 I.23]. Tt is
demonstrated that this zriterion is often much simpler than earlier criteria
for use in robust design because a constructive method for finding robust
solutions is available for the new criterion whereas no such general methol isS available for earlier formulations. A second problem involves the design of
robust systems for ietecting signals in the presence of iependent noise
I l.2 ]. The proposei technique uses a moving-average Toiel for no'se
iecendence and considers designs that are minimax robust to first orier in tie
I -e3gree of dependence. Te thir -effort in this area involves findin1
solutions to the problem of robust discrete-time matched filtering 5-3.53].
I The problem of robust matchei filtering has been considered previously iithin
a general Hilbert space for.ulation -15. ri. ani 715.53, derives specifi"
solutions for the partic ular problem of iscrete-time matc2hel filtering ithin
I a variety of uncertailty mCoels for signals and noise.
.i
90
1 5. COMMUNICATIO1S
15.?.2 Digital Detection 3ystems
The second continuing area of study in letection, estimation, and
filtering involves the analysis and optimum design of ligital signal-detection
systems. This problem is of interest in a wile variety of applications in
which receiver ligitization is increasingly important for efficient
implementation. Recent progress in this general area can be iiviied roughly
into two main topics: optimum data quantization for detection systems and
optimum nonlinear designs for signal detection in dependent noise.
Progress in the first of these areas is reported in 15.41, -15.29', and
n 'I. 1n F5.41, results of an earlier study on optimum iata Icantization
for coherent letection systems are eqtended to systems for the ietection of
purely stochastic signals. Optimum and suboptimum schemes are compared, and
it is seen that significant detection effic.iency can be gained by implementing
in optimum quantization scheme rather than, for example, by using stanlard
minimum-distortion quantization in a qualratic receiver. These results are of
interest in sonar, radio astronomy, and seismological applications in 4hich
signals are often modeled as nondeterministic stochastic processes. k further
study of quantization, , -13.301, considers the lesign and analysis of
quantizers for use in memoryless (recursive) signal detection systems for
detection in m-depenlent noise. It is lemonstratei here that, by utilizing
the dependence structure smong noise samples, one can improve memoryless
Letection performance consilerably without increasing complexity. Aoreover it
i3 iemonstratil that, in the presence of strong dependence, qiuantizer-
detectors utilizing even small numbers of quantization levels ::an be nearly "s
efficient as the optimum ('ieyman-?earson) detector for the case of laussiin
noise. A third problem considered is that of designing optimum input
amplitude compressors for coherent detection systems utilizing unifo r
quantization. It is shown in 715. 291 that the optimum such input 3ompressor
is a scaled version of the locally optimum detection nonlinearity, and that
the performance legradation 'over locally optimum letection) resulting from
the proposed uiantization scheme is negligible for only a molerate number of
quantization bits.
I91
I 15. COMMUNI"AT'1S 3
Procedures for designing effb:ent systems for letecting signals in
lependent noise backgrounds have been discussed in two of the above contexts,
I notably F15.2631 which ieals with robust detaction in weakly lepenlent noise
and 1i 3.301 which deals with optimum quantization for letection in m-dependent
noise. Optimum detection for the weak-dependence :odel of '15.261 is3 considered in 713.2 5 ]. t is shown here that, for a moving-average dependence
model, efficient detection -an be achieved by reshaping the independent-noise
optimum detection nonlinearity with an additive linear correction term. I n
investigation of several commonly used noise models indicates that the
resulting performance improvement is most significant for impulsive tyins of
noise. In r15.461 similar modified structures are considered for letecting
I signals in $-mixing noise. It is demonstrated that the -mixing model "with
only the mixing coefficients given is not sufficiently lescriptive of
lerendence structure to admit a design dhich is uniformly better than the
independent-noise design (This is in contrast to the moving-average model of
5.25 1). However, optimum memoryless designs for 0-mixing and other
j lenendence models can be derived using general results of r15.22] provided
that the second-order iistributions of the noise-Drocess are known. The
I optimum system in this case is the solution to a ?redholm operator equation of
the second kn anid, as such, can be characterized in terms of the
j 3arrett-,imparl expansion of the aoise process. A further result in this
general area involves the comparison of memoryless ietection systems to
systems with linear memory (such as the Neyman-?e3arson detector for the case
of ;aussian noise). In F1 5.34] it is shown that, for a general situation
involving stationary noise, memoryless systems are asymptotically as efficient
as systems with linear memory. This indicates that implementation :omplexity
can be rqduced without sacrifiing performance for such situations involving
large sample sizes.
3 !3. 3 Data 77mpression Theory and Techniqjues
3 }',r research on lata compression has been extendei into several n
2assss of problems including the construction of universal source zoes.
bounds on the redundancy of universal data zompresslon technilxes, and anI ivestigation of the r-e4undancy of '+3iffman codes. Zistence results.
redundancy bounds, and construction technijues for vniversal -oroe ioes fori'A
92
I3. COAM UNI -AT10NS
classes of memoryless sources are given in F15.91. The codes considered are
fixed-length-to-variable-length (FL-VL) source codes. New results have been
obtained on other types of source codes including variable-length-to-
fixed-length (7V-FL) codes. For short block lengths we have found that
properly designed VL-FL universal source codes provide lower redundancy than
our best FL-7L source codes previously reported in p15.91.
Our investigation of FL-VL universal source codes for unifilar .Aarkov
sources has provided several new results. A computational formula was found
for evaluating upper and lower bounds on the redundancy of weighted universal
coding for a binary two-state Harkov source, where the weighting distribution
of the transition probabilities is uniform. This formula was also analyzed
for large block lengths and found to agree asymptotically with the upper bound
on the redundancy for minimax coding. Earlier results on upper bounds on the
red udancy for minimax coding for unifilar sources were extended to cases
here the structure of the unifilar source was unknown.
knother problem which has been investigated is that of the redundancy of
:4uffman coding. uffman"s alorithm is a -ell-cnown procedure for
constructing a variable-length code of minimum possible average length for a
3oarc3 with precisely 'known statistic s. The redundancy of a Huffman code is
iefined to be the difference between its average length and the source
entropy. Ie have developed a powerful general technique for estimating Ihi3
redundancy. :t implies for example that most sources have redundancy near the
value log~log) e- /2 - .029766373, and that apart from a few exceptiona'..
cases, no code has redundancy exceeding I - log. e log, e log, e
.B6071332. This work is reported in r1. 4 -1.
- m
I,93I
3 15. COMMUJNICATION~SReferences
[ai] E. A. Geraniotis and 74. B. Pursley, "Trror Probability for Direct-
Sequence 3pread-Spectrum Multiple-Access Communications -- Part !1:
Approximations," submitted to the IEEE Transactions on Communications.
I [12j B. -a.jek, "Hitting Time and Occupation Time Bounds Implied by Drift
Analysis with Applications." submitted to Alvances in Awolied Prob-
I ability.
[5 1 3. Hajek, "Acknowledgement Based Random Access Transmission Control:
An Equilibrium Analysis," preprint.
I i 1 3. Tajek, "The Proof of a 'olk Theorem on ueueing Delay with Apnlica-
tions to Routing in Networks," preprint.I[R91 A. 3. Plirsley, D. V. Sarwate and W. E. Stark. "Error Probability for
jDirect-Sequence Soread-Spectrum Xultiple-Access Communications -- PartI: Upper and Lower Bounds," submitted to the :EES Transactions on
I Communications.
L'161 T. -tan Loon, "Application of Phase-Type Markov Processes to Xultipl?
Access and Routing for Packet Communications," M.S. Thesis, in finalI preparation.
II
I
94
16. ANALOG AND DIGITAL CIRCUITS
Faculty and Senior Staff
E. I. El-Masry W. K. Jenkins T. N. Trick
I. N. Hajj M. R. Lightner M. E. Van Valkenburg
W. Mayeda
Graduate Students
N. Attaie M. Hanlon V. Rao
W-K. Chia D. Hocevar J. Resman
G. Cortelazzo P. ollaritsch J. Stothoff
R. Davis C. Lee Y-P. Wei
L. 3arza N. Nguyen C-?. Yuan
C. Pahlmeyer
16.1 Computer Algorithms for Integrated Circuit Design*
In this research a study was carried out on the use of partitioning and
tearing algorithms for the analysis of large-scale circuits and systems. .n
this approach the circuit is partitioned into subcircuits, each subcircuit is
solved separately, and then the subcircuit solutions are combined to obtain
the solution of the entire circuit. Tearing provides many computational
advantages, especially in the simulation of large-scale integrated circuits.
First, computer storage can be reduced if identical subcircuits are created.
Secondly, computer time can be saved by exploiting subcircuit latency: that
is, subcircuits that are inactive during a given time interval can easily be
by-passed in the solution process. Thirdly, the use of tearing methods allows
!This work was supported by the Joint Services Electronics Progra('.3. Army, U.S. Nav _, and U3.3. Air Force) under Contract NOOO -79-C-0121,and by the ational Science Foundation under grant TCS BO-01775.
95
16. ANALOG AND DIGITAL CIRCUITS
for the parallel processing of subcircuit equations. Finally, tearing3 facilitates the use of reduced-order or macromodels in the analysis and
permits the application of mixed-mode simulation.
In our research, equation solution techniques for the analysis of
circuits and systems were investigated. In particular, when tearing methods
are employed, there are a number of solution techniques that can be used to
exploit latency. These techniques have been studied and for electronic
Rcircuits the most efficient solution technique has been identified F16.381.
In conjunction with this work a fast and reliable ordering technique for the
modified nodal equation formulation was found F16.61J. Our ordering procedure
eliminates the time-consuming pivot selection process that is usually required
for the solution of the modified nodal equations. in addition to the above
work on the solution methods, various latency criteria for the transient
-analysis of ligital circuits were studied. In order to test our algorithms
the SPICE2 circuit analysis program was modified to include tearing methods.
This new circuit analysis program is called SLATE F16.311; in general, thi3
program runs faster than the SPICE2 program for the transient analysis of
digital logic circuits and yields the same accuracy.
In related work, algorithms for efficient solution updating when Large
changes in the values of some system parameters occur, have been derivedr16.4. FThis problem has also been investigated -When changes occur in large,
partitioned systems r16.22].
In addition, the problem of finding the shortest path in a network, which
occurs in network routing among other applications, has been studied and an
efficient updating algorithm when changes occur in the network parameters has
been derived r16.211.
In many cases electronic device models are represented in tabular form.
This tabular form can be used to approximate the nonlinearities by
piecewise-linear functions. In our research the properties of the solutions
of piecewise-linear circuits have been investigated in detail 716.5,16.20.
Research has also been lone on the use of piecewise nonlinear models.
The use of piecewise nonlinear models for the exponential nonlinearity of a pn
Junction offers great flexibility in modeling 16.29,16.7311. However these
models io not satisfy the requirements of the standard iterative solition
96
16. ANALOG AND DIGITAL CIRCUITS
techniques. In this work we have explored new solution algorithms tailored to
the piecewise nonlinear models and compared, computationally, the new
algorithms and models to many of the traditional methods used in programs for
the calculation of the DC operating point of bipolar circuits. Our results
indicate that the piecewise nonlinear models and algorithms compare very well
to the best methods that are available today r16.30 ,16. 361.
Future work in this area will include theoretical studies of the
algorithms for piecewise nonlinear models. Also the applicability of this
type of modeling for use in MOS circuits will be studied.
16.2 A New Structured Logic .rra
The use of highly structured architectures is becoming necessary to take
advantage of the advent of VLSI circuits. One area that lends itself to
structuring is that of logic arrays. These arrays, such as ROM's, PLA's, gate
arrays, and custom logic cells, are- widely used throughout the 1C industry.
The success of these structured logic arrays is in large part due to the
levelonment of successful CAD tools for their design, layout, and routing.
There are two broad classes of structured arrays: customer
reprogrammable and customer nonreprogrammable. In this work F16.341 we are
investigating a new architecture for a customer reprogrammable logic array.
This array structure, known as REPLICA (Random, Electronically Programmable,
Logic Integrated Custom Array) is a regular structure of NOR gates -*-ith
restricted fan-in and fan-out. We have developed CAD tools for the
minimization of logic realizations in REPLICA similar to the folding
algorithms used for PLA's. We have also developed guidelines for the optimal
form for the logic equations to be implemented in REPLICA. Several examples
of logic functions implemented in REPLICA indicate that it competes well with
PLA's and has some features that may make it more desireable than PLA's in
situations such as realizations of sequential logic. Work is progressing on
comparisons of REPLICA with existing structured arrays and on improving the
CAD tools necessary for realizing larger test examples in the REPLCA
architecture.
.,This work oms supported by the Joint 3ervices Electronics Program'3.s. Arty, U.3. Navy, and U.S. Air Force) under contract NOOO 4 -9-C-D4? .
973 16. ANALOG AND DIGITAL CIRCUITS
16.3 Hierarchical Design Techniques for VLSI Systems*
With the complexity of the systems that can be realized on a single VLSI
chip the design time is, using conventional techniques, on the order of 100
man-years. In order to reduce this unacceptable design time, structured
design techniques are being investigated. In this work F16. 3 7 i the impact of
a structured design methodology on the design of a VLSI system was studied.
The design and testing of the chip were both considered from the beginning of
the design process. As a test example a CORDIC function (used for realizing
trigonometric functions) was designed. The function to be implemented wasIstudied by simulation in order to finalize the architecture of the design.
Then followed a structural design, a physical design including layout, and a
j verification using different simulation tools.
The investigation clearly showed that, with some sacrifice in chip area,
Sa structured design approach was feasible for VLSI. Various other systems are
now being studied for implementation in VLSI using structured design
techniques. The goal of these studies is the refinement of the tools and
techniques necessary in the design hierarcy for VLSI systems.
I 16.4 The Use of Multiple Criterion Optimization for the Desin ofDigital Filters"*IThe design of digital filters, both FIR [16.12,16.321 and I1R, using3 competing criteria F16.31 is being studied. In particular, various
formulations of the design problem which trade-off between magnitude and phase3 performance are being considered. These formulations involve the
specification of a desired phase behavior as well as the standard desired
magnitude behavior.
Performance constraints, such as monotonic magnitude response in the
transition regions, are also imposed on the behavior of the filter. In order
to solve these rather difficult optimization problems we are modifying theU sequential quadratic method for constrained optimization proposed by Han and
This work was supported by the Joint Services Electronics Program(U.S. Army, U.S. Navy, and U.S. Air Force) under Contract N000i4-vg->-,±24.
"This work was supported by the Joint Services Electronics Pro ram,1.3. Army, U.S. Navy, and U.S. Air Force) under Contract '10001i-'9--,D 2 ,3 and by the National Science Foundation under 3rant ECS 90-07095.
98
16. ANALOG AND DIGITAL CIRCUITS
Powell. We are also exploring the use of a dynamic frequency grid which will
allow, theoretically, convergence to the true solution of the design problem
and not the approximation that is found in all existing digital filter design
programs today.
16.5 Analytical and Computer-Aided Design of lew Circuit 3tructures for
Switched Capacitor Filters*
Research is being conducted in optimum switched capacitor (3C) filter
structures and in computer-aided analysis and design (CAD) of SC circuits.
Recent studies of optimum SC filter structures resulted in two different
des{gns using the bilinear z-transformation. One design uses a state space
technique I16.11,16.14,T6.17] while the other uses a follow-the-leader
feedback (FLF) topology F16.10]. Both techniques allow SC filters to be
implemented with stray insensitive first order modules (such as forward Euler,
backward Euler, or bilinear descrete integrators) or second order biquadratic
modules. 3ptimization procedures were applied to maximize dynamic range while
minimizing sensitivity, capacitance ratios, and total capacitance. Impressive
improvements in sensitivity, as analyzed by Monte Carlo analyses, have been
obtained for an example of an optimized fifth-order Chebyzhev LP filter
iesigned with the state space and FLF approaches.
In the area of CAD, a user-oriented program, SCAP II, based on modified
nodal analysis, was completed. Frequency and sensitivity analyses were
demonstrated on a vide variety of useful SC circuits taken from the literature
f16.25,16.26,16.-51. SCAP II is capable of performing frequency analysis and
frequency domain sensitivity analysis on SC circuits that zontain MOS
switches, capacitors, and voltage controlled voltage sources, and which
operate with a 2-phase 501 luty cycle switching sequence. The analysis is
efficiently implemented with sparse matrix numerical methods and a specially
designed equation ordering strategy which minimizes fill-in during computation
of the LU factors. Sensitivity analysis, which is an implementation of first
order relations obtained by differentiating the original modified nodal
equations, is efficiently implemented for capacitors, capacitor ratios, and op
This work is supported by the National Science Foundation unler 3rant
!99
1 16. ANALOG AND DIGITAL CIRCUITS
amp gains by using the same LU factors as already computed for the frequencyI analysis. SCAP I! requires approximately twice the machine time as the
previously reported block-partitioning technique of SCAP I, and hence appears
to be a useful aid in SC circuit design F16.1,16. 2 31.
The most recent activity in SC CAD has been the development of a very
general N-phase SC analysis routine, referred to as SCAPNI 6.13'. In
addition to using sparse matrix methods and proper reordering techniques SCAPN
uses a specially designed "pre-LU factorization algorithm" which minimizes the
amount of computation that must be lone to update the LU factors at each new
frequency. SCAFN also features a very efficient sensitivity analysis based on
the adjoint network concept F16.14, and a nonlinear distortion analysis based
on the Volterra series technique for analyzing the effects of driving MOS
capacitors and amplifiers into nonlinear operation "16.15.
16.6 Automatic Tuning Algorithms for Analog Filters*ITo meet filter response specifications analog filters usually must be
tuned or adjusted, preferably by computer automation if the production level
is high. In this research three generalized tuning algorithms have been
studied on the basis of their architecture, computational complexity, and
effectiveness. Further, a method has been found for the tuning element and
frequency selection problem, a problem relevant to all three methods. Monte
Carlo simulations have been done on several active filter circuits in order to
enhance the comparison and provide a demonstration [16.231.
16.7 Fault Isolation in Analog Circuits**
I The objective of this research is to be able to isolate faulty components
in analog circuits from test point measurements. Presently the only approach
Ithat has had limited success in practice is the fault dictionary method.
1 * This work was supported by the National Science Foundation under Trant ENG73-1 173.
*This work ws supported by the U.3. Naval Air 3n-nineering 3enter uinderContract N6BV35-30-.C--0 and by the Joint Services Electronics Program
SJ.S. Army, U.S. Navy, and U.S. Air ?orceq under Contract NO001 -9-2- 2.
10016. ANALOG AND DIGITAL CIRCUITS
Unfortunately this method requires that one anticipate all possible fault
conditions in advance. This is usually an impossible task, and reliable field
data is required in order to update the dictionary. Typically it is also
necessary to confine the dictionary to single catastrophic faults in order to
avoid excessive storage requirements and long search times for the fault
conditions. Finally, in the fault dictionary it is not easy to determine the
testability of the circuit from a given set of test points.
In our work we have shown that a sensitivity matrix can be used to
isolate both hard and soft faults. The idea behind the method is that if the
columns of the sensitivity matrix are linearly independent, then any fault in
a component will cause a unique shift in the test point voltages. The method
can locate single as well as multiple faults in analog circuits. Furthermore,
the independence of the columns of the sensitivity matrix are a good measure
of the testability of the circuit with respect to the given test points. The
sensitivity matrix is computed for the linearized circuit model at the nominal
parameter values in the pre-test phase. Thus, only a few simple mathematical
operations are required in the test equipment. A sensitivity matrix must be
computed for each test signal, and assuming that the rank of the sensitivity
matrix is equal to the number of test points for any column ordering, then the
number of faults in the circuit must be less than the number of test noints in
order for the method to give reliable results.
101I_17. DECISI)N A2TD 'ONTROLI
I Faculty
T. Basar P. V. Kokotovic P. . Sauer
J. B. Cruz, Jr. D. P. Looze N. Wax
W. R. Perkins
IResearch Associates
B. Avramovic H. Salman
3raduate StudentsIR. eLair D. 3ahutu 3. ?eponiiles
J. Benhabib X. loilakota R. Phillips
Y.. Ian P. Ioannou V. Saksena
J. ". Darrigh D. K. Lindner H. 3alhi
J. S. Freudenberg R. McEwen J. ien
I P. F. Parent
17.1 'ontrol and Decision Strategies for 3ystems Under Imperfect Information*
Uncertainties arise because of unknown system parameters, unknown signal
environments. ani hardware tolerances. Relat el oplexities -ar4e n
situations involving multiple system Derformancq iritria ni multinle
lecision makers. The objective of this project is to -ain a bqsi:3 unlerstanling of the behavior and :ontrol of iomplex systems. During -he jast
year, attention has been iiven to several topios. A ighli-hts are summarizeil
I in the follo'ing.
This work was supported by the Joint 3ervies lectronis Program('U.3. Army, U3.S. avy, and U.S. Air Force) under 2ontract :OOO14-2-- 1o4--.
6
10217. DECISIO AUD CONTROL
3everal results concerning the sensitivity and robustness of control
systems to uncertainties in the plant model have been obtained. W e have
developed bounds on the magnitude of allowel perturbations which guarantee
that the perturbed system performs better than an equivalent open loop system.
These bounds are expressed in terms of the singular values of the return
difference matrix. 'ie have also developed the singular value sensitirity
function as a robustness analysis tool. Singular value sensitivities can be
uised to analyze perturbations such as multiplying incident perturbations or
structur-ed perturbations) for which the guaranteed bounds do not aDply.
In the area of fundamental system theory basic to control strateg y
development, three general topics were the subject of investigation this past
year: (1) the input-output behavior of some nonlinear feedback loops, (?) the
invertibility of input-output systems, and (3) the existence and properties of
'strange attractors" in system theory. We give below a brief account of these
studies, concentrating on the last one, since it is new.
1. 'ie have treated the "locking" of a class of nonlinear feedback loons
to a certain set of input signals; sufficient conditions for the
locking have been obtained. Iurther, sufficient conditions for the
"tracking" of the input signal by the outout have been found. These
results are being extended; a manuscript is in preparation.
2. liven an output, what was the input to the system? it is this which
constitutes the question of invertibility. 3ome partial results have
been achieved, in the sense that sufficiient conditions for the
invertibility of a class of input-output systems have been found.
These conditions are fairly straight-forward extensions of previous
work. Currently, efforts are being made to generalize the conditions
so that they are applicable to a wile class of systems.
3. A variety of systems are suspected of exhibiting 'strange' behavior.
By this is meant something of the following sort (the definitions
varying according to the investigator): There exist trajectories of
the system which, starting close to each other, iiverjg exponentiall.-
initially; these trajectories remain bounded, however, for all time.
?inally, the trajectories approach a set of noints (the "strange
attractor") which are not periodic orbits or a finite set of ing ular
103
17. DECISIN AM CONTROL
?o 4:1 tsa.
The forced van der ?ol equation with very large damping and
iroportionately large sinusoilal forcing and similar such equations
are the only differential equations for which it has been proven that
strange attractor exists. Numerical solutions of the Lorenz system
and of a version of the forced Duffing equation exhibit behavior much
like that of strange attractors, but proofs of this are absent.
-3 It has been suggested that some of the properties of turbulent
mixing and of a number of other phenomena could be explainel if
strange attractors existed for the governing equations.
We have started to study strange attractors, by repeating some
of the numerizal work reported in the literature, and by a review of
the theory. We were unable to duplicate the results on the forced
Duffing- like equation for a long time until we used an
Adams-3ashforth predictor-corrector integration routine for the
numerical solutions of the differential equation; now our results
agree -with those published. We are in the initial stage of what may
Sbe a long term project.
S"Well-posedness of leader-follower strategies obtained from singularly
certurbed reduced order models has been investigated in a manner similar to
our approach previously developed for lash strategies for reduced order
models. We have derived order reduction procedures whic-h lead to well-posed
I formulations of the leader-follower control of singularly perturbed i5rnamic
systems. We have investigated the strong connection between infornation
structure and well-posedness of singularly perturbed Nash games and we have
ilentified "why the natural order reduction is ill-.osed as we reported
earlier. )ther structures w#hic.h preserved information -were lemonstrated to
lead to well-posed solutions. Details are tescribed in the jou.rnal and
meeting papers.
The problem of controlling a linear stochastic system so 1s to Minilize a
I yialratic steady state cost functional with uncertain process and obse.ration
noise statistics was investigated. The problem was fo)-nulateld s a mininax
optimization. We showed that the minimax solution is obtained by the usual
I
10417. DECISION AND CONTROL
linear quadratic deterministic gain multiplying the state estimate provided by
a linear filter. The linear filter is a Kalman-Busy filter for a particular
worst case pair of spectral density matrices. These results parallel those of
the minimax state estimation problem discussed in Area 15.
17.2 Implementation Constrained Decomposition and Hierarchical Control*
In large-scale systems, control tasks may be decomposed and hierarchical
levels may be imposed on the basis of analytically derived strategies. This
project is devoted to the fundamental problem of investigating an analytical
framework for incorporating realistic constraints of a computer network.
'uestions concerning information flow, time scales, and loss of feedback loops
are examined in coordinating the multiple controllers in a network.
We have investigated and developed several classes of hierarchical
optimization algorithms. Classes of algorithms which exploit a weak
structural coupling between variables of an optimization problem were analyzed
asing local convergence theory. Natural decompositions were defined in terms
of splittings which directly exploit the structure of the problem in the
decomposition. We demonstrated that iterations iefined by natural
decompositions would converge whenever the coupling is sufficiently weak. We
also have developed a hierarchical optimization algorithm which can be used to
solve the linear quadratic robust control problem. The algorithm exploits the
structure of the dual problem to define a simple hierarchical algorithm which
does not require a stabilizing initial point and which has guaranteed
convergence properties.
A common feature in large scale system practice is that different
decision makers assume different simplified low-order models of the same large
scale system. This may be due to the necessity of easing the computational
burden in evaluating strategies or the lack of adequately modeled dynamics of
some parts of the large scale system. Based on these simplified models, the
decision makers would then attempt to arrive at low-order decentralizel
strategies which would be economicslly feasible to implement, at the same tine
preserving near-optimality. The problem of designing control strategies for
deci3ion makers under a situation where they ha,ie lifferent models, different
'his work was supported by the Joint Services Electroniis Program(i.a. Ar-y, U.S. Navy, and 'J.3. Air Force) under contract NOOO14-71-C-424.
I105
I 17. DECISTJN AND ,0NTROL
information sets, and different objective functionals, has been addressed
previously within the framework of multiparameter singular perturbations. In
this framework, a large scale system is viewed as consisting of a "slow" core5 coupled to a number of "fast" subsystems. A multimodel situation results when
each decision maker retains the dynamics of one fast subsystem and the slow
core, while neglecting the dynamics of the remaining fast subsystems. This
year the general problem, wherein the fast subsystems need not be weakly
coupled and each fast subsystem might be affected by more than one decision
maker, has been addressed. A procedure has been formulated to obtain near
optimal decentralized Nash strategies in the two player case under a
multimodel situation "Aithin the framework of multiparameter singular! perturbations.
17.3 Control 3trategies for Complex Systems for Use in kerospace Xvionics*
I ;henever model uncertainty is present or a range of operating conditions
is anticipated, engineering analysis and lesign must deal with questions of
sensitivity. Performances of manufactured components 4re necessarily
specified with nonzero tolerances. The parameter values which characterize
these components are therefore uncertain. These parameter values may change
during the operating lifetimes of the comnonents due to aging and due to
j changes in environmental conditions, such as pressure and temperature. :n
addition, mathematical models used for analysis and lesign of actual systems
cannot possibly lead to predicted performance which exactly matches the
performance of the actual systems. It is useful to regard some of the
parameters of the models as uncertain in order to make these models more
realistic. The combined effects of parameter uncertainties on overall system
behavior are of principal consideration in any system design.
1 We investigate uncertainty using three techniques. Dne point of view
regards the parameters as unknown but deterministic, ani studies the
Incertainty situation using parameter sensitivity methods. New results here
include establisment of some relationships between singular value robustness
anproaches and the now-standard comparison sensitivity methols. Design-
oriented techniques are being developed for large parameter variations in
3 linear ant nonlinear systems.
This work was supported by the U.S. Air ?orce .inler 3rant AFOSR-i- 677.
106
17. DECISION AND CONTROL
A second approach models the parameters as random variables. The tools
for this approach are stochastic control and dual control theory. including
stochastic multi-controller situations. The interplay between information
structure and control has been emphasized in our recent work.
We have expanded our basic research to encompass an area of increasing
importance in control research and aerospace applications - adaptive control.
Our approach focuses on the role of singular perturbation theory in the
development of model reference adaptive systems when the plant and reference
model have different orders. An important possible interaction has been
discovered between frequencies present in a "sufficiently rich" adaptive input
signal and neglected high frequency terms in a reduced-order model in a model
reference adaptive system. Details of these results may be found in the
journal articles and conference papers listed in the publications section.
17.4 Large 3cale Systems*
Several new results have been obtained in multimodeling, that is, in
modeling ani control situations in which different decision makers employ
diff3rent models of the same system. Multiparameter singular perturbation
techniques are employed to capture the multimodel nature of fast dynamic
subsystems interconnected through slow variables. In one problem studied, a
decentralized filtering and control scheme was developed for obtaining low
order Nash equilibrium strategies in a multimodel setting. The lecision
makers have decentralized information structures and are constrained to use
finite dimensional compensators. In another study, physical examples (e.g..
networks) are used to reexamine the role of modeling in large scale system
analysis and design. A general property of the systems consilered is that
they are strongly coupled (physically) in the slow time scale, and weakly
counled in the fast time scale. This leads to a multimodel situation in which
every subsystem controller can neglect all other fast subsystems except his
own.
In the area of stochastic multiplayer strategies, an indirect method to
obtain the Stackelberg solution of two-person nonzero-sum differential games
with general dynamics and cost functionals has been developed with the leader
This work was supported by the National Science Foundation under *Irant EC379-1 9596.
I107
i 17. DECISION AND CONTROL
having access to sampled values of the state. This new method involves a
three-step optimization, and converts the original dynamic problem into two
open-loop optimal control problems and one static finite-diaensional
optimization oroblem. Solutions of these optimization problems are obtained
as solutions of a set of differential equations with mixed boundary
5 conditions, which in turn leads to the Stackelberg cost value of the leader,
as well as to a class of strategies that yield that cost level.
3 Chained aggregation, an iterative approach to large scale system model
reduction, has been developed further. The resulting system representation,
I the eneralized Tiessenberg Representation ("HR), has been shown to exhibit
explicitly the system observability structure. Geometric (in the sense of
'Wonham) properties of the 2HR representation have been exposed and exploited
in the development of a three-component iecentralized hierarchical feedback
U design technique.
Finally, coherency, a conceat widely used in power system studies, has
been examined from an aggregation Doint of view. Specifically, the
relationshiD among coherency, aggregation, and obsetTability has been
clarified, and a fundamental relationship between "slow coherency" and weak
coupling has been established. Details of these and other results are given
B in the list of journal articles.
17.5 Power System Normal and Security State kssessment*
U Normal and security state assessment of large scale power systems
requires the analysis of power system performance under all levels of likely
load and contiagencies. Performance is measured in terms of bus voltage
levels, line power flows, stability margins, and their sensitivities to large3 changes i network conditions. xolicit load flow solutions have been
leveloped to proviJe near closed form analysis of network response to total3 load. These explizit forms may be used to produce time varying solution
trajectories and exact stochastic load flow solutions.U
U This work was supoorted by the National Science Foundation under jrant S'I79-05394, and the lrii'ersity of :llinois Power Affiliates Program.I
108
17. DECISION AND CONTROL
Many optimal load flow algorithms require stability margin constraints.
These constraints have typically hinged largely on bus voltage angle limits.
'hen finite VAR control is included, these angle limits have been shown to be
erroneous and in some cases unfeasible. Extension of these realistic
stability margins to an arbitrary network is currently being considered.
17.6 Control of Stochastic Systems Containing Parameter Uncertainty"
The main result of this research is the letermination of algorithms for
Sensitivity Adaptive Feedback with Estimation Redistribution (3AFER) control.
This type of control, which we proposei and developed, has dual properties in
the sense that the control takes into account future Darameter estimation
accuracy. The parameter sensitivities are altered by changing the sensitivity
weighting coefficients in a composite performance index. Our early work on
thi3 problem dealt with a general multivariable state space formulation which
led to quite complicated algorithms. de continually simplified the control
algorithms by first consilering input-output representations, and finally
single-input single-output representations. Much of the application of
adaptive control is for this latter class of problems. Very substantial
reductions in computation are achieved not only because of the minimal number
of unknown parameters in the single-input single-output case, but also because
of our use of only two sensitivity functions to capture the impact of all the
parameters in a SAFER control setting. Our simpler algorithm for SAFER
control has been demonstrated to yield performances which are superior to the
certainty equivalent control.
This project is a cooperative effort between the Coordinated Science
Laboratory of the University of Illinois and the Systems Engineering
Laboratory of the Venezuelan Institute for Scientific Research.
This work was supported by the National Science Foundation under "rant :NT77-20969.
£ 109
1 17. DECISION AND CONTROL
17.7 Aggregated Hierarchical Control*
U The objective of this research is to investigate hierarchical
i ecomposition algorithms for which the problem solved at each level of theU hierarchy is approximated. Standard hierarchical decomposition theory
requires that an optimization problem be decomposed exactly and that the
iteration defined by the decomposition converge to the exact solution. 'We
have relaxed the exactness requirement by allowing the exact problems at each3 hierarchical level to be replaced by an operator theoretic aggregate. The
hierarchical structure is deter~mined by the composition of an exact
I decomposition and an aggregation. Conditions under which the approximate
decomposition iterates converge to an approximation of the exact solution have
been determined for linear problems. The extension of these results to
nonlinear problems through the use of nonlinear aggregation and nonlinear
splitting functions is under current investigation.
I
!I!
I '
This work was supported by the National 3cience Foundation unler srAntUNG-71-03773.
110
19. DIGITAL SIGNAL AND IAGE PROCESSING
Faculty and Senior Staff
E. I. Tl-Masry W. K. Jenkins D. C. Munson
T. S. Huang A. Steinbach
2raduate Students
A. C. Bovik F. Hong A. C. Segal
T. Ii. C. Martin J. H. Strickland
0. A. Hayner J. D. O'Brien T. P. Walker
W E. .- ig gins D. F. Paul R. Y. Tsai
J. Posluszny
19.1 An investigation of Residue Number Architectures for Diital Filters
with Fault Tolerance*
This project involves a study of how the theory of redundant residue
number systems (RNS) can be developed and applied to the isolation and
detection of hardware failures in digital processors, with subsequent "soft
failure", that allows a faulty processor to continue operating with reduced
capabilities. The class of systematic redundant RNS codes was analyzed and
simulated, thereby establishing preferred lesign procedures and demonstrating
failure resistant behavior in recursive ligital filters "19.121. At this
time, the theory of failure resistant Aesign for the systematic RNS :codes is
relatively complete, including bounds on the necessary redundancy and
procedures for iientifying faulty -odules and for reallocating resources
following soft failure. The major issues now concern reliability of the irror
checking hardware itself, as well as new techniques for minimizing the
This work was supported by the lational 9cience Foundation anier 'rantE'NG-9-O !686.1
13. DIGITAL SIGNAL AND IMAGE PROCESSING
complexity and cost of the overhead hardware.
Progress has been reported in the design of complex residue arithmetic
3 for use in spectral analysis, processing of bandpass signals represented in
analytic form, and in complex cepstral analysis for speech3 F1B.13 ,193.o, 3 .32 1. Also a design has been presented which permits the block
elements of a general mixed radix converter to be controlled in such a way
I that the general converter can be electronically adapted to sequentially
generate all projections that are required for error location ri3.261. Since
this is a significant step toward being able to implement an entire error
checker on a single VLSI integrated circuit, a great leal of effort is being
expended to extend this concept in order to bring failure resistant concepts
into ractical usage.
j jA new effort has been started to evaluate the advantages and to extend
tha theory of redundant RNS product codes for digital signal processing.
I Rather than simply adding redundant residue digits, data is encoded in a
product code by multiplication with a code generator. An RNS product -ode is
an aritlmetic code which is preserved ,nder error free arithmetic operations.
'hnile not too much is Inown about the theory of RNS product codes in signal
processing, initial indications are that they have useful properties worthy of
further study.
13.2 Roundoff Noise and Limit Cycles in Digital Filtering*
I It is well known that canonical Lmolementations of narrowband recursive
digital filters can exhibit large error due to arithmetic roundoff. However,
I this error can be reduced at the expense of additional hardware complexity.
For example, Mullis and Roberts have developed a procedure for finding the
minimutr roundoff noise linear state-space (LSS) filter structure _iC.
I Another realization for reducing errors in digital filters is the error
spectrum shaping (ISS) structure R2,R31. The ESS structure utilizes
I uantizer feedback and generally has a simpler implementation than the LSS
structure.
Ihi3 work was supported by the Joint Services 'lectroniis Program(U.S. Army. U.S. Navy, F.S. Air Force) under Jontract N00014-7?-3-0424.
*4
112
19. DIGITAL SIGNAL AND IMAGE PROCESSING
We have compared the noise reduction performance of the ESS structure
with that of the optimal LSS structure for second order digital filter
sections. It was found that optimal direct form 1 and direct form 2 ESS
realizations always have a higher signal-to-noise ratio than the optimal LSS
structure. Several suboptimal ESS structures with simple hardware
implementations were also considered and optimal values for the ESS
coefficients in these structures were derived. For filters with zeroes at z =
-1, it was found that several of the simple ESS structures can outperform the
optimal L3S structure. For elliptic filters it was found that the suboptimal
direct form 2 ESS structures perform poorly, but that the direct form 1 ESS
structures perform well. We are currently investigating high-order ESS
structures and developing ESS implementations with low coefficient
sensitivity.
Our work concerning limit cycles covers two topics. First, we have
studied the concept of accessibility, introduced by Claasen et al. rR41. A
zero-input limit cycle is accessible if it can be reached from a state which
is not a state of the limit cycle. It is important to study accessible limit
cycles, since they are most frequently observed. We have shown that
accessibility places restrictions on the error sequences introduced at the
filter quantizers. In the case of two rounding quantizers, these restrictions
can be used to derive a lower bound on the minimum amplitude accessible limit
cycle. A related, but different, result holds for two truncating quantizers.
We have also shown that only certain states can directly access a limit cycle
and that large limit cycles tend to be accessible.
The second limit cycle topic is that of determining exact maximum
amplitude limit cycles in digital filters. We have recently ievised an
algorithm for finding the maximum amplitude limit cycle of a tarticular
period. This is accomplished by considering all possible initial states,
which can be maximum amplitude states of a limit cycle. Results have been
obtained which significantly reduce the number of states requiring
consideration, thereby resulting in a computationally feasible algorithm.
I113
13. DIMIAL SITNAL AND :4Aj3 .?CES3SN1
13.3 'Tew Techniques in Digital 3ignal Processing for Synthetic kDerture Radar*
Research in synthetic aperture radar (3AR) has been conducted in four
distinct, but related projects: (I ) analysis of inherent phase distortion in
rectangular format FFT algorithms, (2) a SAR computer simulation with polar
I format FFT processing, (3) a study of 2D interpolators for polar-to-
rectangular transformation, and (1) an investigation of number theoretic
concepts for high speed failure resistant digital processors required in
real-time SAR. Recently, special emphasis was placed on analysis of spotlight
mode operation, including requirements for 2D spatially-varying interpolators
for molar-to-rectangular coordinate transformation required in FFT processing.
Five 2D interpolators have been analyzed and simulated: (1) 1st order inverseJ distance (nonseparable), (2) 1st order separable bilinear, '3) 3rd order
separable Lagrange, (4) separable optimal FIR filtering, and (5) zero order
I interpolation with over-sampling. Attention was concentrated on low order
interpolators because they are simple enough to imDlement in practice.
.\ithouigh many questions remain, it appears that the use of optimal
space-invariant FIR interpolation coupled with nearest-neighbor selection
\zero order interpolation) may be the most efficient computational means to
achieve high resolution performance.
It was discovered that in polar format FFT SAR processing, each line of
recorded data is the Fourier transform of a density integral of the tar-et
area, collected at a specified look angle. By interpreting the data recording
in this way, it was possible to establish that the SAR principle does not3 depend on instantsneous motion between the target and the radar, but rather on
the total view angle through which projections are taken. This principle
seems to be obscured in the present literature, where 3AR is discussed in
terms of Doppler frequency analysis, thereby implying a dependence on
instantaneous Doppler frequency. Recent work has concentrated on learning ho:f
reconstruction can best be accomplished from incomplete frequency iomain data,
which results in SAR due to limited view angles.U5 * This work was suoported by the Air Force Office of 1,ientific Research unler
grant AFSSR-19-)1)2,A.
. - .
114
1. DIGITAL SIGNAL AND IMAGE ROCESSING
13.4 New Directions in Multidimensional 3inal Processing*
We have recently begun a program of basic research in multidimensional
signal processing spanning the areas of digital array processing, linear space
varying processing, and nonlinear processing using order statistics. In array
processing, efforts are being concentrated on the development of a common
mathematical framework for computer aided tomography (CAT) and spotlight-mode
synthetic aperture radar (SAR). The underlying principle in CAT is the
projection-slice theorem. We have shown that this same theorem also forms the
basis for spotlight-mode SAR; hence similar processing is required in both
systems. Future work will explore the possibility of applying efficient "AT
algoritms to SAR and vice versa. In the area of linear space varying
processing we have extended recent results concerning required sampling rates
for time varying digital systems. Unlike the time invariant case, the
required sampling rate depends on both the input bandwidth and the desired
analog impulse response. We are presently investigating the use of simple
time varying digital filters to perform time invariant filtering. In the area
of nonlinear filtering we have studied a generalized median filter whose
output is a weighted sum of order statistics. Assuming a constant signal in
white background noise, a procedure has been developed for finding optimal
weight values. Deterministic properties of the generalized median filter and
various possible recursive order statistical filters are also being studied.
13.5 Image Sequence Processing and Dynamic Scene Analysis**
The processing of images involving motion has become increasingly
important. Applications include target tracking, dynamic robot vision, image
bandwidth compression, dynamic medical imaging, and highway traffic
monitoring. Image sequence processing involves a large amount :f iata.
4owever, because of the rapid progress in computer, LS1. and VLSI
technologies, many useful processing tasks for image sequences oan now be
performed in a reasonable amount of time.
* This work was sunoorted by the Joint Services Electronics Program(U.S. Army, U.S. Navy, T.S. Air Force) under Contract N00014-79-C-5421.
This work was supported by the lational Science Foundation under 3rantACS-7-1 3906.
II
1153 13. DT3rTAL SIAL XND IMAGE PROCESSING
A central issue in image sequence processing is motion estimation.3 During the past year we have developed a number of efficient algorithms for
estimating the three-dimensional motion parameters of rigid bodies 'both
translation and rotation) from two time-sequential image frames. The
following uniqueness theorems have also been proven: (i) For a rigid planar
patch, two solutions for the motion parameters dill exist in general.
(ii) For a rigid curved surface, the solution for the motion parameters is
K I unique.
18.6 Underwater Image Processing*
We ha-re been collaborating with the Naval Research Laboratory on a basic
j research project in underwater image processing using both optical and sonar
sensors. Amliations include searching for man-made objects on the ocean
floor and ocean mining. During the past year, our research has concentrated
on studying the effects of ocean water on optical and sonar images and on
finding shade and texture features of images of the ocean floor that 2an
i distinguish between man-made and natural objects. Successful algorithms for
extracting shape features have been developed. We have also studied the
characteristics of various acoustical imaging methods and concluded that
synthetic aperture sonar and holography are the two most promising approaches
for underwater imaging.
imagin_
'References
SI11 , 3. T. Mullis and R. A. Roberts, "Synthesis of Xinim,'m Roundoff Noise
Fixed Point Digital Filters," TSE. Trans. on - iuits and 3SstemsI ZA3-23, 551-562 rSept. 1170').
I
This work -as supported by the )ffiie of Naval Research ziler ontractUo001
!4
116
18. DIGITAL SIGNAL AND LAG3 PROCESSING
rR21 T. Thong and B. Liu, "Error Spectrum Shaping -n Narrowband Recursive
Digital Filters," IEEE Trans. on Acoustics, Speech, and Signal
Processing A3SP-25, 200-203 (April 1977).
LR11 k. I. Abu-El-Haija and A. M. Peterson, "An Approach to Eliminate
ioundoff Errors in Digital Filters," IEEE Trans. on Acoustics, Speech,
and Signal Processing XSSP-27, 195-198 (April 1979).
R41 T. A. C. A. Claasen, W. F. . ecklenbrauker, and J. B. H. Peek, "Some
Remarks on the Classification of Limit Cycles in Digital Filters,"
?hilips Research Report 29, 297-305 'Aug. 1973).
117
1 9. PO P!LATISN /FOOD /CLIMATV, 3TUDT5-,S*
I 7
n ?aculty and Senior 3taff
I IP. Handler J. Kane
Graduate StudentsI3 J. Posluszny M. Tao-Liou J. Zielinski
19.1 Summary of Progress
This group has two primary areas of interest. The first is
computer-assisted education in population and food and their impact on other
social and economic areas of society. The second is concerned with climate
and food and the impact of large interannual variations of climate on the
I availability of food for the world's growing population.
In the first area, population and food, we have designed a
I computer-assisted education program concerned with world population as well as
specific countries. The program is available in BASIC and has been adapted
for use on the Apple computer. The program is used in over 200 educational
institutions each year. Many of the programs are used at the high school
i level where distribution of the programs has just begun luring the past year.
In the second area, food and climate, a retrospective study has been made
of the relationship of crop production to climate anomalies. Various
interesting climate sequences have been observed where a :ertain type of
climate anomaly in one region of the world is very frequently followed by
another anomaly in another part of the world.
* This work was supported by the National Sience Foundation under 3rant SEDI9-1 3501 and Unl ied tates Department of Agriculture Contract SEA
53-1 9?B-0- 334.I1
118
I 3. ?OP'JLATION/FOOD/CLIIATq 3TItES
These sequences of climate anomalies can be usei as a means of crop
forecasting in the region of the later anomaly. Various techniques are being
levelopel to attempt to recognize these sequences in the Iorli climate pattern
of the past twenty years. Yhen special patterns are obser-ei they are testel
agist longer PIta bases. Some of the lata are 2oatainel 1 time seri s
,ri -h ,o back to I36S. A paper lescribiag the reiatLonship of c-rops in
Australia and Inionesia to the climate anomalies of the tropizal Pacifi3 3cean
nas been acceptei for publication by the journal Mlimate Change 19 .1l.
I(continued from front page)
American Society for Information Science International Research and Exchange BoardArco Solar, Inc. Knowledge Science, Inc.Atlantic Richfield Corporation Learned InformationEastman Kodak Company Library ResourceEngineering Index, Inc. Magnavox CompanyFord Motor Company Memorex CorporationGTE Laboratories Shearson HaydenHarris Corporation State of Illinois LibraryHewlett-Packard Company Universal Oil Products Inc.Hitachi LimitedInstitute of Electrical and Electronics
Engineers, Inc.
Reproduction in whole or in part is permitted for any purpose of the United StatesGovernment
a-
(m.
