ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOLUME 64B
Advances in
Electronics and Electron Physics EDITEDBY PETER ...
107 downloads
1204 Views
14MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOLUME 64B
Advances in
Electronics and Electron Physics EDITEDBY PETER W. HAWKES Laboratoire d’Optique Electronique du Centre National de la Recherche Scientijique, Toulouse, France
VOLUME 64B 1985
ACADEMIC PRESS (Harcourt Brace Jovanovich, Publishers)
London Orlando San Diego New York Toronto Montreal Sydney Tokyo
Photo-Electronic Image Devices PROCEEDINGS OF THE EIGHTH SYMPOSIUM HELD AT IMPERIAL COLLEGE, LONDON, SEPTEMBER 5-7, 1983
EDITEDBY
B. L. MORGAN The Bluckett Lahorutory, Imperial College, University of London, London, England
I985
ACADEMIC PRESS (Harcourt Brace Jovanovich, Publishers)
London Orlando San Diego New York Toronto Montreal Sydney Tokyo
COPYRIGHT 0 1985, BY ACADEMIC PRESS INC. (LONDON) LTD. ALL RIGHTS RESERVED. NO PART O F THIS PUBLICATION MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER
ACADEMIC PRESS INC. (LONDON) LTD. 24-28 Oval Road LONDON NWI 7DX
United States Edition published by ACADEMIC PRESS, INC. Orlando, Florida 32887
LIBRARY OF CONGRESS CATALOG CARD NUMBER: 49-7504 ISBN: 0-12-014724-6 PRINTED IN THE UNITED STATES OF AMERICA 85060788
9 8 7 6 5 4 3 2 1
CONTENTS CONTENTS OF VOLUMEA CONTRIBUTORS. . PREFACE. ABBREVIATIONS. .
vii ix xvii xviii
.
Microchannel Plate Intensifiers and Properties of Photocathodesand Phosphors Detection Efficiencies of Far-Ultraviolet Photon-Counting Detectors. By G. R. 299 CARRUTHERS A N D C. B. OPAL. High-Resolution and Large Size Wafer Microchannel Image Intensifier. By B. JEAN, 315 J. P. BOUTOT,V. DUCHENOIS, A N D R. POLAERT . A Two-Dimensional Photon-CountingTube. By M. KINOSHITA,K. KINOSHITA, K. 323 YAMAMOTO,A N D Y. SUZUKI. High Spatial and Temporal Resolution imaging with a Resistive Anode Photon L.SALAS,E. Rufz, G. F. BISIACCHI, Counter. By C. F I R MANI L.,GUTICRREZ, 33 I A N D F. PARESCE . . Output Energy Distribution of a Microchannel Plate. By N. KOSHIDA.M. 331 MIDORIKAWA, A N D Y. KIUCHI. Computer Analysis of the Temporal Properties of a Microchannel Plate Photomulti343 plier. By K. OBAA N D M. ITO. The MCP as a High-Energy Particle Track Detector. By K. OBA,P. REHAK,AND 355 D. M. POTTER . High-Resolution Luminescent Screens for Image intensifier Tubes. By V. 365 . DUCHENOIS, M. FOUASSIER, A N D c . PIAGET. Multialkali Effects and Polycrystalline Properties of Multialkali Antimonide Photo313 cathodes. By WU QUAN-DE A N D LlU LI-BIN Properties of a Photocathode with a Palladium Substrate. By ZHANGXIAOQIU, 385 PANGQICHANG, A N D LEI ZHIYUAN. The Luminous Efficiency of a Phosphor Layer in the Forward and Backward Direc393 tions. By A. G. DU TOITA N D C. F. VAN HUYSSTEEN . Study of ESBl and Sensitivity Characteristics of Ag-0-Cs Systems for Image Tubes. 403 By M. SRINIVASAN, M. D. VAIDYA, D. R. KULKARNI, A N D T. B. BHATIA . 41 1 A Near-infrared Photocathode. By TAOCHAO-MING.
.
Television Systems and X-Ray Intensification
A Magnetic Focus Electrostatic Deflection Compact Camera Tube. By M. 415 . KURASHIGE,S. OKAZAKI, A N D C. Ocusu Pyroelectric Vidicons for Submillimeter Wavelengths. By W. M. WREATHALL. 425 An Amorphous Silicon Vidicon Tube. By B. L. JONES,J. BURRAGE, A N D R. 431 HOLTOM . A Physical Model of Heterostructure Targets for Camera Tubes. By M. JEDLICKA 447 AND F. SCHAUER . A Quality Figure for the Emission System in Camera Tubes. By SHENCHING-KAI, 463 FENGCHIH-TAO, TUNGKUN-LIN, A N D FANGER-LUN . V
vi
CONTENTS
Recent Developments in Real-Time Image Processing. By R. AUBERT,B. BUHLER, A N D W. GEBAUER 469 The Evaluation of Silicon CCDs for Imaging X-Ray Spectroscopy in the Range I to 8 keV. By R. E. GRIFFITHS . 483 X-Ray Imaging and Spectroscopy with CCDs. By D. H. LUMB,G. R. HOPKINSON, 497 A N D A. A. WELLS . Configuration and Performance of a Television X-Ray Detector System for Imaging and Diffraction Applications. By K. KALATA,S. S. MURRAY,A N D J. H. 509 CHAPPELL. . A Solid-state Slit Scan X-Ray Detector in Large Field of View (LFOV) Radiology. By H. ROUGEOT,G . ROZIERE.AND B. DRIARD . 52 I A Gated X-Ray Intensifier with a Resolution of 50 Picoseconds. By A. K. L. . 53 I DYMOKE-BRADSHAW. J. D. KILKENNY, A N D J. WESTLAKE An Experimental TV Camera Tube Sensitive to the Soft X-Ray Region. By SHE . YONG-ZHENG, YANG XlAOWEN, AND DING YlSHAN 54 I Electron Optics and Miscellaneous Applications The Effect of Electron Optics on the Properties of the X-Ray Image Intensifier. By 549 V. JARES . Variational Theory of Aberrations in Cathode Lenses. By XIMEN JI-YE. ZHOULIWEI, A N D AI KE-CONG . 56 1 . 575 A Generalized Theory of Wide Electron Beam Focusing. By ZHOULI-WEI Oblique Magnetic Focus Point Spread Profiles and MTFs. By 3. L. LOWRANCE. 591 Electron Beam Deflection in the Focusing Magnetic Field of a Camera Tube. By Y. KIUCHIA N D T. SAKUSABE. 601 Subpicosecond Chronoscopy Using a Photochron 1V Streak Camera. By M. R. BAGGS,R. T . EAGLES,w. MARGULIS, w. SlBBETT, A N D w . E. SLEAT. 617 A Picosecond Framing Camera for Single or Multiple Frames. By M. R. BAGGS, 627 R. T. EAGLES,w. MARGULIS. W. SIBBEIT, A N D W. E. SLEAT . A Spatial Light Modulator. By T. HARA,M. SUGIYAMA, A N D Y. SUZUKI . . 637 Imaging Characteristics of Rigid Coherent Fiber Optic Tapers. By C. I. COLEMAN649 Developments in S . I Photocathode Image Converters for High Speed Streak/Fram663 . ing Photography. By J. H. GOODSONA N D B. R. GARFIELD Performance of a Picosecond Streak Camera Used in Conjunction with a Photodiode Array Measuring System. By C. CAVAILLER, N. FLEUROT, A. MENS, G. 67 1 KNISPEL,A N D J. A. MIEHE .
INDEX .
.
681
CONTENTS OF VOLUME A Electronography and Imaging Photon-Counting Systems A New Concept in the Development of a Very Large Field Electrographic Camera. By X. Z. Jia and P. J. Griboval. Installation d'une Camera Electronique Grand Champ au Telescope Canada-FranceHawaii. By B. Servan, G. Wlkrick. L. Renard, G . Lelievre, V. Cayatte, D. Horville, et J. Fromage. Photon-Counting Imaging and Its Application. By Y. Tsuchiya, E. Inuzuka. T. Kurono, and M. Hosoda. Interpolative Centroiding in CCD-Based Image Photon-Counting Systems. By A. Boksenberg, C. 1. Coleman. J. Fordham, and K. Shortridge. The Imperial College System for Photon Event Counting. By R. W. Airey, D. J. Lees, B. L. Morgan, and M. J. Traynar. Image Intensifiers Performance and Reliability of Third-Generation Image Intensifiers. By H. K. Pollehn. Third-Generation Image Intensifier. By E. Roaux, J. C. Richard, and C. Piaget. A Proximity-Focused Image Intensifier for Astronomy. By R. H. Cromwell, P. A. Strittmatter, R. G. Allen, E. K. Hege, H. Kiihr. K.-H. Marien. H. W. Funk, and K. Frank. Super Inverter Image Intensifier. By L. K. Van Geest and K. W. J. Stoop. A Large-Area Electron Image Multiplier. By D. Washington, A. J. Guest, and A. G. Knapp. A 512 Channel Parallel-Output Detector. By R. Rudolph, H. Tug, and Th. Schmidt-Kaler. Diode Intensifier Tube with Fast Phosphor Screen, By J. P. Boutot, R. Goret, M. Jatteau, J. Paulin. and J. C. Richard. Intensifier Solid-state Detector for Light Pulse Barycenter Reconstruction. By H. Rougeot, G. Roziere, and B. Driard. The Prototype MOSAIC Detector. By D. Weistrop, J. T. Williams, and R. P. Fahey. Design and Performance of the High-Resolution Spectrograph Sensor Subsystem. By H. J. Eck, E. A. Beaver, and J. L. Shannon. Charge-Coupled Devices Electrographic Detectors versus Charge-Coupled Devices: A Comparison of Two Quality Panoramic Detectors for Stellar Photometry. By H. M. Heckathorn, C. B. Opal, P. Seitzer, E. M. Green, and E. P. Bozyan. Evaluation of the GEC 385 x 576 Charge-Coupled Device Image Sensor for Astronomical Use. By B. Thomsen and E. Sendergaard. Photometric and Spectroscopic Performance of a Thinned RCA CCD Detector. By R. W. Leach. The UCL Charge-Coupled Device Camera at the South African Astronomical Observatory. By D. Walker. P. Sandford. A. Lyons, J. Fordham, D. Bone, A. Walker, and A. Boksenberg. Getting More by Taking Less: A Method Of Summing up Pixels on a CCD Imager. By J. R. Krdmm and H. U. Keller. vii
...
Vlll
CONTENTS OF VOLUME A
A CCD Camera for Cinematographic Use in Astronomy. By B. Fort, J. P. Picat. C. Lours, J. P. Dupin, P. Tilloles, F. Avitabile, G. Bailleul, and J. L. Prieur. Low(es1) Noise Reticon Detection Systems. By G. A. H. Walker, R. Johnson. and S. Yang. Reticon Detector Electronics for the Halley Multicolor Camera on the Giotto Space Mission. By H. J. Meyer, W. K. H. Schmidt, and H. Rosenbauer. Investigation of CCD-Digicon Detector System Characteristics. By R. G. Hier, E. A. Beaver, G. W. Schmidt, and C. E. Mcllwain. An Intensified Photodiode Array Detector for Space Applications. By K. S. Long, C. W. Bowers, P. D. Tennyson, and A. F. Davidsen. A Charge-Sensitive Readout Technique for Infrared Photoconductors. By G. W. Schmidt, R. G. Hier, S. E. Nelson, and R. C. Puetter. Thinned Backside-Bombarded RGS-CCD for Electron Imaging. By M. Lemonier, C. Piaget, and M. Petit. L L L TV Imaging with GaAs Photocathode/CCD Detector. By Y. Beauvais. J. Chautemps, and P. De Groot. A CCD image Sensor Using a Glow Discharge Amorphous Si Photoconductive Layer. By 0. Yoshida, N . Harada, K. Ide. and T. Yoshino. An Improved 2438 Element Three-Phase CCD Linear Image Sensor. By You Zhong-Qiang, Pan Shu-Ren, and Chen Yi-Fei. Improved Diagnostic Radiography and Reduced Radiation Exposure Using a 1024 x 1024 Pixels Linear Diode Array Imaging System. By D. Sashin, J . Horton, E. J . Sternglass, K. M. Bron, B. S. Slasky, J. M. Herron. W. H. Kennedy, J. W. Boyer. B. R. Girdany, and R. W. Simpson.
A1 KE-CONC,Xian Research Institute of Applied Optics, Xian, China (p. 561) R. W. AIREY,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 282, England (p. 49) R. G. ALLEN,Steward Observatory, The University of Arizona, Tucson, Arizona 85721, U.S.A. (p. 77) R. AUBERT,Contraves AG, Schaffauser Strasse 580, Ch 8052 Zurich, Switzerland ( p . 469) F. AVITABILE,Observatoire de Toulouse. 14 Avenue Edouard Belin, 31400 Toulouse, France (p. 205) M. R. BAGGS, The Blackett Laboratory, Imperial College of Science and Technology, London SW7 2BZ, England (pp. 617 & 627) G. BAILLEUL, Observatoire de Toulouse. 14 Avenue Edouard Belin, 31400 Toulouse, France (P.205) Y. BEAUVAIS,Thomson-CSF, Division Tubes Electroniques, 38 Rue Vauthier, 92102 Boulogne-Billancourt, France (p. 267) E. A. BEAVER,Center for Astrophysics and Space Sciences, University of California. San Diego, La Jolla, California 92093, U.S.A. (pp. 141 & 231) T . B. BHATIA,Bharat Electronics Limited, NDA Road, Pashan, Pune 411021, India ( p . 403) G . F. BISIACCHI, Instituto de Astronomia, Universidad Nacional Aut6noma de Mexico, 04510 MPxico D.F. (p. 331) A. BOKSENBERG, Royal Greenwich Observatory, Herstmonceux Castle, Hailsham, East Sussex BN27 IRP, England ( p p . 33 & 185) D. BONE,Department of Physics and Astronomy, University College London, London WClE 6BT. England (p. 185) J . P. BOUTOT,Laboratoires d’Electronique et de Physique Appliquke, 3 Avenue Descartes, 94450 Limeil BrPvannes, France (pp. 113 & 315) C. W. BOWERS,Department of Physics, Johns Hopkins University, Baltimore, Maryland 21218, U.S.A. (p. 239) J . W. BOYER,Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. (p. 289) E. P. BOZYAN,McDonald Observatory, The University of Texas, Austin, Texas 78712, U.S.A. (p. 153) K . M. BRON,Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. (p. 289) B. BOHLER,Contraves AG, Schaffauser Strasse 580, Ch 8052 Zurich, Switzerland ( p . 469) J . BURRAGE,English Electric Valve Company Limited, Waterhouse Lane, Chelmsford, Essex CMI 2QU, England (p. 437) G. R. CARRUTHERS, E. 0. Hulburt Center for Space Research, U.S. Naval Research Laboratory. Washington D.C. 20375, U.S.A. (p. 299) C. CAVAILLER, Commissariat d I’Energie Atomique, Centre dEtudes de Limeil Valenton, 94190 Villeneuve-Saint-Georges, France ( p . 671) V. CAYATTE,Observatoire de Paris, 92190 Meudon, France ( p . 11) J. H. CHAPPELL,Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, U.S.A. (p. 509)
ix
X
CONTRIBUTORS
J. CHAUTEMPS, Thomson-CSF, Division Tubes Elecironiques, 38 Rue Vathier, 92102 Boulogne-Billancourt, France (p. 267) CHENYI-FEI,Hebei Semiconductor Research Instiiute, Shuiazhuang, Hebei, China (p. 285) C. I . COLEMAN, Marconi Space and Defence Systems Limiied, The Grove, Warren Lane, Stanmore, Middlesex HA7 4LY. England (pp. 33 & 649) R. H . CROMWELL, Steward Observatory, The University of Arizona, Tucson, Arizona 85721, U.S.A. (p. 77) A. F. DAVIDSEN, Depariment of Physics, Johns Hopkins University, Baltimore, Maryland 21218, U.S.A. (p. 239) P. DE GROOT,Thomson-CSF, Division Tubes Electroniques, 38 Rue Vauthier, 92102 Boulogne-Billancouri, France (p. 267) DINGYISHAN,Changchun Institute of Optics and Fine Mechanics, Changchun, China (p. 541) B. DRIARD, Thomson-CSF, Division Tubes Electroniques, 38 Rue Vathier, 92102 BoulogneBillancourt, France (pp. 123 & 521) V. DUCHENOIS, Laboratoires d’Electronique et de Physique AppliquCe, 3 Avenue Descartes, 94450 Limeil Brgvannes, France (pp. 315 & 365) J . P. DUPIN,Observatoire de Toulouse, 14 Avenue Edouard Belin, 31400 Toulouse, France (P. 205) A. K. L. DYMOKE-BRADSHAW, The Blackett Laboratory, Imperial College of Science and Technology, London SW7 2BZ, England (p. 531) R. T. EAGLES,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 2BZ, England ( p p . 617 & 627) H . J . ECK,Ball Aerospace Systems Division, P.O. Box 1062, Boulder, Colorado 80306, U.S.A. (p. 141) R. P. FAHEY,Laboraiory for Asironomy and Solar Physics, Goddard Space Flight Center, Greenbelt, Maryland 20771, U.S.A. (p. 133) FANGER-LUN,North Industries Corporation, Beijing, China (p. 463) FENGCHIH-TAO, Kunming Instiiute of Physics, Kunming, China (p. 463) C. FIRMANI, Instiiuio de Astronomia, Universidad Nacional Autdnoma de Mexico, 04510 Mkxico D.F. (p. 331) N . FLEUROT, Commissariat a I’Energie Aiomique. Centre d’Etudes de Limeil Valenton, 94190 Villeneuve-Saint-Georges, France (p. 671) J , FORDHAM, Department of Physics and Astronomy, University College London, London WClE 6BT, England ( p p . 33 & 185) B. FORT, Observatoire de Toulouse, 14 Avenue Edouard Belin, 31400 Toulouse, France (p. 205) M. FOUASSIER, Laboratoires d’Electronique et de Physique AppliquCe, 3 Avenue Descaries, 94450 Limeil Brbvannes, France (p. 365) K. FRANK,Proxitronic-Funk GMbH & Co. KG, Rudolf Diesel Strasse 23, 0-6108 WeiterstadtlDarmstadi, Federal Republic of Germany (p. 77) J . FROMAGE, Observaioire de Paris, 92190 Meudon, France ( p . 11) H.W. FUNK,Proxitronic-Funk GMbH & Co. KG, RudolfDiesel Strasse 23, D-6108 WeiterstadtlDarmstadt. Federal Republic of Germany (p. 77) B. R. GARFIELD, English Electric Valve Company Limited, Waierhouse Lane, Chelmsford, Essex CMI 2QU, England (p. 663) W. GEBAUER, Contraves AG. SchafFauser Strasse 580, Ch 8052 Zurich, Switzerland ( p . 469)
CONTRIBUTORS
xi
B. R. GIRDANY,Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. (p. 289) J. H. GOODSON,English Electric Valve Company Limited, Waterhouse Lane, Chelmsford, Essex CMI 2QV, England ( p . 663) R. GORET,Laboratoires dElectronique et de Physique Appliqube, 3 Avenue Descartes, 94450 Limeil Brivannes, France ( p . 113) E. M. GREEN,Mount Stromlo and Siding Spring Observatory, Australian National University, Canberra, Australia (p. 153) P. J. GRIBOVAL, McDonald Observatory, The University of Texas, Austin, Texas 78712, U.S.A. ( p . 1) R. E. GRIFFITHS,Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, U.S.A. (p. 483) A. J. GUEST,Philips Research Laboratories, Redhill, Surrey RHI SHA, England ( p . 101) L. GUTI~RREZ, Instituto de Astronomla, Universidad Nacional Aut6noma de Mexico, WSlO Mbxico D.F. ( p . 331) T . HAM, Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435, Japan ( p . 637) N. HARADA,Toshiba Research and Development Center, Toshiba Corporation, Kawasaki, Kanagawa 210, Japan (p. 275) H. M. HECKATHORN, E. 0. Hulburt Center for Space Research, U.S.Naval Research Laboratory, Washington D.C. 20375, U.S.A. ( p . 153) E. K. HEGE,Steward Observatory, The University of Arizona, Tucson, Arizona 85721, U.S.A. (p. 77) J. M. HERRON,Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh. Pennsylvania 1S261, U.S.A. (p. 289) R. G. HIER,Center for Astrophysics and Space Sciences, University of California, San Diego, La Jolla, California 92093, U.S.A. ( p p . 231 & 251) R. HOLTOM,English Electric Valve Company Limited, Waterhouse Lane, Chelmsford, Essex CM1 2QU, England ( p . 437) G. R. HOPKINSON, Department of Physics, University of Leicester, Leicester LEI 7RH, England ( p . 497) J. HORTON,Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. ( p . 289) D. HORVILLE, Observatoire de Paris, 92190 Meudon, France (p. 11) M. HOSODA,Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435, Japan ( p . 21) K. I D E , Toshiba Research and Development Center. Toshiba Corporation, Kawasaki, Kanagawa 210, Japan ( p . 275) E. INUZUKA. Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435, Japan (p. 21)
M. ITO, Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 43S, Japan ( p . 343) V. JARES,TESLA Vacuum Technics, Prague, Czechoslovakia ( p . 549) M. JATTEAU, Laboratoires d'Electronique et de Physique Appliqube, 3 Avenue Descartes, 94450 Limeil Brbvannes, France ( p. 113) B. JEAN, Laboratoires d'Electronique et de Physique Appliqube, 3 Avenue Descartes. W S O Limeil Brtvannes, France ( p . 315) M. JEDLICKA, TESLA Vacuum Technics, Prague, Czechoslovakia (p. 447) X. 2. JIA, McDonald Observatory, The University of Texas, Austin, Texas 78712, U.S.A.
(P. 1) R.
Department of Geophysics and Astronomy, University of British Columbia, Vancouver, British Columbia V6T 1 W5, Canada (p. 213)
JOHNSON,
xii
CONTRIBUTORS
B. L. JONES, English Electric Valve Company Limited, Waterhouse Lane, Chelmsford, Essex CMI 2QU, England (p. 437) K. KALATA,Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, U.S.A. (p. 509) H. U. KELLER,Max-Planck-lnstitut fur Aeronomie, 0-3411 Katlenburg-Lindau 3 , Federal Republic of Germany (p. 193) W. H. KENNEDY, Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh. Pennsylvania 15261, U.S.A. (p. 289) J. D. KILKENNY, The Blackett Laboratory, Imperial College of Science and Technology, London SW7 282, England (p. 531) K. KINOSHITA, Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435. Japan (P. 323) M. KINOSHITA. Hamamatsu Photonics K.K., 1126 Ichino-Cho. Hamamatsu 435. Japan (P. 323) Y. KIUCHI,Department of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184, Japan (pp. 337 & 601) A. G . KNAPP,Philips Research Laboratories, Redhill, Surrey RHl 5HA, England (p. 101) G. KNISPEL,Centre de Recherches NuclCaires, Physique des Rayonnements et Electronique NuclCaire, 67037 Strasbourg, France (p. 671) N. KOSHIDA, Department of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184, Japan (p. 337) J . R. KRAMM, Max-Planck-lnstitut fur Aeronomie, 0-341I Katlenburg-Lindau 3, Federal Republic of Germany (p. 193) H. KUHR,Steward Observatory, The University of Arizona, Tucson, Arizona 85721, U.S.A. (P. 77) D. R. KULKARNI, Bharat Electronics Limited, NDA Road, Pashan, Pune 411021. India (P. 403) M. KURASHIGE, NHK Technical Research Laboratories, 1-10-11 Kinuta. Setagaya-ky, Tokyo 157, Japan (p. 415) T. KURONO, Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435, Japan ( p . 21) R. W. LEACH,The University of Texas at Austin, Austin, Texas 78712, U.S.A. (p. 177) D. J. LEES,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 282, England (p. 49) LEIZHIYUAN, Xian Institute of Optics and Precision Mechanics, Academia Sinica, Xian, Shaanxi. China (p. 385) G. LELI~VRE, Observatoire de Paris, 92190 Meudon, France (p. 1 I ) M. LEMONIER, Laboratoires dElectronique et de Physique AppliquPe, 3 Avenue Descartes, 94450 Limeil BrCvannes, France ( p . 257) LIULI-BIN,Department of Radio-Electronics, Peking University, Beding, China ( p . 373) K. S. LONG,Department of Physics, Johns Hopkins University, Baltimore, Maryland 21218, U.S.A. (p. 239) C. LOURS, Observatoire de Toulouse, 14 Avenue Edouard Belin, 31400 Toulouse, France (P. 205) J. L. LOWRANCE, Princeton University Observatory, Princeton, New Jersey 08544, U.S.A. (P. 591) D. H. LUMB, Department of Physics, University of Leicester, Leicester LEI 7RH, England (P. 497) A. LYONS,Department of Physics and Astronomy, University College London, London WCIE 6BT, England (p. 185)
CONTRl BUTORS
xiii
C. E. MCILWAIN.Center for Astrophysics and Space Sciences, University of California, Sun Diego, La Jolla. California 92093 (p. 231) W. MARGULIS,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 2BZ, England (pp. 617 & 627) K.-H. MARIEN, Steward Observatory, The University of Arizona, Tucson, Arizona 85721, U.S.A. (p. 77) A. MENS,Commissariat a I'Energie Atomique, Centre d'Etudes de Limeil Valenton, 94190 Villeneuve-Saint-Georges, France ( p . 671) H. J. MEYER,Max-Planck-lnstitut fur Aeronomie, 0-3411 KatlenburglLindau 3, Federal Republic of Germany (p. 223) M. MIDORIKAWA, Department of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184, Japan (p. 337) J. A. MIEHE,Centre de Recherches Nucle'aires, Physique des Rayonnements et Electronique Nucle'aire, 67037 Strasbourg, France (p. 671) B. L. MORGAN,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 282, England (p. 49) S. S. MURRAY, Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, U.S.A. (p. 509) S . E . NELSON,Center for Astrophysics and Space Sciences, University of California, San Diego, La Jolla, California 92093, U.S.A. (p. 251) K. OBA,Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435, Japan (pp. 343 & 355)
C. OGUSU,NHK Technical Research Laboratories, 1-10-11 Kinuta, Setagaya-ky, Tokyo 157, Japan (p. 415) S. OKAZAKI, NHK Technical Research Laboratories, 1-10-11 Kinuta, Setagaya-ky, Tokyo 157, Japan (p. 415) C. B. OPAL,E. 0. Hulburt Center for Space Research, US.Naval Research Laboratory, Wushingfon D.C. 20375, U.S.A. (pp. 153 & 299) PAN SHU-REN , Hebei Semiconductor Research Institute, Shijiazhuang, Hebei, China (p. 285)
PANGQICHANG, Xian Institute of Optics and Precision Mechanics, Academia Sinica. Xian, Shaanxi, China ( p . 385) F. PARESCE,Space Telescope Science Institute, Johns Hopkins University, Baltimore, Maryland 21218, U.S.A. (p. 331) J . PAULIN,Laboratoires d'Electronique et de Physique Appliquke, 3 Avenue Descartes, 94450 Limeil Brhannes, France (p. 113) M. PETIT,Laboratoires dElectronique et de Physique Appliquke, 3 Avenue Descartes, 94450 Limeil Brkvannes, France (p. 257) C. PIAGET,Laboratoires d'Electronique et de Physique Appliqute, 3 Avenue Descartes, 94450 Limeil Bre'vannes, France (pp. 71, 257, & 365) J. P. PICAT,Observatoire de Toulouse, 14 Avenue Edouard Belin. 31400 Toulouse, France (P. 205)
R. POLAERT, Laboratoires dElectronique et de Physique Appliquke, 3 Avenue Descartes, 94450 Limeil Brkvannes, France (p. 315) H. K. POLLEHN, U.S. Army Night Vision and Electro-Optics Laboratory, Fort Belvoir, Virginia 22060, U.S.A. ( p. 61) D. M. POTTER.Rutgers University, Piscataway, New Jersey 08854, U.S.A. (p. 355) J. L. PRIEUR, Observatoire de Toulouse, 14 Avenue Edouard Belin, 31400 Toulouse, France (P.205)
xiv
CONTRIBUTORS
R. C. PUETTER,Center for Astrophysics and Space Sciences, University of California, San Diego, La Jolla, California 92093, U.S.A. (p. 251) P. REHAK,Brookhaven National Laboratory, Upton, New York 11973, U.S.A. (p. 355) L. RENARD,Observatoire de Paris, 92190 Meudon, France (p. 11) J. C. RICHARD,Laboratoires d'Electronique et de Physique AppliquPe, 3 Avenue Descartes, 94450 Limeil Br&uannes,France (pp. 71 & 113) E. ROAUX,Laboratoires d'Electronique et de Physique AppliquPe, 3 Avenue Descartes. 94450 Limeil Brkvannes, France (p. 71) H. ROSENBAUER, Max-Planck-lnstitut fur Aeronomie, 0-3411 Katlenburg-Lindau 3. Federal Republic of Germany (p. 223) H. ROUGEOT, Thomson-CSF, Division Tubes Electroniques, 38 Rue Vathier, 92102 Boulogne-Billancourt, France (pp. 123 & 521) G. ROZIERE. Thomson-CSF, Division Tubes Electroniques, 38 Rue Vathier. 92102 Boulogne-Billancourt, France (pp. 123 & 521) R. RUDOLPH,Astronomisches Institut, Ruhr-Universitat, 4630 Bochum, Federal Republic of Germany (p. 11 1) E. Ru fz,Instituto de Astronomia, Universidad Nacional Autdnoma de MPxico, 04510 MPxico D.F. (p. 331) T. SAKUSABE, Department of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184, Japan (p. 601) L. SALAS,Instituto de Astronom fa, Universidad Nacional Autdnoma de Mexico, 04510 M4xico D.F. (p. 331) P.SANDFORD, Department of Physics and Astronomy, University College London, London WClE 6BT, England (p. 185) D. SASHIN,Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. (p. 289) F. SCHAUER, Military Academy AZ, Brno, Czechoslovakia (p. 447) G.W.SCHMIDT, Centerfor Astrophysics and Space Sciences, University of California, San Diego, La Jolla, California 92093, U.S.A. (pp. 231 & 251) W. K. H. SCHMIDT,Max-Planck-Institut fur Aeronomie, 0-3411 Katlenburg-Lindau 3, Federal Republic of Germany ( p . 223) TH. SCHMIDT-KALER, Astronomisches Institut, Ruhr-Universitat, 4630 Bochum, Federal Republic of Germany (p. 111) P. SEITZER,Cerro Tololo Inter-American Observatory, La Serena, Chile (p. 153) B. SERVAN,Observatoire de Paris, 92190 Meudon, France (p. 11) J . L. SHANNON, Laboratory for Astronomy and Solar Physics, Goddard Space Flight Center, Greenbelt, Maryland 20771, U.S.A. (p. 141) SHEYONG-ZHENG, Changchun Institute of Optics and Fine Mechanics, Changchun. China (P. 541) SHENCHING-KAI, Zhejiang University, Hangzhou, China (p. 463) K. SHORTRIDGE, California Institute of Technology, Pasadena, California 91 109, U.S.A. (P. 33) W. SIBBETT, The Blackett Laboratory, Imperial College of Science and Technology, London SW7 282, England (pp. 617 & 627) R. W. SIMPSON, Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. (p. 289) B. S. SLASKY, Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. (p. 289) W. E. SLEAT,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 2BZ, England (pp. 617 & 627)
CONTRIBUTORS
xv
E. SBNDERGAARD, Institute of Astronomy, University of Aarhus, 8ooO Aarhus C, Denmark (P. 167) M. SRINIVASAN, Bharat Electronics Limited, NDA Road, Pashan, Pune 411021, India
(P.403)
E. J . STERNGLASS, Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, U.S.A. (p. 289) K. W. J . STOOP, B.V. Devt Electronische Producten, Dwazziewegen 2, 9300 AB Roden, Holland (p. 93) P. A. STRITTMATTER, Steward Observatory, The University of Arizona, Tucson, Arizona 85721, U.S.A. (p. 77) M. SUGIYAMA, Hamamatsu Photonics K.K.. 1126 Ichino-Cho, Hamamatsu 435, Japan (P. 637) Y. SUZUKI, Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435, Japan (pp. 323 8t 637) T a o CHAO-MING, Institute of Electronics, Academia Sinica, Beijing, China (p. 41 1) P. D.TENNYSON, Department of Physics, Johns Hopkins University, Baltimore, Maryland 21218, U.S.A. (p. 239) B. THOMSEN. Institute of Astronomy, University of Aarhus, 8ooO Aarhus C , Denmark (P. 167) P. TILLOLES,Obseruatoire de Toulouse, 14 Avenue Edouard Belin, 31400 Toulouse, France (P. 205) A. G . DU TOIT,National Physical Research Laboratory, CSIR, Pretoria ooO1,South Africa (P. 393) M. J. TRAYNAR, The Blackett Laboratory, Imperial College of Science and Technology, London S W7 282, England (p. 49) Y. Tsucnwa, Hamamatsu Photonics K.K., 1126 Ichino-Cho, Hamamatsu 435, Japan (P. 21) H. TUG, Astronomisches Institut, Ruhr-Universitat, 4630 Bochum, Federal Republic of Germany (p. 1 1 1) TUNGKUN-LIN, Kunming Institute of Physics, Kunming. China (p. 463) M. D. VAIDYA,Bharat Electronics Limited, NDA Road, Pashan, Pune 411021, India (P. 403) L. K. VAN GEEST, B.V. Devt Electronische Producten, Dwauiewegen 2, 9300 AB Roden, Holland ( p . 93) C. F. VAN HUYSSTEEN,National Physical Research Laboratory, CSIR, Pretoria ooO1. South Africa (p. 393) A. WALKER, South African Astronomical Observatory, Cape Town, South Africa (P. 185) D.WALKER,Department of Physics and Astronomy, University College London, London WCIE 6BT, England (p. 185) G . A. H. WALKER,Department of Geophysics and Astronomy, University of British Columbia, Vancouver, British Columbia V6T I W5. Canada (p. 213) D. WASHINGTON, Philips Research Laboratories, Redhill, Surrey RHI 5HA, England (P. 101)
D. WEISTROP,Laboratoryfor Astronomy and Solar Physics, Goddard Space Flight Center, Greenbelt, Maryland 20771, U.S.A. (p. 133) A. A. WELLS,Department of Physics, University of Leicester, Leicester LEI 7RH, England (P. 497)
J . WESTLAKE,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 t B Z , England (p. 531)
xvi
CONTRIBUTORS
J. T. WILLIAMS,Laboratory for Astronomy and Solar Physics, Goddard Space Night Center, Greenbelt, Maryland 20771, U.S.A. (p. 133) G. WLBRICK,Observatoire de Paris, 92190 Meudon, France (p. 11) W. M. WREATHALL, English Electric Valve Company Limited, Waterhouse Lane, Chelmsford, Essex CMl 2QU, England (p. 425) WU QUAN-DE,Department of Radio-Electronics, Peking University, Beijing, China ( p . 373) XIMENJI-YE, Department of Radio-Electronics, Peking University, Beijing, China ( p . 561) K. YAMAMOTO.Hamamatsu Photonics K.K.. 1126 Ichino-Cho. Hamamatsu 435, Japan (P. 323) S.YANG,Department of Geophysics and Astronomy, University of British Columbia, Vancouver, British Columbia V6T 1W5, Canada ( p . 213) YANG XIAOWEN,Changchun Institute of Optics and Fine Mechanics, Changchun, China
(P. 541) 0. YOSHIDA, Toshiba Research and Development Center, Toshiba Corporation, Kawasaki, Kanagawa 210, Japan ( p . 275) T.YOSHINO, Electron Devices Engineering Laboratory, Toshiba Corporation, Yokohama, Japan (p. 275) YOU ZHONG-QIANG, Hebei Semiconductor Research Institute, Shijiazhuang, Hebei, China (p. 285) ZHANG XIAOQIU,Xian Institute of Optics and Precision Mechanics, Academia Sinica, Xian, Shaanxi, China (p. 385) Z n o u LI-WEI, Department of Optical Engineering, Beijing Institute of Technology, Beijing, China (pp. 561 & 575)
PREFACE The Eighth Symposium on Photo-ElectronicImage Devices was held at Imperial College, University of London, from September 5 to 7, 1983. On this occasion the symposium was organized in cooperation with S.P.1.E.-the International Society for Optical Engineering, and was one of three conferences held in the same week at Imperial College. As for the seven previous symposia, the proceedings are here published as volumes in the series, Advances in Electronics and Electron Physics. I would like to express my gratitude to Dr. P. W. Hawkes and Academic Press for making this possible. After each of the earlier symposia the organizers considered the alternatives of publishing the proceedings in this hardbacked form or publishing it by photocopying the authors’ typescripts and off-set printing. The former has the disadvantage of a longer elapsed time between the symposium and the publication of its proceedings, but enables the contents to be presented in a uniform style with consistent units and notation. Moreover, the organizers have often been told that the series of hard-backed volumes provides a comprehensive and permanent reference work for this field. To control production costs some compromises had to be made compared to earlier proceedings, for instance, it was no longer possible to redraw a majority of the diagrams; however, it is hoped that this volume will maintain the standard of the series. The symposium was opened by Professor J. D. McGee and it is a great pleasure to record here my thanks to him. Professor McGee started the Symposia on Photo-Electronic Image Devices in 1958 and has taken an active role in every one since then. I hope and expect to see him at the Ninth Symposium. Finally, I would like to thank the Organizing Committee, Professor Jim Ring, Dr. Harold Ables, and Dr. Geoffrey Towler for planning the program. I would also like to thank all those members of the Astrophysics Group of Imperial College who gave their cheerful and unstinting help in running the symposium.
February 1985
B. L. MORGAN
xvii
ABBREVIATIONS For the most part the Editors have tried to keep to the units and terminology currently accepted and to adopt consistent abbreviations following Systkme Internationale usage wherever possible. References cite journals abbreviated as recommended in Science Abstracts. Citation of earlier Symposia is frequent and the Editors have sought to simplify by the use of “Adv. E.E.P.” for “Advances in Electronics and Electron Physics” followed by the appropriate volume and page numbers.
xviii
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS,VOL. MB
Detection Efficiencies of Far-Ultraviolet Photon-Counting Detectors G.R. CARRUTHERS and C. B. OPAL E . 0 . Hulburt Center f o r Space Research, Naval Research Laboratory, Washington D.C., U.S.A.
INTRODUCTION The sensitivity of most astronomical instruments operating in the optical and ultraviolet spectral ranges is maximized by the use of detectors having (1) the highest possible detective quantum efficiency, and (2) the capability to detect and count individual photon events (photon-counting mode). The theoretical limit on signal-to-noise ratio at low light levels is set by the statistical fluctuation in the rate of arrival of photons at each picture element, or pixel, of a two-dimensional (imaging) detector. However, the use of photon-counting techniques has the added advantage, in comparison to analog integration techniques, of allowing the rejection of very large events (due to ions or cosmic ray interactions in the detector) or very small events (due to readout noise and other electronic noise in the detector). The main objective of our present work is to quantitatively compare the photon-counting detection efficiencies of several detector approaches suitable for use in the far- and extreme-ultraviolet spectral ranges. Applications of these detectors include second-generation instruments for the Space Telescope, as well as instruments for the proposed Far Ultraviolet Spectrographic Explorer (FUSE). This work is supported by NASA's Space Astronomy Ultraviolet Detector Development Program (SAUDDP),with additional support in the CCD detector area by NASA's Planetary Instrument Definition Program (PIDP). It is being carried out jointly by our group at the Naval Research Laboratory and by a group at the Princeton University Observatory which includes Edward Jenkins, John Lowrance, Paul Zucchino, Michael Reale, and John Opperman. It also includes collaboration with J. Gethyn Timothy of the University of Colorado. It has been known for some time*-4that, in the ultraviolet spectral range 299 Copyrighl 0 1985 by Academic Press, Inc. (London) Ltd. All rights of reproduction in any form rexrved. ISBN 0-12-0147246
300
G . R. CARRUTHERS A N D C. B. O P A L
below 2000 A, opaque alkali halide photocathodes provide very high photoemissive quantum yields (see Fig. 1). These provide, typically, a factor of 5 higher quantum efficiency than the more commonly used semitransparent photocathodes; in addition, there is no short-wavelength cutoff imposed by the semitransparent photocathode substrate. Therefore, our current investigation is limited to detectors utilizing opaque alkali halide photocathodes. In principle, a microchannel plate (MCP) detector which has an opaque alkali halide photocathode deposited on its front surface should have nearly the same detection efficiency as one which uses a separate, opaque photocathode and which uses the MCP only to amplify the photoelectrons incident on it from the separate photocathode. However, the actual results obtained by various investigators differ widely, with most reporting substantially lower photon-counting detection efficiencies than expected with MCPs having photocathodes deposited on their front surfaces. Therefore, a more thorough and quantitative comparison of the two types of MCP detectors was a major aim of our investigation.
1 0 0 1 . .
.
. . . .
.
\
\KEr
r S . T ~CSI S.T.
\
\
I
1200
1400
1600
x
I
1
.
Csl
\
\ IS00
2000
tx,
FIG.1. Quantum efficiency versus wavelength in the vacuum ultraviolet for opaque and semitransparent alkali halide photocathodes.
FAR-UV PHOTON-COUNTING
DETECTOR EFFICIENCY
30 I
MICROCHANNEL PLATEDETECTORS The configurations shown in Fig. 2 were used for comparison of MCP detectors; these included MCPs operated uncoated, with a separate opaque photocathode, and with a photocathode deposited on the front face of the MCP. The folding mirrors used with two of the detector configurations shown in Fig. 2 were both coated with Al + MgF2 in the same evaporation, and so should have identical reflectance versus wavelength. However, a measure of this reflectance could be obtained using the alternate, direct-view configurations shown. The direct-view arrangement was necessary to compare the quantum efficiency of the uncoated MCP used with the opaque photocathode with that of the CsI-coated MCP. The folding mirror arrangements were more convenient for comparing the coated MCP with the opaque photocathode + MCP combination. The MCPs were Galileo 25-mm-diameter active area dual-plate (Chevron) arrays, with single metal collecting anodes. The photoevent pulses were detected and counted using a Photochemical Research Associates photon-counting system. A difficulty with the Chevron-configuration microchannel plates was that we were unable to find a “plateau” in the count rate versus voltage curve, even though the plates had been vacuum baked at 350°C for 72 hr prior to installation in the testing vacuum system. However, all of the plates used had very similar curves, with counting thresholds at nearly the same voltages. We typically operated at 2200 V (400 V above threshold), at which the count rate measured was the same as for a single, helical channel electron multiplier (Galileo 4039) for the same intercepted photon flux. Although the absolute photon-counting efficiencies of the Chevron MCP detectors are thus uncertain by as much as a factor of two, the relative efficiencies of the various MCP configurations should be much more accurate (to within 220%). The detector assemblies were set up in a vacuum system equipped with a vacuum monochromator and collimating mirror so that the detectors could be illuminated with a parallel, monochromatic beam of selectable wavelength. The beam intensity was monitored during the tests using two separately calibrated “standard” detectors; these were a channel electron multiplier (Galileo model 4039) for the wavelength range below 1300 A, and a photomultiplier (EMR 541G) for the 1150- to 1800-A range. The apertures in the configurations shown in Fig. 2 were I .27 cm2,whereas the areas of the calibration comparison detectors (PMT and Channeltron) were taken to be 0.7 cm2. Thus, the area ratio of MCP detectors to comparison detectors was I .8 : 1. The MCP for use with integral photocathode, and the substrate for the opaque photocathode, were both coated with CsI (nominal thickness
U N
a
> I
0
: 2
>
0 P
w x
N
m
FAR-UV PHOTON-COUNTING DETECTOR
EFFICIENCY
303
A) in the same evaporation. This was to assure the most accurate possible intercomparisons. The coated MCP and the opaque photocathode were kept in dry nitrogen or dry air after removal from the deposition belljar and immediately transferred to the testing vacuum system with no exposure to ambient laboratory air. On any subsequent occasions during which the vacuum system had to be let up to atmospheric pressure, dry air of less than 5% humidity was used, and any work on the detectors was done in a dry air tent attached to the vacuum chamber door, with relative humidity never exceeding about 15%. Previous tests have shown that, at least with opaque CsI photocathodes, such exposure has little effect on the absolute quantum yield. However, one of the objectives of our investigation was to compare the relative effects of such exposure on Csl-coated MCPs and opaque Csl photocathodes, since (with windowless far-UV detectors) such exposure is nearly unavoidable in practice. We also wanted to assess the effects of long-term exposure to less-than-perfect vacuum environments on the relative and absolute efficiencies, since in many cases it is not possible to have continuous pumping of the detector vacuum housing. As part of our program, we have investigated the effect of accelerating voltage on the detection efficiency of the MCP with a separate, opaque photocathode. We also investigated the effect of a variable retarding (negative) voltage (applied to the aperture plate or folding mirror holder) on detection efficiency, for those configurations in Fig. 2 in which initial photon detection is at the front surface of the MCP. The results we have obtained appear consistent with previous measurement^.^.^ In particular, maximum yield with a separate, opaque photocathode is reached at about 4 kV potential difference, with a gradual decrease toward higher voltages and a sharp decrease below 2 kV. We have not yet done a detailed and complete reduction of our measurements of relative quantum yield and its variation with wavelength. However, the following quick-look results are quite definite, and have been supported by several repeat measurements. (1) At a wavelength of 1216 A, the MCP detector with a separate, opaque CsI photocathode is about 50 times more efficient than a bare MCP, and about 5 times more efficient than a CsI-coated MCP. (The comparisons were made at equivalent operating voltages and MCP-front plate potential differences.) For 4000
FIG. 2. Microchannel plate detector configurations used in photon-counting quantum efficiency measurements at NRL. Configurations 1A and 1B utilize a folding mirror or, alternately, light directly incident on the CsI-coated MCP. Configuration 2A uses a folding mirror with a separate, opaque CsI photocathode, whereas 2B exposes the same, uncoated MCP to directly incident light.
304
G . R. CARRUTHERS A N D C. B. OPAL
equal photon fluxes, the bare MCP count rates were similar to those of a comparison Channeltron detector, having a measured quantum efficiency of 1.75 k 0.25% at 1216 A. (2) The CsI-coated MCP shows a more rapid drop in quantum efficiency toward longer wavelengths than does the MCP with separate, opaque CsI photocathode. For example, the ratio of quantum yields is about a factor of 2.5 at 1026 8, (cf. 5 at 1216 A) but is a factor of 20 at 1614 A. (This result is in disagreement with the results of Martin and Bowyer.6)(3) The opaque photocathode-MCPquantum yield appears to be more stable with exposure to dry air and storage in partial vacuum than the CsI-coated MCP. (These measurements are still in progress.) It is worth considering the possibility that, although the CsI-coated MCP gave less than ideal detection efficiencies in our tests, the quantum efficiency might be considerably higher if the MCP is never exposed to air at atmospheric pressure following deposition of the CsI. The CsI-coated MCP has several important practical advantages over detectors with separate photocathodes and magnetic focusing, such as reduced size, weight, and complexity. Therefore, to investigate this possibility, we plan to borrow a multianode microchannel array (MAMA) detector from J. G. Timothy, which incorporates a curved channel MCP and is equipped with a vacuum gate valve in place of the usual input window. The detector and gate valve will be interfaced with our CsI deposition vacuum system; after deposition of CsI on the MCP, the detector will be closed off under uacuum and then transferred to our measurement vacuum system, where it will be similarly interfaced. An opaque CsI photocathode prepared in the same evaporation will be transferred (in air) for use with a separate, bare Chevron MCP. In addition, our usual standard detectors will be run for comparison. This test will also give us the opportunity to compare the curved-channel MCP used in the MAMA with the Chevron-type MCP. Previous development tests of the MAMA7.*indicate that the curved channel MCP has a much narrower pulse height distribution and better plateau characteristics than the Chevron, although its maximum gain is typically limited to lo6 or less compared to lo7 for the Chevron. Therefore, determination of a true photon-counting quantum efficiency should be much easier for the curved channel MCP than for the Chevron MCP. We plan to do these measurements (and continue the previously described ones) during late 1983 and early 1984. In addition, we hope to further improve the absolute and relative accuracy of the measurements by using the synchrotron storage ring light source (SURF-11) at the National Bureau of Standards. This latter source has an accurately known spectral distribution, but is variable over a very wide range of intensity. This will allow direct comparison of the unity-gain (photodiode mode)
FAR-UV PHOTON-COUNTING
DETECTOR EFFICIENCY
305
quantum yields of opaque and MCP-deposited photocathodes (measured at high light level) with their photon-counting quantum yields (measured at low light level). Since the SURF is a broad-band, continuum light source, an interference filter will be used to limit the wavelength range to a narrow region near 1216 A. ELECTRON-BOMBARDED CCD DETECTORS The Naval Research Laboratory and the Princeton University Observatory have been independently developing far-UV detectors based on the use of opaque alkali halide photocathodes with CCD arrays operated in electron-bombarded mode.’+’’ The NRL detectors are adaptations of our opaque-photocathode electrographic Schmidt and all-reflective-optics cameras3s4whereas the Princeton detectors utilize the oblique-focusing concept for coupling an opaque photocathode to an external optical sysThe NRL camera system electronics, developed under contract by Texas Instruments, is initially being used in analog integration mode, whereas the Princeton camera systems operate in photon-counting mode. Princeton is currently developing a sounding rocket payload, the Interstellar Medium Absorption Profile Spectrometer (IMAPS), which utilizes an oblique focus EBCCD detector system. In the present investigation, NRL and Princeton are jointly involved in the measurement of absolute photon-counting efficiencies of opaque photocathode plus EBCCD detector configurations, and comparing them with similar detectors utilizing MCPs. Tests are being run in a vacuum chamber at Princeton, using the Princeton photon-counting electronics, as well as in facilities at NRL using the Texas Instruments electronics. We are comparing two types of Texas Instruments virtual-phase, 327 x 490-pixel CCD arrays: (1) thinned, backside bombarded, and (2) unthinned, frontside bombarded. These in turn are being compared with RCA thinned, backside-bombarded 320 x 5 12-pixel arrays. Figure 3 is a diagram of an NRL EBCCD Schmidt camera, and Fig. 4 is a block diagram of the camera system with TI-provided camera electronics. This system presently operates in analog integration mode, which provides a capability to operate over a wide range of light intensities by variation of the integration time and/or the electron-bombardment gain (the latter is linear with accelerating voltage above the “dead layer” threshold). With an accelerating voltage of 20 kV, a typical gain of the order of 4000 is achieved, which allows ready detection (but not counting) of single-photon events. However, since there is no event discrimination, ion events and readout noise are significant contributors to the total signal. Figure 5 is a far-UV spectrum of a laboratory discharge light source,
t- DRIVER
PREAMPLIFIER FIG.
ELECTRONICS
/
FOCUSING MAGNET
I-
25 kV
3. Diagram of a Schmidt camera with opaque alkali halide photocathode and electron-bombarded CCD detector under development at NRL.
- REMOTE HEAD ASSEMBLV
- VACUUM
CHAMBER
--
-
-____ --_____CCD - VIDEO --
-
,
VIDEO
OUT1
-
-
I I I
I,
-
POWEl SUPPLIES
ANALOG
VIDEO PROCESSING MODULE
-
_ _-
ADC
-
I I
I
I
1
LOCAL
REMOT
+5
POP- I I COMPUTER
DATA READY
-
VERT ( I l l BLANKING
I
~
HORlZ ( X d BLANKING
~
I
READ LINE
PERIPHERALS
I I
--
s.
1
PIXEL
-I
.-
12 BIT PARALLEL VIDEO I OUT
VIDEO LOGIC MODULE
J
DATA DISPLAY
MONITOR
I C-JMD.
-
-
DATA STORAGE PERIPHERALS
_ I
c
1
1
1
v
FIG.4. Block diagram of the EBCCD camera electronics (provided by TI) and the image data handling and processing system (provided by NRL).
- 1243 N - 1304 0
- 1493 N
- 1744 N
FIG.5: Ultraviolet spectrum of a helium discharge light source, using the NRL/TI camera system with an objective grating in a vacuum test facility. The display was photographed from the XYZ monitor with analog video input.
FAR-UVPHOTON-COUNTING
DETECTOR EFFICIENCY
309
made using the EBCCD camera with an objective grating. This spectrum was photographed from the screen of the XYZ monitor in analog mode, but we are currently interfacing the camera system’s digital output to our PDP 11/44 computer and Ramtek image display system. This will allow quantitative evaluation and processing of the image data, including flat field corrections, background subtractions, and summing of separate frames. Figure 6 is a diagram of an oblique focus, opaque photocathode EBCCD detector similar to that being developed at Princeton. In the IMAPS application, a KBr photocathode is used with a cross-dispersed objective echelle grating optical system. The spectrograph operates in the 900-1200 A spectral range and provides a spectral resolution of 0.01 A, a factor of 5 better than the Princeton telescope and spectrometer which was on board the OAO-3 (Copernicus) satellite. Figure 7 is a diagram and photograph of the EBCCD Schmidt camera and Cerenkov light source which is being used at Princeton for intercomparison of electron-bombarded CCDs in photon-counting mode. The Schmidt camera has a 10-cm-diameter aperture and 15-cm focal length. The Cerenkov source is used because it is highly stable in intensity (the half-life of the strontium-90 beta source is 28 years) and is portable and simple to use. Its intensity is very low, but is more than adequate for use with photon-counting detectors. As with the SURF light source, the r FOCUSING
MAGNET
PREAMPLIFIER CCD ARRAY THERMOELECTRI
PHOTOCATHODE PHOTOELECTRONS
FIG.6. Oblique focus EBCCD detector system with opaque alkali halide photocathode, similar to that under development at Princeton University Observatory.
310
G . R. CARRUTHERS A N D C. B . O P A L CEAENKOV
COLLIMAIINO
LIGHT SOURCE
SCHMIDT
CAMERA
MIRROR
FIG. 7. Cerenkov light source and EBCCD camera assembly used in EBCCD tests at Princeton.
Cerenkov source gives a broad-band continuum. The intensity of this source has been measured using the comparison standard PMT, both broad-band and with a narrow-band interference filter centered near 1220 A. The overall detective quantum efficiency of the Schmidt camera can be estimated (to within 220%) as the product of the clear aperture area, transmission of the corrector plate, reflectance of the primary mirror, and quantum efficiency of the CsI photocathode (all of which can be measured independently as well as in combination). Of these quantities, only the quantum efficiency of the CsI photocathode (and its wavelength dependence) is likely to vary significantly with time. The Schmidt camera was kept purged with dry nitrogen between runs to prevent degradation of the CsI photocathode by atmospheric humidity. However, comparisons of various CCDs were also made as close together in time as possible to maximize the confidence in the relative efficiency measurements. The comparison of EBCCDs with MCPs can also be done using the same setup, simply by placing the MCP detector at the electron focus of the Schmidt camera instead of an EBCCD. As with the EBCCD detectors, the effects of photoelectron energy variation could be measured, but with the MCP detector could be extended to much lower energies (below 2 keV). Preliminary tests with MCPs have given unreliable results, in that large count rate time variations (using the Cerenkov light source) have been observed which are not yet understood. We plan to investigate this further in the coming year. Figure 8 is a photograph of a video monitor, and photographs of oscilloscope traces of single scan lines, from a test run at Princeton. In this test, the Cerenkov source was observed with the EBCCD Schmidt camera using a TI front-side virtual-phase CCD. An integration time of 7 sec was followed by an unshuttered readout which took 10 sec (because the read-
F A R - U V PHOTON-COUNTING DETECTOR EFFICIENCY
31 I
FIG. 8. (a) Image of a the Cerenkov light source, photographed from a video monitor, obtained using the setup shown in Fig. 7 at Princeton. The bright spots are ion events, and the image trailing is due to exposure of the CCD during frame readout. (b) and (c) Oscilloscope tracings of single CCD vertical scan lines through the Cerenkov source image. In (b) the gain is 5 times that in (c), and the amplitudes of single photoelectron events (outside of the central image) are readily discernible.
out rate was limited by the computer to 16,000 pixels sec-1). This results in the image trailing noticeable in Fig. 8. Also, ion events (due to residual gas in the vacuum system) give rise to the bright spots apparent in the image. The oscilloscope tracings show that the intensity in the center of the spot image corresponds to about 20 photoevents per pixel. However, as revealed by stray light events in the outskirts of the spot image, single photon events are readily detectable, and have relatively little spread in amplitude. This latter impression was quantitatively confirmed by measurements of the pulse amplitude distributions, particularly at the higher operating voltages (see Fig. 9). With the TI frontside device, it was possible to see photon events on the oscilloscope at operating voltages as low
312
G . R. CARRUTHERS A N D C. B . OPAL
T1 2118-1-18 8.5 kV
178PATCHES 6/01/83
....
..... ..... ......
...... ...... ....... ....... ....... ....... ....... ........ ........
......... .......... ........... ........... ............ ............ ............ ............ .............
..................... ............. ........................ ...............................
A
"20
0
"40
T1 2118-1-18
"60
"80
............................... ................ .............................................. ................................................................... . . "100
"120
"140
"160
"180
"200
"220
"240
20.0 kV 48 PATCHES 6/01/83
... .... .... .... .... .... ....
.... .... .... .... ....
.....
..... .....
....... ....... ....... ........ ..........
...... ............... ........ . . . . .............................................. .......................................................................... 40 "80 "120 "160 "200 " 2 4 0 "280 "320 "360 "400 ............
"
0
"440
"480
FIG.9. Histograms of photoelectron event amplitudes (pulse height distributions) for a TI frontside-bombarded virtual-phase CCD,obtained in the Schmidt camera tests at Princeton. Accelerating voltages were 8.5 kV (upper figure) and 20 kV (lower figure).
as 5 kV. With thinned, backside-bombarded CCDs, however, threshold voltages in the 8- to 10-kV range are more typical. Computer disk recordings were taken of a 16 x 16 pixel region of the CCD, at a very low light level appropriate for photon counting. This run
FAR-UV PHOTON-COUNTING
DETECTOR EFFICIENCY
3 13
utilized a 0.5-mCi Y3r source and a 5-cm2circular aperture in front of the Schmidt camera aperture. Even so, it was necessary to select the patch position in the outskirts of the spot image to keep the count rate down to an acceptable level. The purpose of this run was to determine the amplitude of single photoelectron events. Following this, a digitized record was made of the full-frame analog signal when the full camera aperture was used with a 5-mCi Y3r source. As noted in Fig. 8, this yields an integrated maximum intensity in the center of the spot image which is well within the full-well capacity of the CCD, although much too high for photon counting. Division of the analog signal (integrated over the image) by that corresponding to a single photoevent (Fig. 9) yields the total photoevents recorded. This, in turn, can be used to determine the detection efficiency of the CCD if the other system efficiencies and light source intensity are known (these reductions are currently in progress). An important motivation for the intercomparison of CCDs is to determine whether the frontside-bombarded virtual-phase CCD is a viable approach, not only from the standpoint of detective quantum efficiency, but also in regard to device stability and lifetime. Therefore, we plan tests in which the devices are operated at relatively high light levels for extended periods, to high accumulated primary electron dosages. Changes in detection efficiency, dark current, and noise level will be monitored. Preliminary measurements'* indicate that the virtual-phase CCD is much more resistant to damage by energetic electrons incident on the frontside than are other types of CCDs, such as the TI and RCA three-phase devices. Since the thinning of CCDs is a difficult technology, and still has a rather low yield of usable devices, frontside bombardment, if viable, is a much more practical approach to routine use in EBCCD detectors. In particular, it greatly increases the feasibility of large format (800 x 800 pixel and larger) EBCCD detectors. CONCLUSIONS This article is a progress report on an ongoing project, for which final results are not yet available. The results obtained to date, however, are consistent with the following conclusions. (1) The detective quantum efficiencies of microchannel plate detectors and EBCCD detectors with separate, opaque alkali halide photocathodes are probably consistent with expectations based on the unity gain photoemissive quantum efficiencies of these photocathodes. (2) The detective quantum efficiencies of microchannel plate detectors with CsI photocathodes deposited on the front faces of the MCPs appear to be significantly lower than expected based on the unity gain photoemissive quantum efficiencies of coated MCPs oper-
314
G . R. CARRUTHERS A N D C. B. OPAL
ated in diode mode, and/or are much more sensitive to degradation by exposure to air than are CsI photocathodes on solid substrates. (3) The frontside-bombarded virtual phase CCD appears to be a promising approach for two-dimensional, photon-counting imaging detectors. ACKNOWLEDGMENTS We acknowledge the assistance of, and many useful discussions with, our co-workers at Princeton. We also thank Dr. R. Daniel McGrath and others at the Texas Instruments Central Research Laboratories for providing, and helping with the development testing of, the virtual-phase CCDs and camera electronics. This work is supported by NASA’s Space Astronomy Ultraviolet Detector Development Program and Planetary Instrument Definition Program, and also by the Office of Naval Research.
REFERENCES 1. Duckett. S. W. and Metzger, P. H.. Phys. Rev. 137, A953 (1965). 2. Metzger. P. H., J . Phys. Chem. Solids 26, 1879 (1%5). 3. Carruthers, G. R., Appl. Opr. 8, 633 (1969). 4. Carruthers, G. R., I n “Adv. E. E. P.,” Vol. 338, p. 881 (1982). 5 . Schagen, P., I n “Advances in Image Pickup and Display,” ed. by B. Kazan, Vol. 1, p. IS. Academic Press, New York (1974). 6. Martin, C. and Bowyer, S., Appl. Opr. 21,4206 (1982). 7. Timothy, 3. G., Rev. Sci. Insfrum. 52, 1131 (1981). 8. Timothy, J. G. and Bybee, R. L., Proc. S.P.I.E. 279, 129 (1981). 9. Lowrance, J. L., Zucchino, P., Renda, G. and Long, D., I n “Adv. E. E. P.,” Vol. 52, p. 441 (1979). 10. Lowrance, J. L. and Carruthers, G. R.. Proc. S.P.I.E. 279, 123 (1981). 11. Carruthers, G. R. and Opal, C. B., AIAA Paper 83-106 (1983). 12. Everett, P., Hyneck, J., Zucchino, P. and Lowrance, J., Proc. S.P.I.E. 331, 151 (1982).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS,VOL. 648
High-Resolution and Large Size Wafer Microchannel Image Intensifier B. JEAN,t J. P. BOUTOT,t V. DUCHENOIS, and R. POLAERT Laboratoires d’Electronique et de Physique Appliqude, Limed Brdvannes. France
INTRODUCTION Owing to their high luminous gain and spatial resolution, microchannel plate (MCP) image intensifier tubes (IIT) are widely used in night vision equipment. Tubes with double proximity focusing (wafer type) have some clear advantages in comparison with the inverter types in regard to distortion-free imaging, light weight, and compactness.2 Tubes with a useful diameter of 18 mm are presently in production. IITs of such a design are particularly well suited for designing low-light level (LLL) film camera equipment. This application requires tubes with a large useful area (diameter > 40 mm) and high image quality. The realization of tubes with these characteristics requires high accuracy in the planicity and parallelism of the photocathode substrate and the MCP input face. This goal implies large technological efforts, mainly due to the poor planicity of large size high-resolution MCPs. The Laboratoires d’Electronique et de Physique Appliqude have designed and constructed a prototype of a compact tube (designated M 44) with a photocathode diameter of 44 mm to be used in LLL airborne reconnaissance equipment. DESIGNOF A LARGEAREACOMPACTIMAGEINTENSIFIER The main features of the M 44 tube design have been determined by carrying out a preliminary study with 18-mm-diameter wafer tubes. As shown in Fig. 1, these features are (1) a double proximity-focusing structure which offers the major requirements of the application (distortionfree image, no focusing adjustment, gating suitability, compactness, and light weight); (2) a high-resolution microchannel plate electron multiplier t Now with Hyperelec S.A., Brive, France. 315 Copyright 0 1985 by Academic Press, Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-0147244
316
8. JEAN E T A L .
9 0 m m diam
525 photocathode
Glass input window
\
Fiber optics output window
P47 (blue) Screen Microchannel plate'
4 4 m m diam
FIG. I . Schematic cross section of the flat double proximity-focused MCP image intensifier tube (M 44).
(channel diameter 12.5 pm) with 44 mm useful diameter (the largest plate usable at the time); and (3) a P * 47 luminescent screen (YzSiOs :Ce) deposited onto a fiber optic window. This has been selected for its very short persistence (decay time to 10% brightness -100 nsec) to prevent image lagging. Its blue spectral emission, centered on A = 400 nm, is well matched to the peak sensitivity of photographic film. The use of a fiber optic window allows us to obtain very good optical coupling between the screen and the film by directly pressing the latter onto the window. The design objective for limiting spatial resolution at 4% contrast was 30 Ip mm-I. A preliminary study carried out with small size tubes has shown that this performance can be obtained with the following geometrical and electrical conditions: (1) the input stage has a 200-pm gap and 200 V between photocathode and MCP input and (2) the output stage has a 1mm gap and 6000 V between the MCP output and the screen. When considering the setting up of a tube with a useful diameter as large as 40 to 50 mm, the difficulty is to obtain the required flatness (+20 pm) of the input electrodes, i.e., the photocathode substrate and the MCP input face, and accuracy in their positioning (parallelism). For the plane glass input window, the necessary flatness can be achieved by optical polishing and by using a window thick enough to limit the deformation produced by air pressure. A deformation of less than 5 p m is achieved at the window center for glass of thickness 12 mm. The standard manufacturing process for high-resolution (12.5-pm channel) MCPs does not give good enough flatness. They generally show a warping effect. LEP have devised quite original solutions to these problems for the M 44 tube.
317
MICROCHANNEL IMAGE INTENSIFIER
TUBETECHNOLOGY Flatness measurements made at different steps of the manufacturing process of an MCP with 54 mm overall diameter have shown that the most important geometrical deformation takes place during the reduction process. Various processes have been tested with the aim of minimizing the deformation. Putting the channel plate under some tensile stress resulted in plates with a flatness better than +20 p m . Furthermore, by sealing the input face plate with indium to a rigid metallic rim (Fe-Ni-Co), any deformation during the baking of the tube is avoided. The rim is also used as one of the electrical connections to the channel plate. The other connection is made by gold wires with a small diameter (0.33 pm) bonded by thermocompression onto the edge of the MCP output face. The technique is similar to that used in semiconductor chips. As for any wafer tube, processing of the M 44 has to be done by the well-known transfer method. The solution selected for making the final sealing is that now commonly used for manufacturing the second-generation 18-mm wafer tubes. It consists ofjoining the input window to the tube body by a hot indium alloy. This technique, used for sealing at the same time the input and the output windows, has made tube construction markedly simpler than previous techniques3 The tube has a very flat and rugged all-glass structure which allows the required spacing between the tube elements to be obtained with high accuracy. The size of the tube envelope has been chosen to allow for the use of MCPs with a useful diameter of up to 60 mm. An unpotted M 44 tube is shown in Fig. 2. CHARACTERISTICS AND PERFORMANCE A few prototype tubes have been made with the following general characteristics: the photocathode is an S * 25 of useful diameter 44 mm on a plane glass window. The MCP has 12.5-pm channels with a length-to,,A 400 nm and a diameter ratio of 60. The phosphor is a P 47 having decay time to 10% of order 100 nsec. It is situated on a fiber optic window. The magnification is 1 : 1 and the dimensions of the potted device are diameter 112 nm and length 28 mm. It weighs approximately 500 g. The following tentative laboratory results were measured with the typical voltages of 200 V at the input stage, 800 to 900 V on the MCP, and 6 kV on the output stage. The radiant power gain is defined as the ratio of radiated watts at the P-47 emission peak, per watt incident at the input at the peak of the spectral response of the photocathode (A = 550 nm). The dependence of the radiant power gain on the MCP voltage is given in Fig. 3. The MCP
-
B. JEAN ET A L .
318
FIG.2. Photograph of an unpotted M 44 image intensifier tube (envelope diameter 90 mm, thickness 26 mm).
electron gain is also shown in the same figure. The efficiency of the P * 47 screen used in conjunction with the 12-mm-thick fiber optic window has been evaluated at about 0.01, which is in good agreement with the results obtained for P - 47 screens deposited onto glass windows,? allowing for the fact that the light transmission of the fiber optic window at 400 nm is no more than 30%. Limiting spatial resolution at 4% contrast IS of order 27 to 30 Ip mm-I. The low distortion and good gain uniformity produced by the tube can be appreciated from Fig. 4. The response curve of an MCP image intensifier tube is mainly defined by the current transfer characteristics of the plate.4 The response curve of an M 44 tube operated at a radiant power gain GL = lo3 W W-1 is given in Fig. 5 . In the continuous operating mode, the response linearity is preserved as long as the mean output current Z, from the MCP does not t See p. I13 (Boutot et
a[.).
319
MICROCHANNEL IMAGE INTENSIFIER
lo4
o3 .-0c
0
-al c
0 0
-
1 c
8
g
f C
1
700
750
I
I
I
I
850 900 950 Channel plate voltage ( V l 800
0'
!OOO
FIG.3. Radiant power gain (GL)of the M 44 tube and MCP electron gain (Gcp) versus channel plate voltage.
exceed about one-tenth of the MCP conduction current I,, corresponding to a maximum photocathode illumination of about 3 X 10-3 lux. When operated in the pulsed mode, for instance, by applying a gating voltage in the input optics, the linearity limit is no longer defined by the conduction
320
B. JEAN ET A L .
FIG.4. Example of picture quality obtained with the M 44 tube.
current I , of the plate but by the maximum output charge Qs(max) which can be delivered by the MCP.- This maximum is about 5 x lo-'" C per C for the square centimeter of MCP operating area, i.e., about 7.5 x M 44 tube. This means that the response of the tube remains linear as long as this charge is not reached. The limit has been plotted in Fig. 5 (curve c) for an MCP gain of 500 (Le., for GL = lo3 W W-I). As an example, for a photocathode illumination of lux giving a peak current J P = ~ 3 X A, the linearity is preserved as long as the exposure time A f does not sec. It is also easy to define the limits of the operating exceed 5 X conditions when the tube is to be used in a recurrent pulsed mode. In that case, the linearity conditions for the continuous operating mode also has to be fulfilled. Thus if we take I , = 6 X lo-" A and Qs = 0. I Qs(max), the total current can be kept below 0.1 1, if the pulse frequency does not exceed 800 frames sec-I.
32 I
MICROCHANNEL IMAGE INTENSIFIER
1
10"
1
1
lo-2 Photocathode illuminance ( Lux 1
1
lo-'
FIG. 5. Response linearity of the M 44 tube: curve a: current transfer characteristics versus photocathode illumination in the continuous operating mode (continuous line) and in the pulse operating mode (dotted line); curve b: MCP conduction current at 880 V (GMCP = 5 0 0 ) ; curve c: limit of linear operating region in the pulse operating mode.
CONCLUSION The compactness of the proximity-focused intensifier tube allows the design of a novel camera for airborne reconnaissance and also for general purposes in physics. The main advantages of this new tube are a large useful format and very high spatial resolution giving impressive image quality reaching 90,000 picture elements per frame, direct optical coupling between the phosphor screen and film yielding the three advantages of efficient optical coupling, low distortion with no vignetting, and reduction of blooming due to the local saturation of the channel plate and the light guiding through the fiber optic window, and, finally, easy frame gating for medium speed cinematography.
322
B. JEAN ET A L . ACKNOWLEDGMENTS
This work was supported by ETCA (Establissement Central de 1’Armement) in a joint program with OMERA. The authors wish to thank Mr.Miquel from ETCA for his support throughout this work as well as J. Dietz, P. Marchal, A. Goutelle, and E. Desvaux for their assistance in preparing and measuring the tubes, and the Hyperelec development team for its efficient collaboration. Finally, the authors are grateful to C. Piaget, M. Fouassier, and J. C. Richard for their advice and comments in the redaction of the article.
REFERENCES 1. Acta Electron. 20, 3 (1977), 2. In “Image Intensifiers,” Technical Information 033. Philips Electronic Components and Materials (1977). 3. Graf, J., Fouassier, M., Polaert, R. and Savin, G., In “Adv. E.E.P.” Vol. 33A, p. 145 (1972). 4. Loty, C., Acra Elecrron. 14, 107 (1971). 5. Audier, M., Delmotte, J. C. and Boutot, J. P., Rev. Phys. Appl. 13, 188 (1978). 6. Eberhardt, E. H., IEEE Trans. Nucl. Sci. NS-28,172 (1981). 7. Fouassier, M., Rosier, J. C. and Dietz, J., Acra Electron. 20, 369 (1977). 8. Polaert, R. and Rodiere, J., Acra Electron. M , 379 (1977).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS,VOL.648
A Two-Dimensional Photon-Counting Tube M. KINOSHITA, K. KINOSHITA, K. YAMAMOTO, and Y. SUZUKI Hamamatsu Photonics K.K., Hamamatsu, Japan
INTRODUCTION A photomultiplier tube has been used as a detector for a photoelectroncounting device making use of its high quantum efficiency and large electron gain coupled with low dark counts. There are demands for a twodimensional photoelectron-counting imaging in many fields. At very low light levels a device should incorporate electron multiplication and signal output preserving the input spatial information. To meet these specifications, a two-dimensionalphotoelectron-counting tube has been developed which has an inverter-type image intensifier structure with a silicon position sensitive device (Si-PSD) instead of a phosphor screen. For applications in the visible wavelength region a multialkali photocathode was used. This article describes the two-dimensional photon-counting tube.
CONSTRUCTION AND OPERATION Figure I is a schematic diagram of the tube. The tube consists of an image section in which an optical image is converted into an electron image, a multiplier section, and a readout section. The maximum diameter of the tube is 65 mm and its total length is approximately 150 mm. The imaging section consists of a multialkali photocathode on a flat face plate, a mesh electrode for photoelectron acceleration, a focusing electrode, and an anode. The electron optics with a flat face plate was especially designed for an ultraviolet image intensifier for which a fiber face plate was not available. The photoelectrons are focused onto the input surface of a three-stage MCP and are thus multiplied by as much as 10’. The output electron crowds are injected into the Si-PSD in the readout section and are there multiplied approximately 100 times. A total electron gain of lo9 is therefore obtained in the tube. Information corresponding to the position of an incident photon on the photocathode is derived by an ordinary charge division method through 323 Copyright 0 1985 by Academic Press. Inc. (London) Lld. All rights of reproduction in any form reserved.
ISBN 0-12-0147244
324
M. KINOSHITA ET AL. Focusing Electrode
Mesh Electrode
I 3-Stage MCP
FIG. 1. Structure of the photon-counting tube.
two pairs of counter electrodes on the Si-PSD. By accumulating such information with an external device, a two-dimensional photoelectroncounting image can be obtained. The operating voltages of the electrodes are listed in Table I. CHARACTERISTICS Spectral Response and Dark Current
An S * 20 multialkali photocathode was chosen for the tube because of its high quantum efficiency and its wide spectral response. In the photocathode deposition process alkali metal vapor is introduced in the tube; if this is absorbed by the MCP and the Si-PSD the dark current will increase and the signal-to-noise ratio of the tube will deteriorate. In the worst case photoelectron counting becomes impossible. Care was therefore taken that alkali metal vapor did not reach either the MCP or the Si-PSD. The luminous sensitivity of the tube is approximately 200 p A lm-1. Figure 2 shows the dark current characteristic of the tube and that of a tube of conventional design. The low dark current corresponds to that of a tube TABLEI Operating voltages of the device Photocathode Mesh electrode Focusing electrode Anode MCP input electrode MCP output electrode PSD output electrode PSD biasing electrode
-8.0 kV -6.5 kV -7.6 kV -5.0 kV -5.0 kV -3.0 kV
ov
+I5 V
A TWO-DIMENSIONAL PHOTON-COUNTING TUBE
325
Acceleration Voltage (kV)
FIG.2. Spectral response and quantum efficiency of the tube.
designed for ultraviolet photon counting in which activation with alkali metal vapor is not used. Characteristics of a Three-Stage MCP in Pulse Mode Operation
The MCP of the multiplier section has an effective area 20 mm in diameter. Its channel pitch and diameter are 15 and 12 pm, respectively. Its resistance is low at 200 Mi2 and is chosen to suppress saturation of the channels. The pulse height distribution and gain of the three-stage MCP are shown in Figs. 3 and 4, respectively. Figure 5 shows the counting rate capability of the multiplier section. At a gain of lo7, the maximum counting rate is approximately lo5 sec-I. The maximum counting rate can be improved by using MCPs with lower resistance or by using MCPs having a specially designed resistance distribution to minimize the potential increase near the output of the MCP.
Pulse Height
- FIG.3. Pulse height distribution of the three-stageMCP operated at 700 V per stage. Gain G is 2.5 x lo6, peak-to-valley ratio is 10.
M. KINOSHITA ET A L .
326
105
L1
w
600 Voltage per stage (V)
FIG.4. Gain of the three-stage MCP versus applied voltage.
Characteristics of the Si-PSD
Figure 6 is a schematic diagram of an Si-PSD and its equivalent circuit. Its sensitive area is 27 x 27 mm2 which covers the effective output area of the MCP. Electron bunches from the MCP are accelerated between the MCP and the Si-PSD and generate electron-hole pairs in the space charge region of the reverse biased Si-PSD after passing through a p+ layer. The p+ layer has a thickness of 300 nm and a uniform resistivity distribution. The Si-PSD is a distributed parameter circuit comprising a junction capacitance and a resistive layer which control its response time and resolution. Figure 7 shows the response time of the Si-PSD when stimulated by light pulse inputs. The pulse response time becomes shorter as the resistance between the electrodes becomes smaller and as the reverse bias voltage on the diode becomes higher. Conversely the higher the
--7oOv 29x106 c
I-
0
al Iz:
mv
1.3XlO'
'9OOV 4.4~107
t3
lo4
to5
Io6
Counting Rate (sec-')
FIG.5. Counting rate characteristics of the three-stage MCP.
A TWO-DIMENSIONAL PHOTON-COUNTING TUBE
327
A
b 5
FIG.6. Schematic diagram of the Si position sensitive device and its equivalent circuit: 1-4, output electrodes; 5, common electrode; R., surface resistance; P, power source; D, ideal diode; Cj, junction capacitance; Roh.shunt resistance.
resistance, the better is the resolution. An Si-PSD having a resistance of 5 k n between the electrodes was chosen as a compromise between resolution and response time. The output signal pulse rise time is 0.3 psec and its duration less than 5 psec at a bias voltage of 15 V. The leakage current is less than 6 pA at the applied reverse bias voltage of 40 V. The pulse
z
0.1
10 I Resistonce between Electrodes (Kn)
FIG.7. Pulse response time of the Si position sensitive device.
328
M. KlNOSHlTA ET A L .
[/I [, .-n
,
":,
.2 .4 .6 I
2
,
L
-
o
650v 700v
s
I
1
4 6 10
Acceleration Voltage (kV)
FIG.8. Electron multiplication in the Si position sensitive device.
height distribution of the pulses from the Si-PSD is similar to that of the input electron pulses. Figure 8 shows the electron multiplication characteristic of the Si-PSD. The dashed line in this figure shows the theoretical gain if an electron-hole pair is generated for each 3.6 eV of incident energy. Performance of the Two-Dimensional Photon-Counting Tube
Figure 9 shows the total electron gain of the tube for various MCP voltages. Figure 10 shows the pulse height distribution of the output signal pulses of the tube. The peak value corresponds to a gain of 1.5 x lo9 and the peak-to-valley ratio is 12, which enables reduction of noise and efficient detection. The maximum counting rate of lo5 sec-' is obtained at this electron gain. The dark count rate of the tube is approximately 240 sec-I, due not to the MCPs but to thermal emission from the photocathode. The dark count can be reduced by cooling the photocathode or by choosing a bialkali photocathode which has a lower dark count characteristic. Figure 11 shows an example of the photoelectron-counting image of a pattern taken by the tube combined with a photon-counting image aquisition system. The image is composed of 5 X lo7 photons, and the counting rate is 5000 sec-I. The average count per pixel is 20.6 and the maximum count is 1185. The spatial resolution is more than 10 Ip mm-I.
A TWO-DIMENSIONAL PHOTON-COUNTING TUBE
Id"
I
. . ., .
329
.VMCP, .. S,.(.:
c .-
0
(3
0
c
c 10'
.I
I
10
Acceleration Voltage (kV)
FIG.9. Total electron multiplication gain of the tube.
FIG. 10. Pulse height distribution of output signal pulses from the tube. Peak value corresponds to a gain of 1.5 x lo9.
330
M . KINOSHITA E T A L .
FIG.1 1 . Example of a photon-counting image; total number of photons: 5 x 107, counting rate 5000 sec-1.
CONCLUSION A two-dimensional photon-counting tube has been developed, which incorporates a three-stage MCP and an Si position sensitive device. Since this tube has large electron gain and requires only a simple readout circuit, it is expected to be useful in a number of fields. ACKNOWLEDGMENTS The authors would like to thank Mr. T. Hiruma, President of Hamamatsu Photonics K . K . , who gave us the opportunity to carry out these experiments, and Mr. Y . Tsuchiya for continuous support of this work.
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. 648
High Spatial and Temporal Resolution Imaging with a Resistive Anode Photon Counter C. FIRMANI, L. GUTIkRREZ, L. SALAS, E. RUfZ, and G . F. BISIACCHI Instituto de Astronomia, Universidad Nacional Authoma de Mexico, Mexico and
F. PARESCE Space Telescope Science Institute, Johns Hopkins University, Baltimore, Maryland, U.S.A.
INTRODUCTION The new detector, Mepsicron (microchannel electron position sensor with time resolution) has been developed in order to overcome some of the main limitations of the existing image photon counter devices.'.* The Mepsicron is an image photomultiplier sensor for high spatial and time resolution, working in a photon-counting regime. High count rates have been obtained especially for point source images. The design of the detector is extremely flexible and appears very promising for the development of a large area photon-counting detector. The Mepsicron has been designed for astronomical use for deep sky photometric pictures, high-resolution spectrophotometry with single or crossed dispersion spectrographs, long slit spectroscopic techniques, high time resolution pictures, the spectrophotometry of fast varying sources such as pulsars and for Fabry-PCrot interferometry. However, it is likely that the Mepsicron will be used outside the boundaries of astronomy. In fact the detector is an excellent sensor of X rays, ions, and electrons, and many applications appear promising in medicine, biochemistry, physics, space research, etc. THEMEPSICRON The first prototype of the Mepsicron has been manufactured for astronomical use. It has a 25-mm-diameter multialkali photocathode. Figure 1 is a schematic diagram of the detector. Three main parts are visible: a 33 I Copyright 0 1985 by Academic Press. Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-014724-6
332
C. FlRMANl E T A L . MICROCHANNEL PLATES
hu
f
OUARTZ FACEPLATE
FIG.1. The Mepsicron detector.
proximity-focused photocathode deposited on a quartz face plate; an assembly of five Varian high current 40 : 1 microchannel plates mounted in two stages with a V and Z configuration, respectively; and a distortionfree resistive anode. A photoelectron from the photocathode is accelerated toward the first microchannel plate which is covered by an ion barrier film. The overall probability for the photoelectron to trigger a cascade is about 60% essentially determined by the open area ratio and the film transmission. The geometry and the electric potentials of the microchannel plate assembly are optimized so as to generate an electron cloud exhibiting highly controlled homogeneity in both the total number of electrons and in their energy distribution. The total gain can be up to 108. A distortion-free resistive anode collects the electron cloud produced by the cascade and divides the charge into four position-dependent output pulses. A pulse position analyzer (PPA) then calculates the values of the x and y coordinates for each count. The spatial resolution is 40 and 52 p m (FWHM) in the red and in the blue, respectively; this spatial resolution is quite uniform over the entire sensitive area. An array of 1024 by 1024 pixels provides sampling of the image compatible with this resolution. Although the uncertainty of the photon arrival time is less than 0.2 psec, the time coordinate is recorded with a precision of 0.1 msec, which is sufficient for the majority of astronomical purposes. The dark current contributes approximately 50 counts sec-I over the entire sensitive area when the detector is cooled to -30°C. This corresponds to 1 count per pixel every 5 hr. The contribution of the dark current noise is quite low and makes the performance of this detector very close to the photocathode quantum noise limit. The dead time of the electronic image processing system is 2 psec,
HIGH SPATIAL A N D TEMPORAL RESOLUTION IMAGING
333
which introduces a maximum count rate of 5 x los counts sec-I. However, the count rate for point images is limited by the recovery time of the microchannel plates. Preliminary tests suggest that 50 count sec-' pixel-' can be considered the maximum count rate: more detailed experiments are necessary to establish the interdependence between pixels in this regime. The geometrical stability of the array has been carefully considered in terms of the electronic design and the upper limit of the geometrical shift over the entire frame is 1 pixel in several days. The dynamic range of the detector is quite wide because of the high permitted count rate for point images and the low contribution of the dark current. The ratio of these quantities, about 106, gives a rough estimate of the dynamic range. Figure 2 is a schematic diagram of the whole system. The Mepsicron pulses are amplified and decoded by the PPA. The x, y, and t coordinate of each count can be recorded during the exposure using a magnetic tape unit. A 2-Mbyte memory, arranged in a matrix of 1024 x 1024 elements of 16 bits, integrates the whole image, recording each count in the respective address. The integrated image is displayed in real time in an image color display. The digitized image can be handled at the telescope by a Nova 1200 computer. ASTRONOMICAL PERFORMANCE The Mepsicron has been coupled to the REOSC echelle spectrograph at the 2-m telescope of the Mexico National Observatory at Baja California. The dispersion obtained with this system is 0.14 8, pixel-' (blue) and 0.25 8, pixel-' (red). The resolution is 0.3 8, FWHM (blue) and 0.4 8, FWHM (red), compatible with a slit width of less than 2 arcsec. The blue spectrum of a 16-magnitude star is obtained in I hr with a signal-to-noise ratio -10. In Fig. 3 the red spectrum of the galaxy MK 201 taken with a 900 groove mm-1 echellette grating is shown. In the second order from the top the two sky lines [OI], A6300, and [OI], A6364, are seen. The width of these two lines is determined by the width of the slit, 10 arcsec, and not by the intrinsic resolution of the detector. H a and the two [NII] lines A6548 and A6583 are visible in the next order. The low density of the gas, less than lo3 cm-3, is suggested by the relative intensity of [SII] lines, A6717 and A673 1, in the fourth order from the top. To give an idea of the resolution of the system, the full-resolution image of the doublet [OII] AA3726-9, for the galaxy MK 35, is shown in Fig. 4. Although the line width is due to the intrinsic velocity field of the galaxy the two lines separated by 2.7 8, are completely resolved. Each small
334
C. FIRMANI ET A L .
MEPSICRON
Lp.-.+?-+-,
IMAGE CRT
IMAGE COLOR DISPLAY
I
PHOTON by PHOTON TAPE UNIT
I
2 M BYTES
FAST IMAGE MEMORY
CURSOR
r3
;3 COMPUTER IMAGE TAPE UNIT
MEMORY CONTROL BOARD
TERMINAL
C
3
I IMAGE PROCESSING LIBRARY I FIG.2. Block diagram of the overall system.
rectangle in Fig. 4 represents one pixel. Figure 5 shows the blue part of the spectrum of BL 3 11 observed with a 200 groove mm-I echelette grating. The first line visible at the top, both in the third and in the fourth order, is Ha. A strong P Cygny profile is evident in this line as in HP (fifth order from the top). The other strong emission lines of the object are He1 A4471 (fifth order), He1 A4921 (tenth order left), and He1 A5015 (tenth order right and eleventh order left). The thin line in the last order at the bottom is the [OI] sky line A5577.
HIGH SPATIAL A N D TEMPORAL RESOLUTION IMAGING
FIG.3. The red spectrum of the galaxy MK 201.
FIG.4. The [OII] AA3726.9 doublet of the galaxy MK 35.
335
336
C. FlRMANl ET A L .
FIG.5. A detail of the blue spectrum of BL-311.
REFERENCES 1. Firmani, C., Ruiz, E., Carlson, C., Lampton, M. and Paresce, F., Rev. Sci. Instrum. 53, 570 (1982).
2. Firmani, C., Gutierrez, L., Ruiz, E., M a s , L., Bisiacchi, G . F., Paresce, F., Carlson, C. and Lampton, M., Asrron. & Asrrophys. (in press) (1984).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. 646
Output Energy Distribution of a Microchannel Plate N. KOSHIDA, M. MIDORIKAWA, and Y. KIUCHI Depariment of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology, Koganei, Tokyo, Japan
INTRODUCTION The energy distribution of the output electrons (EDOE) from a microchannel plate (MCP) is very important in many applications. When an MCP is used as an image intensifier, for instance, the energy spread of the output electrons may have a considerable effect on spatial resolution. Very little has been known, however, about the EDOE of an MCP.' The present article at first summarizes the EDOE characteristics of a singlechannel electron multiplier (CEM) reported previously and then describes those of an MCP with new experimental results.
OUTPUTENERGY DISTRIBUTION OF A CEM Previously, we reported the EDOE characteristics of a CEM with various output end structure^.^-^ The EDOE obtained from a CEM with a conventional output electrode was broad in nature.2 The full width A E at half-maximum of the EDOE, extending up to several tens of eV, strongly depends on the channel length-to-diameter ratio, a,the applied voltage, VA, and the output current. The AE characteristics can be generalized by a set of universal curves in terms of the normalized field VAlaalong the channel. These results show that the EDOE is almost entirely determined by the potential distribution near the output end of the channel. On the basis of these results, two kinds of output electrode were proposed to reduce the unfavorable output energy ~ p r e a d These .~ are the internal and the external electrodes by which the fields along the channel wall near the output end are fixed at zero and a low value, respectively. The average potential gain of the output electrons can therefore be considerably suppressed. The values of AE were then decreased by an order of magnitude compared with those obtained from a CEM with a conventional output electrode. The possible reduction in gain associated with the 337 Copyright 0 1985 by Academic Press. Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12414724-6
N. KOSHIDA E r A L .
338
improvement of the EDOE was apparent in the internal output electrode. A further serious problem in this case was that a broad subpeak appeared in the EDOE curve, the position of the subpeak and its relative strength both depending on the operating conditions of the CEM. These characteristics are caused by the abrupt change in the channel wall potential near the electrode. In the case of the external output electrode, by contrast, the reduction in gain was much less and no subpeak was observed. This is because the channel wall potential is changing smoothly and slowly along the channel. The external output electrode is therefore more advantageous. Moreover, a simultaneous improvement of the EDOE and gain was made by combining the external output electrode scheme and deposition of a thin film of KCI within the channel near the output end.4
-
1.0
-
0.5
.
ul c
.-
J
+2
-v
w
I2
2
0-
E,’ 20
0
ELECTRON
40 ENERGY
VA=l.O(kV)
W
60 80 E (eV)
Ic(A)
dE(eV) ES(eV)
-4.Ox1O9 3.2 0.5
AE
.......
100
2.0X10-6 3.7
26.5 21.5
z
20 ELECTRON
40
60
80
100
ENERGY E (eV)
FIG. I . Typical examples of the EDOE of an MCP. (a) The EDOE curves in the unsaturated mode for two different applied voltages; (b) the EDOE curves in unsaturated and saturated operation modes for a constant applied voltage.
EDOE
339
OF A MICROCHANNEL PLATE
OUTPUTENERGY DISTRIBUTION OF AN MCP
A straight-channel MCP with a useful area 20 mm in diameter, a channel diameter d of 12 pm, a length-to-diameter ratio a of 40, and a bias angle 8 of 5" were used in the experiment. The average penetration depth of the output electrode into the microchannels was 0.9 d. The MCP was mounted in an ion-pumped UHV system whose operating pressure was about 5 x 10-7 Pa. The diameter of the incident electron beam was 3 mm. The EDOE curves were measured by an ac retarding potential method using a lock-in amplifier type LI-574A and hemispherical mesh and collector electrodes.2 The amplitude and frequency of a small ac voltage superimposed on the slowly varying collector voltage were 0.5 V,, and 130 Hz, respectively. The measurements were made in both unsaturated and saturated regions of MCP operation. Figure la shows typical EDOE curves measured in the unsaturated region for two different applied voltages. The EDOE consists of two distinct components: a sharp main peak with a full width at half-maximum of a few eV and a broad subpeak extending over a wide energy range. The subpeak shifts in the higher energy direction with increasing applied voltage, while the main peak remains almost constant in energy and in shape. Under a constant applied voltage, on the other hand, the subpeak shifts in the direction of lower energy with increasing output current as is shown in Fig. Ib. The subpeak energy E, for two applied voltages is shown in Fig. 2 as a function of output current. The dotted line in Fig. 2 shows the onset of the saturated operation mode. As the consequences of such changes of
--
0.8 1.0
r Q
W
a m
3
v,
20
1 10-9
*
I
0
.
I
I
,
.
.
I
*
.
10-a lo-' OUTPUT CURRENT Ic ( A )
I
1
1
*
10-6
FIG. 2. The subpeak energy as a function of output current for two applied voltages. The dotted line shows the onset of the saturated mode.
N. KOSHIDA ET A L .
340 40 -
30 -
-s8
z *O
-
10 -
0'
'
'
. . I
10-9
.
.
'
'
OUTPUT
CURRENT
I,
'
*
10-6
10-7
(A)
FIG.3. The relative number of the output electrons with energies higher than 50 eV as a function of output current. The dotted line shows the onset of the saturated mode.
the subpeak, the relative number of high-energy ( 1 5 0 eV) output electrons N50 varies from about 40 to 10% with the individual parameters as shown in Fig. 3. The dotted line in Fig. 3 corresponds to that in Fig. 2. It should be noted that N50 changes considerably in the unsaturated region which is particularly important in imaging device applications. These results can be explained qualitatively from the potential distribution near the output end of the MCP. The output electrode of an MCP is formed by vacuum evaporation. The output end structure and the wall potential near the electrode are shown schematically in Fig. 4. The penetration of the output electrode into the microchannels produces a constant-potential region and a high-field region similar to those in a CEM with the internal output electrode. The abrupt change in potential between these two regions almost certainly causes the undesirable subpeak in the EDOE. The output electrons emitted from the regions A and B in Fig. 4 contribute to the main peak and the subpeak, respectively. The behavior of the subpeak shown in Figs. 1 and 2 is therefore due to the fact that the channel wall potential in region B varies with the operating conditions of the MCP. At higher applied voltage, the change in potential near the electrode becomes more abrupt and the field in region B is greatly increased. As a result, the subpeak becomes more apparent. As the output current is increased under a constant applied voltage, the field near the electrode is reduced because of the appearance of current flowing against the standing current to supply the net secondary electron emission. The subpeak therefore shifts in the direction of lower energy. In the
EDOE
OF A MICROCHANNEL PLATE
34 I
MCP
ELECTRON ENERGY POTENTIAL
FIG.4. Schematic representation of the output electrode of an MCP, the channel wall potential near the output end, and the form of the EDOE.
limit, a constant field is established near the electrode. Hence the subpeak energy becomes almost constant at very high output current as shown in Fig. 2.
CONCLUSION The fundamental EDOE characteristics of an MCP were measured. This study provides direct information on the output energy spread which is very important for various applications of an MCP, especially for imaging devices. It has been found that the EDOE of an MCP behaves very similarly to that of a CEM with an internal output electrode. This means that the MCP gain is greatly affected by the penetration depth of the electrode into the microchannels (as in a CEM). The EDOE consists of a sharp main peak and an undesirable subpeak extending over a wide energy range. Although the former remains almost fixed, the latter changes in energy and in shape according to the operating conditions of the MCP. This causes a considerable change in the relative number of high-energy output electrons. It would be desirable to obtain a narrower EDOE without the subpeak and without the reduction in gain. Further studies are necessary to satisfy these conditions. ACKNOWLEDGMENTS
The authors would like to thank Dr. K. Oba for his support and Dr. S. Yoshida for his continuing guidance and encouragement.
342
N. KOSHIDA ET A L .
REFERENCES I . Schagen, P., In “Advances in Image Pickup and Display” ed. by B. Kazan, Vol. I , p. 15. Academic Press, New York (1974). 2. Koshida, N . and Yoshida, S., Rev. Sci. Instrum. 50, 177 (1979). 3. Koshida, N., Kunii, M. and Yoshida, S . , Rev. Sci. Insrrum. 51, 365 (1980). I . Inst. Telev. Eng. J p n . 35, 752 4. Koshida, N., Suzuki, S., Kunii, M. and Yoshida, S., . (1981) (in Japanese).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS. VOL. 64B
Computer Analysis of the Temporal Properties of a Microchannel Plate Photomultiplier K.OBA and M.IT0 Hamamatsu Photonics
K.K..Hamamatsu, Japan
INTRODUCTION In photochemistry and photobiology, measurement of fluorescence life time has been recognized as a most useful tool in the investigation of, for example, the onset of photochemical reactions and the relaxation of excited states in molecules and living cells. The most advanced method used for fluorescence life time measurement is time-correlated single-photon counting, using a microchannel plate photomultiplier tube (MCP-PMT) and a laser to produce picosecond light pulses at a high repetition rate. * In this method, a sample is excited by the laser pulses and the fluorescence light corresponding to each pulse is detected by the MCP-PMT operated in single-photon-countingmode. The interval between excitation and the detection of each photon is accumulated until a statistically reasonable decay curve is obtained. The most important characteristic for this application is the transit time spread (TTS) of the detector. Only the MCP-PMT can satisfy this requirement among high-speed PMTs currently available. The MCP-PMT also satisfies other important requirements such as high gain, low dark count, and wide dynamic range. In this article, a computer analysis of the MCP-PMT based on a Monte Carlo simulation is described, emphasizing the temporal characteristics. The calculated results are compared with experimental results obtained in a nonproximity-focusedand a proximity-focused MCP-PMT. The calculation showed that the transit time spread due to the electron multiplication process inside the MCP is approximately 60 psec. On the other hand, the TTSs between photocathode and MCP are 3 psec in the proximity type and 70 to 35 psec in the nonproximity type, depending on the acceleration voltage. Total TTSs for these types are then estimated to be about 60 psec in the proximity type and 92 to 70 psec in the nonproximity types. These results were compared with experimental results obtained in the nonprox343 Copyright 0 1985 by Academic Press. Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12914724-6
344
K . OBA AND M. IT0
imity MCP-PMTs types R1294 and R1645 and the proximity type R1564, showing good agreement. The analysis showed that the TTS of the MCP depends strongly on the emission coefficient, 61, of the first collision of the photoelectrons in the MCP. The results showed that an increase in the value of 6, from 1 to 10 improved the TTS from 60 to 40 psec. The dynamic range in the time domain is another important characteristic and it was studied experimentally in these three tubes mentioned above. ANALYSIS OF THE TEMPORAL PROPERTIES O F THE MCP-PMT The TTS of an MCP-PMT arises in the transit between the photocathode and the MCP and in the multiplication process inside the MCP. The former is mainly due to initial energy and angular distributions of the photoelectrons. The shape of the potential distribution also contributes to the TTS in general, but the effect is small in this case. The second source of TTS is due to the initial energy and angular distributions of the secondary electrons and to the repetition of the collisions. In the following analysis, both nonproximity type and proximity type MCP-PMTs are studied. Figure 1 shows cross-sectional views of the two types. Tube types R1294 and R1645 have a focusing electrode between the photocathode and the MCP. Usually, a tandem type or so-called chevron type MCP is used to obtain gain higher than lo6. Output electrons are collected by the anode disk at the bottom. The difference between the R1294 and the R1645 is an A1 film on the entrance of the MCP in the R1645. This film prevents secondary electrons emitted from the top surface of the MCP entering neighborhood channels and thus starting an avalanche. Photoelectrons are accelerated to 980 V before hitting the MCP in type R1645 in order to penetrate the Al film. The acceleration voltage in the R1294 is 490 V . The type R1564 uses proximity focusing between the photocathode and the MCP. Other parts of the tube are similar to those of the R1645. TTS BETWEEN
THE
PHOTOCATHODE AND
THE
MCP
The TTS of this region was obtained by a Monte Carlo simulation assuming that photoelectrons have a gaussian distribution of initial energy and a cosine distribution of emission angle. The mean energy and the standard deviation of the gaussian distribution were determined to be
COMPUTER ANALYSIS OF T H E
MCP-PMT
345
FIG.I . Cross-sectional views of (a) nonproximity MCP-PMTs types R1294 and R1645 and (b) proximity MCP-PMT type R1564.
1 and 0.3 eV, respectively, for an S * 20 photocathode. For the R1294 and R1645 types, the effect of the focusing electrode was taken into account in calculating the photoelectron trajectory, and it was assumed that the probability of photoelectron emission was uniform over the useful area of the photocathode. The calculated results are shown in Fig. 2. Acceleration voltages are determined according to the specifications and are 490 V in the R1294,980 V in the R1645, and 950 V in the R1564. The mean transit time of photoelectrons is 2585 psec in the R1294, 1845 psec in the R1645, and 160 psec in the R1564. The FWHM of the TTS is 70 psec in the R1294, 35 psec in the R1645, and 3 psec in the R1564. The distributions shown in the figure were constructed with 1000 photoelectrons.
VPC-MCP = 490V d = 2 0 mm Tpc-MCP = 70 psec U W
m 5
3
z W 5.
5 W
a: TIME (psec)
v, t-
1.0
VPC-MCP = 98ov d=20mm T p c - ~ c p = 35 psec
z
3
0 0
4
w
a:
0
20
40
60
L 80 100
I 1 1
140
120
160
TIME (psec)
1
1 .o
C VpC-MCP = 950 V d = 1.5 mm Tpc-McP = 3 psec
2 3
0 0
a: W
m 5
3
z
w
> -
+
4
W
a:
L 0
-
10 TIME (psec)
-
I
12
14
16
FIG.2. Calculated values of TTS between photocathode and MCP (a) for nonproximityfocused type without Al film, R1294, (b) for nonproximity-focused type with Al film, R1645, and (c) for proximity-focused type with Al film, R15W.
COMPUTER ANALYSIS OF THE
MCP-PMT
347
'TTsIN THE MCP The MCP was modeled as two-dimensional parallel plates having 12p m spacing and a length of 480 pm, giving a normalized length of 40. The plate voltage was 800 V. In calculating the trajectory and the multiplication factor, the following assumptions were made. 1. The effects of space charge, charging of the emitting surface, and reflection of incident electrons are not considered. 2. The number of secondary electrons is determined by a Poisson distribution having average value S(8i, Vi) given by Eq. ( 1):2 6(Oi, Vi) = 4Vi6m(ei)/{Vm(ei)[l
+ Vi/Vm(Oi)Iz}
(1)
where Bi and Vi are the incident angle and energy of the electron hitting the wall, S,(Oi) is the maximum value of the yield, and V,(ei) is the incident energy at which the maximum yield S,(ei) is obtained. 3. The dependence of S,(ei) and V,(ei) on the incident angle is given by S,(&) =
exp[a(l
- cos ei)]
v,(ei) = v m o / a
(2) (3)
where 6,o is the maximum yield corresponding to the electrons with incident angle normal to the surface of the channel wall, Vmo is the incident energy which gives maximum yield, and (Y is the material constant. 4. Secondary electrons have a Maxwellian energy distribution and a cosine angular distribution. 5. A single secondary electron is emitted initially at the entrance of the channel. 6. The number of secondary electrons 61 is one. The values of ,,a, V m o , and (Y were determined so that the gain of the MCP operating at 800 V becomes about 500. This was done by changing ,,a between 2 and 4, V,, between 200 and 300, and (Y between 0.4 and 0.6. Finally, the values 6,, = 4.0, V,, = 250 V, and (Y = 0.5 were chosen. The average energy of the secondary electron, which was found to be not critical in the gain calculation, was assumed to be 8.0 eV. The calculation then proceeds as follows: first, the yield 6(8i, Vi) of each electron is calculated from Eqs. (l), (2), and (3) using the values Vi and 8i obtained from trajectory analysis. The number of secondary electrons emitted in the collision is then obtained from a random number based on the Poisson distribution with the average value 6(Bi, Vi). In the next step, the initial energy and emission angle of these secondary electrons are decided by random numbers produced by Maxwellian and cosine distributions, respectively. Trajectory analysis of each electron
K. OBA A N D M. IT0
348
based on these data gives values of Vi and 8i for each secondary electron in the next step. These calculations are repeated for every collision of every electron until all the electrons are outside of the channel. The important point of this simulation is that the total transit time of each electron participating in the process is derived by recording the transit time of each flight between collisions. In this simulation, 2000 such trials were repeated to obtain the overall transit time distribution or TTS. As is well known, the transit time spread is generated mainly in the initial stage of the multiplication process where the number of electrons involved is still small, say less than 100. Once the number becomes larger, the transit time spread does not increase. The data obtained here may therefore be applied to the case of a tandem MCP where the gain is 105 to 106 even though the TTS is obtained at an electron gain of 500 in this analysis. One of the output waveforms obtained in this simulation is shown in Fig. 3. The transit time is 190 psec and FWHM is 55 psec. Two thousand trials gave the pulse height distribution shown in Fig. 4, which is an exponential distribution with a mean value of 490. Since a single electron was used as the start of the multiplication process in this case (6, = I), some electrons failed to initiate an avalanche because of the probability of producing zero secondary electrons at the collision. In this case, 1150 output pulses were counted out of an initial 2000 electrons, giving a detection efficiency of 57.7%. Figure 5 shows the transit time distribution; the FWHM is about 60 psec.
Transit time : 190 (psec) F W H M : 55 (psec)
1 1 . 1
z
3 Z
OfJ
W
: 4 W K
0
40
80
120
160
200
240
280
TIME (psec)
FIG.3. Temporal distribution of output electrons from MCP,6,= 1.
320
COMPUTER ANALYSIS OF THE
MCP-PMT
VVCP
349
: 800 (V)
Gain : 490
Y
0
Y "
20
L
U'
40
n 60
80
loo
120
140
160
CHANNEL NUMBER
FIG.4. Pulse height distribution of output pulses from MCP, 6, = 1.
The total transit time spread of the MCP-PMT is then estimated by following relation: = '6C-MCP
(4)
+M ' CP
for the nonproximity type. In the R1645, the TTS becomes 70 psec. In the proximity type, T p c - ~ c pis negligible compared with TMCP and the TTS becomes T = 60 psec.
: 800 (v) THCP: 60 (psec)
VHCP
z
i3
0 LL
0
n
$
05
5
2
z W
2 + a -I
w
a: 0
40
80
120
160
200
240
TIME (psec)
FIG.5. Transit time spread of the MCP, 61 = 1.
280
320
350
K. OBA A N D M. IT0 1.0,
a
I
8a W
m
52 2 0.5-
VMCP
800 (v)
~ Y C P : 45
(psec)
ZO
U
0
VUCP : 800 (V) TUCP : 40 (psec)
a
w
m Iv)
2g
0 5 ~
wo PO
I-
4
W
a 0
40
Bo
120
160
200
240
280
320
TIME (psec)
FIG.6. Transit time spreads of the MCP (a) for 61 = 5 and (b) for 6, = 10.
100 VUCP
v)
I-
800 (v)
Gain : 7000
Z
2 0 8 10
m
I 3 z
6 W
W
U
20
40
sb
100
140
CHANNEL NUMBER
FIG.7. Output pulse height distribution of MCP for 6,
= 10.
160
COMPUTER ANALYSIS OF THE
MCP-PMT
35 I
As is well known, the TTS may be improved by increasing the number of initial secondary electrons 6 , . Figure 6 shows the results obtained for 6, = 5 and 6, = 10. The FWHM are 45 and 40 psec, respectively. A drastic change appears in the pulse height distribution when 61 is increased; as shown in Fig. 7, the distribution has a clear peak. This phenomenon, together with space charge effects, helps to produce a peaked pulse height distribution in higher gain operation.
EXPERIMENTAL RESULTS The three types of MCP-PMT shown in Fig. 1 were tested using the TTS measuring system shown in Fig. 8. The samples were illuminated with short light pulses (FWHM = 30 psec) at 400 nm from a laser diode followed by a second harmonic generator (SHG). The output pulse of the MCP-PMT is sent to a CFD (ORTEC 583) via a preamplifier (HP 8447F) and produces a stop pulse for the TAC (ORTEC 467). The switching pulse for the laser diode also generates a trigger pulse to start the TAC. The output of the TAC is then transferred to the multichannel analyzer to produce the transit time distribution. Experimental results are shown in Fig. 9. Figure 9a shows the TTS of type R1294; the FWHM is 100 psec. The dynamic range, defined as the amplitude between the top and the foot of the main peak, is a very impor-
MULTICHANNEL ANALYZER PHD) NAIG
FWHM = 30 PSEC
352
K. OBA A N D M. IT0 1
I
I
FW 1/2M = 1001psec) FW l / l O M = 183(psec) v)
z t
lo2
-
10'
-
8 w
I
I
I
a
VPC-MCP : 490 (V) VUCP : 800 (V) VMCP-A : 490 (V)
d
-
..
m
.:
4
.
Z
% I= 4
1
L+.\
loo-
%>:
.a w 10
-
.
* ,fa.&,.'.
.. _._..I
-4
'
%.f& \:: .,.,.
-3
I
I
-2
-1
I
1
0
+1
I
I
+2
+3
+4
b
FW 1/2M = 8l(psec) FW l/lOM = 164(psec) Vn-ucr : 880 (V) vvcr : 800 (v) vucr-r : 480 (V)
f
' I
..
-4
-3
-2
-1
0
+l
+2
+3
+4
5
TIME (nsec)
FIG. 9. Experimental transit time spreads (a) for nonproximity-focused type, R1294, (b) for nonproximity-focused type, R1645, and (c) for proximity-focused type, R1564.
tant characteristic because the curve is usually used to deconvolve the measured fluorescence decay curve to obtain the real decay curve. The dynamic range here covers two orders of magnitude. The subpeak 500
COMPUTER ANALYSIS OF THE
353
MCP-PMT
FW 1/2M = 61 (psec) FW 1/10M= 115(psec) V n - u u : 960 M
10'
v
lool--5
-4
. .- . .. . -3
-2
-1
. I
0
~ : 4m~ (v) -
... - ...---
---. +1
~
+2
+3
+4
. #
TIME (nsec)
FIG.9c. See legend on facing page.
psec after the main peak corresponds to secondary electrons emitted from the front surface of the MCP; these secondary electrons are emitted toward the photocathode and repelled back into neighboring channels and produce an avalanche. The apparent transit time of this pulse is about 500 psec longer than normal pulses. The effect can be reduced by applying a high positive voltage to the accelerating electrode with respect to the MCP input electrode, thus collecting all the secondary electrons emitted from the front of the MCP.4 Another solution is to cover the MCP with a thin A1 film; the secondary electrons producing this effect do not have enough energy to penetrate the A1 film. This method was applied to the type R1645 resulting in the smooth curve having a dynamic range of four orders of magnitude (Fig. 9b). Since the acceleration voltage used in the R1645 is twice that in the R1294, the FWHM of the TTS of the R1645 is 81 psec, shorter than for the R1294. The result for the type R1564 is shown in Fig. 9c. The curve has a dynamic range almost four orders of magnitude and the TTS has FWHM of 61 psec. The small subpeak 200 psec after the main peak is thought to be related to ion feedback effects inside the MCP. Careful examination of the subpeak reveals that it is related to large pulses at the tail of the pulse height distribution which are due to ion feedback inside the MCP.
354
K. OBA A N D M. IT0
CONCLUSION The transit time spread (TTS) of nonproximity-focused and proximityfocused MCP-PMTS has been studied by computer simulation. The results showed that the TTS of the nonproximity type without an Al film coating on the MCP is 92 psec, that of the nonproximity type with Al coating is 70 psec, and that of the proximity type is 60 psec. These results were compared with experimental results and showed good agreement. The study also showed that the TTS of the proximity type MCP-PMT can be improved to 40 psec by increasing the yield of the input surface of the MCP to a factor of 10. This could be realized by coating the MCP with a high yield material such as CsI, MgO, and so on. In the actual tube, the dynamic range of the transit time distribution is limited by factors such as ion feedback effects inside the MCP and secondary electrons emitted from the front surface of the MCP. The former may be solved by improving the vacuum and the latter by using an Al coating on the MCP. ACKNOWLEDGMENTS
The authors wish to thank Mr. H . Kume, Mr. N . Ohishi, and Mr. K. Koyama of HPK for continuous support in taking data and Mr. Y. Watase and Mr.K. Nakatsugawa of HPK for making the tubes.
REFERENCES 1. 2. 3. 4.
Murao, T . , Yamazaki, I. and Yoshihara, K., Appl. Opt. 21, 2297 (1982). Yakobson, A. M . , Rudiotekh. & Elektron. 11, 1813 (1966). Bruining, H., I n “Physics and Applications of Secondary Electron Emission” (1954). Wijnaendts van Resandt, R. W., Vogel, R. H. and Rrovencher, S. W., Rev. Sci. Instrum. 53, 1392 (1982).
ADVANCES IN ELECTRONICS AND E L E n R O N PHYSICS. VOL. 648
The MCP as a High-Energy Particle Track Detector K. OBA Hamamatsu Photonics K . K . , Hamamatsu, Japan P. REHAK
Brookhaven National Laboratory, Upton, New York, U.S.A.
and
D.M. POlTER R w e r s University, Piscataway, New Jersey, U.S.A.
INTRODUCTION Since the discovery of particles carrying the new flavor quantum numbers, charm and beauty, there has been a great interest in the development of high-energy particle track detectors with very high spatial resolution. The predicted lifetime of these particles requires spatial resolution of a few microns for direct lifetime measurement or for particle identification via an observation of the secondary vertex. The principle of a particle track detector consisting of microchannel plates (MCP) and phosphor screen is shown in Fig. 1. A high-energy charged particle passes through the MCP, crossing a large number of channel walls as shown in Fig. lb. Along the particle path, secondary electrons are emitted into some channels and generate avalanches, which produce a chain of spots corresponding to a projection of the particle trajectory on the phosphor screen. Several features of the MCP' make this scheme a suitable candidate for a high-resolution track detector.2They are a small channel diameter of 12 pm, very fast response time less than 1 n ~ e c and , ~ the capability for producing trigger signals for event selection. This article describes the selection of MCP parameters to optimize the performance of an MCP track detector. The performance of such a detector is also reported. 355 Copyright Q 1985 by Academic Press, Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-0147244
356
K. OBA ET A L .
HIGH ENERGY
CHARGED PARTICLE
H 140 mm P-1 I'PHOSPHOR
SCREEN
a
HIGH E N E R G Y CHARGED PARTICLE
MCP 1,
*
*
*\
PHOSPHOR SCREEN IMAGE ON SCREEN
b
FIG.1. (a) Cross section of high-energy particle track detector and (b) principle of operation.
CONSTRAINTS ON THE MCP TRACK DETECTOR Figure 2a shows the well-known chevron MCP configuration used for the particle track detector. This configuration can give gain higher than 105 sufficient to produce a bright spot on the screen corresponding to each avalanche. However, an electron cloud from a single channel of the first MCP spreads into more than seven channels of the second plate, resulting in a degradation of the position resolution and a broadening of the output pulse height distribution (PHD). To preserve the resolution and the narrow PHD, only one, thick MCP is preferable as shown in Fig. 2b. In this configuration, the MCP operates under space charge saturation resulting in a narrow distribution of spot brightness which is preferable for subsequent optical recording. The other parameter related to position resolution is spot size broadening due to the effects of space charge between the MCP and the screen.
MCP
AS HIGH-ENERGY PARTICLE TRACK DETECTOR
357
a
M
b
c
@-SPOT SIZE
FIG. 2. Output spot size in various MCP configurations. (a) chevron configuration of MCP; (b) thick MCP; (c) thick MCP + magnetic field.
Figure 3 shows a theoretical estimate of the broadening of the electron cloud as a function of gain. Above a gain of 104, the broadening of the spot size becomes serious. At a gain of 105 the spot diameter becomes about 160 pm. To overcome this effect, magnetic focusing is effective as shown in Fig. 2c. The focusing condition is obtained approximately by equating the time of flight of an electron from the MCP to the screen to the period of the cycloidal motion of the electron. This calculation shows that the MCP has to be operated in a magnetic field of about 6 kG to realize minimum spot size on the screen.
OPERATION OF THE SINGLE MCP IN A MAGNETICFIELD The magnetic focusing described in the previous section requires the MCP to operate in a high magnetic field. The performance of the MCP in a magnetic field was studied experimentally in order to optimize its bias angle. In the test, the MCP was 1.0 mm thick, had a 0" bias angle, and a channel diameter of 12 pm. Its peak gain and pulse resolution were measured as a function of magnetic field. The results are shown in Fig. 4. The figure shows that both the peak gain and the pulse resolution can be optimized at around 5 kG by choosing a bias angle of 15". When operated at high gain, the output pulse height distribution of the MCP shows a tail toward high gain which is a consequence of positive ion feedback. However, the tail becomes unimportant for magnetic fields
358
K. OBA ET AL.
Channel Dia 12pm Initial energy 30 eV
V s (kV1
Initial beam length 58prn MCP-Screen Imrn
1 o5
10'
10'
MCP Gain
FIG.3. Effect of space charge on spot size.
above 3 kG for nonzero angles between the magnetic field and the channel axis. Thus, the presence of the magnetic field enables a saturated gain to be obtained from a single MCP without the usual problems connected with positive ion feedback.
MECHANISM OF THE MCP STUDYOF THE EXCITATION The theory of secondary emission4 predicts that the probability for a particle to excite a channel of the MCP is proportional to the local specific ionization dEldx. Earlier work2v3showed that the probability of excitation for a minimum ionization particle was in the few percent region which produces a density of a few spots per millimeter for the track inside the MCP. In this study, the channels of the MCP were coated with CsI in order to increase the spot density. A CsI coating is believed to increase the spot density by two different processes. First, by its higher secondary electron emission coefficient and second by acting as an ultraviolet-sensitive photocathode inside the channel. In the latter case, ultraviolet Cerenkov light emitted from the PbO glass channel wall is assumed to excite the CsI coating.
MCP
AS HIGH-ENERGY PARTICLE TRACK DETECTOR
359
#' I
0"
,d'
0
1
2
3
4
5
6
7
Magnetic Field (kG)
FIG.4. Peak gain (continuous lines) and pulse resolution (dotted lines) as a function of magnetic field for various angles between the channel axis and the directionof the magnetic field.
Figure 5 shows the increase of efficiency due to the Csl coating. The data were taken by scanning a 3.5 GeV T - beam across a plate half the area of which was coated with CsI. The efficiency of the area coated with CsI reaches about 90%. On the other hand, the side without the CsI coating shows an efficiency of 70%.
100
s
V %
.-5 t i o .-V
..6
Beam 3.5 G e V ( + )
-5
-4
-3
-2
-1
0
1
2
3
4
5
Beam Position (mm)
FIG.5. Effect of CsI coating on efficiency of a 1-mm-thick MCP.
K. OBA ET A L .
360
I-
I!rn 1.0. W
>
F 0.5.
L
4
W
a
I
1
10
30
8= Plmc
FIG.6. Relative probability of exciting the MCP as a function of the kinetic parameter 7) = Plmc.
Figure 6 shows the dependence of the excitation probability, which is directly related to the spot density, on the kinetic parameter, 7 = P/mc, of the fast particle. (P is the particle momentum, m is the mass of the particle, and c is the speed of light.) The shape of the probability curve follows the well-known dEldx curve showing that the Cerenkov light does not contribute appreciably to the excitation process of the MCP. (The Cerenkov light has a threshold at about 7 = 0.8 and rises sharply at that value .) PERFORMANCE OF THE MCP DETECTOR Based on the studies described in the preceding section, two singlestage MCP detectors were produced. Both detectors had MCPs which were 1 mm thick with a bias angle of 15" and a channel diameter of 12 pm, a Pa 11 phosphor and an MCP phosphor distance of 1 mm. One detector used a standard MCP, the second had channels coated with CsI. Tests were performed using the M1-beam at the Fermi Laboratory. Figure 7 shows the picture-taking system and the trigger scheme. The system consists of an MCP tube, a coupling lens, a gated four-stage image intensifier tube, which serves as an optical switch and an optical amplifier, and a film camera. Passage of a particle through the detector is detected by means of thin plastic scintillators SI and Sz and an anticoincidence counter A, producing a signal B. Coincidence between B and a detector signal D produces a gate signal to the image intensifier and the camera. In operation, the second stage of the image intensifier is normally off and the image on the first stage is not transferred to the camera. When an interac-
MCP
AS HIGH-ENERGY PARTICLE TRACK DETECTOR
36 1
0
U Camera
E!!zFG*TE
At
D
Camera
FIG.7. System for picture recording and trigger system.
tion is detected, the first stage of the image intensifier is switched off in order to stop accepting the signal and the second stage is turned on to transfer the detected image on the long persistence phosphor screen of the first stage to the film. One of the interesting test results is shown in Fig. 8, where the detection efficiency is plotted as a function of the depth of the particle track through the MCP for MCPs with CsI and without CsI. The 100-pm-thick plastic scintillators SI and Sz were scanned along the depth coordinate to define the position of the particle track. The results shows that the CsI layer substantially increases the active depth of the detector. Figure 9 shows the measured spot size on the camera film as a function of the focusing magnetic field intensity for an MCP with a CsI coating. The measurement includes the effects of the image intensifier. If a simple optical system is used, the spot size decreases to 15 p m as was measured by direct photographic recording without the image intensifier. An example of the images of 200 GeVlc T- interactions within the detector is shown in Fig. 10. Examination of a larger number of samples provides a measurement of the spot density for minimum ionizing parti-
1 Scanning
500 p m
Beam position (MCP depth)
FIG.8. Efficiency of detection versus depth in channels of MCP.
MCP WITH Csl
HV 50
t
5.1 k
I 0
1
3.3
6.6 Bii
(kG)
FIG.9. Measured spot size versus magnetic field.
MCP
AS HIGH-ENERGY PARTICLE TRACK DETECTOR
363
FIG. 10. Example of the interaction due to a 200 GeVlc s - I inside the MCP.
cles. The measured spot density was about (3.0 0.5) mm-' and was essentially the same for both detectors. The rms displacement of the dot center from a least squares fitted line is about 5 pm. This number is a fair estimate of the position resolution of the detectors and is basically limited by the size of the channel. The excess of dots unrelated to the interactions on Fig. 10 is due to the relatively long decay time of the phosphor of the image intensifier, thus a few spots from previous events were recorded. The intrinsic time resolution of the detector was of the order of 1 psec, limited by the decay time of the P 11 phosphor. The rate capability of the detector was estimated from the output charge measurement to be in lo5Hz region. For some applications a lower MCP bleeder resistance may be needed.
-
CONCLUSION An MCP was tested as a high-energy particle track detector for application in the direct measurement of very short lifetime of charm and beauty particles. Application of a 1-mm-thick MCP with a 15" bias angle and magnetic
364
K. OBA ET A L .
focusing gave high detection efficiency, small spot size, capability to produce trigger signal, and high rate capability. CsI was tested as a coating material to improve efficiency. It is desirable to develop an MCP with smaller channel diameter to give higher spatial resolution and with lower resistance to give a higher counting rate capability. ACKNOWLEDGMENTS The authors would especially like to thank Mr. M. Montag for the mechanical design of the focusing magnet and Mr. E. Hassel for his technical assistance in the assembly of the magnetic focusing system. They also appreciate the patience of the physicists in experiment E-515 during data taking in the M-1 beam at Fermi Lab. This research was supported by the U.S. Department of Energy under Contract DE-AC02-76CH00016.
REFERENCES 1. 2. 3. 4.
Wiza, V.,Nucl. Instrum. & Methods 162, 587 (1979). Potter, D. M.,Nucl. Instrum. & Methods 189,405 (1981). Oba, K.,Rehak, P. and Smith, S. D., IEEE Trans. Nucl. Sci. NS-28, 705 (1981). Simon, R. E. and Williams, B. F., IEEE Trans. Nucl. Sci. NS-15, 167 (1968).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. 64B
High-Resolution Luminescent Screens for Image Intensifier Tubes V. DUCHENOIS, M. FOUASSIER, and C. PIAGET Laboratoires d’Electronique et de Physique Appliquke, Limeil BrPvannes. France
INTRODUCTION In image intensifier tubes, the picture quality can be characterized by the modulation transfer function (MTF), the output contrast versus spatial frequency for 100% input contrast. This function depends on many parameters such as the photoelectron energies, tube voltages, distance between electrodes, channel diameters (for MCP intensifiers), screen granularity, fiber optic pitch, etc. The goal of this study was the improvement of the modulation transfer function of the screen when it is settled on a fiber optic output window. Luminescent screens are often settled on fiber optics in order to reduce the light spread from the emitting phosphor particles and, consequently, to minimize the veiling glare of the tube. Twisted fiber optics also allow the image to be inverted with low light losses. To increase the contrast, Piedmont and Pollehn proposed some years ago to intagliate the phosphor particles in holes scooped out at the input of each elementary fiber.’ This is the approach taken here. PRINCIPLE In a conventional screen settled on a fiber optic plate (Fig. 1) the geometric positions and the dimensions of phosphor particles and fibers are completely independent of each other. Therefore the modulation transfer function is the combination of the phosphor screen MTF with the fiber optic MTF. Photons are emitted by the phosphor grains in all directions and very often reach the fiber optic far from their emission point (sometimes after diffusion by the phosphor particles and/or reflection on the aluminum layer). In order to avoid this light spread, the idea was to insert the phosphor particles in holes excavated in the elementary fibers (Fig. 2). Photons emitted by the screen will be kept within the corresponding fiber. In order to prevent light reaching the other fibers, the lateral walls of the holes 365 Copyright 8 1985 by Academic Press. Inc. (London) Ltd. AU rights of reproduction in any form reserved. ISBN 0-12-0147244
366
V. DUCHENOIS ET AL.
I
Aluminum layer
\
Electrons
Cmntional phosphor screen
FIG.1. Schematic diagram of a conventional phosphor screen on a fiber optic window.
need to be metallized. Then, each elemental cell acts as a blackbody emitting only into its corresponding fiber. Consequently, the modulation transfer function of the fiber optic screen should not depend on the phosphor but, theoretically, only on the MTF of the fiber optic. The different requirements of this study were (1) to excavate the holes in the fiber optic by elimination of the core glass to a certain depth without attacking the cladding glass; (2) to metallize the walls of the cavities; (3) to select a screen powder having particles small enough to be capable of entering the holes with a good filling factor; and (4) to adapt the screen deposition technique in order to force the particles to enter the cavities.
FIBER OPTICCHEMICAL ATTACK A selective chemical attack technique was studied in order to etch the fiber core glass without dissolving the cladding glass. This technique derives from that used for microchannel plates. The experiments were carried out with various types of fiber optic, from different manufacturers, and with various elementary fiber optic diameters ranging from 6 to 10 p m pitch. The glass compositions are very different from one type to another and the process has to be adapted to each case. Generally, acid solutions give good etching of the core glass. The etching speed may reach values 40 to 70 times higher for the core glass than for cladding glass. However, these speeds are not proportional to time and depend on the type of glass as well as the chemical and mechanical histories of the plate. Aluminurn layer
\, Phosphor screen
I
Electrons
Fiber optic
lntagliated phosphor screen
FIG.2. Schematic diagram of an intagliated phosphor screen on a fiber optic window.
HIGH-RESOLUTION LUMINESCENT SCREENS FOR
IITS
367
FIG.3. Photograph of a &pm Mullard fiber optic plate after chemical etching.
The cavity depth has to be larger than the phosphor particle size to get more than one particle in each hole. This is a necessary condition for high conversion efficiencies. Hole depths have been obtained ranging from 1.5 to 9 pm. Figure 3 shows an example of chemical attack in a Mullard 6-pm pitch fiber optic. The cavity depth is 5 pm. The core glass is very well attacked and the cavity walls made of cladding glass are very clean.
PHOSPHOR SCREEN PRODUCTION The phosphor particles should be small enough to enter the cavities, the sizes of which are about 8 and 4 pm, for 10- and 6-pm fiber optic pitch, respectively. However, these particles should not be too small because the screen conversion yield decreases with their size. For these experiments, screen powders were selected by a slow sedimentation technique. The best compromise between particle size and luminous yield was obtained with a preselected powder, the mean value of the particle size of which was 1.8 pm, only 6% of the particles having sizes of more than 3.5 pm. When a sedimentation technique is used to deposit the phosphor grains, they remain above the entrance of the fiber optic and do not enter the holes. This is due to capillarity forces in the liquid used in the process. It
368
V. DUCHENOIS ET AL.
FIG. 4. Photograph of an intagliated screen on a Mullard fiber optic plate (observed at normal incidence).
is therefore necessary to use a technique which forces the powder to enter the cavities. With the conventional centrifuge technique, the results have not been much better, but it has been possible to adapt this technique and to obtain a good hole-filling factor by a proper choice of mixture compositions and centrifuge cycles. Figures 4 and 5 show an example of an intagliated screen on a 6-pm pitch Mullard fiber optic, observed through a scanning electron microscope; the angles of observation are 90 and 45", respectively. The holes are uniformly filled with phosphor grains and all the excess powder has been eliminated. MEASUREMENTS As for standard screens, where thickness is important, the conversion yield (or cathodoluminescence efficiency) of the intagliated screens depends on the cavity depth. Figure 6 shows this variation for a 10-pm pitch fiber optic. The efficiency is optimum near 3 pm depth. At smaller depths, the yield decreases because too few particles are present in the holes, and some of the incident electrons do not fall onto the powder. If the cavity is deeper, the yield also decreases. As there are several phosphor particles
HIGH-RESOLUTION LUMINESCENT SCREENS FOR
IITS
369
FIG.5. Photograph of an intagliated screen on a Mullard fiber optic plate (observed at 45”).
superposed in each elementary hole, the light emitted by the upper particles is reflected, diffused, and partially absorbed many times by the other phosphor particles and by the metallization before being able to reach the core glass at the bottom of the cavity.
I
-‘B
‘\
X
K
0 0
I
1
1
2
I
I 3
4
I
5
I
I
6
7
I
I 8
9
1
0
Cuvity depth(pm)
FIG.6. Conversion efficiency of an intagliated screen versus cavity depth.
370
V. DUCHENOIS ET A L .
If the side walls of the cavities are not metallized, the total yield may increase by a factor of 2, but this is obtained at the cost of strong degradation of the spatial resolution, the screen then operating in a way similar to conventional screens on fiber optic plates. The screen modulation transfer functions were measured at Philips Research Laboratories by Mr. A. J. Jenkins with equipment described elsewhere.2The screen is introduced into a scanning electron microscope, the beam of which is used to write a thin electron line on the screen. The picture of this thin line is imaged through a microscope objective out of the vacuum system into an ODETA I11 system which enables direct recording of the Fourier transform of the line spread function which is in fact the modulation transfer function. As an example, Fig. 7 gives the MTF of a phosphor screen intagliated in a 6-pm pitch fiber optic. Curve a shows the MTF of the fiber optic plate itself, curve b that of a conventional screen on the same fiber optic plate, and curve c that of the intagliated one. The contrast improvement due to the intagliation of the phosphor grain is very impressive as the resulting MTF is almost identical to that of the fiber optic plate. For example, at 30 Ip mm-I, the fiber optic exhibits a contrast of 83% whereas a screen conventionally deposited reaches 49% and the intagliated one 78%. The intagliation improves the contrast of the screen by a factor of at least 1.5
0
5
10
15 20 25 30 35 40 Spatiol frequency (/'p/rnrn-l)
45
50
FIG.7. MTF of 6-prn pitch fiber optic phosphor screen: (a) fiber optic plate, (b) conventional screen on a fiber optic plate, and (c) intagliated screen on a fiber optic plate.
HIGH-RESOLUTION LUMINESCENT SCREENS FOR
IITS
37 1
at 30 lp mm-I and will similarly improve the contrast of image intensifier tubes. The contrast improves slowly when the cavity depth is increased (about + 10% between 2 and 9 pm). A computation made by J. C. Richard with the programs described elsewhere3 on the basis of the above-mentioned results shows that the range for the detection of green painted objects in a vegetation background can increase by 15 and 25% at starlight and moonlight illuminations, respectively. CONCLUSION It has been demonstrated that it is possible to excavate the core glass of fiber optic plates and to force phophor particles into these cavities. This intagliation technique improves the screen modulation transfer function so much that it closely approaches the MTF of the fiber optic plate. At 30 lp mm-I, the contrast was improved by a factor of 1.5. If such intagliation techniques were adapted for second-generation double proximity image intensifier tubes the contrast improvement at 30 lp mm-' would be 50% and would raise limiting resolution up to or beyond 40 Ip mm-I, increasing detection ranges by 15 to 25%. ACKNOWLEDGMENTS The authors with to thank Mr. A. J. Jenkins of Philips Research Laboratories who made the MTF measurements and Miss E. Desvaux who contributed to making the screens. This work has been partly supported by Direction des Recherches Etudes et Techniques.
REFERENCES 1. Piedmont, J. R. and Pollehn, H. K . , Proc. S.P.I.E. 99, 155 (1976). 2. Jenkins, A. J . , Proc. S.P.I.E. 274, 154 (1981). 3. Richard, J. C., Lamport, D. L., Roaux, E. and Vanneste, C., Actu Electron. 4, 353 ( 1977).
This Page Intentionally Left Blank
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS. VOL. 648
Multialkali Effects and Polycrystalline Properties of Multialkali Antimonide Photocathodes WU QUAN-DE and LIU LI-BIN Department of Radio-Electronics, Peking University, Beving, China
INTRODUCTION The “multialkali effect,” i.e., the fact that photocathodes of high sensitivity may be obtained if antimony is combined with more than one alkali metal, was discovered accidentally by A. H. Sommer in 1955.’ Up to now, however, the physical meaning of the multialkali effect has still not been explained explicitly. Because of the background of its real applications, the characteristics of the multialkali photocathode, especially its integral sensitivity and spectral response in the red and near-infrared light regions, have been improved successively in the recent decade. We believe that this photocathode still has large potential. One of the significant problems is the influence of the polycrystalline grain sizes and grain boundaries on photoemission. Usually a semiconductor model was used for the multialkali photocathode, but it is really a monocrystal model. This model cannot be used to discuss its spectral response, quantum yield, and integral sensitivity, etc. In addition to the influence of the composition and surface structure of the photocathode, the grain sizes and grain boundaries also have obvious influence on the photoemissive properties. The physical basis for the multialkali effect and the influence of the grain sizes and grain boundaries on photoemission are discussed in this article. PHYSICAL BASISFOR
THE
MULTIALKALI EFFECT
Crystalline Structures of Alkali Antimonide Materials
The crystalline structures of alkali antimonides have been determined by X-ray or electron diffraction. They can be classified into two main kinds, i.e., the cubic type and the hexagonal type. The cubic type has fairly good photoemissive properties, but the photoemissive properties of 373 Copyright 8 1985 by Academic Press. Inc.(London) Ltd. All rights of reproduction in any form reSeNcd.
ISBN 0-12-0147244
374
WU Q U A N - D E A N D LIU LI-BIN
TABLEI Lattice constants of hexagonal alkali antimonide materials
c
u
(A) (A)
9.51 5.35
10.636 6.025
11.19 2 0.02 6.32 2 0.02
10.932 5.610
the hexagonal type are relatively poor. Therefore the cubic type of alkali antimonide material has usually been used. Lattice constants of several alkali antimonides collected from some papers are listed in Tables I and 11. From these tables, we note that K3Sb and Rb3Sb have two types of structure, and the cubic type has larger density than the hexagonal one. From Table 11, it can be seen that the lattice constant of Na2KSb is the smallest, and its atom density is the largest; thus the electron density in its valence band is also the largest. Because, however, its electron affinity, E A , is fairly large, it is not a good photoemissive material. Its electron affinity can be lowered by treatment with Cs and Sb. Na2KSb(Cs)has a slightly enhanced lattice constant, but its value of E A is much lower, thus Na2KSb(Cs)has the best photoemissive properties among the materials listed in Table 11. The Crystalline Structures of Tri-Alkali Photocathodes
It is concluded in Somrner2 that there are three crystals, Na2KSb, NaK2Sb, and KzCsSb, present in tri-alkali photocathodes. To obtain sensitivity of more than 300 pA Im-', the Na2KSbcontent must form most of the photocathode, whereas the KZCsSb content must be much l e s 3Thus, theory and experiment have proved that the Na2KSb layer must be made first in order to obtain a fine photoemissive layer. The Multialkali Effect
The reasons for the multialkali effect in Sb-K-Na-Cs are as follows.
photocathodes
1. The lattice constant of NazKSb is the smallest, the atom density is the largest, and the electron density of the valence band is also the largest. The absorption of light and the photoexcitation of electrons are proportional to the electron density of the valence band. Therefore the larger the
375
MULTIALKALI EFFECTS
TABLE I1 Lattice constants of cubic alkali antimonide materials Cs3Sb
RbJSb
K2CsSb
K3Sb
Na2KSb
Na*KSb(Cs)
9.176 k0.003
8.84 20.02
8.615 k0.002
8.493 +0.005
7.727 20.003
+O.W
7,745
electron density is, the more photoelectrons are excited to the conduction band, and the higher is the probability of photoemission. 2. The NazKSb crystal has an ordered structure. So the probability of collision between the photoelectron and a phonon is smaller and the escape length is longer than that in other materials. The threshold energy of production of electron-hole pairs is 3.0 eV in this material, so collision events of this kind cannot occur with visible light or near-infrared light. 3. The electron affinity of NazKSb(Cs)is smaller than that of Na2KSb. Although the crystal constant of the former is somewhat larger, the escape probability of the photoelectrons through its surface to vacuum increases, i.e., the probability of photoemission is enhanced. The physical basis of the so-called “multialkali effect” is the synthesis of advantages of the three features mentioned above. In addition, it is pointed out that photoemission is concerned with the distribution of state densities of electrons in valence and conduction bands. Although the structures of the energy bands of alkali antimonide materials have been calculated to some e ~ t e n tthe , ~ data given were insufficient. Therefore we can discuss the influence of distribution of the state densities on photoemission only approximately, as is shown later. THEPOLYCRYSTALLINE PHOTOEMISSIVE MODELAND QUANTUM YIELD The multialkali photocathode of high sensitivity is composed of a large number of p-type Na2KSb crystalline grains. There may exist different alkali antimonides, some impurities, and an excess of cesium in grain boundaries. The dimension of the grain boundary is so much less than that of crystalline grain, that it can be neglected. A schematic diagram of a semitransparent photocathode is drawn in Fig. 1. The light absorption, photoexcitation, and influence on photoemission in a slab of width dz per unit area which is perpendicular to the z axis are functions of z only (if the influence of grain boundaries in the x and y directions can be neglected, or
376
WU QUAN-DE AND LIU LI-BIN Ro
atmosphere glass
F I G . 1.
R,
photocathode
R2
vacuum
Schematic diagram of a semitransparent photocathode.
these influences can be included in that of z direction). This is the basis of the one-dimensional model adopted in this article. The Potential Barrier of the Polycrystalline Grain Boundary
For convenience to compare theoretical results with experimental results, we assume that the photocathode is composed of grains with the same dimension d. It is also assumed that the impurity atoms are uniformly distributed with a concentration Nd ( ~ m - ~and ) , that the boundary state density is Qt (cm-2), located at energy level Et . Because there exist boundary states energy band bending takes place near the grain boundary. Figure 2 shows the energy band structure along the Oz axis. It is composed of a set of flat band regions and pairs of back-to-back energy barrier regions. The depletion layer approximation is adopted to solve Poisson's equation, and the electric potential V ( z ) in this depletion region can be obtained. If the original 0' of the coordinate system is at the midpoint of the flat region (see Fig. 2), V ( z )becomes
where EEO is the dielectric constant of the cathode, q is the charge of a carrier, 1 is the width of the bending region (region I or 111), and V,, is the maximum level of the valence band at the flat region. The doping density in the multialkali photocathode is so large that it is partly depleted (Nd d > Qt>.The energy barrier height of the grain boundary EB becomes
EB = qvg = q 2 Q ? / 8 & & ~ d It is evident that EB is inversely proportional to Nd.
(2)
MULTIALKALI EFFECTS
377
( n - l l d (n-l)d+l nd-l nd
FIG.2. Energy band structure and final states of photoelectrons.
Photoemission from Polycrystalline Photocathodes
The existence of the grain boundary may produce the following effects on the photoexcited electrons. (1) Recombination between photoelectrons and holes can easily occur at the interfaces of grains. This results in a loss of photoelectrons, when they pass through each grain boundary. (2) Reflection, scattering, or absorption of light at the grain boundaries can be equivalent to an increase in the loss rate of photoelectrons. These two loss rates can be synthesized into an effective grain boundary loss rate, P,. (3) The energies of photoexcited electrons in the region of the potential barrier with barrier height EB is lower than that of those excited in the flat region. When EB increases, the flat region decreases, and the fraction of electrons with high energy decreases. The number of electrons which escape must also decrease. Previously6it was assumed that the scattering probability for electronphonon scattering and the energy loss could be approximately described by Poisson distributions. In this article, it is assumed that the energy loss for electron-phonon scattering can be neglected, i.e. , quasi-elastic scattering occurs; the quantum yield and energy distribution of the photoelectrons are then discussed. The energy distribution of photoexcited electrons in a slab of unit area in the nth crystalline grain with width dz located at a distance z from interface between the glass window and the cathode has the following form: N,(E*, hv, Z )
a G(z,hv)p,(E*
- hv)pc(E*)dz
(3)
where C ( z ,hv) is the generation rate of photoelectrons, pv and pc are the
378
WU QUAN-DE AND LIU LI-BIN
density functions of energy states for valence and conduction bands, respectively, and the subscript n denotes the nth crystalline grain. The energy of the photoexcited electrons is a function of z , and the energy refers to the bottom of conduction band in region I1 in Fig. 2 as zero. The energy of the photoexcited electron, E*(z), in every region can be written as E - (q2Nd/2&&0)[(n - I)d + 1 -
i
zI2,
region I region I1
E*(z) = E, E - (q2Nd/2&&0)[z - (nd
- 012
(4)
region I11
The photoelectrons produced in the nth grain might pass through ( N +
1 - n) grain boundaries ( N is the total number of grains in the z direction),
when they escape from the surface of the photocathode. As defined above, P, is the effective grain boundary loss rate, then the effective grain boundary transmission rate is P, which is equal to (I - P e ) .The energy distribution of photoelectrons excited in dz which get to the surface is then N;[E*(z),hu, z ] d ~= c * PNf'-"G(z,hu)pV(E* - hu)pC(E*)dz ( 5 )
where c is a constant. The escape probability of photoelectrons is determined by the surface function T(E*)'
T(E*) = [I - (E,/E*)"*]
(6)
where EA is the electron affinity. The contribution to the energy distribution of photoelectrons which are excited in dz and can escape into vacuum is then equal to N'[E*(z), hu, Z
) ~ Z= c
*
P N + ' - n G ( h~~, ) p , ( E *- hu)p,(E*)T(E*)dz (7)
The contribution to the energy distribution of photoelectrons in the nth grain can be obtained by integrating Eq. (7), Nn(E*, h ~ =) c . PN+'-nT(E*)G ( z , h ~ ) p v ( E *- h~)p,(E*)dz (8) This integral should be carried out in three different regions. Summing up the contributions of all grains in the z direction, we obtain the following equations:
379
MULTIALKALI EFFECTS N
N d E , hv) = c
r-'
2 PN+'-'T(E*)
n=l
n-IWtI
G ( z , hv)p,(E*
- hv)p,(E*)dz (9)
N
N d E , hv) = c n = l P N + ' - T ( E * ) nd-l G(z, hv)pV(E*- hv)p,(E*)dz
and the expression for the energy distribution of the photoelectrons can be represented as N ( E , hv) = N1
+ N I I+ N ~ l l
(10) Let Zo denote the number of incident photons per unit area per second, then the quantum yield can be written as follows, Y(hv) = (l/Zo)
4
N ( E , hv)dE
(1 1 )
where EA = EO- E- (region 11),EOis the vacuum level, Emax = hv - E G , and EG = E- - E+ , i.e., the forbidden gap. The values Of Em, in regions I and I11 are less than that in region 11. If the effect of inelastic collision is extremely obvious during the motion of the photoelectron from its position of excitation to the surface, for example for ultraviolet light, the above-mentioned equations should be modified.
EXPERIMENTAL RESULTS The above-mentioned formulas can be used in discussion of various polycrystalline alkali antimonide photocathodes. The physical parameters of the Sb-K-Na-Cs photocathode are discussed further in this section.
Comparison between Theoretical and Experimental Results
The generation rate of photoelectrons G(z, hv) for a GaAs photocathode on a sapphire substrate8 is G(z, hv) = ZoA[exp(-az)
+ B exp(az)l
(12)
where A and B are given by the following two equations:
B = R2 exp(-2cuD)
(14)
380
WU QUAN-DE A N D LIU LI-BIN
where D is the thickness of the cathode, and Ro, R I , R2 are the reflection coefficients of the interfaces as shown in Fig. 1. According to thin film optics,
Ro
=
(n, (n, + 112'
(n, - n,)2 + k2 R~ = (n, - n,)2 + k2'
R2 =
(n, - 1)2 + k2 (n, + 1)2 + k2 (15)
where n, and n, are the indexes of refraction of the glass window and the photocathode, respectively, and k is the absorption coefficient. In performing the calculation n, and nc are taken to be 1.5 and 3.0, respectively, and k is taken according to the values measured by K o n d r a s h ~ v . ~ The State Density in the Valence Band pv(Ev) pv can be expressed as 4
pV=
2 A,Evn
n=O
(16)
whereA0 = -4.73, A, = -18.3,A2 = -1.50,A3 = 4.73, andA4 = 1.13.1° The state density in the conduction band can therefore be expressed as a constant. The Surface Escape Function T(E) Equation (6) can be used for T(E). The Size of the Crystalline Grain d and Thickness of the Cathode We suppose that d is 150 A.The thickness D estimated from the color of the cathode is about 1000 to 1200 A.The number of the grains in the z direction is 7 or 8 in our calculation. Polycrystalline Parameters EB, Nd, Qt, I, and P, The potential barrier height EBcan be obtained by using the following formula: IIR = A5T-I" eXp(-EB/kT) (17) According to the isotropic properties of the polycrystalline layer, E B can be derived by measuring the temperature dependence of the resistance R of the layer. The value of Q, is assumed to be 5.5 x 10l2 cm-2.6JoThe
38 1
MULTIALKALI EFFECTS Wavelength (pm)
0.9
0.8
0.7
0.4
0.5
0.6
-
1.3
1.5
1.7
1.9
2.1
2.3
2.5
2.7
2.9
3.1
Photon energy (eV)
FIG.3. Theoretical and experimental spectral responses for quantum yields of I, 375 pA Im-I and 11, 275 pA Im-I.
doping density Nd can be obtained by using Eq. (21, where E = 6." The ,. width 1 of the depletion region can be obtained from EB = q N d l 2 / 2 ~ qThe important parameter P, must be determined through experiment and by comparison with the experimental curve. It can be seen from Fig. 3 that the quantum yield has a tendency to drop in the short wavelength range when P, is large, whereas it may not do so when P, is small. Explanation of Experimental Results
Comparing the theoretical results with the experimental results, we find that the lower the potential barrier height EB is, the larger the quantum yield is, and the effect is more obvious in the range of low photon energies. The reason is that the higher the potential barrier is, the more serious the energy band bending is. It can also be seen that the larger P, is, the more serious the decrease in quantum yield is, especially in the region of high-energy photons. The reason is that the absorption coefficient of the cathode for high-energy photons is larger, so the photoexcited electrons are nearer the substrate window. When these photoelectrons move to the surface, they must pass through more grain boundaries, thus the energy loss must be large.
382
WU QUAN-DE AND LIU LI-BIN
TABLE111 Calculated values of quantum efficiency“ d
(A)
P, (%)
I50
200
300
600
20 15 10
3IOb,‘ 350 455
371 419 544
473 534 694
552 623 810
(1
Ee
= 0.12 eV,
EA = 0.35 eV.
Assumed value. Values, S (PA 1rn-l).
DISCUSSION In general, it is believed that if both the composition and the thickness of the photocathode are appropriate, and EAis also low, its photosensitivity should be high. Photocathodes with high sensitivity can also be classified into three types: (1) EB is small, but Pe is slightly large. In this case, the spectral response in the region of red and infrared is good. (2) Pe is small but EBis slightly large. This photocathode has a better short wavelength response. (3) Both EB and Pe are small. This photocathode has an ideal spectral response curve. The integral sensitivity of a photocathode can be calculated from its quantum yield. Therefore the possibility of increasing the integral sensitivity can also be discussed from the polycrystalline point of view. There may exist some impurities and/or an excess of cesium in the grain boundaries. Pe may be decreased by controlling the amount of cesium and the purity of material of the cathode. EB may be decreased by increasing the doping density appropriately. Increasing the size of the grains, to decrease the number of grain boundaries in the z direction, could also raise the integral sensitivity of the photocathode. Some calculated data are shown in Tables 111 and IV. We can go a step further: let P, = 5%, E g = 0.05 eV, EA = 0.24 eV, d = 300 A, then we obtain a photosensitivity S = 1250 p A lm-I. When we let d = D, P, = EB = 0, EA = 0.24 eV, i.e., the case of a “quasi-monocrystal film,” we could get S = 1400 pA lm-’. We have pointed out in another paper that if a photocathode can be prepared in the form of a “quasi-monocrystal film,” there should be no difference between its integral sensitivity and that of the monocrystal film.
383
MULTIALKALI EFFECTS
TABLEIV Calculated values of quantum efficiency“
0.35 0.30 0.24
455b 568 723
544 681 869
694 870 1110
810 1100 1290
@ E n= 0.12 eV, P, = 10%. Values, S (PA lm-I).
REFERENCES 1. Sommer, A. H., Rev. Sci. Instrum. 26, 725 (1955). 2. Sommer. A. H., In “Photoemissive Materials,” ed. by R. E. Krieger. Wiley, New York (1980). 3. Dowman, A. A., J . Phys. D 8, 69 (1975). 4. Mostovskii, A. A., Izv. Akad. Nauk SSSR, Ser. Fiz. 38, 195 (1974). 5. Seto, Y. W., J . Appl. Phys. 49, 5565 (1975). 6. Fan Yao-Liang and Wu Quan-De, Acta Electron. Sin. 11, 26 (1983). 7. Fowler, R. M., Phys. Rev. 38, 45 (1931). 8. Liu, Y. Z., Appl. Phys. Lett. 17, 60 (1970). 9. Kondrashov, V. E., Bull. Acad. Sci. USSR, Phys. Ser. 28, 1349 (1964). 10. Liu Li-Bin, Master’s Thesis, Peking University (1983). 11. Fisher, D. G., J . Appl. Phys. 45, 487 (1974).
This Page Intentionally Left Blank
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. 64B
Properties of a Photocathode with a Palladium Substrate ZHANG XIAOQIU, PANG QICHANG, and LEI ZHIYUAN Xian Institute of Optics and Precision Mechanics, Academia Sinica, Xian, Shaanxi, China
INTRODUCTION The multi-alkali photocathode has high sensitivity and wide spectral response range. It is often used in streak tubes, but its resistance is very large, in general' about lo6R I T 1 . It is unsuitable for direct use in the tube at high sweep speeds. In order to obtain sufficient brightness of the image on the phosphor screen to take a picture, there must be very large photoelectric current density at the photocathode. Thus, if the resistance of the photocathode is very large, there will be a very large potential fluctuation across the photocathode, leading to distortion of the image and decreasing the spatial res~lution.~J In order to decrease the resistance of the multi-alkali photocathode, we used Nesa with a protective film of SiOa as the substrate. But when the its transparency becomes very resistance of Nesa is lower than 50 0 IT1, poor, the surface resistance is not uniform, and the spray technology is complicated. We therefore stopped using this technology a few years ago. A metal mesh of 50 Ip mm-I and 50% transmission which was embedded into a glass disk in protective gas was then used as the substrate of our tubes. The resistance was lower than 0.1 0 0 - I . Multi-alkali photocathodes prepared on this substrate by standard processing techniques did not change the substrate resistance and its sensitivity was up to 130 pA lm-'. The metal mesh substrate completely suited the requirements of the streak tube and the framing tube used in high-speed photography. In order to improve the time resolution of the streak tube, a spherical grid mesh was incorporated near the spherical photocathode to increase electric field strength to about 27 kV cm-I and thus to decrease the transit time spread caused by the initial photoelectron velocity spread. Because of interference between the substrate mesh and the grid mesh Moire fringes4appeared during static operation and the resolution of the tube 385 Copyright 0 1985 by Academic Press. Inc. (London) Ltd. All rights of reproduction in any form resewed. ISBN 0-12-014724-6
386
ZHANG XIAOQIU ET AL.
was decreased. We therefore introduced a uniform conductive palladium film instead of the metal mesh substrate to eliminate MoirC fringes. PREPARATION AND TESTING OF THE CONDUCTIVE PALLADIUM FILM The test sample is shown in Fig. 1. It is made of a circular zinc Crown glass plate of diameter 60 mm with two platinum wire electrodes welded symmetrically on its opposite sides. Some silver paste is painted between the platinum wire and glass, Aluminum strips are evaporated onto them. The Pd film is sputtered on the area between the two Al strips. The useful area of Pd film is 1 in.2. The sample is sealed in a container joined to a high vacuum system. When it is evacuated to a pressure less than mm Hg, the container is baked, and the resistance and transparency of the Pd film are measured at different temperatures. The results are shown in Figs. 2 and 3. From these curves it can be seen that the resistance of Pd films decreases quickly and its transparency decreases slowly with increase in the baking temperature, but, when the temperature exceeds a certain value, both the resistance and the transparency increase again. The higher the resistance and the transparency of the Pd film, the lower the temperature it can undergo in vacuum baking. It is probable that the form of the sputtered Pd film is a group of islands. During vacuum baking these islands are gradually reduced to uniformity, so the resistance and transparency decrease. Above the critical temperature the Pd film will reevaporate, showing an increase of its resistance and transparency. The characteristics of the sputtered Pd film are stable; even if it is exposed to air for a long time, there is no change in its resistance and transparency. After vacuum baking at high temperature there will be proper adhesion between the Pd film and the glass. Before vacuum baking Al strips
glas; plate
FIG.I . Sample used to test resistance and transparency.
PHOTOCATHODE WITH A PALLADIUM SUBSTRATE
300
I
01
387
Baking for 24 hr
I
100
200 300 400 500 Temperature ("C)
FIG.2. Resistance of the Pd films as a function of vacuum baking temperature.
the film can be easily removed. When we wipe the baked Pd film, some Pd always adheres to the glass, thus the insulation performance of the glass is reduced. CHOICE OF THE RESISTANCEOF THE PALLADIUM FILM If the cathode has a circular configuration, if the emission current density of the photocathode is uniform, and if surface resistivity is constant over the photocathode area, the potential drop V at the center of the photocathode for the framing operation is given by the formulas,6 V = pJR214 (1) where p is the surface resistivity of the photocathode, J is the emission
1
I
01
Baking for 24 hr
100 200 300 400 500 Temperature ("CI
FIG.3. Transparency of the Pd films as a function of vacuum baking temperature.
388
ZHANG XIAOQIU ET AL.
current density, and R is the radius of the photocathode. When the potential drop V, the emission current density J , and the radius R are known, the maximum allowable surface resistivity can be calculated. Usually, the required photocathode surface resistivity is about 50 R U-'.'.s*7-'o For the streak tube, the potential drop V at the center of the photocathode is given by the formula6q7 V = tJpa(a + 2b)
(2)
where a is the half-width of the sweep slit and b is the width of the unexposed part of the conductive substrate. It can be seen from Eq. (2) that, for a particular potential drop V, the allowable photocathode surface resistivity of the streak tube is much larger than that of the framing tube. Some authors suggest that the former is about 30 times larger than the latter. From the curves shown in Fig. 2, it can be seen that if the initial resistance of the sputtered Pd film is about 100-150 R U-I, the final resistance will be less than 50 R 0-l after baking at 380°C for 24 hr. PROCESSING S - 20 PHOTOCATHODES ON PALLADIUM SUBSTRATES S - 20 photocathodes have been prepared on Pd substrates by standard processing techniques. The sensitivities obtained are 60 to 120 p A Im-I. Because the transparency of the Pd substrate is about 30% when its resistance is 50 R I T ' , this sensitivity corresponds to 200-300 p A lm-' in the photocathode itself. It had been found that the resistance remains nearly constant during the S * 20 photocathode processing but the transparency of the substrate increases during K-Sb cathode processing and then remains fixed. The results are listed in Table I. The spectral response of the Pd substrate photocathode is nearly the same as that of the metal mesh substrate photocathode. TABLEI Resistance values during multi-alkali processing on a Pd substrate Tube number 2 3
I
Initial resistance R0-I
After bake-out
n0 - 1
110
45 28 43
50 150
After K
After
n0 - 1
After Na
n0-1
Cs 0-1
Final resistance n 0-1
45 30 43
50 30 45
49 28 45
48 28 45
PHOTOCATHODE WITH A PALLADIUM SUBSTRATE
389
Usually, if the processing of an S - 20 cathode is not successful, a rebake at 380°C can be made to remove the unsuccessful cathode from the substrate and to prepare it again. This cannot be done. The measures are not suitable for photocathodes on Pd films because the resistance of the Pd film will increase by about two orders of magnitude.
RESULTS Framing tubes were constructed whose photocathode substrates were Pd films of resistance 50 0-1and whose photocathode sensitivities were 60 to 120 pA lm-l. A rectangular pulse of width 30 nsec was applied to the shutter electrode of one of the frame tubes and the resolution obtained on the phosphor screen was 6 Ip mm-I, as shown in Fig. 4. The result is similar to that obtained with the metal mesh substrate tube, as shown in Fig. 5. We also made a streak tube with its photocathode on a Pd film of resistance 50 i2 0 - I . The sensitivity was 110 pA lm-I. YAG mode-locked laser pulses of width 30 psec transmitted through a frequency multiplier, a Fabry-Perot etalon and a slit were projected onto the photocathode of this tube and a scanning voltage giving a scanning speed of 4 x lo9 cm sec-l was applied to the deflection plates. The streak image shown in Fig. 6 was obtained. The dynamic spatial resolution of the tube was 7 Ip mm-I, thus the theoretical time resolution was about 3.6 psec. The static spatial resolution of the tube, substrate was about 30.5 lp mm-' and there was no Moire fringing (Fig. 7a). A static resolution photograph produced by a streak tube using a metal mesh substrate shows obvious Moirt fringes (Fig. 7b). In testing the Pd substrate tube the light intensity must be limited; too much energy may burn the cathode irrecoverably.
FIG.4. Framing image taken from tube with Pd substrate; exposure time 30 nsec.
390
ZHANG XlAOQlU ET A L .
FIG.5. Framing image taken from tube with metal mesh substrate;exposure time 30 nsec.
FIG.6. Streak tube image, scanning speed 4 x lo9 cm sec-I; resolution is 7 Ip mm-I.
FIG.7. Static images of streak tubes using (a) Pd substrate and (b) metal mesh substrate.
PHOTOCATHODE WITH A PALLADIUM SUBSTRATE
391
CONCLUSIONS A Pd film can be used as the conductive substrate of the photocathode in high-speed photography image tubes. It does not generate Moire fringes so that the resolution of the tube is increased. The technique is simple, so that it is a more suitable conductive substrate than the metal mesh. ACKNOWLEDGMENTS We would like to thank the Vice-director of the Institute, X. Hou, for his continuous encouragement and guidance, and to thank all the people of fifth department of the Institute for supporting us in the manufacture and measurement of the image tubes.
REFERENCES 1. Engstrom, R. W., Stoudenheimer, R. G., Palmer, H. L.and Bly, D. A., IEEE Trans. Nucl. Sci. NS-6, 120 (1958). 2. Stewart, G. W. and Wainck, P. W., Rev. Sci. Instrum. 34, 512 (1%3). 3. Thomas, B. R., I n “Adv. E.E.P.” Vol. 33B, p. 1119 (1972). 4. Mu, G., I n “Optics,” p. 359. The People’s Educational Publishing House, Beuing, China (1978) (in Chinese). 5. Garfield, B. R. C. and Folkes, J. R., I n “Adv. E.E.P.” Vol. 28A, p. 375 (1%9). 6. Niu, H., Song, Z. X., Ren, Y. A. and Wang, G. H.,I n “First Annual Conference on Optics in China”, September (1978). 7. Ahmad, N . , Gale, B. C. and Key, M. H. I n “Adv. E.E.P.” Vol. 28B, p. 999 (1%9). 8. Garfield, B. R. C. and Bailey, P. C., I n “Adv. E.E.P.” Vol. 33B, p. 1137 (1972). 9. Hou, X., Sibbett, W. and Weekley, B., J . Appl. Phys. 53, 3243 (1982). 10. Hou, X., Sibbett, W. and Weekley, B., Rev. Sci. Instrum. 52, 1487 (1981).
This Page Intentionally Left Blank
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. 64B
The Luminous Efficiency of a Phosphor Layer in the Forward and Backward Directions A. G.DU TOIT and
C.F. VAN HUYSSTEEN
National Physical Research Laboratory, CSIR, Pretoria, South Africa
INTRODUCTION It is of general interest to have some knowledge of the distribution of light from a phosphor screen. Bril and Klasens' derived formulas for this distribution based on the assumption that the screen consists of a thick layer of very small phosphor particles. Equations were derived for the forward and backward components of the phosphorescent light generated in such a screen. For image intensifier tubes, however, the layer of phosphor particles is normally made just dense enough to have a complete coverage of the substrate. The Fountain tube2 presented a unique opportunity to gain more information on the luminous efficiency of such a phosphor layer in the forward and backward directions. In this image intensifier tube (see Fig. 1) the photoelectrons are reflected by a retarding electric field after passing through a small hole at the crossover. This small hole is in a flat metal plate and the phosphor layer is applied to the back of this plate. The screen is viewed from the side on which the electrons are incident. The luminous efficiencies of this screen and of a conventional screen produced by the same techniques, using the same type of phosphor particles were compared. The effect of an aluminum backing layer was also determined in both cases. EXPERIMENTAL TECHNIQUES
Phosphor Powder
A P * 20 phosphor powder, BSE Ultrafine (from Riedel de Haen) was used. Only particle sizes between 1 and 3 p m were selected by means of a sedimentation column. The final particle size distribution is shown in Fig. 2. 393 Copyright 8 1985 by Academic Press, Inc. (London) Ltd. All rights of reproduction in any form reserved.
ISBN 0-12-014724-6
394
A. G . DU TOIT A N D C. F. VAN HUYSSTEEN
-FOCUSING ELECTRODE -PHOSPHOR SCREEN
FIG.1. The Fountain tube.
Screen Forming Technique
All the phosphor layers were applied by means of a wet settling technique using potassium silicate as binder and accelerated sedimentation by centrifuging. The screen densities were 1.2 ? 0.1 mg cm-*. The Fountain Tube Screen
The metal substrate was made of Kovar or equivalent material, and then polished and blackened by an oxidation process. The measured coefficient of reflection was 6%. Half the area where the phosphor layer was to be applied was vacuum coated with aluminum to form a reflecting layer. The phosphor layer was then applied (see Fig. 3). Thus one-half of the layer was on a nearly nonreflecting substrate and the other half was on a highly reflecting substrate. This screen was mounted in a Fountain tube. The tube was processed in a photocathode transfer system3 and used to measure the luminous efficiencies of the two screen types. Conventional Screen
The screen was deposited on a normal glass substrate. The aluminum backing was applied by means of a flotation technique. The thickness of the aluminum film was 130 nm.
LUMINOUS EFFICIENCY OF A PHOSPHOR LAYER
395
PARTICLE SIZE (pm)
FIG.2. Size distribution of phosphor particles.
Screen without Aluminum Backing
In this case a conductive tin oxide layer was formed on the substrate. The tin oxide layer and the opposite glass surface were coated with antireflecting layers of magnesium fluoride. The measured transmittance was better than 96%.
LAYER OF PHOSPHOR PARTICLES
FIG.3. Experimental Fountain tube screen.
396
A. G. DU TOlT A N D C. F . VAN HUYSSTEEN
Measurement of Luminous Eficiencies Screen brightness was measured by means of a Spectra Pritchard Photometer Model 1980A.t In the case of the Fountain tube the photocathode was uniformly illuminated and the photocurrent and the screen brightness measured at screen voltages from 1 to 10 kV. The ratio of the voltages on the electrodes was kept constant at the focusing value to assure that all photoelectrons arrive at the screen. In the case of the screen on the conductive substrate and the conventional screen with aluminum backing, electrons from a multichannel plate excited by ultraviolet radiation were used. The screens were placed in close proximity to the output of the multichannel plate. The screen brightness was measured at acceleration voltages from 1 to 10 kV. In each case the current density was fixed at a value between 30 and 50 pA m-2 (3-5 nA cm-2). All brightness values were corrected for the transmission losses caused by the windows through which the measurements had to be made.
RESULTS The luminous efficiency of the various screens were plotted against electron energy (see Fig. 4). Measurable brightness was obtained for all t From Photo Research, 3000 North Hollywood Way, Burbank, California.
0.20,
I
I
I
I
1
ACCELERATING VOLTAGE (kV)
FIG.4. Luminous efficiencies of various phosphor screens as a function of accelerating voltage.
LUMINOUS EFFICIENCY OF A PHOSPHOR LAYER
397
the screens from approximately 1 kV, except for the conventional aluminum-backed screen for which it was measurable from 3 kV, because of energy losses in the aluminum layer. The slopes of all the efficiency curves increase from zero to a constant value at approximately 5 kV. The efficiency values for the conventional screen with aluminum backing agree well with that obtained by Diakides4for a similar screen. It is noticeable that the effect of the reflecting aluminum layer is much less in the case of the Fountain tube screens than for the conventional screens. It should also be noted that though the Fountain tube screen on the aluminum showed the highest efficiency, the rate of increase for the conventional screen with aluminum backing was higher. The slopes of the straight sections above 5 kV were computed and are given in Table I.
DISCUSSION When a phosphor layer is made by means of a sedimentation technique there is a statistical variation in the thickness of the layer when single particles are observed. In order to avoid voids in the screen the layer has to be several particles thick on most of the covered area (see Fig. 5 ) . (This is also evident from the fact that with a screen thickness of 1.2 mg cm-*, ~ , a packing the density of the phosphorescent material 3.45 g ~ m - and density of about 50%, an average thickness of 7 p m is to be expected.) Electrons with energies of 10 kV or less can hardly penetrate the particles in the top layer if they are 1 p m or larger. It will therefore be assumed that all the phosphorescent light is generated in a surface layer of particles on top of a relatively thick underlayer with a transmission factor of T. For the conventional screen with aluminum backing it can then be stated that
RT(dLBIdV) + T(dLF1dV) = 27.9
(1)
TABLEI Slopes of luminous efficiency curves above 5 kV Screen Fountain tube screen on black substrate Fountain tube screen on aluminum coated substrate Conventional screen without aluminum backing Conventional screen with aluminum backing
Slope of luminous efficiency curve in candelas per watt 18.0
23.9 15.5 27.9
398
A. G . D U TOlT A N D C. F . VAN HUYSSTEEN
FIG.5. Scanning electron microscope photograph of layer of phosphor particles, magnification 3000.
and for the conventional screen without aluminum backing that T(dLF/dV)= 15.5 (2) where R is the coefficient of reflection of the aluminum backing, LBis the light emitted in the backward direction, LF is the light emitted in the forward direction, and V is the acceleration voltage. If it is further assumed that R = 0.8 as is normally the case, and that dLF/dV is the same for both screens, we obtain dLB/dV = (1/T)15.5
(31
and dLF/dV = (1/T)15.5 (4) From this result it can be concluded that the derivatives of the forward and backward components are not only constant but also equal between 5 and 10 kV.Then it can be assumed that between these voltages Lg and LF
LUMINOUS EFFICIENCY OF A PHOSPHOR LAYER
399
is given by the equation
LB = LF = (1/T)15.5 (V
- Vo)
where Vo is a constant for a particular screen (4.15 kV for the conventional screen with aluminum backing and 2.72 kV for the screen without aluminum backing). In the case of the Fountain tube screen the light emitted in the forward direction passes through the underlayer once, is reflected and has to pass through the underlayer a second time before it leaves the screen. In this case we have
(dLBldV) + RP(dLF/dV) = 23.9
(5)
and
(dLeldV) + Rbp(dLF/dV) = 18.0
(6) for the phosphor layer on the aluminum-coated and the black substrate, respectively, where Rb is the coefficient of reflection of the latter, i.e., 0.06. If it is now assumed that dLBldV = dLF/dV as in the previous case, one can easily show that T = 0.67. This agrees well with the value of approximately 0.70 suggested by Donofrio and RehkopP for a similar phosphor layer. Substituting this value of T into Eqs. (3), (4), (3,and (6) one finds that for the conventional screen
and for the Fountain tube screen
The lower values for the Fountain tube screen may be attributed to poisoning by alkali vapor from the photocathode. (If this were not the case the slope of the luminous efficiency curve for the aluminum-backed Fountain tube would have been 36.6 cd W-',) If both the initial assumptions, viz. that the light is generated in a thin surface layer of the phosphor layer and the deduction from Eqs. (3) and (4) that the forward and backward components of the light generated are equal and correct, it can be stated in analogy with Eqs. (5) and (6) that (1
and
+ R P ) L ( V ) = MN(V)
(7)
400
A. G . DU TOlT AND C. F. VAN HUYSSTEEN
where for an acceleration voltage of V
L(V)
=
LB = LF
and MAI(V)and Mb(v) are the measured luminous efficiency in candelas per milliampere of the phosphor layer on an aluminum-coated substrate and on a black substrate, respectively. Therefore
MAI(v)/Mb(v) = (1
+ R p / ( 1 + Rbp) = 1.32 (constant)
The actual values of MA1(v)/Mb(V )for acceleration voltages from 1 to 10 kV are given in Table 11, and are in fairly good agreement, in particular those below 5 kV. These results confirm the assumption made by Bril and Klasens that the forward and backward components of the generated light are equal. It also indicates that the electrons do not penetrate the first layer of particles to any significant extent, at least not below 10 kV. The light generated in a single particle is thus distributed in equal amounts to the front and to the back, independent of the energy of the electrons. Since the depth of penetration into a single particle depends strongly on the energy of the electrons, it appears that the distribution is also independent of the depth of penetration into each particle. This means that lens effects due to the
TABLE I1 Ratio of the luminous efficiencies of a Fountain tube screen on an aluminum-coated and a black substrate Acceleration voltage (kV) I 2 3 4 5 6 7 8 9 10
M,dV)/Mb(V) 1 .27u
I .37 I .36 I .36 I .38 I .35 I .35 I .34 I .34 I .34 ~
Calculated from measurements at end of range of photomer.
LUMINOUS EFFICIENCY OF A PHOSPHOR LAYER
40 1
approximately spherical shape of most of the particles do not play a measurable role. CONCLUSION The results seem to confirm that nearly all the phosphorescent light is generated in a top layer of small particles on top of a layer with rather low transparency. In the case of a conventional screen all the light has to pass through this underlayer, except for the part of the backward component lost by reflection at the aluminum backing. The Fountain tube screen has the advantage that the backward component escapes without any losses but the forward component suffers heavy losses due to double passage through the bottom layer and one reflection at the aluminum-coated substrate. It is clear that both screen types would benefit significantly if a technique could be developed to form a closely packed single layer of 1 to 2-pm particles, directly on the substrate, eliminating the relatively thick underlayer.
REFERENCES 1. Bril, A. and Klasens, H. A , , Philips Res. Rep. 7 , 401 (1952). 2. Evrard. E., In “Adv. E.E.P.” Vol. 52. p. 133 (1979). 3. van Huyssteen, C. F., In “Adv. E.E.P.” Vol. 40A, p. 419 (1976). 4. Diakides, N . A., Proc. S.P.I.E. 42, 83 (1973). 5. Donofrio, R. L. and Rehkopf, C. H . , J . Electrochem. SOC.126, 1563 (1979).
This Page Intentionally Left Blank
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS,VOL. 64B
Study of ESBI and Sensitivity Characteristics of Ag-0-Cs Systems for Image Tubes M. SRINIVASAN, M. D. VAIDYA, D. R. KULKARNI, andT. B. BHATIA Bharat Electronics Limited, Pashan, Pune, India
INTRODUCTION The silver-oxygen-cesium (Ag-0-Cs) photocathode was the first composite surface developed almost at the same time by Koller' and Asao and Suzuki.2Ever since its development various methods for making Ag-0-Cs photocathodes have been reported by many authors. All these attempts were made solely to improve its response in the near-infrared region; but the results have not been very significant in this direction due to the complex nature of the cathode. Since the Ag-0-Cs photocathode has very low quantum efficiency (0.45%), continuous efforts to develop other photoemissive composite surfaces yielded alkali antimonide photocathodes viz. antimony-cesium (S.4 and S-1 I), antimony-potassium, sodium-cesium (S-20, S.25), and 111-V compounds. The multialkali antimonide photocathodes (S-20 and S.251, though possessing higher quantum efficiency (20%), have no useful response beyond 1.O pm. While 111-V photocathodes hold promise for the future, they are still not available freely commercially in image tubes. In this situation the Ag-0Cs photocathode still continues to possess an advantage in near-infrared region applications (e.g., using the YAG-Nd Laser or GaAs and for nearinfrared imaging). However, this presentation is limited to the study of Ag-0-Cs photocathodes pertaining to infrared image converter tubes produced in our laboratory. It is well known that the attractive feature of the Ag-0-Cs photocathode viz. its near-infrared response in turn gives rise to an associated high level of thermionic emission from the photocathode. In the case of image tubes, this manifests itself in the form of a disturbing glow in the absence of any incident illumination or radiation. The field performance of such an image tube for night vision applications is critically linked to its near-infrared response and to its level of background glow. Attempts were therefore 403 Copyright 0 1985 by Academic Press, Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-014724-6
404
M. SRINIVASAN E T AL.
made to study parameters such as conversion efficiency and background glow of a large number of tubes produced in our laboratory, and the results are presented in this article.
EXPERIMENTAL DETAILS The Ag-0-Cs photocathodes were processed in a conventional 6914 image converter tube geometry employing a combination sputter ion pump and sorption pump, following the technique reported by S ~ m m e r . ~ After initial bake-out and degassing of the tube envelope and other parts, a thin layer of silver is deposited on the photocathode substrate at a pressure of Torr to give a transmission of 4% for white light. This takes about 20 sec, corresponding to an average rate of deposition of 0.7 to 1.0 nm sec-I. The silver film is then oxidized in the conventional manner by subjecting it to an RF glow discharge in an oxygen atmosphere at 1.0 Torr. The oxidization is terminated when the light transmission rises to a level in excess of 90%. Silver is evaporated onto the oxidized layer to give a transmission of 60%. This surface is activated by cesium vapor at 150°C. The cesiation process is carried out by employing a voltage of the order of 5 kV and observing the output of the phosphor screen. The photocurrent is monitored simultaneously by connecting a microammeter in series with the power supply. The cesiation is stopped when the photocurrent reaches a peak value. Additional silver is then condensed onto this cesiated oxidized silver surface and a mild bake (heat treatment) is given which enhances the sensitivity of the photocathode. The characteristics of these image tubes such as luminous sensitivity, infrared sensitivity, spectral response, thermionic emission, equivalent screen background input, etc., were measured by following the standard procedures. The measurements employed a tungsten filament lamp operating at a color temperature 2870 K, an infrared transmitting visible abTABLEI Characteristics of image intensifiers Tube number E-974 E-892 E-706 E-868
Infrared sensitivity (PA)
Conversion efficiency
3.2 3.6 4.2 5. I
22.3 26.0 34.8 41.3
Thermionic emission current density (A cm-* at 28°C) x 3.9 1.8 2.3 4.8
Thermionic (lux)
work function (eV)
0.00076 0.00062 0.00054 0.00082
0.94 0.91 0.99 1.05
ESBI
Ag-0-Cs SYSTEMS
FOR IMAGE TUBES
405
sorption filter of Corning CS-7-56 type 2540, a Keithley Electrometer model 602, a Carl Zeiss PMQ I1 spectrophotometer and a photomultiplier tube having an S * 1 1 response. Their results are presented in Table I. The thermionic emission currents of the cathodes were measured at various temperatures in the temperature range of 25 to 60°C. Adequate precautions were taken to ensure accurate measurement of thermionic emission current taking ohmic leakage across the tube into consideration. The results are presented in the form of Richardson’s plots in Fig. 1.
-16
t
FIG. 1. Thermionic current versus temperature for the tested tubes.
406
M. SRlNlVASAN ET AL.
RESULTSAND DISCUSSION Equivalent Screen Background Input (ESBI)
The field performance of an image tube depends critically on its signalto-noise ratio, which in turn relates to the conversion efficiency and the background glow. The screen background glow, expressed in terms of the equivalent screen background input (ESBI), is that level of input luminance which would cause an increase of screen luminance equal to the background luminance. The screen background luminance can be caused by one or more of the following factors: (1) thermionic emission from the photocathode, (2) field emission, (3) electron scintillations, and (4) ion scintillations. All these sources may be controlled efficiently except the thermionic emission which is intrinsic to the photocathode. It should be appreciated that thermionic emission, and as a consequence ESBI, is a very critical function of temperature. The ESBI is measured by illuminating the photocathode by a known light flux (IR filtered) and using a photomultiplier tube to measure the ratio of the luminous intensities in the screen with and without the illumination incident on the photocathode. Then
ESBI
=
E~[(12 ]~)/(II- 12)]
where I , and I2 are the photomultiplier currents in the presence and the absence of the input illumination EL, respectively, and I3 is the photomultiplier dark current. The measurements were carried out in the temperature range 25 to 60°C. The results are presented in Fig. 2. It will be seen from the results (Table I) that the photocathodes fabricated in our laboratory exhibit a thermionic emission current of order A cm-2 at 28°C. The effective work function of these cathodes is in the range of 0.95 to 1.0 eV. It is also observed (Fig. 2) that in the temperature range 25 to 60°C the ESBI is increased about twofold for an increase of temperature of about 7°C. Conversion Efficiency
The infrared conversion efficiency of an image tube is defined as the ratio of the output luminance flux emitted by the tube screen to the input infrared flux incident on the photocathode: CI = F I / F ~ T
where F1 is the total available luminous flux emitted by the phosphor screen, F2 is the unfiltered flux from 2870 K light source incident on the
Ag-0-Cs
SYSTEMS FOR IMAGE TUBES
407
FIG.2. Equivalent screen background input (ESBI) versus temperature for the tested tubes.
408
M. SRINIVASAN ET AL.
WAVELENGTH (nml
FIG.3. Spectral responses of two tubes, A with low conversion efficiency and B with high conversion efficiency.
photocathode of the tube, and T is the filter factor of a calibrated Corning CS-7-56 type infrared filter ( 10.8%).4 The measurements carried out on a large number of image converter tubes produced in our laboratory indicate that there exist two types of
Ag-0-Cs
SYSTEMS FOR IMAGE TUBES
409
tubes, one having conversion efficiency around 35 to 40% and the other group having a conversion efficiency of 20 to 25%. The typical spectral response of these tubes is represented in Fig. 3. It may be clearly seen that tubes having higher values of conversion efficiency exhibit a higher long wavelength response (beyond 0.8 pm). They exhibit different photoelectric thresholds, However, the thermionic currents of these cathodes do not seem to show a large variation. Thus it is found that although the cathodes exhibit different photoelectric thresholds, they seem to possess the same thermionic work functions. The enhanced near-infrared sensitivity of the cathodes and the longer threshold can be attributed to the nature and condition of the original silver film. Among other conditions, the rate of evaporation of the parent silver film plays an important role in determining its structure. It is very well known that cathodes with fine grain structure exhibit a low work function. It is found that an evaporation rate about 1 .O nm sec-I gives rise to the cathodes with enhanced infrared sensitivity. ACKNOWLEDGMENTS The authors wish to thank the management of Bharat Electronics Ltd. for their permission to publish these results and Shri G . K. Bhide for many useful discussions.
REFERENCES I. 2. 3. 4.
Koller, L. R., Phys. Rev. 36, 1639 (1930). Asao, A. and Suzuki, M., Proc. Math. SOC.J p n . 12,247 (1930). Sommer, A. H., In “Photoemissive Materials”, Wiley, New York (1968). Eberhardt, E. H., Appl. Opt. 7, 2037 (1968).
This Page Intentionally Left Blank
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. 64B
A Near-Infrared Photocathode TAO CHAO-MING Institute of Electronics, Academia Sinica, Beding, China
INTRODUCTION Escher has described a transmissive NEA InP/Gao.47Ino.s3Asphotocathode with photoresponse in the region 0.9 to 1.65 pm.' In this article we describe a near-IR transmissive NEA GaP/Gql -,)In,P/Gao.47In0.aAs photocathode with photoresponse in the range 0.71 to 1.65 pm. THEPHOTOCATHODE
A graded baffle layer of Ga+,)InxP is grown on a substrate of GaP by VPE, MOCVD, or MBE. The value of x varies gradually from 0 to 1 . An emissive layer of G~.47In~.s3As is grown successively on the graded baffle layer. The lattice constant of the GqI-&,P varies from the lattice constant of GaP to the lattice constant of InP. The lattice constant of the Ga,,.471no.53As may be matched to the lattice constant of InP. In transmission, radiation passes through the GaP only if A > 0.55 pm. Actually in practice, the Ga(l-Jn,P layer absorbs incident radiation with A > 0.7 pm. It is also known that the bandgap of the Gao.47In0.53Asmaterial is 0.75 eV and its threshold is therefore 1.65 pm. The photoresponse photorange of the transmissive NEA GaP/Ga(l-,)In,P/Gao.47Ino.53As cathode is thus 0.7 to 1.65 pm. PERFORMANCE UNDER FULLMOONA N D CLEAR STARILLUMINATION Figure 1 gives the spectral dependence of full moon and clear star illumination. 2 From the literature, the overall quantum efficiency of the transmissive GaAs cathode is about that of the reflective InP/Gao,47In0.~3As cathode is about 8%.4 In this article, the overall quantum efficiency of the transmissive GaP/Ga,,471no.53As photocathode is taken to be 3%. The re41 I Copyright 0 1985 by Academic Press, Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-014724-6
412
TAO CHAO-MING
10-7
I
1
0.4
0.6
0.8
1.0
1.2
1.4
1.6 1.8
A (clrn) FIG.1.
Spectral distributions of starlight and full moonlight.
sponse in electrons sec-I cm-2 Sr-' pm-L was calculated for both cathodes and the results are shown in Fig. 2. It can be seen that the response of the new photocathode is approximately a factor of 3.4 greater than that of the GaAs cathode in starlight, but is approximately a factor 4.8 times smaller than that of the GaAs cathode in full moonlight. This is because infrared light is dominant in starlight, but visible radiation dominates moonlight.
CONCLUSION A near-infrared transmissive NEA GaP/Ga(l-,)In,P/Gao,471no.s3As photocathode with photoresponse in the range 0.7 to 1.65 pm is discussed. The photoemission of this photocathode is calculated and compared with the photoemission of transmissive GaAs photocathode under clear star and full moon illumination. The calculated results show that the photoemission of the transmissive GaP/Gql -x)In,P/Gao.471no.~3As photocathode is superior to the photoemission of the transmissive GaAs photocathode under clear star radiation. But, under full moon radiation, the photoemission of the transmissive GaAs photocathode is superior to that of the transmissive GaP/Ga(l-x)In,P/Gao.4,1no.s3As photocathode.
413
A NEAR-INFRARED PHOTOCATHODE 10" -
,
E,
0.4
0.6
0.8
1 .o
1.2
1.4
1.6
X (ccrn)
FIG.2. Photoemission of (a) GaAs and (b) G a P I G ~ , - , , I n , P I G ~ . , , I ~ . photocathodes ~~As in starlight and full moonlight.
414
TAO CHAO-MING
REFERENCES I. 2. 3. 4.
Escher, J. S., IEEE Truns. Electron Devices ED-28, 123 (1981). Richards, F. A., In "Adv. E.E.P." Vol. 28B, p. 661 (1%9). Csorba. 1. P., Appl. Opt. 18, 2440 (1979). Escher, J . S., IEEE Truns. Electron DPuices ED-25, 1347 (1978).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS. VOL. 648
A Magnetic Focus Electrostatic Deflection Compact Camera Tube M. KURASHIGE, S. OKAZAKI, and C. OGUSU
NHK Technical Research Laboratories, Tokyo, Japan
INTRODUCTION The basic electron optical merits of a magnetic focus, electrostatic deflection (MS) system, the “deflectron,” particularly its uniformity of resolution, were first reported by Schlesinger et al. under an assumed operational condition of uniform focusing and deflecting fields (hereafter referred to as ideal operational condition).l,* However, uniform fields cannot be created in actual devices of restricted axial length. A practical design method comforming to nonuniform fields has been investigated by Kubota et Despite those attempts, a universal design concept for MS-type camera tubes is still under investigation, because, except for a few approximated calculation^,^-^ the potential distribution inside the deflectron remains unsolved due to the difficulties of three-dimensional potential analysis with such complicated boundary conditions. In this article, the fundamental design concept which leads to high performance and tube compactness will be discussed mainly on an experimental basis. The article also describes the results obtained from the development of a compact 3-in.-diameter Saticon with almost the same high resolution as that of the 1-in.-diameter tube conventionally used in standard television system. A compact MS 1-in.-diameter Saticon with ultra-high resolution developed for HDTV (high definition television) is also described. ~
1
.
~
9
~
FOR OPTIMAL ELECTRON KEYDEFLECTRON PARAMETERS OPTICAL PERFORMANCE
Length-to-DiameterRatio Let 1, and d represent the deflectron length and diameter, respectively, and let y be ld/d. The deflection angle (with respect to the tube axis) can 415 Copyright 0 1985 by Academic Press. Inc. (London) Ltd. All rights of reproduction in any form reserved.
ISBN 0-12-014724-6
M. KURASHIGE ET A L .
416
be expressed by the following equation: 8d = tan(2/y)
(1)
Therefore, ed decreases as y increases. Since, as is usually the case, a larger deflection angle causes larger aberrations, it might be concluded that wide deflection and high performance are incompatible with each other. However, with the MS system, excellent uniformity of resolution can be expected because only a very small field curvature is generated by the electron beam deflection. This is seen in Fig. 1 which represents the ideal case.' The advantage originates from electrostatic deflection itself, in which the axial velocity of the electrons is almost unaffected by the deflection. We examined experimental pilot model tubes of 1 in. diameter to see whether this fundamental advantage can be retained with practical, nonuniform focusing and deflecting fields. Figure 2 shows the dependence of tube performance on the deflectron length Id for the real tube. In the figure, ARC,and AKorepresent amplitude response values at the picture center and at the picture comer, respectively, and Ad stands for raster distortion. The deflection voltage f?d required for the specified scan size is also shown. The pitch L, and the total twist angle ee of the deflectron pattern were fixed at 9 mm, and !No, respectively. The focusing magnetic field distribution was kept constant, and the collimation ratio (the ratio of the field mesh voltage V M to the average dc voltage VD applied to the deflectron electrodes) was optimized for each Id so as to minimize raster distortion.
A
Electrostatic deflection
Mignetic
Image PI ane
Pimary
I
vo
a 0
I
I
I
rimar)) trajectory: Axial distance Z
FIG. 1. Fundamental advantage of MS system.
AN
MS
COMPACT CAMERA TUBE
417
The figure shows that ARC,increases, reaches a maximum at y = 2.5, and then saturates as Id decreases, while A&, reaches a peak value at y = 3. These phenomena can be explained as follows. 1. The magnification of the MS system is determined by the focusing magnetic field distribution as well as by the electrostatic collimation lens. A strong collimation lens as is needed for a small l d contributes to demagnification of the beam spot. However, too strong a lens leads to the generation of large aberrations, particularly at the picture corner. The combined effect of these tendencies results in the observed variation of AR with I d . 2. Raster distortion increases proportionally to 0d3, as is generally the case, but the absolute value of distortion for the MS system is less than that which is usual in M M (magnetic focus, magnetic deflection) or SM (electrostatic focus, magnetic deflection) systems. 3. The deflection voltage ed is proportional to d2 under ideal operational condition,' however, the relationship does not apply to actual tubes which include nonuniform fields inside the deflectron space.
The Collimation Lens The collimation lens in the MS system is formed by the field mesh electrode and by the end of the deflectron whose circumference is divided into four arcs with different deflection voltages applied to each so that a
1'"
h
ed (a.u.1 I
I
2
3
'
0
4
I,/d
FIG.2. Variation of amplitude response AR, distortion Ad, and deflection voltage versus the ratio of length to diameter.
ed
M. KURASHIGE ET AL.
418
ov
-
FIG.3. Typical potential distribution inside the quadrupole lens.
quadrupole lens is formed. The typical potential distribution inside the quadrupole lens is shown in Fig. 3.6 The position of maximum potential, which is nearly equal to the field mesh voltage, is shifted away from the tube axis toward the deflectron element to which a positive deflection voltage is applied. The amount of the shift depends on the deflection voltage. Thus, the potential variation along the axial direction is reduced in the vicinity of the deflectron element where the deflected electron beam will pass. Since the strength of the electrostatic lens depends on the second derivative of the potential distribution with respect to the axial distance, the effective collimating action in the MS system is weaker than would be expected from the externally applied collimation voltage ratio. The optimal collimation voltage ratio depends on the deflectron length I,,, the spacing between the field mesh and the deflectron end, and the focusing magnetic field distribution. These factors should be combined so as to make field mesh voltage V, as high as possible, thus reducing the “beam bending effect.” Dejectron Pattern
Under ideal operational conditions, 8, controls the curvature of the electron trajectory and determines the deflection sensitivity because Oe coils or uncoils the electron path of which the projection on the x,y plane (normal to the axis) is a cycloid.2 However, experiments demonstrated
AN
MS
COMPACT CAMERA TUBE
,
--I,/d=4 1
90
N ,
0
419
N
I
180
Twist angle 8. of deflectron pattern (degrees)
FIG.4. Variation of amplitude response and distortion with respect to twist angle O,, with 1,ld as parameter.
that the deflection sensitivity is hardly affected by 8,. Unexpectedly, 0, has a close relationship to the uniformity of resolution as well as to the geometric distortion, as shown in Fig. 4. The optimal value of 8, was found to be nearly 90". The pattern pitch L, has almost the same effect on the tube characteristics as that of 0,. As shown in Fig. 5 , smaller L, results in better uniformity and less distortion.
0.2
0.4
Pitch of deflectron pattern Lp in Deflectron diameter d
FIG.5. Dependence of tube performance on the pitch of the deflectron pattern.
420
M. KURASHIGE ET A L .
Focusing Magnetic Field Distribution
The authors examined the optimal flux density distribution using a pilot #-in. MS camera tube and experimental coils. The experimental results shown in Fig. 6 indicate that the best compromise to obtain good and uniform resolution at the high field mesh voltage needed for small beam bending is case (2). Moreover, this also gives the smallest raster distortion. NEWCOMPACT CAMERA TUBES
The design concept thus determined was applied to a new f-in. compact camera tube. A schematic configuration of the tube is shown in Fig. 7. The following optimized specifications and manufacturing techniques were used: (1) the length to diameter ratio is 2.8, approximately equal to the optimum, (2) the deflectron pattern is 45 mm long with as short a pitch as possible within the producible limit, and with a twist of 90", (3) the collimation lens configuration has the field mesh voltage as high as possible and the deflectron dc voltage as low as possible so as to reduce both the beam bending effect and the deflection voltage requirement, (4) the tube-end is a construction with a face plate, a short glass cylinder for
Field mesh voltage
: VM(V)
FIG.6. Variation of amplitude response with respect to field mesh vo!tage for four focusing magnetic flux density distributions. Solid lines and broken lines indicate A R , and Aka, respectively. In the order from (a) to (d), the peak of the flux density distribution is shifted toward the cathode.
AN
MS
COMPACT CAMERA TUBE
42 1
FIG. 7. An experimental !-in. MS-type camera tube.
insulation, and a metal disk fitted with the field mesh, hermetically sealed together on the end of the glass envelope by two indium rings, ( 5 ) the electron gun is precision mounted inside the glass envelope, (6) the deflectron pattern is formed by precision photoetching of a vapor deposited chrome layer of thickness about 1500 A, (7) the Saticon photoconductive layer is of thickness 4 pm,and (8) a high-resolution and low-lag electron gun is employed. The amplitude response of the tube is shown in Fig. 8. In spite of the smaller tube length (85 mm long, shorter by 20 mm than conventional tubes), the amplitude response value was as high as 70% at 400 TV lines at
- Experimental 2/3inch compact MStype
(without dynamic focusing) MMtype (with dynamic focusing) ---2/3inch SMtype (with dynamic focusing) -X1 inch MM type
---. 2/3inch
50
50
500
Upper right
Upper left ( T Vlines)
Lower left
500
Center
( T V lines)
Test chart
: RCA-P200
50
Lower right
FIG.8. Comparison of the amplitude response of the experimental MS system with that of other conventional systems.
422
M. KURASHIGE ET A L .
FIG.9. The newly developed compact camera tube for high-definition TV.
the picture center. This is almost equal to that of conventional 1-in. tubes. Its resolution uniformity is excellent and an amplitude response value as high as 50% was obtained at the four picture corners. Additionally, small geometric distortion, less than 1.5 TV lines, was obtained. The total weight of the tube and coil assembly was reduced to 170 g compared to 250 g for conventional MM type 3-in. tubes. The same design concept was applied to a 1-in. MS system. The tube structure is shown in Fig. 9. By installing an ultra-high resolution electron gun which has been used in the DIS (diode gun impregnated cathode Saticon), a compact and low-voltage HDTV (High Definition Television) camera tube with almost the same high-resolution capability (about 40% at 800 TV line@ as that of DIS was obtained (Fig. 10). The tube-and-coil weight is reduced from 1060 to 270 g, the tube length from 160 to 105 mm,
projected in reverse
0)
c
0
a
2 73 0) c ._
E"
a
40 -
20 -
OO
400
I
800
L
1200
1600
T V lines
FIG.10. Comparative measurement of amplitude response for HDTV camera tubes. Solid line, MS type with mesh voltage VM = 800 V. Broken line, MM type with VM = 2 kV.
AN
MS
COMPACT CAMERA TUBE
423
FIG.I I . Part of high-definition (1 125 line) TV monitor picture obtained using the I-in. MS tube.
and the field mesh voltage from 2000 to 800 V. Figure 11 shows a monitor picture obtained using the I-in. MS camera tube.
CONCLUSION The authors have established a basic design concept for compact MStype camera tubes. The inherent high and uniform resolution characteristics of the MS electron optical system are compatible with the highresolution capability of the Saticon photoconductive layer. Using this design concept, the authors developed a %in. compact camera tube for standard broadcasting applications with almost the same resolution capability as that of conventional 1-in. tubes. Furthermore, a 1-in. compact MS tube with ultra-high resolution capability was developed for HDTV applications. ACKNOWLEDGMENTS
The authors wish to express their gratitude to Dr. I. Ohishi, Mr. N. Goto, and Dr. T. Kawamura for affording them the opportunity to conduct this study. Heartful thanks are also due to Mr.Y.Isozaki, Mr.Y.Nagashima, Mr.Y.Ikeda, and Mr. T. Asahina for their helpful advice and assistance.
424
M. KURASHIGE ET A L .
REFERENCES 1. Schlesinger, K., IEEE Trans. Electron Deuices ED-14, 163 (1967). 2. Ritz, E. F., Jr., IEEE Trans. Electron Devices ED-20, 1042 (1973).
3. Kubota, K., Tagawa, S., Sawai, M., Narnba, K. and Kakizaki, T., IEEE Trans. Consum. Electron. CE-24, 114 (1978). 4. Shino, T., Kadota, T., Tsukada, T., Katayama, Y. and Horii, Y., Tech. Rep. Insr. Teleu. Eng. Jpn. ED5, 88 (1981). 5. Hutter, R. G. E., IEEE Trans. Electron Devices ED-19, 731 (1972). 6. Kurashige, M. and Yamagishi, T., Tech. Rep. IECE, Jpn. ED82-33 (1982). 7. Oku, K. and Fukushima, M., Tech. Rep. IECE, Jpn. ED=-37 (1983). 8. Isozaki, Y., Kumada, J., Okude, S., Ogusu, C. and Goto, N., IEEE Trans. Electron Devices ED-28, 1500 (1981).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS.VOL. 648
Pyroelectric Vidicons for Submillimeter Wavelengths W. M. WREATHALL English Electric Valve Company Limited, Chelmsford, Essex, England
INTRODUCTION Until quite recently little use has been made of the submillimeter region of the electromagnetic spectrum, compared with optical and infrared wavelengths and microwaves. One reason for the neglect of the spectrum between 20 p m and 1 mm has been a lack of sufficiently powerful sources, but this has changed since there are now numerous lasers that fill the gap.' These offer potential applications for imaging, most obviously for mapping the mode patterns of the lasers, but also for radar modeling2u3and for detection of concealed object^.^ The pyroelectric vidicon, a television pick-up tube with bolometric action, can exploit these needs. Standard pyroelectric vidicons are optimized for operation with thermal radiation in the 8- to 14-pm band wavelengths. When used at much longer wavelengths these tubes have significant limitations. First, they are relatively insensitive. Second, particularly when used with coherent radiation, they produce spurious images. These arise principally from radiation that is reflected back to the target by the mesh and, to a lesser extent, by radiation that is reflected between the target and the surfaces of the faceplate. This article describes how these limitations have been overcome to produce tubes that are optimized for imaging using wavelengths around one-third of a millimeter, in particular 337 pm, the emission from an HCN laser. ASSESSMENT OF STANDARD TUBES The transmission and reflection spectra of sample components have been measured (as a function of frequency), at the National Physical Laboratory, by Fourier transform spectroscopy. 425 Copyright 8 198s by Academic Press. Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-014724-6
426
W. M. WREATHALL
The DTGS Target
The transmission curves (Fig. 1) relate to a standard target of deuterated triglycine sulfate (DTGS), 18 p m thick. The two curves correspond to mutually perpendicular polarizations of the incident radiation; the target is orientated for maximum dichroism. At 30 cm-I (333 pm) transmission ranges from 42 to 65%. Absorption will be less than the complement of the transmission, because of reflection losses. The Germanium Faceplate
The transmission spectrum (Fig. 2) of the germanium faceplate is dominated by channel fringes due to interference between waves partially
0 9--
08
--
Wavenumber (cm-')
FIG.I . Far-infrared transmission spectrum of standard DTGS target.
PYROELECTRIC VIDICONS
427
reflected at the faces. The transmission at 30 cm-I therefore depends upon the precise thickness of the faceplate, and also on the angle of incidence; however, it would typically be 25%. These measurements show that no more than 20% and possibly less than 10% of the incident radiation at 337 p m would be used by a standard tube. COMPONENT DEVELOPMENT Analysis of Target
Since DTGS is transparent in the far infrared, resort must be made to surface layers to provide absorption. The thermal capacity of these must be low compared with that of the DTGS. This rules out bulk absorbers Wavelength (,urn)
tom
0.40
600500
800
Lm
L
0.30-.
O.ZS-t
.-
In
.-
E In
0.20
..
0.15
-.
0.10
-'
e
c
z3
a"
t
Oo5
I
i1 . : ; ; ; ; ; ; ; ; 5
; : : : : ; : : : : I : : : : ; : : : : ; : : : : I ! : :
10
1s
P
25
30
Wavenurnber (crn-'1
FIG.2. Transmission spectrum of germanium faceplate.
35
t
428
W. M. WREATHALL
and favors metal films. The behavior of the latter has been in terms of the thickness of the dielectric and the conductivities of the metal films. These analyses show that total absorption is possible in a composite with the following properties: (1) a dielectric film with a thickness equal to an odd number of quarter wavelengths of the radiation; (2) the conductivity of the metallization on the front, radiation incident side should be 377 fl 0 - I ; (3) the rear side should be highly conductive. The front metallization serves conveniently as the signal plate, but the conductive rear side would short out the pattern of signal charge. This can be circumvented by making the surface in the form of an array of closely spaced metallic islands separated by insulating strips. Structures of this type are used as filters and beam-splitters in the far infrared. Analysis of their optical behavior7s8shows that maximum reflectivity is obtained when the pitch of the islands is equal to the wavelength of the radiation in the dielectric. Wavelength ( p m )
FIG.3. Transmission spectra of metallized DTGS target.
429
PYROELECTRIC VIDICONS
Target Fabrication
The DTGS target has a thickness rather less than 35 pm. Both surfaces have vacuum-deposited metal layers. The islands on the rear side are defined by evaporating aluminum through a mesh mask in contact with the DTGS. The meshes are made by plating on to a master pattern ruled on glass, with a pitch of 150 lines per inch. Spectral Measurements
Reflection and transmission spectra recorded for sample-metallized targets are shown in Figs. 3 and 4. The dependence of the spectra on polarization, due to birefringence in DTGS, is demonstrated by the pairs of spectra. Transmission and reflection coefficients are given in Table I for wavelengths of 337 and 400 pm; at the latter wavelength this particular target absorbs over 87% of the radiation incident on it, independently of polarization. Wavelength (pm)
4 : : : : : 5:
:::l::::f::::l::::I::::!:::: 10
15
20
25
30
r : : : :Lo: : : :LS: i : : :50: l
35
Wavenumber (cm-’1
FIG.4. Reflection spectra from front surface of metallized DTGS target.
430
W . M . WREATH ALL
TABLEI Transmission and reflection coefficients of metallized DTGS targets Wavelength
Transmission Reflection Front (parallel) Front (perpendicular) Rear
337 pm
400 pm
0.028
0.035
0.044
0.09 0.09 0.80
0.27 0.80
Similar high efficiency is obtained at 337 pm from a thinner target. The high reflectivity of the rear surface ensures that little of the small fraction of transmitted radiation can interfere with the primary image. Reflection by the vidicon mesh is therefore of no consequence. Wavelength ( p m ) XKK)
10 4
800
M)O
3w
400
500
250
200
a
E
e
-
05/ LOW HIGH
RESLUTION RESOLUTMN
:
2m-l
:
0 Scm-'
a
O0 23 I
f::.:
, !
5
: : :
I::::
10
i : : : : :
15
20
::::;::!!I:::' I : : ! ! I : ::: I : !:: I : : 25
30
35
LO
Wavenumber (cm-'1
FIG.5. Transmission spectrum of crystal quartz faceplate.
L5
50
55
1
PY ROELECTRIC VIDICONS
43 1
The Faceplate Crystal quartz is highly transparent in the far infrared. Fused quartz might be preferred on economic grounds but is substantially more absorbent. Faceplates are therefore made from crystal quartz, z-cut to avoid birefringence. The refractive index is quite high, 2.11, so that reflection losses and channel fringes are intrusive as demonstrated by the measured transmission spectrum (Fig. 5 ) . The reflections can be eliminated by using a quarter wavelength of material with a refractive index close to 1.45. It is impractical to achieve the required thickness (nearly 60 ym) by vacuum evaporation. However, polythene has a close match to the required refractive index9and can be obtained in the form of film. Such films have been successfully bonded to crystal quartz. Transmission and reflection spectra of a faceplate with polythene antireflection films are shown in Figs. 6 and 7. These show that the reflections have been almost completely eliminated at 337 pm.
Wavelength
LOW
RESOLUTION
2cm-1
HIGH
RESOLIJTON
o
(pm)
scm-1
FIG.6. Transmission spectrum of polythene bonded crystal quartz faceplate.
432
W . M. WREATHALL
Wavenumber (cm-'1
FIG.7. Reflection spectrum of polythene bonded crystal quartz faceplate
TUBESA N D PERFORMANCE Tubes of 25 mm incorporating 20-mm-diameter metallized targets as described above have been assembled and pumped. Indium rings were used to seal the quartz faceplates to the glass bulb. The tubes were run in a pyroelectric camera operating at CCIR TV scan standards. In focus, the metal islands on the scanned side of the target were well defined and it was found preferable to defocus the reading beam to avoid large signal spikes arising from instantaneous discharge of the islands.
PYROELECTRIC VIDICONS
433
Modulation Transfer Function
MTFs, measured in the near infrared, using a bar pattern in front of a radiant source, imaged by a germanium lens at fll.4, were as follows: 50 TV lines per picture height (1.4 lp mm-I) 0.49; 75 TV lines per picture height (2.1 lp mm-I) 0.20; 100 TV lines per picture height (2.8 lp mm-I) 0.10. These measurements were made with the beam defocused to eliminate modulation by the island structure. The performance achieved is well matched to the diffraction limit of lenses in the far infrared; for example, the cut-off frequency at f/2 is less than 1.5 Ip mm-' at a wavelength of 337 pm. Responsivitj
Responsivity, measured using an HCN laser as a source of known power at 337 pm, exceeds 4 pA W-'. The pyroelectric efficiency, characterized by the responsivity times thickness product, is somewhat greater than that of standard targets operated in the 8- to 14-pm band. APPLICATION TO RADAR MODELING Because radars use coherent radiation, the returns are confounded by interference between reflections from various parts of the target and are sensitive to changes of orientation and of polarization. Any object therefore has a set of radar signatures that can only be compiled by experiment, and, since radars and targets are both unwieldy and expensive, the experiments are most economically conducted with models. As both model and wavelength have to be scaled equally, optical wavelengths demand impossibly precise modeling, hence the preference for submillimeter wavelengths. A typical model is illustrated in Fig. 8a. When this vehicle is illuminated by radiation from an HCN laser, with a wavelength of 337 pm, a typical return is as shown in Fig. 8b. This was recorded from a pyroelectric camera fitted with the long-wave length tube and with a 100-mm focal length lens made of TPX, a plastic that has a low loss in the far infrared. The object is unrecognizable, although with experience it might be possible to memorize the signature. In order to identify the reflecting areas, it is desirable to superimpose them on an outline of the vehicle. As the TPX lens and quartz faceplate are both transparent to visible light, this can be done by placing a lamp behind the model, with the result shown in Fig. 8c. It is found that large changes in signature can result from small shifts of the target vehicle.
434
W . M . WREATHALL
FIG.8. (a) Model road roller; (b) returns from (a); (c) returns against silhouette.
CONCLUSION Pyroelectric vidicons have been designed and manufactured which are highly efficient in the far infrared and demonstrate useful imagery with HCN laser illumination. The same techniques could be applied to produce efficient tubes for any part of the submillimeter spectrum. ACKNOWLEDGMENTS Acknowledgment is due to E.M.I. Ltd. and the National Physical Laboratory (NPL) for the use of HCN lasers and to NPL also for spectral measurements. This work was supported by the Ministry of Defence, Royal Signals and Radar Establishment through the Procurement Executive. D.C.V.D.
REFERENCES 1. Waniek, R. W., Laser Focus 19, 79 (1983). 2. Cram, L. A. and Woolcock, S. C., Radio & Electron. Eng. 49, 381 (1979).
PYROELECTRIC VIDICONS
3. 4. 5. 6. 7. 8. 9.
Waldman, J., Proc. S. P. I . E. 259, 152 (1980). Barker, D.H., Proc. S.P. I. E . 67, 27 (1975). Hilsum, C., J . Opr. SOC.Am. 44, 188 (1954). Stilberg, P. A., J. Opt. SOC.Am. 47, 575 (1957). Ulrich, R., Infrared Phys. 7, 37 (1%7). Durschlag, M. S. and DeTemple, T. A., Appl. Opr 20, 1245 (1981). Birch, J. R., Infrared Phys. 21, 225 (1981).
435
This Page Intentionally Left Blank
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. MB
An Amorphous Silicon Vidicon Tube B. L. JONES, J. BURRAGE, and R. HOLTOM English Electric Valve Company Limited. Chelmsford, Essex, Englond
INTRODUCTION Glow discharge plasma deposited hydrogenated amorphous silicon (a-Si :H), with its low density of gap states' (-lot7 cm-3 eV-I), is fast becoming a useful material for low-cost solar cells,2thin film transistor^,^ and practical imaging device^.^ In particular the high resistivity ( 1OI2 R-cm) and optical absorption coefficient in the visible region (105 cm-I at 520 nm) combined with a large photoconductive response recommends a-Si : H to be potentially useful as the target for a vidicon camera tube. In this article the characteristics of such a tube are presented and it is believed that these are the first reported results of such a tube fabricated in this country. EXPERIMENTAL TECHNIQUE The a-Si : H targets used for this work were prepared by RF glow discharge decomposition of SiH4, using a capacitively coupled reactor system (Fig. 1). Typical deposition conditions are listed in Table I. The structure of the a-Si:H target adopted for this investigation (Fig. 2a) consists of a photoconductive layer of intrinsic material of an amorphous silicon matrix with hydrogen atoms (-2 p m thick) bounded by a holeblocking layer of Si-N on the input side (-300 8, thick) to prevent hole penetration from the signal plate to the photoconductive layer and a beam-blocking layer of similar material and thickness on the electron gun side to prevent electron penetration to the photoconductive layer. This target type produces a tube similar to the conventional Sb& autotarget tube with the signal output dependent on target voltage. The layers were deposited on 26-mm-diameter glass faceplates coated with a transparent SnO2 electrode (-0.2 p m thick). Measurements of the photoelectric properties of these targets were carried out in individual sealed tubes with an electron gun (Fig. 2b). The tin oxide was biased 437 Copyright 8 1985 by Academic Press. Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12014724-6
438
B. L. JONES ET AL.
FIG. I . Glow discharge plasma deposition system.
positively with respect to the scanning electron beam and the scanned area of the target was 9.5 x 12.7 mm2. The characteristics of this tube were measured as functions of illumination and target voltage.
RESULTSAND DISCUSSION Distribution of Hydrogen in a-Si :H
The distribution and type of bonded hydrogen in hydrogenated amorphous films determines the quality of the layers and hence the vidicon tube performance. Therefore an intensive investigation into the hydrogen content has been undertaken. Hydrogen incorporated into these films compensates the dangling Si bonds and thus reduces the density of gap states.2 The bonding configuration of the hydrogenated silicon, i.e., monohydride (SiEH), dihydride (Si=HZ), or trihydride (Si-H3) can be determined by infrared spectra and is a strong function of deposition
TABLEI Deposition parameters RF power density Pressure Substrate temperature Flow rate Deposition rate
0.05 W cm-2 500 mT 250°C 50 sec cm-1 0.04 pm min-1
AN AMORPHOUS SILICON VIDICON TUBE
439
a
Lknl
VOEO
b
I
14K?1
YEW
LUIfcaJ3
&x€Lcumn ’
\
LlHoCf
YyUloI
FIG.2. a-Si : H vidicon target structure (a) and tube structure (b).
temperature. For high substrate temperatures (2200°C)the hydrogen present is of the monohydride form. Low substrate temperature films ( &, T& ;T& and T; are the Maxwell relaxation times of crystalline and amorphous semiconductor, respectively; V , is the interface potential, Ve is the external applied voltage U e ; other symbols are as in Figs. I and 2.
tion has the form
where xois the interface coordinate, x2 is the coordinate of the amorphous semiconductor surface, VZis the potential in the plane given by the coordinate x2, V , and VSoare the interface potentials with and without external applied voltage U B ,L = (ca/e2NF)’”and E* is the permittivity of the amorphous semiconductor.
A PHYSICAL MODEL OF HETEROSTRUCTURE TARGETS
45 1
Figure 4 shows the calculated results of the potential distributions according to this model for a CdSe-As2Se3vidicon target having the following parameters: us= -10l6 m-2, N D = 5 x loz1m-3, N F = 1024 m-3 eV-', m. E~ = E~ = 10-lo F m-l, layer thickness da = dc = The dependence of the interface potential V , on the magnitude of the applied target voltage U Bis shown in Fig. 4a. The state of complete CdSe layer depletion due to the interface charge is indicated by an arrow and corresponds to an applied voltage of about 10 V. This is in good agreement with experimental results in practice as it is known that, for target voltages ranging from 8 to 13 V, the currentholtage characteristics become saturated and the quantum yield in the prevailing part of the spectral response approaches unity: the total depletion of the crystalline layer is a necessary condition for this state. Figure 4b shows the target potential distribution as a function of coordinate x and Fig. 4c shows the corresponding dependence of the electric field strength E = E ( x ) . Dark Current The generation of carriers from surface states is the most important process responsible for the formation of the dark current in heterogeneous semiconductor junctions having band gaps Eg 2 1.5 eV.2 This mechanism is modified in crystalline-amorphous semiconductor structures by the electron and hole generation, G, from the states at the interface of both semiconductors; here the rate-limiting process is electron generation from states in the amorphous semiconductor near the Fermi level to those at the top of the conduction band in the crystalline semiconductor. It may be shown, therefore, that G is given by G=
I" E"
vSnnn,(E)dE= v S , N , ~ ( ~ ~ - ~ ~ ) ' ~ ~ N(2) F~~T
where v is the thermal electron velocity, Sn is the electron capture cross section of the surface states, n is the electron concentration at the edge of the conduction band, n,(E) is the density of trapped holes of energy E , Nc is the effective concentration of states in the conduction band of the crystalline semiconductor, and NF is the density of states at the Fermi level of the amorphous semiconductor. Under the assumption that the carrier drift may be neglected (vdift S v), we obtain for the dark current density Jd = Ja-c
- Jc-a
=
eG = 2 e v S , N , N ~ k T e - ~ b ' ~ ~( e1)~ ~ B (3)
where AEb = AEba - eUa (Fig. I), AEba is the magnitude of the potential
452
M . JEDLICKA A N D F . SCHAUER
XI
1
a
t
gi
",I 6t
0
1
2
3
4
-
1UI.
5
6
7
8
9
K
l
[VI
t
0
1
xlml FIG.4. Potential and electric field strength distribution in CdSe-As2Se3vidicon target. (a) Dependence of the interface potential V. on the external applied voltage U,. The arrow denotes the completely depleted CdSe layer. (b) Distribution of the potential V in the target with external applied voltage CJB as the parameter. (c) Electric field strength distribution E = e(x) in the target for various applied voltages UB.
A PHYSICAL MODEL OF HETEROSTRUCTURE TARGETS
0
-
1
453
2.16’
xlml
FIG.4c. See legend on facing page.
barrier for external applied voltage UB= 0 V. The voltage Ua is given by the potential difference V , - V Z and can be calculated by solving the potential distribution for the steady-state case of the target. The method of calculation is given el~ewhere.~ The resulting relation is
where N D is the shallow donor concentration in the crystalline semiconductor, VD = AED/eis the diffusion potential, AEDis the difference of the Fermi level positions in both semiconductors before the formation of the junction, N F is the density of the states localized near the Fermi level in the amorphous semiconductor, and cCis the permittivity of the crystalline and ea that of the amorphous semiconductor. The voltage Ua is a very important quantity as it determines the lowering of the potential barrier with rising external voltage U B. Figure 5 shows the current/voltage characteristics of the dark current in a CdSe-As2Se3 heterojunction target at a temperature of 300 K for various values of the
FIG.5. The model currentlvoltage characteristics of the heterojunction target for various values of the ratio NDINF:(a) = 0 eV, (b) = 5 x eV, (c) = 5 x lo-’ eV, (d) = 5 x 10-4 eV, (e) = 5 x lo-’ eV, (f)= 5 x lo-* eV, (9) = 5 x 10-I eV. Donor energy ED = 0.3 eV.
-yrv1 FIG.6. The model currentholtage characteristics of the heterojunction target for various values of the interface density ub:(a) = 0, (b) = -3 x loi4,(c) = -lot5, (d) = -3 x 1015, (e) = -5 x 10” m-2.
A PHYSlCAL MODEL OF HETEROSTRUCTURE TARGETS
455
ratio NDINF, The deviation from the saturated state increases with rising ratio NDINF.A family of currentholtage characteristics for the same target is shown in Fig. 6 for various values of the interface state density us.The higher the charge corresponding to these states, the lower the relative magnitude of the dark current. Figure 7 gives a group of currentholtage characteristics of the dark current in a true experimental CdSe-As2Se3 heterojunction measured at various temperatures, the magnitude of the junction area being 10 mm2 and the thicknesses of both layers dc and da being 1 pm.It is evident that these experimental characteristics follow the shape of the calculated curves in Figs. 5 and 6. The influence of the external voltage U Bon the lowering of the potential
16”
16’
- [vl
6
loa
XI1
XI2
UI
FIG.7. The measured curredvoltage characteristics of a real CdSe-As2Se3heterojunction. Temperature T: (a) = 340 K, (b) = 320 K, (c) = 310 K, (d) = 300 K, (e) = 290 K.
456
M . JEDLICKA A N D F. SCHAUER
01
lo-'
6'
16'
loo
10'
J
-lJ,IVl FIG.8. The dependence of the dark current activation energy AEbon the external applied voltage Us;circles are measured points, the continuous line is the calculated result.
barrier is shown in Fig. 8. The curve was calculated using the following parameters: AEbo= 0.68 eV, NDINF= 0.018 eV, us= -4.5 x lOI4 m-2, the diffusion potential VD = -0.026 V. The points along the curve are calculated from the dark current values given in Fig. 7 by means of Eq. (4) and the relation AEb = AEba - eUa. It is known that in case of the CdSe-AszSe3 target, negative interface charge can be introduced and its magnitude influenced by baking the CdSe layer in air whereby the layer surface is strongly affected by oxygen. Figure 9 shows the dependence of the dark current in this target on the time of air-baking its CdSe layer at 400°C.6 First, the magnitude of the negative charge at the interface rises with rising baking time; as a result, the height of the barrier AEb increases and the dark current decreases. After a certain baking time, saturation of the dark current occurs. This may be explained by the possible automatic reduction of the processes occurring during introduction of interface charge by means of sorption or other processes connected with the presence of oxygen.
PHOTOCURRENTS As a result of the absorption of photons in the crystalline semiconductor, electrons and holes are generated. Their transport through the both crystalline and amorphous semiconductors give rise to a photoelectric current. For efficient target operation the lowest losses of carriers by recombination and trapping are required. This condition is fulfilled in the crystalline part of the target if the semiconductor is completely depleted. If this is the case, the low mobility of holes in the amorphous target layer
A PHYSICAL MODEL OF HETEROSTRUCTURE TARGETS
457
FIG.9. Dependence of the dark current Idof the CdSe-As2Se3junction on the air-baking time of the CdSe layer.
due to volume trapping states in the mobility gap becomes the factor limiting the quantum yield.’ The energy position of the respective traps determines the wide range of trapping and release times, giving rise to dispersive transport. A dynamic model of an amorphous semiconductor was calculated to establish the transit time of holes in the amorphous semicondu~tor.~ This model enables the determination, for typical conditions, of the time dependence of the decay of the photoelectric signal created by a light pulse absorbed in the photoconductive crystalline semiconductor. The dependence of the normalized signal current Zs/Zsoon the normalized time is U ~po shown in Fig. 10. The normalized time is tlto where to = ( d a ) 2 / p ~and is the free hole mobility. The signal current in the external circuit is l,(r) = (epo/da)
I”’
E&,
xo
t)dx
(5)
In Fig. 10, three typical regions of the time dependence are observed: the region for which I , z - O . ~ , the region for which I, t-*, and a tail region caused by carrier release from deep traps. The arrow in Fig. 10 indicates a break in the curve corresponding to the transit time of the carrier majority. The results of the theoretical analysis are in accordance with the experi-
-
-
458
M. JEDLICKA A N D F. SCHAUER
FIG. 10. Time dependence of the signal current I , .
mental results. The time dependence of the relative signal current I , in the heterojunction CdSe-AszSe3 is that shown in Fig. 11. The curves measured at various temperatures show the knees indicated by the dashed lines which correspond to the values of 2 , . The examined target had SnOz and Au contacts on the crystalline and amorphous surface, respectively, and was tested in a sandwich arrangement whereby no electron beam was needed. Thereby, troubles connected with lag due to electron beam scanning were eliminated. The hole transit times in the CdSe-AszSe3 target are in the range from a few tenths of milliseconds to a few milliseconds and are temperature dependent. This is in good agreement with the measured results on CdSe-AszSe3 vidicons; the lag of the signal current in these tubes is due to the electron beam scanning process whereas the photoconductive and transport lag component is usually negligible. CONCLUSION The physical model of a camera tube heterojunction target composed of crystalline-amorphous semiconductors enables prediction of some important target properties such as the dependence of potential and electric field strengths, dark current, and lag due to the transport of photoexcited carriers. It also enables the spectral quantum efficiency to be predicted from the stage of target crystalline layer depletion. A CdSe-AszSe3 target was chosen for comparison with the experiment; however, the model has also proved capable of the study of a target structure composed of crystalline and hydrogenized amorphous silicon and apparently it could be used generally.
A PHYSICAL MODEL OF HETEROSTRUCTURE TARGETS
459
FIG. 1 I . The measured signal current response I , = I&) to the light pulse excitation in CdSe-As2Se3 heterojunction for various temperatures: (a) = 360 K, (b) = 340 K, (c) = 320 K, (d) = 300 K, (e) = 290 K .
A note should be emphasized in conclusion: it has been assumed that all the material and geometric properties of the target are identical in all planes which are parallel to the target surface and perpendicular to the x axis direction. As a matter of fact, this is not true in practice: e.g., the thickness of the layers dc and da are not exactly the same over the whole target area. This situation causes local variations of the potential and electric field distribution with consequences related to the current/voltage characteristics. The interface situation is certainly most critical. For technological reasons, inhomogeneity of the interface built-in charge occurs quite readily. Figure 12 shows the effect of the interface state density as on the potential distribution. Changes of the interface density cause not only a change in the potential distribution (effecting the currentholtage characteristic of the dark current) but also a change in the width, b, of the
C
A
FIG.12. The influence of increasing charge interface state density on the potential distribution V = V ( x ) in the junction for an external applied voltage Ue = 0 (V, No are constant, brlI < Ios21< I(+&
loo
10'
5.K)'
UD[VI
FIG.13. The currentholtage characteristics of an otherwise uniformly illuminated CdSeAs2Se3camera tube target containing a white spot; Id is the average target dark current, I, is the average target signal current, Ids is the spot dark current, and I,, is the spot signal current.
A PHYSICAL MODEL OF HETEROSTRUCTURE TARGETS
46 1
depleted region (thus influencing the quantum efficiency of the photoelectric conversion particularly in the shortwave region). Figure 13 shows the current/voltage characteristics of the signal current from a uniformly illuminated image and from a white spot on this image whose size is 0.7% of the picture height. The tube was a CdSe-AszSe3 vidicon. This is an example of area inhomogeneity probably due to local changes in the interface state density rather than to a possible change in the N D / N Fratio. The curves were measured on a standard camera tube testing equipment by means of an oscilloscope with discrete line selection facility. The large dark current of the spot limits the magnitude of the applied target voltage in the spot area. The physical model does not claim to explain how to achieve adequate target area homogeneity but points to its importance.
REFERENCES I . Weimer, P. K., RCA Reu. 36, 385 (1975). 2. Rhoderick, E. H., I n “Metal-Semiconductor Contacts,” Oxford University Press (Clarendon), London and New York (1980). 3. Brodsky, M. H. and Dohler, G. H., CRC Crit. Rev. Solid State Sci. 5, 591 (1975). 4. Schauer, F., JedliEka, M. and Polcer, J., Phys. Status SolidiA 70, 755 (1982). 5. Schauer, F., Jedlitka, M. and Macku, J., Phys. Status SolidiA 71, 335 (1982). 6 . Schauer, F . , HeZa, S. , Jedlitka, M . , Kosek, F. and Cimpl, Z., Phys. Starus SolidiA 61, 349 (1980). 7. Nagels, B . , In “Amorphous Semiconductors” ed. by M. H. Brodsky, “Topics in Applied Physics”, Vol. 36. Springer-Verlag, Berlin and New York (1979).
This Page Intentionally Left Blank
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS. VOL. 64B
A Quality Figure for the Emission System in Camera Tubes SHEN CHING-KAI Zhejiang University, Hangzhou. China
FENG CHIH-TAO and TUNG KUN-LIN Kunming Institute of Physics, Kunrning, China
and FANG ER-LUN North Industries Corporation, Beijing, China
Computation and design of the electron gun in camera tubes can now be carried out to quite good accuracy with the aid of a digital computer. However, a criterion for performance evaluation of the emission system is still lacking. If a figure of merit can be defined, the quality of an emission system can be judged at the design stage. A good emission system for a camera tube must satisfy the following basic requirements: (1) adequate beam current, (2) high resolution, and (3) low discharge lag.
Beam Current The beam current should be large enough to neutralize the positive charge in the high lights of the signal. On the other hand, it should not be so large as to introduce a high noise level and thus lower the signal-tonoise ratio; meanwhile space charge effects will also cause broadening of the beam and low resolution. For vidicons in general, the required beam current under operating condition is about 1 PA.
Resolution From the point of view of electron optics, the beam cross section at some point inside the system acts as the object of the focusing system, 463 Copyright 0 1985 by Academic Press, Inc. (London) Ltd. All rights of reproduclion in any form reserved. ISBN 0-12-0147244
464
SHEN CHING-KAI ET A L .
and is imaged onto the target. The smaller the object size, the higher the resolution; but what is the object of the focusing system? To this elementary question only vague answers can be found in the literature. Some say it is the cross-over, while others say it is the anode aperture. As a matter of fact, both are incorrect. The anode aperture is about 25 to 40 p m in diameter, which is much smaller than the size of the cross-over. Electron current from the cross-over cannot completely pass through the anode aperture. Hence only a part of the cross-over acts as the effective object. The conception of taking the anode aperture as the object is wrong. The focusing field is always adjusted so as to give best picture quality (best resolution). That is to say, the radial current-density distribution of the object corresponds to the sharpest curve shape. Obviously this is the case at the cross-over. If we keep the anode voltage constant, and adjust the magnetic focusing field to get the best image, in so doing we are really projecting the current-density distribution of the object at different positions on to the target. As the magnetic field is increased, the object-image distance is shortened. The object moves in the direction of the target. When the object moves to the position of the cross-over, the electron beam passing through the anode aperture has the sharpest current-density distribution in the radial direction. If the magnetic field is further increased, the anode aperture may be imaged onto the target. However, the resolution in this case becomes lower. As to the question of which part of the cross-over acts as the effective object, we suggest the following answer. Suppose several principal rays start from the cross-over, one of which just touches the edge of the anode aperture, then the radial coordinate of the starting point of that principal ray indicates the effective radius of the object. Computation can be carried out by solving the following equation of motion:
z.. = -e- d V
m az’
e dV
e
m dz
2m
where B is the magnetic focusing field, and ro is the initial value of the r-coordinate. As shown by computation, for ordinary electron guns in camera tubes with anode apertures 30 to 40 p m in diameter, the value of ro is about 10 pm. It should be noted, however, that the above definition of the object is used to define only the object size. As to the radial currentdensity distribution, the values at that part of the cross-over do not correspond to those obtained after the beam passes through the anode aperture. This is because the wall of the anode aperture intercepts a part of the beam-edge current.
EMISSION SYSTEM IN CAMERA TUBES
465
Discharge Lag Coulomb force interaction between electrons in the neighborhood of the cross-over causes broadening of the axial energy spread and increase of the discharge lag. Several analytical treatments have been published, such as those by K. H. Loeffler,' W. Knauer,2and E. de Chambost and C. authors deduced expressions for mean-square axial enH e n n i ~ nThese .~ ergy spread A E 2 . Owing to differences in physical models their results do not agree. However, there is an important result in common, i.e., all these expressions include three important parameters: the cross-over current density ( j J , the cross-over diameter (d,) and the inclination of the electron ray at the cross-over (a).There are only slight differences in their results: Loeffler': AE2 a j,dc/a2 Knauer2: AE2 jcd,la (2) Chambost and Hennion3: ajc3'4dcl~1/2 In the expressions of Loeffler and Knauer other factors are also included, but these factors are not significantly affected by variation of cross-over parameters. As a first-order approximation, we may neglect these factors. We adopt the result given by Knauer, replace the crossover diameter by its radius r, , and use the rms value of ao.In Knauer's analysis, the current density was assumed to be uniform along the crossover. This is not so in practical cases. We take the central current density as j , , and define r, as the radius where the current density drops to lle of its maximum value. For a Maxwell distribution, the mean axial energy spread is directmroportional to the beam temperature while = (AE2/2)'/2.Hence AE2 may be used to measure the relative value of discharge lag. We define the lag coefficient as F I where
FI = jcrc/ao
(3)
Both the object radius ro and the lag coefficient FI vary with grid voltage. Comparisons should be made at a specified value of beam current, say, l PA. A good emission system should have a low value of ro and a low value of F I . Computation shows that high resolution and low lag are incompatible. However, it is possible to make a trade-off by proper design. It is suggested that I / ( r o F Ibe ) taken as the quality figure for emission systems in camera tubes. As emphasis may be laid on either of the two parameters according to the different requirements, two weighing factors, rn and n are introduced, so that the quality figure becomes I l(rornF,").
466
SHEN CHING-KAI E T A L .
0.08 0.06 0.04
0.02 60
40
20
0
FIG.1 . Typical computed curves: (a) beam current versus grid voltage, (b) object radius versus grid voltage, and (c) lag coefficient versus grid voltage.
EMISSION SYSTEM IN CAMERA TUBES
467
A computer program has been written in which the physical model is essentially that proposed by C. Webe# and modified by K. N i n ~ m i y a . ~ Computed results for the modulation characteristics of two types of camera tubes agree well with the experimental data. More than 20 types of emission systems with different electrode geometries have been computed. An interesting result is noted. As the grid voltage is lowered so that the beam current is reduced, the curve of object radius versus grid voltage passes through a flat minimum (Fig. lb). This result has been verified experirnentally.'jAs the beam current was varied by raising the grid voltage from the cut-off value, and the modulation at 400 TV lines was observed, a flat maximum appeared in the modulation grid voltage curve. This fact confirms our physical model of the object radius. ACKNOWLEDGMENTS
The authors wish to thank the directors of The Kunming Institute of Physics and the Computer TQ-16 team for their support and assistance in this work.
REFERENCES LoefHer, K . H., Z . Angew. Phys. 27, 145 (1969). Knauer, W., Optik 54, 211 (1979). de Chambost, E. and Hennion, C., Optik 55, 357 (1980). Weber, C., Philips Res. Rep., Suppl. 6, (1967). 5 . Ninomiya, K., Trans. Inst. Electron. Commun. Eng. Jpn. 54-B,490 (1971). 6. Cheng Yutsai, private communication.
1. 2. 3. 4.
This Page Intentionally Left Blank
ADVANCES I N ELECTRONICS A N D ELECTRON PHYSICS, VOL. M B
Recent Developments in Real-Time Image Processing R. AUBERT, B. BUHLER, and W. GEBAUER Contraves AG, Zurich, Switzerland
INTRODUCTION Our principal application of real-time image processing is the detection, the measurement of angular position, and the tracking of targets such as aircraft and missiles for the purpose of trajectory measurement on test ranges and for fire control in land- or sea-based systems. The video tracker (FTT) receives as input the video signal of a TV or infrared camera (FLIR) fixed on a tracking mount. The task of the video tracker is to detect a possible target in the camera field of view in every frame of the video signal and to measure its angular position relative to the center of the field of view. With this information the system is capable of moving the tracking mount in such a way that the target, with small deviations and despite target manoeuvres, remains virtually in the center of the image. The measured deviation, added to the angular position of the mount, is a high-precision measurement of the target position in azimuth and elevation. The first use and development of such real-time video trackers at Contraves goes back to the 1970s. These first trackers were based on analog video technology and on simple contrast threshold methods for target/ background discrimination. They proved to work very well in favorable tracking conditions, i.e., homogeneous background and relatively high target contrast. Processing of the binary image was later digitized, and many new features were added to the tracker in order to handle special situations in target tracking, some of which arose because of the poor performance of these contrast trackers in high-clutter background. Every new feature and every modification made it necessary to develop new hardware, which for operation at video data rates is not trivial. With this experience, and with the tremendous advances in high-integration digital signal processing, microprocessor technology and digital signal and image processing theory in the past decade, we found it necessary to begin with a totally new video 469 Copyright 0 1985 by Acsdemic Press. Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-0147?4-6
470
R. AUBERT ET AL.
tracker development. The new project “FTT” (FLIRITV-Tracker) was started in 1979/1980. We have also studied and simulated many different tracking algorithms and have applied them to synthetic and real digitized video sequences. From those results and from studies of different potential user requirements1V2we have derived a very flexible, modular hardware and software structure for the implementation of a real-time video tracker. Finally we have realized a prototype FTT including a target detection unit (TDU) and a target tracking unit (TEU). First trials were run by closing the gate loop of the FTT prototype on FLIR and TV images in real time. What remains to be done is integration of the FTT into the loop of a tracking system, where the FTT output controls the tracking mount. IMAGE SEQUENCE ANALYSIS A N D SIMULATIONS Figure 1 shows the tools used for image analysis and tracker simulations. It is composed of a VAXl1/780 computer system with peripherals, and, in principal, a video image input/output unit formed by the De Anza IP 6400 image digitizer, storage, processing, and display device. With this system real-world TV or IR images and image sequences can be analyzed. Synthetic image sequences can be generated with the measured parameters, and both the real and synthetic image sequences can be fed to computer-simulated image processing algorithms for detection and tracking purposes. This system allowed us to evaluate many different algorithms and to optimize them for a wide variety of tracking situations without having to worry about the problems of hardware implementation from the first moment. Figure 2a and b shows the digitized image and the intermediate result of a simulated segmentation algorithm (without noise cleaning operation). A N D CONCEPT THEFTT DESIGNOBJECTIVES
The FTT was designed with the following objectives: (1) high tracking reliability (low track loss probability) even in high background clutter conditions and during target obscurations, (2) high tracking accuracy despite rapidly changing target aspect (little or no aimpoint wander), (3) fully automatic operation capability (intelligent strategies controlled by host computer, no direct operator intervention), (4) automatic target detection/ designation capability, ( 5 ) built-in test equipment (BITE) for detection and location of system failures, (6) dual target tracking capability, and (7) modular, flexible, and expandable structure (accommodation of special user requirements, adaptation to different signal sources, interchangeable image processing modules). The key to the solution of these requirements
47 1
REAL-TIME IMAGE PROCESSING
I I
GRUNDIO VIDEO-RECORD.
CVS 517
TIMEBASE CORR
I
T
I
I
DEC
HOST COMPUTER
I
C
-
NET
O
A
X
)
-
PDP 11I 3 4 FRONTEND
IMAGE MEMORY OE
I P 6400
0
I
TV -MONITOR
COMPUTER SYSTEM
I
I
VIDEO
PICTURE
INPUT I OUTPUT
FIG.1 . The image processing system.
lies partially in the image processing algorithms and partially in the hardware/software structure of the FTT. The FTT is considered as a subsystem. The FTT comprises an (expandable) number of compatible subunits or modules, some of which are necessary in every FTT realization, and others which are optional or interchangeable. The FTT configuration is set by the overall system requirements. DESCRIPTION FTT PROTOTYPE
Figure 3 shows the hardware structure of the prototype FTT. The input signal is a CCIR standard video signal. The FTT synchronizes the image processing of this input signal by the timing generator module. The ADC module converts the entire image field into a digital 8bit-per-pixel version of the picture, sampling the video signal at the rate of
412
FIG.
R . AUBERT E T AL.
2. (a) An example of a tracker simulation input and (b) the tracker simulation output.
-.-*----
*-----.---.----
TRACKER PROCESSOR
TRACKIK, OATA
IHPUT~~UTPUT
SENSOR CONTROL
DC SOPPLY W I T
-*-.-.*--.-. COMP. SYNC. INPUT
ANALOG VIDEO OUTPUT
OISPLAY PROCESSOR
FIG.3. Structure of the FIT hardware.
474
R . AUBERT E T AL.
8 MHz. Its output is the digital video bus, which feeds all the digital image processing units, i.e., any of the target detection units (TDU) or tracking units (TEU). Every image processing unit is equipped with and controlled by a microprocessor (Z80 in the prototype). If more than one of these units is required in an FTT, then the microprocessors are linked together and to the tracker (control) processor (TP) by the FTT bus, a dual serial multiprocessor bus system. The tracker processor controls all the image processing units, handles input and output data from and to the user system, transforms data from one coordinate system to the other, and performs a big part of the built-in test, controlling therefore the test image generator unit and checking the results from the other units.
THEFTT IMAGEPROCESSING UNITS Two digital image processing units (IPUs) have been realized for the FTT so far: (1) a target detection unit (TDU) and (2) a general purpose tracking unit (TEU). The TDU has been developed for special applications to detect one or multiple small, but unknown, targets within the full video field of view. In combination with a FLIR sensor automatic search (including scan motion of the sensor), detection and lock-on to distant targets far beyond the actual sensor field of view is possible and has been tested successfully. When using the tracking unit, the presence of a target within the field of view is usually assumed. In the automatic tracking mode the position of that target is known so that only a part (the tracking gate) of the image needs to be processed by the tracking unit. But, since the target size on the one hand can vary a great deal, and the target position in the acquisition phase on the other hand may not be known precisely (joystick, radar, or TDU data), the tracking gate must be of variable size. In the automatic tracking mode the TEU automatically controls the gate position, size, and shape. To reduce aimpoint wander the prototype TEU always tries to detect the whole target area (notjust parts of it) and to track its center of gravity. Therefore maximum gate size up to almost the full camera field of view is necessary. Target detection and recognition in real-world images is a very complex pattern recognition problem. In addition, for real-time tracking the computations must be done every 20 msec and with the shortest possible delay.
REAL-TIME IMAGE PROCESSING
475
There is no one single image feature, which always would allow discrimination between target and background pixels. Clearly, to make it possible to detect an object in an image at all, there must be some intensity variations in the image. But this does not mean that target and background pixels can always be distinguished just by their different intensities. That is why even the most sophisticated algorithms using the intensity information only for image segmentation must fail in certain tracking situations. The ETU therefore uses multiple features for image segmentation. Figure 4 shows a block diagram of the TEU image processing: the histogram processing is a statistical evaluation of the information in two intensity histograms, taken from the inside and the surroundings of the tracking gate, respectively, to determine “costs” for target and background classification of any pixel within the gate as a function of its gray level. Similar “costs” are derived by the image difference processing using the motion information for target/background discrimination (background motion being the inverse of the sensor motion). These two features have proved in some way to be complementary: against a homogeneous background the motion information cannot help, but this is the optimal case for the use of intensity information. In high background clutter conditions the roles are reversed. A gradient processing method has also been added in the simulations, but, so far, has not been implemented in hardware. To combine all those features, their different costs for target and background classification (corresponding to the negative logarithm of the probability of belonging to one or the other class) are added pixel-wise for every gate pixel. In the dynamic optimization process, which is described elsewhere,*the pixel classification is carried so that the sum of the costs for the chosen classification is minimized over a certain area (one gate line in the prototype). Since changes from target to background classification (and vice versa) in neighbor pixels are penalized, the effect is that of optimal noise cleaning. Memory can also be added to this classification process by penalizing changes of corresponding pixels in consecutive images. At this stage we have the binary image of the target, which can be displayed in real time in the FTT video output as shown in Fig. 5 . In the next stage the zeroth, first, and second moments of the binary target image are calculated to characterize the size, position, and shape/ orientation of the target. The measured position, together with data qualification, is sent to the user system, and the optimal gate position and size are predicted for the next image.
476
R. AUBERT ET AL.
DIGITAL VIDEO V
BAYES THRESHOLD
IMAGE SHIFT VECTOR
-
c
L
-
IMAGE OIFFERENCE PROC ESSlN G
HISTOGRAM PROCESSING
- 1 I
1
TARGET IMAGE STORAGE I k 1 )
-
ZKH HKH
ZKB HKB COSTS SUMMATION
c
I I I
I
I I I I I I
I
WEIGHT
~~~~
-
I
I I I I I
DYNAMIC OPTIMIZATION CLASSIFICATION
I
MOMENTS COMPUTATION
I I
?
(1
*
GATE PREDICTION
I
OUTPUT: AX,AY,O.
FIG.4. The FTT image processing system.
RESULTSAND CONCLUSIONS The FTT has been tested on recorded FLIR- and TV-tracking sequences and in field trials. Its ability to detect small targets and automatically lock on them has been demonstrated. The histogram processing has by itself brought a big improvement in tracking reliability compared to contrast threshold methods.
FIG. 5. Video output of the FTT showing a synthetic target in high background clutter without difference image processing (a) and with difference image processing (b).
..
0.
5
MEHO AUTO
6
.
7
Pt
15
-
t csl
ox ox
(>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
t csl
FIG.6. FIT tracking error and status results for a synthetic target in high background clutter. (a) and (b) Without difference image processing; (c) and (d) with difference image processing. (a) and (c) Top, tracking error: bottom, tracking status. (b) and (d) Histogram of Y-error.
479
REAL-TIME IMAGE PROCESSING
-16
-12
-8
-4
0
4
8
12
16
Y
AVERAGE VALUE : .o STANDARO OEVIATION : 1.S
FIG.6b. See legend on facing page.
Since in all the trials the FTT has not yet been fully integrated into the tracking system, the tracking mount position information, which is a necessary input for the image difference processing, was not available to the FTT.This method has therefore so far only been tested on images of a real background, in which a synthetic target has been superimposed (another facility of the test image generator in the FTT).Since the synthetic target position is known exactly, the tracking error introduced by the FTT itself can be calculated separately. Figure 5 shows the video output of a tracked synthetic target in high clutter background. The target is moving with constant speed from right to left at about 8 pixels per 20 msec. Figure 6a, b, c, and d shows the tracking errors without and with the use of the
0
s
r7
4-
u
48 1
REAL-TIME IMAGE PROCESSING
-16
-12
-8
-4
0
AVERAGE VALUE : STANDARD DEVIATION :
.
8
4
12
16
-
Y
.o .2
FIG.6d. See legend on p. 478.
image difference processing, respectively. In this case the optimal use of the motion information in addition to the intensity information allows further reduction of the tracking error standard deviation by a factor of 6 to 7 (Fig. 6b and d). Another measure for the improvement in tracking reliability is the relative persistence time in the tracking mode. As can be seen from Fig. 6a and c this figure was improved from 75 to 94%. With the FTT project the feasibility of a hardware realization of highly sophisticated and powerful image processing algorithms for automatic real-time video target detection and tracking has been demonstrated.
482
R. AUBERT ET A L .
Moreover with the FTT prototype we have a flexible tool for further developments in the area of real-time image processing.
REFERENCES I . Reischer, B., Proc. S. P. I . E. 178, 67 (1979). 2. Grossmann, J . , Dissertation No. 6489, ETH Zurich (1979).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS,VOL. 646
The Evaluation of Silicon CCDs for Imaging X-Ray Spectroscopy in the Range 1 to 8 keV R . E. GRIFFITHSt Haruard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts, U.S.A.
INTRODUCTION At the time of the Seventh Symposium on Photo-Electronic Image Devices (1978), charge-coupled devices had become established for use in optical imaging, but only preliminary work had been done to explore their sensitivity to X rays’” and the first paper had just been published on single-photon X-ray r e ~ p o n s eThe . ~ motivation for such experiments has come from the need for improved X-ray imaging in the fields of biology, medicine, laboratory plasma diagnostics, and X-ray astronomy, of which the latter has provided perhaps the greatest thrust for developments within the past 5 years. A number of research groups has been involved in the evaluation and development of CCDs for single-photon X-ray imaging and spectroscopy in the energy range -1 to 8 keV, i.e., to cover the energy range of present and future X-ray telescopes. Developments for biological research have centered on improved X-ray quantum detection efficiency at energies of 8 keV and greater.5 The reasons for intensive effort in the development of CCDs for these application areas lie in the following proven and potential factors: (1) high spatial resolution, i.e. 520 pm; (2) good spectral resolution for singlephoton detection (not more than one X ray per pixel): AEIE 5 0.3 at 1 keV, AEIE 5 0.1 at 3 keV where AE = FWHM of photopeak response; (3) high quantum detection efficiency (250%) over a broad energy range (50.5 keV to 2 8 keV); (4) low internal background with long count lifetime and high count rate capability (?lo3 cts sec-I); and ( 5 ) high geometric efficiency, with capability of mosaicking to practicably large areas (one CCD - 1 cm2). t Now at Space Telescope Science Institute, Homewood Campus, Baltimore, Maryland, U.S.A. 483 Copyright 0 1985 by Academic Press. Inc. (London) Lid. All rights of reproduction in any form reserved. ISBN 0-12-014724-6
484
R. E. GRIFFITHS
In the applications considered, the CCD offers a near-ideal instrument, being unique at the present time in providing a combination of factors 1-3. THEORETICAL PERFORMANCE PARAMETERS A number of independent parameters can be identified in order to quantify the performance of CCDs for imaging X-ray spectroscopy, viz. quantum detection efficiency (QDE), charge transfer efficiency (CTE), and readout noise. Other factors such as geometric efficiency and area may also be of fundamental importance, but do not impact directly on the physics of the basic device. Single-Pixel QDE
An important figure of merit for X-ray sensing with CCDs is the quantum detection efficiency for single-pixel detection, i.e., the photoelectric absorption and collection of resulting charge within one pixel. The singlepixel QDE is controlled at low energy ( s 2 keV) by absorption in the entrance layers (e.g., electrodes and insulating layers for front illumination and “dead layers” for back illumination), and at higher energies, 2 3 keV, by the absorbing depth of depleted silicon, i.e., QDE = e-XPeJe(1 -
e-Pdxd)
where p is the linear absorption coefficient and x the thickness of each layer, and subscripts e and d refer to entrance layers and depleted collection layers, respectively. Figure 1 shows calculated QDE curves for CCDs with effective entrance dead layers of 0.1,0.4, and 2.0 p m of silicon (where the latter includes front-illuminated electrodes), and depleted silicon depths of 3, 10, and 75 pm. High QDE at the lowest energies is achievable by back illumination of a fully depleted device, such that the entrance dead layer consists of S O . 1 p m of SiO2. In this back-illumination mode, an electric field must be present just below the effective “dead layer,” otherwise diffusion and recombination of carriers will occur (see discussion below with regard to deep depletion). It is also noteworthy, perhaps, that progress has been made within the past few years in the minimization of front electrode thickness. EastmanKodak, for example, has developed devices with electrode thicknesses of -0.2 pm with a total effective entrance thickness of 10 p m is required (see Fig. l), compared with 3 to 5 p m in some commercial devices, e.g., TI and GEC, and 7 to 10 pm in Fairchild CCDs. X Rays which penetrate the depletion region may be absorbed in the underlying substrate (which may be epitaxial or “bulk” material), where the resulting cloud of minority carriers where D is the diffusion length and T the diffuses to a radius r time for the center of the diffusing ball to reach the nearest potential A full calculation of the total amount of charge collected over the several or many pixels involved can only be done with knowledge of the electric field distribution within the substrate silicon. A discussion of this calculation is outside the scope of the present article. The approximation of a null electric field is discussed by Peckerar et al.’ and by Seib,9following the calculations of Crowell and Labuda.’O
-
486
R. E. GRIFFITHS
From the point of view of X-ray spectroscopy, however, any loss of charge by recombination, or any spread into more than one pixel, will result in degraded performance, i.e., loss of energy resolution and a trailing of event amplitude toward the low-energy part of the spectrum. The relevant figures of merit for CCDs in this application are therefore the single-pixel QDE and single-pixel energy resolution ; summing of pixels does not achieve the desired goals. In order to increase the depletion depth, a greater resistivity, i.e., lower impurity material has to be employed. In the approximation of an infinite slab, D a pl" where D is the depletion depth and p the material resistivity. p is inversely proportional to impurity concentration, e.g., 4 kn-cm is equivalent to 5 3 x lot2impurity atoms ~ m in- p-type ~ material, and 20 ka-cm material results from 5 7 x 10" impurities ~ m - The ~ . achievement of deep depletion in CCDs thus requires the conservation of high resistivity during processing of ultra-pure starting wafers." Care must be taken that oxygen does not form aggregates or generate stacking faults, implying a maximum temperature during processing of -750°C. For conventional multielectrode CCDs, this may imply that high-pressure oxidation be used in the formation of insulating layers. Charge Transfer Eficiency (CTE)
Theoretical studies of CTE in buried-channel CCDs have been made by Brewer,'* and also by Jack and Dyck,13with results indicating that charge For transfer inefficiency, E , may be reduced to values approaching a CCD of size N 2 pixels, and a readout noise of u rms electrons it is desirable that S , , ~ N E < u,where S is the signal from an X ray at the peak of the useful energy range, e.g., for cr = 10 electrons and N = 500, with S = 2100 (-8 keV), it is desirable that E C 5 X With E = the 8 keV signal would be degraded by an average of -20 electrons in transfemng through 500 to 500 pixels. However, calibration of the device using monochromatic X rays over the full area can be used to correct for small charge transfer inefficiency of this kind, depending on geometric location.
-
Readout Noise and X-Ray Energy Resolution
For a CCD readout noise of u electrons rms, the corresponding contribution to X-ray energy resolution (FWHM of photopeak response) is FWHM = 2.35 x r) (eV per electron-hole pair) x cr, e.g., 10 electrons rms noise leads to a contribution of 87 eV in the X-ray FWHM, for r ) = 3.7. The intrinsic X-ray energy resolution in silicon is FWHM (eV) = 2.35 x
SILICON
CCDS
FOR IMAGING
X-RAY
SPECTROSCOPY
487
(FqE)’” where E is the X-ray energy (eV) and the Fano factor F is of order 0.08. The intrinsic detector resolution at 3 keV, for example, is -70 eV FWHM, roughly equal to the contribution from readout noise, and combining in quadrature to give a total resolution of 112 eV FWHM, equal or better than that of most “solid-state detectors’’ (note that the latter devices, although employing a cooled off-chip JFET, for example, are limited in noise performance by the input capacitance, controlled by the detector size; in the CCD, charge can be transferred onto a node capacitance as small as 0.1 pF, thus reducing the amplifier noise). Note also, however, that several effects may contribute to a degradation of energy resolution, and may dominate both the amplifier noise and the intrinsic detection noise. These effects have been mentioned above, and include (1) nonunity CTE, (2) charge diffusion into more than on pixel, necessitating summation of pixels, and (3) recombination of charge in undepleted material, resulting in a trailing of events to the low energy side of the photopeak.
RESULTSOF EVALUATION
Devices tested for imaging X-ray spectroscopy during the period 19791983 were those listed in Table I. Imaging Characteristics-MTF,
CTE, Spatial Resolution
Initial tests on an RCA device (SID 52501) showed poor charge transfer and a relatively high output noise level (-80 electrons rms). Bulk-state traps and nonoverlapping gates were identified as the sources of the poor CTE, and these devices were not tested further. A Fairchild 21 1 CCD was used as a pilot device to test an early version of the camera system.I4The particular chip used had an unstable floating gate amplifier, but useful results were obtained between 3 and 6 keV, and demonstrated the imaging and spectroscopic capabilities of these devices. A Fairchild development of the 21 1, namely the larger format 221 and 222 devices, were characterized in greater detail.Is All Fairchild devices tested had an aluminum stripe covering the vertical shift registers, and these registers were not used as photosites during data collection (such use is possible at X-ray energies, with modification to the clocking scheme). Single-photon images were built up using the Fairchild devices, such as that illustrated in Fig. 2. This image was obtained using 3 keV X rays incident on the proximity-focus nickel mask. Some loss of CTE occurred in the 221 devices over a large fraction of the area (that part which exhibited a “swiss cheese” pattern in “dark” images, attributed to
TABLEI Performance parameters of tested CCDs Active pixel size (pm)
Device
Format
Fairchild 211 22 11222
199 x 240 388 x 480
RCA SID 52501
512 x 320 30 x 30
GEC P8600
385 x 576 22 x 22
Westinghouse (5040 series) Bell Northern Research
200 x 100 20 x 20
Active area (mm2)
Amp. noise Readout method
i7l-l.S
electrons
CTE at 50 kHz
Single-pixel QDE at 6 keV
Comments on
X-ray response
~~
Commercial
Developmental
W62A
48 x 256
14 x 18 14 x 18
26 x 26
4.4 X 5.7 Interline, 8.8 x 11.4 2-phase
15.4
X
9.8
Frame, 3-phase
30
20.99994
40
50.99
8.5 x 12.7 Frame, 20 (eng.) 3-phase 10 (astron.) 4.0 x 2.0 Frame, 40 4-phase 1.2
X
6.6
Frame, 2-phase
40
20.99998
0.3 0.3
0.2 Deep depletion
-0.99
0.99993 (7°C) -0.999 (-48°C)
Aluminum stripes cover vertical shift registers (4 area) Preliminary evaluation OdY
0.8
Deep depletion, 4 kn-cm material
SILICON
CCDs
FOR IMAGING X-RAYSPECTROSCOPY
489
FIG.2. Single-photon image in 3 keV X rays, obtained with a Fairchild 221 array. The displayed image covers 2 x l(r pixels, i.e., -0.1 of the full frame readout. Bright pixels have collected charge from a 3 keV photon. The grid pattern has slits 25 pm wide, comparable to the 14 x 18 pm’ active pixel area.
an incorrect anneal during processing as discussed by Murphy.I6 This problem was remedied in the manufacture of the Fairchild 222 devices, however, and tests with one of the latter devices did not show such an effect. The GEC P8600 devices use frame transfer readout, with no loss of geometric area. These devices exhibited high CTE, and reasonably good imaging performance for single-photon X rays (see the discussion below, however, on charge spread). The engineering devices tested had some light-emitting defects which limited the useful integration times, but the “astronomer-grade” devices did not have such defects. The developmental Bell Northern Research devices (manufactured under contract with Itek Corp.) were operated by frame-transfer readout, although they have usually been used for applications involving timedelay-and-integration. The devices were clocked so that the pixel size was 26 x 26 pm2, with a channel stop width of -8 pm, and an effective geometric collection area of at least 70%. These devices were tested especially for single-pixel QDE at X-ray energies of 2 to 6 keV (see
’’
490
R. E. GRIFFITHS
below), but single-photon X-ray images of the HEAO-B mask pattern were also obtained at different energies, especially at 4.5 and 6 keV (with results similar to Fig. 2, allowing for larger active pixel sizes). Spectroscopic Capabilities
The Fairchild 221 arrays were those first used to demonstrate X-ray energy resolution and linearity (Figs. 3 and 4), with a readout noise of about 30 electrons rrns. The engineering grade GEC devices had a read noise of -20 electrons rms (see Fig. 5 for a histogram of AlK events), and the astronomer-grade devices a noise level of 10 electrons, corresponding to a 90 eV contribution to the X-ray energy resolution FWHM (see Fig. 7). These latter GEC
I I
MnK
Energy (keV) FIG. 3. Pixel amplitude histograms for monochromatic X rays incident on a Fairchild 221 device, with X rays from aluminum K (1.5 keV), silver L (3 keV), and manganese K (5.9 keV).
Energy (keV)
FIG.4. Linearity of pixel output amplitude versus X-ray energy for the Fairchild 221.
1
Energy (keV)
Energy (keW
FIG.5. Pixel amplitude histograms for (a) AIK X rays, and (b) MnK X rays incident on a GEC P8600 engineering grade device.
492
R. E. GRIFFITHS
30-
-
u)
c
E
$ 20-
0)
a 10-
I
I
I
1
2
3
1
I-J-
1
4 5 Energy (keV)
I
I
6
7
FIG. 6. Pixel amplitude histogram for MnK X rays collected with a BNR W62A (deep depletion) CCD. (a) 7°C; (b) -48°C.
devices have a readout noise comparable with that of the best previously manufactured devices (the TI three-phase CCDs)I8 but their overall performance is marred by the small depletion depth (see section below). The developmental Westinghouse and BNR devices both had reasonably low electronic readout noise levels, considering the output amplifiers alone, viz. 30 to 40 electrons rms. Both devices, however, had charge transfer problem (unrelated to the fact that the devices had greater than normal depletion depths). The Westinghouse tin oxide electrode technology resulted in poor CTE, and the BNR devices had a temperaturedependent transfer problem in the output amplifier. X-ray energy resolution in the BNR (see Fig. 6) device was dominated by the contribution of dark current (or fat zero) shot noise necessary to effect good CTE. The
SILICON
CCDs
FOR IMAGING
X-RAYSPECTROSCOPY
493
Output amplifier noise (electrons rms)
FIG.7. Contribution to FWHM energy resolution from CCD output amplifier noise only, Le., with CTE = 1, dark current = 0, and not including intrinsic detector resolution. Values labeled TI are taken from Blouke el d.'* and Janesick er al.'
resulting total noise level was of the order of 100 electrons rms. At this level of X-ray energy resolution, the BNR device is not competitive for performing X-ray spectroscopy, but the resolution was sufficient to measure the QDE of these deep-depletion devices. Single-Pixel QDE
The single pixel QDE, i.e., the quantum efficiency for absorption and collection of charge from an X-ray event within a single pixel, is one of the most important figures of merit for X-ray imaging spectroscopy using CCDs (along with the factors discussed above). The first devices tested, the Fairchild 211, 221, and 222, all had similar single-pixel QDE values, with typical measurements shown in Fig. 1. The indicated values are consistent with collection of charge from a depletion depth of pm, agreeing with the depletion depth estimated from MTF measurements in the optical and near-infrared bands19 (note that the absorption length for 6 keV X rays is the same as that for -800 nm light). The measurements made with these devices used the usual photosite area, with a resulting loss of -50% in geometric efficiency (this factor is not included in the values shown). X Rays absorbed below the depletion layer in the neutral bulk silicon diffused into more than one pixel (with
494
R. E. GRIFFITHS
some loss of charge by recombination). These spread events appear in the histogram in Fig. 3 as a low-energy tail, with a relatively large number of pixels collecting an amount of charge equivalent to 1 keV X rays. (One single event may give rise to several of the histogrammed pixel counts at low apparent energies.) The Fairchild histograms (Fig. 3) are typical of those taken from areas of the F221 arrays where no swiss cheese pattern is present, such as the 20-pixel wide borders, and thus have high CTE values. Histograms such as these were used to estimate the CTE values shown in Table I. (The 21 1 and 222 chips did not exhibit the swiss cheese problem.) At the low-energy end, the QDE is dominated by absorption in the electrodes and insulating layers, and measurement of the QDE was consistent with an effective dead layer of -2 p m silicon. The Fairchild QDE is moderate in the energy range 2 to 6 keV, and the easy availability of 222 devices may make them suitable for some applications. The GEC P8600 devices tested (engineering grade bulk devices and astronomer grade epitaxial devices) exhibited better amplifier noise than the Fairchild chips, but had a more severe problem of event spreading (see the histogram from 6 keV events in Fig. 5b). In fact, a large fraction of events could be seen to be spread over at least -10 to 20 pixels, and these events contribute to the low-energy pile-up just above the electronic noise peak. Such events originate with X rays absorbed at, say, 30 p m beneath the surface, where the electric field is extremely weak. The histograms shown in Fig. 5 were taken with a “bulk” device, and the performance of the epitaxial device was very similar, except for a somewhat smaller fraction of events spread over many pixels and contributing to the low-energy pile-up. Note that the doping concentration in the epitaxial device is roughly the same as the material of the “bulk” device, so that the depletion depth is approximately the same in each case. This depletion depth was found, from X-ray QDE measurements on the bulk device, to be approximately 5 -+ 2 p m . A considerable fraction (-20%) of events was spread over 2 or more pixels even at I .5 keV (see histogram in Fig. 5a, where a secondary peak occurs at -0.75 keV from those events spread over 2 pixels). The BNR devices were manufactured on (100) silicon of resistivity 4.2 kn-cm, with a design goal of at least SO-pm depletion depth. As discussed above, these devices had poor CTE (probably caused by bulk traps in the vertical and horizontal registers, and an undefined problem in the output amplifier), and either had to be operated above O’C, where dark current provided the necessary bias charge (“fat zero”), or at temperatures down to - 100°C, where an optical input was provided for the bias charge. The
-
SlLlCON
CCDS F O R IMAGING X-RAY SPECTROSCOPY
495
QDE at 6 keV was measured (Fig. 1) to be at least 80%, in accordance with a depletion depth of 250 pm. It should be noted that control devices manufactured on low-resistivity silicon (in the same batches as the highresistivity wafers) also exhibited the poor CTE characteristics, so that deep depletion was achieved without the introduction of any deleterious effects. The histogram in Fig. 6, accumulated from one readout after exposure to 6 keV events, shows a shallow “trough” around 3 keV, i.e., a cleaner photopeak than either the Fairchild or GEC devices.
CONCLUSIONS Over the past few years, a laboratory program has demonstrated that charge-coupled devices can be used for X-ray imaging spectroscopy over the energy range -1 to 8 keV, in applications where the flux of X rays (taking into account integration time and readout frequency) is such that not more than one X ray is incident per pixel per frame. The spatial resolution is controlled by pixel size (-20 X 20 pm2) and the energy resolution by on-chip amplifier readout noise, of the order of 10 electrons rms in the best current devices (contributing 90 eV to the X-ray energy FWHM), with results approaching those of conventional solid-state detectors. The concept of a deep depletion CCD has been demonstrated, for full single-pixel collection of charge from X rays of up to 6 keV, with quantum detection efficiency of at least 80%. ACKNOWLEDGMENTS
This work was supported at SAO by NASA Grant NGS-7615 and NASA contract NAWS3000. It is a pleasure to acknowledge the engineering and technical assistance of G. Polucci, A. Mak, A. Roy, E. Dennis, and J. Logan together with the software programming of Maureen Conroy and Ellen Ralph, and the helpful advice and cooperation of J. Gerdes, D. A. Schwartz, S. S. Murray, and H.Tananbaum.
REFERENCES Peckerar, M. C., Baker, W.D. and Nagel, D. J., J . Appl. Phys. 48,2565 (1977). Koppel, L. N., Rev. Sci. Instrum. 48, 669 (1977). Renda, G. and Lowrance, J. L., JPL SP 43-21,91 (1975). Catura, R. C. and Smithson, R. C., Rev. Sci. Instrum. 50, 219 (1978). Peckerar, M. C., McCann, D. M. and Yu, L., Appl. Phys. Left. M,55 (1981). 6. Anagnostopoulos, C., Garcia, E., Lubberts, G., Moser, F.and Losee, D., hoc. IEEE Custom Integrated Circuits Conference, p. 78 (1980). 7. Janesick, J. R., Hynecek, J. and Blouke, M. M.,Proc. S.P. I. E. 290, 165 (1981).
1. 2. 3. 4. 5.
496
R. E. GRlFFlTHS
8. McCann, D. H.,Peckerar, M. C., Mend, W.,Schwartz, D. A., Griffiths, R. E., Polucci, G. and Zornbeck, M. V., Proe. S . P . I . E . 217, 118 (1980). 9. Seib, D. M., IEEE Trans. Electron Devices ED-21, 210 (1974). 10. Crowell, M. K. and Labuda, E. F., Bell Syst. Tech. J . 48, 1481 (1969). 11. Peckerar, M. C., McCann, D., Blaha, F., Mend, W.and Fulton, R., Proc. I.E.D.M., Dig. Tech. Pap. 144 (1979). 12. Brewer, R. J., IEEE Trans. Electron Devices ED-27, 401 (1980). 13. Jack, M. D. and Dyck, R. M., Proc. IEEE Trans. Electron Devices ED-23, 228 (1976). 14. Griffiths, R. E., Polucci, G., Mak, A,, Murray, S. S.. Schwartz, D. A. and Zombeck, M. V.,Proc. S. P . I . E. 244, 57 (1980). 15. Grifiths, R. E., Polucci, G., Mak, A., Murray, S. S. and Schwartz, D. A., Proc. S.P.I.E. 290, 62 (1981). 16. Murphy, H . E., Proc. S . P . I . E. 203, 80 (1979). 17. Ibrahim, A. A., Yu,K. Y.,Sallant, D., White, J. J., Bradley, W. C., Calvin, D. W., and Sandorti, C. J., Proc. Int. Conf. on Application of CCDs (NOSC, San Diego) p. 1-25 ( I 978). 18. Blouke, M. M., Janesick, J. R., Hall, J. E., Cowens, M. W. and May, P. J., Opt. Eng. 22, 607 (1983). 19. Dyck, R. M. and Steffe, W.,Proc. Int. Conf. on Application of CCDs (NOSC, San Diego) p. 1-55 (1978).
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS. VOL. 646
X-Ray Imaging and Spectroscopy with CCDs D.H.LUMB,G . R. HOPKINSONJ and A. A. WELLS Deppcirlmenl of Physics, University of Leicester, Leicester, Engliind
INTRODUCTION To date the most widely used imaging detectors for the 0.1 to 10 keV Xray waveband have been the imaging proportional counter (IPC) and the microchannel plate (MCP) detector. Traditionally the MCP has suffered from poor quantum efficiency (-5%) and no energy resolution. Conversely the moderate (-15%) energy resolution offered by the IPC has been obtained only with a spatial resolution of >300 pm, more than an order of magnitude worse than the MCP. The application of CCDs to X-ray imaging was first discussed in 1977'~~ when it was realized that the imaging capabilities of CCD TV cameras might be combined with the excellent energy response of silicon to provide simultaneous imaging and spectroscopy exhibiting the best features of the conventional devices described above. Subsequently a number of groups3q4have confirmed the feasibility of the technique, but revealed deficiencies in performance which have indicated that conventional CCD architectures are not well matched to the needs of X-ray imaging.
COMMERCIAL DEVICES The highest quality CCD imagers, when operated in a cooled slow scan mode, with low noise system electronics, have been shown to exhibit noise levels as low as -10 electrons rms. This is comparable with the Xray photoelectron shot noise in silicon (e.g., at 6 keV, with a Fano factor of 0.12, CT = 14 electrons rms), and consequently the energy resolution of CCDs may be very good, providing the collection of X-ray signal charge is complete. t NOWat SIRA Ltd., South Hill, Chislehurst, Kent, England. 491 Copyright 0 198s by Academic Press, Inc. (London) Ltd. All rights of reproduction in any form reserved. ISBN 0-12-014724-6
498
D. H. L U M B ET AL.
Charge Collection
Experimental measurements on samples of GEC CCDs have demonstrated deviations from the target 100 eV resolution due to incomplete collection of charge. The devices supplied by GEC for this work were designated research grade, i.e., they had been rejected as inadequate for commercial applications. The noise performances were in general inferior to current commercially available devices, and were constructed on an obsolete architecture (MA 328 type, 150 x 120 pixels). One such MA 328 CCD was illuminated by 22.1-keV photons emitted by a "Wd radioactive source, and the resulting pulse height distribution of CCD signals is shown in Fig. 1. The first peak correspond to the zero charge level and associated system noise, and the expected K a and KP peaks are clearly visible to the right. However, it is evident that a large number of events of intermediate energies have also been recorded. These events are manifestations of incomplete charge collection. The MA 328 devices were fabricated on a substrate on which an epitaxial layer of opposite conductivity type was created. The depletion layer is
Channel number
FIG.1 . Output pulse height distribution from MA328 CCD exposed to 22.1-keV photons: raw data.
X-RAY IMAGING A N D SPECTROSCOPY WITH
CCDS
499
established wholly within the epitaxial layer, which is itself biased such that the substrate-epi junction acts as a sink for the signal electrons. Therefore there is an undepleted volume within the epitaxial layer which may absorb photons. The photoelectrons generated may recombine in the substrate, but some will reach the depletion layer by diffusion. In diffusing to the depletion layer, the signal charge suffers recombination and may be split between two or more pixels. Consequently the charge collected is only a fraction of that generated by the X-ray photon, and this partial collection results in such events appearing in anomalously low energy channels in the pulse height distribution. The substrate doping is arranged to produce a low minority carrier lifetime, so that charge collected within the substrate is unable to diffuse to the depletion layer. Hopkinson5has considered a quantitative model for this diffusion process which predicts the amount and distribution of the charge collected in these devices. Two methods of analyzing the data were implemented in software: 1. Clustering: In this mode, when an X-ray event was found, any charge collected in neighboring pixels was added to the main event. The aim was to recover the “true” pulse height for each event. 2. Discrimination: In this analysis, all events which were spread over more than 1 pixel were removed from the data set. The remaining data were then assumed to be single events corresponding to photons absorbed in or close to the depletion layer.
Figure 2 shows the effect of the two methods. Better energy response is seen in the discriminated case, but clustering more accurately represents the total charge collected. Different devices with two epitaxial thicknesses were available 12(+2.4)p m and 30(&6)pm, corresponding to field-free layers of qk2.4) and 23(26) pm. The W d pulse height spectra for the thinner device are shown in Fig. 3. It can be seen that as expected there are relatively fewer diffusion events, and also the effect of the discrimination analysis is not significant. The latter effect is due simply to the fact that in diffusing only 5 p m to the depletion layer, the lateral diffusion cannot be much greater than 5 p m and most events are confined to a single pixel. Measurements were also made at 5.9 keV, and Table I compares the predicted and measured event rates for both devices. Note that the error in the manufacturer’s data is worse than those in the experimental figures, indicating that X-ray measurements may lead to a better estimate of some of the device parameters in future.
500
D. H . LUMB ET AL.
x 104
Channel number
FIG.2. Data for thick epitaxial device after processing: discriminated (hatched) and clustered (open) events, together with theoretical predictions.
Quantum Efficiency and Linearity
For single-photon absorption the quantum efficiency was measured at a number of different energies, by comparing the count rates from different X-ray sources in a CCD and a xenon proportional counter.6 Single-photon peak events only were recorded and the efficiency agrees well with the predictions for a 7-pm depletion layer. It can be seen from Fig. 4 that for energies >3 keV the efficiency falls off rapidly due to the relatively small extent of the depletion layer. The efficiency at 2.3 keV is dominated by the opacity of the electrode structure (0.12 p m of silicon dioxide, 0.05 p m of silicon nitride, and overlapping layers of 0.5 p m of polysilicon electrodes and oxide interphase isolation). Again there is good agreement between predicted and measured efficiency. For X-ray imaging spectroscopy of astronomical objects it is the efficiency at energies above 3 keV which is of most concern, for example, an important application envisaged for the CCD might be to map iron line features at 6.7 keV. The measured signal charge for each energy was noted, in order to derive an estimate for the linearity of energy response. From Fig. 5 it can
X-RAYIMAGING
AND SPECTROSCOPY WITH
CCDs
501
Channel number
FIG.3. Processed data for thin epitaxial device. Note that discrimination (hatched) and clustering (open) events give similar response, as predicted.
be seen that the response was linear from 2.3 to 24.8 keV, and with a measured response of 3.6 eV/charge pair, as predicted for silicon. It has been shown then, that the energy resolution of CCDs is degraded by the effects of charge diffusion from a field-free layer. Furthermore, the TABLEI Comparison of predicted and measured charge collection characteristics ~~
~
Thin epi
Thick epi
0.71
3.3
Fe event ratios
Cd event ratios
Ratio of free :depletion depths
Predicted Thin epi Thick epi
0.57
&
2.0
&
0.26 0.3
Measured
Predicted
Measured
0.59 2 0.06 1.8 2 0.2
0.71 k 0.3 3.3 ? 0.8
0.97 2 0.1 2.9 2 0.2
D. H. LUMB ET A L .
502
0.2-
Ag
(K d
FIG.4. Comparison of predicted and measured single-photon event efficiency. Solid line, MA 328; broken line, device on 100 0-cm substrate. In each case I represents the measured value.
quantum efficiency is limited by the relatively small extent of the depletion layer. Thus, as might be expected, CCD imagers designed for optical imaging are not optimized for X-ray imaging. The latter problem has been addressed by several groups who have fabricated devices on high resistivity substrates, in which the depletion layer extends to much greater depths (the depletion depth scaling as the square root of the resistivity). This has the dual advantage of enhancing the efficiency above 1 keV and minimizing the field-free layer. While the feasibility of the technique has been demon~trated,~ it has proved difficult to produce fully depleted devices while maintaining good charge-coupling operation. The approach chosen at Leicester, in collaboration with GEC, is to investigate devices with a range of resistivities and to note the change in depletion depths, as revealed by increasing X-ray efficiency. Should an optimum depletion depth be found, epitaxial substrates matched to this thickness may then be employed to ensure that a negligible field-free region exists, and the energy resolution maximized.
X-RAY IMAGING A N D SPECTROSCOPY WITH CCDS
Energy (keV)
503
-
FIG.5. Linearity of energy response of the CCD.
100 R-cm DEVICES
The first of these devices, fabricated on 100 R-cm substrates (conventional CCDs are fabricated on 20 to 30 a-cm material) have been manufactured by GEC Hirst Research Centre. Four devices of acceptable image quality have been produced, and are being tested at Leicester. To date preliminary testing has been performed on one of these (that with the worst image quality, #2043/18/11). Figure 6 depicts an image taken with this device. A copper mask with etched features of 120 pm was placed -1 mm from the CCD and illuminated with 5.9 keV radiation from a flight source. Several frames of data were accumulated to form this image, on a storage scope. At this temperature of -100°C the images are very clean, with very few light-emitting defects, a uniform background level (