Advances in
Electronics and Electron Physics EDITEDBY
PETER W. HAWKES Laboratoire d’Optique Electronique du Centre National de la Recherche ScientiJique. Toulouse, France
VOLUME 74 1988
ACADEMIC PRESS Harcourt Brace Jovanovich, Publishers London San Diego New York Boston Sydney Tokyo Toronto i
This Page Intentionally Left Blank
Photo-Electronic Image Devices PROCEEDINGS OF THE NINTH SYMPOSIUM HELD AT IMPERIAL COLLEGE, LONDON, 7-1 1 SEPTEMBER 1987
EDITEDBY
B. L. MORGAN The Blackett Laboratory, Imperial College, University of London, London, England
1988
ACADEMIC PRESS Harcourt Brace Jovanovich, Publishers London San Diego New York Boston Sydney Tokyo Toronto
...
111
ACADEMIC PRESS LIMITED 24/28 Oval Road LONDON NWI 7DX United States Edition published by ACADEMIC PRESS INC. San Diego, CA 92 101
Copyright 0 1988, by ACADEMIC PRESS LIMITED All Rights Reserved No part of this book may be reproduced in any form by photostat, microfilm, or any other means, without written permission from the publishers
British Library Cataloguing in Publication Data Symposium on Photo-Electronic Image Devices 9th : 1987 : London, England). 1. Photoelectronic imaging I. Title 11. Morgan, B.L. 111. Series 621.3815’42 ISBN 0- 12-014674-6
Typeset and printed in Great Britain at the Alden Press, Oxford
iv
CONTENTS CONTRIBUTORS . PREFACE. PROFESSOR J. D. MCGEE.
ix xv xvi
LLL TV Sensors with 18 mm Useful Input Diagonal Using Demagnifying Image Intensifiers. L. K. VAN GEFST . D. RIOUAND Low-light-levelTV with Image Intensifier Tubes and CCDs. J. C. RICHARD, M. VITTOT . Low-light-level TV with II/CCD Coupled Devices: Relative Merits of Different . Approaches. H. ROUGEOT AND P. GIRARD. A fibre-optically Coupled Intensified CCD Camera System. P. R. TOMKINS. A 3-stage 80: 7 mm Image Intensifier Combination for the CERN UA2 Scintillating Fibre Detector. L. BOSKMA, A. BAKKER, J. A. C. COCHRANE AND K. W. J. STOOP . A 40 mm MCP Intensifier for Photon Counting. T. J. NORTON,R. W. AIREY, B. L. MORGAN, P. D. READAND J. R. POWELL . Development ofa CCD-Digicon Detector System. R. G. HIER,W. ZHENG,E. A. BEAVER, C. E. MCILWAIN AND G. W. SCHMIDT . Increased Gain ofchannel Intensifier Tubes by Pulsed Biasing. B. W. NOEL,M. R. CATES AND L. A. FRANKS . Influence of Output Electron Energy Distribution of Microchannel Plates on the . AND Y. KIUCHI Resolution of Image Intensifiers. N. KOSHIDA MCP-PMTs as Ultra-fast Wide-band and Infrared-sensitive Detectors. K. OBA, AND K. NAKATSUGAWA . H. KUME,K. WAKAMORI Performance of a Photon-counting Microchannel Plate Intensifier with Wedge and Strip Image Readout. 0.H. W. SIEGMUND, C. J. HAILEY, R. E. STEWART AND J. H. LUPTON A Multichannel Detector for Photon Correlation. D. N. Qu AND J. C. DAINTY Application of Image Intensifier-Vidicon Systems to Low-light-level Phenomena in A. EISEN,A. J. WALTON AND L. A. CRUM. Physics and Biology. G. T. REYNOLDS, Cooled CCD Systems for Biomedical and Other Applications. C. D. MACKAY. Utilisation Astronomique de la Cambra Electronique Grand Champ-11. G. WLERICK, G. LELIEVRE, L. RENARD,B. SERVAN D. HORVILLE, J. FROMAGE, J. M. LEFLOHIC ET D. . BAUVUIN, A. BIlAOUl ET G. COURTES Image Recording in Electron Microscopy. D. MCMULLAN . A ;-inch 792 (H) x 492 (V) Pixel Colour Synchro Vision CCD Image Sensor. N. HARADA, Y. ENW, C. TANUMA, M. IFSAKA,Y. EGAWA,H. NOZAKI,S. UYA, S. SANADA, A. FURUKAWA, s. MANABEAND 0.YOSHIDA . Thinned Rear-face Electron-bombarded FT CCDs for LLL TV Imaging. L. BERGONZI. M. LEMONIER AND M. PETIT. A 2 x 2048 Pixel Bilinear CCD Array for Spectroscopy (TH 7832 CDZ). J. L. COUTURES AND G. BOUCHARLAT . Development of EBCCD Cameras for the Far Ultraviolet. G. R. CARRUTHERS, H. M. HECKATHORN, C. B. OPAL,E. B. JENKINS ANV J. L. LOWRANCE . Recent Developments in Solid-state Arrays for Infrared Astronomy. I. S. MCLEAN . V
1
9 17 27
35 41
55 69 79 87 97 107
1 I9 129 135 147 I57 165 I73 181
20 I
vi
CONTENTS
Multiple-frame UV/X-Ray Picosecond Framing Camera. R. T. EAGLES,W. SIBBETT, 209 W. E. SLEAT,D. R. WALKER, J. M. ALLISON AND N. J. FREEMAN . Evaluation of PVOOl and P-100 Tubes for Multiple-channel Streak Cameras. S. MAJUMDAR, P. Y. KEY, M. YA SCHELEV,Y. SURDYUCHENKO, W. SEKA, 219 AND R. KECK . M. C. RICHARDSON, P. YAANIMAGI Evaluation of a Photon-counting Streak Camera with CCD Recording. S. MAJUMDAR, 22 1 P. Y. KEY,V. PLATONOV AND A. RIDCLEY. A New Method for Observing High-speed Luminous Phenomena. Y. KIUCHI, 223 N. KOSHIDA AND T. SAKUSABE . An X-Ray Streak Tube. with Demountable Photocathodes. B. E. DASHEVSKY, v . A. PODVYAZNIKOV,A. M. PROKHOROV, A. v . PROKHlNDEEV AND v . K. CHEVOKIN. 233 A Subnanosecond Multi-framing Camera. V. V. LUDIKOV, A. M. PROKHOROV AND V. K. CHEVOKIN . 239 A CsI(Na) Scintillation Plate with High Spatial Resolution. K. OBA, M. ITO, 247 M. YAMACUCHI AND M. TANAKA. X-Ray Imaging Sensor Using a CdTe/a-Si: H Heterojunction. Y. HATANAKA, s. G. MEIKLE,Y.TOMITAAND T. TAKABAYASHI. 257 Low-noise Solid-state Linear Detectors for Large-field-of-view X-Ray Radiology. 269 AND P. PRIEUR-DREVON . H. ROUGEOT,B. MUNIER,G. ROZIERE An Image-intensified CCD Area X-Ray Detector for Use with Synchrotron Radiation. 215 R. H. TEMPLER, S. M. GRUNER AND E. F. EIKENBERRY . Further Developments of an X-Ray Television Detector. U. W. ARNDTAND G. A. IN’T VELD 285 Development of Large-format Photon-counting Array Detectors for the Lyman 291 Ultraviolet Space Telescope. E. H. ROBERTS, I. R. TUOHY AND M. A. DOPITA . Image Intensifier Tubes with Intagliated Screens. M. FOUASSIER, V. DUCHENOIS, J. DIETZ, E. GUILLEMET AND M. LEMONIER. 315 Photocathodes on Polycrystalline CsI/Na. Y. ARAMAKI 323 . High Performance with Trialkaline Antimonide Photocathodes. P. DOLIZY,F. G R O L ~ R E AND M. LEMONIER. 33 1 Relationship Between Microstructure and Photoelectric Quantum Yield (PQY) of S I 339 Photoemitting Surfaces. C. W. BATESJR, Q. Y. CHENAND N. V. ALEXANDER . S.20 Photocathode Stability Considerations. E. A. BEAVER, L. ACMN, D. DOLIBER, E. DOZIER AND H. WENZEL . 341 Physical Model and Optimization of a Heterostructure Vidicon Target Based on Amorphous Hydrogenated Silicon. F. SCHAUER, M. JEDLIEKA AND J. K&KA . 359 A New Extended Infrared Vidicon. T. KAWAI,K. SUGA,K. MURAMATSU, T. OTAKA, K. ATSUMI AND R. NISHIDA . 369 Avalanche-mode Amorphous Selenium Photoconductive Target for Camera Tube. K. TANIOKA, J. YAMAZAKI, K. SHIDARA, K. TAKETOSHI, T. KAWAMURA, T. HIRAI 379 AND Y.TAMSAKI . An Electrostatic Deflection, Electromagnetic Focusing Pick-up Tube for High-definition Television. H. ROUGEOT AND J.-L. RICAUD . 389 Surface Temperature Measurement of Small Objects by the Microthermovision Technique. V. RYQANEK . . 391 Anti-veiling Glare Windows for Third-generation Image Intensifiers. J. R. HOWORTH. 405 The Design of the Image Intensifier for the Faint Object Camera of NASA’s Space Telescope. R. P. RANDALL AND B. WILD . 413 Evaluation of Photon-event-counting Intensifiers. R. W. AIREY, T. J. NORTON, B. L. MORGAN, P. D. READAND J. L. A. FORDHAM 425
-
vii
CONTENTS
Calculations of the Electron Optics of Image Intensifiers Taking Account of Deviations from Rotational Symmetry. W. M~ILLER. . 435 Optimization of the Imaging Properties of an Image Inverter Tube. M. F. CALITZ, A. G. DU TOITAND C. F. VAN HWSSTEEN. . 457 An XUV Image Sensor for Rowland-circle Spectrographs. J. L. LOWRANCE AND C.L. JOSEPH . . 465 Properties of Imaging Electron-optical Systems for Image Tubes. V. JARES . . 475 Numerical Evaluation of Spread and Transfer Functions for Image Intensifiers with Polychromatic Illumination. J. M. WOPNICKI . . 483
INDEX
.
.
497
This Page Intentionally Left Blank
CONTRIBUTORS L.ACTON,Science Applications International Corporation. La Jolla, California, U.S.A. (p. 347) R.W . AIREY, The Blackett Laboratory, Imperial College of Science and Technology,London S W7 2BZ, England (pp. 41 & 425) N . V . ALEXANDER, Department of Physics, Stanford University. Stanford, C A 94305-4060, U.S.A. (P.339) J. M. ALLISON,A W R E Aldermaston. Reading, Berks. RG7 4PR, England (p. 209) Y. ARAMAKI, Horikawacho Works, Toshiba Corporation, Kawasaki, Japan (p. 323) U. W . ARNDT,M R C Laboratory of Molecular Biology, Cambridge, England (p. 285) K. ATSUMI,Hamamatsu Photonics K.K., Hamamatsu. Japan (p. 369) A. BAKKER, Delft Electronische Produkten, Roden, Holland (p. 35) C. W . BATES, JR., Department of Materials Science and Engineering, Stanford University,Stanford. C A 94305-2205, U.S.A. (p. 339) D.BAUDUIN,Observatoire de Paris, 75014 Paris, France (p. 135) E. A. BEAVER, Centre for Astrophysics and Space Sciences, University of California. San Diego, C-011, La Jolla C A 92093, U.S.A. (pp. 55 and 367) L.BERGONZ~, Laboratoires d’Electronique et de Physique Appliquie, 3 , avenue Descartes. 94450 Limeil Brivannes, France (p. 165) A. BIJAOUI, Obsewatoire de Nice, Nice, France (p. 135) L. BOSKMA, Derft Electronische Produkten, Roden. Holland (p. 35) G. BOUCHARLAT, Thomson-CSF Electron Tubes Division, 38. rue Vauthier BP 305, 92102 Boulogne-Billancourt, France (p. 173) M. F. CALITZ,National Physical Research Laboratory, CSIR, Pretoria, South Africa (P. 457) G. R. CARRUTHERS, E.O. Hulbert Center for Space Research, U.S. Naval Research Laboratory. Washington. D.C. 20375. U.S.A. (p. 181) M.R. C A ~Enrichment , Technology Applications Center. ORGDP. P.O. Box P, Oak Ridge T N 37831. U.S.A. (p. 69) Q.Y . CHEN,Department of Materials Science and Engineering, Stanford University. Stanford, C A 94305-2205, U.S.A. (p. 339) V. K. CHEVOKIN,General Physics Institute of the USSR Academy ofsciences, Vavilov Street, 38. II7942, Moscow, U S S R (pp. 233 & 239) J. A. C. COCHRANE, Delft Electronische Produkten, Roden, Holland (p. 35) G. COURTES, Laboratoire dAstronomie Spatiale. Marseille. France. {p. 135) 1. L.COUT~RES,Thomson-CSFElectron Tubes Division, 38. rue Vauthier BP305.92102 BoulogneBillancourt, France (p. 173) L.A CRUM, Physical Acoustics Laboratory. University of Mississippi. U.S.A. (p. 119) J. C. DAINTY, The Blackett Laboratory. Imperial College of Scienceand Technology. London SW7, 2BZ. England (p. 107) B.E. DASHEVSKY, General Physics Institute of the U S S R Academy of Science, Vavilov Street. 38, II 7942, Moscow, U S S R ( p . 233) J. DIETZ, Laboratoires dElectronique et de Physique AppliquCe, 3 , avenue Descartes. 94450 Limeil Brivannes. France (p. 315) D. DOLIBER, Science Applications International Corporation, La Jolla, California. V.S.A. (p. 347)
ix
X
LIST OF CONTRIBUTORS
P. D~LIZY, Luboratoires d'Electronique et a2 Physique Appliquke, 3, avenue Descartes, 94450 Limeil Brkvannes, France (p. 331) E. DOZIER, Science Applications International Corporation. La Jolla, California. U.S.A. (p. 347) M . A. DOPITER, Mount Stromlo and Siding Spring Observatories. Woden, ACT 2606. Australia (P. 397) A. G. DU TOIT,National Physical Research Laboratory, CSIR. Pretoria. South Africa (p. 457) V. DUCHENOIS, Laboratories d'Electronique et de Physique Appliquie, 3 , avenue Descartes, 94450 Limeil Brhannes, France (p. 315) R. T. EAGLES,University of St. Andrews, Department of Physics, North Haugh, St. Andrews, Fife KY16 9SS. Scotland (p. 209) Y. EGAWA,Toshiba Corporation, Research and Development Center, Toshiba-cho, Komukai, Saiwai-ku, Kawasaki-City, Japan (p. 157) A. EISEN,National Institutes of Health. U.S.A. (p. 119) Y. ENW, Toshiba Corporation, Research and Development Center, Toshiba-cho, Komukai. Saiwaiku, Kawasaki-City. Japan (p. 157) E. F.EIKENBERRY, Department of Pathology, UMDNJ-Robert Wood Johnson Medical School. Piscataway. NJ 08854 U.S.A. (p. 275) J. L. A. FORDHAM, University College London, Gower Street, London WC1 England (p. 425) M.FOUASSIER, Laboratoires dElectronique el de Physique Appliquke, 3 , avenue Descartes, 94450 Limeil Brhannes. France (p. 315) L. A. FRANKS, EGPG Energy Measurements. Inc.. I30 Robin Hill Road, Goleta. C A 93117, U.S.A. (P. 69) N . J . FREEMAN, A W R E Aldermaston. Reading, Berks. RG7 4PR, England (p. 209) J. FROMAGE, Observatoire a'e Paris, 75014 Paris, France. (p. 135) A. FURUKAWA, Toshiba Corporation, Research and Development Center. Toshiba-cho, Komukai. Saiwai-ku. Kawasaki-City, Japan (p. 157) P. GIRARD,Thomson-CSF Electron Tube Division, 38, rue Vauthier BP 305, 92102 BoulogneBillancourt. France (p. 17) F. GROLERE, Laboratoires d'Elec tronique et de Physique Appliquie. 3 , avenue Descartes. 94450 Limeil Brhannes, France (p. 331) S. M . GRUNER, Department of Physics, Princeton University, Princeton, NJ 08544. U.S.A. (P. 275) E. GUILLEMET, Laboratoires, dElectronique et de Physique Appliquie. 3, avenue Descartes, 94450 Limeil Brwannes. France (p. 315) C. J . HAILEY, University af California. Lawrence Livermore National Laboratory, Livermore, C A 94550, U.S.A. (p. 97) N . HARADA,Toshiba Corporation, Research and Development Center, Toshiba-cho, Komukai, Saiwai-ku, Kawasaki-City, Japan (p. 157) Y. HATANAKA, Research Institute of Electronics, Shizuoka University. 3-5-1 Johoku, Hamamatsu 432, Japan (p. 257) H. M . HECKATHORN, E. 0. Hulbert Center for Space Research, US Naval Research Laboratory, Washington DC 20375, U.S.A. (p. 181) R. G. HIER,Centre for Astrophysics and Space Sciences, University of California. San Diego. C-011, La Jolla. CA 92093, U.S.A. (p. 55) T.HIRAI,Central Research Laboratory, Hitachi Ltd, Kokubunji. Tokyo 185. Japan (p. 379) D. HORVILLE, Observatoire de Paris, 75014 Paris, France (p.135) J. R. HOWORTH, 21 Victoria Road, Maldon. Essex. England (p. 405) M . IESAKA, Toshiba Corporation, Research and Development Center, Toshiba-rho, Komukai Saiwai-ku, Kawasaki-City. Japan (p. 157) G . A. IN'T VELD, M R C Laboratory of Molecular Biology, Cambridge, England ( p . 285)
LIST OF CONTRIBUTORS
xi
M . ITO,Hamamatsu Photonics K. K. Shimokanzo. Toyooka-mura. Iwata-gun, Shizuoka-ken, Japan (P. 247) V. JARES, TESLA-Vacuum Engineering, Prague, Czechoslovakia, (p. 475) M . JEDLICKA TESLA-Vacuum Engineering, Prague, Czechoslovakia (p. 359) E. B. JENKINS,Department of Astrophysical Sciences, Princeton University. Princeton. NJ 08544, U.S.A. (p. 181) C. L.JOSEPH, Princeton University Observatory, Princeton, NJ 08544, U.S.A. (p. 465) T.KAWAI,Hamamatsu Photonics K.K.. Hamamatsu. Japan (p. 369) T.KAWAMURA, NHK Science and Technical Research Laboratories, Setagaya. Tokyo 157, Japan (P. 379) R. KECK,Laboratory for Laser Energetics. University of Rochester, U S A . (p. 219) P.Y.KEY,Delli Delti Ltd, 56 Mattock Lane, London W13, 9LA, England (pp. 219 & 221) Y.KIUCHI, Department of Electronic Engineering, Faculty of Technology. Tokyo University of Agriculture and Technology. Koganei. Tokyo 184. Japan (pp. 79 & 223) J. KOCKA,Institute of Physics, Czechoslovak Academy of Sciences, Prague, Czechoslovakia (p. 359) N. KOSHIDA, Department of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology. Koganei, Tokyo 184, Japan (pp. 79 & 223) H.KUME,Hamamatsu Photonics K. K.. Shimokanzo Toyooka-Mura, Iwata-gun. Shizwoka-ken, Japan (p. 87) J. M . LE FLOHIC, Observatoire de Paris, 75014 Paris, France (p. 135) G. LELIEVRE, Obsewatoire de Paris, 75014 Paris, France (p. 135) M. LEMONIER, Laboratoires d’Electronique et de Physique Appliquke. 3, avenue Descartes. 94450 Limeil Brhannes. France (pp. 165, 315 & 331) J. L. LOWRANCE, Department of Astrophysical Sciences, Princeton University. Princeton, NJ 08544, U.S.A. (pp. 181 & 465) V. V. LUDIKOV, General Physics Institute of the USSR Academy of Sciences, Vavilov Street. 38, I1 7942, Moscow, U S S R (p. 239) J. H.LUPTON,KMS Fusion Inc., Ann Arbor, M I 48150, USA (p. 97) C. D.MACKAY, Institute of Astronomy, University of Cambridge, Cambridge CB3 OHA. England (P. 129) S . MAIUMDAR, Delli Delti Ltd, 56 Mattock Lane. London W13 9LA, England (pp. 219 & 221) S. MANABE,Toshiba Corporation. Research and Development Center, Toshiba-cho, Komukai. Saiwai-ku, Kawasaki-City. Japan (p. 157) C. E. MCILWAIN, Centerfor Astrophysics and Space Sciences, University of California. San Diego. C-011, La Jolla. C A 92093, U.S.A. (p. 55) I . S . MCLEAN,Royal Observatory. Edinburgh, U.K. Infrared Telescope, 665 Komohara St. Hilo. Hawaii 96720, U S A . (p. 201) D. MCMULLAN,Cavendish Laboratory. Cambridge University, Cambridge CB3 OHE, England (P. 147) S . G. MEIKLE, Research Institute of Electronics. Shizwoka University, 3-5-1 Johoku. Hamamatsu 432, Japan (p. 257) B. L. MORGAN,The Blackett Laboratory, Imperial College of Science and Technology, London SW7 2BZ. England (pp. 41 & 435) W .M~LLER Siemens , A,G. Medical Division, Erlangen, West Germany (p. 435) B. MUNIER,Thomson-CSF Electron Tubes Division, 38, rue Vauthier BP 305, 92102 BoulogneBillancourt. France (p. 269) K . MURAMATSU, Hamamatsu Photonics K.K., Hamamatsu. Japan (p. 369) K. NAKATSUGAWA, Hamamatsu Photonics K. K., Shimokanzo, Toyooka-mura, Iwata-gun, Shizwoka-ken. Japan (p. 87) R.NISH~DA Hamamatsu Photonics K.K.. Hamamatsu, Japan (p. 369)
xii
LIST OF CONTRIBUTORS
B.W . NOEL,Los Alamos National Laboratory. P.O. Box 1663, Los Alamos. N M 87545, U S A . (P. 69)
T . J . NORTON, The Blackett Laboratory, Imperial College of Science and Technology,London S W7 2BZ. England (p. 425) H . NOZAKI, Toshiba Corporation, Research and Development Center, Toshiba-cho, Komukai. Saiwai-ku, Kawasaki-City, Japan (p. 157) K. OBA,Hamamatsu, Photonics K.K., Shimokanzo, Toyooka-muva. Iwata-gun, Shizuoka-ken, Japan (pp. 87 & 247) C. B. OPAL,Department of Astronomy, University of Texas. Austin. TX 78712. U.S.A. (p. 181) T . OTAKA, Hamamatsu Photonics K.K., Hamamatsu, Japan (p. 369) M . PETIT, Laboratoires dElectronique et de Physique Appliquhe. 3 , avenue Descartes. 94450 Limeil Brwannes, France (p. 165) V. PLATONOV, General Physics Institute, 38, Vavilov Street, Moscow, U S S R (p. 221) V. A. PODWAZNIKOV, General Physics Institute, 38. Vavilov Street, Moscow, U S S R (p, 233) J. R. POWELL, Instrument Technology Ltd, Hastings. Sussex. England (p. 41) P. PRIEUR-DREVON, Thomson-CSF Electron Tubes Division, 38, rue Vauthier BP 305, 92102 Boulogne-Billancourt, France (p. 269) A. V. hOKHINDEEV, General Physics Institute of the U S S R , Academy of Sciences, Vavilov Street, 38, II 7942, Moscow. U S S R (p. 233) A. M. PROKHOROV,General Physics Institute of the USSR,Academy of Sciences, Vavilov Street 38. II 7942, Moscow, USSR (pp. 233 & 239) D. N. Qu, The Blackett Laboratory, Imperial College of Science and Technology, London S W 7 2BZ, England (p. 107) R. P.RANDALL,Thorn EMI Electron Tubes Lid.. Ruislip, Middlesex, England (p. 413) P. D. READ,Royal Greenwich Observatory, Herstmonceaux, Sussex, England (pp. 41 & 425) L.RENARD,Observatoire de Paris. 75014, Paris, France (p. 135) G. T. REYNOLDS, Department of Physics, Princeton University, Princeton, U.S.A. (p. 119) J.-L. RICAUD, Thomson-CSF Division Tubes Electroniques, 38. rue Vauthier BP 305, 92102 Boulogne-Billancourt. France (p. 389) J. C . RICHARD,Laboratoires dElectronique et de Physique Appliquke, 3, avenue Descartes, 94450 Limeil Brkvannes, France (p. 9 ) M . C. RIcnmDmN, Laboratory for Laser Energetics, University of Rochester, U.S.A. (p. 219) A. RIDGLEY, Rutherford Appleton Laboratory, Chilton, England (p. 221) D. RIOU, Laboratoires dElectronique et de Physique Appliquke, 3 , avenue Descartes, 94450 Limeil Brhannes. France (p. 9 ) E. H . ROBERTS,Auspace Limited, P.O. Box 1992, Canberra, ACT, 2601, Australia (p. 297) H.ROUGEOT, Thomson-CSF Electron Tubes Division, 38, rue Vauthier BP 305, 92102 BoulogneBillancourt, France (pp. 17,269 & 389) i G. ROZIERE, Thomson-CSF Electron Tubes Division, 38. rue Vauthier BP 305, 92102 BoulogneBillancourt. France (p. 269) V . RYUNEK,Czech Technical University. Prague, Czechoslovakia (p. 397) T. SAKUSABE, Department of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184. Japan (p. 223) S. SANADA,Toshiba Corporation, Research and Development Center, Toshiba-cho, Komukai, Saiwai-ku, Kawasaki-City, Japan (p. 157) F. SCHAUER, Technical Academy. Bfno, Czechoslovakia (p. 359) G. W . SCHMIDT, Centre for Astrophysics and Space Sciences, University of California. San Diego, C-011. La Jolla. C A 92093, U.S.A. (p. 55) W. SEKA,Laboratory for Laser Energetics, University of Rochester U S A . (p. 219) B. SERVAN, Observatoire de Paris, 75014 Paris, France (p. 135)
LIST OF CONTRIBUTORS
...
Xlll
K. SHIDARA,NHK Science and Technical Research Laboratories, Setagaya, Tokyo 157, Japan (p. 3 79) W. SIBBETT, University of SI. Andrews. Department of Physics, North Haugh, St. Andrews, Fife, KY16 9SS. Scotland (p. 209) 0.H . W. SEGMUND, Space Sciences Laboratory, University of California, Berkeley, California. U.S.A. (p. 97) W. E. SLEAT,University of St. Andrews. Department of Physics. North Haugh, St. Andrews, Fife, KY16 9SS. Scotland (p. 209) R.E.STEWART, University of California. Lawrence Livermore National Laboratory, Livermore. C A 94550, U.S.A. (p. 97) K . W. J. STOOP, Deljt Electronische Produkten, Roden. Holland (p. 35) K.SUGA,Hamamatsu Photonics K.K., Hamamatsu, Japan (p. 369) Y. SURDYUCHENKO, General Physics Institute. 38, Vavilov Street. II7942, Moscow, U S S R (p. 219) T. TAKABAYASI, Research Institute of Electronics. Shizuoka University, 3-5-1 Johoku. Hamamatsu 432, Japan ( p . 257) Y.TAKASAKI,Central Research Laboratory. Hitachi Ltd, Kokubunji, Tokyo 185, Japan (p. 379) K . TAKETOSHI, NHK Science and Technical Research Laboratories, Setagaya, Tokyo 157, Japan (P. 379) M.TANAKA, Hamamatsu Photonics K.K., Shimokanzo, Toyooka-mura. Iwata-gun. Shizuoka-ken, Japan (p. 247) K . TANIOKA, NHK Science and Technical Research Laboratories. Setagaya. Tokyo 157, Japan (P. 379) C. TANUMA,Toshiba Corporation, Research and Development Center. Toshiba-cho, Komukai, Saiwai-ku, Kawasaki-City. Japan (p. 157) R. H. TEMPLER, Department of Physics, Princeton University. Princeton NJ 08544, U.S.A. (P. 275) Y. ToMIT.4, Research Institute of Electronics, Shizuoka University. 3-5-1 Johoku, Hamamatsu 432, Japan (p. 257) P. R.TOMKINS, Photonic Science Ltd., Tunbridge Wells. Kent, England (p. 27) I . R. TUOHY, Mount Stromlo and Siding Spring Observatories, Woden, ACT 2606, Australia (P. 297) S. UYA, Toshiba Corporation. Research and Development Center. Toshiba-cho, Komukai. Saiwaiku. Kawasaki-City, Japan (p. 157) L. K . VAN GEEST,Devt Electronische Produkten, Roden. Holland (p. I ) C. F. VAN HUYSSTEEN, National Physical Research Laboratory. CSIR. Pretoria, South Africa (P. 457) M. VITTOT,Laboratoires d'Electronique et de Physique Appliquke, 3, avenue Descartes, 94450 Limeil Brhannes, France (p. 9 ) K. WAKAMORI, Hamamatsu Photonics, K.K.. Shimokanzo, Toyooka-mura, Iwata-gun. Shizwokakan, Japan (p. 87) D.R. WALKER,University of St. Andrews. Department of Physics, North Haugh, St. Andrews. Fifee, KY16 9SS. Scotland (p. 209) A. J. WALTON,Cavendish Laboratory. Cambridge, England (p. 119) H. WENZEL,EG and G, San Ramon, California. U.S.A. (p. 347) B. WILD,Thorn EMI Electron Tubes Ltd.. Ruislip. Middlesex. England (p. 413) G. WLERICK, Observatoire ak Paris, 75014 Paris, France (p. 135) J. M . WO~NICKI, Institute of Microelectronics and Optoelectronics, Warsaw University of Technology, Warsaw, Poland (p. 483) M . YA SCHELEV, General Physics Institute, 38 Vavilov Street, Moscow, U S S R (p. 219) P.YAANIMAGE, Laboratory for Laser Energetics. University of Rochester. U S A (p. 219)
xiv
LIST OF CONTRIBUTORS
M.YAMAGUCHI, Hamamatsu Photonics K.K., Shimokanzo, Toyooka-mura. Iwata-gun, Shizuokaken, Japan (p. 247) J. YAMAZAKI, NHK Science and Technical Research Laboratories. Setagaya. Tokyo 157,Japan (P. 379) 0.YOSHIDA, Toshiba International Ltd.. Uxbridge, Middlesex UBl l l A R , England ( p . 157) W. ZHENG,Centre for Astrophysics and Space Sciences, University of California, San Diego, C-011, La Jolla. C A 92093.U.S.A.( p . 55)
PREFACE The Ninth Symposium on Photo-Electronic Image Devices was held at Imperial College, University of London, from September 7 to 11,1987. As for the previous eight symposia, the proceedings are here published in the series “Advances in Electronics and Electron Physics”. I would like to express my gratitude to Dr P. W. Hawkes and to Academic Press for making this possible. The Eighth Symposium was opened by Professor J. D. McGee, who had inaugurated the series of meetings in 1958. In the preface to the proceedings of that symposium, I expressed the hope that he would attend the Ninth Symposium. Unfortunately, that was not to be, since he died in February 1987. The present conference opened with a tribute to Jim McGee given by Dr Dennis McMullan, following which delegates suggested that the symposia should in future be known as “The McGee Symposia on Photo-Electronic Image Devices”. A brief appreciation of McGee’s life and work appears below. As always, it was a great pleasure to renew old friendships and to make new ones with colleagues from all over the world. Many spoke kindly of the value of these meetings and of the hard-backed proceedings which, they say, form a comprehensive and permanent reference work for the field of photoelectronics. In editing the papers, I have tried to achieve a reasonable degree of uniformity in matters such as the system of units and the style of presentation. I have not interfered with the scientific content, which is, of course, the responsibility of the authors. I hope that the resulting volume does not fall too far below the standard set in the proceedings of earlier symposia. Finally, I would like to thank the members of the Astrophysics Group of Imperial College who collaborated so magnificently in running the symposium. The tenth symposium in the series is planned for September 1991.
B. L. MORGAN
December 1987
xv
J. D. MCGEE(1903-1987)
xvi
PROFESSOR J. D. McGEE The series of symposia of which these are the proceedings was inaugurated in 1958 by Professor James Dwyer McGee, OBE, FRS, who died in February I987 at the age of 83. His life had been devoted to the field of photoelectronics, ranging from his contributions to the development of television in the 1930s to his work on astronomical detectors in the 1960s and 1970s. McGee was the sixth of seven children. He was born in 1903 at Tuggeranong near Canberra, Australia, where his father was the local schoolmaster. The family supplemented their income to maintain a simple but comfortable existence by running a small farm that was attached to the school. His father was an active sportsman who respected scholarship and loved music. McGee absorbed all these passions and retained them throughout his life. McGee’s secondary education was taken at St Patrick’s College, Goulburn, from where he went to St. John’s College, Sydney University. In 1926 he obtained a first-class degree in Physics and Mathematics and undertook research into the motion of electrons in gases, for which he was awarded a master’s degree. The John’s College magazine described him as “the most distinguished graduate of John’s for many years”. In 1928 he won an 1851 Exhibition Scholarship and moved to the Cavendish Laboratory, Cambridge, where he was awarded his doctorate for research into nuclear physics carried out under the supervision of Chadwick. In 1932 McGee accepted a post in the EM1 laboratories at Hayes, Middlesex, to work on the development of television. At that time the only demonstrations made of television had been by mechanically-scanned systems that extended the pioneering work of John Logie Baird. The limitations of these mechanical systems were very apparent but, as early as 1908, A. A. Campbell-Swinton had suggested an all-electronic system for television; however, the then-contemporary electronics technology fell far short of the requirements. McGee and his immediate superior, Tedham, were convinced that an all-electronic system for television was then possible and that it would supersede mechanical systems. To prove this they worked without authorization, in their own time, to design and construct an electronic television camera tube. The device worked briefly, and this, plus the news that RCA were working on electronic systems for television, was sufficient to convince Shoenberg, the head of research, who redirected EMI’s efforts into electronically-scanned systems. Within a few years Shoenberg was able to xvii
xviii
PROFESSOR J . D. MCGEE
show the BBC high-definition television pictures in a 405-line system using signals obtained by the Emitron television camera tube which had been developed by the group by then headed by McGee. This system was the standard adopted by the BBC in the first scheduled high-definition television service in 1936. McGee’s work on teIevision camera tubes was interrupted in 1939 by the war, during which he worked on night-vision devices, eventually running a section of EM1 in which infrared sensitive “image converters” were manufactured. It was largely for this work that he was awarded an OBE in 1952. In 1945 McGee resumed his work on television camera tubes but also investigated various ideas for imaging intensifiers based partly on the nightvision devices he had developed during the war. This work was noticed by P. M. S. Blackett, who had recently become Head of the Department of Physics at Imperial College. Blackett foresaw the importance of imageintensifying devices and, in 1954, invited McGee to accept the Chair of Instrument Technology (later Applied Physics) at Imperial College. McGee’s achievements at Imperial College were remarkable. He and his group were responsible for developing many now-familiar photoelectronic devices including the cascade image intensifier, which is the primary detector in the Image Photon Counting Systems used at the La Palma and Anglo-Australian Observatories and on the Hubble Space Telescope; the secondary electron conduction (SEC) television camera tube (which was also developed independently by Westinghouse in the USA); high-speed camera-tubes; and the Spectracon electronographic image tube. These devices were not developed in isolation, but were used in joint programmes with many collaborators in fields such as astronomy, medicine and nuclear physics. The success of those who worked with McGee is also remarkable. For instance, in 1962,apart from himself, the academic staff in his group were D. J. Bradley, L. Mandel, N. D. Twiddy and W. L. Wilcock-within a few years all four held Chairs and three subsequently became heads of their departments. In 1958 McGee organized the first “Symposium on Photoelectronic Image Devices” at Imperial College. McGee’s reputation ensured international interest and the meetings were an immediate success. They have since come to be regarded as perhaps the most important in the field. At the ninth symposium, of which these are the proceedings, delegates proposed that, as a tribute to its founder, the series should be continued under the general title “The McGee Symposium on Photoelectronic Image Devices”. In 1966 McGee was elected a Fellow of the Royal Society. He retired in 1971 but continued his research at Imperial College as Emeritus Professor. In 1980 he moved to Sydney, from where he made numerous trips abroad to deliver lectures and to attend conferences. McGee was always courteous, charming and scrupulously fair, but he was
PROFESSOR J. D. MCGEE
xix
very firm and maintained close control of all research carried out by members of his group. Nevertheless, he would listen to every proposal and, when convinced by a well-organized argument for a particular course of action, would give it his whole-hearted support. He enjoyed entertaining and he and his wife were excellent hosts. His main interests other than his work were principally gardening, sport and, above all, music. He loved opera, especially Wagner, and guests would frequently be invited to hear the latest recordings played through fine "hi-fi" equipment. In his youth he had been an active sportsman playing football, cricket and tennis and ski-ing well; he was a strong swimmer and new research students were sometimes surprised to see him plunging into the Imperial College swimming pool during his lunch hour. In 1944 he married Hilda Winstone of Auckland, New Zealand, who survives him. They had no children. B. L. MORGAN
This Page Intentionally Left Blank
LLL TV Sensors with 18 mm Useful,Input Diagonal Using Demagnifying Image Intensifiers L. K. VAN GEEST B. V. Devt Electronische Producten. Roden. Holland
INTRODUCTION Standard TV cameras using solid state image sensors (SSIS) have sensitivities down to about 1 lux scene illumination. To allow them to be used at lower illumination levels, some form of amplification has to occur. This can be achieved by preceding them with an image intensifier. The most commonly used SSISs have a rectangular image area (3:4) with a diagonal of either 11 mm (5 inch format) or 7 mm (t inch format). The standard input diameters of the majority of image intensifiers are 18,25 and 40 mm. The most popular choice for use in intensified SSISs is the 18 mm proximity-focused wafer tube which therefore has an 18 mm output. The matching of this 18 mm output diameter to an 1 1 mm or a 7 mm sensor can be done by either optical demagnification or tapered fibre-optics, or by using an electron-optically demagnifying image intensifier. Optical Coupling By Lenses
The output image of the intensifier is imaged onto the sensor using an optical lens and the appropriate magnification. This means of coupling is characterized by a good MTF, but involves a high loss of light and is rather bulky. Fibre-optic Coupling
Very efficient coupling can be achieved by using fused fibre-optics between the fibre-optic output window of the intensifier and the chip surface. If a straight fibre-opticis used, only that part of the intensifier that corresponds to the image area of the sensor is utilized. A better approach is to use a demagnifying tapered fibre-opticwhose input and output match respectively, the output of the intensifier and the input of the sensor. The advantages of this way of coupling are low light losses plus a sturdy, compact construction. The 1 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyrighi 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
2
L. K . VAN GEEST
disadvantages incurred are the introduction of more distortion and vignetting, and an increase in the number of blemishes. Demagnifying Image Intens$er
Both image intensification and matching of image sizes are obtained in a single step when an image intensifier is used that has the proper electronoptical demagnification. At DEP, two special first-generation image intensifiers have been developed for this purpose: (i) the 1817 tube, which demagnifies from an 18 mm input to a 7 mm output for coupling to an SSIS with a 7 mm image diagonal; and (ii) the 18/1 I tube, which demagnifies from an 18 mm input to an 11 mm output for coupling to an SSIS with an 11 mm image diagonal. Both tubes have fibre-optic input and output windows, S.25 photocathodes, and either P.20 or Pa43 phosphors. They are used either for single-stage intensification or as the second stage of a two-stage super-inverter that has as its first stage an 18 mm Gen. I1 microchannel plate wafer tube. The super inverter has a very high gain, typically 100.000 cdm-* lx-', which makes possible TV images at very low light levels. An external gain-control enables the user to vary the gain of the micro-channel plate tube, so that an optimum match can be obtained between gain and signal-to-noise ratio (Van Geest and Stoop, 1985). Using a GaAs photocathode (as in the Gen. I11 wafer tube) should further improve the performance under night conditions. In Fig. 1, the cross-section of the XXl570 LLL TV sensor is shown. The XXI 570 is a super-inverter with an 18 mm input and a 7 mm output coupled to a Philips M A 1011 frame-transfer sensor. The XX1540, which has a single stage 18/7 image intensifier coupled to a CCD, has a similar construction at the output
FIG.1. Cross-sectional view of the XX1570 tube.
LLL TV SENSORS USING DEMAGNIFYING IMAGE INTENSIFIERS
3
FIG.2. Two examples of LLL TV sensors: XXI 540 (left) and XX1570 (right)
end. Two completed LLL TV sensors, types XX1540 and XXl570,are shown in Fig. 2.
SPECIFICATIONS The main parameters of four different LLL TV sensors are given in Table I. The intensifier sections have integrated power supplies operating from an input of 2.5 V. The standard single-stage intensifiers are operated at 15 kV and have earthed anodes. This makes the coupling of the SSIS possible without any extra precautions. Automatic brightness-control is not incorporated. The so-called hybrid tubes have automatic brightness-control incorporated as standard. During operation, the anode is at 12 kV high voltage. The sensor is isolated from the high voltage by the fibre-optic output window, whose external surface is earthed via a transparent conductive coating that is deposited on the fibre-optic window of the sensor. The hat-shaped fibre optic window of the sensor is coupled to the chip surface, using immersion oil, and the brim is mechanically sealed to a metal plate which, in turn, is sealed to the chip carrier. The complete sensor assembly is coupled to the intensifier’s output window with immersion oil and is held in place by a spring between sensor and metal cap at the back. The metal cap is attached to the intensifier housing with screws and contains a connector block for the electrical connections to the sensor. With this construction and proper earthing, no electrical interface problems occurred.
4
L. K. VAN GEEST
TABLE I
Specifications of different LLL TV sensors _____
Parameter
Intensifier section Type Input diameter (mm) Cathode sensitivity (PA lm-') at 800 nm (mA W-I) at 850 nm (mA W I ) Output diameter (mm) Output Phosphor Luminance gain (cd m - 2 lx-') Resolution (output) (lp mm-') Magnification E.B.I. (PIX) ~
Sensor section Type Image area (mm2) Pixels
XX1540
XX1570
XX1550
XX1560
Gen. I 18 300 20 15 7 P-20 320 75 0.37 0.1
Hybrid 17.5 300 24
Gen. I I8 300 20
Hybrid 17.5 300 24 18
18
15
7 P.43
I1 P.20 130
105 55
0.37 0.1
Philips NXA 1011 6.0 x 4.5 604 H x 576 V
11
0.57
P.43 I 05 42 0.57
0.1
0.1
15
Hitachi HE98222A 8.8 x 6.6 384Hx485V
PERFORMANCE The performance of the LLL TV sensors is measured in terms of resolution and S/N ratio. Resolution
Resolution is measured by using an optoliner in front of the camera which images a high-contrast TV test target, at different light levels, on to the sensor's input surface. One type of camera (HTH MX-camera) has been tested with three different sensors: (a) the standard Philips CCD N X A 101 1 (604 H x 576 V picture elements); (b) a single-stage intensified CCD XX1540; and (c) a super-inverter intensified CCD XX1570. The horizontal resolution in MHz is given as a function of the sensor illumination in lux in Fig. 3. In order to compare fibre-optic demagnification with electron-optical demagnification by an extra intensifier, resolution has been measured using a Teledyne camera with the following sensors: (a) a Hitachi HE 98222 AW MOS Image Device, 8.8 x 6.6 mm2, 384 H x 485 V picture elements; (b) the same sensor, but with tapered fibre-optic window (18: 11) and intensified with
Illuminance on Sensor (Lux)
FIG.3. Measured resolution as a function of the sensor illumination for three different sensors: (a) the Philips CCD NXA 101I; (b) the single-stage intensified CCD (XX1540); and (c) the superinverter intensified CCD (XX1570).In all three cases an HTH MX-camera has been used.
0
v) 0
2,
/
/ ~~
FIG.4. Measured resolution as a function of the sensor illumination for three different sensors: (a) the Hitachi HE 98222 AW MOS image device; (b) the XX1450 intensified Hitachi device with tapered fibre-optics; and (c) the XX1560 intensified Hitachi device. In all three cases a Teledyne camera has been used.
6
L. K . VAN GEEST
an 18 mm Gen. I1 wafer tube (XX1450); and (c) the same sensor intensified with a demagnifying hybrid tube 18/18 Gen. 11+ 18/11 Gen. I (XX1560). The resolution as a function of the illumination of these sensors is given in the corresponding curves shown in Fig. 4. Although the curves in Figs. 3 and 4 are based on subjective measurements, it can be seen that the super-inverter combinations offer the widest dynamic range. It is also clear that the limiting resolution is decreased when an extra intensifier stage is added. Signal-to-Noise Ratio
Another method of measuring the performance of image-forming systems is based on signal-to-noise ratios. This method has been developed by Geluk (1983). A 6% contrast grey and white vertical bar pattern is observed with the camera under test at various illumination levels. The dimensions of the bar pattern and total magnification are chosen such that a video signal with a frequency of 1 MHz is obtained. The noise signal is measured over a reference area with the same bar pattern, but now with horizontal bars (Fig. 5). After some electronic manipulations the signal-to-noise ratio meter gives the square of the signal-to-noise ratio (S/N)*. A second measurement is done with interchanged vertical and horizontal bar patterns and the average of both measurements is taken. This is necessary to balance out differences in image quality. In Fig. 6 the quantity (S/N)2 is shown as a function of target illumination for the HTH MX-camera using a Tarcus 75 mmfll.3 objective with the same three sensors which were measured in the resolution tests. From the curves it can be seen that at high light levels the signal-to-noise ratio is limited by structural noise. The structural noise increases as the
FIG.5. Picture of the target used for the signal-to-noise measurements, illustrating the principle.
Scene illuminance (Lux)
FIG.6. The square of the signal-to-noise ratio as a function of target illumination for an HTH MX-camera using the following sensors: (a) the Philips CCD NXA 101I ; (b) the single-stage intensified CCD (XXI 540); and (c) the super-inverter intensified CCD (XX1570).
Scene illuminance (Lux)
FIG.7. The square of the signal-to-noise ratio as a function of target illumination for a Teledyne camera using the following sensors: (a) the Hitachi HE 98222 AW MOS image device; (b) the XX1450 intensified Hitachi device with tapered fibre-optics; and (c) the XX1560 intensified Hitachi device.
8
L. K. VAN GEEST
number of components such as fibre-optics, MCP and phosphor screens is increased. At low light levels the ratio is limited by constant background noise as, for example, preamplifier noise, photocathode dark current and by quantum noise. In Fig. 7 similar curves are shown for the Teledyne camera using the same three sensors. It can be seen that, at very low light levels, the performance of the XX1560 super-inverter is better than that of the single Gen. I1 tube with tapered fibre optics. Another advantage of the high gain of the XX1560 and XXI 570 super-inverters is that faster phosphors with lower efficiency than those used in standard intensifier tubes can be used, and that shorter decay times are achievable than with standard combinations. CONCLUSION Using a solid-state image sensor intensified by a single demagnifying Gen. I image intensifier extends the range to light levels approximately two orders of magnitude lower. By adding a Gen. I1 wafer tube as preamplifier, very high gain can be achieved, allowing respectable TV images to be obtained at scene illuminations of less than lx. ACKNOWLEDGEMENTS The author wishes to thank Mr. H. Janssen, A. Meinen and G. Henneveld for their contributions to the development of these low light level sensors and Dr. L. Bosch for the performance measurements.
REFERENCES Van Geest, L. K. and Stoop, K. W. J. (1985). In “Adv. E.E.P.” Vol. MA, pp. 93-100 Geluk, R. J. (1983). Proc. S.P.I.E. 467,42
Low-light-level TV with Image Intensifier Tubes and CCDs J. C. RICHARD, D. RIOU and M. VITTOT Laboratoires d'Electronique et de Physique Appliquie, Limeil Brhannes. France
INTRODUCTION
Electronic image tubes such as image intensifiers show high performance in sensitivity, resolution and speed, but to analyse the detected information, a complementary stage must be added to temporarily store the image. Among the numerous possibilities, the use of a solid-state image sensor is of major interest for image recording, processing and display. This paper aims to compare calculated and experimental results for two types of device based on the combination of a first- or second-generation image intensifier with a fibre-optic (FO) input window CCD. After summarizing the calculation method, results for sensitivity, gain, signal-to-noise ratio and image quality are compared. This theoretical and experimental work extends a previous theoretical study comparing a large number of possible low-light-levelTV devices (Lemonier et al., 1985). The two devices are considered under normal operational conditions and the comparison is presented in terms of signal-to-noise ratio, taking the area of one pixel as the reference area.
CALCULATION METHOD Each system is a succession of detection, amplification or conversion devices (photocathode, MCP, screen, FO, CCD) having their own statistical properties which can be combined to determine the overall statistical distribution. Input signals are given either in photometric units or in photons per pixel and per frame in order to distinguish between the basic capability of detection and the availability of geometrical information (the device with the largest sensitive area will give the highest signal-to-noise ratio at a given illumination). All calculations imply the following assumptions or conventions: 9 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
10
J. C. RICHARD, D . RIOU AND M. VITTOT
(i) The unit of area corresponds to the area of a CCD pixel referred to the input face of the device by taking the demagnification factors into account. (ii) The integration time is taken equal to 20 ms. (iii) Conversion from photometric or energetic units into quantum units is required and we take 1 lux=3.9 x 10l6photons m-2s-' = lop2Wm-* (black body at 2854K, radiation in the 0.4 to 0.9 pm wavelength range). All the input parameters are converted into quantum units. The input and dark signals, as well as the physical features of each detection stage, are characterized by their two first statistical moments. Average values and variances are combined to obtain the mean output signal and its fluctuation at given illumination levels on the input face of the device-usingformulae due to Albrecht (1965). The gain of the system is assessed by computing the number of electrons collected in a pixel per incident photon on the photocathode. The CCD pixel saturation is another important parameter: taking into account the pixel CCD maximal charge-handling capability, the light level corresponding to saturation can be evaluated. The minimal input light level illumination corresponding to a signal-to-noise ratio (S/N) equal to 1 can be determined by successive iterations. The dynamic range is obtained by dividing the input signal for CCD saturation by the minimal detectable input signal at S/N = 1. Values calculated from the measured parameters corresponding to each system are listed in Table I.
EXPERIMENTAL EQUIPMENT
Figure 1 is a block diagram of the apparatus used for collecting the experimental data. The tube assembly is placed on an optical bench with the following facilities: light level adjustment; white and monochromatic light illumination; square-wave MTF measurement facilities; and a microcomputer-based analysis system for processing and displaying statistical data combining a programmable multichannel analyser, a video digitizer and a video memory. The most relevant parameters for this comparison are the read-out noise data evaluated as the signal-to-noise ratio at the output versus the input illumination level and the modulation transfer function (MTF) of the complete device. The MTF of the image intensifier plays a role in smoothing the noise effect and therefore has to be taken into account (Zucchino, 1976). A specific factor K will be introduced to fit calculated and experimental data.
11
LLL TV WITH IMAGE INTENSIFIER TUBES AND CCDS
TABLE I Parameters of the tested systems FO-CCD Philip N XA lOlI/FO Number of pixels Pixel area (A) Integration time ( I ) CCD quantum efficiency (P)(560 nm) CCD noise equivalent signal (uo) FO Lambertian transmission (T) Demagnification (MI)
2 x 288 lines x 604 points 15.6~ 10=156pm2 20 ms 0.17 el ph-' 70 el-' pixel frame-' 0.18 ph ph-' 2.33
Philips image intensifier tubes Is1 generation
2nd generation XX 1410 SPS I8 mm 18 mm I 350 pA Im-' 0,055 el ph-' 75 ph el-' 21 Im W-' 109 el el-' 1
xx losot
Photocathode diameter Screen diameter Demagnification (M2) Photocathode sensitivity
50 mm 16.5 m m 3.15 330 pA Im-' 0.052 el ph-' 307 ph el-' 34 Im W - '
Screen efficiency (Y) (1 l m - 4 x 10'5phs-')
-
MCP gain (G) Reduced variance ( V )
-
Complete system (calculated values) Total gain GT Input signal at S/N = 1 Number of photoelectrons at S/N= 1 Light level on the photocathode at S/N=I
0.49 el ph-' 154 ph pixel-' frame-'
8 el frame2 . 4 10-5 ~
13.8 el ph-' 50 ph pixel-' frame-' 2.7 el frame-'
I
7.5 x 10-5 lux
IUX
6 x lo5ph pixel-' frameInput signal at saturation 9.4 x 10-2 lux Light level at the photocathode for saturation Dynamic range 4000
I
2 x 10 ph pixel-' frame-' 3.3 x 10-2 lux 440
f Inverter diode.
$ Proximity-focused microchannel intensifier.
RESULTS The results of the laboratory tests and computer analysis, as well as the characteristics of the image intensifier and FO-CCDs, are given above. The parameters given in Table I apply to standard operating conditions without any optimization. Some of them, such as photocathode sensitivity or MCP gain, may differ between samples.
12
J . C. RICHARD, D. RIOU AND M. VITTOT
source
1
\ Iris diaphragm neutral and
'Fdgenemtion
n\,h
monochromatic filters
Multi__channel analyser
-
II
+ FO CCD I
V
selector Oscilloscop
Video -digitizer memory
vv
Video Monitor
Computer
L
-
FIG.1. Block diagram of the experimental apparatus.
The gain (GT) of the system is defined as the number of CCD signal electrons per incident photon. Its value is measured as the ratio between the CCD reset drain current ZRD and the photocathode current IPK, referred to the same surface area. The resolution capabilities of an image device can be defined by the modulation amplitude of the video signal corresponding to a sine-wave test pattern, i.e. by the MTF. In usual practice, the spatial resolution measurements are done with square-wave test patterns (black and white bars). The sine MTF can be obtained by applying the correction formula given by Coltman (1954):
"L
r(3n) r(5n) r(7n). . . R ( n ) = - r(n) + --+4 3 5 7
]
where R(n) and r(n)are respectively the sine- and square-wave response factor to the different frequencies n. For the second-generation system, the test patterns are put into contact with the FO input window of the photocathode. In the first-generation tube, the photocathode is spherical; it is then necessary to use a lens between the test
LLL TV WITH IMAGE INTENSIFIER TUBES AND CCDS
o
i
2 3 4 Spatial frequency (Mhz)
5
13
6
(a)
0
1
2
3
4
Spatial frequency ( Mhz 1
5
6
b) FIG.2. Resolution of the two LLL TV systems consisting of a first-generationdemagnifier image or a second-generationMCP image intensifier (- - -) each coupled to an FOintensifier (-) CCD; (a) experimental square-wave response and, (b) deduced sine-wave response.
patterns and the photocathode. The measured MTF is then smaller than the MTF of the system but, in this case, the MTF of the lens can be neglected. Figure 2 shows the measured square-wave MTFs and the deduced sinewave MTFs for the two combinations. Using the multichannel analyser, the signal from a given CCD pixel can be recorded for different light levels incident on the photocathode. The analyser gives a histogram of the pulse height distribution, which enables the determination of the signal-to-noise ratio. The variation of this ratio with input light level is shown in Fig, 3. Theoretical results are also shown in Fig. 3. They are deduced as follows, where the parameters are as defined in Table I.
14
J. C. RICHARD, D . RIOU AND M. VITTOT I
I
I
io2
4.10’
lo2
to4
103 Photons/pixel /frame
(a)
-2
Light level incident (Lux)
(b) FIG.3. Comparisonof the calculated(- - -) and experimental (-) variation of the signal-tonoise ratio versus the light level incident at the photocathode; (a) with light level given in incident photons per pixel per frame to eliminate the factor due to demagnificationand, (b)with light level given in lux at the photocathode.
The average signal is S=nGT, where n is the number of photons per pixel per frame. The variance a* of the signal is such that a2= as+ ci, where a is a constant depending on the system. Thus a = YPT for the first-generation system and a = (1 + V)G YTP for the second-generation system. If we now assume that, when one photon is incident, spatial resolution, the persistence of the phosphor in the image intensifier and the bandwidth of the video amplifier combine to produce a signal in K pixels, the addition of independent random variables gives SK= KS and &= Ka2.
LLL TV ‘WITH IMAGE INTENSIFIER TUBES AND CCDS
15
The variance of the signal in one pixel is then:
We can then deduce the expression for the signal-to-noise ratio:
Comparison of the theoretical and experimental for each device gives a determination of the value of K: for the first-generationsystem K = 5.7 and for the second-generation system K = 11.6. The variation of K between the two systems can be explained by the difference in the resolution capabilities of the parts of the system: the better the MTF of the tube, the smaller the factor K . The smoothing effect in the second-generationimage intensifier is thus larger than in the first-generation one. CONCLUSION Measurements of the signal-to-noise ratio of the two types of LLL TV imagers based respectively on first- or second-generation image intensifiers and CCD read-out are in good agreement with the calculated values if the resolution capabilities of the intensifier tube are taken into account. Such a comparison between measurements and simple calculations should be useful in designing night-vision equipment. ACKNOWLEDGEMENTS The authors wish to thank all the team involved in this work, J. Dietz, M. Petit, M. Vittot and especially M. Lemonier and L. Bergonzi for their valuable advice and useful discussions.
REFERENCES Albrecht, C. (1965). In “Diagnostic Radiologic Instrumentation,” (ed. Moseley and Rust), pp. 291-31 1 Coltman, J. W. (1954). J. Opt. Soc. Am. 44,6,468-471 Lemonier, M., Richard, J. C. and Piaget, C. (1985). In “Proc. Electronic Imaging Conference, Boston” Zucchino, P. (1976). In “Adv. E.E.P.“ Vol. MA, pp. 239-252
This Page Intentionally Left Blank
Low-light-level TV with II/CCD Coupled Devices: Relative Merits of Different Approaches H. ROUGEOT and P. GIRARD Thomson-CSF Electron Tube Division. Boulogne-Billancourt. France
INTRODUCTION
Charge-coupled devices (CCDs) are nowadays increasingly used for TV image pick-up by virtue of their good detection performance and their excellent reliability. However, the sensitivity of the CCD is not sufficient for use at very low illumination. One approach to low-light-level TV imaging consists in coupling an image intensifier (11) to a CCD (II/CCD) so as to combine the sensitivity of the intensifier with the sampled, low-impedance output of the CCD. Many different combinations are possible with such a coupling, calling for a careful choice of intensifier, coupling mode, image format and CCD. The aim of this paper is to explain the parameters that come into play for an optimum use of an II/CCD system, and to present the overall performance characteristics of several systems implementing different generation LIIs. THE CHARGE-COUPLED DEVICE IMAGESENSOR Theoretical and experimental evaluations described here were based on the TH 786 1 CCD produced by Thomson-CSF. The sensor is a CCIR TV mode, frame-transfer type CCD compatible with S inch optics. Its principal characteristics are shown in Table I. ELECTROOPTICAL PERFORMANCE OF II/CCD COUPLINGS
Since the usefulness of a CCD is limited at low illumination levels, the coupling of an I1 at the front end of a CCD raises its gain and improves overall sensitivity. The performance will depend on the intensifier chosen and the coupling used. We exclusively considered fibre-optic (FO) coupling since it enables the CCD to collect the photons emitted by the intensifier screen more efficiently. Moreover, this also reduces the overall size of the IIjCCD. The main parameters defining the performance of an II/CCD coupling are 17 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited
All rights of reproduction in any form reserved ISBN 0-12-014674-6
18
H. ROUCEOT
AND
P. GIRARD
TABLEI Characteristicsof TH7861 CCD Number of pixels vertical horizontal Pixel dimensions (pm) Photosensitive area Sensitivity (v 4 - I cm-z) (mv lux-’) Saturation voltage (mV) RMS noise in darkness (mV) Illumination range
2 x 288 384 23 x 23 6.62 x 8.83 (finch format) 3.3 100
500 0.1 5000
The TH 7861 is useable in the illumination range lo-* to 25 lux.
therefore light sensitivity (V lux-’); signal-to-noise ratio (S/N) for a given illumination; number of resolvable TV lines; input image diameter and component system weight and size. Sensitivity
The sensitivity (So)of an II/CCD system is the ratio between the CCD output and the system’s input illumination (measured at the photocathode). It depends above all on the intensifier’s gain and the coupling mode chosen, and is expressed as r
s ~ = sm2 ~K,,G,, where s is the CCD’s sensitivity to the 11’s output screen spectrum, T is the transmissivity of the FO coupling, m is the magnification, GL is the 11’s luminance gain (Cd mPz lux-’) and KO is the coefficient expressing the radiation index at the tube’s output. Signal-to-noise ratio
Noise in the CCD output signal has two distinct origins: noise directly linked to the CCD (A VCCD),composed of readout noise which is temperature independent, and dark signal, which is a function of temperature; and quantum fluctuation noise (A Vphot),which depends directly on illumination and detection performance.
LLL TV WITH II/CCD COUPLED DEVICES
19
In low-light-level television, CCD noise can be considered negligible compared with photon noise if the I1 gain is sufficiently high. In this case, the output signal-to-noise ratio due to photons (S/N),h, can be written as
where Spis the photocathode sensitivity (A lm-’), A is the pixel area (m2), t is the integration time (s), E is the input illumination (lux), q is the electronic charge (C), F is the intensifier tube noise factor and m is the magnification, i.e. the ratio of the pixel dimension to the corresponding photocathode dimension. The operation of the system at a given illumination level will depend on the probability of detecting an image element, e.g. a pixel, and is bound by a condition of the type: S/N > K.The minimum useable illumination Eminfor the system can thus be expressed as
Emln . =
K q Fm2 SpA t
This means that the lower illumination limit for operation can be extended by raising the photocathode sensitivity, by demagnifying the image size, and by having the lowest possible noise factor. However, if the intensifier gain is not very large, the CCD noise can become as important as photonic noise, if not more so, and must be taken into account.
Number of TV Lines
At relatively strong light levels image quality will depend on the systems’ spatial resolution and thus on the number of resolvable TV lines. For the CCD alone this is equal to the number of vertical pixels. With an II/CCD the modulation transfer function (MTF) of the intensifier and coupling should be taken into account, and can in certain cases degrade the CCD’s limiting resolution. In the present example this is 22 Ip mm-I, equal to the Nyquist frequency. Input Image Format, Dimensions and Weight
For a given CCD, the image input format is directly linked to the magnification of its system and affects the choice of objective lens, and hence the field of view. The system’s mechanical characteristics (dimensions, weight) are determining factors and depend on the choice of both the intensifier and coupling, as well as the input optics and power supply.
20
H. ROUGEOT AND P. GIRARD
TYPESOF COUPLING EVALUATED
A number of different II/CCD couplings using three I1 generations were studied. Their performance characteristics were first calculated theoretically and then measured experimentally. In all, five systems were studied. These are listed in Table 11.
TABLEI1 The five II/CCD combinations I . Generation I intensifier with 30/10 demagnification+ CCD Sp=300pA Im-' F=1, m=1/3 2. Two-stage cascaded generation I intensifier with 25/8 demagnification +CCD S,= 300 pA 1m-I. F= I , m= 1/3 3. Generation 2 intensifier+CCD Sp=300 pA lm-', F e 2 . 5 , m= 1 4. Generation 2 intensifier + 1/2 demagnifying fibre-optics+CCD Sp=300 pA lm-I, F ~ 2 . 5 , m= 1/2 5. Generation 3 intensifier + 1/2 demagnifying fibre-optics+CCD S,=lOOO pAIm-', F z ~ ,m=1/2
COMPARATIVE PERFORMANCE OF THE DIFFERENT IIjCCD COUPLINGS Sensitivity
The sensitivities of the II/CCD systems studied depend on the 11's operational gain, but are typically 5 V lux-' for combination 1 and 100 V lux-' for all the others. Note that the use of demagnifying fibre-optic coupling does not significantly increase sensitivity, since the transmissivity of the fibres is low (T=0.25 for a 1/2 demagnification). Signal-to-Noise Ratio (SIN}
Figure 1 shows the S/N variations as a function of illumination of each of the five II/CCDs. When the CCD noise is negligible, which is the case when the intensifier gain is large, the curves are straight lines, on a logarithmic scale, of gradient 1/2 whose respective positions depend only on S,,Fand m.The knee towards the bottom of the curves at low light levels is particularly noticeable with II/CCD system 1, whose sensitivity is low, and is due to the dominance of the CCD noise.
21
LLL TV WITH II/CCD COUPLED DEVICES
t
lo4 103
la2 lo-’
1
PHOTOCATHODE ILLUMINATION (LUX) FIG.I . Signal-to-noiseratio versus incident illumination for the five combinations.
Operating Illumination Range
In order to determine the operating illumination ranges of the different 11/ CCD couplings. Figure 2 shows the variations in signal ( S ) and noise ( N ) versus photocathode illumination for each of the five II/CCDs. The operating illumination range is the ratio of the II/CCD system’s maximum operating illumination to its minimum level. The minimum operating illumination is determined by the condition S/N = K and its maximum value corresponds to the CCD’s saturation level. Note that high gain II/CCDs have a rather low illumination range, since they are quickly limited by photonic noise. Image Resolution
This parameter is not critical since the number of resolvable TV lines under strong illumination is equal to the CCD’s vertical resolution, i.e. 300 TVL. The only exception is with II/CCD coupling 3, whose performance is slightly degraded by the MTF losses in the intensifier and coupling; here the value
22
H. ROUGEOT AND P. GIRARD
drops to around 220 TVL. However, if the use of CCDs with a greater number of pixels is envisaged, it will be necessary to make a distinction between the different II/CCD systems. There will then appear an advantage with generation 1 intensifiers on the one hand and a demagnification on the other. Other Parameters Overall dimensions and weight. There is a very clear advantage to be gained by using double proximity-focused generation 2 or 3 intensifiers, which are particularly compact. Reliability and lifetime. Microchannel plate intensifier tubes limit the lifetime 1 om’?
/
/
/
/
/
I
CCD SATIJRATION,LEVEL
// / /
/
..
-PHOTOCATHODE ILLUMINATION (LUX) CCD SIGNAL (mV1 lOOOa//, /CCO ,SATURATION, LEVEL
,
, ,,
/ //
CCD NOISE LEVEL
10-5
10” lo-* 10‘’ 1 PHOTOCATHODE ILLUMINATION (LUX)
lo4
(a) FIG.2a. CCD output signal level versus incident illumination for systems 1 and 2.
CCD SIGNAL ( m V 1 1o o o ' y / /
/
CCD NOISE LEVEL
ao SIGNAL
PHOTOCATHODE ILLUMINATION (LUX)
CCD NOISE LEVEL
PHOTOCATHODE ILLUMt NATION ( LUX 1
(b) CCD SIGNAL ( m V )
,,/!',O1OO
/// LEVEL/'/ Lcc p ' /SATURATION /
/ / / /
1-.-
--__-___-
-
CCO NOISE LEVEL
PHOTOCATHODE ILLUMINATION ( L U X ) (c) FIG.2. CCD output signal level versus incident illumination for (b)systems 3 and 4 and (c) system 5.
24
H. ROUGEOT AND P. GIRARD
of II/CCD systems since they are susceptible to after-image with prolonged fixed scenes. Generation 1-based systems therefore have a better lifetime characteristic. Over-illumination, gain control. Generation 2 and 3 intensifiers provide good gain and their microchannel plate additionally gives partial over-illumination protection. Generation 1 intensifiers should preferably be used with CCDs having antiblooming protection.
OVERALL II/CCD
PERFORMANCE
The overall performance characteristics of the five II/CCD systems evaluated are as follows. IZ/CCD 1. Image definition is excellent, and the absence of a microchannel plate eliminates all risk of after-image and ensures a good lifetime. On the other hand, gain control is more difficult to achieve and an antiblooming CCD is necessary. ZZ/CCD 2. This system gives good performance but the same remarks as above apply for gain control and over illumination protection. The overall size is large. TABLE111 Summary of results 1 2 First Gen First Gen One Stage Two Stage
Input diameter (mm) Magnification Sensitivity (V lux-’) Direction limit (lux) Saturation (lux) Illumination range SIN at lux Limiting resolution (TV lines) Microchannel plate
30
113 5 2.10-4 lo-’ 500
25 113 100 2.10-5 5.IOW’ 250
40
40
300
250
No
No
3 Second Gen
I1 1 100
4
Second Gen FO 112
22 112 100
5 Third Gen FO 112
18 1I2 100 3.5 10-5
5.10-4 1.25 10-4 S.lO-’(AGC)* 5.10’ (AGC)* 5.103(AGC)* 10 (+AGC) 40 (+AGC) 150 (+AGC) 8 16 28 200 250 250 Yes
*Automatic Gain Control
Yes
Yes
LLL TV WITH II/CCD COUPLED DEVICES
25
ZZICCD 3. This system can be made very compact and gain control can be obtained through the integral power supply. However, the illumination dynamic is low at high gain. Resolution is inferior to that of the other systems. IZICCD 4. This system is preferable to the previous one because of its lower minimum operating illumination. ZZICCD 5 . This system exploits the high sensitivity of the GaAs photocathode and forms a very compact unit of high performance. However, its lifetime is lower than that of the other systems and its cost remains high. The results are summarized in Table 111. CONCLUSIONS From the evaluations of the different II/CCD systems, the following two conclusions can be drawn: (i) generation 2 and 3 intensifiers produce highperformance II/CCD systems and should preferably be used with a demagnifying coupling. Overall size is small and their integral power supply allows easy gain control as a function of illumination. The microchannel plate generally protects against over-illumination, it makes the system susceptible to after-image and limits lifetime. The use of generation 3 tubes increases the cost; (ii) for a number of applications, generation 1 tubes provide a very attractive solution and can produce high-performance II/CCDs when used with demagnification. The latter reduces the effects of photon noise and tube noise. Such systems have good lifetime characteristics but require an antiblooming CCD and do not allow simple gain control. Overall size for such systems is larger, but their cost can be low. The choice of intensifier should be considered case-by-case and future improvements can be expected with the use of new CCDs (e.g. Thomson-CSFs TH 7864 and TH 7866) offering substantially higher resolution.
This Page Intentionally Left Blank
A Fibre-optically Coupled Intensified CCD Camera System P. R. TOMKINS Photonic Science Ltd.. Tunbridge Wells. Kent, England
INTRODUCTION
The blue response of most detectors is poor and in the near-UV the response is virtually zero. It is in this spectral range that the photocathode-based tube still has a useful role. Some very sensitive prototype systems have been developed (IPCS, IPD, etc.), but most are photon-counting systems, limited in dynamic range and in their ability to handle any but the lowest light levels meaningfully. To fill this gap, a new detector system has been designed to function at light levels above the photon-counting regime and thus work in the analogue mode. The system consists of a first-stage proximity diode tube with an S.20 photocathode on a fused-silica window offering a spectral response from longer than Hcl, through the visible up to the UV atmospheric cut-off. The gain of the proximity diode was too small to enable the system to have a useful dynamic range, so a gain stage consisting of a commercial secondgeneration image intensifier was placed behind the proximity diode. Recognizing the excellent characteristics of the CCD led us to use such a device, working in the conventional video mode, as the output stage of the system. The CCD was coupled to the fibre-optic output screen of the second intensifier by a coherent fibre-optic taper, thus “projecting back” the CCD’s active area to fit just inside the circular format of the image intensifier. This arrangement produced a sensitive input area with an I8 mm diagonal and a standard TV 4:3 format. The use of proximity focusing in both tubes resulted in distortionless imaging. SYSTEM DESCRIPTION
The input stage consisted of a 25mm proximity diode with an S.20 photocathode on a fused-silica window. The cathode-phosphor gap was 3.5 mm, the phosphor being a P-20 type with an aluminium backing deposited on a fibre-optic output window. The peak wavelength of emission of the phosphor was 530 nm. The device has an integrated high-voltage power 27 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copynght 0 1988 Academic Press Limited All rights of reproducbon in any form reserved ISBN 0-12-014674-6
28
P. R. TOMKINS
supply to provide the necessary EHT up to a maximum of 21 kV. The device alone showed a resolution of 38 lp mm-’. The radiant power gain at 530 nm could be varied by altering the input voltage, and hence in proportion the applied EHT, from zero to approximately 45 W W-I. The gain stage was an EEV P8304A “night-vision’’ type 18 mm microchannel plate intensifier. This had a fibre-optic input window with an s-25 photocathode which had a good sensitivity to the output wavelength of the phosphor of the first stage. A variant of the conventional night vision type tube was used, the device having a thin, plano-plano output fibre-optic window. Once again a P.20 output phosphor was used, its peak emission being well suited to the CCD response. The gain of this stage was fixed at about 11 000 (luminous). This corresponded to a radiant power gain (measured) of roughly 1100 W W-’ at 530 nm. The limiting resolution of this stage was measured at 28 Ip mm-I. The fibre-optic taper was made to order by Schott Fibre Optics. This component had a useful input diameter of 18 mm, tapering to 7.5 mm diameter over a length of 14 mm. The fibres were 10 pm in diameter at the large end, and extramural absorbing material (EMA) was included to preserve contrast. The resolution was not specified, but was quoted as being better than 80 Ip mm-’ by the manufacturer. The lambertian transmission of the taper in the demagnifying mode was measured as 10%. The CCD was a Phillips NXA 10 10 image sensor. This device has an image diagonal of 7.5 mm, its format being 604 horizontal elements by 576 vertical. It is a frame transfer device with a 4: 3 aspect ratio. A “B” grade device was used, the cosmetic quality of which was very good. The sensor’s architecture leads to a spectral response rather different from the previously available devices, having a peak response at 600 nm (typically 900 mA W-I) and about 90% of peak at 530nm, the peak emission wavelength of the intensifier phosphor. This CCD is therefore ideally suited as a readout device for image intensifiers using P.20 phosphors. The chip was supplied with a glass input window that was subsequently removed. Using a preparatory technique, the chip-carrier was joined to a ceramic carrier plate, and a 6 mm thick fibre-optic coupling plug was bonded directly to the silicon surface of the CCD chip. The coupling plug was made of fibre-optic material using 6 pm fibre size with EMA, and selected for high optical quality (freedom from sheers, chicken wire effects etc). The camera provided a composite video output capable of driving a 20 m cable which could be brought from the telescope to a control unit. The control unit enabled manual video gain-control or AGC selection and intensifier gaincontrol or switching. Three independent video outputs and a sync output were available for connection to video recorders, monitors etc. The specially developed drive electronics used the Phillips chip set provided to support their
INTENSIFIED CCD CAMERA SYSTEM
29
Proximity MCP CC D Tapper CCDdrive circuitry diode tube e n 7 7
y
1 FIG. 1. The detector head assembly.
CCD sensors. Considerable additional circuitry was necessary, but the use of surface mount techniques enabled the whole camera to be fitted inside a 50 mm length tube of 45 mm diameter. The camera was designed for unity gamma and provided CCIR-compatible video. The detector head is shown schematically in Fig. 1. It measures 150 mm long and 80 mm in diameter at its widest section. The fibre-optic taper was joined to the CCD coupler with a UV-setting photopolymer, and was checked for an extremely small and parallel optical film of adhesive prior to curing. All other optical couplings were made with an optical grease whose refractive index was matched to that of the core glass of the fibre-optic employed. PERFORMANCE
The monochromatic sensitivity is a function of the first-stage photocathode sensitivity, for which a figure can be calculated most easily at 530 nm as the product of radiant power gain of the intensifier and the coupling efficiency of the fibre-optic, i.e. 45 x 1100 x 0.05 = 2475. This sensitivity related to the unintended chip. The increase in sensitivity at other wavelengths is simply calculated from this number in proportion to the relative photocathode response at the wavelength of interest, compared to that at 530 nm. Luminous sensitivity is harder to calculate but was measured at a government laboratory, which found that an input illumination of 300 plux would produce peak white in the video signal with the first intensifier at full gain.
30
P. R. TOMKINS
The two intensifiers had limiting resolutions of 38 and 28 Ip mm-' respectively, giving a value of 22.5 lp mm-' for the combination. Assuming the fibre-optic taper had a resolution of 80 lpmm-', this degrades the resolution slightly to 21.7 lp mm-' measured at the small end of the taper which had a demagnification of 2.4, which corresponds to 52.1 lp mm-'. Passing through the coupler which had 6 pm fibres, this is further reduced to 49.7 lp mm-'. In television terms it is more usual to talk of lines than line pairs, so the system at this point can be considered to give 99.4 lines mm- or a line width of 10 pm. Convolving this with the pixel width of 10 pm we get 424 TV lines as the horizontal resolution. The vertical resolution is obtained by convolving the 10 ,um line width with the 7.8 pm vertical pixel spacing of the sensor producing 370 vertical TV lines. Practical measurements of horizontal resolution produce a value of only 340 TV lines. This difference from theory is in part attributed to the non-zero thickness of the coupling layers between the various fibre-optic components, which has been ignored in the above calculations. At maximum gain setting it was quite easy to see discrete photon events on a monitor, though the ion events, seen as large white spots saturating the channel plate of the second intensifier, became annoyingly high at about 30 per second spread randomly across the whole field. The ion-event rate fell to less than 1 per second at intensifier gain settings less than about half of the maximum value. No attempt was made to measure the pulse-height distribution for photon and ion events, though this may be undertaken in the future, since there is clearly the possibility of using the system as a photoncounting detector when run at high gain.
'
ASTRONOMICAL EXPERIMENTS Field trials were undertaken in both the UK and Brazil on astronomical telescopes. The observing system comprised the intensified camera system, a low-band U-matic video recorder, an integrating video frame store and three TV monitors. In addition, an automated filter wheel carrying narrow-band interference filters was mounted in front of the intensified camera head to allow narrow-band imaging to be carried out. The observing system is shown schematically in Fig. 2. It was immediately apparent that the camera provided an excellent aid to object acquisition, and real-time integration was possible with the frame store, enabling very faint objects to be located rapidly. COMMERCIAL VERSION Several improvements were made for the commercial version. (i) The NXAlOlO sensor was replaced by an NXA1011 sensor. These CCD chips are monochrome versions of a colour CCD which has a
31
INTENSIFIED CCD CAMERA SYSTEM
FIBRE-
DIODE
POWER SUPPLIES
TAPER
IRCUIT
GAIN CONTROL6
r MONITOR
VIDEO RECORDER
VIDEO AMPLIFIERS AND SYNC BUFFER
MONITOR
FIG.2. Schematic of system used in astronomical observations.
three colour strip filter in front of the sensor and corresponding output nodes are multiplexed together for monochrome use. The NXAlOlO sensor produced an unattractive vertical patterning as a result of this process. With an NXA1011 sensor this is greatly reduced. (ii) A direct coupling was made from the taper to the CCD. This refinement of the bonding process, eliminating the coupling plug, doubles the light transfer efficiency from intensifier to sensor. (iii) Additional control facilities were included: a control unit was produced for the camera allowing easy user control of intensifier gain, video gain, black level and a choice of sync input enabling phase locking to mains frequency, internal quartz crystal, or the use of gen. lock input. THEISIS CAMERA SYSTEM
The UV two-stage system generated considerable interest, but it became apparent that many applications required a more economical version of the detector and did not require UV sensitivity. It was decided to build another version in which the two intensifiers were replaced with one high-gain “nightvision” type. A Mullard XX1500 series intensifier was chosen as being most suitable, offering a luminous gain approaching 100 000. This device has an S -25 photocathode on a fibre-optic window, which severely limits the UV response, but the use of such a window enabled coupling to other fibre-optic components for X-ray imaging and a wide range of other applications. PERFORMANCE OF THE
ISIS CAMERA
The monochromatic sensitivity at 530 nm compared to an unintensified
P. R.
32
TOMKINS
sensor is given by the product of the radiant power gain of the intensifier and the coupling efficiency of the fibre-optic taper. These numbers, measured from typical samples are approximately 8000 and 0.1 respectively, giving a value of 800 for the increase in sensitivity. A sample measured for luminous sensitivity produced a peak white in the video signal at around 700 plux. The limiting resolution of an XX1500 intensifier is 34 lp mm-' on average, and gives rise to a theoretical 500 lines horizontally. Practical measurements have regularly produced results in excess of 450 horizontal lines. COMPARISON OF
THE
Two SYSTEMS
Properties of the two intensified CCD camera systems are summarized in Table I. A noticeable difference between the two systems was that the twostage system produced a visually more pleasing picture than the ISIS camera (ignoring the ion spots) when the two systems were set to similar gains and viewing the same scene. The ISIS camera gave a considerably more grainy picture. The difference can be explained by the relative detective quantum efficiencies (DQE) of the two systems. The DQE is defined as the fraction of input events (in this case discrete photons) that give rise to detected output events. We also define responsive quantum efficiency (RQE) of photocathodes as the fraction of input photons that give rise to primary photoelectrons, this being a function of wavelength. Consider now two hypothetical systems with identical photocathodes (same RQE), along the lines of the two systems described above. The twostage system has a proximity diode at its input, and approximately 80% of the primary photoelectrons leaving the first photocathode penetrate the aluminium backing layer of the phosphor screen, each giving rise to several hundred output photons (the other 20% are either reflected or absorbed by the aluminium film). The DQE of the second stage can be shown to be easily high TABLE I Properties of the two camera systems
Input window Spectral response lntensification Luminous sensitivity for peak white video Intensifier/CCD coupling Horizontal resolution
UV System
ISIS
Fused silica 170-800 nm Proximity diode and secondgeneration wafer tube 300 plux
Fibre-optic 400-920 nm Inverting MCP tube
FO taper and straight coupler 370 TV lines
FO taper 450 TV lines
700 plux
INTENSIFIED CCD CAMERA SYSTEM
33
enough to detect unambiguously such multiphoton events as arise from the first intensifier, and so the overall DQE can be considered as being equal to that of the first stage, i.e. about 80%of the RQE of the first photocathode. In the second system, based around an XX1500, the primary photoelectrons strike the microchannel plate with an energy of several keV, which is beyond the optimum energy for peak electron detection efficiency.This combines with the effect of open area ratio for the MCP to reduce the usage of primary photoelectrons to an estimated 30%. This makes the DQE for such an intensifier only 30% of the RQE of its photocathode. The effect is in no way apparent from any of the published data for the intensifiers. This arises from the fact that both luminous and radiant power gains for intensifiers are the product of two components, namely the DQE of the system and the photoelectron gain, the latter being defined as the number of photons produced at the screen per detected photoelectron. What this means is that a reduction of DQE can be compensated by increasing the electron amplification in the tube without changing either the luminous or radiant power gain values. The output image will, however, be noticeably different. Consider the following three hypothetical intensifiers, all with the same luminous gain, all being fed with a pulse of one thousand photons: the first has a DQE of 0.1 % and a photoelectron gain of 10 000, producing one output event of 10 000 photons in a “single flash”; the second has a DQE of 1% and a photoelectron gain of 1000,producing 10 output events each of 1000 photons; the third has a DQE of 10%and a photoelectron gain of 100, producing 100 output events of 100 photons. Each intensifier gives the same number of output photons per input photon and therefore has the same gain. One images well, one poorly and one not at all when fed a pulse of 1000 photons. All three devices are practical, and could even have the same cathode sensitivity, so in terms of the published specification would appear identical! The ISIS camera, in a search for economy, has been forced to use a tube with a poorer DQE than the twostage version, with a corresponding reduction in performance. CONCLUSIONS
The use of direct fibre-optic bonding to CCDs has proved reliable over a two-year period, giving improved achievable sensitivity compared to the use of relay lenses. The UV and the ISIS intensified camera systems work well when linked to a frame store for real-time integrations. Both systems enable raw data to be recorded onto a video recorder for later processing. This permits the user to choose at a later date precisely which information is to be processed. Both types of detector are now commercially available and have been successfully used in applications ranging from telescope acquisition and guidance to looking through microscopes at cell-level structures.
This Page Intentionally Left Blank
A 3-stage 80:7 mm Image Intensifier Combination for the CERN UA2 Scintillating Fibre Detector L. BOSKMA, A. BAKKER, J. A. C. COCHRANE and K. W. J. STOOP Devt Electronische Produkten,Roden. Holland
INTRODUCTION At CERN, interactions between protons and antiprotons are observed with
a number of new systems (Ansorge, 1987). One of the main motivations for physicists is the search for certain electrons as a signature for production of top quarks. Protons and antiprotons travel in opposite directions through a beam pipe. A 2 m long, cylindrical detector, consisting of 24 layers of 1 mm diameter scintillating-plastic fibres (a total of about 60 000 fibres) encompasses the zone of collision. A resulting electron, passing through these fibres, causes scintillations forming a track with a wavelength peaking at 440 nm. At both ends the fibres are brought together in 16 bundles, 32 in total, each bundle being coupled to an image intensifier chain. These image intensifiers transfer the light information to CCDs. This article concerns the work involved in the development at DEP of these image intensifier chains.
THESYSTEM The intensifier chains had to be gateable together with specific requirements for photon gain, resolution, decay times and frame magnification. In close cooperation with scientists from CERN, an image intensifier chain was envisaged consisting of three tubes (Fig. 1). The first stage is an 80: 16 type (that is to say, with an 80 mm input diameter and a 16 mm output diameter), electrostatically focused demagnifying tube. The second stage is an 18 : 18 second-generation tube. The third stage is an 18 :7 first-generation tube. All three tubes have fibre-optic input and output windows. The spectral response of the photocathode of the 80: 16 tube has to match that of the scintillating fibres (Fig. 2). For the coupling between the first and the second stages, and for the coupling between the second and the third stages, a combination of P-47 fast blue phosphor and an S-20 photocathode was chosen (Fig. 3). With this combination a higher photon gain is achievable
35 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
FIG.1. The image intensifier chain.
50
n
I
Photocathode sensitivity 40
-
I
c
I
3 30-
\
U
-IE
g 2cVI
d fibres (a u 1
\ L
h 1
Wavelength A ( nm )
FIG.2. Spectral responses of 80: 16 intensifier photocathode and scintillating fibres.
A-
50
40
-
2 30 I
E %
t
H 20 [r
10
0
\ c
3 m
500
600
700
800
900
Wavelength X(nrn)
FIG.3. Spectral responses of S.20 photocathode and P.47 phosphor.
2kd-Bd-k Wavelength,X(nrn)
FIG. 4. Spectral responses of
P.46 phosphor and CCD.
38
L. BOSKMA E T A L .
TABLE I
Characteristics of individual tubes and the whole chain I Tube Frame magnification Cathode quantum efficiency (%) Photon gain Phosphor type Decay time (ns) Centre resolution at 50% MTF EBI (FIX)
I1
80: 16 18: 18 0.2 1 .o > 13 16 - 10 -800 Pa47 P.47 -50 - 50 > 27 >9 < 0.2 <0.2
I11
I+II+III
18: 18
0.39 16 -6 Pa46 -100 > 30 < 0.2
0.08
48000
than with a combination of a fast green P.46 phosphor and an S.25 photocathode. Ideally, the final output should spectrally match the CERN CCD. The best compromise available was a Pa46 phosphor (Fig. 4). Typical data for the individual tubes and for the chain, are given in Table I. The frame magnification of the complete chain is 0.08 reducing from 80 mm down to 6.5 mm. (CERN uses a somewhat smalfer rectangle in this 80 mm diameter.) Shown in the table are the quantum efficiencies of the photocathodes including their fibre substrates. These are equivalent to sensitivities of 50 to 65 mA W-I for 440 nm. The MCP is not run at its full voltage as there is already enough photon gain available, thus increasing the working life of the chain. It was required that the system be fast because CERN needs to gate immediately after an interesting event, allowing time to read the CCD without the image being spoilt by incoming light from further particle interactions. SPECIAL REQUIREMENTS
The DEP range of tubes includes the 80 : 18 tube as used for medical X-ray purposes in the Delcalex system. This standard tube uses glass input and output windows, so that a redesign was necessary to allow the use of fibreoptic plates. Also to change format from 80 : 18 to 80: 16, a recomputation of the electron optics was required. The dimensions of the second stage are basically identical to a DEP standard tube. The third stage required only a special fibre output to accommodate the CCD. The S.20 Photocathode
Though sensitivities at 440 nm of 60 mA W-l for the smaller tubes were
3-STAGE SO :7 MM IMAGE INTENSIFIER COMBINATION
39
achieved without difficulty, in the 80 mm tube the target of a minimum sensitivity over the complete input area of 45 mA W-’ required a great deal of investigation. It was found that the conditions of the process environment, such as substrate preparation and the wall materials, were more critical when growing a blue-sensitive cathode for the larger tubes than for the smaller tubes. Gain
In optimizing the screen efficiency, it was found that the classic sedimentation technique gave better results than the normal “brushing” technique, but also required that the optimum thicknesses of both the phosphor screen and the aluminium backing layer had to be found. A further improvement came from the use of fibre-optic plates from Francke Optics in Giessen (Germany) that have a significantly higher transmission for blue light. High- Voltage Problems
CERN wished to have both the input and the output planes of the chains at earth potential. This meant that there was 44 kV across two fibre-optics. This problem was not eased by the additional requirement that the cathode should not be exposed to the influences of high electric fields that could lead to deterioration in cathode stability. The solution chosen was to deposit a conducting layer on the outside of the cathode fibre-optic, electrically connected to the cathode flange. Thus there remain two fibre-optics, each with 22 kV across them, and this can lead to scintillations. To prevent these scintillations, the fibre-optics were aged. Finally, CERN required tighter mechanical tolerances than normal for a production tube. For example the parallelism of the output plane of the 80: 16, referred to the input plane should be better than 0.02 mm. ACKNOWLEDGEMENT
The authors wish to thank the many DEP employees for their help, particularly J. Hofman, H. Janssen and M. Zoethout.
REFERENCE Ansorge, R.E. (1987) CERN-EP/87-63 “Performance of a scintillatingfibre detector for the UA2 upgrade”
This Page Intentionally Left Blank
A 40 mm MCP Intensifier for Photon Counting T. J. NORTON, R. W. AIREY, B. L. MORGAN The Blackett Laboratory, Imperial College. London, England
P. D. READ Royal Greenwich Observatory, Herstmonceux, Sussex, England
and
J. R. POWELL Insrrument Technology Ltd, Hastings. Sussex, England
INTRODUCTION A 40 mm dual microchannel plate intensifier has been developed for use in high-resolution, centroiding, photon-counting systems for astronomy. The device has been designed in particular for incorporation into a CCD-based interpolative centroiding photon-counting system developed at University College, London (Fordham et al., 1986) and is intended ultimately to replace an EM1 99 12 4-stage magnetically focused cascade intensifier as the front-end detector in the above read-out system. A collaborative programme was set up involving Imperial College, The Royal Greenwich Observatory and Instrument Technology Limited to design and produce such an intensifier to the specification outlined in Table I. It is the purpose of this paper to describe the construction and manufacturing techniques required to produce an intensifier of the specification outlined above. Device performance and overall photon-counting system data have been given elsewhere in these proceedings? and by Read et al. (1987).
DETECTOR CONSTRUCTION The device incorporates an S.20 photocathode using proximity focus to a dual, double-length microchannel plate stack in the chevron configuration. The resulting electron cloud is proximity-focused onto an aluminized P.20 phosphor screen deposited on a fibre-optic window (Fig. 1).
t This volume, p. 425. 41 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any iorm reserved
ISBN 0-12-014674-6
42
T. J. NORTON ET A L .
TABLE I
Performance of 40 mm photon-counting intensifier Parameter
Design aim
Event gain (photons) RQE 420 nm (YO) DQE (%) Resolution (centroided FWHM) (pm) Ion count rate (counts cm-* s - ] ) SIB (Yo) Lifetime (10%RQE,,k decay) Phosphor decay (ps)
I
Achieved so far 10’ 12
107
20 12 25 1 10
5 33 0.5
1000 h 3
> 1000 h? 2.5
9
-
I
Spacing register
Photocathode
Intermediate electrode
FIG. 1. 40 mm photon-counting image intensifier.
The intensifier body is of metal and ceramic construction, metal rings providing the relevant electrode feedthroughs into the vacuum envelope for the microchannel plates, phosphor screen and tube getter. The S.20 photocathode is formed onto an input window which has high optical transmission in the blue. The input optic is sealed into a metal flange incorporating a knife-edge which is used to effect an indium cold weld to the intensifier body. Two, double-length (length-to-diameter ratio of 80 :I), “long life” microchannel plates are used in the “chevron” configuration. The MCPs are made by Galileo Electro-optics and have channel diameters of 12 pm with 15 pm
MCP INTENSIFIER FOR PHOTON COUNTING
43
channel spacing and an MCP bias angle of 8”. Such an MCP configuration was chosen in order to provide a saturated output pulse height distribution of modal gain 1 x lo7photons per photoelectron and at the same time to prevent positive ion feedback from the rear MCP and optical feedback from the output phosphor to the photocathode. The chevron arrangement provides a “line-of-sight trap” to positive ions liberated from the high-electron-flux regions of the rear MCP and also facilitates adequate optical attenuation to prevent any optical feedback. As a precaution against ion feedback from the front MCP, a 50 A Si02ion barrier film is deposited onto the front surface of the input MCP. Since the resistance of any two MCPs can vary considerably from plate to plate, an electrode in the form of a metal spacer of thickness 50 pm is placed between the two channel plates, thus relaxing the requirement for platematched resistance selection. Also, the ability to apply different bias voltages to the MCPs is very useful in enhancing the degree of saturation and modal gain of the output pulse height distribution as well as in optimizing the MCP conditioning procedure (see below). The radial emission energy of electrons from the front MCP leads to a degradation of device resolution in the interplate gap: however, this may be recovered in the final system centroiding. It has been possible to minimize MCP dark noise and arcing by careful design of the microchannel plate-retaining mechanism devised at Instrument Technology, Hastings. Finally a Pa20 phosphor is deposited onto a plane fibre-optic output window. The screen is then aluminized to produce an efficient and pinhole-free phosphor screen.
CHAMBER THE PROCESSING Once the intensifier has been assembled, the tube body and cathode optics are placed into a large UHV transfer chamber specially designed for processing the tube (Fig. 2). Using this chamber, it is possible to process two photocathodes and select the better for the sealed-off device. Removal of asperities such as small crystal growths on the photocathode, which could cause field-emission hot-spots in the completed tube, can be carried out using a “spot knocking” facility. This involves placing the cathode substrate over an anode plate and applying a large field gradient, typically 20 kV over 1 mm. The microchannel plates and phosphor screen can be monitored during MCP conditioning and under operation prior to seal-off. The various components in the chamber are manoeuvred in uucuo by means of a system of wobble-stick manipulators.
44
T. J . NORTON ET A L .
MICROCHANNEL PLATECONDITIONING STUDIES Before the S.20 photocathode can be sealed to the intensifier body, the two double-length MCPs must undergo a mandatory “scrubbing” process. This is designed to remove most of the residual gas species which can be both adsorbed and chemisorbed onto the large surface area presented by the microchannel plates (typically 0.5 m2). If this contamination is not removed, the device will have a very short operational lifetime owing to loss of photocathode sensitivity under bombardment from both neutral and ionized gas species released from the MCPs. In addition, the tube will display high noise and poor pulse height statistics (Fraser el al., 1987a). However, if the device is subjected to excessive input signals, the large induced flux of ions released from the MCP channel walls may cause premature gain fatigue. This is possibly due to modification of the semiconducting surface layer of the rear MCP reducing the secondary emission coefficient of the MCP glass (Authinarayanan and Dudding, 1976). Since a review of recent scientific literature concerning optimum MCP conditioning schedules yielded only ad hoc procedures concerning a large range of different channel plate structure regimes (Timothy, 1981, 1985; Sandel et al., 1976), studies were undertaken to optimize MCP scrubbing techniques for a dual, chevron MCP intensifier.
EXPERIMENTAL SET-UP The scrubbing signal was provided by an electron flood gun within the UHV chamber as well as an ultraviolet lamp shiningthrough a window from outside the chamber. Whilst the bias voltages applied to the MCPs were incremented upwards throughout the conditioning procedure, the voltage between the rear MCP and the phosphor was held constant. The following parameters were recorded at intervals during tube scrubbing: (i) pulse-height distribution-shape and modal gain; (ii) extracted charge from rear MCP; (iii) outgassing species-uantitative monitoring; (iv) visual inspection of phosphor screen. A lens and fibre-optic feed from the UHV chamber to a photomultiplier and pulse-height analyser allowed pulse height amplitude distributions to be recorded during MCP processing. An electrometer was used to measure the average anode current and a quadrupole-head mass spectrometer (Fig. 2) which was mounted in the chamber enabled the gases released by the MCP to be monitored. The resulting mass spectra were stored on disc by means of a microcomputer.
MCP INTENSIFIER FOR PHOTON COUNTING
45
FIG.2. UHV processing chamber.
Residttal Gas Analysis Figure 3 shows the residual gas background of the UHV chamber before MCP scrubbing. It can be seen that the main residuals in order of descending partial pressure consist of: (i) hydrogen from the stainless steel walls of the chamber; (ii) water from outgassing of the walls of the chamber and the mass spectrometer filament-reduced by chamber bake; (iii) methane from cracked hydrocarbons; (iv) carbon monoxide and carbon dioxide from reactions of system water with carbides on the hot filament; (v) noble gases which are not pumped by ion pump; (vi) at very small partial pressures cracked hydrocarbon
I Moss (n m u )
FIG.3. Residual gas background of processing chamber (hydrogen partial pressure= 5 x Torr).
T. J. NORTON E T A L
46
I
I
.
I
I
I
I
I
I
20 30 40 50 60 70 80 90 100 Mass (a.m.u.1 LM
FIG.4. Mass spectral scan prior to start of scrubbing.
groups from chamber components which have been leak-tested with rotary and diffusion pumped systems and chamber cleaning fluid contaminants, etc. With this background subtracted, the MCP scrubbing was initiated and the main outgassing components were recorded (Figs. 4 and 5). The principal gas species evolved was hydrogen, followed by methane, carbon monoxide, water and carbon dioxide. However the scrub-induced partial pressure of hydrogen was found to be greater than that of all of the other species by an order of magnitude. This is to be expected since, during manufacture, MCPs are reduced in an atmosphere of hydrogen and one would therefore expect this gas to diffuse out of the bulk glass owing to the resistive heating induced by heavy surface electron bombardment. In addition, any chemisorbed water from exposure of the MCP to the atmosphere will be dissociated under electron
-60 f O
FM
Mass (a.m.u.)
80 90 100 LM
FIG.5. Mass spectral scan after scrubbing for 5 minutes.
47
MCP INTENSIFIER FOR PHOTON COUNTING
bombardment, yielding hydrogen and oxygen. The oxygen will quickly react with other system residuals to form carbon monoxide and carbon dioxide. Dissociation of any hydrocarbons present will produce methane.
Device gain and pulse height distribution Figure 6 shows pulse-height distributions taken at intervals during the entire MCP conditioning schedule. It can be seen that the position of the saturated peak moves to the left, i.e. the modal gain decreases during the scrub. Plot 1 was taken at scrub start and plot 5 at the end. Figure 7 shows modal gain versus extracted charge for the entire conditioning schedule for tube number “MCP 3f”. At a particular MCP bias voltage setting, the ratio between the gains at the start and finish of scrubbing is approximately 1 1 : 1 . A total of 3000 mC was extracted from the rear plate of the chevron pair during the scrubbing cycle. 200
160
120 c u 0
$
80
d
40
0
50
I00 i 50 PULSE A b I P L I T U 0 E
200 (Channel
250
300
350
no.)
FIG.6 . Pulse-height distributions at different points in scrubbing process for intensifier MCP 3g.
Physical model of scrubbing process A model for reversible secondary emission yield in MCP glass (Authinarayanan and Dudding, 1976) has been proposed to explain the relationship between the gain and the extracted charge (Fig. 7). An electropositive material such as hydrogen adsorbed at the MCP glass surface would serve to reduce the surface barrier potential and increase the secondary-electron escape probability. The removal of this contaminant under electron bombardment will cause
T. J. NORTON E T A L X
I
clean
UP
phase
X
0
0 0
X
X
O
.
O
*
0
.
0
0
O O
0
L 100
ma*
0
I
,,a0
2000
2100
1DOI
Extracted Charge (mC)
FIG.7. Modal gain versus accumulated extracted charge for intensifier MCP 3f
the energy band to bend upwards and subsequently reduce the secondary emission yield. A loss in MCP gain will then result. This model is proposed for the “clean up” phase seen in Fig. 7. If, however, one were to continue bombardment of the MCPs indefinitely, the gain would eventually fall at some large value of accumulated charge. This non-reversible gain decay is thought to be due to migration of alkali metals into the bulk of the MCP glass, causing a drop in secondary emission yield. However, it is not though to be significant over the input signal levels for which this type of photon-counting intensifier is designed. Conclusion From MCP Conditioning Studies
It is clear from experience gathered during the processing of very high-gain MCP image intensifiers that in order to produce a device with stable operational characteristics and useful lifetime, the following recommendations should be carried out: store MCPs before use at pressures < lop6Torr; apply a minimum vacuum bake-out of at least 48 hours at 350°C; use a slowly rising MCP bias voltage ramp during conditioning; use both UV and electron sources as scrub signal; ensure that the device is scrubbed to its gain plateau, i.e. its region of stable operation; avoid ion-feedback-induced self-scrubbing; finally, ensure that the UHV processing environment is free of hydrocarbons.
MCP INTENSIFIER FOR PHOTON COUNTING
49
We have found that electron scrubbing is particularly efficient and also seems to reduce significantly both the intensity and number of switched-on channels observed in the MCP stack. However, it seems that some degradation in counting efficiency may result from excessive electron bombardment during conditioning. Since we have exclusively used Galileo “long-life plates” throughout the intensifier manufacturing programme, we have no information concerning the lifetime characteristics of MCPs fabricated by other manufacturers. However, comparison of the above data with those published elsewhere (Fraser et al., 1987a) for MCPs fabricated by other manufacturers seems to suggest that the gain stability and outgassing properties of various MCP glasses are fairly similar. All of the phosphor screens so far incorporated into the intensifiers have undergone rigorous pre-ageing schedules. This is possibly why we have detected no outgassing due to electron bombardment of the phosphor screen.
PHOTOCATHODE SEAL-OFF After MCP conditioning the photocathode can be sealed to the intensifier body. This is effected by a hydraulic ram press acting upon the UHV chamber bellows system (Fig. 2). This has proved to be a very reliable method for vacuum seal-off. DETECTOR PERFORMANCE The achieved performance parameters of the detector are summarized in Table I. Gain Of the devices produced so far, all have exhibited gains between 0.5 and 1.5 x lo7 photons per photoelectron. This is more than adequate for efficient
photon-event detection using the CCD-based read-out system (Fordham et al., 1986). In addition, narrow, peaked pulse-height distributions have been achieved (Fig. 8). These enable efficient noise discrimination as well as minimizing the dynamic range requirements of the read-out system. Both these are very desirable features in photon-counting detectors. Resolution
Using the CCD-based read-out system, centroided resolution of 33 pm full width at half-maximum has been achieved (Read et al., 1987). However, we
50
T. J. NORTON E T A L . 800,
u) +
5 480 8
c 0
2 320 160
0
100 200
300 400
500 600 700 800
Pulse amplitude (channel no.)
900 1(
FIG.8. Optimum pulse-height distribution of intensifier MCP 3g.
expect to exceed this figure in future tubes since emission point defects have limited the maximum voltage which can be applied between the first MCP and the photocathode, limiting the achievable resolution. Improved spot-knocking of the photocathode should remove such defects, allowing field strengths of up to 1.6 kV mm-’ to be applied. With this field strength a resolution of up to 25 pm should be possible. Lifetime
The lifetime of this type of image tube is primarily determined by the MCP conditioning procedures discussed in a preceding section. The responsive quantum efficient (RQE) of the photocathode at 649 nm was monitored as a function of time and integrated output signal. Figure 9 shows the RQE of the tube MCP 3f for 150 days after seal-off. It is evident that there has been no discernible decay in sensitivity during this period.? It would be desirable to subject the tubes to accelerated lifetests at relatively large input signal levels, providing a more accurate measure of possible lifetime constraints. Absolutely no MCP gain decay has been observed in any device so far. All of the above leads us to believe that these devices can be manufactured with satisfactory lifetimes.
t No
deterioration in sensitivity occurred over a further period of 150 days after these measurements.
MCP INTENSIHER FOR PHOTON COUNTING
51
FIG.9. RQE of intensifier MCP 3f as function of days elapsed following seal-off.
Uniformity
No geometrical distortion is observed with the present device. Under flatfield illumination small filament-like areas of enhanced sensitivity can be observed in a region of the output of one of the intensifiers, MCP 3g. This is due to “crazing” of the ion barrier film, that is tears in the silica film on the front MCP through which electrons can pass unimpeded. These are thought to be caused by hydration and subsequent bowing of the MCP. It is also possible that either mishandling or inadequate provision of outgassing paths during vacuum system roughout can cause these film fissures. However, careful handling and storage of the channel plates will preclude these cosmetic defects. Dark Noise The inherent dark noise of an MCP is at a very low level. Typically 0.1 counts cm-2 s-I. One of the main causes of MCP dark noise is field emission, primarily from point defects within a channel (Fraser et al., 1987b). We typically observe about 3 to 6 hot spots or switched-on channels over the full detector area; however, we find that electron bombardment during MCP conditioning can reduce this number. More serious, however, are fieldemission points on the photocathode. Cleaner substrates and effective “spot knocking” can eliminate these point defects altogether. The ion count rate of the devices produced so far has been very low, of the order of 0.5 counts cm-2s-’; we believe that this may be a reflection of the efficiency of our conditioning schedule as well as the effectiveness of the ion barrier film.
52
T. J. NORTON E T A L
CONCLUSIONS An extremely compact, very high-gain, photon-counting image intensifier has been manufactured for use in high-resolution astronomical photoncounting systems. It is hoped in the future to be able to optimize the MCP conditioning of these tubes to such a degree that the present ion barrier film may be eliminated. This would enhance the DQE of the device by allowing the collection of electrons scattered by the MCP web and at the same time removing cosmetic defects which arise from the barrier film. The complete detector head, incorporating the dual MCP intensifier, demagnifying fibre-optic boule and CCD read-out camera is shown in Fig. 10. This complete system has been used successfully in astronomical trials a t the Royal Greenwich Observatory and as the detector on the University of Belfast echelle spectrograph at the La Palma Observatory. It is to be used in observations on the Anglo-Australian telescope in 1988.
FIG.10.40 mm photon-counting intensifier mounted in the UCL CCD-based read-out system.
ACKNOWLEDGEMENTS The authors wish to thank the staff of Instrument Technology, Hastings for their assistance and encouragement. We would also like to express our thanks to Drs J. Fordham and D. Bone at University College, London for help with the trials of the complete photon-counting system and for many useful discussions. T. J . Norton is grateful for a SERC Case Studentship.
MCP INTENSIFIER FOR PHOTON COUNTING
53
REFERENCES Authinarayanan, A. and Dudding, R. W. (1976). In “Adv. E.E.P.” Vol408, pp. 167-181 Fordham, J. L. A. F., Bone, D. A. and Jorden, A. R. (1986). Proc S.P.I.E.627,206-212 Fraser, G. W., Pearson, J. F. and Lees, J. E. (1987a). Presented at IEEE Nuclear Science Symposium, San Francisco, Ca., 21-23 October 1987 Fraser, G. W., Pearson, J. F. and Lees, J. E. (1987b). Nucl. Instrum. Methods A254,447-462 Read, P. D., van Breda, 1. G., Norton, T. J., Airey, R. W., Morgan, B. L. and Powell, J. (1988) In “Instrumentation for Ground-based Optical Astronomy”, Proc. 9th Summer Workshop in Astronomy and Astrophysics, University of California, Santa Cruz (To be published) Sandel, B. R., Broadfoot, A. L. and Shemansky, D. E. (1976). Appl. Opt. 16, 1435-1437 Timothy, J. G. (1981). Rev. Sci.Instrum. 52, 1131-1142 Timothy, J. G. (1985) Opt. Eng. 24, 1066-1071
This Page Intentionally Left Blank
Development of a CCD-Digicon Detector System R. G. HIER, W. ZHENG, E. A. BEAVER, C. E. McILWAIN and G. W. SCHMIDT Center for Astrophysics and Space Sciences. University of Calijornia. San Diego, La Jolla, California, USA
INTRODUCTION The CCD-Digicon, an intensified charge-coupled device (ICCD) detector system based on the incorporation of a thinned backside-illuminated CCD as the anode of a Space Telescope design Digicon tube, is being developed for application to astronomical observations with stringent requirements for accuracy, rapid temporal response, low background, and high two-dimensional resolution (Hier et al., 1982). This photon-counting system, through the use of magnetic deflection substepping and charge pulse centroiding read-out techniques, offers an excellent combination of enchanced resolution (over the number of picture elements inherent in the CCD) and uniformity of elemental response (Hier et al., 1985).The addition of a mesh photocathode extends this system’s applicability to the extreme ultraviolet waveband (Jia et al., 1984). Our main goal in this portion of the development programme is to demonstrate the feasibility of attaining sub-CCD-cell resolution with such a system by the use of the centroiding procedures and techniques discussed earlier. In order to investigate and optimize the important characteristics of system component design in detail before commitment to the production of a full-up system, a number of electron-bombardment imaging experiments have been performed, utilizing existing backside-illuminated CCDs in a demountable Digicon test stand. First, a modified slow-scan drive electronics system was used to investigate in detail the suitability of the charge distributions resulting from electron-optical imaging for centroiding operation. Next, an interim fast-scan system was assembled to verify that system performance operating at full video rates remained consistent with the goals of this effort. In this paper, we present results of these ongoing electron bombardment imaging experiments, demonstrating the potential for sub-CCD-cell resolution that this system offers. CCD drive and support electronics are currently running at the video rates (60 fields s-I) required for practical operation in the centroiding mode, and continued refinements of the control and datahandling system are underway. 55 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
56
R. G . HIER ET AL.
SLOW-SCAN SYSTEM TESTING
Experimental Set-up The basic foundation o f the CCD-Digicon test stand has been described in more detail elsewhere (Hier et al., 1979; Hier and Schmidt, 1985). The system has been upgraded to utilize an IBM PC-AT and associated peripherals for control, data storage, reduction, and display capabilities. For these experiments investigating the details of fundamental deposited charge behaviour in the CCD, the emphasis is on accuracy and low-noise performance. Data acquisition occurs in a slow-scan mode (roughly 1 0 0 p per cell) over an optionally definable area of interest on the CCD, utilizing a correlated double sampling scheme and a precision 12-bit A/D converter. The CCD employed here is an RCA SIDSOI-DS, a thinned device with 320 x 512 cells each 30 pm square, cooled to roughly 0°C to reduce thermal generation of charge in the cells, operated in a non-double-buffered mode with the ability to clock the area outside the region of interest at a rapid rate (a factor of 16 faster) in order to reduce the overall read-out time. The demountable Digicon configuration used here, with an open-ended Digicon tube, permanent-magnet focus assembly, trim focus and deflection coils, CCD, pre-amp, photocathode, light source, etc., all housed in a laboratory bell jar for vacuum operation of the entire system, provides considerable flexibility and interchangeability of components for experimentation purposes. Many of these components have been made available through Hubble Space Telescope programmes (Brandt et al., 1979; Harms et al., 1979) for testing in this system. For the present programme, ultraviolet images were provided by a range of targets photo-etched in aluminium on quartz face-plates, overcoated with a semitransparent palladium photocathode, and illuminated by a pen lamp producing the 253.6 nm Hg line. Slowscan testing was done primarily with a tube operating potential of 17 kV, where electron-optical focus for this demountable Digicon system occurs with no trim focus current. Operating in a more traditional uncentroided “analogue” CCD mode (many photoelectron events in each CCD cell), a number of images were obtained, and the system was found to perform quite well over a wide range of illumination and exposure conditions. Figure 1 shows blow-ups of four separate single-frame images of a standard US Air Force test pattern, taken over a range of a factor of 100 in exposure. In these frames, and particularly in the bright-exposure image in the upper left quadrant, aliasing becomes quite evident as the spatial frequency of the bar pattern approaches the sampling limit imposed by the CCD cell size (as seen in the groups of bars proceeding down the right-hand edge of each frame-the bars in the inner pattern here are
DEVELOPMENT OF A CCD-DIGICON DETECTOR SYSTEM
57
FIG.1. The central portion of four single-frame 17 kV test pattern images, taken at various levels of exposure.
completely unresolved). Clearly, integrating such frames of data in the traditional way would produce improved statistics for better overall images, but would do nothing to improve this fundamental resolution limit, which is something greater than cell size. However, data at individual frame exposure levels corresponding to that of the fainter frame in the lower right quadrant or less (all the way down to practical limits of around 10 to 3 counts per cell s-I, which is a factor of 10 above the expected CCD-Digicon dark noise-see Hier et al., 1982), where single-photoelectron events are clearly and unambiguously distinguishable, offer the possibility of substantially increased resolution by centroiding each distinct photoelectron event and then integrating these much more precisely determined events. This would of course be in addition to the other potential benefits of photon counting and electronic deflection sub-
58
R. G . HIER E T A L .
stepping in providing greater noise immunity and uniformity of elemental response (Hier et al., 1985).
Photoelectron Charge-spreading Results In order to investigate the potential for attaining sub-cell resolution by centroiding, a special photocathode was constructed by ruling a variety of scratches (ranging in width down to roughly 12pm as measured with a microscope) in the aluminium with a diamond scribe prior to palladium overcoating. Figure 2 shows a section of a relatively bright (single-frame) exposure, with a 12 pm scratch intentionally oriented almost parallel to the rows of CCD cells so that the scratch image passes slowly from one row to the next. The three basically orthogonal scratches are considerably wider and thereby also brighter. The faint dots that can be seen randomly scattered through the otherwise “blank” areas of the frame are single-photoelectron events from large scratches elsewhere in the image. These fell there during CCD read-out because, for these tests, the system was not shuttered or gated during read-out owing to a mechanical shutter failure-in future systems, this may be accomplished electrostatically or electromagnetically. In Fig. 3 are plotted the data values corresponding to the response along each of the CCD rows between the wider pair of vertical scratches, one solid line on the plot for each row; in addition, the dotted lines shown are copies of other rows’ plots,
FIG.2. A portion of a bright single-frame image,showing a nearly horizontal 12 pm scratch used for these centroiding studies.
59
DEVELOPMENT OF A CCD-DIGICON DETECTOR SYSTEM
3 145
I50
155
160
165
CELL COLUMN NUMBER FIG.3. CCD response, plotted by row of CCD cells, along the scratch image in Fig. 2.
shifted horizontally for easy comparison with the central row. The rather smooth and continuously varying overlap of these profiles, corresponding to most of the vertical extent (30 pm) of the cells, results in part from the finite size of the image of the 12 pm scratch (estimated to be 16 pm when convolved with the effects of the electron-optics) and in part from the charge-spreading. Figure 4 shows this central region of the narrow scratch in a stack of four different low-exposure frames; single-photoelectron events from the narrow scratch (and a few other random ones during read-out) are evident here, and numerous cases of the charge (a few thousand electrons) from a single event being spread (both horizontally and vertically) over neighbouring cells are discernible. Earlier work (Hier et al., 1979) predicted a flattened pulse-height distribution due to this spreading effect, dependent on the ratio of effective diffusion depth to cell width in the CCD. The pulse-height distribution resulting from the data of the type taken. here matches the predicted distributions quite well, in correspondence with the present overall CCD thickness of about 10 pm. In fact, the results from these tests seem to indicate that, if anything, slightly greater charge-spreading would be closer to optimal for centroiding purposes (especially in the vertical dimension, since vertical spreading in this CCD appears to be somewhat less than horizontal spreading), e.g. hit locations near the centre of a cell may be difficult to centroid as accurately as those nearer the edges due to a plateauing effect, with too much of the deposited charge remaining in the central cell. Therefore, the use of a CCD with a slightly greater depth-to-width ratio (a less-thinned substrate, greater diffusion depth,
60
R. G . HIER ET A L .
smaller cells, etc.) would be beneficial in this sense. However, even the present somewhat suboptimal situation is amenable to significant gains in resolution by centroiding, as evidenced by the following.
Preliminary Centroiding Results
A plot of several rows’ worth of data from a frame similar to those in Fig. 4 may be seen in Fig. 5. In this data we see four single-photoelectron events, along with some roughly centroided estimates of their actual hit locations. (We can also clearly see that the magnitude of noise inherent in the CCDDigicon system is small compared to the charge deposited by single photoelectrons.) Continuing this rough centroiding process over a number of acquired frames, the results in Fig. 6 are obtained. We see that the resulting grouping of hit locations is substantially narrower than CCD cell size, and appears to be in basic agreement with the expected size of the scratch image. (It should also be noted that, owing to the nature of the process of creating these photocathodes, the scratches do contain some irregularities, kinks, bumps, etc., which are not taken into account here.) In Fig. 7, the distribution of distances of centroided hit locations from the assumed centre line of the scratch image is shown, along with the (substantially wider) expected
F I G .4. Four frames of single-photoelectronevents along the narrow-scratch image.
61
DEVELOPMENT OF A CCD-DIGICON DETECTOR SYSTEM 1200
1100
L
1000 145
147
149
151
155
153
157
159
163
161
COLUMN NUMBER
FIG.5. Data showing four single-photoelectron events, with hit locations centroiding from charge-spreading. Solid line, row 1 12; crosses, row 113; diamonds, row 1 14; triangles, row 1 IS.
distribution for the uncentroided case, where the distribution would be due to the convolution of a 16 pm image with the 30 pm cell size. Even with these relatively poor statistics, the result indicates that resolution is likely to be limited by the size of the image.
t
~
log I10
117 I18
3 x 30pm0 CELL SIZE
,
146
148
150
1
152
1
154
1
156
1
158
1
160
~
162
~
164
CELL COLUMN NUMBER
FIG.6. Compilation of centroided hit locations along the narrow-scratch image.
~
62
R. G. HIER E T A L .
$
& n
s
2
-
lo-
n l h
8-
-
64 -
Expected without
\
-30
-15
I
I
0
15
I
30
45
Distonce from centre ( p m )
FIG.7. Centroided hit locations of Fig. 6 , shown relative to the centre of the narrow-scratch image.
VIDEO-RATE OPERATION Drive Electronics
In order to have a practical usable device which takes advantage of centroiding in this way, CCD read-out must occur sufficiently frequently that the density of photoelectron events within each single scan of the CCD is low enough to keep pulse overlap from significantly complicating the processing electronics. Read-out at normal TV-video rates (60 fields per second) would be suitable for application to many astronomical observations; for current CCD sizes, this corresponds to read-out speeds something in the neighbourhood of 140 ns per cell. The development of circuitry to provide CCD readout and centroiding in real-time operation at these rates has been discussed (Hier et al., 1985); as an intermediate step, to investigate and evaluate performance characteristics at video rates as a guide to optimizing the details of construction of such specialized circuitry, a video-rate CCD control and data-capture system was assembled for initial testing, using as many off-theshelf components as possible. Figure 8 shows a block diagram of the system used for this initial video-rate testing. Drive electronics for the CCD are provided by a commercially available RCA (TC2855C) CCD monochrome video camera intended for closed-circuit applications, modified by us to accept the slightly different architecture of the RCA CCD. The CCD is driven in double-buffered mode,
- -- - _ DIGICON Tube
CCD
Pre-Amp
Tar et
Chamber
El+ v l *Ref
- - -~- - - -
I
h
I
-
--
Composite mdeo
Modified Video Frame Grabber with LUT Display
Color
M a s s Storage U n i t s ME Hard Disk and EMS Tape Backup
80
CAMAC
Dciketihn
-
Control Unit 525 Line 30 Frames/sec
Diode
Graphics Display Card
Tv
CCD Camera
CAMAC Controller Interface
PC-AT Computer Image Acquisition Software Image Processing Software FORTH Programming System DOS Operating System
FIG.8. Block diagram of CCD-Digicon control and data-handling system for video-rate testing.
64
R . G . HIER ET AL.
producing 320 x 256 cell images at 60 fields per second using the separate onchip image and storage areas to acquire a new field while the previous one is being read out. This camera also produces video interlace (somewhat akin to Digicon substepping) by causing charge to be integrated into the potential wells formed by different sets of image area vertical electrodes on alternate fields. A low-noise pre-amp/driver circuit is employed at the CCD (in this case, inside the vacuum chamber) to deliver the signal to the camera unit. The RCA CCD camera includes some intriguing circuitry, used for the suppression of l/f amplifier noise and floating diffusion reset noise. This replaces the correlated double-sampling techniques commonly employed at slower scan rates, with claimed equivalent performance (Levine, 1985). The circuit uses a differentiator to remove base-band video and l/f noise, followed by a sampleand-hold (requiring only a single clock pulse per pixel for this purpose) to synchronously detect and restore the signal to base-band while suppressing reset noise. The camera then outputs the information from the CCD as standard RS-170 video, which is passed through a video amplifier to a commercial video frame-grabber (Datacube IVG- 128) installed in the PC-AT for image acquisition. The frame-grabber card also is used for display purposes, both for real-time viewing of the stream of photoelectron images as they come from the CCD-Digicon and for looking at reduced data off-line. During these experimental stages, the existence of the data in standard video format turns out to be quite advantageous in a number of ways. CCDDigicon output can be viewed directly on a standard video monitor even without the use of the computer data-acquisition system. Real-time viewing, both direct to monitor or through the frame-grabber, considerably simplifies system set-up and adjustment. It also provides important insights into system performance, in that the effects of adjustments to the system (e.g. CCD clocking waveforms and operating voltages, high voltage, focus, deflection, positioning, illumination, etc.) can be seen immediately and continuously. The format also makes for ready compatibility with a number of available video accessories, ranging from videotape recorders to complex real-time image-processing systems such as DigiVision’s FluoroVision (Hier and Schmidt, 1985). These accessories are used to enhance the visualization of subtle image details during manipulation of CCD-Digicon system operating parameters.
Preliminary Data A number of video-rate CCD-Digicon photoelectron images have been acquired in this fashion. Figure 9 shows an integrated, backgroundsubtracted full-frame image of the Air Force test pattern, taken at 20 kV and relatively high illumination (the shading effects result from non-uniform optical illumination of the photocathode in this particular test set-up). In
DEVELOPMENT OF A CCD-DIGICON DETECTOR SYSTEM
65
FIG.9. Full frame 20 kV bright test pattern image from CCD-Digicon operating at video rates.
addition to such specific images, considerable insight into system performance can be gained through real-time “playing” with the system. For example, the qualititative effects of, for example, image rotation and geometric distortion for out-of-focus conditions can be viewed dynamically. Or, by varying the high voltage below a few kV, one can obtain images corresponding to multiple-loop focus in the electron optics. We have seen recognizable images with as much as a 7-loop focus, at tube potentials of only a few hundred volts (for a given magnetic field, the electric field required for a focus condition varies as the inverse square of the number of loops). With reduced illumination levels, the single-photoelectron statistical nature of these CCD-Digicon images is readily apparent even in real-time viewing. Figure 10 is a single-frame image similar to that in Figure 9 but at these reduced levels. Again, single-photoelectron events are clearly discernible, particularly in the large, weakly illuminated block area of the test pattern seen at the left of the image. There appears to be somewhat more system noise (predominantly coherent noise seen in “rippling” of the background) than was seen with the slow-scan system, resulting from the commercial-grade electronics used here. (We also find that the single events experience some additional horizontal spreading by the anti-alias filter on the input to the commercial frame-grabber.) Pulse-height distributions obtained from data of this sort, while showing this additional noise, again agree with earlier predictions for such distributions including the effects of both noise and
66
R. G. HIER ET
AL.
FIG.10. Same as Fig. 9, but at reduced illumination to show single-photoelectron events.
charge-spreading, and examples of vertical charge-spreading can again be found in the data. CONCLUSIONS We have performed a number of investigations into the feasibility of employing charge-pulse centroiding in a CCD-Digicon detector system to improve actual resolution capabilities by a factor of a few in each dimension, utilizing both slow-scan and video-rate read-out techniques. The slow-scan work demonstrated basic electron-optical and CCD system performance with a good signal-to-noise ratio commensurate with our goals; the video-rate data, while somewhat noisier than desirable for optimal centroiding purposes owing to the commercial electronics used, still clearly show single-photoelectron performance at these rates. Based on the experimental results from these studies, as well as our earlier experimental and theoretical work, we feel confident that, with suitably developed electronics, a viable CCD-Digicon detector system utilizing real-time centroiding can be developed for astronomical uses. ACKNOWLEDGEMENT This work is sponsored by NASA under the Space Astronomy Ultraviolet Detector Development Program, contract NASW-3667.
DEVELOPMENT OF A CCD-DICICON DETECTOR SYSTEM
67
REFERENCES Brandt, J. C., Boggess, A., Heap, S. R., Maran, S. P., Smith, A. M., Beaver, E. A,, Bottema, M., Hutchings, J. B., h a , M. A., Linsky, J. L., Savage, B. D.,Trafton, L. M. and Weymann R. J. (1979). Proc. S.P.I.E. 172,254-263 Harms, R. J., Angel, R., Bartko, F., Beaver, E. A,, Bloomquist, W., Bohlin, R., Burbridge, E. M., Davidsen, A. F., Flemming, J. C., Ford, H. and Margon, B. (1979).Proc. S.P.I.E.183,74-87 Hier, R. G. and Schmidt, G. W. (1985). Proc. S.P.I.E. 535,298-301 Hier. R. G., Beaver, E. A., Schmidt, G. W. and Schmidt, G. D. (1979) In “Adv. E.E.P.”,Vol. 52, pp. 463-480 Hier, R. G . , Beaver, E. A., Bradley, S. E., Burbidge, E. M., Harms, R. J., Mcllwain, C. E., ’ Schmidt. G. W. and Smith, R. D. (1982). Proc. S.P.I.E. 363, 57-65 Hier, R. G. Beaver, E. A. and Schmidt, G. W. (1985). In “Adv. E.E.P.” Vol. 64A, pp. 231-238 Jia, X. Z . , Beaver, E. A. and Hier, R. G. (1984). Proc. S.P.I.E.501, 103-110 Levine, P. A. (1985). IEEE Trans. Electron Devices ED-32, 1534-1537
This Page Intentionally Left Blank
Increased Gain of Channel Intensifier Tubes by Pulsed Biasing B. W. NOEL Los Alamos National Lnboratory. New Mexico, U S A ,
M. R. CATES Enrichment Technology Applications Center. Oak Ridge, Tennessee, U S A ,
and
L. A. FRANKS EG & G Energy Measurements, Inc.. Goleta. California, USA
INTRODUCTION Although the most common use of proximity-focused channel intensifier tubes (CITs) is as optical image amplifiers, electrical gating allows them to be used as fast shutters. Except for gating, however, the effects of imposing transient or pulsed changes on CIT biases have received little attention in the literature. It would be desirable in many applications to attain the maximum possible gain, but not at the expense of reduced signal-to-noise ratio (S/N). In the experiments described here, we used pulsed biasing to increase gain; our measurements show a marked increase in gain over the DC gain without the increase in electronic noise and risk of damage that higher DC potential create.
CURRENT FLOWAND GAINMECHANISMS By using a simplified analysis (Noel et al., 1987), we can show that the current density at the input to the microchannel plate is
In Equation (l), ng is the volume density of the electrons, e is the electronic charge, m is the electron mass, V k is the photocathode bias, and V, is the “stopping” voltage required to prevent all photocathode electrons from entering the MCP at zero bias. The absolute value of V,, typically about 20 V (Levi, 1980), is a small fraction of Vk for normal applied.bias. 69 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright (3 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
70
B. W. NOEL, M. R. CATES A N D L. A. FRANKS
For our purposes, the first stage of a CIT consists of the photocathode, the MCP input up to the first point where essentially all of the entering electrons have caused secondary emission, and the gap between the photocathode and this point. Geometrically similar to a vacuum photodiode, this stage also acts electrically like a photodiode, except that the current leaving the MCP input is not identical to the photocathode current (as it would be for a photodiode anode). Instead, the current is multiplied by the first-stage secondary-emission ratio, of the MCP material; 6, is a function of the primary-electron energy, e vk. Eberhardt (1979) has shown that MCPs behave somewhat like discretedynode photomultipliers of n stages, such that the electron gain is G = d l P 1 , where 6 is the secondary-emission ratio of each stage after the first. Under normal operation conditions, the gain relationship can be approximated by log G = k (n - 1) log( VnZcp/n V,)
+ log y( VJ
J‘Jk
(2)
which is linear in a log-log plot of G versus V,, for constant v k . In equation (2), k is a constant describing the curvature of the secondary-emission function, V,, is the MCP bias, V, is the first crossover potential (where the secondary-emission ratio is unity), and y is the fraction of the MCP input area that will accept the electrons from the photocathode. The functional dependence of equation ( 2 ) holds until the MCP starts to saturate. Such saturation may be caused either by a sufficiently high current flow (spacecharge build-up) to reduce the MCP‘s electric field significantly or by charge depletion in the walls of the microcapillaries. The “dynode” spacing is a constant given by z = L/n, where L is the length of the channels. Note from equation (2) that, if Vmcp is fixed, the overall gain of the n - I stages following the first stage is also fixed; the first-stage gain depends on vk. The luminescent intensity L of a relatively thick layer of phosphor? is related to the accelerating voltage between the MCP output and the phosphor Vfh,by the expression
L =f(9(Vph- VOlP (3) where i is the instantaneous current striking the phosphor, V , is a threshold potential known as the dead voltage, and p is a number between 1 and 2 describing the exponential dependence of the particular material response (Curie, 1963). The function f ( i ) is linear for small currents and tends to saturate at high current densities, especially for low-energy electrons.
t “Relatively thick” layers are more than about 2 pm thick; such layers cause the impinging electrons to give up an appreciable fraction of their energy in the phosphor material. We used a CIT whose P.20 phosphor stratum was 20-50 prn thick.
INCREASED GAIN OF CHANNEL INTENSIFIER TUBES BY PULSED BIASING
71
Bias and Current Limits
The CIT requires close spacing between the photocathode and MCP input to achieve good focus. For a typical spacing of about 0.18 mm, vk is limited to about 200 V. This value is less than the breakdown voltage of an ideal gap because of non-uniform spacing caused by irregularities in the photocathode surface. Field emission from the “high” points can give enough electrons to damage the MCP input. The phosphor requires high accelerating potential for good luminescent efficiency, so the MCP output-to-phosphor spacing is larger, typically about 1.25 mm. This spacing limits Vphto about 5 kV. Neither charge depletion nor field collapse in the MCP-to-photocathode gap due to large space-charge current is likely to cause the photocathode current to saturate in CITs. Well before either of these effects occurs, the resulting MCP output current will become excessive. Destructive current flow in the MCP can result from either excessively large photocathode current at a normal gain or excessive gain. The latter restricts the maximum DC voltage across the MCP to about 800 V. Excessive current flow can create a dead spot in the tube by destroying several adjacent microcapillaries. In an extreme case, the entire MCP can be destroyed. Such damage may be caused by localized over-heating, probably near the output end where the current density is highest. One reason we pulse the MCP (or the photocathode) to increase the system gain is that MCPs operating at low duty cycles can tolerate output current densities that would be destructive if the tubes were operating continuously at those current densities. TESTPROCEDURES In our experiments, we used an ITT type F4111 (18 mm diameter active area) with a P.20 (green) phosphor and an S.20 photocathode. To determine the relative CIT gain, we measured the output with and and without pulsed bias voltage under three operating schemes: (1) fixed Vmcp Vphwith pulsed bias added to vk while varying the DC level of Vk; (2) fixed V k with pulsed bias added to the MCP DC bias while varying the DC level of Vmcp and Vph (their sum remaining constant); and (3) fixed Vk and V,, while varying Vph.Figure 1 is a schematic diagram of the CIT test configuration. A short-duration (approximately 9 ns full width at half-maximum) nitrogen-discharge lamp served as the light source. The pulse generator that provided primary timing for the test system served both to trigger the lamp and, through a second slaved pulse generator, to trigger a gate pulser and an oscilloscope. We adjusted the timing so that the light-output pulse occurred in the centre of the gate pulse. A radiometer monitored the phosphor output. The pulsed-bias source was a model 357A gate pulser designed and
0,
m
U t
0 >-
r m .c
2
Q t
J I 0,
0-
c o s t
u U t 1- J
U Q O
.--
c-
.-
C .-
OI t
Lv) U C
0
C .- J cy-t
o , o , 9
-Lo,
L L - .
--cf
k-
c,
L
t
3
rI
I
00
LD
+
I
.- .-
+-(I-
c _ _
-+
O J U
t
mnL o C
o, 0,
a t
U
U a t
x3
u a -t
X Y
U Q
I
Q
Y
v) i -
C .-
c
U
v)
.-0 U
.-U L
t
o
> r
t
L
U I J
-L
.- .-
c m n mc L
0)
0)
+ L X t
o
0
P U v)
0
.-
-0 0
v)
FIG.I . Diagram of the experiment setup used to acquire pulsed-gain data.
INCREASED GAIN OF CHANNEL INTENSIFIER TUBES BY PULSED BIASING
73
fabricated at the Los Alamos National Laboratory. It is an avalanchetransistor pulse generator whose pulse amplitude is - 260 V and whose pulse width is determined by an attached charge line. We fixed the pulse width at 680 ns throughout the test?. In practice, the gate-pulser output was connected to either the photocathode or the MCP. In the former case, the negativepolarity pulse was square and flat-topped. When connected to the MCP, the pulse was inverted to positive by a pulse transformer. The resulting pulse had a sloping top because of the non-DC response of the inverting transformer.
RESULTSAND DISCUSSION Pulsed Bias Added to
v k
The upper curve in Fig. 2 is a plot of signal amplitude versus vk for the CIT. consists of two parts: the D C bias, which was varied from -200 to 300 V, and the - 260 V pulsed bias from the 357 A pulser. The abscissa is labelled with both the DC value, which was determined accurately, and the total value, which included an estimate of the 357 A output. For very large total negative vk, the CIT was operating well beyond the bias limit established v k
+
A
v,
.= 40 C 3
2 30 .9
-
L
c L
u (0
-
20
250 I
0
200
150
100
50
0
-50 -100 -150
dc photocathode bias (V) -100
-200
-300
,
1
I ,
-400
Total photocathode bias (V)
FIG.2. Results of applying pulsed bias to the photocathode. Upper curve: raw data; lower curve: raw data divided by the secondary-emission ratio of the first stage of the MCP.
t This pulse width was long enough to ensure steady-state turn-on of any sample of CIT we were likely to test. The tube we actually used was turned on fully in about 10 ns.
74
B. W. NOEL, M. R . CATES AND L. A. FRANKS
by the tube component spacing and photocathode surface-uniformity estimates. In fact, with the 357 A pulse added to the fixed bias, a positive DC reverse bias was required to bring the total to the calibrated value of - 180 V. In this mode, the tube was gated on only during the 680 ns of the 357 A pulse. Comparing the output signal at Vk = 180 V and that from the highest total bias achieved, -460 V, shows a gain increase of about 50%. We observed no SIN degradation, nor was a background glow imposed over the image plane. The for the pulsed part of the bias provided little low duty cycle (about 5 x energy for additions to the noise characteristics established by the DC biases. Recall that the forward section of a CIT should act like a photodiode multiplied by the first-stage gain: namely, the secondary-emission ratio, which is a function of V,. If we divide the upper curve in Fig. 2 by 6,, we should obtain a photodiode curve. We did this using 6 versus accelerating potential data for SiOz glass (Sackinger, 1971) combined with data on 6 versus accelerating potential at 10” grazing incidence (Goff and Hendee, 1967). The resulting lower curve in Fig. 2 agrees very well with the prediction. That part of the curve to the left of - 10 V total photocathode bias follows the shape predicted by equation (1) up to the point where the photocathode current saturates. The current density is proportional to the output signal in the test setup, provided the experiment geometry is fixed and provided the tube remains linear through its electron multiplication and electron-to-light conversion stages. Equation (1) can be rewritten in terms of signal as S = BI V,+ Vk(”*,where S is the signal output and B is a proportionality constant. By fitting a square-root function through the data, we find the value of B to be about 1.8. Pulsed Bias Added to Vmcp
Because the MCP is the electron multiplier for the CIT, the dominant effect on gain by pulsed biasing should occur when one varies VmCp.When we operated the MCP of our particular tube at a DC bias greater than the factoryspecified nominal value of 686 V, we found a steady increase, with increasing bias, in background glow from the phosphor. It was obvious that, under these conditions, further increases in the DC bias would not only cause the S/N ratio to deteriorate quickly, but would also cause the tube to fail. For these reasons, we limited the DC component of V,, to 775 V, which is 25 V less than the absolute maximum specified by the manufacturer. Figure 3 summarizes the features of the pulsed Vmcpdata. The lower curve shows that increased DC bias (but with no pulsed component), results in gain increases somewhat greater than a factor of 10 over the practical range of MCP bias. We cannot directly compare the shape of the gain function with the expression of equation (2), because the value of V,, decreases as Vmcp
INCREASED GAIN OF CHANNEL INTENSIFIER TUBES BY PULSED BIASING
F
'C
/'
a
200-
g
100-
E v
4 -
-
v)
'5
0 6
75
550
660
I
I
800
850
650
750
7hO
dc bias on MCP (V) 900
950
1000
8AO I
1050
Total bias on MCP (V) (upper curve only)
I 1100 1100
FIG.3. Results of applying DC and pulsed bias to the MCP.
increases. However, the shape does show the greater-than-linear response typical of electro-optical devices with electron-multiplying stages. The upper curve in Fig. 3 shows a significant gain increase when 260 V of pulsed bias is added to Vmcp.Although the DC bias was varied over a range of only 225 V, the trend of the pulsed gain increase appears consistent with that of the lower curve; i.e. an extrapolated estimate of the value of the lower curve at 810 V agrees with the actual value, on the upper curve, where the total bias is 810 V. Also, when we added the 357 A pulse to Vmcp, the background, including the increased noise and the glow normally observed at higher DC biases, did not change measurably. Like the photocathode, then, the MCP can be operated well beyond its DC bias range with no noise increase and no measurable device deterioration. The upper curve reaches a limiting value of signal above a D C bias of 675 V (total bias about 935 V). Because the value of Vph decreases steadily as the value of V,, increases, the saturation of the signal cannot be attributed unequivocally to MCP saturation; it could be caused by saturation of the phosphor or by space-charge limiting in the phosphor-to-MCP gap. For this reason, we explored the effects of varying Vph while keeping the other biases fixed. Effect of Varying
Vph
The almost-linear curve in Fig. 4 is a plot of the light output when Vph is varied while the total value of Vmcpis fixed at 950 V (686 V DC plus the 357 A pulse). Vk was kept at - 180 V, whereas we selected for Vphan upper limit 10%
B. W. NOEL, M. R. CATES AND L. A. FRANKS
76
I
I
I
I
I
1
n
.-3 300
f
Y
a
200 .-6 v)
I
/ Decreasing “mcp
/
-
53 1 0 0 I
I
I
I
I
I
3
3.5
4
4.5
5
5.5
Phosphor bias,Vph(kV) FIG.4. Results of applying pulsed bias to the phosphor
above the tube specifications. When Vk and Vph are fixed, the current at the output of the MCP is constant for constant input-light amplitude, so any increase in phosphor output signal from pulsed bias is caused by increased electron energy that results from increased vph.The figure shows that over a range of about 1.5 kV the phosphor responds nearly linearly to the increased electron energy. Thus, for V,, fixed at 950 V, there is no saturation, not even for Vph as low as 3.7 kV. Equation (3) describes the CIT’s response to increased accelerating potential across the MCP-to-phosphor gap. The linearity displayed in Fig. 4 implies that the exponent p is approximately unity for this device, so the equation can be rewritten in terms of signal as S= C( Vph - VO),where C is constant. We also assume Vph > VOfor accelerating potentials in the kilovolt range (Curie, 1963). We then have SzCVph. For these data, C=200 units kV - I . In Fig. 4,the data plotted for “varying Vmcp”,are the data from Fig. 3 for which both V,, and Vphhad been varied. Like Fig. 3, Fig. 4 includes the 375 A pulse as part of the total Vmcp. The data (now plotted as a function of phosphor voltage) still show the saturation. The crossing point of the two curves is where the tube conditions are identical for both data sets; the correspondence of all three parameters (S, Vmcp,Vph) indicates correct normalization. The data below and to the right of the “varying VmCp” curve display a rapid decrease in S
INCREASED GAIN OF CHANNEL INTENSIFIER TUBES BY PULSED BIASING
77
as Vmcp decreases. This behaviour is consistent with that predicted by equation (2). For this example, V k and V, are constant and k varies only slightly over the range of Vmfpencountered here, so the equation can be rewritten S
= D( V,,,Jk‘”
- ‘I,
(4)
where D incorporates the constant terms. Equation (4)assumes no change in Vphrand there is no change for the pair of values at each position along the abscissa in Fig. 3. The only parameter difference between the abscissa values (one from the upper and one from the lower curve) is the 357 A pulse present on Vmcpin the “with pulser” case and not in the other. We can estimate the exponential dependence of the expression by taking the ratios of several of the pairs of values, at various Vphvalues along the abscissa of Fig. 3, to determine k(n - 1). We did this for values near the low- Vphend of the curves where we observed no flattening in the upper curve, and found k(n - 1) x 9. This ninthpower dependence of the signal on Vmcp completely dominates the first-power dependence on V p h . This can be seen clearly in the lower-right data (V,, < 686 V) in the “varying Vmcp)’curve of Fig. 4.In the upper-left portion ( VmLp > 686 V) of that curve, however, the signal limits, not increasing despite increased Vmcp. Our conclusion, based on the strong dominance of V,, on the signal, is that the MCP gain was limiting. Further Discussion and Possible Applications
The use of pulsed biasing to increase the gain is limited to relatively short duty cycles. We used only one pulse width (680 ns) and a low repetition rate (10 pps). We did not try to find the point at which the duty cycle becomes excessive. We also did not have available sufficient pulse amplitude to extend the total bias to saturation (or tube failure); only the MCP appeared to saturate. Indeed, the phosphor output was linear up to the maximum available bias. In practice, one could use pulsed bias in as many permutations as it is possible to superimpose pulses on the DC biases. For example, one could see low-duty-cycle optical transients fall within the “window” created during the pulsed bias. As another example, one could take short-duration slices out of long-duration light sources by shuttering the photocathode during the timepulsed bias is applied to the MCP and/or phosphor.
SUMMARY OF STUDYCONCLUSIONS We added pulsed bias to both V k and Vmcpof a CIT for configurations where the DC potentials on Vmr,,,Vk and V,,, were varied. Following are the main results of our experiments: (i) The photocathode responded, as predicted, like
78
B. W. NOEL, M. R. CATES AND L. A. FRANKS
a photodiode. Inclusion of a pulsed component of V k allowed it to be operated beyond its normal bias range. In this mode, the CIT gain increased by about 50% without SNR degradation; (ii) When the MCP operated in the lower part of its bias range, its gain followed about a ninth-power dependence on VmCp. The gain did not saturate, but only for total biases well above the limit of DC bias that the MCP could tolerate. The pulsed-bias component on V,, added about a factor of 30 to the system gain for the nominal DC bias settings of the CIT; (iii) The output signal was linearly dependent on V,, for nominal values of Vmcpand Vk.Jncreasing the value of Vphby 10% did not damage the tube; (iv) Adding pulsed components to both Vmcpand Vk and operating V,h 10% higher than normal can increase the CIT gain by at least a factor of 50 without significantly affecting the S / N . This technique is applicable to the study of transients that are shorter in duration than the biasing pulses. CITs specially designed to exploit pulsed-biasing techniques will probably operate at even larger gain increases. ACKNOWLEDGEMENTS We wish to thank John Cuny, ITT Electro-Optics Division, for participating in helpful discussions and supplying data on the CIT. We also wish t o thank James D. Schoenborn for providing technical support.
REFERENCES Curie, D. (1963). In “Luminescence in Crystals,” Methuen, London. pp. 291-292 Eberhardt, E. H. (1979). Appl. Opr. 18, 1418-1423 Goff, R. F. and Hendee, C. F. (1967). “Proc. 27th Annual Conference on Physical Electronics”. Massachusetts Institute of Technology Levi, L. (1980). Appl. Opt. 2,465-466 Noel, B. W., Cates, M. R., and Franks, L. A. (1987). Unpublished report Sackinger, W. M. (1971). In “Photoelectronic Imaging Devices” (ed. L. M. Biberman, and S. Nudelman), Plenum, New York. Vol I , p. 181
Influence of Output Electron Energy Distribution of Microchannel Plates on the Resolution of Image Intensifiers N. KOSHIDA and Y. KIUCHI Department of Electronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology. Japan
INTRODUCTION When a microchannel plate (MCP) is employed in an image intensifier the energy distribution of its output electrons (EDOE) is a very important factor affecting the resolution. Previously (Koshida and Hosobuchi, 1985; Koshida et al., 1985; Koshida, 1986) we showed that the EDOE of an MCP consists of a sharp main peak and a long tail, the latter component having a considerable effect on the resolution, In this paper, the resolution characteristics of proximity-focused intensifiers are evaluated by using the EDOE data reported previously. A method for suppressing the tail component is also presented.
OF MODULATION TRANSFER FUNCTION (MTF) SIMULATION
The output end structure of a proximity focused intensifier is schematically shown in Fig. 1. Because of the energy and angle distributions of the output electrons, the spot corresponding to one channel, as detected on a fluorescent screen in close proximity to the channel, spreads out beyond a channel diameter. This is shown by I ( y ) in Fig. 1 . The spread of the spot was simulated on the basis of some EDOE data using the Monte Carlo technique. The exit energy was sampled from the EDOE curve at an interval of 0.25 eV. The exit angle distribution of the output electrons was assumed to be Gaussian with a mean value of 12" (the bias angle) and a full width at half-maximum of 5" (Kato et al., 1986). With the emergence conditions for each electron thus determined, the computer can calculate its path in two dimensions from the ballistic equations, yielding information on the resultant spot for typical values of the MCP-screen spacing (1 mm) and potential difference V, (13 kV).The number of s a q l e s was 5 x lo4. The spot size can be estimated from the spatial distribution of the points of impact of the electrons striking the 79 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright (01988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
80
N. ROSHIDA AND Y. KIUCHI
MCP
SCREEN 1 mm
FIG.1 . Schematic of the output end structure in a proximity focused MCP intensifier. The spot corresponding to one channel on an output screen, I(y), is also shown. The origin of the y-axis is set to the lower output end of a starting microchannel.
screen (Schagen, 1974). The spot data were used as input to a subsequent calculation of the effective MTF for bar patterns.
RESULTSOF
SIMULATION
Figure 2 shows a typical EDOE observed in unsaturated operation (Koshida and Hosobuchi, 1985). The MCP used in this experiment was of the straight-channel type with a channel diameter d of I 1 pm, a length-todiameter ratio a of 40, a channel pitchp of 14 pm and a bias angle 4 of 1 2 O . t The penetration depth of the output electrode o was I .Od. The applied MCP voltage VAand output electron current Z, are shown in the figure. A typical example of the spot simulated from the EDOE in Fig. 2 is shown in Fig. 3. The spot size at half-maximum is about 35 pm, and the tail of the spot extends to about 100 pm. In Fig. 4, the MTF curves calculated from the spot data are shown as a function of V,. The MTF decreases rapidly with increasing spatial frequency of bar patterns, owing to a superposition of the adjacent spots. The MTF behaviour in Fig. 4 is very similar to that reported in the, literature (Csorba, 1985) with conventional proximity-focused intensifiers. This supports our assumption that the resolution is limited mainly by the
t Provided courtesy of the Hamamatsu Photonics K. K.
OUTPUT ELECTRON ENERGY DISTRIBUTION OF MCP
n
1.o
ul
c .-
C
g
--
0.5
81
Va = 1.0 kV
Ic =10nA W = 1.0d
w
Y
Z
0 0
20
60
40
ELECTRON
80
100
E (eV )
ENERGY
F I G .2. A typical EDOE of an MCP operated in the unsaturated mode.
UI
1.0 -
W =1.0d
c .-
V, = 1.0 kV V5 23.0 kV
C
3
$
-v
.3
0.5 -
1
0
I
0
I
I
DISTANCE
I
2 00
100 y (pm)
F I G .3. One channel spot simulated from the EDOE in Fig. 2.
output energy and angular distribution. When the exit energy is fixed at a certain value, the MTF depends more weakly on the spatial frequency than it does when calculated under thecondition that the exit angle is fixed. Therefore the influence of the energy distribution, especially of the tail component, on the MTF is more important than that of the angle distribution.
IMPROVEMENT IN THE EDOE
AND
MTF
To suppress the tail component in the EDOE, a thin film of KBr was deposited on the microchannel wall near the output end. The penetration depth of KBr into the microchannels was adjusted to the value w. The
82
N. KOSHIDA AND Y. KlUCHI
100 W = 1.0d V,(kV) VA = 1.0 kV 1 0 2-&
3-0
0 ' 0
I
10 SPATIAL
I
I
1
20
30
40
FREQUENCY ( I p l m m )
FIG.4. The MTF characteristics calculated from the spot data for different screen potentials.
-
I 1.0
VI
.c _
C
--
tn
VA = l.0kV W = 1.0d
0.5
w
v
z 0 0
20 ELECTRON
40
60
ENERGY
80
100
'E (eV)
FIG.5. Comparison of two EDOE curves in the unsaturated mode before and after the KBr deposition.
usefulness of this technique has been confirmed by preliminary tests with a single-channel multiplier (Koshida et al., 198 1). The experimental results for the MCP used above is shown by the dashed curve in Fig. 5 , together with the original EDOE. The details of the measurement are the same as described previously (Koshida and Hosobuchi, 1985). It can be seen that the tail component near lOeV is reduced. The corresponding spot size becomes somewhat smaller as shown in Fig. 6 , reflecting an improvement in the EDOE. Figure 7 compares the MTF characteristics before and after the KBr
OUTPUT ELECTRON ENERGY DISTRIBUTION OF MCP
-
W
=1.0d
C
VA
= 1.0 kV
G
Vs = 3.0 kV
h
2
.-
1.0
83
m
v
-
2 0.5 v
0 0
200
100
DISTANCE
y (pm)
F I G .6. One channel spot simulated from the dashed curve in Fig. 5.
0 ’ 0
I
I
I
10 20 30 SPATIAL FREQUENCY ( lplrnrn
I
40
FIG.7. Effect of a thin KBr film on the MTF behaviour.
deposition. At high spatial frequencies near the limiting value (1 / 2 p ) ,the MTF after deposition is about 10% higher than that before. In addition, the gain in the unsaturated mode was increased by a factor of 2. As previously reported (Koshida, 1986), an increase in w is also useful for suppressing the EDOE tail. The solid curve in Fig. 8 shows the EDOE for w=2.Od. After the KBr deposition, the tail component near 10 eV can be further suppressed, as shown by the dashed curve. In this case, the gain was several times higher than the original value. The MTF at high spatial frequencies was, however, similar to that shown by the dashed curve in Fig. 7. This is due to the fact that the tail component in the energy range beyond
84
N. KOSHIDA AND Y. KIUCHI
-
1.0
ffl
c ._
C
3
$
-
tn
,V = 1.0 kV
w = 2.0 d
0.5
-z h
W
0
0
20
40
ELECTRON
80
60
ENERGY
100
E ( eV )
F I G .8. Two EDOE curves in unsaturated mode for an MCP with a deep output electrode penetration before and after the KBr deposition. In both cases, the output current was about 3 nA.
-20 eV remains unchanged in relative intensity even after the KBr deposition. To obtain a substantial improvement in the MTF, more precise control of the deposition depth of KBr (or some other efficient secondary emission material) is necessary. CONCLUSIONS It has been shown that the resolution of proximity-focusing MCP intensifiers is limited by the EDOE tail extending to energies higher than l00eV. The tail component can be suppressed to a certain degree by depositing an efficient secondary emission material on the microchannel wall near the output end. This is also useful for enhancing the gain. When both the output electrode penetration and the deposition technique are optimized, it should become possible to obtain a well-peaked EDOE without effect on the gain; the most important limiting factor for the resolution would then be removed. ACKNOWLEDGMENTS The authors would like to thank Dr. K. Oba for his support and Professor S. Yoshida for his encouragement.
REFERENCES Csorba, I. P. (1985) In "Image Tubes", p. 94. Howard W. Sarns & Co., Indianapolis Kato, T., Kinoshita, K. and Suzuki, Y.(1986) J. Inst. Tefeu.Eng. Jpn. 40, 386 (in Japanese)
OUTPUT ELECTRON ENERGY DISTRIBUTION OF MCP
85
Koshida, N. and Hosobuchi, M. (1985). Rev. Sci. Instrum. 56, 1329 Koshida, N., Midorikawa, M. and Kiuchi, Y. (1985). In “Adv. E.E.P.” 64B,pp. 337-342 Koshida, N. (1986). Rev. Sci.Instrum. 57, 354 Koshida, N., Suzuki, S., Kunii, M. and Yoshida, S. (1981). J . Inst. Telev. Eng. Jpn 35, 753 (in Japanese) Schagen, P. (1974). In “Advances in Image Pickup and Display” (ed. B. Kazan), Vol. 1, p. 16. Academic Press, New York
This Page Intentionally Left Blank
MCP-PMTs as Ultra-fast Wide-band and Infrared-sensitive Detectors K. OBA, H. KUME, K. WAKAMORI and K. NAKATSUGAWA Hamarsu Photonics, K.K., Shizuoka-ken. Japan
INTRODUCTION Strong interest in ultra-fast, wide-band and infrared-sensitive de :C Drs has been continuously growing in the fundamental research fields of physics, chemistry and biochemistry as well as in many other applications. The situation has accelerated the development of new type of photomultipliers, where effort has been focused mainly on microchannel plate photomultiplier tubes (MCP-PMT) because of the MCP’s fast time response. Conventional MCP-PMTs have already achieved transit time spreads (TTS) of 50-60 ps, generating many new results from measurements of fluorescence decay times by time-correlated photon counting (TCPC) (Yamazaki et a/., 1985). Frequency-domain fluorometry has also been developed to investigate fluorescence decays and the MCP-PMT has also been employed in this field, where its wide bandwidth characteristics, extending to 2 GHz, although at low modulation (Lakowitz et al., 1986), are useful. Optical communication is one of the most important and rapidly growing fields, leading to a strong demand for a new type of detector with fast time response and infrared sensitivity up to 1.6 pm. This area is covered by the synchronous scanning streak camera (Tsuchiya et al., 1982). However, the MCP-PMT must also be considered to be a candidate in laboratory applications because of its wide bandwidth and the availability of S . 1 photocathodes. This paper describes new types of MCP-PMTs which have better time response, bandwidth and spectral response than conventional MCP-PMTs. The first is a proximity-focused MCP-PMT that incorporates an MCP with 6 pm diameter channels instead of the standard 12 pm diameter channels. Analysis of the electron motion inside the channel indicates that the time response of a 6 pm MCP is half that for a 12 pm MCP (Oba et al., 1976). The TTS of the tube is less than 30 ps. The second new MCP-PMT is the so-called triode type which has very wide bandwidth. The tube has about 160 ps rise time and 250 ps fall time and the frequency response shows good modulation up to 2 GHz. This makes the device very useful for frequency-domain 87 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 01988 Academic Press Limited All rights of reproduction in any ronn reserved ISBN 0-12-014674-6
88
K . OBA ET A L .
fluorometry and for characterization of the frequency response of optical communication components, especially, fibre cables. For infrared operation a MCP-PMT with an S . 1 photocathode has been developed. Although the sensitivity of the photocathode is very low in the region 1.3-1.6 pm, accumulation of signals by time-correlated photon counting facilitates good results. This device also has fast time response feature and is therefore a useful tool for measurements on laser diodes, optical fibres and so on in the infrared and subnanosecond regions. AN MCP-PMT INCORPORATING Two 6 p m MCPs Analysis of the simple parabolic motion of a secondary electron inside a channel shows that the transit time of the electron between collisions is proportional to the diameter of the channel. An improvement in the time response of a MCP-PMT has therefore been obtained by using a 6 pm channel diameter MCP instead of the standard 12 pm MCP. The experimental tube was made by incorporating two 6 pm MCPs into a proximity type device as shown in Fig. 1. The tube was accommodated in a specially designed housing including a coupling capacitor and SMC connector together with the resistor chain. Photographs of a 6 p m MCP and a 12pm MCP are shown for comparison in Fig. 2. The tube was tested in a TCPC system to measure its intrinsic TTS. Singlephoton excitation of the tube was made by illuminating it with second harmonics of a short light pulse (6 ps) from a synchronously pumped cavity dumped dye laser. The output of the tube was fed to a time-to-amplitude converter through a tuned Tennelec TC454 constant fraction discriminator. The TTS obtained in this test is shown in Fig. 3, where the FWHM is 28 ps. This curve is used as an instrument response function to deconvolute
ANODE
(a) FIG. I . (u) Cross-sectional view of the proximity type MCP-PMT.
MCP-PMTS AS DETECTORS
89
FIG. 1 . ( b) Photograph of the prototype MCP-PMT.
experimental decay curves of samples to obtain their intrinsic decay curves. The accuracy of this method is approximately 1/10 of the instrument response function, so that the time resolution of the system is about 3 ps which is almost comparable to the time resolution of a standard streak camera system. A TRIODETYPEMCP-PMT The rise time of the output signal from an MCP-PMT is determined by the spread of the electron cloud in the MCP and its transit time t between the MCP and anode. The value of z in nanoseconds can be derived from
z = 0.8 x 3.37 x d/JV
(1)
where d millimetres and V volts are the separation and applied voltage between MCP and anode respectively. If the rise time of the output electron cloud from the MCP is t then, the output signal from the anode would have a rise time of T, where
T = p - T 7 . (2) Unfortunately, f is not known precisely, but is estimated by computer simulation to be less than 100 ps. The rise time t is of course proportional to the channel diameter d. Equation (1) applies to a diode electrode construction as shown in Fig. 4(a), where the electrons from the MCP are uniformly accelerated towards the anode. The output current in the anode circuit therefore changes with the
90
K. OBA E T A L
FIG.2. Comparison of the structures of a 12 pm channel MCP and a 6 pm channel MCP.
electron velocity. A conventional MCP-PMT thus gives a longer rise time than that derived from equation (2). To improve the rise time of the MCP-PMT, the triode type shown in Fig. 4(b) has been tested. In this case, the output electrons are accelerated between the MCP and the mesh electrode and then travel between the mesh and the anode at constant velocity. The induced current between the mesh and the anode circuit then depends only on the amount of charge crossing this space. If the separation and applied voltage are the same in these two types, the transit time of the electron cloud between the mesh and the anode in triode type is a half of that in the diode type, giving an improvement in the rise time of factor 2.
MCP-PMTS AS DETECTORS 10'~
I
,
I
I
,
,
,
,
.
,
,
,
,
91 ,
~
,
,
TIME
FIG.3. Comparison of the TTS of the new MCP-PMT with two 6 pm MCPs (R2809) and a conventional 12 pm MCP type (R1564).
P
FIG.4. Schematics of (a) diode type MCP, (b) triode type MCP.
The structure and a photograph of the triode type MCP-PMT are shown in Fig. 5. The distance between the MCP and the mesh is chosen to be 2 mm, to reduce voltage breakdown. On the other hand, the separation between the mesh and the anode is 1.5 mm, to reduce the transit time. Because of the construction itself, the capacitance of the anode circuit in the triode is less than
92
K. OBA ET AL.
FIG.5. (a) Cross-sectional view and (b) photograph of a triode type MCP-PMT.
a half of that of a conventional MCP-PMT, resulting in considerable improvement of the fall time as shown later. The time response of the triode has been tested by exciting the tube with a 20 f 10 ps light pulse at 400 nm from a laser diode. The output was observed on a Tektronix 7854 sampling oscilloscope. The output waveform is shown in Fig. 6 , which demonstrates a rise time of 160 ps and fall time of 310 ps at an accelerating voltage of 2000 V. The rise time of 160 psis about a half of that of a conventional MCP-PMT and the fall time is also less than a half of the 700 ps observed with conventional MCP-PMTs. Figure 7 shows curves of rise time against the accelerating voltage together with a dotted line with slope - 1/2 as derived from theory. The large deviation of the curves from the dotted line means that the rise time is limited by the electron multiplication process inside the MCP. This idea is supported by the improvement in rise time that is obtained by increasing the voltage applied to the MCP.
FIG.6. Output waveform of a triode type MCP-PMT (R2566)with 12 pm MCPs.
500
I
I
1000
1500
20vc)
3( 0
Accelerating Voltage (MCP-Mesh) in V
FIG.7. Rise time of the triode type MCP-PMT as a function of acceleratingvoltage. Circles, V ~ c p 800 , V per stage; open triangles, V ~ c p 900 , V per stage; solid triangles,conventional MCPPMT; dashed line, slope
-t.
94
K . OBA ET A L .
10' k
R2566
$ loo; B 2
lo-' f
Since, in the triode type, both rise and fall times are considerably smaller than for the diode type, an excellent frequency response is to be expected. Frequency response curves for both types are shown in Fig. 8. The curves were obtained from the Fourier transformation of the output waveform when illuminated by a 30 ps laser pulse. MCP-PMT
WITH
S. 1
PHOTOCATHODE
Combining an MCP with an S . 1 photocathode gives a high-speed, IRsensitive PMT. Demand for such a detector seems to be growing, for example in such fields as optical communication, biophysics and semiconductor measurement. In particular, in the optical communication field, the range from 1.3 to 1.6 pm is of great interest. Figure 9 shows the spectral response of an S-1 photocathode. The quantum efficiencies at 1.3 pm and 1.5 pm are 6x and 2 x respectively. In a conventional device, this photocathode does not have a sufficient response in this region for satisfactory operation. However, if an S. 1 MCP-PMT is used in a time-correlated photon counting (TCPC) approach at high repetition rate, the signal can be detected satisfactorily. To improve signal-to-noise ratio and to reduce accidental coincidence in the measurement, the photocathode should be cooled to about -50°C, where the dark count due to thermal emission becomes less than 103 cps. The output waveform when illuminated by 1.55 pm laser diode was observed using a TCPC system and the results are shown in Fig. 10. Because of the good time resolution, the fine structure of the pulse is clearly observed.
t l o + ,
I
,
I
300 500
,
700
I
,
I
I
I
,
I
900 1100 1300 1500
WAVELENGTH [nm]
FIG.9.Typical spectral quantum efficiency of an S - 1 photocathode.
I " " " " ' I x103
PMT ;R1564(Sl) Input ; 1.55 pm (Laser Diode) 10 ns Pulse width 8 x 10' PhotonslPuise 5 x lo3CPS Temp. ; -30°C integration : 3000 sac
10 -
rz
$
8-
,' 1
::
6-
4 2 4 -
2-
k
0
2
4
6
8
10
12 Time (n sec)
14
18
I
18
20
FIG.10. Output waveform of S - 1 MCP-PMT when illuminated by the light pulse from a 1.55 pm laser diode operating at 5 kHz with 10 ns pulse width.
96
K. OBA ET A L .
CONCLUSION For advanced applications an MCP-PMT with a 6 pm MCP, a triode type MCP-PMT and an MCP-PMT with an S.1 photocathode have been developed. Because of the smaller transit time jitter in the 6 pm MCP, the new MCP-PMT attained about 28 ps TTS in a TCPC experiment. This stimulates a variety of applications involving fluorescence decay time measurements. The triode type MCP-PMT, which operates on a quite new principle, shows a 160 ps rise time. This tube also realized a shorter fall time than conventional types because of its smaller capacitance. Its frequency response reaches 2 GHz. The prototype tube still uses a 12 pm MCP; however, the introduction of a 6 pm MCP is expected to further improve the rise time to 100 ps. An MCP-PMT using an S. 1 photocathode showed its capability for measurement of light pulses at wavelengths up to 1.55 pm. Because of the fast response of the tube, the fine structure of the pulse waveform related to the selfpulsation effect was clearly observed. ACKNOWLEDGEMENT The authors would like to express many thanks to Dr. Yamazaki of the Institute for Molecular Science for helpful discussions and experiments.
REFERENCES Lakowicz, J. R. et al. (1986). Rev. Sci. Instrum. 57,2499-2506 Oba, K. et al. (1976). In “Adv. E.E.P.” Vol. 40A, pp. 123-139 Tsuchiya, Y. et al. (1982) Proc. S.P.I.E. 348,245-250 Yamazaki, I. el al. (1985). Rev. Sci. Instrum. 56, 1187-1194
Performance of a Photon-counting Microchannel Plate Intensifier with Wedge and Strip Image Readout 0.H.W. SIEGMUND Space Sciences Laboratory, University of California. Berkeley, California. USA
C . J. HAILEY, R. E. STEWART University of California, Lawrence Livermore National Laboratory. Livermore. California USA
and J. H. LUPTON KMS Fusion Inc., Ann Arbor, Michigan. USA
INTRODUCTION We have developed a 40 mm aperture sealed-tube imaging detector (Fig. 1) for photon-counting applications in astronomy and atomic physics. The device has a fibre-optic input window with a semitransparent bialkali photocathode deposited on its inner surface. The photocathode is proximityfocused onto a stack of three microchannel plates (MCPs), each having an 80: 1 ratio of channel length to diameter, placed in a back to back (Z) configuration. The signal levels on the wedge and strip electrodes of the readout anode, behind the MCP stack, are linearly proportional to singlephoton event X and Y positions. The event positions are deduced by direct digitization of the individual wedge, strip and zig-zag amplified signals, followed by calculation of the X and Y positions in software.
MICROCHANNEL PLATEOPERATION The MCP configuration used in this detector has previously been utilized in a number of other devices, both open-face (Siegmund et al., 1986), and sealed tube (Siegmund et al., 1987). Whilst manufacturing the image tube the MCPs were subjected to a high-temperature bake (> 35OoC), and a UV scrub (0.1 coulomb cmP2extracted). During these processes the MCP gain, for a given voltage, reduced by approximately an order of magnitude. Nevertheless, as with the other devices, we have been able to achieve saturated gains in excess of lo', with narrow pulse amplitude distributions (50% FWHM, Fig. 2).
97 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
FIG. 1 . Photograph of the completed detector.
Gain
FIG.2. Pulse amplitude distribution for the detector at a modal gain of lo’, illuminated with white light.
99
PERFORMANCE OF A PHOTON-COUNTING MCP INTENSIFIER
DETECTOR SENSITIVITY Quantum Eficiency
The quantum efficiency of the photocathode combined with the fibre-optic transmission is shown in Fig. 3. The quantum efficiency was determined by comparing the current collected at the microchannel plate input with the current detected by a standard photodiode using the same illumination. This method does not assess the microchannel plate detection efficiency for photoelectrons. The sharp rise in quantum efficiency from 3500 to 4000 A is a result of the short-wavelength cutoff in the fibre-optic transmission. A relatively thin bialkali cathode was fabricated, which gave a rapid drop off in quantum efficiency for longer wavelengths. The fibre-optic used was 10 mm thick, with EMA, and the transmission above 4000 A is only 50%. The peak quantum efficiency for the photocathode alone is therefore 15%, which is a typical value for bialkali cathodes on fibre-optic substrates. We expect the total detective quantum efficiency of the device to be -70% of the values shown in Fig. 3, based on the photoelectron detection efficiency of the microchannel plates (Fraser, 1983). No thin ion-barrier layer was used on the top MCP, but ion feedback effects have not been observed. Over the seven months since the device was completed, no significant change in the quantum efficiency has been detected.
-
N
$
'
.08 '
l
w
""~""~""~""'~""~"~~""
0 3000
3500
4000
4500
5000
5500
6000
6500
Wavelength ( 8, )
FIG.3. Quantum efficiency as a function of wavelength for the cathode and fibre-optic.
100
0. H. W. SIEGMUND E T A L
Detector Background
The detector background, in photon-counting mode, is dominated by two components, the MCP internal background and the photocathode thermal background. The pulse amplitude distribution for background events shows a peak, at the same amplitude as the single photoelectron peak (Fig. 2), which is due to thermal photoelectrons from the cathode. However, there is also an exponentially decreasing background component due to the MCP internal events. Biasing the cathode at the same potential as the top of the MCP stack, the pulse amplitude distribution of the MCP internal events alone was determined. The total background rate, and the MCP background rate, are shown in Fig. 4 as a function of the lower level threshold. The modal gain represents the signal amplitude at the peak of the pulse amplitude distribution. The MCP background rate is consistent with the rates measured for the same Z stack configurations in other applications (Siegmund el al., 1986). It is thought that this level of MCP internal background is a result of 40K decay in the MCP glass (Fraser et al., 1986). The contribution of the MCP background events is less than 4 events s - I over the whole detector, for a threshold of 20% of the modal gain. Using the same threshold, the photocathode background rate is less than 4.5 events s - I at room temperature. The low photocathode background rate is due to the low red response of the cathode.
s
,
w?!
.01
0
10
20
30
40
SO
60
70
80
90
100
110
Threshold ( X modal gain)
FIG.4. Background event rate as a function of the lower level counting threshold, for the cathode biased on and biased off.
PERFORMANCE OF A PHOTON-COUNTING MCP INTENSIFIER
101
IMAGINGCHARACTERISTICS Image Linearity The image linearity of the detector has been assessed by examination of an image of a pinhole array mask. The mask contains 50 prn diameter pinholes spaced at 2 mm and 4 mm intervals in the Xand Y directions respectively, and a central line of holes with 25 pm diameter. The distortion over the central part of the image is low, the most noticeable distortion being within 2 mm of the edge of the field of view. Figure 5 shows an expanded view of a quadrant of the image, demonstrating the edge effects. The barrel type distortion at the very edge is attributable to variations in the field strength, between the MCP stack and the readout anode, due to the finite thickness of the MCP back contact electrode.
.
..
.
.
*
I
C
&
0
* . I
a
.
c
.
. . . . .
..-
a
. -
.
*
.
.
*
*
H 2mm
.
C
.
X Position
FIG.5. Image of a section of the detector, illuminated with white light through a pinhole shadow mask, showing edge distortion effects.
Resolution The resolution of the detector is determined by several factors. The wedge and strip anode resolution and the lateral photoelectron spread are the major components. The MCP pore size (12.5 pm) and the ADC bin width are minor contributions at the level of resolution achieved here. The photoelectron spread (Siegmund et al., 1987) is 33 pm FWHM for the 300 pm proximity gap and the 300 V mm-' gap field used. The wedge and strip anode resolution
-
102
E
0.H. W. SIEGMUND E T A L .
h
100
I
10
I
I
1E6
I
I
t
,
,
1
I
I
,
lE7
Microchannel Plate Gain
FIG.6 . Measured resolution as a function of gain compared with the predicted resolution.
is determined by the amplifier noise and the partition noise, as described by Martin et al. (1 981) and Siegmund et al. (1986). The partition noise occurs as a result of statistical jitter in the charge distribution of the electron cloud, causing variations in the division of charge among the three electrodes. Since the position-encoding algorithms are continuous functions, the resolution is characterized by the width of the detector point-spread function, rather than a defined pixel format. By digitizing the wedge, strip and zig-zag signals to an accuracy of several more bits than the expected resolution, it is possible to examine the shape of the detector point-spread function. The single-photon event positions were electronically binned to an accuracy of 12 bits, giving a bin width of 10 pm. The value of the FWHM of the pointspread function was obtained from the individual pinhole results and is compared with the predicted resolution components in Fig. 6 . The experimental results agree well with the predictions. At the highest gain used, a resolution of 70 pm FWHM is achieved, which corresponds to 575 x 575 resolution elements.
-
-
-
Position Sensitivity
The position sensitivity, as distinct from the position resolution, is the ability to detect changes in the position of any given image feature.
PERFORMANCE OF A PHOTON-COUNTING MCP INTENSIFIER
103
a
E8
P
0
Expected Position Shift (Microns)
FIG.7. Observed pinhole image centroid position displacement as a function of the expected image shift.
Specifically, the position sensitivity limit 6s is the accuracy with which the centroid of the point-spread function can be determined, so that (1)
6 S g SR/(N)’.’
where 6R is the FWHM position resolution and N is the number of photons detected in the image feature. When the image feature is larger than the detector resolution, or when there is a quantization of the image by an element of the detector, the position sensitivity may be degraded. One such effect, the spatial binning of the microchannel plate pores, is significantly reduced because the photoelectron spread is significantly larger than one pore. The position sensitivity of the detector has been assessed by measuring the position shift of pinhole images when the light source was displaced by known amounts. The mean detected pinhole image shift, and the experimental errors, for a number of pinholes is shown in Fig. 7 compared to the predicted image shift. The errors correspond closely to the expected error, estimated using 10 pm is demonstrated, largely equation (1). Position sensitivity of dominated by the statistical errors rather than any limitation of the detector. We also found the image positions to be stable to within 10 pm over periods of hours when the source was not moved.
-
-
104
0. H . W. SIEGMUND E T A L .
Flat Field Many characteristics of the detector play a part in determining the flat field, including the quantum efficiency, the resolution, the image linearity, and any defects in the optical components. We have investigated the flat field by illuminating the detector with a pinpoint light source and accumulating an image with a large number of events (Fig. 8). The source is not particularly uniform, but it does suffice to show significant features in the detector flat field. The flat-field image has a number of prominent anomalies, the most striking of these being the oval zone. Histogram slices taken through the image reveal that the oval zone is totally insensitive. The other patches are, at worst, a factor of 2 less sensitive than the surrounding areas. Examination of the MCP background images suggests that these areas are solely due to variations in the quantum efficiency of the photocathode.
FIG. 8. Image of a flat field accumulation showing a dead spot and other low-sensitivity patches.
PERFORMANCE OF A PHOTON-COUNTING MCP INTENSIFIER
105
ACKNOWLEDGEMENTS We would like to thank the Electronic Vision Systems Division of Science Applications International for their participation in the fabrication of this device, Mr. H. Bissinger and the optics shop for anode substrate fabrication and also Mullard Ltd, England for supplying the microchannel plates and Mr. J. Hull for fabricating the anode. Thework was performed under the auspices of the US Department of Energy by Lawrence Livermore National Laboratory under contract #W-7405-Eng-48.
REFERENCES Fraser, G. W. (1983). Nucl. Instrum. Meth. 206,445-449 Fraser, G. W., Pearson, J. F. and Lees, J. E. (1986). Nucl. Instrum. Meth. (preprint, accepted for publication) Martin, C., Jelinsky, P., Lampton, M., Malina, R. F. and Anger, H. 0.(1981). Rev. Sci. Instrum. 52, 1067-1074 Siegmund 0. H. W., Lampton, M., Bixler, J., Chakrabarti, S., Vallerga, J., Bowyer, S. and Malina, R. F. (1986). J . Opt. SOC.Am. 3,2139-2145 Siegmund 0. H. W., Lampton, M., Bixler, J., Vallerga, J. and Bowyer, S. (1987). IEEE Trans. Nucl. Sci. NS-3441-45
This Page Intentionally Left Blank
A Multichannel Detector for Photon Correlation D. N. QU and J. C. DAINTY Blackerr Laboratory, Imperial College of Science and Technology, London, England
INTRODUCTION With the advent of the microchannel plate (MCP) photomultiplier tube (PMT), much attention has been given to the applications and performance evaluation (e.g., Oba et al., 1979) of this device. The Hamamatsu R2519 photomultiplier tube is a MCP-PMT with a matrix of 100 anodes. Each individual anode can detect the photon events incident on a small corresponding region of the photocathode. Therefore, by making full use of these anodes, both temporal and spatial information of photon events can be picked up simultaneously. In analysing time-varying images such as speckle patterns and scintillation of starlight due to atmospheric turbulence, the space-time correlation can be measured and used to deduce certain features of both light sources and turbulent media (Dainty et al., 1981). In this paper, we examine the operating performance of this MCP-PMT and present measurements of the detected autocorrelation and space-time correlation of a randomly varying speckle pattern. MCP-PMT EVALUATION The Hamamatsu R2519 MCP-PMT has an S-20 photocathode with a maximum sensitivity of 64 mA W-' at the wavelength of 420 nm. Its focusing method is electrostatic and the system magnification is 0.8. We have concentrated on several parameters which are important for photon correlation measurements. The evaluation results of these parameters are shown below. Quantum Counting Eficiency (QCE) The QCE is the combination of the quantum efficiency (QE) of the photocathode and the collection efficiency (CE) of both the focusing system and the MCP. The photons incident on the photocathode cause a certain number of photoelectrons to be emitted, dependent on the QE of the 107 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rlghts of reproduction in any form reserved ISBN 0-12-014674-6
108
D . N . QU AND J. C. DAINTY
QCE (Yo)
0.51 -
0.5
fl
-
0.4
Central anode Sideanode Side anode
0.3 -
-
0.2 0.1
-
0.0
1.6
. . r ' 1.7
1.8
1.9
2.0
2.1
2.2 2.3
(kV)
FIG.1. Measured QCE versus voltage applied across the MCP at room temperature (18°C);the QCE of the central anode is slightly higher than that of an anode located at the side.
photocathode. Owing to inefficiencies of collection, a certain number of the photoelectrons are lost within the tube and only a fraction of the electrons are finally collected by the anodes. In our case, the QCE is found to be substantially lower than the QE of the photocathode. As Fig. 1 shows, the QCE was measured against voltage applied across the MCP for central and side-located anodes when the incident light was at the wavelength of 632.8 nm. At the typical operating point (2.1 kV), the QCE is about 0.167'0, while the QE of the photocathode is given by the manufacturer as about 4.8% at this wavelength. Dark Counts and Afterpulsing
Figure 2(a) shows that the dark counts are quite low, and that some of them can be removed by thresholding. For different anodes, the dark counts are slightly different. A low dark count may be achieved at the expense of low QCE and long dead time, as shown later. Afterpulses are spurious output pulses that are time-correlated with the signal pulses. Afterpulses are generally of lower amplitude than the signal pulses and occur soon after the initiating pulses. A time autocorrelation peak can be observed in Fig. 2 (b)caused by the afterpulsing effect. This correlation function was measured using a very short sample time of 100ns for the incident light field with Poisson statistics, which implies that the light field has constant classical intensity and has an autocorrelation value of zero. The maximum afterpulsing value (Brown et al., 1987) C(O)*N, where C(0) is the
A MULTICHANNEL DETECTOR FOR PHOTON CORRELATION
109
CountstS
50
-
30 20 -
40
10
0
Central anode Sideanode
-
01.0
1.0
1.2
1.4
1.6
1.8
2.0
2.2
(kV)
-
0.5 -
0.0 O.Oe+O
1 .Oe-5
5.08-6
t (sec)
(b) F i e . 2. ((I) Dark counts versus voltage applied across the MCP at 18°C. (h) Correlation function showing afterpulse effect. Sample time 100 ns. Peak value is C(0) = 0.38. Average number of photon counts per sample time is N = 1.1 x
peak value of the correlation function (normalized in the sense of equation ( I ) below) and N is the average number of photon counts per sample time, is found to be 0.0420/;,,which is adequate for correlation measurements. The correlation length is spread over the first three channels of delay time, corresponding to 300 ns, or 3.3 MHz. Therefore, to avoid the afterpulse effect in correlation measurements, this tube should not be used to measure frequencies higher than 3 MHz.
D. N. QU AND I. C. DAINTY
110
Pulse Height Distribution
For an MCP-PMT operating in linear gain mode, the output pulse height distribution is exponential. This causes difficulties in choosing the threshold voltage of the discriminator, for example to eliminate low-amplitude noise afterpulses. To overcome these difficulties, it is necessary to operate in gainsaturated mode so that the smaller signal pulses are amplified to the same level as large saturated signal pulses while a gap still remains between the heights of signal and noise pulses. Therefore, all the signal pulses are expected to have similar pulse height at the output. By taking advantage of the difference between signal and noise pulses, a threshold voltage can be chosen to eliminate noise pulses and a higher signal-to-noise ratio can be obtained.
Counts/S
10000
EI
6000 4000 2000 -
A
8000
0 15
+ 0
I
I
I
I
I
I
I
20
25
30
35
40
45
50
2.0kV 2.lkV 2.2kV 2.3kV
Threshold(mV)
FIG.3. Pulse counting rate versus discriminator threshold at various voltages applied across the MCP.
TO evaluate the pulse-height distribution and obtain the best operating parameters, we measured the pulse counting rate versus voltage applied to the MCP for different threshold voltages of the amplifier/discriminator (Fig. 3). A fairly good pulse height distribution can be inferred from the fact the curve has a large plateau. This plateau should be taken as a reference for setting optimal operating parameters. In our case, an applied voltage of 2.1 kV for the MCP and a threshold voltage of 30 mV for the amplifier/discriminator are chosen as the best operating parameters.
111
A MULTICHANNEL DETECTOR FOR PHOTON CORRELATION
Maximum Counting Rate
When the tube is operated in a photon-counting mode, the output counting rate decreases if the incident power is too high. Figure 4 shows the relationship between counting rate and intensity incident on the photocathode in the cases of white and red (A=632.8 nm) light. In both cases, the maximum counting rate was found to be 45 kHz. It is found that the counting rate has a linear relationship with the incident power up to a rate of approximately 30 kHz. As the incident power increases,
20000
50000
1
~
1
~
1
~
1
F I G 4. . Counting rate versus incident intensity for ((I) white light and ( b ) light at 1 = 632.8 nrn.
1
~
112
D. N. QU AND J. C. DAINTY
the counting rate reaches a maximum and decreases thereafter because of the overlap effect. An incident power range which will keep the tube operated in a photon counting mode can be derived from the results. The range is rather small compared with that of an ordinary PMT.
Dead Time The dead time of this MCP-PMT is significant owing to the fact that a long time is needed to restore the charge to the saturated channels of the MCP. The dead time was determined by measuring the photon-counting distribution P ( n , T ) ,which is the probability distribution ofcounting n photons during the sample time T. A function F(n) = (n
+ 1)P(n + 1, T ) / P ( n ,T )
can then be calculated. It has been pointed out (Johnson et al., 1966) that for coherent incident light with Poisson statistics and with the sample time T much longer than dead time t,
F(n) = - ( 2 N t / T ) n
+ N(l + Nt/T)
where N is the observed average number of counts per sample time T. This relationship allows us to calculate dead time t if Na n d Tare known. Typical results are shown in Fig. 5, showing a measured dead time of about 4.4p.
0.1
0.0 0
4
8
12
(a) F I G .5 . ( a ) P(n, T )measured with sample time 2.0 x sample time, N = 6.8.
s and average number ofcounts per
A MULTlCHANNEL DETECTOR FOR PHOTON CORRELATION
0
4
8
12
16
113
(n)
(b) F I G .5. (b) F(n) calculated from P(n, T) giving a best fit to the line F(n) = 0.03 n implied dead time is T = 4.4 ps.
+ 7.12. The
Cross- talk
When the photocathode is illuminated by light restricted by a fine pinhole to cover a small area corresponding to one anode, signal pulses are detected not only from this anode but also from the adjacent anodes. The cross-talk is defined as the ratio of signals detected from adjacent anodes to that from the illuminated anode. There are two main causes of cross-talk: one is the spreading of photoelectrons inside the tube, the other is the capacitance between the anodes. The measurement shown in Fig 6(a) was made by scanning a pinhole image across the photocathode and recording the photon counts in a single anode. The result indicates that the anode is collecting signals from a circular area of about 2 mm diameter at the photocathode. The spatial averaging effect, as discussed later, may severely degrade the contrast of a measured correlation function. In addition to the source of cross-talk mentioned above, the capacitance between anodes also gives rise to cross-talk. As Fig. 6(6) shows, the cross-correlation between two adjacent anodes was measured at a very high sampling frequency (2 MHz). The incident light was spatially uniform and had Poisson statistics so that the light field did not itself introduce crosscorrelation between the two channels and the observed correlation was that between the original signal from one anode and the capacitance-induced signal from the other anode. There appears a sharp peak a t very high frequencies and a rather long tail. This peak covers the first two channels of
114
D. N . QU AND J.
C.
DAINTY
counts (“Yo) 100
-
60 40 20 80
-8
-6
-4
0
-2
2
6
4
8
(a)
c (%5>
5= 1.25 mm
0.0
:
O.Oe+O
I
I
5.09-5
2.5e-5
2
(sec)
(b) FIG.6. (a) Result of scanning a 0.2 mm diameter pinhole image across photocathode; (b)crosscorrelation with Poisson incident light beam showing capacitative cross-talk.
delay time, which corresponds to a frequency of 1 MHz. It is therefore clear that the capacitance between anodes is quite large and capacitance-induced cross-talk becomes dominant at frequencies above 1 MHz.
PERFORMANCE OF MCP-PMT
ON
PHOTON CORRELATION MEASUREMENTS
We are using this MCP-PMT to measure the correlation function of timevarying images. By considering the statistics of the secondary emission
A MULTlCHANNEL DETECTOR FOR PHOTON CORRELATION
f"
115
f3 AmpJdixrim
Rotating disc (Ground glass)
F I G .7. System for measuring correlation functions.
process for a random incident light field, the following equation can be derived (Jakeman, 1973). (n(t, T ) d t + u, T I ) - W t , T)l(r+ u, TI) , u f O (1)
2 (02 where n(t, T ) is the photon counts at time t over a period of T, Z(t, r ) is the intensity at time t, and u is the time delay. The angle brackets represent an ensemble average. Equation (1) has a very important implication: an unbiased estimate of the intensity correlation function can be obtained by means of correlating photon counts. In our case, the MCP-PMT is used to count photon events and these counts are then correlated by a digital correlator (Langley-Ford DC128). By making full use of an array of anodes, both autocorrelation and space-time correlation can be measured. The experimental setup is shown in Fig. 7; the correlation function of the speckle pattern produced by a rotating disc is measured at far field by the MCP-PMT system. The results are shown in Fig. 8. As mentioned earlier, the contrast was severely degraded by the finite detecting area. Figure 8(b) shows that the contrast could be enhanced to unity by mounting a pinhole over one element. Figures 8(c) and ( d )shows the cross-correlation in the x and y directions. Two dimensional cross-correlations can be easily determined by the tube. A zoom lens can be mounted in front of photocathode to adjust the distance between the two points to be measured.
CONCLUSIONS The performance of this tube for photon correlation has been evaluated. In general, its main advantage is that both spatial and temporal information can be measured with a single tube. The tube also has a good pulse height distribution and low dark count if it is operated optimally. The main drawbacks of the device are due to the fundamental characteristics of the MCP, such as long dead time and low detection efficiency.
116
D. N. QU AND J. C. DAINTY
Together with the long dead time, both afterpulsing and cross-talk limit the tube to measurement of frequencies up to 1 MHz. Furthermore, the finite detection area can severely degrade the contrast of photon correlation functions. This can be overcome by mounting a pinhole array in front of the photocathode.
C
"."
,
O.Oe+O
I
I
1 .Oe-3
2.0e-3
z (sec)
(b) FIG.8. (a) Temporal correlation function measured using the MCP-PMT; spatial averaging has degraded the contrast C(0). (h) Temporal correlation determined by placing a small pinhole over the photocathode; the contrast C(0) increases to unity.
A MULTICHANNEL DETECTOR FOR PHOTON CORRELATION
117
1.0-
FIG. 8.(c) The spatial correlation in the x-direction. (d) The spatial correlation in the ydirection.
ACKNOWLEDGEMENTS The authors are grateful to Hamamatsu Photonics K.K. for donating the tube for evaluation. The authors also wish to thank Dr. A. A. D. Canas for advice on the electronic circuits required for the evaluation. This work was supported by UK Science and Engineering Research Council (SERC) under grant GRjC 78940.
118
D. N. QU AND J. C. DAINTY
REFERENCES Brown, R. G. W., Jones, R., Rarity, J. G. and Ridley, K. D. (1987). Appl. Opt., 26, 2383-2389 Dainty, J. C., Hennings, D . R. and O’Donnell, K . A. (1981). J . Opt. SOC.Am., 71,490-492 Jakeman, E. (1973). In “Photon Correlation and Light Beating Spectroscopy” (ed. H . Z. Cummins and E. R. Pike) pp. 75-149 Johnson, F. A., Jones, R., Mclean T. P. and Pike, E. R . (1966). Phys. Rev. Lett., 16, 589-592 Oba, K., Sugiyama, M., Suzuki, Y. and Yoshimura, Y. (1979). IEEE Trans. Nucl. Sci., 26,346355
Application of Image Intensifier-Vidicon Systems to Low-light-level Phenomena in Physics and Biology G. T. REYNOLDS Department of Physics, Princeton University. USA
A. EISEN National Institutes of Health, USA
A. J. WALTON Cavendish Laboratory, University of Cambridge, Cambridge, England
and
L. A. CRUM Physical Acoustics Laboratory, University of Mississippi, U S A
INTRODUCTION
A high-gain image intensifier (EM1 9912) has been used in experiments investigating a number of low-light-level phenomena of interest in current research problems in physics and biology. The intensifier has been variously coupled with a microscope or spectrometer, and the output recorded by means of either a SIT vidicon or conventional cameras. In its various forms, the system provides spatial, temporal, and spectral information on the processes studied. The purpose of this paper is to review briefly the significant results of recent research in the areas of sonoluminescence, crystalloluminescence, and detection of calcium transients in certain cellular processes. SONOLUMINESCENCE
When high-intensity acoustical waves traverse certain liquids, small bubbles form (cavitation). In some cases, there is a weak light emission from the bubbles, termed sonoluminescence (SL). Observations to date have been made primarily with photomultipliers, providing no detailed spatial information. Because the light is weak, spectra have been difficult to obtain using conventional techniques. We have observed the bubbles responsible for the light emission using the 119 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
120
G T. REYNOLDS ET A L .
PROBE LIQUID
MIRROR
I
CAMERA
LENS
FIG. 1. Experimental arrangement for observing the spatial and temporal distribution of sonoluminescence Sonicator: Bronson “Sonifer, Cell Durupter 200,” 20 kHz, probe $ inch diameter, liquid container (5 x 5 x 6) inches deep. (a) Side viewing; (b) bottom viewing.
experimental arrangement shown in Fig. 1 . Lens coupling permits viewing of luminescence from the side or from below. Figure 2 shows some typical results of these observations. It is clear that photomultiplier results do not provide important spatial information. The patterns of SL shown in Figs. 2(a) and 2(b) were stable in time, and occurred in standing waves in the tank at relatively modest levels of acoustic pressure amplitude (power delivered: 3-10 W cm-2). For Fig. 2(c) the power was approximately 30 W cm-2. A calibrated probe hydrophone was used to measure the acoustic pressures. The best estimate of the pressure for an intermediate power radiated to the liquid was 0.17 k0.03 MPa (1.4-2 bar). The light emission in Figs. 2(a) and 2(6) indicates that gas bubbles oscillating for a number of cycles generate sufficiently intense pressures and temperatures during the compression phase to produce light. These observations, taken with others in which SL was observed from a single bubble levitated in a transducer consisting of two PZT cylinders sandwiching a glass viewing section (Reynolds et al., 1986; Crum, 1980), corroborate the findings of Saksena and Nyborg (1970) that light output occurs under conditions in which the power input is below that required for transient cavitation. Combined with spectral results these observations provide evidence that the light emission mechanism is associated with free radical production and recombination at elevated temperatures. In view of the
APPLICATION OF IMAGE INTENSIFIER-VIDICON SYSTEMS
121
(c) FIG.2 (a) Photograph of a stable configuration of cavitation bubbles emitting sonoluminescence. Tank is viewed from the side, as in Fig. I(a). The input power level is relatively low (stable cavitation). (6) The cavitation configuration of (a) viewed from below, as in Fig. I(b) (stable cavitation). (c) Result of viewing as in ( b ) but at an increased power level (transient cavitation).
widespread use of ultrasound systems in medicine, and the fact that bubbles have been shown to form in certain biological materials, the production of SL should be given serious attention, since it is an indication of potentially harmful chemical activity. Because the light from sonoluminescence is very weak, it is difficult to obtain spectra by conventional means, requiring in some instances exposures of as much as 24 hours. Using an image-intensified spectrometer, spectra suitable for microdensitometry can be obtained in intervals from 10 seconds to 2 minutes with a spectral resolution approximately 20 A. The spectral response of the entire system was calibrated using a standard
122
G T. REYNOLDS ET AL.
lamp input. The gain of the image tube was 1.5 x lo5. The following detector characteristics were accounted for in the calibration process: (i) transmission characteristics of the optical elements in the spectroscope; (ii) vignetting effects in the geometry of the spectroscope; (iii) spectral response of the image intensifier cathode (approximately s 20); (iv) non-uniformity of cathode efficiency as a function of position; (v) film characteristics, and development conditions.
-
Repeated exposures to the same spectrum indicated measurement errors of the order of 5%. Thus, on the results to be discussed, the error bar on each spectral position is given as +7%, attributing error uncertainties to the determination of the calibration curve and to measurement error on the particular SL spectrum measured. A representative spectrum is shown in Fig. 3. Within the error bars, spectra from water saturated with air and argon are indistinguishable. The relative intensities of the spectra as determined from the densitometer traces at A=4300 A are shown in Table I. These results indicate that the
FIG. 3. Spectrum determined for argon-saturated distilled water (22 approximately 10-15 W cm-2.
I T ) . Power levels
APPLICATION OF IMAGE INTENSIFIER-VIDICON SYSTEMS
I23
TABLE I Relative intensities of SL Solution
Relative intensityat 4300 A
Air+distilled water (29+ 1°C) Air+distilled water (16+ 1°C) Argon+distilled water (22+ 1°C) Argon+distiiled w a t e r ( l 3 i 1°C)
1 .o 1.5 9.4 4.5
intensity of SL depends on the dissolved gas and agrees with previous investigations (Walton and Reynolds, 1984). In addition to interest in SL for its own sake, data can provide useful information relevant to possible effects of medical applications of ultrasonic devices (National Council on Radiation Protection and Measurements, 1983). Spectral data are particularly significant, since they can provide information on the physical and chemical conditions within the bubbles of the cavitating medium. The literature on spectra provides evidence for a complex phenomenon on which further careful study remains to be done. It is now generally believed that the light does not result solely from blackbody radiation, and that chemiluminescence at high temperatures is an important source. With conventional recording techniques, a long series of observations becomes prohibitive. Using image-intensifier techniques, recording times can be reduced to tens of seconds, or a few minutes, depending on the nature of the source. Because the possible contribution of black-body radiation to SL has frequently been discussed in both theoretical and experimental papers it is instructive to investigate the SL spectra reported here by comparing the ratio of the number of photons per wavelength interval at two selected wavelengths (4200 A and 5000 A). For these wavelengths this ratio can never exceed 2.0 for purely black-body radiation, for any temperature. For very high temperatures the ratio goes as 1k3 and for the wavelengths selected its value is 1.7. The spectra obtained in the present experiments yield ratios 3.16 (T= 20°C) and 2.95 ( T = 16°C) for air; and 3.15 ( T = 22°C) and 2.95 ( T = 13°C) for argon, which are all the same within the errors assigned, and in all cases significantly larger than could be explained by black-body radiation alone. These spectra were obtained at 20 kHz and at a power levels that resulted in transient, rather than stable cavitation (The present results are restricted to wavelengths between 4000 A and 6000 A because system corrections become very large beyond these limits. Observations in the near UV would be very important, but special optical components would be required.)
124
G T. REYNOLDS ET A L .
The conclusions from these observations are as follows. (i) The advantage of an image-intensified spectroscope for recording SL spectra has been demonstrated. Spectral details will provide important information as to the nature of SL and the complexity of the spectra will require extensive series of measurements. The significant reduction in time required for recording data, made possible by the intensifier, makes such a study feasible. (ii) The data obtained in the spectral range 4200-5500 A are not consistent with pure black-body radiation.
CRYSTALLOLUMINESCENCE Crystalloluminescence is the term applied to the light that is emitted in the early stages of the crystallization process of certain inorganic and organic substances. The phenomenon has been known for approximately two hundred years (Harvey, 1957). Although attempts to develop a theory explaining the effect date back approximately 150 years, no satisfactory explanation of the experimental observations appears to exist. In the past 25 years many observations have been made using photomultipliers to detect the light emitted. Reports include observations of inorganic salts as well as of certain organic binary mixtures. In view of the demonstrated advantages of image intensification over photomultipliers in studies of sonoluminescence and current interest in the structure of crystal growth, this technique has been applied to imaging the distribution of luminescence during rapid crystallization. A saturated solution of NaCl was placed in a cup a t the focal plane of the two-lens system, designed to provide approximately 1 : 1 imaging at the tube cathode. The output of the intensifier was focused on a SIT vidicon camera, providing simultaneous recording on tape and real-time viewing on a monitor. The image intensifier was operated at a gain of 2 x lo5. Crystallization was initiated by injecting concentrated HC1 into the cup. Proportions of HC1 and NaCl solution were always close to 1 : 1. The results obtained made it clear that impurities play an important role in the process. For example, a saturated solution of reagent-grade NaCl gave very little light, compared with a solution of table salt. Garten and Head (1963, 1966) report in detail the effect of various impurities. The image intensifier-vidicon system provides a capability to record the dynamic characteristics of the crystalloluminescence process. The most dramatic impression of the process is obtained from viewing the videotape recording in real time, when the light comes from apparently randomly situated centres. Although crystal growth may be proceeding in a dendritic growth pattern, this does not seem to be characteristic of the light emission, at
APPLICATION OF IMAGE INTENSIFIER-VIDICON SYSTEMS
125
F I G . 4. (a) and ( b) Two successive frames approximately 2 s after the onset of the crystalloluminescenceprocess.
least within the limitations of the time resolution of the intensifier-vidicon system. Figure 4 shows Polaroid pictures of pairs of successive single frames taken from the video record at an early stage of a representative run. The intensifier gain was 2 x los, so that individual cathode electrons could be recorded. In addition to the large spots of light, individual photons are evident, due to scattered light from the solution-crystals mix. The duration of the lightemitting process in this particular run was approximately 2 minutes. Other sequences varied from seconds to minutes. Results obtained to date show the following. (i) The intensifier system used is suitable for studying the dynamic process of luminescence during crystal growth from saturated solutions. (ii) The role of impurities is very significant. The present study was not
126
G T. REYNOLDS ET AL.
sufficiently controlled to enable quantitative results to be obtained on this aspect. (iii) The large light spots in the records show irregular boundaries, indicating that their size is not due to over-exposure or “blooming” in the detector, but possibly to asymmetries in the crystal patterns. (iv) Several reports in the literature describe observations of the spectral distribution of the emitted light. Using conventional methods, exposures of several hours are required. The image-intensified spectrometer, which has proved useful in studies of sonoluminescence, is also able to provide useful data on the spectral distribution of crystalloluminescence, in times short enough to make systematic studies feasible, leading to a better understanding of the phenomenon. OBSERVATIONS OF CA’
+
IN
BIOLOGICALSYSTEMS
It is useful to call attention to the fact that image intensification has made important contributions to a range of biological studies. Since the first microscope-intensifier-vidicon system was described over 20 years ago (Reynolds, 1964) numerous applications have been made, mainly in studies of
PHOTOMULTIPLIER
AMPLIFIER
\-
-
OCULAR,
-
CHART
FILTERS
V I D E O TAPE RECORDER
INTENSIFIER/VIDICON
L IGHT PATH
I
SPECTROSCOPE,~’INTENSIFIER,I‘ CAMERA
-EPI
-
FLUORESCENCE ILLUMINATOR
OSCILLSCOPE FILTERS
MONITOR
STAGE
FIG.5. Image-intensified fluorescence microscope system used to study Ca+ luminescence and pyridine nucleotide fluorescence in the fertilization of single eggs. The intensified spectroscope was used to determine the fluorescence emission spectrum; the image intensifiervidicon system determined the spatial and temporal changes in the C a + transients and the weak intracellular fluorescence. +
+
APPLICATION OF IMAGE INTENSIFIER-VIDICON SYSTEMS
127
bioluminescence, providing information on the spatial, temporal and spectral characteristics of light emitted from a wide range of organisms (Reynolds, 1972, 1978). More recently, important results have been obtained concerning the details of free calcium (Ca++) release, transport and uptake in the fertilization process in eggs of marine organisms, using the photoprotein aequorin. A typical experimental arrangement for such studies is shown in Fig. 5. By means of time-related observations of the outputs of imageintensified vidicon, photomultiplier, and membrane potential sensor, supplemented by intensified spectrometry and direct photography of the microscopic image, it has been possible to determine, for single eggs of several marine organisms, the sequence of events in egg fertilization. In a typical sequence, a wave of surface contraction is coincident with membrane depolarization, followed by the onset of a rapid, propagated increase in cytoplasmic free calcium which is in turn followed by an increase in pyridine nucleotide fluorescence and membrane elevation. Recent results have been reported in detail elsewhere (Eisen et al., 1984; Eisen and Reynolds, 1984, 1989, demonstrating the utility of image intensifier-vidicon systems in continuing studies of biological importance involving very low-light-level emissions. ACKNOWLEDGEMENT
This work was supported in part by DOE grant number DE-FG02-87ER 60522-A000
REFERENCES Crum, L. A. (1980). J. Acoust. SOC.Am. 68,203-21 I Eisen, A,, Kiehart, D. P., Wieland, S. J. and Reynolds, G . T. (1984). J . Cell. B i d . 99, 1647 Eisen, A. and Reynolds, G . T. (1984). J . Cell. Biol. 99, 1878 Eisen, A. and Reynolds, G. T. (1985). J . Cell. Biol. 100, 1522 Garten, V. A. and Head, R. B. (1963). Phil. Mag. 8, 1793 Garten, V. A. and Head, R. B. (1966). Phil. Mag. 14, 1243 Harvey, E. N . (1957). “A History of Luminescence from the Earliest Times until 1900.” The American Philosophical Society, Philadelphia National Council on Radiation and Measurements. ( I 983). ‘‘Biological Effects of Ultrasound: Mechanisms and Clinical Implications”. NCRP Report #74 Reynolds, G . T. (1964). IEEE Trans. Nucl. Sci. NS-11,147 Reynolds, G, T. (1972). Q.Reu. Biophys. 5,295 Reynolds, G . T. (1978). Photochem. Photobiol. 21,405 Reynolds, G . T., Crum. L. A. & Walton, A. J. (1986). IEEE Trans. Nucl. Sci.NS-33, 240 Saksena, T. K. and Nyborg, W. L. (1970). J. Chem. Phys. 53, 1722 Walton, A. J. and Reynolds, G. T. (1984). A h . Phys. 33, 595-660
This Page Intentionally Left Blank
Cooled CCD Systems for Biomedical and Other Applications C. D. MACKAY lnstiiute of Astronomy, University of Cambridge. Cambridge, England
INTRODUCTION There are very many areas of scientific effort that depend on imaging at low light levels. In most fields it is common to use image intensifiers in front of television camera tubes in order to work at the lowest light levels. Optical astronomers have these same requirements yet they seldom use intensified TV cameras (apart from telescope guiding applications). The reasons for this are twofold. Firstly, the light levels attained by these cameras or l o p 6lux) are not, in fact, very low at all since they correspond to perhaps 105photons mm-2 s-I. Secondly, in many applications the imaging quality is rather poor. Astronomers have stars and galaxies to observe that cover a range in brightness of 1O'O and so it is important that they be able to relate measurements of brighter and fainter objects with some accuracy. Dynamic range and linearity of response are photometrically of the greatest importance to astronomers and they are becoming more so for users of imaging systems in other fields. In astronomy the detector of choice is now the cooled slow-scan chargecoupled device (CCD). The details of the operation of these systems have been described (Mackay, 1986). Briefly they use the same sort of CCD that is used in many commercial TV frame-rate CCD cameras. At typical room temperatures they have a peak signal capacity of around 500 000 photons per picture element (pixel). The lowest signals detectable are limited by the device dark current which is typically a few thousand photons equivalent per TV frame time. This gives a dynamic range in one exposure of a few hundred. If the CCD is cooled, the dark current is reduced (typically by a factor of 10 for each 20°C drop). Then the system noise is limited by the noise associated with the CCD on-chip amplifier at a level of 200-300 photons equivalent. Astronomers then read out the CCD slowly enough (5-20 s for a full frame of 250 000 pixels) to allow more subtle analogue signal processing that brings the system noise floor down to only 5 or 6 photons equivalent in the lowest-noise devices. 129 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copynght 0 1988 Academic Press Limited All rtghis of reproduction in any form reserved ISBN 0-12-014674-6
130
C. D. MACKAY
The cooled slow-scan CCD is then able to work with long image integrations (many hours if necessary) to achieve imaging at very low levels (down to one photon per pixel per minute or lo-" lux) and with very wide dynamic range, up to 100 000: 1 in the best cases. This dynamic range is greatly superior to that of any other imaging system. As far as the low-lightlevel performance is concerned it is possible for the best photon-counting systems to come close in terms of low-light-level sensitivity. However, in general the much lower quantum efficiency of photocathodes more than offsets the ability to detect each photon and count it, The principal difficulty with photon-counting systems is their inability to handle the relatively high photon arrival rates that are essential if a photometric accuracy of better than a few per cent is to be achieved. PERFORMANCE OF
COOLEDCCDS
The performance that can now be achieved by standard commercially available cooled slow-scan CCD systems is remarkable. They are within a small factor of being theoretically perfect detectors. Typical performance figures for one such system (Astromed Ltd., UK) are given below. These systems use the P8603 CCD manufactured by EEV Ltd. (UK), probably the best available scientific CCD supplied tested at low temperatures and in slow scan mode. Performance Figures Format and resolution. 385 x 576 pixels, 22 x 22 pm, 8.5 x 12.5 mm2 sensitive area. Resolution is set only by the pixel spacing. Detective quantum eficiency. Peak 50 per cent at 700 nm usable sensitivity from 400 to 1050 nm. Overall system noise and dark current. Read-out noise 5-10 photons equivalent RMS. Dark current at 120 K less than 1 photon equivalent per pixel per hour. Cosmic ray event rate about two per minute over the whole device. Cosmic ray events cover a small number of pixels (1-4) and have total signals of about 2000 photons equivalent. Charge transfer eficiency. Operation at the lowest light levels in cooled slow-scan CCD systems is often compromised by the difficulties of reading out all the charge without leaving any behind and so giving rise to image smearing. The charge transfer efficiency (CTE) is often improved if there is a uniform background signal to work against. In some systems this signal is added
CCD SYSTEMS FOR BIOMEDICAL APPLJCATIONS
131
artificially to the desired image to keep the CTE at an acceptable level. For the P8603 CCD there is excellent CTE at the lowest levels (peak signals of around 10 electrons per pixel). Dynamic range. This is the ratio of the peak signal per pixel ( - 500 000 electrons) to the read-out noise (5-10 photons equivalent or electrons) so is 50 000 : 1 to 100 000 : 1 per pixel. Real images cover several pixels, so dynamic range in excess of 100 000: 1 is routinely achieved. Exposure times. Exposure times are limited by shuttering mechanisms only. Values from milliseconds to hours are routinely practicable. Illuminance range. Illuminance range that can be detected by the CCD is in therange lo-” lux to 10 lux (equivalent to 1-1013 photons mrn-*s-’ incident on the CCD). Cosmetic quality. Devices are available at modest cost and in reasonable volume that are blemish-free and have top noise and CTE performance.
APPLICATIONS The possible applications of cooled CCD imaging systems are very wide indeed. The main disadvantages of using these systems are (i) in some cases, their relative bulk, though developments with fibre-optic re-imaging systems help considerably, (ii) their unsuitability to applications requiring fast readout for high time-resolved imaging applications. Three examples will be given here. Immunoassays
The detection of biologically significant amounts of molecules such as antibodies is vitally important to the rapid diagnosis of diseases, both physiological diseases such as cancer and infectious diseases such as AIDS. Most depend on the use of monoclonal antibodies and enzyme amplification methods coupled with dyes that produce colour changes in the samples of interest. However, the concentrations needed to produce an observable colour change are considerable. Work done with the Imperial Cancer Research Fund (London, UK) and Corolab Research (Leaback, 1987) has developed a double-antibody assay that leads to the emission of light from an enzyme in the vicinity of malignant cells in breast tumours. When used with a cooled slow-scan CCD camera the technique is capable of a much more rapid diagnosis of malignancy at a much earlier stage in the development of the tumour.
132
C. D. MACKAY
Two-Dimensional Gel Electrophoresis of Proteins
The detection of the faintest spots on a conventionally stained 2-D gel depends on detecting a faintly absorbing region against the bright, transparent background. The wide dynamic range of a cooled slow-scan CCD system enables the rapid routine detection of the faintest spots. Much greater sensitivity may be obtained by tagging the proteins with a fluorescent marker either before electrophoresis or after the isoelectric focusing stage. Then the faintest spots are visualized as fluorescing against a dark background, giving an enormous improvement in sensitivity. Exceptionally weak spots containing only a few thousand protein molecules may be detected and quantified (Jackson, 1988). The fluorescence approach to visualizing 2-D gels has many other advantages. The gel does not need to be stained and so it may be visualized without being unpacked. If it is not to be unpacked it can be thinner and smaller, dramatically reducing gel running time. In addition the fluorescence is a linear process, so that accurate quantitation of the material in each spot, one relative to another, becomes a practicality for the first time. Also, because fluorescence is so much more sensitive a technique, it is possible to reduce considerably the amount of protein loaded onto the gel. This has the desirable effect of greatly reducing the apparent geometric errors seen when different gels are compared. In some cases radio-labelled gels need to be visualized rapidly. This can be achieved by incorporating a scintillator in the gel (for any radio-label, but especially for tritium) or by contacting the gel against a scintillating screen. The light generated is optically coupled to the cooled slow-scan CCD for the most rapid visualization of radio-labelled gels. X-RAYIMAGINGAND MICROSCOPY Any X-ray image may be converted into a visible image by using an intermediate phosphor screen. The visible image on the phosphor screen is then optically coupled (magnified or demagnified with lenses) onto the detector. The coupling efficiency for large images such as a thoracic X-ray image onto the small CCD is inevitably poor but for smaller images the coupling efficiency is excellent. With a demagnification factor of a few or smaller, a cooled, slow-scan CCD will give an excellent signal-to-noise ratio for the detection. Under the right conditions the sensitivity is such that the Xray dosage may be reduced, thus limiting potential damage. Alternately, images with much better signal-to-noise ratios may be obtained at the same Xray dosage level, The wide dynamic range of the system makes it uniquely able to work with very low-contrast images. This opens up the possibility of X-ray
CCD SYSTEMS FOR BIOMEDICAL APPLICATIONS
133
imaging of very soft tissues without greatly increasing the X-ray dosage. It also permits low-contrast features in other specimens to be detected for the first time. In many situations, separated pairs of images permit the generation of stereo pairs of images, which permits real three-dimensional analysis of the specimen under test. The excellent resolution of the system enables X-ray diffraction patterns to be measured reliably even at rather low X-ray flux levels. Weak diffraction spots from complex crystals can be detected much more readily. The recent development of microfocal X-ray sources has opened the whole subject of X-ray microscopy. Microfocal X-ray sources permit X-ray images to be obtained with a resolution of down to around 1 pm. The excellent sensitivity of the cooled slow-scan imaging systems allows excellent images to be obtained even with the very low X-ray fluxes that microfocal X-ray sources inevitably give.
CONCLUSIONS Astronomers are not able to adjust or change the objects they want to look at in the way that is possible in most laboratory sciences. As a result they have to put all their effort into making the very best of the signals from space that are available. The consequence of this is that they have developed the art of imaging detector manufacture more rapidly than has happened in other fields. The technology developed around the use of cooled slow-scan CCD systems is now beginning to have a considerable impact in a wide range of other applications. These now include transmission and fluorescence microscopy, one- and two-dimensional electrophoresis of proteins, DNA sequencing, Xray and electron-beam imaging, Raman spectroscopy and X-ray crystallography. Coupling this technology with modern, sophisticated computers with image-processing facilities, the possibilities are limitless.
REFERENCES Jackson, P. (1988). Elecrrophoresis (in preparation) Leaback, D. (1987). J. Analyt. Cyrometry (in press) Mackay, C. D. (1986). Annu. Rev. Astron. Astrophys. 24, 255-283
This Page Intentionally Left Blank
Utilisation Astronomique de la CamCra Electronique Grand Champ-IIt G. WLERICK, G. LELIEVRE, L. RENARD, B. SERVAN, D. HORVILLE, J. FROMAGE, J. M . LE FLOHICJ D. BAUDUIN Observatoire de Paris, Paris, France
A. BIJAOUI Obseruatoire de Nice, Nice, France et
G . COURTES Laboraroire d‘AsIronornie Spatiale, Marseille, France
INTRODUCTION Lors du preckdent Symposium, nous avons decrit l’installation d’un ensemble de camkras au telescope de 3,6 m Canada-France-Hawaii (Servan et al., 1985)et les premikres observations astronomiques effectukes (Wlkrick et al., 1984a). Nous indiquons les ameliorations rkaliskes depuis 1983, et signalons les observations difficiles qui ont ttk faites. Nous dkcrivons kgalement, I’kvolution de l’ensemble utilisk 6 I’Observatoire de Haute-Provence, et, en particulier, son adaptation a un rtducteur focal, permettant de travailler avec un champ angulaire plus grand sur le Ciel. AMELIORATION DE LA CAMERA ET
DES
PROCEDURES D’OBSERVATIONS
Nous rappelons brikvement les caractkristiques du tube: diamktre utile de la photocathode 8 1 mm; grandissement 1,O; 2 x lo6 klkments-images (pixels); cathode de type S . 11 (possibilitk d’utiliser des S-20); kmulsion electronographique Kodak Electron Microscope; chaque tube permet de faire 9 poses (bient6t-24 poses). Les amkliorations ont porte principalement sur la qualitk des photocathodes.
t Les observations
ont etk effectuees I I’Observatoire de Haute-Provence du CNRS et avec l’instrurnent de la Socitte du Telescope Canada-France-Hawaii. $ Cet article est dedii li J . M. Le Flohic, decbdk au Mauna Kea, a Hawaii, le 20 Juin 1987.
135 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 14
Copyright Q 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
136
G. WLERICK ET A L .
Rendement Quantique des Cathodes
Les couches photoemissives sont produites par le Laboratoire des Photocathodes de I’Observatoire de Paris. Au cows des trois dernieres anntes, ce laboratoire a rkussi a augmenter le rendement quantique des couches S - 11 produites pour la grande camera; cette augmentation s’est d’abord accompagnee d’un accroissement de l’kmission parasite; maintenant le laboratoire produit des couches a la fois sensibles et sans bruit; les rendements typiques sont les suivants: domaine 350-400 nm, R = 15-20%; domaine 400-500 nm, R=20-25%. Difauts de la Surface Photoimissive
I1 existe au moins trois types de dkfauts. I1 y a d’abord les variations a grande kchelle de la sensibilitk de la photocathode; elles sont likes a la gkometrie de l’ampoule dans laquelle la cathode est prkparke. On observe souvent un maximum de sensibilitk dans la rkgion centrale. Ces variations a grande kchelle ne perturbent pas I’analyse des clichks, si I’on peut disposer d’une carte de la sensibilitk de la cathode obtenue avec une source d’eclairement uniforme.
FIG.I . Champ Sandage-VCron,couleur U (ultraviolet), pose 260 min, 20 Mars 1985, Tblescope C.F.H. (cliche G. Wlerick, G. Lelievre, Ch. Vanderriest, D. Horville).
UTILISATION ASTRONOMIQUE DE LA CAMERA ELECTRONIQUE
137
Un autre type d’imperfection est constitue par des rayures ou des piqures; ces defauts sont de petite dimension (100 pm a 1 mm); ils sont Crees, soit lors de la fabrication de la cellule, soit lors de sa mise en place dans la camera. 11s sont peu ggnants, si leur nombre est limitt. Les defauts les plus ghants, quand ils existent, sont des defauts fins qui peuvent s’ttendre sur une grande partie de la cathode; ils ressemblent 2 un peignage et on ne les aperCoit vraiment que lorsque la densite des cliches est Clevee (D> 1); ils produisent une faible modulation de la densite, mais cette modulation complique nettement la mesure des astres tres faibles. Heureusement ces defauts se produisent rarement. Une partie des cathodes utilisees ne prtsentent pratiquement aucun defaut. D’autre part, quand il y a des imperfections, I’importance de celles-ci diminue quand la longueur d’onde de travail diminue. A Hawaii, la plupart des cliches ont t t t pris dans I’une ou I’autre des bandes larges du systeme U, B, V de Johnson et Morgan (ultraviolet, bleu, jaune); les cliches en ultraviolet presentent des defauts relativement faibles (Fig. I ) .
Essui de Correction des Dtfuuts sur les Clichts Astronomiques On peut graver des traits sur le support des photocathodes. Lors d’une mission d’observation, a Hawaii, en juin 1987, nous avons essaye deux types de gravure: deux traits diametraux, perpendiculaires entre eux, ou quatre croix situtes a la ptripherie du cliche. A la fin de la nuit d’observation, nous prenons un cliche court sur le Ciel, a l’Aurore, quand le soleil est a environ-10” au-dessous de I’horizon. Ce cliche tient lieu de “lumiere uniforme” et I’approximation est bonne car les ttoiles interviennent faiblement pour une pose courte. On obtient ainsi une bonne estimation des variations a grande Cchelle de la sensibilite de la cathode. Les reperes gkomttriques permettent de comparer le cliche astronomique et le cliche “ciel a I’aurore”, en chaque point de la cathode. En particulier, on peut, avec un ordinateur, diviser le fichier des donnees du premier cliche par celui des donntes relatives au Ciel. Ceci supprime completement ou attenue fortement les dkfauts a grand contraste. La Fig. 2 montre l’effet sur les gravures elles-mgmes, c’est-a-dire sur un “defaut” tres contrast&. I1 faut analyser les limites de la methode; celle-ci repose sur l’hypothese que les emulsions tlectronographiques sont bien homogenes, ce qui semble verifit avec l’tmulsion Kodak Electron Image que nous employons rtgulierement. D’autre part, lorsqu’on effectue la division d’un fichier par un autre, on additionne les bruits: il convient donc que le cliche “Ciel” soit relativement peu bruite; ceci suppose que sa densite soit superieure a celle du cliche a corriger, par exemple D = 1,5 pour le cliche “Ciel” et D=0,6 pour le cliche astronomique.
138
G. WLERICK ET A L .
FIG.2. (a)Partie centrale d’un cliche de calibration de la Selected Area 57, couleur V (jaune), 4 poses de 1 min, 2 min, 5 min et 10 min. (b).Resultat obtenu en corrigeant le cliche pre&dent a l’aide d’un cliche pose sur le Ciel, a l’aurore, quand le Soleil est a - 10”au-dessous de l’horizon.
I1 reste a examiner si on aura assez de resolution pour corriger les dtfauts de type peignage. Par ailleurs, les traits diamttraux graves sur la cathode, permettent de mesurer la distorsion en S de la camkra.
UTILISATION ASTRONOMIQUE DE LA CAMERA ELECTRONIQUE
139
A l’heure actuelle, nous utilisons le m&metelescope, le mtme foyer mais avec un reducteur focal qui permet de travailler a l’ouverturef/8, correspondant a une longueur focale equivalente de 15 m. Ce montage, qu’on peut utiliser dans les bandes B et V, permet d’augmenter le champ angulaire de la camera: 18’; Cgalement, la focale est mieux adaptke a la valeur moyenne de I’ktalement des images, dii a la turbulence atmospherique. Le probleme ktait difficilecar il fallait rtaliser un reducteur focal compatible avec la grande distance frontale de la camera, 160mm. La solution a etk trouvee par G. Courtts et D. Kohler. Les deux pieces principales du reducteur sont deux objectifs astronomiques travaillant ttte-beche; on a donc un faisceau parallele entre les deux objectifs. Ce montage se rtvele tres souple, car on peut introduire differents elements dans le faisceau parallbie et dans le plan focal primaire: avec un rtseau dans le faisceau parallde-spectroscopie sans fente; en ajoutant une fente dans le plan focal-spectroscopie a fente; en plaGant plusieurs fentes dans le plan focal-spectroscopie multifentes. On peut aussi placer des tlkments optiques dans le faisceau convergentfl8, situe devant la camera.
PROBLEMES RENCONTRES DANS L’ETUDE DES ASTRESFAIBLES Pour observer de plus en plus loin dans I’Univers, il faut mesurer des astres de plus en plus faibles, en particulier des galaxies. L’enregistrement de ces astres tres peu lumineux ne poserait pas de problemes s’ils Btaient seuls Cmettre dam le Ciel: avec un recepteur lintaire (pas de seuil), il suffirait d’effectuer une pose suffisamment longue. Malheureusement, le signal dii a un astre faible est noyk dans un signal beaucoup plus fort, l’emission du Ciel nocturne, et le bruit propre de cette emission (essentiellement le bruit de photons) impose une limite a la detection des galaxies lointaines. D’autre part, I’Univers est heterogine a petite kchelle; pour obtenir des resultats significatifs, il faut donc disposer d’un champ angulaire suffisant sur le Ciel. Grice a sa capacitt de stockage et a son grand format, la camera electronique grand champ permet de rksoudre, au moins partiellement, ces problkmes. UTILISATION DE
LA
CAMERA A L’OBSERVATOIRE DE HAUTE-PROVENCE
Jusqu’en 1985, la camera a ett utilisee avec le Telescope de 193 cm, au foyer Cassegrainfl15, donc avec une longueur focale equivalente de 29 m (Wlerick er ai., 1979). Dans ces conditions, le champ sur le Ciel est 10’. Ce dispositif est toujours utilise dans I’ultraviolet (bande U). En particulier, on peut utiliser une lame de Savart qui dkdouble I’image de chaque ttoile en deux images
140
G. WLERICK ET AL.
FIG.3. Champ de la Radiosource 21 13+29. Telescope de 193 cm de I’0.H.P.. avec reducteur focal /78 et lame de Savart, pose 60 min, 9 Septembre 1986. La lame de Savart fournit, pour chaque &toile, deux images presentant des directions de polarisation orthogonales (cliche K. Meisenheimer, G. Wlerick, G. Courtes, B. Servan).
correspondant a des directions de polarisation orthogonales (Roser, 1981). I1 est possible, ainsi, de detecter ceux des astres d’un champ qui emettent une lumiere polarishe. Le rkducteur focal et la camera ont ete utilises avec succbs pendant six nuits en septembre 1986 pour plusieurs programmes de G . Court& (spectres sans fente, images monochromatiques) et de K. Meisenheimer et H. J. Roser (recherche d’objets polarists). La Fig. 3 correspond au champ qui entoure la radiosource 21 13+ 29. Chaque ttoile donne deux images et on observe que la qualitt optique du systeme reducteur-lame de Savart-camera reste la meme dans tout le champ de 18’.
OBSERVATIONS AVEC
LE
TELESCOPE CANADA-FRANCE-HAWAII
De decembre 1982 a juin 1987, la camera grand champ a ett utilisee, au cours de huit missions d’observations, pendant un total de 5 1 nuits; 300 clichts ont Ctt. obtenus.
UTILISATION ASTRONOMIQUE DE LA CAMERA ELECTRONIQUE
141
Les proprietks de la camera et la qualiti astronomique du site (Mauna Kea, 4200 m) ont permis d’observer, avec succes, differents types d’astres. Des clichts, a haute resolution spatiale (v0,6’’)ont Cte obtenus sur les jets optiques de la radiogalaxie M87 et du quasar 3C 273 (Lelibvre et al., 1983, 1984), sur l’environnement du noyau de la radiogalaxie 3C 120 (WlCrick et al., 1986), sur le mirage gravitationnel PG 11 15 +080, dont on a trouvt la variabilitk (Vanderriest et al., 1986). I1 semble qu’aujourd’hui les domaines privilkgiks d’emploi de la camtra grand champ sont ceux od les autres recepteurs manquent actuellement, soit de champ (cas des CCD), soit de dynamique (cas du comptage de photons). Deux types d’observations necessitant un grand champ ont itC entrepris: d’une part I’btude detaillee d’une galaxie proche rksolue en Ctoiles, d’autre part des sondages de I’Univers plus profonds que ceux realisis a ce jour avec la plaque photographique.
Etude des Etoiles de Messier 33 Madame R. Dubout de I’Observatoire de Lyon Ctudie les populations stellaires de la galaxie Messier 33. Cette galaxie est situCe a environ deux millions d’annkes-lumierede notre Voie LactCe: c’est une spirale vue de face. L’ttude porte sur le bras spiral Sud. Un grand champ est ntcessaire pour couvrir une region suffisante dans la direction du bras et pour examiner I’kvolution des populations dans la direction perpendiculaire au bras. Sur un cliche en couleur B ou V, avec pose de deux heures, on enregistre plusieurs dizaines de milliers d’ktoiles. On concoit que I’extraction des donntes soit difficile: d’abord, il y a le trks grand nombre d’astres et le fait que leurs images empietent les unes sur les autres; en outre, ces Ctoiles sont noykes dans des nebulositks qui forment I’equivalent d’un fond de ciel variable en chaque point. Pour analyser les cliches, des logiciels tres Claborts ont ttC mis au point. Un apercu sur le premier logiciel utilist a Cte donne dans WlCrick et al. (19846). Un prochain article par Debray et al., dtcrira completement les techniques de reduction utilisees a I’heure actuelle.
Sondages Profonds prks du PBle Galactique Ce travail a commence en 1985,dans deux champs situes Si moins de 10”du P6le galactique Nord: le champ Sandage-Veron et la Selected Area 57. Le but est de recenser le contenu en astres dans cette direction du Ciel: combien d’ttoiles, quels sont leurs types spectraux? Combien de galaxies, quelles sont leurs formes et leurs couleurs? Combien de quasars?
142
G . WLERICK ET A L
FIG.4. Champ Sandage-Vtron,couleur B (bleu), pose 90 min, Mars 1985, Telescope C.F.H. I1 s’agit d’un agrandissement de la partie Est du champ (cf. Fig. 1). La grande tchelle du clicht original (1” = 145 pm)et la qualid des images du site de Mauna Kea permettent de distinguer les ttoiles des galaxies.
Champ Sandage- Vgron
Nous avons present&(Fig. 1) une pose longue, 260 min, en couleur U. Nous montrons, Fig. 4, un agrandissement d’une pose de 90min, en couleur B, d’une partie du m2me champ. On voit qu’avec les conditions d’observations utilisees (longueur focale 29 m, emulsion Kodak Electron Image), on distingue bien les Ctoiles des galaxies. Selected Area 57
En mars 1985, nous avons obtenu une pose longue dans chacune des couleurs U, B et V et nous avons pris tgalement des cliches de calibration. La Fig. 5 represente une pose de 240 minutes en couleur U. Comme dans 1’Ctude de la galaxie Messier 33, on est confronte a un important problkme de traitement de donnees; le travail a ktk effectuk a I’Observatoire de Nice avec Albert Bijaoui; l’ordinateur disponible, au debut du travail, etait relativement petit; par suite, nous nous sommes limit& a des fichiers 2000x2000; ceci nous a conduits a enregistrer les cliches, au microdensitomktre PDS, avec un pas de 40 pm et une fente de 40 pm x 40 pm, soit, sur le Ciel, 0,28” x 0,28”.
UTILISATION ASTRONOMIQUEDE LA CAMERA ELECTRONIQUE
143
FIG.5. Selected Area 57, couleur U, pose 240 min, 21 Mars 1985, TClescope C.F.H. (clichk G. WICrick, G. Lelievre, Ch. Vanderriest, D. Horville).
Un grand travail d’klaboration de logiciels a ete effectuk (Bijaoui et Wlirick, 1986). Deux points ont CtC particulierement abordes: la soustraction correcte de la lumikre du Ciel nocturne (fond du ciel) et l’appreciation correcte de la contribution des astres voisins; nous preparons un catalogue des magnitudes U, B, V de galaxies faibles jusqu’a la magnitude, dans le bleu, B = 25,4. Nous disposons dCja d’un catalogue partiel concernant les objets “isoles”, c’est-a-dire ceux pour lesquels il n’y a pas de perturbation par un astre voisin “brillant”. Koo (1986) a ttudie les galaxies jusqu’a B.u 2 3 3 et trouve qu’elles sont d’autant plus bleues qu’elles sont plus faibles. Nos rksultats confirment ceux de Koo et Ctendent cette propriete jusqu’a B = 25 (Figs. 6 et 7). Durant la dernitre periode d’observation, en juin 1987, nous avons effectut une pose de 140 min, en couleur B, sur la Selected Area 57. D’apres un sondage effectue dans la region situke au Sud-Est du centre du champ (Fig. 8), nous estimons qu’environ 4400 astres doivent Gtre mesurables sur l’ensemble du cliche. Si les galaxies ktaient espactes regulierement les unes des autres, cela correspondrait a une distance moyenne entre deux galaxies de 7” seulement; comme les galaxies sont des astre ttendus, on concoit que leurs images empiktent, souvent, les unes sur les autres. D’autre part le chiffre de
144
G . WLERICK ET AL.
. . . .. ... .. . - . ..* :. .*
-2
-1
0
1
2
Indice [B-V] FIG.6. Diagramme Indice de couleur (B -V), Magnitude V, pour 172 astres relativement isol6s de la Selected Area 57. On remarque que I’indice (B-V) dkroit quand la magnitude V croit, autrement dit, plus les astres sont faibles, plus ils sont bleus.
4400 astres, pour l’ensemble du cliche, correspond une densitt de galaxies de 230.000/[1°]2. Cette valeur est plus grande que celles publikes a ce jour. L’analyse de ce cliche est en cours; on peut estimer provisoirement que I’on peut mesurer des astres un peu au-deli de B =26. CONCLUSION A notre connaissance, la camera Clectronique grand champ est le seul rkcepteur d’images, en service actuellement, qui possede, a la fois, les qualites suivantes: une grande sensibilitk, en particulier dans le bleu et l’ultraviolet; une
UTILISATION ASTRONOMIQUE DE LA CAMERA ELECTRONIQUE
.. . . *
....’ :.’ *
-2
-1
0
1
145
-1 2
lndice [B-V]
FIG.7. Diagramme Indice de couleur (U-B), Magnitude B, pour les mdmes astres que ceux present& Fig. 6. La conclusion est la mime: I’indice (U -B) decroit quand la magnitude B croit.
dynamique raisonnable; un grand format exprime en mm-81 et, un grand nombre d’elkments-images (pixels)-22 x lo6. Naturellement, les dimensions de la cathode pourraient &treaugmentkes et une vanne pourrait ttre incorporee au recepteur. REMERCIEMENTS Nous remercions le Laboratoire des Photocathodes, 1’Atelier de Verrerie et I’Atelier de Mkanique de l’observatoire de Paris pour leur coopdration trks efficace. Nous sommes redevables Cgalement au GRECO “Rtcepteurs” de I’Institut National des Sciences de l’Univers pour son soutien humain et financier.
146
G. WLERICK ET A L .
FIG. 8. Partie d’un cliche de la Selected Area 57, couleur B, pose 140 min, 24 Juin 1987, Tilescope C.F.H. On remarque le grand nombre d’astres enregistres (cliche G. Wkrick, J. Arnaud, D. Horville).
REFERENCES Bijaoui, A. et Wlkrick, G. (1986). Proc S.P.I.E. 702, 241-244 Debray, B., Dubout, R., Llebaria, A. et Petit, M. “The Cappella Bidimensional Software for Accurate Photometric Reductions of Crowded Stellar Fields in Nearby Galaxies” (a paraitre) Koo, D. C. (1986). Astrophys. J. 311, 651-679 Lelievre, G., Nieto, J. L., Wlerick, G., Servan, B., Renard, L. et Horville, D. (1983). C.R. Acad. Sci. Paris, 296, Skie 11, 1779-1786 Lelitvre, G., Nieto, J. L., Horville, D., Renard, L. et Servan, 8. (1984). Astron. & Asfrophy. 138, 49-56 Roser, H. J. (1981). Astron. & Astrophys. 103, 374-381 Servan, B., Wlbrick, G., Renard, L., Lelievre, G., Cayatte, V. Horville, D. et Fromage, J. (1985). In “Adv. E.E.P.” Vol64 A, pp. 11-20 Vanderriest, Ch., Werick, G., Lelievre, G., Schneider, J., Sol, H., Horville, D., Renard, L. et Servan, B. (1986). Asrron. & Astrophys. 158, L5-L8 Wlerick, G., Lelievre, G., Servan, B., Renard, L. et Leftvre, B. (1979). In “Adv. E.E.P.”, Vol52, pp. 295-303 Wlerick, G., Lelitvre, G., Servan, B., Cayatte, V., Michet, D., Renard, L. et Horville, D. (1984a). Proc. S.P.I.E. 445, 143-150 Wlbrick, G., Renard, L., Horville, D., Lelievre, G., Bijaoui, A,, Llebaria, A,, Dubout, R. et Petit, M. (1984b). Proc. Colloque UAI no 79, “Very Large Telescopes, their Instrumentation and Programs”, Garching, 9-12 Avril 84 (td. M. H. Ulrich et K. Kjlr), pp. 639-657 Wlerick, G., Soubeyran, A., Servan, B.,Renard, L., Horville, D., Bijaoui, A,, Lelitvre, G. et Bouchet, P.(1985). Proc. Symposium UAI no 119, “Quasars” (id. par G. Swarup et V, K. Kapahi), pp. 129-130. Reidel, Dordrecht
Image Recording in Electron Microscopy D. McMULLAN Cavendish Laboratory, Universiry oJ Cambridge, Cambridge. England
INTRODUCTION The invention of the electron microscope by M. Knoll and E. Ruska (1932) took place at about the same time as the first attempts to develop image intensifiers. The problem of recording the images produced was the same in both types of instrument. Initially, photography of the image on a phosphor was adopted but it was soon realized that a very much more efficient detection could be achieved by allowing the electrons to fall directly on to a photographic emulsion. This method had in fact been employed for many years for recording traces in high-voltage cathode-ray oscillographs (Dufour, 1914) and the characteristics of emulsions had been investigated by a number of workers, including Becker and Kipphan (193 1) who demonstrated the linear relationship between developed density and exposure. The pioneer of image tubes using this method was A. Lallemand (1936) with the electronographic camera and his reason for choosing it was the same as that of the electron microscopists: to record as high a proportion of the electrons as possible. In Lallemand’s case it was because of the faintness of astronomical images, while in electron microscopy there was the need to minimize the electron dose to the specimen and so avoid radiation damage. As is well known, for high-energy electrons the detective quantum efficiency (DQE) of photographic emulsions can approach unity and this recording method is still used in most transmission electron microscopes (TEMs), particularly because of its high spatial resolution and convenience. The rather limited contrast range (- 100) is generally acceptable. A remaining problem is the real-time display of the image for selection of the area of interest and for focusing. Phosphors and low-light-level TV cameras (SIT or SEC) are used but are often inadequate from the points of view of both DQE and spatial resolution. As is described later in this paper, a single-crystal YAG scintillator with a charge-coupled device (CCD) appears to be an attractive alternative. Although electron microscopes are often used just to produce an image of the specimen, a most important technique is “analytical electron microscopy”, in which a crystallographic or chemical analysis of a specimen can be 147 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 14
Copyright Q 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
148
D. MCMULLAN
made point by point. This is done by illuminating a very small area of the specimen and adjusting the electron optics to produce a diffraction pattern at the photographic emulsion or, for chemical analysis, by using an electron spectrometer to measure the energy losses of the electrons that have passed through the specimen. Characteristic energy losses due to the interaction of the electrons with the core electrons of the atoms can be identified and accurate estimates of the total mass of an element in the electron path can be made. ANALYTICAL ELECTRON MICROSCOPY A type of electron microscope that is specially adapted for analytical work is the scanning transmission electron microscope (STEM). The electron optics are designed to produce a very small-diameter focused spot (-0.5 nm in the VG Microscopes HB501 STEM, see Fig. 1) that is scanned in a raster over the specimen, and the transmitted electrons are detected either on-axis to produce a bright-field image or by an annular detector for dark field. If this scan is stopped, the diffracted electrons from the specimen can be recorded by a phosphor screen and film or TV camera as a microdiffraction pattern. Alternatively, the beam from the specimen can be analysed in the electron spectrometer. In both cases the electron beam will have interacted with an extremely small volume of the specimen, typically .c 10 nm3,corresponding to a few thousand atoms.
MICRODIFFRACTION The phosphor screen used to record microdiffraction patterns in the standard VG HB501 STEM has a rather poor resolution and the DQE is low unless a low-light-level TV camera is used. Both deficiencies would be overcome by direct recording on photographic emulsion, but the HB501 has Torr) and outgassing from the an ultra-high vacuum in the column (< emulsion would impair it. This low pressure is necessary for high-resolution imaging and analysis in order to avoid the build-up of contaminating layers on the specimen. A solution to this problem has been drawn from the Lenard window electronographic camera, the Spectracon, which was developed by J.D. McGee at Imperial College. The Spectracon (McGee and Wheeler, 1962) was a single-stage image tube provided at the output end with a mica window that was thin enough for the majority of 40-keV electrons to pass through, but thick enough to withstand atmospheric pressure and be completely vacuumtight. McGee used mica 4-pm thick and the photographic emulsion on a thin base was pressed against it. Only small, slot-shaped windows were strong enough and the main use of this detector was for recording spectra.
IMAGE RECORDING IN ELECTRON MICROSCOPY
-
149
Objective lens
FIG. 1. Schematic of the VG HB501 scanning transmission electron microscope.
At the Royal Greenwich Observatory, electronographic cameras were developed having much larger mica windows, 90 mm diameter, for recording star fields (McMullan et al., 1987).The mica was still 4 pm thick and with such a large window the differential pressure across it had to be kept to below about 100 Torr by loading the photographic film through an air-lock. In a similar way, the ultra-high vacuum in the column of the STEM can be protected from the outgassing of a photographic emulsion by means of a
150
D. MCMULLAN
Lenard window. In this case, instead of mica that disintegrates into small flakes if it breaks, a 7-pm film of aluminized Kapton polyimide is used. The Kapton is pulled into close contact with the emulsion by an electrostatic field, a technique which was originally developed by P. Griboval (1979) for a largefield electronographic camera. The 35-mm film is introduced through an airlock because the 23-mm diameter window would be damaged by exposure to atmospheric pressure. A cross-section of the window assembly is shown in Fig. 2. ,
10rnm
Rough vacuum Prfr
mounting
HT
Kopton Viton 0 r i n g
plate
L- i UHV
Incident electrons
FIG.2. Kapton window assembly used in the STEM microdiffraction camera. Before exposure, the HTplate is lowered to bring the photographicfilm close to the Kapton; the potential of the HT plate is then raised to 1 kV, pulling the Kapton into close contact with the emulsion.
With this recording method, the exposure time is about 500 times less than with the phosphor plus film camera, and 10-pm features in the patterns can be distinguished, which represents an improvement in resolution of about 10 times. It is now possible to obtain good-quality microdiffraction patterns from the minute volume of matter that corresponds to the resolution element of the microscope (-0.5 mm diameter); in conventional TEMs the smallest area of the specimen that can be illuminated for convergent-beam electron diffraction is at least 5 nm diameter. Figure 3 shows a pattern from a diamond specimen obtained with an exposure of 250 ms. The greatly reduced exposure times facilitate the study of radiation-sensitive materials and also structural features comparable in size with atomic spacings. The microdiffraction camera built at the Cavendish (Rodenburg and McMullan, 1985) is shown in Fig. 4 mounted on the HB501 STEM.
FIG. 3. Example of a microdiffraction pattern from diamond (very close to a [IOO] pole); 100 keV, direct recording, exposure 250 ms.
FIG.4. The microdiffraction camera in position on the column of the HB501 STEM. The long projecting cylinder contains the device for inserting a loop of 35-mm film through the air-lock gate valve up to the Kapton window which is on the axis of column. The film and TV cameras for recording diffraction patterns from the phosphor screen are on the left side of the column.
152
D. MCMULLAN
ELECTRON ENERGY-LOSS SPECTROSCOPY Again, an imaging detector is most desirable for electron energy-loss spectroscopy (EELS), in which the inelastically scattered beam of electrons from the specimen is analysed in a spectrometer. The classical way of recording an EELS spectrum is to deflect it across a narrow slit and count the electrons using, for example, a scintillator and photomultiplier. But because of the very low count rates at high-energy-losses,it can take a long time (many minutes) to accumulate sufficient counts; parallel recording without a slit is therefore highly desirable. One of the problems in implementing this is the very large dynamic range of the counts: it is desirable for quantitative work to record the low-energy-loss part of the spectrum; the ratio of the zero-loss counts to those at, say, 285 eV (the carbon K edge) is about 5000. Detectors that have been used for the parallel recording of EELS spectra include a scintillator coupled to a silicon intensified target (SIT) vidicon television camera or a self-scanned silicon photodiode array, and direct exposure of a diode array to the electron image. There is a problem with all these detectors because, with high-energy electrons, the image scale of a reasonably sized spectrometer is very small (typically -2 pm per eV). A magnifying electron lens following the spectrometer is therefore essential for high-resolution work. Although a SIT vidicon lens-coupled to a scintillator can give good results (Shuman and Kruit, 1985) because of the gain in the silicon target (up to -2000, which is sufficient for even single electrons to be detected), it has a number of disadvantages: the dynamic range is limited by the rather high read-out noise and the limited storage capacity of the target; the geometrical accuracy depends on the stability and linearity of the vidicon scanning fields and there may be local distortions due to the surface charges on the target, which will prejudice accurate calibration of the camera sensitivity; a SIT vidicon is easily damaged by too high a light level and is expensive to replace. Self-scanned diode arrays used to detect photons from a scintillator have several advantages: high responsive quantum efficiency; well-defined geometry that simplifies calibration; and resistance to damage by an excessive light input. Their main drawback is the high read-out noise. Krivanek et af. (1987) have described a parallel detector in which the optical coupling between a YAG scintillator (Autrata et al., 1984) and a 512-diode self-scanned array is by a fibre-optic plate (Fig. 5). To obtain reasonable resolution, the scintillator must be very thin (- 50 pm) because there is no focusing of the light and the diode pitch is 50 pm. Krivanek has found that a single 100-keV electron produces about 160 charges in a photodiode, and calculations show that, since the read-out noise is 3500 charges RMS, the minimum signal per read-out is 480 high-energy electrons (for a DQE of not less than 0.5). The
IMAGE RECORDING IN ELECTRON MICROSCOPY
153
FIG.5. Schematic of the parallel detection system developed by Krivanek et al. (1987) showing the magnetic electron energy analyser, quadrupole lenses 41-3 for magnifying the spectrum, and the YAG+ photodiode detection system.
saturation charge of the diodes is very high, typically 1.4 x lo8, and therefore the dynamic range for a single read-out is about 2000. At very low electron rates, say 10 s-' per channel, the signal must be integrated for about 1 min before reading out, and therefore the chip is cooled to reduce the dark current. With the fastest read-out (10 ps per channel) the maximum count rate is 1.5 x lo8electrons per second per channel, which may not be enough to record the zero-loss peak. An alternative to detecting the photons from a scintillator with a diode array is to expose the array directly to the high-energy electron beam (Bourdillon et al., 1985). This overcomes the disadvantage of the high readout noise because of the EBIC gain in the silicon diodes: a single 100-keV electron produces 2.5 x lo4 charges, giving a signal-to-detector-noise ratio of about 7. Single electrons are therefore readily detected, but the maximum counting rate is very limited: with 6 ms between read-outs the maximum rate is lo6 s-I per channel. The most serious problem is radiation damage to the diodes by the electron bombardment: either almost immediate destruction if the zero-loss beam is accidentally incident, or slow deterioration with normal use; the cost of a replacement is high.
-
-
CCD+YAG DETECTOR FOR EELS A type of optical detector that appears to be very suitable as a detector for EELS is the charge-coupled device (CCD), which combines the high responsive quantum efficiency of a silicon diode with extremely low read-out
154
D. MCMULLAN
noise, making it possible to realize a high DQE. With a cooled CCD the RMS read-out noise can be only a few electrons and lens-coupling to a YAG scintillator is therefore feasible in spite of the low optical transfer efficiency; the magnification can be set to match the size of a scintillation to the CCD resolution. A proposal for using a CCD for parallel EELS recording has been made by Strauss et al. (1987) at the Argonne National Laboratory; at the Cavendish Laboratory a detector of this type is being developed based on a CCD camera manufactured by Wright Instruments*, that uses an English Electric P8600 CCD sensor. The P8600 CCD has an imaging area 8.5 x 6.4 mm2 (385 x 288 pixels) and if it is cooled the read-out noise is less than 10 electrons RMS. Calculations-tests show that witha2 lens coupling only about three electrons will be generated in the CCD per high-energy electron on the YAG, but even so, because of the low read-out noise, the DQE is reasonable at low count rates. The DQE is given by
where E is the number of incident high-energy electrons and n is the read-out noise referred to the scintillator, i.e. the CCD read-out noise divided by the number of CCD charges produced by a single electron. In the present example, the minimum signal for a DQE >0.5 would be only 10 highenergy electrons (30 CCD charges), so that the integration time for a 10electrons per second per channel signal (DQE=0.5) would be 1 s. The CCD saturation charge per pixel is 3 x lo5 and the dynamic range would be lo4. These calculations are summarized in Table 1 and compared with the selfscanned diode array used by Krivanek. The P8600 is a frame-transfer device with an imaging area and a storage area. The spectrum will be focused on the imaging area and will cover about 200 pixels at right-angles to the dispersion. At the end of an exposure period, the charges will be shifted down rapidly (in about 250 p s ) and binned into the first row of pixels in the storage area; these charges will then be shifted down one row so that the first row is ready to receive another spectrum from the imaging area. It will thus be possible to store up to 290 spectra before a readout is necessary and this will take about 5 s to be completed. There are 385 columns, so that there will be only this number of channels per spectrum; but the length of the spectrum can be increased by changing the potential of the spectrometer drift tube between exposures and recording up to, say, five 385channel sections cyclically. The Wright CCD camera will flat field, drift correct, and add the spectra to give a spectrum of nearly 2000 channels, each with a maximum count of about 5 x lo6 electrons.
-
* Wright Instruments Ltd., 4 Chalkwell Park Road, Enfield, Middlesex EN1 2AJ, England.
IMAGE RECORDING IN ELECTRON MICROSCOPY
155
TABLE I
Comparison of self-scanned diode array and CCD detectors Self-scanned array + YAG Minimum count per channel’ Maximum count per channel? Dynamic range Integration time at 10electrons- rate$ Number of channels
’
480 (8.7 105)
2000 48 512
CCD + Y AG 9 electrons
lo5electrons I04 Is 385
* for DQE=0.5; per integration period.
t per integration period. $ for DQE=0.5.
The CCD has the advantage over a diode array of being a two-dimensional sensor, which will be useful in the adjustment of the spectrometer and for ensuring that the spectrum always falls exactly on the same area. A large CCD sensor coupled to a YAG scintillator would also be able to replace the photographic emulsion and other detectors in many other electron microscope applications, including recording diffraction patterns and direct images in TEM as well as providing a very sensitive on-line viewing system. For the latter, the CCD will have to be scanned at TV rates and the read-out noise will be much larger, typically 70 electrons RMS. Equation (1) shows that by fibre-optically coupling the scintillator to the CCD (using a tapered optic), and thus producing perhaps 50 CCD charges per high-energy electron, it will be possible to obtain DQE approaching unity for images having a minimum of about 10 electrons per pixel per frame. At present the largest available sensors have about 200 000 pixels, but sensors having as many as four million are reported to be under development. In applications not requiring high resolution, existing CCDs can of course be used. ACKNOWLEDGEMENTS I wish to acknowledge the important contributions made by Drs S. Berger and J. Rodenburg of the Cavendish Laboratory, and by Dr J. Wright of Wright Instruments Ltd., and to thank Professor A. Howie FRS and Dr L. M. Brown FRS of the Cavendish for stimulating discussions.
REFERENCES Autrdta, R., Schauer, P., Kvapil, J. and Kvapil, Jo. (1984). In “Electron Microscopy 1984” (8th European Congress on EM) (ed. A. Csanady, P. Rohrlich, and D. Szabo), pp. 617-625 Becker, A. and Kipphan, E. (1931). Ann. Phys. 10, 15-51
156
D. MCMULLAN
Bourdillon, A. J., Stobbs, W. M., Page, K., Home, R., Wilson, C., Ambrose, A,, Turner, L. J. and Tebby, G. P. (1985). Inst. Phys. Conf. Ser. No. 78, 161-164 Dufour, A. (1912). C.R. Acud. Sci. 158, 1339-1341 Griboval, P. J. (1979). In “Adv. E.E.P.” Vol. 52, pp. 305-314 Knoll, M. and Ruska, E. (1932). Z. Phys. 78, 318-339 Krivanek, 0. L., Ahn, C. C. and Keeney, R. B. (1987). Uftramicroscopy22, 103-1 16 Lallemand, A. (1936). C.R. Acud. Sci. 203,243-244 McGee, J. D. and Wheeler, B. E. (1962). In “Adv. E.E.P.” Vol. 16, pp. 47-59 McMullan, D., Powell, J. R. and Curtis, N. A. (1976). In “Adv. E.E.P.” Vol. 40B, pp. 627-640 Rodenburg, J. M. and McMullan, D. (1985). J . Phys. E 18,949-953 Shuman, H. and Kruit, P. (1985). Rev. Sci. Instrum. 56, 231-239 Strauss, M. G., Naday, I., Sherman, I. S. and Zaluzec, N. J. (1987). Uftrumicroscopy22,117- 123
A $Inch 792(H) x 492(V) Pixel Colour Synchro Vision CCD Image Sensor N. HARADA, Y. ENDO, C. TANUMA, M. IESAKA, Y. EGAWA, H.NOZAKI, S. UYA, S. SANADA, A. FURUKAWA, S. MANABE Tmhiba Corporation, Research and Develapment Cenler, Kawasaki-City, Japan
and
0. YOSHIDA Toshiba In lernalional LId., Uxhridge, Middlesex, England
INTRODUCTION An improvement in resolution is desirable for high-quality television images. Generally, the resolution limit is determined by the number of pixels in the sensor. CCD swing operation (Harada et al., 1983a; Yoshida et al., 1985) is a unique method for meeting this requirement by enhancing resolution without increasing the pixel density in the sensor. In swing operation a CCD chip is swung in the horizontal direction. A black and white Synchro Vision (SV)-CCD with CCD swing operation realizes a doubling in resolution (Yoshida et al., 1985). CCD swing operation was also applied to a one-chip colour CCD with a checkerboard pixel layout (Harada et al., 1983b). However, it was difficult simultaneously to enhance resolution for luminance and colour signals with this device because the optimum swing distances for these signals were different. A new high-resolution SV-CCD image sensor using a +-inch 792(H) x 492(V) pixel CCD chip has been fabricated. This device has achieved simultaneous resolution enhancement for both luminance and colour signals. In addition, aliasing and Moire effects have been significantly reduced without using an optical low-pass filter.
SWINGCCD OPERATION Figure 1 shows the principle of swing CCD operation for a black-and-white SV-CCD. An interline transfer CCD chip is used in this device. The pixel unit is composed of a photodiode and a vertical CCD register which is optically shielded. The area of the photodiode is the aperture for incident light. The 157 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL 74
Copyright 0 1988 Academic Press Limited All rights of reproduction i n any form reserved ISBN 0-12-014674-6
1 A
(b)
nit picture element Photosensing section
B
1 t
Swing width
Signal charge reod-out pulses
FIG.1. Black and white Synchro Vision CCD image sensor principle.
el
2
x, i z i x?)
xq
FIG.2. Swing operation for a one-chip colour CCD image sensor.
A PIXEL COLOUR SYNCHRO VISION CCD IMAGE SENSOR
159
CCD chip is periodically swung right and left in a horizontal direction at the 30 Hz frame frequency. The swing travel is one-half of the picture element centre-to-centre distance, PH/2. Signal charge packets, stored in the photodiode, are transferred to the vertical CCD register during vertical blanking intervals. A full-frame element comprises the signal stored at the first site during one field and that stored at the second, adjacent, site during the next field interval. This is equivalent to an increase in the number of pixels in the horizontal direction. The pixel signals obtained in fields A and B are moved with respect to each other in the horizontal direction on the monitor so as to correspond spatially with the sampling point on the CCD chip, resulting in a twofold enhancement of horizontal resolution. Figure 2 shows CCD swing operation for a one-chip colour CCD image sensor. White, cyan and yellow colour filters are arranged in a checkerboard pattern on the CCD chip. The most suitable swing distance for this device to enhance the resolution in the luminance signal is PH/2. On the other hand, the optimum distance for colour signals is 3PH/2.A two-step swing was therefore applied, as shown in Fig. 2. This brought about a resolution improvement for both luminance and colour signals. However, Moirk and aliasing phenomena could not be adequately suppressed.
DEVICEFABRICATION Figure 3 shows a schematic diagram of the CCD chip that is used in the new colour SV-CCD image sensor. The chip is of a f-inch format, 792(H) x 492(V) Colour filter
Green Red
R
Blue
I
4 +-l
2nd H-CCD 3rd H-CCD
9H3 M 2 +HI FIG.3. Schematic diagram of a colour Synchro Vision CCD chip.
N. HARADA ET A L
c3
T
t tt
R BG
E E
e
Unit pixel size : IOpm ( V lx8.5pm ( H l
cd
FIG.4. Photograph of the Synchro Vision CCD chip.
I
pixels, interline transfer type with a triple read-out register. Organic R, G and B stripe colour filters are mounted directly on the chip. Signal charge packets, corresponding to the R, G and B pixels, are read-out by the first, second and third horizontal CCD registers, respectively. Each read-out register is driven by 4.7 MHz clock pulses. Figure 4 shows a photograph of the CCD chip. The pixel size is 8.5 x 10 pm2 and the chip size is 8.2 x 6.4 mm2. Figure 5 shows the construction of the colour SV-CCD image sensor. The CCD chip is mounted on two piezoelectric bimorph actuators arranged in
CCD chip
bimorph FIG.5. Colour Synchro Vision CCD image sensor construction.
A PIXEL COLOUR SYNCHRO VISION CCD IMAGE SENSOR
161
parallel with each other. Swinging of the CCD in the horizontal direction is operated by applying 30 Hz voltage pulses to the actuators. The entire SVCCD image sensor is a box-shaped package, with a 22-pin dual in-line format, whose dimensions are 34(L) x 23(W) x 9(H) mm3. COLOUR IMAGINGPRINCIPLE Figure 6 shows the swing operation principle for a colour SV-CCD image sensor with R, G and B stripe filters. The swing distance is 3 P ~ / 2In. this swing operation, the numbers of spatial sampling points for the R, G and B colour signals are simultaneously increased by a factor of 2. This means that the resolution for the luminance signals is also improved by a factor of 2. This may be understood by reference to Fig. 7 which shows a comparison of the pixel layouts of a conventional CCD without swing operation and of a colour SVCCD with swing operation. The SV-CCD is operated in a filed integration mode, in which signal charges stored in all pixels are transferred simultaneously to the vertical CCD register during the vertical blanking period. The R, G and B output signals for fields A and B are therefore shifted not only in the horizontal direction but also in the vertical direction. By swing operation, the R,G and B pixels for a frame are rearranged from a stripe layout to a checker board pattern, so that the horizontal limiting resolutions for both luminance and colour signals are simultaneously improved by a factor of 2.
I 194
I
Swing width
I
r
Photosensing section
Signal charge read-out pulses
FIG.6. Colour swing operation principle.
- I b 3512
FIG.7. Pixel layout comparisonbetween a conventional CCD without swing operation (upper) and a colour Synchro Vision CCD with swing operation (lower).
genemtor
FIG.8. Block diagram of circuitry used for colour image reproduction.
163
A PIXEL COLOUR SYNCHRO VISION CCD IMAGE SENSOR
IMAGEREPRODUCTION
Figure 8 shows a circuit block diagram for colour image reproduction. By using 3 P ~ / delay 2 lines, the picture element signals obtained in fields A and B are relatively shifted on a monitor so as to spatially correspond to the
(b) FIG.9. RETMA resolution chart pictures obtained by (a) a CCD image sensor without swing operation and (b)a colour Synchro Vision CCD image sensor.
164
N. HARADA ET A L .
FIG.10. Bar-pattern pictures ( a ) without and (b) with swing operation.
sampling points on the CCD. In the camera the starting times of the clock pulses applied to the horizontal CCD register are shifted for fields A and B. Figure 9 shows pictures obtained by a conventional CCD imager without swing operation and by the colour SV-CCD image sensor. The horizontal limiting resolution of the SV-CCD sensor is 380 TV lines for both luminance and colour signals, which is twice that without the swing operation. Figure 10 shows CZP pictures made without and with swing operation. As shown in this figure, the colour Moire effect is significantly suppressed.
CONCLUSIONS A high-resolution one-chip colour CCD image sensor has been realized by using swing CCD operation and a novel +-inch format CCD chip with triple read-out registers. The CCD chip has 792(H) x 492(V) pixels, on which R, G and B organic stripe filters are mounted directly. The colour SV-CCD image sensor has achieved simultaneous resolution enhancement for both luminance and colour signals without increasing the number of pixels packed in the sensor. A horizontal limiting resolution of 380 TV lines has been realized. ACKNOWLEDGMENTS The authors would like to thank H. Iizuka, S. Sano, T. Kubo and M. Wada for their encouragement. The authors also wish to thank the members of the research group who developed the device fabrication technologies.
REFERENCES Harada, N., Endo, Y., Hayashimoto, Y., Egawa, Y., Tanuma, C., Yokoyama, K. and Yoshida, 0. (1983a). Japan Display Tech. Dig. (Oct.) 424-427 Harada, N., Endo, Y., Hayashimoto, Y., Egawa, Y., Tanuma, C., Yokoyama, K. and Yoshida, 0. (1983b). J. Insf. Telev. Eng. Jpn. 37 (10). Yoshida, O., Endo, Y., Egawa, Y. and Harada, N. (1985). IEEE Trans. Electron Devices, ED-32 (8).
Thinned Rear-face Electron-bombarded FT CCDs for LLL TV Imaging L. BERGONZI, M. LEMONIER and M. PETIT Laboratoires d’Electronique et de Physique Appliqute, Limeil Brtvannes, France
INTRODUCTION Progress in the field of image tubes, for example those involved in low-lightlevel (LLL) imagers, streak cameras or oscilloscopes, includes developments in their electronic read-out which, besides conventional image display on a phosphor screen, allows features such as storage, transmission and processing of the information obtained in the form of a video signal. Charge-coupled devices (CCDs) are small and sensitive image sensors. They are therefore particularly well adapted to convert a two-dimensional transient image into a “video” signal. In the case of LLL TV applications, two methods can be used for obtaining electronic read-out of the image intensifier tube (Richard and Lemonier, 1986): optical coupling between the tube and the CCD; and integration of an electron-sensitive CCD inside the tube, in place of the phosphor screen. The former solution is analysed in detail by Richard et al. (1987).t The latter is described here, first with a presentation of the principle and technology of the electron-bombarded CCD, then with comments on some new results obtained in our laboratories.
CCD ELECTRON-BOMBARDED When an electron with an energy of several keV impinges on silicon, it creates electron-hole pairs through the EBS cascade process, which exhibits two major features (van Roosbroeck, 1965): firstly the average EBS gain (the number of electron-hole pairs created on average by a single electron) is E;/3.6. related to the energy Ei (eV) of the ionizing particle by the relation Thus, as distinct from visible photon detection, the number of pairs created is much larger than unity and can reach several thousands. Secondly the statistical fluctuations of the process are low and the gain variance o2can be
m
m=
t See p. 9 this volume. 165 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
166
L.
BERGONZI, M.
LEMONIER AND M. PETIT
expressed as o2= Ffl, where the Fano factor (I;)is approximately equal to 0.12 in silicon. For example, an average gain fl of 1000 corresponds to a standard deviation o < 1 1. In an EB CCD, the signal electrons can be collected by the closest pixels, giving rise to a signal at that location. Let P be the collection efficiency of the EBS signal electrons inside the CCD. The statistical variance of the global process becomes 02=PlV[l-P(l-F)].
m=
With an EBS gain 2000 and a collection efficiency P= 50%, the average gain will be P N = 1000 with a standard deviation a < 2 4 . Since the noise equivalent signal (NES) inside the CCD is about 100 electrons RMS per pixel and per video frame, the inequality # % N E S % c clearly shows the EBS capabilities for single-event detection and electron counting operation. However, these ideal results can only be obtained if the following conditions are fulfilled: (i) the incident energy is actually dissipated in the silicon material itself; (ii) no incident electron is lost; (iii) the signal electrons are properly collected; (iv) the lateral diffusion of the charge under a given pixel does not overlap the neighbouring pixels; and, (v) the electron bombardment does not disturb the normal CCD operation. OF EB CCDs TECHNOLOGY
In the case of front-face electron bombardment of a CCD, the incident electrons generate signal electrons after they have passed through the top MOS structures. Their energy is mostly absorbed in these layers. The EBS gain versus energy exhibits a high threshold energy (some tens of keV) below which the detection gain is almost zero. This is due to the energy lost in the surface layers. Moreover, because of charge trapping and surface state generation, front-face bombardment can rapidly damage the top oxide layers. In virtual phase CCDs manufactured by Texas Instruments (Everett et al., 1985), one-half of the pixel area is free of gate and there is only a thin oxide layer. The threshold energy should therefore be lower, but only half of the incident electrons are detected. This implies a decrease of the signal-to-noise ratio by a factor On the other hand, rear-face bombardment makes the detection of every incident electron a possibility (Barton, 1975). Since the EBS cascade process takes place close to the rear face (Fig. I), it becomes necessary to thin the CCD in order to conserve both sensitivity and resolution. The final thickness (about 10 pm) must be comparable to the diffusion length of the signal electrons and to the pixel size; it has also to remain larger than the depletion depth under each pixel.
fi.
THINNED REAR-FACE ELECTRON-BOMBARDED FT CCDS
h
I
a,
1 1 5 keV
c
-
Y
f!
167
0
1
3
2
Depth(pm FIG.1. EBS generation rates for different fast electron energies.
Nevertheless, such a thinned CCD exhibits, when operated in an electronbombarded mode, a signficant energy threshold similar to that caused by the surface dead layer. This phenomenon is due to the presence of a depletion zone at the rear face which collects the majority of the signal electrons (Fig. 2). (This zone is some microns thick: the higher the dopant concentration, the thinner the depletion zone.) A passivation layer, for example a highly doped pf layer on the CCD p substrate, is therefore necessary to prevent surface recombination. Its major advantages are the following. (i) It creates a potential barrier that prevents back-diffusion of the signal electrons and therefore enhances the
a
I
b
FIG.2. Band diagram inside the EB CCD (b) with, and (u) without a p t passivation layer, showing (1) electron-hole pair generation;(2) surface recombination;(3) bulk recombination;(4) collection of signal electrons; and (5) rear-face dark current generation.
168
L. BERGONZI, M. LEMONIER AND M. PETIT
collection yield. This barrier also prevents the dark current generated on the rear face from being collected by the pixels. (ii) It reduces the thickness of the depleted zone because of its higher dopant concentration. (iii) Being more conductive than the initial substrate, it secures the potential stability of the rear face. (iv) It improves the collection of the holes generated at the same time as the signal electrons. Regarding point (iv), it is possible to distinguish three zones in the nonpassivated substrate of the EB CCD (Fig. 3): in the first zone (I), the equivalent EBS generation rate is equal to zero because no signal charges are collected from this depleted zone; in the second zone (11) with no electric field, the signal electrons diffuse towards both front and rear faces. The probabilityp(x) for an electron created at position x to reach position Xd is then P(X) =
WxSh[(Xd-
WlLI WILI ’
where L is the diffusion length of the electrons; in the third zone (111), which represents the potential wells under the pixels, all the signal electrons are collected. When used with Klein’s generation function (Klein, 1966), this model fits the experimental results provided that L= 5 pm (compared to about 100 pm for free electrons in p silicon). During the EBS generation cascade, the EB CCD substrate is in a strong injection mode, in which the majority carriers (holes) created by EBS cannot be neglected because of the very small generation volume. This greatly reduces the diffusion length of the electrons because of their reciprocal interaction with holes. The collection of holes at the
Rear face
Front face
I
o
w
I
* Xd
d
x
FIG.3. Model showing three distinct areas in a non-passivated CCD: (I) depleted zone at the rear face; (11) central zone without electric field; and (111) storage zone (CCD potential well).
THINNED REAR-FACE ELECTRON-BOMBARDED FT CCDS
169
rear face is therefore of utmost importance to separate them from the signal electrons. In order to achieve the thinning and passivation operations in a single process, a special FTCP sensor, derived from the Philips 604 x 596 Frame Transfer NXA 1010 CCD (van de Steeg et al., 1985) has been designed by the SSIS Group at Philips Research Laboratories. This sensor is processed on a 15 pm thick epitaxial p layer (5 x lOI4boron atoms cmP3)grown on a (100) p+ ~ m - wafer. ~ ) Selective chemical etching is used to remove the p+ substrate under the sensitive area so that a thin layer (some tenths of a micron) of p + material (about 10l8~ r n - remains ~) at the rear face. The self-supported matrix is being held in an open package, and it can then be used in a rear-face electron bombardment mode.
RESULTS EBS gain Experiments were first performed on an RGS (resistive gate sensor), then on FTCP sensors, both types prepared as described previously. Some experiments concerning front-face bombarded sensors show a short lifetime (a few minutes) and weak EBS gain (about 10 for 10 keV). When operated in rear-face bombarded mode, the sensors show no limitation of their lifetime at least up to 1 O 1 O electrons per pixel (Cheng et al., 1978). Figure 4 shows the variation of the EBS gain versus the energy of the incident electrons for three RGS samples with differences in thickness and rear passivation and a
c0
3000-
1
(r
''m
w
nGs
15 jrni without p f
2 RGS 5.7111r1without Pt 2500-3 RGS 20111nwith Pt 4 F T 4 - P 131iin with P c
/''
0
4
fi
8
10
12
14
16
Energy ( k e V ) FIG.4. EBS gain versus electron energy for different samples.
170
L. BERGONZI, M. LEMONIER AND M. PETIT
passivated FT4-P sensor. The importance of the thickness of the silicon substrate compared to the electron diffusion length is clearly demonstrated by curves 1 and 2. The comparison between curves 2 and 3 exhibits the tremendous importance of the passivation layer; the EBS gain is much higher in the presence of a p f layer, even in the case of a thicker sample, showing the effect of hole collection at the rear face. Curve 3 effectively reaches the slope 1/(3.6 eV) for incident energies larger than 8 keV. Curve 4 was obtained with the FT4-P sensor. The frame transfer structure of this sensor, with separate imaging and storing areas, is particularly well adapted to rear-face electron bombardment. Since no lag occurs during the EBS process, no smearing effect is perceptible provided the electron bombardment is stopped during transfer of the frame. Curve 4 shows a perfect fit with the expected 1/(3.6 eV) slope. The energy threshold is as low as 4.5 keV. A single 10 keV electron creates about 1500 signal electrons.
Dark Current Figure 5 shows the influence of the p + layer on both dark current and EBS gain in the case of a badly passivated sample. The video image in Fig. 5(a)was generated during 10 keV uniform bombardment and that in Fig. 5(b) was obtained at 60°C without bombardment. When the p + layer is too thick (at the edges of the samples), neither dark current nor EBS signal can be seen; when the p+ layer has the right thickness (a thin rectangular ring around the middle of the sensor), it stops the spurious dark current signal and gives high EBS gain; when the passivation layer has been completely removed, the EBS gain is very weak and the dark current is high.
Electron Counting As demonstrated by the EBS gain curves, single-electron detection capability is expected for a frame transfer EB CCD under 10 keV bombardment. In an SEM, the electron beam is pulsed in phase with the video signal and the spot is focused accurately on one pixel. The pulse height distribution of the signal from this single pixel is plotted in Fig. 6. The peaks corresponding to 0, 1 and 2 electrons are clearly visible, showing that the system is able to detect single 10 keV electrons. The EB CCD can also be operated in an integration mode owing to the fact that a signal corresponding to up to about one hundred 10 keV electrons can be stored in each pixel’s potential well. Grey level determination as well as single-electron detection are therefore possible with the same component. The single-electron peak of Fig. 6 corresponds to about 800 signal electrons in the chosen pixel and 700 in the neighbouring ones (the sum being equal to
THINNED REAR-FACE ELECTRON-BOMBARDED FT CCDS
171
FIG.5. Smoothed images given by a non-homogeneously thinned FT4-P sensor during (a) a uniform 10 keV electron bombardment, and (b) heating (60°C) without bombardment.
1500, the value of the EBS gain at 10 k e y . These values show that, when more than one pixel is bombarded, a correction for the lateral diffusion of the charges inside the substrate has to be performed in order to achieve twodimensional single-electron detection.
CONCLUSION Experiments have demonstrated the capability of the FT4-P frame transfer EB CCD to record electronic images with a very low number of electrons (one or more) impinging on each pixel. These promising results lead to various applications, such as: (i) LLL TV sensors in which an EB CCD can replace the microchannel plate, the screen, the fibre-optic taper and the image pick-up
172
L. BERGONZI, M. LEMONIER AND M. PETIT n
.
=!
1000
0 U
t
c
800-
3
0 0
+ 0
600-
+
L
Q)
n E
400-
0
200
100
Signal (a.u.1 FIG.6. Pulse-height distribution of the signal from a single pixel under point bombardment with 10 keV electrons.
device. It offers both gain and imaging with enhanced signal-to-noise ratios compared to the present LLL TV sensors. (ii) Fast oscilloscopy, where an EB CCD can be used in place of the screen to detect signal spots a t a level as low as one electron. Image processing can then be used for determination of the centre of gravity of the spot and the elimination of gaps in the trace, allowing higher performance for future fast oscilloscope tubes. ACKNOWLEDGEMENTS The authors are very grateful to Messrs M. G. Collet, H. L. Peek, A. J. P. Theuwissen (Philips Research Laboratories, Eindhoven, NL), M. Vittot, J. C. Richard (LEP)and C. Piaget (Portenseigne, Louviers), for their support and helpful collaboration.
REFERENCES Barton, J. B., Curry, J. J. and Collins, D. R. (1975). In “Proc. Int. Conf. on the Application of CCD, San Diego,” pp. 133-145 Cheng, J. C., Tripp, G. R. and Coleman, L. W. (1978). J. Appl. Phys. 49, 5421-5426 Everett, P., Hynicek, J., Zucchino, P. and Lowrance, J. (1985). Opr. Eng. 24, 360-362 Klein, C. A. (1966). Appl. Opi. 5, 1922-1924 Richard, J. C. and Lemonier, M. (1986). In “Proc. Int. Topical Meeting on Image Detection and Quality,” pp. 13-16 Van de Steeg, M. J. H., Peek, H. L., Bakker, J. G. C., Pals, J. A., Dillen, B. M. G. H. and Oppers, J. M. A. M . (1985). IEEE Trans. Electron Devices ED-32, 1430-1438 Van Roosbroeck, W. (1965). Phys. Rev. 139, A1702-A1716
A 2 x 2048 Pixel Bilinear CCD Array for Spectroscopy (TH 7832 CDZ) J. L. COUTURES and G. BOUCHARLAT THOMSON-CSF Elecrrvn Tubes Diiiision. Boulogne-Billancourt, France
INTRODUCTION Photodiode linear arrays are perfectly adapted for spectral analysis, since the two types of information contained in the spectrum, the light intensities of spectral rays and their positions on axis, are directly converted into an electrical signal. Among linear devices with photodiode pixels presently manufactured, we find photodiodes associated with CCD registers that have the advantage of low read-out noise (less than 200 electrons), but with too small a photodiode size (about I5 pm-30 pm) for suitable recording of spectral energy along the length of each spectral line. Other devices, with large sizes (up to 2.5 mm), associated with shift registers have both a higher readout noise level (more than 1000 electrons) due to the large capacitance of the read-out bus, and a bad read-out efficiency for low level signals. The TH 7832 CDZ bilinear array is a new device specially adapted to low-level detection (exposures < 7 nJ cm-*) with a photodiode signal read-out efficiency better than 97% over the entire dynamic range which exceeds 70 dB.
GENERAL DESCRIPTION The bilinear CCD array comprises two identical parts (see Fig. 1). Each part is composed of one line of 2048 photodiodes with its associated circuits. The width of the sensitive area is 750 pm with a 13 pm pixel pitch, corresponding to a resolution of 77 pixels mm-'. The overall length is 26.6 mm. The gap between the two sensitive lines is 500 pm, There are two identical sensitive lines to allow processing of two pieces of information at the same time, e.g. for background signal suppression. Two CCD registers are associated with each photodiode line, one to read information from odd pixels and the other from even pixels. As usually used, the output stage a t the end of each CCD register converts the signal charge packets into analogue voltages. One of the special features of this device is an electrical input on each CCD register which 173 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any Corm reserved ISBN 0-12-014674-6
S
1B
750 pm Photodiode area
500
750
P
pn
c.
'1A
'
2A
Photodiode area
ZB
t
t e l e c t r i c a l inputs
FIG.1. Structure of the 2 x 2048 bilinear CCD array.
output s t a g e s
A PIXEL BILINEAR CCD ARRAY FOR SPECTROSCOPY
175
generates bias charges. These bias charges, added to the signal charges, increase the reading efficiency of the signal. BIASCHARGE
AND
SIGNAL CHARGE TRANSFER
Each input stage converts the bias voltage into charge packets at the same frequency as the CCD register transfer. These charge packets, or bias charges, are shifted into each CCD register stage during the integration time. At the end of the integration time, each bias charge is simultaneously transfered to each photodiode so that the bias charge is added to the signal charge. After that, the added charges are transfered simultaneously from all the photodiodes to the CCD register. They shift along the register to the output stage. One burst of summed charge packets is followed at the register by one burst of bias charge packets which is used for the next integrated signal. The adjustable input voltage thus generates an added bias voltage on the signal samples at the output.
EFFECTSOF
THE
BIAS CHARGE
Reading Eficiency
The effect of the bias charge on the signal reading efficiency is apparent in the time response to a transient light. Figure 2(a) shows the transient response at one output (odd or even pixels) without the bias charge. The light has a quasi-uniform level after the transient. The output signal level reaches the If the light has a duration less correct value after several integration times (Ti). than one integration time, the read-out efficiency is given by the ratio of the first step value to the asymptotic value. In Fig 2(a), the read-out efficiency is about 15%. Figure 2(b)shows the response with a bias charge. In this case, the output signal has the correct value after the first integration time interval. Usually, with photodiode arrays of large length read by CCD registers or by switches to a bus, the signal reading efficiency decreases for low-level signals. The bias charge reduces this effect. With a bias charge corresponding to 1 V bias voltage read at the output, and for signal levels of 1.5% to 50% of the saturation level, the reading efficiencies are respectively 98% and 99% (see Fig. 3). Device Noise Level
In order to have a large dynamic range, the detector must have the lowest possible detection noise. Table I gives all the noise source values (in electrons) of the bilinear detector, from experimental results taken at room temperature,
176
J. L. COUTURES AND G. BOUCHARLAT
(b)
FIG.2. Response to a transient light source (a) without bias charge and (6) with bias charge. Each step represents the signal read at one output in the integration time.
5 ms integration time and 1 MHz output sample frequency. The optical input stage contributes the highest noise. This noise is proportional to the squareroot of the photodiode capacitance, so the larger the photodiode is, the higher its capacitance is, and the greater is the optical stage reset noise. Figure 4 shows experimental results for the effect on noise levels of varying the added bias charge. The solid curve gives the noise variation under standard transfer conditions. The noise level increases with the bias charge level up to an asymptotic value of about 330 electrons. In order to prove that the noise variation is not due to the use of a bias charge, the dashed curve represents the noise variations when the bias charge is transfered from the electrical input to the output stage, without a transfer between the photodiodes and the register. It can be seen that the read-out efficiency (Fig. 3) and
A PIXEL BILINEAR CCD ARRAY FOR SPECTROSCOPY
-
100 -
177
Transient signol values:
I
1
Bias charge voltage ( V ) FIG.3. Read-out efficiency versus bias charge for several signal levels.
,--.-Standard
cn
conditions
-- --*- -- Without photodiode
*--*-----b-
reading
I 0.5
I
1 Bias charge voltage ( V 1
FIG.4. RMS noise versus bias charge.
the noise level (Fig. 4) vary in the same way with bias charge level. Bad transfer efficiency from photodiode to CCD register therefore acts as a low-pass filter for signal and noise. When the read-out efficiency is increased by increasing the bias charge, the noise level increases with the cut-off frequency of the filtering. The known temperature variation of noise in the CCD register, photodiode
178
J. L. COUTURES AND G . BOUCHARLAT
TABLE I Noise source values of bilinear detector Reset noise of optical input stage Reset noise of electrical input stage Reset noise of reading output stage Shot noise of leakage current in photodiode Shot noise of leakage current in CCD register Transfer noise in CCD register Output amplifier noise
295e70e1 10e53e1 le45eIOOe-
RMS total noise
345e-
leakage current and reset noise allow the output noise at low temperature to be calculated. Figure 5 shows the computed noise variation for several integration times. The noise floor represents the noise value without leakage currents. Depending on the application, the integration time may be chosen to suit the light intensity and the necessary dynamic range. The device is cooled until its noise reaches the noise floor.
GENERAL ELECTRO-OPTICAL CHARACTERISTICS Table I1 gives typical electro-optical characteristics of the bilinear CCD array, at room temperature, 5 ms integration time and 1 MHz output sample frequency.
-100
-80
-60
-40
-20
0
20
Temperature ("CI FIG.5. RMS noise versus temperature for various integration times (indicated in seconds).
A PIXEL BILINEAR C C D ARRAY FOR SPECTROSCOPY
179
TABLE I1 Electro-optical characteristics of bilinear array Output conversion factor Response to a light source (2854 K tungsten filament lamp + infrared filter) Saturation output signal (with 0.8 V added bias charge) Saturation exposure Average dark current (with respect to the integration time) Quantum efficiency at 400 nm wavelength Quantum efficiency at 700 nm wavelength Contrast transfer function (at 500 nm wavelength and Nyquist point) Minimum integration time
1.6 pV electron-' 300 V pJ-' 2v I nJ cm-2 1 mV ms-I 55% 75% 68 % 250 ps
CONCLUSIONS The bilinear CCD array TH 7832 CDZ is packaged in a dual in-line integrated circuit package with quartz window or with fibre-optic window. The detector geometry is suitable for coupling to spectrographs. Integration time reduction is possible because of the good read-out efficiency over the full 70 dB dynamic range. Only a single integration time interval is necessary to suppress background. The bias charge could not be used in two situations. First, when the signal has variations which are slow compared to the integration time repetition interval (200 Hz maximum signal frequency for the minimum integration time); because the noise is then filtered and falls to 110 electrons; second, when the signal contains a constant background level which acts in the same way as the bias charge.
This Page Intentionally Left Blank
Development of EBCCD Cameras for the Far Ultraviolet G. R. CARRUTHERS, H. M. HECKATHORN E. 0. Hulbert Center for Space Research. Nauul Research Laboratory. Washington D C , USA
C. B. OPAL Deparimeni of Asrronomy, Uniuersiiy of Texas, Austin, Texas. USA
E. B. JENKINS and J. L. LOWRANCE Department of Astrophysical Sciences, Princeton University, Princeton. New Jersey, USA
INTRODUCTION Princeton University and the Naval Research Laboratory have ongoing programmes to develop imaging detectors with high quantum efficiency and photon-counting capability. These are based on the use of charge-coupled devices (CCDs) in electron-bombarded mode, with the inputs being photoelectrons emitted by opaque alkali halide photocathodes (Lowrance et al., 1979; Lowrance, 1979; Lowrance and Carruthers, 1981; Zucchino et al., 1981; Carruthers and Opal, 1985).Two primary detector systems of this type are the oblique magnetic focus and the Schmidt camera systems (Lowrance and Carruthers, 1981; Carruthers and Opal, 1985). Both make use of the high quantum yields of opaque alkali halide photocathodes, as compared to the more conventional semi-transparent photocathodes. Opaque alkali halide photocathodes have very high quantum yields in the far ultraviolet (Duckett and Metzger, 1965; Metzger, 1965; Carruthers, 1969), exceeding 50% in some wavelength ranges. These yields are typically a factor of 4 better than obtainable with semi-transparent photocathodes of these same materials in the wavelength range longward of the LiF cut-off of l050A. In addition, they can be operated in a windowless mode with sensitivity at much shorter wavelengths. In principle, this advantage should also apply to alkali halides deposited on the front surfaces of microchannel plates, but previous studies (Carruthers and Opal, 1985; Siegmund et al., 1986) have indicated that the quantum yields and photon-counting quantum efficiencies of CsI-coated MCPs are a factor of 3-5 less than those of opaque CsI photocathodes deposited on flat substrates (e.g. aluminized glass), used in photodiode mode or as the photoelectron source for a pulse-counting MCP 181 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL 14
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
182
G . R . CARRUTHERS E T A L .
detector. We recently completed a more detailed study and comparison of CsI-coated MCPs versus MCPs fed with photoelectrons from a separate, opaque CsI photocathode (Carruthers, in press), confirming a sensitivity better typically by a factor of 4 for the latter configuration. An important objective of the investigation described here was to directly determine the effective quantum detective efficiencies of EBCCD sensors, for comparison with those expected and those of the two previously-discussed MCP detector approaches. In the EBCCD sensors, photoelectrons are accelerated to energies of 1020 keV before impacting on the CCD. The photoelectron energy is converted to electron-hole pairs at the rate of 3.6 eV per pair; hence the theoretical signal pulse produced by a 10 keV photoelectron is of the order of 2700 electrons. There is some loss of photoelectron energy in traversing inactive surface “dead layers”, from which generated secondary electrons are not efficiently collected, typically resulting in threshold energies of 3-8 keV. Even so, photoelectrons of 12-16 keV energy produce charge pulses of more than 2000 electrons, in an area of the CCD comparable to the pixel size. This signal is many times the typical CCD read-out noise levels of 10-100 electrons per pixel; hence it is relatively easy to detect and count individual photoelectron events.
COMPARISON AND TESTING OF EBCCD SENSORS Methods of Analysis
Three methods, which are described below, can be used to determine the quantum detection efficiency of an EBCCD sensor. Only the latter two were used in the results discussed here. In the first method, the EBCCD sensor is operated in true photon-counting mode: that is, each signal pulse (often involving more than one pixel) within a specified range of pulse amplitude is counted as one detected photoelectron. This can be compared directly with an independent measure of incident photon flux to yield a quantum efficiency. In practice, this method requires that the integrated signal level in an exposure be low enough (about 0.1 photoevent per pixel or less) to avoid superimpositions of more than one “hit” in or near a pixel. This is because over-large events are rejected or at best counted as a single event. This mode is preferred in actual low-light-level applications of EBCCD (and MCP) sensors since spurious events due to ions or cosmic rays can be discriminated against. However, it is difficult to calibrate such a detector because it is difficult to measure accurately the very low light levels with typical comparison standard detectors. A first step in developing a photon-counting mode of operation is to
DEVELOPMENT OF EBCCD CAMERAS FOR THE FAR U V
183
determine the average signal level associated with single-photoelectron events. This is obtained from a histogram of event amplitudes (pulse height distribution) recorded over a period of time with the detector illuminated at a low light level (to minimize overlapping events). Examples of pulse-height distributions for an EBCCD with various incident electron energies are shown in Zucchino et al. (1981). In a real-time photon-counting system, the read-out electronics are set to reject events having amplitudes much below or much above this average photoevent amplitude. However, the average single-event amplitude can also be used, in postacquisition data processing, as the basis of the second method for determining the number of recorded photoevents. This method can also be used when the detector is operated at a light level too high for true photon-counting mode, i.e. in the range above 0.1 photoevents per pixel per read-out, up to the full well capacity (in the range 30-200 photoevents per pixel per read-out). In this latter, “analogue integration” mode, the average signal level per photoevent can be divided into the integrated analogue to yield the number of detected photoevents. The third method for determining the quantum detection efficiency is based on measurements of the integrated analogue signal read-out from the CCD, plotted as a function of accelerating voltage. Theoretically, each 3.6 eV lost by the primary photoelectron should yield one electron-hole pair (i.e. a signal of 1000 electrons should be produced per 3600 eV primary energy above “dead layer” threshold). If the absolute signal output (in electrons per frame, integrated over the image area) at a given voltage is known, this signal divided by the theoretical gain yields the number of detected photoelectrons. The disadvantage of this method is the lack of an independent verification of the theoretical yield of one electron-hole pair per 3.6 eV. The actual yield could be lower, owing to recombination of electron-hole pairs before charge collection.
EBCCD Tests at Princeton In a series of runs at Princeton University in April 1984, we compared the outputs of an RCA CCD and two TI CCDs operated in electron-bombarded mode, and a two-stage (chevron) MCP. The RCA CCD was a thinned, backside bombarded, three-phase device, having 320 x 5 12 30-pm pixels. The TI CCDs were unthinned, front-side bombarded, virtual-phase devices having 328 x 490 23-pm pixels. The tests were made using the same electron-optical assembly, a Schmidt camera with an opaque KBr photocathode fed by a Cerenkov light source (Fig. 1). The detectors were all placed, sequentially, at the electron focus of the Schmidt camera. A KBr photocathode was used instead of the Csl
184
G . R . CARRUTHERS E T A L . CERENKOV LIGHT S O U R C E
,
,
SCHMIDT
-
CAMERA
1 1
‘I-
I
I .
I
I
M
COLLIMATING
~
WINDOW^ F ~
MIRROR
C s I PHOTOCATHODEA
gosf SOURCE
C C D H E A D ASSEMBLY
FIG. I . Setup of the EBCCD Schmidt camera and Cerenkov light source used in the tests at Princeton University. Not shown is an insertable mask containing a 2.54 cm diameter aperture, for runs at a lower light intensity.
photocathode used in our previously reported results, because (i) the KBr photocathode is less susceptible to changes in its sensitivity and spectral response with time when exposed to air between runs; and (ii) it provides a narrower wavelength range of response of the detector, allowing more accurate comparison of the signal with that of a comparison standard photomultiplier. The EBCCD read-out electronics were operated in analogue read-out mode, with signal levels per pixel digitized and recorded on tape. A liquidnitrogen cooling system maintained the CCDs at -80°C or less during operations. The chevron MCP was operated with a single-anode (nonimaging) read-out to a single-channel photon-counting system as in previous tests at NRL. The MCP detector was operated with a photocathode-to-MCP accelerating voltage of 4 kV, in the optimal range for MCP detection (Carruthers, in press). Theoretically, the single-photoelectron responses of the two types of CCD differ in that in the thinned back-surface bombarded CCD there is likely to be significant lateral spread of the generated secondary charge before collection at the front-side electrodes. Therefore, the charge produced by a single photoelectron may be spread over several pixels. For the front-side bombarded CCD, on the other hand, the secondary charge is produced closer to the collecting electrodes and less spreading into adjacent pixels is expected. Charge transfer inefficiency also spreads charge, during the read-out process, from the packet originating in the “target” pixel into those corresponding to trailing pixels. The Cerenkov source yielded a spot image 0.66” in diameter, which corresponded to 1.73 mm diameter at the electron focus of the EBCCD Schmidt camera. Its intensity was measured before and after the runs using a separately calibrated EMR 541G photomultiplier. Using a 4.5 mCi 90Sr
DEVELOPMENT OF EBCCD CAMERAS FOR THE FAR UV
185
source and MgF2window, this was typically 14 photons cmP2sec-’ k ’near 1220A. To allow operation at reduced light level for photon-counting, a motorized aperture stop (25 mm diameter) could be placed in front of the 100 mm camera aperture to effect intensity reduction by a factor of 16. Runs were made with high and low light intensities and also with different accelerating voltages (up to 17.5 kV, and including zero volts to allow dark current and read-out noise determination). Integration times used were typically 30 s, as the system used required 23 s for read-out of the CCD. Weak exposures were expected to give better determinations of single-event amplitudes (by minimizing the number of multiple hits in single pixels), and were necessary for post-run photon-counting data analysis with the hardware and software used at Princeton. Stronger exposures, on the other hand, gave better statistical accuracy in integrated signal measurements. The data recorded on tape were later processed on the PDP-11/44 and Ramtek image-processing system at NRL. The NRL data system displays pixel intensities in “data numbers” or DN, whose absolute relation to signal levels (electrons per pixel) in the data obtained for the RCA and TI CCDs at Princeton were not known a priori. Individual pixel signal amplitudes were used to generate histograms of number of pixels versus amplitude. These are not true pulse-height distributions, of the type shown by Zucchino et al. (1 98 I), because the sharing of single-photoevent charge among two or more pixels was not taken into account. For a spot image, a “density volume”, the summed product of the number of pixels times their D N amplitudes, can also be determined. If the average density volume of a single photoevent is known, this can be divided into the image density volume to derive the number of photoevents. The variation of image density volume with exposure yields a gain curve, as shown in Fig. 2. This curve is linear (with a theoretical slope of 278 electrons per primary keV) except near the threshold voltage (corresponding to penetration of the CCD dead layer). With the TI CCD (Fig. 2(b)) at 16 kV, the threshold was about 5 kV and the “effective” potential 1 1 kV. The TI devices exhibited high “spurious charge” (similar to dark current but not eliminated by cooling), which caused high read-out noise and broadened the peaks in the pixel amplitude distributions. The RCA device, on the other hand, was much quieter when cooled, and showed more sharply peaked pixel amplitude distributions. The sharp single-event peak was superimposed on a broad background due to split events, i.e. events in which not all of the charge is deposited in one pixel. This is due to spreading of the photoelectron-produced charge before collection at the pixel electrodes, and, to some extent, trailing along the read-out direction due to charge transfer inefficiency. The charge spreading effect before collection is smaller in the case of the front-side bombarded TI chip, where nearly all the charge is produced in the depletion region close to the electrodes (and is less likely to diffuse into
186
G . R. CARRUTHERS ET A L
Accelerating voltage ( k V )
FIG.2. Integrated density volume versus accelerating voltage, for the Cerenkov light source spot image on an RCA CCD (top) using the setup shown in Fig. 1, and for a TI CCD (bottom) using a gas-discharge light source in a test setup at NRL.
neighbouring pixels). However, as mentioned, the pixel amplitude distributions were significantly degraded by spurious charge noise. Although it was not possible to determine accurately an average singleevent amplitude for the TI CCDs in the runs at Princeton, the spot-integrated density volumes were about the same as for the RCA CCDs. Therefore, since the gain curves were also similar, we conclude that the photoelectron detection efficiencies of both EBCCD types were comparable. The pixel amplitude histogram for the RCA CCD indicated a peak at a single-event amplitude of 175 DN when operating at 15 kV. As mentioned, this is not a true event pulse-height distribution. If the charges corresponding to individual photoelectron events are properly summed, the peak should
DEVELOPMENT OF EBCCD CAMERAS FOR THE FAR UV
187
FIG.3. Video display of data obtained in a low-light-level test of the RCA CCD, with 15 kV accelerating voltage. The full data frame is displayed at the top, and a small portion is shown with higher magnification at the bottom. Note that in the spot image, the density of events is too high to discriminate individual photoevents. Outside the spot, the background events show the effects of charge-transfer inefficiency in that the events are “trailed” in the direction of read-out.
become narrower and be shifted to a larger D N value, and the “continuum” of event amplitudes should be greatly reduced (Zucchino et al., 1981; Janesick et al., 1984). Unfortunately, as can be seen in Fig. 3, even our “low-light-level” images had too many events per unit area to allow true photon-counting. It is also apparent that charge-transfer inefficiency at the low operating temperature of - 80°C trails the events in the read-out (left-right) direction. Another difficulty was the apparent high background level in the areas of the image outside that of the spot. If we naively used the 175 D N peak derived from the low-light-level RCA CCD imagery as representative of the single-event pulse amplitude, we derive a count rate of slightly more than 1000 counts s - ‘ at high light level. On the
188
G . R . CARRUTHERS ET AL.
other hand, the chevron MCP recorded a somewhat higher rate, of about 1300 counts s-', when used a t the electron focus of the Schmidt camera. Hence, the detection efficiencies of the EBCCD sensors were inferred to be about 75%,at most, of the MCP detector in these runs. If, as expected, the true single-event pulse amplitude is higher than 175 DN, the discrepancy is worsened because the integrated charge measured would correspond to a smaller number of photoevents. EBCCD Tests at NRL
Additional measurements were made in a vacuum collimator facility at NRL in October 1984, using a Texas Instruments 328 x 490 pixel device at the electron focus of a Schmidt camera (see Fig. 4), with the TI-provided electronics described by Carruthers and Opal (1985). The collimator light source consisted of a discharge in a mixture of 10% 0 2 in He. A narrow-band Acton interference filter centred near I300 formed the window of the light source. Hence, the light output was nearly monochromatic 01 1304 A radiation. With the 25 cm aperture, f / 8 collimator, this yielded a circular diffuse spot for imaging by the camera having an angular diameter of 0.27". The intensity of the light source was monitored by a separately calibrated EMR 541J (semi-transparent KBr photocathode) photomultiplier. Prior to the EBCCD runs, the Schmidt camera was operated as a
FIG.4. EBCCD Schmidt camera detector under test at NRL. At the left is shown the camera assembly, which contains an opaque KBr photocathode; at the right is the CCD sensor head containing a Texas Instrument 328 x 490 pixel virtual-phase CCD.
DEVELOPMENT OF EBCCD CAMERAS FOR THE FAR UV
189
photodiode, using a collecting anode at the electron focus as input to a picoammeter. The light source was operated at a much higher intensity for this run than for the later EBCCD runs, but the photomultiplier count rate was still in the linear range. The collecting anode was then replaced by the EBCCD sensor head, and the camera was then operated with a much lower light level. Assuming linearity of the photomultiplier, and no change in the efficiency of the Schmidt camera and photocathode, it was possible to determine the number of photoelectrons per second incident on the CCD. The EBCCD signal could be read into our PDP-11/44 computer and Ramtek imaging processing system, or displayed in real time on an analogue oscilloscope display. When the Ramtek image processing system is used with the TI CCD camera system at NRL, an indirect calibration can be obtained based on the camera electronics specifications provided by TI. The output of the camera system is via a 12-bit analogue-to-digital converter; all bits set to 1 corresponds to a fullscale reading of our Ramtek image-processing system of D N = 4095. According to the TI specifications, this also corresponds to a single pixel signal of 512 000 electrons (i.e. 125 electrons per DN). The TI CCDs used had fullwell capacities in the range 530 000 to 740 000 electrons per pixel. As for the Princeton runs, pixel-amplitude histograms, spot image density volumes, and gain curves were generated. In a typical run, comparison of the spot-integrated density volume with the photomultiplier count rate indicated a value of 15.7 D N per photoelectron, if all photoelectrons are detected by the CCD. In contrast to the previous tests, however, we were able to detect a single-event peak in the TI CCD histogram, at about 20.8 DN. (Note that since different read-out electronics were used, this should not be compared with the 175 DN of the Princeton RCA CCD.) Dividing this value into the spot-integrated density volume indicates that about 75% of the expected photoevents were detected, in agreement with the results obtained at Princeton with the RCA CCD. Using the gain curve for the TI CCD (Fig. 2), assuming the theoretical slope of 278 electrons per primary keV, and using the nominal value of 125 signal electrons per DN, we expect 2.22 D N per primary keV or 24.4 D N per photoevent at 16 kV (1 I kV effective).This would indicate a detection efficiency of only 64%. As yet another test, the EBCCD analogue signal output was viewed directly on an oscilloscope, and the TI-quoted conversion factor of charge per pixel to analogue video signal level (500 000 electrons per pixel per volt) was used in conjunction with the EBCCD gain factor to yield the detected number of photoevents per second, integrated over the image. Comparison of this with the Ramtek-displayed density volume for an equal exposure time yielded a value of 16.4DN per photoelectron, close to the value derived from the photomultiplier calibration, assuming that all photoelectrons are detected.
190
G . R . CARRUTHERS E T A L .
However, uncertainties in the conversion factor and oscilloscope calibration could easily be as much as f30%. As discussed above in connection with the RCA CCD runs at Princeton, the positions of the “peaks” in the amplitude histograms are probably not accurate representations of the average single-event pixel amplitudes. Other likely sources of error, in the runs at Princeton and NRL, include (i) loss of photocathode sensitivity from one run to the next; (ii) in the runs at NRL, the relatively small size of the spot image did not provide a large enough number of counts per image when the light level was low enough to avoid overlapping of events; (iii) uncertainty of the conversion factor of signal electrons per volt in the Texas Instruments camera electronics; and (iv) difficulty of accurately integrating the total spot signal owing to relatively high and spatially-varying background event levels. In the runs a t Princeton, a chevron MCP was used as a comparison standard. Our previous measurements on MCP detectors indicated that the pulse-counting detection efficiencies of chevron MCPs with separate opaque photocathodes were nearly equal to the photoemissive quantum yields of the photocathodes: however, the uncertainties in the absolute values of the former are not inconsistent with the measurements of electron detection efficiencies of MCPs obtained by others (see, e.g. Fraser, 1983) which indicate a maximum photoelectron detection efficiency of the order of 80%. However, we have observed that MCPs operated in a poor vacuum or without sufficient prior degassing can exhibit spuriously high apparent count rates, apparently due to ion feedback. In the runs at Princeton, the pressures in the vacuum chamber were typically in the range 5-10 x Torr, which is higher than recommended for operation of chevron MCPs. Hence, the apparent detection efficiency of the Schmidt camera plus MCP may have been too high, resulting in an under-estimate of the EBCCD photoelectron detection efficiency. The spot image in the NRL EBCCD runs covered a diameter of about 26 pixels, and was not uniform in intensity (it was peaked at the centre with a nearly triangular profile). The Cerenkov source used at Princeton yielded a spot diameter of 75 TI-CCD pixels, but was also centrally peaked. It is possible that photoevent statistics at low light levels are degraded when using the Cerenkov source, owing to the fact that each energetic p-particle interacting in the source produces many UV photons almost simultaneously, rather than uniformly time distributed as with gas-discharge light sources. In addition to tests aimed specifically at determining overall detection efficiencies, we have operated EBCCD sensors in spectrographic mode at NRL, using the TI 328 x 490 pixel devices, with Schmidt and oblique focus detector configurations. Figure 5 shows spectra taken with the EBCCD Schmidt camera, using a 1200 line mm-’ objective grating to observe a collimated UV light source. Figure 6 shows an oblique focus sensor, which
0
5
0
1
0
0
X-PIXEL
1
5
0
NUMBER
m
o
a
o
P
o
~
0
5 0 l Q ) i S D m o a o m ~
X - PIXEL NUMBER
FIG.5. Spectra of helium and hydrogen discharges, in the wavelength range 1160-1370 A, obtained with the EBCCD camera in the laboratory using an objective grating. At the top is a video display representation, and at the bottom are computer-generatedintensity profiles along the spectra.
192
G . R . CARRUTHERS E T A L .
FIG.6 . Oblique-focus sensor, using Texas Instruments 328 x 490 pixel EBCCD, opaque CsI photocathode, and permanent-magnet focusing assembly. The sensor is used in a laboratory Rowland-configuration spectrograph with a 100 cm radius concave grating.
’
was placed on the Rowland circle of a 100 cm radius, 1200 line mm- concave grating. The resulting higher resolution spectrum of a hydrogen discharge is shown in Fig. 7. These laboratory spectra were used to develop imageprocessing and frame-addition/subtraction techniques for the NRL data processing system. DISCUSSION The measurements of overall detective quantum efficiency at both Princeton and NRL, using two types of CCDs in electron-bombarded mode, mostly indicate significantly lower detection efficiencies than the photoelectron yields of the optical systems plus photocathodes. This indicates that 30% or more of the primary photoelectrons are not effective in producing detectable events in the CCDs. This result was not in accord with our initial expectations. Although an MCP can occasionally fail to detect an energetic primary photoelectron owing to failure to produce secondary electrons, it was difficult to imagine a mechanism by which the energetic electron incident on a CCD could fail to produce at least a moderate number of electron-hole pairs in the silicon. However, very recently we became aware of the work by Daud et al. (1987) in which the quantum efficiencies of CCDs for electrons was measured using a
oBFoc9c ( w # I l m
-)I
LIEES 3w
em 351 SmXlMD (m€
Fms)
FIG.7. Spectrum of a hydrogen discharge obtained using the oblique-focusEBCCD sensor and Rowland spectrograph. The wavelength range covered is roughly 1220-1280 A. Also shown is a plot of Ramtek density number (vertical sum of two horizontal lines and smoothed horizontally) versus horizontal position.
194
G. R. CARRUTHERS E T A L .
scanning electron microscope. They showed that, at the 15-16 kV energy we typically used, about 25% of the primary electrons are backscattered from the CCD with an average of 50% of the primary energy, and that the overall primary electron energy loss efficiency is about 65% for untreated, back-side bombarded CCDs. This result is in good agreement with our measurements. In retrospect, this effect could have been anticipated from measurements of the backscattering of electrons from solid surfaces made by several previous researchers (although not in the context of electronic imaging), for example the work of Sternglass (1954). Although electrons backscattered from the CCD with less than their incident energy will eventually return to the CCD, they are not re-imaged by the magnetic field and most will land well outside the original impact area. Thus, these backscattered electrons probably contribute to the background of low-energy events outside the image area. Similarly, MCP detectors operated with electrons incident from a separate photocathode should also suffer detection inefficiency resulting from backscattering. However, the MCP face is not smooth and flat but is composed of adjacent channels, and the analysis of the effect of backscattering would be much more complex. For example, electrons incident in channels would have a much lower probability of being backscattered out of the channel, and failing to produce a count, than would electrons incident on the surface between channels. This, in combination with the fact that the MCP detector in our tests at Princeton was operated in a non-imaging mode (so that backscattered electrons returning to the MCP a second time but outside the image area could contribute to the detected signal) probably explains, at least in part, the higher count rate of the MCP versus the EBCCD. However, as mentioned, we were unable to achieve very high vacuum in the runs at Princeton, and the MCP count rates may have been spuriously high owing to ion feedback. CURRENT AND FUTURE DEVELOPMENTS We have modified our calibration facility at NRL to provide a capability of imaging a diffuse source of larger angular size (1" diameter) and more uniform illumination (Carruthers et al., 1987). It will also use a monochromator, allowing more accurate and detailed calibrations of EBCCD and other photon-counting detectors. We have now converted the TI camera electronics and data-handling system at NRL to operate with the TI 1024 x 1024pixel arrays (see Fig. 8), and tests of these arrays in the EBCCD Schmidt camera are in progress (see Figs. 9 and 10). The short-term emphasis of our measurements is to obtain repeat measurements of overall detective quantum efficiency and pulse-height
DEVELOPMENT OF EBCCD CAMERAS FOR THE FAR UV
195
FIG.8. EBCCD Schmidt camera sensor head with 1024 x 1024 pixel Texas Instruments virtualphase CCD installed.
distributions. These are not expected to be significantly different for the larger TI arrays currently in use versus the 328 x 490 pixel arrays used previously. In our most recent tests, we have discovered that noise generated in the read-out electronics is a major factor that limits our capability to determine true singlephotoevent pulse-height distributions (and probably also was in our previous measurements). Therefore, we are presently attempting to find and eliminate these sources of noise. In addition, we are developing improved dataprocessing software to allow more accurate and efficient determinations of pulse-height distributions and overall detection efficiency. As a parallel effort, we are developing a totally new camera electronics system, which is intended as a prototype of a flight-worthy system. It will be capable of a maximum read-out rate of 400000 pixelss-' with 16-bit digitization, versus a maximum of 100 kHz for the current TI electronics. It will also be adaptable to operation in true photon-counting mode. The system is patterned after one currently in operation a t the McDonald Observatory, with modifications to allow a higher read-out rate. A remote head assembly includes the CCD, thermoelectric cooler, pre-amplifier, and bias regulators. The main electronics assembly includes the clock driver, analogue signal processor (with high-speed Analogic ADAM-826- 1 AID converter), and a microsequencer based on an AM2910 microprocessor. It is being designed from the start for space-Right capability. Applications of the 1024x 1024 EBCCD include two proposed flight measurements. The first is a sounding rocket instrument for direct imagery of
196
G . R . CARRUTHERS E T A L .
FIG.9. Full-frame image (upper) and enlarged image (lower) of a test pattern light source (1; diameter) obtained with the Schmidt camera of Fig. 5 with a 1024 x 1024 pixel EBCCD.
star fields and galaxies in the 912-1350 A and 1350-2000 8, wavelength ranges. For this instrument, an oblique magnetic focus arrangement, similar to that used in the Princeton IMAPS instrument, would be used (see Fig. 11). The sensor is to be mounted at the prime focus of a 40 cm aperture, f/4 telescope mirror. Each pixel of the EBCCD would then correspond to a 2.5 arc-second square field, and the total array field of view would be 38 arc-
DEVELOPMENT OF EBCCD CAMERAS FOR THE FAR UV
197
FIG. 10. Image as in Fig. 9 (lower) without test pattern, operating at a lower light level and enhanced contrast in the display system, showing detection of individual photoevents by the EBCCD.
minutes square. Interchangeable (in flight) opaque photocathodes of KCl and CsI are used for the short- and long-wavelength ranges, respectively; also, an insertable BaFz filter is used with the CsI photocathode to limit the longerwavelength range. The unique capability of the rocket instrument, in comparison with other UV imaging instruments (such as the Goddard Space Flight Center’s UV Imaging Telescope on the Astro shuttle payload) is its short-wavelength imaging capability. The relative brightness of hot stars are much more temperature-sensitive in the 912-1 350 A range than at longer wavelengths. Therefore, the proposed two-colour imaging system will provide improved temperature measurements, and capability for detecting very hot but faint objects, in comparison with imagery only in longer-wavelength bands. The second proposed application is of an EBCCD Schmidt camera for imagery of the terrestrial night air-glow from a high altitude (preferably in geosynchronous orbit) satellite. Nightglow emissions of atomic oxygen at 1304 A and 1356 A wavelengths are produced in the recombination of ionospheric O+ with electrons. The intensity of the recombination nightglow is proportional to the square of the O+ density in the F2 region of the ionosphere, where O+ is the dominant ion. Therefore, measurements of the nightglow intensity (and its spatial distribution) provide a means for remote sensing of the night ionosphere. Imagery of the O + nightglow emissions was obtained with NRL’s Far UV
FIG.1I. Diagram (at top) and photograph (at bottom) of an oblique-magnet-focusEBCCD sensor, planned for use in NRL’s Far Ultraviolet Imaging Telescope rocket instrument. In the photograph, the sensor is shown with a phosphor screen at the electron image plane, for preliminary tests of the electron optics.
DEVELOPMENT OF EBCCD CAMERAS FOR THE FAR UV
199
Camera on the Apollo 16 mission in April, 1972 (Carruthers et al., 1987) using an electronographic Schmidt camera. The proposed instrument would replace the electronographic film with a 1024 x 1024 pixel EBCCD, allowing remote read-out of the images, and would be ten times closer to Earth, yielding ten times better resolution with the same optical system. The proposed Global Imaging Monitor of the Ionosphere (GIMI) will cover a field of view 13.75” square (8600 km square at Earth from geosynchronous orbit altitude) with 3 arc-minutes (30 km) resolution. CONCLUSIONS
We have operated electron-bombarded CCD arrays in Schmidt imaging systems with opaque KBr photocathodes. These were compared with a chevron MCP detector operated in pulse-counting mode, or a collecting electrode for direct photocurrent measurement, placed interchangeably at the Schmidt camera electron focus, with simultaneous light-source monitoring by calibrated photomultipliers. The results indicate that possibly only 70-75% of the expected photoelectron events are detected by the CCD arrays. This is in agreement with recent JPL results in which the sensitivities of CCDs to electrons were measured with a scanning electron microscope. However, further measurements with improved test facilities and data-reduction software are required to verify these results, and to improve the accuracies of our detection efficiency measurements. This work is now in progress, using 1024 x 1024 pixel TI arrays in place of the previously used 328 x 490 pixel devices. We are also developing an improved camera electronics system which is the prototype of a version planned for sounding rocket and satellite applications of the 1024 x 1024 pixel EBCCD sensors. ACKNOWLEDGEMENTS This work was supported by NASA’s Space Astronomy Ultraviolet Detector Development Program and Planetary Instrument Definition Program, and by the Office of Naval Research. We thank our co-workers, in particular Paul Zucchino, John Opperman, and Michael Reale at Princeton,and Stephen Andersen and Timothy Seeley at NRL, for their assistance; and personnel at RCA and Texas Instruments for useful discussions.
REFERENCES Carruthers, G. R. (1969). Appl. Opr. 8,633 Carruthers, G. R. and Page, T. (1976). J . Geophys Res. 81,483 Carruthers, G. R. and Opal, C. B. (1985). In “Adv. E.E.P.” Vol. 64B, pp. 299-314 Carruthers, G . R., McCoy, R. P. and Opal, C. B. (1987). In “Proceedings of the Eighth Workshop
200
G . R. CARRUTHERS ET A L .
on the Vacuum Ultraviolet Radiometric Calibration of Space Experiments. Joint Institute for Laboratory Astrophysics” (in press) Daud, T., Janesick, J., Evans, K. and Elliott, T. (1987) (in press) Duckett, S. W. and Metzger, P. H. (1965). Phys. Rev. 137, A953 Fraser, G . W. (1983). Nucl. Instrum. Methods 206,445 Janesick, J. R., Elliott, T., Collins, S., Marsh, H., Blouke, M. M. and Freeman, J. (1984). Proc S.P.I.E.501, 1-31 Lowrance, J. L. (1979). Proc. S.P.I.E., 172,232-238 Lowrance, J. L.,Zucchino, P., Renda, G. and Long, D. (1979). In “Adv. E.E.P.”Vol. 52,pp. 441452 Lowrance, J. L. and Carruthers, G. R. (1981). Proc. S.P.I.E. 279, 123-128 Metzger, P. H. (1965). J . Phys. Chem. Solids 26, 1879 Siegmund, 0.H. W., Everman, E., Vallerga, J. V., Labov, S., Bixler, J. and Lampton, M. (1986). Proc. S.P.I.E. 687, 117 Sternglass, E. J. (1954). Phys. Rev. 95, 345 Zucchino, P., Long, D., Lowrance, J. L., Renda, G., Crawshaw, D. D. and Battson, D. F. (1981). Proc. S.P.I.E. 290, 174
Recent Developments in Solid-state Arrays for Infrared Astronomy I. S. McLEAN Royal Observatory, Edinburgh, U K Infrared Telescope. Hawaii, USA
INTRODUCTION One of the most exciting developments in photoelectronics in the past five to six years has been the rapid improvement of solid-state imaging devices for low-light-level astronomical applications at infrared wavelengths. Of course, we are all familiar with the impact made by silicon charge-coupled devices on a wide range of topics including astronomy. Indeed, astronomical imaging was one of the early driving disciplines. The availability of silicon CCDs, with useful sensitivity over the wavelength interval from 300 to 1100 nm greatly extended the traditional “optical” window for astronomers. An obvious and astrophysically important way to extend this kind of semiconductor technology was into the infrared. Unfortunately for infrared astronomers, this was easier said than done. Apart from difficulties of manufacture associated with the use of less well-developed low band-gap semiconductors with intrinsic IR sensitivity, the important military uses of solid-state infrared imaging devices inevitably resulted in the technology becoming classified, thereby delaying the development of devices suitable for scientific and other applications. Recently, however, small-format IR arrays with very respectable sensitivities have started to appear on the commercial market, with astronomy once more providing one of the stimuli. There is little doubt that the availability of these arrays is causing a revolution in infrared techniques and sending repercussions through the entire subject of astronomy. In this paper I will describe a few of the most significant devices. More detailed discussions can be found in McCreight (1986) and in the proceedings of the Hilo Workshop of Ground-Based Astronomical Observations with Infrared Array Detectors edited by Wynn-Williams and Becklin (1987). DETECTOR TECHNOLOGY A variety of semiconductor materials, read-out schemes, formats and performance specifications can be found in the literature. The basic photo20 1 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copynght 0 1988 Academic Press Limiled All rights of reproduction in any form reserved ISBN 0-12-014674-6
202
I. S. MCLEAN
TABLE I
Cross-section of current array technology Material
HgCdTe
InSb
Si:In
Si:Ga
Si:Bi
Si:As
Wavelength (pm) 2.5 PV Type Mux CCD Format 64x64 Source RI
5.1 PV DRO
7.5 PC CCD
17.5 PC DRO
18.7 PC CID
24.0
IBC SWIFET
16x16
10x50
62x58 32x32 62x58
SBRC
RI
SBRC AEROJET
RI
detection process is almost always electron-hole pair production in an “intrinsic” photovoltaic semiconductor such as indium antimonide (InSb) or cadmium mercury telluride (HgCdTe), or in an “extrinsic” photoconductor of doped silicon. Schottky-barrier diodes have also been used to make fairly large-format arrays but with much lower quantum efficiencies than either PV or PC detectors. More recently, blocked-impurity-band (BIB) detectors have been employed for the arrays sensitive to the longer IR wavelengths. Among the selection of dopants used with silicon are In, Ga, Bi, As and Sb. Infrared arrays are usually “hybridized” to silicon read-out structures. These may be CCDs (most commonly surface channel), CIDs or arrays of switched MOSFETs. Most, but not all, manufacturers use some kind of “bump-bonding’’ to electrically interconnect the array of IR sensors to the silicon multiplexer. Formats range from 16 x 16 to 128 x 128 at present, and there are working examples of all of the above array types, covering the entire infrared spectrum from 1 to 30 pm. Table I lists some of the more successful devices used in ground-based astronomy applications. Before considering the front-runners in the field it is worthwhile to review the detector properties required for astronomy.
IR ARRAYPERFORMANCE REQUIREMENTS In infrared astronomy the photo-signal from the astrophysical source is generally much less than the signal from the warm telescope and other optics together with radiation from the terrestrial atmosphere above the telescope. As the spectral bandwidth is reduced, device dark current and multiplexer read-out noise both enter the signal-to-noise calculation. When the detector dark current is very much less than the photocurrent due to the background radiation, and when the square of the read-out noise is very much less than the accumulated background photocharge (and hence much less than the full-well capacity of a pixel), the detector is background-limited.
RECENT DEVELOPMENTS IN SOLID-STATE ARRAYS
203
For very low backgrounds this condition requires the ability to integrate ONCHIP for as long as necessary without degradation, implies the need for low dark currents and low read-out noise, and demands the highest possible quantum efficiency. Since the background signal is a very strong function of wavelength, it may be difficult to achieve these conditions over a wide interval of wavelength. There has been a tendency to split the infrared domain into three regions, namely, 1 to 2.5 pm, 2.5 to 5.0 pm and 5 to 30 pm, and to use different semiconductor materials in each regime. For the 1 to 2.5 pm interval the current choice is CMT, for 2.5 to 5.0 pm the best choice is InSb and this important material is also very sensitive in the 1 to 2.5 pm region also. Extrinsic silicon is used for longer wavelengths with most of the recent attention going to gallium-doped silicon for response out to 17.8 pm. Table I1 TABLE I1 Typical performance figures for currently available arrays InSb
Si:Ga
64x64
64x64
62x58
62x58
RI
HE0
SBRC
SBRC
2.5 70”h 800-1000 2x106 10-50
2.5 70% 1000 1x106 50
5. I 70% 400-500 1x106
17.5 50% 500 3x106 1 x 104
HgCdTe HgCdTe
Cut-off ( p n ) QE Read noise (e-) Capacity (e-) Dark current (e- s-I)
100
lists the published array performance figures for the most important devices in use at present. Because of its excellent noise performance, high QE and low dark current, the front-runner at present in the 1 to 5 pm window is the InSb/DRO array from SBRC. This detector array has recently been installed in a camera for use on the UK Infrared Telescope in Hawaii. Indeed, a major stimulus for SRBC was UKIRT’s commitment to this project; the first devices were bought by UKIRT and the first astronomical results were obtained on Mauna Kea. The new infrared camera system, called IRCAM, was developed at the Royal Observatory, Edinburgh and went into operation October 1986. IRCAM became a “common-user” facility instrument for UKIRT on 1 September 1987. The remainder of this paper is devoted to presenting some results from this instrument to illustrate the immense potential for IR arrays in astronomy.
COMPUTER ROOM
1,
DOME
II
Serial commands
CONTROL ROOM
Array C o n t r o l Unit IACUI
P i 0 c e s so I O E C LSI 1 1 / 2 5
I
DRY 11-1
secondary mirror
control
:Sync
m i r r o r position
-5Hz
had paiallrl port Seqwnrer
~I I-
Array D a t a
erial commands
I [oll,matar
Array Electronic
FP e t a l o n IRCAn 0 tlCS
Array drive drive Array signals and bias voltages R V 11-8 DMA
-
Analogue llbit
I
p-$$q
D R V 11-8 011A notor c o n t r o l U"lt
IEEE 488 BUS
I
I '
Pream
QrOCeSSlng
D i f f e r e n t i a l array d a t a
I
'Uiuohk'
ells
l n r t r u m e n t n o u n t i n g Platforr IltlPi
notoidrlve I
FIG.1 . A block diagram of the UKIRT Infrared Camera, IRCAM. The instrument is controlled from the UKIRT Vax computer running the ADAM software environment. Two LSI 11 processors are used in the system; one of these is the control processor for the sequencer, drive electronics and A/D unit and the other is a data pre-processor which is used to buffer the Vax and allow exposures to be co-added in memory.
RECENT DEVELOPMENTS IN SOLID-STATE ARRAYS
205
THE UKIRT INFRARED CAMERA IRCAM has been described in detail elsewhere (by McLean ef al., 1986). Briefly, it is a general-purpose imaging system comprising the following components, A LHe/LN2cryostat with cold filters and a cold re-imaging lens; the cryostat can accept a collimated beam from external optics. The camera operates in a way very similar to an optical CCD system and so the drive electronics, sequencer, analogue signal processing chain and A/D unit are a familiar feature of the system. A system block diagram is shown in Fig. 1 and Fig. 2 shows how the instrument is attached to the Cassegrain focus of UKIRT. There are basically three data-acquisition modes called STARE, CHOP and SNAPSHOT. IRCAM can be configured to operate as a direct imager, an imaging polarimeter or an imaging spectrometer using a FabryPerot etalon. One of the main features of this instrument is its excellent software package for image display and analysis, and its easy-to-use interface to the astronomer.
FIG.2. The I to 5 pm infrared camera IRCAM, attached to the 3.8 m UK Infrared Telescope on Mauna Kea, Hawaii, The LHe/LN2 cryostat containing the 62 x 58 InSb array is mounted in a side-looking configuration on an optical bench attached to the north port of the UKIRT Instrument Support Unit. Drive electronics are located close to the camera on the underside of the mounting platform and on the mirror cell.
206
I. S. MCLEAN
RESULTSWITH
A N INFRARED CAMERA
One of the most important aspects of infrared astronomy is the study of star formation. Figure 3 is a 2.2 pm image of the well-known star-formation region in Orion. The brightest object is called BN and it is totally invisible at optical wavelengths owing to the extremely strong extinction of optical light caused by the large quantities of dust in such regions. The exposure time was just 30 s, and the image reveals faintly for the first time the extreme infrared source called IRc2 which is believed to be a major power source driving giant outflows of gas. Extragalactic astronomy also benefits greatly from the new IR array technology. To study a whole galaxy by earlier IR techniques required a tedious and crude process of point-by-point mapping with a single-element photometer. Figure 4 shows a 5 min exposure of a distant galaxy (NGC660) at the excellent resolution of 0.6 arc-sec per pixel. Infrared observations at 2.2 pm penetrate the dense dust clouds in the spiral arms of this galaxy and enables astronomers to “see” the true morphology of the stellar distribution.
FIG.3. A 30 s exposure at a wavelength of 2.2 pm of the BN region of star formation in the Orion Nebula. The passband is 0.4 pm, the pixel scale is 0.6 arc-sec per pixel and the field-of-view is 37 x 35 arc-sec. Just to the lower left of the bright, optically invisible source known as BN lies a faintly visible source called IRc2 over 1500 times fainter than BN.
RECENT DEVELOPMENTS IN SOLID-STATE ARRAYS
207
FIG.4. A 5 min exposure of a distant spiral galaxy showing complex optical structure. In the IRCAM image at 2.2 pm complexity caused by obscuration due to dust lanes is greatly reduced.
In conclusion, it is worth pointing out that IR array technology for scientific applications is probably at about the same level of development as optical CCDs were about 10 years ago. Provided that we can maintain the interest of the detector manufacturers, there is an excellent decade ahead. ACKNOWLEDGEMENTS It is a pleasure to acknowledge and thank all of my colleagues on the IRCAM Project Team and all those at ROE and UKIRT for their support.
REFERENCES McCreight, C. R. (1986) In “Proc. of the Second Infrared Detector Technology Workshop,” NASA Technical Memorandum 88213 McLean, I. S., Chuter, T. C., McCaughrean, M. J. and Rayner, J. T. (1986). Proc. S.P.I.E.627, 430-437 Wynn-Williams, C. G. and Becklin, E. E. (1987). “Ground-based Astronomical Observations with Infrared Array Detectors.” University of Hawaii
This Page Intentionally Left Blank
Multiple-frame UV/X-Ray Picosecond Framing Camera R. T. EAGLES, W. SIBBETT, W. E. SLEAT, D. R. WALKER University of St. Andrews, Department of Physics. St Andrews, Fqe, Scotland
J. M. ALLISON and N. J. FREEMAN A WRE Aldermaston, Reading, England
INTRODUCTION The requirement for a sensitive two-dimensional imaging diagnostic with picosecond time resolution, particularly in the study of laser-produced plasmas, has already been discussed by Baggs et al. (1985). A temporal sequence of framed images would provide usefyl supplementary information to that provided by time-resolved streak images across a spectral region of interest from visible to X-ray. To fulfil this requirement the Picoframe camera system has been developed. Results pertaining to the operation of a camera having S 20 photocathode sensitivity are reviewed and the characteristics of a UV/X-ray sensitive version of the Picoframe system are presented. THE PICOFRAME CAMERA The Picoframe I camera system illustrated in Fig. 1 is based around an electron-optical image-converter tube that has been described in detail elsewhere by Sibbett et al. (1982). Briefly, in operation the beam of imageforming photoelectrons is swept rapidly across a small aperture by applying a time-varying linear ramp to a pair of parallel-plate deflectors, the framing deflectors. This provides a short exposure and when a linear voltage ramp of the opposite sense is applied to the compensating deflectors the swept motion of the electrons is cancelled. If precise compensation is achieved, a stationary image is displayed on the phosphor screen of the image tube. To ensure adequate light gain in dynamic operation and to limit detrimental spacecharge defocusing effects due to high photocurrents, a microchannel plate image intensifier (Mullard type XXl330A) is fibre-optically coupled to the output phosphor. This is gated on for 200 p s to allow intimately coupled photographic film to be used for image recording without undue accumulation of integrated noise. 209 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
R. T. EAGLES ET A L .
210 FRAMING CAMERA
FRAMING IMAGE
INPUT OBJECT APERTURE
PHOTOGRAPHIC
MICROCHANNEL PLATE INTENSIFIER
FIG. 1. Schematic of the Picoframe camera with deflection waveforms for single-frame operation.
Evaluation of the S - 2 0 Picoframe I camera has indicated that high-quality framed images having a spatial resolution of 8 lp mm-' a t the photocathode, with exposure durations between 100 ps and 200ps (FWHM) can be recorded. The experimental camera has been operated in both single-frame and double-frame modes and comparable spatial and temporal resolution has been observed. For two sequential images, interframe times of 1.2 ns have been achieved (Eagles et al., 1986). Ultra-violet Imaging
To facilitate greater flexibility in future framing camera development and specifically for imaging studies on X-ray luminous events, a vacuum demountable Picoframe variant has been fabricated at AWRE Aldermaston. Initial dynamic testing has been carried out using the frequency-quadrupled radiation from a mode-locked, Q-switched, ND :YAG oscillator and single amplification stage. Ultra-violet (266 nm) test pulses of 50 ps (FWHM) duration have been used to excite a 10nm thick gold photocathode. To evaluate the dynamic spatial resolution characteristics, the gold photoemissive layer was evaporated onto a chromium negative USAF resolution test target, fabricated on a fused silica substrate. The experimental image tube has an electron-optical magnification of - 1.4 with a static limiting resolution of 20 Ip mm- at the photocathode. In combination with the image intensifier, the camera system exhibited a static resolution of 14 Ip mm-' at the input photocathode. As already stated, the single-frame operation of the camera is achieved by
'
MULTIPLE-FRAME UV/X-RAY PICOSECOND FRAMING CAMERA
2 11
application of time-varying linear voltage gradients to the deflectors. These can be generated by laser illuminated semiconductor devices (Margulis and Sibbett, 1984), or more conveniently by a stack of avalanche transistors. A circuit incorporating two synchronized transistor chains has been developed at AWRE (Jackson et al., 1986). Each circuit provides complementary outputs of plus and minus 3 kV in 1.5 ns into a 50 R resistive load. In framing operation, two circuits are employed such that the framing and compensating deflectors are driven independently. An adjustable transmission line delay unit is provided to alter the trigger signal delay to the second circuit, thus allowing the relative phasing of the deflectors to be varied. Differences in the deflector responses can be nullified by control of the applied voltage gradient to yield spatially resolved images. This is achieved by incorporating the deflector plate capacitance into a capacitative voltage divider network. Single framed images of the USAF test chart, occupying a 6 mm x 6 mm cathode area were recorded on Ilford HP5 film processed for a speed rating of 1600 ASA. A typical recording is reproduced in Fig. 2(a). Visual inspection of the recorded framed image reveals that element 1 of group 3 can be resolved by eye, which corresponds to an input spatial frequency of 8 Ip mm-I. A microdensitometer trace across the vertical bars of this element is presented in Fig. 2(b). Owing to the non-linear film response, a density step-wedge was recorded onto the same film before processing. This enabled the film response to be characterized such that the density axis of the line scan could be related to intensity. The intensity baseline was set arbitrarily to the intensity equivalent fog level of the film. Depth of modulation measurements could then be taken directly from the line scan using Modulation
=
rmax -1min Imax
+ Imin
For an input spatial frequency of 8 lp mm-1 at the input photocathode, a modulation depth -20% is observed. The frame open time has been determined using 50 ps (FWHM) laser pulses to probe in time the instrumental transmission function of the camera. A modified Tektronix sampling arrangement (Eckart et al., 1986), successively shifts the camera trigger time relative to the incident optical probe pulse. At each instant the intensity level on the screen is monitored by a contacted photodiode. The transmission profile of the camera is then displayed as an intensity-versus-time trace on an oscilloscope. To reveal the true frame duration the 50 ps (FWHM) laser pulse duration must be deconvolved from this profile. Using this scheme, the single-frame duration was determined to be 100 ps (FWHM). In the prototype Picoframe I with single framing aperture and no shift
POSITION (mml
FIG. 2. (a) Reproduction of single-frame image produced by the UV-sensitive Picoframe camera, exposure time is 100 ps (FWHM); (6) horizontal line scan across vertical bars of group 3, element 1, input spatial frequency is 8 Ip mm-I.
MULTIPLE-FRAME UV/X-RAY PICOSECOND FRAMING CAMERA
2 13
deflectors, the generation of two time-sequenced frames is effected by the application of triangular-shaped voltage waveforms to the deflectors. Shuttering action is achieved once on the rising voltage gradient and again on the falling voltage gradient. A dephased opposite-polarity compensating waveform applied to the compensating deflectors results in two spatially distinct images at the screen (Eagles et al., 1986). Appropriate triangular waveforms can be generated by propagating a ramp voltage profile into a pulse-forming network. The network, consisting of a pair of equal-length shorted stubs placed on a transmission line, has been described in detail elsewhere (Margulis and Persson, 1985). If the round-trip time in each stub is less than or comparable to the signal risetime, a triangularshaped waveform of half or less of the input ramp amplitude will be available at the output. With stub lengths between 10 cm and 20 cm, triangular voltage profiles with FWHM durations of between 1 ns and 2 ns have been generated. A Michelson type optical delay configuration was arranged to provide a pair of variable-separation subpulses from each laser pulse. With this scheme, the temporal separation or interframe time of the two images could be determined. A reproduction of a double-framed image demonstrating a dynamic limiting spatial resolution of 8 lp mm-' at the input photocathode is shown in Fig. 3. The exposure duration was measured to be 250 ps (FWHM) for each frame, with an interframe time of 2 ns. The relatively long open time
FIG.3. Double-frame format produced by UV-sensitive Picoframe camera. Exposure time is 250 ps (FWHM) with interframe time of 2 ns. (First image-right; second image-left.)
214
R . T. EAGLES ET A L .
is due mainly to the voltage gradient of the triangular deflection waveform being half of that on the input waveform. Additionally, the moderately high capacitative load presented by the pulse-forming network slows the risetime of the sweep circuit. X-ray Imaging
The Picoframe camera system has been designed specifically to cover the visible to soft X-ray spectral sensitivity range, through appropriate choice of photocathode. Preliminary characterization of the camera imaging performance when viewing light from an X-ray luminous event has been carried out in the single-frame mode. The dynamic camera resolution was first optimized during the ultraviolet test pulses from the frequency-quadrupled Nd :YAG laser system. The gold photocathode on fused silica substrate was then replaced by a 9 mm diameter, 10nm thick gold photocathode on a 25pm beryllium foil substrate. A moderately coarse copper wire mesh was overlaid behind the photocathode to provide a resolution test pattern. This consisted of 150 pm diameter copper wires arranged in a square grid with a period of 450 pm. Picosecond X-ray bursts were generated from a laser-produced plasma. Frequency-doubled (532 nm) pulses from an amplified Nd: phosphate glass laser system delivering approximately 4 J in 100 ps (FWHM) were used to irradiate foil targets. The beam was brought to a 30 pm diameter focus onto 25 pm thick gold targets at the centre of a 1.0 m diameter evacuated target chamber. The camera was secured to an access port at a small angle from normal to eliminate direct X-ray stimulation of the phosphor. The sensitivity of the camera was such that it was necessary to incorporate a 20pm aluminium filter to attenuate the incident X-ray flux to a level at which wellexposed images were obtained for moderate intensifier gains. Single framed images were recorded, an example of which is reproduced in Fig. 4(a). A
(a) FIG.4. (a) Frame X-ray image of overiayed copper mesh, exposure time 120 ps (FWHM).
MULTIPLE-FRAME UV/X-RAY PICOSECOND FRAMING CAMERA
21 5
POSITION h n l
(b) FIG.4. (b) Horizontal line scan across image diameter.
horizontal line scan across the diameter of the image is presented in Fig. 4(b). The frame duration was measured to be 120 ps (FWHM) using the frequencyquadrupled Nd: YAG facility.
ANALYSIS The unequal bar-space widths of the wire grid used as a test object in evaluating the X-ray sensitive Picoframe made direct comparison to the results obtained with the equispaced bar pattern of the USAF chart rather difficult. To determine the relative instrumental resolutions for these cases, a basic computer model has been developed at AWRE. The code assumes a Gaussian instrumental response to a delta function input where the full width at half maximum characterizes the spatial resolution. A series of Gaussian response functions of different widths are convolved with a square-wave input function which is defined to correspond to the experimental conditions. For each simulated output response the modulation (or visibility) is evaluated and plotted as a function of the instrumental resolution (Gaussian line width). A line scan of the experimentally recorded output is used to determine the modulation depth for a specific input spatial frequency. Comparison of the experimentally determined modulation to the computer-generated curve yields an estimate of the instrumental resolution. Since the code can accept any periodic square-wave input function, not necessarily of equal spacing,
216
R. T. EAGLES ET AL
direct comparison may be made between the UV and X-ray spatial response characteristics. The results obtained from the UV-sensitive Picoframe I demonstrate a dynamic spatial resolution of 8 Ip mm-' with 20% modulation. This indicates an instrumental resolution of 90 pm at the input photocathode. Analysis of the grid image obtained from the X-ray sensitive Picoframe I is complicated owing to the inconsistent image quality across the field of view. It is noted that some areas are more distinct than others. This is attributed to irregularities and contamination on the beryllium substrate of the photocathode. When line scans across the image are analysed, modulation depths ranging from 40% to 85% are observed. Using the computer code this implies full width at halfmaximum Gaussian response functions of 300pm to 120pm which is somewhat lower than that observed under UV illumination conditions. However, it is suspected that photocathode defects accounted for a significant degradation in the image modulation and that improved X-ray photocathode fabrication techniques will lead to enhanced instrumental functions.
CONCLUSION
A provisional framing capability throughout the spectral range from visible to soft X-ray has been demonstrated by the Picoframe I camera system. Some degree of image degradation under X-ray illumination might be anticipated from the broader spread in secondary electron emission energies from the photocathode, but our data imply that potentially good spatial resolution can be achieved in this regime. Future work will include the characterization of the Picoframe I in doubleframe mode under X-ray conditions. Alternative multiple-image schemes are also under preliminary evaluation and a demountable twin-aperture Picoframe I1 camera has been fabricated. This latter design permits the generation of two sequenced frames from a single sweep, and by applying an appropriately timed staircase waveform to a pair of orthogonally orientated shift plates, four (six, eight, etc.) sequential framed images can be obtained. ACKNOWLEDGEMENTS The financial support of the Paul Instrument Fund of The Royal Society and the Science and Engineering Research Council is gratefully acknowledged. One of us (R.T.E.) is supported by AWRE, Aldermaston and another (D.R.W.) is supported by an SERC CASE Studentship in collaboration with AWRE. Particularthanks are due to Mr M. C. Jackson and Mr R. D. Long for their assistance with the electronic equipment and to Mr S. D. Rothman for his work on developing the computer program. We would also wish to thank the staff of the MERLIN laser facility. This work has been carried out with the support of Procurement Executive, Ministry of Defence.
MULTIPLE-FRAME UV/X-RAY PICOSECOND FRAMING CAMERA
2 17
REFERENCES Baggs, M. R., Eagles, R. T., Margulis, W., Sibbett, W. and Sleat, W. E. (1985). In “Adv. E.E.P.” Vol. 64B, 621-636 Eagles, R. T., Sibbett, W. and Sleat, W. E. (1986). Opf. Commun. 57,423 Eckart, M. J., Hanks, R. L., Kilkenny, J. D., Pasha, R., Wiedwald, J. D. and Hares, J. D. (1986). Rev. Sci. Instrum. 57, 2046 Jackson, M. C., Long, R. D., Lee, D. and Freeman, N. J. (1986). Laser andParticie Eeams4,145 Margulis, W. and Persson, R. (1984). Rev. Sci. Instrum. 56, 1586 Margulis, W. and Sibbett, W. (1984). Opr Commun. 51,91 Sibbett, W., Baggs, M. R. and Niu, H. (1982). Pror. S.P.I.E. 348,267-270
This Page Intentionally Left Blank
Evaluation of PVOOl and P-100 Tubes for Multiple-channel Streak Cameras S. MAJUMDAR, P. Y. KEY DeNi Delti Ltd, London, England
M. YA SCHELEV, Y. SURDYUCHENKO General Physics Institute, Moscow, USSR
W. SEKA, M. C. RICHARDSON, P. YAANIMAGI and R. KECK Laboratory for Laser Energetics, University of Rochester. U S A
ABSTRACT Streak cameras are known to suffer from intensity-dependent non-linearity of response. The exact mechanism is still not fully understood, although it is clear that both vacuum space-charge and internal bottlenecking in the streak tube photocathode play an important part in the process. Recently, there have been demands for the use of a streak camera to record in parallel as many separate channels of information as possible; preferably up to 100 completely independent channels. As this type of operation will need a far greater loading of the photocathode of the streak camera, it is important that the streak tube and intensifiers are optimized to operate in the most sensitive mode and with optimum signal to noise ratio. In the present paper data were presented from a Dellistrique DS-3 camera, which was used with two different streak tubes: one made in the UK, the Picotron P-100, and one made in the USSR, the PV-001. In both cases, the camera had single-photoelectron detection capability. The P-100-based system was used to record single photoelectrons in a 12channel operation and showed a channel dynamic range of 250, with no channel cross-talk. As only half of the photocathode was exposed for this experiment, it is expected that 25 completely independent channels (cross-talk less than 0.l'X) should be resolvable in the system based on the P-100 tube. The PV-001 tube has a smaller photocathode, and its phosphor was different from the P-100. The PV-001-based DS-3 streak camera showed that a comparable channel density on the photocathode was available from this system, but the overall number of channels would be limited to around 15 owing to the smaller photocathode. 219 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
This Page Intentionally Left Blank
Evaluation of a Photon-counting Streak Camera with CCD Recording S. MAJUMDAR, P. Y. KEY Delli Delti Ltd, London, England
V. PLATONOV General Physics Institute. Moscow, USSR
and A. RIDGLEY Rutherford Appleton Laboratory, Chilron, England
ABSTRACT Streak cameras with CCD recording have definite advantages over those with film recording. The CCD system is directly amenable to computer analysis of the streak camera results and offers better quantitative data reduction than does film. The poor calibration of streak cameras has so far excluded them from many quantitative absorption spectroscopic studies and limited their use to relative measurements, even though they are capable of quantitative measurements. The calibration results from a Dellistrique DS-3 streak camera and a liquidnitrogen-cooled CCD camera combination were presented. The system detected photoelectrons from the streak tube, which was fibre-optically coupled to one- or two-channel plate intensifiers. The characteristics of the CCD camera, the channel plate intensifier and CCD camera combination, and the streak camera plus CCD camera combination were separately evaluated. The system was then compared with Ilford HP5 film for recording of the data. The system dynamic range was limited by the noisy channel plate intensifier to around 500, while the CCD camera and the streak tube were capable of operating with a dynamic range of 10 000 and 1000 respectively, when used in the streak mode. The system was calibrated to work as a single-channel photometer with better than 10 picosecond accuracy and with single-photoelectron counting ability. The usefulness of such a system for picosecond absorption and emission spectroscopy was discussed. 22 1 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
This Page Intentionally Left Blank
A New Method for Observing High-speed Luminous Phenomena Y . KIUCHI, N. KOSHIDA and T. SAKUSABE Department of Elecrronic Engineering, Faculty of Technology, Tokyo University of Agriculture and Technology, Tokyo, Japan
INTRODUCTION With advances in research on ultra-short laser pulses and their applications to various scientific experiments, the requirement for means of observing high-speed luminous phenomena have increased. The image-tube streak camera has been extensively investigated and may be one of the best systems to satisfy such requirements, especially since it is now able to detect subpicosecond light pulses (Bradley and Sibbett, 1975; Baggs et al., 1985; Takiguchi et al., 1984; Kinoshita et al., 1984). A new technique for observing high-speed luminous events (Nozaki et al., 1985) is described in this paper, although its present performance is no match for the streak camera. The fundamental idea and some results of preliminary experiments are reported here.
FUNDAMENTAL CONSIDERATIONS Photoelectron Train For simplicity, let us suppose that photoelectrons produced by a light pulse are travelling through the equipotential region due to a drift electrode without interaction due to collisions and/or space-charge effects. These assumptions require that the photoelectrons are travelling along paraxial trajectories and have a very low current density. Because the duration of the light pulse is extremely short, the photoelectrons are in a group and the length of this photoelectron “train” 1 is
I=vT, =[$(Vd-VP)]l”
T,,
where e is the electron charge, m is the electron mass, V, and Vpare the drift electrode and the photocathode voltages respectively, v is the velocity of the 223 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copynght 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
224
Y. KIUCHI N. KOSHIDA AND T. SAKUSABE
photoelectron and TI is the time width of the light pulse. If we adopt values of 1000 V for V d - V,,and 1 ps for T,,the length of the photoelectron train is only 19 pm. It would seem very difficult to deflect such a short photoelectron train by generating high deflection voltages in the picosecond domain, although this is, in fact, what is achieved in the streak tube. STRETCHING THE
TRAIN
If the photocathode voltage can be rapidly raised during the light pulse, the velocity of each photoelectron in the drift electrode region can be varied depending on its time of emission from the photocathode. The first photoelectron of the train moves faster and the later ones run slower so that, after passing through the drift electrode region, the train is stretched out as illustrated in Fig. 1. It is also possible to stretch the train by reducing the drift electrode voltage rapidly. However, the application of a quickly falling voltage to the drift electrode is made very difficult by its high capacitance. PHOTOCATHODE
I iNC I DENT
D l i l F T ELECTRODE
I
PH3TOELECTOIIS
ooooLIGHT PULSE
COLLECTOR
O ~ O - t O + O
osc I LLO-
?
SCOPE
PHOTOELECTRON T i i k l N
FIG.1. Principle of operation of the tube.
CONSTRUCTION AND OPERATION OF
THE
TUBE
The tube consists of a photocathode, an accelerator electrode (the first mesh), a decelerator electrode (the second mesh), a long cylindrical drift electrode, a chevron-type microchannel plate (MCP) and a collector electrode, all enclosed in a vacuum envelope as shown in Fig. 2. The structure of the tube bears some resemblance to a streak tube but there are no deflector plates and no phosphor screen. The amplitude of the voltage step at the photocathode may be as high as 20 V within the duration of the luminous event. The photoelectrons are firstly accelerated through about several thousand volts by the first mesh and then
OBSERVING HIGH-SPEED LUMINOUS PHENOMENA
225
~ _ _ _ _ _ _ _ _
FIG.2. Structure of the experimental tube.
decelerated by the second mesh. Usually the second mesh is electrically connected to the drift electrode, which is held at a constant voltage of about 10-100 v. Collection of the photoelectrons by the MCP is ensured by applying an axial magnetic field of appropriate strength. The axial magnetic field does not affect the axial velocity of the photoelectrons.
CALCULATION OF THE PULSESTRETCHING
Approximate Analysis The pulse width To,,of the output signal is equal to the interval between the arrival times of the first and last photoelectrons at the input face of the MCP:
To,, = T ,
+ ATp, + ATd,
(2)
where TI is the length of the incident light pulse and ATpmand ATd are the transit time differences of the first and the last photoelectrons in travelling from the photocathode to the first mesh and through the drift electrode region respectively. If the incident light pulse is very short and the voltage of the first mesh is very high, T I and ATpmcan be neglected comparing to ATd, then
To,, ==ATd = L 12);’(
[(2y2 -(L)1’ vd-
vp2
vd-
(3)
Vpl
where L is the length of the drift electrode region, Vpl and Vp2 are the respective values of the photocathode voltage when the first and last photoelectrons are emitted.
22b
Y. KIUCHI N . KOSHIDA AND T. SAKUSABE
Arrival Times of the Photoelectron
The arrival times of photoelectrons at the end of the drift electrode region were calculated by computer for the various conditions of applied electrostatic and magnetic fields. It was found that there is little difference between the arrival times of photoelectrons which leave the centre of the photocathode and ones simultaneously emitted from a point 5 m m from the centre. The time difference is no more than 6.3 ps compared to the stretched pulse length of order 10 ns (see next section). There will therefore be little error if, in this experimental tube, the area of the light pulse is restricted to a circle of 10 mm diameter. Figure 3 shows experimental and computed values of the retardation of arrival times as a function of voltage difference V d - V,, between the drift electrode and the photocathode. The results show that if Vd- Vp is changed from 30 to 10 V during the incidence of the light pulse, the output signal can be stretched to about 50 ns.
300
200
100
0
20
40
VOLTAGE DIFFERENCE VD-Vp
60 (
I30 V
)
FIG.3. Retardation of the photoelectron arrival time versus applied voltage difference.
227
OBSERVING HIGH-SPEED LUMINOUS PHENOMENA
RESULTS OF PRELIMINARY EXPERIMENTS The experimental arrangement is shown in Fig. 4. A conventional semiconductor laser is used as a source of the light, giving a pulse of FWHM of about 700ps. The oscilloscope used is a TEKTRONIX 7834 storage oscilloscope with 7A19 plug-in amplifier having 400 MHz bandwidth. Figure 5 shows some oscilloscope traces. Trace (a) is the input light pulse of a semiconductor laser determined separately. Trace (b) is the output signal obtained when the photocathode voltage is held constant at 0 V and the drift electrode voltage is held at 173 V. In this case all photoelectrons travel through the drift electrode region with the same velocity so that there is no pulse stretching. The pulse shape is almost similar to that of the input light pulse with a little widening due partly to the frequency response of the LASER
-j
EXPERIMENTAL TUBE
osc I LLO SCOPE
7834 PULSE OUT r--O-
- - - PULSE GENERATOR - - - -
-I
I
I
FIG.4. Experimental arrangement for testing tube.
oscilloscope and partly to secondary electron dispersion in the MCP. The time delay of the entire signal from the moment of incidence of the light pulse is about 40 ns. Trace (c) shows the waveform of the photocathode step voltage and trace ( d ) is the waveform of the output signal obtained when the photocathode voltage is raised from 0 to 50 V whilst the drift electrode voltage is held at 60 V. The FWHM of the output pulse is about 10 ns, so that the time extension ratio is about 15. This confirms in principle the validity of the new method. LIMITATION OF
THE
METHOD
This method is based on velocity differences of the photoelectrons. It is therefore affected by the initial velocity dispersion of the emitted photoelectrons. The statistically averaged value of the initial energy is about 0.1 eV, the photocathode step voltage should therefore be much larger than 0.1 V, at least 10 V, say, during the incidence of the light pulse.
228
Y. KIUCHI N . KOSHIDA A N D T. SAKUSABE
(c)
(d)
FIG.5. Examples of oscilloscope traces. (a) Input light pulse from semiconductor laser, 500 ps per division. (b) Output signal when the photocathode voltage is constant and the drift electrode voltage is 173 V, 1 ns per division. No pulse stretching is obse&ed. (c) Waveform of the photocathode step voltage. (d) Output signal when the photocathode voltage is raised from 0 to 50 V and the drift electrode voltage is 60 V, 5 ns per division.
The electron temperature of the photoelectrons was measured by a retarding potential method using He-Ne gas-laser irradiation. The measured temperature was 631 K, corresponding to an initial energy of 0.54 eV, larger than the theoretical value given above. The pulse widening due to the measured initial velocity dispersion alone is shown in Fig. 6 . The highest slew rate of the step voltage that can be obtained at present is about 1 V ps-', so that the shortest measurable pulse width for a step voltage of 10 V is about 10 ps. The initial velocity is a statistical quantity, so that there is the possibility of improving this method by deconvolution. There exist, however, other practical difficulties such as noise and jitter associated with the signal and the trigger pulses.
DATAACQUISITION AND PROCESSING The oscillographs of Fig. 5 show that the output pulse waveform is asymmetric even if the input pulse is almost symmetric. This is due to the non-
229
OBSERVING HIGH-SPEED LUMINOUS PHENOMENA
ELECTRON TEMPERATURE 631 K V : VOLTAGE DIFFERENCE L T W E E N V p A M VD
1
W I
1 GO
50
ARRIVAL TINES
150
200
( ns )
FIG.6. Calculated pulse widening produced by the initial velocity dispersion.
TIX
f
VOLTAGE PHOTOCATHODE VOLTAGE CHAWGE
ARRIVAL OIFFERENCE
T2 T1
-
1 v2
v1
VOLTAGE D I FFERENCE BETWEEN V p AND VD
CORRECTED PULSE
FIG.7. Principle of pulse-shape correction.
linearity of the arrival time versus voltage difference shown in Fig. 3. To determine the true waveform of the incident light pulse, it is necessary to correct the oscillograph trace. Figure 7 shows the principle of the correction. In this figure, the voltage between photocathode and the drift electrode at the
230
Y. KIUCHI N . KOSHIDA A N D T. SAKUSABE
EXPER I MENTAL
TV CAMERA
osc I LLOSCOPE
PULSE GENERATOR
MEMOZY SYNC.
COMPUTER PC 9801F
-
CPU
I .Fa 4
CRT
FIG.8. Block diagram of the data-acquisition and signal-processing system.
FIG.9. An example of the effect of data processing to remove jitter.
beginning of the incident light pulse is Y , and the arrival time of the first photoelectron is t l . A block diagram of the data-acquisition and processing system that is now under development is shown in Fig. 8. Generally the high-speed oscilloscope trace is very faint, so that the pick-up camera must have high sensitivity. Data processing may be able to correct for the influence of noise and jitter as well as the above-mentioned initial velocity dispersion. Figure 9 shows an example of data processing used to remove the jitter seen in Fig. 5(d); noise and nonlinearity are still present.
CONCLUSIONS A new method for observing high-speed luminous phenomena has been demonstrated in principle, but at present the pulse stretching is not as large as previously expected and the technique is currently limited to light pulse lengths of the order hundreds of picoseconds. There are still many difficulties: the initial velocity dispersion and interactions between photoelectrons are inherent, but may be improved by data processing to some extent. The main practical difficulties are noise and
OBSERVING HIGH-SPEED LUMINOUS PHENOMENA
23 1
jitter associated with the signal and the trigger pulses. These must be eliminated by step-by-step improvements to the circuits and the system. ACKNOWLEDGEMENTS The authors would like to thank Professor D. J. Bradley for giving us the motivation to do this work while discussing the streak camera and we also thank Dr. Y. Suzuki, Dr. K. Oba, Dr. Y. Tsuchiya and other members of Hamamatsu Photonics Co. Ltd. for their valuable contributions and many helpful discussions.
REFERENCES
Baggs, M. R., Eagles, R. T., Margulis, W., Sibbett, W. and Sleat, W. E. (1985). In"Adv. E.E.P." Vol. 648, pp. 617-625 Bradley, D. J. and Sibbett, W. (1975). Appl. Phys. Lett. 27, 382-384 Kinoshita, K., Hirai, N. and Suzuki, Y. (1984). Proc. S.P.I.E. 491, 63-67 Nozaki, M., Sakusabve, T. and Kiuchi, Y. (1985). Tech. Rep. Inst. Electron. Commun. ED85-19, 13-19 (in Japanese) Takiguchi, Y.,Kinoshita, K., Inuzuka, E. and Tsuchiya, Y. (1984). Proc. S.P.I.E. 491,224-229
This Page Intentionally Left Blank
An X-Ray Streak Tube with Demountable Photocathodes B. E. DASHEVSKY, V. A. PODVYAZNIKOV, A. M. PROKHOROV, A. V. PROKHINDEEV and V. K. CHEVOKIN General Physics Institute of the USSR Academy of Sciences, Moscow, USSR
INTRODUCTION Time-analysing X-ray streak tubes (XRST) are reliable diagnostic instruments in experiments on the interaction of high-power laser radiation with matter. Currently, several types of time-analysing XRST have been described (Bradley et al., 1975; Kinoshita et al., 1983; Kauffman et al., 1982; Boutry et al., 1982; Bryukhnevich et al., 1978; Chevokin et al., 1974), which differ in their electron optics and their input windows. All these instruments except those described by Bryukhnevich et al. (1978) and Chevokin et al. (1974) are of the open type, i.e. they are connected to a vacuum chamber to allow manufacture of a very thin substrate with good transparency not only in the soft X-ray region but also in the UV range of the spectrum. The possibility of photocathode exchange makes it possible to conduct comparative tests of photocathode X-ray quantum efficiency. If the photocathode is destroyed, it can be replaced, which is impossible in the sealed-off devices. The present paper reports on the development of an open-type timeanalysing XRST, intended for studying fast processes in the soft X-ray and UV ranges. The design of the instrument and radiation sources for adjustment of the XRST’s static regime are considered, and the results of previous tests of its static and dynamic characteristics are presented.
DESIGN AND ELECTRONIC PROPERTIES The time-analysing XRST uses an electron-optical system which produces a quasi-spherical focusing field. The electron-optical magnification is unity. The XRST has a special flange at its input connecting it to a vacuum chamber, In the centre of this flange there is a hole for installation of the various photocathode holders. The flange also contains holes for evacuation of the XRST volume. In the centre of the photocathode holder there is a slit of width 100 pm and height 8 mm which is covered by 1000 A thick nitrocellulose film, acting as a substrate for the photocathode. Evaporated layers of Au and CsI 233 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
234
B. E. DASHEVSKY ET A L .
FIG.I . Photograph of the X-ray streak tube.
with thicknesses 300A and IlOOA, respectively, are used as the photocathodes. The ability to remove the photocathode makes it possible to carry out tests, which are impossible in sealed-off tubes. In some experiments the nitrocellulose film was burned out and measurements were performed on the unsupported photocathode film. If the photocathode is damaged by an accident, it can be changed very easily and rapidly. For time analysis of the image there are two perpendicular conic streak plates, separated by an aperture diaphragm at a cross-over point. The diaphragm hole diameter is 0.3 mm. At a photocathode potential of 5 kV,the electrical plate sensitivity is 4 k V across the screen. A pair of vertical deflection electrodes (sweep electrodes) streaks the signal across the 25 mm diameter screen. This image is then amplified by an MCP image intensifier tube. A photograph of the demountable photocathode, X-ray streak tube is shown in Fig. 1.
EXPERIMENTAL INVESTIGATIONOF XRST
PROPERTIES
The main difficulty for experimental investigation of the XRST is that its Xray photocathodes are not sensitive to visible light so that development of appropriate radiation sources has been necessary. The principal requirements for such sources are small size, simplicity of design, well-know spectral range, and sufficient radiation power. To measure the static characteristics of the XRST we therefore tested sources of both soft X-ray and UV radiation as well as of electron-beam radiation. In all cases the distance between the XRST photocathode and the source was 400 mm. The first source of soft X-ray radiation consisted of a direct-filament
AN XRST WITH DEMOUNTABLE PHOTOCATHODES
235
thermocathode, a grid fixed near the thermocathode and an anode. A tungsten filament was used as a thermocathode to increase its electron emission; this was oxidized in a vacuum. The grid was supplied with a pulsed voltage of duration 10 ms and amplitude up to 2.5 kV. The anode voltage was about 15 kV. The source operated with a repetition frequency up to 50 Hz. The anode current was 400 mA, which corresponds to an electron-beam power of 6 kW in the pulse. The advantage of this source is that at a fixed anode voltage the bremsstrahlung X-ray radiation spectrum is well known. Thus, by varying the anode voltage, the XRST parameters can be measured in a required spectral range. Thus, for example, the limiting X-ray wavelength l o is related to the anode voltage u by 1
l o = 1.24 x 10V6U
and the maximum wavelength of X-ray radiation, A,,,, depends on l o as A,, = f l o . The low coefficient of electron energy conversion into bremsstrahlung radiation (in our case it reaches- 1%) is a disadvantage of this type of source. As a second source we used radiation from a spark gap situated in a vacuum chamber. A 5 kV pulse with a repetition frequency up to 10 Hz was applied to one of the electrodes and the other electrode was grounded. During the pulse duration of 1-3 ps the current was 50-100 A. The experiments showed that the radiation spectrum of this source has a maximum in the UV range, and the power was sufficient to observe a bright image of the slit on the screen. The low energy stability from flash to flash is a disadvantage of spark sources. X-ray photocathodes, both metallic and dielectric, are known to be very sensitive to electrons. Therefore, it should be possible to carry out static measurements on the X-ray streak-tube by irradiating the photocathode with an electron beam. A tungsten filament similar to that employed in the bremsstrahlung X-ray source was used as the electron source. A positive voltage was applied to the XRST to produce an accelerating potential for the thermoelectrons. After oxidation, the maximum emission current from the cathode was 20 mA and this did not change when air was allowed into the vacuum chamber followed by re-evacuation. An electric field was created near the cathode by a pulsed voltage of amplitude 2.5 kV and frequency 50 Hz, which was applied across the grid. The 100 mA current reaching the photocathode of the XRST was quite sufficient for observing the input image on the screen. The electron-beam source therefore makes it easily possible to carry out qualitative calibration of the XRST. Correct operation of the X-ray streak-tube was determined by means of the radiation sources. In all measurements the XRST was connected to the
236
B. E. DASHEVSKY ET A L .
FIG.2. Test of spatial resolution of the X-ray streak tube.
intensifier. An optimal focusing voltage was selected. To measure the spatial resolution of the XRST, we used a special photocathode holder consisting of a container in the form of honeycomb grid with 95% transmission hexagonal cells. A 1000 A nitrocellulose film and a vacuum-evaporated 300 A gold film (which forms the photocathode) were successively deposited onto the grid. Applying the substrate as a honeycomb grid enables one to produce an unsupported 8 mm diameter gold film. Outside of the honeycomb grid were placed metallic grids with different steps which were irradiated by one of the sources mentioned above. Figure 2 illustrates the 150 pm step grid as well as the honeycomb structure. The measured spacial XRST resolution was not less than 10 Ip mm-’.
DYNAMIC TESTSOF
THE
XRST
The dynamic tests were performed in experiments recording laser plasma Xray radiation. This was produced by focusing laser radiation on the surface of a flat target which was placed in the vacuum chamber. The laser was a modelocked Nd:YAG generator with a single pulse selector and a two-stage Nd :YAG amplifier. A KDP crystal for second-harmonic generation was placed in front of the vacuum chamber, giving a conversion coefficient of 30%. A copper target was installed in the vacuum chamber.
-
AN XRST WITH DEMOUNTABLE PHOTOCATHODES
237
9
ir)lX-Ray
\
A=O.! ,
A = 1.06p-n
50 ps ,30MJ
18
A
11
11 20
FIG.3. Experimental set-up. 1,7,8,21,22-100?/0 mirrors; 2 d y e c e l l ; 3,9, I G Y A G : Nd rods; 5-50Y0 mirror; &system for extraction of a single laser pulse; 1 I-light delay; 12-light diode; 13-vacuum chamber; 14-X-ray source; IS-opper target; I G X R S T ; 17-image intensifier; 18-pulse electronics for XRST; 19-power supply; ZGstreak camera with S-1 photocathode for laser monitoring.
The circuitry for the X-ray streak-tube was manufactured using avalanche transistors switched according to the Marx scheme, and two high-frequency metal-ceramic valves. The control circuit generated two pulses at rates from lo8 cm s-' to lo9 cm s-'. The delay of the circuit was 15 ns with a jitter of L- 100 ps. The required synchronization was obtained by varying the corresponding optical delay line. A schematic of the set-up is shown in Fig. 3. The temporal resolution of the XRST depends on the photocathode potential. With higher potential, the field strength near the photocathode increases, thereby increasing the temporal resolution. The sensitivity of the streak plates falls, however, making the design of the control circuit more difficult. As was mentioned above, the photocathode voltage was 5 kV. Estimates of the temporal resolution, taking the axial potential distribution into account, shows that it is 40 ps, quite sufficient for most experiments on laser plasma X-ray diagnostics. Figure 4 shows a temporal scan of the X-ray laser plasma radiation; one section of the photocathode is coated with a gold, the other with CsI.
238
B. E. DASHEVSKY ET A L .
0
-
t [nsl
1
1
2
FIG. 4. Response to X-ray radiation from laser plasma: (a) Au photocathode; (b) Csl photocathode. The interval between the two pulses is 1 ns.
At present we are engaged in development of measurement techniques for the dynamic properties of the X-ray streak-tube.
REFERENCES Boutry, B., Cavailler, C. and Fleurot, N. (1982). Proc. S.P.I.E.348,766-771 Bradley, D. J., Roddie, A. G., Sibbet, W., Key, M. H., Lamb, M. T., Levis, C. L. S. and Sachsenmaier, P. (1975). Opr. Commun. 15,231-236 Bryukhnevich, G. I., Zak, E. I., Kil’pio, A. V. and Klepov, A. F. (1978) In “Proc. 13th International Congress on High Speed Photography and Photonics, Tokyo”, pp. 492-495 Chevokin, V. K., Kas’yanov, Yu. S., Korobkin, V. V., Malyutin, A. A., Prokhorov, A. M., Schelev, M. Ya. and Richardson, M. C. (1974). Laser Elektro-Opt. 4,40-42 Kauffman, R., Stradling, G. and Medecki, H. (1982). Proc. S.P.I.E. 348,752-759 Kinoshita, K., Inuzuka, E., Takiguchi, V., Okada, H., Suzuki, K., Hayashi, I., Tsushiya, Yu., Oba, K., Medecki, H. and Stradling, G. (1983). Proc. S.P.I.E.427, 3 6 4 4
A Subnanosecond Multi-framing Camera V. V. LUDIKOV. A. M. PROKHOROV and V. K. CHEVOKIN General Physics Institute of the USSR Academy of Sciences. Moscow. USSR
INTRODUCTION Research in the field of multi-framing image-converter cameras (ICC) has been in progress for many years, and several versions of multi-frame ICCs with subnanosecond exposure times have been developed (Lieber and Sutphin, 1971; Clement, 1972; Kalibjan, 1976; Sibbett et al., 1982; Kinoshita, 1985; Majumdar, 1984; Huston and Harris, 1972; Fleurot et al., 1982). These developments differ in the various ways of forming the images on the screen of an image-converter tube, in the range of the cameras’ spectral sensitivity, in the number of frames, in the time interval between the frames, etc. The importance of multi-frame cameras has grown recently because they are needed for laser plasma studies, such as studies of the dynamics of plasma formation and scattering in various spectral regions from the near IR up to the hard X-ray regions. The present paper reports the development at the General Physics Institute of a four-frame ICC with less than 400 ps shutter time. We present the results of dynamic tests on the camera and its principal technical specifications.
FOUR-FRAME ICC,
THE
“SAPPHIRE-FRAME”
A four-frame ICC, the “Sapphire-Frame”, has been developed from the basis of a PIM-IOSB image-converter tube. The camera consists of a tube which incorporates a photocathode, a shutter grid, a focusing system, an anode diaphragm and two pairs of mutually perpendicular deflecting plates. The shutter grid is made as a 50 R strip-line with transmission up to 1.5 GHz and is formed by a photocathode and a fine-mesh grid 1.8 mm away from it. The grid cut-off voltage is not more than 1OV. The PIM-lO5B tube incorporates an MCP, providing an intensification of not less than 103.Two pairs of deflecting plates are made in the form of 50 R strip-lines connected with a coaxial cable. The screen is formed on an optical fibre disk, the screen diameter is 40 mm. The electron-optical magnification of the PIM-lO5B tube is 1.8. The photocathode is an S.20 of useful area 15 x 15 mm2.A general view of the tube is shown in Fig. I .
239 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited
All rights of reproduction in any form reserved ISBN 0-12-014674-6
FIG.1. Photograph of the ICT PIM-105 B.
rlOnsl FIG.2. Block-diagram of the “Sapphire-Frame” ICC.
A SUBNANOSECOND MULTI-FRAMING CAMERA
24 1
A block diagram of the “Sapphire-Frame’’ ICC is shown in Fig. 2 . In the static mode the tube is closed by making the photocathode voltage positive with respect to the grid. The grid potential may be as high as 7.5 kV. The tube is switched on by making the photocathode voltage negative with respect to the grid. Spatial resolution in the static mode is 15 Ip mm-’. A static voltage of 850 V may be applied to the MCP without giving rise to dark glow at the screen, so that pictures may be recorded by a contact technique. Appropriate voltage supplies are provided for the static requirements of the ICC and control circuit. To measure dynamic performance of the “Sapphire-Frame”, pulsed electronics have been developed (see Fig. 3). It consists of two independent parts; one produces pulses for the grid shutter, the other produces staircase pulses which are applied to the deflecting plates. Both parts of the control circuitry consist of avalanche transistors and metallic-ceramic triodes. A high-frequency metal-ceramic GI-25 valve is used in the shutter pulsegenerating circuit. Pulses from the avalanche transistors with about 100 V amplitude and 5 ns duration reach the grids of both triodes simultaneously. The output pulse from the first triode is fed immediately to the circuit output, and that from the second valve passes via the cable delay line providing a 10 ns pulse delay. Thus, two pulses are formed with an amplitude of about 120 V and 5 ns duration. The pulses are directed via a high-voltage separating capacitor and 50 R cable to the shutter cathode-grid connection of the ICC. A 50 R cable is connected to the output of the cathode-grid section. The length of the cable is chosen so that the reflected pulses return in 5 ns after travelling via the ICC’s shutter section. Four shutter pulses are thus formed, which travel along the cable to trigger the cathode-grid unit with 5 ns intervals. The number of the frames in the “Sapphire-Frame’’ camera is therefore four. A GI-41-1 triode is used in a staircase generating circuit. A pulse of amplitude 400 V and duration 25 ns pulse is applied to the input of a forming line consisting of sections of 75 R cable 2, 3, and 4 m in length, respectively. The cable sections are connected in parallel, thus providing pulse delays of 10, 15, and 20 ns, respectively. The delayed pulses are summed at the output cable to form a staircase pulse with 400 V amplitude, about 2 ns rise time at each step and a 3 ns flat section from step to step. This pulse is directed to one of the deflecting plates via a 50 R cable. A preliminary deflection voltage brings the image to the screen edge. The second plate is grounded. Initially, a pulse which prevents reversal of the beam was applied to the second pair of the plates, but experiments showed that the shutter grid does not trigger because of reflection of the blocking pulses from the tube; thus, it was considered more appropriate not to cut-off the reverse beam. Figure 4 shows the two blocking pulses with a 10 ns interval between them and a staircase voltage pulse recorded from the screen of a high-speed oscillograph. Both control circuits are triggered from
FIG.3. Control circuit of the “Sapphire-Frame”ICC.
Elapsed time ( n s )
FIG.4. (a) Two blocking pulses with a 10 ns interval, recorded from the screen of a high-speed oscillograph; (6) the staircase voltage pulse; (c) timing diagrams of the blocking pulses and staircase pulse.
FIG.5. Photograph of a part of the test chart recorded in framing mode.
244
V. V. LUDIKOV, A. M. PROKHOROV AND V. K . CHEVOKIN
the leading edge of the incoming pulse via a transformer wound on a ferrite ring. To bring the two parts of the control circuit into step, the shutter pulse is triggered via a delay line made of a 50 R cable. Figure 5 is a photograph of a test chart taken from the ICC screen of the “Sapphire-Frame” camera operating in the framing mode. The spatial resolution of the test chart is 5 lp mm-’ and this resolution is seen to be preserved in each frame. The frame size on the screen is 7 x 7 mm2. The test chart image was illuminated using a flash-lamp operating in a self-breakdown regime.
I
0
I
1
5 10 Elapsed time (ns)
I
15
FIG.6. (a)Streak records of four frames; half of the slit is attenuated by a neutral density filter; (b) and (c) show the corresponding microdensitometer traces.
A SUBNANOSECOND MULTI-FRAMING CAMERA
245
TABLE I Major specifications of the “Sapphire-Frame’’camera Spectral sensitivity (S-20 photocathode) Number of frames Frame size on the screen Distance between frames Speed of photograph Spatial resolution in each frame Exposure time of each frame Trigger signal (R=50 a)amplitude Jitter Delay Size Weight Power supply
0.4-0.8 pm 4
7x7mm2 5 ns 2 x 10’ frames per second +5Ipmm-’ 400 ps 5+100V 1 0 . 1 ns >30 ns 255 x 280 x 330 mm3 12 kg 40 W
To measure the frame duration a slit was illuminated with the flash lamp and a linearly increasing pulse was applied to the first pair of plates. This pulse was formed by an auxiliary circuit not built into the camera and used only for camera tests. Four images of the slit were observed on the camera screen, corresponding to the four blocking pulses. Each image was swept across the screen in the sweep direction. To restrict the signal to the region of linear blackening of the film, half of the slit was covered by a neutral density filter. The results are shown in Fig. 6. It can be seen that, firstly, the frame duration is far less than that of the applied blocking electric pulse and, secondly, the shape of the blocking pulse and the shape of the corresponding frame pulse differ greatly. Both of these facts are associated with peculiarities of the cathodegrid shutter and at present we are attempting to reduce the frame duration. However, preliminary estimates indicate that the frame duration depends non-linearly on the blocking pulse: for very short pulses, the frame duration is almost 3 to 5 times shorter. When the pulse duration is 10 ns or more, there is no marked difference between the pulse and frame duration. CONCLUSIONS A multi-frame image converter camera, the “Sapphire-Frame” has been developed, providing a framing rate of 2 x lo8 frames per second with about 400 ps duration in each frame. The frame duration of 400 ps is easily achieved and we expect that in the very near future our experience will enable us to attain a framing rate of about 4 x 10’ frames per second. The authors hope that the “Sapphire-Frame” camera will be widely used in laser experiments. Its principal features are summarized in Table I.
246
V. V. LUDIKOV, A. M. PROKHOROV AND V. K. CHEVOKIN
ACKNOWLEDGEMENTS The authors are grateful to V. P. Degtyareva for help in reducing the results, V. K. Belyaev and V. A. Podvyaznikov for help in camera fabrication, V. E. Postovalov and Yu. N. Serdyuchenko
for useful discussions of some problems in the electronic circuits, and M. Ya. Schelev for interest in the work.
REFERENCES Clement, G. (1972). In “Adv. E.E.P.” Vol. 33B, pp. 1131-1136 Fleurot, N., Gex, J. P., Rostaing, M. and Sauneuf R. (1982). Proc. S.P.I.E. 348,772-776 Huston, A. and Harris, R. (1972). In “Adv. E.E.P.” Vol. 33B, pp. 1109-1117 Kalibjan, R. (1976). Proc. S.P.I.E. 97, 269-274 Kinoshita, K. (1985). Proc. S.P.I.E. 569,2-8 Lieber A. and Sutphin H. (1971). Rev. Sci.Instrum. 21, 1663-1667 Majumdar, S. (1984). Proc. S.P.I.E. 491,913-916 Sibbet, W., Baggs, M. R. and Niu, H . (1982). Proc. S.P.I.E. 348,267-270
A CsI(Na) Scintillation Plate with High Spatial Resolution K. OBA, M. ITO, M. YAMAGUCHI and M. TANAKA Hamamatsu Photonics K.K. Shimokanzo. Toyooka-mura. Iwata-gun. Shizuoka-ken, Japan
INTRODUCTION A two-dimensional segmented scintillation plate and a scintillation fibre bundle have been considered as useful converters suitable for soft X-ray and yray imaging and charged-particle detection. Applications have included X-ray image intensifiers, high-energy particle track detectors, neutron imaging devices, autoradiography, electron microscopy and so on. A typical application of segmented CsI(Na) scintillators is the X-ray image intensifier, in which a closely packed array of CsI(Na) thin capillaries serves as a two-dimensional X-ray-to-light converter with which 3-4 lp mm- spatial resolution has been obtained (Sano et al., 1986). In the field of high-energy particle track detectors, several trials have been reported, including bundles of scintillation fibres comprised of scintillation glass (Ruchti et al., 1986), plastic scintillator (Konaka et al., 1987) or CsI scintillator (Smith, 1975). Conventional phosphor material has also been used as a two-dimensional fl-ray detector in autoradiography (Kume et al., 1987);phosphor particles were intagliated to a chemically etched fibre plate to improve spatial resolution. The important points of these scintillators may be summarized as follows: (i) segmentation is as small as possible to increase resolving power; (ii) optical cross-talk is as low as possible to decrease smearing of the image; (iii) transmission of the light is efficient enough to obtain good signal-to-noise (S/N) ratio. This paper describes a new type of CsI(Na) scintillation plate which has a completely different construction and a different operational principle from the segmented scintillators mentioned above. The scintillation plate consists of segmented CsI(Na) scintillators and a fibre plate. In the fabrication of the plate, the CsI(Na) crystals are grown on the fibre plate, which has a special surface configuration so that the CsI(Na) crystals grow as extensions of each glass core of the fibre plate. Accordingly, the light image produced in the CsI(Na) crystal array is transferred to the fibre plate and appears on the other
'
247 ADVANCES IN ELECTRONICS AND ELECTRON P w s m VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in m y form reserved ISBN 0-12-014674-6
248
K . OBA ET AL.
side of the plate without loss of spatial information. The sizes of the fibres tested were 20 pm and 6 pm. Because of the special configuration of the scintillation plate, the requirements mentioned above are almost satisfied. A disadvantage of the CsI(Na) scintillation plate is degradation of efficiency due to the hygroscopic effect. However, this effect has been overcome by enclosing the plate in a sealed housing with a beryllium window. In the following sections, fundamental properties of the plate such as detection efficiency, output light intensity, linearity and spatial resolution are described, and the results of life tests on the sealed plate are given.
PREPARATION OF
THE
CSI(NA)SCINTILLATION PLATE
The new type of CsI(Na) scintillation plate has the structure shown in Fig. 1. In the preparation of the plate, first one side is processed in such a way that the clad glass is etched selectively. The depth of etching is less than the diameter of the core. Figure 2 shows a top view of the fibre plate after the etching process. Two types of fibre plates have been used having 20 pm cores and 6 pm cores, respectively. The type shown in Fig. 2 has 20 pm cores. CsI(Na) crystals were grown on these plates by vacuum deposition. The heights of the crystals were chosen to be 10 pm to 150 pm, depending on the application. An SEM sectional view of the crystal is shown in Fig. 3. Looking at the bottom of the crystal carefully, it seems clear that the growth of each pillar starts from the top of the core glass. Furthermore, a very narrow space is seen between each pillar. This is an important feature which reduces cross-talk effects. Aluminium is evaporated on the top of the pillars to give a reflective coating. The CsI(Na) crystal is well known to be very hygroscopic and shows rapid degradation of scintillation efficiency if operated in air. To prevent this degradation, a special air-tight housing with a Be window has been developed y-photon
I
b
Csl (Na) Crystal /
i
FIG.1. Structure of the CsI(Na) scintillation plate.
FIG.2. SEM picture showing top view of the etched fibre plate.
FIG.3. SEM picture showing sectional view of a scintillation plate of the 20 pm type.
2 50
K. OBA
ET AL.
FIG.4. Air-tight housing with beryllium window.
as shown in Fig. 4. Results of life tests of the crystal accommodated in this housing are given later.
EXPERIMENTAL RESULTSAND DISCUSSION Light Emission and Linearity The amounts of light emitted from the CsI(Na) scintillation plate when excited by prays from 55Fe(5.9 keV), '291(29.0keV) and 24'Am(59.5keV) were measured using a photon-counting photomultiplier (Hamamatsu R878). The results are plotted in Fig. 5 for both 20 pm and 6 pm types with thicknesses of about 100 pm. Typical pulse height distribution for a plate excited by 1291is shown in Fig. 6 . Contrary to expectations, larger numbers of photons were observed using the 6 pm plate than the 20 pm type. It became clear from SEM studies that the size of the individual crystals was larger than the diameter of the fibre in the case of the 6 pm plate. Furthermore, the transmittance of the 6 pm type was about twice that of the 20 pm type. The crystal size would lead to the transmission of the light through the crystals being almost the same in both types; however, the difference of transmittance results in lower output from the fibre in the 20 pm type. Linearity of the light emission to the pray energy seems fairly good except in the lower energy region. This is because the light emission occurs near the input surface region in the case of 55Feand the transmission loss becomes
CSI(NA) SCINTILLATION PLATE WITH HIGH SPATIAL RESOLUTION
1500 . .
25 1
20 pm Fiber plate A 6 pm Fiber plate 'output from the Flber Plate 0
=
'Thlckness
100 pm
~1000-
6
c
I
0
10
20
.
I
30
I
I
40
I
I
.
L
L
60
50 rray energy (keV)
FIG.5. Number of photons emitted from the scintillation plate excited by y-rays from "Fe, '*'I and 241Am.
'2m W
pSource '"1 Max. 420 c. ihotons
2a
00
100mm4w500600
700
800 800
1m
CHANNEL NUMBER
FIG.6. Typical pulse-height distribution of the output light from a 6 pm type scintillation plate excited by y-rays from !*'I.
large. For low energies a sample as thin as 10 pm on the 6 pm type fibre plate was tested. About 50 photons were obtained from this plate for each y-ray from 55Fe.This type also offers better spatial resolution as described later. So far, attempts to extend these 6 pm crystals to 100 pm thickness have failed.
Detection Eficiency The detection efficiency of the 100 pm thick crystals for y-rays from '291was measured and compared with that of the bulk crystal. The results show that
252
K. OBA ET A L
the 6 pm type gives an efficiency of 22% and the 20 pm type gives 16% instead of the 25% obtained from the bulk crystal. Spatial Resolution The spatial resolution has been tested using both a 4 pm slit-shaped electron beam and a 20 pm slit-shaped X-ray beam. The energies of the electron beam and the X-ray beam were 20 keV and 90 keV respectively. (90 keV was the voltage on the X-ray tube.) The electron beam produces light near the surface of the scintillator because of the beam’s small penetration depth. It is therefore useful to evaluate crosstalk through the transmission process. Figure 7 shows slit images of the electron beam. The pictures were taken using an SIT camera coupled to the scintillation plate via a magnifying lens. Figure 7(a)shows an image obtained from the 20 pm type with a thickness of 100 pm. It is clear that only one line of 20 pm crystal is excited by the electron beam and the FWHM is 20 pm. Figure 7(6) shows the 6 pm type with a 10 pm thick layer. The FWHM is 10 pm. If the layer thickness is increased to 100 pm, the crystal diameter itself becomes larger and the FWHM rises to 19 pm as shown in Figure 7(c). In the case of X-ray excitation, the spread becomes a little wider than expected, as shown in Fig. 8. Figure 8(a) shows an image obtained from the
FIG.7. Images of a 4 pm slit-shaped electron beam taken using (a) the 20 pm type with 100 pm thick crystal, ( b ) the 6 pm type with 10 pm thick crystal, (c) the 6 pm type with 100 pm thick crystal. FWHM values are (a)20 pm, (b) 10 pm, (c) 19 pm.
CSI(NA) SCINTILLATION PLATE WITH HIGH SPATIAL RESOLUTION
253
FIG.8. Image of a 20 pm slit-shaped X-ray beam taken using (a)the 20 pm type and (b)the 6 pm type. FWHM values are (a) 45 pm and (b) 30 pm.
20 pm type and the FWHM is 45 pm. At some positions, more than two lines are scintillating. In the case of the 6 pm type (Fig. 7(b)),the FWHM is 30 pm. Life Test
It is well known that CsI(Na) suffers from degradation of scintillation efficiency due to the hygroscopic effect. The efficiency, however, can be recovered by baking, for instance, at 300°C for about 30 minutes in air. This would be only a temporary solution. A better solution would be to accommodate the crystal in an air-tight housing with a suitable input window. Figure 4 shows an example of the housing, which has a 500 ,urn Be window and which is filled with dry nitrogen gas. In Fig. 9 scintillation efficiency is plotted against time. The crystal accommodated in the housing has shown no
c
100' G 90;..80-
o
Sealed type
70-
.-6
5
$
0
.-B 2
-
60 5040-
3020 10-
,
254
K. OBA E T A L .
FIG.10. Output image from the image intensifier corresponding to 120 pm pitch w-slit pattern irradiated with 45 keV X-rays.
degradation after 750 hours; however, the naked crystal, which was kept in a desiccator filled with dry nitrogen gas, shows a very rapid degradation. In both cases, the measurement was done in a dark box filled with dry nitrogen gas. The life test of the crystal in the housing is not yet completed, but the results obtained so far look promising. COMBINATION OF THE PLATE WITH AN IMAGE INTENSIFIER
In the previous section, slit images taken using a combination of the plate and an SIT camera coupled through an optical lens were shown to demonstrate the intrinsic spatial resolution of the plate itself. However, in practical applications, an image intensifier (11) would be used to amplify and make detectable the very faint image produced in the plate by X-rays or charged particles. Figure 10 shows a picture from an SIT camera that is looking at the output screen of an I1 (Hamamatsu V1366) with a 20 pm type scintillation plate on the input window. The plate is excited by X-rays from an X-ray tube operated at 45 keV. The pattern has 120 pm pitch and 50% open area. The material is tungsten with a thickness of 50 pm. The slit is clearly seen with modulation higher than 10%.It is estimated that the FWHM of the line spread in this system is less than 80pm. The degradation of the spatial resolution compared with the intrinsic resolution occurs mainly at the interface of the plate and the I1 window.
CSI(NA) SCINTILLATION PLATE WITH HIGH SPATIAL RESOLUTION
255
CONCLUSION
A new type of two-dimensional CsI(Na) scintillation plate has been developed based on a special fibre plate processing and crystal growth technique. The 6 pm and 20 pm fibre size plates with a thickness of 100 pm have been fabricated and their fundamental properties such as light emission, linearity, detection efficiency and spatial resolution have been measured. The 6 pm type seems promising, having good light-emitting properties and spatial resolution. A fundamental disadvantage of such a plate is its hygroscopic nature; however, it has been shown that this problem can be solved by use of a special air-tight housing. ACKNOWLEDGEMENTS
The authors wish to express thanks to Mr. Endo, Mr. Uchiyama and Mr. Sugawara for their help in making this work successful.
REFERENCES Konaka, A. et al. (1987). Nucl. Instrum. Methods A256, 10-75 Kume, H. et ai. (1987). In “Proc. Pittsburgh Conference”, p. 658 Ruchti, R. et al. (1986). IEEE Trans. Nucl. Sci.33, 151-154 Sano, T. et al. (1986). Denshi Tokyo 25, 133-137 Smith, R. W. (1975). RCA Rev. 36,632-637
This Page Intentionally Left Blank
X-Ray Imaging Sensor Using CdTe/a-Si :H Heterojunction Y. HATANAKA, S. G. MEIKLE, Y. TOMITA and T. TAKABAYASI Research Institute of Electronics, Shizuoka University, Hamamatsu. Japan
INTRODUCTION Hydrogenated amorphous silicon (a-Si :H) having high resistivity and high photoconductivity has been extensively studied for imaging devices such as camera tubes (Imamura et al., 1979; Jones et al., 1985), overlaid solid-state image devices (Tsukada et al., 1981) and linear array input devices (Kanoh et al., 1981). Its small carrier diffusion length is suitable for the high resolution and anti-blooming that are most important for imaging devices. For X-ray imaging, there are sensing systems that use visible light conversion in a phosphorescent screen or direct X-ray photoconduction. Direct photoconduction without a light-optical system has been used in vidicon-type imaging devices, resulting in very high resolving power. Vidicontype camera tubes with photoconductive targets made from materials such as PbO (Heijine et al., 1954; Nishida and Okamoto, 1966), amorphous Se (Cope and Rose, 1966; Smith, 1960; Mithel and Rhoten, 1962), crystalline Si (Chester, 1969; Ashikawa and Takemoto, 1971), amorphous SeAs (Kawamura, 1982) and a-Si:H (Hatanaka et al., 1985) have been investigated. PbO vidicon tubes that are highly X-ray sensitive were found to be the most practical. PbO, however, is severely affected by environment, making mass production of devices difficult. Low-atomic-mass materials such as Si, a-Si :H and Se are poor for X-ray absorption and thus require a thick layer for device fabrication. In this paper, we describe the properties of single-crystal (sc-) and polycrystalline (pc-) CdTe/a-Si :H heterojunction devices constructed in an attempt to exploit the photoconductive properties of CdTe. The visible range optical band gap (1.5 eV) and high absorption coefficiency of CdTe make it ideal for solar cell applications as well as X-ray and y-ray detectors. The most promising use of the CdTe/a-Si :H heterojunction is not only in an X-ray camera tube but also in a large area sensor such as a solid-state panel, because it can function over the long term in the ambient ar. 257 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
258
Y . HATANAKA ET A L
X-RAY ABSORPTION COEFFICIENT OF CDTE AND A-SI:H FILM X-ray absorption coefficients of a-Si :H and pc-CdTe films were measured, with the results as shown in Fig. 1. a-Si :H films and pc-CdTe were deposited on a beryllium (Be) plate 0.5 mm thick and 26mm in diameter by a capacitively coupled glow discharge and by RF sputtering in argon gas respectively. The deposited films covered half the plate. The transmitted values of X-ray flux were measured for a plate with and without the film. The X-ray detector used had no window and consisted of two parallel plate electrodes of area 1 cm2 spaced 1 cm apart and filled with one atmosphere of dry air (Knoll, 1979). The electrodes were biased with a 1 kV DC voltage. The ion currents generated by the X-rays were measured after calibration with a Victorin radiometer (RADOCON model 500). The a-Si:H film has an absorption coefficient of about the same order as the crystalline silicon. Soft X-ray absorption is fairly strong, but hard X-ray absorption is poor. In comparison with the silicon material, CdTe has an absorption coefficient one decade larger. Most significant is the fact that the hard X-ray absorption coefficient is large.
-1 L ld
'E
c
u
Y
0
Z
Cd Te
k o
m
a
0
a-s;:H
10 20 30 40 50 X RAY TUBE VOLTAGE(HV1
FIG.1. Absorption coefficient versus X-ray tube voltage.
259
X-RAY IMAGING SENSOR
SINGLE-CRYSTAL CDTE/A-SI :H HETEROJUNCTION The sc-CdTe/a-Si:H heterojunction consisted of a high-resistivity (1 031O5 R cm) n-type CdTe substrate onto which p-type (boron-doped) a-Si :H was deposited. Depositions of a-Si:H were performed at 250°C in an R F capacitively coupled system with SiH4and B2H6 as source gases. Immediately prior to deposition the CdTe substrates were etched in 1 % and 0.1 % bromine in methanol solutions for 2 minutes in each. Contacts were provided by evaporated indium and gold for the CdTe and a-Si :H respectively. Figure 2 shows the current density versus voltage (Z-V) characteristics of a device with the a-Si:H layer doped using a B2H6/SiH4ratio of Forward current showed a steep rise with increased applied voltage and a diode factor that was larger than the value between 1 and 2 which had been expected. The large diode factor is considered to be a result of the high bulk resistivity of the CdTe and of space-charge effects in the a-Si: H film. The reverse current was low ( cm-2) and showed saturation characteristics with an activation energy of approximately 0.8 eV. This is considered to be a recombinationgeneration current at the interface, since neither JO nor E, showed a strong dependence on the a-Si :H doping level.
0
280
4.O 6 .O VOLTAGE ( volt 1
8 .O
FIG.2. I- V characteristics of an sc-CdTe/a-Si: H heterojunction. Open circles show forward current, solid circles dark current and solid triangle X-ray response.
260
Y . HATANAKA ET A L .
B I A S VOLTAGE 0 A 0
500
o v
1.0 V 10.0
v
6 00 7 00 8 00 WAVELENGTH ( n m )
FIG.3. Spectral response of an sc-CdTe/a-Si:H heterojunction.
In Fig. 2 are also shown the data obtained when X-rays of 100 R min-’, 40 kV were used. The photoexcited current showed a three-decade increase, corresponding to several microamperes in reverse bias saturation current. Since the X-ray absorption coefficient for CdTe is approximately ten times larger than that for a-Si :H, it is considered that absorption in CdTe governs the sensitivity characteristics. Figure 3 shows the spectral photoresponse when light was shown through the a-Si:H layer. A wide spectral response down to 1.5 eV indicates that a depletion region forms in the sc-CdTe and photoexcited carriers are efficiently injected into the a-Si :H film.
POLYCRYSTALLINE CDTE/A-SI:H HETEROJUNCTION The sc-CdTe described above is difficult to use in imaging devices because it is difficult to make a high-quality thin film. For imaging devices, we used a polycrystalline CdTe film prepared by the R F sputtering method. The pcCdTe film was prepared by sputtering of a CdTe polycrystalline target (purity 99.999%), 100 mm in diameter. The deposition conditions were: a substrate
26 1
X-RAY IMAGING SENSOR
C(220) C(311)
z - 0 20
60
C (511)
80
20(deg) FIG.4. X-ray diffraction spectrum of pc-CdTe prepared by the RF sputtering method.
temperature of 310"C, argon gas pressure of 0.06Torr and R F power of 150 W. The a-Si :H film was prepared by capacitively coupled glow discharge deposition. Conditions were approximately the same as for the case of the scCdTe/a-Si :H heterojunction. Figure 4 shows the X-ray diffraction pattern of a typical CdTe film deposited by the sputtering technique. The structure of this film was of the zinc blende type with a preferential orientation of the (1 1 1) planes parallel to the substrate. From FWHM (full width at half-maximum) of the (1 11) peak, grain size was estimated to be about 1000 A, which was smaller than that for films grown by the HWVE (hot wall vacuum evaporation) technique (Mimura et al., 1985). Electrical resistivity was about 1O'O l2 cm and was one order higher than that for the HWVE films. Electrical properties of pc-CdTe were measured with coplanar electrodes. Figure 5 shows the dependence of the light and dark current characteristics of the pc-CdTe on the sputtering substrate temperatures. Dark electrical resistivity and photoconductivity were highest at a temperature of 340°C. At temperatures higher than 340"C, a multi-random-axis domain grew in the films and the dark resistivity fell rapidly. The heterojunction characteristics for sandwich cells such as ITO/CdTe/Au (sample I), and ITO/n+a-Si:H, undoped a-Si : H/CdTe/Au (sample 2) were investigated. Figure 6 shows the I- V characteristics. Solid triangles represent sample 1; open circles and solid circles represent forward and reverse current of sample 2, respectively. The pc-CdTe/a-Si :H heterojunction shows diode behaviour, although the ratio of forward to reverse current of this junction is less than that for the sc-CdTe/a-Si :H heterojunction.
Y. HATANAKA ET A L .
262
h
a C
v
1
I
I I I
b
I
d
I
I I
I
I
I
II
I
/I /' / /
I I
I
1'
/
/
//
' 6'
Auo:23OoC A:29@C O O : 34OoC
n
a c
/
v
0.4 IZ a CT
3 0
0.2&
a
Q
D
0
10 20 30 APPLIED VOLTAGE ( V )
0
FIG.5. Dependence of substrate temperature on the pc-CdTe prepared by the RF sputtering method.
IMACMGDEVICES For an applications experiment, an X-ray imaging pick-up tube was fabricated. The target configuration was as shown in Fig. 7. For the substrate plate illumination window, a Be plate 0.5 mm thick and 26 mm in diameter was used for X-ray imaging and borosilicate glass 2 mm thick and 26 mm in diameter was used for visible-light imaging. The n+ a/Si :H and undoped aSi :H on the substrate plate served as blocking layers against holes from the substrate electrode. The pc-CdTe layer was grown to be intrinsic and was intended as a generation layer for the X-ray photocarriers. A weak p-type aSi :H 1 pm layer was deposited on the pc-CdTe. The a-Si :H and pc-CdTe layers formed the heterojunction. Holes injected from CdTe were carried to the a-Si :H surface without lateral diffusion. The 500 A layer of Sb2S3 was Torr. This layer functioned as deposited under vacuum at a pressure of protection against electron injection from the electron beam. Figure 8 shows the resolution chart reproduced by the glass-face-plate
100
-a -
70
I
+7z/ d.
i
C
I
F-
z
20
/
/
0
a Q
I
50
Y
[II
I-
/
3
z W
I
u
I
Q
a a 103
a
W rl)
53 30
0
a W >
U
W
ot
O,
1
2 3 4 5 APPLIED VOLTAGE ( V )
f
FIG.6. pc-CdTe/a-Si: H heterojunction characteristics.
X RAY
Be(0.5mm)
u
/
L
SbZS3( sood 1
FIG.7. Configuration of pc-CdTe/a-Si: H heterojunction target.
Y. HATANAKA ET A L
264
FIG.8. Image of visible-light resolution chart.
vidicon. High resolution of over 800 TV lines was obtained. Sensitivity to light input was about 200 nA lx-' as shown in Fig. 9. For X-ray imaging, pc-CdTE/a-Si :H heterojunction vidicons with a Be face plate were constructed as shown in Fig. 7. Target signal current characteristics are shown in Fig. 10. The CdTe/a-Si: H heterojunction target is
-
N
E,
z
a
cl
+
Z
w
= 5 10- 7 L)
0 I-
0
I
a
1o *
1 I L L U M I N ANCE
10 (Lx)
L
I
10 X-RAY
RADIATION
100 (Rlmin)
FIG.9. Light (triangles) and X-ray (circles) conversion characteristics.
26 5
X-RAY IMAGING SENSOR
0
5
10
TARGET
15
20
25
VOLTAGE (V)
FIG.10. Target signal current versus target voltage.
one decade higher in X-ray sensitivity than the a-Si: H target. This observation agrees with the absorption coefficient data shown in Fig. 1. Figure 9 also shows that the signal current versus X-ray intensity had a power dependence of 0.65. Here the X-ray induced current was about 300 nA for 100 R min-', 40 kV soft X-rays. Measurements were performed in a standard TV system with a field time of 16 ms. Figure 11 shows the X-ray monitor image of an IC package. Clearly resolved gold bonding wire images of an IC package were observed, though they included some imperfections in the form of white spots. After-image lag was quite large. These problems can possibly be overcome by improvement of the film deposition process.
266
Y . HATANAKA ET A L .
FIG. 1 I . Reproduced image of IC package observed using X-ray radiation.
SUMMARY
The visible-range optical band-gap and high absorption coefficient of CdTe make it ideal for solar cells as well as X-ray and y-ray detectors. The most promising use of the CdTe/a-Si :H heterojunction is not only in X-ray camera tubes but also in large-area sensors such as solid-state panels. The X-ray absorption coefficient of CdTe was measured as about 3000 cm-I, one decade larger than that for c-Si or a-Si :H. A sc-CdTe/a-Si:H heterojunction consisting of an n-type single-crystal CdTe substrate and p-type (B-doped) aSi: H prepared by the glow discharge method exhibited good diode behaviour with a low reverse bias saturation current of a few nA cm-*. The wide spectral response down to 1.5 eV indicates that a depletion region forms in the sc-CdTe and photoexcited carriers are injected efficiently into the a-Si :H film. A pcCdTe/a-Si :H heterojunction consisting of a CdTe film prepared by RF sputtering and glow discharge a-Si:H also showed clear diode characteristics and was used in a vidicon type tube. This device showed resolution greater than 800 TV lines and a sensitivity of 300 nA per 100 R per minute, i.e. one decade higher than that of a-Si:H based devices. ACKNOWLEDGEMENTS This work was in part supported by the Research Development Corporation of Japan. The authors wish to thank Dr. R. Nishida for his useful advice on the X-ray sensor and Mr. T. Kawai
X-RAY IMAGING SENSOR
267
of Hamamatsu Photonics Co. for providing the Be face-plate and facilities for tube assembly, and finally, Dr. K. Hirata of Nippon Mining Co. for providing the single-crystal CdTe wafers.
REFERENCES Ashikawa, M. and Takemoto, I. (1971). J. Inst. Telev. Eng. Jpn. 20, 715 Chester, A. N. (1969). Bell Sysf.Tech. J . 48, 345 Cope, A. D. and Rose, A. (1966). J. Appl. Phys. 25, 192 Hatanaka, Y., Zeng Bai Chuang and Mimura, H. (1985). Jpn J . Appl. Phys. 24, L129 Heijine, L., Schagen, P. and Bruining, H. (1954). Philip Tech. Rev. 16, 23 Imamura, I., Takasaki, Y.,Kusano, C., Hirai, T. andMaruyama, E. (1979). Appl. Phys. Lerr. 35, 349-35 1 Jones, B. L., Burrage, J. and Holtom, R. (1985). In “Adv. E.E.P.” Vol. 64B, pp. 437-445 Kanoh, Y., Usui, S., Sawada, A. and Kiuchi, M. (1981). In ”Tech. Dig. Int. Electron Devices Meet.”, 313 Kawamura, T. (1982). In “Proc. Nat. Conv. Inst. Telev. Eng. Jpn”, pp. 3-7 Knoll, G. F. (1979). In “Radiation Detection and Measurement”, p. 123. Wiley, New York Mimura, H., Kajiyama, S., Nogami, M. and Hatanaka, Y. (1985). Jpn J . Appl. Phys. 24, L717 Mithel. J. P. and Rhoten, M. L. (1962). SMPTEJ. 71,444 Nishida, R. and Okamoto, S. (1966). J . Insf. Telev. Eng. Jpn 20, 192 Smith, C. W. (1960). In “Adv. E.E.P.” Vol 12. pp. 345-361 Tsukada, T., Baji, T., Shimomoto, Y.,Sasano, A., Tanaka,,Y.,Maruyama, E., Takasaki, Y., Koike, N. and Akiyama, T. (1981). In “Tech. Dig. Int. Electron Devices Meet.”, 479
This Page Intentionally Left Blank
Low-noise Solid-state Linear Detectors for Large-field-of-view X-Ray Radiology H. ROUGEOT, B. MUNIER, G. ROZIERE and P. PRIEUR-DREVON Thomson-CSF Division Tubes Electoniques, Boulogne-Billancourt. France
INTRODUCTION Solid-state linear X-ray detectors consist of a linear array of silicon photodiodes having an overlying scintillator coating to convert the X-rays into visible photons which are sensed by the silicon. The radiological image of an object is obtained by scanning the object in the X-ray beam. We shall describe successively the operating principle of solid-state linear X-ray detectors, their normal operating conditions and their main characteristics. We shall then outline some of their applications.
A SOLID-STATE LINEARX-RAYDETECTOR
The main functions of a solid-state linear X-ray detector are shown schematically in Fig. 1: a linear silicon photodiode array covered by a scintillator detects X-rays that have passed through the object being examined. The signal from each photodiode is read out successively by CCD multiplexers. The multiplexed signal is subsequently processed by an electronic unit which corrects the DC offset and the gain of each channel by digital processing. The corrected signal can be exploited either for direct display through a frame memory or for data processing, e.g. for an automated control process. Figure 2 shows the detection and read-out principle used with solid-state X-ray detectors. A scintillator coating converts the X-ray photons into visible photons. The scintillator material is Tb-doped Gd202S. Its emission wavelength is 540 nm. The absorption curve for this scintillator coating for different X-ray energies and thicknesses is given in Fig. 3. It can be seen that absorption is very high (greater than 70%) for X-ray energies below 30 keV and between 50 and 80 keV. It is possible to increase the coating thickness for energies above 80 keV, but this would be at the expense of spatial resolution. The 540 nm emission band is detected by an array of silicon photodiodes. 269 ADVANCES IN ELECTRONLCS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12.014674-6
-
L
L .
-
-
DC offset
and g i n corrections
) .
\
--SCINTILLATING
-Sl
LATERAL INPUTS
LAYER
PHOTODIODES
I 1
t FIG.2. Principle of X-ray detection and electronic read-out in a linear X-ray detector.
LOW-NOISE SOLID-STATE LINEAR DETECTORS
27 1
X-ray energy(KeV1 I
20
I
30
I
40
I
50
I
60
I
70
FIG.3. Absorption curve for Gd202S scintillator.
For the detector described here, the array is an assembly of 16 modular arrays of 64 photodiodes. These photodiodes are formed by a p-type implantation on an n-type silicon substrate. The signal detected by the diodes is accumulated in the photodiode capacitances. The main diode characteristics are as follows: dimensions, 0.45 x 0.6 mm2; pitch, 0.45 mm; operating mode, 2-3 V reversebias voltage. Each array of 64 photodiodes is multiplexed by a silicon CCD. The CCD multiplexers provide the following functions: transfer of the photodiode signal charges to the CCD shift register (this transfer is operated in synchronism for the whole array); read-out of the charges by transfer in the shift register; charge-to-voltage conversion in the CCD output stage and amplification (the butput voltage range is 3 V across 1 kR).The main features of the CCD multiplexers are low noise ( < 0.5 mV RMS) and high dynamic range ( > 4000). A schematic diagram of the pre-processing electronics module is shown in Fig. 4. Its successive functions are as follows. (i) Sample-and-hold on the multiplexed signal. A memory stores the addresses of any sub-standard pixels so that their sampling can be inhibited. The corresponding signal is then replaced by that of the nearest neighbour. (ii) Coarse DC offset and gain corrections. These are carried out by
272
H. ROUGEOT E T A L . Corrfzt. of
Fodty Pixel odress
Bbit
DIA
8 bit DIA
Oc offset
5
8 bits to frame memory or computer
Gain
-
L
/
Fine mrecttons
-
/
Coarse corrections
FIG.4. Pre-processingelectronics.
analogue processing using signals stored in digital memories and enable the largest non-uniformities in the diodes’ dark current and Xray flux dispersions across the beam to be corrected. (iii) Analogue-to-digital conversion. (iv) Fine DC offset and gain corrections. The correction terms are stored in a RAM and can be updated at any time. The DC offset correction term is obtained by triggering an image-acquisition sequence with the Xrays off, and calculating for each pixel the average over 16 successive read-outs. The acquisition time for this correction term varies between 0.6 and 0.8 s. The gain correction term is obtained with the X-rays on and no object in the way of the beam. Here, the moving average is calculated over 32 successive samples, and the acquisition time for this correction term is between 1.2 and 1.5 s. These fine corrections are essentially carried out to compensate for thermal drifts in the detectors. (v) Finally, a gamma correction can be obtained on the output signal. This correction is adapted to each application, depending on the nature of the observed image. The output signal is digitized in 8 bits. If used for display on a TV monitor, this signal is fed to a 1024 x 512 pixel frame memory module (designation THX 1063) which gives a direct video output.
DETECTOR CHARACTERISTICS The main characteristics of the TH 9583 solid-state linear detector are as follows: number of sensitive elements, 1024; pitch, 0.45 mm; and sensitive length, 460 mm.
LOW-NOISE SOLID-STATE LINEAR DETECTORS
273
Operating Conditions
The normal operating conditions for the detector are as follows: X-ray integration time per line, 8 ms minimum; number of lines per image, 512; number of pixels per image line, 1024; image acquisition time, 4 s; object scanning speed in X-ray beam, 6 cm s-’. (This speed may be raised at the expense of a loss in resolution in the scanning direction.) The normal X-ray beam characteristics are as follows: a tungsten anode tube using 10 kV to 160 kV peak accelerating voltage (higher accelerating voltages are possible; however, the scintillator efficiency decreases with increasing high voltage); current, 1 mA to 20 mA; and focal spot size, 0.3 x 0.3 mm2; filtering is by 2 mm Al; the source-object distance is 75 cm. Detection Performance
The detection dynamic range, defined as the ratio between the saturation level and the RMS noise level, is 4000. This dynamic range is compatible with a digital processing over 12 bits. With the pre-processing unit described above, with an 8-bit output, it is necessary to window these 8 bits inside the total available dynamic range. The electronic noise is smaller than the X-ray quantum noise. It corresponds to 2-5 X-ray photons at 60 keV. Owing to this low noise level, it is possible to detect low contrasts in objects: for example, thickness differences of 2% in 40 mm aluminium samples can be observed. Applications
The main applications for solid-state linear X-ray detectors are in X-ray non-destructive testing (NDT) systems. Here, such detectors have major advantages compared with existing systems based on radiological film or Xray image intensifiers. (i) Large field of view. The detector presented here has a photosensitive length of 460 mm, and much greater lengths can be achieved without difficulty. (ii) Rejection of scattered radiation, one of the major characteristics of these detectors. As the X-ray beam is defined by a slit, major parts of the radiation scattered by the object are rejected. This provides an improvement in image contrast. (iii) Direct, digitized video output. This signal may be exploited either for direct display on a TV monitor or for digital processing of the image data. These detectors are therefore directly compatible with automated N D T processes.
274
H. ROUGEOT ET A L .
(iv) High dynamic range, allowing object viewing in widely varying radiation conditions, for example highly absorbing zones immediately next to low absorption zones. Applications for solid-state linear X-ray detectors cover all aspects of inspection in which different thicknesses of material or different densitieshave to be detected or visualized in an object. Such applications are very numerous in non-destructive testing, including such applications as detection of foreign bodies in foodstuffs, quality control of welded metallic workpieces, structural control on composite or plastics structures. Each application calls for a definition of the optimum operating conditions for the detector in terms of incident X-ray beam energy, acquisition of reference signals, output signal processing, etc. CONCLUSION
Solid-state linear X-ray detectors are the new tools of radiological imaging. They offer numerous advantages over classical solutions and are perfectly suited for non-destructive testing by X-ray examination, especially where automated process controls are required. They can also be used in industrial scanners. Finally, additional developments can lead to dual-energy detectors, sensitive to two different X-ray ranges. Such a detection mode would considerably improve system performance, allowing different materials to be discriminated through digital processing.
An Image-intensified CCD Area X-Ray Detector for Use with Synchrotron Radiation R. H. TEMPLER, S. M. GRUNER Department of Physics, Princeton University. Princeton, New Jersey. USA
and
E. F. EIKENBERRY Departmen6 of Pathology, UMDNJ-Robert Wood Johnson Medical School, Piscataway, New Jersey, U S A
INTRODUCTION Two-dimensional area detectors capable of handling high input X-ray rates must be used in order to take full advantage of the high photon fluxes at synchrotron sources. For this reason, photon-counting devices, such as multiwire proportional counters, are not appropriate since they are countrate-limited well below the scattered X-ray fluxes encountered (Gruner, 1987; Tate et al., 1987). By contrast, many integrating detectors do not suffer from count-rate limitations and are, therefore, the natural choice for synchrotronbased diffraction and scattering experiments. Although film is undoubtedly the best known of the two-dimensional integrating X-ray detectors, two-dimensional electro-optical X-ray detectors, such as the silicon intensified-target vidicon devices described by Gruner et al. (1982), have a number of significant advantages in their use at a synchrotron source. In particular, the electro-optical detectors have a greatly increased detective quantum efficiency at low dose, wide dynamic range, and a linear dose response. Furthermore, the diffraction pattern is immediately available for inspection with a digitizing electro-optical instrument, which facilitates monitoring and interaction with the experiment. The first two generations of electro-optical detectors developed in the Princeton Physics Department used silicon diode and a silicon intensifiedtarget vidicons, respectively, and had the disadvantage of requiring extensive custom electronics. By contrast the third-generation prototype detector (Eikenberry et al., 1986), which used a charge-coupled device (CCD), utilized commercially available electronics to control it. Experience gained with this prototype aided in the design of the present CCD detector, described below, based on a newly available CCD and lens coupling. 275 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
276
R. H. TEMPLER, S. M. GRUNER A N D E. F. EIKENBERRY
THE DETECTOR
A schematic diagram of the detector system is shown in Fig. 1. Incoming Xrays were converted into visible light using an yttrium oxysulphide (P 45) phosphor. The phosphor was deposited at a density of 10 mg cm-* on a low residual radioactivity, 8 mm thick fibre-optic plate (6 pm fibres). This fibreoptic plate was in turn coupled to the fibre-optic input window of the twostage, electrostatically focused, 40 mm diameter Varo image intensifier (tube type 4214-1, S 20 photocathodes, P 20 intermediate phosphors). The tube was selected for low dark current and was distortion-corrected, having less than 3% distortion from edge to edge. Ideally this tube should have used lowpersistence phosphors, but such a tube was not available for this work. Image tubes with low-persistence phosphors will be evaluated in the future. To protect the CCD from stray electrostatic fields arising from the image tube’s internal power supply, the tube was wrapped in an earthed copper sheath. The intensifier output was imaged on a 13.8 x 13.8 mm2 CCD, using a 30 mm focal length, y 2 . 8 micro-projector lens (D. 0. Industries, Inc., Rochester, NY). The image reduction performed by this lens transformed a 1 mm displacement at the input phosphor to a 0.42 mm displacement at the CCD. The CCD was a front-illuminated, 512 x 512 pixel chip, with a 27 x 27 pmZelement size (type TK512M-31, Tektronix, Beaverton, OR). At the time of testing only a grade-3 chip was available and this was used in the
-
-
I
CRYOSTAT (-53”)
’Z
PHOSPHOR
FIBRE OPTIC
:-r
I I CCD CONTROLLER
I
IBM AT COMPUTER
TV MONITOR
FIG.1 . Schematic diagram of the CCD-based X-ray detector.
AN IMAGE-INTENSIFIED CCD AREA X-RAY DETECTOR
277
study. Tektronix is now able to supply grade-1 chips, which will be used in the X-ray detector in the near future. All the preceding components were housed within a temperature-controlled cryostat (- 53 f 0.5"C) in a dry nitrogen atmosphere. The image tube, lens and CCD were each mounted on a carrier riding on a cylindrical sled. External control of the lens and CCD carrier positions allowed magnification and focusing adjustments to be made while the detector was cold. Detector control, image acquisition, read-out and storage was performed by the CCD camera head electronics (Princeton Scientific Instruments, Inc., Model V CCD camera electronics) and associated computer (IBM-AT). A variety of CCD detector parameters, such as pixel read-out rate, on-chip amplifier control and chip format may be set with software control. The CCD was read in a slow-scan mode (1 pixel every 7 ,us; total read-out time of 5 s), which gives lower read-out noise than a video-rate scan. The on-chip CCD amplifier gain was 0.68 pV per electron; the amplifier current source was turned off during X-ray exposure. The control electronics were connected to the CCD by a short (1 m) length of cable. The video signal was digitized using a 16-bit A/D converter before being transmitted to the computer via a fast serial bus. All image measurements and manipulations took place within the computer. Images were displayed on a high-resolution monitor using a video display board (Image Technology, Woburn, MA). OF DETECTOR PERFORMANCE MEASUREMENT
A standard set of tests (Gruner and Milch, 1982) was used to assess the detector's performance. The test set was designed to obtain data of generic TABLE I Summary of detector performance characteristics Sensitivity Saturation dose System magnification Distortion Point-spread function CCD temperature Dark current Zero-dose noise (5 x 5 pixel, 10 s integration) Dynamic range for X-rays (5 x 5 pixel)
t X-rays were of 5.9 keV energy.
65 electrons per X-ray? 8 x lo6 X-ray mrn-z at phosphor 15.6 pixels mm-' at phosphor 135 prn RMSdeviationat phosphor (0.3%) 2.6 pixels (FWHM) or 170 pm at phosphor - 53°C 0.05 electrons per second per pixel 130 electrons 3
105
278
R . H. TEMPLER, S. M. GRUNER A N D E. F. EIKENBERRY
relevance to detectors used in X-ray diffraction studies; in particular it was intended to provide an objective means of comparing instruments. In this work, the reported measurements are sensitivity, linearity of response, uniformity of response, spatial distortion, spatial resolution and noise. Data from these measurements are summarized in Table I. Sensitivity A source of reasonably monochromatic radiation was required in order to evaluate the detector's sensitivity to X-rays. We used a well-collimated 55Fe source, providing a beam roughly 4 m m in diameter. The source was calibrated using a NaI scintillation counter, beryllium and mylar windows being interposed in order to mimic the input window of the cryostat. The total X-ray flux was 10267 'r 33 counts s-'. The CCD-based X-ray detector allowed software selection of two settings of the amplifier gain, denoted as low and high gain, prior to digitization of the signal. Overall system sensitivity was determined by summing the analogueto-digital converter units (ADU) over the entire X-ray spot image. At low gain, each X-ray yielded 1.6 ADU and at high gain each X-ray gave 4.6 ADU. Using the manufacturer's specifications it was estimated that this represented approximately 65 electrons for each incident X-ray. In all the measurements that follow (with the exception of the linearity of response test) the high gain setting was used.
Linearity of Response
The detector was shown to be linear within experimental error over a dose range of 5 x lo3 to 5 x lo6 X-ray mm-* by varying the exposure time to the 55Fesource. Saturation occurred at roughly 2 x lo4ADU per pixel (low gain) which was equivalent to 8 x lo6 X-ray mm-2 at the phosphor. It should be noted that in practice one would use the CCD only up to 75% of its saturation capacity in order to avoid high-dose problems, such as charge bleeding between pixels. Uniformity of Response
The uncorrected uniformity of the detector was measured by illuminating the input phosphor with a uniform X-ray source. The source was a small Xray generator 75 cm from the phosphor. A contour map of the response across the entire CCD area is shown in Fig. 2. The non-uniformity is stable and can be readily corrected for in subsequent computation. Most of the nonuniformity arises in the image intensifier, which was optimized for low distortion at the expense of uniformity.
AN IMAGE-INTENSIFIED CCD AREA X-RAY DETECTOR
279
500
400
300 200 100
0 0
100
200 300 row (pixel no.)
400
500
FIG.2. Contour map of detector responseto uniform X-ray illumination. Contour levels are in ADU, with dashed lines representing 10 000 ADU intervals.
Spatial Distortion
Information on the spatial distortion of the detector was obtained by uniform X-ray illumination of a hexagonal matrix of holes in a metal plate. The projected X-ray spots were 3 10 pm in diameter and spaced on 2.08 mm centres. To determine the detector spatial distortion, we compared the hexagonal pattern of spots before and after passing through the detector. The centroids of 208 spots in the CCD image were found and the locations were used to construct a least-squares best fit hexagonal lattice to the array. This was done using the normal equations for the coordinates of the lattice origin and the coordinates of the unit cell vector. From the unit cell vector length determined in this way, the average system magnification was found to be 15.6 pixels mm-’, referred to the input phosphor. In Fig. 3, the spatial distortion of the detector is illustrated by overlaying centroid spots for the ideal (or input) hexagonal lattice and the observed centroid positions. The RMS deviation from the ideal lattice corresponded to 140 pm at the phosphor or 0.3% of the full input screen diameter. This small distortion is stable and can be corrected by computation, if required. From previous observations, we believe the distortion is principally due to the
280
R . H. TEMPLER, S. M. GRUNER A N D E. F. EIKENBERRY
500
. .
1
.
.
_
.
+
_
. . . . ., ,
,
-
-
: .
. _
I
. -
.
I _
I
.
I
““1..- , .. . .. . . . ’. . .. . .. , . , - I . , , . . . .. . . . ‘
’
0 1 ’ I ’ 1 . I 0 100 200 300 400 500 I
,
.
,
,
1
’
row (pixel no.) FIG.3. Detector distortion. Centroids of an ideal hexagonal grid are superimposed upon measured centroid positions.
electron-optics of the image intensifier tube (Gruner and Milch, 1982; Gruner et al., 1982). Resolution
The data from the hexagonal grid exposures were used to determine the point-spread function (PSF) for the detector. Since every spot in the array was both well resolved (up to levels where some “blooming” occurred) and highly symmetrical, it was possible to create an average spot. This was done by 1.o
0.8 0.6 0.4
0.2 0.0 0
1 2 radial distance (pixel)
FIG.4. Point-spread function of the detector.
3
AN IMAGE-INTENSIFIED CCD AREA X-RAY DETECTOR
28 1
translation of each spot to a common origin using the previously determined centroid positions. This composite spot was cylindrically averaged and deconvoluted for the 310pm projected hole diameter to obtain the PSF (Fig. 4). The PSF has a full width at half-maximum of 2.6 pixels, corresponding to 170 pm at the X-ray phosphor.
Measurement Precision
In quantitative measurements of X-ray diffraction data, the experimenter must measure the X-ray dose incident within the area of interest on the input phosphor. The accuracy of the dose determination depends upon the noise introduced by the detector. The detective quantum efficiency D may be used to express the efficiency with which the X-ray detector passes dose information through the system (Gruner et al., 1978). The detective quantum efficiency is given by
where S is the signal, a is the RMS noise in the signal and i and o subscripts pertain to the detector input and output respectively. For an ideal detector D is 1 and, with the addition of degradation by detector noise, D decreases. Measurement of the uncertainty p in the X-ray dose,
is a more intuitive means of quantifying a detector's efficiency. Given the condition that the incident X-ray signal obeys Poisson statistics the two are related by
In general p and D will depend on the dose and a number of variables which depend on the detector. For our system we expect the noise would be a function of integration area, dose rate, detector temperature, and detector read-out rate. In this work we report measurements of p as a function of dose and integration area, all other parameters being held constant. Estimates of the noise were obtained by taking a sequence of five exposures of equal duration of the 55Fesource along with five paired, contemporaneous background exposures. The background-subtracted outputs of each exposure were then used to determine image-to-image RMS variation, ao,in the signal,
282
R. H. TEMPLER, S. M. GRUNER AND E. F. EIKENBERRY
from identical series of boxes of fixed area. It was found that the data could be fitted by an equation of the form 6, =
,Js;+ p.
The zero-dose noise p was found from the RMS variation in shortbackground exposures and a was found from a least-squares fit to the RMS variation in the data for exposure at various durations. The time dependence of p was ignored since at - 53°C background (dark current) for this detector required more than 60 hours to reach saturation. In Fig. 5 curves are presented describing the uncertainty in determination of dose for 5 x 5 and 15 x 15 pixel areas of integration. The 5 x 5 pixel area, corresponding to 0.095 mm2 at the phosphor, was chosen because the PSF spreads each X-ray event over this area. (Smaller regions show “better” than ideal accuracy because of smoothing of the data due to the PSF.) For a 5 x 5 pixel region at 109 X-ray mm-2 s-’ a minimum of 2 X-rays can be measured at a signal-to-noise ratio of 1. With the upper end of detection set at 75% of saturation, one would be able to measure 6 x lo5X-rays, giving a dynamic range for X-rays of 3 x lo5in the 5 x 5 pixel area. The dynamic range for Xrays in a 15 x 15 pixel area (0-85 mm2) is estimated to be lo6.
-
-
N
10,
0.001 ! 1o2
I
I
1o3
1o4
total dose (x-rays)
.
1o5
FIG.5. Fractional uncertainty in dose measurement. The CCD-based detector is compared to X-ray film over two areas, 0.1 and 0.9 mm2. The film behaviour is calculated by assuming a chemical fog level of 105grains mm-2, that 60%of X-rays are absorbed, and that each absorbed X-ray produces one darkened film grain.
AN IMAGE-INTENSIFIED CCD AREA X-RAY DETECTOR
283
SUMMARY
The SIT vidicon and CCD X-ray detectors are quantum-limited devices (their noise response is virtually identical), but the CCD achieves this with much lower front-end gain. This is possible because of the inherently low readout noise of the CCD (13 electrons per pixel at -54°C). The low read-out noise and deep wells of the CCD combine to enhance the dynamic range of this detector over the SIT vidicon by a factor of 300. Because of the use of a stage of intensification, neither the uncorrected spatial distortion nor uniformity of response has been improved in the CCD detector. However, a factor of 2 improvement in resolution at the phosphor has been achieved and the dark current is 300 times smaller than in previous instruments, allowing day-long exposures. Performance characteristics aside, the operational improvements using a CCD for read-out have been significant. The detector electronics are more straightforward (the SIT tube requires multiple, dynamically stabilized voltages), there is less susceptibility to mechanical vibration, read-out and preparation are more rapid, the entire detector apparatus is more compact than the SIT system, and the cost is significantly lower. In summary, the CCD X-ray detector is a useful quantum-limited device both for use with a standard X-ray source and at a synchrotron source where the very high count-rate limitations prevent the use of photon-counting devices. ACKNOWLEDGEMENTS We thank Princeton ScientificInstruments for the CCD controller, for the generous loan of the CCD chip used in these tests, and for the help given throughout in the development of the detector. We are further indebted to Professor George Reynolds, Mr Paul Botos and Mr David Turner for their help and advice. The detector development was supported by the US Department of Energy under grant DE-FG02-87ER60522. Partial funding was also received from NIH grant AR34235 and from the New Jersey Center for Advanced Biotechnology and Medicine.
REFERENCES Eikenberry, E. F., Gruner, S. M. and Lowrance, J. L. (1986). IEEE Trans. Nucl. Sci. 83,542-545 Gruner, S. M. (1987). Science (in press) Gruner, S. M. and Milch, J. R. (1982). Trans. Am. Crystallogr. Assoc. IS, 249-267 Gruner, S. M., Milch, J.R. and Reynolds, G. T. (1982). Rev. Sci. Instrum. 53, 1770-1778 Tate, M. W., Gruner, S. M., Shyamsunder, E. andd’Amico. K. L. (1987). BUN.Am. Phys. SOC.32, 843
This Page Intentionally Left Blank
Further Developments of an X-ray Television Detector U. W. ARNDT MRC Laboratory of Molecular Biology, Cambridge, England
and G . A. IN’T VELD B. V. Enraf-Nonius, Delft. Netherlands
INTRODUCTION
The electronic area detector used in the Enraf-Nonius FAST (fast-scanning area-sensitive television) diffractometer is an analogue detector: the numerical output produced by the detector is a measure of the incident X-ray intensity but this output is not expressed directly in terms of X-ray photons per unit area per unit time. The purpose of this paper is to explain the functioning of the detector and the way in which its electronic performance is optimized for the particular crystallographic measurements to be made with it. The detector is shown in schematic form in Fig. 1. An X-ray diffraction pattern, such as that from a rotating single crystal, is recorded on an X-ray phosphor which is viewed by an intensifier-low-lightlevel television camera. The design concepts have been discussed by Arndt and Gilmore (1979a,b) and by Arndt (1982). The standard thickness of the Gd202S X-ray phosphor is 10mgcm-2. When the detector is to be used predominantly with radiation of wavelength less than 1 A, the deposit thickness can be increased to at least 30 mg cm-2 with a consequent gain in absorption at the cost of a small loss in spatial resolution. The fraction of Xrays absorbed in these layers is shown in Fig. 2. The necessary demagnification between the 80 mm diameter input screen and the 18 mm diameter silicon intensifier target (SIT) TV camera tube is provided in part electron-optically in the image intensifier and in part by the tapered fibre-optics cone. The TV camera is operated in accordance with the European interlaced 625-line standard; the duration of each line is 64 ps, 5 1.2 p s of which constitute the active line period, the remainder being needed for the line fly-back. The active frame consists of 512 lines and each active line is sampled at 512 points, that is, at a rate of 10 M-samples s-l by means of an 8-bit flash ADC. The digitized output is integrated in an 18-bit-wide mass
28 5 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
u. w. ARNDT
286
AND G.
A.
IN’T VELD
Gd,02S Phosphor
FO disc
I
Image Intensifier
*,
Amplifier
Clock
Generator
i&
C0”IlDl
ConlrOl D l 5,”.1,1 (lh.Il
m0101.
.,5
OP Disc
FIG.1. Schematic illustration of the FAST detector. The detector is mounted on the arm of a single-crystal diffractometer. It can be translated along the arm which, itself, can be swung in the horizontal plane about the specimen.The latter can be rotated about three non-orthogonal axes.
store. As the crystal rotates about one or more axes different X-ray spectra, or “reflections”, are recorded; their positions on the detector face and the angular coordinates at which they occur can be calculated from the unit cell of the crystal and from its initial orientation. A typical data collection run from a large-unit-cell crystal, e.g. from a protein crystal, involves a rotation of about 100” during which several hundreds of thousands of reflections have to be measured. The integrated intensities of the reflections are collected from the mass store by the associated computer and either the live or the integrated pattern can be displayed on a TV monitor.
FURTHER DEVELOPMENTS OF AN X-RAY TV DETECTOR
+*
pq
287
+*
@
100 CA 0 ”
cu
‘
2
CL
8
90 80 70 60 50
40
9
30
&“-
20
0.5
1.0
1.5
2.0
h (A”)
FIG.2. Absorption of X-rays in the input phosphor as a function of wavelength.
DEFINITIONS Pixel (picture element, abbreviated px): an element of the 2-dimensional area representing the digitized image (51 2 x 512 points). The size of the active area of the detector is 64 mm (vertical) x 48 mm (horizontal). The nominal pixel size, therefore, is 0.125 x 0.094 mm2. MSU (mass store unit): this is the digital unit in which the mass store contents are expressed as seen by the host computer. Of the total mass store word width of 18 bits, the 16 most significant bits are available to the host computer. A full-scale output from the ADC thus results in the addition of 64 MSU per frame into the mass store. Frame period (abbreviated fp): time period of 40 ms, being the time to scan one complete image; each pixel value is read out once a frame. Thus the frame frequency is 25 Hz. Dark current: the current from the camera tube in the absence of light on the input screen of the tube. This current originates in the leakage of the target diodes in the camera tube. The statistical fluctuations of this leakage current contribute to the total detector noise. Equivalent background illumination (EBI) : the small output signal of the image intensifier in the absence of an input due to the random emission of electrons from the photocathode which becomes visible in the final output as individual scintillations. The time-integral of these scintillations contributes to
288
u. w.
ARNDT AND G. A. IN’T VELD
the total offset which is removed in difference measurements. ThefIuctuations in this signal contribute to the total detector noise and are discussed further below. ADC ofset: DC offset is added to the signal before it is digitized in the analogue-to-digital converter (ADC). The reason for this offset is that noise spikes in the signal may be of either polarity, so that at small signal levels the peak of these spikes may be positive. The ADC gives a zero digital value for a positive analogue input; a zero offset would thus lead to a “rectification” of the noise instead of a long-term cancellation of positive and negative noise spikes. The DC offset, which does not contribute to the noise in the detector output, is eliminated during processing. DETECTOR SENSITIVITY The gain of the video amplifier that precedes the ADC is factory-adjusted to give a 2.0 V output for an output current of the camera tube of 300 nA. This corresponds to the tube manufacturer’s specification of a “peak-white” signal, i.e. the top end of the linear range of the camera tube transfer curve. The X-ray sensitivity of the FAST detector can be varied in three ways under computer control. (i) By varying the light gain of the image intensifier. This is done by changing the high voltage applied to the intensifier in steps between 4.5 kV and 15 kV. The ratio of the gains at maximum and minimum settings is about 15. (ii) By varying the gain of the intensifier stage of the SIT TV camera tube by changing the high voltage applied to the camera tube in steps between 3 kV and 10 kV. The gain varies by a factor of about 100 between the minimum and maximum settings. (iii) By varying the reference voltage, of the ADC and thus its span. The sensitivity of the FAST detector for a given combination of the three parameters is expressed in MSU per X-ray photon. In principle, the sensitivity of a given detector can be varied from a maximum of about 2.5 MSU per keV of photon energy to about 1/1000 of this value, but in practice, for precision measurements, the choice of values is restricted. The image intensifier high voltage should be chosen so as to minimize the ratio of total electronic noise to sensitivity (see below). SIT tube high voltages below 2 kV should be avoided in the interest of maximum stability. The ADC span should normally have a value of 2 V for a full-range signal current of 300 nA. The sensitivity of the detector is measured with the help of a radioisotope Xray source (e.g. 55Fe) masked with a lead plate with a small (0.1-1 mm diameter) pin-hole rigidly attached to the source. The number of quanta emerging through the pin-hole is first measured with an X-ray scintillation counter of known efficiency.The source is then mounted in front of the FAST detector and the intensity over the entire spot image in MSU is measured by
FURTHER DEVELOPMENTS OF AN X-RAY TV DETECTOR
289
integrating for a known time. The sensitivity is defined in terms of absorbed photons: it is quite different from the absorbed fraction shown in Fig. 2. DETECTOR NOISE
The crystallographer using the FAST system is primarily interested in the total noise of the detector. This may be determined by carrying out a large number ( > 20) of integrations in the absence of X-rays and calculating the standard deviation from the mean. Each integration should be over a box size equal to that used for a typical diffraction spot (10 x 10 pixels). The noise is approximately Poissonian, so that the standard deviation is proportional to the square-root of the integrating time. As a result of spatial variations of the dark current and equivalent background illumination (EBI), the standard deviation depends on the selected position in the image; because of spatial correlation between neighbouring pixels due to a finite point-spread function, it also depends on box size and box shape. The individual components of the noise must be known to appreciate their relative importance and the way in which detector performance can be improved, e.g. by cooling. These components are (i) amplifier noise; (ii) statistical fluctuations in the TV camera tube dark current; (iii) statistical fluctuations in the dark emission of the image intensifier; (iv) large-amplitude scintillations corresponding to an equivalent incident energy of about 1 MeV, probably due to cosmic radiation and radioactivity in the first fibre-optics face-plate. Provided that the thorium content of the glass from which this face-plate is made is less than 1 part in lo6,these are rare and random events, but they may result in a very occasional “rogue” measurement. Our worst, and unuseable, detector exhibited about lop4mmP2s-’ of these scintillations: this would correspond to a thorium content of 250 parts in lo6 in the face-plate. The manufacturer’s analyses of a recent glass sample was only 5-20 parts in lo6.We are investigating this matter further. Finally (v) synchronized electrical interference resulting in a fixed-pattern noise was formerly quite serious since even a very small synchronous signal is integrated. This is now reduced to negligible proportions by improved grounding and shielding and is almost completely eliminated in difference measurements. The first three of these components are the most important. They are Poissonian and add in quadrature. They will be discussed below. INDIVIDUAL NOISE COMPONENTS Amplijier Noise
Amplifier noise is largely due to thermal noise in the pre-amplifier. The signal-to-noise ratio of the amplifier should be at least 50 dB, i.e. the RMS
290
u. w.
ARNDT AND G. A. IN'T VELD
noise is about 1 nA fp-i px-f =0.22 MSU fp-f px-;. This noise is Poissonian, so it will increase with the square-root of the integration time and of the number ofpixels in the integration box. It can be measured by determining the standard deviation from the mean of a large number of integrations for a given number of frame periods, with the scanning beam of the camera disabled. In principle the amplifier noise is proportional to the square-root of the absolute temperature, so that it cannot be reduced very much by a moderate amount of cooling. Shot Noise in the SIT Tube Dark Current
The RMS shot noise for a dark current of Idk amps is 2eI&B, where e is the A at electronic charge = 1.6 x C, Idk is the DC dark current (- 12 x 35"C), and B is the Bandwidth= 5 x lo6Hz. This gives a shot noise of 0.14 nA=0.03 MSU fp-: px-f. Recently supplied RCA SIT tubes exhibit a dark current lower by a factor between 2 and 3; in addition, the detector is now operated at about 15"C,so that this noise is now negligible.The shot noise and the amplifier noise add in quadrature so that the shot noise may be measured by determining the total noise with the scanning beam enabled but with the image intensifier high voltage off and by then correcting for the amplifier noise. The dark current is strongly dependent upon the temperature (see below), and hence the shot noise is reduced by cooling. Fluctuations in the Image Intensi$er Dark Emission The dark emission, usually expressed as an equivalent background illumination (EBI), is different for individual image intensifiers; it varies with position on the detector, being highest near the centre; and it is non-linearly dependent on the high voltage. The intensity may be measured by carrying out a background-subtracted integration for a period of several minutes in the absence of X-rays; the background is determined with the image intensifier high voltage switched off. A knowledge of the EBI is useful as a test of the image intensifier. The statistical performance of the detector is determined only by theJuctuations in this intensity. The latter can be derived from the total noise variance, i.e. the square of the standard deviation, measured as described above, by subtracting the variance due to dark current and amplifier shot noise. The mean amplitude of a dark current scintillation appears to vary more rapidly with high voltage than does the mean amplitude of an X-ray photon flash. At maximum voltage the average scintillation produces a signal equivalent to that from a photon with an energy of about 10 keV, at lower voltage it is the equivalent of only about 0.5 keV. These values are consistent with the assumption that the EBI is due to the emission of single electrons
29 1
FURTHER DEVELOPMENTS OF AN X-RAY TV DETECTOR
from the photocathode at low gain and to multiple-electron events at high gain. It will be apparent that the scintillations have a smaller relative effect when working with harder X-rays. DETECTIVE QUANTUM EFFICIENCY
Typical experimentally determined noise components are summarized in Table 1. At low to medium intensifier voltages, the noise is dominated by that due to the amplifier. In the future we intend to replace the SIT tube with a solid-state television sensor that is capable of being operated with a very much lower read-out noise. Only then will it become worthwhile to cool the image intensifier photocathode. The parameter which determines the detector performance is the ratio of the total noise standard deviation (in MSU fp-fpx-f- ) to the sensitivity in MSU per X-ray photon. This ratio is the standard deviation expressed in equivalent X-ray photons fp-f px-f and is denoted by 0. Its numerical value is inversely proportional to the photon energy. As discussed above, cr varies over the image surface but we can take as a comparison standard the value of cr measured 5 mm from the centre of the detector using a 10 x 10 box size. The defective quantum efficiency (DQE) E of a photon detector is defined as the ratio of the signal-to-noise ratio at the output to that at the input. Since the input signal-to-noise ratio is determined by photon counting statistics, E can be expressed as E
= (Rn) -
',
where R is the relative variance in the output signal produced by n photons, that is, the ratio of the variance to the square of the signal. An analogue detector with a DQE of E thus behaves like a perfect Poissonian detector, or ideal counter, which responds to a fraction E of the incident photons. TABLE I Noise, DQE and dynamic range Detector sensitivity MSU photon-' Amplifier noise (MSU fp-f px-;) Dark current shot noise (1 5OC) (MSU fp-f yx-f)
Intensifier EBI fluctuations MSU fp-r px-' Total noise (MSU fp-l px-l) Total noise (a)(photons fp-r px-f) Count rate for c = O S (photons s - ' px-') Max count rate (photons s- * px-I) Dynamic range
20
10
1
0.22 0.01 0.01 0.55 0.028 0.019 80 4200
0.22 0.01 0.05 0.23 0.023 0.013 160 12300
0.22 0.01 0.00 0.22 0.22
0.1
0.22 0.01 0.00 0.22 2.2 1.21 121 1600 16000 1300 132
292
U. W. ARNDT AND G. A. IN'T VELD
>. C
.-U E L
w
Count Rate [Photons p x - l s - ' )
FIG.3. Detective quantum efficiency as a function of count rate at a detector sensitivity of 10 MSU photon-'.
For the FAST detector, assuming that the photon variance and the noise variance are additive, we can write R = (1
+02/An)/An
and
E =A / ( 1
+rr2/An),
where n is the incident number of photons fp-' px-' and A is again the fraction of absorbed photons; E thus varies with the count rate. Figure 3 shows E for Cu K, X-rays as a function of count rate as measured for one particular FAST detector when operated at the optimum values of high voltage. The count rates at which E =0.5 are given in Table I together with the maximum count rate possible at the different detector sensitivities. The TABLE I1 Background count rates from a typical protein crystal
Source
Count rate (photons px-' s-I)
800 W microfocus tube 1600 W fine focus tube 5000 W rotating anode tube Synchrotron beam line
0.02 0.05 0.15 20
FURTHER DEVELOPMENTS OF AN X-RAY TV DETECTOR
293
maximum rates are those which produce a quantity of 64 MSU px-' fp-I. These count rates may be compared with the mean background intensities for a typical protein crystal with an 80 A unit cell for different X-ray sources, as indicated in Table 11. The weakest X-ray reflections that must be measured are only a little above background level; the strongest are more than a thousand times more intense. These figures show that from the point of view of noise and counting statistics, the FAST detector behaves almost like an ideal detector for the more intense sources; they also show that for more weakly diffracting specimens (smaller crystal size or larger unit cells), intense X-ray sources are essential.
EFFECTOF TEMPERATURE ON DETECTOR PERFORMANCE The dark current of the SIT tube is reduced by cooling; on older SIT tubes it was about 12 nA (-2.6 MSU fp-l px-I) at 35°C and less than I nA (-0.2 MSU fp-I px-l) and 0°C. On more modern tubes operated at 15"C, the dark current is about 2 nA. The statistical fluctuation in the dark current ("shot noise") is proportional to the square-root of the current: the standard deviation due to this component can, therefore, be expected to vary by a factor of 4 between 35" and 0". Since this component is only a small part of the total noise, the effect is small. However, it is essential for the DC offset to remain constant during the integration of one frame or during the scanning of any given reflection, otherwise large errors may be introduced in the background correction. It is, therefore, more important for the temperature to be stable than for it to be particularly low, and the temperature of the detector is now thermostatically controlled to & 0.2"C. The amplifier thermal noise is proportional to the square-root of the absolute temperature, so that this type of noise can only be reduced by a few per cent by the amount of cooling that is possible in practice.
MAXIMUM COUNTING RATEAND DYNAMIC RANGE The maximum rate at which the television detector can operate, that at which the ADC produces an output of 64 MSU px-l in one frame period, is inversely proportional to the chosen sensitivity. We have seen above that the minimum sensitivity useable in practice is about 0.02 MSU (8 keV per photon). The maximum counting rate, therefore, is 64/0.02 = 3200 photons px-' fp-'=8 x lo4 photons px-' s-I. It is correspondingly lower for higher-energy X-ray photons. Unlike, for example, a multi-wire chamber, the television detector could function at the maximum rate simultaneously in all of its 262 144 pixels. The limitations are local and not global count rates. In addition, overflows of the ADC affect only those pixels where they occur: it is
294
u. w.
ARNDT AND G. A. IN’T VELD
perfectly possible to collect accurate single-crystal data even though the detector saturates on a few of the strongest reflections. In the interests of achieving the highest detective quantum efficiency for the weakest reflections, the detector should always be operated at the highest sensitivity that avoids saturation on the strongest reflections it is desired to measure. The dynamic range of the detector may be defined as the range of counting rates over which the DQE has an acceptable value: we shall arbitrarily set this minimum value as 50%. The dynamic range defined in this way is shown in Table I. SPATIAL DISTORTION
The output image of the television detector is distorted by imperfections in the electron-optical and the light-optical components and the TV camera raster scan. These distortions are very stable with operation of the detector under constant conditions, so that an infrequent distortion calibration is all that is necessary. However, they vary slightly with intensifier and camera tube high voltage; because of local variations in the external magnetic field they may vary with position of the detector on the detector arm and with the swing angle of this arm. For the highest accuracy the distortion calibrations must, therefore, be carried out at the position at which the detector is used for data collection and at operational voltages and temperature. Calibration is carried out with the help of an accurate shadow mask fixed in front of the detector. A material which fluoresces strongly in the incident X-ray beam, i.e. one which has an excitation potential just below the photon energy of this beam, should be mounted in the crystal position to act as an isotropically emitting source of radiation. Suitable materials are compounds of iron for Cu K, radiation and of strontium for Mo K, radiation. The positional variation of the distortion is greatly reduced by demagnetizing those parts of the detector arm that are made of hard magnetic material (the rails), but the variation can never be totally eliminated. NON-UNIFORMITY OF RESPONSE The measurements made with the detector must be corrected for spatial variations of its response, which may amount to a factor of 2 between the centre and the corners of the image. The major cause of the non-uniformity is a variation in the pixel size as a result of spatial distortions. A fluorescent material mounted in the sample position is a suitable isotropic source for this purpose also. Corrections must, of course, be made for the variation of the path length with position, which causes an inverse-square-law variation of intensity and a change in the air absorption. Oblique-incidencecorrections are
FURTHER DEVELOPMENTS OF AN X-RAY TV DETECTOR
29 5
not necessary as they affect calibration and data equally for the same detector position. Polynomial correction functions have recently been found preferable to look-up tables for both types of calibration.
SPATIAL RESOLUTION The resolution of the composite detector is characterized by its pointspread function. This function is very nearly circularly symmetrical, that is, the resolution along the television lines is approximately equal to that perpendicular to the lines. The FWHM of the point-spread function is about 2 pixels, but the curve is far from Gaussian and has long tails at the 2% level where the full width is about 10 pixels. It is thus easy to resolve neighbouring spots visually, but for the quantitative measurements of close neighbours, particularly of weak reflections close to strong ones, the experimental pattern must be deconvoluted using precisely determined point-spread function curves. A program for the profile analysis of diffraction spots is in course of development (Thomas, 1986). The X-ray phosphor and the SIT tube contribute to the broadening of the point-spread function in about equal measure; the modulation transfer functions of the de-magnifying image intensifier and of the fibre-optic coupling extend to considerably higher spatial frequencies than those of the phosphor and the SIT tube. QUALITY OF DATA Many hundreds of data sets have been collected from protein crystals and several structures have been solved using the FAST diffractometer in conjunction with stationary anode and rotating-anode X-ray generators and with synchrotron radiation beam lines. Under the best conditions, the quality of the data, as determined by internal and external consistency, is close to that to be expected from counting statistics and a precision of better than 2% has been achieved in intensity measurements. Routine measurements from very weakly diffracting materials are still limited in precision by factors other than the characteristics of the detector itself. Further software improvements are in progress to improve the refinement of the crystal orientation, and of its lattice parameters, the analysis of the reflection profiles and the correction for distortion and non-uniformity of response. These problems are shared by all electronic area detectors and are being studied intensively (Filhol et al., 1986). The use of these detectors is bringing about a revolution in crystallographic data collection procedures,
296
u. w.
ARNDT AND G. A. IN’T VELD
especially for macromolecular crystals where they are speeding up data collection by factors of more than 100.
REFERENCES Arndt, U. W. (1982). Nucl. Instrum. Methods. 201, 13-20 Arndt, U. W. and Gilmore, D. J. (1979a). J. Appl. Crystnllogr. 12, 1-9 Amdt, U. W. and Gilmore, D. J. (1979b). In “Adv. EEP” Vol. 52, pp. 209-216 Filhol, A,, Maier, B., Mason, S. A., McIntyre, G. H., Roth, M., Lewit-Bentley, A., Bentley, G. A., Bricogne, G. and Kahn, R. (eds.) (1986). Proc. Internat. Workshop on Evaluation of SingleCrystal Diffraction Data from 2-D Position-Sensitive Detectors. J. de Physique 47, Coll. C5, Suppl. No. 8 Thomas, D. J. (1986) J. de Physique 47, Coll. C5. Suppl. No.8, pp. 69-73
Development of Large-format Photon-counting Array Detectors for the Lyman Ultraviolet Space Telescope E. H. ROBERTS Auspace Limited, Canberra, ACT, Australia
I. R. TUOHY and M. A. DOPITA Mount Stromlo and Siding Spring Observatories. Woden. ACT, Australia
INTRODUCTION The Lyman mission will open up to detailed study the as-yet largely unstudied but scientifically critical window of the electromagnetic spectrum between 90 and 120 nm. In order to obtain the maximum output from the mission, high-performance imaging detectors are required. This paper examines the application of the large-format photon-counting array detector technology developed at Mount Stromlo and Siding Spring Observatories to the Lyman requirements. THE LYMANMISSION The Lyman programme will involve the flight of a grazing-incidence telescope and up to three spectrographs, on a dedicated spacecraft in high Earth orbit. Pending mission approval by ESA, Lyman is scheduled for launch in 1995 on an ESA Ariane rocket. The programme is currently undergoing parallel phase A studies within ESA and Australia. The latter work is funded by the Australian Department of Industry, Technology and Commerce. Studies have also been performed in the United States, but the level of participation by NASA is currently uncertain in the wake of the Challenger Shuttle disaster. The prime spectrograph will have a resolution of 30 000 over the range 90120 nm, with a simultaneous coverage above 10nm. It may be either a Rowland circle spectrograph, with a single optical component, or an Cchelle spectrograph, with three. The construction of the two spectrographs, and their performances, are quite different, as are their detector requirements. In addition to the prime spectrograph, two additional spectrographs will be flown, to cover the EUV range from 10 to 90 nm, and the FUV range from 120 to 200 nm. The former will allow overlap with the near X-ray range studied by 297 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 14
Copynght 0 1988 Academic Press Limited All nghts of reproduction in any form reserved ISBN 0-12-014674-5
298
E. H. ROBERTS, I. R. TUOHY AND M. A. DOPITA
the ROSAT and EUVE satellites. The latter will allow overlap with the Hubble Space Telescope, and comparison of results with the International Ultraviolet Explorer, which is still providing important data well after the expiration of its design life. Lyman Astronomy
The ultraviolet waveband between 90 and 120 nm is important because it contains the Lyman series of hydrogen and deuterium, the Lyman bands of their molecules and a host of resonance lines covering a wide range of ionization conditions such as CIII, NI, NII, NIII, NV, OVI, OIV, SIII, SIV, SVI, AlI, A111 and FeIII. Hydrogen is the most abundant element in the universe. The ratio of hydrogen to its isotope, deuterium, is an important clue to the conditions of the Big Bang, the birth of the universe. This can be determined directly using the R 30 000 prime spectrograph of Lyman. Lyman can observe processes in plasmas over the range of temperatures lo3 to 5 x lo7 K, including targets from the nearby planets to the most distant quasars. The number of spectral absorption lines in the prime wavelength range is far higher than in the FUV or the visible. This in turn means that far more information is available to astrophysicists to determine the nature of processes in astronomical targets.
-
Lyman Design
The baseline Lyman design consists of a 0.8 m diameter, 1.7 m focal length, f l l 0 grazing incidence telescope, directly feeding a high-resolution spectrograph. The secondary EUV and FUV instruments will be separately selected by a focal plane mirror mechanism. The performances of the instruments are indicated in Table I. The instrument package will be built onto a bus derived from an earlier ESA TABLE I Lyman instrument requirements EUV
Prime
Wavelength coverage (nm) 10-30t 90-120 Simultaneous coverage (nm) $ 33 Spectral resolution (dA/A) 300 30 000 Slit length (arc-sec) >I0 >to Minimum sensitivity (an2) 5: > 10
t With extension to 90 nm desirable. $ Best effort.
FUV
120-200 $ 10 000 >10
1
LARGE-FORMAT PHOTON-COUNTING ARRAY DETECTORS
299
i
36s-
FIG. I . Layout of the Lyman spacecraft within an Ariane Type 01 fairing. The Rowland spectrograph is shown in side elevation below the grazing incidence telescope assembly.
mission such as SOHO. The satellite will be launched as the minor partner of a dual launch into geostationary transfer orbit, in order to save launch costs. From there it will be boosted into a highly elliptical orbit which will allow excellent observing efficiency for users. The overall layout of Lyman, assuming a Rowland spectrograph, is shown in Fig. 1. Lyman Detector Requirements
ESA has specified the use of low-noise photon-counting detectors for Lyman. These are required to have a two-dimensional format and up to one million pixels, with a pixel size of less than 25 pm. Such detectors are required for the prime, EUV and FUV spectrographs. We have also investigated the use of photon-counting detectors for the fine error sensors in Lyman, in contrast to the standard cooled-CCD arrays.
300
E. H. ROBERTS, 1. R. TUOHY AND M. A. DOPITA
TABLE I1 Lyman baseline detector requirements Prime Rowland Wavelength sensitivity Quantum efficiency Count rate Noise rate Pixel size Format Prime Ccheiie Format Other performances
FV V Cchelle Wavelength sensitivity Format Other performances EVV detector Wavelength sensitivity Format Pixel size Other performances
90- 120 nm 30% goal, 20% required >0.3 Hzpixel-' los counts s - ' total Hzpixel-I < 25 x 25 pm2 8 192 x 64 pixels 26 x 26 mm2 1024 x 1024 pixels As above 120-200 nm 26 x 26 mmz 1024 x 1024 pixels As above 10-30 nm 80 mm x 3.5 mm2 2048 x 64 pixels < 50 pm As above
The nature of the prime wavelength detectors depends on the technology used for the spectrograph. The Cchelle spectrograph leads to a compact focal plane format and a relatively straightforward detector requirement. The Rowland, however, results in an extended linear format, with a 2.1 m diameter curvature and 3 mm image height, up to 450 mm long. Such a spectrograph requires a novel detector design. The Rowland spectrograph is of interest in spite of its focal plane format because it offers a much higher efficiency than an Cchelle spectrograph in the waveband of interest. The EUV and FUV detector requirements are more straightforward. The detector requirements are summarized in Table 11.
LARGEFORMAT PHOTON COUNTING ARRAYDETECTOR We propose the application of the large-format photon-counting array (LFPCA) detector to the Lyman mission. These detectors have high
LARGE-FORMAT PHOTON-COUNTING ARRAY DETECTORS
30 1
resolution, low noise and a long history of use on ground-based telescopes. They are modular and multichannel, making them well suited to space missions requiring good reliability and low cost. An ultraviolet version of the detector is currently being prepared for a flight on a Space Shuttle, and has already successfully undergone vibration and thermal cycling to conditions more extreme than those expected in a space mission. Operation of LFPCA Detectors
The operation of LFPCA detectors has been described in earlier papers (Gorham et al., 1982; Roberts et al., 1986), and is briefly reviewed here. The LFPCA is an intensified CCD system. Incoming photons liberate photoelectrons at a photocathode, which are amplified in a microchannel plate array and then reconverted to light at a phosphor. The light is coupled into chargecoupled devices (CCDs) via coherent fibre-optic bundles. The output of the CCDs is digitized and stored in a high-speed frame memory. The incoming video data are subtracted from data in the previous CCD frame, to give a difference signal. This signal is applied to comparators, and events within a window are counted as photons. Photon events are then centroided in high-speed, dedicated electronics, in order to locate the centre of a “photon event” to one-quarter of a photosite of the CCD. The CCD photosite dimensions are (in the current LFPCA detectors) 30 x 36 pm’ ( W x H ) , giving a final pixel size of 15 x 18 pm2. This information is fed to a high-speed, autoincrementing memory, in which the image is accumulated. The data may, alternatively, be recorded in time-tagged form for speckle observations. A block diagram of a typical detector system is shown in Fig. 2. The basic single-channel system can be expanded into several channels by the use of multiple fibre-optic bundles feeding separate CCDs from a single image tube assembly. The system consists of independent channels from the CCD back to the memory, so high data rates and large image areas are simultaneously available. Development of Detectors
The LFPCA detectors are the outcome of a detector development programme originating at MSSSO in the mid-1 970s. Several generations of detectors have evolved from this programme, and the resultant detectors have been standard observatory instruments for the last eight years. Initial detectors used up to six electrostatic image tubes in tandem to obtain the gain required for photon counting, and one-dimensional Reticon arrays for read-out. The electrostatic tube stacks were replaced with a chevron
ELEmRONICS
SUBTRAIXION
IMAGE TUBE
1mrrt.r. PROCESSING ELECTRONICS
-
ADDITIONAL CCD CEANNELS
VOLTAGE
ACCUMULATING MEMORY
AND PRUlllCTION
VIDEO DISPLAY
COMPUTER
FIG.2. Schematic layout of the large-format photon-counting array detector chain.
COMPUTER INTERFACE
LARGE-FORMAT PHOTON-COUNTING ARRAY DETECTORS
303
microchannel plate tube and one electrostatic tube. These have been replaced with a tandem pair of standard night-vision microchannel plate image tubes for current observatory work. Early detectors used a single CCD or Reticon, with lens coupling of the intensified image at the tube to the sensor. Newer observatory detectors use two CCDs and pairs of solid fibres per set of image tubes. In parallel with ongoing development of the observatory detectors, a programme was begun in 1981 to develop space-qualified forms of the detector, with funding from the Australian Department of Science, and later the Department of Industry, Technology and Commerce. Initially these were designed for the STARLAB programme, a collaboration between NASA, Canada and Australia. Following the termination of the STARLAB programme, Australia was invited to participate in the LYMAN programme, and to investigate the application of detectors to the programme. The first detector to be developed in this programme has been a fourchannel system called the “Functional Model”. This uses a single, high-gain 40 mm image tube made by ITT, and four Fairchild CCD222 sensors. The Functional Model has been used for laboratory testing and astronomy on the MSSSO 40-inch telescope. Recognizing the necessity of flight-qualification of the detector in order to apply it to a major mission, the Endeavour programme was developed, to fly a four-channel LFPCA detector on the Space Shuttle in a low cost “Get Away Special” mission. Endeavour uses a single, high-gain 40 mm image tube, a 9-way flexible-fibre image dissector and four active CCDs. The detector is coupled to a binocular pair of telescopes with coatings optimized for the FUV, centred on 154 nm. The image tube uses an opaque photocathode, directly deposited onto the first of a pair of plates, and has a quantum efficiency of 17% peak, on a plate with a bias angle of 7” (Fig. 3). The Endeavour experiment is currently undergoing final integration and test. The structural model of the detector has been subjected to temperature cycling over the range -40°C to 30°C and vibration at up to 40 g peak, well in excess of the levels expected for the Shuttle flight.
+
Performance of Detectors
In order to fully exploit the potential of the Lyman mission, the detectors must have high performance. Particularly important parameters are the noise rate, the peak count rate, the resolution and the quantum efficiency (QE). The performance characteristics of LFPCA detectors have been established during several years of operation as standard astronomical instruments at MSSSO, and during the development of the Functional Model and Endeavour detectors.
304
E. H. ROBERTS, 1. R. TUOHY AND M. A. DOPITA 1
0
2
L
x
.'I
.
,
l
]
I
I
I
I
[
x
ReferencePM.
0-
Image tube
-s
c
,
x *
X
10' r: -
~
-
;
ZI
c ..-0 r w
-
-
loor -
5
:
+
a
3
-
lo-'
-
r X
10-2
1200
I
I
1400
I
I
1600
I
I
1800
I
.
.
1
2000
I
2200
FIG. 3. Quantum efficiency of Endeavour 40mm image tube. This tube has an opaque photocathode deposited on the first plate of a chevron pair and was made by ITT.
Noise rate The dark rate observed in LFPCA detectors is very low, in common with other forms of MCP-based detectors. Noise may be generated from thermionic emission from the photocathode, radiation in orbit, radiation from the MCP glass materials, field-emission in the image tubes and electronic noise within the detector chain. In image tubes for the visible, the dominant noise is thermionic emission from the photocathode. By cooling the photocathode, rates of less than lop5Hz pixel-' are routinely seen at MSSSO with S.20 photocathode material. In the UV, typical photocathode work functions are sufficiently high that cooling is not necessary, and the other factors become relatively important contributors. The LFPCA detector chains are sufficiently quiet that no noise events are recorded from the electronics (i.e. with the tube powered down) at a threshold of 1 LSB, or 0.4% of peak. The in-orbit noise due to radiation and tube defects is expected to be approximately 1 count cm-2 s-', which is much lower than the baseline requirement.
LARGE-FORMAT PHOTON-COUNTING ARRAY DETECTORS
305
Peak count rate The peak count rate required for Lyman is 0.3 Hz/pixel-'. Current MSSSO LFPCA systems achieve peak count rates for uniform illumination of greater than 0.5Hz/pixel-' for a 10% loss. For line illumination conditions the equivalent figure is 4 Hz per pixel. The peak count rate obtainable is dominated by coincidence loss. At higher count rates, the probability of two photons being physically coincident within the one frame increases. The expected peak count rate in Lyman will be higher, owing to the use of faster phosphors, single-tube assemblies producing smaller electron event sizes and (optionally) the elimination of frame subtraction. Resolution element The resolution of the detector is better than that due to either the CCD or the image tube output spot size individually, owing to the use of centroiding. The final pixel size at the detector is 15 pm, and the FWHM of the resolution element is approximately 18 pm. This is significantly better than is required for the baseline Lyman programme. The resolution may be improved with the advent of new, finer-pored MCP plates, and CCDs with smaller photosite dimensions. The spatial resolution of a MSSSO PCA detector has been investigated using arc spectra obtained with the CoudC spectrograph and 16-inch camera on the 1.9 m telescope at Mt. Stromlo Observatory. With the 600 line mm-' grating used, this system gave a dispersion of 5 nm mm-' and a format of 750 spectral pixels by 41 spatial pixels. A long-exposure spectrum (300 s) from a high-pressure argon-thorium lamp was obtained in connection with the SN1987a data discussed below. The 41 spatial channels were co-added after wavelength calibration to produce a single arc spectrum representative of the entire format. Individual emission lines were then fitted with Gaussian profiles to characterize their widths. A typical pair of profiles is depicted in Fig. 4. Both lines have a FWHM of 0.125 nm, which is equivalent to a spatial FWHM of 24.6 pm. This figure, however, includes a contribution from the spectrograph optics (12 pm FWHM) and from the finite (but unknown) widths of the emission lines from the high-pressure arc. The figure may also contain a contribution due to focusing error in the manual focusing procedure. Quadratic deconvolution of the spectrograph contribution therefore leads to an upper limit to the PCA resolution in this case of 21.5 pm FWHM. We note that the fundamental limit to the resolution of the PCA is set by the 15 pm pore spacing of the microchannel plates of the tubes used in the detector. It is therefore clear from our upper limit that very little resolution is lost in the centroiding process. Lyman will use MCP plates with a pore-to-pore spacing of 12 pm or less.
306
E. H. ROBERTS, 1. R . TUOHY AND M. A. DOPITA 0
m
I
I
I
I
I
: Angstroms
FIG.4. A Gaussian fit to emission lines from a high-pressure argon-thorium arc recorded using the photoncounting array (PCA) and the Coudl spectrograph on the 1.9 m telescope at Mt. Stromlo Observatory. The pixel size is 15 pm and the FWHM of the lines is equivalent to 24.6 pm, which leads to a deconvolved upper limit to the PCA spatial resolution of 21.5 pm (see text).
Quantum eflciency ( Q E ) The QE of the LFPCA is determined by the photocathode of the image tube. The Lyman detectors will use an opaque photocathode deposited directly onto the first microchannel plate. The prime candidate materials are caesium iodide, and potassium bromide. Both materials have good QE from the extreme ultraviolet to about 185 nm, tailing to 200 nm in the FUV. Siegmund (1987) has reported values for KBr in excess of 50%, with good QE stability upon exposure to nitrogen or air after coating.
Detective quantum efficiency (DQE) The LFPCA detectors used at MSSSO exhibit a DQE of 90%. Since internal noise sources are negligible, this figure implies that 90% of the photoelectrons generated at the photocathode are recorded as counts in the accumulating memory. The DQE has been derived both from variance analyses and measurements of star count rates. The 10% loss can be attributed
LARGE-FORMAT PHOTON-COUNTING ARRAY DETECTORS
307
to the dead space between the microchannel plate pores and a loss of photoelectrons resulting from the Poisson distribution of the secondary electron emission at the first stage of amplification. High-gain MCP materials and the use of repeller fields in UV tubes can be expected to increase the DQE by recovering some of these photoelectrons (Martin and Bowyer, 1982; Hurwitz and Bowyer, 1986). Photometric error
The photometric error in LFPCA data is typically less than 110% of the Poissonian photon noise, as determined by variance analysis. Two major contributors to photometric error are missed counts at high count rates and double counting. By using high clock rates on the CCDs and processing electronics, the count rate is maximized. By using high-speed phosphors and frame subtraction, double counting is minimized. Image stability
The LFPCA detectors exhibit an image stability of better than 2 pm RMS on a night-to-night basis, as determined by flat-field division of images. This stability is due to the short free-electron flight paths in image tubes, and the use of fibre coupling between the CCD and image tube. SAMPLE DATA:SN1987~ As topical example of the capabilities of the LFPCA, we present spectra of SN1987a obtained at Mt. Stromlo Observatory using the Coudi spectrograph. Figure 5 shows a broad-band spectrum of the supernova between 400 and 780 nm derived by concatenating 13 overlapping spectra aquired in July 1987. Note in particular the B and A atmospheric absorption bands near 690 and 760nm, the latter of which is superposed on an 01 resonance-line absorption feature intrinsic to the supernova. The complex structure in these regions is vividly revealed in the expanded spectra shown in Fig. 6. The Rbranch series is fully resolved in both bands. DESIGNOF DETECTORS FOR LYMAN Several detectors are required for the Lyman mission, including the FUV tchelle detector, the EUV detector, the prime spectrograph detector, and possibly fine error sensor detectors. The latter are not considered here. If some or all of the instruments and sensors utilize LFPCA technology, they may share a common set of image-processing electronics within the
E. H. ROBERTS, I. R. TUOHY AND M. A. DOPITA
308 1
"
~
'
1
'
~
'
'
1
"
'
l
"
'
'
~
'
"
'
I
"
'
~
1
I
'
"
I
'
1
10
0
SN1987a July 1987
(D
10
XE! 3 %
G!*
W
t c
4
W
u: In
2 N
4000
4500
5000
5500 Wavelength
6000
6500
7000
7500
Angstroms
FIG. 5. A broad-band spectrum of SN1987a in the Large Magellanic Cloud obtained by concatenating 13 overlapping spectra acquired in July 1987. Experimental details are as in Fig. 4.
spacecraft, thereby saving mass, volume and power. Moreover, such commonality will help increase the overall reliability of the mission by providing redundancy at little extra cost. Since only one instrument is selected at any one time in Lyman, such sharing does not compromise the science that may be done. Prime Spectrograph Detector
As noted above, two detector formats may be developed for the prime (90120 nm) mission. The Rowland format is the more challenging, because it requires an unusual image tube format, coupled to several channels of processing electronics. The tchelle format requires a fairly conventional image tube assembly, coupled to four CCDs and four channels of image processing. Rowland Spectrograph Detectors
As noted in Table 11, the Rowland focal plane requires a detector with a length of approximately 150 mm, a height of 5 mm, and approximately 8000
7500
7550
7600
7650
7700
7750
Wavelength Angstroms
1
I
6860
6880
6900
6920
6940
Wavelength Angstroms
FIG.6. Expanded spectra from Fig. 5 of atmospheric absorption bands in the spectrum of SN1987a. The A-band at ~ ~ 7 6 A 0 0is superposed on an 0 1 absorption feature produced in the supernova itself. Note that the R-branch series is fully resolved in both absorption bands.
310
E. H. ROBERTS, I. R . TUOHY AND M. A. DOPITA
pixels in order to cover 10 nm of the prime wavelength range. The focal plane is in fact curved, with a diameter of 2.1 m. It is proposed to build a series of linear sub-tubes, mounted as secants to approximate the Rowland circle, and to couple each sub-tube, via solid coherent fibre bundles, to channels of CCDs and processing electronics. Each sub-tube and its associated CCDs would form an electrically independent assembly, to minimize the loss of image in the event of failure of a single tube or CCD channel. The design as proposed has two unusual features, open-windowed and open-ended rectangular image tubes, and “multiplexing” of a single CCD onto two adjacent sections of the image tube. It is necessary to use open-ended tubes in order that several may be butted together at the focal plane with minimum loss of image between the tubes. It is proposed to separate each sub-tube from its neighbours by a thin plastic (e.g. Teflon) insulating layer, with an intertube gap of less than 2 mm. The subtubes will be mounted onto a common support assembly which provides vacuum interfaces and a sealing front door, to protect the photocathodes. This common assembly will provide the mechanical support to the tubes during launch. During integration the tubes will be supported via their fibre output plates, which will be bonded or brazed to the tube ceramics. A crosssection of a typical sub-tube is shown schematically in Fig. 7. As noted above, the image height in the spatial direction is less than 5 mm, whereas the height of the CCDs is 8.8 mm (for a CCD222). It is thus proposed to map two adjacent sections of the intensified image onto the one CCD via two solid fibre-optic bundles, each approximately 4 mm high by 11.2 mm long. The resulting detector uses only one-half as many CCDs as would otherwise be necessary, and hence only one-half as many image-processing channels. This results in considerable mass, power and volume savings in the spacecraft. BONDED COLLAR
PRELOAD SHIMS
SOLID FIBRE
VACUUM SEAL
F I B R L OUTPUT P L A T E DETECTOR STRUCTURE
ZED
MCP P L A T E S
CERAMIC SPACERS
TUBE CERAMIC
VACUUM SEAL
REPELLER G R I D
FIG.7. Schematic cross-section of an open-windowed sub-tube for the Rowland spectrograph.
311
LARGE-FORMAT PHOTON-COUNTING ARRAY DETECTORS CCDM’S I S )
DCE/l?E
CONNECTDR
FLEX-RIGID PCB COOLING I / F
SOLID F I B R E S VACUUM BOULE S E A L
CCD COOLING BRAID DETECTOR STRUCTURE TUBE (open windowed)
VACUUM DOOR
C C D CHIP ( d ; r e c t bonded)
\
DETECTOR/OPT. BENCH I/F
I
100 mm
FIG.8. Schematic cross-section of the Rowland detector-headassembly showing an image tube, two CCD modules, coupling fibre-opticsand the vacuum assembly.
The proposed detector will consist of four such sub-tubes, each 44.8 mm long, and each with two CCDs. Eight image-processingchannels are required for the basic detector. Each sub-tube will use its own high-voltage supplies, for electrical independence. The design of the baseline Rowland spectrograph is such that, by selection of backup gratings, the spectral image may be shifted by one full sub-tube, thereby allowing data to be recovered in the event of a tube failure. The detector-head assembly for such a detector is shown in Fig. 8. During storage and integration the detector head will be sealed around its protective door, and the photocathodes will be kept under vacuum by ion or getter pumping. Magnesium fluoride windows in the door assembly will allow the tubes to be tested in the FUV range without having to open the protective door. Prime Echelle Detector
The 6chelle spectrograph provides a compact, flat focal plane, with nominal dimensions of 25 x 25 mmz (1000 x 1000 x 25 pm pixels). The detector will consist of an open-windowed,circular image tube, with flat MCP plates, and four CCDs coupled by short solid fibre bundles. Using an LFPCA detector with four CCD222s, the format becomes 17.6 mm (H) x 22.4 mm (L) with 960 x 1520 pixels, each 18 pm (H) x 15 pm (W), with a total of 1.48 million pixels. As noted above, a four-CCD detector has been demonstrated with the
312
E. H.
ROBERTS,
I. R. TUOHY AND M. A. DOPITA
required performances. A nine-channel ultraviolet detector, with four active CCD channels has been built and is undergoing testing prior to a Shuttle flight in 1988/9. The development of the prime Cchelle detector, therefore, is an extension of existing work. EUV Detector
The format of the EUV spectrograph has yet to be decided by the study team. Among the current options are grazing incidence kchelle systems, which provide a compact image format, and variable line space grating spectrographs (Hurwitz and Bowyer, 1986), which provide a linear, extended format. In either case, the image tube is required to be open-windowed, with an opaque photocathode, and repeller grid. The EUV khelle format requires a single CCD and an 18 mm image tube with open window, and some form of sealing door assembly. Since the pixel size required is less than 50 pm, it is possible to use a larger field of view at the image tube, and minify the pixels with a fibre-optic taper, in order to potentially cover a larger waveband than the 10-30 nm baseline. In particular, it is desirable to extend the coverage to 90 nm. The linear VLS spectrograph format may be up to 100nm long, and approximately 3 mm high. The image tube required for such an instrument would be similar to that required for the prime Rowland spectrograph, with similar design constraints. Between four and six image-processing channels would be used for such a system.
FUV Detector The FUV detector operates in the waveband from 120 to 200 nm, and hence can use an image tube with a window. A standard high-gain MCP tube may therefore be used, with an active diameter of greater than 32 mm. The 40-mm tubes used in the Endeavour programme, coupled to a four-CCD detector chain as used in the current Functional Model, will be adequate. The FUV is required to have one million pixels, with a size of 25 pm square or less. The Functional Model provides 1.4 million pixels, with dimensions of 15 x 18 pm2, and is hence fully compliant with these requirements. CONCLUSION
The Lyman ultraviolet space telescope requires high-performance photoncounting detectors. LFPCA detectors are well suited to the programme, and will provide the required performances. The development of image-tube assemblies, particularly for the prime Rowland spectrographs, is a challenging
LARGE-FORMAT PHOTON-COUNTING ARRAY DETECTORS
313
area. Layouts of suitable tube and detector head designs are presented. The current detector development programme underway within MSSSO and Auspace Limited gives a high degree of confidence that space-qualified LFPCA detectors can be developed for Lyman within the mission schedule. ACKNOWLEDGEMENTS The authors wish to thank Steven Rawlings for his assistance in the data reduction, and Alex Rodgers for discussions regarding the performance of the Mt. Stromlo PCA detectors. Auspace Limited acknowledges the role of the Department of Industry, Technology and Commerce in funding the Australian Lyman Phase A Study and the Endeavour experiment, and (with the Department of Science) the earlier phases of the STARLAB and Endeavour programmes.
REFERENCES
Gorham, R., Rodgers, A. and Stapinski, T. (1982). Proc. S.P.I.E.,331,490497 Hurwitz, M. and Bowyer, S. (1986). Proc. S.P.I.E.627, 371-378 Martin, C. and Bowyer, S. (1982). Appl. Opt. 21,4206 Roberts, E., Stapinski, T.and Rodgers, A. (1986). J . Opt. SOC.Am. 3, 2146 Siegmund, 0. (1987). Appl. Opt. (in press)
This Page Intentionally Left Blank
Image Intensifier Tubes with Intagliated Screens M. FOUASSIER, V. DUCHENOIS, J. DIETZ, E. GUILLEMET and M. LEMONIER Laboratoires d'Electronique et de Physique AppliquPe. Limeil Brevannes Cedex, France
INTRODUCTION
Image intensifier tubes (IITs) are used in many applications, such as image detection at low light levels (night vision) or X-ray imaging. The image quality can be characterized by the modulation transfer function (MTF): output contrast versus spatial frequency for 100% input contrast. This function depends on several parameters, including photoelectron energy distribution, electron optics, microchannel pitch, screen granularity, and fibre-optic pitch. Various improvements have already been studied and applied to tubes. When the output screen is settled on a fibre-optic window, a further MTF improvement can be performed, as described in this paper. Usually, screens are settled on fibre-optic plates in order to minimize veiling glare, and also to couple the tube directly with an image sensor such as photographic film, TV pick-up tube or CCD (Richard et al., 1987). Here, the spatial resolution is enhanced by intagliating phosphor particles in holes excavated in each elementary fibre at the input of the fibre-optic plate. This technique has been already described by Duchenois et al. (1985). After briefly recalling the principle and the techniques, the main MTF results obtained on these intagliated screens, as well as on several types of tubes equipped with them, are given.
INTAGLIATED SCREENS Principle
In a conventional screen settled on a fibre-optic plate (Fig. l), photons are emitted by phosphor grains in all directions and often reach optical fibres located far from the emission point, possibly after diffusion by the phosphor particles and/or reflection on the aluminium layer. As the geometrical positions and the dimensions of the phosphor particles and fibres are completely independent of each other, the MTF of the screen on the fibre315 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14
Copyright Q 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
316
M. FOUASSIER ET AL.
I
Aluminium layer
\
Electrons
Fibre optic
FIG. I . Schematic diagram of a conventional phosphor screen on a fibre-optic window.
Aluminium layer
\
1
Electrons
Lateral ,metallisation
Phosphor screen
Fibre optic
FIG.2. Schematic diagram of an intagliated phosphor screen on a fibre-optic window.
optic is a combination of the phosphor screen MTF and the fibre-optic MTF. In order to avoid any light spread, the idea is to insert the phosphor particles in holes excavated in all the elementary fibres (Fig. 2). To prevent the light passing into adjacent fibres, the lateral walls of the holes are metallized. As a result, photons emitted by the screen powder are kept inside the corresponding fibre. Consequently, the MTF of the intagliated screen should not depend on the phosphor but, theoretically, only on the fibre-optic MTF.
Manufacturing Technique Holes are excavated in the fibre-optic plate by elimination of the core glass to a certain depth without etching the cladding glass. In our case, this is carried out by selective chemical attack, since the two glasses react very differently to acid solutions. Figure 3 shows an example of a fibre-optic plate after etching of the core glass to a depth of 5 pm. The cladding glass is not attacked and the walls of the cavities are very clean. The cavities’ lateral walls must be metallized as previously stated. The incidence angle of the metallization must be correct: if it is too small, the bottom of the cavity is metallized and the luminous yield decreases; if it is too large, the walls are not completely metallized and the contrast improvement can remain low. The particles must also be of the right size: if too large, they will not enter
IIT WITH INTAGLIATED SCREENS
317
FIG.3. Photograph of a 6 pm pitch Mullard fibre-optic plate after chemical etching to S jtm depth.
FIG.4. Photograph of an intagliated screen in a 6 pm pitch Mullard fibre-optic plate.
the cavities; if too small, the screen’s luminous yield will remain low, since it decreases with the grain size. In practice, more than one particle is necessary in each hole, with dense compacting for a high conversion efficiency. The powder is intagliated in the fibre-optic by centrifuging to force the particles into the cavities. Figure 4 shows an example of powder intagliation. The holes are uniformly filled with phosphor grains, a necessary condition for a good screen uniformity, and all the excess powder has been eliminated. Finally, the intagliated screen is metallized with an aluminium layer.
318
M. FOUASSIER ET AL.
Resuits The luminous yield of these intagliated screens depends on the type of fibreoptic, especially the diameter of each elementary fibre and its open-area ratio, as well as on all the parameters described above. It is generally optimum at a cavity depth of about 3 pm, for a 10 pm pitch fibre-optic plate. Smaller depths lead to lower yields, because too few particles are present in each hole and some of the incident electrons do not fall onto the powder. If the cavity is deeper, the yield also decreases because, as there are several phosphor particles superposed in each elementary hole, the light emitted by the upper particles, at which the incident electrons arrive, is reflected, diffused and partially absorbed many times by the other phosphor particles and by the metallizing before reaching the core glass at the bottom of the cavity. In comparison with a standard screen, the luminous yield of an intagliated screen is generally reduced by a factor of about 2 at 6 keV primary electron energy. The screen MTF measurements were made at Philips Research Laboratory, Salford by A. J. Jenkins (1981). The screens are introduced into the vacuum system of a scanning electron microscope (SEM) and electron beam scanning is used to obtain the equivalent of a line or slit less than 1 pm thick. The image obtained on the screen is relayed out of the vacuum system by a microscope objective, on an ODETA I1 equipment giving the MTF directly as the Fourier transform of the line-spread function. The advantage of this measurement technique is that the screen is operated under the same conditions as in image intensifier tubes, i.e. in the electron detection mode and with its aluminium layer. It is important to notice that the MTF results are very different when the measurements are made with ultraviolet light and without an aluminium layer on the screen. Figure 5 shows the MTFs of various intagliated screens made on 10 pm pitch fibre-optic plates in comparison with the MTFs of the fibre-optic plate alone and of a standard screen deposited on the same type of fibre. Though the intagliated screen MTFs are slightly lower than the fibre optic MTF, they are much better than that of the standard screen. At 30 Ip mm-', the intagliation technique improves the contrast from 0.35 to about 0.65. Figure 6 shows the same type of results obtained on 6 pm pitch fibre-optic plates. The influence of the fibre-optic pitch is obvious both for a standard screen and for an intagliated screen. Even for a standard screen, the contrast, at 30 lp mm-', is improved from 0.35 to 0.5 when the pitch decreases from 10 pm to 6 pm. With the intagliation technique the contrast, at 30 Ip mm-I, is enhanced from 0.5 to around 0.75. At high spatial frequencies (601pmm-'), the contrast rises from 0.15 to 0.55. Intagliated screens will be of particular interest for tubes in which the main MTF limitation is due to the screen itself as, for example, first-generation image intensifier tubes (diode tubes).
319
IIT WITH INTAGLIATED SCREENS
Contrast
0
20
40 60 Spatial frequency (Ip/rnm)
80
100
FIG.5. MTF of various intagliated screens, a standard screen and the fibre-optic alone for a 10 pm pitch fibre-optic plate. Contrast
?\
t I
0
20
Intaghated s c r e e n s ' 7
I I l l 1 1 40 60 80 Spatial frequency (Ip/ mrn] 1
400
FIG.6. MTF of various intagliated screens, a standard screen and the fibre-optic alone for a 6 pm pitch fibre-optic plate.
IMAGEINTENSIFIER TUBESWITH INTAGLIATED SCREENS Some image intensifier tubes with or without microchannel plates have been produced with intagliated screens. Their characteristics are given below in comparison with tubes equipped with standard screens.
320
M. FOUASSIER ET AL.
Second Generation Image Intensifier Tubes
XX 1410 second-generation IITs from Philips, produced at RTC Brive (France), were used for this comparison. They are 18 mm double proximityfocused intensifiers equipped with S * 20R photocathodes deposited on fibreoptic input windows, microchannel plates and P 20 luminescent screens settled on fibre-optic output windows. In order to show the influence of the intagliated screens on the tube performance precisely, the standard tube chosen for this comparison has the same physical characteristics as the intagliated tubes (i.e. microchannel plate, fibre-optic plates, interelectrode distances, voltages, etc). Only the screens are different. The substrate is a 10 pm pitch fibre-optic plate. The screen luminous yield, measured at the normal screen voltage of 5.7 kV, is around 10 lm W-' for the intagliated screens and around 20 lm W-' for the standard screens. This loss of luminous yield leads to a tube luminous gain decrease that can be recovered by an increase of the MCP gain (higher voltage). The signal-to-noise ratio might then be affected. However, for the usual tube gains (2500-5000 cd m-* lux-'), a gain increase of a factor 1.5-2 generally leads to a maximum reduction in signal-to-noise ratio of 5%. It is then possible to recover the standard luminous yield without significant damage to signal-to-noise ratio. Figure 7 shows the MTFs obtained on two tubes with intagliated screens in comparison with those obtained on a standard tube. The contrast improve-
-
Contrast
0
5
10
15
20
25
30
35
40
Spatial frequency (Ip/mrn)
FIG.7. MTF of XX 1410 image intensifier tubes with intagliated screen and standard screen.
32 I
IIT WITH INTAGLIATED SCREENS
ment produced by the intagliation is very significant and fits the calculated one well. At 30 Ip mm-I, the contrast is improved by a factor of about 2. The limit of resolution is visually measured at a contrast of about 4%. This is around 34 Ip mm-' for a standard tube and around 39 lp mm-' for tubes with intagliated screens. First Generation Image Intens$er Tubes
These tubes include only photocathode and luminescent screen. The tubes used for this comparison are XX 1080 inverter diode tubes (electrostatically focused) made by Philips. They are used in Driverscopes made by TRT (OB 31), Philips USFA (UA 9630) and PST (BM 8005). The photocathode diameter is 50 mm and the screen diameter is 16 mm giving a demagnification of 0.3. The screen is settled on a 6 pm pitch fibre-optic output window. As before, the tube performance is compared to that of a standard tube (Fig. 8). The screen luminous yield at 14 keV electron energy is 5.1 lm W-' compared to 24 Im W-' for a standard screen. This loss cannot be recuperated. It is due, in part, to the open-area ratio (0.5) of the fibre-opticused for the intagliated screen, which is lower than that for the standard screen (0.7). It is also due to use of a different screen powder (ZnS :Cu instead of ZnCdS :Ag), which has a lower luminous yield, and probably also to too deep metallization. The MTF includes the effects of electron focusing as well as the screen being Contrast
0
20 40 60 80 Screen spatial frequency (Ip/mrn 1
100
FIG.8. MTF of XX 1080 image intensifier tube (diode tube) with intagliated and standard screens.
322
M. FOUASSIER ET AL.
settled on the fibre-opticplate. As the electron focusing MTF is almost perfect here, the tube MTF is almost equivalent to that of the screen on the fibreoptic. Thus, if an intagliated screen is used for such a tube, the tube MTF will be almost equivalent to that of the intagliated screen, which according to the results given above, should significantly improve the contrast. The contrast improvement shown in Fig. 8 is very high: the tube MTF is almost equal to the fibre-optic MTF. At 60 lp mm-’, the contrast increases from 0.16 to 0.68, a still higher contrast than expected (0.55). This implies that these intagliated screens are optimized for resolution improvement rather than for luminous yield, which is lower here.
CONCLUSION The intagliation technique allows the screen MTF to be improved so much that it approaches that of the fibre-optic plate. At 30 Ip mm-’, the contrast is improved by a factor of 1.5-2. The luminous yield is reduced by a factor 2, approximately. These intagliated screens can be used in image intensifier tubes to improve the image sharpness. Nevertheless, it is necessary to make a compromise between gain and resolution depending on the application. In secondgeneration image intensifier tubes, the resolution improvement is around 5 lp mm-’ and the lower gain of the tube can be compensated by higher gain in the microchannel plate. In a first-generation tube, if gain is not too important, there is a large improvement in image quality due to the much higher contrast (increased by a factor about 4 at 60 Ip mm-I). Many other devices, such as X-ray image intensifiers, in which a contrast improvement is desirable could take advantage of this technique. ACKNOWLEDGEMENTS The authors wish to thank all the persons who havecontributedto this work, and specially A. J. Jenkins (PRL Salford) for the screen MTF measurements, F. Stijnis (Elcoma Eindhoven) for producing the XX 1080 tube, and J. Dupuy (RTC Brive) for producing the XX 1410 tube. This work was supported in part by the Direction des Recherches, Etudes et Techniques (DRET).
REFERENCES Duchenois, V., Fouassier, M. and Piaget, C. (1985). In “Adv. E.E.P.” Vol 64B, pp. 365-371 Jenkins, A. J. (1981). Proc. S.P.I.E. 274, 154-158 Richard, J. C., Lamport, D. L., Roaux, E. and Vanneste, C. (1977). Acra Electron. 20,4,353-368
Photocathodes on Polycrystalline CsI/Na Y. ARAMAKI Horikawacho Works, Toshiba Corporation, Kawasaki, Japan
INTRODUCTION It is difficult to make semitransparent photocathodes on polycrystalline CsI/Na (hereafter denoted p-CsI/Na) having both high sensitivity and long life. When a photocathode is actually made on p-CsI/Na, the photocurrent, which is initially high, often decreases during operation. There is little doubt that the cause of this reduction in photocurrent is an interaction between the substrate and the photocathode. This report describes Auger analysis of two improved types of photocathode.
PHOTOCATHODE AN IMPROVED
Methods of Photocathode Production on p-CsrjNa This report describes two ways of producing photocathodes with high sensitivity and long life on p-CsI/Na substrates. Basically, the processing of these photocathodes consists of two steps. The first step is to create a base layer on an intermediate layer just under a photoelectron conversion layer. This is done by introducing abundant alkali metal vapour; the process is called doping. The second step is to make the photoelectron conversion layer. The two ways of producing the photocathode differ in the first process. The first step of one method is introduction of abundant alkali metal vapour to the p-CsI/Na. This process, called doping, is different from that used in forming conventional photocathodes. In order to make K-Cs-Sb bi-alkali photocathodes, potassium or other alkali metals are used for doping. The potassium is introduced in a heated dispenser at a constant substrate temperature and the system is evacuated. The alkali metal is fully absorbed in the substrate whilst the photocurrent is monitored. After doping with the alkali metal, the photoelectron conversion layer is formed. Antimony is deposited first. The antimony deposition is followed by the introduction of potassium and then caesium. The photocathode is completed by further depositions of antimony and caesium in turn. 323 ADVANCES I N ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright Q 1988 Academic Press Limited All rights of reproduction in any form reserved
ISBN 0-12-014674-6
324
Y. ARAMAKI
Another method for bi-alkali photocathodes will be described briefly. An oxide layer of alkali metal is first formed on the substrate, and then the photoelectron conversion layer is formed. The oxide layer of alkali metal is made as follows. As described above, alkali metal is introduced while keeping the pre-determined temperature of the substrate constant and while evacuating. The composition of each oxide layer is variable according to which alkali metal is introduced and whether it is made of composite layers. It is then essential that, after deposition of alkali metal on the substrate is completed, oxygen is introduced to fully oxidize the alkali metal. After oxidation of the alkali metal, a photoelectron conversion layer is formed by the method described above.
The Intermediate Layer The photocathode was made on an intermediate layer created by vacuum evaporation onto the p-CsI/Na. One type of intermediate layer was composed of A1203 and In203,and the other was made in In203alone. The A1203 was deposited in order to cover discontinuities in the p-CsI/Na which make it unsuitable as a substrate for the photocathode.
AUGERANALYSIS OF IMPROVED PHOTOCATHODES The operating conditions of the Augur apparatus were as follows: base Torr; beam diameter, 0.5 pm; beam current, vacuum pressure, I x 0.1 pA; sputtering gas, xenon; sputtering rate, 3 A min-' (for Si02). Data were obtained from Auger analysis by sputter etching of the photocathode surface.
Auger Analysis of Photocathode Surfaces In carrying out Auger analysis of the photocathode surface great care was taken to minimize the movement of alkali metal by using only low probe beam currents. Figure 1 shows the results of Auger analysis conducted on the surfaces of three types of photocathodes. It can be seen that, in the case of the alkali metal oxide photocathode, the spectrum of carbon was hardly detected. This is believed to have been due to oxygen introduced in making the photocathode. Each spectrum in Fig. 1 is different in the peak height of each element. Another feature of these spectra is that sodium was detected despite its being separated from the p-CsI/Na by the intermediate layer.
PHOTOCATHODES ON POLYCRYSTALLINE CSI/NA
325
Distribution of Elements on Surface
The Auger analysis probe spot was scanned over the surface of the photocathode in order to obtain information on the distribution of elements over the surface. Figure 2 shows the results of a composition analysis of the photocathode surfaces in one direction. As was shown by the spectra in Fig. 1, each distribution is characterized by the ratio of alkali metals to Sb. Fig. 2(a) shows that Cs is two orders of magnitude higher than K and Sb. This is abnormal for conventional glass substrate photocathodes. In Fig. 2(6) and (c), the ratios of alkali metals to Sb seem acceptable. These facts are also clear from the averages and standard deviations calculated on the assumption of normal distributions and shown in Table I. In Fig. 2, Na was detected although it had not been introduced. In Fig 2(b), Na is present only where A1 exists, which suggests some correlation between these two elements. Correlation coefficients of elements were obtained. Values
t
L
I 0
0 0
1
I000 E l e c t r o n E n e r g y I eV
2000
I
FIG.I . Auger spectra of photocathodes: (a)alkali-metal-richphotocathode; (6) oxidized alkali metal photocathode on A1203 and 111203;(c) oxidized alkali metal photocathode on InzO3.
Distance .
a
0
Distance
(b) FIG. 2. Elemental distribution in photocathodes: (a) alkali-metal-rich photocathode; ( b ) oxidized alkali metal photocathode on A1203 and In203.
327
PHOTOCATHODES ON POLYCRYSTALLINE CSI/NA
(c)
'
I
0.1
I
I
1
I
I
Distance
FIG.2. Elemental distribution in photocathodes: (c) oxidized alkali metal photocathodes on 111203.
were derived for each element together with data on the depth profile; these are shown in Table 11. As is evident in Table 11, there is a close correlation between A1 and Na and it seems that Na has diffused from substrate of p-CsI/Na. The diffusion coefficient of Na in A1203 is known to be large (Freer, 1980), which is consistent with this observation. TABLE I Calibrated average peak values ( f )and deviations (3) of elements Element Photocathode
R
Sb
cs
K
Na
0.0429 0.138
42.0 1.67
0.123 0.296
71.4 7.84
Potassium-rich layer base
d
Potassium oxide layer base
8
(In203+A1203)
B
17.2 11.5
f 6
18.3 12.5
Potassium oxide layer base
1.08 1.91 52.0 25.6
14.1 1.77
13.7 12.4
12.3 10.9
69.2 57.6
Y. ARAMAKI
328
TABLE I1 Relationship between elements Element
Photocathode
Analysis direction
Potassium-rich layer base
Surface Depth
Potassium oxide layer base (In203+AlzOd
Surface Depth
@
~($1~($1
-@ 0 o(--@) @
Potassium oxide layer base (In204
Surface Depth
@ @
0
0
Sb-K
Sb-Cs
Sb-Na
Na-In
Na-A1
x
X
X
X
X
x
X
X
-@
@
O(@)x
X
x(@)
0 0
~
-@
@ Reliability 99%; 0reliability 95%; x low or no reliability; - negative (others are positive). Data in parentheses were obtained from other specimens.
Profile in the Depth Direction The distributions of elements in the photocathode were also measured in the depth direction. Results are shown in Fig. 3(a), (b) and (c).
0
(a)
10
20
30
Sputtering Time (min.)
FIG.3. Depth profile of photocathodes: (a) alkali-metal-rich photocathodes.
t
L
0 (b)
1
10
I
I
20
30
Sputtering Time (rnin.)
t s
-
. vo 0
i -I
I
Sb
In
I
FIG.3. Depth profile of photocathodes:(b) oxidized alkali metal photocathode on A1203 and (c) oxidized alkali metal photocathodes on InzO,.
In203;
330
Y. ARAMAKI
Alakali metal rich layer photocathode
Figure 3(a) shows that there is more Cs than Sb and that it is present not only on the surface of the intermediate layer but also deep down. There is not so much K as Cs, but it is present at a greater depth than Sb. Na is abundant. Oxidized alkali metal layer photocathode
This photocathode consists of an oxide layer under the photoelectron conversion layer, the composition of which varies in accordance with the type of intermediate layer. The photocathode was applied to both types of intermediate layers. In either case, the ratio of Cs to Sb is near to unity and is smaller than in the case of alkali-metal-rich layer photocathode shown in Fig. ] ( a ) . PHOTOCATHODE SENSITIVITY
If the sensitivity of the alkali-metal-rich photocathode formed on an intermediate layer consisting of layers of A1203 and In203 was taken as 1, the sensitivity of the oxidized alkali metal photocathode formed on a double layer of A1203 and In203 was 1.5, while that of the oxidized alkali metal photocathode on a single intermediate layer of In203was 1.8. These sensitivity values might change with conditions of formation of the intermediate layers. CONCLUSION
In this experiment we prepared two kinds of K-Cs-Sb photocathodes on polycrystalline CsI/Na and carried out an Auger analysis. In one type of photocathode the ratio of Cs to Sb is high and the Cs forms a deep layer, while in the other type the alkali metal oxide is formed under a photoelectron conversion layer and the ratio of K to Sb is high, while that of Cs to Sb is low. In the presence of A1203,Na diffuses easily towards the photocathode from the p-CsI/Na substrate. ACKNOWLEDGEMENTS The author wishes to acknowledge the contribution made by Mr Y. Ueda and Mr T. Suzuki who carried out the Auger analysis.
REFERENCE Freer, R.(1980). J. Marer. Sci. 15, 803-824
High Performance with Trialkaline Antimonide Photocathodes P. DOLIZY, F. GROLIeRE and M. LEMONIER Laboratoires d’Electronique et de Physique Appliqute, Limed Brkvannes Cedex. France
INTRODUCTION NazKSb,Cs photocathodes have been known for a considerable time (Sommer, 1968). Nevertheless, work on new types of photoemitters, such as 111-V compounds, has given a new impetus to improving our understanding and thereby enabling the technology to achieve so-called standard photoemissive films. Photocathodes with white-light sensitivities in excess of 500 pA lm-’ are now a reality in industrial production. The homogeneity in composition of these films in a direction perpendicular to their surface is a property of great importance in obtaining very high photoemissive properties. The escape length of photoelectrons becomes so large that it offers new possibilities.
SOMEFEATURES OF PHOTOCATHODES
It has been already shown (Dolizy et al., 1977) that the crystalline structure and texture of NazKSb are strongly influenced by the initial growth conditions. Let the initial step in the growth of NazKSb be the deposition of K3Sb as a thin film. It has to be transformed into Na2KSb by subsequent evaporations of antimony and sodium. The crystallinity of the initial K3Sb as a thin film is very poor and the crystallinity of the corresponding Na2KSb film appears to be very poor also. Now let the initial film be made of Na3Sb. In contrast, this material, deposited as a thin film onto a glass substrate, appears to be very well crystallized and exhibits a very strong texture with its (001) plane parallel to the surface. This material can be transformed as a thin film into Na2KSb by subsequent evaporations of alkali metal (potassium) and antimony. Chemical reaction yields thin-film NazKSb that is generally very well crystallized, exhibiting a strong texture with the (1 11) plane parallel to the surface (Fig. 1). The photoemissivity resulting from the latter method has been observed to be better than that from the former. From this point of view, 33 1 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any farm reserved ISBN 0-12-014674-6
332
P. DOLIZY, F. GROLIERE AND M. LEMONIER
3 I
=c
v)
5u”
e e
y
uN
$o z5 -0 5
-i
I
Y
N
P
om
5
n
‘0
%
1
5
10
15 Bragg angle (degrees)
20
FIG.I . X-ray diffraction pattern of a (Na,Sb, Na2KSb) thin film.
it is concluded that the initial growth condition is of interest for the crystallization of the Na2KSb film. Nevertheless, as explained below, the initial Na3Sb thin film has to be completely transformed into Na2KSb in order to produce high photoemissivity. X-ray diffractograms have shown that during the growth of the NazKSb material, a maximum of two phases coexist within the bulk of the material deposited as a thin film. A rule that has been formulated about the coexistence of phases within the thin film is that there can be present only one or two “neighbour” compounds taken from the list: Na3Sb, NazKSb, NaK2Sb, and K3Sb. X-ray diffraction patterns are useful in studying the growth of films. Nevertheless, the sensitivity of this tool is limited, especially when the thickness of the material under study becomes thinner than a few hundred angstroms. In this situation, it becomes quite difficult to analyse the composition of the film. Appropriate optical measurements have therefore been applied to very thin films to obtain information about the heterogeneity in composition within their volume in a direction perpendicular to their surfaces. OPTICAL
HETEROGENEITY IN THINPHOTOCATHODES
Heterogeneity Related to the Photocathode Substrate
The heterogeneous character of a thin Na3Sb film has been discussed elsewhere (Dolizy and Groli&re, 1987). The optical reflectance and transmittance measured at one wavelength have been determined by Auger electron spectroscopy to derive data about the heterogeneity of the thin film in a direction perpendicular to its surface. From this study, it has been concluded that Na3Sb deposited on a glass substrate ( S 747-01 or S 801-51) is often
TRIALKALINE ANTIMONIDE PHOTOCATHODES
333
heterogeneous. This appears to be the result of chemical reactions between alkali metals and glass and to the presence of impurities. The heterogeneity corresponds to a stratified optical system in which the thickness of the stratum has been measured to be as much as 200A with an optical extinction coefficient k 1 at a wavelength A = 632.8 nm.
-
Heterogeneity Related to Photocathode Growth
Photocathodes which contain sodium, potassium and antimony are often composed of two phases such as Na3Sb and NaaKSb (Fig. 1). This situation appears particularly when the growth of the Na2KSb film begins with the deposition of a thin Na3Sb film on the glass substrate as described above. The coexistence of these two phases within the bulk of the film in a direction perpendicular to its surface can be analysed by using appropriate optical techniques (Dolizy, 1987). Two physical situations have been considered. The two phases Na3Sb and Na2KSb can be homogeneously mixed together, or one phase can be on the top of the other. As an example, the spectral reflectance of a 1540 A film deposited on a glass substrate and made of pure Na2KSb, or containing 1020% of Na3Sb, has been computed and is shown in Fig. 2(a). Moreover, if the film is made of Na2KSb deposited on the top of Na,Sb, the corresponding computed spectral reflectance is as shown in Fig. 2(b). Comparison between theoretical and experimental spectral reflectance curves give indications of the behaviour of the two phases within the thickness of the film. First, we have to make sure that the initial Na$b film is homogeneous in bulk, as defined by Dolizy and Groliere (1987). Then, the NajSb film is progressively transformed into Na2KSb by deposition of additional Sb and K. Before this chemical transformation has been completed, the film corresponds to a stratified optical system which can be interpreted as Na3Sb covered by Na2KSb. The spectral reflectance is very sensitive to the heterogeneous character of such a film. A Na3Sb film as thin as 50 A, located at the interface between the glass substrate and the NaZKSb layer, produces large variations of the optical reflectance measured over the visible spectrum, as shown in Fig. 2(b). The experimental result given in Fig. 2(b) relates to a Na2KSb film, obtained after completion of the reaction. This indicates that the Na3Sb film first deposited on the glass substrate has been totally transformed into Na2KSb. Effects of Heterogeneity on Photoemission
The physical nature of the interface between the substrate and the photocathode is rather complicated. This interface can be made of Na3Sb but
334
P. DOLIZY, F. GROLI~REAND M. LEMONIER
(0)
experimental points
\
I
0
500
600
700
000
Wavelength (nm)
(a)
OOL
400
1
500
1 600
I
700
000
Wavelength (nrn
(b) FIG.2. Theoretical spectral reflectancesfor (a)differentcontents of NaSb in NazKSb and (b)a stratified film (50 A of Na3Sb and 1490 A of Na2KSb).
it can also contain more complex absorbing materials resulting from chemical attack of the substrate by the alkali metals or it may exhibit some alkali metal in excess (Goho, 1983). It is therefore of interest to look for the effectsof such a Na3Sb under-layer on the photoemission of the NazKSb upper-layer. The optical absorption of the N a S b phase is very high compared to that of the NaZKSb phase. As a result a large amount of light is absorbed at the interface of the film. The relative photoelectric quantum yield computed for such a system is given in Fig. 3 as a function of the NajSb under-layer thickness with a constant total film thickness (1200 A). The photoelectric yield is computed using the usual three-step process: absorption of photons, transport and escape into vacuum of the photoelectrons (Dolizy, 1982). In green light, the
335
TRIALKALINE ANTIMONIDE PHOTOCATHODES
0
I
I
100
I
I
I
400
300
200
I
500
f 10
Phase thickness (8)
FIG.3. Relativequantum yield of a NazKSb/Na$b photocathodeas a function of thickness of the NaJSbphase.
photoelectric yield of a NatKSb film of thickness 1200 A is reduced by about 30% with an intermediate Na3Sb under-layer of thickness 150 i\. One must also realize that the Na3Sb material is an n-type semiconductor. The energy band diagram for the film including the interface between the substrate and the Na2KSb film is assumed to be as shown in Fig. 4. The width of the band-bending region located at this interface is unknown, but it surely reduces the apparent escape length of the photoelectrons. It can be expected that larger photoelectron escape lengths will be possible when the film is free of Na3Sb and made of a homogeneous NazKSb material within its bulk. I
e-escape
_ _ _ - ---Limit -Loss
Le-
--
fore- e- interactions
of electron energy by interaction with valence electron escape
---
El = 0.3eV
~-Vacuum level E,, = 1.3eV
-_ Electron hole pair production
FIG.4. Electron energy band diagram for Na2KSb,Csp-type material with n-type surface and interface states.
336
P. DOLIZY, F. GROLIERE AND M. LEMONIER
ESCAPE LENGTHOF
PHOTOELECTRONS IN
HOMOGENEOUS PHOTOCATHODES
The photoelectric yield of NazKSb becomes high when it is covered with about one monolayer of caesium deposited on its surface (Dolizy and Grolikre, 1987; Dolizy, 1982). Before determination of the escape length of the photoelectrons in this material, its surface was Cs-treated. Experimental determination of the escape length of the photoelectrons at one given wavelength is usually made by using the ratio of the photoelectric yields measured in reflection and transmission modes (Dolizy et al., 1977). It was not always possible to carry out this procedure. In such cases, the probability for one photoelectron to escape into the vacuum has been assumed to be 0.5. The escape length of the photoelectrons is then deduced by fitting experimental spectral photoemission curves to theoretical ones (Dolizy, 1987). The values of the escape length determined for homogeneous NazKSb film in a direction perpendicular to its surface are given in Fig. 5. Several homogeneous films of different thicknesses have been considered. The escape length of the photoelectrons increases with the thickness of the film and can be as high as 850A in green light for a 2400A thick material. At 800nm wavelength, the photoelectron escape length reaches a value as large as 600 A in this material. Similar results have been reported previously (Denisev and Klimin, 1978). The shapes of the curves are similar to previously published ones (Dolizy, 1982). They are understood as follows: for photon energies higher than 2.3 eV, the electron energy losses are due to electron-phonon and electron-electron interactions. For lower photon energies, the electron energy losses can only be due to electron-phonon interactions. Hence, the escape length as a function of the wavelength of the incident radiation has a
1UUL
Wavelength (nm)
FIG.5. Escape length L of photoelectrons in homogeneous films as a function of wavelength.
TRIALKALINE ANTIMONIDE PHOTOCATHODES
2o
331
r-
t eot
I
1
1000
2000
3( )O
Thickness ( A )
FIG.6 . Photoelectric yield versus the thickness of the NazKSb,Cs layer for different values of the escape length L.
maximum at 540 nm (2.3 eV), which is slightly greater than twice the band gap of this material. The photoelectric quantum yield of this material can be computed as a function of its thickness, taking the escape length L of the photoelectrons as a parameter. Figure 6 summarizes some of the results computed for red light (800 nm), for escape lengths successively 300, 400 and 600 A. For normal photocathodes, the photoelectric yield at this wavelength is maximum for a film of thidcness 1200 A. This quantum yield will reach its maximum value only if the parameter L becomes large with respect to the thickness of the material. It can be also seen that the quantum yield of this material, with thickness 1200 A and L = 300 A, is comparable to that of a film 2400-2600 A thick with a larger escape length of 600 A. This property should be of interest for low-light-level image detection. CONCLUSIONS
Very large photoelectron escape lengths have been found in NazKSb,Cs material that is homogeneous in bulk. The photoelectron escape length increases with thickness and, consequently, photocathodes as thick as 24002600 A have a very high photoelectric quantum yield in red light. The blue light absorption of so thick a film is large and this film acts as a filter for blue
338
P.
DOLIZY, F. GROLIZRE
AND M. LEMONIER
Wavelength (nml
FIG. 7. Spectral photoelectric yield curves of NazKSb,Cs photocathodes: (i) 1300 A thick, 700 pA lm-I; (2) 2400 A thick, 500 pA lm-I.
light. A thick photocathode is therefore very photosensitive to red light but less sensitive to blue. These properties appear very clearly in Fig. 7, where curve (1) refers to the measured spectral quantum yield of a photocathode having a thickness of 1300 A and curve (2) corresponds to a photocathode of thickness 2400 A. The spectral response of the latter is considerably reduced in the blue range, which is of interest for improving the contrast of night-vision tubes (Richard et al., 1977). REFERENCES Denisev, V. P. and Klimin, A. I. (1978) Sou. Phys.-JETP 23,1388-1389 Dolizy, P. (1982). Philips Tech. Rev. 40, 19-28 Dolizy, P. (1987). Actu Electron. 27, 3-4 Dolizy, P. and Grolitke, F. (1987). Appl. Opt. 26,2401-2406 Dolizy, P., De Luca, 0. and Deloron, M. A. (1977). Acru Elecfron.20, 265-296 Goho, S. M. (1983). PhD Thesis, University of Florida Richard, J. C., Lamport, D. L., Roaux, E.and Vanneste, C . (1977). Acru Elecfron.20,4,353-368 Sommer, A. H. (1968). In “Photoemissive materials”. Wiley, New York
Relationship Between Microstructure and Photoelectric Quantum Yield (PQY) of S 1 Photoemitting Surfaces C. W. BATES, JR.,Q. Y. CHEN Department of Materials Science and Engineering, Stanford University. Stanford, California. USA
and
N. V. ALEXANDER Department of Physics, Stanford University, Stanford, California, USA
INTRODUCTION
Recently, we presented a theory for the photoelectric quantum yield (PQY) of S * 1 photocathodes that gives excellent agreement with experiment and involves essentially no adjustable parameters (Bates, 1981; Bates and Alexander, 1985; Chen and Bates, 1987). The theory is based on a combination of a dynamic effective medium approximation (DEMA) used to characterize the absorption of electromagnetic radiation in inhomogeneous composite systems (Alexander and Bates, 1984), which from electron micrographs is applicable to the S * 1 with the silver existing in particulate form, and a description of photoexcited electrons within the silver particles that involves elastic scattering of them from the particles’ boundaries, providing additional opportunities for them to reorient themselves to have sufficient perpendicular energy to escape near threshold and above (Chen and Bates, 1987). The specific microstructure we use to model the S - 1 photocathode consists of two basic units, silver particles surrounded by matrix and matrix surrounded by silver particles. In the base layer the matrix is Cs20 and in the surface layer of the S. 1 the matrix is vacuum. This microstructure is consistent with electron micrographs taken of S . 1 films and has been wed to successfully describe the transmission of Au-SiO:! cermets. We find that the broad peak centred at 800 nm in the infrared is due to the second unit of Ag clustered around the matrix (vacuum). This is consistent with some recent studies (Curtin et a/.,1985) which have shown that the anomously high absorption of small metal particles in the infrared is due to metal clusters. By varying the 339 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14 Copyright 0 1988 Academic Press Limited
340
C. W.
BATES, JR.,
Q.
Y.
CHEN A N D
N.
V. ALEXANDER
volume percentage of silver, particle size and shape and the relative portions of the two microstructural units, it is possible to vary the PQY in a prescribed fashion. This work suggests that in order to increase the PQY in the infrared it is necessary to change the microstructure in the base layer. BASICMICROSTRUCTURE AND THEORY The microstructures of the surface and base layers and the corresponding energy band diagrams are shown in Figs l(a) and l(b) respectively. The presence of the Cs1103, in addition to the Cs20, has been verified by X-ray photoelectron spectroscopy (XPS) (Bates, 198l), ultraviolet photoelectron spectroscopy (UPS) (Ebbinghaus et al., 1976) and surface enhanced Raman spectroscopy (SERS) (von Raben et al., 1984). The coating of the Ag particles by the CsI1O3was deduced from the SERS results. Extending a theory first proposed by Stroud and Pan (1978) using the condition that the forward-scattering amplitude of particles embedded in a medium should vanish on the average, we reformulated the treatment of composite media to optical frequencies and to a system of metal island spheres suspended in a dielectric. Our results were in good agreement with experimental transmission data from cermets of Au in Si02 for volume fractions from 10% to 80% Au and Si02 (Cohen et al., 1973). In modelling these composite media, we have found that in order to get good agreement with the experimental results, we had to take into account two types of particle: metal completely surrounded by dielectric and dielectric completely surrounded by metal. Our theoretical S . 1 was modelled by assuming the base layer (Csl coated Ag microparticles in a matrix of Cs20) to be 30 nm thick (Bates, 1981). This is consistent with S. 1 film thicknesses of 30-40 nm. In this treatment we assumed that the Csl 1 0 3 coating served the function of a low-work-function coating and did not contribute to the dielectric function of the matrix. This is reasonable considering its metallic behaviour (Simon, 1975) and will be true if it is thin enough. We used an &Dr& for Ag that included interband contributions from the work of Johnson and Christy (1972), along with a size, a is the particle radius (2.5 nm) dependent relaxation time given by a / v ~where and vF is the Fermi velocity (1.39 x lo8cm s-l). We set the volume fraction occupied by these 5 nm diameter particles to be 50% and set E = 4 for the Cs20. The 50% volume fraction was obtained from examining micrographs of Asao (1940), Soboleva ( I 959) and Wu (1 979). The top layer was given as Csl103-coatedAg microparticles ( 5 nm in diameter) in a medium with E = 1, of 10 nm thickness, and occupying 50% of the volume. The same &Dr& was used here for the Ag as in the base layer. In the base layer we assumed that half of the units were Ag surrounded by Cs20 and
Surface Layer I
(3nm to Snm d i m e t r )
A
ev
2.Oev
I 5
C 2 = l surface
-
FIG.I . (a) The microstructure of the S 1 photocathode composed of a surface layer residing on a base layer. (b) The energy band diagram of the separated layers of the S.1 photocathode. &= Fermi level. Eb = Schottky barrier height, E,= vacuum level, E,= bottom of conduction band of Cs20; y is the enhancement factor due to multiple scattering.
342
C. W. BATES, JR., Q. Y. CHEN AND N. V. ALEXANDER
0.BI
I\
Ag
I\
= 0.96 ( 1, = am) = 0.96 (f, = 0.02) yf,= l . O I f * = 0 ) f,
-
f,
0.5
15
Wavelength in Micrometers
FIG.2. The second unit effect on the absorption of the S.1 surface with microstructural parameters D=5OA, H=1001\, P=50%,f2=0,2,4,20% and &,=I.
that the other half were Cs20 surrounded by Ag. This was again obtained from the previous micrographs. We counted those structures that were almost completely surrounded by Ag as completely surrounded units. The total results (quantum yield) are somewhat insensitive to this choice. But in the top unit in the medium with E = 1, the final results are more sensitive to the choice of the fractions of units used, as shown in Fig. 2, though a fair range of choice is still possible. For a system with 98% of the units being Ag surrounded by air, one obtains an absorption that is a good replica of the quantum yield from 300 to 1200 nm. We note from Fig. 2, wheref, and f2 are the volume fractions of the coated Ag particles surrounded by matrix (vacuum) and matrix surrounded by Ag respectively, that f o r b = 0 (no clustering) the absorption at 800 nm is negligible as well as the PQY. Hence, the broad peak in the PQY at 800 nm is a structural effect due to clustering of the Ag particles, whereas the sharp absorption peak centred near 400 nm is due to the dielectric anomaly caused by the isolated Ag particles. We note that in the top layer the particles are small (3-5 nm diameter) and, as mentioned previously, a replica of the PQY. This is certainly not true for the base layer, where the absorption that is due only to 5 nm Ag particles does not reproduce the absorption of S. 1 photocathodes, as given by Asao (1940). This only points to the fact that whereas the quantum yield may be determined by particles of one size, the absorption takes place in particles of all sizes. This was noted by Wu (1979), who attempted to calculate the quantum yield of Ag particles in a matrix of
RELATIONSHIP BETWEEN MICROSTRUCTURE AND PQY
343
FIG. 3. The optical absorption of the S-1 photocathode. D=SOA, H=100A,P=SO%, E,,,= 1 for surface layer. R,= 700 A, H = 100 A, P=5O%,f2= 30%,~,,,=4 for base layer. D ranges from m 100 to 1000 A corresponding to a = 4 in the y-distribution.
f2=2%.
Cs20 using several adjustable parameters and found an “equivalent particle” size of 3 nm to fit his data. Figure 3 shows the absorption of the base and top layers as well as the total absorption of the S.1 photocathode using the parameters given previously, calculated by means of DEMA. In the top layer we assumed a 5 nm diameter particle size and a variation of particle size in the base layer, ranging from 10 nm to 100 nm diameter, that obeys the r-type size distribution function with mode radius Rm= 70 nm (Chylek and Srinastava, 1983). The r-type size distribution function is p ( R ) = Ra exp( - UR/Rm),
-
where a is a distribution factor taken as 4 in the size distribution for the base layer mentioned. The total absorption given in Fig. 3 is in excellent agreement with the result of Asao (1940). PHOTOELECTRIC QUANTUM YIELD(PQY)
Using a three-step process of photoemission (Fig. l(b)) in a free-electron model, we calculated the PQY of the S . 1 photocathode (Chen and Bates, 1987).In this calculation we assumed that the photoexcited electrons in the Ag particles are scattered elastically from the boundaries of the Ag particles and hence have the opportunity to reorient their motions to ones more favourable for emission into the CszO. This introduces an enhancement factor y due to
344
C. W. BATES, JR., Q. Y. CHEN A N D N. V. ALEXANDER
m
FIG.4. Photoemission (a) from a planar surface and (b) from a sphere.
multiple scattering (Fig. l(6)). Previous theories assumed an inelastic process. This may be viewed as a geometrical enhancement factor y due to the Ag particles considered as spheres, as shown in Fig. 4. The total PQY including contributions from the base and surface layers is shown in Fig. 5 and is in excellent agreement with experiment. This shows that the major contribution to the PQY comes from the surface layer. Any attempts to improve the infrared response must come from the base layer, as CszO has a work function of about 0.6 eV whereas the work function of the Csl,03-coated Ag particles on the surface is 1.1 eV (Fig. l(b)).
FIG.5. Comparison of the contribution of PQY from the surface layer and the base layer.
RELATIONSHIP BETWEEN MICROSTRUCTURE AND PQY
345
ACKNOWLEDGEMENTS This work was supported by the National Science Foundation under Grant No. ECS 85-20580. Facilities at the Center for Materials Research at Stanford University were used in this investigation.
REFERENCES Alexander, N. V. and Bates, C. W. Jr. (1984). Solid State Commun. 51, 331 Asao, S . (1940). Proc. Phys. Math. Soc. Jpn 22,448 Bates, C. W. Jr. (1981). Phys:Rev. Lett. 47, 204 Bates, C. W. Jr and Alexander, N. V. (1985) J. Opt. SOC.Am. 2, 1848 Chen, Q.Y.and Bates, C. W.Jr. (1987). Phys. Rev. Lett. 57,2737 Chylek, P. and Srinastava, V. (1983). Phys. Rev. B27,5089 Cohen, R. W., Cody, G. D., Coutts, M. D. and Abeles, B. (1973). Phys. Rev. B8,3689 Curtin, W .A., Spitzer, R. C., Aschroft, N. W. and Severs, A. J. (1985). Phys. Rev. Lett. 54,1071 Ebbinghaus, G., Braun, W. and Simon, A. (1976). Phys. Rev. Lett. 37, 1770 Johnson, P. B. and Christy, R. W. (1972). Phys. Rev. B6,4370 von Raben, K. U.,Lee, K. C., Chang, R. K. and Benner R. E. Jr. (1984). J. Appl. Phys. 55,3907 Simon, A. (1975). In “Crystal Structure and Chemical Bonding Inorganic Chemistry,” p. 60. North-Holland, Amsterdam Soboleva, N. A. (1959). Radio Eng. Electron. 4, 204 Stroud, D. and Pan, F. P. (1978). Phys. Rev. B17, 1602 Wu, Q. D. (1979). Acta Phys. Sin. 28,553
This Page Intentionally Left Blank
S 20 Photocathode Stability Considerations E. A. BEAVER University of California, San Diego, La Jolla. California. USA
L. ACTON, D. DOLIBER, E. DOZIER Science Applications International Corporation, La Jolla. California, U S A
and H. WENZEL EG&G. San Ramon, California, U S A
INTRODUCTION The history and development of the multi-alkali (S -20) photocathode has been widely reported in the literature (Dolizy, 1982; Sommer, 1968; Spicer, 1958). Here we report on stability measurements and techniques for the S-20 process used to produce the Faint Object Spectrograph (FOS) Digicon tubes for NASA's Hubble Space Telescope programme (Harms et al., 1984). The FOS S -20 photocathode requirements are somewhat different from the normal night-vision S -20 produced by the major image intensifier manufacturers. For night-vision applications the S.20 process has been adjusted to yield maximum quantum efficiency (QE) of > 5% in the 800 nm to 900 nm region of the spectrum. Generally there is a falling response shortward of 400 nm for this photocathode. The FOS photocathode specifications are: (i) high UV QE ( 2 2 5 % at 185 nm); (ii) modest red QE (0.1% I Q E I 2 . 5 % a t 800 nm); (iii) low dark current ( I 20 photoelectrons s - I cmP2at - 10°C); (iv) stability at all wavelengths (185 nm s 1 I 8 0 0 nm) for many years; (v) no QE deterioration after 100 hours of 51 "C bake. Requirement (v) is set by the elevated cure temperature required during tube encapsulation with the urethane potting material. Note also that the FOS dark current specification determines the limit on the maximum red QE. From Rome's (1 969) work, the FOS dark current specification requires a red cut-off wavelength of I940 nm. The red cut-off wavelength is defined to be that wavelength at which the QE is 0.001%. In practice this background specification constrains the FOS photocathode QE at 800 nm to be I 2.5%. Figure 1 shows the quantum efficiency (QE) profile for three FOS Digicon
-
341 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
348
E. A. BEAVER ET AL.
10
-
w 0
l 0 F8 DATA t
FII D A T A
- THEORY 0.10.
I
0
F12 DATA
x FOS SPEC € A = 0.45 ev I
I
I
I
1
J
FIG. I . The QE of FOS Digicon Tubes F8, F11 and F12 at fabrication as a function of wavelength. The solid curve is calculated from Spicer’s QE formula for EA=0.45 eV.
tubes. Also shown overlaid is a theoretical profile for the S.20 photocathode QE based on the formula derived by Spicer (1958) for an electron affinity EA = 0.45 eV. At wavelengths above 800 nm the photocathodes of Fig. 1 rapidly decrease in QE such that all have red cut-off wavelengths of less than 940 nm and meet the FOS background specification (iii) above. Even though the QE profiles shown in Fig. 1 are very similar and have essentially the same EA,long-term QE measurements show that while F11 is stable at 22”C, FS is unstable at 22°C. Also, of the three tubes, only F12 is stable at elevated temperatures up to 51°C. An analysis of the photocathode process records indicate significant process differences between the three photocathodes of Fig. 1. The increasing stability from F8 to F11 to F12 will be shown to be correlated with improving process techniques designed to enhance QE stability. The difficulty of long term S . 20 QE stability has periodically been reported by other researchers. Hoene (1972) stresses that too rapid a “synthesis” of the S.20 photocathode can lead to instability. McGee et al. (1969) discuss the S.20 QE decay problems associated with early Spectracon tubes. During high-voltage operation, surface bombardment by spurious background electrons caused internal evolution of absorbed gases in the Spectracon. This led to photocathode decay caused by enhanced
s * 20 PHOI’OCATHODE STABILITY CONSIDERATIONS
349
ion bombardment during high-voltage operation and slow chemical reactions of the photocathode with the evolved gases during storage. Oliver (1972a) points out that transferred photocathodes (as in the Digicon) are more sensitive to decay from gases than those made in situ. This characteristic, he states, is due to the general lack of free alkali metals present in a vacuum-transferred photocathode tube. Evidently, without excess alkali metals, the S.20 photocathode is the sole “getter” and hence suffers an accelerated QE loss. McMullan and Powell (1976) analyse the extreme sensitivity of the S.20 photocathode to atmospheric oxygen and water vapour. They set a photocathode life criterion for an S-20 image tube at 10%loss in photocathode QE at 640 nm over a period of three years. This is based on realistic atmospheric leak rates for their large format electronographic image tube. Presumably lifetimes for a more simply constructed tube (such as a Digicon) would be greater. It is the experience of one of the authors (Wenzel) of this paper, based on the fabrication of thousands of image tubes, that there are S.20 photocathode process techniques that lead to improved photocathode stability. Sommer (1968) discusses the wide variation in processes possible when developing the S -20 photocathode. In order to satisfy the FOS programme requirements, we have had to optimize the S.20 process for stability.
THEMODELFOR QE INSTABILITY Before covering techniques to stabilize the S * 20 photocathode, we discuss the character of the unstable photocathode. Figure 2 displays the F8 decay characteristic. Except for the first 0.6 year of life (during which time F8 accumulated 150 hours at 5 1 “C of accelerated ageing from the encapsulation cycle), F8 was stored at I 22°C with occasional (low-intensity) operation. From 0.6 to 2.7 years one notices that the QE decay is approximately fitted by a power-law function for any particular wavelength. Also, increasing wavelengths yield increasing rates of QE decay. This characteristic form of S.20 QE decay was pointed out by Decker (1969) in his work on 5 - 2 0 photocathodes degraded by contaminants. To understand the physics behind this phenomenon, we consider Spicer’s (1958) model formula (equation l),
where Q is quantum efficiency in %, EAis the electron affinity, Epis the incident photon energy, EG is the band gap energy, C=29, y = 1.57, Ec= 1 and all energies are given in electron volts, with the parameter of electron affinity Ea
3 50
E. A. BEAVER ET A L .
increasing with time. This is drawn as the solid curves in Fig. 2 for a model of EA linearly increasing from 0.58 eV at 0.6 year to 0.67 eV at the 2.7-year QE measurement. The fit of this simple model to the decay data is a reasonable characterization of the decay phenomena. Note that the decay at 800 nm for F8 is so rapid that the data beyond 0.6 year is outside the bound of the Spicer model. Here we are dealing with electron transitions to vacuum from impurity levels above the valence band but below the conduction band. In Fig. 2 the QE decay at 800 nm is the most sensitive indicator of QE stability. The decay rate at 800 nm is used by us as a criterion of specific tube stability. It is of interest to compare decay purely as a function of wavelength. This is plotted in Fig. 3 for various EA values in the Spicer model formula. Also plotted is the F8 data from the 0.6-year region (fitted by E~=0.58eV). The curves have all been “normalized” to the McMullan and Powell life criterion (10% loss at 640 nm). Thus at 800 nm the QE loss is 50% or five times greater than the 10% loss at 640 nm for an EA=0.45 eV. In other words, the McMullan and Powell life criterion translates to a 50% loss at 800 nm in 3 years for the FOS photocathodes. The physical model for the S -20 photocathode is that of a sodium bi-alkali photocathode (probably only several hundred iingstroms thick for the FOS S.20) with an added surface monolayer of caesium (Csorba, 1985). The
I
10 A
0.10
I
I
I
vdpx *-x-x
X
I V X
X
1 C 8OOnm
DATA
0 700nm DATA
0.01 L
0
X 600nm DATA - THFORY
1
0.4
1
L
,
0.8
1.2
I
I .6
t 7 6 0 n m DATA 660nm DATA V 560nm DATA
A
-2
2.4
2.8
Y E A R S F R O M MANUFACTURE
FIG.2. The QE decay of Digicon F8 for various red wavelengths. The solid curves are from Spicer’s QE formula for a decay model of EA linearly increasing with time.
s * 20 PHOTOCATHODE STABILITY CONSIDERATIONS
351
-THEORY A F8 D A T A
580
I
I
620
660
I
roo
I
I
740
780
!O
WAVELENGTH (nm) FIG.3. The solid curves represent the theoretical QE decay as a function of wavelength for several electron-affinity values. The F8 data points with 10 errors are deduced from the decay characteristic for F8 in the 0.6-year region (EA-0.6 eV).
addition of the caesium monolayer reduces the original bi-alkali photocathode electron affinity of 1 eV to -0.5 eV for the S.20 case. Note that the band gap energy for both cathodes is 1 eV. The caesium is believed to combine with the photocathode bulk material to form a surface dipole extraction field at the S 20 photocathode-vacuum interface. Oliver (1972b) states that the accepted mechanism for S.20 photocathode shelf-life instability is the slow chemical reaction of the photocathode caesium with the internal residual tube gases. Although room-temperature caesium evaporation from the S -20 photocathode is discounted by Oliver, undoubtedly significant caesium evaporation occurs at the elevated temperatures during the photocathode face-plate hot indium seal and final urethane potting encapsulation. It is these caesium reaction and loss mechanisms that lead to the increase in electron affinity seen in Fig. 2 for the F8 tube. From this model one sees the possibility that stability might be improved if an excess amount of caesium could be "loaded" into the S.20 photocathode. This excess caesium would allow for reaction and loss of caesium during final seal and encapsulation and subsequent tube life, while still maintaining the vital caesium surface layer. This additional stability may come at the expense of some other photocathode parameter such as efficiency. Also the added caesium must be suitably 0
3 52
E. A.
BEAVER ET AL.
bound to the photocathode to reduce caesium migration to other tube components. Otherwise this improved stability could be at the expense of increased tube background noise. FOS S.20
PROCESS
DEVELOPMENT
The basic process developed for the FOS Digicon is shown in the flow diagram presented in Fig. 4. Despite the considerable number of theoretical studies done on the S 020 it is still necessary to develop the process empirically. Although the process may look similar to a standard S *20photocathode-type process it has been modified in several key areas in order to produce a photocathode that will meet FOS requirements. The FOS Digicon is processed using a transfer-station-type technology. The photocathode substrate (Supracil) and tube body are loaded into a vacuum chamber and subjected to a high-temperature vacuum bake to degas
FOS S - 2 0 PROCESS FLOW DIAGRAM
u Co-Evaporation
Conversion to bi-alkali phase
11 . b A K + Sb C o -A Ed vd ap ed o r taot i o n N a K S b Layer
I
1
i Has t h e 4 0 0 nm response peaked ~
C s t NaKSb Conversion to
cooled to 1 6 0 ° C
C S t S b YO-YO
FIG.4. The key elements of the FOS S.20 process.
Yes
s - 20 PHOTOCATHODE STABILITY CONSIDERATIONS
353
and clean the substrate and tube body. The photocathode is then sensitized and the photocathode substrate is transferred and sealed to the tube body using a hot indium alloy sealing technique. The FOS S.20 process was developed by processing a number of experimental photodiodes and then reviewing the results obtained from each experiment.The positive results were then incorporated into the processing of the next photodiode. The S.20 process is begun by heating the transfer station to 200°C. An initial potassium-antimony “layer” is formed by the co-evaporation process. A potassium partial pressure is formed in the photocathode forming well and antimony is evaporated through this potassium “cloud”. The initial potassium layer is “n”-type with a hexagonal structure and chemical composition K3Sb. It has been widely reported that photoemitters that exhibit the highest sensitivity have a cubic structure and are “p”-type. This conversion can be accomplished by exposing the K3Sb layer to sodium. Some of the potassium is replaced with sodium and the layer changes both its chemical composition and structure. The bi-alkali photocathode has a cubic structure and is “p”-type with the chemical composition of NaZKSb. The Na2KSb layer is then exposed to an excess of sodium. It is our experience that by varying the ratio of sodium to potassium in the NazKSb layer, the spectral sensitivity can be adjusted. S-20 photocathodes with an excess of sodium seem to exhibit higher longwave response than those with an excess of potassium. The results obtained on the experimental photodiodes indicate that one of the factors influencing the long-term stability of the photocathode is the amount of sensitivity multiplication that occurs when the K3Sb layer is converted to NazKSb. We found that an acceptable sensitivity multiplication can be obtained if the K3Sb is antimony-rich. We were able to obtain the desired result by adding an additional antimony evaporation after the formation of the K3Sb. After the Na3KSb layer has “stabilized”, an additional potassium antimony co-evaporation is carried out to bring the bi-alkali phase back to the desired ratios of potassium, sodium, and antimony. Since the FOS S.20 must have excellent QE at 400 nm, the sensitivity is measured at 400 nm using a narrowband-pass filter. The potassium-antimony co-evaporations and sodium exposures are continued until the 400 nm response has peaked. The bi-alkali phase of the process is terminated on a potassium-antimony co-evaporation. The bi-alkali phase is then exposed to caesium. The addition of caesium converts the NazKSb photocathode to the S.20 photocathode. This conversion is marked by a dramatic increase in the photoresponse of the photocathode. This increase is particularly evident at the longer wavelengths. Prior to the addition of caesium, the photocathode does not exhibit much sensitivity beyond 650 nm. After the caesium reaction, the photocathode sensitivity extends beyond 900 nm. At the same time, the transfer station cooling is started. At 200°C the vapour pressure of caesium is too high and the
354
E. A. BEAVER ET AL.
binding energy is too low to allow the caesium to remain bound to the photocathode. The photocathode is repeatedly exposed to caesium depositions as the system cools. At approximately 160°C the photocathode has cooled sufficiently to allow the caesium to remain bound to it and a “stable” photoresponse is obtained. Once the S-20 has stabilized, it is exposed to alternating antimony and caesium evaporations. This phase of the process is known as the “caesium-antimony yo-yo”. The photocathode process is then ended with a final antimony deposition. The addition of antimony during the final phase of the process serves two purposes. (i) The additional antimony has proven to yield an increase in the long-wave length response of the S.20. This increase is particularly noticeable at 800 nm and beyond. (ii) The additional antimony seems to tie up any excess caesium that might be loosely bound to the photocathode. Also, this additional antimony at the end of the process yields a photocathode better able to withstand the additional heating necessary for the hot indium alloy seal and the tube encapsulation. It should be noted that the S.20 process should not be viewed as separate from the rest of the tube process, but should be considered as part of the entire tube process from the initial parts cleaning to the final product. If the tube needs any additional heating after the S-20 formation (such as during substrate sealing or high-voltage encapsulation), this needs to be taken into consideration during the development of the photocathode process. STABILITY CONSIDERATIONS
Because S. 20 stability problems were encountered with some of the early FOS Digicon tubes, it was necessary to carry out an investigation into the S.20 photocathode process as related to the long-term stability. The first step was to review the tube data and process carried out on the early tubes EM4, F8, F9 and F11 (see Table I). All of these tubes had exhibited some photocathode QE decay to varying degrees. At room temperature, EM4 was very unstable and over a period of a year decayed to a bi-alkali QE response function. F8 dropped in sensitivity at 800 nm by 50% in 3 months at 22°C. F9 dropped by 50% at 800 nm in 9 months at 22°C. The F11 QE was stable at 22°C (even out to 900 nm) over a year of measurements. After reviewing the photocathode processing records, the effects on the stability of the photocathode of several areas of the process came into question for the early tubes. There were three major areas of consideration: (i) the relatively poor first sodium reactions; (ii) the relatively high temperature at which the process was terminated; (iii) the lack of a final caesium-antimony yo-yo. In all the above tubes (except the F11 stable at 22°C) the photocathode process was terminated at system temperatures above 170°C. Indeed for these four tubes there is a perfect correlation between lower termination tempera-
s * 20 PHOTOCATHODE STABILITY CONSIDERATIONS
355
TABLE I S.20 tube stability performance7
QEat 800 nm
51°C QE change at 800 nm (Yn A/lOO hrs)
Na reaction
Tube
(%)
EM4 F8 F9 F11 753-7 755-7 757-7
0.115 I .05 0.19 0.57 2.0 2.2
- 75 -40 - 13 -2.4 - 1.9 12
+
Poor Poor Poor Poor Poor Good Good
758-7 77 1-7 772-7 F12 Goal
I .8 1.2 1.4 0.81 >0.1
-8 20 0 +2 20
Bad Good Good Good Good
1 .o
+
No.of No. ofCs Cs-Sb depositions yo-yos 1 6
185 180 170 150 151 145 148
9 14 13 11 15 9 16 16 16
Until stable
Process end temp. ("C) Comment
3 3 3 3 Until stable
141 150 150 147 5 I50
End on Cs End on Cs End on Cs End on Cs End on Sb End on 2 Sb Sb excess End on Sb End on Sb Same as 751-7 Same as 757-7 Same as 757-7 Sb excess End on Sb
t Tubes EM4, FS, F9 and FI 1 are Digicon tubes. Tubes 753-7 to 771-7 are smaller photocathode verification tubes. Tube 772-7 is a Digicon tube without the silicon diode array target. tures and greater stability. If the process is terminated at temperatures above 160°C, the photocathode will lose caesium as it cools and the resulting photocathode will be caesium-deficient. Also, if the final caesium-antimony yo-yo is omitted, the remaining caesium may not be sufficiently bound to yield a photocathode with long-term stability. Poor photocathode stability can be caused by a number of problems; only some of these are related to the photocathode process itself. The lack of cleanness of various tube components, including the photocathode substrate, the cleanness of the transfer processing station, and the handling of the tube after the photocathode has been processed can lead to QE decay in the photocathode. To determine whether the QE decay was caused by processingrelated problems, several experimental photodiodes were processed with various changes in the S-20 process. The results of these experiments can be seen in Table I for photodiode tubes 753-7 to 771-7. In Table I EM4, F8, F9, F11 and F12 are Digicon tubes of 304 cm3volume with a silicon target. 772-7 is a Digicon tube of volume 304 cm3 without the silicon target array. 753-7 through 771-7 are small photodiode tubes of 41 cm3 volume without silicon target array. These tubes are made from two ceramic
356
E. A. BEAVER ET A L .
rings of the 15-ceramic-ring Digicon tube body and are thus the same construction as in 772-7, except that they are shorter. Also, column 3 of Table I is the fractional percentage change in QE at 800 nm due to a 100-hour tube bake at 51°C. The first major process change was to continue the caesium reactions until the system temperature had cooled to approximately 150°Cand to introduce a final caesium-antimony yo-yo reaction to stabilize the photocathode and yield an increase at the longer wavelengths. To determine the long-term stability of these photodiodes, they were subjected to a 51°C bake for a long period. Since caesium is very active, the baking would accelerate chemical processes and any migration of caesium that was not sufficiently bound to the photocathode. Although there was some QE decay noted after a few hundred hours at 51”C, these photodiodes were much more stable than the early FOS Digicons (see column 3, Table I). The one photodiode (758-7) that exhibited QE decay at a rate higher than desired had a sodium reaction that was considered bad. As we have stated previously, our experience has shown there is a relationship between the amount of sensitivity increase during the sodium reaction and the resulting stability of the finished photocathode. Photodiodes 757-7 and 771-7 were processed using an extra antimony deposition after the first K + Sb co-evaporation. This was done to insure that the first antimony layer was antimony-rich. The S * 20 photocathode process remained the same except for the above changes. Both the photocathodes exhibited greatly improved stability when they were subjected to the 5 1°C accelerated life test. In fact, both of these photodiodes exhibited an increase in the 800nm response with time; photodiode 757-7 increased 26% in 236 hours at 51°C. During baking, these “over-processed” photocathodes appeared to return to a more efficient configuration. There was some concern about the influence of the volume of the Digicon tube on the photocathode stability, since the volume of the photodiodes was considerably less than that of the Digicon. For example, a different tube volume could lead to different partial pressures of the residual gases and alkali vapours, possibly resulting in different caesium reaction and loss rates for the photocathode. Prior to processing the FOS Digicon F-12, an additional photodiode (772-7)was processed using an FOS Digicon tube body instead of the smaller photodiode body. The same photocathode process that was used on photodiode 771-7 was used on 772-7. When subjected to the 51°C accelerated life test, 772-7 exhibited superior stability to the previous FOS Digicons, but did not exhibit the magnitude of the increase at 800 nm that was noted with photodiode 771-7. This might be explained by the increase in tube volume, but there are many other variables that could affect this and more experimentation will be necessary to determine whether the change in volume has an effect on the photocathode stability. We
s ’ 20 PHOTOCATHODE STABILITY CONSIDERATIONS 40
2
1
I
I
I
1
I
I
357 I
1
20
I-
a
a
t v)
I
0
0
0:
LL
W 0
z
a -20 I
0
-40
-
+ 771-7 800nm OE D A T A A F12 8DOnm D A T A
-
3
I
I
I
I
I
I
I
FIG.5. The fractional percentage change in QE at 800 nm wavelength as a function of time at 51°C for a few of the tubes listed in Table I.
are tracking the long-term 22°C stability of tubes 753-7 and 772-7; the stability of tube 753-7 dropped by 2% in 9 months at 800 nm, whereas that at 772-7 rose by 2% over 5 months at 800 nm. Figure 5 displays the percentage change in QE at 800 nm for these final process tubes. The tubes of Fig. 5 were heated to 51°C for the indicated number of hours. QE measurements are made at 22°C. Also plotted in Fig. 5 are the poorer stability results from the most stable early tube, F1 I . FOS Digicon F- 12 was processed as the final step of this development. The photocathode process was identical to that used on photodiode 772-2. The resulting S.20 was quite acceptable (see Fig. 5 ) and initial life test results indicate that the photocathode response at 800 nm is increasing slightly with time. CONCLUSION
Caesium-deficient photocathodes can lead to excessive shelf-life decay. Over-processing the S * 20 photocathode appropriately with excess antimony, sodium and caesium leads to improved QE stability with little reduction in QE for the FOS photocathode. The S.20 process must be viewed as a part of the entire tube design, process and life cycle. Our experienceis that these S.20 stability techniques can also be applied to
358
E. A. BEAVER ET AL.
the night-vision, extended-red photocathode, although in this case increased stability may come with the trade-off of lower sensitivity. ACKNOWLEDGEMENTS This research was supported by National Aeronautics Space Administration, Goddard Space Flight Center, Contract NAS 5-24463 and Science Applications International Corporation (SAIC) internal research funds. Melisa Galba and Steve Crumb of SAIC performed the massive series of QE measurements required.
REFERENCES Csorba, I. P. (1985). In “Image Tubes.” Howard Sams, Indianapolis Decker, R. W. (1969). In “Adv. E.E.P.” Vol. 28A, pp. 357-365 Dolizy, P. (1982). Philips Tech. Reu. 40, 19 Harms, R. J., Beaver, E. A., Burbidge, E. M., Hier, R., Allen, R., Angel, R., Bartko, F., Bohlin, R., Ford, H., Davidsen, A. and Margon, B. (1984). Proc. S.P.I.E. 4415,410426 Hoene, E. L. (1972). In “Adv. E.E.P.” Vol. 22A, pp. 374-375 McGee, J. D., McMullan, D., Bacik, H. and Oliver, M. (1969). In “Adv. E.E.P.” Vol. 28A, pp. 6 1-80 McMullan, D. and Powell, J. R. (1976). In “Adv. E.E.P.” Vol MA, pp. 427439 Oliver, M. (1972a). In “Adv. E.E.P.” Vol. 33A, pp. 27-35 Oliver, M. (1972b). In “Residual Gases in Electron Tubes”, (ed. T. A. Giorge and P. della Porta). Academic Press, London Rome, M. (1969). Proc. S.P.I.E. 14,42 Sommer, A. H. (1968). In “Photoemissive Materials”. Wiley, New York Spicer, W. E. (1958). Phys. Rev. 112, 114
Physical Model and Optimization of a Heterostructure Vidicon Target Based on Amorphous Hydrogenated Silicon F. SCHAUER Technical Academy, Bino, Czechoslovakia
M. JEDLICKA TESLA - Vacuum Engineering, Praha. Czechoslovakia
and J. KOCKA Instituie of Physics. Czechoslovak Academy of Sciences, Praha. Czechoslovakia
INTRODUCTION As in some other laboratories (Imamura et al., 1979; Shimizu et al., 1980; Jones et af., 1985), a vidicon with an amorphous hydrogenated silicon (a-Si :H) target has been developed by TESLA - Vacuum Engineering with the co-operation of the Physical Institute of the Czechoslovak Academy of Sciences (JedliEka et al., 1984). The tube, 25 mm diameter, has a promising performance: the target voltage for a saturated signal current is about 10 V for white light; dark current is a few nanoamperes (according to the applied voltage); the average luminous sensitivity to a 2856 K tungsten light source is 2100 pA lm-l, the wavelength of the spectral response curve maximum is about 660 nm with a quantum yield of 90%; the long-wavelength threshold is about 850 nm; the “gamma” of the transfer characteristic for signal levels between 3 and 300 nA is unity. The lag depends on the dark current value and varies from 5% to 15%. No blooming or other spurious signal has been observed. A comparison of spectral response curves of our amorphous silicon vidicon with those of other camera tubes targets is shown in Fig. 1. As usual, the a-Si:H target consists of three layers. An SnOz-covered glass face-plate has been used as the substrate. First, a properly doped holeblocking a-Si,N, -,film (1 5-30 nm) is deposited. The photoconductive aSi :H layer (2-4 pm)is deposited as the second layer by radiofrequency glow discharge decomposition of a suitable gas mixture (SiH4, B2H6, PH3 and N2). The last layer is the electron-blocking film of boron-doped silicon nitride (20 nm) or silicon carbide (400 nm). 359 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
360
F. SCHAUER, M.
JEDLICKA AND
J.
KOCKA
Wavelength (nm )
FIG. 1. Spectral response curves of a-Si:H and other types of vidicon camera tubes.
At the last symposium in 1983, we reported on a physical model of the heterostructure target for camera tubes and we used a model of the amorphous-crystalline CdSe-AszSe3 junction as an example (JedliEka and Schauer, 1985). Now, we analyse the fully amorphous heterostructure using a-SiN, :H,P/a-Si :H,B/a-Sic,: H,B target system. (See the schematic energy band diagram in Fig. 2.) THEORY
Let us examine the properties of a single heterojunction; the extension to the second junction, forming the three-layer target, will then be obvious. Fully amorphous heterojunctions, suitable for camera-tube targets, exhibit several important differences in comparison with their amorphous-crystalline
HETEROSTRUCTURE VIDICON TARGET
361
7€: c
AE;
E:
1
counterparts (JedliEka and Schauer, 1985). (i) There is a strong indication of the reduced probability of surface states in the fully amorphous heterojunction (Robertson and Powell, 1985). (ii) The potential distribution within the amorphous semiconductor is completely determined by the electron localized states distribution around the Fermi level, whereas in the crystalline semiconductorsthe Debye approximation, based on the exhaustion of mobile carriers, is a good approximation (Schauer et al., 1982). (iii) The Maxwell relaxation time is, in suitable amorphous semiconductors, several orders of magnitude longer than in useful crystalline semiconductors. A comparison of schematic energy band diagrams of the completely relaxed crystalline-amorphous and amorphous-amorphous heterojunctions without interface states is shown in Figs. 3(a) and (b),respectively. Also included are the potential V(x)and field strength, &x), distributions. The alignment of the Fermi levels and with the condition of charge neutrality were used as the main criteria to determine the junction properties. Variables describing the lefthand amorphous semiconductors in Fig. 3 are denoted by the symbol I; those describing the right-hand ones by the symbol 11. The discontinuities of both the conduction band AE, and the valence band AE,, which are important in blocking the injection of electrons and holes, are obvious in Fig. 3. POTENTIAL DISTRIBUTION In a previous paper (JedliEka and Schauer, 1985) we discussed in detail the possible stages of the relaxation processes in the heterojunction. The detailed calculations of the properties of the amorphous-amorphous heterojunctions
362
F. SCHAUER, M.
CRYSTALLINE 1s.]- 1
JEDLIEKAAND
J. K O ~ K A
AMORPHOUS s-11 AMORPHOUS s-n]
FIG.3. Comparison of the schematicalenergy band models and of the correspondingpotential V(x)and electricfield strength b(x)distributionsin (a) crystalline-amorphousheterojunctionand (b) fully amorphous heterojunction.
are given elsewhere (Schauer et al., 1987),but for the completely relaxed case, which is not too different from the actual regime of a real vidicon target, the potential distribution (for x < xo) is given by V = y e X p ( - - ) -x-x, l], (a- 1)
where a = exp[(xo-xl)/L] and L = (.c/e2NEF)"2,where E is the permitivity and NEFisthe density of states at the Fermi level. From equation (1) it is clear that the potential distribution in the amorphous semiconductor exhibits exponential dependence, whereas that in a crystalline semiconductor with shallow donors exhibits a parabolic dependence [(xg -x - 1)12,1 being the Debye screening length). This difference is of great importance for proper extraction of photogenerated carriers. The surface potential V,(x = xo) is
where V2 = V , + V , ( V , is the diffusion potential, and V , is an externally applied potential) is the value of V at x = x2. The band-bending voltage, responsible for lowering the barrier energy, UL!,,(see Fig. 2) may be calculated from the surface potential V,:
HETEROSTRUCTURE VIDICON TARGET
363
(3) DARKCURRENT The most probable process responsible for the formation of the dark current in heterogeneous semiconductor junctions with band gaps E g 2 1.5 eV is the generation of carriers from interface states (Rhoderick, 1980). Then the current density in the dark may be written as
J = J'-"-J"-'
= 2 evS!, IVL NkFexp (-AEb/kT) [exp(eU,lkT)-
11, (4)
where u is the thermal electron velocity, S, is the electron capture cross section, N , is the effective concentration of states in the conduction band of amorphous semiconductor, N E Fthe ~ ~density of states at the Fermi level and
AEb = Ebo-eUym,
(5)
Figure 4 shows the model current-voltage relationship and activation
m4Y los
10'
10'
18
10' loa Voltage V 1
Id
FIG.4. The model current-voltage I ( U ) and activation energy-voltage E,(U) dependenciesof the model heterojunction corresponding to equations (4) and (5) with (3). Parameter is /3 = &'L1/&''L1', T = 300 K.
364
F. SCHAUER, M.
JEDLICKAAND
J.
KOCKA I
-71
n
i
-8
c
a
Y
. I -
C
g
a
-9-
7
Y
3 -10
-12-
-1
10.5
0
2 1 Log Cvoltoge ( V 11
FIG.5. The measured current-voltage I( U ) and activationenergy-voltageEo(U )characteristics of the a-SiN,:H,P/a-Si:H,B/a-Sic,: H,B. (BO, B10 and B20 denote the &HQ concentration in SiH4 expressed in p.p.m. when producing the a-Si :H,B layer, BO is the light-soaked sample, see text.)
energy voltage E,(U) as calculated using equations (4) and (5) combined with equation (3), the parameter fi being given by
/?= E'N&/E"N!&!,,u/(u- 1) x 1, T = 300 K. In Fig. 5 are shown the functions I ( U ) and Eu(U ) measured at T= 300 K in the dark on targets composed of a-SiN,: H,P/a-Si: H,B/a-Sic,: H,B (20 nm, 2.5 pm and 300 nm). The parameter is the degree of boron doping (i.e. the p.p.m. concentration of B2H6 in SiH4) in the a-Si :H,B layer expressed as BO, B10 and B20. The functions were measured on samples provided by Sn02 and Cr electrodes, both in the steady-state and simulating by electron-beam scanning with a fast pulse generator and picking up the response by a sampling adaptor. The I ( U ) and Eu(U ) dependences were identical in both regimes, indicating that the built-in potential is approximately the same in both cases. There is a general trend for the current in the dark to increase with the extent of a-Si:H boron doping. It is interesting that the E,(U) dependence also changes with the boron doping and in Fig. 6 the E,(U) dependence for the B20
365
HETEROSTRUCTURE VIDICON TARGET
I
1.01
FIG.6. The measured activation energy-voltage E,(U) characteristic of the B20 sample taken from Fig. 5.
sample is replotted in linear co-ordinates. Three distinct regions are obvious: (1) with parameter /II = 1.3 x lo-*, (2) with p2 = 6.7 x (Supposing &I1/ d = 3 and N&= lOI9 eV-’ (Schmidt et al., 1985), we obtain quite reasonable density of states for a-Si:H, 4 x 10”and 2.4 x lo” eV-’ (Schauer and KoEka, 1985).) Region (3) starting at U = 35 V corresponds to a quasi-depleted amorphous semiconductor. In Fig. 5 there is also included a measurement of the sample BO’ after light-soaking by a heat-filtered halogen lamp with 550 nm low-pass filter at about 30 mW cm-2 for 50 h. This creates metastable states at energy 0.6-0.8 eV and shifts the Fermi level to deeper energies (Staebler and Wronski, 1977; Schauer and KoEka, 1985). Both Z(U) and &(U) behave according to the models and an extensive examination of the effect on vidicon performance is under way. PHOTOCURRENT
The spectral response curve (Fig. 1) corrected for the glass window reflectance is slightly depressed at short wavelengths by the absorption of the comparatively thin a-SiN wide-gap layer, and also by the lower value of the electric field strength at the interface. The long-wavelength threshold is due to the falling absorption coefficient of the photoconductive a-Si :H layer. Shifting this threshold towards longer wavelengths would be possible using a lower bandgap material, an especially promising one being the alloy a-Si,Ge,: H. The Z( U ) characteristics measured by steady-state and pulse methods are shown in Fig. 7; 450 nm-filtered white light was used and the parameter is
3 66
F. SCHAUER, M. JEDLIEKA AND J. KOCKA
FIG.7. Current-voltage I ( U ) characteristics under illumination by blue light (450 nm) for various doping levels (as in Fig. 5); Us is voltage value for saturation.
again the boron doping in the photoconductive central layer. At the energy of the light used, holes are responsible for the primary photocurrent, and the dependence may be described by a modified Hecht’s curve (Schauer et al., 1987). The modification is needed to allow for the non-homogeneous electric field distribution in the layer. From the figure it is obvious that the mobility-
TABLE I Comparison of properties of crystal-amorphous and fully amorphous target Properties
Crystal-Amorphous ~
Production Spectral response Potential distribution Relaxation of photoconductor within frame time Junction forming Mobility-lifetime in photoconductor
Fully Amorphous
~~
Rather complicated (two technologies) Difficult to optimize
Easy (one technology) Easy to optimize (changing bandgap)
Optimal (parabolic). Easy to deplete
Not optimal (exponential). Difficult to deplete
Yes
No
On side of crystal by free carriers
Exclusively by localized carriers
High
Low
HETEROSTRUCTURE VIDICON TARGET
367
lifetime product, reflected in the saturation voltage Us, is a strong function of the degree of boron doping (Ishioka et al., 1983). CONCLUSION
Using experience gained in our previous studies on crystalline-amorphous heterostructures, we have studied the behaviour of a fully amorphous heterostructure both theoretically and experimentally. The main qualitative results of the comparison are collected in Table I.
REFERENCES Imamura, Y.,Ataka, S., Takasaki, Y.,Kusano, C., Hirai, T. and Maruyama, E. (1979). App. Phys. Lett. 35,349-351 Ishioka, S . , Imamura, Y., Takasaki, Y., Ch. andNobutoki, S. (1983). Jpn J . Appl. Phys. 22, S221,461-464 JedliEka, M., KoEka, J., Kubelik, I., Stika, 0. and Stuchlik, J. (1984). In “Proc. 11th IMEKO Photon-Detectors Symp.” (ed.J. Schanda, K. H. Herrmann and I. Ungvari), pp. 72-77, IMEKO Secretariat, Budapest JedliEka, M. and Schauer, F. (1985). In “Adv. E.E.P.” Vol. 64B, pp. 447461 Jones, B. L., Burrage, J. and Holtom, R. (1985). In “Adv. E.E.P.” Vol64B, pp. 437-445 Rhoderick, E.H. (1 980). In “Metal-Semiconductor Contacts”, Oxford University Press, London and New York Robertson, J. and Powell, M. J. (1985). J . Non-Crysr. Solids, 77, 1007 Schauer, F., JedliEka, M. and Polcer, J. (1982). Phys. Status Solidi A70, 755-761 Schauer, F. and KoEka, J. (1985). Phil. Mug. BS2, L25 Schauer, F., JedliEka, M. and Polcer, J. (1987). (to be published) Schmidt, M. P., Bullot, J., Gauthier, M., Cordier, P., Solomon, I. and Tran-Quoc, H. (1985). Phil. Mag. B51,581 Schimizu, I., Oda, S.,Saito, K. and Inoue, E. (1980). J . Appl. Phys. 51, 6422 Staebler, D. L. and Wronski, C. R. (1977). Appl. Phys. Lett. 31,292
This Page Intentionally Left Blank
A New Extended Infrared Vidicon T. KAWAI, K. SUGA, K. MURAMATSU, T. OTAKA, K . ATSUMI and R. NISHIDA Hamumatsu Photonics K . K . , Hamamatsu, Japan
INTRODUCTION Demand for infrared vidicons having sensitivity in the 2-3 pm range has been growing rapidly. Their applications are not limited only to the wellknown thermography but are also expanding into new fields such as semiconductor materials inspection, integrated circuit inspection, observation of the light from laser diodes and so on. The reason the applications to semiconductors are growing is that Si and GaAs, now commonly used in the semiconductor industry, are transparent at these wavelengths. Conventional infrared vidicons use specially grown Pbs-PbO polycrystalline targets to cover the infrared wavelengths up to 2.4pm. This paper describes a newly developed infrared vidicon which accommodates a new type of target using an increased ratio of PbS to polycrystalline PbO compared with the conventional type. This target has been developed using a new sulphurization method and shows sensitivity as high as 200 pA mW-I or more at wavelengths up to 2.7 pm. Measurement of this vidicon has been made using an infrared spectrometer and an F-centre laser excited by a krypton laser. The results are presented below. The paper also discusses applications of the vidicon to, for example, inspection of integrated circuits, oscillating laser patterns, and so on.
TUBECONSTRUCTION The infrared vidicon (IR vidicon) is an image pick-up tube having as the sensor a photoconductive target. This is operated in the same way as a conventional vidicon with magnetic focusing and a deflection-coil assembly. The electron gun consists of a cathode (K), a beam control electrode (G),an accelerating electrode (G2), a beam focusing electrode (G3)and a field mesh (G4). Stray light from the cathode cannot reach the sensitive target because of a screen attached to the focusing electrode. 369 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright (01988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
370
T. KAWAl ET AL.
PHOTOCONDUCTIVE TARGET
The photoconductive target consists of a black, porous film of polycrystalline PbO and PbS. The target is formed as follows. A borosilicate glass faceplate is coated with a transparent SnOz electrode on which a PbO layer (bandgap 1.9 eV) is deposited under an oxygen pressure of about lop3Torr. It is then sulphurized by heat treatment at a temperature of 150°C in ambient sulphur vapour. X-ray diffraction analysis shows that the polycrystallinePbO is deposited as needle- or plated-like crystals which are a few micrometres in length and are preferentially oriented in the (1 10) direction. The polycrystalline PbS shows various orientations at random. Auger electron spectroscopy (AES) reveals that there is a twofold increase of sulphur and a little decrease of oxygen. The depth profile of sulphur shows that only a small quantity is distributed beneath the surface. The sensitivity and spectral response depend mainly on the sulphurization method. It is important to realize that the PbS diffuses uniformly and deeply into the polycrystalline PbO. As is well-known, pure polycrystalline PbS (band gap 0.3 eV) has too low a resistivity to operate in the vidicon mode, while a pure PbO substrate is not sensitive to longerwavelength radiation. The features of both crystals have therefore to be properly combined by substituting PbO by polycrystalline PbS. Heat treatment for several hours at 150°C is used to establish an ideal fusion instead of the more conventional heat treatment of a few minutes at a temperature higher than 300°C. The new sulphurization method helps to increase the content of PbS while keeping the high resistivity (about 10" R cm) needed to store the electric charge. The PbS content determines the sensitivity and the maximum detectable wavelength.
CHARACTERISTICS Spectral Response
Figure 1 show the spectral transmission of the vidicon face-plate with and without an SnOzcoating film. The spectral response of the IR vidicon is shown in Fig. 2. These measurements were made using an infrared spectrometer with a 150 W halogen lamp. Incident powers are measured using a thermocouple radiometer. The borosilicate glass and SnOz film make the transparency drop abruptly at 2.7 pm. It is very important to find a new window material for the IR vidicon which has a better transparency and MTF. Light-transfer Characteristics
The light-transfer characteristic of the vidicon is shown in Fig. 3. The
37 1
A NEW EXTENDED INFRARED VIDICON
Detector : Themcouple Radiometer Slit Width : Entrance
800
1200
1600
2 . b
2000
Wavelength
2800 (IIUL]
FIG.1. Spectral transmittance of the borosilicate glass used as the substrate of the vidicon target.
.3
a Wavelength (nm)
FIG.2. Spectral response of the IR vidicon. The parameter is incident radiation energy per cm2 at the surface of vidicon target.
parameter is the wavelength of the incident radiation. This characteristic was measured using the infrared spectrometer described above. The infrared vidicon was tested up to a wavelength of 2.6 ,urn. As shown in Fig. 3, the target gives y-values of about 0.8 in the wavelength region 800-2600 nm. Figure 4 shows a similar light-transfer characteristic measured using an F-centre laser
1
2
10
5
20
50.
100
200
Incident Enerey (vw)
FIG.3. Light-transfer characteristic. The parameter is the wavelength of the incident radiation.
O
L
I
.
'
I
'
500
I
'
I
'
1000
Pumping energy in Kr laser (mW)
FIG.4. Light-transfer characteristic. The wavelength of the incident radiation was 2.6 pm. Horizontal axis shows the pumping power of the krypton ion laser.
373
A NEW EXTENDED INFRARED VIDICON Crystal chamber
Chopping unit
L'
Turning arm
1
Input pump beam
-
Kr laser
T V monitor
Camera head
FIG.5. System for measuring light-transfer characteristic using an F-centre laser.
(Burteigh Instrument Co.).The set up is shown in Fig. 5. The output gives yvalues of about 1 .O, which is the same as that of the InSb detector used for monitoring. In general, the distribution of PbS affects the uniformity of spectral sensitivity, which will become significant when the camera is installed in industrial measuring instruments. y-Values vary from 0.6 to 0.8 from tube to tube and also become smaller when the incident light level exceeds that at which the signal output current is 200 nA.
Dark Current
The PbO-PbS target gives inherently large dark current. This depends mainly on the applied target voltage and the temperature of the photoconductive target. The target structure is not a blocking type like a Plumbicon but an injection type like a conventional vidicon. From our experimental results, dark current shows only a small drift over long time intervals provided that the temperature of the target is lower than 25°C and that the initial dark current is lower than 10 nA. In normal use, the temperature in the camera is 7 4 ° C higher than room temperature. If precise measurements are needed, it is necessary to maintain the temperature inside the camera constant.
374
T. KAWAI ET AL.
APPLICATIONS Semiconductors
Si and GaAs are currently the most important semiconductors in electronics. They are almost transparent to infrared radiation. Usually the surface, edge, back surface and inside of Si wafers are analysed by conventional methods. For integrated circuits, there are many necessary inspections of the circuit networks themselves and of the diffusion and oxidation processes. The IR vidicon is a very useful tool to investigate these parameters in a non-destructive manner. Figure 6 shows circuit networks observed using the IR vidicon camera. The A1 connections, the p- and n-type areas (diode, transistor, resistance), the A1-Au ball bonding and the Al-Si eutectics are clearly identified. These are observed through the back surface of the chips using an infrared microscope. Changing the wavelength by only 0.1 pm as shown in (a) and (b), enables the IR vidicon to distinguish between the energy levels of the conduction and valance bands because of the change in the transmittance of the infrared radiation. Infrared Lasers
Patterns obtained using an F-centre laser excited by a krypton laser (910 mW) are shown in Fig. 7. Intensity distributions are displayed on the left. The faint cross-shaped background is interference fringes due to the mesh electrode. There are many other applications in detecting light from LEDs, LDs and various lasers. Optical Communications
Infrared vidicons have been applied to measure the diameter of the mode field in single-mode optical fibres. These fibres are as small as 10pm in diameter and have a small refractive index, so that it is difficult to detect the interface between the core and cladding. This infrared technique is then useful to measure near-field patterns directly as the intensity on the output surface of the optical fibres. CONCLUSIONS The wavelength range of an infrared vidicon in a conventional camera tube operating at room temperature has been extended to 2.7pm. The new sulphurization method introduces a higher content of polycrystalline PbS throughout the PbO photoconductive target. The spectral response and lighttransfer characteristic measured by an infrared spectrometer and F-centre
FIG.6. Internal construction of integrated circuit obtained using an infrared microscope and infrared TV camera. The photographs show observations made at different infrared wavelengths.
FIG.7. Laser patterns emitted from an F-centre laser.
A NEW EXTENDED INFRARED VIDICON
377
laser has verified the results of the sulphurization process. This vidicon is useful for applications in the wavelength range 1 .O-2.7 pm. Applications include imaging and inspection of the characteristics of semiconductors, observation of optical components in communications systems, and characterization of various lasers. ACKNOWLEDGEMENTS The authors wish to express their thanks to Dr H. Tokiwa for the use of the F-centre laser. We would like to especially note the contribution of Messrs I. Shimizu, R. Nakamura, M.Ishikawa and Sawaya for their assistance during production and measurement of the camera tubes.
This Page Intentionally Left Blank
Avalanche-mode Amorphous Selenium Photoconductive Target for Camera Tube K. TANIOKA, J. YAMAZAKI, K. SHIDARA, K. TAKETOSHI, T. KAWAMURA NHK Science and Technical Research Laboratories, Setagaya. Tokyo, Japan
T. HIRAI and Y. TAKASAKI Central Research Laboratory. Hitachi Ltd, Kokubunji, Tokyo, Japan
INTRODUCTION It is well known that there are two types of photoconductive target layers: the injection type and the blocking type. In the case of injection target layers, a quantum efficiency as great as unity or more can be achieved, but it has the disadvantages of time lag and high dark current. On the other hand, blocking target layers produce little time lag and low dark current, both necessary for good picture quality; hence these are widely used in colour camera tubes. The only disadvantage of this type of target layer is its limited maximum sensitivity. In principle, its quantum efficiency is less than or equal to unity. The authors have studied in detail how to build a more sensitive camera tube using a blocking photoconductive target layer. It is found for the first time that an amorphous photoconductive layer of selenium produces avalanche multiplication. This has been used as a target layer in experimental 18 mm camera tubes. In spite of its blocking-type target, the camera tube shows high sensitivitya quantum efficiency greatly in excess of unity, the upper limit of the sensitivity of a conventional blocking target layer. Deterioration of the lag and resolution characteristics have not been observed. The additional noise produced by avalanche multiplication is negligibly small. This experimental target layer is expected to prove especially suitable for high-performance camera tubes. In this article, the characteristics of the experimental 18 mm tubes for standard television systems are presented. TARGETSTRUCTURE AND CURRENT-VOLTAGE CHARACTERISTICS A schematic diagram of the experimental tube target is shown in Fig. 1. The
379 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copynghi 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
K. TANIOKA ET A L .
380
SIGNAL ELECTRODE (SnOz) FACEPLTE ,-CeOZ SbzS3
-
i
I
INCIDENT LIGHT
FIG. 1 . Schematicdiagram of the experimental tube target.
photosensitive layer is of evaporated amorphous selenium. The thickness of the target layer is 2 pm.A thin layer of antimony trisulphide is deposited on the scanningside of the photoconductor to prevent electron injection from the scanning beam and to reduce the emission of secondary electrons. Between the selenium layer and the signal electrode, a thin layer of CeO:! (about 20 nm) is interposed to make the hole-blocking contact stable. These layers are deposited using an ordinary vacuum-evaporation method. This target can be operated as a blocking target.
TARGET VOLTAGE (V)
FIG.2. Signal current and dark current versus target voltage.
PHOTOCONDUCTIVE TARGET FOR CAMERA TUBE
38 1
Figure 2 shows the signal current and the dark current versus the target voltage (V,).Blue light was incident on the target. The signal current rapidly increases and eventually saturates with increasing target voltage. If the target voltage is increased still further, the signal current again increases rapidly. The quantum efficiency q of an amorphous selenium layer for blue light is 0.9 for an applied electric field of 0.8 x lo8V m-', as demonstrated by Pai (1975). Since the thickness of the target layer is 2 pm, this electric field corresponds to a target voltage of 160 V. Therefore, as shown in Fig. 2, ?= 10 at V,=240 V and q =40 at V,= 260 V, indicating very high sensitivity. In the region in which the sensitivity again increases rapidly, the dark current also increases. However, at a target voltage of 240 V, namely at a quantum efficiency of 10, the dark current is as little as 0.2 nA. These characteristics imply that some multiplicative phenomenon occurs in the amorphous selenium layer. PROPERTIES OF THE
MULTIPLICATIVE PHENOMENON
Figure 3 shows the effective storage capacitance versus target voltage for an experimental target. The capacitance was calculated from the signal current, which was generated by changing the cathode potential with pulses of 1 V for each target voltage. The capacitance does not depend on the target voltage. If electron injection occurred from the scanning beam to the photoconductor, the capacitance would increase with increasing target voltage. This implies that the multiplicative phenomenon is not caused by electron injection. Figure 4 shows the signal current for photogenerated holes and electrons versus the target voltage in an experimental tube. When the target is illuminated through the face-plate by blue light, holes are the dominant
0
TARGET VOLTAGE (V)
FIG.3. Effective storage capacitance versus target voltage.
382
K. TANIOKA E r A L .
1000 FACEPLATE
' 7 0 100 200 3 I TARGET VOLTAGE (V) FIG.4. Signal current for photogenerated holes and electrons versus target voltage.
carriers in the photogenerated signal current. On the other hand, when the target is illuminated from the scanning-beam side, electrons are the dominant carriers. Because the signal electrode is biased positively with respect to the scanning electron beam, the carriers are generated in the surface region; this is thin compared with the selenium layer, because of its large optical absorption coefficient for blue light, as demonstrated by Hagen (1984). Both the incident light intensities were adjusted so as to obtain the same values of the signal current at a target voltage of 60 V. The hole current and the electron current show almost the same characteristics up to a target voltage of 160V. However, at target voltages of more than 160 V, the hole current is much larger than the electron current. Figure 5 shows the current-field characteristics of experimental tubes with selenium layers of different thicknesses. The target layers are 1 pm, 2 pm and 3 pm in thickness respectively. At fields greater than 0.8 x lo8V m-I, the signal current as a function of the field increases with increasing seleniumlayer thickness. From all the experimental results mentioned above, it may be concluded that the multiplicative phenomenon in an experimental tube observed at target voltages of more than 180V in Fig. 2 results from avalanche multiplication in the amorphous selenium layer.
PHOTOCONDUCTIVE TARGET FOR CAMERA TUBE
1r ,
I
1
1
,
,
,
,
1
1
1
1
383
1
FIG.5. Current-field characteristics of experimental tubes with selenium layers of different thicknesses.
CHARACTERISTICS OF THE EXPERIMENTAL TUBE All the characteristics of the tube were measured at a target voltage of 240 V, under incident blue light, or at a quantum efficiency of 10, except for the light-transfer characteristics measurement and a colour camera test. Figure 6 shows the build-up and decay-lag characteristics of the experimental tube, with a signal current of 200 nA, a beam current of 600 nA, but without a bias light. Under these conditions, the decay-lag in the third field after turning off the incident light was 4.6%. This agrees with the capacitive lag of 4.6% calculated from a target-layer storage capacitance of 1600 pF and an electron-beam temperature of 3000 K. Thus, it may be concluded that the operating mode does not cause a photoconductive lag. Figure 7 is a monitor photograph of the EIAJ test chart-A obtained from the experimental tube. The limiting resolution is more than 800 TV lines, indicating that the same resolution is obtained as for a conventional 18 mm camera tube, e.g. the SATICON H4286D. The resolution of the tube does not depend on the operating conditions of the target layer. In other words, the resolution of the tube is limited by the size of the beam, so that an experimental
384
K. TANIOKA E T A L .
FIG.6. Build-up and decay-lag characteristics. Signal current: 200 nA. Beam current: 600 nA. The spot interval represents 1 field or 1/60 second.
FIG.7 . Monitor photograph of the EIAJ test chart-A obtained from the experimental tube.
tube using electron optics for HDTV (high-definition television) showed a limiting resolution of more than 1400 TV lines. According to the wave-form of the ITE grey-scale chart-I produced by the experimental tube, shown in Fig. 8, the difference in noise performance, both in the white and the black parts, is very slight. It is believed that the additional noise produced by the experimental tube is negligibly small. In the light-transfer characteristics measurement, it was found that y was
FIG.8. Wave-form of the ITE grey-scale chart-I. Vertical: 40 nA per division; horizontal: 5 ps per division. Video signal bandwidth: 4.5 MHz.
SCANNED AREA 6.6mm x 8.8rnm 1 3200 K
0.0I
0.I
I
ILLUMINATION ON TUBE FACE ( lux )
FIG.9. Light-transfer characteristics.
10
386
K. TANIOKA ET AL.
close to unity (0.95) up to a signal current of 300 nA with a light level incident on the tube face of 1.2 lx, but this decreased slightly when the intensity of the incident light was increased still further, as shown in Fig. 9. Generally speaking, a selenium evaporated layer is almost insensitive to red light, because of its low quantum efficiency in a low electric field (Pai, 1975). However, in a high electric field, of the order of lo8V m- ', high sensitivity can be obtained for light wavelengths up to 620 nm, which corresponds to the amorphous selenium band gap of 2.0 eV. This is due to increased photodetection efficiency at long wavelengths and to avalanche multiplication.
IMPURITY DOPING The experimental target layer operating in the avalanche mode tends to increase the number of white blemishes in the video image because of its large electric field (of the order of lo8 V m-I), To solve this problem, a change in the form of the internal field in the selenium layer was attempted, using a doping technique. A thin region of the selenium layer next to the signal electrode was doped with a small amount of lithium fluoride. As a result, the number of white blemishes decreased greatly. It may be concluded that the positive space-charge formed in the selenium layer by doping with lithium fluoride weakens the internal field between the signal electrode and the selenium layer. From a practical point of view, the selenium layer is also doped with arsenic to suppress crystallization. COLOURCAMERA TEST
The performance of experimental tubes using targets doped with impurities was evaluated in an ordinary three-tube colour camera. However, the target voltage supply and illumination were changed to Vl= 240 V and 180 Ix (iris settingfl4) respectively. As shown in Fig. 10, it was confirmed that the picture quality was almost equal to that of conventional studio colour cameras, which are usually used at a higher illumination (2000 lx, j74). CONCLUSION
The phenomenon of avalanche multiplication in an amorphous selenium layer was found for the first time at an applied field of greater than 0.8 x lo8 V m-'. This phenomenon was exploited in the target layer of an experimental camera tube. The experimental tube produced high sensitivity, about 10 times as high as that of conventional camera tubes, without any degradation of picture quality, i.e. with good dark current, lag characteristics, resolution and noise performance. This target layer will give greatly improved
PHOTOCONDUCTIVE TARGET FOR CAMERA TUBE
387
FIG.10. Monitor picture produced by a three-tube colour camera equipped with experimental tubes. The illumination is 180 Ix and the lens iris is at f/4.
performance not only in HDTV cameras but also in cameras for any other television system. The authors call an amorphous photoconductive layer operating in the avalanche mode “HARP”-High-gain Avalanche Rushing amorphous Photoconductor. ACKNOWLEDGEMENTS The authors would like to thank Mr T. Hirashima and Dr H. Matsumura for affording them the opportunity to conduct this study. They also express their deep gratitude to Dr M. Kurashige, Mr E. Hiruma, Mr Y. Ikeda and Mr N. Egami for their helpful advice and assistance.
REFERENCES Hagen, S. H. and Derks, P. J. A. (1984). J . Non-Crysf.Solids 65,241-259 Pai, D. M. and Enck, R. C. (1975). Phys. Rev. B11,5163-5174
This Page Intentionally Left Blank
An Electrostatic Deflection, Electromagnetic Focusing Pick-up Tube for High-definition Television H. ROUGEOT and J.-L RICAUD Thomson-CSF Division Tubes Electroniques, Boulogne Billancourt Cedex. France
OPERATING PRINCIPLES AND STRUCTURE OF TV PICK-UPTUBES In a vidicon-type pick-up tube, the image of the object is optically formed on a photoconductive layer which creates a pattern of electric charges corresponding to the distribution of brightness. The photoconductive layer is scanned with an electron beam emitted from an electron gun to sequentially discharge this pattern, and a charging current corresponding to the sequential discharging is extracted from the tube as a signal. The electron gun comprises two main parts: the electron gun per se, which produces the narrow electron beam, and the focusing and deflecting structure which generates the television scanning. The electron gun includes a cathode, indirectly heated by a tungsten heater, a first electrode which controls the beam current, and a second accelerating electrode which limits the beam section by means of a very small hole called limiting aperture. The second part of the electron gun is a combination of electrostatic and magnetic fields designed to focus the beam into a very small spot on the photoconductive target and to deflect the beam in the horizontal and vertical directions in such a way that the spot moves on the target according to the TV scanning. Thus we distinguish three main parts in a pick-up tube: the tube head (the photoconductive layer with its surrounding structure); the focusing and deflection mechanism; and the electron gun. HD TV Requirements
As the next milestone after colour TV, high-definition TV aims to raise image sharpness to cinema level. This requires two basic improvements. (i) An increase in spatial resolution, to typically twice the number of lines (1250 instead of 625 in Europe) and twice the number of points per line, and thus to four times the total number of pixels: this will allow a larger viewing angle for a given angular resolution, giving real interest in large-size TV screens because the effect will be more like that in a cinema, where the viewing distance is 389 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copynght 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
390
H. ROUGEOT AND J.-L. RICAUD
usually three times the picture height instead of five times for present TV systems. (ii) An increase of temporal resolution, which means better motion portrayal and suppression of defects such as flicker and interline twitter (these phenomena are detailed below). In addition, other aspects such as improved colour rendition, separate colour-difference and luminance signals, a wider aspect ratio (to fit to a cinema aspect ratio), and multi-channel high-fidelity sound have to be considered. The quality of portrayal of motion is limited on one hand by the sampling rate (and therefore the frame frequency)and on the other by the response time of the pick-up tube (due to both the photoconductive layer and the electron gun) which causes lag (loss of resolution for moving parts of the image) and comet-tailing (blurring of brightest parts, especially when moving). Flicker is the annoying detection of the sampling rate by the human eye, which detects a pulsed image; the solution is to increase the line rate. Interline twitter is particularly visible on horizontal parts of outlines, which are subject to vertical oscillation. The solution is to replace interlaced scanning by progressive scanning (where lines are sequential). As the pick-up tube is not a limitation for the scanning features such as the number of lines, the aspect ratio and the choice of interlaced or progressive mode, only two of its properties are relevant. These are its resolution for static parts of the image-higher resolution means better horizontal resolution (more points per line); and its dynamic resolution, that is to say its resolution for moving parts-an increase in dynamic resolution implies lower lag and reduced comet-tail effect. The increase of horizontal resolution has to be uniform over the image format, and requires a video bandwidth ( B ) of more than 20 MHz instead of 5 MHz presently. A simplified expression for the signal-to-noise ratio is signal S -- knoise
CB3I2’
where S is the sensitivity and Cis the output capacitance of the camera tube. It can be seen that the tube has to be improved towards higher sensitivity and lower output capacitance. Finally, the increased number of lines requires more precise coincidence of the images produced by the three tubes of the camera; this means that better registration is necessary.
BASICFEATURES OF THE THX 898 PRIMICON
To meet HDTV requirements, we have fitted our prototype tube with the most relevant technique advances: a tube head including an Se-Te-As
39 1
ELECTROMAGNETIC FOCUSING PICK-UP TUBE
TABLE I
Technical choices for the Primicon THX 896
Electrostatic deflection
Se-Te-As Photoconductive layer
Diode gun
~
High resolution Better resolution uniformity High sensitivity Low lag Better registration
X
X
X X
X
X
X
amorphous photoconductive layer and a low output capacitance structure, electrostatic deflection and magnetic focusing (Ricaud and Calais, 1986)and a diode-mode electron gun. Table I summarizes the contribution of each element to HDTV requirements.
TECHNICAL DEVELOPMENT Improvement of the Se-Te-As Photoconductive Layer
The Se-Te-As photoconductive layer, first introduced in Japan under the name Saticon, features intrinsic advantages for high-performance colour cameras: high resolution, due to its amorphous high-resistance structure; very low flare, due to its small surface reflectivity (this gives good image contrast); almost no shading, due to small dark current; illuminance response with gamma equal to 1, and well-balanced spectral response across the entire visible light range (this means that three-tube colour cameras would not need a dedicated tube for each red, green and blue channel); and low lag and little after-image. All these characteristics make the Se-Te-As layer a fairly good candidate for HDTV, but several points need to be improved particularly after-image and high light decay lag. The after-image effect is the retention of static images, especially when highly contrasted. High light decay lag is the retention of local high-intensity signals, which causes both after-image and comet-tail disturbances. Improvement in these features requires optimization of the layer internal structure. To reduce lag for better dynamic resolution, we have developed an integrated light bias device in our tube. This provides the following result: at 1000 nA beam current and 40 nA signal current, the decay lag at the third field (after 60 ms) is only 4%, compared to 8% without light bias. This is due to a 30 nA dark current induced by the bias light.
H. ROUGEOT AND J.-L. RICAUD
392
Optimization of the Twisted Arrow-pattern Deflection Electrodes
The arrow-pattern is the most widely used pattern for electrostatic deflection pick-up tubes because this sinusoidally curved arrow-tip pattern gives quite uniform crossed deflection electrostatic fields even near the edges (Hutter and Ritterman, 1972). Further improvements have been introduced, particularly twisting around the longitudinal axis of the tube (Ritz, 1973). A uniform twist causes a helical distribution of the arrow tips. This gives better deflection sensitivity and, most importantly, the value of this twist can be adjusted (for given focusing coil and deflection electrode dimensions) to reduce astigmatism. This is very important in improving resolution uniformity. We have investigated this approach and found that a variable twist provides better results because a local twist combines with the local strength of the magnetic focusing field, which has a non-uniform distribution along the tube axis. Figure 1 shows two examples of sinusoidal twists, but other types of shape could be advantageous.
mm
BE
70
68
58
40
30
28
I0
8 E '
98 * 180 *
278
a
E '
38
mm
88
78
68
58
48
38
20
IE
L
FIG.I. Two examples of sinusoidal twist.
8
'
393
ELECTROMAGNETIC FOCUSING PICK-UP TUBE
Computer-aided Design of the High-resolution Diode-Mode Electron Gun
A diode-mode electrode gun is characterized by a positive voltage applied to the first electrode, rather than a negative voltage as in a triode-type gun. This more recent type of gun features lower lag and higher resolution, but needs a more efficient cathode, such as a dispenser cathode. As HDTV requires higher resolution, the electron gun must have a smaller limiting aperture, without decreasing the beam current or increasing the beam divergence angle. In addition, the energy dispersion of beam electrons has to be minimized in order to reduce lag. The design of such a gun is quite intricate because the electron-optical effect of the electrodes depends not only on the applied voltages and geometric shapes, but also on space-charge. Fortunately, we now have computer simulation facilities which take this into account. We are now developing a high-performance electron gun by means of this computer-aided design, but there is also a need for technological improvements in the cathode and in the electrode-mounting process,
PERFORMANCE Table I1 summarizes the performance results we have obtained. Compared to a standard tube the sensitivity is 20% higher and the resolution at the centre of the image is 50% higher. The resolution uniformity, measured as the ratio between corners and centre, is raised from 33% to 56% at 800 TV lines frequency corresponding to 42 Ip mm-I on the photoconductive layer. The lag at a low signal level (40nA) is significantly better because of the improvement in the layer. Geometrical distortion is only 0.5% of picture TABLE I1 Performance
Feature Sensitivity (PA lm-') Resolution at 800 TVL (42 Ip mm-') Centre (%) Corners of the image (%)) Lag at low signal (%) Image distortion (registration) (%) Output capacitance (pF)
Classical 1-inch tube
THX 898
400 standard Se-Te-As
480
30 10 6 (light bias) I I
45 25 4 (light bias) 0.5 4
394
H. ROUGEOT AND J.-L. RICAUD
height becuase of the use of electrostatic deflection. Finally, the output capacitance is reduced to 4 pF by improvements in the tube head structure. HD CAMERA PERFORMANCE Table 111 compares the main performance features of H D cameras with those of classical cameras. The choice of scanning standard has a critical influence on these features, and several trade-offs must be considered. Basically, the listed values show that HDTV provides better static resolution and registration, but lower signal-to-noise ratio at a given optical aperture. There is therefore a trade-off between resolution and signal-to-noise ratio. Discussions are now taking place inside the European HDTV development TABLE 111 Comparison of HD cameras and standard cameras Feature Scanning standard Aspect ratio Sensitivity Signal-to-noise Static resolutionat 700 TVL horizontal vertical Dynamic resolution Registration
Classical camera
HD camera
625/50/2: 1 4/3 1000-2000 lux at A4 55-60 dB
1250/50/2:1 (1 : 1) 16/9 1000-2000 (3000) atfl4 40-46 dB (3 1)
15-25% 20%
40-50% (idem) 40% (60)
4
idem
0.4%
* (improved) 0.03% (idem)
project (EUREKA EU95 PROJECT) concerning the choice between interlaced scanning and progressive scanning. The progressive mode degrades the signal-to-noise ratio but significantly improves the dynamic resolution because of the doubled updating frequency. The question is whether the signal-to-noise decrease is acceptable, taking into account the eye’s response to this noise and the possibilities of raising the signal-to-noise ratio by filtering or other signal processing. CONCLUSION As Table I1 shows, pick-up tubes have far from reached their limits and significant progress is taking place owing to technological improvement and
ELECTROMAGNETIC FOCUSING PICK-UP TUBE
395
computer modelling. For the moment, the pick-up tube is the most suitable image sensor for HDTV cameras, by virtue of its very high resolution. Nevertheless, HD CCDs are already under development and will be used in the second generation of HDTV cameras. ACKNOWLEDGEMENT This work is carried out with French government support (DAII-SERICS) in the context of the EUREKA HDTV Project.
REFERENCES
Ricaud, J-L. and Calais, E. (1986). Proc. S.P.I.E. 702,67-70 Ritz, E. F. (1973). IEEE Truns. Electron Deoices ED-20,( 1 1) Hutter, R. G. E. and Ritterman, M. B. (1972). IEEE Truns. Electron Deoices ED-19 (6)
This Page Intentionally Left Blank
Surface Temperature Measurement of Small Objects by the Microthermovision Technique
v. RYSANEK Czech Technical University. Prague, Czechos[ovakia
INTRODUCTION
The decrease in dimensions of microelectronic elements has concentrated attention on methods of measuring their properties. Temperature is a very sensitive indication of non-uniformities and it is possible in many cases to find the source of a problem from temperature measurements. The basic difficulty is the accuracy of the measurements over very small areas. Such measurements cannot be made by contact methods, because the high mass of any thermosensor completely changes the heat dissipation of the measured,object. The thermographic method has a better chance, but has limitations in resolution and accuracy. The resolution depends on the optics of the thermographic system and on the accuracy of its thermosensitive detectors. For objects down to 15 pm it is possible to use a professional thermovision system with microscope attachments. Despite its very sophisticated construction, direct measurements are only qualitative, because there are no quantitative figures relating the output signal and the temperature of objects. In our research we used an AGA 680 Thermovision system with microscope attachments with magnification 5, 50 and 125 and a special attachment for calibration of temperature measurements.
TECHNIQUE THE MEASUREMENT The calibration attachment consists of two parts. The first was used for calibration of the sensitivity of the microscope and the second for temperature correction for small object angles. The units for calibration of sensitivity are depicted in Fig. 1. This comprises two pointed copper ends with two resistors R,, Rz built into the copper masses. The temperature of each point was measured by copper thermocouples th,, thz. The difference in temperature was controlled by two electrical power sources. The gap between these points was 50 pm and the copper was covered by black paint. 397 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
398
rnV
DC2
rnV
FIG.1. Equipment for calibration of sensitivity.
The temperature correction for small objects was made using a planar Gunn diode as a source of heat and an adjustable diaphragm, as depicted in Fig. 2. The diaphragm consists of two razor-blades with thin aluminium layers having small emissivity.The gap between the blades was adjustable from 10 to 250 pm. The width of the active gap in the Gunn diode was constant and was 200 pm.
W
3 15 p m
FIG.2. System for correction of temperature of small objects.
SURFACE TEMPERATURE MEASUREMENT BY MICROTHERMOVISION
399
RESULTS
The sensitivity calibration was measured for a microscope magnification of 50 x and for a temperature range of 20-150°C. The dependence of sensitivity on temperature is depicted in Fig. 3. It can be seen that the sensitivity depends on temperature and is between 1.6"C (at 60°C) and 1.2"C (at 150°C) per division. Similar results were found for 25 x and 125 x magnification (O.l°C and 9°C respectively). Noise was very significant when using the higher sensitivities (i.e. 1 or 2 per division). The higher differences of temperature may therefore only be measured with lower accuracy.
FIG.3. Sensitivity versus temperature.
Gap between blades ( p m )
FIG.4. Temperature correction versus gap width.
400
v. RYSANEK Correction of Temperature of Small Objects
The dependence of measured temperature on the gap between the blades is shown in Fig. 4 for a magnification of 50 x and a substrate temperature of 60°C. It can be seen that the correction below 100 pm is higher than 20% and
FIG.5. Thermograph of planar Gunn diode.
FIG.6.Normal thermogram of Gunn diode.
SURFACE TEMPERATURE MEASUREMENT BY MICROTHERMOVISION
40 1
FIG.7. Temperature profile across Gunn diode
becomes 65% at a gap of 20pm. In this measurement we used both the isotherm system of measurement and the profile temperature adaptor. The accuracy using the profile temperature adaptor was higher. Figure 5 shows an example of an isotherm measurement. Figures 6 and 7 show temperature distributions of the planar Gunn diode. Figure 8 is a microphotograph of the Gunn diode.
FIG.8. Photograph of planar Gunn diode
v. RYSANEK
402
TEMPERATURE DISTRIBUTION Using this technique we found anomalies in the gap between the electrodes of the planar Gunn diode. These anomalies were found to be higher than 3°C at the corner of the electrode system, as shown in Fig. 9. Because we did not find the maximum temperature to be in the centre of the gap, we calculated the temperature distribution profile by computer. The results are shown in Fig. 10. Good agreement therefore exists between the 70 -
Temperature
I
correction
t
68-
!!
3
c
E
b
E
$! 64-
’\
-
\
\
\
i
\
/-
6ZL
,Po0
‘0
I
I
I
I
I
150
Position ( p m )
FIG.9. Corrected temperature profile across Gunn diode.
-
FIG.10. Calculated temperature distribution.
SURFACE TEMPERATURE MEASUREMENT BY MICROTHERMOVISION
403
measured results and the theoretical model. The anomalies at one corner of the electrode system indicate unsuitable construction and technology in the Gunn diode device. Similar results were found by applying this technique to VLSI circuits to determine the optimum thickness of the interconnecting aluminium layers and to the power thyristor to study the emitor electrodes.
CONCLUSIONS Thermography of small microelectronic devices enables optimization of their construction and technology. The results are valid only if the limitations in resolution and sensitivity of commercial devices are carefully considered.
This Page Intentionally Left Blank
Anti-veiling Glare Windows for Third-Generation Image Intensifiers J. R. HOWORTH 21 Victoria Rd, Maldon, Essex. England
INTRODUCTION The earliest image intensifiers used glass input windows and S - 1 photocathodes. The development in the 1960s of the much more sensitive S.25 photocathode and of fibre-optic face-plates enabled the 3-stage electrostatic focus intensifier to have sufficient gain and resolution for many night-vision tasks. In the early 1970s, these tubes started to be replaced by microchannel plate intensifiers. These were called Generation 2 (Gen. 2) to distinguish them from the Gen. 1 3-stage tubes. Although many Gen. 2 tubes are electrostatically focused, a new category, the wafer tube, is merely plano-plano, with a high field to minimize the spread of electrons. One would imagine that the wafer tube would have been designed with a plain glass cathode window, for low cost and high quantum efficiency, but the fibre-optic, which is an essential design feature of electrostatic focus tubes, was carried over into Gen. 2 wafer tubes, because it offers an easy solution to potential problems with veiling glare and in lens design. The crucial breakthrough for GaAs photocathodes came in 1971, when Antypas (1972) discovered that GaAs could be bonded to Corning 7056 glass. While it was proved much later that GaAs could be bonded to a properly designed fibre-optic (Howorth, 1981), the design of virtually all GaAs photocathode image intensifiers (Gen. 3) has been “frozen” with Corning 7056. The optical transmission of fibre-optic face-plates optimized for GaAs is about 75%; cathode sensitivity on Corning 7056 is about 30% higher. The glass face-plate of the tube also serves as the final optical element of the objective lens, and special lens designs were needed for Gen. 3 tubes. Prototypes were made and tested in the mid 1970s and it was found that, although performance of the new Gen. 3 goggles systems was superb, there was a problem with veiling glare. 405 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rishts of reproduction in any form reserved ISBN 0-12-014674-6
406
J. R. HOWORTH
VEILINGGLARE Veiling glare was found to be particularly objectionable when a bright light was just off the edge of the field of view. A schematic sketch of a night-vision goggle is shown in Fig. 1. The field of view is generally chosen to be 40” from considerations of human factors. It is important to note that this field of view is determined by the cathode diameter and not by the lens; a larger cathode diameter would result in a wider field of view. At large angles, the lens hood reduces the light falling on the tube (vignetting) and various aberrations become much more significant. A typical goggle lens may only be transmitting 20% of the light at 45” compared with the axial case (Slegtenhorst, 1981). A thin metal film is used as an electrical contact between the photocathode and the outside of the tube; this behaves as a mirror which can reflect light from outside the normal field of view back towards the photocathode, as shown in Fig. 2. Many more reflections are possible than those sketched, and it is easy to see that a bright light outside the normal field of view can give rise to objectionable ghosts, phantoms and glare. The bull’s-eye or anti-veiling glare window was invented to counter these reflections (Stowe, 1983). Much later, in 1985, an objective test was devised and criteria of acceptability were defined in MIL-1-49428 (ER). The test set is shown in Fig. 3. A 100% signal on the telephotometer is set up with a continuous integrating sphere, after which a central portion of the sphere (having a 3 1” field of view) is removed and replaced with a “black hole”. The signal in the centre of this should be less than 1.5% as shown in Fig 4. This 1.5% is an overall figure with
CATHODE MCP ANODE
EYEPIECE FIG. 1. A night-vision goggle system
407
ANTI-VEILING GLARE WlNDOWS
/
A
Direct reflection
FIG.2. Reflection possibilities using plain glass windows.
1.2.10 - 4 GEL (4Cd I m2)
Lamp
FIG.3. Veiling glare test set.
contributions from the lens, tube face-plate and internal effects of the tube. While the remainder of this paper is specific to Gen. 3 wafer tubes in goggles systems, similar effects occur in all image intensifier and LLTV systems. The separate contributions of various components are discussed in the following sections.
Veiling Glare in the Objective Lens The veiling glare in state-of-the-art night-vision goggles is shown in Fig. 5. The test pattern is a black and white board providing even illumination to half
J. R . HOWORTH
408
100~/0 LIGHT INPUT 10%
or 14mm mm from the centre 7 6 5 4 3 2 1 1 2 3 4 5 6 7 FIG.4. Veiling glare criterion MIL-I-49428(ER) Jan 85.
100%
10% .*
1010
PROBE
=.. -I I+*...
LENS 0.1
0
DISTANCE mm 1 2 3
4
5
6
7
FIG.5 . Veiling glare in night-vision goggles objective lens.
the lens and a test probe of 0.5 mm aperture was scanned across the image plane. Compared with the requirements outlined in Fig. 4, the objective lens is contributing about 0.6% of the allowed veiling glare. This figure is probably better than average for afj1.2 lens and is a result of state-of-the-art antireflection coatings, computer-aided design and careful development over a period of 4-5 years.
ANTI-VEILING GLARE WINDOWS
409
Veiling Glare in the Intensijier
Figure 6 shows the veiling glare performance of a typical wafer image intensifier, using a half-tube illumination as for the objective lens (Stefanik, 1976). It can be seen from Fig. 6 that the veiling glare of a wafer image intensifier can be divided into two regions. In the first region, glare falls away by a factor of two every 0.2 mm or so. This is due to MTF effects and electron scattering effects. For a typical microchannel plate (MCP) with 60% openarea ratio, there is a 40% chance that primary photoelectrons will strike the NiCr-plated web of the MCP. Some of these will be lost, but a proportion, about 40%, will be reflected with a range of angles. The percentage reflected depends primarily on the atomic weight of the atoms at the surface of the MCP. Simple electrostatics dictate that all these scattered and reflected electrons will be returned to the MCP within a radius that is double the gap between cathode and MCP (Csorba, 1979). The standard production tube has a gap of about 0.25 mm, and hence it is not surprising to see the signal dropping away by a factor of 2 for each 0.2 mm that we move away from the source of primary photoelectrons. Further, the fact that about 8% of the signal remains at 0.5 mm is consistent with the observation that 40% of the primary photoelectrons have a 40% chance of being scattered at any distance up to 0.5 mm in a normal wafer tube. For a tube with a filmed MCP, which is standard for Gen. 3, scattering is increased because the whole surface of the MCP is covered, but it is reduced because the high-atomic-number NiCr has been replaced by a low-atomic-
Distance
ycl
4
6
lmm
Distance mm
.01
0
1
2
3
5
7
FIG.6. Veiling glare performance of a typical wafer image intensifier.
410
J.
R.
HOWORTH
number film of A1203 or SiOz. Short-range veiling glare of Gen. 2 and 3 wafer tubes is practically indistinguishable. Similar electron scattering occurs at the phosphor screen, with approximately the same probability as the input because the phosphor is covered with aluminium, which naturally has a surface coating of aluminium oxide. The gap between MCP and phosphor is typically 0.7mm, so that one expects to see glare falling away with a characteristic distance of 0.7 mm. In fact, veiling glare falls away more slowly. Three other factors have been neglected. (i) The backing layer of aluminium on the phosphor screen may have pinholes or cracks in it that allow light from the illuminated portion of the tube to feed back into the dark portion. (ii) Not all of the input light is absorbed at the photocathode, some of this light may be scattered into the dark portion of the tube. The GaAs photocathode of a Gen. 3 tube absorbs much more of the input light than does an S.25 photocathode. The importance of this effect can be explored by performing the veiling glare test with monochromatic light of different wavelengths, since photocathode absorption is wavelength-dependent. (iii) Electrons in the tube striking solid surfaces have a finite chance of generating X-rays. X-rays may be absorbed in the photocathode, giving rise to electron-hole pairs in proportion to the X-ray intensity. In Gen. 3 intensifiers, bright scintillations can be observed at distances of several millimetres from a signal. The relatively thick GaAs photocathode is expected to be a much more efficient X-ray detector than an S-25 photocathode. The combination of all these factors gives the wafer image intensifier an inherent veiling glare of about 0.2% compared with the requirement set-out in Fig. 4. Adding this to the contribution of the objective lens ( O h % ) , we can deduce that the maximum allowable contribution from the face-plate of the tube is 0.7%. ANTI-VEILING GLARE WINDOWS There are various different methods of manufacturing anti-veiling glare windows in use by several companies (Stowe, 1983; Sink, 1982; Howorth, 1984). That described here relies on the presence of various metal oxides in Corning 7056 which can be reduced by baking in hydrogen to form colour centres. The spectral transmission of Corning 7056, Schott ZKN-7 and Sovirel 747 after identical hydrogen reduction cycles is shown in Fig. 7. All three glasses have similar physical properties and can successfully be bonded to GaAs to make Gen. 3 photocathodes.
ANTI-VEILING GLARE WINDOWS
100
41 1
SChOtt ZKN-7
.........
10
.**'
Sovirel 747
.. 1 Corning 7056 01 -01
Wavelength in nm
500 600 700 800 900 FIG.7. Optical transmission of various glasses after similar hydrogen reduction cycles.
The necessary steps to make an anti-veiling glare window by hydrogen reduction are shown in Fig. 8 (Howorth, 1984). It is clear that an anti-veiling glare window can be produced by this technique very cheaply, and that the blackened light-absorbing area is inherently concentric and self-aligned to the photocathode region. The degree of blackening, density and depth of blackening is related to the temperature-time cycle of the hydrogen reduction process. The shape of the spectral transmission curve is related to the metallic oxide composition of the glass. For general production quality control, a maximum transmission of 0.3%for white light against an S.20 photocell was rather arbitrarily chosen. This corresponds to the curve labelled Corning 7056 in Fig. 7. It is anticipated that this level of attenuation for off-axis illumination, which is in any case reduced by vignetting in the objective lens, Shape
Blacken
Remove Unwanted Black Glass L----i
FIG.8. Production sequence for anti-veiling glare windows.
412
J . R. HOWORTH
will reduce the veiling glare contribution of the tube face-plate to much less than the 0.7% referred to in the previous section. RESULTS
Results obtained by customers are not always reported in a detailed way. So far, measurements at Oldelft are the most comprehensive, and indicate that this type of anti-veilingglare window is at least as successful in reducing glare as those of other competitors, and that there are certain advantages for processing in the monolithic construction. REFERENCES
Antypas, G . (1972). US Patent No. 3,769,536 Csorba, I. (1979). Appl. Opt. 18,2440-2444 Howorth, J. R. (1981). UK Patent No. 2,094,056 Howorth, J. R. (1984). UK Patent No. 2,165,691 Sink, R. (1982). US Patent No. 4,475,059 Slegtenhorst, R. (198 1). Private communication Stefanick, R. (1976). In “Proc. Electro-optical Design Conference”, Chicago, pp. 654-659 Stowe, G . (1983). US Patent No. 4,406,973
The Design of the Image Intensifier for the Faint Object Camera of NASA's Space Telescope R. P. RANDALL and B. WILD Thorn EMI Electron Tubes Ltd., Ruislip, Middlesex, England
INTRODUCTION In the Faint Object Camera (FOC) of the Hubble space telescope, images from the primary mirror are directed through two optical systems with respective apertures offl96 andfl48. At each focus, Thorn EM1 3-stage magnetically focused image intensifiersare used as prime detectors and imagebrightness amplifiers. The output signal from each intensifier is lens-coupled to a Westinghouse type WX 34323 40 mm EBS camera tube. The video signal from the camera tube is processed to extract and locate in the image single photoelectron events occurring in the first stage of the image intensifier. The assembly of the image intensifier, lens, camera tube and associated power supplies is referred to as the Photon Detector Assembly (PDA). ESA are contracted by NASA to build the FOC and the prime contract for the PDA development was placed by ESA with British Aerospace. Thorn EM1 were subcontracted to produce the image intensifier assembly and subsequently to encapsulate the Westinghouse camera tube. Figure 1 shows a photograph of the completed PDA unit. The principal virtue dictating the use of these intensifiers is their capability for VUV detection, as well as substantial detection efficiency at visible wavelengths. The specification for VUV detection required a photocathode quantum efficiency goal of 10% at a wavelength of 120 nm. Other essential features of the performance specification are a dark count of less than 10 electrons cm-* s- and a photoelectron gain exceeding lo6.The input window also has to provide an accurate reseau array for image location calibration. The image intensifier has to be encapsulated to meet space-approved conditions. This includes a careful choice of materials, especially to ensure that optical components in the telescope do not become contaminated by condensates; this is particularly relevant at VUV wavelengths. The tube assembly must also survive the environmental conditions of Shuttle launch. This can only be checked by ground tests which include acoustic and vibration
'
413 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
414
R. P. RANDALL AND B. WILD
FIG. 1. Photograph of Photon Detector Assembly on test at British Aerospace
qualification. General vibration tests were performed using a random noise spectrum as detailed in Table I, for a period of 1 minute for each axis of the tube. It also had to be demonstrated that the encapsulated assembly would tolerate a space environment for a considerable period of time without deterioration. This is particularly relevant to the methods adopted to contain the high voltages on the tube and is checked by extended thermal vacuum and partial discharge tests. A critical requirement was found to be that of ensuring that any charge leakage from the high-voltage elements of the assembly was conducted satisfactorily to an external grounding scheme. TABLE I Results of vibrational tests Frequency band (Hz) 2040 40-1 30 130-190 190-400
400-600 600-1200 1200-2000
Level (g2 Hz-’)
0.02-0.20 (roll on) 0.2 0.6 0.2 0.1 0.1-0.05 ( - 3 dB roll off) 0.05-0.02 (- 5.8 dB/octave)
41 5
DESIGN OF THE IMAGE INTENSIFIER FOR THE FOC GLASS
SWIVEL TARGET
FIXED TARGET
A
FOCUS ELECTRODE
MAGNESIUM
I
1 s t stage
I
2nd s t a g e
3rd s t a g e
FIG.2. Diagram of 3-stage image intensifier.
TUBEDESIGN Figure 2 shows a simple diagram of the intensifier, the elements of which have been fully described at a previous Symposium (Randall, 1966). Changes made to the basic tube design to meet the project requirements were as follows. The Input Window Magnesium fluoride was adopted as the input window material, providing the required transmission and detection sensitivity at a wavelength of 120 nm. The window is glass frit-sealed to a stainless steel flange. An array of opaque reseau marks is deposited on the inner surface of the window to be coplanar
All positional tolerances to 2 microns
Q
$.bmm
0.075mm 0.075mm Enlarged scrap view of reseaus FIG.3. Diagram of magnesium fluoride face-plate and detail of reseau array.
416
R . P. RANDALL AND B. WILD
with the photocathode subsequently formed on this surface. The window assembly is then argon-arc welded to the tube body. A detail of part of the reseau is shown in Fig. 3.
The Mica Target Supports The interstage elements of phosphor and photocathode are supported on 4 pm thick mica discs sealed to a metal flange using glass frit. The shape of the metal support rings was modified to reduce resonance of these membranes under vibration, as shown in Fig. 4. Steps also had to be taken to ensure adequate adhesion of the phosphor screens to the mica. Mica support flange
\
Sealingarea Optical’ image area
w
FIG.4. Shape of mica target support ring.
The Photocathode Photoactivation proceeds in two chambers, one generated by rotating the first target so that the photocathode surface faces the input window. Subsequent development involves activation in each of the three stages so that target rotation and a locking mechanism are no longer required. This simplifies the build of the tube, especially against the vibration hazard. It also enables higher-quality photocathodes to be produced. The progress of each photocathode can be monitored during activation and a longer throw from the antimony bead is possible because this is now introduced through a sidearm situated on the glass ring that is third away from the photocathode of each stage, resulting in improved cathode uniformity. To achieve the specified dark count, SbNaK photocathodes are used throughout the tube. No caesium is introduced, even in the latter stages, as this migrates to the first stage, causing unacceptable dark current. Typical achieved photocathode quantum efficiencies are shown in Table 11. Higher red quantum efficiencies can be achieved with this type of photocathode but have to be avoided to control dark count.
DESIGN OF THE IMAGE INTENSIFIER FOR THE FOC
417
TABLE I1 Typical achieved photocathode quantum
efficiencies. Photocathode sensitivity: 75 pm lm-’ Specification
(nm) 120 250 300 350 400 450 500 550 600 650 660
QE (min) 1
20 20 18 13
9 4 2.5 1.O 0.1 0.04
Actual QE 18.2 22.5 21.1 18.3 15.9 12.4 8.23 4.95 2.42 0.73 0.05
Phosphors A standard commercial Pa 11 ZnS phosphor is employed, but crystal size distribution is severely restricted so that both resolution and screen deposition characteristics are controlled. The electron/photon gain of each screen has to be at least 100 to meet the overall gain requirements demanded of the tube. This calls for exacting performance from the phosphor screens, with a high degree of protection from the photocathode-activation process. This is largely achieved by the production of a nearly impervious aluminium backing film on the phosphor layer. This is a very difficult technique, since the quality of the aluminium film depends on the starting quality of the nitrocellulose films on which the aluminium is deposited. These combined films have to be pin-holefree, but a controlled permeability to gases must be introduced to allow the escape of gas molecules evolved during removal of the nitrocellulose film or when outgassing the tube.
INTERNAL CHARGE CONTROL The relatively open design of the internal electrode structure in each stage of the tube requires that the potential of the glass walls be strictly controlled to minimize image drift, which is required to be less than 2 pm per hour. Some of the causes of internal wall charging are shown in Fig. 5 ( 4 : i, spurious electron
Photocathode
Phosphor
(a)
Bonded paper Silicone rubber
Conducting paint
-
Acrylic tube
Air gap
Silicone rubber conformal coating
Plated copper film Acrylic tube
FIG. 5. (a) Wall charging process in the intensifier; (b) early encapsulation scheme; (c) commercial tube encapsulation scheme; ( d ) encapsulation scheme adopted for space hardware.
DESIGN OF THE IMAGE INTENSIFIER FOR THE FOC
419
emission in the photocathode plane; ii, ion emission due to impact from spurious electrons emitted from various segments of the tube; iii, primary and secondary electron emission. These emission sources also contribute to the generation of spurious signals in the tube but can be eliminated by control of the internal wall potentials. In this respect, the charge flow in the medium surrounding the glass envelope is of major consequence. It was considered that, ideally, a near-uniform voltage distribution along the inner glass wall was required to invoke the desired image stability. In early tubes, charge-control rings were painted on theexternal walls of the glass rings and the tube was encapsulated in silicone rubber, as illustrated in Fig. 5(b). Redesign of the tube to reduce its diameter involved a reduction of the metal flange protusions into the silastomer. This led to an immediate problem in wall potential control that resulted in image drift and poor geometry. To overcome these problems, the scheme shown in Fig. 5(c) was adopted and resulted in excellent performance. This relies on an acrylic tube, grounded externally, to contain the high voltages placed on the intensifier and an air gap to stabilize the wall potential. Any gross longitudinal potential differences are rapidly dissipated by discharges in the air gap, as operating voltages are applied to the tube. This prevents local charge build-up and yields the required stable internal wall potential distribution. There is no problem with breakdown in the air gap along the tube because potential gradients are less than 2 kV cm-I. Since this arrangement utilizes discharges in the air gap for stability, the
TABLE 111 Properties of Feldex Two-part, pour in place, polyurethane curing at 40°C Low out-gassing elastomer Excellent low-temperature flexibility Low expansion coefficient; high breakdown voltage Tolerates long out-gassing prior to cure Low resistance material for controlledleak path in high-voltagesystems ESTEC space-approved Vibrational attenuation matched to launch vibration environment Suitable for optical systems subject to high thermal shock Specific gravity Refractive index Volume resistivity Breakdown voltage Coefficient ofexpansion
1.1 1.464 1 x 1OLo m 10.2 kV mm-I 50 p.p.m./”Cat - 50°C 182 p.p.m./”Cat 40°C
420
R. P. RANDALL AND B. WILD
system gave excellent performance up to an altitude of 5000 m and removed the need for any further attention to combat insulator breakdown. Similar charge-control concepts were required for space applications, but, clearly, air could not be used. This was replaced by a polyurethane encapsulant having a very low level of electrical conductivity, to provide controlled charge dispersion. This was developed at Thorn EMI, available under the trade name Feldex. Some of the properties of this space-approved material are shown in Table 111. The acrylic tube was retained, copper plated externally as part of the charge-control scheme, as shown in Fig. 5(4. THE HIGHVOLTAGE SCHEME The electrical scheme for the tube is shown in Fig. 6. In the PDA unit, the tube is operated in a permanent magnet. Tube focus is achieved by varying the potential of stage 1 with respect to stages 2 and 3 and by magnetic focusing across the third stage. For this purpose, a low-wattage focus trimming coil is incorporated into the magnet assembly. As shown in Fig. 6 , the interface connections from the intensifier to the PDA unit were required to be only three: earth, 12 kV and 36 kV. Originally it was intended that the dividing resistors to provide voltages for the other electrodes should be contained in the tube encapsulation. This proved to be unacceptable, since the transverse field gradients across the resistors caused resistance instability. This was thought to be due to migration of the conducting constituents in the inks forming the resistance element. These resistors were therefore modified in design by adding an electrostatic shield Output Intensifier
__
I
I
I
Input I
Resistor box
R, to R:, 500 Megohms each R, to Rs, : 300 Megohms each
C ZOO picofarads
FIG.6. Electrical scheme for the intensifier and resistor box.
DESIGN OF THE IMAGE INTENSIFIER FOR THE FOC
42 1
and were encapsulated in a separate box. This also had to contain some added passive components which, owing to the capacity of high-voltage cables between the tube and the box, were then needed to control the maximum potential excursions across each resistor in the event of an inadvertent shortcircuit of the 36 kV power supply.
CHARGE CONTROL ON
THE
OUTPUTWINDOW
Charge control at the output window proved to be a difficult problem. The tube operates with the input grounded and 36 kV, nominally, at the output. Overload criteria called for operation at up to 43 kV. This voltage, on the output window, had to be contained within the general grounding scheme. As all 13electrode-feed leads emerged from the input end of the tube, the problem was simply to close off the acrylic cylinder across the output window of the intensifier. Various schemes failed that involved sealed inserts of glass windows into the end of the acrylic tube to provide optical access to the output. Despite attempts to minimize field gradients at various interfaces and
+
Acrylic tube
-Feldex -Glass -Feldew
R3 optical window
R3
Acrylic optical window Tube output window
R7 Feldex (b)
FIG.7. (a) Acrylic “pot” for high-voltage containment; (b) charge-leakage control window assembly.
422
R. P. RANDALL AND B. WILD
to maximize seal path length, breakdown could not be prevented. It was eventually concluded that no system with cemented components and interfaces would prove satisfactory. The design that was finally adopted relied on the use of a totally enclosing acrylic “pot”, turned from solid, as shown in Fig. 7(a). The optical area was formed by high-grade turning and polishing, but imperfections in these surfaces were still unacceptable for optical transfer of the output image. The mounting method for the tube involved cementing the output window directly to the output charge-control window incorporated into the acrylic tube. Because the tube assembly has to resist thermal cycling over the range - 50°C to +40”C, thermal expansion differences between the glass output window and the acrylic window that had been evolved were a major problem. Both these problems were overcome by interposing as the cementing agent a 0.5 mm thick layer of Feldex between these components. This material is sufficiently flexible to absorb expansion differentials and its refractive index is such as to minimize optical effects due to the poor optical quality of the acrylic window. A further glass window was added externally in a similar manner and the electrical conductivity of the Feldex layer was utilized to conduct charge leakage to the ground plane provided at the periphery of this window. This assembly is shown in Fig. 7(6).
FINAL ASSEMBLY After completion of the previous stage, the void between the tube and acrylic cylinder is filled with the Feldex encapsulant. This assembly is then mounted in a titanium tube having bearing points at each end that are used to clamp the assembly in the magnet, as shown in Fig. 8. The intensifier assembly in the acrylic tube is rigidly attached to the titanium tube only at the output Sealing point for tube assy.
-+-+FIG.8. Assembly of encapsulated intensifier into titanium case.
1
DESIGN OF THE IMAGE INTENSIFIER FOR THE FOC
423
FIG.9. Photograph of completed intensifier.
end but is restrained from lateral movement at the input under vibration by several small silastomer pads inserted between the titanium cylinder and the acrylic tube. This provides compensation for differentials in thermal expansion along the axis of the tube so that the input image plane remains substantially fixed during temperature excursions. The equipment is completed by making electrical connections and encapsulation of the resistor box. The finished intensifier assembly is shown in Fig. 9. The tube then undergoes vibration tests followed by thermal vacuum, during which some performance tests are carried out. Finally, the tube is subjected to rigorous performance tests, after which it can be delivered to the contractors for further integration in the Faint Object Camera build. Two flight tubes are now integrated into the space telescope systems awaiting launch. ACKNOWLEDGEMENTS The authors wish to thank ESA, British Aerospace plc, and the Directors ofThorn EM1 plc for permission to publish this paper. This project was started in 1977 and has involved a number of engineers who contributed to the development of this equipment. Although we have not named them, the authors would like to acknowledge that the foundations for the achievements reported in this paper are based on their excellent team work over a period of eight years.
REFERENCE Randall, R. P. (1966). I n “Adv. E.E.P.” Vol. 22A, pp. 87-99
This Page Intentionally Left Blank
Evaluation of Photon-event-counting Intensifiers R. W. AIREY, T. J. NORTON, B. L. MORGAN The Blackett Laboratory, Imperial College, London, England
P. D. READ Royal Greenwich Observatory. Herslmoneeux, Sussex. Engiand
and
J. L. A. FORDHAM University College London, Gower Street. London, England
INTRODUCTION The design and manufacturing procedures for a 40 mm microchannel plate image intensifier for use with an image photon-counting system have been described elsewhere in these proceedings.? This detector has been developed jointly in a collaboration between Imperial College, London, The Royal Greenwich Observatory and Instrument Technology Ltd of Hastings, Sussex. The later stages of the programme have involved the participation of University College London in measurements and observing trials using the intensifier. It is proposed to consider here some of the measurements that have been carried out to evaluate the performance of the detector for photon-event counting.
CALIBRATED FAINTLIGHT SOURCE A test laboratory has been set up to measure the characteristics of low-lightlevel detectors. Device operating parameters such as gain, resolution, pulseheight distribution, noise and detective quantum efficiency can be measured. Most of the equipment required to make these measurements is readily available, but it was necessary to develop a special light source to provide the low light-flux required for detectors operating at photon-event-counting levels. If a small tungsten lamp, operating at a colour temperature equivalent to
t See p. 41. 425 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copynght 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
426
R . W. AIREY ET AL.
white light is used as the primary source, it is necessary to provide an attenuation of loioin order to produce a sufficiently low light-flux (Jelley, 1982). Neutral-density filters can be used to cut down the light, but there will be an uncertainty as to the effective density of the combination owing to scattering and multiple reflections between filters. An integrating-sphere attenuator (Goebel, 1967; Schneider and Goebel, 1981) has been designed to provide the required low light-flux without the use of a neutral-density filter stack. The system is shown in Fig. 1. A small quartz halogen lamp (30 W at 6 V) run from a stabilized power supply is used as the primary light source. These lamps have been found to be highly stable and may be run for more than a thousand hours without changing their output. The colour temperature of the lamp must be maintained at 3000 K instead of the more usual 2854 K in order to retain the action of the halogen filling that prevents discoloration of the walls. Light from the standard lamp enters the integrating sphere and strikes a low-reflectivity Lambertian scattering disc supported on a thin rod at the centre of the sphere. The reflected light from the disc impinges on the walls of the inside of the sphere, which are coated with a high reflectivity matt barium sulphate paint. Luminous flux diffused within the sphere illuminates the rear face of the disc, which in turn acts as a source to illuminate a small output aperture. The required attenuation is easily achieved by suitable choices of scattering disc coating and the sizes of the input and output apertures. The sphere provides an attentuation of lo7and the light
-
-
U
TUBE
INTEGRATING SPHERE
ING
DISC N D. F I L T E
H
OUARTL-HALOGEN LAMP
FIG.1. Integrating-sphere low-light-level source.
427
EVALUATION OF PHOTON-EVENT-COUNTING INTENSIFIERS
-
output can be adjusted to lower levels by the use of a single neutral-density filter. A light flux lo5photons cm-* s-’ can be obtained with the system. PARAMETERS OF PHOTON-EVENT-COUNTING SYSTEMS
Measurements have been made of a number of critical parameters that are of particular importance for photon-event-counting systems and that are often not quoted by commercial manufacturers. Pulse-height Distribution
In photon-event counting each scintillation event is recorded in position only and no account is taken of the relative weight of detcted events. However, it is still necessary for a detector to have a narrow distribution in the weight with which output events are recorded. These events should be clustered around a modal value well separated from the system noise to allow a discriminator level to be set in the electronics so that most of the recorded counts are signal events. Measurements have been made of the pulse-height distribution of the 40 mm MCP intensifier running with the microchannel plates operating at saturated gain. These measurements were made using the image photon-counting system (IPCS), developed for the La Palma observatory, using special software that re-configures the instrument so that it behaves as a pulse-height analyser. A well-peaked pulse-height distribution is obtained (Fig. 2). This result has been confirmed, by measurements with an
zto-
I
1I
I/
m
c 2120 0 U
70
9% 60
0
FIG.2. Pulse-heightdistributionfor 40 mm MCP intensifier showing number ofcounts plotted against event amplitude expressed on an arbitrary scale in ADU (analogue digital units).
428
R. W. AIREY ET AL.
5 0
160
150
200
FIG. 3. Pulse-height distribution of EM1 9912 four-stage magnetically focused cascade intensifier.
optical-fibre bundle attached to a photomultiplier and a pulse-height analyser. For comparison, Fig. 3 shows the pulse-height distribution of an EM1 9912 four-stage magnetically focused cascade intensifier measured with the image photon-counting system. There is no well-defined saturated gain peak. Counting Eficiency
The counting efficiency of a photon-event-counting detector is directly related to the detective quantum efficiency and has been measured for the 40 mm MCP intensifier by counting detected events with a photomultiplier when a known low light level is applied to the tube input. Figure 4 shows the results obtained for two prototype intensifiers. It should be noted that the two tubes show very different MCP counting efficiencies. This is due to differences in the thickness of the silicon dioxide barrier membrane applied to the input surface of the first microchannel plate of the chevron pair. It is expected that improvements to the microchannel plate conditioning procedures adopted during tube processing will enable MCP intensifiers to be made without the use of an ion-barrier membrane. This will lead to a dramatic improvement in MCP counting efficiency, since, for an MCP tube with proximity focus between the photocathode and the MCP, counting efficiencies as high as 6070%can be obtained (Eschard, 1976; Fraser, 1983). Photoelectrons scattered from the MCP web tend to enter adjacent channels in the channel plate and are not lost. This is not the case for plates coated with a barrier membrane. For
429
EVALUATION OF PHOTON-EVENT-COUNTING INTENSIFIERS
I l a 20
z
I-z 3 0
10
0
ioa
200
300
4ao
son
600
700
am
INPUT S T A G E VOLTAGE
FIG.4. MCP counting efficiency plotted against the input stage voltage for two prototype 40 mm MCP intensifiers.
a four-stage cascade intensifier, measurements of counting efficiency indicate that about 60% of the photoelectrons from the primary photocathode produce detectable events. Signal-induced Background An important factor affecting the detective quantum efficiency of a detector, and the linearity with which signal brightness is recorded, is the signal-induced background (SIB) (Delori et al., 1972). This is an increase in noise at the output of the device in the presence of an input signal and may arise from a number of dissimilar causes.
Optical signal-induced background
Light entering the device may suffer several reflections in the intensifier input face-plate or inside the vacuum envelope. This light returning to the photocathode gives rise to the emission of photoelectrons which are now no longer spatially related to the image information, and hence are noise. Imperfections in the aluminium backing on a phosphor screen will permit feedback of light that can find its way back to the photocathode. In a microchannel plate image intensifier using a pair of chevron microchannel plates, with each plate cut at 8" bias angle, the microchannel plates behave as
430
R. W. AIREY ET AL.
an optical attentuator with an equivalent density of three for light at normal incidence. Optical feedback from the phosphor screen in these tubes is virtually negligible for screens with well-prepared aluminium backing layers. Scattered electrons
Signal electrons may be scattered from a phosphor screen, or from other structures within an image intensifier by elastic and inelastic scattering processes. These electrons will appear as noise in a different part of the detector output.
Ion events When an electron strikes an internal surface in an image intensifier, positive ions can be desorbed which are accelerated back to the photocathode and contribute to signal-induced background by producing ion-induced electron emission. SIGNAL-INDUCED BACKGROUND MEASUREMENT Signal-induced background is difficult to measure and may not be uniform across the output of the detector. Furthermore, the result of any particular measurement may depend upon the measurement technique adopted. For the 40 mm MCP image intensifier, a method of measurement due to Cromwell et al. (1985) has been used. In this method, a 2 m m wide strip across the photocathode is blacked out and uniform low-level illumination is applied to the intensifier input. Using a small fibre-optic rod coupled to a photomulti-
u PULSE
HEIGHT
I A Id P I 1 F I E R
FIG.5. The system for signal-induced background measurement.
EVALUATION OF PHOTON-EVENT-COUNTING INTENSIFIERS
43 1
plier (Fig. 5 ) , three measurements are made of the intensifier output: (i) in the dark region corresponding to the image of the blacked out strip ( S d ) ; (ii) in the and (iii) with the input illumination switched off (SJ. illuminated region (sb); The signal-induced background ratio 2 is given by
In general, signal-induced background causes a loss of contrast and resolution in intensified images and reduces the detective quantum efficiency of photonevent-counting detectors by a factor (1 z). The narrow strip measurement procedure approximates fairly closely to the signal-induced background produced when the whole surface of the photocathode is subjected to uniform illumination. For the 40 mm MCP intensifier, a value of the signal-induced background index ( Z x 100%) of 9% is obtained. Table I shows a comparison of signal induced indices for various image intensifiers.
+
TABLE I Signal-induced background measurements
Tube type.
Signal-induced background index (%)
Reported by
4-Stage cascade (magnetic focus) 3-Stage cascade (magnetic focus) 4-Stagecascade(electrostaticfocus) 40 mm MCP (proximity focus)
41 36 4 9
Cromwell et al. (1985) Delori et al. (1972) Cromwell et al. (1 985) Airey et al. (loc. cit.)
SIGNAL-INDUCED BACKGROUNDIN PROXIMITY-FOCUSED INTENSIFIERS In the case of the 40mm MCP intensifier, some of the signal-induced background is the result of multiple reflections between the input surface of the first microchannel plate and the photocathode. The glass web between the channels of a microchannel plate is usually coated with an evaporated metal film to provide a low-resistance conducting electrode across the face of the channel plate. This metal film has relatively high reflectivity. Figure 6 shows a graph of the spectral reflectivity of a microchannel plate with a nichrome conducting film compared with a similar curve for a plate with bare microchannel plate glass. It is clear that the use of a low-reflectivity or a transparent conducting coating will considerably reduce multiple reflection effects at the input of the intensifier. Observing trials with the IPCS on the 30inch telescope at the Royal Greenwich Observatory indicate that, in the
432
R. W. AIREY ET AL. 50
-
40
-
$30 I->
-
N I C H R O M E COATED MCP
?
IU 20
W
A lL
W
300
4CO
5GO
6t0
WAVELENGTH
(nm1
700
8C3
FIG. 6. Reflectivity measurements for an MCP with nichrome conducting film and for an uncoated MCP surface.
prototype intensifiers, multiple reflections in the photocathode-MCP gap do not seriously distort spectra, although the effect is just discernible. Figure 7 shows atmospheric oxygen absorption lines across the flat-fielded spectrum of the star Spica. It is to be noted that the spectral lines are quite sharp with very little infilling due to localized signal-induced background.
150-
100
400
600
800 Channels
1000
FIG.7. Spectrum of the star Spica, showing atmospheric oxygen absorption lines; obtained with the IPCS using a 40mm MCP intensifier on the 30-inch telescope at the Royal Greenwich Observatory.
EVALUATION OF
PHOTON-EVENT-COUNTING
INTENSIFIERS
43 3
CONCLUSION Measurements carried out on the prototype 40 mm MCP image intensifiers which have been produced as part of a programme to develop an improved front-end detector for photon-event-counting systems indicate that the tubes have several desirable characteristics for use in this application. The pulseheight distribution is well-peaked and compares very favourably with that for a magnetically focused cascade intensifier and a much lower level of signalinduced background is observed. MCP counting efficiency can be improved by modifications to the manufacturing procedures. REFERENCES Crornwell, R. H., Strittrnatter, P. A,, Allen, R. G., Hege, E. K., Kuhr, H and Marien, K.-H. (1985). In “Adv. E.E.P.” Vol. 64A, pp. 77-92 Delori, F. C., Airey, R. W. and McGee, J. D. (1972). In “Adv. E.E.P.” Vol 33A, pp. 99-1 16 Eschard, G., Graf, J. and Polaert, R. (1976). In “Adv. E.E.P.” Vol. 40A, pp. 141-152 Fraser, G. W. (1983). Nucl. Instrum. Methods 206,4455449 Goebel, D . G. (1967). Appl. Opt. 6, 125-128 Jelley, J. V. (1983). Proc. S.P.I.E. 445, 569-576 Schneider, W. E. and Goebel, D. G. (1981). Proc. S.P.I.E. 262,74-83
This Page Intentionally Left Blank
Calculations of the Electron Optics of Image Intensifiers Taking Account of Deviations from Rotational Symmetry
w.MULLER Siemens AG. Medical Division, Erlangen. W. Germany
INTRODUCTION In the middle 1970s, Lange and Schweda (1976) experimentally investigated the deviations from rotational symmetry of the focusing electrodes inside an X-ray image intensifier. Their intention was to study the astigmatism of the image on the output screen caused by perturbations such as ellipticity, tilting and radial shift. Electron-optical imaging devices generally suffer simultaneously from a number of image defects and it is relatively difficult to decide which defect results from which perturbation. To make a unique assignment, Lange and Schweda incorporated electrodes with comparatively large perturbations. Meanwhile, the development of X-ray image intensifiers has made significant progress and the question arises: how will very small deviations (0.2-0.5 mm) from rotational symmetry affect the image? Apart from the time-consuming nature of the experiments, the technical equipment necessary for a detailed investigation would be relatively expensive. We therefore decided to generate a software package that would enable us to simulate our problems on a computer. In this way we are also able to determine the kind of geometrical deformation that is still tolerable to guarantee an image of high quality. The present paper is divided into six parts. In the first two sections we explain the mathematical background of perturbation theory and the discretization formulae used in the solution of the boundary-value problem. Calculation of the electrical field and of the electron trajectories is carried out in the third and fourth sections. The last two sections deal with the grouping of individual rays into beams and discuss the results. PERTURBATION THEORY
By analogy with the proposals of Sturrock (1951), we assume that the electrodes inside an X-ray image intensifier are equipotential surfaces that will 43 5 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction In any form reserved ISBN 0-12-014674-6
w.
436
MULLER
I
tY
I
i
FIG.1. The x-y cross-section of an axially symmetric electrode (circle) and a perturbed electrode (dashed line). The deviation from the ideal configuration is indicated by the vector d(rA)n. A is an arbitrary point on the unperturbed electrode. B is the corresponding point on the perturbed electrode.
undergo certain deformations. The influence of these deformations on the potential field is derived from the knowledge of the unperturbed field U ( r , z ) and the geometrical deviation d(rA)from the axially symmetric electrode. For a better understanding, we now cite some basic ideas of Janse’s (1971) work. The geometrical deviation from axial symmetry (Fig. 1) is expressed by the term
+
d(rA)= MyA) O(A’),
(1)
where L(0 I L I 1) represents a strength parameter. Since the investigated deviations are very small compared with the size of the electrodes, we linearize the perturbed field @(r, 8, z ) . @(r, 8, Z ) = U ( r , Z)
+ 1 V(r,8, z ) + o(L2),
(2)
where U(r, z ) represents the axially symmetric field and the perturbation field LV(r, 8 , z ) accounts for the deformation. As the geometrical deviations will not change the potential on the electrodes, we can write @(.B)
=
UPA).
(3)
In a further step we expand the potential @(r) at r = rA as @(r,) = @@A + d(r,) n) = @(rA) + d(r,J ( n V 1 @(r)lr= rA9
(4)
where n is the direction of the geometrical deviation. Substituting equation (3) into equation (4) we get u(rA) = @(‘A)
+ d(p,)
(n * V 1@(r)lr= r A *
With the help of equation ( 2 ) , equation ( 5 ) is transformed into
(5)
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
U(r,) = U(r,)
+ 1W,) + d(r.4) ( n
*
437
V) ( U ( r ) + A W)>l,=,,,
or, neglecting terms of the order A2, we find
(n * V) u(r),r=r,* (64 Equation (6(a)) is valid for all values of A and represents the boundary condition for the perturbation field V ( r , 8, z), which must satisfy the Laplace V(rA) = -f(rA)
equation:
(6b) The geometrical deviationf(r, 8, z ) and the perturbation field Y(r, 8, z ) are now replaced by a Fourier series:
c (F'"(r, sin(m8) + F2m(r,z ) cos(m8)); V(r, 8, z ) = c P(Vlm(r, sin(m8) + V2m(r,z ) cos(m8)). M
f ( r , 8, z ) =
z)
(7)
m=O
M
z)
(8)
m=O
From the mathematical point of view M is infinite, but for our investigations M I 6 is sufficient. With the help of equations (7) and (8) we are able to reduce our 3-dimensional problem equation (6) to a set of 2-dimensional boundary value problems: (9) with the boundary conditions V"(r,)
= - r A p m Fim(r,, z,) (n
- V ) U(r),,=,,
(10)
and for r -+ 0
v:p=o= 0.
(1 1)
For more details see Janse (197 1). is performed The determination of the normal derivatives (n V ) U(r), with the aid of Newtonian polynomials. Since the gradient of the potential is different inside and outside the perturbed electrode, we have to use different derivatives on both sides of the electrode.
DISCRETIZATION OF THE
2-DIMENSIONAL PARTIAL
DIFFERENTIAL
EQUATIONS In the vicinity of curved elements (i.e. the cathode) the formulae of Janse
w.
438
MULLER
boundary
boundary
T
T
r
r
I- hz-l FIG.2. Mesh point Po in the vicinity of a curved boundary: (a) off-axis situation; (b) on-axis situation. The rectangle (ABCD) represents the integration area G. h, and h, denote the mesh widths of the grid.
fail. Owing to this we generalize his concept. For simplicity we replace Vimby We multiply equation (9) by rZm and obtain equation (12):
D.
+
r Z m +](or,
+ 2m+r ~
1
v,
+ u,,)
= ( r Z m +b,),
+ (rZm+
IU,)~
= 0.
(12)
Applying Stokes theorem, this elliptic partial differential equation is transformed into
j
( --rZm+ 'ur dz
+ rZm+'v, dr) = 0,
(13)
B(G)
where B ( G ) denotes the closed boundary of an arbitrary region G in the r-z plane. According to Janse we divide the contour B ( G )into four parts, along which the normal derivatives v, or v , are assumed to be constant. Figure 2 shows a mesh point Po close to a curved boundary. Pois neighboured by the nodes P I , P,, P, and P,. The dashed rectangle (ABCD) represents the region G. For the evaluation of the integrals (equation (13)) the derivatives v, and u, are replaced by the followingfinite differences: v, = (vo-v4)/s4 h, v, = (uz-vo)/sz h, v, = ( v l -vo)/sI h, v, = (vo-v3)/s3 h,
from A to B, from C to D, from B to C, from D to A,
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
439
where sIh,, s2 h,, s3h, and s4 h, denote the distances from the mesh point Po to the adjacent points Pi ( i = 1, . . ,4) respectively. uirepresents the potential at the node Pi.In this way the integral equation (13) is changed into a linear equation n
1 ci ui = 0. i=O
If Po is 08-axis (n = 4), the coefficients ci are substituted by c , =-
1
2m-i-2
co =
-(c1
((j
+ ~ , / 2 ) ~ " -( j-s4/2)2" +
+ + + c3
c2
h
+ 2)
2
hz s 1 '
c4).
In the on-axis case (n = 3), we obtain 1 2m+2
c, =-
c2
= (s,
+ -, s3)
$2 h r
We can set up a linear equation of type (14) for each node in the grid and finally get a system of linear equations, which is solved by the successive overrelaxation (SOR)method: C v = b (Cis the coefficient matrix, v, b are vectors).
(1 7)
The discretization error is of third order and reduces to fourth order if Pois an interior point (away from any boundary). A further reduction of the discretization error (O(h6))can be achieved with Kasper's formulae (1 984).
w.
440
MULLER
Here, the potential at a mesh point Po is expressed by the potentials at eight neighbouring points (see Kasper, 1984; Miiller, 1988).
CALCULATION OF THE ELECTRIC FIELD The total potential O(r, 8, z ) consists of two parts (equation (2)). Owing to the linearity of @(r, 8, z ) , the gradient V @ ( r ,8, z ) is also linear: V O(r, 8, z ) = V U ( r , z )
+L V
V ( r , 8, z )
Using Taylor’s expansion, we find
+ A , Az + A , Ar + 4 ( A , Az2 + 2A4 AzAr + A , A?) + O(h3), Ur(r,z ) = A , + A , Az + A , Ar, UZ(r,z ) = A , + A , Az + A , Ar, U ( r ,z ) = Uo
with Az
= z-zo,
Ar = r-r,,,
A I = ( VJO = ( UI - U,)/% A2 = (Urh = (u3- U7)/2hr, A3 = (UJO = ( U , - 2 u o U,)/hS, = (u2- u, u,-U&/4hrh,, A4 = (urz)o A , = (Urr)o= (U3-2Uo U,)/h?.
+ +
+
FIG.3. Distribution of the mesh points Pi ( i = O , . . . , 8) for the calculation of the derivatives U,,4, V,, Vo and V, at the point P,. P, is the position of the electron.
44 1
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
See Fig. 3. If the mesh point Po lies on the z axis, U,
=
U , = U7 and
U,,
(22)
U , = U,.
The following considerations are applicable to each component of the Fourier series equation (8). For simplicity, we single out one term of equation ( 8 )and write w(r, 8, z ) = f ( r , z ) cos(m8) with f(r, z ) = rrnVlm(r,z ) . Analogously to equations (19), (20) and (21) we find w(r, 8, z ) = (f o B , Az B, Ar 4 (B, Azz 2B4 Az Ar B, Ar2)
+
+
+
+
+
+ O(h3))cos(mO),
(23)
(24)
with
fo
= r r Vlm(ro~o),
+ +
w,(r, 8, Z ) = (B, B, AZ + B, Ar) cos(m8), wz(r, 8, z ) = ( B , B4 Ar B, Az) cos(m8), wg(r,8, z ) = -m sin(m8) (fo + B , Az + B, Ar (B, Azz + 2B4 Az Ar + B, Ar2)).
+
+4
(25) (26) (27)
Replacing Ui by f , (i = 0, . . . , 8) with f;. = rim Vlm(ri,zi), we change the coefficients Ai into Bi. Adding the expressions w, for all relevant components of the Fourier series, we will get the derivative V,. The derivatives V, and V, are created in the same manner. With the relations cos(m(8 sin(m(8
+ 180)) = s cos(m8) I s = 1 for me2n + 180)) = s sin(rn8) j s = I for m ~ 2 n+ 1 -
n&Z$
(28)
we find for nodes Po being situated on the z-axis (r = 0)
f,
= sf2,
f7
= s f 3 and
f6
= sf,.
(29)
A slightly different method for calculating the expressions w,, w, and w, is proposed by Miiller (1988). CALCULATION OF
ELECTRON TRAJECTORIES
For numerical calculation of electron trajectories we use a 3-dimensional relativistic equation of motion:
m
w.
442
MULLER
where e is the charge of the electron, m = m, [ 1 - (u’/c’)]is the relativistic electron mass, m, is the electron rest mass, v is the velocity of the electron, c is the velocity of light and V O = (Ox,O,,, OJ is the gradient of the potential. The dotted i designates the derivative with the respect to time. In contrast to Wu (1986) we apply Cartesian co-ordinates, which allow us to calculate oblique electron trajectories as well. With x = r cos 8, y = r sin 8,
e = arctan (ylx),
r = (x’
+ y2)1’2,
and a, are substituted by
MATHEMATICAL DESCRIPTION OF ELECTRON BEAMS The individual electron trajectories starting from a point P on the cathode form a beam. According to Schwierz (1 973), each single trajectory is weighted by two uncorrelated factorsf;. and &.A accounts for the emission direction i (i = 1, . . . ,n) of the electron and gk for the possible emission energy k (k = 1, . , . ,rn) (for more details, see Schwierz). The encounter of the (i, k)th electron trajectory with the plane of the output screen (z = z,) is described by the vector where Xi&,Yi,k are Cartesian co-ordinates. Averaging all single trajectories of the beam with respect to the weighting distribution Pi,&=f;. gk, we obtain the weighted centroid P’ of the beam. P’ is called the image point of the cathode point P. The co-ordinates of P are calculated with the help of equation (33): =
c i,k
Wi,k
Pi.k,
= (X, Y).
(33)
With the aid of X’/R and y’/R (R = (X’’ + Y2)IL2),we transform the vectors Wi.k into the x-y co-ordinate system (Fig. 4(a)). The following
443
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
b
x 4
t Y’
Z
Z
Zscreen
.= Zscreen
FIG.4. (a) Location of the x-y co-ordinate system relative to the x’-y’ system. P denotes the centroid of the beam. PI, . . . , P5 represent the penetration points of the single-electron trajectories through the output screen. (b) Elliptic image spot in the plane of the output screen. The sizes of the semi-axes are given by stan W I and stan W2.
considerations refer to the x-y system. We introduce the standard deviation stan Wj ( j= 1, 2) from the weighted centroid W by
In a first approximation the image spot of the beam in the plane of the output screen has the shape of an ellipse. The semi-axes of the ellipse, represented by the standard deviations stan W , and stan W , are called meridional and sagittal semi-axes (Fig. 4(b)).In the vicinity of the output screen we encounter a virtually field-free anode space and therefore substitute the electron trajectories by their tangents to obtain, at time t: Xi,k
( t )= ( Vi , k) x = (Ui,k)y z = (ui,k)r t
Yi,k ( t )
+
Xi,kr
+ Yi,k?
+
Zsr
w.
444
MULLER
or
or, as a vector
where Si,k
= (('i,k)x,
<'i,k)y>/(ui,k>z
and (Ui,k)x, (ui,k)y, ( u i , J z
represent the velocities of the electron (i, k) in the plane of the output screen (z = zs).
Analogously to equations (33) and (34), we derive the expressions S and stan S,, where S is the mean value of the electron velocities in the plane of the output screen. With the help of S a n d wi,Jz), we are now able to calculate the centroid of the beam for each plane z = const:
W(z)= w + (Z-Z,)S.
(36)
Using equation (36) together with the expressions for stan S,, stan W, and the magnitude covariance cov( W ,S),, COV( W,S>j = i,k ((Wi,k)j - wi> ((si,d,- Sj) Pi,k
(37)
7
we find the standard deviations stan W(z),: stan W(z),= (stan'
wj + 2 cov( W,S),(z-zz,)
+ stan'
S,(z-
z,)')'~*.
(38)
With the help of equation (38) we find two planes parallel to the plane of the output screen ( z = 23, in one of which (z = z,,,) the meridional semi-axis has minimum length. The minimum for the sagittal semi-axis is encountered in the plane z = zSag,The corresponding positions of the centroid X,,, = X(z,,,) and Xsag= X(z,,,) are called the meridional focus and the sagittal focus. Connecting first the meridional focuses and then the sagittal focuses of all beams by cubic splines, we create two surfaces, which we call focal or image surfaces (Fig. 5).
DISCUSSION OF RESULTS The first model of an X-ray image intensifier in which we study astigmatism is shown in Fig. 6 . It consists of a parabolic cathode, three focusing electrodes
- meridional
--- sagittal 1.2 1.0
5 .
.6
X
.4
.2 0 25.0
25.2
25.4
zlcrn
25.6
.o
FIG.5. Meridional (solid curve) and sagittal (dashed curve) image surface. The positions of the centroids are illustrated by straight lines.
10
. 2
-2
-4
-6
-0
-10
FIG.6. Electron trajectories and equipotential lines in a meridian section of an X-ray image intensifier (three-electrode system).
w. WLLER
446
and an anode (three-electrodesystem). At first we calculate the image surfaces for the elliptically deformed electrode E3 (cylinder) (Fig. 7). The semi-axes of the ellipse are
R, = R
+ 0.2 mm,
R,
= R-0.2
mm,
where R is the radius of the axially symmetric electrode.
'I
FIG. 7. Elliptic deformation of an electrode. The geometrical deformation is demonstrated by the dashed ellipse. R, and Rb are the semi-axes of the ellipses. R is the radius of the.unperturbed electrode.
x/cm
E3
E3
---
-
E3
7
perturbed unperturbed
rneridional
25.0
25.5
26.0
25.0
25.5
z/cm
25.0
25.5
26.025.0
z/cm
26.0
25.0
25.5
z/cm
25.5
26.025.0
z/cm
26.0
z/cm
25.5
26.0
z/cm
FIG. 8. Perturbed (dashed curves) and unperturbed (solid curves) meridional and sagittal image surfaces for electrons starting in the o", 30" and 45" meridian section (electrode E3 (cylinder) elliptically deformed).
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
447
The resulting perturbation field V ( r , 8, z ) contains three terms, V ( r ,8, Z ) = Vl0(r,Z)
+ r2VI2(r,Z) cos 28 + r4VI4(r, z ) cos 48,
(39)
where the second term (m= 2) wiil have the greatest influence on the electron trajectories. Terms with m 2 5 are negligible. Figure 8 shows the resulting meridional image surfaces (upper row) and the corresponding sagittal image surfaces (lower row) for electrons starting in a 0", 30" and 45" meridian section. The solid curves indicate the unperturbed and the dashed curves the perturbed image surfaces. We first turn to the meridional image surfaces (upper row). We notice an image shift away from the cathode of about 0.7 mm for the 0"meridian section (left figure). The shift is less for the 30" section and disappears for electrons starting in the 45" meridian section. For the sagittal image surfaces (lower row), we encounter an analogous image shift, now, however, towards the cathode, which is again smaller for the 30" section and zero in the 45" section. For electrons starting in the 60"and 90" meridian section (Fig. 9), the effect is the opposite. The meridional image surfaces are now shifted towards the cathode, while the corresponding sagittal image surfaces are shifted away
E3
x/cm
___
E3
-
perturbed unperturbed
1.0
meridional
.5
90
~
'
0. 25.0
25.5
26.0
25.0
25.5
z/cm
26.0 z/cm
x/cm L
L
sagittal
25.0
25.5
26.0
z/cm
25.0
25.5
26.0
z/cm
FIG.9. Perturbed (dashed curves) and unperturbed (solid curves) meridional and sagittal image surfaces for the 60" and 90" meridian section (electrode E3 (cylinder) elliptically deformed).
w.
448
MULLER
TABLE I Differences between perturbed and unperturbed image surfaces for the meridional (Axmer, Azmer)and sagittal (Axsag, Azsag)directions. (E3 (cylinder) elliptically deformed.)
0.1
1.0
2.0
4.0
Az,,, (mm) -0.59 AzSag(mm) 0.77 Axmer(mm) 0.00 0.00 Axsag (mm)
-0.58
-0.58 0.60
-0.59
0.60
-0.60 0.60
-0.01 0.00
-0.03 0.00
-0.04
-0.05 0.00
Azmer(mm) 0.72 AzSag(mm) -0.61 0.00 Axmer(mm) 0.00 Axsag(mm)
0.61 -0.58
0.60 -0.58 0.03 0.00
3.0
5.0
7.0
8.0
-0.58 0.59 -0.07 0.02
-0.59
-0.61
-0.09 0.04
0.58 -0.11 0.07
0.56 -0.58 0.07
0.56 -0.57
0.59 -0.59
0.09
-0.02
-0.04
0.10 -0.07
6.0
90"-emission-plane
0.00
0.58
-0.59 0.58 -0.06
0.01
0.61
0"-emission-plane
0.01 0.00
0.58
-0.58 0.04 0.00
0.55 -0.58 0.05 -0.01
0.59 -0.59 0.06 -0.01
Azmer= zmer(perturbed) - zmer unperturbed; (analogous expressions stand for (Azsag, Axmerand (Axsad.
from the cathode (see Table I). This behaviour results from the leading ~ of the perturbation field, where cos(20) is responcomponent r2 V Icos(20) sible for astigmatism changing polarity when we pass beyond the 45" meridian section. The extent of this component in the 0" meridian section is plotted in Fig. 10, where the hatched region is responsible for the image shift, since electrons passing through this area will experience a similar perturbation. In the foregoing, elliptic deformation was applied to the electrode E3 (cylinder) only. If we deform the electrode E3 (cylinder) and electrode E2 , a
4
2
0
2
4
6
8
10
12
14
16
I8
20
22
24
26
z/cm
FIG. 10. Extent of the perturbation component S = I? Vlz(r, z) cos(28)1 in the 0" meridian section. The hatched area is responsible for the image shift. -, 0.1 V < S < O S V ; 0, 0.5 V $ 6 < 1 V ; A, 1 V Q S < 10 V ; A ,10 V $ 6 < 50 V; 0 , 5 0 V < 6 < 100 V; 0 , a> 100 V .
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
449
1 FIG. 11. Two elliptically deformed electrodes (E2 and E3 (cylinder)). a denotes the angle between the principal axes of the two ellipses.
simultaneously, the resulting image surfaces depend on the angle M between the corresponding principal axes of the two ellipses (Fig. 11). Figure 12 shows the image's surfaces for parallel ellipses (M = 0). In the left column we find the image surfaces for the elliptically deformed electrode E2. The central column represents the situation for the electrode E3 (cylinder) and the right column x/cm
E2
E2II E 3
E3
---
perturbed unDerturbed
rneridional
25.0
25.5
26.0
25.0
25.5
26.0
25.0
25.5
z/cm
z/cm
26.0 z/cm
1.0
sagittal
.5
0. 25.0
25.5
26.0
z/cm
25.0
25.5
26.0
z/cm
250
25.5
260
z/cm
FIG. 12. Perturbed and unperturbed meridional and sagittal image surfaces for electrons starting in the 0" meridian section: (a)electrode E2 elliptically deformed; (6) electrode E3 (cylinder) elliptically deformed; (c) E2 and E3 (cylinder) elliptically deformed. (The angle a between the principal axes is O".)
w.
450
MULLER
TABLE I1 Differences between perturbed and unperturbed image surfaces for the meridional (AxmerrAzme,)and sagittal (Axsagr AzSag)directions. (E2 parallel to E3 (cylinder).)
0. I
1 .o
2.0
3.0
4.0
5.0
6.0
7.0
8.0
0"-emission plane Azmer(mm) 1.04 AzSag (mm) -1.04 Ax,,(mm)O.OO Axslln(mm) 0.00
1.03 -0.91 0.01 -0.01
1.02
-0.98 0.03 -0.02
0.96 -0.90 0.04 -0.03
1.00
1.27
-0.96 0.06 -0.04
-1.00 0.08 -0.06
0.88 -0.87 0.08 -0.07
0.97 -0.99 0.08 -0.13
1.25
-0.92 0.15
-0.17
shows us the image surfaces, where both electrode E2 and E3 (cylinder) are deformed. In the present investigation all electrons started in the 0" meridian section. Consider first the meridional image surfaces of the electrode E2. We again encounter an image shift away from the cathode, but now we notice considerable variation of the perturbed, dashed image surface in its upper half. This variation results from the fact that electrons starting from different positions on the electrode will pass through different perturbation areas. The image surfaces for a combined deformation of the electrode E2 and E3 (cylinder) shown in the right-hand figure may be compared with the first two figures in the same row; we find a reinforcement effect. For the sagittal image surfaces a corresponding reinforcement effect is detected, now, however, towards the cathode (see also Table 11). If the angle tl between the principal axes is 90" we get a cancellation effect (Miiller, 1988). A further perturbation that occurs in the electron optics of imaging devices is the radial shifr of an electrode (Fig. 13). To study its influence on the image
FIG.13. Radial shift of an electrode. The dashed circle denotes the position of the shifted electrode.
45 1
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
surfaces, we shifted the electrode E3 (cylinder) by d=0.2 mm. The relevant components of the perturbation field V ( r , z ) are V ( r , 8, Z )
Vl,,(r, Z )
=
+ r V,,(r,
Z)
cos 0
+ r2 V12(r,
Z)
cos 28,
(40)
where r V l l ( r z, ) cos 8 is dominant. Terms with m 2 3 are negligible. Figure 14 shows the image surfaces for electrons starting in the 180" meridian section (upper image halves) and for those starting in the 0" section. In addition we have incorporated the straight lines of the centroid to illustrate the radial shift of the image surfaces. Considering the two figures, we observe an image shift, which coincides with the direction of the electrode shift and results from the component r Vll(r, z ) cos 8 (see Table 111). Furthermore, we detect a small variation of the image surfaces especially in the off-axis region, which is due to the fact that electrons starting from an outer area of the
sagittal
meridional
x/cm
180
-.2
-
-.4
- - - - - - - _ _ _ __ _ _ _ _
-.6
-_--------------_- __ __ __ __ __ __ __
-.a -1.0
-
-1.2
-
____
-----.._____
0
I
I - - _ - - - 2 ~
----____
,
----_ _ _ ----____ ,
.------]I
,
25.0
---I i I i I I 1 , I C
25.2
25.4
25.6
25.8
z/cm
---
250
25.2
25.4
25.6
158
z/cm
unperturbed perturbed
FIG.14. Perturbed and unperturbed meridional and sagittal image surfacescombined for the 0" and 180" meridian section. Additionally the straight lines of the centroids are incorporated.
w.
452
MULLER
TABLE I11 Differences between perturbed and unperturbed image surfaces for the meridional (Ax,,, Azmer)and sagittal (Axsagr Azsag)directions. (E3 (cylinder) radial shifted.) remission (cm) I .o
0. I
2.0
3.0
4.0
5.0
6.0
7.0
8.0
-0.01 0.12 0.13 0.15
-0.03 0.00 0.13 0.13
0.06 0.17 0.15 0.16
-0.12 0.01 -0.13 -0.12
0.10 -0.02 -0.12 -0.14
0.32 -0.15 -0.08 -0.15
1 8O0-emission-plane
Az,,, (mm) -0.03 Azmg(mm) 0.18 Axme,(mm)0.I3 AxSag(mm) 0.13
0.05 0.00 0.13 0.13
-0.06 0.01 0.13 0.13
0.01
-0.01 0.13 0.13
0.02 -0.02 0.13 0.13
-0.01
0.01 0.13 0.13
0"-emission-plane
-0.01
Az,,(mm) -0.11 Azsag(mm) -0.01 Ax,,, (mm) -0.13 Axsag(mm)-0.13
0.08 0.08 -0.13 -0.13
-0.05 -0.01 -0.13 -0.13
0.02 -0.13 -0.13
-0.08
-0.02 0.02 -0.13 -0.13
0.00 -0.14 -0.13
cathode experience a perturbation force that is slightly different from that experienced by electrons emitted closer to the axis. In the case of a tilted electrode, we detect a radial shift as well as comparatively strong astigmatism in the off-axis region of the image surfaces (Muller, 1988). Finally, we considered the influence of different deformed sections of an electrode on the image surfaces. To study this we elliptically deformed the rlcm
cathode
:B
screen
2
4-1
\
8 610-
y+ n
'
I
I
I
I
!
i
I
f i
j r j I
i
i
0
I
anode
i
j r '
! C / I
I
0
I
I
I
,I
zicm
FIG.15.0"meridian section of a two-electrode system.The deformed forward,central and rear sections of the electrodes El and E2 are indicated by the charactersf, c and r .
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
453
forward, central and rear section of the electrodes El and E2 of a twoelectrode image intensifier (Fig. 15). The structure of the resulting perturbation fields is the same as derived in equation (39). The semi-axes of the ellipse now differ from the radius of the cylinder by d = f0.5 mm. The results are given in Fig. 16, where we find the image surfaces for a deformed rear section of electrode E 1. We encounter relatively strong astigmatism, where the meridional image surface is shifted away from the cathode (upper figure) and the sagittal image surface towards the cathode (lower figure). For the case of a deformed forward section of electrode E2 (see the right column) the astigmatism has changed its polarity. That means that the meridional image surface now reveals a slight shift towards the cathode, while thecorresponding sagittal image surface is shifted away from the cathode. To understand the change in polarity, we plotted the dominant component r2 V12 cos(28) of the perturbation fields for the 0" meridian section. Figure 17 represents the component for the rear section of the electrode E l. We notice a cloud-like spread of the perturbation component in the vicinity of the deformed rear section, which is of positive character as demonstrated by the perturbation sign pattern in the lower figure. The situation changes when we study the
x/cm
El-rear
_ _ _ perturbed - unperturbed
E2-forward
1.0
meridional
.5
0. 27.5
28.0
28.5
27.5
28.0
28.5
z/cm
z/cm
x/cm
sagittal
27.5
28.0
28 5
z/cm
27.5
28 0
28.5
z/cm
FIG.16. Perturbed and unperturbed meridional and sagittal image surfaces for the 0"meridian section: (a)ellipticallydeformed rear section of the electrode El; (h)elliptically deformed forward section of the electrode E2.
454
w.
MULLER
FIG. 17. (a) Extent of the perturbation component 6 = ( r 2 V12cos(20)(for the elliptically deformed rear section of the electrode El in the 0" meridian section: -, 0.1 V < 6 <0.5 V; 0 , 0.5 V < 6 < 1 V; A , 1 V < 6 < 1OV; A , 1 0 V < 6 < 5 0 V; 0,5 0 V < 6 < 1OOV; 0 , 6 > 100 V . (b) The corresponding sign pattern. -, component < -0.1 V ; +, component20.1 V.
perturbation component produced by the deformed forward section of the electrode E2 (Fig. 18). The perturbation again spreads in a cloud-like manner, but now its sign gradually changes from plus to minus. This is the reason why the resulting astigmatism changes its polarity as well. In the case of a deformed central section of the electrode E2, astigmatism once more changes its polarity. The remaining sections are of secondary priority. The trend mentioned above fully agress with experimental data gained by Lange and Schweda (1976).
CONCLUSION Among the various geometrical perturbations that were studied, elliptic deformation is dominant. Depending on the voltage of the deformed
iv
455
THE ELECTRON OPTICS OF IMAGE INTENSIFIERS
cathode
........................... .......................... ......................... ....................... ...................... .................. ............... ....... z/cm
.....................
El
++++++++++++++ t i +++ti+ +++i++++++++++t++f+++++++ tt'
.. .. .. .. .. .. .. .. .. .. +++++++++++ .. .. .. .. .. .. .. .. .. ..
E2
+i+*+i++++*++++++++++
.................................
+t::::::::::::::Y+~++~~~ t+++++t++++++++ ...........
+++,++++++++++- . . .. .. ....... .. .. ....... .. . . ++,++++,+,++ ,++++++++++ . ..... .. .. .. .. .. .. .. .. .. ...... ................ . . . . . . . . . . . . . ++++++tt+++++
FIG.18. (a) Extent of the perturbation component 6=lr2 V12 ( r , Z)cos(20)1 for the elliptically deformed rear section of the electrode E2 in the 0' meridian section: *, 0.1 V < 6 10.5 V; 0, 0.5 V C 6 < I V; A, 1 V < 6 < I O V ; A,l O V < 6 < 5 O V ; O , 5 O V G 6 < IOOV;O,6> I00 K(6)The corresponding sign pattern. -, component < -0.1 V; + ,component > 0.1 V.
electrodes 0.2-0.5 mm of ellipticity are sufficient to shift the image surfaces up to 0.7 mm towards or away from the cathode. The separation almost doubles if there exists another electrode with a deformation that reinforces the first. In contrast, the separation is reduced if the perturbations are orthogonal to one another. The second type of perturbation we must pay attention to is tilting. We encounter strong astigmatism, especially in the off-axis domain of the image. In the case of partly deformed electrodes, the shift of the image surfaces depends on the position and the size of the deformation. Radial shift of an electrode causes less severe perturbation. The data reveal that in this case the image suffers from only slight radial shift and negligible astigmatism. Our investigations revealed that elliptic deformations are the main cause of severe astigmatism. Whether the deformation is tolerable finally depends on the voltage of the electrode, the size and the position of the electrode, and of
456
w.
MULLER
course, on the size and the position of the deformation. Detailed studies are therefore necessary in each individual case. So far we have merely investigated the astigmatic property of an X-ray image intensifier under certain geometrical perturbations. In reality, however, image intensifiers suffer from a combination of several geometrical defects. Evaluation of such combined defects will provide further interesting study.
REFERENCES Janse, J. (1971). Optik 33,270-281 Kasper, E. (1984). Optik 68, 341-342 Lange, F. W. and Schweda, S. (1976). In “Adv. E.E.P.” Vol. 40A, pp. 507-517 Miiller, W. (1988). Thesis (in preparation) Schwierz, G., (1973). Siemens Forsch. Entwicklungsber. 2,305-3 15 Sturrock, P. A. (1951). Philos. Trans. R. Soc. London A243,387-429 Tong, Lin-su and Wu, Chang-jin (1986). NTG Fachber.95
Optimization of the Imaging Properties of an Image Inverter Tube M. F. CALITZ, A. G. DU TOIT and C. F. VAN HUYSSTEEN National Physical Research Laboratory. CSIR, Pretoria, South Africa
INTRODUCTION The electron optics of an existing image intensifier tube design was investigated by means of computer simulation with a view to achieving optimal focusing. The main purpose of this investigation was to assess the sensitivity of the imaging properties to small variations in the dimensions of the electrode structures. Maximum error tolerances for the structural parameters can then be determined. A method for correcting for small errors in the electrode dimensions by adjusting the position of the anode is proposed. A semi-analytic model is proposed to analyse the paraxial focusing properties and their dependence on certain critical tube parameters.
SEMI-ANALYTIC
MODELOF
THE
ELECTRON OPTICS
The electrode structure of the image inverter tube investigated is shown in Fig. 1. The lens action of this tube was initially modelled by regarding it as a combination of a two-concentric-spheres lens with a pierced anode and a double-cylinder lens, indicated by regions I and 2 respectively. The length of the tube and the final magnification are given by fP L = -&fs M = 2 (~--f)(n- 1 ) R,’ (s-f) ’
where with the axial position of the cathode as the origin s =I-(p+R,-R,),
p = 2R,
(n - 1)[27c(n - 1) - kn]
2n(n - l)(n - 2 ) - nn(n - 1 ) -kn(n - 2) ’ 457
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
458
M. F. CALITZ, A. G. DU TOIT AND C. F. VAN HUYSSTEEN
I is the position of second lens, l=R2/(O.986{(--) =+p 0- 1
2p2 1
[lno+(pz-l)
In
R, is the radius of outer sphere; R, is the radius of inner sphere; n = RJR,; k = d/R,; d is the diameter of aperture in anode; IS= V3/V,; p = R3/Rt. Further details regarding these formulae can be found elsewhere (Schagen et al. 1952; Paszkowski, 1968). As a first stage in designing the tube with fixed voltages on the electrodes, the approximate values of tube length for a given magnification or the values of the magnification for a given tube length can be estimated using the above formulae. The choice of R, seems to be critical in the results obtained and can be determined by back-calculating from a given magnification or tube length. The relative sensitivity of the position of the image plane on-axis to variations in certain structural parameters can then be assessed directly from these formulae. Defining sensitivity as dzildai where z i is the position of the image plane along the axis and ai is the structural parameter, Table I shows the sensitivities for small changes in some structural parameters as determined from the above formulae. The effects of varying the voltages on the second and third electrodes can also be estimated from the formulae. A decrease in the voltage ratio results in a decrease in the tube length and magnification.
IMAGING PROPERTIES OF AN IMAGE INVERTER TUBE
459
TABLE I
Sensitivity of image plane position to structural parameters Parameter
Sensitivity
Anode position Anode aperture radius Anode cylinder radius Screen cylinder
-0.65
14.15 0.35 0.2
NUMERICAL METHOD A rudimentary CAD system was developed on a supermicrocomputer using an array-processor, specifically to investigate the electron optics of image inverter tubes (Calitz and du Toit, 1987). The calculation of the electric field was done using the boundary element method (Renau et al., 1982). The trajectories were calculated by solving the equations of motion in the tangential plane using a predictor-corrector method (Kasper, 1985). Sufficient trajectories are then run to determine the polychromatic line-spread function at the image plane, assuming a Lambertian angle distribution and a Maxwellian energy distribution (Csorba, 1972). The edge-response function was then determined so as to integrate any asymmetry in the line-spread function especially off-axis. The width of the edge response function is then related to the modulation transfer function so that a measure of electron optical resolution can be obtained (Johnson, 1973). Magnification is taken as the position of the maximum intensity point of the line-spread function divided by the initial position of the object. Distortion is then defined as the variation in the magnification over the image plane. The sensitivity analysis was initially conducted by varying the value of a particular geometrical parameter and determining the consequent variations in the width of the edge-response function, magnification and distortion. As the calculations required a large amount of computing time, a shorter means of analysing sensitivity was tried which proved adequate. The position of the focal plane was determined as the cross-over point of the two trajectories originating at an angle of 45" to the normal and with a most probable energy of 0.1 eV. The magnification was determined b y the position on the image plane of the principal ray. The variations in the position of the focal plane with changes in any particular geometrical parameter were then taken as a measure of the sensitivity. Some results are shown in Table 11. Figure 2 illustrates focal plane variations with changes in the position of the
460
M. F. CALITZ, A. G . DU TOIT AND C. F. VAN HUYSSTEEN
TABLE I1 Effect of changes in geometrical parameters on focal plane
0
1
Parameter
Sensitivity
Anode position Anode aperture radius Anode cylinder radius Screen cylinder radius
-6 3.5 2.0 0.5
2
3 4 5 6 Position of initial electron ( mm )
7
8
FIG.2. Focal plane versus anode position.
anode aperture. Small variations in the other tube parameters also produce similar shifts in the focal plane of differing magnitudes. Only variations in the cathode radius of curvature appear to cause a change in the shape of the focal plane. Using the above methods, none of the aberrations are specifically isolated except for distortion. The modulation transfer function implicitly incorporates the other geometric and chromatic aberrations. A method of relating the focal plane position to the final resolution seems impossible, but the proximity
46 1
IMAGING PROPERTIES OF AN IMAGE INVERTER TUBE
of the focal plane to the image plane provided a suitable qualitative means of finding optimum parameter values.
OPTIMIZATION METHOD According to the numerical calculations, the position of the anode proved to be the most sensitive parameter. In addition, the anode is one of the last components to be installed in the final assembly. Consequently, a method of correcting for any construction errors in the rest of the geometric parameters was suggested, whereby the position of the anode is adjusted to compensate for these errors. The adjusted position of the anode is then determined by adjusted position =optimum position +&ui where Eiisthe deviation of parameter from optimum value; aiis the adjustment factor for particular parameters = Sd;/Sa;Sdiis the sensitivity for parameter di; S, is the anode sensitivity. Theoretically, an optimum geometry for the tube was determined. Certain errors were created in this geometry, after which the width of the edge response function and the magnification were calculated. The anode position was adjusted according to the above formula and the performance parameters were recalculated. The results in Table 111, for initial radial object positions of 2 and 8 mm, clearly show an improvement in resolution, but a slight worsening in magnification and distortion, although not serious enough to make the tube unusable. The sensitivity analysis and optimization method employed assume linear independence between the various geometric parameters, which is probably a valid assumption for small variations in the geometric parameters. For large variations, the interdependence of the geometrical parameters needs to be determined, but this requires considerably more computation. TABLE I11 Effect of anode position on edge-response and magnification Width of edge-response function Anode position 22.8 23.23
r =2
rnrn
20 9
Magnification
r=8rnm
r = 2 rnrn
r = 8 rnrn
17 7
1.4755
I .475
1.448
1.452
462
M. F. CALITZ, A. G. D U TOIT AND C. F. VAN HUYSSTEEN
COMPARISON OF MODELWITH NUMERICAL RESULTS Quite clearly, the analytic model proposed originally does not result in the same sensitivities as obtained numerically. The model is also only confined to the paraxial region. If one uses the model in a self-consistent way, a qualitative idea is obtainable of the effects ofchanges in certain critical dimensions on the position of the focal plane and the magnification. Figures 3 and 4 show equipotential plots of the actual tube and the concentric-spheres plus doublecylinder model. The equipotential map shows that the contours around the double-cylinder lens are very similar to the contours in the actual tube. The spacing of the contours in the concentric-sphere region does not match up with those in the actual tube, although they approach each other near the anode aperture. Preliminary basic dimensions of a similar tube with a required magnification or tube length can be obtained using the formulae in the model. Numerical ray-tracing can then be undertaken to make the necessary adjustments to these dimensions. For example, if you wished to make a tube with a similar structure that gave lower magnification, the formulae indicate that the voltage ratio between the second and third electrodes would have to be reduced and the tube shortened.
1
1I I I
I
FIG.3. Equipotentials in experimental tube.
FIG.4. Equipotentials in theoretical model.
IMAGING PROPERTIES OF AN IMAGE INVERTER TUBE
463
CONCLUSION It appears that a reasonably simple and accurate analytic model of the image inverter tube is not really possible, as was categorically stated by Schagen and Woodhead (1967). The model proposed does give some insight into the electron optics of the tube and can be used in a qualitative sense to assess the effects of certain critical dimensions. It can also be used to investigate designs at the preliminary stage. The numerical method has proved successful in sensitivity analysis, where relative accuracy is quite adequate. For the determination of real geometrical values for the tube, the results of the numerical calculations need to be very accurate in an absolute sense. The numerical model itself may not prove sufficiently accurate in this respect, but this aspect is under further investigation.
REFERENCES Calitz, M. F. and du Toit, A. G. (1987). Compumug (in press) Csorba, 1. P. (1972). RCA Rev. 33,393-398 Johnson, C . B. (1973). Appl. Opt. 12, 1031-1033 Kasper, E. (1985). Oprik 59, 117-125 Paszkowski, B. (1968). In “Electron Optics”, p. 181. Iliffe Books Ltd Renau, A,, Read, F. H. and Brunt, J . N. H. (1982). J . Phys. E. 15,347-354 Schagen, P., Bruining, H. and Francken, J. C. (1952). Philips Res. Rep. 7 , 119-130 Schagen, P. and Woodhead, A. W. (1967). Philips Tech. Rev. 28, 161-182
This Page Intentionally Left Blank
An XUV Image Sensor for Rowland-circle Spectrographs J. L. LOWRANCE and C. L. JOSEPH Princeton UniverrsifyObservatory, Princeton. New Jersey, USA
INTRODUCTION In space astronomy there is considerable interest in the spectral region shortward of 1200 A. From the standpoint of efficiency, measurements in this spectral region require that the detectors be windowless and that the number of optical reflections be minimized because of the relatively low transmissivity and reflectivity of optical components. The spectrograph of choice for many of these applications is the Rowland-circle type in which the slit, grating and focal plane all lie on a circle. The grating, ruled on a concave aspheric surface, provides the focusing as well as the dispersion with a single reflection. Optical aberrations normally confine the focal plane to a narrow band along the circle. This paper reports on the preliminary design of a windowless oblique magnetic-focus image sensor similar to existing sensors (Picat et a/., 1972; Johnson and Hallam, 1976; Lowrance et al., 1979; Lowrance, 1985; see also Carruthers et al., 7) but configured to match the long narrow image format of a Rowland-type spectrograph. The principal advantage of this image sensor is its use of an opaque photocathode on a smooth surface to obtain the highest quantum efficiency available. A schematic of the detector and spectrograph is shown in Fig. 1. The focusing magnetic field is generated by a long “C”shaped permanent magnet assembly. Since the magnetic field in the gap of such configurations is intrinsically less uniform than fields inside cylindrically shaped magnets normally employed to focus image tubes (Coleman, I979), the magnetic focusing characteristics have been simulated for a suitable distribution of photoelectron energies in order to demonstrate the feasibility of this design. Although simulations of the electromagnetic focusing for a variety of sensor parameters have been made, none of these parameters has yet been optimized to achieve the best possible image quality. The model uses software developed several years ago to design a 2-dimensional windowless intensified charge-coupled device (WICCD) for the Interstellar Medium Absorption
t This volume, p. 18 I . 46 5 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
466
J. L. LOWRANCE A N D C. L. JOSEPH
___.-FIG.1. Rowland-circle spectrograph configuration showing windowless image intensifier with a long narrow format matching the spectrograph image.
Profile Spectrograph (IMAPS) sounding rocket programme. The results of those computer simulations turned out to be a very accurate representation of the operating characteristics of the now flight-proven IMAPS WICCD. In order to have a set of well-defined goals, this paper focuses on the preliminary design of a WICCD for a particular envisaged Rowland-circle spectrograph. IMAGE FORMAT
Studies in support of the next generation of ultraviolet astronomy satellite (the proposed Lyman Mission) have pointed out the high optical efficiency in the XUV of a Wolter-Schwarzschild Type I1 grazing-incidence telescope in combination with a Rowland-circle type spectrograph. It has been proposed that the 910-1250 A spectral range be divided into three 120 A bands to allow the grating efficiency to be optimized for each band. At a spectral resolution of AjAA = 3 x lo4,each band would contain 3600 spectral images of the slit and thus, would require at least 7200 pixels in the spectral dispersion direction to sample properly each 120 A wide spectrum without aliasing.
AN
xuv
IMAGE SENSOR FOR ROWLAND-CIRCLE SPECTROGRAPHS
u)
c
C
3
s
467
I 20
190.1 0
Position (mm)
Position (mm)
FIG.2. (a) Optical ray tracing of 1.2 m diameter Rowland-circle spectrograph with an aspheric grating, A/AA = 30 000.Shown are two wavelengths separated by one part in 15 000; (b)histogram showing the spectrum of Fig. 2(a) convolved with an ideal detector having 20 p n wide pixels.
An 80 cm f/lO telescope has an image scale of 39 pm per arc-sec. The telescope resolution is nominally one arc-sec, dictating a spectrograph slit width of approximately 40 pm and a Rowland circle diameter of 2 m for AjAA = 3 x lo4. The image sensor should have a line-spread profile that is small compared to the 40 pm slit and fit an arc that is 14.4 cm long on a circle 2 m in diameter. The magnetic lens simulation presented below, having a full width at half-maximum (FWHM) along the dispersion of 19 pm compares favourably with this requirement. Ray-tracing of the envisaged spectrograph for the Lyman Mission shows that the spatial resolution along the slit is limited by optical aberrations, as shown in Fig. 2(a) (Moos, 1986; Cash, 1984), while Fig. 2(b) shows this image convolved with an ideal detector, having 20 pm wide pixels and 100% modulation transfer function (MTF).
-
FOCUS REQUIREMENTS MAGNETIC The magnetic focal length between the photocathode and target for a uniform field is expressed by the following equation (Johnson and Hallam, 1976; Beurle and Wreathall, 1962):
where V is the total voltage drop experienced by the photoelectrons and B is the strength of the magnetic field. While a magnetic field does not have to be uniform to provide focusing, it is
468
J. L. LOWRANCE AND C. L. JOSEPH
important that all the photoelectrons experience essentially the same magnetic field as they travel from the photocathode to the target. The field between two magnetic pole faces is not uniform. In this case, however, there is uniformity in the long, dispersion direction, except near the ends. Also, the telescope’s 60 arc-sec usable field of view, limiting the image at the photocathode to only 2.4 mm perpendicular to the plane of dispersion, somewhat eases the off-axis magnetic focus requirements. Increasing the magnetic field generally reduces the point-spread function of the imaged photoelectrons. On the other hand, for a fixed electric field strength, equation (1) indicates that increasing the strength of the magnetic field reduces the focal length, and correspondingly the voltage drop experienced by the photoelectrons. Each photoelectron must be accelerated sufficiently to produce a measurable gain of signal electrons. Hence the problem becomes one of balancing an improved point-spread function of the photoelectrons caused by larger B fields against the corresponding decreased signal strength. The electric field is limited by high-voltage breakdown considerations. For the Rowland-circle spectrograph described above, the magnetic field must fill a volume approximately 3 x 45 x 150 mm3in extent. The electric field was taken to be 405 kV m-’ a t an angle 30” to the magnetic field (see Fig. 3). Table I shows the on-axis variation of the magnetic field along Z , the distance from the centre of the magnet gap to the photocathode or target, for the dimensions shown in Fig. 3.
/d>EB Incident light
FIG. 3. Sketch of cross-section of the “C”-shaped permanent magnetic assembly showing location of photocathode and focal plane of oblique-focus image intensifier electron optics. The distance from the nearest pole face to the centre of the photocathode is 33.5 mm. The distance from the nearest pole face to the centre of the target is 24 mm.
AN
xuv
IMAGE SENSOR FOR
ROWLAND-CIRCLE SPECTROGRAPHS
469
TABLE I Magnetic field across the air gap of the magnetic assembly shown in Fig. 3
-35.0 - 30.0 - 25.0 -20.0 - 15.0 - 13.5
- 10.0 -5.0 0.0 5.0 10.0 15.0 19.6
20.0 25.0 30.0 35.0
574 496 44 1 403 377 371 (at photocathode) 360 35 I 348 351 360 377 40 1 (at focus) 403 441 496 574
THECCD The CCD employed in the IMAPS intensified CCD detector is the RCA 501, which is back-illuminated and has 256 x 320 pixels of 30 x 30 pm2. This Rowland-circle version of an electron-bombarded WICCD detector would require chips from other vendors, since these CCDs are no longer manufactured. The actual CCD configuration could be made up of a long narrow row of CCD chips closely spaced such that there would be small gaps in the format that could be filled when necessary by a second exposure. Alternatively, these small gaps might be eliminated by custom-designing the CCD to be butted at the ends or built as one long narrow CCD comparable in length to the diagonal of the Tektronix 2048 x 2048 pixel CCD (80 mm). In any event, each 40 pm slit image (resolution element) should be sampled at least twice in the image read-out process to reduce abiasing artefacts in the data.
MAGNETIC Focus POINT-SPREAD PROFILES The reader is referred to the proceedings of the Eighth Symposium on
470
J. L. LOWRANCE A N D C . L. JOSEPH
FIG.4. Photoelectron energy distribution of opaque Csl photocathode excited by 11.8 eV and 1487 eV photons.
Photo-Electronic Image Devices for a detailed description of the computer modelling and electromagnetic ray-tracing programme used in this design study to evaluate the focusing properties of a field between two pole faces of a permanent magnet (Lowrance, 1985). Figure 4 shows the photoelectron energy distribution of an opaque CsI photocathode, excited by 1 1.8 eV photons (Jenkins, private communication). This energy distribution was simulated by choosing five discrete values of photoelectron energy and assigning each a weight in calculating the focused point-spread profile (Lowrance, 1985). For this study, the photocathode-to-target distance L was taken to be -33 mm and the accelerating voltage 12 kV at an angle 4 of 30" to the magnetic field. The magnetic field used in calculating electron trajectories is summarized in Table I. The photocathode is located 13.5 mm from the centre of the gap and the focus lies 19.6 mm from the centre on the other side. These are not necessarily optimum positions, merely the best of two or three trials. Figures 5 and 6 show the resultant line-spread profiles in the dispersion direction and along the slit, respectively. The FWHM is approximately 19 pm wide (and 64 pm along the slit). The point-spread function is astigmatic, but considerably less than the astigmatism of the spectrograph's optical image, shown in Fig. 2. This is understandable, since the magnetic field is uniform in the direction normal to the slit and non-uniform along the slit. It should be noted that the magnetic field uniformity can probably be improved somewhat by shaping the magnet pole faces, by increasing the distance between the pole faces, or by increasing the height of the pole faces. This would undoubtedly reduce the astigmatism along the slit. The position of
AN XUV IMAGE SENSOR FOR ROWLAND-CIRCLE SPECTROGRAPHS
47 1
x
.B
-5900
-5850
-5800
Y ( microns) FIG.5. Line-spread profile in dispersion direction.
x
.B
FIG.6. Line-spread profile along the spectrograph slit.
the photocathode and focal plane in the gap is not necessarily optimum either. However, given the spectrograph optical aberrations, such optimization appears to be unnecessary for the application. Figure 7 shows the point-spread profile and the point-to-point mapping of the image in 0.5 mm steps, off-axis. All of the point-spread profiles appear to be essentially the same as the on-axis one over a range of k 2.5 mm, which is
472
J. L. LOWRANCE AND C. L. JOSEPH
0
2000 -
*
E
0
FWHM spr' size compareo. t o a 40 micron slit
c
0
+
5
a
1000
-
-
0
0 0 -
-1000
0
0 1000 2000 Dispersion Direction (microns)
FIG.7. Variations in the point-spread function and the mapping of points in 0.5 mm steps offaxis. The full width at half-maximum for a typical spot is compared to the projected 40 pm slit width for the Lyman Mission.
more than adequate to accommodate the limited off-axis optical performance of the Rowland-circle spectrograph and telescope described above. Figure 7 also compares the point-spread function to the slit width as seen at the detector. The depth of focus of the electromagnet (Fig. 8) is deep, providing a margin for variation in the magnetic and electric fields and, for example,
60
1 14
RldS orthognal to dispersion
.
18
18
20
Position (rnm 1
FIG.8. Electromagnetic depth of focus for the photoelectrons.
AN XUV IMAGE SENSOR FOR ROWLAND-CIRCLE SPECTROGRAPHS
473
making the detector insensitive to the orientation of the Earth’s magnetic field. The preliminary results presented in this paper demonstrate the feasibility of this design as an image sensor for the proposed Lyman Mission. Configurations yielding higher resolution and less-astigmatic images than those presented here, are likely to be found, since a number of parameters have latitude for a change. In an earlier study the spatial frequency response of an oblique magnetic-focus lens was analysed and found to be more than 90% at 25cyclesmm-‘ for an XUV photocathode and a uniform focus field (Lowrance, 1985). This high resolving power was substantiated by laboratory tests and IMAPS flight data (Jenkins, private communication). ACKNOWLEDGEMENTS The authors are indebted to David Brown and David Rutzel for substantial assistance in the development of the electron-optical ray-tracing programme. This work was funded in part by NASA grants NAG5-616 and NAS5-30110.
REFERENCES Beurle, R. L. and Wreathall, W. M. (1962) In “Adv. E.E.P.” Vol 16, pp. 333-340 Cash, W. C. (1984). Appl. Opt. 23, pp. 4518-4522 Coleman, C. I., Delamere, W. A,, Dionne. N . J., Kamminga, W., Long, D., Lowrance, J. L. and van Zuylen, P. (1979). In “Adv. E.E.P.” Vol 16, pp. 89-99 Johnson, C. B. and Hallam, K. L . (1976). In “Adv. E.E.P.” Vol. 40A, pp. 69-82 Lowrance, J. L. (1985). In “Adv. E.E.P.” Vol. 64B, pp. 591-599 Lowrance, J. L., Zucchino, P., Renda, G. and Long, D. C. (1979). In “Adv. E.E.P.” Vol. 52, pp. 44 1-452 Moos, H. W. (1986). NASA Proposal for Lyman Mission
This Page Intentionally Left Blank
Properties of Imaging Electron-optical Systems For Image Tubes
v. JARES Tesla-Vacuum Engineering, Prague, Czechoslovakia
INTRODUCTION Investigation of the electron-optical characteristics of high-resolution vacuum devices such as image converters and image intensifiers plays an important role in improving the performance of these devices. The shape of the electrodes, their alignment, the interelectrode distance and the applied fields co-determine the nature of the electron lens. The electron lens sections of image converters and image intensifiers are generally designed by using a computer model. The model developed in the Tesla-Vacuum Engineering Electron Optics Laboratories has proved very accurate in predicting the major lens characteristics. The electric fields are determined by describing the boundary of the problem in terms of position and potential of the electrodes. An iterative numerical method is then used to solve the Laplace equation. The trajectory programs use a step-by-step integration process to obtain the positions of the electrons.
COMPUTATIONAL METHOD The numeric solution of the Laplace equation makes use of methods based on approximating derivatives by differences. T o speed up the convergence of the iteration process in the sub-program for computing the field, the YoungFrankel super-relaxation technique was used, adopting an automatic approximate optimization of the iteration process. The sub-program for computing the electron trajectory involves the numeric solution of the motion equations by the Runge-Kutta technique in an axially symmetric electrostatic field. The program was set up in the FORTRAN language for computing on a HewlettPackard 1000 computer. Owing to memory limitations of the computer, three separate programs were used for studying the properties of the imaging system. Program FIELD conducts the computation of unit fields in accordance with the given input data. These include the geometric configuration of the system, the definition of the node point mesh and the fixed potentials of the 47 5 ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL. 74
Copyright 0 1988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
476
v. JARES
individual electrodes. The computed field can be expressed graphically in the form of equipotential lines, by printing a table of node point potentials or by transferring onto punched tape or disk for subsequent processing. Program TRAJECTORIES 2 serves for establishing the focusing voltage Ur of the required electron-optical system. To ensure optimal focusing of the electronoptical system for a point on the axis of the system, the intersection points of the trajectory sets emerging from a given point on the photocathode with the plane of the luminescent screen must exhibit minimum dispersion. The input data for program TRAJECTORIES 2 include the computed electrostatic field and a set of input parameters of the electron trajectories formed by the coordinates of the investigated point, of the energy and the output angle of the electron. The intersection points of the individual trajectory set with the plane of the screen are computer-processed and the standard deviation defined as the dispersion ring d,i of the image point is determined. The magnitude of the dispersion ring is investigated for a number of voltages with the aim of establishing the minima of d,i =f( Ur). The results obtained are tabulated and stored on disk or punched tape for graphical processing. By using TRAJECTORIES 1 for a particular value of the focusing voltage, the shape of the individual electron trajectories can be expressed graphically or in table form. The input data of TRAJECTORIES 1 are identical with those of TRAJECTORIES 2. The results thus obtained are then used for determining the imaging properties of the required electron-optical system (magnification and distortion of the image, shape of the image plane and resolution). A TRIODEIMAGING SYSTEM The programs mentioned above for computing the distribution of the electrostatic field and the shape of the electron trajectories were adopted to investigate some properties of a simple triode imaging system. In operation, the system is focused by varying the focusing voltage Ur. The desired properties of the focused image can be obtained by altering the geometric configuration of the electron-optical system. Figure 1 shows the initial configuration of a triode imaging system with a spherically curved photocathode and a plane luminescent screen. This configuration was used for studying the effect of altering the curvature Rfk of the photocathode, the distance A of the anode aperture for the centre of the photocathode and of the position F of the focusing electrode for a constant photocathode-screen distance L (JareS, 1986).The computations were carried out for an anode voltage of 10 kV. The values of the trajectory input parameters (YO, W, a) were chosen with regard to the extent of the computations in such a way that the variations in the imaging parameters could be studied. Under these conditions, the electrons were assumed to leave the photocathode in a normal direction and with deviations
IMAGING ELECTRON-OPTICAL SYSTEMS
A
477
'
L
FIG.I . Cross-section of a triode imaging system.
of & 30°,& 60" in the meridional plane. The output energy of the electrons was assumed to be 0.5 eV. This value depends on the type of the photocathode; if necessary the program can be employed for processing the trajectories of any number of values of electron output energies and for studying their effect on the imaging parameters of the system (chromatic aberration).
EFFECTOF
PHOTOCATHODE CURVATURE ON IMAGING PARAMETERS
Using the program FIELD, the unit electrostatic fields were computed for systems having photocathode curvature radii of Rfk = 13, 15, 17,20 mm and plane photocathodes. TRAJECTORIES 2 was used to compute the dependence dSi=f(U,-)for the following input parameters of the electron trajectory: ro=O, 2, 5.4, 8 mm; W=OS eV; u=O", k 30", )60". By processing the minimum values of the dependences d,i = ( L I T ) , relations were obtained for the dependence of the focusing voltage of the focused system Um=f(r0) and the dependence Ufo =f (Rfk).Table I gives the values of the focusing voltage Vfo for the chosen radii of the photocathode curvatures. TRAJECTORIES 1 was TABLE I
Focusing voltages for specific photocathode radii of curvature Rrk (mm)
U',(V)
13 192
15 152
17 114
20
plane
73
-10
478
v. JARES
FIG.2. Determination of the photocathode curvature Rfk for minimal image plane curvature.
used to express, in tabular and graphical form, the shape of the individual electron trajectories for a given value of the focusing voltage. This program yielded the values dsi, d,, F, (for a trajectory with u=O) of the individual trajectories for focusing voltages listed in Table I. The computed values were used to assess the effect of the photocathode curvature on the imaging parameters of the system. The meridional cross-section of the image plane was approximated by a parabolic function of the form z =ar2+b. For the chosen focused system b = 50 mm; then a =f(Rfk) defines the optimum for a = 0. This dependence is shown in Fig. 2 where it can be see that a minimum curvature of the image plane is obtained for Rfk = 16.4 mm. This results in a minimum resolution drop in the peripheral regions of the image. In Fig. 3 the meridional cross-sections of the image planes for individual curvature Rfk are presented. The values dsi,F, listed in the tables of the trajectories computed by program TRAJECTORIES 1 were arranged into the following functions: dsi = p x lo4 Ir,),
(1)
M=mxri+n,
(2)
where the parametersp, q. m, n, which are dependent on R,,, were determined by linear regression. These parameters are given in Table 11. From equations (1) and (2) a relationship may be derived for the resolution as a function of radius r,:
479
IMAGING ELECTRON-OPTICAL SYSTEMS
ZObr ( rnrn 1
FIG.3. Meridional cross-section of image planes.
TABLE I1 Parameters used to compute resolution Rfk (mm)
P 4 m x lo3 n
13 2.6 0.1622 -0.147 -0.9678
15 2.6 0.0978 -0.142 -0.9537
17 2.6 0.1156 -0.115 -0.9360
20 2.6 0.1733 -0.134 -0.9240
plane 4.5 0.3111 -0.335 -0.9390
The results obtained were plotted in a diagram showing S versus R n (Fig. 4). This may be used to estimate the value of Rfk for maximum resolution of the image in the range from 15 to 17 mm. Similarly, by using equations (1) and (2) in the relation for the geometrical image distortion D = (M/Mo- 1) x 100% and by modifying
the minimum in the region Rfk=17mm may be estimated from the
v. JARES
480 2001
1
;13
14
15
16
li
18
19
20
6)
R,, ( mm 1
FIG.4. Resolution of the image in the meridional plane.
0010
13
14
15
16 17 Rfk(mm)
18
19
20
FIG.5. Estimate of the geometrical image distortion minimum.
relationship between C, and Rfk (Fig. 5). The results obtained may be used to determine the optimal photocathode curvature Rfk. CONCLUSION
Basic relations have been presented governing the performance of a triode
IMAGING ELECTRON-OPTICAL SYSTEMS
48 1
imaging system assembled within tolerances currently encountered in normal production processes. These relations can be used with advantage if certain parameters of the imaging system are to be altered or optimized.
REFERENCE JareS, V. (1986). ‘In “Proc. International Symposium on Electron Optics”, Beijing, China, Sept 9-13
This Page Intentionally Left Blank
Numerical Evaluation of Spread and Transfer Functions for Image Intensifiers with Polychromatic Illumination J. M.WOZNICKI Institute of Microelectronics and Optoelectronicst. Warsaw University of Technology. Warsaw, Poland
INTRODUCTION Over the past 30 years a great deal of effort has been spent to develop methods of using digital computers to solve electron-optical problems. Many publications have discussed image quality and, particularly, the theory of electron-optical focusing and aberrations. However, an improvement in the methods for numerical calculation of the field distribution and for electron ray-tracing makes the numerical evaluation of system properties of real devices more practical. Various parameters have been applied for assessment of imaging systems. In the case of linear systems, spread and transfer functions seem to be the most important. The basic element in the formation of an image is the response to a point source, that is the point-spread function (PSF). The closely related line-spread function (LSF) can be obtained by passing a narrow slit across the field. We can also scan the field with a knife-edge instead of a slit and obtain the edge-spread function (ESF). The LSF may be computed as the integral of the PSF or as a derivative of the ESF. A powerful technique for analysing spread functions is the Fourier transform, which determines the optical transfer function (OTF). Thus we can analyse the modulation for any frequency present in the image using the modulation transfer function (MTF), and the phase shift for that frequency can be analysed by means of the phase transfer function (PSF). In the case of an image intensifier system the spread of a point-source image results from the initial energy and direction distributions of emitted photoelectrons. The initial-energy distribution curves for a given photocathode depend on the wavelength A of the incident radiation. It results in a spectral dependence of spread functions and in consequence, of transfer
t Formerly known as the Institute of Electron Technology 483 ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS VOL. 14
Copyright 01988 Academic Press Limited All rights of reproduction in any form reserved ISBN 0-12-014674-6
484
J.
M.WOZNICKI
functions. With polychromatic illumination these functions and thus the performance of the image intensifier tube are affected by the spectral detector responsefll) = s ~ D Iwhere , SI is the spectral energy of the incident radiation and D1 is the photocathode sensitivity. The influence of polychromatic illumination on image intensifier characteristics has been analysed by a few authors (McDowell et al., 1973; Nijhawen et al., 1983; Woinicki, 1986, 1987). However, spectral broadening effects on spread functions and OTF-response, especially for real systems with PSF asymmetry have not so far been discussed in detail. The aim of this paper is to present methods for full numerical evaluation of spread and transfer functions of image intensifiers under polychromatic illumination. Rather than the PSF, which is difficult to describe, the LSFs are considered in two mutually perpendicular directions, i.e. sagittal and tangential. The methods used are based on ray-tracing data. BASICFORMULAE Generally, the image transfer properties of an imaging system can be fully described by means of a two-variable OTF-response. It is more common to treat the OTF as a one-variable function of spatial frequency N for an arbitrarily chosen image orientation with an orthogonal axis u. Linearity of the imaging system is assumed, which enables calculation of the polychromatic OTF = MTF exp (J'PTF) as a weighted average of the monochromatic OTFs: OTF(N) =
I f l l ) OTF(N,l) d l Jfl4 d l '
where OTF(N,A) =
SLSF(u,l) exp( --j27cNu) du ILSF (UJ) du
or OTF(N,I) =
Jd/du ESF(u,l) exp( -j27cNu) du jd/du E S F ( u , l ) du
9
(3)
u is equal to x or y for the sagittal or tangential orientations of the line respectively. Substituting equations ( 2 ) and (3) in equation (I), changing the sequence of the integration and also, in case of equation (3) using the formula for the Fourier transform of the derivative we find:
OTF(N) = f [ f f l l ) A ( l )LSF(u,l) dAjfll) dl] exp( --j2nNu) du
(4)
EVALUATION OF SPREAD AND TRANSFER FUNCTIONS
48 5
and
OTF(N) = fj2nN[fflA)A(1) ESF(u,l) d1fflA)dAI exp( --j2nNu) du
(5)
where
In equation (6) the LSF(u,l) is assumed to be zero or at least negligibly small at some finite limits denoted by k urnax.The urnaxvalue is defined as a sufficiently large number, the same for all the monochromatic LSFs considered. Obviously, in case of polychromatic illumination the OTF can be expressed in terms of the polychromatic spread functions:
ur
LSF(u) exp( -J?nNu) du
OTF(N) =
”T”LSF(u) du
3
(7)
- urnax
”$j 2 n N ESF(u) exp( -j2nNu)
du
By comparison of equations (4) and (5) with equations (7) and (8) respectively, we obtain formulae for the polychromatic spread functions as the weighted average of the monochromatic ones:
f LSF(u)du, - urnax
ESF(u) =
ffl1)44ESF(u,4 d V f f l 4 d l ESF(urnax) - ESF( - urnax)*
(10)
Equations (9) and (10) as well as equation (1) are the basis for the numerical formulae.
NUMERICAL PROCEDURES
Numerical evaluation of the monochromatic spread functions is carried out by counting the photoeIectron trajectories separately in properly shaped regions on the screen. The ESF-procedure uses a set of half-planes; the LSFprocedure uses strip-shaped areas (see Fig. 1). The calculated spread functions
486
J . M. W O ~ N I C K I
Y
FIG.1. Illustration of:..L
main concept of the numerical method.
are tabulated, optionally smoothed, and the monochromatic OTF is computed as a Fourier transform according to equations (2) or (3). This method is evolved from the procedure used by Stark et al. (1969). Photoelectrons issued from a cathode point P are selected by determining the values of the initial parameters, energy Woand angles q , 4 (see Fig. 2). This enables the weight-function values for a given photocathode to be determined (Csorba, 1972). All these values are assumed to be equal and the increments of the initial parameters of consecutive selected photoelectrons are calculated according to that assumption. Since electron-ray tracing is still a timeconsuming procedure, interpolation of the end-point 2-dimensional distribution is also applied. The use of these procedures is very practical and gives good numerical efficiency for computation of the spread and transfer functions (Woinicki, 1987). The errors of the end-point position calculation must be at least an order of magnitude less than the spot-size diameter. Our programs, which satisfy this condition, and apply to systems with axial symmetry, use the SOR (successive over-relaxation) method and various two-dimensional interpolation formulae of the fourth order. To get sufficient accuracy in the end-point position calculation, predictor-corrector methods are useful. In order to improve
EVALUATION OF SPREAD AND TRANSFER FUNCTIONS
487
FIG.2. Choice of the initial electron parameters.
numerical efficiency of the programs, a special optimization process for the parameters of the mathematical procedures was introduced. This relates to tolerances and other executive parameters. Polychromatic spread functions are calculated for the sagittal and tangential line orientations according to equations (9) and (10) on the basis of monochromatic LSFs (and ESFs respectively) using the following numerical integration formulae with a,-coefficients for various wavelength values li (i'l, 2, . . . imax):
c afi
imnx
LSFsag/tang(xj/yj) =
42g/tang
LSF(,'?,,t,,, (XjlY,) imax
=
Asagjtang
c aif,
i= I
where
(1 1)
J. M. WOZNICKI
488
1
1
I-
and i = 1, 2, . . . imax; 11"
.
i
I
A,,A,, . . . , li,. . . Aimax
;j = 1,2,
...
.
jmax;
f, = SAi DAi=f(A,).
The polychromatic transfer functions are calculated from the monochromatic MTFs and PTFs according to equation (1) and the definitions MTF(N) = IOTF(N)I, PTF(N)
= arg[OTF(N)]
(14)
where the OTF(N) is given by the following numerical formula: 1aif;MTFXN7Ji)exp[~PTF~(N,11~)1 OTF(N)=
(15)
Caif,(iiI i
Note that the polychromatic MTF cannot be generally computed as a weighted average of the monochromatic MTFs because the effects of the phase transfer function have to be included. Obviously the phase shift vanishes on-axis and for the sagittal line image orientation.
;ooo
10,000
20,000
30,000
40,000
50,000
FIG.3. The electron-optical system of the diode image intensifier taken for the illustrative calculations.
EVALUATION OF SPREAD AND TRANSFER FUNCTIONS
489
RESULTSAND DISCUSSION
The mathematical procedures described above were used as a basis for the computer programs, which were applied to obtain the results published here.
I"\
I
t
I x 1 0 'Z~ n DOPED 60 A s
cs+ ( 0 +Cr )
01
v) C
0
I P n
!i?
1.5
2.0 2.5 Electron energy ( e V )
FIG.4. Electron energy distributionsfor type (I) photocathode for photon energy increments of 0.2 eV (James and Moll, 1969).
FIG.5. Theoretical energy distribution statistics using a cosine law (type I1 photocathode).
490
J . M. W O ~ N I C K I LSF
soplttol
1987-08-29
2209
4,OE-01
LSF 3.21-01
3.K-01 3.C-01
3.OE-01
3.4E-01
2.8E-01
3.2E-01 3.OE-01
2.4E-01 2.2E-01 2 OE-01 LEE-01
2.8E-01
2.6E-01 Z.*E-Ol 28E-01 2.m-01 1,BE-a 1.6E-01 ME-01
1.6E-01 1.4E-01 1.x-OI
1.PE-01
1.OE-01 8,OE-02 6.OE-02
1.OE-01 8.OE-02 6.E-02
4.OE-02 2.K-02 .OE+OO -2.OE-02 -1.E-02 12nss.xt.c7
MONOCHROMATIC .OE+OO
1.E-02 wp
2,OE-02
2.E-02 .1 V .
.OE+OO tnn1
wp =
12n55cxt.c11
.I ev
MONOCHROMATIC
1.OE+00
8.OE-01 WE-01
4.OE-01
L
I
4.OE-01
2.E-01
2.OE-Ol
OE+OO -2.0-02 -1.OE-02 12n55cxt.c7
.OE+OO
1.OE-02
wp =
MONOCHROMATIC
2.OE-02 [nnl .1 ev
.OE+OO
It
.OE+OO
12n55sxt.cll
wp=
.lev
MONOCHROMATIC
FIG.6. Selected LSF and ESF curves for photocathode 11.
The programs were comprehensively examined using typical computerprogram testing methods. The problem of two concentric equipotential spheres, for which the complete analytical solution is known, was selected as the best test for the ray-tracing program. As an example of the application of the methods and programs presented, calculations were made for a diode image intensifier with a centrally symmetric electrostatic image-inverting lens (Fig. 3). The photocathode was assumed to be a polyenergetic electron emitter obeying Lambert's law with two different spectral photoelectron-energy distributions: (i) experimentally obtained, normalized and smoothed curves for Zn-doped GaAs crystal with Cs (0 Cs) surface treatment (Fig. 4), taken for increments of 0.2 eV over a photon energy range of 1.4-3.0 eV, i.e. for the I-range of412483 pm (James and Moll, 1969); (ii) analytically defined energy distribution statistics obeying a cosine law (Csorba, 1972) taken here as a discrete set of curves (Fig. 5 ) calculated from the formula
+ +
49 1
EVALUATION OF SPREAD AND TRANSFER FUNCTIONS LSF 7.OE-01
tanQcntlol
1987-08-29
11154
LSF 2.4E-01
6.OE-01
1987-08-29
srlgiitta.i
7
-
2.2i-01
'
'
I
'
10193
'
I
2.OE-01 -
1.8E-01
5.OE-01
-
1.6E-01 i . 4 ~ - n 1-
4.OE-01
-
1.2E-GI 3.OE-01
1.OE-01
-
mi-02
2.01-01
-
6.OE-02
~ . o E - Q ~ tungsten
ME-01 .OE+OO
-1OE-01 12n55ert.cll
.OE+OO
Inn1
LOE-01 wp-
M 0 N0 C H R 0MA T I C
cmFJ
,le~j
wp =
12n55rxt.cll
P 0L 'I' CHR 0M A T I C
tw
1 ev
sogittci
1.OE+00
8.OE-01
6.OE-01
4.OE-01
2.OE-01
M0 1.10C HR OM A T I C
-WE-01 12n55ert.cll
.Oil00
I.OE-01 wp
.Oi+OO C"d
cnn1
.OE+OO
12n55ext.cll
.1 eV
wp =
.I
F'/
PULY C H R O M A T I C
FIG.7. Selected LSF curves for photocathode 11.
where
over the same incident photon energy range as (i). In equation (17) EAand EG are the electron affinity and the forbidden band gap energy, respectively, of the photocathode material. Taking the assumption made by Nijhawen et al. (1983), the work function EA EC for the S 25 photocathode was calculated from the condition V,,, = 0 at 1= 910 nm. P * 20 phosphor and tungsten 2870 K detector response curves were also assumed.,Optical frequency v was
+
-
492
J. M. W O ~ N I C K I
used instead of wavelength 1 in the calculation. Monochromatic spread and transfer functions were evaluated for nine v-values. The calculations were carried out for the two directions given by sagittal and tangential orientations of the image line for two different regions of the screen. The position of these regions was determined by the ray endpoints of electrons emitted from photocathode points with radial co-ordinates & = 0 mm, RO= 7 mm and Ro = 1 1 mm. The monochromatic spread and transfer functions calculated on an IBM AT microcomputer are shown in Figs. 6 to 12. It can be seen that the spectral-spread effect can be significant when white-light illumination is applied. This effect combines with the chromatic and geometrical aberrations. The contribution of the spectral spread to the increase in spot size varies between 10% and 100% of the monochromatic spot diameter on the screen (Figs. 6 , 7). The influence of the spectral effects on the transfer functions (especially the MTF) is therefore significant.
M F I.OE+W
l.OE+OO
8.OE-01
B.OE-01
6.OE-01
6.OE-01
4.E-01
4.OE-01
2.OE-01
2.OE-01
.OE+00
,OE+OO
2,OE+01 12nJSext.bO
4.OE+OI
8.OE+-01w n n 1 up = 1.2 eV
6.0€+01
.OE+00 .OE+OO 2.OEt01 12nS5ext.bO
MONOCHROMATIC rrr
tsngentlol
1987-OE-E9
t0nQentlol
13846
6aE+O1 8.OE+OI1 Cl/""I u p = 1.2 ev
4.0E+OI
POLY CHROMATIC tangentla1
FTF
1713Q
lII87-OE-CI
6,OE-02
1987-08-29
17611)
" " ' I ' " '
4.OE-02 2.Oi-02 -1.9E-09
-2.E-02 -4,OE-02
-6.OE-02 -8.OE-02
.oi+oo
, , , I , I I I I I I
1.OEcOl 12n55ext.bll
I
2.OEc01 up = 1.2
MUNUCHROMATIC
,
W""1
ev
-BOE-02 ,OE+OO
, , ,
1
l.OE+Ol
12n55ext.blI
, , ,
, , , ,-
I
2.OEt01
w p = 1.2 ev
PDLYC H R D M A T I C
FIG.8. Selected MTF and PTF curves for photocathode I .
c1/nn3
493
EVALUATION OF SPREAD AND TRANSFER FUNCTIONS MTF
1.OEiOO
l.OE+OO
8.OE-01
8.OE-01
6.0E-Ul
6.OE-01
4.0E-01
4.OE-01
2.OE-01
W""1
1 0
.01
l.OEt02 wp = .1 ev
,55eut.c7
MUNUCHRUMATIC
PTF
1987-09-24
tamentlol
[I/""l
POLY C H R O M A T I C 23803
wrs
I%~-o~-cY
rangentiat
FfF
1.9E-09
1.OE-01
7%-09 2870 K
-1.OE-01 -1.2E-01
-2.OE-01
-ME-01 -3.OE-01 -4,OE-01 Il^".lc/L,L7
.OE+OO 12n55ext.c7
l.OEt02 Up = .1 eV
,
-2.2E-011
I
-5 OE-01
wnn1
.OE+OO
12n55ext.c7
,
,
, 4
,
L.OEt02 wp = .I ev
C1innI
POLY C H R O M A T I C
MONOCHROMATIC
FIG.9. Selected MTF and PTF curves for photocathode 11.
There are important differences between the two spectral photoelectronenergy distributions. This implies that the problem of photocathode spectral sensitivity variations which can occur between cathodes produced by different manufacturers, or even by the same manufacturer, may be significant. This problem seems to be rather difficult to analyse experimentally and has not been attempted in detail so far. The methods of computer simulation described here are fully applicable to such problems. Examples of the calculated polychromatic spread and transfer functions for two detector response curves relating to a P 20 phosphor and tungsten 2870 K radiation (Nijhawen et al., 1983) are shown in Figs. 7 to 10. Differences in the polychromatic spread and transfer functions for the two given sources of illumination are apparent. The results obtained for sagittal and tangential orientations of the line image on the screen distinctly show the astigmatic asymmetry of the PSFs for
-
tanyentior
NTF
1987-08-t9
12
l.OE+OO
a.oE-oi WE-01
4.K-01
BE-UI .OE+OO
.OE+OQ L.OE+OL 12n55ext.cIl
2.OE101
3.OE+Ol wp = .I ev
WFi
I I I I I I I I I ( ( I I I J I I I / 2,OE+Ol 3.OE+OI
.OE+OO .OE+OO I.OE+OI 12n55cxt.cll
HTF
soglttol
1987-08-29
.I
4
mr
1143
l.OE+OO
l.OE+OQ
8.OE-DI
8.DE-01
6.OE-Dl
6.K-01
4.OE-01
wp =
MONOCHRUMATIC
MONOCHROMATIC
\
\
4.OE-01
iungsien 2.370 K
.OE+OD L.OEtO1 12n55.xt.cIl
tungsten 2870 K
2BE-01
2 . N-01
2.OE+Ol
3.OE+OI wp =
[IlIW 12n55rxt.cIl
.I P V
FIG.10. Selected MTF curves for photocathode 11.
PTF
vp
f
POLY CHROMATIC
POLY CHROMATIC
tanaentlal
1987-08-25
17101
POE+OO
-2 OE+OO -3.OE *OD .OE+Oo 2.0E*01 12n55rxt.bO
4.OEt01
-
b,OE+Ol 8.OEcOl W n d wp 1.2 ev
MONOCHROMATIC FIG.1 I . Selected PTF curves for photocathode I.
.I -v
49 5
EVALUATION OF SPREAD AND TRANSFER FUNCTIONS PTF
Z.OE-02
.OE+OO
.OE+OO l.OEt01 12"ssext.cll
2.0€+01
3.OEiOl wp = .1 ev
MONOCHROMATIC
Cl/R"I
tongentlQ1
1987-08-29
11850
-
-ax-02
-
-4.OE -02
-
2870 K
l.GEtO1 12n55ext.cll
.OE+OO
2,0E+Ol
3.OE+OI wp =
[l/FIFlI .1 e v
PULYCHROMATIC
FIG.12. Selected PTF curves for photocathode 11.
the greater ROvalues (see for instance Figs. 8 and 10 with Ro = 1 1 mm). This distortion is slightly dependent on the spectral effect. The calculated PTF-curves (for example Figs. 1 1 and 12) show the expected phase differences. Because of the rotational symmetry of the electron-optical system, the PTF for the axial image point is zero, or possibly & II when the real OTF-value passes through the value zero (Fig. 11). For off-axial image points, especially for large ROvalues (e.g. Ro= 1 1 mm), the PSF is highly asymmetric and the PTF varies strongly with spatial frequency. The effect of polychromatic illumination on the PTFs, although more obvious in the figures than for the MTFs, is less significant because of the small values of the phase shifts. Moreover, the phase shifts arise in aberrated systems, which are usually spectrally less sensitive with real photocathodes. CONCLUSIONS
The procedure and programs described enable calculation of subtle spectral effects in image intensifier spread and transfer functions. The IBM AT microcomputer has been used succesfully for the calculations. The importance of using polychromatic illumination to specify the performance of image intensifier electron-optical systems has been stressed. The influence of polychromatic illumination on spread functions and transfer functions is apparent. The most important result is the considerable spectral dependence of MTFs. Polychromatic spread functions can be computed as a weighted average of monochromatic ones. Polychromatic transfer functions (MTF, PTF) must be generally determined by means of the polychromatic OTF obtained from the weighted monochromatic OTFs. The effect of various sources of illumination and of the spectral sensitivity of the
496
J. M. W O ~ N I C K I
detector on image intensifier performance is considerable for both sagittal and tangential orientation of the line image in the same local region of the screen. The influence of spectral effects on image intensifier electron-optical systems is determined above all by the spectral sensitivity of the photocathode.
REFERENCES Csorba, I. P. (1972). RCA Rev. 33,393-398 James, L. W. and Moll, J. L. (1969). Phys. Rev. 183,740-743 McDowell, M. W., van Rooyen, E., Dicks, L. and Boettcher, A. J . (1973). Proc. S.P.I.E.42,59-69 Nijhawen, 0. P., Gupta, S. K. and Hradaynath, R. (1983). Appl. Opt. 22,2453-2455 Stark, A. M., Lamport, D . L. and Woodhead, A . W. (1969). In “Adv. E.E.P.”Vol.28B, pp. 567575 Woznicki, J . M. (1986). Proc. S.P.I.E. 702, 159-162 Woznicki, J. M. (1987). Proc. S.P.I.E. 818 (in press)
Index
A
B
Acton interference filter, 188 ADC offset, 288 Afterpulses, 108-109 AGA 680 Thermovision system, 397 Aliasing, 154, 159 Amorphous photoconductive layer in avalanche mode, see Se photoconductive target Se-Te-As,see Pick-up tubes for HDTV Amplifier noise in FAST detector, 289290,293 Analogue integration mode, 183 Anti-veiling glare windows, 410-412 Arrays for IR astronomy, see Infrared astronomy large-format photon-counting, see Lyman Ultraviolet Space Telescope silicon photodiodes, 269, 271 see also Bilinear CDD array Astigmatism, 444-446, 470 Astronomy cooled CCD systems for, 129-130 extragalactic, 206-207 intensified CDD camera system, 30,3 1 Lyman, 298 MCP intensifier for photon counting, see Photon counting wide field camera for see Wide field astronomical camera see also Faint Object Camera (FOC): Infrared astronomy Atomic physics, photon-counting in, 97105 Auger analysis of surfaces, CsI/Na photocathodes, 324,325-328 Avalanche multiplication, see Se photoconductive target Average single-event amplitude test, 183
Beryllium windows, 248, 250, 262, 264 Bias charge, bilinear CCD array and, 175-179 BIB, 202 Bilinear CCD array, 173-179 bias charge, 175-179 electro-optical characteristics, 178, 179 noise level, 175- 179 read-out efficiency, 173, 175, 177 Biological studies cooled CCD system for, 131-133 Gel electrophoresis of proteins, 132 image intensifier-Vidicon systems, 126-127 immunoassays, 131 X-ray imaging and microscopy, 132133 Black body radiation, somnoluminescence and, 123 Black hole, veiling glare and, 406,407 Blocked-impurity-band (BIB) detectors, 202 Blocking pulses, in Sapphire-Frame, 241, 243 Blocking target layers, see Se photoconductive target Blue response, 27 Brushing technique, 39 Bump-bonding, 202
c Cameras FOC, see Faint Object Camera (FOC) HDTV, see Pick-up tubes for HDTV HTH MX, 4 , 6 IRCAM, 203,204,205-207 picosecond framing, see Picoframe camera
497
498
INDEX
Cameras (cont.) Schmidt camera system, see Schmidt camera system streak see Streak cameras subnanosecond multi-framing, see Sapphire-Frame Teledyne, 4 TI, 194 wide field, see Wide field astronomical camera see also CCD camera system: Electron bombarded CCD Cascade image intensifier, 428,429 CCD camera system fibre-optically coupled intensified, 2733 astronomical experiments, 30, 3 1 commercial version, 30-3 1 performance, 29-30 resolutions, 30 system description, 27-29 ISIS camera system 31-33 CCD-Digicon, 55-67 dark noise, 57 slow-scan system testing, 56-62 centroiding, 58, 60-62 experimental set-up, 56-58 photoelectron charge spreading, 5860 video-rate operation, 62-66 drive electronics, 62-64 frame-grabber, 64 CCDs bilinear array, see Bilinear CCD array cooled for astronomy, 129-130 for biomedical applications, 13 1133 performance of, 130- 131 electron bombarded, see Electron bombarded CCD in X-ray detector, see Electro-optical X-ray detector, with CCD LLL TV with, 9-1 5 NXA 1010,28, 30, 31 NXA 1011, 30, 31 photon-counting streak camera with, 22 1 RCA, 183 RCA SIDSOl-DS, 56
swing operation, see Synchro Vision (SV)-CCD TH 7861, 17, 18 TH 7864,25 TH 7866,25 TI, 183 used with image intensifiers see 11/ CCD system see also Synchro Vision (SV)-CCD CdTe/a-Si :H heterojunction use, 257267 absorption coefficient of CdTe and a-Si:H film, 258 electrical properties of CdTe, 261 imaging devices, 262-266 polycrystalline heterojunction, 260262 single crystal heterojunction, 259-260 sputtering technique, 260, 261 Centroiding, 58, 60-62, 301 Cerenkov source, 183, 184, 190 CERN Scintillating Fibre Detector intensifier chain system, 35-38 screen efficiency, 39 special requirements, 38-39 spectral responses, 37 Channel intensifier tubes, pulsed biasing, see Pulsed biasing Charge-coupled devices, see CCDs Charge spreading, photoelectron, 58-60, 66 Charge transfer efficiency, 130-1 3 1, 175 Chemiluminescence, 123 CHOP data acquisition mode, 205 Coherent noise, 65 Colour imaging, SV-CCD image, 161164 Cooled CCDs, see CCDs Corning 7056 glass, 405 Cross-talk, 113-1 14 Cryostat, 277 Crystalloluminescence, 124- 126 CsI/Na photocathodes, 323-330 Auger analysis of surfaces, 324, 325328 element distribution on surface, 325328 element profile in depth direction, 328329 intermediated layer, 324
INDEX
production methods, 323-324 sensitivity, 330
499
charge transfer, 130-1 3 1 quantum see Quantum efficiency reading, 173, 175, 177 D signal charge transfer, 175 Electric field calculation, 440441 Dark counts, 108 Electro-optical X-ray detector, with Dark current CCD, 275-283 definition, 287 detector, 276-277 in heterogeneous semiconductor junclinearity of response, 278 tions, 363-365 measurement precision, 28 1-282 see also Noise noise estimates, 281-282 Dark emission, 290-291 performance measurement, 277-282 see also Equivalent background illumiresolution, 280-28 1 nation (EBI) sensitivity, 278 Dark noise, 51 spatial distortion, 279 CDD-Digicon, 57 uncertainty in X-ray dose, 281-282 Debye screening length, 362 uniformity of response, 278-279 Deflection electrodes, twisted arrow- Electron bombarded CCD, 165-166 pattern, 392 collection of holes, 168-169 Degradation, hygroscopic, 253-254 dark current, 168, 170 DEMA, see Dynamic effective medium depletion zone, 167 approximation (DEMA) electron counting, 170-171 Demagnifying image intensifiers, 1-8 for far UV, 181-200 Detective quantum efficiency (DQE), 147 analysis methods, 182-183 cooled CCD, 130 applications, 195-199 determination, 182-183 comparison and testing of, 182-192 electro-optical X-ray detector, with developments current and future, CCD, 275 194-199 FAST detector 291-293 DQE determination, 182-183 LFPCA, 306-307 flight-worthy system, 195 relative, 32, 33 nightglow, 197, 199 signal induced background and, 429 space-flight capability, 195 Digicon, see CDD-Digicon tests at NRL, 188-192 DigiVision’s FluoroVision, 64 tests at Princeton, 183-188 Diode-mode electron gun, 393 front-face bombardment, 166, 184 Diode tube, proximity, 27 gain 169-1 70 Diodes, Schottky barrier, 202 passivation layer, 167 Dynamic effective medium approximarear-face bombardment, 166, 184 tion (DEMA), 339 technology of, 166-169 Electron counting, 170-171 E Electron energy-loss spectroscopy, 152153 Echelle detector, 31 1-312 CCD and YAG detector for, 152, 153Edge-spread function (ESF), 483 155 EDOE Electron microscopy improvement 8 1-84 image recording in, 147-1 56 of MCP and resolution see Resolution microdiffraction, 148-1 51 EELS, see Electron energy-loss spectroSTEM, 148 see also Electron energy-loss spectroscopy Efficiency XOPY
500
INDEX
Electron-optical characteristics for image tubes computational method, 475-476 photocathode curvature and, 477-480 triode imaging system, 476-477 Electron scrubbing, 49 Electron temperature measurement, 228 Electron trajectories calculation, 441-442 Electrons, signal induced background and scattered, 430 Electrooptical performance, II/CCD couplings, 17-18 Electrophoresis of proteins, 132 EM1 9912 image intensifier, 119,428 EMR 541G photomultiplier, 184 Endeavour programme, 303 Energy distribution of output electrons, see EDOE Equivalent background illumination (EBI), 287-288,290-291 EUREKA EU95 Project, 394 Extragalactic astronomy, 141-144, 206207
F Fabry-Perot etalon, 205 Faint light source, 425-427 integrated-sphere attenuator, 425-427 Faint Object Camera (FOC), 413-423 charge control on output window, 42 1-424 Feldex layer, 419, 420, 422 final assembly, 422-423 high voltage scheme, 420-421 input magnesium fluoride window, 415-416 internal charge control, 417-420 mica target supports, 416 phosphors, 417 SbNaK photocathode, 416-417 Faint Objects Spectrograph Digicon tubes, see FOS Digicon tubes FAST detector, 285-296 data quality, 295-296 definitions, 287-288 detector sensitivity, 288-289 DQE, 291-293 dynamic range, 293-294
maximum counting rate, 293 noise, 289-291 non-uniformity of response, 294-295 spatial distortion, 294 spatial resolution, 295 temperature effects, 293 Fast oscilloscope, 172 Fast shutters, 69 Feldex layer, 419, 420, 422 Fibre optic aging, 39 Fibre optic coupling, 1-2, 17 Fibre optic taper, 28, 29, 32 Fibre optic windows, 41 input windows, 9-15 see also CCD camera system FIELD program, 475,477 Flat field, 104 Fluorescence microscopy, 126-127, 133 FOC, see Faint Object Camera (FOC) FOS Digicon tubes S-20 photocathodes, 347-358 process development, 352-354 quantum efficiency (QE) instability, 349-352 quantum efficiency (QE) profile, 348 requirements, 347-348 Spectracon Tubes, 348-349 stability considerations, 354-357 Fourier transformation, 483 Frame period, defintiion, 287 FT4-P sensor, 169
G
GaAs photocathodes, 2,405 Galactic north pole, 141-144 Sandage-Vkron field, 142 selected area 57, 142-143 Gamma-ray imaging, 247,25 1 see also Scintillation plate, CsI(Na) Gel electrophoresis of proteins, 132 Geometrical distortion, see Uniformity GI-41-1 triode, 241 Global Imaging Monitor of the Ionosphere (GIMI), 199 Goddard Space Flight Centre, 197 Gunn diode, planar, 398, 401,402,403
INDEX
H
50 1
for Hubble telescope, see Faint Object Camera (FOC) for X-ray, see X-ray imaging Hamamatsu R2519 photomultiplier MTF of, 10 tube, 107 Mullard XX1500,31, 32, 33 HARP see Se photoconductive target photon counting for astronomy, see HB501 STEM, 148-151 Photon counting Heterojunctions, see CdTe/a-Si :H proximity-focussed, signal induced heterojunction use: Vidicon with background in, 431-432 amorphous a-Si :H target resolution, see Resolution second High definition television, pick-up tubes generation, 15 for see Pick up tubes for HDTV used with CCDs, 17-25 Hubble Space Telescope, 347 veiling glare, see Veiling glare FOC see Faint Object Camera (FOC) XX 1330A, 209 Hybrid tubes, 3, 6 see also EIectron-optical characterHydrogenated amorphous silicon, 257 istics for image tubes: Spread and see also Vidicon with amorphous transfer functions a-Si :H target Hygroscopic effect degradation, 253-254 Image inverter tube electrode structure, 457 electron optics, semi-analytic model, I 457-459 model comparison with numerical II/CCD system, 17-25 results, 462 coupling types and performance, 20-25 modulation transfer function, 460 electrooptical performance, 17-1 8 numerical analysis, 459461 input image format, 19 optimization method, 461 number of TV lines, 19 optimization of imaging properties, operating illumination range, 21 457-463 over-illumination and gain control, 24 sensitivity analysis, 459, 461 overall performance, 24-25 Image recording, solid-state sensor, 9-1 5 reliability and lifetime, 22, 24 IMAPS, see WICCD, for IMAPS soundresolution, 21-22 ing rocket sensitivity, 18, 20 Immunoassays, 131 signal to noise ratio, 18-19, 20-21 Impurity doping, 386 Illumination range, 21 Infrared astronomy, 201-207 Image instability, LFPCA, 307 array performance requirements, 202Image intensifier chains, for CERN scin203 tillating Fibre Detector see CERN detector technology, 201-202 Image intensifier-Vidicon systems IRCAM, 203,204,206-207 biological studies, 126-127 results with IRCAM, 207 crystalloluminescence, I 2 4 126 semiconductor materials, 201-202 somnoluminescence, 1 19-124 UKIRT, 203 Image intensifiers Infrared lasers, 374, 376, 377 18mm proximity-focussed wafer tube, Infrared sensitive detectors, MCP-PMTs 1 as ultra-fast wide-band 87-96 cascade, 428,429 Infrared vidicon, 369-377 CCDs and, 9- 15 dark current, 373 demagnifying, 1-8 IR lasers, 374, 376, 377 EEV P8304A, 28 light-transfer characteristics, 370, 372EM1 9912, 119,428 373
502
INDEX
Infrared vidicon (cont.) optical communications, 374 photoconductive target, 370 semiconductor applications, 374 spectral response, 370, 371 tube construction, 369 see also Vidicon with amorphous a-Si :H target Injection target layers, 379 Input windows fibre-optic, 41 FO 9-15,27-33 fused-silica, 27 Instability of image, LFPCA, 307 Intagliated screens, 315-322 IITs with, 319-322 first generation, 321-322 second generation, 320-321 manufacturing technique, 3 16-3 17 principle, 3 15-3 16 results, 318-319 Integrated analogue signal read-out, 183, 189 Interstellar Medium Absorption Profile Spectrograph (IMAPS), see WICCD Ion events, signal induced background and, 430 IRCAM, 203,204, 205-207 ISIS camera system, 31-33
L La Palma observatory, 427 Large format photon-counting detector, see Lyman Ultraviolet Space Telescope, LFPCA Lasers He-Ne gas-laser, 228 infrared, 374, 376, 377 mode locked Nd :YAG generator, 236 pulses, 21 1, 227 Nd:YAG system, 214 Lenard window, 150 LFPCA (large format photon-counting array), see Lyman Ultraviolet Space Telescope, LFPCA Line-spread function (LSF), 483 Linearity
of image, 101 of light emission, 250-251 of response, electro-optical X-ray detector, with CCD, 278 Low light level applications, see Infrared astronomy Low light level TV electon-bombarded CCDs, see Electron bombarded CCDs image intensifiers and CCDs, 17-25 performance, 4-8 Philips NXA 1011 frame-transfer, 2, 4 resolution, 4-6 signal to noise ratio, 6-8, 13-15 specifications, 3-4 with demagnifying image intensifiers, 1-8
with image intensifiers and CCDs, 9-15 XX1540 sensor, 2,3,4 XX1560 sensor, 6, 8 XX 1570 sensor, 2, 3, 4 Luminescent intensity, 70 Luminous phenomena observation, 22323 1 construction and operation of tube, 224-225 data acquisition and processing, 228230 limitations, 227-228 photoelectron arrival times, 226 photoelectron train, 223-224 pulse stretching calculation, 225 Luminous sensitivity, 29 Lyman Ultraviolet Space Telescope, 297313 design, 298-299 design of detectors, 307-312 detector requirements, 299-300 EUV detector, 312 FUV detector, 312 LFPCA, 300-3 12 development of, 301, 303 DQE, 306-307 Endeavour programme, 303 functional model, 303 image instability, 307 noise rate, 304 operation of, 301 photometric error, 307 quantum efficiency, 306
503
INDEX
required performance, 303 resolution, 305 Lyman astronomy, 298 Lyman mission, 297-298 prime Echelle detector, 31 1-312 prime spectrograph detector, 308 Rowland spectrograph, 299,300,308311 WICCD for, 465-473
M McDonald Observatory, 195 Magnesium fluoride windows, 31 1,415416 Magnetic focus, 465, 467-469 magnetic field uniformity improvement, 470-471 point-spread profiles, 469-473 Mass store unit (MSU), definition, 287 MCP chevron configuration, 41,42 conditioning studies, 44-49 EDOE and resolution, see Resolution intensifier XX1330A, 209 scrubbing process, 44,41-4A wedge and strip image readout, 97-105 detector background, 100 detector sensitivity, 99-100 flat field, 104 image linearity, 101 MCP operation, 97-98 position sensitivity, 102-103 quantum efficiency, 99 resolution, 101-102 MCP-PMTs afterpulses, 108-1 09 cross-talk, 113-1 14 dark counts, 108 dead time, 112-1 13 incorporating 6 pm MCPs, 88-89 maximum counting rate, 1 1 1-1 12 photon correlation measurements, 114-115 pulse height distribution, 110 quantum counting efficiency (QCE), 107-108 triode type, 89-94 ultra-fast IR detectors, 87-96 with S-1 photocathode, 94-96
Measurement precision, electro-optical X-ray detector, with CCD, 281282 Messier 33 galaxy study, 141 Michelson type optical delay configuration, 213 Microchannel plates, see MCP Microdiffraction, 148-1 51 Microstructure of S-1 photoemitting surfaces, see Photocathodes Microthermovision technique, 397-403 AGA 680 Thennovision system, 397 calibration, 397 correction for small objects, 400-40 1 measurement technique, 397-399 noise, 399 planar Gunn diode, 398,401,402,403 sensitivity calibration, 399 temperature distribution, 402 MIL-1-49428 (ER), 406 Modulation transfer function (MTF), 10, 13, 315, 316, 318,483 improvement 8 1-84 polychromatic, 488, 495 stimulation of 79-81 Moire effect 154, 159, 164 Monochromatic sensitivity, 29, 31-32 MTF, see Modulation transfer function (MTF) Multi-channel streak cameras, 21 9 Multi-framing ICC, see Sapphire-Frame Multichannel analyser, 13 Multiplication, avalanche, see Se photoconductive target
N NazKSb photocathodes, 331-338 Cs treated, 336-337 escape length, 336-337 features of, 331-332 optical heterogeneity growth and, 333 photoemission and, 333-335 substrate and, 332-333 photoelectric quantum yield, 334, 336, 337 X-ray diffractograms, 332 NA :YAG generator (Laser), 236 Night-vision goggles, veiling glare in, 407-408
504
INDEX
Nightglow emissions, 197, 199 Noise bilinear CCD array and, 175- I79 coherent seen in rippling, 65 cooled CCD, 130 dark current in EB-CCDs, 168, 170 in IR vidicon, 373 in semiconductor junctions, 363-365 electro-optical X-ray detector, with CCD, 281-282 FAST detector, 289-291 amplifier noise, 289-290, 293 dark current and temperature, 293 dark current, shot noise in, 290,293 dark emission, 290-291 in microthermovision technique, 399 LFPCA, 304 spurious charge, 185 see also Dark noise: Signal to noise ratio Non-destructive testing, 273-274
in CDD camera system, 27, 29, 31, 32 irradiation with electron beam, 235 KBr, 183-184 NaKSb, see NakSb photocathodes opaque alkali halide, 181 polycrystalline CsI/Na 323-330 s-1 basic microstructure and theory, 340-343 microstructure and PQY, 339-345 PQY, 343-344 S-6 in MCP-PMTs, 94-96 $20, 38-39,41,44, 239 see also FOS Digicon tubes SbNaK, for FOC, 416-417 trialkaline antimonide, see Na2KSb photocathodes Photoconductive layers, Se-Te-As, 39 1 Photoconductive target, IR vidicon 370 Photocurrent, vidicon with amorphous a-Si: H target, 365-367 Photodiodes linear arrays, see Bilinear CCD array silicon, array, 269, 271 0 Photoeleciric quantum yield (PQY) NazKSb, Cs photocathodes, 336 Oblique focus detector configurations, Na2KSb photocathodes, 334, 336, 337 190, 192 of S-1 photocathodes, see PhotoObservatoire de Haute-Provence, L’, cathodes 135- 146 Opaque alkali halide photocathodes, 181 Photoelectron charge spreading, 58-60 vertical, 66 Optical communications, IR vidicon and, Photoelectron train, 223-234 374 arrival time of photoelectron, 226 Optical delay configuration, 213 stretching the train, 224 Optical transfer function (OTF), 483,495 see also Luminous phenomena obserOptoliner, 4 vation Oscilloscope, fast, 172 Output electron energy distribution, see Photoelectrons, electron temperature measurement, 228 EDOE Photomultiplier tubes, 87 EMR 541G, 184, 188 P see also MCP-PMTs Photon correlation mesurements, MCPPerturbation theory, 435-437 PMT and 1 1 4 1 I5 Phase transfer function (PTF), 483, 495 Photon counting, 30, 182 Phosphors, for FOC, 417 for astronomy, 41-53 Photocathodes calibrated faint light source, 425curvature of, 477-480 427 demountable, XRST with, 233-238 counting efficiency, 428-429 GaAs, 2,405 dark noise, 51 gold, 210, 214 detector construction, 4 1 4 3
INDEX
505
experimental set up, 44,45 Q gain, 49 lifetime, 50 Quantum counting efficiency (QCE), MCP conditioning studies, 44-49 107-108 performance, 42, 49-51 Quantum efficiency photocathode seal-off, 49 detective, see Detective quantum processing chamber, 43 efficiency (DQE) pulse-height distribution, 427-428 instability, 349-352 residual gas analysis 45-47 LFPCA, 306 resolution, 49-50 MCP with wedge and strip image uniformity, 51 readout, 99 see also CDD-Digicon profile, 348 in atomic physics, 97-105 responsive (RQE), 32, 33, 50, 51 signal induced background, 429432 streak camera with CCD recording, R 22 1 see also MCP, wedge and strip image RADOCON model 258, 500 readout Ramtek image-processing system, 185, Pick-up tubes for HDTV, 389-395 189 diode-mode electron gun, 393 Reading efficiency, bilinear CCD array, H D camera performance, 394 173, 175, 177 HDTV requirements, 389-390 Relative detective quantum efficiencies, performance, 393-394 32, 33 portrayal of motion quality limit, 390 Residual gas analysis 45-47 Se-Te-As photoconductive layer, 391 Resistive gate sensor, 169- 170 THX 898 Primicon, 390-391 twisted arrow pattern deflection Resolution capabilities, 4-6, 12 electrodes, 392 CCD camera system, 30 Picoframe camera, 209-2 17 cooled CCD, 130 UV imaging, 210-214,216 EDOE of MCP and 79-85 X-ray imaging, 214-216 MTF improvement 81-84 PIM-IO5B image-converter tube, 239, MTF stimulation 79-8 I 240 electro-optical X-ray detector, with Pixel, definition, 287 CCD, 280-28 1 Point-spread function (PSF), 483 II/CCD system, 21-22 Position sensitivity, 102-103 LFPCA, 305 PQY see Photoelectric quantum yield low light level TV sensors, 4-6 Princeton University, 183-188, 275 MCP with wedge and strip image readProximity diode, 27 out, 101-112 Proximity focus, 41 spatial see Spatial resolution signal induced background in, 43 1-432 Response Pulse stretching calculation, 225 non-uniformity in FAST detector, Pulsed biasing, 69-78 294-295 current flow and gain mechanisms, 69uniformity, 51, 278-279 71 Responsive quantum efficiency (RQE), current limits, 71 32, 33, 50, 51 results, 73-77 RGS see Resistive gate sensor test procedure, 71-73 Pyridine nucleotide fluorescence, 126- Rotational symmetry deviations, 435456 127
506
INDEX
Scrubbing process, 44,47-48,49 Se photoconductive target, 379-387 avalanche multiplication, 381-383,386 colour camera test, 386 current-voltage characteristics, 380381 dark current, 381 experimental tube charcteristics, 383386 HARP, 387 hole and electron currents, 382 impurity doping, 386 sensitivity, 381 target structure, 379-380 Sedimentation technique, 39 Semiconductors for IR astronomy arrays, 201-202 IR vidicon application, 374 see also Solid state Sensitivity S electro-optical X-ray detector, with CCD, 278 FAST detector, 288-289 Sapphire-Frame, 239-246 Shot noise in dark current, FAST detecblock-diagram, 240 tor 290,293 blocking pulses, 241, 243 Signal charge transfer, 175 control circuit, 242 dynamic performance measurement, Signal induced background, 429-432 24 1 in proximity-focussed intensifiers, 43 1432 principal features, 245 ion events, 430 test chart image, 243, 244 measurement, 430-43 1 SATICON H4286D, 383 SbNaK photocathode, for FOC, 416-417 optical, 429-430 scattered electrons, 430 Schmidt camera, 181, 183-188, 188-189, Signal to noise ratio, 13- 15 190-191, 194, 197, 198 II/CCD system, 18-19,20-21 Schottky-barrier diodes, 202 LLL TV sensors 6-8 Scintillating fibres, see CERN ScinSilicon, amorphous hydrogenated, see tillating Fibre Detector Vidicon with amorphous a-Si :H Scintillation plate, CsI(Na), 247-255 detection efficiency, 251-252 target hygroscopic effect degradation, 253- Silicon photodiode array, 269,271 SNAPSHOT data acquisition mode, 205 254 life test, 253-254 Solid-state light emission and linearity, 250-251 arrays, see Infrared astronomy preparation, 248-250 see also Semiconductors SIT camera and, 254 Solid-state linear X-ray detectors, 269spatial resolution, 252-253 274 Scintillations, 287-288 detection performance, 273-274 detector characteristics, 272 Screens operating conditions, 273 efficiency, 39 silicon photodiode array, 269, 27 1 intagliated, see Intagliated screens
Rotational symmetry deviations (cont.) astigmatism, 444-446 electric field calculation, 440-441 electron beams, mathematical description, 442-444 electron trajectories calculation, 4 4 442 elliptically deformed electrode, 446450,452-454 perturbation theory, 435-437 radial shift or electrode, 450-452 results, 444-454 2-dimensionalpartial differentialequations, 437-440 Rowland spectrograph for Lyman telescope, 299,300,308-3 1 1 XUV sensor, see WICCD, for IMAPS sounding rocket
507
INDEX
Somnoluminescence, 119-1 24 black body radiation and, 123 chemiluminescence and, 123 Spatial distortion electro-oDtical X-ray detector, with CCD, 279 FAST detector, 294 Spatial resolution FAST detector, 295 LFPCA, 350 of scintillation plate, CsI(Na), 252-253 Spectracon Tubes, 148,348-349 Spectroscopy, bilinear CCD array, see Bilinear CCD array Spread and transfer functions, 483-496 basic formulae, 484-485 diode image intensifier, 490 monochromatic spread functions, 485487,492-493 monochromatic transfer functions, 492-493 numerical procedures, 485-488 polychromatic spread functions, 487488,493 polychromatic transfer functions, 488, 495 results and discussion, 489-495 sagittal and tangential orientations of image, 492,493,495 Spurious charge, 185 Sputtering technique, 250, 261 Staircase generating circuit, 241 Star formation, 206 STARE data acquisition mode, 205 STEM see Electron microscopy Streak tubes Dellistrique DS-3 camera, 221 multi-channel, 219 photon-counting, with CCD recording, 221 see olso X-ray streak tubes (XRST) Super-inverters, 4, 8 Swing operation, see Synchro Vision (SV)-CCD Synchro Vision (SV)-CCD, 157-164 colour imaging principle, 161-162 device fabrication, 159-161 image reproduction, 163-164 swing CCD operation, 157-159 Synchronization, 237
Synchrotron radiation, X-ray detector for use with, 275-283
T Targets amorphous hydrogenated Si, see Vidicon with amorphous a-Si:H target amorphous Se see Se photoconductive target Tektronix sampling arrangement, 21 1 Television, high-definition, see Pick-up tubes for HDTV Temperature measurement electron, 228 see also Microthermovision technique TESLA, 359 TH 7832 CDZ, 173-179 TH 9583 solid-state linear detector, 272 THX 898 Primicon, 390-391 TI camera, 183, 194 TRAJECTORIES I program, 475, 477478 TRAJECTORIES 11 program, 475,477 Transfer functions, see Modulation transfer and Spread and transfer functions Triodes GI-41-1 triode, 241 imaging system, 476-477 Twisted arrow pattern deflection electrodes, 392
U UK
Infrared Telescope in Hawaii (UKIRT), 203 Ultra violet EB-CCD cameras for, see Electron bombarded CCD, for far UV Picoframe camera, 210-214, 216 response, 27, 32-33 sensor for Rowland-circle spectrographs, 465-473 Ultrasonic devices, medical applications, 123 Uncertainty in X-ray dose, 281-282 Uniformity, 5 1
508
INDEX
Uniformity (cont.) electro-optical X-ray detector, with CCD, 278-279 V
Veiling glare, 405-412 anti-veiling glare windows, 410-412 black hole, 406, 407 in intensifier, 409-410 in objective lens, 407-408 Victorin radiometer, 258 Video-rate operation, CDD-Digicon, 6266 Vidicon, IR, see Infrared vidicon Vidicon-type pick-up tube, see Pick-up tubes for HDTV Vidicon with amorphous a-Si :H target, 359-367 dark current, 363-365 photocurrent, 365-367 potential distribution, 361-363 properties of single heterojunction, 360-361
W Wedge and strip image readout, 97-105 WICCD, for IMAPS sounding rocket, 465-473 image format, 466467 magnetic focus, 465,467-469,469-473 Wide field astronomical camera, 135-146 Canadian/French camera in Hawaii, 140-144 correction of faults on the slides, 137139 galactic north pole, in depth study, 141-144 improvements and observation procedures, 135- 139 I'Observertoire de Haute-Provence, 139-140 Messier 33 galaxy study, 141 photoemissive surface faults, 136-1 37 weak star sources, 139 Windows anti-veiling glare windows, 410412 beryllium, 248, 250, 262, 264
input, see Input windows Lenard, 150 magnesium fluoride, 3 1 1,415-41 6 out, charge control on, 421-422 windowless intensified charge-coupled device see WICCD Wolter-Schwarzschild Type I1 grazingincidence telescope, 466
X X-ray detectors solid-state, see Solid-state linear X-ray detectors use with synchrotron radiation, see Electro-optical X-ray detector with CCD, see Electro-optical X-ray detector, with CCD X-ray diffractograms, NazKSb photocathodes, 332 X-ray imaging, 252-253 absorption coefficient of CdTe and a-Si :H film, 258 and microscopy, 132-1 33 CdTe/a-Si :H heterojunction use, see CdTe/a-Si :H heterojunction use Picoframe camera, 214-216 rotational symmetry and, see Rotational symmetry deviations see also Scintillation plate, CsI(Na) X-ray streak tubes (XRST) design and electronic properties, 233234 dynamic tests, 236-238 experimental investigation of properties, 234-236 sources of X-rays, 234-235 temporal resolution, 237 with demountable photocathodes, 233-238 X-ray television detector, see FAST detector XX 1080 inverter diode tube, 321 XX 1330A MCP intensifier. 209 Y
YAG scintillator, 152, 153-1 55
