sign in sign up
<<
Home , Counting efficiency, Dynode, NEP Telephone, Photocathode, Photocurrent, Photodetector

Contents

1. lntroduction...... 3 Early development, photoemitter and secondary-emitter development, applications development, photomultiplier and solid-state detectors compared

2. Photomultiplier Design ...... , ...... 10 Photoemission, practical materials, opaque and semitransparent , glass transmission and spectral response, , , time tag in photoemission and secondary emission

3. of Photomultipliers , ...... 26 Electron-optical design considerations, design methods for photomultiplier electron optics, specific photomultiplier electron-optical configurations, configurations

4. Photomultiplier Characteristics ...... , . . 36 Photocathode-related characteristics, gain-related characteristics, and , time effects, pulse counting, scintillation counting, liquid scintillation counting, environmental effects

5. Photomultiplier Applications ...... 80 Summary of selection criteria, applied considerations, mechanical considerations, optical considera- tions, specific photomultiplier applications

Appendix A. Typical Photomultiplier Applications and Selection Guide ...... 119

Appendix B.Glossary of Terms Related to Photomultiplier lubes and Their Applica- I ...... 125

Appendix C.Spectral Response Designation Systems ...... 132

Appendix D. Photometric Units and Photometric-to-Radiant Conversion...... I 37

Appendix E.Spectral Response and Source-Detector Matching...... I 43 Appendix F.Radiant and Sources...... 148 Appendix G. Statistical Theory of Noise in Photomultiplier Tubes...... I 60

Index...... 177

information furnished by BURLE INDUSTRIES, INC. is believed to be accurate and reliable. However, no responsibility of liability is assumed by BURLE for its use, nor for any infringement of patents or other rights of third parties which may result from its use. No license is granted by implication or otherwise under any patent or other rights of BURLE INDUSTRIES, INC. BURLE ® and BURLE INDUSTRIES, INC.® are registered trademarks of BURLE TECHNOLOGIES, INC. Marca(s) Registrada(s). Copyright © 1980 by BURLE TECHNOLOGIES, INC. All rights reserved. No part of this book may be copied or reproduced in any form or by any means -- graphic, electronic, or mechanical or incorporated into any information storage or retrieval system, without the written permission of the copyright owner. Supersedes PMT-62, 8-80 Printed in U.S.A./ 10-89 TP-136

2 1. Introduction

The photomultiplier is a very versatile and without need for additional signal amplifica- sensitive detector of in the tion. Extremely fast time response with rise , visible, and near regions times as short as a fraction of a nanosecond of the . A schemat- provides a measurement capability in special ic diagram of a typical applications that is unmatched by other is given in Fig. 1. The basic radiation radiation detectors. is the photocathode which is located inside a envelope. Photoelectrons are emit- ted and directed by an appropriate electric EARLY DEVELOPMENT field to an or within the The development (history) of the photo- envelope. A number of secondary multiplier is rooted in early studies of secon- are emitted at this dynode for each imping- dary emission. In 1902, Austin and Starke1 ing primary photoelectron. These secondary reported that the metal surfaces impacted by electrons in turn are directed to a second rays emitted a larger number of elec- dynode and so on until a final gain of trons than were incident. The use of secon- perhaps 106 is achieved. The electrons from dary emission as a means for signal the last dynode are collected by an anode amplification was proposed as early as which provides the signal current that is read 1919.2 In 1935, Iams and Salzberg 3 of RCA out. reported on a single-stage photomultiplier. The device consisted of a semicylindrical photocathode, a secondary emitter mounted on the axis, and a collector grid surrounding the secondary emitter. The tube had a gain of about eight. Because of its better frequen- cy response the single-stage photomultiplier was intended for replacement of the gas- filled as a sound pickup for movies. But despite its advantages, it saw only a brief developmental sales activity PHOTOELECTRONS before it became obsolete.

92cs-32288 Multistage Devices Fig. 1 - Schematic representation of a photo- multiplier tube and its operation In 1936, Zworykin, Morton, and Malter, all of RCA4 reported on a multistage photomultiplier. Again, the principal con- For a large number of applications, the templated application was sound-on-film photomultiplier is the most practical or sen- pickup. Their tube used a combination of sitive detector available. The basic reason electrostatic and magnetic fields to direct for the superiority of the photomultiplier is electrons from stage to stage. A photograph the secondary-emission amplification that of a developmental sample is given in Fig. 2. makes it possible for the tube to approach Although the magnetic-type photomultiplier “ideal” device performance limited only by provided high gain, it had several dif- the statistics of photoemission. Amplifica- ficulties. The adjustment of the magnetic tions ranging from 103 to as much as 108 field was very critical, and to change the gain provide output signal levels that are com- by reducing the applied voltage, the patible with auxiliary electronic equipment also had to be adjusted.

3 Photomultiplier Handbook

Another problem was that its rather wide Snyder 8 and by Janes and Gloverg, all of open structure resulted in high dark current RCA. The basic electron-optics of the cir- because of feedback from and cular cage was thus well determined by 1941 developed near the output end of the device. and has not changed to the present time For these reasons, and because of the although improvements have been made in development of electrostatically focused processing, construction, and performance photomultipliers, commercialization did not of the 931A product. follow. The success of the 931 type also resulted from the development of a much improved 10 photocathode, Cs3Sb, reported by Gorlich in 1936. The first experimental photo- multipliers had used a Ag-O-Cs photocath- ode having a typical peak quantum efficien- cy of 0.4% at 800 nm. (The Ag-O-Cs layer Fig. 2 - Magnetic-type multistage photomul- was also used for the .) The new tiplier reported by Zworykin, Morton, and Cs3Sb photocathode had a quantum effi- Malter in 1936. ciency of 12% (higher today) at 400 nm. It was used in the first 931’s, both as a photocathode and as a secondary-emitting material for the dynodes. The design of multistage electrostatically focused photomultipliers required an analysis of the equipotential surfaces be- tween and of the electron trajec- PHOTOEMITTER AND SECONDARY- tories. Before the days of high-speed com- EMITTER DEVELOPMENT puters, this problem was solved by a Photocathode Materials mechanical analogue: a stretched rubber membrane. By placing mechanical models of Much of the development work on the electrodes under the membrane, the photomultiplier tubes has been concerned height of the membrane was controlled and with their physical configuration and the corresponded to the electrical potential of related electron optics. But a very important the electrode. Small balls were then allowed part of the development of photomultiplier to roll from one electrode to the next. The tubes was related to the photocathode and trajectories of the balls were shown to cor- secondary-emission surfaces and their pro- respond to those of the electrons in the cor- cessing. RCA was very fortunate during the responding electrostatic fields. Working 1950’s and 60’s in having on its staff, prob- with the rubber-dam analogue, both J.R. ably the world’s foremost photocathode ex- Pierce5 of Bell Laboratories and J.A. pert, Dr. A.H. Sommer. His treatise on Rajchman6 of RCA devices linear arrays of Photoemissive Materials1 1 continues to pro- electrodes that provided good focusing prop- vide a wealth of information to all erties. Although commerical designs did not photocathode process engineers. result immediately from the linear dynode Sommer explored the properties of array, The Rajchmann design with some numerous photocathode materials-par- modifications eventually was, and still is, ticularly alkali-antimonides. Perhaps his used in photomultipliers-particularly for most noteworthy contribution was the high-gain wide-bandwidth requirements. multialkali photocathode (S-20 spectral response). This photocathode, Na2KSb:Cs, First Commercial Devices is important because of its high sensitivity in The first commercially successful the red and near infrared; the earlier Cs3Sb photomultiplier was the type 931. This tube photocathode spectral response barely ex- had a compact circular array of nine tends through the visible, although it is very dynodes using electrostatic focusing. The sensitive in the blue where most first such arrangement was described by emit. Zworykin and Rajchman.7 Modifications Bialkali photocathodes were also de- were later reported by Rajchmann and veloped by Sommer and have proven to be

4 Introduction better in some applications than the Cs3Sb APPLICATIONS DEVELOPMENT photocathode. Thus, the Na2KSb photo- and cathode has been found to be stable at higher than Cs3Sb and, in addition, Early applications of the photomultiplier has a very low dark (thermal) emission. It were in astronomy and spectroscopy. has been particularly useful in oil-well- Because the effective of logging applications. Another bialkali pho- the photomultiplier was at least ten times tocathode, K2CsSb, is more sensitive than that of photographic film, astronomers were Cs3Sb in the blue and is, therefore, used by quick to realize the photomultiplier tube’s RCA to provide a better match to the advantage. Furthermore, because the output NaI:Tl crystals used in scintillation count- current of the photomultiplier is linear with ing. incident radiation power, the tube could be used directly in photometric and spec- trophotometric astronomy. The type 1P28, a Dynode Materials tube similar to the 931 but having an The first secondary-emission material ultraviolet-transmitting envelope was par- used practically by RCA was the Ag-O-Cs ticularly useful in spectroscopy. The size and surface. But with the development of the shape of the photocathode were suitable for Cs3Sb material for photocathodes, it was the detection and measurement of line spec- found that this material was also an excellent tra and the very wide range of available gain secondary emitter. Other practical secondary proved very useful. 14 emitters developed during the early years of photomultiplier development were MgO:Cs Radar Jammer (often referred to as “-”) A totally unexpected application for the and BeO:Cs (“-”). new photomultiplier tube occurred during In the early 1960’s, R.E. Simon12 while World War II. The development of radar for working at the RCA Laboratories developed detecting and tracking aircraft led to the his revolutionary concept of Negative Elec- simultaneous need for wideband electronic- tron Affinity (NEA). is the noise sources as radar jammers. Although energy required for an electron at the other sources of noise were tried, the conduction-band level to escape to the photomultiplier proved to be most suc- vacuum level. By suitably treating the sur- cessful. The advantage of the tube was its face of a p-type material, the high gain (107) and wide band width (several band levels at the surface can be bent hundred MHz). As a noise source the tube downward so that the effective electron af- was operated with a non-modulated input finity is actually negative. Thermalized elec- light source and with high gain. The output trons in the conduction band are normally photoelectric was repelled by the electron-affinity barrier; the “white” and thus indistinguishable from advantage of the NEA materials is that these natural noise sources. This application of electrons can now escape into the vacuum as photomultiplier tubes resulted in production they approach the surface. In the case of of thousands per month compared with secondary emission, secondary electrons can previous production measured in only hun- be created at greater depths in the material dreds per year. and still escape, thus providing a much greater secondary-emission yield. In the case Scintillation Counting of photoemission, it has been possible to A proliferation of photomultiplier designs achieve extended-red and infrared sen- followed the invention of the scintillation sitivities greater than those obtainable with counter shortly after World War II.15,16 The any other known materials. The first prac- photomultiplier tubes were designed with tical application of the NEA concept was to semitransparent photocathodes deposited on secondary emission. An early paper by an end window which could be coupled Simon and Williams l3 described the theory directly to the . The principal and early experimental results of secondary- scintillator used, NaI doped with , emission yields as high as 130 at 2.5 kV for was discovered by Hofstadter17. Much of GaP:Cs. the development work on photomultiplier

5 Photomultiplier Handbook tubes during this period was reported by depending upon the individual signals from RCA and its competitors in the biannual each of the photomultipliers. Counting is meetings of the Sym- continued until several hundred thousand posium. These symposia were reported fully counts are obtained and the organ in ques- in the IRE (and later the IEEE) Transactions tion is satisfactorily delineated. The location on Nuclear Science beginning with the of each scintillation is represented by a point meeting in Washington, January 1948. The on a cathode-ray-tube presentation. scintillation counter became the most impor- The Computerized Axial Tomographic tant measurement instrument in nuclear (CAT) scanner was introduced to this coun- , nuclear medicine, and radioactive try in 1973. The device uses a pencil or fan- tracer applications of a wide variety. beam of X-rays which rotates around the pa- Headlight Dimmer tient providing X-ray transmission data During the 1950’s, RCA collaborated with from many directions. A scintillator coupled the General Motors Company (Guide-Lamp to a photomultiplier detects the transmitted Division) on a successful headlight dimmer. beam-as an average photomultiplier cur- The photoelectric headlight dimmer-first rent-and a computer stores and computes made available only on Cadillacs and the cross-section density variation of the pa- tient’s torso or skull. The photomultipliers Oldsmobiles-basically used a tube similar 3 to the 931A, but redesigned and tested to the are1/ or /4-inch end-on tubes which auto manufacturer’s particular require- couple to the scintillator, commonly BGO ments. The optical engineering problem was ( germanate). Each unit is equipped to sense the oncoming headlights or tail- with as many as 600 photomultipliers. being followed without responding to PHOTOMULTIPLIERS AND SOLID- street and house lights. Vertical and horizon- STATE DETECTORS COMPARED tal angular sensitivity was designed to match the spread of the high beams of the automo- In some applications either a photomulti- bile. A red filter was installed in the optical plier or solid-state detector could be used. path to provide a better balance between sen- The user may make his choice on the basis of sitivity to oncoming headlights and to tail- factors such as cost, size, or previous ex- lamps being followed. The device achieved a perience. In other applications, the choice remarkable success, probably because of the may be dictated by fundamental properties novelty, and thousands of photomultiplier of the photomultiplier or the solid-state tubes were used. But today, one rarely sees a detector. A discussion follows of some of the headlight dimmer. common applications favoring one or the other detector with reasons for the choice. A Medical Diagnostic Equipment summary presents the principal considera- tions the user must apply in making a choice In recent years two medical applications in an application for which he requires a have used large numbers of photomultiplier . This information should be tubes and have spurred further develop- ments and improvements. The gamma cam- particularly useful to the designer who is not era18 is a sophisticated version of the scin- well acquainted in this field. tillation counter used medically for locating Photomultiplier Features tumors or other biological abnormalities. A The photomultiplier is unique in its ability radioactive isotope combined in a suitable to interface with a scintillation crystal and compound is injected into the blood stream not only count the scintillations but measure or ingested orally by the patient. The their magnitude and time their arrival. Most radioactive material disintegrates and gam- scintillators emit in the blue and near ultra- ma rays are ejected from preferential loca- violet. This spectral output obviously favors tions such as tumors or specific organs. A the photomultiplier having a photocathode large crystal intercepts the gamma rays and with high quantum efficiency in the short scintillates. Behind the crystal are photo- wavelength range. On the other hand a sili- multiplier tubes, perhaps 19, in hexagonal con p-i-n is relatively poor in this part array. The location of the point of scintilla- of the spectrum but does best in the red and tion origin is obtained by an algorithm near infrared. The most important factor,

6 Introduction probably, is the gain of the photomultiplier As a result of increasing concern about en- which permits the measurement of the very vironment, pollution monitoring is becom- small signals from individual scintillations ing another important application for photo- with a good signal-to-noise ratio, limited multiplier tubes. For example, in the moni- primarily by the statistics of the number of toring of NOx the gas sample is mixed with photoelectrons per pulse. Finally, the short O 3 in a reaction chamber. A chemilum- rise time of the photomultiplier using fast inescence results which is measured using a scintillators permits time-of-flight measure- near-infrared-pass filter and a photomulti- ments to be made in . plier having an S-20 spectral response. Al- though the radiation level is very low, NO Although the CAT scanner equipment can be detected down to a level of 0.1 ppm. also uses photomultipier tubes to detect the The advantage of the photomultiplier in this scintillations in bismuth germanate (BGO) application is again the high gain and good crystals, the situation is somewhat different signal-to-noise ratio (the photomultiplier is from the scintillation counting applications cooled to 0°C to reduce dark-current noise) discussed above. In the CAT scanner the even though the radiation spectrum is ob- X-rays produce a broad band of pulse served near the threshold of the S-20 spectral heights and no attempt is made to single out range. and detect single scintillation events. The In another pollution-monitoring applica- photomultiplier is used in an analog mode to tion, SO2 is detected down to a level of 0.002 detect the level of radiation incident on the ppm. Here, the sample containing SO2 is ir- crystal. In the CAT scan operation the radiated with ultraviolet and the excited SO2 typical machine scans the patient in a few molecules fluoresce with blue radiation that seconds and the level of from the is detected with a combination of a narrow- crystal onto the photomultiplier is relatively band filter and photomultiplier. Very weak high so that only a relatively low gain photo- signals are detected and again it is the high multiplier is required. Furthermore, the gain, good signal-to-noise ratio and, in addi- speed of response requirement for the tion, good blue sensitivity which makes the photomultiplier is relatively modest-per- detection and measurement of small contam- haps a few hundred microseconds. Still, the inations of SO2 possible. principal advantage of using a photomulti- Spectroscopy is one of the very early ap- plier in this application for the detection of plications for photomultipliers. The wide the radiant signal is its good signal-to-noise range of radiation levels encountered is ratio. This ratio is very important to the pa- readily handled by the approximately loga- tient because a reduction in its signal-to- rithmic gain variation of the photomultiplier noise ratio would have to be made up for with voltage. At very low signal levels, the with an increased X-ray dose. Nevertheless, signal-to-noise capability of the photomulti- there is interest and development activity plier is essential. Because photomultiplier aimed at replacing the photomultiplier with spectral response (with quartz or ultraviolet- p-i-n detectors. Two factors could transmitting-glass windows) covers the range favor the alternate use of a silicon cell: (1) a from ultraviolet to near infrared, the better scintillator (BGO is almost an order of photomultiplier is the logical choice for spec- magnitude less sensitive than NaI:Tl; (2) a troscopic applications, except in the infrared faster scanning machine (a very desirable region of the spectrum. technological advance because is would minimize effects of body motions). Both of Photocell* Features these factors would result in a larger Because of their small size and low cost, and could bring the signal level CdSe and CdS type photocells are the logical for the silicon detector to the point where the selection for applications such as automatic fundamental signal-to-noise ratio from the exposure control in photographic cameras or X-ray source would not be degraded. Such various inspection and counting require- developments may be anticipated because ments. *“Photocell” is used here to indicate a photosensitive the silicon detector would also have the ad- device in which the charge transport takes place through vantage of smaller size and perhaps lower a solid as compared with “phototube” in which the cost. charge transport is through a vacuum.

7 Photomultiplier Handbook

Many p-i-n silicon cells are used in com- diode can be of the order of 100, but the sen- bination with lasers or LED’s (light emitting sitive area is small-about 0.5 square ). Here, one of the principal advantages millimeter. of the silicon cell is its good response in the Sensitive Area. Photomultiplier tubes are near infrared out to 1100 nm. In combina- made in a variety of sizes so that many dif- tion with the Nd:YAG laser emitting at 1060 ferent optical configurations can be accom- nm, the silicon cell is used widely in laser modated. The largest photocathode area ranging and laser tracking. A similar ap- available in commercial RCA photomulti- plication utilizes an LED emitting near 900 plier tubes has a nominal diameter of 5 in- nm with a silicon cell for automatic ranging ches and a minimum useful area of 97 square 1 for special camera equipment. Size and in- centimeters. By way of contrast, the /2-inch frared sensitivity are again the important side-on photomultiplier has a projected pho- qualifications. tocathode area of 0.14 square centimeter. A rapidly growing application for photo- Silicon p-i-n diodes are available with sen- cells is for fiber-optic communication sitive areas generally not larger than 1 square systems. LED’s are coupled to the fibers and centimeter; and avalanche silicon cells, 0.005 the detector may be a p-i-n diode or, for a square centimeter. In many applications, a better signal-to-noise ratio, a silicon ava- fairly large area is required, e.g., coupling to lanche diode. The qualifying attributes for a cathode-ray tube or a large scintillator. the choice of detector are size, near infrared This requirement generally indicates the use sensitivity, adequate speed of response, and of a photomultiplier tube. Silicon cells are at good signal-to-noise ratio. an advantage when the source is small for Smoke detectors now use large numbers of direct coupling or for lens imaging. LED’s and p-i-n silicon cells. Again size, . Photomultipliers are gener- cost, and infrared sensitivity are the impor- ally not rated for operation at temperatures tant qualifications. higher than 75° C. Exceptions are photomul- Characteristics Comparison Summary tipliers having a Na2KSb photocathode. This Spectral Response. Photomultipliers can bi-alkali photocathode can tolerate temper- be obtained with good spectral sensitivity in atures up to 150° C or even higher for short the range 200 to 900 nm. Silicon cells have cycles. In oil- measurements this rather poor blue sensitivity, but are excellent consideration is important. Photocathode out to 1100 nm. In general, then, the photo- sensitivities and gain change very little with multiplier is to be preferred for applications temperature, but dark current does increase involving the shorter wavelengths, although rapidly. Dark currents at room temperature other factors may override this considera- are of the order of 10 - l5 ampere at the pho- tion. tocathode and double about every 10° C. Speed of Response. If very fast response is Silicon cells are rated from - 50 to 80° C. required, the photomultiplier is usually the Sensitivities are also relatively independent best choice of a detector. Photomultipliers of temperature. But dark current which may are available with rise times (10 to 90%) of 1 be 10-7 ampere at room temperature, also or 2 nanoseconds using a 50-ohm load. The tends to double about every 10° C. inherent rise time of silicon cells may be in Signal-to-Noise Ratio. At very low light the range 10 to 20 nanoseconds, depending levels, the limitation to detection and upon the area of the cell. However, because measurement is generally the signal-to-noise of the cell’s , the effective rise ratio. One way of describing the limit to time is much longer depending upon the detection is to state the Equivalent Noise In- choice of load resistance. For example, with put (ENI) or the Noise Equivalent Power a 1-megohm load resistance, the rise time (NEP). The NEP is the power level into the may be of the order of 20 microseconds. A device which provides a signal just equal to fairly large load resistance must be chosen to the noise. Most often the bandwidth is maintain good signal-to-noise characteristics specified as 1 hertz and the wavelength of the for the silicon cell. Silicon avalanche photo- measurement is at the peak of the spectral diodes can have rise times as short as 2 responsivity. ENI is the same type of specifi- nanoseconds. Gain for an avalanche photo- cation except the unit instead of power may

8 Introduction

be luminous . 2. J. Slepian, U.S. Patent 1, 450, 265, For a photomultiplier such as one used for April 3, 1923 (Filed 1919). spectroscopy, the NEP at room temperature 3. H. E. Iams and B. Salzberg, “The at 400 nm is about 7 x 10-16 watts, or the secondary emission phototube,” Proc. IRE, EN1 is about 7 x 10 - l3 lumens. Both Vol. 23, pp. 55-64 (1935).

specifications are for a l-hertz ban dwidth. 4. V.K. Zworykin, G.A. Morton, and L. For a p-i-n silicon photocell, the NEP at 900 Malter,"The secondary-emission multipli - nanometers may be of the order of 2 x er-a new electronic device,” Proc. IRE, 10- watts, or the EN1 of 1.5 x 10- l1 Vol. 24, pp. 351-375 (1936). lumens. Both values are for a l-hertz band- 5. J.R. Pierce, “Electron-multiplier width. Thus, the photomultiplier is clearly design,”Bell Lab. Record, Vol. 16, pp. superior in this category. Also it should be 305-309 (1938). pointed out that the silicon diode must be 6. J.A. Rajchman,“Le courant residue1 coupled into a load resistance of about 5 dans les multiplicateurs d’electrons elec- megohms in order to avoid noise domination trostatiques,”These L’Ecole Polytechnique from the coupling . Unfortunately, Federale (Zurich, 1938). this large resistance then increases the effec- 7. V.K. Zworykin and J.A. Rajchman, tive rise time of the silicon device to about “The electrostatic , Proc. 100 microseconds. The NEP of a silicon ava- IRE, Vol. 27, pp. 558-566 (1939). lanche is about 10-14 watt at 8. J.A. Rajchman and R.L. Snyder, “An 900 nanometers or the ENI is 8 x 10 - l3 electrostatically focused multiplier lumens, both for a l-Hz bandwidth. The phototube,”Electronics, Vol. 13, p. 20 lumen in these descriptions is that from a (1940). source operating at 2856 K color 9. R.B. Janes, and A.M. Glover, “Recent temperature.Peak emission for such a developments in ,” RCA Review, source is near 1000 nm and thus closely Vol. 6, pp. 43-54 (1941). Also, A.M. Glover, matches the spectral peak of the silicon “A review of the development of sensitive devices. phototubes,” Proc. IRE, Vol. 29, pp. Gain. A photomultiplier can have a gain 413-423 (1941). factor, by which the fundamental photo- 10. P. Gorlich, “Uber zusammengesetzte, cathode signal is multiplied, of from 103 to durchsichtige Photokathoden,” 2. Physik, 108. Silicon avalanche have a Vol. 101, p. 335 (1936). gain of about 100. Silicon p-i-n diodes have 11. A.H. Sommer, Photoemissive no gain. The high gain of the photomulti- materials, John Wiley and Sons; 1968. plier frequently eliminates the need of 12. R.E. Simon, Research in electron special , and its range of gain con- emission from , Quarterly trolled by the applied voltage provides flex- Report, Contract DA 36-039-AMC-02221 ibility in operation. (E) (1963). Stability. Photomultiplier tubes are not 13. R.E. Simon and B.F. Williams, noted for great stability although for low “Secondary-electron emission,” IEEE anode currents and careful operation they Trans. Nucl. Sci., Vol. NS-15, pp. 166-170 are satisfactory. When the light level is (1968). reasonably high, however, the very good 14. M.H. Sweet,“Logarithmic photomul- stability of the silicon p-i-n cell is a con- tiplier tube ,” JOSA, Vol. 37, p. siderable advantage. The silicon cell makes a 432 (1947). particularly good reference device for this 15. H. Kallmann, Natur u Technik (July reason. In fact, the National Bureau of Stan- 1947). dards has been conducting special calibra- 16. J.W. Coltman and F.H. Marshall, “A tion transfer studies using p-i-n silicon photomultiplier radiation detector,” Phys. diodes. Rev., Vol. 72, p. 582 (1947). 17. R. Hofstadter,“Alkali halide scintilla- REFERENCES tion counters,”Phys. Rev., Vol. 74, p. 100 1. H. Bruining, Physics and applications (1948). of secondary electron emission, (McGraw- 18. H.O. Anger,“Scintillation camera”, Hill Book Co., Inc.; 1954). Rev. Sci. Instr., Vol. 29, pp. 27-33 (1958).

9 2. Photomultiplier Design

PHOTOEMISSION In the energy diagram for a metal shown The earliest observation of a photoelectric in Fig. 3, the represents the effect was made by Becquerel in 1839. He energy which must be given to an electron at found that when one of a pair of electrodes the top of the energy distribution to raise it in an electrolyte was illuminated, a voltage to the level of the potential barrier at the or current resulted. During the latter part of metal-vacuum interface. the 19th century, the observation of a in led to the development of selenium and cuprous oxide METAL photovoltaic cells. The emission of electrons resulting from the action of light on a photoemissive sur- face was a later development. Hertz dis- covered the photoemission phenomenon in 1887, and in 1888 Hallwachs measured the photocurrent from a plate subjected to FERMI ultraviolet radiation. In 1890, Elster and ENERGY Geitel produced a forerunner of the vacuum phototube which consisted of an evacuated glass bulb containing an and an auxiliary electrode used to collect the 92CS - 32289 negative electrical carriers (photoelectrons) emitted by the action of light on the alkali Fig. 3 - Energy mode/ for a metal showing metal. the relationship of the work function and the . Basic Photoelectric Theory The modern concept of photoelectricity stems from Einstein’s pioneer work for According to the quantum theory, only which he received the . The one electron can occupy a particular quan- essence of Einstein’s work is the following tum state of an atom. In a single atom, these equation for determining the maximum states are separated in distinct “shells”; nor- kinetic energy E of an emitted photoelec- mally only the lower energy states are filled. tron: In an agglomeration of atoms, these states are modified by interaction with neighboring (1) atoms, particularly for the outermost elec- Eq. (1) shows that the maximum energy of trons of the atom. As a result, the outer the emitted photoelectron mv2/2 is propor- energy levels tend to overlap and produce a tional to the energy of the light quanta hv continuous band of possible energy levels, as shown in Fig. 3. must be given to an electron to allow it to The diagram shown in Fig. 3 is for a escape the surface of a metal. For each temperature of absolute zero; all lower metal, the is character- energy levels are filled. As the temperature is increased, some of the electrons absorb ther- ed in electron-. mal energy which permits them to occupy

10 Photomultiplier Design scattered states above the maximum level for absolute zero. The energy distribution of electrons in a particular metal is shown in Fig. 4 for several different temperatures. At absolute zero, all the lower states are oc- cupied up to the Fermi level. At higher temperatures, there is some excitation to up- per levels. The electron density at a par- ticular temperature is described by the Fermi-Dirac energy-distribution function, which indicates the probability of occupa- tion f for a quantum state having energy E: 1 f = (2) fig. 4 - Energy distribution of conduction electrons in at temperatures of 0, When E is equal to Ef, the value of f is 1/2. 200, and 1033 degrees Kelvin based on It is customary to refer to the energy of level elementary Sommerfeld theory. (ref. 19) Ef, for which there is a 50-per-cent prob- ability of occupancy, as the Fermi level. At Work Function and Spectral Response absolute zero, the Fermi level corresponds to the top of the filled energy distribution. Measurement of the work function and If the energy derived from the radiant spectral response for clean metal surfaces energy is just sufficient to eject an electron at has been of considerable importance in the the Fermi level, the following relation exists: development of photoelectric theory. Work functions for pure metals are in the (3) range 2 to 5 electron-volts. (See A.H. Som- mer, Ref. 11, Table 3.) Fig. 5 shows spectral-response curves for radiation, is related to the long-wavelength the alkali metals. The curves indicate a regular progression of the wavelength for maximum response with atomic number. The most red-sensitive of these metals is (4) cesium, which is widely used in the activa- The relationship may be rewritten to relate tion of most commercial phototubes. the long-wavelength limit to the work func- The energy distribution of emitted photo- electrons has been measured for a number of tion, as follows: metals and photosurfaces. Typical results are shown in Fig. 6 for a potassium film of (5) 20 molecular layers on a base of silver.20 The maximum emission energy corresponds to Because some of the electrons occupy that predicted by the Einstein photoelectric states slightly higher than the Fermi level, as equation. shown in Fig. 4, excitation of these electrons produces an extended response at the red Quantum Efficiency threshold of the spectral-response character- Because a quantum of radiation is neces- istic. As a result, there is no abrupt red sary to release an electron, the photoelectric threshold at normal temperatures, and the current is proportional to the intensity of the true work function cannot be obtained in a radiation. This first law of photoelectricity simple manner from the spectral-response has been verified experimentally over a wide measurement. However, a universal func- range of light intensities. For most materials, tion devised by Fowler can be used to predict the quantum efficiency is very low; on the the shape of the spectral-response curve near best sensitized commercial photosurfaces, the threshold; the work function can then be the maximum yield reported is as high as one calculated from these data. electron for three light quanta.

11 Photomultiplier Handbook

WAVELENGTH-NANOMETERS 92CS-32291 Fig. 5 - Spectral-response characteristics for the alkali metals showing regular pro- gression in the order of the periodic table. (ref. 19)

An ideal photocathode has a quantum ef- from to electron, (2) motion of the ficiency of 100 per cent; i.e., every incident electron toward the material-vacuum inter- photon releases one photoelectron from the face, and (3) escape of the electron over the material into the vacuum. All practical potential barrier at the surface into the photoemitters have quantum efficiencies vacuum. below 100 per cent. To obtain a qualitative Energy losses occur in each of these steps. understanding of the variations in quantum In the first step, only the absorbed portion efficiency for different materials and for dif- of the incident light is effective and thus ferent wavelengths or photon , it is losses by transmission and reflection reduce useful to consider photoemission as a pro- the quantum efficiency. In the second step, cess involving three steps: (1) absorption of a the photoelectrons may lose energy by col- photon resulting in the transfer of energy lision with other electrons (electron scat- tering) or with the lattice ( ). Finally, the potential barrier at the surface prevents the escape of some elec- trons. Metallic and Semiconductor Materials The energy losses described vary from material to material, but a major difference between metallic and semiconducting materials makes separate consideration of each of these two groups useful. In metals, a large fraction of the incident visible light is reflected and thus lost to the photoemission process. Further losses occur as the photo- electrons rapidly lose energy in collisions with the large number of electrons in the metal through electron-electron scattering. As a result, the escape depth, the distance from the surface from which electrons can reach the surface with sufficient energy to Fig. 6 - Energy distribution of photoelec- overcome the surface barrier, is small, and is trons from a potassium film. (ref. 20) typically a few nanometers. Finally, the

12 Photomultiplier Design work function of most metals is greater than ty. At temperatures higher than 0 K, some three electron-volts, so that visible electrons in the valence band have sufficient which have energies less than three electron- energy to be raised to the conduction band, volts are prevented from producing electron and these electrons, as well as the holes in the emission. Only a few metals, particularly the valence band created by the loss of electrons, alkali ones, have work-function values low produce electrical conductivity. Because the enough to make them sensitive to visible number of electrons raised to the conduction light. Because of the large energy losses in band increases with temperature, the con- absorption of the photon and in the motion ductivity of semiconductors also increases of the photoelectron toward the vacuum (the with temperature. Light can be absorbed by first and second steps described above), even valence-band electrons only if the energy of the alkali metals exhibit very low quantum the photon is at least equal to the band-gap efficiency in the visible region, usually below energy EG. If, as a result of light absorption, 0.1 per cent (one electron per 1000 incident electrons are raised from the valence band photons)l As expected, higher quantum effi- into the conduction band, ciencies are obtained with higher photon is achieved. For photoemission, an electron energies. For 12-electron- photons, in the conduction band must have energy quantum yields as high as 10 per cent have greater than the electron affinity EA. The been reported. additional energy EA is needed to overcome the forces that bind the electron to the solid, or, in other words, to convert a “free” elec- tron within the material into a free electron in the vacuum. Thus, in terms of the model of Fig. 7, radiant energy can convert an elec- tron into an internal photoelectron (photo- conductivity) if the photon energy exceeds EG and into an external photoelectron (photoemission) if the photon energy ex- ceeds (EG + EA). As a result, photons with total energies Ep less than (EG + EA) cannot produce photoemission. The following statements can therefore be made concerning photoemission in semicon- Fig. 7 - Simplified semiconductor energy ductors. First, light absorption is efficient if band model. the photon energy exceeds EG. Second, energy loss by electron-electron scattering is The concept of the energy-band models low because very few free electrons are pres- that describe semiconductor photoemitters is ent; thus, energy loss by phonon scattering is illustrated in its simplest form in Fig. 7. Elec- the predominant loss mechanism. The trons can have energy values only within well escape depth in semiconductors is therefore defined energy bands which are separated by much greater than in metals, typically of the forbidden-band gaps. At 0 K, the electrons order of tens of nanometers. Third, the of highest energy are in the so-called valence threshold wavelength, which is determined band and are separated from the empty con- by the work function in metals, is given by duction band by the bandgap energy EG. the value of (EG + EA) in semiconductors. The probability that a given energy level may Synthesis of materials with values of be occupied by an electron is described by EG + EA) below 2 electron-volts has Fermi-Dirac statistics and depends primarily demonstrated that threshold wavelengths on the difference in energy between the level longer than those of any metal can be ob- under consideration and Fermi level. As a tained in a semiconductor. Semiconductors, first approximation, it may be said that any therefore, are superior to metals in all three energy levels which are below the Fermi level steps of the photoemissive process: they ab- will be filled with electrons, and any levels sorb a much higher fraction of the incident which are above the Fermi level will be emp- light, photoelectrons can escape from a

13 Photomultiplier Handbook greater distance from the vacuum interface, the conditions of Fig. 8 has greatly increased and the threshold wavelengths can be made escape depth. Under such circumstances, the longer than those of a metal. Thus, it is not photosensitivity is significantly enhanced. surprising that all photoemitters of practical Substantial response is observed even for importance are semiconducting materials. photons with energies close to that of the Negative-Electron-Affinity Materials 20a band gap where the absorption is weak. Effi- In recent years, remarkable improvements cient photoemission in this case results only in the photoemission from semiconductors because of the greater escape depth. have been obtained through deliberate The reduction of the electron affinity is modification of the energy-band structure. accomplished through two steps. First, the The approach has been to reduce the elec- semiconductor is made strongly p-type by tron affinity, EA,and thus to permit the the addition of the proper “” agent. escape of electrons which have been excited For example, if arsenide is the host into the conduction band at greater depths material, zinc may be incorporated into the within the material. Indeed, if the electron crystal lattice to a concentration of perhaps affinity is made less than zero (the vacuum 1,000 parts per million. The zinc produces level lower than the bottom of the conduc- isolated energy states within the forbidden tion band, a condition described as gap, near the top of the valence band, which “negative electron affinity” and illustrated are normally empty, but which will accept in Fig. 8), the escape depth may be as much electrons under the proper circumstances. The p-doped material has its Fermi level just above the top of the valence band. The se- cond step is to apply to a semiconductor a surface film of an electropositive material such as cesium. Each cesium atom becomes ionized through loss of an electron to a p-type energy level near the surface of the semiconductor, and is held to the surface by electrostatic attraction. 92CS-32294 The changes which result in the energy- Fig. 8 - Semiconductor energy-band model band structure are two-fold. In the first showing negative electron affinity. place, the acceptance of electrons by the p-type impurity levels is accompanied by a as 100 times greater than for the normal downward bending of the energy bands. material. The escape depth of a photoelec- This bending can be understood by observ- tron is limited by the energy loss suffered in ing that a filled state must be, in general, phonon scattering. Within a certain period below the Fermi level; the whole structure of time, of the order of 10-12 second, the near the surface is bent downward to ac- electron energy drops from a level above the complish this result. In the second place, the vacuum level to the bottom of the conduc- potential difference between the charged tion band from which it is not able to escape electropositive layer (cesium) and the body into the vacuum. On the other hand, the charge (filled zinc levels) results in a further electron can stay in the conduction band in depression of the vacuum level as a result of the order of 10-10 second without further a dipole moment right at the surface. loss of energy, i.e.,without dropping into Another way to describe the reduction of the valence band. If the vacuum level is the electron affinity is to consider the surface below the bottom of the conduction band, of the semiconductor as a . The the electron will be in an energy state from charge on one side of the capacitor is which it can escape into the vacuum for a represented by the surface layer of cesium period of time that is approximately 100 ions; the other charge is represented by the times longer than if an energy above the bot- region of filled acceptor levels. The reduc- tom of the conduction band is required for tion in the electron affinity is exactly equal escape, as in the materials represented by to the potential difference developed across Fig. 7. Therefore, a material conforming to the capacitor.

14 Photomultiplier Design

In a more rigorous analysis, the amount The photocathodes most commonly used by which the energy bands are bent is found inphotomultipliers are cesium- to be approximately equal to the band-gap, (Cs3Sb), multialkali or trialkali (Na2KSb: and the vacuum level is lowered until the ab- Cs),* and bialkali (K2CsSb). Another sorption level of the electropositive material bialkali photocathode (Na2KSb) is par- is essentially at the top of the valence band. ticularly useful at higher operating tempera- tures because of its stability. Recently, the PRACTICAL PHOTOCATHODE -cesium-antimony (probably Rb2 MATERIALS Cs Sb) photocathode has been introduced Research on commercially useful photo- because of its favorable blue sensitivity. emitters has been directed primarily toward Typical spectral response curves for these developing devices sensitive to visible radia- materials are shown in Figs. 9 and 10. Addi- tion. The first important commercial photo- tional information about these and other surface was silver-oxygen-cesium. This sur- photocathodes of practical importance is face, which provides a spectral response shown in Table I. designated S-l, is sensitive throughout the entire visible spectrum and into the infrared. Although it has rather low sensitivity and high dark emission, the good response in the red and near-infrared still recommends its use in special applications although other photocathodes are more generally used in photomultipliers today.

300 400 500 600 700 WAVELENGTH-NANOMETERS

92CM-32296 Fig. 10 - Typical spectral-response curves for various photocathodes useful in scintilla- tion counting applications. The variation in the cutoff at the low end is due to the use of different envelope materials.

The long-wavelength response of the mul- 200 400 600 800 1000 tialkali photocathode has been extended by WAVELENGTH - NANOMETERS 92CM-32295 processing changes including the use of an Fig. 9 - Typical spectra/-response curves, increased photocathode-film thickness at the with 0080 lime-glass window for (a) silver- expense of the short-wavelength response. oxygen-cesium (Ag-O-Cs), (b) cesium- Fig. 11 shows two typical spectral-response antimony (Cs3Sb), (c) multialkali or trialkali curves of the ERMA types (Extended Red (Na2KSb:Cs Multi-Alkali) II and III compared with the *The terminology “:Cs” indicates trace quantities of S-20 response of the conventionally pro- the element. cessed multialkali photocathode.

15 Photomultiplier Handbook

Table I Nominal Composition and Characteristics of Various Photocathodes

Conversion Luminous Wave- Dark JEDEC Factorb Respon- length of Respon- Quantum Emission Nominal Photo- Envelopea Response (lumen/ sivitv Maximum sivitv Efficiency at 25° C Composition cathode Material Designation watt) (uA/lumen) Response (mA/watt) (percent) (fA/cm2)

Cs3Sb 0 0080 S-4 950 40 380 38 12 0.2

Cs3Sb 0 9741 S-5 1244 40 340 50 18 0.3 Cs3Sb S 0080 S-11 857 70 400 60 19 3

Na2KSb S 7056 S-24 1250 40 380 50 16 .0003 Na2KSb S Sapphire 1400 43 380 60 19 .0003

K2CsSb 0 0080 938 60 380 56 18 K2CsSb 0 7740 1083 60 400 65 20 .02 K2CsSb S B270 1111 90 380 100 33 .02 K2CsSb S 0080 1120 80 380 90 29 .02 K2CsSb S 7740 1140 71 420 82 24 K2CsSb S 9741 1240 56 380 70 23 .02 Rb2CsSb S 0080 948 100 420 95 28 .08 Na2KSb:Cs 0 9741 510 75 380 38 12 Na2KSb:Cs S 7740 ERMA IIIc 160 180 575 29 6 .3 Na2KSb:Cs S 0080 ERMA IIc 250 200 550 50 11 Na2KSb:Cs S 0080 S-20 480 135 390 65 21 0.4 Na2KSb:Cs Sd 0080 230 300 530 70 16 1.2

Na2KSb:Cs S 7056 432 117 420 51 15 GaAs: Cs-0 O e 9741 115 720 800 80 12 92.

a Numbers refer to the following glasses: c A BURLE designation for “Extended-Red Multial- 0080 - Corning Lime Glass kali.” 9741 - Corning Ultraviolet Transmitting Glass d Reflecting substrate. 7056 - Coming 7740 - Corning Pyrex Glass e Single crystal. B270 - Schott BK270 0 = Opaque bThese conversion factors are the ratio of the radiant responsivity at the peak of the spectral response char- S = Semitransparent acteristic in amperes per watt to the luminous respon- sivity in amperes per lumen for a tungsten lamp oper- ated at a color temperature of 2856 K.

16 Photomultiplier Design

The negative-electron-affinity materials for 1060 nm, the wavelength of the Nd:YAG described earlier are used in both opaque laser. For comparison of sensitivities at this and semitransparent photocathodes. Spec- wavelength, the spectral response of the Ag- tral response curves for GaAs:Cs-O and 0-Cs photocathode is also shown. The InGaAs:Cs-O are shown in Fig. 12. There InGaAs:Cs-O photocathode has the higher has been considerable interest in detectors responsivity at 1060 nm. The NEA photo- are generally fairly small compared with the large semitransparent photocath- odes used for scintillation counting. Stabili- ty, especially for the longer-wave-length NEA photocathodes, is a problem unless the photomultiplier output current are kept low.

OPAQUE AND SEMITRANSPARENT PHOTOCATHODES Photocathodes may be classified as opaque or semitransparent. In the opaque photocathode, the light is incident on a thick photoemissive material and the electrons are emitted from the same side as that struck by the radiant energy. In the second type, the semitransparent photocathode, the photo- emissive material is deposited on a transparent medium so that the electrons are emitted from the side of the photocathode opposite the incident radiation. Because of the limited escape depth of photoelectrons, the thickness of the semitransparent photocathode film is

80

0 200 400 600 800 1000 1200 WAVELENGTH -NANOMETERS 92CS-32298 92cs -32299 Fig. 12 - Spectra/ response characteristics Fig. 13 - Spectral absorptance of the Ag-O- Cs and the K2CsSb semitransparent photo- pared with the S-1 characteristic (Ag-O-Cs). cathodes.

17 Photomultiplier Handbook critical. If the film is too thick, much of the The radiant spectral flux absorption of a incident radiant energy is absorbed at a semitransparent photocathode varies with distance from the vacuum interface greater wavelength as illustrated in Figs. 13 and 14. than the escape depth; if the film is too thin, The ordinate in Fig. 13 is absorptance which much of the incident radiant energy is lost by is defined as the ratio of the radiant flux ab- transmission. sorbed by the layer to that incident upon it. (Absorptance plus reflectance plus transmit- tance add to unity.) The data shown in Fig. 14 are spectral absorption coefficients. The

of the photocathode layer is given by

(6)

that the data in Fig. 14 are given in units of micrometers - l so that in Eq. 6, d must be given in micrometers. Typical thickness of a Na2KSb:Cs photocathode is about 0.030 Thus, at 400 nm the photocathode ab- sorbs 87% of the flux which is not reflected. 0 200 300 400 500 600 700 800 900 The spectral response of semitransparent WAVELENGTH- NANOMETERS photocathodes can be controlled to some ex- 92CS-32300 tent by thickness variation. With increasing Fig. 14 - Spectral absorption coefficient data thickness in the case of alkali antimonides, blue response decreases and red response in- tocathodes. creases.

WAVELENGTH - NANOMETERS 92CS-32301

Fig. 15 - Ultraviolet transmittance cut off of various glasses and crystals used in photomultiplier photocathode windows. Data are all for 1 mm thickness.

18 Photomultiplier Design

GLASS TRANSMISSION AND tion. Photomultiplier tubes made with fused SPECTRAL RESPONSE quartz windows are useful in Cerenkov Although photocathode spectral response counting applications where the spectral is determined primarily by the nature of the energy distribution increases with decreasing photocathode surface, especially in the visi- ble and long wavelength cut-off regions, the terval. is also a useful material short-wave length cut-off characteristic of in liquid scintillation counting because of its 40 all photocathodes is determined by the minimum contamination with K which can transmission of the window to the photo- cause unwanted background counts. cathode. Utraviolet cut-off characteristics Less expensive than synthetic fused quartz of a number of glasses or crystals which have is Corning 9741 glass, which is frequently been used in photomultiplier fabrication are used in photomultipliers designed for the shown in Fig. 15. near ultraviolet. It is a Kovar-sealing glass The data presented in this figure are all for but has the disadvantage of possible 1 mm thickness. The data also include losses weathering over long periods of exposure to from reflection. For most glasses with an in- the atmosphere. dex of refraction of about 1.5, the reflection Another glass for the near ultraviolet is loss is about 4% at each surface. The loss is Corning 9823, which seals to 0120 glass higher in high index-of-refraction material and which can be used with Dumet metal such as sapphire. Although some photo- . This glass is somewhat inferior to 9741 multipliers, especially those of small size at the shortest wavelengths, but is better for may have window thickness of 1 mm or less, wavelengths longer than 240 nm. larger face plates are generally thicker in Glass type 7056 is selected for its good op- order to provide adequate strength. The tical quality. It is a hard glass which seals to transmittance varies with thickness accord- 7052 and Kovar. Pyrex type 7740 is selected ing to the following relationship primarily for its low content of 40K and is used in liquid scintillation-counting applica- (7) tions. Lime glass is the least expensive; it is a soft glass which seals to lead glass, type 0120. where k is a factor (approximately 0.92 for Corning 9025 is a special non-browning most glasses) dependent upon the surface glass. It is doped with and resists darkening from exposure to ionizing radia- tion, and t is the thickness. tion. One application is in satellites which The window extending the furthest into must pass through space regions of high- the ultraviolet is LiF. A few tubes are made intensity . with this material but, because its fabrica- tion is difficult, LiF face plates are only used for special applications of spectroscopy. THERMIONIC EMISSION Transmission extends to wavelengths shorter Current flows in the anode circuit of a than the Lyman-alpha limit of 121.5 nm. photomultiplier tube even when it is Sapphire windows (ultraviolet grade) are operated in complete darkness. The dc com- good down to about 150 nm and are easier to ponent of this current is called the anode use than LiF. Sapphire can be sealed to dark current, or simply the dark current. Kovar by a metalizing and brazing tech- This current and its resulting noise compo- nique. Special photomultiplier tubes having nent usually limit the lower level of sapphire windows were used in the HEAO photomultiplier light detection. As a result, program (High Energy Astronomical Obser- the anode dark-current value is nearly vatory). always given as part of the data for any tube. Except for the extended short wavelength There are several sources of dark current cutoff of sapphire, Suprasil, an ultraviolet in a photomultiplier tube: ohmic leakage, grade of synthetic fused quartz, has a better thermionic emission, and regenerative ef- transmission characteristic. Sapphire suffers fects. Ohmic leakage may result from con- some loss in transmission by reflection taminations on the insulators within the because of its relatively high index of refrac- tube, on the outside of the tube envelope, or

19 Photomultiplier Handbook

on the base. Thermionic emission generally separation of valence and conduction originates from the photocathode itself and bands). For an intrinsic semiconductor, ther- is amplified by the gain of the multiplier sec- mionic emission originates from the valence tion. Some emission may also come from the band, as does photoemission, but the “work secondary-emission dynode surfaces. Regen- function” is not the same as for photoemis- erative effects can occur in the tube par- sion. In the case of an intrinsic semiconduc- ticularly if it is operated with high voltage tor, thermionic-emission density can be ex- and high gain. (Regenerative effects are dis- pressed as cussed in more detail in a later section on Dark Current and Noise.) The following dis- cussion relates to the origin of thermionic emission current and gives practical values for this current in photomultiplier photo- cathodes. In a metal, the electrons which escape as thermionic emission are generally from the For an impurity semiconductor where top of the conduction band (see Fig. 3). thermionic emission originates from the im- Thus, the work functions for photoemission and for thermionic emission are the same. Thermionic emission as a function of work

ture T in degrees Kelvin is given by the familiar Richardson equation:

where j is the thermionic ; e, the electron charge; m, the electron mass; k, Boltzman’s constant; and h, Planck’s con- stant. If the constants before the exponential expression are given in mks units, the equa- tion (8) expressed in amperes per meter2 becomes

For semiconductor photocathodes, the work functions of photoemission and ther-

mionic emission may be quite different. The IO work function for photoemission (see Fig. 7) I is the electron energy corresponding to the height from the top of the valence band to the vacuum level, or EA (the electron affini- ty) plus EG (the forbidden gap, i.e., the

*An electron volt is the energy acquired by an electron in being accelerated through a drop in potential of one volt. In equations such as (8), the value of kT in the ex- ponent must also be expressed in electron volts. It is the Fig. 16 - Variation of thermionic-emission current density from various photocathodes in volts must be multiplied by the electron charge, e, in used in photomultiplier tubes as a function of coulombs (1.6 x 10 - 19) to obtain the work function in reciprocal temperature. Thermionic emission joules. The value of kT may then also be expressed in multiplied by the gain of the photomultiplier joules. is a principal source of anode dark current.

20 Photomultiplier Design purity centers, the equation for thermionic Fundamentals of Secondary Emission emission may be written as follows21: The physical processes involved in secon- dary emission are in many respects similar to those already described under Photoemis- sion. The main difference is that the impact of primary electrons rather than incident photons causes the emission of electrons. The steps involved in secondary emission can be stated briefly as follows: where EF is the Fermi level energy referenced from the top of the valence band, and no is 1. The incident electrons interact with the impurity concentration. electrons in the material and excite them to Typical dark-emission current-density higher energy states. characteristics for various photocathodes are 2. Some of these excited electrons move shown in Fig. 16 as a function of reciprocal toward the vacuum-solid interface, temperature. The current density is plotted on a logarithmic scale to show the 3. Those electrons which arrive at the sur- exponential-like character of the emission. face with energy greater than that repre- Anode dark current of the photomultiplier sented by the surface barrier are emitted into results from the cathode emission multiplied the vacuum. by the gain of the tube. Thermionic dark When a primary beam of electrons im- emission varies from tube to tube of the pacts a secondary-emitting material, the same type, probably because of the variation primary-beam energy is dissipated within the in no. Note that some of the curves show material and a number of excited electrons substantial curvature at the low temperature are produced within the material. The end. Some of this curvature is explained by numbers of excited electrons produced are the variation of T in the T2/T3/4 = T5/4 indicated in Fig. 17 for primary energies term. But an explanation for the greater part varying from 400 to 2200 electron-volts. The of the curvature may be the presence of total number of excited electrons produced patches of different impurity level or con- by a primary is indicated by the area of the centration, or it may be that the impurity individual rectangles in the figure. These ap- concentration itself is a function of tempera- proximate data are based on experimental ture. Exposure to temperatures above the data assuming that the range of primary normal operating range sometimes results in electrons varies as the 1.35 power of the permanent reduction in dark current. Most primary energy and that the number of elec- photocathodes are p-type semiconductors trons excited is uniform throughout the one result of which is a lower dark emission primary range. than for n-type semiconductors because of the reduced Fermi-level energy. 0.5

SECONDARY EMISSION When electrons having sufficient kinetic energy strike the surface of a material, secondary electrons are emitted. The fined as follows:

(12) -NANOMETERS 92cs-32303 where NS is the average number of secondary electrons emitted for Ne primary electrons Fig.17 - Theprocesses of secondary emis- incident upon the surface. sion.Seetextfor explanation.

21 Photomultiplier Handbook

As an excited electron in the bulk of the creases with voltage to much higher values. material moves toward the vacuum-solid in- Even though electrons are excited rather terface, it loses energy as a result of colli- deep in the negative-electron-affinity sions with other electrons and optical material and lose most of their excess energy . The energy of the electron is very as a result of collisions, many still escape in- rapidly dissipated as a result of these col- to the vacuum because of the nature of the lisons, and it is estimated that the energy of surface barrier. such an electron will decay to within a few times the mean thermal energy above the bottom of the conduction band within 10-12 second. If the electron arrives at the vacuum-solid interface with energy below that required to traverse the potential bar- rier, it cannot escape as a secondary elec- tron. Therefore, only those electrons excited near the surface of the material are likely to escape as secondary electrons. The probabil- ity of escape for an excited electron is as- sumed to vary exponentially with the excita- tion depth, as indicated in Fig. 17. If the pro- duct of the escape function and the number PRIMARYENERGY - keV of excited electrons (which is a function of 92cs -32304 primary energy and depth) is integrated, a Fig. 18 - Typical experimental curve of sec- ondary-emission yield as a function of . primary-electron energy in GaP:Cs and MgO. obtained, as indicated in the insert at the top Also shown in calculated curve for GaP:Cs of Fig. 17. The model which has been as- from Simon and Williams.13 sumed thus explains the general characteristics of secondary emission as a function of primary energy. Secondary- Because the GaP:Cs material is more dif- emission yield increases with primary ficult to handle than more conventional energy, provided the excited electrons are secondary emitters, and, therefore, results in produced near the surface where the escape higher cost, its use as a dynode has been probability is high. As the primary-electron restricted to applications where the very high energy increases, the number of excited elec- secondary emission is particularly advan- trons also increases, but the excitation oc- tageous. It is used as, for example, in curs at greater depths in the material where photomultiplier applications benefitting by escape is much less probable. Consequently, the reduction in statistical noise, or in the the secondary-emission yield eventually design of photomultipliers having fewer reaches a maximum and then decreases with stages for a given amplification. The use of primary energy. fewer stages also reduces the variation of gain with voltage changes. Secondary Emitter Materials In the development of photomultiplier Experimental secondary-emission-yield tubes it has been found that the photocath- values are shown as a function of primary- ode material may also be useful as a secon- electron energy in Fig. 18 for MgO, a tradi- dary emitter22.Such was the case in some of tional secondary-emission material, and for the first photomultipliers developed which GaP:Cs, a recently developed negative- used a Ag-O-Cs photocathode and Ag-O-Cs electron-affinity material. Also shown is a dynodes. Because of its high dark-emission calculated curve from Simon and Williams13 current and its instability, especially at based on their model of the GaP:Cs emitter. moderate current-density levels,this Although both MgO and GaP:Cs display the material is no longer used. general characteristics of secondary emission Other photocathode materials which also as a function of primary energy, as expected serve as secondary emitters are Cs3Sb, Rb- from the model illustrated in Fig. 17, the Cs-Sb, K CsSb, and Na2KSb:Cs. Because secondary-emission yield for GaP:Cs in- the processing2 of these secondary emitters is

22 Photomultiplier Design not identical in most cases to that of the cor- A very practical secondary emitter can be responding photocathode, the particular made from an oxidized silver-magnesium chemicalformulations specified here may alloy 25,26 containing approximately 2 per not be accurate. Secondary emission ratios cent of magnesium. Although silver- for these and other materials are shown in magnesium dynodes do not have as high a Fig. 19. secondary-emission ratio as some of the materials mentioned above (see Fig. 18), the material is easily processed and is more stable at relatively high currents. In addi- tion, it can tolerate higher temperatures. This surface has a low thermionic background emission which is important in applications requiring detection of low-level light. When it is activated with cesium, gain is somewhat higher. Without the cesium ac- tivation,the oxygen-activated silver- magnesium layer has been used effectively in demountable systems for detecting ions and other . A material having characteristics very similar to those of silver-magnesium is an oxidized layerof copper-beryllium alloy27,28,29in which the beryllium compo- Fig. 19 - Secondary emission ratios for a nent is about 2 per cent of the alloy. Secon- number of materials which have been used as dary emission is usually enhanced by the dynodes in photomultipliers as a function of bake-out in cesium vapor. A secondary- accelerating voltage of the primary electrons. emission characteristic of the cesium- activated copper-beryllium material is shown in Fig, 19. Because of the advantages in Very high secondary-emission yields have 23,24 handling and the manufacturing cost, the been reported for Na2KSb:Cs, the copper-beryllium is largely taking the place multi-alkali photocathode (S-20 response). of silver-magnesium in applications requir- Some photomultiplier with this ing low dark emission and stability at material as the secondary emitter although relatively high current densities. The lower its processing is complex. The very high secondary-emission yield is usually compen- yields, particularly at high primary energies, sated for in photomultiplier design by ap- would suggest that the material has an effec- plication of higher voltage or by an increase tive negative electron affinity similar to that in the number of dynode stages. of GaP:Cs. This explanation may also hold true for K-Cs-Sb, which is used is some tubes having this type of photocathode. A material that has been very commonly used is Cs3Sb corresponding to the photocathodes with S-4 or S-l1 spectral responses. It has good secondary emission in the practical working range near 100 volts. Rb-Cs-Sb is a rather new material which is just coming into use because the corresponding photocathode has good properties.All of the alkali an- timonides mentioned here have limitations. They cannot tolerate exposure to air and they are damaged by temperatures in excess 92CS - 32306 of 75 degrees C. In addition, stability suffers Fig. 20 - Typical secondary-electron energy distribution; peak at right is caused by re- cm-2. flected primary electrons.

23 Photomultiplier Handbook

When secondary electrons are emitted into second for an MgO layer34 formed on the the vacuum, the spread of emission energies surface of an AgMg alloy. may be quite large, as illustrated in the curve The upper limit for the time lag in photo- of Fig. 20 for a positive-electron-affinity emission, however, is not well established. emitter. The peak at the right of the curve From careful measurements of the time per- does not represent a true secondary, but formance of fast photomultipliers it can be rather a reflected primary. Data are not inferred that the limit must be less than available for the emission energies from a 10 - 10 seconds. negative-electron-affinity material, but they While these limits are a useful guide to the are expected to be considerably less than for type of time performance to be expected in positive-affinity materials. present photomultipliers, they will probably have less significance as photomultipliers us- TIME LAG IN PHOTOEMISSION AND ing new semiconducting photoemitters and SECONDARY EMISSION secondary emitters are developed. Semicon- ductors having minority-carrier lifetimes of Because both photoemission and secon- the order of microseconds are now available. dary emission can be described in terms of Probably, by combination of this character- the excitation of electrons within the volume istic with negative electron affinity, higher of the solid and the subsequent diffusion of gains and quantum efficiencies can be these electrons to the surface, a finite time achieved, but at a sacrifice of time response interval occurs between the instant that a or band width. However, the first generation primary (photon or electron) strikes a sur- of negative-electron-affinity emitters (e.g., face and the emergence of electrons from the Gap) has actually resulted in photomulti- surface. Furthermore, in the case of secon- pliers having better time performance dary emission, the secondaries can be ex- because a smaller number of stages pected to reach the surface over a period of operating at higher voltage can be used. At time. Within the limitations of a mechanistic this time it can only be concluded that in the approach to a quantum phenomenon, time future photomultipliers will probably be intervals for metals or insulators of the order designed to match in more detail the re- of 10-13 to 10-14second may be estimated quirements of a particular use. from the known energy of the primaries, their approximately known range, and the approximately known diffusion velocities of the internal electrons. In negative-electron- REFERENCES affinity semiconductors, it is known that the 19. V.K. Zworykin, and E.G. Ramberg, of internal “free” electrons having Photoelectricity and its Application, John quasi-thermal energies (i.e., electrons near Wiley and Sons, Inc., New York, (1949). the bottom of the conduction band) can be 20a. H. Rougeot and C. Baud, “Negative of the order of 10-10 second. Electron Affinity Photoemitters,” Advances Thus far, experiments have provided only in Electronics and Electron Physics, Vol. 48, upper limits for the time lag of emission. In Edited by L. Marton, Academic Press, 1979. the case of secondary emission, a variety of 20. J.J. Brady,“Energy Distribution of experiments have established limits. Several Photoelectrons as a Function of the 30,31,32 investigators have deduced limits Thickness of a Potassium Film,” Phys. from the measured performance of electron Rev., Vol. 46, (1934). tubes using secondary emitters. Others, 21. D.A. Wright, Semi-conductors, making direct measurements of these limits, Methuen and Co., New York (1955). have determined the time dispersion of 22. A.H. Sommer,“Relationship between secondary emission by letting short electron photoelectric and secondary electron emis- bunches strike a target and comparing the sion, with special reference to the Ag-O-Cs duration of the resulting secondary bunches (S-l) photocathode,” J. Appl. Phys., Vol. with the measured duration of the primary 42, pp. 567-569, (1971). bunch. Bythis means an upper limit of 23. A.A. Mostovskii, G.B. Vorobeva, and 6x10-l 2 second was determined for G.B. Struchinskii,Soviet Physics-Solid platinum33 and an upper limit of 7 x 10-11 1 State, Vol. 5, p. 2436, (1964).

24 Photomultiplier Design

24. C. Ghosh and B.P. Varma, “Secon- 29. A.H. Sommer,“Activation of silver- dary emissionfrom multialkali photo- magnesium and copper-beryllium dynodes,” cathodes,”J. Appl. Phys., Vol. 49, pp. J. Appl. Phys., Vol. 29, pp. 598-599, (1958). 4554-4555, (1978). 30. G. Diemer and J.L.H. Jonker, “On 25. V.K. Zworykin, J.E. Ruedy, and E.W. the time delay of secondary emission,” Pike,“Silver-magnesium alloy as a secon- Philips Research Repts., Vol. 5, p. 161 dary emitting material,” J. Appl. Phys.; (1950). Vol. 12, (1941). 3 1. M.H. Greenblatt and P.A. Miller, Jr., 26. P. Rappaport,“Methods of processing “A secondary-electron multi- silver-magnesium secondary emitters for plier,” Phys. Rev., Vol. 72, p. 160 (1947). electron tubes,”J. Appl. Phys., Vol. 25, pp 32. C.G. Wang,"Reflex Oscillators using. 288-292, (1954). secondary-emission current,” Phys. Rev., 27. J.S. Allen,“An improved electron Vol. 68, p. 284 (1945). multiplier counter,” Rev. Sci. Instr., 33. I.A.D. Lewis and F.H. Wells, Milli- Vol. 18, (1947). Microsecond Pulse Techniques, Pergamon 28. A.M. Tyutikov, “Structure and secon- Press (1959). dary emission of emitters of activated 34. K.C. Schmidt and C.F. Hendee, beryllium bronze, "Radio Engineering and “Continuous-channel electron multiplier Electronic Physics, (Translated from the operated in the pulse saturated mode,” Russian and published by the IEEE), Vol. 8 IEEE Trans. Nucl. Sci., Vol. NS-13, No. 3, No. 4, pp. 725-734, (1963). p. loo (1966).

25 Photomultiplier Handbook

3. Electron Optics of Photomultipliers

ELECTRON-OPTICAL DESIGN dynode electron optics. If all of the emitted CONSIDERATIONS photoelectrons are not guided properly to the first dynode, the signal-to-noise ratio of One of the primary design considerations the photomultiplier is degraded and poorer in a photomultiplier tube is the shaping and pulse-height-resolution characteristics result. positioning of the dynodes (usually in a re- The region between the last dynode and current geometrical pattern) so that all the the anode is also of special electron-optical stages are properly utilized and no electrons concern. The withdrawal field at the last are lost to support structures in the tube or dynode should be large to minimize space deflected in other ways. Although it is not charge effects which limit the linearity and necessary that the electrons come to a sharp magnitude of output current pulses. Another focus on each succeeding stage, the shape of consideration in high-speed photomultipliers the fields should be such that electrons tend is to provide a structure which is matched to to return to a center location on the next appropriate transmission lines. dynode, even though the emission point is The complete electron-optical configura- not at the location of the preceding tion of the photomultiplier must also be such dynode. If this requirement is not met, the as to avoid regenerative effects. For exam- electrons increasingly diverge from the ple, there should not be an open path in the center of the dynode in each successive tube through which occasional ions or light dynode stage. This effect in turn can lead to could feed back from the output end to the the skipping of stages and loss of gain. photocathode. Magnetic fields may be combined with elec- trostatic fields to provide the required elec- tron optics, although today most photomul- tipliers are electrostatically focused. In DESIGN METHODS FOR PHOTO- addition to providing good collection of MULTIPLIER ELECTRON OPTICS secondary electrons from stage-to-stage, it is important for some applications to minimize Before the days of high-speed computers, the time spread of electron trajectories. For photomultiplier electron-optical problems this purpose is is useful to provide strong were often solved by means of mechanical electric fields at the surfaces of the dynodes analogues such as a stretched rubber mem- to assure high initial acceleration of the elec- brane. When mechanical models of the elec- trons. Also important may be the design of trodes were placed under such a membrane configurations which provide nearly equal and their height adjusted to correspond to transit times between dynodes regardless of the desired electrical potential, the height of the point of emission on the dynode. the membrane in the spaces between the elec- In scintillation counting applications, fair- trodes corresponded to the equivalent elec- ly large photocathode areas are required for trical potential. Small balls were then al- efficient scintillator coupling. Ideally, the lowed to roll from one electrode to the next. photocathode should be semitransparent The trajectories of the balls corresponded to and located on the flat window of the tube. those of the electrons in the photomultiplier This requirement poses a special problem in structure. With appropriate model design, design of efficient photocathode-to-first- friction and depression of the membrane by

26 Electron Optics of Photomultipliers the ball were made negligible. This model, electrode configurations. Collection efficien- however, was only valid for geometries in cy and time response may be predicted from which the electrodes could be assumed to be an analysis of the electron trajectories. Col- cylindrical surfaces (generated by a line lection efficiency at the first dynode is de- parallel to a fixed direction and moving fined as the ratio of the number of photo- along a fixed curve) sufficiently long to be electrons which land upon a useful area of considered infinite in extent. Application the dynode to the number of emitted photo- was made to numerous dynode configura- electrons. If all the photoelectrons begin tions. their trajectories at the surface of the photo- Another useful analogue was the resis- cathode with zero velocity, 100% collection tance network such as a two-dimensional would be possible. Because of the finite in- array of connectors having equal spacing itial velocities, however, some electrons vertically and horizontally. This array repre- begin their trajectories with unfavorable sented a cross-section plane through cylin- angles of launch and are not collected on a drical surfaces again assumed to be infinite useful area. in length. of equal value were con- In modern photomultiplier structures, nected between adjacent connectors both first-dynode collection efficiencies range vertically and horizontally. Points in the ar- from 85 to 98 per cent. Ideally, the emitted ray corresponding to an electrode were all photoelectrons should converge to a very connected to the same potential. Equipoten- small area on the first dynode. In practice, tial lines between electrodes could then be this electron-spot diameter is usually less determined by observing the potential values than 1/4 of the cathode diameter, depending on the connectors between the electrodes. upon the tube type and focusing structure. When the equipotential lines had been deter- mined, electron trajectories could be readily calculated between closely spaced equipoten- tial surfaces. In another variation of the resistance - SPECIFIC PHOTOMULTIPLIER work, the vertical distribution of resistance ELECTRON-OPTICAL values was made logarithmic instead of uni- CONFIGURATIONS form. This array then corresponded to an Circular-Cage Structure with axially symmetrical system with the axis ap- Side-On Photocathode proximated across the top of the board. An The first commercially successful photo- example of a relevant problem is the region multiplier design was based on a circular ar- between the photocathode and first dynode ray of photocathode and dynodes-as in the for end-on-type photomultipliers where the 931A. This design is depicted schematically axis passes through the center of the photo- in Fig. 21. The photocathode is of the cathode. “opaque” type; i.e., electrons are emitted Only very simple electron-optical systems from the same side as the photocathode is il- can be solved in closed form, which requires luminated. This particular type of tube is relatively inexpensive and is useful for ap- However, by using relaxation techniques plications such as spectroscopy where high with a computer, the potential distribution sensitivity is required but the photocathode on a set of points confined within defined area need not be large. In this case the best boundary conditions can be determined. collection is from the side of the photocath- Secondly, using the force equations, electron ode near the first dynode. In the part of the paths can be traced. Computer programs ex- photocathode near the apex formed by the ist for solving various cases with or without grill and the photocathode, the electrostatic symmetry and with irregular potential boun- fields are small and collection efficiency is daries. poor. The time of response of the tube is An electron-optical design for a photo- short because of its small size and high inter- multiplier may be arrived at from computer- dynode field strengths, and there is relatively developed equipotential lines and electron small feedback from the anode end of the trajectories plotted on a plan showing the tube back to the photocathode.

27 Photomultiplier Handbook

Circular-Cage Structure with useful first-dynode area in the end-on con- End-On Photocathode struction. In addition, a skewed field is pro- Fig. 22 illustrates the use of a circular cage vided by the focusing electrode which results coupled to a flat semi-transparent photo- in a more favorable pattern of impacting cathode. Such a configuration is useful in photoelectrons on the first dynode. This scintillation counting where the flat photo- end-on construction still has a relatively cathode is coupled to the scintillating crystal. short time response because of the focused In order to improve the collection efficiency dynode arrangement, although the transit- of photoelectrons by the first dynode, the time spread of electrons traversing the electrode which was the photocathode in photocathode-to-first-dynode space in- Fig. 21 has been modified to provide a larger creases the time of response as compared with that of the simple circular side-on struc- ture of Fig. 21. Box-and-Grid Structure with End-On Photocathode The box and grid dynode structure shown in Fig. 23 is also useful in scintillation count- ing because of the relatively large flat semi- transparent photocathode. This configura- tion has the advantage of providing a rather large entrance area to collect photoelectrons so that collection efficiency is very nearly 100%. This feature is important in providing good pulse-height resolution in scintillation counting. The individual dynode boxes are open at the exit end and have a grid at the en- trance. The grid provides an l-10: DYNODES which penetrates the preceding dynode re- 92CS-32308 I I : ANODE gion and aids in the withdrawal of secondary fig. 22 - An end-on photomultiplier structure electrons. The grid also eliminates a retard- utilizing a circular dynode arrangement. This ing field that would be caused by the poten- type of tube would be useful as a detector for tial of the preceding dynode. Because the scintillation counting. field penetration is rather weak, the electron

28 Electron Optics of Photomultipliers

INCIDENT ‘RADIATION

INCIDENT RADIATION

Fig. 23 - The box-and-grid multiplier structure.

FACEPLATE

INCIDENT RADIATION

Fig. 24 - The Venetian-blind multiplier structure. transit time between dynodes is relatively electrons are lost because of the interposition slow and has a rather large time spread. of the grids between stages, as with the box- and-grid construction. Venetian-Blind Structure Another structure which provides good Tea-Cup Structure collection of photoelectrons, but again is A recent innovation in front-end design is rather slow in response time, is the venetian- the so-called “tea-cup” photomultiplier, blind photomultiplier shown in Fig. 24. named after its large first dynode. Secondary Good collection of photoelectrons is aided electrons from the first dynode are directed by the size of the photoelectron collecting to an opening in the side of the tea-cup and area which can be even larger than that of thence to the second dynode. Fields between the box-and-grid construction. The relatively the photocathode region and the first slow time response is the result of the weak dynode region are separated by a very fine electric field at the surfaces of the dynode grid structure. The particular advantage of vanes. The structure is very flexible as to the this design is that the collection efficiency of number of stages. Some of the secondary the large first dynode is good not only for

29 Photomultiplier Handbook

92CM-32311 Fig. 25 - Tea-cup photomultiplier showing photoelectron paths directly from the photocathode and those initiated by light transmitted through the photocathode and striking the side wall which also has an active photoemissive layer. Also in- dicated are equipotential lines in the region between photocathode and first dynode. photoelectrons emitted from the photocath- In-Line Dynode Structure with ode but for photoelectrons emitted from the Curved Photocathode activated side walls between the first dynode The planar-photocathode design, such as and the photocathode as shown in Fig. 25. that shown in Figs. 22-25, provides excellent The side-wall photoemission results from coupling to a scintillation crystal, but its light which passes through the front semi- time response is not as good as that of a transparent photocathode. The increased spherical-section-photocathode design. The photoemission and collection efficiency im- spherical-section photocathode shown in proves the pulse-height resolution in scin- Fig. 26 when coupled to a high-speed elec- tillation counting applications. Indicated on tron multiplier provides a photomultiplier the diagram are equipotential lines and pho- having a very fast time response. Some tube toelectron paths showing the collection of types utilize a spherical-section photocath- electrons from the front surface and from ode on a plano-concave faceplate to facili- the side walls. tate scintillator coupling. This design is Electron Optics of Photomultipliers

FRONT- END REGION

92CM-32312

Fig. 26 - Photomultiplier design with curved faceplate and in-line dynode structure to provide a minimum transit time and transit-time spread.

PHOTOCATHODE

TYPICAL ELECTRON TRAJECTORIES

TYPICAL EQUIPOTENTIAL LINES

DYNODE NO. I

92CM-32313

Fig. 27 - Cross section of a photomultiplier showing equipotential lines and electron trajectories that were plotted by computer.

31 Photomultiplier Handbook

usually limited to faceplate diameters of two shorter electron paths and stronger fields inches or less because of the excessive thick- from dynode to dynode to produce nearly ness of the glass at the edge. The plano- equal total transit time. concave faceplate may also contribute to Certain materials have advantages over some loss in uniformity of sensitivity be- other materials in providing good multiplier cause of internal reflection effects near the time performance. For example, a standard thick edge of the photocathode. dynode material, copper beryllium, has a The performance of the front-end struc- maximum gain per stage of 8 at a 600-volt in- ture shown in Fig. 26 has been determined by terstage potential difference. In contrast, the use of a computer to trace equipotential gallium phosphide exhibits a gain of 60 or lines and electron trajectories which have more at a 1200-volt interstage potential dif- then been superimposed on a schematic dia- ference. The advantage in time performance gram of the tube structure as shown in Fig. of the GaP dynode over one of the CuBe 27. type is that the number of stages, and thus A time parameter of interest is the photo- the total transit time may be reduced, and cathode transit-time difference, the time dif- that the energy spread of the secondary elec- ference between the peak current outputs for trons is less. simultaneous small-spot illumination of dif- ferent parts of the photocathode. In a planarcathode design, the transit time is longer for edge illumination than for center illumination because of the longer edge tra- jectories and the weaker electric field near the edge of the photocathode. The center-to- edge transit-time difference may be as much as 10 nanoseconds. The spherical-section photocathode affords more uniform time re- sponse than the planar photocathode be- cause all the electron paths are nearly equal in length; however, the transit time is slightly longer for edge trajectories than for axial trajectories because of the weaker electric field at the edge. The photocathode transit-time difference is ultimately limited by the initial-velocity distribution of the photoelectrons; this distribution causes time-broadening of the electron packet during its flight from the photocathode to the first dynode. The

broadening effect can be minimized by in- 92cs-32314 creasing the strength of the electric field at the surface of the photocathode. Fig. 28 - A compensated-design multiplier. Because the energy spread of secondary electrons is even larger than that of photo- Continuous-Channel Multiplier Structure electrons, initial-velocity effects are the ma- The continuous-channel multiplier struc- jor limitation on the time response of the ture35 shown in Fig. 29 is very compact and electron multiplier. Multiplier time response utilizes a resistive emitter on the inside sur- is usually improved by the use of high face of a cylinder rather than a discrete electric-field strengths at the dynode surfaces number of dynodes. The lack of discrete and compensated design geometries. In a dynodes causes the electron-multiplication compensated design, such as that shown in statistics to be poor because of the variable Fig. 28 and as used in the type of photomul- path lengths and the variable associated tiplier shown in Fig. 26, longer electron . The gain of a continuous-channel paths and weaker fields alternate with multiplier is determined by the ratio of the

32 Electron Optics of Photomultipliers channel length to inside diameter; a typical photoelectrons measured at full width at half value of this ratio is 50 but it may range from maximum (FWHM) is less than 300 pico- 30 to 100. seconds.37 Because of the construction and high electric fields employed, the sensitivity to external magnetic fields is much reduced, a fact which is important in some nuclear physics experimentation. Crossed-Field Multiplier Structure As mentioned earlier, a crossed-field 92cs-32315 photomultiplier was first reported in earl 3g Fig. 29 - The continuous-channel multiplier 1936 by Zworykin, Morton and Malter. structure. Their tube, shown schematically in Fig. 31, used a combination of electrostatic and mag- netic fields to direct electrons to repeated Very high gain can be achieved with such a stages of secondary emission. Above each single-channel multiplier, provided the chan- emitter was a field plate whose potential was nel is curved or bent to avoid line-of-sight set to be equal to that of the next emitter feed-back paths. Numerous special purpose down the line. As a result of this configura- photomultiplier tubes have been built using tion, electrons emitted from the photocath- this concept. ode or from one of the secondary emitters, Recently,36 very-high-speed photomulti- were caused to follow approximately cy- plier tubes have been designed utilizing cloidal paths to the next electrode. This early microchannel plates in proximity with the development was not carried into large-scale photocathode and anode. Such a device is il- manufacture because of the critical lustrated in Fig. 30. A microchannel plate is magnetic-field adjustments needed to change an array of parallel channels, each perhaps the gain. Also, the rather wide open struc- ture resulted in high dark current because of ray is mounted close to the photocathode feedback from ions and light developed near with the plane of the channel plate parallel to the output end of the device. the photocathode. A high voltage, 1 kilo- Recently, however, similar, but improved volt, may be applied between photocathode crossed-field photomultipliers39 have been and microchannel plate, and similar voltages designed to provide perhaps the fastest rise applied across the microchannel sandwich time of any photomultiplier. The electron and between the plate and the anode. These trajectories are essentially isochronous in the voltages assure short transit times. Time crossed magnetic and electric fields so that a resolution for pulses initiated by single very short rise time, 250 picoseconds40, is

MICROCHANNEL PHOTOCATHODE PLATE / /FACEPLATE \

I 1

ANODE

92cs-32316 Fig. 30 - Microchannel-plate photomultiplier.

33 Photomultiplier Handbook

achieved. Unfortunately, the photocathode is rather small and inaccessible by the nature of the design. A schematic drawing of the re- cent crossed-field photomultiplier is shown in Fig. 32. Note the single field electrode in contrast to the multiple field the plates of the Zworykin tube. The stepped arrangement is required in order to provide uniform electric field. A 50-ohm coaxial output connector provides coupling to the high-speed photo- multiplier. POTENTIAL BETWEEN LAST DYNODE AND ANODE-VOLTS 92cs -32319 A FIELD PLATES Fig. 33 - Constant-current characteristics of a photomultiplier anode.

CATHODE COLLECTOR 92CS - 32317 GRID-LIKE Fig. 31 - Schematic of the Zworykin crossed- TRAJECTORY field photomultiplier reported in 1936. A mag- netic field, perpendicular to the plane of the drawing, and an electrostatic field, produced by the upper fieldplates, combine to bend the electron paths in the cycloidal trajectories il- Iustrated.

Fig. 34 - The simplest anode structure, a grid-like collector.

In applications where large output pulse currents are required it is important to pro- vide a design having reasonably high with- drawal fields to avoid space-charge develop- ment which could limit the output current. FIRST may actually limit the output at DYNODE the next to the last dynode, which is the case 92CS-32318 in the circular-cage structure, Fig. 21, Fig. 32 - Schematic arrangement of a because the withdrawal field at the last modern static crossed-field photomultiplier. dynode is significantly higher than at the eighth dynode. In special applications a tapered divider network may be used to pro- ANODE CONFIGURATIONS vide higher inter-stage voltages in the last The primary function of the anode is to several stages of the tube (see Chapter collect secondary electrons from the last 5-Photomultiplier Applications). dynode. The anode should exhibit a Where fast time response is important, the constant-current characteristic of the type type of design shown in Fig. 34 may be at a shown in Fig. 33. The simplest anode struc- disadvantage because electrons from the last ture, shown in Fig. 34, is a grid-like collector dynode are not necessarily collected on the used in some Venetian-blind structures. The first pass through the anode grid structure. It secondary electrons from the next-to-last is also important for fast time response that dynode pass through the grid to the last the be designed with matched- dynode. Secondary electrons leaving the last impedance transmission lines or with short dynode are then collected on the grid-like connecting support leads. Most high-speed anode. circuits are designed to utilize a 50-ohm im-

34 Electron Optics of Photomultipliers pedance, which requires a suitable connector 36. Ph. Chevalier, J.P. Boutot and G. or lead geometry outside the tube to permit Pietri, “PM of new design for high speed proper impedance matching. physics,” IEEE Trans. Nucl. Sci., Vol. Anode configurations and internal trans- NS-17, No. 3, pp. 75-78 (June, 1970); G. mission lines may be analyzed by use of the Pietri , “Contribution of the channel elec- standard methods of cavity and transmis- tron multiplier to the race of vacuum tubes sion-line analysis. These methods yield ap- towards picosecond resolution time,” IEEE proximate design parameters41, which are Trans. Nucl. Sci. , Vol. NS-24, No. 1, pp. optimized experimentally by means of time- 228-232, (Feb. , 1977). domain reflectometry, TDR42’43. TDR pro- 37. C.C. Lo, P. Lecomte and B. Leskovar, vides information about the discontinuities “Performance studies of prototype micro- in the characteristic impedance of a system channel plate photomultipliers," IEEE as a function of electrical length and is an ex- Trans. Nucl. Sci., Vol. NS-24, No. 1 , pp. tremely useful approach in the design of 302-311 , (Feb. , 1977). voltage-divider circuits and mating sockets 38. V.K. Zworykin, G.A. Morton, and L. which do not readily lend themselves to Malter , “The secondary emission multi- mathematical analysis. plier-a new electronic device”’ Proc. Certain anode structures in fast-rise-time I.R.E. Vol. 24 p. 351-375 (1936). photomultipliers exhibit a small-amplitude 39. R.S. Enck and W.G. Abraham, pulse (prepulse) that can be observed a “Review of high speed communications nanosecond or two before the true signal photomultiplier detectors," Proc. Soc. pulse. In grod-like anode structures this Photo Optical Instrumentation Engineers, prepulse is induced when electrons from the Vol. 150; Laser and Fiber Optics Communi- next-to-last dynode pass through the anode cations’ San Diego CA; 28-29 Aug. , 1978 grid. In more sophisticated photomultiplier (Bellingham , WA; Soc. Photo-Optical In- designs this phenomenon may be suppressed strumentation Engineers , 1978) , p 3 l-8. by auxiliary grids that shield the anode from 40. B. Leskovar and C.C. Lo, “Time Res- the effects of the impinging electron cloud. olution Performance Studies of Contempo- rary High Speed Photomultipliers,” IEEE Trans. Nucl. Sci., Vol. NS-25, No. 1, pp. 582-590, (Feb. 1978). REFERENCES 41. I.A.D. Lewis and F.H. Wells, Milli- Microsecond Pulse Techniques, Pergamon 35. K.C. Schmidt and C.F. Hendee, Press (1959). “Continuous-Channel Electron Multiplier 42. “Time-Domain Reflectometry,” Operated in the Pulse-Saturated Mode,” Hewlett Packard Application Note 62. IEEE Trans. Nucl. Sci., Vol. NS-13, No. 3, 43. “Time-Domain Reflectometry,” p. 100 (1966). Tektronix Publication 062-0703-00.

35 Photomultiplier Handbook

4. Photomultiplier Characteristics

PHOTOCATHODE-RELATED CHARACTERISTICS Current-Voltage Characteristics Photomultiplier tubes may be operated as photodiodes by utilizing the first dynode and focusing electrode (if present) tied together as an anode. In this mode of operation the photocurrent is linear with light flux except that, for semitransparent photocathodes, the resistivity of the thin photocathode layer limits the current which can be drawn. In the VOLTAGE-VOLTS case of a resistive photocathode, when the 92CM-32321 light is directed to the center of the sensitive Fig. 35 - Current-voltage characteristics for area, there is a drop in potential between the 75-mm diameter photomultiplier tubes oper- outer conductive ring and the illuminated ated as photodiodes. The poor collection effi- area. Because the illuminated area becomes ciency shown for the tube with the resistive positively charged with respect to the photo- photocathode is caused by the voltage drop cathode contact, only the more energetic from the edge of the photocathode to the cen- photoelectrons emitted will overcome the ter illuminated spot and the resulting electro- slightly repelling electric field near the static field distortion between the photocath- photocathode surface. ode and the first dynode and focus electrode Fig. 35 shows a pair of current-voltage used together as an anode. characteristics for 75-mm-diameter photo- multipliers operated as photodiodes. In both cases the illuminated area is approximately 5 mm in diameter and located at the center of ode current to avoid non-linear operation. the photocathode. The multialkali photo- Resistivity characteristics of several types of cathode (S-20 response) layer is fairly con- photocathode are shown in Fig. 36. Here, ductive so that even for a photocurrent of the resistance per square is shown as a func- over 200 nanoamperes, good collection effi- tion of temperature. The increase in resistivi- ciency is achieved for a collection voltage as ty with decreasing temperature is expected low as 50 volts. On the other hand, the because of the semiconductor nature of bialkali photocathode , K2CsSb , is very photocathode materials. Resistance per resistive causing poor collection efficiency square is the resistance of a surface layer be- over a wide range of voltage even for photo- tween conductors at opposite sides of a emission currents as low as 10 nanoamperes. square of the layer. Note that, if the resistivi- For resistive photocathodes it is thus necessary to limit the maximum photocath- square of side dimension d, and thickness t,

36 Photomultiplier Characteristics

the resistance R is given by Photocathode resistive effects can be avoided at manufacture by the use of nearly transparent conductive undercoatings. Opaque types of photocathode, such as Cs3Sb on a solid substrate of , do not where d l t is the cross-section area. Thus, have a resistivity problem. Table II provides the resistance per square is independent of guidance as to the maximum dc current that the side dimension. On photocathode types can be utilized with various semitransparent that are very resistive , it is advisable to main- photocathodes at room temperature. It tain operating temperatures above -100° C should be realized that these data are only depending upon photocathode diameter , the typical; individual photocathodes may vary light-spot diameter, and the photocathode considerably from the guideline provided in current. The larger the photocathode diam- their capability of delivering current without eter and the smaller the light-spot diameter, resistive blocking. the more severe the effect. Table II - Maximum Recommended Pboto- cathode DC Currents for Various Semitrans- parent Photocathode Types as Determined by Surface Resistivity. Photocathode Maximum Recommended Current for T = 22°C

2

In the case of pulsed photocathode cur- rents, the peak current values may be higher than those shown in Table II. Average cur- rents, however, are still limited, as shown, by the resistance of the photocathode layer. For a current pulse, the local surface poten- tial of the cathode is sustained for a time by the electrical charge associated with the pho- tocathode capacitance. Thus, a square cen- timeter of photocathode may have a capaci- tance of about 0.5 picofarad. A 5-volt change in potential is about the maximum that could occur without blocking of the photoemission current. The equivalent charge is therefore 2.5 picocoulombs. For a pulse duration of t seconds, a pulse current of 2.5/t picoamperes could be maintained. For t = 1 microsecond, the current maximum Fig. 36 - Resistance per square as a function of temperature for various semitransparent could be 2.5 microamperes, as determined photocathodes. These data were ob- by stored charge alone. tained with special tubes having connections Pulse current maximum can be increased to parallel conducting lines or between con- by the capacitance of the photocathode centric conducting circular rings on the Sur- layer. For example, if a ground plane is pro- face of the photocathode. vided in contact with the outside of the

37 Photomultiplier Handbook photocathode faceplate, the capacitance of a temperature is illustrated in Fig. 38, also for square centimeter of photocathode may be Cs3Sb. According to Spicer and Wooten47, increased to about 2 picofarads (assuming a the increase in response in the blue at low dielectric constant 6.75 and a glass thickness temperatures is the result of a decrease in of 3 mm). The ground plane must, of course, energy loss from lattice scattering; the de- be reasonably transparent to the light signal. crease in the red at low temperatures results A conducting mesh is a suitable solution. from a decrease in occupied defect levels in The ground plane should also be set at or the forbidden energy band because of an in- near photocathode potential, or noisy opera- crease in band gap and possibly because of tion and deterioration of the photocathode an unfavorable change in band bending. It may result. See the section below under may also be the case that, in photocathodes “Dark Current and Noise.” having significant impurity levels, the ex- tended red sensitivity at higher temperatures Variations in Spectral Response with is partly the result of thermally assisted Temperature. photoemission. The change of slope at the Minor variations in the spectral response elevated temperature in Fig. 38 is suggestive characteristics of photocathodes occur with of this mechanism. changes in temperature. Data are presented here on the typical variations which can be expected. Deviations from these data may be expected on individual photocathodes be- cause of variations in thickness and process- ing. When the temperature is decreased, the response of photocathodes usually improves in the shorter-wavelength region and worsens in the longer-wavelength region near the threshold. An illustration of this trend is provided in Fig. 37 in which the

0.1 1 I I I 1 400 500 600 WAVELENGTH -NANOMETERS

92cs-32324 Fig. 38 - Dependence of responsivity of a Cs3Sb photocathode on temperature; taken from Spicer and Wooten.

Fig. 37 - Temperature coefficient of cesium- The data shown in Fig. 39 is of the same antimony cathodes as a function of wave- type as that in Fig. 38, but for a GaAs:Cs length at 20 “C. Note the large positive effect photocathode. Note, again, the increase in near the threshold. sensitivity with decrease in temperature ex- cept for the shift in the long wavelength temperature coefficient of responsivity of a threshold. In GaAs:Cs the photoexcitation is cesium-antimony photocathode is plotted as from the valence band; there is essentially no a function of wavelength.46 The difference excitation from the forbidden energy band. in spectral response from room to liquid-air Note the sharp long-wavelength cutoff.

38 Photomultiplier Characteristics

cidence. This conclusion is at odds with some reported measurements. Experimental measurements for a semitransparent Cs3Sb

Fig. 39 - Relative spectral responsivity shift with temperature for a GaAs:Cs photocath- 92CM-32326 48 ode. Fig. 40 - Theoretical photoexcitation deter- mined by optical factors for different polariza- Variation of Photocathode Response with tion orientation as a function of angle of Angle of Incidence and Angle of Polariza- incidence. Radiation is incident on a semi- tion. transparent photocathode such as Cs3Sb In a theoretical paper49 Ramberg has through a glass substrate of index of refrac- evaluated the optical factors which deter- tion 1.5. Data are taken from Ramberg. mine the variation of photoresponse as a function of angle and angle of incidence. One interesting conclusion of Ramberg’s work is that for normal incidence and for a certain regime of cathode thickness and escape depth, the photoemission “may be expected to be about 1.4 times as great for illumination from the glass side as for il- lumination from the vacuum side.” Photo- multiplier tubes with semitransparent 0 photocathodes such as are used in typical scintillation counting applications are all designed for radiation incident on the glass substrate which supports the photocathode. 2.0 The two curves of Fig. 40 are calculated from Ramberg’s formulae for incidence through the glass faceplate, and for polariza- tion perpendicular and parallel to the plane of incidence. (i.e. “parallel” implies that the electric vector lies in the plane defined by the incident ray and the normal to the glass sur- face.) These curves assume an index of (b) refraction for the glass of 1.5, for the photo- 92cs-32327 sensitive layer of 3.25, and the absorption in- dex for the photosensitive layer, K, of 3.25. Fig. 41 - (a) Photoresponse for a semitrans- The photocathode layer is assumed to be parent Cs3Sb photocathode as a function of thin with respect to the wavelength of light- (0°) and perpendicular (90°) to the plane of in- taken as 550 nanometers for Fig. 40. cidence. (b) Photoresponse ratio for polariza- Note that Ramberg’s prediction is that response should be greater for perpendicular sured at an angle of incidence of 70°(from polarization, especially at large angles of in- Hoenig and Cutler50).

39 Photomultiplier Handbook

photocathode on Corning 9741 glass are reported by Hoenig and Cutler50. Their data, are shown in Fig. 41 a and b. Their curve, a, shows the variation of photosen- for two different angles of polarization. The parallel to the plane of incidence. The curve, b, shows the ratio of the photoresponse for polarization angle 0° to the photoresponse both measured at 70° angle of incidence. Note that the response is greater for 92CS-32328 polarization parallel to the plane of in- Fig. 42 - A glass quadrant in optical contact cidence . with the photocathode window, permitting in- Measurements at RCA support the general coming radiation to approach the glass- conclusion that the response is higher for photocathode interface at any angle of in- parallel polarization for Cs3Sb, K2CsSb and critical angle for the glass-air interface (about Na2KSb:Cs semitransparent photocathodes 42°) all of the radiation is reflected, permit- at large angles of incidence. On the other 51 ting multiple excitations of the photosen- hand, Hora finds the response for a sitive surface. “trialkali antimonide” to be higher for perpendicular polarization. D.P. Jones56 on the behavior of a semitrans- It is likely that the discrepancies noted parent Na2KSb:Cs photocathode with angle above are related to the differences in of incidence, wavelength, and polarization photocathode substrates (between the glass angle. The optical arrangement is similar to and the photocathode), photocathode thick- that shown in Fig. 42 except that a full 180° ness, and photocathode processing which cylindrical lens is cemented to the faceplate may result in non-isotropic and structured and half of the surface is coated with layers, in contrast, for example, to the aluminum to reflect the rays back to the assumption by Ramberg of a very thin and photocathode. His data and theory favor isotropic photocathode layer. perpendicular polarization, particularly at Optical Devices to Enhance Photoresponse longer wavelengths. Numerous optical devices have been de- Photocathode Uniformity vised to enhance photoresponse by utilizing The uniformity of photocathode response multiple paths in the faceplate and photo- over its area may be of importance in some cathode. Experimental data and theoretical applications. For example, in scintillation analyses have been reported by a number of counting, light pulses which excite different different authors.52-56 Fig. 42 shows a areas of a non-uniform photocathode would typical arrangement permitting the light then result in variations in measured pulse beam to enter the glass substrate so that height. Such an effect would be more ap- reflected rays approach the glass-air inter- parent in the case where the scintillating face at an angle greater than the critical crystal is thin. In flying-spot scanners, non- angle and thus do not permit a refracted ray uniform photocathode sensitivity would to escape the glass. In this manner multiple result in picture shading unless precaution is excitations of the photosensitive layer are taken to avoid even approximate imaging on possible. Enhancement of the quantum effi- the photocathode surface. ciency for an alkali-antimonide photocath- Photocathode non-uniformity may result ode may be something less than 2:1 in the from a variety of causes. Fig. 43 shows some blue where the absorption of the photocath- tests of photocathode uniformity. These ode layer is high, but may be as high as 4: 1 in curves represent photoresponse for a small the red where the absorption is low. focused (l/16-inch) spot as it is scanned An interesting series of measurements and across the diameter of a semitransparent theoretical predictions have been reported by photocathode. Curve (a) represents the

40 Photomultiplier Characteristics uniformity ( ± 3%) typical of a well-pro- creased emission, or there may actually be cessed photocathode in a 2-inch photomulti- some photosensitivity on the shoulder plier designed for scintillation counting. In because of the presence of processing curve (b), the variation (1) across the photo- materials. cathode is the result of a non-uniform Anomalies in the photocathode uniformi- evaporation of antimony during the process- ty such as shown in Fig. 43(c) can be largely ing of the photocathode. Curve (c) shows eliminated by a fabrication technique57 some peculiar non-uniformities in a 3-inch creating a diffusing layer on the photocath- photocathode. The peaks (2) are the result of ode substrate. The fabrication of a diffusing light entering the rounded corner of the tube layer can be easily accomplished by sand- envelope and increasing emission by reflec- blasting the inside surface of the photomulti- tions in the glass faceplate. The peaks (3) plier faceplate. The effect of the diffusing result when light,which is transmited layer is to scatter the incoming light so that through the photocathode, strikes the re- light transmitted through the photocathode entry shoulder of the glass envelope. The in- is spread out in many directions, thus side shoulder of the envelope is aluminized avoiding direct correlation between the posi- as part of the photocathode contact. Some tion of the incoming light spot and any inter- light striking the shoulder may be reflected nal reflections. The diffusing layer also has a back to the photocathode and cause in- desirable effect of enhancing the photocath- ode sensitivity. Uniformity of photocathode response, however, is not sufficient to assure a uniform output response from the photo- multiplier. All of the emitted photoelectrons may not be properly directed to the first dynode by the electron optics. In the case of a Venetian-blind dynode structure, some of the photoelectrons may strike parts of the dynode which do not provide good collec- tion fields to the second dynode. Fig. 44 is a

DISTANCE ACROSS PHOTOCATHODE-INCHES 92cs-32329 92CS-32330 Fig. 43 - Photocathode uniformity patterns Fig. 44 - Relative collection efficiency for observed by scanning a 1/16-inch spot across photoelectrons in a venetian-blind type photo- a diameter of a semitransparent photocath- multiplier tube.58 ode in 2-inch and 3-inch photomultiplier tubes used in scintillation counting. (a), scan on a plot of collection uniformity in a 3-inch typical well-processed 2-inch tube; (b), scan Venetian-blind photomultiplier tube. The on a tube with non-uniform antimony evapora- structure in the pattern for the scan perpen- tion on the faceplate (1); (c), scan on a typical 3-inch tube showing peaks (2) due to entry of dicular to the length of the dynode is related light on the edge of the faceplate and (3) to the individual slats in the Venetian-blind peaks due to reflection or photoemission structure. By way of contrast, Fig. 45 shows from transmitted radiation through the photo- similar collection uniformity plots for a cathode onto the shoulder of the bulb. 3-inch “teacup”design where the large

41 Photomultiplier Handbook open structure of the “tea-cup” first dynode If a photomultiplier were to contain any eliminates the electron-optical effects significant trace of gas, ionization of the gas observed in the Venetian-blind type tube. could occur , depending upon the applied voltages and level of photocurrent. bom- bardment of the photocathode can readily cause damage and loss of sensitivity in pro- portion to the ion current. Photomultiplier tubes, however, are generally so well evacu- ated that ion damage is essentially only a theoretical possibility. Electrolytic effects can bring about serious fatigue in Cs3Sb photocathodes (Ref. 11, p. 79) and presumably in other alkali- antimonide photocathodes. An electrolytic effect is caused when a potential gradient is maintained across the cathode surface. DISTANCE ACROSS PHOTOCATHODE- INCHES Photocathodes generally only have one 92CS-32331 physical contact, but a gradient can be Fig. 45 - Relative collection efficiency for developed because of the resistive nature of photoelectrons in a “teacup” type RCA pho- 58 the photocathode layer if a large photocur- tomultiplier tube. rent is drawn, for example, from a center spot on the cathode. Actual electrolytic Uniformity of response of a photomulti- decomposition takes place, which can be plier may also be affected by the voltage at recognized by a color change in the cathode which the focusing electrode is operated. material. Many photomultipliers are equipped with a focusing electrode, between the photocath- ode and the first dynode to provide optimum collection of the photoelectrons emitted from the photocathode. The focusing- electrode voltage is usually set at the point at which maximum anode output current is ob- tained. In some applications, spatial unifor- mity i.e., the variation of anode current with position of photocathode illumination, may be more important than maximizing output current. In such cases, however, the final ad- justment of the focusing-electrode potential should not differ significantly from the ad- justment that provides optimized collection efficiency. Fig. 46 shows a typical focusing- electrode characteristic. Photocathode Stability Stability and life of a photocathode is usually related approximately inversely to 92CS-32332 the current drawn from it. More stable and Fig.46 -Atypicalfocusing-electrodecharac- reliable performance results if small areas of teristic. concentrated illumination on the cathode surface are avoided. In normal operation of It is also possible to damage a photocath- photomultipliers, of course, the photocath- ode by maintaining a difference in potential ode current is usually small and damage to through the faceplate supporting the photo- the tube is more likely to be caused by the cathode, particularly at elevated tempera- amplified secondary emission currents in the tures. For example, if the photocathode in a latter dynode stages. photomultiplier is maintained at - 1000

42 Photomultiplier Characteristics volts and a ground plane at 0 volts is radiation for several photocathode types. established on the outside surface of the Some of this effect is the result of phosphor- glass, ionic conduction takes place through escence in the glass faceplate of the photo- the glass. Photocathode sensitivity will be multiplier, but a larger part of the effect is gradually deteriorated by the ionic apparently an excitation phenomena in the Photomultiplier tubes should be stored in the dark when not in use. Blue and ultra- violet radiation, especially, can cause photo- chemical changes in the photocathode which result in changes in sensitivity. It is especially important to avoid exposure to intense il- lumination such as sunlight even when no voltage is applied to the tube. Permanent damage may also result if the tube is exposed to radiation so intense that it causes ex- cessive heating of the photocathode. Tubes should not be stored for long periods at temperatures near the maximum rating of the tube; high temperatures almost always Fig. 47 - Variation of dark current following result in loss of sensitivity in the photocath- exposure of photocathode to cool white fluo- ode. rescent-lamp radiation. The various photo- A photomultiplier having a multialkali cathodes are identified by their spectral- photocathode (S-20 spectral response) tends response symbols. to lose sensitivity especially in the red por- tion of its spectral response upon extended When semitransparent photocathodes exposure to high ambient room lighting; the such as Cs3Sb and Ag-Bi-O-Cs (and prob- change is usually permanent. Contrary to ably others) are kept in the dark for many this behavior, photocathodes of the ex- hours, they become very resistive59, especial- tended red multi-alkali (ERMA) type ap- ly at reduced temperatures. The effect may parently do not exhibit this loss. be so great that the photomultiplier may ap- The Ag-O-Cs photocathode (S-l) spectral pear to have lost sensitivity. Apparently, response) also suffers a decrease in sensitivi- passage of current through the photosen- ty, particularly during operation, when ex- sitive layer as a result of normal operation posed to high radiant-energy levels normally restores the photocathode conductivity. At not harmful to other types of photocathode reduced temperatures, the process may re- materials. The decreased sensitivity occurs quire minutes or hours of operation to re- primarily in the infrared portion of the spec- turn the photocathode to a reasonable con- trum. Loss of infrared sensitivity may also ductivity. occur following long periods of storage. Photomultiplier tubes have been exposed The GaAs:Cs photocathode is particularly to gamma and X-radiation to an intensity of sensitive to responsivity loss even for 1010 roentgens per hour by the Naval Another effect related to the exposure of photocathode damage was noted except that photocathodes to excessive blue or ultra- faceplate discoloration was observed for ex- violet radiation, as from fluorescent room posure in excess of 104 roentgens. Glass lighting, is a temporary increase in photo- fluorescent effects were also noted during cathode dark emission current. The increase the tests. For applications where excessive may be as much as three orders of magnitude radiation may be present, it may be noted even from relatively short exposure. This in- that Corning has developed a non-browning crease in dark current occurs even though glass, Ce-doped No. 9025, which has been voltage is not applied to the tube and may used for special photomultipliers. persist for a period of from 6 to 24 hours As a result of enhanced dark count rates after such irradiation. Fig. 47 illustrates the observed in photomultiplier tubes used in recovery after exposure to fluorescent-lamp various earth-orbiting satellites, an in-

43 Photomultiplier Handbook vestigation was made by Viehmann et al61 In the design or operation of a multiplier on fluorescent and phosphorescent effects in phototube having a fixed supply voltage, the windows used in photomultiplier tubes when number of stages can be chosen so that the bombarded by beta rays. Source of the beta gain of the tube is maximum. For this pur- rays was an 0.8 millicurie beta emitter, pose, the optimum voltage per stage is that value at which a line through the origin (uni- to 2.23 MeV (megaelectron volts). A number ty gain on the log-gain scale) is tangent to the of special windows were tested. For Coming curve, as shown in Fig. 48. This point is 9741 glass, two decay constants were ob- served in the phosphorescense: 4.2 and 57 minutes. For an exposure of 9.5 x 1010 elec- trons per square centimeter in 30 minutes, an steradian)- 1 was observed in a photomulti- plier with an S-20 response, closely coupled to the glass plate.

GAIN-RELATED CHARACTERISTICS Gain vs. Voltage When several secondary-emission stages are coupled together, so that the secondary electrons from one become the primary elec- multiplier phototube is given by

92cs-32334 Fig. 48 - Log of gain as a function of volts In practice, some of the electrons may skip per stage for a tube (1P21) with Cs-Sb stages, or become lost to the amplification dynodes and for a tube (6342A) with Cu-Be process by impinging upon nonproductive dynodes. secondary-emission areas. It is customary to describe the gain of the multiplier phototube as a function of the ap- identified on the graph as the point of max- plied voltage. Fig. 48 shows two such curves imum gain per volt. (Note that this argument on a semilog scale. These curves illustrate the neglects the voltage used between the last wide range of amplification in a multiplier dynode and the anode and any discrepancy phototube. They also indicate the necessity resulting from nonuniform distribution of of providing a well regulated voltage supply voltage per stage). In most applications of for the dynode stages. multiplier phototubes, the tubes are It is possible to operate a four-to-six stage operated above the point of maximum gain photomultiplier so that each stage is at the per volt. When both the gain and the voltage voltage required for maximum secondary are presented on a logarithmic scale, the emission, as shown in Fig. 19. In such cases, resultant curve is then closely approximated the gain could be made practically indepen- by a straight line. Fig. 49 describes the anode dent of voltage over a small range. However, sensitivity in amperes per lumen and the such a condition would require approximate- typical amplification characteristics of a ly 500 volts per stage; thus the total voltage photomultiplier as a function of the applied required would be very high for the amount voltage. Curves of minimum and maximum of gain achieved. sensitivity are also shown.

44 Photomultiplier Characteristics

I-

SUPPLY VOLTS (E) BETWEEN ANODE AND CATHODE D- 92cs-32335 Fig. 49 - Typical anode sensitivity and ampli- fication characteristics of a photomultiplier tube as a function of applied voltage. Note the log-log scaling.

External Magnetic and Electrostatic Fields All photomultipliers are to some extent sensitive to the presence of external magnetic and electrostatic fields. These fields may deflect electrons from their normal path be- tween stages and cause a loss of gain. Tubes designed for scintillation counting are generally very sensitive to magnetic fields because of the relatively long path from the cathode to the first dynode; consequently, such tubes ordinarily require electrostatic 92CS-32336 and magnetic shielding. Magnetic fields may Fig. 50 - Curves for 3/4-inch diameter type easily reduce the anode current by a much as 4516 photomultiplier showing the effect on 50 per cent or more of the “no-field” value. anode current of magnetic fields parallel to the main axis of the tube and perpendicular The three curves in Fig. 50 show the effect to the main axis in the directions parallel and on anode current of magnetic fields parallel perpendicular to the dynodes. (Units for mag- to the main tube axis and perpendicular to netic field intensity are shown in both Sl the main axis in the directions parallel and units, ampere turns per meter, and conven- perpendicular to the dynodes. The curves are tional cgs units, oersteds. Note that 1 oersted

45 Photomultiplier Handbook

High-mu material in the form of foils or plier tubes contain some parts which have preformed shields is available commercially magnetic properties, but if they are for most photomultipliers. When such a neglected, the magnetic induction-or mag- shield is used, it must be at cathode poten- netic flux density B-measured in units of gauss would be numerically the same as the tial. The use of an external shield may pre- oersted values. in the case of Sl units for sent a safety hazard because in many ap- magnetic induction, 1 weber per square meter plications the photomultiplier is operated = 1 tesla = 104 gausses.) with the anode at ground potential and the cathode at a high negative potential. Ade- usually provided for one or more values of quate safeguards are therefore required to over-all applied voltage and indicate the prevent personnel from coming in contact relative anode current in per cent as a func- with the high potential of the shield. tion of magnetic field intensity. Fig. 51 It is possible to modulate the output cur- shows the variation of output current of rent of a photomultiplier with a magnetic several photomultiplier tubes as a function field. The application of a magnetic field of magnetic-field intensity directly parallel generally causes no permanent damage to a to the major axis of the tube. The magnitude photomultiplier although it may magnetize of the effect depends to a great extent upon those internal parts of a tube that contain the structure of the tube, the orientation of ferromagnetic materials (tubes are available the field, and the operating voltage. In which contain practically no ferromagnetic general, the higher the operating voltage, the materials). If tube parts do become magne- less the effect of these fields. tized, the performance of the tube may be

`-2400 -1600 -800 0 800 1600 2400 3200 MAGNETIC FIELD INTENSITY - AMPERE TURNS PER METER

I 1 I I I I I 1 -30 -20 -10 0 IO 20 30 40 OERSTEDS

92CM - 32337 Fig. 51- Variation of output current of several photomultiplier tubes as a function of magnetic-field intensity directly parallel to the major axis of the tube. Positive values of magnetic field are in the direction of the tube base. Operating voltages are indicated. The 931A is a circular-cage, side-on type. The 4902, 8053 and 8575 are 2-inch end-on types: teacup; Venetian-blind dynode; and in-line dynode structure, respectively.

46 Photomultiplier Characteristics degraded somewhat; however, the condition Table III provides guidance as to the max- is easily corrected by degaussing, a process in imum pulse current which can be drawn which a tube is placed in and then gradually from the anode of various types of photo- withdrawn from the center of a coil operated multipliers before spacecharge effects limit at an alternating current of 60 Hz with a linearity. Note that the type 8575, which has maximum field strength of 8000 ampere a focused dynode structure, provides the turns per meter (100 oersteds). highest withdrawal fields and highest linear output current. On the other hand, venetian- Linearity blind or box-and-slot dynodes have relative- Because the emission rate of photoelec- ly low withdrawal fields. Higher currents can trons is proportional to the incident radiant be obtained from all types of photomulti- flux, and the yield of secondary electrons for pliers by unbalancing the voltage distribu- a given primary electron energy is propor- tion to provide higher fields at the critical tional to the number of primary electrons, last stages. the anode current of a photomultiplier is Some photomultipliers used in applica- proportional to the magnitude of the inci- tions requiring high output pulse current use dent radiant flux. A linearity plot over a an accelerating grid between the last dynode wide range of light level is shown in Fig. 52 and the anode to reduce the effects of space- charge limiting. The potential of such a grid is usually between that of the last and the next-to-last dynode and is adjusted by ob- serving and maximizing the value of the anode output current. Another factor that may limit anode- output-current linearity is cathode resistivi- ty; cathode resistivity is a problem only in tubes with semitransparent photocathodes, particularly of the Cs-Sb or bialkali type. Linear behavior is not always obtained from photomultipliers even at low current levels. For example, if the test light spot on a 931A is not directed close to the center of the LIGHT FLUX -LUMENS active area of the photocathode, disturbing 92CS-32338 effects may arise from the proximity of the Fig. 52 - Range of anode-current linearity as ceramic end plates. Near the end plates, the a function of light flux for a 931A photomulti- fields are not uniform and are affected by plier. charge patterns on the insulator spacers, which change with the current level. An ex- for a type 931A. The limit of linearity occurs terior negative shield placed around the bulb when space charge begins to form. Space- wall may improve tube linearity by elimi- charge-limiting effects usually occur in the nating bulb charging effects. The passage of space between the last two dynodes. The excessive current may change the sensitivity voltage gradient between anode and last of the tube and cause an apparent non- dynode is usually much higher than between linearity. dynodes and, therefore, results in a limita- Some photomultipliers exhibit a tem- tion at the previous stage, even though the porary instability in anode current and current is less. The maximum output cur- change in anode sensitivity for several rent, at the onset of space charge, is propor- seconds after voltage and light are applied. tional to the 3/2 power of the voltage gra- This instability, sometimes called hysteresis dient in the critical dynode region. By use of because of cyclic behavior, may be caused by an unbalanced dynode-voltage distribution electrons striking and charging the dynode increasing the interstage voltages near the support spacers and thus slightly changing output end of the multiplier, it is possible to the electron optics within the tube. Sensitivi- increase the linear range of output current. ty may overshoot or undershoot a few per

47 Photomultiplier Handbook cent before reaching a stable value. The time behavior may be a problem in applications to reach a stable value is related to the such as where a photomultiplier resistance of the insulator, its surface is used in a constant-anode-current mode by capacitance, and the local photomultiplier varying the photomultiplier voltage as the current. This instability and non-linear light input changes.

Table III Maximum Anode Currents That Can be Utilized in Various Photomultiplier Types Without Serious Loss of Linearity from Space-Charge Build-up.

Tube Type Supply Voltage Maximum Volts Distribution* Output Current (mA) with Linearity Loss Less Than 10%

931A 1000 1 1 (side-on, circular cage) 4524 1500 5 (3-inch, venetian- blind dynodes)

4900 1200 5 (3 -inch, Tea-cup type) 8575 1760 30 (2-inch, linear fo- cused dynode struc- ture)

8575 2100 120 C-31059 1050 4 (1 1/8-inch, box and slot dynodes)

C-31059 1150 12

*These number series represent the relative voltage applied between cathode and first dynode, first dynode and second dynode,. . . . . ,last dynode and anode. Note that in the case of the 8575 and the C-31059, a second voltage distribution is given with increased voltage drop in the last stages providing higher output currents before the onset of space-charge-limiting conditions.

48 Photomultiplier Characteristics

Hysteresis has been eliminated in many with Temperature. The gain data were ob- tubes by coating the dynode spacers with a tained by measuring both the anode response conductive material in the manufacturing variation and the photocathode response process and maintaining the coating at a variation with temperature and dividing out fixed potential. Tubes treated by this method the latter to obtain only the gain variation. It assume final sensitivity values almost im- is probable that the increase in gain at low mediately upon application of light and temperatures is the result of a decrease in voltage. energy loss from lattice scattering. Although Peak linear and saturation currents are the variation in over-all gain is fairly large, it usually measured by pulsed methods. One should be remembered that the variation common method makes use of a cathode-ray results from a small variation in each stage tube with a P15 or P16 phosphor. The grid is compounded by the 9 stages. At room tem- double-pulsed with pulses of unequal ampli- perature, the average gain per stage for the tudes but fixed amplitude ratio. As the 8571 is 4.5 at 100 V per stage and the percen- amplitude of the two pulses is increased, a tage change in stage gain with temperature is point is observed at which the amplitude of approximately -0.06% per °C. Thus the the larger of the anode pulses does not in- secondary emission temperature coefficient crease in the same proportion as the smaller is generally less that the photoresponse pulse. At this point the tube is assumed to temperature coefficient, both for Cs-Sb sur- become non-linear. The current value at this faces. See Fig. 37. point is then measured by means of an oscilloscope and load resistor. The max- LOX imum saturation current is found when a further increase in radiation level yields no further increase in output. Although very high anode current can be drawn from photomultiplier tubes, it should be emphasized that stability cannot be expected at such levels. These high cur- rent levels are principally of interest in light- pulse applications. In this case, the stability of the tube is approximately that of the tube operated at the integrated average current level t Gain Variation with Temperature It is often advisable to reduce the ambient temperature of a photomultiplier in order to reduce dark emission from the photocathode and improve the signal-to-noise ratio of the measurement. Variations of the spectral characteristics of photocathodes with temperature have been noted in the previous section on Photocathode Stability. There are TEMPERATURE - °C numerous reports on the over-all variation 92cs-32341 of the photomultiplier output current with Fig. 53 - Typical variation of gain with tem- temperature including the effect of tempera- perature for a 9-stage photomultiplier tube (type 8571) with Cs-Sb dynodes operating at ture on photocathode sensitivity62-64. Fig. 100 volts per stage. 53 is a typical characteristic for type 8571 photomultiplier having Cs-Sb dynodes showing the variation of gain with A few words of caution regarding the am- temperature. These data do not include the bient temperature of photomultipliers are variation of photocathode response with pertinent. A maximum ambient tempera- temperature which has been described in the ture, and in some instances a minimum tem- section on Variations in Spectral Response perature, is specified for all photo-

49 Photomultiplier Handbook

multipliers. The specification of maximum maximum anode current is restricted to a ambient temperatures reduces the possibility few milliamperes when the tube is operated of heat damage to the tube. Cesium, for ex- at 100 to 200 volts between the last dynode ample, is very volatile and may be and the anode. Many photomultipliers are redistributed within the tube causing loss of rated for only 0.1 mA or less. secondary emission gain or loss of cathode Operating a photomultiplier at an ex- sensitivity. cessively high anode current results in an in- It is recommended that photomultipliers creased fatigue that occurs as the average be operated at or below room temperature so anode current increases. The loss in sensitiv- that the effects of dark current are mini- ity occurs as a result of a reduction in the mized. The variation of dark current, or secondary emission, particularly in the last noise, is most important because of its effect stages of the photomultiplier where the cur- on ultimate low-light-level sensitivity. rents are the highest. Various cryostats and solid-state thermionic Tube fatigue or loss of anode sensitivity is coolers have been designed that reduce dark a function of output-current level, dynode current at low temperatures in low-light-level materials, and previous operating history. applications. An important consideration in The amount of average current that a given the use of these devices is to prevent conden- photomultiplier can withstand varies widely, sation of moisture on the photomultiplier even among tubes of the same type; conse- window. A controlled low-humidity atmos- quently, only typical patterns of fatigue may phere or special equipment configuration be cited. may be necessary to prevent such condensa- The sensitivity changes are thought to be tion. the result of erosion of the cesium from the Another reason for avoiding the operation dynode surfaces during periods of heavy of photomultipliers at extremely low temper- electron bombardment, and the subsequent atures is the possible phase change that this deposition of the cesium on other areas type of operation may cause in some of the within the photomultiplier. Sensitivity losses metal parts. These changes are particularly of this type, illustrated in Fig. 54 for a 1P21 probable when Kovar is used in metal-to- glass seals. Tubes utilizing Kovar in their construction should not be operated at tem- peratures below that of liquid nitrogen (-196°C). In some tubes, particularly those with multialkali photocathodes, it is sometimes observed that the noise actually increases as the temperature of the photocathode is re- duced below about -40°C. The reason for this noise increase is not understood. How- ever, most of the dark-current reduction has already been achieved at temperatures above -40°C. In general, it is recommended that all and connections to the tube be encap- sulated for refrigerated operation. Encap- sulation minimizes breakdown of insulation, TIME-MINUTES especially that caused by moisture condensa- 92cs-32339 tion. Fig. 54 - Short-time fatigue and recovery Stability and Gain characteristics of a typical 1P21 operating at 100 volts per stage and with a light source ad- All photomultipliers have a maximum justed to give 100 microamperes initial anode anode current rating. The primary reason current. At the end of 100 minutes the light is for such a rating is to limit the anode power turned off and the tube allowed to recover dissipation to approximately one-half watt sensitivity. Tubes recover approximately as or less. Consequently, the magnitude of the shown, whether the voltage is on or off.

50 Photomultiplier Characteristics operated at an output current of 100 micro- amperes, may be reversed during periods of non-operation when the cesium may again return to the dynode surfaces. This process of recovery may be accelerated by heating the photomultiplier during periods of non- operation to a temperature within the max- imum temperature rating of the tube; heating above the maximum rating may cause a permanent loss of sensitivity. Sensitivity losses for a given operating cur- rent normally occur rather rapidly during in- itial operation and at a much slower rate after the tube has been in use for some time. Fig. 55 shows this type of behavior for a

TIME -HOURS

Fig. 56 - Typical responsivity variation on life for a photomultiplier having silver-magne- sium dynodes. Initial anode current was 2 mil- liamperes and was readjusted to this operat- ing value at 48, 168, and 360 hours. 2 0 100 200 300 400 5 TIME -HOURS operation, after which a very gradual de- 92cs-32340 crease takes place, as illustrated in Fig. 56. Fig. 55 - Typical sensitivity loss for a 1P21 The operating stability of a photomulti- operating at 100 volts per stage for a period of plier depends on the magnitude of the 500 hours. Initial anode current is 100 micro- average anode current; when stability is of amperes and is readjusted to this operating prime importance, the use of average anode value at 48, 168, and 360 hours. currents of 1 microampere or less is recom- mended. 1P21 having Cs-Sb dynodes, operating at an In addition to the life characteristics, output current of 100 microamperes. Tubes which are probably the result of changes in operated at lower current levels, of the order the dynode layer itself, other changes of a of 10 microamperes or less, experience less temporary nature also occur. Not all these fatigue than those operated at higher cur- changes are well understood; some are rents, and, in fact, may actually recover charging of insulators in the tube. from high-current operation during periods Fig. 57 illustrates one of the peculiar in- of low-current operation. stabilities which are sometimes observed in Fatigue rates are also affected by the type photomultiplier tubes. When the light is first of dynode materials used in a tube. Copper turned on, the current apparently overshoots beryllium or silver magnesium dynodes are and then decays to a steady value. This par- generally more stable at high operating cur- ticular phenomenon is probably the result of rents than the cesium antimony types. The the charging of the supporting insulator for sensitivity for tubes utilizing these dynodes the dynodes. The effect is observed to occur very often increases during initial hours of more rapidly at higher currents, presumably

51 Photomultiplier Handbook

because of the greater charging currentIn order. to investigate the phenomena of Electrons striking the insulator probablpulse-heighty variation with pulsecount rate, result in secondary emission and a resultana purposelyt exaggerated experiment was positive charge. The change in potentialdevised. af- Instead of a scintillating crystal, a fects the electron optics in the space between pulsed cathode-ray tube was used as a light dynodes. The effect is observed as source.an in- Two pulse rates were studied: 100 crease in some tubes and as a decreaseand 10,000 in pulses per second. Pulse duration others. When the photomultiplier is designed was one microsecond; decay time to 0.1 with metal shields or with conductivmaximume was 0.1 microsecond. The experi- coatings on the critical areas of thement in- was devised to study the rate at which sulators, this effect is eliminated. the photomultiplier tube output response changed when the pulse rate was suddenly switched between the two rates. Tubes such as the 6342A, and especially the 8053 and 8054, showed practically no effect (of the order of one per cent or less). Fig. 59 shows

TIME - SECONDS

92cs-32343 Fig. 57 - Sudden shift in anode current prob- ably as the result of insulator spacer charg- ing. Observation was made using an experi- mental photomultiplier in which the effect was unusually large. Fig. 59 - Variation of output pulse height as the rate of pulsing is changed in a poorly A related phenomenon is the variation ofdesigned experimental tube. Light pulses pulse height with pulsecount rate inwere provided from a cathode-ray tube. At the scintillationcounting applications. Thusleft, of the graph, which shows the pulse- when a radio-active source is brought closeramplitude envelope with time for the output to a scintillating crystal a greater rate of scin-of the photomultiplier tube, the pulses are at 100 per second. The pulse rate is increased tillations should be produced, all having thesuddenly to 10,000 per second and again re- same magnitude. In a particular photomulti-duced as indicated. Changes in amplitude are plier a few per cent change in amplitude mayprobably the result of insulator charging. result and cause problems in measurement. Fig. 58 shows the typically minor variation of pulse height with pulse-count rate for thea pulses during the switching procedure for type 6342A multiplier phototube. an experimental tube. The phenomena were completely reversible and were observed (to a lesser extent) on many different tube types. The time-decay period of several seconds suggests the charging of an insulator spacer to a new potential as the result of the in- creased charge flow and the subsequent

0.4 0.6 0.8 modification of interdynode potential fields. COUNTING RATE In scintillation counting it is particularly 92cs-32344 important that the photomultiplier have very Fig. 58 - Typical variation of pulse height good stability. There are two types of gain with pulsecount rate for a 6342A. (13 3Cs stability tests which have been used to source with a Nal:Tl source). evaluate photomultipliers for this applica-

52 Photomultiplier Characteristics tion: (1) a test of long-term drift in pulse- Life Expectancy height amplitude measured at a constant The life expectancy of a photomultiplier, counting rate; and (2) a measure of short- although related to fatigue, is very difficult term pulse-height amplitude shift with to predict. Most photomultipliers will func- change in counting rate. tion satisfactorily through several thousand In the time stability test, a pulse-height hours of conservative operation and propor- analyzer, a 137Cs source, and a NaI(T1) tionally less as the severity of operation in- crystal are employed to measure the pulse creases. Photomultipliers do not have height. The 137Cs source is located along the elements which “burn out” as in the case of major axis of the tube and crystal so that a a filament in a . Furthermore, count-rate of 1000 counts per second is ob- loss of sensitivity which occurs with opera- tained. The entire system is allowed to warm tion tends to recover during idle periods or up under operating conditions for a period during conservative operation. of one-half to one hour before readings are Factors which are known to affect life recorded. Following this period of stabiliza- adversely are high-current operation, tion, the pulse height is recorded at one-hour excessive-voltage operation, high photocath- intervals for a period of 16 hours. The drift ode illumination, and high temperature. rate in per cent is then calculated as the mean Operation of photomultipliers in regions gain deviation (MGD) of the series of pulse- of intense nuclear radiation or X-rays may height measurements, as follows: result in an increase in noise and dark cur- rent as a result of fluorescence and scintilla- tion within the glass portions of the tubes. Continued exposure may cause darkening of the glass and a resultant reduction in transmission capability. DARK CURRENT AND NOISE where p is the mean pulse height, pi is the pulse height at the ith reading, and n is the The lower limit of light detection for a total number of readings. Typical maximum photomultiplier tube is determined in many mean-gain-deviation values for photomulti- cases by the electrical noise associated with pliers with high-stability Cu-Be dynodes are the anode dark current. There are several usually less than 1 per cent when measured sources of dark current in a photomultiplier. under the conditions specified above. Gain These sources are described below, stability becomes particularly important Sources of Dark Current when photopeaks produced by nuclear Dark current in a photomultiplier tube disintegrations of nearly equal energy are be- may be categorized by origin into three ing differentiated. types: ohmic leakage, dark or “thermionic” In the count-rate stability test, the photo- emission of electrons from the cathode and multiplier is first operated at 10,000 counts other elements of the tube, and regenerative per second. The photopeak counting rate is effects . then decreased to 1000 counts per second by Ohmic leakage, which results from the im- increasing the source-to-crystal distance. perfect insulating properties of the glass The photopeak position is measured and stem, the supporting members, or the plastic compared with the last measurement made base, is always present. This type of leakage at a counting rate of 10,000 per second. The is usually negligible, but in some tubes it may count-rate stability is expressed as the become excessive because of the presence of percentage gain shift for the count-rate residual metals used in the processing of the change. It should be noted that count-rate photocathode or the dynodes. Condensation stability is related to the hysteresis effect of water vapor, dirt, or grease on the outside discussed above. Photomultipliers designed of the tube may increase ohmic leakage for counting stability may be expected to beyond reasonable limits. Simple precau- have a value of no greater than 1 per cent tions are usually sufficient to eliminate this gain shift as measured by this count-rate sort of leakage. In unfavorable environmen- stability test. tal conditions, however, it may be necessary

53 Photomultiplier Handbook

to coat the base of the tube with moisture- be measured on a dc meter, is usually the resisting materials, which may also prevent principal dc component of the dark current external arc-overs resulting from high at normal operating voltages. The limitation voltage. to the measurement of very low light levels is Ohmic leakage is the predominant source the variable character of the thermionic of dark current at low-voltage operating dark-current component. It is not possible to condition. It can be identified by its propor- balance out this wide-band noise component tionality with applied voltage. At higher of the photomultiplier tube, as it might be to voltages, ohmic leakage is obscured by other balance out a steady ohmic-leakage current. sources of dark current. Nevertheless, it is usually advantageous to Fig. 60 shows the typical variation of dark operate the photomultiplier tube in the range where the thermionic component is domi- nant. In this range, the relationship between sensitivity and noise is fairly constant as the voltage is increased because both the photo- electric emission and the thermionic emis- sion are amplified by the same amount. Typical dark emission current densities for various photocathodes are given in Table I (page 16). The resulting anode dark current may be estimated by multiplying the dark emission per unit area at the photocathode by the photocathode area and by the gain of the photomultiplier tube at the desired operating voltage. The thermionic component of the dark current varies in a regular way with temperature as illustrated in Fig. 16, and because the thermionic component of the dark current is a source of electronic noise in the anode circuit, it is frequently advan- VOLTS PER STAGE tageous to cool the photomultiplier and take 92CS-32346 advantage of the reduced dark current and Fig. 60 - Typical variation of dark current noise.Various cryostats have been de- with voltage for a multiplier phototube. signed62,66providing low temperature operation of photomultipliers. One practical current of a photomultiplier tube as a func- consideration is the prevention of condensa- tion of applied voltage. Note that in the mid- tion of moisture on the window. In a Dewar- range of voltage, the dark current follows type arrangement, condensation may not be the gain characteristic of the tube. The a problem; in simpler set-ups moisture con- source of the gain-proportional, dark cur- densation may be prevented by a controlled rent is the dark or thermionic emission of low-humidity atmosphere at the external electrons primarily from the photocathode. window. On some types of photocathode, Because each electron emitted from the too cool a temperature may result in the photocathode is multiplied by the secondary- photocathode becoming so resistive that the emission gain of the tube, the result is an photoemission is blocked by a drop in poten- output pulse having a magnitude equal to the tial across the photocathode surface. See charge of one electron multiplied by the gain earlier section on Current-Voltage Charac- of the tube. (There are statistical amplitude teristics. Commercial cryostats or cooled variations which will be discussed later.) photomultiplier chambers are available Because the emission of thermionic electrons designed especially for photomultiplier is random in time, the output dark current operation67. consists of random unidirectional pulses. At higher dynode voltages, a regenerative The time average of these pulses, which may type of dark current develops, as shown in

54 Photomultiplier Characteristics

Fig. 60 . The dark current becomes very er- measurement. If the type of modulation and ratic, and may at times increase to the prac- bandwidth used in the measurement is tical limitations of the circuit. Continued known, an equivalent noise input, ENI, can flow of large dark currents may cause be calculated from the signal-to-noise ratio, damage to the sensitized surfaces. Some Equivalent noise input is defined as the value possible causes of the regenerative behavior of incident luminous or radiant flux which, will be discussed in more detail later. All when modulated in a stated manner, pro- photomultiplier tubes eventually become duces an rms output current equal to the rms unstable as the gain is increased. noise current within a specified bandwidth, Dark-Current Specification usually 1 Hz. Dark-current values are often specified at a particular value of anode sensitivity rather than at a fixed operating voltage. Specifica- tions of dark current in this manner are more closely related to the actual application of the photomultiplier. The best operating range for a given photomultiplier can usually be predicted from the quotient of the anode dark current and the luminous sensitivity at which the dark current is measured. This quotient is identified as the Equivalent Anode Dark Current Input (EADCI) in the Technical Data for individual photomultiplier tubes; and is the value of radiant flux incident on the photocathode required to produce an anode current equal to the dark current observed. The units used in specifying EAD- CI are either lumens or watts at the wave- length of maximum cathode responsivity or watts at a specified wavelength. 92c3-32347 The curves in Fig. 61 shows both typical Fig. 61 - Illustrative data showing the varia- anode dark current and equivalent anode tion of anode dark current and the equivalent dark current input (EADCI) as functions of anode-dark current input (EADCI) as a func- luminous sensitivity. The optimum operat- tion of luminous sensitivity for a type 8575. ing range occurs in the region of the Operation of the tube at voltages higher than that for the minimum of the EADCI character- minimum on the EADCI curve, the region in istic does not provide the best signal-to-noise which the signal-to-noise ratio is also near its ratio. maximum. The increase in the EADCI curve at higher values of sensitivity indicates the onset of a region of unstable and erratic Noise Equivalent Power-Another way to operation. Many curves of this type also in- categorize the limit of detection of a clude a scale of anode-to-cathode supply photomultiplier is by noise equivalent power voltage corresponding to the sensitivity (NEP) which is essentially the same as EN1 scale. except the units are always in watts. NEP is Equivalent Noise Input-The dark current the radiant flux in watts at a specified wave- in a photomultiplier is the average current length incident on the detector which gives a value of the output pulses occurring at ran- signal-to-noise ratio of unity. The frequency dom intervals plus the dc leakage current. bandwidth (usually 1 Hz) and the frequency Fluctuations or noise associated with these at which the radiation is chopped must be pulses limit precision of measurement, specified as well as the spectral content of rather than the particular dark current value. the radiation (most often, monochromatic Noise from a photomultiplier may be radiation at the peak of the detector evaluated in terms of a signal-to-noise-ratio response). It should be noted that NEP is

55 Photomultiplier Handbook

frequently specified in units of watts effect of various voltages applied to an exter- HZ-1/2. The numerical value of this for- nal shield around the tube envelope of a mulation of NEP is the same as that given in 1P21 photomultiplier. The graph shows the units of watts but with a specified bandwidth equivalent noise input decreasing as the of 1 Hz. shield is made negative toward the-potential Detectivity-Detectivity (D) is the recipro- of the photocathode. It should be noted also cal of NEP; it is expressed in W - 1. Detectiv- that an actual contact is not always required ity is a figure of merit providing the same in- to produce the effects noted in Fig. 62. The formation as NEP but in the reverse sense so proximity of a positive potential near the that the lower the radiation level to which glass bulb can cause a noisy operation. the photodetector responds, the higher the detectivity. 4 Regenerative Effects Dynode Glow. Although photomultipliers are designed to minimize regenerative ef- fects, at some high voltage and gain almost all photomultipliers exhibit breakdown phenomena. One source of regeneration in photomultipliers is the glowing of the dynodes under electron bombardment.68 The glow has a blue spectral emission and of course is most prominent in the latter stages where the current is highest. The regenera- tive effect occurs when the light from the Fig. 62 - Effect of external-shield potential dynode glow is scattered and reflected back on the noise of a 1P21 photomultiplier. Note to the photocathode. Dynode cage shields the desirability of maintaining a negative bulb and opaque support wafers minimize this ef- potential. fect . Glass Charging Effects. Regenerative pho- Operation of a photomultiplier tube with tomultiplier currents may also be triggered an improper external shield may not only by the electrostatic potential of the bulb cause an increase in noise or lead to an elec- walls surrounding the dynode or photocath- trical breakdown of the tube, but can result ode structure. Particularly when the poten- in damage to the photocathode and reduced tial of the bulb is near anode potential, stray tube life. In order to prevent these effects, electrons may be attracted to the bulb and the envelope wall should be maintained near cause the emission of light on impact, de- photocathode potential by wrapping or pending upon the nature of the glass surface painting it with conductive material and con- and the presence of contamination. Secon- necting this material to cathode potential. dary electrons resulting from the impact of The connection is usually made through a stray electrons on the glass surface are col- high impedance to reduce the shock hazard. lected by the most positive elements in the If a cathode potential shield is not provided, tube and help maintain the positive potential the glass surface in the vicinity of the photo- of the inner surface of the glass. Under these cathode must be insulated from any source circumstances, it is possible to observe the of potential difference so that leakage cur- formation of glowing spots on the inside of rents to the bulb are less than 10 -12 ampere. the glass bulb, provided the eye is dark In photomultiplier tubes in which the pho- adapted and the applied voltage is sufficient- tocathode is of a transmission type, on the ly high. Some of this fluorescent light may inside surface of the glass bulb, it is par- be reflected back to the photocathode and ticularly important to avoid a positive result in an increase in the photocathode voltage contact on the external surface of the dark (or light) emission. photocathode window. In this case, ionic Shielding. The effect just described can be currents can flow through the glass and pro- minimized by controlling the external poten- duce a fluoresence and an accompanying tial of the glass envelope. Fig. 62 shows the noisy photocathode current.

56 Photomultiplier Characteristics

It should also be noted that continued ion; this afterpulse occurs approximately 300 operation of a photomultiplier tube with a nanoseconds after the main pulse in a tube positive voltage contact to the glass in the of type 7850 construction. One source of photocathode area can cause a permanent hydrogen in the tube is water vapor absorbed damage to the photocathode. The damage is by the multiplier section before it is sealed to reported to be the result of ionic conduction the exhaust system. Other gases which may through the glass and poisoning of the pho- cause afterpulsing may be present as a result tocathode by ions.69 of outgassing of the photomultiplier parts Afterpulsing. Afterpulses, which may during processing or operation. Present pho- be observed when photomultipliers are used tomultiplier processing techniques are to detect very short light flashes as in scin- designed to eliminate or at least to minimize tillation counting or in detecting short laser the problem of afterpulsing. pulses, are identified as minor secondary Helium Penetration71,72Another effect pulses that follow a main anode-current must be considered in relation to the sources pulse. There are two general types of after- of dark current-the penetration of helium pulses; both are characterized by their time through the glass of photomultiplier tubes. of occurrence in relation to the main pulse. When the photomultiplier is operated or The first type results from light feedback stored in an environment where helium is from the area of the anode, or possibly cer- present, helium will gradually permeate tain dynodes, to the photocathode; the in- through the glass envelope. Because helium tensity of the light is proportional to the tube is inert, it does not react with the photocath- currents. When this light feedback reaches ode or dynode surfaces. But tubes subjected the cathode, the afterpulse is produced. to such an environment will exhibit a noise Afterpulses of this type, characterized by a increase and an increase in afterpulsing delay in the order of 40 to 50 nanoseconds, because of the ionization of the helium by may be a problem in many older photomulti- electron impact. Depending upon the degree pliers having open dynode structures. The of permeation, a point will be reached at time delay experienced with this type of which complete ionization and electrical afterpulse is equal to the total transit time of breakdown occurs making the tube unus- the signal through the photomultiplier plus able. the transit time of the light that is fed back. Other Noise Sources The second type of afterpulse has been Excess noise or dark current can also re- shown to be the result of ionization of gas in sult from field emission occurring within the the region between the cathode and first tube and from scintillations in the glass dynode. The time of occurrence of the after- envelope of the tube caused by radioactive pulse depends upon the tube dimensions, the elements within glass (most glasses contain type of residual gas involved and the mass of some radioactive 40K). Fused silica is the gas ion, but usually ranges from 200 sometimes utilized in photomultiplier face- nanoseconds to well over 1 microsecond plates to minimize these effects. after the main pulse. When the ion strikes Noise in photomultiplier tubes can also the photocathode, several secondary elec- result from the proximity of nuclear sources trons may be emitted; thus, the resulting or from cosmic rays which result in glass afterpulse has an amplitude equal to several scintillations. Andrew T. Young73, * has electron pulse-height equivalents. These identified large pulses which originate from pulses appear to be identical to the larger Cerenkov light flashes produced by cosmic dark-current pulses, and it is suspected that rays traversing the window of end-on photo- many of the dark-current pulses are the re- multipliers. The flashes correspond to sult of photocathode bombardment by gas ions. *Andrew T. Young also has written a chapter, “Photo- Several gases, including N2 + and H2 + , multipliers: Their Cause and Cure,” in Volume 12 of are known to produce afterpulses. Each gas Methods of : , L. Marton, Editor, Academic Press, 1974. As well as being produces its own characteristic delay follow- a good general reference on photomultipliers, this chap- ing the main pulse. The most troublesome, ter contains a further discussion of the Cerenkov-light- perhaps, is the afterpulse caused by the H2 + flash effect.

57 PhotomultIplier Handbook

photoemission pulses of 50 electrons and or near the maximum operating voltage. larger. The number of pulses is greater when This process, called dark aging, may require the face of the photomultiplier is upward several hours to several days. After such a rather than downward because the Cerenkov process of aging, it is recommended that a radiation is emitted in a conical pattern away photomultiplier be operated for several from the direction of entry. Young reports minutes at the reduced voltage before mea- the total number of pulses from this source surements are attempted. to be about 1.2 min - 1 cm -2. Dark Noise Reduction with Cooling. Another phenomenon which deserves Because the dark emission is reduced as the mention in connection with dark current is temperature of the photocathode is reduced, the effect of previous exposure to light, the dark noise output of the photomultiplier especially blue or near ultraviolet. Large in- may also be reduced by cooling. Fig. 63 creases in photocathode dark emission may occur as discussed in section on Photocath- ode Stability with rather slow recovery, as il- lustrated in Fig. 47. It is advisable, there- fore, that photomultipliers be kept in the dark at all times, or at least for many hours before they are used for making low-level measurements. Noise Output of a Photomultiplier The output current of a photomultiplier consists of a train of unidirectional electrical pulses whether the tube is in the dark or with illumination on its photocathode. Each pulse is the result of an electron emitted from the photocathode and amplified by the secon- dary emission of the tube. (Some pulses may TUBE originate also from light striking the first dynode or from thermionic emission from Fig. 63 - Equivalent noise input in lumens for a 1P21 photomultiplier as a function of tem- the first or other dynodes.) In a system hav- perature. ing a wide bandwith, the individual pulses may be measured and counted. Their spac- shows the variation of equivalent noise in- ing in time will have a statistical variation as puts, ENI, for a 1P21 (opaque Cs3Sb photo- will their height. There variations constitute cathode) over a wide range of temperature. noise and limit the precision of measure- The implication of these data is that by cool- ments. ing from room temperature to - 150 °C the If the bandwidth is not sufficient to low light level limit of detection can be resolve the individual pulses, particularly reduced by two orders of magnitude. when the light level is relatively high, the output will exhibit noise or rapid variation in Photomultiplier Noise Characteristics signal level about an average value. The The following paragraphs describe the noise level relative to the signal level will noise and signal levels from both a pulse and decrease with increasing signal level or with a dc point of view. The noise frequency spec- decreasing bandwidth. The noise may be de- trum is also described and data are presented scribed by giving the rms value of the current showing the dark noise spectrum. variation in a specified bandwidth. At very low signal levels, the detection limit will be shown to be determined by the Dark-Current and Noise Reduction dark current of the photomultiplier and its Dark-Current Reduction. If care is taken associated noise. At high levels, the precision to avoid damage to the photomultiplier by of measurement is limited by the statistical operation with excessive current, the dark variation in the signal pulses-or the noise current can often be reduced by a process of which is always present to some degree in the operating the photomultiplier in the dark at dc signal level. The signal-to-noise ratio can Photomultiplier Characteristics be improved by decreasing the bandwidth in distribution of heights even though each dc measurements or by the equivalent of in- pulse is initiated by a single electron. Fig. 65 creasing the time in pulse-counting applica- tions. Noise Spectrum. The width of an output current pulse initiated by an electron from the photocathode is determined by the varia- tions in transit time through the tube. The noise associated with these pulses of electron current is flat with frequency out to frequen- cies corresponding to the width of the in- dividual pulses. (The frequency spectrum of a delta function is flat.) A power spectrum of the noise from type 931 has been calcu- lated by R. D. Sard74 and is shown in Fig. 64. Sard’s calculation was done by considering

2

PULSE HEIGHT- PHOTOELECTRON EQUIVALENTS

92CM-32350

FREQUENCY - MHz 92cs-32349 Fig. 65 - Typical Dark-Pulse Spectrum

Fig. 64 - Spectral energy distribution of the shows a differential dark-noise pulse spec- noise from type 931A operated at 100 volts per trum-the number of pulses counted per stage, as calculated by R. D. Sard.74 unit of time as a function of their height. A differential dark-noise spectrum is ob- the anode current from a single electron tained with a multichannel pulse-height traversing the space between the last two analyzer. The calibration of the single- dynodes and anode and the distribution photoelectron pulse height is determined by of electron arrival times due to differences in illuminating the photocathode with a light transit time between stages; then, the Fourier level so low that there is a very low probabili- transform of the pulse shape was taken to ty of coincident photoelectron emission. The obtain the frequency spectrum. dark-pulse distribution is then subtracted The bandwidth of the noise spectrum dif- from the subsequent combination of dark fers in different tube designs, depending pulses and single-photoelectron pulses, so upon transit-time variations. Usually, high- that the remainder represents only that speed photomultipliers are characterized by distribution resulting from single-photoelec- rise and fall times of the output pulses rather tron events. By adjusting the gain of the than by bandwidth. This subject is discussed pulse-height analyzer, the single-electron in more detail in a later section. photopeak can be placed in the desired chan- Dark Noise Pulse Spectrum. Because of nel to provide a normalized distribution. the statistical nature of the secondary emis- The dark-pulse spectrum of Fig. 65 is sion gain at each stage of the photomulti- characteristic of photomultipliers intended plier, the output pulses have a fairly wide for use in scintillation counting and other

59 Photomultiplier Handbook low-light-level pulse applications. The curve where e is the charge on the electron and B is shown is idealized and represents an average the bandwidth of the observation, or typical spectrum. An actual spectrum The photocathode dark emission and its shows statistical variations depending upon associated noise are both amplified by the the length of the count. The slope of the curve for the pulse-height for the moment that the amplification pro- region between 1 and 4 photoelectrons is as cess is noise free, and that all of the current expected for single-electron emission, when emitted from the photocathode is collected the statistical nature of secondary-emission by the first dynode, the anode dark current is multiplication is considered. The number of given by pulses in this region may be reduced by cool- ing the photomultiplier. Below a pulse height (17) of one photoelectron equivalent, the curve is determined partly by the statistical spread and the anode rms noise current is given by due to the multiplication process and partly from emission from some of the dynode surfaces. (18) The slope of the curve for the pulse-height Actually, noise is introduced by the secon- region greater than 4 photoelectrons is dary emission process. (This subject is dis- presumed to be caused by multiple-electron- cussed at length in Appendix G.) If one as- emission events. These multiple pulses are sumes Poisson statistics for the secondary caused by processes such as ionic bombard- emission process, the anode noise current ment of the photocathode. Other mecha- shown in Eq. 18 should be increased by a nisms contributing to the noise spectrum in- clude cosmic rays,field emission, and dary emission ratio per stage, assumed to be radioactive contaminants that produce scin- the same for all stages. If a typical value of tillations within the glass envelope. Cooling has little effect upon reduction of the factor of 1.15. Actual measurements indi- number of these multiple electron pulses, but cate a somewhat higher figure. On the other extended operation of the tube may improve hand, higher voltage on the first stage, or the performance. Operation of the tube may use of a GaP:Cs first dynode having a very result in erosion of sharp points and reduce high secondary-emission ratio would reduce the possible contribution of field effects. In this factor. For the purpose of the present addition, improvement occurs because discussion, this increased noise from secon- residual gases are absorbed within the tube, dary emission will be neglected and the ion bombardment of the photocathode is photomultiplier anode noise will be taken as reduced, and the resulting multiple electron the amplified photocathode shot noise as emission is lessened. given by Eq. 18. For many applications it is useful to have A fundamental advantage of a photomul- a summation of the total number of dark tiplier as compared with a photodiode, is the pulses. In Fig. 65, for example, the sum of high gain and the fact that the secondary dark-pulse counts from 1/8th electron emission process contributes very little to the equivalent height to 16 electron equivalent relative noise output of the tube. The high . heights is 4 x 104 . gain of the photomultiplier permits the use Analytical Model of Noise and Signal-to- of a relatively small load resistance (R) Noise Ratio-Consider the lower limit of without deterioration of the photomultiplier light detection capability of a photomulti- signal-to-noise ratio by the Johnson thermal plier as determined by the fluctuation in the noise of the load resistance. The small load thermionic dark emission. If the dark emis- resistance permits an operation of the photo- sion current from the photocathode is id, the multiplier with the very high bandwidths in- rms shot noise associated with this current is herent in the photomultiplier design. A large given by load resistance shunted by the finite tube and . lead capacitance would otherwise seriously (16) limit the effective bandwidth of the system.

60 Photomultiplier Characteristics

In order to maintain the fundamental tion and cable used in pulse applica- signal-to-noise capability of the photomulti- tions-the two noise sources are about plier, the Johnson noise of the load resis- equal. Therefore, at this point in the exam- tance must be less than the photomultiplier ple chosen there is some, but not a serious, output noise. Johnson noise (rms) is given by reduction of signal-to-noise ratio. Noise in Signal. When the photocathode current representing the signal is larger than the photocathode dark emission, the domi- where k is Boltzmann’s constant and T is the nant noise is the noise in the signal. For a temperature in Kelvin units. In order to com- photocathode signal current, is, the anode pare this noise with the photomultiplier out- signal current Is, is given by put noise we may convert it to an equivalent . current noise by dividing by the load resis- (21) tance: and, paralleling equation 18, the anode rms I (20) noise current associated with the signal cur- rent is given by Fig. 66 illustrates the relative magnitudes Irms (22) But, one must consider both the dark emis- sion noise and the noise in the signal current. Because there is no correlation between the two sources, the noises may be added by summing their squares, or for total rms noise current:

(23) An expression for the signal-to-noise ratio may now be written:*

SNR = iS Fig. 66 - Comparison of the magnitudes of Johnson noise in the load resistance and of the output photomultiplier dark noise, both in Fig. 67 illustrates the variation of signal- units of rms amperes per square root hertz. to-noise according to equation 24 as a func- For the case illustrated, both sources of tion of bandwidth and photocathode signal noise are equal for a load resistance of 50 current. The cathode dark emission is as- ohms. sumed to be 10 - 15 ampere. For signal cur- rents less than the dark emission, the dark noise is the limit to the signal-to-noise ratio. In this case, a narrow bandwidth is a require- ment for detection. When dark noise is the limit to detection, it is useful to reduce dark emission by cooling. At wider bandwidths and higher photocathode signal levels, the limit to the signal-to-noise ratio is in the

*In this discussion, the signal is treated as though it were dc. In an actual application, the signal may be modulated and signal-to-noise figures would involve the rms value of the modulated signal. Note also that the noise calculation neglects the factor by which the noise is increased by the secondary emission statistics.

61 Photomultiplier Handbook noise of the signal so that cooling would not terms used in these time characterizations be advantageous. In Fig. 67, it has been as- are defined below. In these definitions it is sumed that the load resistance is sufficient assumed that the photomultiplier is activated (greater than 50 ohms) so that Johnson noise with a delta-function light pulse. Fig. 68 il- is not a factor. lustrates the definitions of some of the more In applying these signal-to-noise concepts common terms. * to an actual case, the photocathode dark current may differ greatly from the 10- 15 ampere figure assumed. (See the data, Fig. 16.) The photocathode signal currents may be calculated from a knowledge of the pho- tocathode responsivity as published in a tube data sheet. For example, a photocathode may have a responsivity of 160 micro- amperes per lumen for tungsten irradiation, or 70 milliamperes per watt at the wave- length of maximum responsivity (i.e., 420 nanometers). A photocathode current of 1 femtoampere would then correspond to

10-15/70x 10-3= 1.43x 10-14 watt flux incident on the photocathode. Fig. 68 - The various time relationships in a photomultiplier output pulse, assuming a delta excitation function. Illustrated are tran- sit time, rise time, fall time, and full width at half maximum (FWHM).

Transit time† “is the mean time difference between the incidence of the light upon the photocathode (full illumination) and the oc- currence of the half-amplitude point on the output-pulse leading edge.” Rise time “is the mean time difference be- tween the 10- and 90-percent amplitude points on the output waveform for full cathode illumination and delta-function ex- citation.” Fall time “is the mean time difference be- tween the 90- and 10-percent amplitude points on the trailing edge of the output- pulse waveform for full cathode illumination and delta-function excitation.” TIME EFFECTS Terminology †This definition differs from the definition appearing in Photomultiplier tubes may be character- the IEEE Standard Dictionary of Electrical and Elec- ized for time response in various ways. The tronics Terms (ANSI/IEEE Std. 100-1977: “The time interval between the arrival of a delta-function light pulse at the entrance window of the tube and the time at *These definitions follow the standards appearing in which the output pulse at the anode terminal reaches “IEEE Standard Test Procedures for Photomultipliers peak amplitude.”). This latter definition is closer to the for Scintillation Counting and Glossary for Scintillation actual transit time which might be defined to the Counting Field.”ANSI N 42.9 - 1972; IEEE Std 398 average time of arrival of the electrons at the output. -1972. Published by The Institute of Electrical and Elec- However, the definition given to the half-amplitude tronics Engineers, Inc., 345 East 47 St., New York, point may be more useful in an application where the N.Y. 10017. half-amplitude point is used for timing purposes.

62 Photomultiplier Characteristics

Full-width-at-half-maximum (FWHM) is tipliers with large photocathode areas. The the mean elapsed time between the half- large photocathode necessitates a fairly long amplitude points on the output waveform path to the first dynode to provide good for full cathode illumination and delta- photoelectron collection from the entire function excitation. photocathode. Delta-function light pulses. In tests related to time characterization of photomultiplier tubes it is useful to have light pulse sources available which approach characterization as a delta-function pulse. A delta-function pulse is one whose duration is significantly shorter than that of the output pulse to be measured. It approaches the mathematical concept of a function whose area is finite but whose width approaches zero. Various light sources have been used to generate delta-function pulses for time- testing of photomultipliers. A reverse-biased light-emitting diode (Fer- ranti type XP-23) has been used by Leskovar75 to obtain light pulse widths of as short as 200 ps. -wetted spark SUPPLY KILOVOLTS BETWEEN sources have also been used and have pro- ANODE AND CATHODE vided pulses having rise times of 500 ps. 92CS-32406 Pulses of radiation having a duration of 50 Fig. 69 - Transit time as a function of the ps or less can be obtained with a mode- square foot of reciprocal applied voltage for a locked Nd:YAG laser. The laser wavelength type 8053 photomultiplier tube. radiation at 532 nm by means of a nonlinear crystal. Random pulses can be produced with fast scintillators. A rise time of 400 ps can be obtained for Naton 136 and a 60 Co source.81

Transit Time Transit Time is expected to increase as the inverse half power of the applied voltage, provided the effects of initial velocities and secondary-emission delay are negligible. Fig. 69 shows the transit time measurements for an 8053 photomultiplier plotted so as to display the reciprocal square root voltage relationship. The intercept of the line on the time axis is probably due to the distortion of the simple relationship by the magnitude of the initial secondary emission velocities. Fig. 70 shows the transit times for a Fig, 70 - Transit time as a function of supply number of photomultiplier tubes plotted voltage (log scales) for a number of photomul- over a range of operating voltages. The tiplier tubes. larger part of the transit time is just the ac- cumulation of times for electrons to traverse Photo- and secondary-emission times may from stage to stage. Usually, the time for the be as short as 10 ps, although for negative- photocathode-to-first-dynode transit is the electron-affinity (NEA) materials, the times largest component, especially for photomul- may be as long as 100 ps. One reason for the

63 PhotomultiplIer Handbook high performance of NEA materials is that Pulse Width long diffusion paths for electrons are ob- The width of the output pulse is deter- tained. Although this increase in diffusion mined by the variation in transit time path length also increases the emission time, through the secondary emission chain of the emission time is not a significant part of the photomultiplier.The variations arise be- transit time in most commercial photomulti- cause of variations in emission energies and pliers. In general, the transit time of a photo- directions of the secondary electrons as multiplier is not as important as its rise time related to the tube structure. The measure- or as variations in the transit time which ment of pulse width is generally the full would cause uncertainty in time measure- width at half maximum. The output pulse ments made with the photomultiplier. A width follows the inverse half power of the fixed delay time is easily compensated for by applied voltage as does the average transit circuit design. time. Fig. 72 illustrates the pulse width Rise Time (FWHM) for an 8053. For timing experi- ments, it is generally desirable to have a nar- Fig. 71 shows the rise time-from the 10 row pulse width for good timing precision capability.

Fig. 71 - Anode-pulse rise times as a func- tion of anode-to-cathode applied voltage (log scales) for a number of photomultipliers.

KILOVOLTS to 90% amplitude points-for a number of 92CS-32409 photomultipliers plotted over a range of Fig. 72 - Pulse width (full width at half max- typical operating voltages. No correction imum) for a type 8053 as a function of the in- was made for the finite rise time of the light verse half power of the applied voltage. pulse and measuring equipment, which is estimated to be of the order of 0.8 ns. In us- ing photomultipliers for high-resolution- Pulse Jitter (Time Resolution) time spectroscopy, the ideal point on the Although pulse timing is done on the ris- output pulse to use as an indicator would be ing characteristic of the output pulse and is at the half maximum of the rising character- more precise for a fast rise time, the ultimate istic. The rising characteristic is generally limit to time measurement is the variation in faster than the fall and thus provides the pulse timing, or pulse jitter. Suppose single highest precision in timing. However, photoelectrons initiate pulses. Variations in because output pulse heights vary, a fixed transit time of photoelectrons to the first discriminator level results in a loss of preci- dynode will occur because of variations in sion. When a fixed discriminator level is the initial velocity and electric field resulting used, the highest precision is obtained by use from the electrode geometry. The same con- of a discriminator level between 10 and 20% siderations apply to the secondary electrons. of maximum. A superior method is the use If a number of pulses initiated by single elec- of a constant fraction of the pulse height as a trons are observed, a histogram can be devel- trigger .76 oped showing the number of pulses having a

64 Photomultiplier Characteristics given transit-time difference. Such a histo- Data on time resolution for single photo- gram, measured by Birk, Kerns, and electrons for a variety of photomultipliers Tusting,77 is shown in Fig. 73. The time of are given in Table IV. Values are given for each pulse was measured by using the full photocathode illumination as well as for leading-edge half-height point. The full illumination at the center point. The time width at half maximum of this distribution is spread for full photocathode illumination is about 360 ps and is a measure of the time larger than for a single point because it in- resolution capability of this particular tube. cludes the difference in transit time from the photocathode to the first dynode for dif- ferent photocathode locations.

TRANSITTIMEDIFFERENCE - ns 92CS-32410 Fig. 73 - Histogram of transit-time difference for single-photoelectron pulses from an RCA developmental type photomultiplier. (From 77 Birk, Kerns, and Tusting .) 92CS-32411 Fig. 74 - Time resolution of a microchannel- If a number, N, of simultaneous photo- plate photomultiplier as a function of the electrons is emitted, the pulse jitter is re- number of photoelectrons per pulse, mea- duced by the square root of N, simply by the sured with light pulse width of 2.6 ns, for full statistical averaging process. This relation- photocathode illumination. Data are from 75 ship has been demonstrated by Leskovar and Leskovar and LO The dashed line showing Lo75 for a microchannel-plate photomulti- the inverse square root relationship has been plier as shown in Fig. 74. added.

Table IV - Time Resolution (FWHM)* for Single Photoelectrons for Various Photomultipliers

65 Photomultiplier Handbook

The improvement of the 8852 over the time of flight for the over-all tube is given by 8575 is the result of the use of the high-gain first dynode. The difference between the (27) 8850 and the 8852 may be due to the larger photoelectron emission energies of the multi- Eq. 27 refers to the variance of the centroid alkali photocathode. The poorer time resolu- of the output pulse from a single photoelec- tion of the 8854 is caused by the very large tron input. If timing is done by a point on cathode area and the consequent reduced the rising characteristic, an additional time electric field strengths and increased path variance might be included, that for the lengths. variation in pulse width. Gatti and Svelto81 A general treatment of single photoelec- have shown that this term would add a tron detection and timing has been provided negligible amount to the variance as given by Eq. 27. tion for a photomultiplier is very similar to that for noise in a photomultiplier as given in PULSE COUNTING Appendix G. Variations in transit time in the One effective way to use a photomultiplier early stages of the photomultiplier are most for measuring very weak signals is to detect important to the over-all time resolution. and count pulses resulting from single The large number of electrons in the latter photoelectrons-sometimes referred to as stages bring about an averaging process “”. (Actually, the highest which reduces the time variation. quantum efficiencies for photocathodes are Summary of Time-Resolution Statistics such that at best one in every three photons would be detected, assuming a spectral The summary of time resolution statistics match to the peak quantumefficiency below follows the work of Gatti and wavelength.) Counting of single electrons assumes that the rate of arrival of photons is If the secondary emission of each stage is such that it would be unusual for more than one photoelectron to be emitted in a time electron flight times including all of n stages equivalent to the output pulse width for a is given by single electron input to the multiplier. Single-electron pulse counting is an impor- tant technique in applications such as , astronomical photom- etry, and bio-luminescent measurements. counting has an advantage equivalent to a variance of the flight time from the ith factor of 1.2 in quantum efficiency over dynode to the (i + 1)th dynode. The last current-measurement techniques. term represents the variance in flight time from the last dynode to the anode. This ex- Output Pulse Height Distribution pression is actually a simplification because An important consideration in photon of the induction effect of the electrons in counting is the distribution of pulse heights flight between the last stage and anode. at the output of the photomultiplier. Statis- tics of the variation in pulse heights for is negligible. If n is large and the variances of single electron inputs is discussed in Appen- flight times for each stage are assumed to be dix G. (See Fig. G-6). Pulse height distribu- tions are obtained with a multichannel pulse- follows: height analyzer. Fig. 75 shows (1) a differen- tial pulse-height distribution in which the (26) number of pulses in a given time interval and in a given channel (between height, h and h When variations in secondary emission are taken into account, assuming all stages have an integral pulse-height distribution in which the total number of pulses occurring in the a Poisson distribution, the total variance in given time interval with a height of h or

66 Photomultiplier Characteristics greater is plotted.83 Also shown on the necessary to measure the count in the dark graph is a rectangle whose intercept on the and to subtract the rates to obtain that due abscissa is the equivalent of the output pulse to the light alone. The optimum time which associated with one photoelectron and an should be spent on each measurement is average photomultiplier gain of G. The rec- discussed in Appendix G-see for example, tangle is determined by setting its height Eq. G-103. If the signal count is much less equal to the intercept of the integral-pulse- than the dark count, however, about equal height distribution curve on the ordinate times should be spent on both measure- axis, and by setting the area of the rectangle ments. At higher signal levels, more time equal to the area under the integral-pulse- should be spent on the signal count. height curve. This area is equal to the total charge of all the pulses counted: the total Copper-Beryllium Dynodes number of pulses multiplied by the average Differential pulse-height distributions are multiplier gain and the charge of one elec- shown in Fig. 76 on a developmental photo- tron. This equality may be demonstrated by integrating the integral-pulse-height curve along the ordinate axis and comparing the result with the integral of the product of the differential pulse-height-distribution or- dinate (number of pulses) and the abscissa value (pulse height or charge associated). The pulse-height resolution for single elec- trons in this case is FWHM = 1.6 electron equivalents. Note that the data of Fig. 75

-LIGHT FOR

PULSE HEIGHT -ARBITRARY UNITS 92cs-32413 Fig. 76 - Differential pulse-height distributions obtained with type Dev. No. C-70101B. (Simi- lar to type 8575, but having an S-20 response. From R. M. Matheson84.)

92CS-32412 multiplier having copper-beryllium dynodes. Fig. 75 - Single electron (1) differential and (2) integral pulse-height distribution curves. Type Note that the dark pulse-height distribution 4501, photocathode K2CsSb, counting time 10 does not show a peak representing single minutes (from G. A. MortonB3). Different scales electrons. The probable explanation is that for (1) and (2). dark emission originates not only from the photocathode, but from the dynodes as well. were taken on a tube, 4501, that has copper- Electrons originating from the first dynode beryllium dynodes and a modest secondary- would show a distribution similar to that of emission gain. The pulse-height resolution the photoelectrons, but with a pulse height for single electrons in the case of gallium- lower by the secondary-emission ratio of the phosphide dynodes is much better, as will be first dynode. Pulse-height distributions also discussed below. usually show an extended foot, as in Fig. 76, When a very low light flux is measured by for heights greater than 7 on the figure’s ar- the pulse-counting technique, it is also bitrary scale. These counts represent pulses

67 Photomultiplier Handbook

of more than one electron per pulse and are are much improved. Consequently, such probably caused by secondary mechanisms tubes are particularly advantageous for such as ion feedback, scintillations in the single-electron pulse counting. The improve- glass envelope initiated by cosmic rays or by ment in multiplication is demonstrated by radioactive traces in the glass or in the tube the differential pulse-height distribution for environment. Note that the numbers of these single electrons shown in Fig. 78. Note the large pulses are usually several orders of magnitude less than the numbers in the single-electron distribution. When a single- electron integral count is made, the upper discriminator level could be set to exclude these larger pulses. In setting the lower level of the discriminator, an optimum level is deter- mined by the different character of the light shown that the minimum error in deter- mining the signal pulse count is obtained by adjusting the lower-level discriminator set- ting so that for an integral count distribution

(28) where Ns represents the number of signal pulses counted and Nd represents the number of dark pulses counted. In Fig. 77, the value of the pulse height that satisfies the relation shown in Eq. 28, is given by hl.

PULSE HEIGHT-PHOTOELECTRON EQUIVALENTS

92CS-32415 Fig. 78 - Resolution of a single electron peak having a measured FWHM of 63%. The first dynode is GaP:Cs. (Data from Morton, Smith,

improvement in resolution as compared with that of Figs. 75 or 76. When the intensity of the light flashes is increased, a pulse-height spectrum such as illustrated in Fig. 79 results, again for a GaP:Cs first dynode. In 3 this case the light level has been adjusted so that the peak counts are nearly equal for Fig. 77 - Single electron response (I) and dark one, two, and three electron pulses. The pulse distribution (2) of tube type 4501 for small pulse distribution to the left of the counting time of 10 minutes; integral distribu- single-electron peak distribution is probably h tion curves. h 1 and 2 are discriminator set- an artifact caused by equipment noise at the tings (From G. A. MortonB3). lower channel settings. Peak-to-Valley Ratio Gallium-Phosphide Dynodes Another important consideration for a For tubes having the high-gain GaP:Cs photomultiplier used in pulse counting is the first dynodes, the statistics of multiplication peak-to-valley ratio. Referring to Fig. 79,

68 Photomultiplier Characteristics

for single electrons, the peak-to-valley ratio is the value of the peak of the single-electron distribution divided by the minimum value of the count distribution between the first and second peaks-in this case a ratio of about 2.3.

92CS-32417

taken with a NaI:Tl crystal and a two-inch “teacup” photomultiplier (type 4902). Pulse-height resolution is illustrated for full Fig. 79 - Pulse-height spectrum showing width at half maximum. The peak at the ex- peaks corresponding to one, two, and up to treme left is the K X-ray peak at 32 five electrons. The first dynode is GaP:Cs. keV. The region between the two peaks is the result of Compton scattering. From Appendix G, Eq. G-111, the pulse- SCINTILLATION COUNTING height resolution is given by Pulse Magnitude In scintillation counting, the problems are similar to those of counting single photoelec- trons but the numbers of equivalent photo- electrons per pulse can vary from a few to a The following values may be assumed: fairly large number. For example, in the case of a soft-beta emitter, using a coincidence liquid scintillation counter (for an un- mc = number of photons corresponding to quenched standard, PPO and POPOP in toluene), the total number of photoelectrons = variance in the number of photons developed in the pair of photomultipliers per photopeak pulse imately 2.5 per keV of beta-ray energy. For the case illustrated in Fig. 80, the Because dark noise from single electrons can FWHM = 7.17%. Substituting the above be effectively discriminated against by using coincidence techniques, a relatively efficient 325,000. The FWHM related to the crystal count of low-energy beta emission can be statistics alone is 6.33%. The FWHM related made. On the other hand, in the case of a to the photomultiplier alone is 3.37%. NaI:Tl crystal coupled to a photomultiplier having a K2CsSb photocathode, the number of photoelectrons developed is approximate- Fig. 81 shows a pulse-height distribution ly 8 per keV of gamma-ray energy. Thus, for the isotope energy of 5.9 keV. Note that for this spec- energy is 662 keV, the pulse height would be trum, the FWHM is approximately 40%. the equivalent of 5300 photoelectrons. Following the same type of analysis as used above to evaluate the relative contributions Isotope 137Cs Sources Fig. 80 shows the pulse-height distribution for the photomultiplier alone is 36% and for the scintillator, 18%.

69 Photomultiplier Handbook

Narayan and Prescott86 have presented figuration so that all or nearly all of the data illustrating a trend in which the photo- emitted photoelectrons effectively impact multiplier statistics dominate the pulse- the first dynode. The statistics of secondary height resolution at low gamma-ray energies emission, especially in the first stage, also af- and the scintillator statistics dominate at fect the ultimate resolution. It is desirable, high gamma-ray energies. This trend is il- therefore, to provide a fairly high secondary- lustrated in the above analysis of the data in emission yield either by the nature of the Figs. 80 and 81. dynode surface material or by applying a rather high voltage between the photocath- ode and the first dynode. Finally, it is generally advantageous that the photoemis- sion be uniform across the photocathode and that a uniform collection of all the photoelectrons be provided. This latter characteristic is particularly important in the case of thin scintillators where the light from a scintillation is more localized on the photocathode than it would be for thicker scintillators. Light pipes may be used to im- prove the uniformity of light on the photocathode and therefore improve the pulse-height resolution in the case of thin scintillators. For a crystal whose length is comparable to its diameter, the output light- flux distribution is generally uniform so that photocathode uniformity becomes statisti- I I I I cally less significant. Peak-to-Valley Ratio As in pulse-counting applications, another Fig. 81 - Pulse-height distribution for 5sFe way of classifying the performance of photo- taken with type 8850 photomultiplier and multipliers in scintillation counters is by the Nal:TI crystal. peak-to-valley ratio of the distribution. This way is illustrated in Fig. 81. The peak will be As the energy is increased, of higher, of course, if the resolution of the course, the relative pulse-height resolution tube is better. Good resolution also results in attributable to the photomultiplier decreases a lower valley point before the pulse-height as the square root of the number of photo- distribution increases on the low energy side electrons per pulse or as the square root of as a result of photomultiplier dark current. the gamma ray energy. The relative pulse- The peak-to-valley ratio is a particularly height resolution attributable to the crystal valuable parameter for measurements with also decreases with gamma ray energy, but sources characterized by low emission not as rapidly. The light-pulse statistics of energy. the crystal also varies from crystal to crystal. Plateau Concept Key Photomultiplier Characteristics for A plateau curve for a scintillation counter Good Pulse-Height Resolution is illustrated in Fig. 82. The curve is devel- Several characteristics of a photomulti- oped by using a source such as 137Cs and plier tube determine its suitability for obtain- counting all pulses higher than a fixed ing good pulse-height resolution. Of prime discriminator level. The total counts are then importance is the quantum efficiency of the plotted as a function of the photomultiplier photocathode, especially as it matches the voltage. The development of the curve in generally blue spectrum of scintillators. Fig. 82 may be understood by referring to Secondly, of course, the photomultiplier the pulse-height distribution curve of Fig. must have a suitable electron-optical con- 80. At the lowest value of photomultiplier

70 Photomultiplier Characteristics

voltage, the pulse heights are less than the plateau would also be shorter as the number equivalent of the discriminator setting. As of stages is increased. the voltage is increased, the discriminator Photomultiplier plateau is of particular in- setting moves in effect from right to left on terest in the application to oil-well logging. Fig. 80. The rising portion of Fig. 82 then A rather intense gamma-ray source such as corresponds to the discriminator moving 137Cs is mounted in the sonde near the through the photopeak, through the Comp- photomultiplier and crystal assembly, but ton scattering region of the distribution, and shielded from it. The gamma-rays result in through the barium X-ray peak. The plateau Compton scattering in the materials near the corresponds to the dark background87 of the probe. The scattered radiation is detected photomultiplier which would appear at the and measured with an integral count with the extreme left of Fig. 80 except for the discriminator set to correspond to a point on scale-the barium X-ray peak at 32 keV cor- the plateau so that essentially all of the in- responds to about 250 photoelectrons. The tercepted radiation is counted. As the depth dark background pulse distribution of a of the measurement in the drill hole in- photomultiplier is shown in Fig. 76. As the creases, the temperature does also. It is, voltage on the photomultiplier is increased, therefore, important that the counting char- the dark current of the tube increases be- acteristic of the scintillation detector be cause various regenerative effects increase, essentially independent of temperature. But, and the plateau is terminated at the upper as the temperature increases, both the dark end of Fig. 82. background count and the instability of the photomultiplier increases and, as a result, the length of the plateau decreases. Fig. 83 il-

PHOTOMULTlPLlER VOLTAGE-VOLTS 92CS-32419 92cs-32420 Fig. 82 - Typical plateau is defined as por- Fig. 83 - Plateau characteristics at room tion of integral-bias characteristic in which temperature and at 150 °C for a photomulti- change of counting rate per 100-volt interval plier having a Na KSb photocathode. The is less than a selected value (Engstrom and 2 87 photomultiplier’s principal application is in Weaver ). oil-well-logging. The concept of utilizing a plateau charac- teristic probably stemmed from the plateau lustrates plateau characteristics taken at of Geiger-Muller tubes. In the case of photo- 22°C and 150°C for a photomultiplier multipliers, the length of the plateau is deter- having a Na2KSb photocathode. The curves mined by the stability of the tube at high indicate a range of photomultiplier voltage gain (a minimum of regenerative effects) and that would be satisfactory for this range of by the characteristic variation of gain with temperature. voltage. That is, a tube having Cs-Sb In scintillation counting applications dynodes would generally have a shorter where gamma-ray energy resolution is im- plateau than one having Cu-Be dynodes portant, the plateau concept is not par- simply because of the more rapid increase in ticularly applicable. Direct measurement of gain with voltage for the Cs-Sb dynodes. The pulse-height resolution would be more use-

71 Photomultiplier Handbook ful. But in special applications such as the A block diagram of a liquid-scintillation oil-well logging, the plateau characterizes the spectrometer is shown in Fig. 85. If a photo- useful range of operating voltage. multiplier having a fast time characteristic is used in the spectrometer, the coincidence resolving time may be as small as 10 nano- LIQUID SCINTILLATION COUNTING seconds. Liquid scintillators are most commonly used for the evaluation of low-energy beta emitters because of the generally short range of beta particles and the need to provide an intimate contact between the source and the scintillator .88,89 There are numerous fluors or solutes which may be used, such as PPO (2,5 diphenyloxazole). A common solvent is toluene which must be miscible with the li- quid radioactive sample. The emission of the typical liquid scintillator is in the near ultra- violet and blue spectrum so that a photomul- tiplier having a“bialkali” photocathode (K2CsSb) provides a good spectral match. Counting Techniques When liquid-scintillation counting was first introduced, the technique involved the use of a single photomultiplier to observe the liquid scintillator .Because of the low energies involved and the need to measure samples of relatively low activity, the background count of the photomultiplier was a severe limitation. An improved ap- proach is the use of a pair of photomultiplier Fig. 84 - Pulse-height spectrum representing tubes facing a common scintillator chamber. the beta energy for tritium (18.6 keV max- Because of the number of photons involved imum) from a liquid-scintillation counter. Maximum and minimum discriminator levels when a scintillation occurs in the sample, it is are indicated (From E.D. Bransome88). quite likely that both photomultipliers will sense the flash, thus permitting the use of coincidence circuitry. Events that initiate background counts in one photomultiplier are only infrequently coincident with those in the second tube and so a high degree of discrimination against background counts is possible. The three most commonly used radioiso- topes in liquid-scintillation counting are tritium, 3H; , 14C; and phosphorus, 32P. The beta-ray energies from these radio- isotopes may vary over a fairly broad range; for example, 3H emits betas having energies varying from zero to a maximum of 18.6 keV. Maximum beta-ray energies for 14C and 32P are 156 keV and 1.71 MeV respec- tively. Fig. 84 illustrates the distribution of 92cs-32422 pulse heights (which are proportional to energy) for tritium. (The average energy is Fig. 85 - Block diagram of a liquid-scintilla- about 6 keV.) tion spectrometer.

72 Photomultiplier Characteristics

Counter Efficiency coincident. The coincident rate may be A conventional figure of merit for a liquid predicted by the following equation: scintillation coincidence counter is the ratio, E2/B, where E is the counting efficiency ex- (29) pressed in per cent, and B is the number of background coincident counts per minute. The efficiency, E, is measured utilizing a where C is the chance coincidence rate per standard sample whose disintegration rate is minute, N1 is the dark-noise count rate in known. The recorded count rate is compared counts per minute from tube No. 1, N2 is the with that of the standard, taking into ac- dark-noise count rate in counts per minute count the exponential decay in the disinte- gration rate. The background count value, of the coincidence circuit in seconds. In a B, is measured by utilizing a blank sample of liquid-scintillation spectrometer employing the scintillator . two tubes each having dark-noise rates of Note that E2/B is proportional to the 30,000 counts per minute each and a coin- square of a signal-to-noise figure for the cidence circuit having a resolving time of 10 system. Thus, the efficiency E is propor- nanoseconds, the number of accidental coin- tional to the signal; and the square root of B cidences is approximately 0.3 count per is the expected standard deviation in the minute. statistical count. Another source of background count is Efficiencies of the scintillators and the cross talk between the two photomultipliers photomultipliers in liquid-scintillation- as a result of light flashes in one tube which counting equipment (using the most efficient are sensed by the other. liquid scintillator and photocathodes) are Cosmic rays and other natural radiation such that about 2.5 photoelectrons are emit- can result in flashes in the scintillator or ted per keV of energy of the beta ray. The possibly in the photomultiplier envelope. best counting efficiency for a given radioiso- Shielding the equipment with lead can tope is obtained when a highly efficient scin- reduce the background from these sources. tillator and a photomultiplier having a high However, the photomultipliers themselves quantum efficiency in the spectral emissivity also contain radioactive isotopes. A com- range of the scintillator are used. The optical mon contaminant is 40K, a naturally oc- system containing the two photomultipliers curring isotope of potassium (0.1 %), that is and the counting vial is also of major impor- present in many glasses. Photomultipliers tance. The system should be designed so designed for use in liquid-scintillation that, as far as possible, the photons pro- counters may utilize quartz face plates or duced in a scintillation are equally divided thin face plates of a glass having a minimum between the photomultipliers. This division potassium content and with a low yield of assures that a coincidence pulse results from scintillations from gamma rays. as many scintillations as possible. In some When a vial filled only with a scintillator is cases two or more different isotopes may be placed between the photomultipliers, and the counted simultaneously. It is desirable, output from the coincidence circuit is ex- therefore, that the photomultipliers have amined by use of a multichannel analyzer, a matched gains and good energy (pulse- pulse-height distribution such as that shown height) resolution capability to provide best in Fig. 86 is obtained. Clearly, not many of isotope separation. A typical value for E is the background pulses shown are caused by about 60%. Quantum efficiency of the the accidental coincidences of dark-noise bialkali photocathode for the scintillation pulses from the photomultipliers, but are radiation is approximately 25 %. caused by cosmic rays of scintillations in the material of the vial and photomultiplier Sources of Background Counts envelope resulting from the presence of There are various sources of the coin- radioisotopes in the materials of which they cident background counts. One source is the are constructed. random dark-noise pulses in each of the two Typical values for B, the number of photomultipliers which are occasionally background counts,in coincidence liquid-

73 Photomultlpller Handbook

scintillation counters are in the range 15 to ronmental limitations or to means for op- 20 counts per minute, and E2/B values are timizing performance under adverse condi- typically in the range 150 to 250 for tritium. tions. Temperature The maximum temperature to which a photomultiplier should be exposed during either operation or storage is determined primarily by the characteristics of the photo- cathode. Temperature ratings vary with tube type depending upon the particular photo- cathode. For a specific type, its published data should be consulted. In general, how- ever, a temperature of 75 °C should not be exceeded. Excessive exposure to higher temperatures will alter the delicate balance of chemicals of the photocathode and result in a loss of responsivity and/or a shift in spectral response.Another reason for 92CS - 32423 avoiding operation of photomultipliers at Fig. 86 - Background distribution obtained elevated temperatures is that it causes an in- from a liquid scintillator using a coincidence system. crease in the dark current. (See Fig. 16.) There are also hazards in cooling photo- multiplier tubes to reduce the dark emission Photomultiplier Selection for from the photocathode. Rapid cooling Liquid Scintillators should be avoided to minimize thermal The E2/B figure of merit is, however, not shock which could result in cracking of the necessarily the best way of evaluating all bulb or stem. Temperatures below - 50°C systems. It is valid at very low counting rates should be avoided because the difference in where the background count is the dominant expansion between the glass and the base factor, but is not of great help at high- material could result in cracking of the glass. counting rates where the background count This hazard is also present for photomulti- of the system becomes less important than plier tubes having metal envelopes or for the efficiency in determining the merit of the tubes having stiff lead glass stems when system. socketed in a material of different expansion In selecting a photomultiplier tube for a coefficient. liquid-scintillation application, the following In the cooling of photomultiplier tubes items are of major importance: high photo- care must also be taken to avoid moisture cathode quantum efficiency, low dark-noise condensation. One method is to provide a count rate, minimum internal-light genera- dry atmosphere for the tube and, especially, tion, low scintillation-efficiency envelope, its base and socket. Another method is to fast time response, and good energy resolu- cover the critical areas of the tube with a tion. The 4501V3 photomultiplier has been silicone rubber such as specifically designed to meet these require- RTV-11. The tube should first be cleaned ments. and then primed (4004 primer for RTV-11) before the silicone rubber coating is applied. ENVIRONMENTAL EFFECTS Differential cooling in which only the pho- Although photomultiplier tubes are not tocathode end of the tube is cooled is not overly sensitive to their environment and can recommended. This practice may result in be handled without exercising undue cau- loss of photocathode sensitivity because of tion, there are various considerations the the vapor transport of alkali metals from the user should be aware of in order to maximize dynodes to the photocathode. tube life, avoid permanent damage, and pro- vide the optimum performance. This section Voltage provides guidance relating to specific envi- The published data for photomultiplier

74 Photomultlpller Characteristics tubes contain information on the maximum Light Level supply voltage as well as information on the It is generally advisable to store photomul- specific maxima for voltages between adja- tiplier tubes in the dark and to avoid ex- cent electrodes such as between dynode No. cessive exposure of the tubes to any light rich 1 and the photocathode. Too high a voltage in blue or ultraviolet such as that from fluor- between closely spaced electrodes can result escent lamps or sunlight. Exposure of the in electrical breakdown, particularly across photocathode to such radiation will general- insulator surfaces. The over-all tube voltage ly cause a very substantial increase in the maximum is determined by considerations of dark current originating from the photocath- the maximum gain which a tube can tolerate. ode. Complete recovery from this effect may Depending upon the construction of the take as long as several days. See Fig. 47. Ex- photomultiplier, at some gain in excess of posure of the photocathode to sunlight even 107 or 10 8, there may be sufficient feed- though no voltage is being applied may also back-perhaps by light generated in the out- result in photocathode response changes put section of the tube-to result in a sus- such as loss of infrared response in a tube tained electrical current. This current having a multialkali photocathode. breakdown, if sufficiently large, can per- Before a photomultiplier tube is used for manently damage the tube by causing an in- critical low-light-level measurements, it is crease in the dark current and a decrease in recommended that it be aged for a period of responsivity . 24 hours in the dark with voltage applied. Ground potential and shielding of the This precaution is particularly recommended photomultiplier can also be a problem. In after a long period of idleness even if the order to minimize regenerative effects it is tube has not been exposed to ambient generally recommended that the wall of the lighting. photomultiplier envelope be maintained at a When the photomultiplier has the ap- potential at or near cathode potential. This propriate voltage applied, it is of course pru- recommendation offers no problem in a cir- dent to avoid any excessive light exposure. cuit in which the cathode is at ground poten- Particularly, if the has a tial. But when the anode is at ground poten- substantial current flow, an excessive current tial, it may be advisable for safety reasons to may flow in the photomultiplier which could provide a double shield: the inner shield at result in loss of gain and an increase in dark cathode potential, surrounded by an in- current. Photomultiplier data sheets gener- sulator, and an outer shield at ground poten- ally contain ratings of the maximum anode tial. The inner shield should be maintained current which may be drawn. at cathode potential through a high resis- Some semitransparent photocathodes are tance (usually 5 to 10 megohms) to avoid very resistive. See Fig. 36. Even though the hazard through accidental contact or light flux on the photocathode may be breakdown of the insulation. relatively small, it is possible that the drop in It is strongly recommended that no dif- voltage across the photocathode between the ference of potential be maintained between electrical contact and the light spot may in- the semitransparent photocathode layer of a hibit the photoelectron current flow and photomultiplier and its outer glass faceplate. cause a non-linear behavior that could upset A potential on the outside of the glass that is careful measurements. positive with respect to that of the photo- When measurements are made at very low cathode can result in damage to the photo- light levels, it is important that the tube be cathode by ionic transport through the glass totally shielded from unwanted stray light to faceplate.69 (See section on “Shielding” avoid an increase in background noise out- earlier in this Chapter.) put. A person’s photopic vision is not a good In supplying the tube voltage by means of judge of the presence of light leakage that a resistive voltage divider, care must be the photomultiplier might sense readily. For taken to avoid physical contact between any example, light may leak through an open- of the resistors. Such contact can cause ended coaxial cable connector such as BNC miniature electrostatic discharges resulting or through certain bases or sockets. It may in noise spikes in the photomultiplier output. be noted that light-tight caps are generally

75 Photomultlpller Handbook commercially available to terminate unusual ber as described earlier under “Tempera- coaxial connectors. ture.” Again, a word of caution about the prob- Magnetic Fields lem of helium penetration of glass. Even at Care should be exercised to keep the room temperature, if there is a significant magnetic field environment of a photomulti- helium content in the ambient photomulti- plier to a minimum. The electron optical plier atmosphere, some helium will penetrate operation of a photomultiplier can be con- the glass envelope. The result is an increase siderably altered by magnetic fields, as in dark current and after pulsing because of shown in Fig. 51. It is also possible to induce ionization of the helium gas. Eventually, a permanent magnetization of some photo- with sufficient helium pressure a complete multiplier parts such as dynodes or dynode electrical breakdown can occur with voltage side rods constructed of nickel. If applied to the photomultiplier tube. magnetization occurs, degaussing may readi- If helium cannot be avoided, it may be ly be accomplished by placing the tube at the desirable to use a separate enclosure for the center of a coil operated at an alternating photomultiplier through which a small flow current of 60 Hz with a maximum field of purging gas such as dry nitrogen is pro- strength of 8000 ampere turns per meter and vided. It is also reported90 that helium then gradually withdrawing the tube from penetration can be blocked by a thin layer of the coil. an epoxy-Epon 828 resin (registered trade- Magnetic shields are generally available mark of Shell Chemical Company) and commercially for photomultiplier tubes of Belsamid 125 hardening agent (registered various constructions. In providing such trademark of General Mills Incorporated). protection, it may be important to use a Shock and Vibration shield which extends beyond the semitrans- parent photocathode a distance of at least Most photomultipliers will survive only a half the diameter of the photocathode. reasonable amount of shock or vibration (less than 10-g shock, depending on shock Atmosphere duration and direction). Although special In some applications it may be necessary photomultipliers have been designed to sur- to operate the photomultiplier in other than vive in extreme environments (shock values normal atmospheric pressure. Most photo- from 30 to 1500 g), the user should make multipliers will tolerate pressures to three at- every effort to avoid excessive shock or vi- mospheres. For the larger tubes, however, bration, possibly by the use of special this pressure may present an implosion vibration-isolation fixtures. The photomulti- hazard. In these special cases, it would be plier tube should be handled as the delicate well to check with the tube manufacturer. instrument that it is. Excessive shock or Pressures less than one atmosphere also pre- vibration can actually cause physical damage sent a problem in that electrical breakdown to the tube to the point of shorting out some can more readily occur. In such cases it may of the elements or even resulting in breakage be necessary to coat all exposed connections of the envelope and loss of vacuum. A lesser and wiring with an insulator such as silicone degree of shock may cause deformation of rubber. the tube elements and can result in loss of Corrosive atmospheres must be avoided, gain or deterioration of other performance especially on photomultipliers having metal parameters. If measurements are being made envelopes. Corrosion could destroy the while a tube is vibrated, it is likely that the glass-to-metal seal and result in loss of output will be modulated by the vibration vacuum. not only because the light spot may be High-humidity conditions should be deflected to different positions on the avoided if possible because condensation on photocathode but also because some of the the base and socket can result in additional dynodes may actually vibrate and cause electrical leakage or even breakdown. If the modulation of the secondary-emission gain. moisture situation is unavoidable, it may Many photomultipliers have been de- again be advisable to coat exposed connec- signed for use under severe conditions of tions with an insulator such as silicone rub- shock and vibration and, in many cases, spe-

76 Photomultiplier Characteristics cifically for use in missile and rocket applica- tubes may be subjected to high levels of tions. Such tubes, however, find uses in radiation such as occur in the Van Allen many other applications including oil-well belts. A summary of the effects of radiation logging or other industrial control applica- on photomultiplier tubes may be found in a tions where the tube may be subjected to paper by S.M. Johnson, Jr.90a. Temporary rough usage. These tubes are available with effects of intense radiation are principally an most of the electrical and spectral character- increase in background current and noise. istics typical of the more general-purpose Continued exposure, however, will also types. These types differ primarily in cause permanent damage to the face-plate mechanical construction in that additional glass, and to a much lesser extent the photo- supporting members may be employed and cathode, and the dynodes. an improved cathode connection may be The origin of the increased background used to assure positive contact when the tube current in photomultipliers exposed to is subjected to these environments. Rug- nuclear radiation is a fluorescence or scin- gedization of tubes using the glass envelope tillation in the glass faceplate that causes has also been accomplished by moving the electron emission from the photocathode. dynode cage close to the stem (thereby For example, irradiation with 60Co gamma drastically shortening the lead lengths and rays produces the equivalent of 10-12 raising their mechanical resonant frequen- W/cm2 of 420-nm flux for an input of 10 - 3 cy), by using heavier leads and extra spacers rad/s. to hold the dynode cage in place, and by a The major damage to a photomultiplier is special heavy-duty welding process on the the browning of the faceplate glass. Signifi- metal-to-metal joints. cant changes in transmission are observed Tubes recommended for use under severe for 9741 and 7056 glass with an exposure of environmental conditions are usually de- 105 rads of 60Co gamma radiation. Fused signed to withstand environmental tests silica shows much less change for the same equivalent to those specified in the ap- irradiation. Lime glass is reported to be very plicable portions of MIL-E-5272C or MIL- susceptible to browning. Sapphire is the least STD-810B in which the specified accelera- degraded by gamma radiation of windows tions are applied directly to the tubes. used in photomultipliers. Of the far-ultra- Sinusoidal vibration tests are performed on violet windows, LiF is reported to be very apparatus that applies a variable sinusoidal susceptible to radiation damage. MgF2 is vibration to the tube. The sinusoidal fre- recommended down to the Lyman-alpha quency is varied logarithmically with time wavelength level in the presence of high from a minimum to a maximum to a mini- radiation exposures. mum value. Each tube is vibrated in each of Although most of the damage to photo- the three orthogonal axes. multipliers from nuclear radiation is in the Random vibration tests are performed by faceplate, some change may be expected in subjecting the tube to a specified spectral the photocathode. The damage is not great density (g2/Hz) in a specified frequency compared with that to the glass, probably band. because of the very small absorption of the Shock tests are performed on an apparat- very thin photocathodes. The same may be us that applies a half-wave sinusoidal shock said for dynode material where, again, the pulse to the photomultiplier tube. The tube surface layer is most important. is subjected to the shock in each of the three orthogonal axes. The shock pulse is ex- REFERENCES pressed in terms of the peak acceleration of the pulse and the time duration of the pulse. 44. W.J. Harper and W.J. Choyke, “The The tubes may receive more than one shock resistance of semitransparent photo- pulse in each of the orthogonal axes. cathodes,”Jour. Appl. Phys., Vol. 27, p. 1358 (1956). 45. R.W. Engstrom, R.G. Stoudenheimer, Nuclear Radiation H.L. Palmer, and D.A. Bly, “Recent Work In specialized applications such as the use on photoemission and dark emission pro- of photomultiplier tubes in satellites, the blems,”IRE Trans. Nucl. Sci., Vol. NS 5,

77 Photomultiplier Handbook

No. 3, p. 120-124, (Dec. 1958). 59. W. Widmaier and R.W. Engstrom, 46. M. Lontie-Bailliez and A. Messen, “Variation of the conductivity of the semi- “L’influence de la temperature sur les transparent cesium-antimony photo- photomultiplicateurs,” Annales de la cathode,”RCA Rev., Vol. 16, No. 1, p. Scoiete Scientifique de Bruxelles, Vol. 73, 109-115 (March 1955). Series 1, p. 390 (1959). 60. H.A. Zagorites and D.Y. Less, “Gam- 47. W.E. Spicer and F. Wooten, “Photo- ma and X-ray effects in multiplier photo- emission and photomultipliers,” Proc. tubes,” Naval Radiological Defense IEEE, Vol. 51, p. 1119-1126, (Aug. 1963). Laboratory, USNRDL-TR-763, (7 July 48. H. Martin, Cooling photomultipliers 1964). with III-V photocathodes,” Electro-Optical 61. W. Viehmann, A.G. Eubanks, G.F. Systems Design, Vol. 8, No. 11, p. 16-20 Pieper, and J.H. Bredekamp, “Photomulti- (Nov. 1976). plier window materials under electron ir- 49. E.G. Ramberg,“Optical factors in the radiation: fluorescence and phosphores- photoemission of thin films,” Appl. Opt., cence”Appl. Opt., Vol. 14, No. 9, p. Vol. 6, No. 12, p. 2163-2170 (Dec. 1967). 2104-2115 (Sept. 1975). 50. S.A. Hoenig and A. Cutler III, 62. R.B. Murray and J.J. Manning, “Polarization sensitivity of the RCA 6903 “Response of end-window photomultiplier photocathode tube,” Appl. Opt., Vol. 5, tubes as a function of temperature,” IRE No. 6, p. 1091-2 (June 1966). Trans. Nucl. Sci., Vol. NS-7, No. 2-3, p. 51. H. Hora,“Experimental evidence of 80-86 (June-Sept . 1960). angular dependence of electron excitation in 63. A.T. Young,“Temperature effects in photoemission”, Phys. Stat. Sol., Vol. (a) photomultipliers and astronomical photo- 5, p. 159-166 (1971). metry,”Appl. Opt., Vol. 2, No. 1, p. 52. W.D. Gunter, Jr., E.F. Erickson, and 5 l-60-(Jan. 1963). G.R. Grant, “Enhancement of photomulti- 64. R.E. Rohde, “Gain vs. temperature ef- plier sensitivity by total internal reflection,” fects in NaI(T1) photomultiplier scintillation Appl. Opt., Vol. 4, No. 4, p. 512 (April detectors using 10 and 14 stage tubes,” Nucl. 1965). Intr. and Methods, Vol. 34, p. 109-115 53. J.R. Sizelove and J.A. Love, III, (1965). “Analysis of a multiple reflective translucent 65. D.F. Cove11 and B.A. Euler, “Gain photocathode,”Appl. Opt., Vol. 6, No. 3, shift versus counting rate in certain p. 443-446, (March 1967). multiplier phototubes,” USNRDL-TR-521, 54. J.B. Oke and R.E. Schild, “A practical U.S. Naval Radiological Defense multiple reflection technique for improving Laboratory, San Francisco (1961). the quantum efficiency of photomultiplier 66. C.S. Wiggins and K. Earley, tubes,”Appl. Opt.,Vol. 7, No. 4, p. “Photomultiplier refrigerator,” Rev. Sci. 617-622 (Apil 1968). Instr., Vol. 33, p. 1057-8 (1962). 55. W. D. Gunter, Jr., G.R. Grant, and 67. For example, among others: Products S.A. Shaw,“Optical devices to increase for Research, Inc., 78 Holten Street, photocathode quantum efficiency,” Appl. Danvers, Massachusetts 01923 Opt., Vol. 9, No. 2, p. 25 l-257 (Feb. 1970). (617-774-3250). 56. D.P. Jones,“Photomultiplier sensitiv- Pacific Precision Instruments, 1040 ity variation with angle of incidence on the Shary Court, Concord, California 945 18 photocathode,”Appl. Opt., Vol. 15, No. 4, (415-827-9010). p. 910-914 (April 1976). 68. H.R. Krall, “Extraneous light emission 57. F.A. Helvy and R.M. Matheson, from photomultipliers,” IEEE Trans. Nucl. “Photosensitive cathode with closely adja- Sci. Vol. NS-14, No. 1, p. 455-9 (Feb. cent light-diffusing layer,” U.S. Patent 1967). 3 242,626, (Mar. 29, 1966). 69. Louis Lavoie, “Photomultiplier cath- 58. R.W. Engstrom, “Improvement in ode poisoning,”Rev. Sci. Instr., Vol. 38, photomultiplier and TV camera tubes for No. 6, pp 833-4, (1967). nuclear medicine,”IEEE Trans. Nucl. Sci., 70. G.A. Morton, H.M. Smith and R. Vol. NS-24, No. 2, p. 900-903 (Apil 1977). Wasserman, “Afterpulses in photomulti-

78 Photomultlpller Characteristics pliers,” IEEE Trans. Nucl. Sci., Vol. NS-14, York, (1972). No. 1, pp 443-448, (1967). 81. E. Gatti and V. Svelto, “Review of 71. V.O. Altemose, “Helium diffusion theories and experiments of resolving time through glass, ” J. Appl. Phys., Vol. 32, No. with scintillation counters,” Nucl. Instr. and 7, pp 1309-1316, (1961). Methods, Vol. 43, pp 248-268, (1966). 72. W.C. Paske,“He + afterpulses in 82. W.A. Baum,“The detection and mea- photomultipliers: Their effect on atomic and surement of faint astronomical sources,” molecular lifetime determinations,” Rev. Astronomical Techniques, Edited by W .A. Sci. Instr., Vol. 45, No. 8, p. 1001-3 (1974). Hiltner, The University of Chicago Press, 73. A.T. Young,“Cosmic ray induced (1962). dark current in photomultipliers,” Rev. Sci. 83. G.A. Morton,“Photon Counting,” Instr., Vol. 37, No. 11, pp 1472-1481 (1966). Appl. Opt., Vol. 7, No. 1, pp l-10, (1968). 84. R.M. Matheson, “Recent photomulti- 74. D. Sard, “Calculated frequency spec- plier developments at RCA,” IEEE Trans. trum of the shot noise from a photo- Nucl. Sci. Vol. NS-11, No. 3, pp 64-7 1, multiplier tube,”J. Appl. Phys., Vol. 17, p. (1964). 768-777 (1946). 85. G.A. Morton, H.M. Smith, and H.R. 75. B. Leskovar and C.C. Lo, “Transit time Krall, “Pulse-height resolution of high gain spread measurements of microchannel-plate first dynode photomultipliers,” Appl. Phys. photomultipliers,”European Conference on Lett., Vol. 13, No. 10, pp 356-7, (1968). Precise Electrical Measurements, Brighton, 86. G.H. Narayan and J.R. Prescott, Sussex, England, 5-9 Sept. 1977 (London, “Line-widths in NaI(Tl) scintillation count- England: IEEE (1977), pp 41-3.) ers for low energy gamma-rays,” IEEE 76. D.A. Gedcke and W.J. McDonald, Trans. Nucl. Sci.,Vol. NS-13, No. 3, pp “Design of the constant fraction of pulse 132-7, (1966). height trigger for optimum time resolution, ” 87. R.W. Engstrom and J.L. Weaver, “Are Nucl. Instr. and Meth., Vol. 58, No. 2, pp plateaus significant in scintillation count- 253-60 (1968). ing?” Nucleonics, Vol. 17, No. 2, pp 70-74, 77. M. Birk, Q.A. Kerns, and R.F. Tusting, (1959). “Evaluation of the C-70045A high-speed 88. The Current Status of Liquid- photomultiplier,”IEEE Trans. Nucl. Sci. Scintillation Counting, Edited by Edwin D. Vol. NS-11, No. 3, pp 129-138 (1964) Bransome, Grune and Stratton, New York 78. F. de la Barre,“Influence of transit time and London, (1970). differences on photomultiplier time resolu- 89. Liquid-Scintillation Counting, Recent tion,” IEEE Trans. Nucl Sci., Vol. NS-19, Developments, Edited by Philip E. Stanley No. 3, pp 119-121 (1972). and Bruce A. Scoggins, Academic Press 79. B. Leskovar, C.C. Lo, “Performance (1974). studies of photomultipliers having dynodes 90. Yasunori Nikaido et al, “Electron tube with GaP(Cs) secondary emitting surface,” for photoelectric conversion,” Patent IEEE Trans. Nucl. Sci., Vol. NS-19, No. 3, Publication No. SHO 42-16002 (Shimizu Pa- pp 50-62 (1972). tent Office, Fukiai-Ku, Kobe, Japan). 80. S.K. Poultney,“Single Photon Detec- 90a. S.M. Johnson, Jr., “Radiation effects tion and Timing: Experiments and Techni- on multiplier phototubes,” IEEE Trans. ques,” Advances in Electronics and Electron Nucl. Sci., Vol. NS-20, No. 1, pp 113-124, Physics, Vol. 31, Academic Press, Inc., New Feb. 1973.

79 Photomultiplier Handbook

5. Photomultiplier Applications

SUMMARY OF SELECTION CRITERIA APPLIED VOLTAGE A comparison of the relative advantages CONSIDERATIONS of photomultipliers and solid-state detectors Proper operation of a photomultiplier is given in the introductory chapter of this depends critically upon the applied voltage manual. The following is a summary of the and the voltage distribution to the dynode application requirements that indicate when, stages. A small variation in voltage can in general, a photomultiplier is the most result in a much larger percentage change in suitable detector. the tube gain. For example, see Fig. 49 which The most important consideration is the shows the log of the gain as a function of the light or radiation level to be detected. The log of the voltage. For a 9-stage tube the use of photomultipliers is recommended slope of the log-gain vs log-voltage curve is when the level of light flux is very low. If approximately 7; for a 14-stage tube the levels are relatively high, it may be simpler to slope is about 12. (If the secondary emission use a solid-state detector. Furthermore, a ratio were linear with voltage, see Fig. 19, high level of light flux could overload and one would expect the slope of the log-gain vs damage the photomultiplier. A photomulti- log-voltage curves to be equal to the number plier may be used if the light flux is 100 of stages.) For a 12-stage tube, a 1% change microlumens or less or of the order of 0.1 in voltage results in a 10% change in gain. microwatt or less at the peak of the spectral Thus, there is need for more than ordinary response. precaution to provide a well-regulated power When a fairly large detector area is re- supply l quired, a photomultiplier is also recom- mended, as is the case in scintillation count- Power Supply, Regulation, Polarity, and ing where a crystal scintillator may be several Shielding centimeters in diameter. Silicon avalanche Commercial power supplies are generally photodiodes have good low-light-level available that have a line-regulation of capability but their area is generally limited 0.005% with some available at 0.001% for a to a few square millimeters. 10% change in line voltage when working in- Spectral response of the photomultiplier, to a fixed divider network. Voltage variation of course, must be a reasonable match to with temperature might be of similar magni- that of the source of radiation. Photomulti- tude for a 1 °C change in temperature. It pliers are useful in the range 120 nanometers would not be too difficult, therefore, to pro- to 1100 nanometers, depending upon the vide voltage supplies that would result in type of photocathode and window material. gain stability of the photomultiplier tube due Responses further in the infrared than 1100 to line voltage and temperature changes of nanometers require an infrared photocon- better than 0.1%. For requirements not this ductive detector or some other infrared- critical, BURLE manufactures a simplified sensitive device. compact power supply that includes a socket When a fast response time is an important and voltage-divider network for 1-l/8inch requirement, the photomultiplier is usually diameter, 9-stage, side-on photomultipliers. the most suitable detector. Photomultiplier Also manufactured by BURLE are Inte- tubes have response-time capability down to grated Photodetection Assemblies (IPA’s) the nanosecond range and even better in the that package a photomultiplier tube, optical case of specially designed tubes. filter, power supply, signal-conditioning

80 Photomultiplier Applications amplifier, and electrostatic/magnetic shielding. The IPA operates from a 12-volt dc supply. The recommended polarity of the photo- multiplier power-supply voltage with respect to ground depends largely on the application intended. Of course, the cathode is always negative with respect to the anode. In some pulsed application, however, such as scin- tillation counting, the cathode should be grounded and the anode operated at a high positive potential with a capacitance-coupled output. In this case the scintillator and any magnetic or electrostatic shields should also be connected to ground potential. In applications in which the signal cannot be passed through a coupling capacitor, the positive side of the power supply should be grounded. The cathode is then at a high negative potential with respect to ground. When this arrangement is used, extra pre- - cautions must be taken in the mounting and 92CS - 32424 shielding of the photomultiplier. If there is a Fig. 87 - Schematic diagram of a resistive potential gradient across the tube wall, scin- voltage divider. tillations occurring in the glass will increase dark noise. If this condition continues for a sufficiently long period of time, the photo- cathode will be permanently damaged by the ionic conduction through the glass. To pre- vent this situation when a shield is used, the shield is connected to photocathode poten- tial. Light-shielding or supporting materials used in photomultipliers must limit leakage currents to 10- 12 ampere or less. Fig. 62 shows a curve of the effect of external-shield potential on photomultiplier noise. To reduce the shock hazard to personnel, a very high resistance should be connected between the shield and the negative high voltage. The following formula may be used to Voltage Divider Design calculate the voltage between stages: The interstage voltage gradients for the photomultiplier elements may be supplied by individual voltage sources. The usual source, however, is a resistive voltage divider placed across a high-voltage source, as shown in Fig. 87. A resistive voltage divider must be de- signed to divide the applied voltage equally or unequally among the various stages as re- The voltage-divider resistor values re- quired by the electrostatic system of the quired for each stage can be determined tube. The most common voltage between from the value of the total resistance re- stages is usually referred to as the stage vol- quired of the voltage divider and the voltage- tage and the voltage between other stages as divider ratios of the particular tube type.

81 Photomultiplier Handbook

The interstage resistance values are in pro- portion to the voltage-divider ratios as follows:

where Rj is the resistance between elements Dyj - 1 and Dyj.The recommended resis- tance values for a photomultiplier voltage divider range from 20,000 ohms per stage to 5 megohms per stage; the exact values are usually the result of a compromise. If low values of resistance per stage are utilized, the power drawn from the regulated power sup- ply may be excessively large. The resistor power rating should be at least twice the calculated power dissipation to provide a Fig. 88 - The relative response of a 931A pho- safety margin and to prevent a shift in tomultiplier as a function of the light flux us- ing the circuit of Fig. 87 with equal stage resistance values as a result of overheating. voltage (at zero light level). (From Engstrom The highest suitable value of stage resistance and Fischer91) (after consideration has been given to average anode current as described below) is The decrease in sensitivity that occurs dictated by leakage currents in the photo- beyond region A of Fig. 88 results from the multiplier and socket wiring. extension of voltage losses to the last two or One criterion for the selection of a suitable three dynode resistors causing defocusing range of voltage-divider resistance values is and skipping in the associated dynode the expected maximum anode current that stages. In order to prevent this loss and may be drawn from the photomultiplier. assure a high degree of linearity, the current When the anode current is of the same order through the voltage-divider network should of magnitude as the divider current, non- be at least ten times the maximum average linear response results. This non-linearity is anode current required. In calculating the illustrated in Fig.88 which shows the voltage-divider current, the average anode response of a 931A photomultiplier as a current must first be estimated; this estimate function of light level using a conventional requires knowledge of the value of the input voltage divider such as shown in Fig. 87 with (light) signal and the required output (elec- equal voltage per stage.91 (The value of RL trical) signal. was essentially zero for this measurement.) Photo-multiplier noise or a shift in gain The anode current is shown relative to the may result from heat emanating from the divider current at zero light level. The voltage-divider resistors. The divider net- superlinearity region is explained by a work and other heat-producing components, change which takes place in the voltage therefore, should be located so that they will distribution between stages from uniform to not increase the tube temperature. Resis- non-uniform as the light level is increased. tance values in excess of five megohms Thus, the electron current flow from the last should be avoided because current leakage dynode to the anode causes less current to between the photomultiplier terminals could flow through the voltagedivider resistor, cause a variation of the interstage voltage. The type of resistor used in a divider voltage between dynodes and a decrease in depends on the dynode structure with which the collection voltage between the last the divider will be used. Close-tolerance dynode and the anode. The reduced collec- resistors, such as the laser-trimmed metal- tion voltage tends to reduce the output cur- film types, are normally required with the rent slightly, but the increased dynode focused structures. On the other hand, inter- voltages more than compensate for this dynode voltages in the Venetian-blind struc- reduction by an increased gain. tures can vary widely with but little effect on

82 Photomultiplier Applications

the photomultiplier. For this reason, the overload the output stages of the photomul- resistors used with a Venetian-blind structure tiplier, it is possible to eliminate these stages can be of a less stable variety, such as com- from the circuit. For example, the last position. several stages of the tube and the anode may be tied together electrically to form an anode. The tube is then operated with the The over-all performance of a tube that reduced gain of the remaining stages. In an has a cathode-to-first-dynode region essen- extreme case, the photomultiplier may be tially electrostatically isolated from the re- operated just as a photodiode using only the maining dynode region can be improved by photocathode and several or all of the re- maintaining a high electric field, in the maining elements together as an anode. It cathode-to-first-dynode region to reduce the must be realized in operating the photomul- transit-time spread of photoelectrons arriv- tiplier in such a manner that many photo- ing at the first dynode and minimize the ef- cathode types are very resistive (see Figs. 35 fect of magnetic fields. A high first-dynode and 36) and, therefore, the photocurrent will gain, which implies a high not be linear with light at the high level which may be available. Photocathodes of the opaque type with a conductive substrate, or the semitransparent type with a conduc- tive undercoating will tolerate a much higher light level without loss of linearity.

have the disadvantage of reducing the gain in

Intermediate Stages In applications in which it is desirable to control the anode sensitivity without changing the over-all voltage, the voltage of a single dynode may be varied. Fig. 89 shows the variation of anode current for a 931A photomultiplier when one of the dynode voltages is varied while the total supply voltage is held constant. The dynode should 700 800

be selected from the middle of the dynode 92CS-32426 string because a variation of dynode poten- Fig. 89 - The output-current variation of a tials near the cathode or anode would have a 931A when the voltage on one dynode (No. 6) detrimental effect on photomultiplier opera- is varied while the total supply voltage re- tion. mains fixed. Operation with Fewer Stages A photomultiplier need not be operated Voltage Dividers for Pulsed Operation with all dynode stages, although the design In applications in which the input signal is of the tube has been optimized for this con- in the form of pulses, the average anode cur- dition. In some situations where the light rent can be determined from the peak pulse level being detected is so high that it would current and the duty factor. The total resis-

83 Photomultiplier Handbook

tance of the voltage-divider network is then decays exponentially through the anode calculated for the average anode current. load resistor with a time constant of RLCL. In cases in which the average anode cur- To prevent pulses from piling up on each rent is much less than a peak pulse current, other, the maximum value of RLCL should dynode potentials can be maintained at a be much less than the reciprocal of the nearly constant value during the pulse dura- repetition rate. tion by use of charge-storage at An important example of this type of the . The voltage-divider current operation is scintillation counting, for exam- need only be sufficient to provide the ple with NaI:Tl. The time constant of the average anode current for the photomulti- scintillations is 0.25 microsecond and, be- plier. The high peak currents required during cause the integral of the output current pulse the large-amplitude light pulses are supplied is a measure of the energy of the incident by the capacitors. radiation, the current pulse is integrated on The capacitor values depend upon the the anode-circuit capacitance for a period of value of the output charge associated with about 10 microseconds. Because the scin- the pulse or train of pulses. The value of the tillations occur at random, the maximum final-dynode-to-anode capacitor C is given average counting rate is limited to about 10 by kHz. If circumstances require a higher counting rate, the integration time must be reduced accordingly. Fast-Pulse Applications In fast-pulse light applications, it is where C is in farads, q is the total anode recommended that the photomultiplier be charge per pulse in coulombs, and V is the operated at negative high voltage with the voltage across the capacitor. The factor 100 anode at ground potential. A typical voltage- is used to limit the voltage change across the divider circuit with series-connected capaci- capacitor to a l-per-cent maximum during a tors is shown in Fig. 90. The parallel con- pulse. Capacitor values for preceding stages figuration of capacitors may also be used, as should take into account the smaller values shown in Fig. 91. The parallel arrangement of dynode currents in these stages. Conser- requires capacitors of higher voltage ratings. vatively, a factor of approximately 2 per Regardless of the configuration, the stage is used. Capacitors are not required capacitors must be located at the socket. The across those dynode stages at which the peak capacitor arrangements just described may dynode current is less than 1/10 of the also be applied to negative-ground applica- average current through the voltage-divider tions. network. For pulse durations in the 1 to 100 ns range, consideration should be given to the inherent stage-to-stage which are in the order of 1 to 3 pF. Medium-Speed Pulse Applications In applications in which the output cur- rent consists of pulses of short duration, the capacitance CL of the anode circuit to ground becomes very important. The capaci- tance CL is the sum of all capacitances from Fig. 90 - Series-connected capacitors in the anode to ground: photomultiplier-anode voltage-divider circuit using positive ground capacitance, cable capacitance, and the in- for pulse-light applications. put capacitance of the measuring device. For pulses having a duration much shorter than The wiring of the anode or dynode “pick- the anode time constant RLCL, the output off” circuit is very critical in pulse applica- voltage is equal to the product of the charge tions if pulse shape is to be preserved. Most and 1/CL because the anode current is simp- pulse circuitry uses 50-ohm characteristic im- ly charging a capacitor. The capacitor charge pedance cables and connectors because of

84 Photomultiplier Applications

92cs-32428 Fig. 91 - Parallel-connected capacitors in voltage-divider circuit for pulsed-light ap- plication.

their ready availability, although 75- and 0.5-nanosecond rise-time light pulse. The 92-ohm components are also used. Careful pulse shape is most easily seen with the aid of wiring is required. Figs 92 and 95 illustrate a repetitive light pulser and a high-speed schematically and pictorially the best loca- real-time oscilloscope. Fig. 94 indicates the tion of pulse bypass capacitors that return general type of distortion encountered with the anode pulse current to Dyn and Dyn _ 1 the use of improperly wired or excessively in- by paths of minimum residual inductance. It ductive capacitors.The output pulse il- should be noted that these bypass capacitors lustrates the ringing that may occur in an im- also serve the function of charge-storage properly wired socket. capacitors, as shown in Fig. 91.

Fig. 93 - Pulse shape obtained from photo- multiplier excited by a light pulse having an 0.5-nanosecond rise time.

92cs-32429 Fig. 92 - Bypass capacitors used to make the last two dynodes appear as a ground plane to a fast-pulsed signal.

Wiring Techniques Good high-frequency wiring techniques must be employed in wiring photomultiplier sockets and associated voltage dividers if 92cs-32430 pulse-shape distortion is to be minimized. Fig. 94 - Pulse shape distortion (ringing) en- Fig. 93 illustrates the pulse shape obtained countered with improperly wired or excessive- from an 8575 photomultiplier excited by a ly inductive capacitors.

85 Photomultiplier Handbook

Fig. 95 shows a socket wired for a negative photomultiplier current pulses is established high-voltage application. The disk-type at low pulse amplitudes where linear opera- bypass capacitors are mounted in series with tion is assured. Next, the light flux is in- minimum lead length because the self- creased in a series of increments. A constant- inductance of the lead wires becomes critical current pulse amplitude ratio indicates linear in nanosecond-pulse work. Care should also operation. A pair of light-emitting diodes be taken in dressing the bypass capacitors may be used to provide the two light pulses. and coaxial cable. - The resistors for the The over-all control of the light levels may voltage divider are shown mounted on the be done with a series of neutral-density socket. In applications requiring minimum filters. Exact filter calibration is not required dark current, the resistors should be remote because the system is self-calibrating in the from the photomultiplier to minimize pulse current measurements. heating effects. Tapered Dividers Some applications require that photomul- tipliers sustain high signal currents for short time intervals, tens of nanoseconds or less. In general, photomultipliers are capable of supplying 0.2 ampere or more into a 50-ohm load for short durations. However, the vol- tage divider must be tailored to the applica- tion to allow a photomultiplier to deliver these high currents. The principal limitation on current output

92cs-32432 (into a 50-ohm system) is space charge at the Fig. 95 - Anode detail of socket wired for a last few stages. This space charge can be negative high-voltage application showing overcome if the potential difference across location of charge-storage capacitors. the last few stages is increased by use of a tapered divider rather than an equal-volts- per-stage divider. The tapered divider places Checking Socket and Tube Performance 3 to 4 times the normal interstage potential Some photomultipliers typically display a difference across the last stage. The pro- reflected-pulse rise time of the order of 1.5 gression leading to the 4-times potential dif- nanoseconds for an initiating pulse of the ference should be gradual to maintain prop- order of 50 picoseconds when the tube and er electrostatic focus between stages; a pro- socket are tested with a time-domain reflec- gression of 1, 1 . . .1, 1, 1.5, 2.0, 3, 4.2 is tometer, the instrument used to test the recommended. anode-pin region of the socket. Another application of the tapered divider During tube operation, the output pulse is to provide a large signal voltage across a can be viewed with a real-time high-speed high-impedance load. Voltage excursions of oscilloscope. The pulse shape can be in- 400 to 500 volts can be obtained from photo- spected for signs of clipping or ringing. multipliers. A possible application is in driv- The effectiveness of the charge-storage ing an electro-optical modulator. Because capacitors can be verified with a simple the photomultiplier is nearly an ideal con- linearity test. The output of a photomulti- stant , its output voltage signal is limited only by the potential difference between the last dynode and the anode. If plot of pulse amplitude as a function of the the potential difference is 100 volts, the logarithm of the voltage value should yield a anode cannot swing through more than a straight line. 100-volt excursion. By impressing a much A two-pulse technique may be employed higher potential difference between the to test linearity at constant operating anode and last dynode by means of a tapered voltage. Consider two sequential pulses hav- divider, greater voltage swings can be ob- ing amplitude ratios in the range from 2:1 to tained. Tube data sheets should be consulted 10:1. The ratio of the amplitude of the two for maximum voltage ratings.

86 Photomultiplier Applications

Dynamic Compression of Output Signal tection for the tube. If overexposure is ex- Most photomultipliers operate linearly pected frequently,interdynode currents, over a of six or sevenorders which can be quite excessive, may cause loss of magnitude, a range few monitor devices of gain. In some applications it may be can accommodate without requiring range worthwhile to protect against dynode changes. When compression of the dynamic damage by using resistors in series with each range is desired, a logarithmic amplifier is dynode lead .91 sometimes used. The photomultiplier may Active Divider Network also be operated in a compressed-output Although normally, the divider network mode, however, without the need for addi- current should exceed the maximum photo- tional compression circuitry. multiplier output current by a factor of 10 or For example,in liquidscintillation more, it is possible by using the emitter- counting where several different isotopes follower characteristics of to pro- may be present, the output pulse height vide a power supply requiring much less might normally vary by as much as 100: 1. By divider current. Fig. 96 is an example of such limiting the potential difference between the an active divider network devised by C. R. last dynode and anode to, for example,10 Kerns92 As the dynode current increases, volts, space charge will limit the maximum the added current is diverted from the high- current that can be drawn. The anode pulse beta transistors rather than from the divider- can then be used for the coincidence timing resistor string, thus improving the voltage with a more convenient range of pulse regulation by a large factor. The capacitors heights. Energy measurements can then be shown are the usual ones for high-frequency made using the current at an earlier dynode bypassing. The circuit also retains a current- where space-charge saturation is not present. limiting action that prevents damage to the Current Protection of Photomultiplier photomultiplier. If a photomultiplier is accidentally ex- MECHANICAL CONSIDERATIONS posed to an excessive amount of light, it may be permanently damaged by the resultant Handling high currents. To reduce this possibility, the Because most photomultipliers have glass resistive voltage-divider network may be envelopes, they should be handled with care designed to limit the anode current. The to avoid damage to the tube seals and other average anode current of a photomultiplier parts. This caution is especially important cannot much exceed the voltage-divider cur- for tube types having graded-seal envelope rent; therefore, the zero-light level voltage- construction. The pins or leads of the tube divider network serves as an overload pro- should also be treated with care.

ANODE

Fig. 96 - An active divider network for an 8575 photomultiplier designed to minimize voltage changes at the dynodes. (From C.R. Kerns92)

87 Photomultiplier Handbook

Basing attached bases, or stiff leads. Semiflexible Photomultiplier tubes may have either a leads may be soldered, resistance (spot) temporary or a permanently attached base. welded, or crimp connected into the Dimensional outline diagrams such as those associated circuitry. When soldering or shown in Fig. 97 are provided in the pub- welding is employed for such connections, lished technical data for individual tubes. In- care should be taken to prevent tube destruc- dicated on the diagrams are the type of base tion due to thermal stress of the seals at the employed, maximum mechanical dimen- stem. A heat sink, such as locking forceps, sions, radii of curvature where applicable, should be placed in contact with the semi- pin/lead details, location and dimensions of flexible leads between the point being magnetic parts used (in tubes utilizing mini- soldered or welded and the tube seals. If mum number of magnetic materials), and soldering is employed, only a soft solder notes regarding restricted mounting areas, (e.g., 60% Sn, 40% Pb) should be used. again where applicable. Heat should be applied only long enough to Photomultiplier tubes intended to be permit the solder to flow freely. By the term soldered directly to circuit boards or hous- semiflexible, it is implied that excessive ings are supplied with semiflexible or “fly- bending may break the leads, most common- ing” leads and a temporary base, intended ly at the stem surface. Some photomulti- for testing purposes only, that should be re- pliers are supplied with insulating wafers at- moved prior to permanent installation. tached to the stem to prevent such an occur- A lead-terminal diagram that shows rence. The semiflexible leads are normally photomultiplier-tube lead orientation with made of dumet or Kovar and are usually the temporary base removed, shown in Fig. plated to facilitate soldering. 98, provides a lead indexing reference. A Photomultipliers supplied with perma- lead-connection diagram, such as the one nently attached bases or stiff leads should shown in Fig. 99, relates terminal to elec- use only high-grade, low-leakage sockets to trode. Care must be exercised in interpreting minimize leakage currents between adjacent basing and lead-terminal diagrams to insure electrode terminals. Teflon and mica-filled against possible damage to the photomulti- sockets should be used. plier resulting from incorrect connections. Mounting and Support Terminal Connections Photomultipliers having permanently at- BURLE photomultipliers are supplied with tached bases normally require no special either semiflexible leads, semiflexible leads mounting arrangements. When special attached to temporary bases, permanently mounting arrangements are used, however,

PHOTOCATHODE

B

92CM - 32434 Fig. 97 - Typical dimensional-outline drawings showing the type of base supplied with each tube and pertinent notes.

88 Photomultiplier Applications

metal flanges employed in the construction of a tube. Such flanges, when present, are part of the tube’s vacuum enclosure and any undue force or stress applied to them can damage the seals and destroy the tube. The use of resilient potting compounds or rubber washers is recommended when pho- tomultipliers are clamp-mounted. If a pot- ting compound is used, its characteris- tics-over the temperature range in which INDEX the tube is to be operated-must be such that its resilience is maintained at low temperature and its expansion, in confined space, is not excessive at high temperature. Fig. 98 - Lead-orientation diagram. The electrical insulation properties of any materials supporting or shielding the photo- multiplier should be considered. If such PIN I - DYNODE NO. I PIN 2 - DYNODE NO. 3 materials come into contact with high PIN 3 - DYNODE NO.5 PIN 4 - DYNODE NO.7 voltage with respect to photocathode, PIN 5 - DYNODE NO. 9 minute leakage currents can flow through PIN 6 - ANODE PIN 7 - DYNODE NO.10 the material and the tube envelope to the PIN 8 - DYNODE NO. B PIN 9 - DYNODE NO. 6 photocathode. Not only does this condition PIN IO - DYNODE NO 4 introduce excessive noise at the tube output PIN II - DYNODE NO. 2 PIN 12 - PHOTOCATHODE but it can also permanently damage the photocathode sensitivity of the tube through LEAD I - DYNODE NO. I electrolysis of the glass envelope. This cau- LEAD 2 - DYNODE NO. 3 LEAD3 -DYNODENO.5 tion is only true when the tube is operated at LEAD4 -DYNODENO.7 LEAD5 -DYNODENO.9 high negative potential with respect to LEAD6 - ANODE ground. Under this operating condition, a LEAD7 - DYNODE NO.10 LEAD 8 - DYNODE NO. 8 decrease in sensitivity can occur if the LEAD9 - DYNODE NO.6 LEAD 10 - DYNODE NO.4 faceplate of the tube comes into contact with LEAD II - DYNODE NO.2 ground. Cathode sensitivity does not recover LEAD 12 - PHOTOCATHODE after such an occurrence. Photocathode

BOTTOM VIEWS 92CS-32436 “poisoning” due to envelope electrolysis can destroy the usefulness of a photomultiplier Fig. 99 - Lead-connection diagram: (a) with in a very short time. Therefore, the in- base connected, (b) with temporary base removed. sulating property of materials supporting the tube should be such that leakage current to the tube envelope is limited to 1 x 10 -12 the envelope, especially that region near the ampere, or less. photocathode, must be maintained at cathode potential. Care should also be taken Shielding so that tube performance is not affected by Electrostatic and/or magnetic shielding of extraneous electrostatic or magnetic fields. most photomultipliers is usually required. Side-on photomultipliers should be mounted When such shields are used and are in con- to allow rotation of the tube about its major tact with the tube envelope, they must axis to obtain maximum anode current for a always be connected to photocathode poten- given direction of incident radiation. An tial. angular tolerance with respect to incident In applications where the dc component of light direction is normally specified in tube the signal output is of importance, the data sheets. cathode is normally operated at high nega- Direct clamping with non-resilient tive voltage with respect to ground. As a materials to the envelope of tubes not having result, the shield is at high negative voltage permanently attached bases is not recom- and precautions must be taken to avoid mended nor should clamping be made to any shorts to ground and to prevent shock

89 Photomultiplier Handbook hazard to personnel. A 10-megohm resistor dark current include external leakage caused should be placed between the negative high by condensation on the tube base and/or voltage and shields to avoid such hazards. socket when conditions of high humidity ex- In scintillation counting applications, it is ist and contamination of the tube base recommended that the photocathode be and/or socket by handling. operated at ground potential. In this case, Moisture condensation can be minimized the shields should be operated at ground by potting the tube socket assembly in potential. silicone rubber compounds such as RTV-11, Magnetic shielding of most photomulti- or equivalent. If a tube is suspected of pliers is highly desirable. Characteristic having high ohmic leakage as a result of curves showing the effects of magnetic fields handling, it is recommended that it and its on anode current are provided in many data socket be washed with a solution of alkaline sheets. cleaner such as Alconox*, or equivalent, and Storage de-ionized or distilled water having a tem- Photomultipliers should be stored in the perature not exceeding 60 °C. The base or the socket should then be thoroughly rinsed in dark. Storage of tubes in areas where light is de-ionized or distilled water (60 °C) and then incident on the tube results temporarily in a air-blown dry. higher than normal dark-current level when the tubes are placed in operation. This in- OPTICAL CONSIDERATIONS crease in dark current is primarily due to phosphorescence of the glass and can persist Incident Light Flux for about 24 hours. Additionally, storage of Photomultiplier tubes are capable of tubes designed for operation in the near IR operating usefully over a very wide range of region of the spectrum (above 700 nano- incident light flux. At the very lowest levels, meters) in illuminated areas may decrease the limit is determined by the useable signal- the “red” sensitivity of the tube. to-noise ratio. Upper range of usefulness is The phototube should never be stored or determined by the saturation of the photo- operated in areas where there are concentra- cathode or by the magnitude of the output tions of helium because helium readily current levels which either result in space- permeates glass. The composition of the charge limitations or cause damage to the envelope material is a major factor govern- secondary emission dynodes. ing the rate of helium permeation. As the The lower limit of detection is discussed in silica content in the glass is reduced, the rate detail in Chapter 4, Photomultiplier Charac- of permeation decreases. Accordingly, the teristics, in the section “Dark Current and rate of permeation is greatest for fused silica Noise.”See in particular Fig. 67. The lower and decreases to a minimum in lime glass. It limit in a current-measurement mode is de- is also to be noted that the rate of permea- termined by dark emission and bandwidth. tion is proportional to temperature and Photocurrent levels as low as 10 - 16 ampere varies directly with pressure.691 92a 92b (625 photoelectrons per second) can be Moisture Condensation measured. This photocurrent level cor- responds to 5 x 10 - 13 lumens for a A very small anode current is observed 200-microampere/lumen photocathode. In a when voltage is applied to the electrodes of a photoelectron-counting mode (see Appendix photomultiplier in darkness. Among the G), the limit is determined by the statistics of components contributing to this dark cur- discriminating between photoelectrons and rent are pulses produced by thermionic emis- thermally emitted electrons. For a Na2KSb sion, ohmic leakage between the anode and photocathode, the dark emission from the adjacent elements,secondary electrons photocathode at room temperature may be released by ionic bombardment of the pho- of the order of 10 - 17ampere or 62 electrons tocathode, cold emission from the elec- per second. Counting for one minute in the trodes, and light feedback to the photocath- dark and one minute in the light, one should ode. Other conditions contributing to anode be able to detect a photocurrent of only a *Distributed by Arthur H. Thomas Company, few photoelectrons per second. Vine Street and 3rd, Philadelphia, PA 19105. For resistive photocathodes (See Table II

90 Photomultiplier Applications

in Chapter IV, Photomultiplier Characteris- has the advantage of providing a reduction tics, in the section “Photocathode-Related factor that is essentially independent of Characteristics.“) such as K2CsSb, the max- wavelength. Such neutral density screens can imum recommended photocurrent is only of be obtained from the Varian Instrument Di- the order of 10 -9 ampere which cor- vision. ** responds to a flux of the order of 10 -5 Large reduction factors can be achieved lumen. by the use of opal glass, which scatters trans- When the photocathode resistivity does mitted light in an approximate Lambertian not limit the input light flux, the maximum (cosine) distribution (although a bit more light flux is determined by output currents in concentrated near the normal angles than the the photomultiplier. For example, the 931A cosine prediction). Unfortunately, the scat- maximum average anode current is 1.0 tered flux is not neutral, providing more red milliampere which corresponds to an input than blue. See Fig. 101. light flux of about 10 -5 lumen with a total applied voltage of 1000 volts. If the over-all voltage were reduced to 500 volts, the tube could tolerate an input flux of 10 - 3 lumen. At times it may be useful to reduce the light level on the photomultiplier by a calibrated amount. Neutral-density filters can be useful for this purpose. It should be appreciated, however, that such filters are frequently not neutral and their stated density may be only an approximation. A comparison of the spectral transmittance of a metalized filter and a gelatin type is shown in Fig. 100. Wratten filters are reasonably

W WAVELENGTH - NANOMETERS 92cs-32430 Fig. 101 - The effective spectral transmit- tance (for scattered light) of an opal-glass filter, 3-mm thickness, into a solid angle of approximately 0.01 steradian.

Convex front-surface spherical aluminum mirrors can also be used to provide a calibrated light-reduction factor. Aluminum mirrors have the advantage of having a Fig. 100 - Comparison of spectral transmit- reasonably flat spectral reflectance-from tance of metalized- and organic-type neutral- 92.3% at 300 nanometers to 86.7% at 800 density filters (ND 1.0). nanometers.93 If a lamp of CP candelas is located at a distance a from the surface of neutral in the visible range but become the mirror having a reflectivity e and a transparent in the infrared, which could radius r, the flux (L) in lumens through a test cause a problem when they are used with a aperture of area A located at a distance b photomultiplier having a near-infrared from the mirror surface is given by94 response. This effect is much more pro- nounced for the more dense neutral-density filters. Another method for reducing light level is by the use of a mesh or screen. This method l *670 E. Arques St., Sunnyvale, Calif. 94086 Photomultlpller Handbook

Angle of Incidence nent rays on the mirror. Photocathode response varies somewhat with the angle of incidence (see Fig. 41). It is Calibration also possible to increase the effective quan- For some purposes it may be useful to pro- tum efficiency of a photocathode by the use vide a calibration of the photomultiplier of a specially constructed optical coupling to anode or photocathode responsivity. A con- the photocathode window that utilizes at the venient test source is the tungsten lamp (see glass-air interface angles of incidence greater Appendix F). Lamps with candle power and than the critical angle. (See Fig. 42). color temperature (2856 K) traceable to the Light Modulation National Bureau of Standards can be ob- For some purposes, it may be advan- tained from BURLE, Lancaster, PA (Type AJ2239). Some method of reducing the light tageous to modulate the light signal with a flux such as discussed above would normally light chopper. A synchronous motor may be be required. used to rotate a chopper disk which could then provide an approximate square-wave Spectral response measurements generally modulation. The output signal could then be require a sophisticated monochromator set- analyzed with a narrow-band-pass amplifier up. Reasonably good results, however, can tuned to the chopper frequency. When the be obtained by use of calibrated narrow- modulated signal is in the presence of an un- band-pass color filters and a calibrated modulated background whether from the tungsten lamp. The power through a par- dark current of the tube itself or from an ex- ticular filter may be obtained by integrating ternal source of light, a significant improve- the product of the filter transmission and the tungsten irradiance over the wavelength ment in signal-to-noise may be achieved. band of the filter. Spectral irradiance for a Regarding the testing of photomultiplier tubes with delta-function light pulses, see the tungsten lamp calibrated to a color tempera- section on “Time Effects” in Chapter 4, ture of 2856 K may be obtained from Appen- Photomultiplier Characteristics. Cerenkov dix F, Table F-I, with a multiplying factor radiation also provides very short pulses of appropriate to the particular total luminous light. flux. Details are given in the footnote ac- companying Table F-I. SPECIFIC PHOTOMULTIPLIER APPLICATIONS Spot Size This section discusses various applications It is generally advisable to utilize a large of photomultipliers and some of the special part of the photocathode area rather than to considerations for each application. This focus a small spot of the light flux on the catalogue of applications is not complete photocathode. If the light flux is fairly high, even for presently known applications, and concentrating it on a small area could cause new ones are being continually devised. The damage to the photocathode or result in applications discussed, however, are some of non-linearity effects because of the resistiv- the major ones and the information given ity of the photocathode layer. Furthermore, can be readily adapted to other applications variations in photocathode sensitivity across or to new ones. the surface area could cause some uncer- tainty in the measurement if the light spot is Scintillation Counting too small. In scintillation-counting applica- A scintillation counter is a device used to tions, spot size is not generally a problem detect and register individual light flashes because of the diffuse nature of the flux and caused by ionizing radiation, usually in the the size of the crystal. But, in flying-spot- form of an alpha particle, beta particle, scanner applications it is particularly impor- gamma ray, or neutron, whose energy may tant that the image being scanned should not be in the range from just a few thousand be in focus on the surface of the photomulti- electron-volts to many million electron- plier because any non-uniformities of volts. The most common use of scintillation photocathode sensitivity would cause fixed counters is in gamma-ray detection and spec- pattern modulation. troscopy.

92 Photomultiplier Applications

The gas, liquid, or solid in which a scin- also results. To satisfy the conditions of con- tillation or light flash occurs is called the servation of energy and momentum, there is scintillator. A photomultiplier mounted in a maximum energy that can be transferred to contact with the scintillator provides the the electron. This maximum energy, known means for detecting and measuring the scin- as the Compton edge, occurs when e in Fig. tillation. Fig. 102 is a diagram of a basic given by

the speed of light. The resultant energy im- parted to the electron can then range from zero to a maximum of TCM. In pair production, the energy of a gamma ray is converted to an electron-positron pair in the field of a nucleus. The gamma ray Fig. 102 - Diagram of a scintillation counter. must have energy at least equal to two times the rest-mass-energy equivalent of an elec- scintillation counter. The three most prob- able ways in which incident gamma radiation energy is transferred as kinetic energy. When can cause a scintillation are by the photoelec- the positron is annihilated, two photons are tric effect, Compton scattering, or pair pro- produced 180 degrees apart, each with an duction. The reaction probabilities associ- energy of 0.51 MeV. The photons are then ated with each of these-types of interaction subject to the normal probabilities of in- are a function of the energy of the incident teraction with the scintillator . radiation as well as the physical size and In neutron detection, unlike alpha- or atomic number of the scintillator material. beta-particle or gamma-ray detection, the In general, for a given scintillator, the primary interaction is with the nuclei of the photoelectric effect predominates at small scintillator atoms rather than its atomic elec- quantum energies, the Compton effect at trons. The interaction may consist of scatter- medium energies, and pair production at ing or absorption; in either case, some or all energies above 1.02 MeV. of the energy of the neutron is transferred to Scintillation Processes. In the photoelectric the recoil nucleus which then behaves effect, a gamma-ray photon collides with a similarly to an alpha particle. bound electron in the scintillator and im- In each interaction between a form of parts virtually all its energy to the electron. ionizing radiation and a scintillator, an elec- In the Compton effect a gamma-ray photon tron having some kinetic energy is produced. A secondary process follows which is in- tron in the scintillator and transmits only dependent of both the kind of ionizing radia- part of its energy to the electron, as shown in tion incident on the scintillator and the type Fig. 103. A scattered photon of lower energy of interaction which occurred. In this secon- dary process, the kinetic energy of the ex- cited electron is dissipated by exciting other electrons from the valence band in the scin- tillator material into the conduction band. When these excited electrons return to the valence band, some of them generate light or scintillation photons. The number of pho- tons produced is essentially proportional to 92CS-32440 the energy of the incident radiation. In the Fig. 103 - Compton-effect mechanism. photoelectric interaction described above, all

93 Photomultiplier Handbook

of the incident photon energy is transferred the fluorescence, or , centers in the to the excited electrons; therefore, the generation of excitons or bound hole- number of photons produced in this secon- electron pairs. These pairs behave much like dary process, and hence the brightness of the hydrogen atoms and, being electrically scintillation, is proportional to the energy of neutral, can wander freely through the the incident photon. crystal until captured by fluorescence Scintillation Mechanism. The exact mecha- centers. The emission of a photon from a nism of the scintillation or light-producing phosphorescence center is a relatively slow process is not completely understood in all or delayed process. Of the three types of types of materials; however, in an inorganic centers through which an excited electron scintillator, the phenomenon is known to be can return to the valence band, the first, caused by the absorption of energy by a fluorescence, is that sought for in the valence electron in the crystal lattice and its preparation of scintillators. The second subsequent return to the valence band. Fig. type, quenching, tends to lessen the effi- 104 shows a simplified energy- ciency of the scintillator because it does not cause the emission of photons, and the third, phosphorescence, produces an undesirable background glow. Scintillator Materials. The most popular scintillator material for gamma-ray energy IMPURITY - _ - LEVELS spectrometry is thallium-activated sodium iodide, NaI:Tl. This material is particularly good because its response spectrum contains a well-defined photoelectric peak; i.e., the material has a high efficiency or probability of photoelectric interaction. In addition, the light emitted by the material covers a spec- 92CS-32441 tral range from approximately 350 to 500 Fig. 104 - A simplified energy diagram of a nanometers with a maximum at about 410 scintillation crystal: ionized electron-hole pairs; an exciton; an activator center through nanometers, a range particularly well which an electron may return to the ground matched to the spectral response of conven- state causing fluorescence; or through which tional photomultipliers, NaI:Tl does not, an electron may return to the ground state by however, have a fast decay time in com- a thermal, non-radiative process; impurity parison to other scintillators, and, therefore, levels that can trap an electron for a time is generally not used for fast-time resolution. causing phosphorescence. As a comparison, the decay time constant for NaI:Tl is approximately 250 nano- of a scintillation crystal. The presence of seconds, while for a fast plastic scintillator energy levels or centers between the valence the dominant decay time constant is in the 1 and conduction bands is the result of im- to 4 nanosecond range. perfections or impurities in the crystal lat- The decay time of a scintillator involves tice. Three types are important: (1) fluor- the time required for all the light-emitting escence centers in which an electron, after luminescence centers to return excited elec- excitation, quickly returns to the valence trons to the valence band. In some of the band with the emission of a photon; (2) better scintillators the decay is essentially ex- quenching centers in which the excited elec- ponential, with one dominant decay time tron returns to the valence band with the constant. Unfortunately, most scintillators dissipation of heat without emission of light; have a number of components each with dif- and (3) phosphorescence centers in which the ferent decay time. excited electron can be trapped in a Collection Considerations. Because scin- metastable state until it can absorb some ad- tillations can occur anywhere in the bulk of ditional energy and return to the valence the scintillator material and emit photons in band with the emission of a photon. An im- all directions, there exists the problem of col- portant process in the transfer of energy to lecting as many of these photons as possible

94 Photomultiplier Applications on the faceplate of the photomultiplier. If it cases, a flexible fiber-optics bundle can be is assumed that the fluctuation in the used. An optically transparent silicone-oil number of photons resulting from a single coupling fluid should be applied at the scin- ionizing event follows simple Poisson statis- tillator (-light guide, if used)-photomultiplier tics*, the relative standard deviation in the interface regardless of the light-conduction number arriving at the face plate is given by method used. The next important consideration in scin- (32) tillation counting is the conversion of the photons to photoelectrons from the photo- where Np is the average number of photons cathode. The photocathode should have the arriving at the faceplate of the photo- greatest quantum efficiency possible over the multiplier per incident ionizing event. Np is spectral range defined by the spectral emissivity curve of the scintillator. The ber of photons per unit energy for the scin- method of determining the quantum effi- tillation, E is the energy of the incident ciency of a photocathode as a function of radiation, and t is the fraction of the total wavelength is explained in Appendix E. Of number of photons produced which arrive at the various photocathodes available, bialkali the faceplate of the photomultiplier. types, K2CsSb and Rb2CsSb, having peak Eq. (32) implies that it is highly desirable quantum efficiencies in excess of 25%, pro- that all emitted photons be collected at the vide the best spectral match to the emission photomultiplier faceplate. This collection from most scintillators. problem can be simplified by careful selec- The uniformity of the photocathode, i.e., tion of the shape and dimensions of the scin- the variation in quantum efficiency at a tillator to match the photomultiplier photo- given wavelength as a function of position cathode dimensions. The coating of all sides on the photocathode, is also important. of the scintillator except that which is to be Because the number of photoelectrons exposed to the photomultiplier faceplate emitted for a constant number of photons with a material that is highly reflective for incident on the photocathode is proportional the wavelengths of the photons emitted by to the quantum efficiency any variation in the scintillator also proves helpful. Because quantum efficiency as a function of position NaI:Tl is damaged by exposure to moist air, results in an undesirable variation in the it is usually packaged in an aluminum case number of photoelectrons emitted as a func- lined with highly reflective MgO or Al2O3 tion of position. powder; the NaI:Tl scintillator is provided When the scintillation is fairly bright, i.e., with an exit window of glass or quartz. To when a large number of photons are pro- avoid total internal reflection, it is important duced per scintillation, and when the scin- that the indices of refraction of the scin- tillator is thick in comparison to its tillator material, its window, any light guide, diameter, as shown in Fig. 105(a), the and the photomultiplier faceplate match as photocathode is approximately uniformly il- closely as possible. luminated during each scintillation. How- If it is not convenient for the photomulti- ever, if the scintillator is thin in comparison plier to be directly coupled to the scintillator, to its diameter, as shown in Fig. 105(b), or if as when the photomultiplier entrance win- the scintillation is very weak, the illumina- dow is not flat, light guides can be used. tion of the photocathode as a function of Again, care should be taken in the design of position is closely related to the position of the light guide to assure maximum light origin of the scintillation; therefore, photo- transmission. The outer side or surface of cathode uniformity is much more important the light guide should be polished and coated when a thin scintillator is used. In this case, with a highly reflective material. In some the use of a light guide may be advan- tageous. Photocathode uniformity also be- *Actually, the variance in the number of photons ex- comes more important as the energy of the iting from the crystal per event is much higher than ex- pected from the simple expression of Eq. (32). See the incident radiation becomes less and the discussion in Chapter 4 on “Scintillation Counting” number of photons per disintegration is related to Eq. G-l 11. reduced.

95 Photomultiplier Handbook

REFLECTIVE either by reflected light or by excitation of ,-COATING photoelectrons from photocathode deposited on the side walls of the envelope. At high count rates, tubes having copper- beryllium dynodes generally provide greater stability than tubes having cesium-antimony dynodes, although for low count rates the latter prove to be satisfactory. A tube having good stability may be expected to shift in gain by no more than 7% in several months of continuous operation at a count rate of 10,000 per second. Variation of pulse height REFLECTIVE or gain with count rate is also of importance. /-COATING Well-designed tubes should show a variation in pulse height of less than 1% between COUPLING FLUID count rates of 1000 per second and of 10,000 per second, related to the photopeak, using 137Cs and a NaI:Tl crystal. (b) Photomultiplier dark noise is of particular 92CS -32442 importance in scintillation-counting applica- Fig. 105 - Scintillator geometries: (a) thick in tions when the energy of the ionizing radia- comparison to diameter; (6) thin in com- tion is small, or when very little energy is parison to diameter. transferred to the scintillation medium; in short, when the flash per event represents In some photomultipliers, the collection only a few photons. If a photomultiplier is efficiency for photoelectrons decreases near coupled to a scintillator and voltage is ap- the outer edges of the photocathode; plied, the composite of all noise pulses therefore, best results are obtained when the coming from the photomultiplier is referred scintillator is slightly smaller in diameter to as the background of the system. The plot than the photocathode. of a frequency distribution of these pulses as a function of energy is shown in Fig. 106. Significant Photomultiplier Characteristics. In scintillation-counting applications (See also the section on “Pulse Counting” in Chapter 4, Photomultiplier Characteristics), a photomultiplier should be selected to pro- vide a photocathode diameter matching that of the crystal to be used. The most important characteristic of the tube to be used in scin- tillation counting is then the effective photo- cathode sensitivity to blue and near-ultra- violet wavelengths. The effective photocath- ode sensitivity includes the basic quantum PULSE HEIGHT efficiency and the collection efficiency of the electron-optical system. Tubes having 92CS-32443 Venetian-blind dynodes provide a fairly large Fig. 106 - Distribution of radiation back- ground pulses as a function of energy. A cali- opening to the first dynode area and thus bration spectrum showing the 137CS photo- frequently have better collection efficiency peak is included for reference. than tubes having a focused-dynode struc- ture, although the focused structure is much better for time resolution. Photomultipliers The well-defined peaks seen in the figure are having the recent “tea-cup” type of first caused by external radiations and can be dynode generally have very good collection reduced by placing the photomultiplier and efficiency and, in addition, are designed to scintillator in a lead or vault to reduce provide increased photocathode emission background radiation.

96 Photomultiplier Applications

It is desirable that the background count with a velocity greater than the velocity of of the scintillation counter be as low as light in the dielectric (i.e., v greater than c/n, possible. Photomultipliers have been devel- where n is the index of refraction of the oped in which the background count is kept dielectric). Polarization of the dielectric by low by the use of radioactively clean metal the particle results in the development of an cans instead of glass-bulb enclosures and electromagnetic wave as the dielectric faceplates of either Lucalox* or sapphire in- relaxes. If the velocity of the particle is stead of glass. As an additional aid in reduc- greater than the velocity of light, construc- ing background, the scintillation counter can tive interference occurs and a conical wave- be surrounded by another scintillator, such front develops, as illustrated in Fig. 107. as a plastic sheet equipped with its own photomultipliers. The output from this scin- with respect to the direction of the particle is tillation shield is then fed into an anticoin- given by the expression cidence gate with the output from the scin- tillation counter. The gating helps to reduce the background contributions from both in- (33) ternal and external radioactive con- taminants. where n is the index of refraction of the Liquid Scintillation Counting In liquid scintillation counting, the scin- velocity to the velocity of light. The spectral energy distribution of the radiation increases tillation medium is a liquid and the ionizing radiations to be detected are usually low- limited by the absorption of the medium. energy beta rays. Because low-energy beta rays cannot penetrate a scintillation con- Cerenkov radiation is the electromagnetic tainer, the radioactive material is dissolved counterpart of the shock wave produced in a in a solution containing the scintillator. The gas by an object traveling faster than sound. scintillator is generally one or more fluores- It is highly directional and occurs mostly in cent solutes in an organic solvent. The beta- the near-ultraviolet part of the elec- ray energy is transferred by ionization and tromagnetic spectrum. Because the radiation by excitation that in turn results in the emis- is propagated in the forward direction of sion of photons in the near ultraviolet. In motion of the , as shown in some cases, a wavelength shifter is also used Fig. 107, Cerenkov detectors can be made to to provide radiation at somewhat longer detect only those particles that enter the wavelengths more compatible with the spec- system from a restricted solid angle. tral responsivity of the photomultiplier. Cerenkov radiation produced in an Important characteristics of the photo- aqueous solution by beta emitters can be multiplier in this application are again high useful in radioassay techniques because it is responsivity in the spectral region of the unaffected by chemical quenching and scintillation and low background count rate. because it offers the advantages, over liquid- A useful figure of merit is the square of the scintillation counting techniques, of simpli- counting efficiency divided by the back- fied sample preparation and the ability to ac- ground coincident count rate in a paired commodate large-volume samples. Because photomultiplier arrangement. Further a fast particle is required to produce details of liquid scintillation counting ap- Cerenkov radiation, rather high-energy beta plications are discussed in the section “Pulse rays are required; e.g., the threshold for Counting” in Chapter 4, Photomultiplier Cerenkov radiation is 261 keV for electrons Characteristics. in water. Because the photon yield for Cerenkov Radiation Detection Cerenkov light is usually very low, the same considerations concerning photomultiplier Cerenkov radiation is generated when a selection apply for the Cerenkov detection as charged particle passes through a dielectric for liquid-scintillation counting. The tube selected should also be equipped with a *Registered Trade Name for General Electric Co. faceplate capable of good ultraviolet trans- material. mission. In some experiments it may be im-

97 Photomultiplier Handbook

portant to select photomultipliers for their be as short as possible, and the geometry and speed of response. Depending upon the reflective coatings of the scintillator should dispersion of the medium, the duration of be selected so that variations in path lengths the Cerenkov flash can be very short. of photons from the scintillator to the photocathode of the photomultiplier are minimized. The photocathode of the photo- multiplier selected should have a high quan- tum efficiency. In addition, the transit-time dispersion or jitter (variations in the time re- quired for electrons leaving the photocath- ode to arrive at the anode of the tube) should be small over the entire photocathode area. The major contribution to transit-time spread occurs in the photocathode-to-first- dynode region and may be a result of the in- itial kinetic energies of the emitted photo- electrons and their angle of emission. Focus- ing aberrations,and the single-electron response or rise time, i.e., the output-pulse shape at the anode for a single photoelectron impinging on the first dynode, may also be of some importance. Although the single- electron response theoretically does not have much effect on time resolution, it does change the triggering threshold at which the best time resolution can be obtained. The most commonly used time-spectro- scopy techniques include leading-edge tim- ing, zero-crossover timing, and constant- Time Spectroscopy fraction-of-pulse-height-trigger timing. The technique used depends on the time resolu- In addition to the energy spectroscopy tion and counting efficiency required and the described above, there are occasions when it range of the pulse heights encountered. A is of advantage to measure time differences block diagram of a basic time spectrometer such as between a pair of gamma rays or a is shown in Fig. 108. combination of gamma rays and particles in Leading-edge timing makes use of a fixed cascade de-exciting some level in a nucleus. threshold on the anode-current pulse and In time spectroscopy, some special con- provides good time resolution over a narrow siderations must be made in selecting the range of pulse heights. The fractional pulse scintillator, photomultipliers, and technique height F at which the triggering threshold is of analyzing the signals from the photomul- set is defined as follows: tipliers. In a photomultiplier, time resolution is proportional to (n)- 1/2, where n is the (34) average number of photoelectrons per event. It is therefore important to choose a scin- tillator material that provides a high light where Vt is the discriminator threshold, and yield for a given energy of detected radia- Va is the peak amplitude of the anode- tion. It is also important that the variation of current pulse. The fractional pulse height the time of interaction of the radiation with has a considerable effect on the time resolu- the scintillator be as small as possible. This tion obtained; best results are usually ob- minimum variation is assured by attention to tained with F equal to 0.2. scintillator thickness and source-to-detector In fast zero-crossover timing, the anode geometry. The decay time constant of the pulse is differentiated. This differentiation light-emitting states in the scintillator should produces a bipolar output pulse that triggers

98 Photomultiplier Applications

TIMING PULSE DETECTOR I NUMBER TIME OF EVENTS SPECTRUM

SOURCE OF COINCIDENT RADIATION

t

MULTICHANNEL DETECTOR 2 (PULSE HEIGHT) ANALYZER

Fig. 108 - Generalized block diagram of a time spectrometer. the timing discriminator at the zero A variety of sondes are used in selected crossover, the time required to collect ap- combinations to determine various aspects proximately 50 per cent of the total charge in of the lithology (the character of a rock for- the photomultiplier pulse. Zero-crossover mation), including density, of the media timing is second to leadingedge timing for along the bore hole. The combinations of time-resolution work with narrow pulse- sondes used depend very much on the bore- height ranges, but is better than the leading- hole media. For example, when a formation- edge method for large pulse-height ranges. density sonde is used in combination with a In constant-fraction triggering, the point neutron sonde in liquid-filled bore holes, on the leading edge of the anode-current both lithology and porosity can be deter- pulse at which leading-edge timing data in- mined. The same pair of sondes allows the dicate that the best time resolution can be measurement of gas and liquid saturation in obtained is used regardless of the pulse bore holes drilled through reservoirs of low- height. For this reason, constant-fraction- pressure gas. The use of the formation- of-pulse-height timing is the best method for density sonde, along with either an induction obtaining optimum time resolution no or a sonde, permits a similar type of matter what the pulse-height range. determination. The final result of the logging activity is information concerning the existence of hydrocarbons and other geological media of interest in establishing an oil field. Oil-Well Logging The formation-density sonde is one of the Logging is the term given to the method of more sophisticated logging devices. Its determination of the mineral composition operational elements are encased in a rugged and structure a few miles under the earth’s cylindrical housing, as shown in Fig. 109. It surface. Oil-well logging companies gather is designed to withstand the high tempera- data by means of probes, or sondes, that ex- ture and shock encountered in probing bore amine the geological media along very deep holes miles deep under the earth’s surface. bore holes. The probes determine various The sonde contains a gamma-ray source, physical and chemical characteristics of the such as radioactive cesium 137, a detector material in their vicinity. Measurements consisting of a sodium iodide crystal and a made by the probes comprise the log. photomultiplier tube, a gamma-ray shield

99 Photomultiplier Handbook

ELECTRONICS

PRESSURE FOOT

GAMMA SHIELD

Fig. 109 - Formation-density sonde sends gamma rays into rock formation and then detects returns with photomultiplier.

for the detector, a pressure foot that presses lustrated in Fig. 110. Most of the decrease is the sonde against the bore-hole wall, and the result of photocathode sensitivity loss, operating electronics. The objective of this but some is associated with the crystal. Be- sonde is to determine the bulk density of the cause of the loss in photocathode response material in the region of the probe. and the increase in thermionic emission with Gamma rays from the source interact with temperature, the desired signal is finally lost the atoms in the geological medium. in the background noise at a temperature Compton-scattered gamma rays are detected near 200°C. Permanent damage to the pho- by the crystal and photomultiplier in the tocathode may also be expected after many sonde. A knowledge of gamma-ray penetra- cycles of operation at 200°C. tion of bulk media, mass absorption data, and the special effects resulting from the chemical nature of various geological media allows the information imparted by the pulses to be deciphered by the electronics into bulk-density data, which becomes the substance of the geological formation- density log. Because of the increase in temperature with the depth of the bore hole-to perhaps 150°C at 15,000 feet-a most important characteristic of the photomultiplier is its 92cs - 32447 resistance to high temperature. Of the Fig. 110 - Pulse-height resolution and pulse numerous photocathodes which have been height as a function of temperature for a pho- tomultiplier having a Na2KSb photocathode developed, the Na2KSb “bialkali” photo- 137 cathode is the most stable at elevated tem- and used with a Nal:TI crystal and a Cs peratures.95 (See Fig. 83 and related text.) source. As the temperature is increased on a pho- tomultiplier with a Na2KSb photocathode Gamma-Ray Camera coupled to a NaI:Tl crystal, the pulse height The gamma-ray camera originally de- (137Cs source) gradually decreases, as il- scribed by Anger96is a more sophisticated

100 Photomultiplier Applications version of the scintillation counter, and is on the left side of the organ are caused to im- used for locating tumors or other biological pact the left side of the scintillation crystal, abnormalities. The general principle of the etc. The crystal covers an area about 10 is illustrated in Fig. 111. A inches or more in diameter. Behind the crystal are, perhaps, 19 photo- multiplier tubes in a hexagonal array. The light of the individual scintillation is not col- limated but spreads out to all of the 19 tubes. The location of the point of scintillation origin is obtained by an algorithm depending upon the individual signals from each of the photomultipliers. Resolution is obtained in this manner to about 1/4 inch. Each scintilla- tion is then correspondingly located by a single spot on a cathode-ray tube. Counting is continued until several hundred thousand 92CS-32448 counts are obtained and the organ in ques- Fig. 111 - Structure of gamma-ray camera. tion is satisfactorily delineated. Fig. 112 is a reproduction of such a scintigram. The par- radioactive isotope combined in a suitable ticular advantage of the gamma camera over compound is injected into the blood stream other techniques such as the CT scanner of the patient or is ingested orally. Certain (described below) is that the gamma camera compounds or elements are taken up prefer- provides functional information. For exam- entially by tumors or by specific organs of ple: Tc-99m polyphosphate is used to reveal the body, such as iodine in the thyroid gland. bone diseases;123I is used in thyroid studies; As the radioactive isotope disintegrates, 127Xe is inhaled to provide information on gamma rays are ejected from the location of lung ventilation. the concentration. Of great concern to the designer of gamma A lead collimator permits gamma rays to cameras and, of course, to the ultimate pass through it only when they are parallel to customer is the inherent resolution capability the holes in the lead; gamma rays at other of the device. Generally, with more photo- angles are absorbed in the lead. In this way, multiplier tubes sampling the scintillation the location of the gamma-ray source may be distribution, the delineation becomes more determined because gamma rays originating precise. Most of the original cameras utilized

Fig. 172 - Scintigrams obtained by Lancaster General HospitaI with a gamma camera. The scin- tiphoto on the left shows multiple emboli in the right lung; photo on right shows lungs after clearing. The isotope technitium-99m was used to tag albumin microspheres-a colloidal form of the albumin protein, with particles ranging in size from 2 to 50 microns. These particles are injected into the bloodstream and are filtered and trapped in the lung capillary bed; the scan can then determine those areas where the capillary bed is intact. Areas of diminished blood flow show as “cold spots.”

101 Photomultiplier Handbook a hexagonal array of 19 tubes. By adding This recalibration could be done periodically another circumferential row, an array of 37 (once a day) and would contribute signifi- provided better resolution. In the same man- cantly to consistent, distortion-free opera- ner cameras are now available with 61 tubes tion. and even 91. (Note the numerical progres- Computerized Tomographic sion of these hexagonal arrays: 1 + 6 + 12 + X-Ray Scanners 18 + 24 + 30.) Different photomultiplier dimensions also are used to provide in- The computerized tomographic (CT) scan- struments with appropriate portability or ner produces an X-ray image in an entirely coverage. Most common has been the 3-inch different manner from that of conventional photomultiplier,but a large number of radiography.* The standard X-ray picture is 2-inch tubes are also used and there is con- a shadowgraph. Thus, a chest X-ray pro- sideration of the use of 1 1/2-inch tubes. Hex- duces an overlay of shadows from the rib agonal 2- and 3-inch tubes have been cage and internal body structure. Interpreta- developed to provide better space utilization, tion is frequently difficult because of the in- and a photomultiplier with a square terfering images. The CT scanner, on the faceplate is now available. other hand, provides a density image that represents a cross section of the patient-a Resolution of the gamma-ray camera de- tomograph. Thus, a CT scan of the head pends fundamentally upon the pulse-height would show the outer bone structure, the resolution of the photomultiplier and, folds of the brain, and possibly a tumor in- hence, the quantum efficiency of the photo- side the skull as though a complete thin slice cathode. The most common photocathode had been taken through the middle of the selected for gamma-ray-camera application head. This tomographic image is produced is the bialkali (K2CsSb). Also important in by exposing the head to X-rays at many dif- the determination of good pulse-height ferent angles of entrance. The multiplicity of resolution is collection efficiency of the shadow-type images thus formed are ana- photoelectrons. Large first dynodes are ap- lyzed by a computer, which produces a propriate and the “tea-cup” configuration is reconstructed cross-section density image. used to advantage. A diagram of a typical CT scanner is Because each scintillation is sensed by a shown in Fig. 113. An X-ray source pro- number of the photomultipliers in the array, ducing a fan beam rotates around the patient the spatial uniformity of the photomultiplier in a few seconds. (Short periods are desirable response and its angular response become to minimize artifacts in the final recon- important especially as related to the structed image of the patient’s cross section algorithm of the gamma-ray-camera design. caused by unavoidable motions.) An array The camera design may also incorporate of several hundred detectors in an outer cir- modifications in the light pipe such as cle surrounding the patient provides the data grooves or etched patterns to provide im- from which a computer derives a tomo- proved spatial location of the scintillations. graphic view of the patient’s body or head. Finally, the stability of the photomulti- A typical tomograph is shown in Fig. 114. plier is important in maintaining the resolu- Three different X-ray detector systems tion and spatial integrity of the display. have been developed for use in CT scanners. Although the photomultiplier currents are All are in use at present by the various CT- generally small in this application, changes scanner manufacturers. Each detector can occur in the photomultiplier gain which system has its problems, but all seem com- would then result in positional errors in the parable in ultimate performance. CRT display.Some gamma-ray-camera The original CT-scanner development designs include a means of recalibrating the utilized a crystal and photomultiplier. A array of photomultipliers so that the pulse- subsequent development used a tube con- height “window” is the same for each tube. taining gas at several atmospheres. The density of the gas and the high atomic "For a more detailed discussion, see the article in Scientific American, October, 1975, p 56, “Image weight of xenon provide sufficient absorp- Reconstruction from Projections,” R. Gordon, G.T. tion so that a large fraction of the X-rays are Herman, and S.A. Johnson. detected by the resulting ion current. More

102 Photomultiplier Applications

ARRAY OF STATIONARY DETECTORS ENCIRCLING PATIENT

92cS-32662 Fig- 113 - Diagram of a typical CT scanner. X-ray source producing a fan beam rotates around the patient. Array of several hundred detectors in outer circle pro vides the data from which a computer derives a tomographic (cross-section) view of the patient’s body or head. recently, detector packages have been The photomultiplier system uses a BGO developed using silicon photocells and a (bismuth germanate) crystal usually coupled CdWO4 crystal. This latter detector package to the photomultiplier by means of a light appears to have cost advantages. Detection pipe. Spectral emission from the crystal is il- with the silicon photocell was not feasible in lustrated in Fig. 115. It is rich in the blue- the early development of CT scanners, but a combination of higher-speed machines and the improved light output of the CdWO4 crystal, relative to the Bi3Ge4O12 (BGO) crystal, has made this combination an economic possibility.

I I I I I 300 400 500 600 WAVELENGTH-NANOMETERS

92CS-32452 Fig. 115 - Fluorescence spectrum of Bi4Ge3O12 (Data from Weber and Mon- champ97).

Fig. 114 - Tomograph showing a “slice” of green region and thus is a good match for the thoracic cavity. Lung tissue, blood ves- bialkali photocathodes. Recently, it has been sels, air passages, ribs, and spine may be ob- found that the Rb2CsSb photocathode is served. (Courtesy Pfizer Medical Systems, more suitable for use with the BGO crystal Inc.). than the K2CsSb photocathode because of

103 Photomultiplier Handbook the somewhat higher response and better spectral match. See Fig. 10. The Rb2CsSb photocathode is also more suitable because it has a lower surface resistance than the K2CsSb photocathode. See Fig. 36. In the CT-scanner systems, pulse-height discrimination is not used nor are individual scintillations counted. The photomultiplier is used as an integrator of the light flux in each density measurement. It is important that the photomultiplier provide a linear translation of the light flux into output cur- rent and also that there be no overshoot or undershoot in as the X-ray 92CS -32453 beam path changes from outside to through fig. 116 - Principle of positron coincidence the patient resulting in a signal variation of detection. A disintegration at “a ” results in 1000:1 or more. Typical maximum photo- oppositely directed gamma rays which are cathode current (X-ray path through air detected in coincidence by the pair of scintil- only) is of the order of 2 nanoamperes de- lator-photomultiplier detectors. An event at "b” results in a count in only one of the detec- pending on operation conditions. The pho- tors. (From G.L. Brownell and C.A. Burn- tomultiplier is usually operated at a modest ham98) gain figure of the order of 60,000. High photocathode sensitivity is important in pro- one plane may be utilized to provide three- viding minimum noise in the signal so that dimensional data. the inherent signal-to-noise ratio in the Reponse time of the photomultiplier is of transmitted X-ray beam is retained and the particular importance in positroncameras X-ray dose to the patient in minimized. The because the discrimination againstspurious photomultiplier tubes used are either 1/2 inch coincidences improves as the resolution time or 3/4 inch in diameter, depending upon the of the system decreases. Photomultiplier configuration of the particular CT scanner. transit time spreads of the order of 2 Positron Camera nanoseconds or less are advantageous in these systemsalthough other time- factors, vides tomographic presentations based on related to thecrystal scintillator (CsF has a coincident gamma-ray emission accompany- time constant of 5 nanoseconds) and cir- ing annihilation of a positron and an elec- cuitry may limit the system discrimination 11 13 time to perhaps 10-20 nanoseconds. Photo- tron. Tracer radionuclides such as C, N, 1 or 15O emit a positron upon disintegration. multiplier tubes of 3/4-inch and 1 /2-inch In the presence of matter such as the brain, diameter have been the sizes preferred. Of the positron interacts almost instantly with course, high quantum efficiency matching an electron resulting in the simultaneous the spectral emission of the scintillators emission of two gamma rays each having an (NaI:Tl, CsF, Bi4Ge3O12) is also important. energy of 511 keV, but moving in nearly op- Photometric and Spectrometric Applications posite directions. When a pair of detectors, The side-on photomultiplier has in gener- one on either side of the patient, observes al, been the most widely used tube in photo- coincident events, the point of radionuclide metric and spectrometric applications. The disintegration lies on a line joining the two side-on tube is relatively small and has a detectors. See Fig. 116. rectangularly shaped photocathode that A number of positron cameras have been matches the shape of the light beam from an designed and built, some commercially. exit slit. Spectrometric applications require Various geometries and reconstruction tech- tubes with good stability, high anode sen- niques have been utilized. Complications in sitivity, low dark current, and broad spectral design arise when each detector can be in sensitivity. A high signal-to-noise ratio is im- coincidence with any of several detectors on portant because of the generally small signal the opposite side of the patient. More than levels.

104 Photomultiplier Applications

Before the development of the new measure of the density. The exposure time negative-electron-affinity type photoemit- using white light and the lens opening used ters, such as , it was to obtain the satisfactory print are noted. necessary to use more than one detector in Next, an area of the production negative the measurement of radiant energy from the similar to that on the master negative is near-ultraviolet to the near-infrared part of chosen and the relative proportions of the the spectrum. A photomultiplier having a red, blue, and green light transmitted are gallium arsenide photoemitter can now be measured again. By use of magenta and used to detect radiant energy from the cutoff yellow correcting filters, the color trans- point of the photomultiplier window in the mitted by the production negative can be near-ultraviolet to 910 nanometers. balanced with that of the master negative. . Spectrophotometers When the production negative is used, the measure the optical density of materials as a lens opening is adjusted so that the exposure function of wavelength and require photo- time with white light is the same as it was for multipliers having a broad spectral range. exposure of the master negative. When the The results of measurements of the absorp- values of color-correcting filters and ex- tion characteristics of substances are fre- posure times thus determined are used, quently expressed in terms of optical density. prints from the production negative can be This logarithmic method correlates with the obtained which are very nearly as good as way the human eye discriminates differences those obtained with the master color in brightness. negative. The transmission density D is defined by Most color-balancing employ steady light sources and handle small- amplitude signals. Consequently, the photo- multiplier used must have low values of dark current and good stability. The life expectan- cy of tubes used in this application is long because the small signal levels reduce the ef- where PO is the radiant flux incident upon a fects of fatigue which might otherwise sample, Pt is the radiant flux transmitted by adversely affect measurement accuracy and a sample, and T is the transmission figure repeatability. Low-current operation also equal to Pt/Po. Density measurements are provides for linear operation where the useful in various applications to films and anode current is proportional to the input other transparencies in addition to chemical flux over the range of transmission values analysis where concentration of a solution is measured. The intensity range of a color- studied as a function of wavelength. balancing photometer, given in terms of the ratio of the radiant flux incident upon a Color-Balancing Photometry. A color- sample to the radiant flux transmitted by a balancing photometer is used to determine sample, is usually of the order of 1000 to 1 or color balance and exposure times necessary more. It must be remembered that as in most to produce photographic color prints from photomultiplier applications, the power sup- color negatives. Such a device is shown in plies used must be capable of providing block diagram form in Fig. 117. It allows the voltages sufficiently regulated and free from matching of the relative proportions of red, ripple to assure minimum variation in sen- blue, and green light transmitted by a pro- sitivity with possible line-voltage variation. duction negative to those of a master color negative. As the first step in the matching Densitometry. Although techniques such process, the master negative is used to make as those employed in the color-balancing an acceptable print. This first print is pro- photometer may be used successfully to duced through trial and error by measure- measure density over an intensity range of 10 ment of the relative proportions of red, blue, or 100 to 1 (density 1 to density 2), their use and green light transmitted through a key becomes increasingly difficult as the range is area of the master negative; the key area increased to 1000 to 10,000 to 1 (density 3 to usually consists of a flesh tone or a gray density 4) or more. The large dynamic ranges area. The amount of light transmitted is a encountered in color-film processing place

105 Photomultiplier Handbook

92CS-32454

Fig. 117 - Block diagram of a color-balancing photometer. severe requirements upon the photomulti- tometer to be equipped with a meter or read- plier because it must be operated in a out scale that is linear and provides precise constant-voltage mode. Problems develop in readings even at high optical densities. this mode at high-density values at which the A simplified circuit of a logarithmic dark current may become a significant pro- photometer capable of measuring film portion of the signal current. The random density with high sensitivity and stability and nature of this dark current, which precludes of providing an appropriate logarithmic its being “zeroed out”, may lead to output- electrical response and linear meter indica- signal instability. As a result of the type of tion of density over three or four density operation needed to produce the dynamic ranges is shown in Fig. 118. range required in density measurements, the The circuit of Fig. 118, in which the photomultiplier anode current is high at low photomultiplier operates at a constant cur- density values. These high currents may rent, minimizes photomultiplier fatigue and result in excessive fatigue, and, depending eliminates the need for a regulated high- upon the operating point, perhaps non- voltage supply. The feedback circuit il- linear operation. As with the color-balancing lustrated develops a signal across Rl propor- photometer,a well regulated low-ripple tional to the anode current. This signal con- power supply is needed to assure accurate trols the bias applied to the control device measurements. and automatically adjusts the current in the voltage-divider network. By this means, the Logarithmic Photometry. In the measure- dynode voltage is maintained at a level such ment of absorption characteristics, the that the anode current is held constant at a changes in brightness levels vary over such a value selected to minimize the effects of large range that it is advantageous to use a fatigue. At optical densities of 3, the dynode photometer whose response is approximately supply voltage may be 1000 volts; at optical logarithmic. This response enables the pho- densities of less than 1, the dynode voltage

106 Photomultiplier Applications

COLOR FILTERS

92CM-32455 Fig. 118 - Block diagram of a logarithmic photometer. may be as low as 300 volts. Dynode voltage Spectrometric applications include absorp- is translated into density by means of the tion, emission, Raman, solar, and vacuum scale on voltmeter V1 . spectrometry, and fluorometry. Because there is an approximately ex- Absorption Spectrometry. Absorption ponential variation of sensitivity of a spectrometry, used to detect radiant energy photomultiplier with applied voltage (see in the visible, ultraviolet, and infrared Fig, 48), in a constant-current mode, the ranges, is one of the most important of the voltage varies almost linearly with the instrumental methods of chemical analysis. logarithm of the input light flux. The voltage It has gained this importance largely as a is thus close to a linear measure of the result of the development of equipment density. A compensating circuit for the lack employing photomultipliers as detectors. of linearity is indicated in Fig. 118 in the The principle underlying absorption spec- form of a variable and automatic shunt trometry is the spectrally selective absorp- across the voltmeter V1 . As the density tion of radiant energy by a substance. The values increase, the effective value of the measurement of the amount of absorption shunt resistance is reduced. This circuit not aids the scientist in determining the amount only compensates for photomultiplier non- of various substances contained in a sample. linearity, but also for the non-linearity of the The essential components of an absorp- optical system employed. tion spectrometer are a source of radiant energy, a monochromator for isolating the Spectrometry desired spectral band, a sample chamber, a Spectrometry, the science of spectrum detector for converting the radiant energy to analysis, applies the methods of physics and electrical energy, and a meter to measure the physical chemistry to chemical analysis. electrical energy. The spectral ranges of the

107 Photomultiplier Handbook source and detector must be appropriate to teraction are so few in number that only the the range in which measurements are to be highest-quality photomultipliers can be used made. In some cases this range may include as detectors. The tube should have high col- one wavelength. In others, it may scan all lection efficiency, high gain, good multipli- wavelengths between the near-ultraviolet cation statistics, low noise, and high quan- and the near-infrared. tum efficiency over the spectral range of in- terest. Fig. 119 shows a Raman spectrum. To Raman Spectroscopy. There are two types reduce the effects of Rayleigh scattering, a of molecular scattering of light, Rayleigh source of noise in Raman spectroscopy, the and Raman. Rayleigh scattering is the elastic photomultiplier is placed at right angles to collision of photons with the molecules of a the light beam, whose wavelength has been homogeneous medium. Because the scat- chosen as long as possible within the range tered photons do not gain energy from or of interest. lose energy to the molecule, they have the classic example of Rayleigh scattering of light from gas molecules is the scattering of the sun’s light rays as they pass through the earth’s atmosphere. This scattering accounts for the brightness and blueness of the sky. Raman scattering is the inelastic collision of photons with molecules that produces scattered photons of higher or lower energy than the initial photons. During the collision there is a quantized exchange of energy that, depending on the state of the molecules, determines whether the initial photon gains 92CS-32456 or loses energy. The differences in energy Fig. 119 - Typical Raman spectrum. levels are characteristic of the molecule. If Fluorometry. The fluorometer is another emitted after the interaction has a frequency instrument which utilizes the photomulti- plier’s capability for low-light-level detec- tion, its high gain, and its good signal-to- noise ratio. There are numerous applications of this instrument in the fields of from the molecule that was in the excited biochemistry,medical research, and in- state. In the reverse case, the initial photon dustrial toxicology. Typical applications in- has given up energy to the molecule in the clude the detection and measurement of unexcited state. minute quantities of air pollutants and of Early Raman instruments had a number components of the blood or urine. Some of disadvantages and were difficult to use. It materials are detected by their own was difficult to find a stable high-intensity fluorescence; others by their quenching ef- light source and to discriminate against fects on the fluorescence of other materials. Rayleigh scattering of the exciting line. With Illustrative of fluorometer operation is the the recent development of high-quality optical design of a model designed by G. K. monochromators and the advent of the laser Turner associates103 shown in Fig. 120. The light source, a renewed interest in the Raman sample is irradiated by an ultraviolet source effect as an analytical method of chemical filtered to eliminate longer wavelength com- analysis has taken place. Raman spec- ponents of the lamp irradiation. Fluorescent trophotometers are generally used to in- spectra at a longer wavelength pass through vestigate the structure of molecules and to a second filter which eliminates scattered supplement other methods of chemical ultraviolet radiation. analysis, particularly infrared-absorption A secondbeam fromthe ultraviolet source spectrometry. provides acalibratedreference flux. Both The scattered photons from a Raman in- the fluorescent beam and the reference beam

108 Photomultiplier Applications

are directed to the photomultiplier and alter- Detection is limited by noise in the anode nately sampled by means of the light inter- current. At low light levels the noise is rupter. The output ac signal from the caused primarily by the fluctuations in dark photomultiplier is processed to provide a current of the photomultiplier (as discussed null signal by means of the light cam and the in the section, “Dark Current and Noise” in attached dial records the fluorescence level. Chapter 4 Photomultiplier Characteristics The null feature of the device eliminates er- and in Appendix G, “Statistical Theory of rors from changes in the photomultiplier or Noise in Photomultiplier Tubes. ") The noise from the ultraviolet source. A third flux in- may be minimized by reducing the band- dicated by the FORWARD LIGHT PATH is width of the measuring system. For example, provided so that even for an absence of a dc system may be used with a bandwidth of fluorescence a balance is always possible. only a few hertz if an appropriate low-level The BLANK KNOB provides an adjustment current meter is selected. Bandwidth can also so that the calibration can be adjusted to the be reduced by some technique of averaging zero fluorescence point. the current fluctuations over a period of Low-Light-Level Detection time. Systems for the detection of low light Another technique of charge integration is levels make use of two basic techniques: to chop the light signal with a motor-driven charge integration, in which the output chopper disk having uniformly spaced holes photocurrent is considered as an integration or slots. The output current is then fed of the anode pulses which originate from the through an amplifier having a narrow band- individual photoelectrons, and the digital width tuned to the frequency of chop. Band- technique in which individual pulses are widths of the order of 1 Hz are typical. counted. Digital Method. In the digital method for Charge-Integration Method. In the the detection of low light levels a series of charge-integration method, either the output pulses, each corresponding to a pho- transit-time spread of the photomultiplier or toelectron leaving the photocathode of a the time characteristics of the anode circuit photomultiplier, appears at the anode. All of cause the anode pulses to overlap and pro- the output pulses from the tube are shaped duce a continuous, though perhaps noisy, by a preamplifier before they enter a pulse- anode current. The current is modulated by amplitude discrimination circuit. Only those turning the light off and on by means of pulses having amplitudes greater than some some mechanical device such as a shutter or predetermined value and having the proper light “chopper”. The signal becomes the dif- rise-time characteristics pass through to the ference between the current in the light-on signal-processing circuits. The digital tech- and light-off conditions. nique is superior to charge integration at

109 Photomultlpller Handbook very low light levels because it eliminates the ratio can also be improved by decreasing the dc leakage component of the dark current as number of dark-noise pulses. Because these well as dark-current components originating pulses originate at the photocathode surface, at places other than the photocathode. Fig. the number can be reduced by reducing the 121 shows a digital system in block form. area of the photocathode or by reducing the In the special case in which the digital effective photocathode area by using elec- technique is used to count single photons in- tron optics to image only a small part of the cident upon the photocathode of a photo- photocathode on the first dynode. It may multiplier, signal pulses appear at the anode also be desirable to cool the photomultiplier with an average pulse amplitude PH equal to and thereby reduce the thermionic emission em, where e is the electron charge and m is from the photocathode. the photomultiplier gain. The number of In some applications not requiring the use signal pulses Na arriving at the anode is of the full area of the semitransparent given by photocathode, it may be useful to restrict the active area of the photocathode by specially (36) arranged magnetic fields104 and thus reduce the collection of thermally emitted electrons where Nis the number of photons incident from non-utilized areas of the photocath- on the ph ode. Fig. 123 illustrates the design of such quantum efficiency of the photocathode at a magnetic defocusing system type PF- the photon wavelength including a factor for 1011 which includes an 8852 photomulti- the loss of light by reflection and absorption, plier (2-inch diameter, ERMA photocath- and a factor for the loss resulting from im- ode, and GaP first dynode). The basic ele- perfect electron-collection efficiency of the ment is a permanent magnet in the form of front end of the photomultiplier. an annular ring (1) polarized so that the The dark-noise pulses present in addition S-pole or N-pole is toward the photocath- to the signal pulses originate mainly from ode. Pole pieces, spacers, and a magnetic- single electrons and have a pulse-height shield cylinder complete the arrangement. distribution as shown in the simplified dark- The resulting magnetic field is such that only noise pulse-height-distribution spectrum of electrons emitted from the central portion of Fig. 122. Region A of Fig. 122 includes the photocathode arrive at the first dynode. circuit-originated noise, some single-electron Typical variation of photocathode response pulses, and pulses caused by electrons as a function of the position of the incident originating at the dynodes in the multiplier light is shown in Fig. 124. Note that the ef- section. Pulses originating at the dynodes ex- fective diameter of the photocathode has hibit less gain than the single-electron pulses been reduced to about one eighth of an inch. from the cathode. Region B represents the Total dark emission at room temperature single-electron pulse-height distribution and from the magnetically controlled photo- is the region in which the single-photon cathode is reduced by approximately 80:1. If signal pulses appear. Region C is caused by the tube is then cooled to - 20°C, the dark cosmic-ray muons, after-pulsing, and radio- count is still further reduced to about two active contaminants in the tube materials electrons per second. and in the vicinity of the tube. To maximize the ratio of signal pulses to noise pulses in Very-Low-Light-Level Photon-Counting single-photon counting, lower- and upper- Technique. Before beginning very-low-light- level discriminators should be located as level photon counting, the following special shown in the figure. For a discussion of precautions must be taken: signal-to-noise ratio and the statistics related 1. The power supply and interconnecting to pulse counting, see “Pulse Counting Sta- circuits must have low-noise characteristics. tistics” in Appendix G. 2. The optical system must be carefully The signal-to-noise ratio may be improved designed to minimize photon loss and to pre- by the square root of the total count time. vent movement of the image of the object on Increasing the time of count is analogous to the photocathode, a possible cause of error decreasing the bandwidth in the charge- if the cathode is non-uniform. It is generally integration method. The signal-to-noise good practice to defocus the image on the

110 Photomultiplier Applications

ADJUSTABLE PHOTOMULTIPLIER - - PULSE AMPLITUDE DISCRIMINATOR

92cs-32458 Fig. 121- Block diagram illustrating a digital system for detection of low light levels. photocathode, especially in the case of a point source, to minimize problems which may result from a non-uniform photocath- ode. 3. The photomultiplier must be allowed to stabilize before photon counting is begun. The tube should not be exposed to ultraviolet radiation before use and should, if possible, be operated for 24 hours at the desired voltage before the data are taken. 4. The photomultiplier should be operated with the cathode at ground poten- tial if possible. If the tube is operated with the photocathode at negative high voltage, care must be taken to prevent the glass envelope of the tube from coming into con- tact with conductors at ground potential or noisy insulators such as bakelite or felt. 92CS-32460 Without this precaution, a very high dark Fig. 123 - Magnetic defocusing system noise may result as well as permanent (PF1011) which includes type 8852 photomul- damage to the photocathode. tiplier. This system permits collection of elec- 5. Large thermal gradients must not be trons only from the central part of the pho- permitted across the tube as it is cooling. In tocathode and thus substantially reduces dark addition, care must be taken to avoid ex- emission. Shown are (7) ceramic magnet, (2) cessive condensation across the leads of the spacer-insulator, (3) magnet pole piece of tube or on the faceplate. black iron, and (4) magnetic shield cylinder. The assembly is designed to fit over the 8852 envelope.

Photomultiplier Selection for Photon Counting. In the selection of a photomulti- plier for use in photon counting, several im- portant parameters must be considered. First, and most important, the quantum effi- ciency of the photocathode should be as high as possible at the desired wavelength. To minimize the thermionic dark-noise emis- sion, the photocathode area should be no larger than necessary for signal collection; the multiplier structure should utilize as large a fraction as possible of the electrons from the photocathode. The over-all tube PULSE HEIGHT 92cs-32459 should have as low a dark noise as possible. In some applications, the rise time and time- Fig. 122 - Dark-noise pulse-height-distribu- resolution capabilities of the tube may also tion spectrum. be important.

111 PhotomultIplier Handbook

DISTANCE FROM CENTER OF PHOTOCATHODE - INCHES 92CS-32461 Fig. 124 - Typical variation of photocathode sensitivity as a function of incident-light spot position for an 8852 photomultiplier with the magnetic defocusing system of Fig. 123 affixed to the tube.

A number of more recent developments in wavelength increases with increasing photomultiplier design are of considerable content. Spectral-response curves for some significance in photon counting. These of the more recent photocathodes are shown developments fall into two major categories, in the section “Photoemission” in Chapter 2 secondary-emission materials and photo- Photomultiplier Design. cathodes. Because of the superior statistics of gallium phosphide (one of the more A recently developed secondary-emission materials discussed in the section “Secon- dary Emission”in Chapter 2 Photomulti- plier Design), a tight single-electron distribu- tion curve can be obtained, as shown in Fig. 79. The tight distribution permits easy loca- tion of the pulse-height discriminator, as is particularly evident when some of the signal pulses are originated by two or more photo- electrons leaving the cathode simultane- ously. If the average number of photoelec- trons leaving the photocathode per pulse were three,a pulse-height distribution 92CS-32462 similar to that shown in Fig. 125 would be Fig. 125 - Pulse-height distribution obtained obtained. Gallium phosphide provides a with a photomultiplier having a gallium higher signal-to-noise ratio than would con- phosphide first dynode. Light level is such ventional secondary-emitting materials such that three photoelectrons per pulse time is as Be0 or Cs3Sb when used in the same tube. the most probable number. Photocathode developments include ERMA (Extended Red Multi-Alkali), a Astronomy105. Astronomers were among semitransparent photocathode having a the first users of photomultiplier tubes. The response to 940 nanometers, and several high quantum efficiency, low background negative-electron-affinity materials. Perhaps noise, and high gain recommend the use of the most significant of these materials is photomultipliers in various astronomical ap- , whose threshold plications. Requirements are often similar to

112 Photomultiplier Applications those for low-level photometry or spec- tive run-away dark currents. But, by pulsing trometry as discussed above. Applications the applied voltage, regenerative effects that include the guidance of telescopes, intensity depend upon transit time of ions are measurements, stellar spectrophotometry, eliminated. Post’s circuit is shown in Fig. Doppler measurement of radial velocities, 126. Not all tubes he tried could tolerate this and the measurement of stellar magnetic high voltage without field emission and fields by measuring Zeeman displacements. break-down effects, but on those tubes Side-on photomultipliers such as the 1P21 which were satisfactory initially or which he are frequently used in these applications. was able to stabilize by application of Refrigeration may be utilized to reduce repeated high-voltage pulses, he found some background. Photon counting may be very interesting advantages. Peak pulse adopted for very low signal measurement. amplitudes of 0.7 ampere were possible and For applications requiring a wide spectral provided a 70-volt pulse to drive an range, tubes with the GaAs:Cs photocathode oscilloscope directly. Thus, a single are very useful. photoelectron resulted in a 10-to-15-volt out- Pulsed Photomultipliers put pulse; the tubes were operated at gains in the neighborhood of 109 The measured rise Some years ago, Post106 demonstrated time he found to be slightly more than one improved characteristics of photomultiplier nanosecond. Refer to Fig. 71. Note in Fig. tubes by operating them with a pulsed 126 the termination of the coaxial output voltage supply. He used side-on photomulti- line at the photomultiplier end which re- pliers types 931A and 1P21 with an applied duced the output RC time constant. It is voltage of between 4 and 5 kilovolts pulsed pointed out by Post “the loss of a factor of for 2.5 microseconds or less. At these two in voltage by terminating at the multi- voltages, 400 to 500 volts per stage, the plier is compensated by voltage doubling at secondary emission for Cs-Sb dynodes is at a the unterminated end.” maximum (see Fig. 19) so that the gain becomes relatively insensitive to the varia- Laser Range Finding tion of the maximum voltage applied. Nor- A simplified block diagram of a laser mally, these tubes could not be operated at range finder is shown in Fig. 127. Such a such high voltages without incurring destruc- device may be utilized by the military for

-4 kV PULSE

Fig. 126 - Post’s voltage divider and bypass capacitor connections of photomulti- plier to output coaxial line for pulsed high-voltage operation.105

113 Photomultiplier Handbook

92CS - 32464 Fig. 127 - Simplified block diagram of a laser range finder. tank fire control or in various industrial scattered laser return beam. The signal-to- surveying type applications. Typically, a noise ratio of such a pulse is proportional to laser is pulsed in a time range of 25 the square root of the product of the number nanoseconds. Measurements are made of of incident photons on the photomultiplier beam travel time with a photomultiplier pick and the quantum efficiency of the photoelec- tric conversion. If the atmospheric attenua- may be used with a 3/4-inch diameter photo- tion is neglected and it is assumed that the multiplier having an S-20 spectral response. laser spot falls entirely within the target, the maximum range in this case would be in- and solid-state junction lasers such as creased as the square root of the quantum ef- ficiency of the photocathode. This increase of the longer wavelengths, the detectors are follows because the number of photons col- more often silicon avalanche diodes. An in- lected by the aperture of the receiving system terference filter is used to pass the varies inversely as the square of the distance wavelength of radiant energy of the laser from the target. with a minimum of background radiation. In cases where the laser pulse must be Photomultipliers used in laser range finding detected against the background of a may have a relatively small photocathode, daylight scene, it is said to be background but they must exhibit high quantum efficien- limited. If atmospheric attenuation again is cy, low dark noise, and fast rise time or an neglected, the signal is proportional to the equivalent large bandwidth. Most photomul- number of incident photons times the quan- tipliers can provide bandwidths exceeding tum efficiency of the photocathode. The 100 MHz and at the same time maintain number of incident photons from the return relatively large output signals. The band- beam is inversely proportional to the square width of a photomultiplier can be limited by of the range, again assuming that the target the RC time constant of the anode circuit. is larger than the laser spot. The noise, The range of a laser-range-finding system however, is independent of the range and is depends on system parameters and opera- determined by the square root of the product tional environment. The maximum range of of the incident background radiation and the a given system may be signal-photon limited quantum efficiency. Thus, to a first approx- or background limited. The photon-limited imation the range is increased in the case exists when the background and de- background-limited case only by the fourth tector noise can be considered negligible. root of the quantum efficiency. Because the The maximum range in this case is deter- photomultiplier current caused by the mined by the signal-to-noise ratio in the background radiation is proportional to the photoelectron pulse corresponding to the solid angle of the scene from which the

114 Photomultiplier Applications photomultiplier collects radiation, the pliers with video amplifiers, and (c) ap- background current in the photomultiplier propriate filters, color operation is possible. may be minimized by the use of a photocath- In color operation, the primary wavelengths ode having a small area or by the use of a are filtered after separation by the light- limiting aperture on the faceplate of the absorbing filters before being focused upon photomultiplier. No loss in collected laser each of three photomultipliers, one for each light need result because the return beam color channel. The output of each photo- may generally be considered as originating multiplier is then fed to a separate video from a point source. The system aperture, of amplifier. course, must be large enough to avoid op- Flying-spot video-signal generators are tical alignment problems. used in the industry primarily for Scanning Applications viewing slides, test patterns, motion-picture A number of photoelectrically sensed film, and other fixed images. Systems have scanning systems have been devised such as been developed for the home-entertainment facsimile scanners in which the material to industry that allow slides and motion-picture be transmitted-a photograph or printed film to be shown on the picture tube of any message-is mounted on a drum that is type of commercial television receiver. rotated to provide scan in one dimension. A A similar system is utilized by the scanning head moves parallel to the drum photographic industry to provide accurate axis to provide the other dimension. The pic- exposure control for film copying. By means ture element is illuminated with a focused of the three color controls, the operator may light spot and the scattered light is picked up produce a color television display represent- by a photomultiplier, providing high-speed ing the film and can adjust and measure each transmission utilizing the modulated output color component to produce a visually satis- electrical signal. fying balance. Two more recently developed scanning Several important considerations must be systems are the flying-spot scanner for the taken into account if the cathode-ray tube in development of television signals and the the system is to produce a light spot capable supermarket checkout system which recog- of providing good resolution. The cathode- nizes a coded symbol on each product. ray tube should be operated with as small a Flying-Spot Scanning. The elements of a light spot as possible. The cathode-ray-tube flying-spot scanning system are shown in faceplate should be as blemish-free as Fig. 128. A cathode-ray tube, in conjunction possible and the tube should employ a fine- with its power supplies and deflection cir- grain phosphor. Blemishes adversely affect cuits, provides a small rapidly moving light signal-to-noise performance and contribute source which forms a raster on its face. This to a loss of resolution. raster is focused by the objective lens in the The spectral output of the cathode-ray- optical system onto the object being tube phosphor should match the spectral scanned, a slide transparency or a motion- characteristic of the photomultiplier. This picture film. The amount of light passing match can be rather loose in a monochro- through the film varies with the film density. matic flying-spot generator. The spectral This modulated light signal is focused upon output of the phosphor of the cathode-ray the photomultiplier by means of the tube used in a three-color version, however, condensing-lens system. The photomultiplier must include most of the visible spectrum. converts the radiant-energy signal into an Phosphors used in monochromatic systems electrical video signal. The amplifier and its may provide outputs in the ultraviolet region associated equalization circuits increase the of the spectrum and still perform satisfac- amplitude of the video signal as required. torily. Phosphors such as P16 and P15, The flying-spot scanning system is capable when used with an appropriate ultraviolet of providing high-resolution monochromatic filter, display the necessary short persistence performance. With the addition of (a) ap- required in a monochromatic system. propriate dichroic mirrors which selectively The visible portion of the P15 or P24 reflect and transmit the red, blue, and green phosphor is used in color systems. The P15 wavelengths, (b) two additional photomulti- and P24 phosphors are, however, much

115 VERTICAL HORIZONTAL SCANNING SCANNING

92CM -32465 Fig. 128 - Elements of a flying-spot scanning system. slower than the P16 phosphor and cause a between the cathode-ray tube and the lens lag in buildup and decay of output from the and to use a second photomultiplier to sense screen. any non-uniformities in the cathode-ray tube The phosphor lag results in trailing, a con- raster display. This sensed signal may then dition in which the persistence of energy be used to eliminate this source of signal output from the cathode-ray tube causes a distortion in the primary photomultiplier continued and spurious input to the photo- pickup. multiplier as the flying spot moves across the The spectral characteristics of the picture being scanned. The result is that a photomultiplier (or photomultipliers in the light area may trail into the dark area in the case of the three-color system) and the reproduced picture. cathode-ray-tube phosphor should match. Similarly, the lag in buildup of screen out- Usually, a photomultiplier having an S-4 or put causes a dark area to trail over into the S-l1 (Cs3Sb) spectral response is suitable for light area. The result of these effects on the use in a monochromatic system or as the reproduced picture is an appearance similar detector for the blue and green channels. An to that produced by a video signal deficient S-20 (Na2KSb:Cs) response is very often in high frequencies. Consequently, high- utilized for the red channel. The bialkali frequency equalization is necessary in the cathode (K-CS-Sb) is also well suited for the video amplifier. blue channel. The speed of the detector must The objective lens used in a flying-spot be sufficient to provide the desired video generator should be of a high-quality bandwidth. Most requirements do not ex- enlarger type designed for low magnification ceed 6 to 8 MHz, a figure well within photo- and, depending upon the cathode-ray-tube multiplier capabilities. light output,should be corrected for The anode dark current of the photomulti- ultraviolet radiation. The diameter of the plier should be small compared to the useful objective lens should be adequate to cover signal current. The signal-to-noise ratio will the slide to be scanned. An enlarging f/4.5 be maximized by operation of the photomul- lens with a focal length of 100 millimeters is tiplier at the highest light levels possible. If suitable for use with 35-millimeter slides. necessary, the over-all photomultiplier gain The optics should not image the film on the should be reduced to prevent excessive anode photomultiplier because shading effects current and fatigue. would result from non-uniformities of the The amplitude of the light input is usually photomultiplier response. In some cases it a compromise between an optimum signal- may be useful to employ a beam splitter to-noise ratio and maximum cathode-ray-

116 Photomultiplier Applications tube life. Tube life may be reduced because gamma-correction amplifier is required in of loss of phosphor efficiency at high beam- each channel. The gammacorrection ampli- current levels. The signal-to-noise ratio can fier assures maximum color fidelity by mak- also be improved by the selection of photo- ing the linearity or gamma of the system uni- multipliers having the highest photocathode ty. sensitivities possible. However, because the spread of photocathode sensitivities is Supermarket Checkout Systems. Fig. 129 seldom greater than two or three to one, the shows the elements of a supermarket check- improvement afforded by such selection is out system. The Universal Product Code limited to two or three dB, an improvement (UPC), which is to be marked on all prod- difficult to detect during observations of a ucts, is scanned by laser beam by means of television display but desirable and necessary an oscillating mirror and a rotating mirror in some critical applications. system. The pattern of the scan is a network Photomultiplier gain need only be suffi- of overlapping sine waves. The reflected pat- cient to provide a signal of the required level tern contains the modulation of the UPC to the succeeding video-amplifier stages. symbol and is converted by the photomulti- These stages, in addition to providing the plier to an electronic signal that is then necessary amplification and bandwidth to analyzed for the content of the code. assure good picture quality, incorporate The laser used in the supermarket equalization circuits composed of networks checkout system is usually a low-power He- having different time constants. The Ne type with the principal emission line at relatively long decay time of these circuits 633 nanometers. A suitable photomultiplier generally results in appreciable reduction of for this application is a two-inch end-on type the useful signal-to-noise ratio. Therefore, having an S-20 (Na2KSb:Cs) spectral the use of short-persistence phosphors is response or an extended-red multialkali recommended to reduce the required amount type. Stability and good signal-to-noise ratio of equalization. are important characteristics of the photo- In addition to the video amplifier, a multiplier tube.

ROTATING MIRROR POLYGON

92CM - 32466 Fig. 129 - Point-of-sale supermarket checkout system. At left is Universal Product Code (UPC) symbol that provides 10-digit product identification. At right is sche matic of optical system for scanning the product symbol and detecting the modula- tion with a photomultiplier.

117 Photomultiplier Handbook

REFERENCES plication of annihilation coincidence detec- 91. R.W. Engstrom and E. Fischer, “Ef- tion to transaxial reconstruction tomogra- fects of voltage-divider characteristics on phy,” J. Nucl. Medicine, Vol. 16, No. 3, pp multiplier phototube response,” Rev. Sci. 210-224, 1975. Instr., Vol. 28, pp 525-527, July, 1957. 100. S.E. Derenzo, H. Zaklad, and T.F. 92. C.R. Kerns, “A high-rate phototube Budinger , “Analytical study of a high- base,” IEEE Trans. Nucl. Sci., Vol. NS-24, resolution positron ring detector system for No. 1, pp 353-355, Feb. 1977. transaxial reconstruction tomography,” J. 92a. V.O. Altemose, “Helium diffusion Nucl. Med., Vol. 16, No. 12, pp 1166-73, through glass,” J. Appl. Phys., Vol. 32, No. Dec. 1975. 7, pp 1309-1316, 1961. 101. M.E. Phelps, E.J. Hoffman, Sung- 92b. F.J. Norton, “Permeation of gases Cheng Huang, and D.E. Kuhl, “ECAT: a through solids,” J. Appl. Phys., Vol. 28, computerized tomographic imaging system No. 1, pp 34-39, 1957. for positron-emitting radiopharmaceuti- 93. American Institute of Physics Hand- cals,” J. Nucl. Med., Vol. 19, No. 6, pp book, Third Edition, McGraw-Hill, 1972. 635-647, June 1978. 94. R.W. Engstrom, “Luminous Micro- 102. G.L. Brownell, C. Burnham, B. flux standard,” Rev. Sci. Instrum. Vol. 26, Ahluwalia, N. Alpert, D. Chester, S. No. 6, pp 622-623, June, 1955. Cahavi, J. Correia, L. Deveau, “Positron 95. D.G. Fisher, A.F. McDonie, and A.H. imaging instrumentation,” IEEE Trans. Sommer, “Bandbending effects in Na2K Sb Nucl. Sci., Vol. 24, No. 2, pp 914-916, Apr. and K2Cs Sb photocathodes,” J. Appl. 1977. Phys., Vol. 45, No. 1, pp 487-8, Jan. 1974. 103. Operating and Service Manual, Model 96. H.O. Anger, “Scintillation camera,” 111 Fluorometer,” G.K. Turner Associates, Rev. Sci. Instrum., Vol. 29, pp 27-33, 1958. 2524 Pulgas Ave., Palo Alto, Calif. 94303. 97. M.J. Weber and R.R. Monchamp, 104. G.Y. Farkas and P. Varga, “Reduc- “Luminescence of Bi4Ge3O12: spectral and decay properties,” J. Appl. Phys., Vol. 44, tion of dark current in transparent cathode No. 12, pp 5495-99, 1973. photomultipliers for use in optical measure- 98. G.L. Brownell and C.A. Burnham, ments,” J. Sci. Instrum., Vol. 41, pp. “Recent developments in positron scin- 704-705, 1964. tigraphy,” Instrumentation in Nuclear 105. Astronomical Techniques, Edited by Medicine, Vol. 2, Ed: G.J. Hine and J.A. W.A. Hiltner, The University of Chicago Sorenson, Chapter 4, pp 135-159, Academic Press, 1962. Press, 1974. 106. R.F. Post, “Performance of Pulsed 99. M.E. Phelps, E. J. Hoffman, N.A. Photomultipliers,” Nucleonics, Vol. 10, No. Mullani and M.M. Ter-Pogossian, “Ap- 5, pp 46-50, May 1952.

118 Typical Photomultiplier Applications and Selection Guide

Appendix A- Typical Photomultiplier Applications and Selection Guide

The many and varied requirements of medical applications to map the location of equipment designers and experimenters isotope disintegrations. preclude the recommendation of a single High-Temperature Environments: Applica- photomultiplier as the optimum device for tions such as the logging of deep oil wells, or any given application category. In most ap- geological exploration, and steel-mill process plications, some trade-offs must be made in controls. electrical characteristics; tube size must be Imaging Devices: A cathode-ray tube or considered; the environment in which the moving mirrors can be used as a light source device is to be operated can be an influential to sequentially illuminate a film positive or factor; and of course, over-all cost is impor- negative or a printed page. This system is tant. Each of these constraints can be best used in (1) optical character recognition, (2) evaluated by the individual designer for the scanning or printed or written material for specific application. transmission by telephone, (3) parts inspec- This Appendix defines a number of the tion, and (4) reproduction of motion pic- more common applications of photomulti- tures, slides, and educational material on a pliers and lists tube types which are suitable television receiver (color or black and white). or are frequently used for the particular ap- Inspection, High-Speed: Small objects such plication. The listing is not all-inclusive and as fruits, vegetables, seeds, candy, toys, is intended to serve only as a general guide paper products and even glass, metal, and for initial type selection. Other photomulti- other industrial parts can be examined for pliers may be satisfactory for the specified color and defects as they move at high speed applications when all system requirements past one or more photomultiplier tubes. are considered. Laser Detection: Lasers provide unique light sources; they are spectrally pure and CATALOGUE OF PHOTOMULTIPLIER produce very narrow collimated beams. APPLICATION CATEGORIES They can be very intense and can be made to Astronomy: The guidance of telescopes, in- produce light pulses of extremely short dura- tensity measurements, stellar spectropho- tion. The photomultiplier provides time tometry, and the like. resolution in the nanosecond and subnano- Colorimetry: The quantitative color com- second ranges and is capable of detecting parison of surfaces (reflectance) and solu- very low light levels such as those received tions (transmission). from weak reflected laser light pulses. C-T Scanners: A medical X-ray equipment Photometry: The measurement of illumina- that provides a cross section density map tion or luminance. Levels of light flux vary (tomograph) of a patient. A photomultiplier over a wide range in photography, is used to detect and measure the light flux astronomy, television, and other applica- from a scintillating crystal. tions. Densitometry: The measurement of optical Photon Counting: A method of detecting density of photographic negatives, neutral photons by counting single photoelectrons density filters, and similar materials. released from the photocathode. Gamma-Ray Cameras: A scintillation Pollution Monitoring: The analysis of the counter having a single large crystal and a level and the nature of contaminants in solu- number of photomultiplier tubes used in tions, gases, and other waste materials.

119 Typical Photomultiplier Applications and Selection Guide

Positron Camera: A scintillation-counter- Spectrophotometry type device providing tomographic presenta- BURLE Photo- Window No. Dia tions based on coincident gamma-ray emis- Type cathode Material1 of meter2 sion accompanying annihilation of a Material Stages positron and an electron . Medical applica- 1P21 Cs3Sb 0080 9 1-1/8"s tions utilize tracer radio nuclides such as 1P28 Cs3Sb 8337 9 1-1/8"s 11C, 13N, and 15O. 1P28A Cs3Sb 8337 9 1-1/8"s Process Control: The measurement of 1P28B K2CsSb 8337 9 1-1/8"s transmitted or reflected light in continuous 931A GaAs 0080 9 1-1/8"s flow processes using solids, liquids, or gases. 931B GaAs 0080 9 1-1/8"s Detects flaws, improper marking, and 4526A Na2KCsSb 0080 10 1-1/2"d changes in color and optical density. By C31034 GaAs 8337 11 2"e C31034A GaAs 8337 11 2"e using radioactive sources and scintillators, S83063E Na KCsSb 7056 10 1-1/8"e photomultipliers can be used for the control 2 S83068E Rb2CsSb 7056 10 1-1/8"e of the weight and thickness of opaque 83089-600 Na KCsSb 7056 10 1-1/8"e materials. 2 83101-600 Rb2CsSb B270 10 3/4"e Radioimmunoassay (RIA): A technique 1 0080, B270 – lime glass; 8337 - Schott, 7056 – borosilicate that enables the measurement of minute 2 e – end-window; s – side-window; d – dormer window quantities of substances in biological fluids as small as 10-12 gram. Compounds are TLD tagged with radioactive isotopes and are BURLE Photo- Window No. Dia measured with a liquid scintillation counter Type cathode Material1 of meter2 for beta-emitting isotopes or a solid crystal Material Stages scintillation counter for gamma-emitting 8575 K2CsSb 7740 12 2"e isotopes. 8850 K2CsSb 7740 12 2"e S83062E Rb CsSb 7056 10 1"e : The measurement of irra- 2 83087-100 Rb2CsSb B270 10 3/4"e diance or . 83101-600 Rb2CsSb B270 10 3/4"e Raman Spectrometry: Measurement of the 1 B270 – lime glass; 7740 – Pyrex ; 7056 - borosilicate wavelength shift of scattered photons from a 2 e – end-window highly monochromatic source such as a laser provides information on molecular structure Gamma -Ray Cameras and bonding energy. BURLE Photo- Window No. Dia 1 2 Scintillation Counting: The measurement of Type cathode Material of meter nuclear radiation by detecting light or single Material Stages scintillations emitted from a scintillation 4900 K2CsSb B270 10 3" e material receiving nuclear radiation. 6199 Cs3Sb 0080 10 1-1/2" e S83010E Rb2CsSb 0080 10 1-1/2" e Severe Physical Environments: Applica- S83013F K CsSb 0080 10 3-1/2" e tions such as oil-well logging, satellites, and 2 S83019F K2CsSb B270 10 2" e military vehicles subject to severe shock and S83020F K2CsSb B270 10 60mm h,e vibration. S83021E K2CsSb B270 10 3" e Thermoluminescent Dosimetry (TLD): Cer S83022F K2CsSb B270 10 2" h,e tain phosphors emit light when they are S83025F K2CsSb B270 10 3" h,e heated after having been exposed to ionizing S83049F K2CsSb B270 8 3" e radiation. Devices using this principle afford S83053F K2CsSb B270 8 60mm h,e personnel protection by determining dosage S83054F K2CsSb B270 8 2" e levels in medical and biological treatments S83056F K2CsSb B270 8 3" h,e and studies. Energy stored in TLD’s is pro S83069E K2CsSb B270 8 35x46.5mm, mh portional to dosage over a very wide range. S83079E K2CsSb B270 8 3" sq 1 0080, B270 – lime glass Time Measurement: In nuclear experiments 2 the “time of flight” of nuclear particles is e – end-window; h – hexagonal; mh – modified hexagonal; important. Photomultipliers permit time sq - square measurements down to a fraction of a nonosecond.

120 Typical Photomultiplier Applications and Selection Guide

High Temperature Environments Laser Detection BURLE Photo- Window No. Dia BURLE Photo- Window No. Dia Type cathode Material1 of meter2 Type cathode Material1 of meter2 Material Stages Material Stages

C31000AP Na2KSb 7056 12 2"e 4526A Na2KCsSb 0080 10 1-1/2"d C31016G Na2KSb 7056 10 1"e 8575 K2CsSb 7740 12 2"e C83051 Na2KSb sa 10 1"e 8850 K2CsSb 7740 12 2"e C83060 Na2KSb sa 10 1-1/4"e 8852 Na2KCsSb 7740 12 2"e C83065 Na2KSb 7056 10 1"e S83063E Na2KCsSb 7056 10 1-1/8"e 83103 100 Na2KSb 7056 10 3/4"e 83087-100 Rb2CsSb B270 9 3/4"e 83101-600 Rb CsSb B270 10 3/4"e 1 sa – sapphire, 7056 – borosilicate 2 2 e – end-window C31034 GaAs 8337 11 2"e C31034A GaAs 8337 11 2"e Flying-Spot Scanners 1 0080, B270 – lime glass; 8337 - Schott; 7740 – Pyrex BURLE Photo- Window No. Dia 7056 – borosilicate Type cathode Material1 of meter2 2 e – end-window; d – dormer window Material Stages Photometry 4552 K CsSb 0080 9 1-1/8"s 2 BURLE Photo- Window No. Dia 1P21 Cs Sb 0080 9 1-1/8"s 3 Type cathode Material1 of meter2 931A Cs Sb 0080 9 1-1/8"s 3 Material Stages 931B K2CsSb 0080 9 1-1/8"s 6199 Cs3Sb 0080 10 1-1/2"e 1P21 Cs3Sb 0080 9 1-1/8”s 83087-100 Rb2CsSb B270 10 3/4"e 931A Cs3Sb 0080 9 1-1/8”s 83090-600 Rb2CsSb B270 9 3/4"e 931B K2CsSb 0080 9 1-1/8”s 83101-600 Rb2CsSb B270 10 3/4"e 4552 K2CsSb 0080 9 1-1/8”s S83010E Rb2CsSb 0080 10 1-1/2"e 1 0080 – lime glass 2 1 0080 – lime glass; 8337 - Schott s – side-window 2 e – end-window; s – side-window Photon Counting Inspection, High Speed BURLE Photo- Window No. Dia BURLE Photo- Window No. Dia Type cathode Material1 of meter2 Type cathode Material1 of meter2 Material Stages Material Stages 8575 K2CsSb 7740 12 2"e 8850 K2CsSb 7740 12 2"e 1P21 Cs3Sb 0080 9 1-1/8" s 8852 Na2KCsSb 7740 12 2"e 4552 K2CsSb 0080 9 1-1/8" s 8854 K2CsSb 8337 14 5"e 6199 Cs3Sb 0080 10 1-1/2" e 931A Cs Sb 0080 9 1-1/8" s C31034 GaAs 8337 11 2"e 3 C31034A GaAs 8337 11 2"e 931B K2CsSb 0080 9 1-1/8" s S83062E Rb2CsSb 7056 10 1-1/2"e S83062E Rb2CsSb 7056 10 1" e 83089-600 Na2KCsSb 7056 10 1-1/8"e S83010E Rb2CsSb 0080 10 1-1/2" e 83090-600 Rb2CsSb B270 9 3/4"e 83087-100 Rb2CsSb B270 10 3/4"e 83101-600 Rb2CsSb B270 10 3/4"e 83090-600 Rb2CsSb B270 9 3/4" e 1 83101-600 Rb2CsSb B270 10 3/4" e 0080, B270 – lime glass; 8337 - Schott; 7740 – Pyrex 7056 – borosilicate 1 0080, B270 – lime glass, 7056 – borosilicate 2 e – end-window 2 e – end-window; s – side-window Positron Cameras Raman Spectroscopy BURLE Photo- Window No. Dia BURLE Photo- Window No. Dia 1 2 1 2 Type cathode Material of meter Type cathode Material of meter Material Stages Material Stages 83087-100 Rb2CsSb B270 10 3/4"e 8850 K2CsSb 7740 12 2”e 83090-600 Rb2CsSb B270 9 3/4"e 8852 Na2KCsSb 7740 12 2”e 83101-600 Rb CsSb B270 10 3/4"e C31034 GaAs 8337 11 2”e 2 83102 100 Rb2CsSb 0080 9 1"e C31034A GaAs 8337 11 2”e S83062E Rb CsSb 7056 10 1"e 1 8337 - Schott; 7740 – Pyrex 2 2 e – end-window 1 0080, B270 – lime glass, 7056 – borosilicate 2 e – end-window

121 Typical Photomultiplier Applications and Selection Guide

Process Control BURLE Photo- Window No. Dia Time Measurement Type cathode Material1 of meter2 BURLE Photo- Window No. Dia Material Stages Type cathode Material1 of meter2 1P21 Cs3Sb 0080 9 1-1/8"s Material Stages 1P28 Cs3Sb 8337 9 1-1/8"s 8575 K2CsSb 7740 12 2"e 1P28A Cs3Sb 8337 9 1-1/8"s 8850 K2CsSb 7740 12 2"e 1P28B K2CsSb 8337 9 1-1/8"s 8852 Na2KSb 7740 12 2"e 931A Cs3Sb 0080 9 1-1/8"s 8854 K2CsSb 8337 14 5"e 931B K2CsSb 0080 9 1-1/8"s S83062E Rb2CsSb 7056 10 1"e 2060 Cs3Sb 0080 10 1-1/2"e S83068E Rb2CsSb 7056 10 1-1/8"e 4856 K2CsSb 0080 10 2"e 83087-100 Rb2CsSb B270 10 3/4"e 4900 K2CsSb 0080 10 3"e 83101-600 Rb2CsSb B270 10 3/4"e 6199 Cs Sb 0080 10 1-1/2"e 3 1 0080, B270 – lime glass; 8337 - Schott; 7740 – Pyrex, S83010E Rb CsSb 0080 10 1-1/2"e 2 7056 – borosilicate S83019F K2CsSb B270 10 2"e 2 e – end-window S83021E K2CsSb B270 10 3"e S83049F K2CsSb B270 8 3"e Scintillation Counting S83054F K2CsSb B270 8 2"e BURLE Photo- Window No. Dia S83079E K2CsSb B270 8 3"sq,e Type cathode Material1 of meter2 1 0080, B270 – lime glass; 8337 - Schott Material Stages 2 e – end-window; s – side-window 2060 Cs3Sb 0080 10 1-1/2"e 4900 K2CsSb 0080 10 3"e Severe Physical Environments 6199 Cs3Sb 0080 10 1-1/2"e BURLE Photo- Window No. Dia 83103 100 Na2KSb 7056 10 3/4"e 1 2 Type cathode Material of meter 8575 K2CsSb 7740 12 2"e Material Stages 8850 K2CsSb 7740 12 2"e 83101 100 Rb2CsSb B270 10 3/4"e 8852 Na2KCsSb 7740 12 2"e C31016G Na2KSb 0080 10 1"e 8854 K2CsSb 8337 14 5"e C83051 Na2KSb sa 10 1"e C31000AP Na2KSb 7056 12 2"e C83060 Na2KSb sa 10 1-1/4"e C31016G Na2KSb 0080 10 1"e C31016H Na KSb 0080 10 1"e 1 0080, B270 – lime glass; sa - sapphire 2 2 e – end-window C31034 GaAs 8337 11 2"e C31034A GaAs 8337 11 2"e Radiometry S83006F K2CsSb 0080 10 5"e BURLE Photo- Window No. Dia S83062E Rb2CsSb 7056 10 1"e Type cathode Material1 of meter2 S83010E Rb2CsSb 0080 10 1-1/2"e Material Stages 83087-100 Rb2CsSb B270 10 3/4"e 83092-500 Na KSb 0080 10 1"e 1P21 Cs3Sb 0080 9 1-1/8"s 2 83101-600 Rb CsSb B270 10 3/4"e 1P28 Cs3Sb 8337 9 1-1/8"s 2 1P28A Cs3Sb 8337 9 1-1/8"s 1 0080, B270 – lime glass; 8337 - Schott; 7740 – Pyrex 1P28B K2CsSb 8337 9 1-1/8"s 7056 – borosilicate 2 931A Cs3Sb 0080 9 1-1/8"s e – end-window 931B K2CsSb 0080 9 1-1/8"s Radioimmunoassay 4526A Na2KCsSb 0080 9 1-1.2"d 2060 Cs3Sb 0080 10 1-1/2"e BURLE Photo- Window No. Dia 1 2 6199 Cs3Sb 0080 10 1-1/2"e Type cathode Material of meter 8575 K2CsSb 7740 12 2"e Material Stages 8850 K2CsSb 7740 12 2"e 4856 K2CsSb 0080 10 2"e 8852 Na2KCsSb 7740 12 2"e 4900 K2CsSb 0080 10 3"e S83010E Rb2CsSb 0080 10 1-1/2"e 6199 Cs3Sb 0080 10 1-1/2"e S83062E Rb2CsSb 7056 10 1"e S83068E Rb2CsSb 7056 10 1-1/8"e 83087-100 Rb2CsSb B270 10 3/4"e S83010E Rb2CsSb 0080 10 1-1/2"e 83101-600 Rb2CsSb B270 10 3/4"e S83019F K2CsSb B270 10 2"e S83054F K CsSb B270 8 2"e 1 0080, B270 – lime glass; 8337 - Schott; 7740 - Pyrex 2 7056 – borosilicate 1 0080, B270 – lime glass; 7056 – borosilicate 2 e – end-window; s – side-window; d – dormer window 2 e – end-window

122 Typical Photomultiplier Applications and Selection Guide

Densitometry BURLE Photo- Window No. Dia Pollution Monitoring Type cathode Material1 of meter2 BURLE Photo- Window No. Dia Material Stages Type cathode Material1 of meter2 1P21 Cs3Sb 0080 9 1-1/8"s Material Stages 1P28 Cs3Sb 8337 9 1-1/8"s 4526A Na2KCsSb 0080 9 1-1/2"d 931A GaAs 0080 9 1-1/8"s S83063E Rb2CsSb 7056 10 1-1/8"e 931B GaAs 0080 9 1-1/8"s 8850 K2CsSb 7740 12 2"e 4552 K2CsSb 0080 9 1-1/8"s 8852 Na2KCsSb 7740 12 2"e 1 0080 – lime glass; 8337 - Schott 83089-600 Na2KCsSb 7056 10 1-1/8"e 2 s – side-window 1 0080 – lime glass; 7740 – Pyrex, 7056 – borosilicate 2 e – end-window; d – dormer window Colorimetry BURLE Photo- Window No. Dia Cerenkov Radiation Type cathode Material1 of meter2 BURLE Photo- Window No. Dia Material Stages Type cathode Material1 of meter2 1P21 Cs3Sb 0080 9 1-1/8"s Material Stages 1P28 Cs3Sb 8337 9 1-1/8"s 8850 K2CsSb 7740 12 2"e 931A GaAs 0080 9 1-1/8"s 8854 K2CsSb 8337 14 5"e 931B GaAs 0080 9 1-1/8"s S83062E Rb2CsSb 7056 10 1"e 83089-600 Na2KCsSb 7056 10 1-1/8”e 83090-600 Rb2CsSb B270 9 3/4"e 4552 K2CsSb 0080 9 1-1/8"s 83101-600 Rb2CsSb B270 10 3/4"e 1 1 0080 – lime glass; 8337 – Schott; 7056 – borosilicate 0080, B270 – lime glass; 7740 – Pyrex, 8337 - Schott 2 s – side-window; e – end-window 7056 – borosilicate 2 e – end-window Astronomy BURLE Photo- Window No. Dia Type cathode Material1 of meter2 Material Stages 1P21 Cs3Sb 0080 9 1-1/8"s 1P28 Cs3Sb 8337 9 1-1/8"s 1P28B K2CsSb 8337 9 1-1/8"s 8575 K2CsSb 7740 12 2"e 8850 K2CsSb 7740 12 2"e 8852 Na2KCsSb 7740 12 2"e C31034 GaAs 8337 11 2"e C31034A GaAs 8337 11 2"e 1 0080 – lime glass; 8337 - Schott; 7740 - Pyrex 2 e – end-window; s – side-window

123 Typical Photomultiplier Applications and Selection Guide

This page is blank.

124 Glossary of Terms

Appendix B- Glossary of Terms Related to Photomultiplier Tubes and Their Applications

This Appendix contains a glossary of anode is operated at a voltage positive with terms frequently used in connection with respect to that of the last dynode and collects photomultiplier tubes and their applications. the secondary electrons emitted from the last Where suitable definitions were available dynode. (112) from established standards or other reliable Anticoincidence circuit: A circuit that pro- sources, they were used, possibly with minor duces a specified output signal when one modifications. For these cases, the sources (frequently predesignated) of two inputs of the definitions are indicated by the receives a signal and the other receives no reference number at the end of the defini- signal within an assigned time interval. (112) tion. In many cases, however, the definitions Background counts (in radiation counters): were prepared by the writers of this Manual. Counts caused by ionizing radiation coming In these latter cases, no source reference is from sources other than that to be measured. indicated after the definition. (112) Absorptance: The ratio of the radiant flux Bandwidth: In electrical measurements, the absorbed in a body of material to that inci- difference between limiting frequencies in a dent upon it. If the absorptance is a, the frequency band, expressed in hertz (cycles per second). The noise equivalent bandwidth is the bandwidth of a rectangular equivalent spectrum and may be defined as Acceptor: An impurity element in a p-type semiconductor that may become ionized by taking an electron from the valence band and induce conduction by holes. For exam- ple, with a valence of three is a possi- ble acceptor impurity in silicon (valence, circuit, A is the maximum absolute value of four). angular frequency. (110) Afterpulse: A spurious pulse induced in a photomultiplier by a previous pulse. (112) Bleeder: Resistive voltage divider. Angle of incidence: The angle between a ray Cage: Part of a photomultiplier including of light striking a surface and the normal to the dynodes, focusing structure, anode, and that surface. support members. Angstrom unit,Å:10-10 meter, or 0.1 Candela, cd: The SI unit of luminous in- nanometer. In the International System of tensity . Units (SI units), the nanometer or microm- Cathode: See Photocathode. eter is the preferred unit for use in specifica- CAT-Scanner or CT-Scanner: Computer- tion of wavelengths of light. ized Axial Tomography - scanner; a medical Anode: An electrode through which a prin- X-ray equipment which provides a cross- cipal stream of electrons leaves the inter- sectional density map of a patient. In a electrode space. In a photomultiplier, the typical device, an X-ray fan beam incident

125 Photomultiplier Handbook on and rotating around the patient is Crystal: See Scintillator. detected by a large array of scintillation Curie, Ci: A unit of radioactivity defined as crystals and photomultiplier tubes. the quantity of an radioactive nuclide in Cerenkov radiation: Radiation generated by which the number of disintegrations per a high-energy charged particle moving second is 3.7 x 1010. (B4) through a dielectric with a velocity greater than the velocity of light in the dielectric Current amplification (photomultipliers): The ratio of (1) the output signal current to (i.e., greater than c/n where n is the index of refraction of the dielectric). (2) the photoelectric signal current from the photocathode. (112) Channel, channel number: In scintillation counting, a number proportional to pulse Dark current: That current flowing in the height specifying a generally narrow range of cathode circuit (cathode dark current) or in pulse heights. the anode circuit (anode dark current) in the Channel multiplier: A tubular electron- absence of light or radiation in the spectrum multiplier having a continuous interior sur- to which the photomultiplier is sensitive. face of secondary-electron emissive material. Delay line: A transmission line for intro- (107) ducing signal time delay. (112) Coincidence circuit: A circuit that produces Delta function (Dirac Delta Function): A a specified output signal when and only when a specified number (two or more) or a 1, when the integration is specified combination of input terminals carried out over the full range of the receives signals within an assigned time inter- variable. Pulsed light sources are sometimes val. (112)-- -- referred to as delta light sources when the Collection efficiency: The fraction of elec- length of the pulse is short with respect to the trons emitted by the photocathode of a response time of the photomultiplier or photomultiplier that lands on the first detecting instrument. dynode. Or more generally, the fraction of Detectivity, D: Reciprocal noise equivalent electrons emitted by one electrode that lands -1 on the next electrode (dynode or anode). power, NEP; it is expressed in W . Detec- tivity is a figure of merit providing the same Color temperature: The temperature of a information as NEP but describes the black body radiator such that its chromati- characteristic such that the lower the radia- city is the same as that of the light under con- tion level to which the photodetector can sideration. (109) respond, the higher the detectivity. See Noise Conduction band: A partially filled energy Equivalent Power. (108) band in which the electrons can move freely, allowing the material to carry an electric cur- Discriminator,constant-fraction pulse- rent. The term is usually restricted to height: A pulse-height discriminator in semiconductors and insulators, where the which the threshold changes with input conduction band is normally empty and is amplitude in such a way that the triggering separated by an energy-gap from the full point corresponds to a constant fraction of bands below it. (109) the input pulse height. (112) Count (in radiation counters): A single Discriminator, pulse-height: A circuit that response of the counting system. (112) produces a specified output signal if and Counting efficiency (scintillation counters): only if it receives an input pulse whose The ratio of (1) the average number of amplitude exceeds in one case or is less than photons or particles of ionizing radiation an assigned value in another case. (112) that produce counts to (2) the average Donor: An impurity element in an n-type number incident on the sensitive area. (112) semiconductor that may become ionized by Crosstalk: As applied to photomultipliers losing an electron to the conduction band used in liquid scintillation counting, light and induce conduction by electrons. For ex- originating internally in one photomultiplier ample, phosphorus with a valence of five is a and transmitted to another, causing coinci- possible donor impurity in silicon (valence, dent background pulses. four).

126 Glossary of Terms

Dynode: An electrode that performs a incident upon a surface. It is recommended, useful function, such as current amplifica- however, that the use of the term exitance be tion, by means of secondary emission. (107) restricted to emission from a surface because EADCI, Equivalent Anode Dark Current terms such as irradiance, spectral irradiance, Input: The input flux in lumens or watts at a and illuminance are commonly used to in- specific wavelength which results in an in- dicate flux density incident on a surface. crease in the anode current of a photomulti- Extended Red Multi-Alkali, ERMA: A plier tube just equal to the anode dark cur- designation of a Na2K Sb:Cs photocathode rent. processed to obtain increased red-near- E2/B: A figure of merit used to evaluate infrared response. performance of liquid scintillation counters. Fall time: The mean time difference of the E is the counting efficiency in per cent. B is trailing edge of a pulse between the 90- and the number of background coincident counts 10-per cent amplitude points. per minute. Fatigue: The tendency of a photomultiplier Electron affinity, EA: The energy, usually responsivity to decrease during operation. expressed in electron volts, required to move Most commonly, responsivity loss is the an electron from the bottom of the conduc- result of a lowering of secondary emission, tion band to the vacuum level. particularly in the latter stages of a photo- Electron multiplier: That portion of the multiplier. Recovery may or may not occur photomultiplier consisting of dynodes that during a period of idleness. produce current amplification by secondary Fermi level: The value of the electron energy electron emission. (112) at which the Fermi distribution function has Electron resolution: The ability of the elec- the value one-half. (107) tron multiplier section of the photomulti- Focusing electrode: An electrode whose plier to resolve inputs consisting of n and potential is adjusted to control the cross- n + 1 electrons. This ability may be expressed sectional area of the electron beam. (112) as a fractional FWHM of the nth peak, or as Flying-Spot Scanner: A system for generat- the peak-to-valley ratio of the nth peak to the th ing video signals for a television display. In a valley between the n and the (n+ 1)th typical conception, the scanned raster of a peaks. (112) cathode-ray tube is focused onto a photo- Electron volt: The energy received by an graphic transparency. The modulated trans- electron in falling through a potential dif- mitted light signal is directed onto the photo- ference of one volt. (108) cathode of a photomultiplier tube which provides the electrical video signal. Equivalent Noise Input, ENI: That value of input radiant or luminous flux that produces Footcandle: A unit of illuminance equal to an rms signal current that is just equal to the one lumen per square foot. The SI unit of il- rms value of the noise current in a specified uminance, the lux (lumen per square meter), bandwidth (Usually 1 Hz). See Noise is preferred. (108) Equivalent Power. (108) Footlambert: A unit of luminance equal to Exitance: The density of radiant flux emit- the candela per square meter (nit), is pre- ted from a surface. is the ferred. (108) integral of radiant flux over all wavelengths with units of watt per square meter. Lumi- Forbidden band: In the band theory of nous exitance is the total of all luminous flux solids, a range of energies in which there are from a surface with units of lumen per no electronic levels. (109) square meter. Spectral radiant exitance is the Full Width at Half Maximum, FWHM: The exitance at a particular wavelength for a full width of a distribution measured at half the maximum ordinate. For a normal distri- specified wavelength interval; units are watt 1/2 per square meter and micrometer. The term bution it is equal to 2(2 ln 2) times the stan- emittance (now deprecated) is synonomous with exitance. Sometimes the term exitance Gain (photomultipliers). See Current Ampli- is used to indicate the density of radiant flux fication.

127 Photomultiplier Handbook

Gamma-ray camera: A device used in It is expressed in lumens per watt. For exam- nuclear medicine to image distributions of ple, the maximum luminous efficacy of a gamma-ray emitters. In a typical instrument, black body (which occurs at about 6600 K) is gamma rays emanating from tracer elements 95 lumens per watt. (108) introduced into the patient are collimated Luminous efficiency: The ratio of the and cause scintillations in a single large, thin luminous efficacy of a given source to the sodium iodide crystal. An array of photo- maximum spectral luminous efficacy. For multiplier tubes views the crystal and pro- example, the maximum luminous efficiency vides addressing information with which an of a black body (which occurs at about 6600 output image of dots is constructed on a K) is 95 lumens per watt/680 lumens per watt cathode-ray tube. = 0.14. Hertz: Hz, the SI unit of frequency Luminous intensity: The luminous flux per equivalent to cycle per second. (108) unit solid angle in the direction in question. Hysteresis: A borrowed term to describe a It is expressed in candelas (lumens per stera- cyclic gain variation in photomultipliers dian). (107) sometimes observed as a result of insulator Lux, lx: The SI unit of illuminance equal to charging and discharging. the flux of one lumen uniformly distributed Illuminance: The density of the luminous over an area of one square meter. flux on a surface; it is the quotient of the Microchannel plate: An array of small flux by the area of the surface when the lat- aligned channel multipliers usually used for ter is uniformly illuminated. The SI unit is intensification. (107) the lux, lumen per square meter. (109) Multichannel Analyzer, MCA: See Pulse- Irradiance: The density of radiant flux on a height analyzer. surface; it is the quotient of the flux by the Negative Electron Affinity, NEA: A term area of the surface when the latter is referring to an electron emitter whose sur- uniformly irradiated. The SI unit is the watt face has been treated with an electropositive per square meter. material in such a way that the conduction Lambert’s cosine law: A law stating that the band minimum lies above the vacuum level. flux per solid angle in any direction from a Nit: The name recommended by the Inter- plane surface varies as the cosine of the angle national Commission on Illumination for between that direction and the perpendicular the unit of luminance equal to one candela to the surface. (107) per square meter. Note: Candela per square Light pipe: An optical transmission element meter is the unit of luminance in the Interna- that utilizes unfocused transmission and tional System of Units (SI). (107) reflection to reduce photon losses. (112) Noise (photomultiplier tubes): The random Liquid scintillation counter: The combina- output that limits the minimum observable tion of a liquid scintillator and one or more signal from the photomultiplier tube. (112) (usually two) photomultiplier tubes used to Noise Equivalent Power, NEP: The radiant measure or count radioactive disintegra- flux in watts incident on a detector which tions. The most common application is the gives a signal-to-noise ratio of unity. The counting of beta rays emanating from a bandwidth and the manner in which the sample mixed with the liquid scintillator. radiation is chopped must be specified as Lumen, lm: The SI unit of luminous flux. It well as the spectral content of the radiation. is equal to the flux through a unit solid angle The most common spectral specification is (steradian) from a uniform point source of for monochromatic radiation at the peak of one candela. (107, 108) the detector response. (Some detector manufacturers rate their detectors in terms Luminance: The luminous intensity per pro- of an NEP having units of watts Hz - 1/2. jected area normal to the line of observation. Assuming that the noise spectrum is flat Formerly called photometric brightness or within the range of the specification and that brightness. (108) NEP is normally specified for a bandwidth Luminous efficacy: The quotient of the of one hertz, the two forms of NEP are total luminous flux by the total radiant flux. numerically equal.) (108) Glossary of Terms

Noise&signal, (photomultiplier tubes): The Pulse-height resolution, PHR: The ratio of noise output resulting from the statistical the full width at half maximum of the pulse- variation in the signal current itself as con- height-distribution curve to the pulse height trasted with that which may be present when corresponding to the maximum of the distri- the detector is in the dark. bution curve. In scintillation spectroscopy, it Opaque photocathode: A photocathode is customary to state pulse-height resolution wherein photoelectrons are emitted from the as a percentage. (112) same surface as that on which the photons Pulse jitter: A relatively small variation of are incident. (Also called reflective photo- the pulse spacing in a pulse train. In cathode.) (112) photomultipliers, pulse jitter is the result of Peak-to-valley ratio: In a pulse-height- electron transit time variations. (109) distribution characteristic, the ratio of the Pulse width: The time interval between the counting rate at the maximum to that at the first and last instants at which the instan- minimum-usually preceding the maximum. taneous amplitude reaches a stated fraction Photocathode: An electrode used for ob- of peak pulse amplitude. (109) taining photoelectric emission when ir- Quantum efficiency (photocathodes): The radiated. (112) average number of electrons photoelectric- Photocell: A solid-state photosensitive elec- ally emitted from the photocathode per inci- tron device in which use is made of the varia- dent photon of a given wavelength. (107) tion of the current-voltage characteristic as a Rad: A unit of absorbed radiation equal to function of incident radiation. (112) 100 ergs per gram-O.01 J/kg in SI units. Photomultiplier, PMT: A phototube with (109) one or more dynodes between its photocath- Radiance: The radiant flux per unit solid ode and output electrode. (112) angle per unit of projected area of the Photon counting: The technique, using a source. The SI unit is the watt per steradian photomultiplier, of counting output pulses and square meter, Wsr - 1 m - 2. (109) originating from single photoelectrons. : The radiant flux pro- Photopic vision: Vision mediated essential- ceeding from the source per unit solid angle ly or exclusively by the cones. It is generally in the direction considered. The SI unit is associated with adaptation to luminance of watt per steradian. (107) at least 3 candelas per square meter. (107) Reflectance: The ratio of the radiant flux Phototube: An electron tube that contains a reflected from a body of material to that in- photocathode and has an output depending cident upon it. (See Absorptance.) at every instant on the total photoelectric Reflective photocathode: A photocathode emission from the irradiated area of the wherein photoelectrons are emitted from the photocathode. (112) same surface as that on which the photons Plateau: (counter): The portion of the are incident. (Also called opaque photocath- counting-rate-versus-voltage characteristic ode.) curve in which the counting rate is substan- Rem: Abbreviation for roentgen equivalent tially independent of the applied voltage. man. (1) In older usage, the dose (absorbed) (109) of any ionizing radiation that will produce Pulse-height analyzer, PHA: An instrument the same biological effect as that produced capable of indicating the number or rate of by one roentgen of high voltage x-radiation. occurrence of pulses falling within each of (2) The unit of the RBE (relative biological one or more specified amplitude ranges. effectiveness) dose that is equal to the ab- (112) sorbed dose in rads times the RBE. (109) Pulse-height distribution: A histogram dis- Resistance per square: The resistance of a playing the pulse count versus channel num- square of a thin conductive coating mea- ber as obtained with a multichannel ana- sured between opposite sides of the square. lyzer, particularly as applied to scintillation The value is independent of the size of the counting. square.

129 Photomultiplier Handbook

Responsivity: The ratio of the output cur- is 680 lumens per watt at a rent or voltage to the input flux in watts or wavelength of 555 nm. (107) lumens. For example, as applied to photo- Spectral luminous efficiency (radiant flux): , multipliers: radiant responsivity expressed in The ratio of the spectral luminous efficacy mA W - 1 at a specific wavelength or for a given wavelength to the maximum spectral luminous efficacy. Accordingly, the lm - 1. (108) spectral luminous efficiency, V(X), is the Rise time: The mean time difference of the ratio V(X)= K(X)/680 and is identical to the leading edge of a pulse between the 10- and standard visibility factor of the photopic 90-percent amplitude points. human eye and to the y-tristimulus value Roentgen: A unit of X- or gamma-radiation exposure such that the associated secondary tristimulus system. (107, 111) ionizing particles produce, in air, ions carry- Spectral radiant intensity: Radiant intensity ing one electrostatic unit of charge of either per unit wavelength interval; for example, sign per 0.001293 gram of air. This quantity watts per (steradian-nanometer). (107) is the equivalent of 2.58 x 10 -4 coulomb per Stage: One step of a multiplier, as one kilogram of air. (109) dynode stage. Scintillation counter: The combination of Stem: The portion of a photomultiplier scintillator , photomultiplier, and associated envelope containing the leads to electrodes. circuitry for detection and measurement of Steradian: The unit of solid angle which ionizing radiation. (112) subtends an area equal to the square of the Scintillator: The body of scintillator radius. (107) material together with its container. (107) Tea-cup: Descriptive term for an RCA Scotopic vision: Vision mediated essentially photomultiplier type having a large cup- or exclusively by the rods. It is generally shaped first dynode. associated with adaptation to luminance Time jitter: See Transit-time spread. below about 0.03 candela per square meter. Time-to-amplitude converter, TAC: An in- (107) strument producing an output pulse whose Secondary emission: Electron emission amplitude is proportional to the time dif- from solids or liquids due directly to bom- ference between start and stop pulses. (112) bardment of their surfaces by electrons or Traceability: Process by which the assigned ions. (107) value of a measurement is compared, di- Secondary emission ratio (electrons): The rectly or indirectly, through a series of average number of electrons emitted from a calibrations to the value established by the surface per incident primary electron. U.S. national standard. (107) Note: The result of a sufficiently large Transit time: For a discussion of the several number of events should be averaged to en- definitions of this term, refer to the section sure that statistical fluctuations are negli- “Time Effects” in Chapter 4. Photomulti- gible. (107) plier Characteristics. Semitransparent photocathode: A photo- Transit-time spread: The FWHM (full- cathode in which radiant flux incident on width-at-half maximum) of the time distri- one side produces photoelectric emission bution of a set of pulses each of which cor- from the opposite side. Synonymous with responds to the photomultiplier transit time Transmission-mode photocathode. (112) for that individual event. (112) Sensitivity: See Responsivity, the preferred Transmission-mode photocathode: A pho- term. tocathode in which radiant flux incident on Spectral luminous efficacy (radiant flux): one side produces photoelectric emission The quotient of the luminous flux at a given from the opposite side. Synonymous with wavelength by the radiant flux at that Semitransparent photocathode. (112) wavelength. It is expressed in lumens per Transmittance: The ratio of the radiant flux watt. The maximum spectral luminous ef- transmitted through a body of material to Glossary of Terms that incident upon it. See Absorptance. REFERENCES Vacuum level: The minimum potential ener- 107. IEEE Standard Dictionary of Electrical gy level an electron must reach to escape en- and Electronics Terms, Wiley-Interscience. tirely from the attraction of a solid or an 108. RCA Electro-Optics Handbook, atom. EOH-11. Valence band: The range of energy states in 109. The International Dictionary of the spectrum of a solid crystal in which lie Physics and Electronics, D. Van Nostrand. the energies of the valence electrons that 110. Information Transmission, Modula- bind the crystal together. (107) tion, and Noise, Mischa Schwartz, McGraw- Hill, 1959. Venetian-blind dynode: A descriptive term 111. Leo Levi, Applied Optics, Vol. 1, John for a photomultiplier dynode structure. The Wiley and Sons, Inc., 1968. dynode is constructed with a number of 112. An American National Standard, parallel slats with space between permitting IEEE Standard Test Procedures for Photo- secondary electrons to be directed to the next multipliers for Scintillation Counting and stage. Glossary for Scintillation Counting Field. Voltage divider (photomultiplier): A series ANSI N42.9-1972; IEEE Std 398-1972. string of resistors across which a voltage is applied providing an appropriate voltage drop per stage. Work function: The minimum energy re- quired to remove an electron from the Fermi level of a material into field-free space. (107)

131 Photomultiplier Handbook

Appendix C- Spectral Response Designation Systems

The spectral response of a photomultiplier mation is of value to the photomultiplier tube depends primarily on the chemical com- user because it will help in the interpretation ponents of the photocathodes. Differences of the published spectral characteristics data in spectral response, however, can result for a specific tube type regardless of which from variations in the processing of response designation system is used. photocathodes even for the same chemicals. Some photocathodes are of the “opaque” or Numerical System “reflective” type where the photoemission is In 1971 RCA/BURLE introduced a from the same side as the excitation. Others, numerical system of spectral response des- having essentially the same chemistry but of ignations supplementing and overlapping the “semitransparent” type where the the JEDEC S-designation system. The rea- photoemission is from the side opposite son for this extended system was the prolif- from that of the excitation, may have a eration of spectral responses resulting from somewhat different spectral response. The the combinations of different glasses, new short-wavelength-cutoff characteristic of the photocathode types, and processing varia- spectral response is the result of the tions. Table C-II is a catalogue of this 1971 transmission characteristic of the particular numbering system providing an identifica- glass used for the photocathode window. tion and description of each number. S-Designation System Alphanumeric Coded System In the early 1940’s, the JEDEC (Joint In 1976, RCA/BURLE changed its number Electron Devices Engineering Council) in- system for spectral-response characteristics to dustry committee on photosensitive devices an alphanumeric combination coded system. developed the “W-system of designating The new designations were combinations of spectral responses. The philosophy included alphanumerics based on (1) the photocath- the idea that the product user need only be ode material, (2) the window material, and concerned about the response of the device, (3) the photocathode operating mode. Table not by how it might be fabricated. And, in C-III provides the code for the spectral fact, the chemical compositions of some of response designation in four columns. The the photocathodes and dynodes were con- first two digits in the designation number sidered proprietary by various manufac- (Column I) indicate the photocathode turers. S-numbers were registered from S-l material; the following alphabetic character through S-40. See Table C-I. Subsequently, (Column II) indicates the window material; the lack of activity of the industry committee the next alphabetic character (Column III) on photosensitive devices resulted in in- indicates the photocathode operating mode. dependent assignment of codes for spectral Where required, the letter “X” is used as a responses. In other cases, the spectral suffix to the designation to indicate an ex- response was described by a curve and an ac- tended response in the red or near infrared. tual description of the photocathode com- As an example of the usage of this system, position and of the window material. tube type 931A has a spectral response that The following material is a brief account was previously designated as 102 (S-4) in of the different designation systems and Table C-II. This tube type has a Cs3Sb their relationship and meaning. This infor- photocathode, a 0080 lime glass window,

132 Spectral Response Designation Systems and a reflection-type photocathode. Its few are sufficiently acquainted with the code designation according to the 1976 code to make good use of it. Instead, most system is 20AR. knowledgeable customers prefer to be in- Similarly, a tube type having a Cs3Sb formed directly of the nature of the photocathode, a 0080 glass window, and a photocathode and the window. As a result, transmission-type photocathode is desig- therefore, BURLE has discontinued the use of nated 20AT. This response was previously coded spectral response designations except designated 107 (S-l 1). for occasional reference to some of the more Current Practices common JEDEC S-numbers, which have become well established. Instead, reference When the coded spectral response designa- is made to the photocathode material, the tion system was devised in 1976, it was an- window material, and any special processing ticipated that all the information provided information. In addition, typical spectral by the code would be useful to customers in response and other related data are pro- specifying the type of photomultiplier vided . needed. Experience has indicated that very

133 Photomultiplier Handbook

Table C-I Spectral Response Designations as Specified by JEDEC Photocathode S-Designation Composition Window* Notes S-l Ag-O-Cs lime glass semitransparent or opaque S-2 (Obsolete; formerly similar to S-l) S-3 Ag-O-Rb lime glass opaque

S-4 Cs3Sb lime glass opaque S-5 Cs3Sb Corning 9741 opaque S-6 Na uv transmitting opaque S-7 Ag-O-Rb-Cs borosilicate opaque S-8 Cs3Bi lime glass opaque S-9 (Obsolete; formerly similar to S-l1) S-10 Ag-Bi-O-Cs lime glass semitransparent

S-11 Cs3Sb lime glass semitransparent S-12 (CdS - a photoconductive crystal) S-13 Cs3Sb fused silica semitransparent S-14 (Ge - photovoltaic) S-15 (CdS-CdSe - a photoconductor) S-16 (CdSe - a photoconductor)

S-17 Cs3Sb lime glass opaque with reflecting substrate S-18 (Sb-S - photoconductor, camera tubes)

S-19 Cs3Sb fused silica opaque S-20 Na2KSb:Cs lime glass semitransparent S-21 Cs3Sb Corning 9741 semitransparent S-22 (Not used) S-23 Rb-Te fused silica semitransparent S-24 (Not used) S-25 Na2KSb:Cs lime glass semitransparent, processed for ex- tended red response S-26 (InSb - photovoltaic) S-27 (Ge:Au - photoconductive) S-28 (InAs - photovoltaic) S-29 (PbSe - photoconductive) S-30 (Ge:Cu - photoconductive) S-31 (PbS - photoconductive) S-32 (PbS - photoconductive) S-33 (PbS - photoconductive) S-34 (InAs - photovoltaic) S-35 (InSb - photoconductive) S-36 (GaAs - photovoltaic) S-37 (Si - photovoltaic) *Window may be of material other than that specified, but it will have equivalent spectral S-38 (PbSe - photoconductive) characteristics. S-39 (PbSe - photoconductive) S-40 (Ge:Hg - photoconductive) References: 113, 114 134 Spectral Response Designation Systems

Table C-II RCA/BURLE 1971 Spectral Response Numbering Code

JEDEC Photocathode Number S-Designation Composition Window Notes 101 S-l Ag-o-cs lime glass

102 S-4 Cs3Sb lime glass 103 Cs3Sb Corning 9741 glass uv transmitting

104 S-5 Cs3Sb Corning 9741 glass uv transmitting 105 S-8 Cs-Bi lime glass 106 S-10 Ag-Bi-O-Cs lime glass

107 S-l 1 Cs3Sb lime glass

108 S-13 Cs3Sb SiO2

109 S-19 Cs3Sb SiO2 110 S-20 Na2KSb:Cs Borosilicate glass 111 Na2KSb:Cs lime glass 112 Na2KSb:Cs Corning 9823 glass uv transmitting 113 Na2KSb:Cs Pyrex 114 Na2KSb:Cs SiO2

115 K2CsSb Lime or boro- silicate glass

116 K2CsSb Pyrex 117 K2CsSb Corning 9823 glass uv transmitting

118 K2CsSb Corning 9741 glass uv transmitting 119 Na2KSb:Cs Pyrex extended red: ERMA III

120 K2CsSb Sapphire, uv grade 121 Cs-Te SiO2

122 K2CsSb Al203

123 Cs3Sb Sapphire, uv grade 124 Cs3Sb Corning 9741 glass 125 Cs-Te LiF 126 K2CsSb Borosilicate or lime glass 127 Ag-Bi-O-Cs Corning 9741 glass 128 Ga-As Corning 9741 glass 129 Ga-As-P Corning 9741 glass 130 Na2KSb:Cs Borosilicate or lime glass 131 Na2KSb:Cs Borosilicate glass extended red: ERMA III 132 Na2KSb:Cs Lime or extended red: borosilicate glass ERMA II

133 K2CsSb SiO2 134 Ga-As Sapphire, uv grade 135 Ga-As-P Sapphire, uv grade

136 K2CsSb Lime glass

135 Photomultiplier Handbook

Table C-II RCA/BURLE 1971 Spectral Response Numbering Code (cont’d)

JEDEC Photocathode Window Notes Number S-Designation Composition

137 Na2KSb:Cs Corning 9741 glass extended red: ERMA II 138 Na2KSb:Cs Corning 9741 glass extended red: ERMA I 139 Na2KSb Borosilicate glass high temperature

140 In.06-Ga.94-As Corning 9741 glass Type I Corning 9741 glass Type II 141 In.12-Ga-.88-As Corning 9741 glass Type III 142 In.18-Ga.82-As

Table C-III RCA/BURLE 1976 Coded System for Spectral Response Designation

Column I Column II Column III 10 = AgOCs A = 0080 (lime glass) or D = Dormer-window type 15 = AgBiOCs 7056 (Borosilicate glass) R = Reflection Type 20 = CsSb C = 7740 (Pyrex) T = Transmission Type 25 = CsBi E = 9741 (UV transmitting 30 = CsTe glass) 35 = KCsSb (Bialkali) G = 9823 (UV transmitting 40 = NaKSb (High glass) temperature bialkali) J = SiO2 (Fused silica) 45 = RbCsSb M = UV-grade Sapphire 50 = NaKCsSb (Multialkali) P = LiF 51 = NaKCsSb (ERMA I) 52 = NaKCsSb (ERMA II) 53 = NaKCsSb (ERMA III) 60 = GaAs Column IV 71 = InGaAs (Type I) X = Extended Response 72 = InGaAs (Type II) 73 = InGaAs (Type III)

REFERENCES 114. “Typical characteristics of photosen- 113. “Relative spectral response data for sitive surfaces,” JEDEC Publication No. 61, photosensitive devices (“S” curves)“, Electronic Industries Association, Engineer- JEDEC Publication No. 50, Electronic In- ing Department, 2001 Eye Street, N. W., dustries Association, Engineering Depart- Washington, D.C. 20006. (1966) ment, 2001 Eye Street, N.W., Washington, D.C. 20006 (1964)

136 Spectral Response Designation Systems

Appendix D- Photometric Units and Photometric-to-Radiometric Conversion

Photometry is concerned with the mea- energy spectrum for each wavelength in the surement of light. Because the origins of the visible range, assuming foveal vision. An ab- photoelectric industry were associated with solute “sensitivity” figure established for the visible spectrum, the units first used for the standard eye relates photometric units evaluating photosensitive devices were pho- and radiant power units. At 555 nanometers, tometric. Today, however, even though the wavelength of maximum sensitivity of many of the applications of photosensitive the eye, one watt of radiant power cor- devices are for radiation outside the visible responds to approximately 680 lumens. The spectrum, the photometric units are still re- quotient of the luminous flux at a given tained for many purposes. Because these wavelength by the radiant flux at that units are based on the characteristics of the wavelength is referred to as the Spectral eye, this discussion begins with a considera- Luminous Efficacy , tion of some of these characteristics. D-l CHARACTERISTICS OF THE EYE Various determinations of the maximum The in the of the human eye value, K (555 nanometers), have varied are of two kinds: “cones” which predomi- somewhat from the nominal value of 680 nate the central (or foveal) vision and lumens per watt. “rods” which provide peripheral vision. The For the dark-adapted eye, the peak sen- cones are responsible for our color vision; sitivity increases and is shifted toward the rods provide no color information but in the violet end of the spectrum. A tabulation of dark-adapted state are more sensitive than the relative scotopic vision is also given in the cones and thus provide the basis of dark- Table D-I. The peak luminosity for scotopic adapted vision. Because there are no rods in vision occurs at 511 nanometers and is the the fovea1 region, faint objects can more equivalent of 1746 lumens/watt. Fig. D-l readily be observed at night when the eye is shows the comparison of the absolute not exactly directed toward the faint object. luminosity curves for scotopic and photopic The response of the light-adapted eye (cone vision as a function of wavelength. vision) is referred to as the Photopic eye The sensitivity of the eye outside the response; the response of the dark-adapted wavelength limits shown in Table D-I is very eye (rod vision) is referred to as Scotopic eye low, but not actually zero. Studies with in- response. tense infrared sources have shown that the Although characteristics of the human eye eye is sensitive to radiation of wavelength at vary from person to person, standard least as long as 1050 nanometers. Fig. D-2 luminosity coefficients for the eye were shows a composite curve given by Griffin, defined by the Commission Internationale Hubbard, and Wald115 for the sensitivity of d’Eclairage (International Commission on the eye for both foveal and peripheral vision Illumination) in 193 1. These standard C . I. E. from 360 to 1050 nanometers. According to luminosity coefficients for photopic vision Goodeve116 the ultraviolet sensitivity of the are given in Table D-I. They represent the eye extends to between 302.3 and 312.5 relative luminous equivalents of an equal- nanometers. Below this level the absorption

137 Photomultiplier Handbook

Table D-I Relative Luminosity Values for Photopic and Scotopic Vision Wavelength (nm) (B>3 cd m-2) (B<3 x 10-5cdm-2) - 0.0003 - 0.0008 - 0.0022 0.00004 0.0055 0.00012 0.0127 0.0004 0.0270 410 0.0012 0.0530 420 0.0040 0.0950 430 0.0116 0.157 0.023 0.239 450 0.038 0.339 460 0.060 0.456 470 0.091 0.576 480 0.139 0.713 490 0.208 0.842 500 0.323 0.948 510 0.503 0.999 520 0.710 0.953 530 0.862 0.849 540 0.954 0.697 550 0.995 0.531 560 0.995 0.365 570 0.952 0.243 580 0.870 0.155 590 0.757 0.0942 0.631 0.0561 610 0.503 0.0324 620 0.381 0.0188 630 0.265 0.0105 640 0.175 0.0058 650 0.107 0.0032 660 0.061 0.0017 670 0.032 0.0009 680 0.017 0.0005 690 0.0082 0.0002 700 0.0041 0.0001 710 0.0021 - 720 0.00105 - 730 0.00052 - 740 0.00025 - 750 0.00012 - 760 0.00006 - 770 0.00003 -

138 Photometric Units and Photometric-to-Radiometric Conversion of radiation by the proteins of the eye lens apparently limits further extension of vision into the ultraviolet. Light having a wave- 2 I length of 302 nanometers is detected by its -PERIPHERY fluorescent effect in the front part of the eye.

-8

-10 300 . 500 700 900 1100 WAVELENGTH - NANOMETERS

92CS- 32468 Fig. D-2 - Relative spectral sensitivity of the dark-adapted foveal and peripheral retina. A suitable standard for practical photo- electric measurements is the AJ2239 cali- brated lamp, which operates at a current of about 4.5 amperes and a voltage of 7 to 10 volts. A typical lamp calibrated at a color temperature of 2856 K provides a luminous intensity of 55 candelas. The luminous inten- sity of a tungsten lamp measured in candelas is usually numerically somewhat greater than

WAVELENGTH-NANOMETERS the power delivered to the lamp in watts. A color temperature of 2870 K served as 92CS- 32467 the basic test standard in this country for Fig. D-7 - Absolute luminosity curves for about 30 years. A change had been made to scotopic and photopic eye response. agree with C.I.E. illuminant A, a more wide- spread standard that at first required a color temperature of 2854 K, but has more recent- PHOTOMETRIC UNITS ly been adjusted to 2856 K to accommodate Luminous intensity (or candlepower) de- to the international practical temperature scribes luminous flux per unit solid angle in a scale of 1968. See also Appendix F. The dif- particular direction from a light source. The ference between the old lamp standard at measure of luminous intensity is the fun- 2870 K and at the new temperature is damental standard from which all other pho- generally negligible. 117 tometric units are derived. The standard of Luminous flux is the rate of flow of light luminous intensity is the candela; the older energy, the characteristic of radiant energy term candle is sometimes still used, but that produces visual sensation. The unit of refers to the new candle or candela. luminous flux is the lumen, which is the flux The candela is defined by the radiation emitted per unit solid angle by a uniform from a black body at the temperature of point source of one candela. Such a source solidification of . A candela is one- sixtieth of the luminous intensity of one A radiant source may be evaluated in square centimeter of such a radiator. terms of luminous flux if the radiant-energy

139 Photomultiplier Handbook

Table D-II is the total radiant power in watts per unit Typical Values of Natural Scene Illuminance wavelength, total radiant power over all Approx. Levels of Sky Condition Illuminance- as follows: lux (lm m-2)

D-2 Direct sunlight ...... l-l.3 x 105 Full daylight (not direct sunlight) . l-2 x 104 Overcast day ...... 103 efficiency. The lumen is the most widely Very dark day ...... 102 used unit in the rating of photoemissive Twilight ...... 10 devices. For photomultipliers, the typical Deep twilight ...... 1 test levels of luminous flux range from 10 -7 Full moon ...... 10-1 to 10-5 lumen (0.1 to 10 microlumens). Quarter moon ...... 10 -2 Illuminance (or illumination) is the density Moonless, clear night sky...... 10 -3 of luminous flux incident on a surface. A Moonless, overcast night sky ...... 10 -4 common unit of illuminance is the footcan- dle, the illumination produced by one lumen uniformly distributed over an area of one square foot. It follows that one candela pro- duces an illuminance of one footcandle at a distance of one foot. The preferred SI unit (International System of Units) of illumi- nance is the lux, which is the illumination produced by one lumen uniformly distri- buted over an area of one square meter. (1 lx = 1 lm m -2) It also follows that one candela produces an illuminance of one lux at a distance of one meter.

1 lux = 0.0929 footcandle

Table D-II lists some common values of il- luminance. Further information concerning natural radiation is shown in Fig. D-3 which indicates the change in natural illumination at ground level during, before, and after sunset for a condition of clear sky and no

92CS-32469 Photometric luminance (or brightness) is a Fig. D-3 - Natural illuminance on the earth measure of the luminous flux per unit solid for the hours immediately before and after angle leaving a surface at a given point in a sunset with a clear sky and no moon. given direction, per unit of projected area. The term photometric luminance is used to For a surface that is uniformly diffusing, distinguish a physically measured luminance luminance is the same regardless of the angle from a subjective brightness. The latter from which the surface is viewed. This con- varies with illuminance because of the shift dition results from the fact that a uniformly in spectral response of the eye toward the diffusing surface obeys Lambert’s Law (the blue region at lower levels of illuminance. cosine law) of emission. Thus, both the emis- The term luminance describes the light emis- sion per unit solid angle and the projected sion from a surface, whether the surface is area are proportional to the cosine of the self-luminous or receives its light from some angle between the direction of observation external luminous body. and the surface normal.

140 Photometric Units and Photometric-to-Radiometric Conversion

The SI unit of luminance is the candela per square meter or a lumen per steradian and square meter. This unit is called the nit. A commonly used unit of luminance is the and the total flux from the area, A, is given by per square foot. 1 nit = 0.2919 footlambert

The relationship between luminance and total luminous flux from a uniform diffuser is illustrated by the use of Fig. D-4.

Fig. D-4 - Diagram illustrating Lambert’s law and the calculation of total luminous flux from a diffuse radiator.

Consider an elementary portion of the dif- fusing surface having an area, A, and a On the other hand if a perfectly diffusing luminance of L. The projected area at the and reflecting surface is illuminated with 1 sented by the differential area on the surface square meter (nits). Table D-III provides data on luminance luminance, L, represents the luminous flux values of various sources. Table D-IV is a per unit solid angle leaving the surface per conversion table for various photometric unit of projected area. Therefore, the flux units.

Table D-III Luminance Values for Various Sources

Luminance Luminance Source (Footlamberts) (Candelas m -2 )

Sun, as observed from Earth’s surface at meridian . . . . . 4.7 x 108 . . . . . 1.6 x 109 Moon, bright spot, as observed from Earth’s surface. . . 730 ...... 2500 Clear blue sky ...... 2300...... 7900 Lightning flash ...... 2 x 1010 ...... 7 x 1010 Atomic fission bomb, 0.1 millisecond after firing, 90-feet diameter ball ...... 6 x 1011 ...... 2 x 1012 Tungsten filament lamp, gas-filled, 16 lumen/watt . . . . 6 x 10 6 . . . . . 9 x 106 Plain carbon arc, positive crater ...... 4.7 x 106 . . . . . 1.6 x 107 Fluorescent lamp, T-12 bulb, cool white, 430 mA, medium loading...... 2000 ...... 7000 Color television screen, average brightness ...... 50 ...... 170

141 Photomultiplier Handbook

CALCULATION OF RADIANT Table F-I in Appendix F. The integrations RESPONSIVITY FROM indicated in Eq. D-7 have been performed LUMINOUS RESPONSITIVITY for most photocathode spectral responses. A Specification of photocathode responsi- vity is most frequently given in terms of watt, is provided in Table I. This factor lumens from a tungsten source at a color represents the ratio of the peak radiant temperature of 2856 K. If a relative spectral responsitivity in amperes per watt to the response is known, it is possible to calculate luminous responsivity in amperes per lumen. the absolute radiant responsivity of the photocathode as follows. Let the relative spectral response (with a Table D-IV maximum value of unity) of the photocath- Conversion Table for Various Photometric Units absolute radiant response at the peak of the SI Units Other Units response characteristic is then given by Luminous Intensity (I) lamp radiation striking the photocathode be 1 candela (cd) = 1 lumen/steradian (lm sr - 1) The response of the photocathode (in am- peres) to the total radiation is then given by

D-4 uniform point source of 1 candela The integration is done over the complete Illuminance (E) range of wavelengths either as limited by 1 lux (lx) = 1 lumen/ 1 footcandle (fc) = The light flux (in lumens) represented by 1 lux = 0.0929 footcandle the total radiant flux is given by Luminance (L) D-5 1 nit (nt) = 1 candela/ 1 footlambert (fL) =

0.ciency as given in Table D-I and 680 lumens 1 nit = 0.2919 footlambert per watt is taken as the maximum spectral luminous efficacy. The luminous responsivity of the photo- cathode in amperes per lumen is then given by the ratio of expressions D-4 and D-5: REFERENCES 115. D.R. Griffin, R. Hubbard, and G. D-6 Wald, “The sensitivity of the human eye to infrared radiation,” JOSA, Vol. 37, No. 7, pp 546, (1947). From Eq. D-6, the maximum radiant 116. C.F. Goodeve, “Vision in the ultra- amperes per watt: violet,” Nature (1934) 117. R.W. Engstrom and A.L. Morehead, “Standard test lamp temperature for photosensitive devices-relationship of ab- solute and luminous sensitivities,” RCA Note that the absolute magnitude of the Review, Vol. 28, pp 419-423 (1967) 118. I.E.S. Lighting Handbook, Illuminat- lamp operated at a color temperature of ing Engineering Society, New York, New York (1959).

142 Spectral Response and Source-Detector Matching

Appendix E- Spectral Response and Source-Detector Matching

This appendix covers the significance of the spectral response of photomultiplier of a photon is tubes; describes some of the methods used E-l for measuring this response; and discusses in some detail the calculations and other con- siderations useful for matching the radiation quency of the incident radiation, c is the source and the photomultiplier tube type for a specific application. the incident radiation. If the quantum effi- of emitted photoelectrons to the number of incident photons), the responsivity is given SPECTRAL-RESPONSE by CHARACTERISTICS A spectral-response characteristic is a display of the response of a photosensitive =-- device as a function of the wavelength of the hc E-2 exciting radiation. Such curves may be on an absolute or a relative basis. In the latter case in units of coulombs per joule or amperes the curves are usually normalized to unity at per watt where e is the charge on the elec- the peak of the spectral-response curve. For tron. Solving for the quantum efficiency: a photocathode the absolute radiant sen- sitivity is expressed in amperes per watt. Curves of absolute spectral response may also be expressed in terms of the quantum ef- ficiency at the particular wavelength. If the curve is presented in terms of amperes per watt, lines of equal quantum efficiency may be indicated for convenient reference. Typical curves are usually included in the E-3 published tube data. It should be understood that because of variations in processing, deviations from these typical curves may be expected. It would not be unusual for the MEASUREMENT TECHNIQUES wavelength for peak response to vary by 30 In order to determine a spectral-response nanometers from the typical value. The same characteristic, (1) a source of essentially variation may be expected in the long- monochromatic radiation, (2) a current- wavelength cutoff as judged by the wave- sensitive instrument to measure the output length for which the response is 10% of the of the photocathode, and (3) a method of maximum. On the other hand, the short- calibrating the monochromatic radiation for wavelength cutoff is more closely held by the its magnitude in units of power are required. glass transmission characteristic of the Monochromatic radiation is often pro- envelope. vided by a prism or a grating type of mono- The relationship between the spectral chromator. Interference filters may also be used to isolate narrow spectral bands.

143 Photomultiplier Handbook

Although interference filters do not provide greater than 1000 nanometers because of its the flexibility of a monochromator, they uniform and stable spectral-emission charac- may be indicated in situations in which teristic. A mercury vapor discharge lamp repeated measurements are required in a par- provides a high concentration in specific ticular region of the spectrum. radiation lines and thus minimizes the The width of the spectral transmission background scattered-radiation problem. band in these measurements must be narrow The mercury lamp is particularly useful in enough to delineate the spectral-response the ultraviolet end of the spectrum where the characteristic in the required detail. How- tungsten lamp fails. Another useful source ever, for the most part, spectral-response for the ultraviolet is the deuterium lamp. characteristics do not require fine detail and (See the discussion on radiant energy sources generally have broad peaks with exponential in Appendix F.) cutoff characteristics at the long-wavelength limit and rather sharp cutoffs at the short- MEASUREMENT OF wavelength end. For spectral measurements, RADIANT POWER OUTPUT therefore, a reasonably wide band is used. A radiation thermocouple or thermopile Such a band has the following important ad- having a black absorbing surface is com- vantages: (1) because the level of radiation is monly used to measure the radiation power higher, measurements are easier and more output at a specific wavelength. Although precise; and (2) spectral leakage in other these devices are relatively low in sensitivity, parts of the spectrum is relatively less impor- they do provide a reasonably reliable means tant. Spectral leakage is a problem in any of measuring radiation independent of the monochromator because of scattered radia- wavelength. The limitation to their accuracy tion, and in any filter because there is some is the flatness of the spectral absorption transmission outside the desired pass band. characteristic of the black coating on the A double monochromator may be used and detector. Throughout the visible and near- will greatly reduce the spectral leakage out- infrared regions there is usually no problem. side the pass band. The double monochro- There is some question, however, as to the mator is at a disadvantage in cost and com- flatness of the response in the ultraviolet plexity. If a pass band of 10 nanometers is part of the spectrum. used, spectral leakage can be insignificant The output of the thermocouple or for most of the spectral measurements. At thermopile is a voltage proportional to the the same time, this pass band is narrow input radiation power. This voltage is con- enough to avoid distortion in the measured verted by means of a suitable sensitive spectral-response characteristic. It is often voltmeter to a calibrated measure of power advisable to vary the pass band depending in watts. The calibration may be accom- upon what part of the curve is being plished by means of standard radiation measured. For example, at the long- lamps obtained from the National Bureau of wavelength cutoff where the response of the Standards. It is theoretically possible to photocathode may be very small, the leakage calculate the monochromatic power from a spectrum may play an important part; thus it knowledge of the emission characteristic of is advisable to increase the spectral bandpass the source, the dispersion characteristic of of the measurement. Wide pass-band color the monochromator or the transmission filters that exclude the wavelength of the characteristic of the filter, and from the measurement and include the suspected spec- transmission characteristic of the various tral leakage region, or vice versa, are used in lenses. This procedure is difficult, subject to checking the magnitude of the possible error, and is not recommended except leakage spectrum. perhaps in the case of a tungsten lamp source combined with a narrow-band filter system. ENERGY SOURCES Another useful reference standard is the Various radiant-energy sources are used to pyroelectric detector. In this case, the radia- advantage in spectral-response measure- tion to be measured must be interrupted by ments. A tungsten halogen lamp is useful means of a light chopper at about 15 hertz. from 350 nanometers to wavelengths much Absolute calibration might proceed in a

144 Spectral Response and Source-Detector Matching manner similar to that of the thermopile. MATCHING CALCULATIONS Special pyroelectric detectors have been The average power radiating from a light fabricated that can be self-calibrated by source may be expressed as follows: means of electric power input. 119 E-4 MEASUREMENT OF PHOTOCATHODE OUTPUT where PO is the incident power in watts per unit wavelength at the peak of the relative For measuring the spectral characteristic of a photocathode, a very sensitive ammeter which is normalized to unity. is required. When the output of the photo- If the absolute spectral distribution for the cathode of a photomultiplier is measured, light source and the absolute spectral the tube is usually operated as a photodiode response of the photomultiplier tube are by connecting all elements other than the known, the resulting photocathode current photocathode together to serve as the anode. Ik when the light is incident on the detector When the photomultiplier is operated as a can be expressed as follows: conventional photomultiplier, the output is very easy to measure. It is necessary, how- E-5 ever, to be careful to avoid fatigue effects which could distort the spectral-response photocathode in amperes per watt at the measurement. It should be noted that the spectral response of a photomultiplier may sents the relative photocathode spectral be somewhat different from that of the response as a function of wavelength nor- photocathode alone because of the effect of malized to unity at the peak. When Eq. E-4 initial velocities on collection efficiency at is solved for the peak power per unit wave- the first dynode and because of the possibili- length, PO, and this solution is substituted in- ty of a photoeffect on the first dynode by to Eq. E-5, the cathode current is expressed light transmitted through the photocathode, as follows: especially if the first dynode is a photosen- sitive material such as cesium-antimony. SOURCE AND DETECTOR MATCHING The ratio of the dimensionless integrals can One of the most important parameters to be defined as the matching factor, M. The be considered in the selection of a photo- matching factor is the ratio of the area under multiplier type for a specific application is the curve defined by the product of the the photocathode spectral response. The relative source and detector spectral curves spectral response of BURLE photomultiplier to the area under the relative spectral source tubes covers the spectrum from the ultra- curve. violet to the near-infrared region. In this range there are a large variety of spectral responses to choose from. Some cover nar- row ranges of the spectrum while others cover a very broad range. The published Fig. E-l shows an example of the data in- data for each photomultiplier type show the volved in the evaluation of the matching fac- relative and absolute spectral-response tor, M, as given in Eq. E-7. curves for a typical tube of that particular If the input light distribution incident on type. The relative typical spectral-response the detector is modified with a filter or any curves published may be used for matching other optical device, the matching-factor the detector to a light source for all but the formulas must be changed accordingly. If most exacting applications. The matching of the transmission of the filter or optical detector to source consists of choosing the photomultiplier tube type that has a spectral response providing maximum overlap of the spectral distributions of detector and light source.

145 Photomultiplier Handbook

When M is substituted for the integral ratio in Eq. E-6, the photocathode current becomes

E-10 In any photomultiplier application it is desirable to choose a detector having a photocathode spectral response that will maximize the photocathode current, Ik, for a given light source. Maximizing the cathode current is important to maximize the signal- to-noise ratio. From Eq. E-9, it can be seen that the product of the matching factor M and the peak absolute photocathode sensitiv- cathode current. The importance of taking into account the absolute photocathode sensitivity, as well as the matching factor, is illustrated by a com- parison of the S-l and S-20 photocathodes with a tungsten light source operating at a color temperature of 2856 K. The S-l and the S-20 matching factors are 0.516 and 0.112, respectively. From the matching fac- tors alone it appears that the S-l is the best WAVELENGTH-NANOMETERS choice of photocathode spectral response. The S-l has a peak absolute sensitivity of 2.3 92cs-32470 milliamperes per watt and the S-20 has a Fig. E-l - Graphic example of factors used in peak absolute sensitivity of 64 milliamperes evaluation of matching factor, M. per watt. Then, from Eq. E-9 the expected photocathode currents are Table E-I shows a number of matching fac- I (S-1)=0.0023 x 0.516 P=0.00119 P tors calculated for various light sources and k spectral response characteristics. When the Ik(S-20)=0.064 x 0.112 P=0.00717 P spectral range of a source exceeded 1200 These calculations show that the S-20 nanometers, the integration was terminated photocathode will provide a response to the at this wavelength. Because none of the tungsten lamp six times that of the S-l photoresponses exceed 1200 nanometers, photocathode. conclusions as to the relative merit of various combinations are still valid. It should also be noted that since these data were originally published there has been REFERENCES a proliferation of photocathode develop- 119. W.M. Doyle and B.C. McIntosh (Laser ment. Many of the new photocathodes, how- Precision Corp., Irvine, California) and Jon ever, have spectral responses similar to those Geist (National Bureau of Standards, Wash- listed in Table E-l. As a result, the spectral ington, D.C.), “Implementation of a system matching factors given could also be used of optical calibration based on pyroelectric for many of the new photocathodes. For radiometry,” Optical Engineering, Vol. 15, example, the data on S-l 1 could well be sub- No. 6, pp 541-548, (Nov.-Dec. 1976) stituted for photocathodes such as Rb2CsSb, 120. E.H. Eberhardt, “Source-detector K2CsSb or Na2KSb, with only moderate spectral matching factors,” Applied Optics, error. Vol. 7, p 2037 (1968)

146 Spectral Response and Source-Detector Matching

Table E-1 Spectral Matching Factors** Other Detectors Light Photocathodes Pho- Sco- topic topic S1 S4 S1O S11 S17 S20 S25 eye eye Source Notes k k k k k k k 1 m Phosphors P1 a 0.278 0.498 0.807 0.687 0.892 0.700 0.853 0.768 0.743 P4 a,b 0.310 0.549 0.767 0.661 0.734 0.724 0.861 0.402 0.452 P7 a 0.312 0.611 0.805 0.709 0.773 0.771 0.882 0.411 0.388 P11 a 0.217 0.816 0.949 0.914 0.954 0.877 0.953 0.201 0.601 P15 a 0.385 0.701 0.855 0.787 0.871 0.802 0.904 0.376 0.495 P16 a 0.830 0.970 0.853 0.880 0.855 0.902 0.922 0.003 0.042 P20 a 0.395 0.284 0.612 0.427 0.563 0.583 0.782 0.707 0.354 P22B C 0.217 0.893 0.974 0.960 0.948 0.927 0.979 0.808 0.477 P22G C 0.278 0.495 0.807 0.686 0.896 0.699 0.855 0.784 0.747 P22R C 0.632 0.036 0.264 0.055 0.077 0.368 0.623 0.225 0.008 P24 a 0.279 0.545 0.806 0.696 0.827 0.725 0.869 0.540 0.621 P31 a,d 0.276 0.533 0.811 0.698 0.853 0.722 0.868 0.626 0.65 1 NaI e 0.534 0.923 0.885 0.889 0.889 0.900 0.933 0.046 0.224 Lamps 2870/2856 std f 0.516* 0.046* 0.095* 0.060* 0.072* 0.112* 0.227* 0.071†* 0.040* Fluorescent g 0.395 0.390 0.650 0.496 0.575 0.635 0.805 0.502 0.314 Sun In space h 0.535* 0.308* 0.388* 0.328* 0.380* 0.406* 0.547* 0.179* 0.172* +2 air masses 0.536* 0.236* 0.348* 0.277* 0.315* 0.360* 0.513* 0.197* 0.175* Day sky 0.537* 0.520* 0.556* 0.508* 0.589* 0.581* 0.700* 0.170* 0.218* Blackbodies 6000K - 0.533* 0.308* 0.376* 0.320* 0.375* 0.397* 0.521* 0.167* 0.159* 3000K - 0.512* 0.053* 0.102* 0.067* 0.080* 0.120* 0.232* 0.075* 0.044* 2870 K - 0.504* 0.044* 0.090* 0.057* 0.069* 0.106* 0.216* 0.067* 0.038* 2856 K - 0.500* 0.042* 0.088* 0.055* 0.068* 0.103* 0.211* 0.065* 0.037* 2810 K - 0.493* 0.039* 0.081* 0.051* 0.062* 0.097* 0.150* 0.061* 0.034* 2042 K - 0.401* 0.008* 0.023* 0.011* 0.014* 0.033* 0.090* 0.018* 0.007* *Entry valid only for 300-1200-nm wavelength j From Gates (Science, Vol. 151, p 523 interval. (1966)) between 300 nm and 530 nm. bet- †For the total wavelength spectrum this entry ween 300 nm and 530 nm. would be 0.0294. A 12,000 K blackbody spectra) dis- Notes: a Registered spectral distribution. tribution was assumed between 530 nm Data extrapolated as required. and 1200 nm. b Sulfide type. k Registered spectral distribution. Data c BURLE data. extrapolated as required. d Low brightness type. l Standard tabulated e Harshaw Chemical Co. data. photopic visibility f Standard test lamp distribution. distribution. g General Electric Co. data. m Standard tabulated h From Handbook of . scotopic visibility i Approximately noon sea level flux at distribution. 60° latitude. ** Data from E.H. Eberhardt, 120.

147 Photomultiplier Handbook

Appendix F- Radiant Energy and Sources

Radiant energy is energy traveling in the hole. The radiation from the hole ap- form of electromagnetic waves. It is mea- proaches that from a theoretical black sured in joules, ergs, or calories. The rate of radiator if the cross-sectional area of the flow of radiant energy is called radiant flux, cavity is large compared with the area of the and it is expressed in watts (joules per exit hole. The characteristic of 100-per-cent second). absorption is achieved because any radiation Planck’s equation for the spectral radiant entering the hole is reflected many times in- exitance of a black body in a vacuum is side the cavity. The radiation distribution for a source which is not black may be calculated from F-l the black-body radiation laws provided the -2 -1 emissivity as a function of wavelength is (in W m m ), known. Spectral emissivity is defined as the where ratio of the output of a radiator at a specific wavelength to that of a black body at the h = Planck’s constant (J s) same temperature. Tungsten sources, for -1 which tables of emissivity data are c = velocity of light (m s ) 121 k = Boltzmann’s constant (J K-1) available, are widely used as practical T = absolute temperature (K) standards, particularly for the visible range. Tungsten radiation standards for the visible BLACK-BODY RADIATION range are frequently given in terms of color temperature, instead of true temperature. As a body is raised in temperature, it first The color temperature of a selective radiator emits radiation primarily in the invisible in- is determined by comparison with a black frared region. Then, as the temperature is in- body. When the outputs of the selective creased, the radiation shifts toward the radiator and a black body are the closest shorter wavelengths. A certain type of radia- possible approximation to a perfect color tion called black-body radiation is used as a match in the range of visual sensitivity, the standard for the infrared region. Other color temperature of the selective radiator is sources may be described in terms of the numerically the same as the black-body true black body. temperature. For a tungsten source, the rela- A black body is one which absorbs all inci- tive distribution of radiant energy in the visi- dent radiation; none is transmitted and none ble spectral range is very close to that of a is reflected. Because, in thermal equilibrium, black body, although the absolute tempera- thermal radiation balances absorption, it tures differ. However, the match of energy follows that a black body is the most effi- distribution becomes progressively worse in cient thermal radiator possible. A black the ultraviolet and infrared spectral regions. body radiates more total power and more power at a particular wavelength than any other thermally radiating source at the same TUNGSTEN SOURCES temperature. Although no material is ideally black, the equivalent of a theoretical black Tungsten Lamps body can be achieved in the laboratory by Tungsten lamps are probably the most im- providing a hollow radiator with a small exit portant type of radiation source because of Radiant Energy and Sources their availability, reliability, and constancy length. Methods of computing the response of operating characteristics. Commercial of a given photodetector to a particular photomultiplier design has been consider- radiation source are outlined in Appendix E, ably influenced by the characteristics of the Spectral Response and Source-Detector tungsten lamp. A relative spectral-emission Matching. characteristic for a tungsten lamp at 2856 K The relative spectral irradiance from a color temperature is shown in Fig. F-l. standard tungsten test lamp operated at 2870 K or 2854 K color temperature has been cal- culated by Engstrom and Morehead.123 Their data utilized the black body character- istic, the tungsten spectral emissivity, and the transmission of the lamp envelope. Tabulated data (ST-3340) are available from BURLE Application Engineering. Their data in the wavelength range 300 to 1200 nanometers has been adjusted to a tempera- ture of 2856 K color temperature and are shown in Table F-I. The data are normalized to unity at the maximum. The conversion was made by multiplying by the appropriate WAVELENGTH-NANOMETERS Planckian functions with a value for the se- 92cs-32471 cond radiation constant, C2 = 14,387.86 µm Fig. F-l - Relative spectra/emission charac- K . 124 teristic for a tungsten lamp at a color temper- ature of 2856 K. Tungsten-Halogen Lamps A variation of the tungsten lamp is the As a result of the work of industry com- tungsten-halogen lamp which is remarkable mittees, virtually the entire photosensitive- in that it can be operated at relatively high device industry in the United States uses the temperatures with increased life and with tungsten lamp at 2856 K color temperature* practically constant light output until the as a general test source. The lamp is cali- end of life. 125 Darkening of the envelope is brated in lumens and is utilized in the infra- virtually prevented in these lamps by a reac- red spectrum as well as the visible. Typical as tion of the halogen gas and the evaporated well as maximum and minimum photosensi- tungsten. In a typical application the lamp tivities are quoted in microamperes per may contain between 0.2 and 0.3 micro- lumen. moles/per cubic centimeter of I2 which is The principal disadvantages of using the broken down to atomic iodine by the heat of tungsten lamp as an industry standard test the filament. The atomic iodine reacts with are that it does not provide a direct measure the evaporated tungsten on the envelope wall of radiant sensitivity as a function of wave- to form WI,. The volatile WI2 diffuses to the length and that it is a somewhat misleading filament where it is decomposed, depositing term when the response of the photomulti- W on the filament and freeing I to repeat the plier lies outside the visible range. To assist cycle. the scientist in using photomultipliers, The temperature of the envelope wall must technical specifications for BURLE photomul- be in the range 250 to 1200°C for the iodine tiplier types include photocathode spectral- cycle to operate successfully. It is common, response curves which give the sensitivity in therefore, for the envelopes to be made of absolute terms such as amperes per watt and quartz. Because the envelopes are small, quantum efficiency as a function of wave- even for high power lamps, the wall attains the proper temperature. If the envelope wall *Formerly 2870 K, but changed to agree with C.I.E. is at too low a temperature, the WI2 will not designated Illuminant A at 2854 K, and again more be desorbed and a brown deposit of WI2 will recently to 2856 K because of the adoption of the inter- be formed. At too high a wall temperature, 122 national practical temperature scale of 1968. the reaction, WI2 - W + 2I will occur and

149 Photomultiplier Handbook

Table F-I Relative Spectral Irradiance from a Tungsten Test Lamp Operated at a Color Temperature of 2856 K Relative Relative Relative Wavelength, spectral Wavelength, spectral Wavelength, spectral nm Irradiance nm Irradiance nm Irradiance 300 .0004 610 .5179 910 .9916 310 .0017 620 .5449 920 .9945 320 .0044 630 .5717 930 .9965 330 .0078 640 .5984 940 .9977 340 .0117 650 .6249 950 .9990 350 .0164 660 .6500 960 1.0000 360 .0221 670 .6747 970 .9985 370 .0285 680 .6988 980 .9963 380 .0361 690 .7219 990 .9961 390 .0450 700 .7438 1000 .9934 .0551 710 .7649 1010 .9909 410 .0663 720 .7860 1020 .9865 420 .0789 730 .8054 1030 .9823 430 .0928 740 .8233 1040 .9787 440 .1082 750 .8402 1050 .9747 450 .1249 760 .8557 1060 .9692 460 .1428 770 .8719 1070 .9647 470 .1623 780 .8861 1080 .9582 480 .1830 790 .8993 1090 .9529 490 .2048 800 .9114 1100 9457 500 .2278 810 .9228 1110 .9412 510 .2517 820 .9346 1120 .9321 520 .2766 830 .9444 1130 .9257 530 .3025 840 .9545 1140 .9174 540 .3286 850 .9624 1150 .9111 550 .9692 1160 .9037 560 ..3826 870 .9751 1170 .8948 570 .4097 880 .9809 1180 .8868 580 .4368 890 .9862 1190 .8796 590 .4636 900 .9893 1200 .8704 600 .4906

= 680 lumens per watt, the luminous flux repre- sented by the tabular values is 2797 lumens. (There is some uncertainty about the value 680 lumens per watt; relative spectral luminous efficiency values. If the max- various slightly differing values have been reported in imum luminous efficacy at 555 nanometers is recent years.)

150 Radiant Energy and Sources the W will not be removed from the wall. Because of the higher operating tempera- tures and the long life with minimum envelope darkening, the tungsten-halogen lamps are useful as standard test lamps. 126 Robert Saunders, Jr., of the Optical Radiation Section of the National Bureau of Standards has developed an empirical for- mula* representing the spectral irradiance of 1000-watt quartz-halogen type DXW lamps. Saunders formula is 0’ 400 500 600 700 WAVELENGTH-NANOMETERS 92CS-32472 Fig. F-2 - Typical spectra/emission curve for a water-cooled mercury-arc lamp at a The fit of this formula to the actual data is pressure of 130 atmospheres. of the order of 0.1% at each wavelength in the range of 350 to 900 nanometers for each character of the light emitted from a mer- of four lamps tested. Rounded-off averages cury arc varies with pressure and operation of the constants in Eq. F-2 are A=0.867; conditions. At low pressure, the spectral out- put consists of sharp lines, which are very useful as reference spectra. Table F-III pro- pressed in nanometers. Ts is the apparent vides a list of some of the mercury lines in black-body temperature. The term ea repre- the visible and near-visible spectral range. In sents a magnitude. Saunder’s data were the case of germicidal lamps, most of the taken at a temperature, Ts , of approximate- energy radiated is in one spectral line, 253.7 ly 3025 K. nanometers. For various purposes it is useful to have a At increasing pressures, the spectral- tabulation of the relative radiant spectral energy distribution from the arc changes flux from a tungsten-halogen lamp operating from the typical mercury-line spectral char- at 3200 K color temperature. Saunder’s data acteristic to an almost continuous spectrum were used to extrapolate to this temperature. of high intensity in the near-infrared, visible,

A value of Ts = 3184 K was found to provide and ultraviolet regions. Fig. F-2 shows the a spectral distribution most closely fitting a spectral-energy distribution from a water- black body at 3200 K and thus providing a cooled mercury arc at a pressure of 130 at- color temperature of 3200 K. Using a value mospheres. 7 of 1.4388 x 10 nm K for the second radia- Deuterium Lamps tion constant, C2, the relative spectral ra- diant flux values from such a lamp were These lamps provide a continuous, line- determined from Eq. F-2 and are tabulated free spectrum in the ultraviolet range. The in Table F-II. lamps are useful in spectrophotometry and related applications. A typical relative spec- ARC AND GAS-DISCHARGE SOURCES. tral energy distribution from a deuterium Mercury Lamps lamp is shown in Fig. F-3. Deuterium lamps may also be obtained with calibrated spectral Of the various types of electrical discharge irradiance over the wavelength range of 180 that have been used as radiation sources, the to 400 nanometers. mercury arc is one of the most useful. The Concentrated-Arc Lamps *Private Communication A very useful point source is the zir- @More detailed information on arc and gas-discharge conium concentrated-arc lamp. Concen- sources may be found in Handbook of Optics, trated-arc lamps are available with ratings sponsored by the Optical Society of America, W. G. Driscoll and W. Vaughan, editors, McGraw-Hill, 1978. These lamps may be obtained from Cenco Company.

151 Photomultiplier Handbook

Table F-II Relative Spectral Irradiance from a Tungsten-Halogen Lamp in a Quartz Envelope Operated at a Color Temperature of 3200 K Relative Relative Relative Wavelength, Spectral Wavelength, Spectral Wavelength, Spectral nm Irradiance nm Irradiance nm Irradiance 300 .0109 610 .6714 910 .9969 310 .0151 620 .6966 920 .9950 320 .0204 630 .7211 930 .9924 330 .0269 .7444 940 .9893 .0347 650 .7669 950 .9859 350 .0440 660 .7883 960 .9820 360 .0548 670 .8086 970 .9778 370 .O671 680 .8280 980 .9733 380 .0811 690 .8464 990 .9683 390 .0967 700 .8636 1000 .9629 .1139 710 .8796 1010 .9572 410 .1327 720 .8945 1020 .9515 420 .1531 730 .9083 1030 .9454 430 .1749 740 .9213 1040 .9392 440 .1981 750 .9327 1050 .9327 450 .2224 760 .9434 1060 .9259 460 .2480 770 .9530 1070 .9194 470 .2745 780 .9618 1080 .9125 480 .3018 790 .9694 1090 .9052 490 .3299 800 .9763 1100 .8984 500 .3585 810 .9820 1110 .8911 510 .3875 820 .9870 1120 .8838 520 .4169 830 .9912 1130 .8766 530 .4463 840 .9943 1140 .8693 540 .4757 850 .9969 1150 .8621 550 .5048 860 .9985 1160 .8544 .5338 870 .9996 1170 .8472 570 .5625 880 1.0000 1180 .8399 580 .5908 890 .9996 1190 8326 590 .6183 900 .9985 1200 .8254 600 .6450

=680 lumens per watt, the luminous flux sent watts per 10-nanometer interval, the sum, represented by the tabular values is 3861 lumens. (There is some uncertainty about the value 680 lumens per relative spectral luminous efficiency values. If the max- watt; various slightly differing values have been imum luminous efficacy at 555 nanometers is reported in recent years.)

152 Radiant Energy and Sources

Table F-III CARBON-ARC LAMP Some of the Principal Spectral Lines 8 Characteristic of a Low-Pressure Mercury Discharge (in nanometers) 237.8 365.0 253.7 404.7 265.3 435.8 296.7 546.1 302.1 577.0 313.2 579.0 334.1 (A more complete table may be found in American Institute of Physics Handbook, Third edition, 1972, McGraw-Hill, pp 7-92 -7-96.) 92cs-32474 Fig. F-4 - Typical spectra/emission curve for a dc high-intensity carbon-arc lamp. from 2 to 300 watts, and in point diameters from 0.08 mm to 2.9 mm. These lamps re- quire one special circuit to provide a high starting voltage and another well-filtered and ballasted circuit for operation. Arc Lamps The carbon arc is a source of great inten- sity and high color temperature. A typical energy-distribution spectrum of a dc high- intensity arc is shown in Fig. F-4. Figs. F-5 and F-6 show relative spectral-emission characteristic curves for xenon and

I 1000 1200 WAVELENGTH -NANOMETERS 92cs-32475 DEUTERIUM LAMP Fig. F-5 - Typical spectralemission curve for a xenon-arc lamp.

n

WAVELENGTH -NANOMETERS 92cs-32473 WAVELENGTH-NANOMETERS 92CS-32476 Fig. F-3 - Typical relative spectral energy distribution from a deuterium lamp. (From Op- Fig. F-6 - Typical spectral-emission curve for tronic Laboratories, Inc., Silver Spring, Md.) an argon-arc lamp.

153 Photomultiplier Handbook

Fluorescent Lamps SUMMARY OF TYPICAL SOURCES The common fluorescent lamp, a very ef- Table F-IV provides typical parameters ficient light source, consists of an argon- for the most commonly used radiant energy mercury glow discharge in a glass envelope sources. internally coated with a phosphor that con- verts ultraviolet radiation from the discharge into useful light output. There are numerous LASERS AND LIGHT- types of fluorescent lamps, each with a dif- EMITTING DIODES ferent output spectral distribution depending In recent years the development of various upon the phosphor and gas filling. The spec- types of lasers and p-n light-emitting diodes tral response shown in Fig. F-7 is a typical with very high modulation frequencies and curve for a fluorescent lamp of the daylight short rise times has increased the types of type. sources that photomultipliers are called upon to detect. Although many of these in- teresting devices have their principal wave- lengths of emission in the infrared beyond the sensitivity range of photomultiplier tubes, some do not. Because of the growing importance of laser applications and the use of photomultipliers for detecting their radia- WAVELENGTH-NANOMETERS tion, Tables F-V through F-IX are provided 92cs-32477 as reference data on crystalline, gas, and li- Fig. F-7 - Typical spectra/emission curve for quid lasers,and on p-n junction light- a daylight-type fluorescent lamp. emitting diodes.

Table F-IV Summary of Typical Sources/Parameters for the Most Commonly Used Radiant Energy Sources Lamp Type DC Arc Luminous Luminous Average Input Dimensions Flux Efficiency Luminance Power (mm) (lm) (lm W-1) (cd mm- 2) (watts) Mercury Short Arc (high pressure) 200 2.5 x 1.8 9500 47.5 250 Xenon Short Arc 150 1.3 x 1.0 3200 21 300 Xenon Short Arc 20,000 12.5 x6 1,150,000 57 3000 (in3 mmx6mm) Zirconium Arc 100 1.5 250 2.5 100 (diam.) Vortex-Stabilized Argon Arc 3x10 422,000 17 1400 Tungsten - 79 7.9 10 Light - 1630 16.3 Bulbs - 21,500 21.5 25 Fluorescent Lamp Standard Warm White 40 - 2,560 64 - Carbon Arc Non-Rotating 2,000 =5x5 36,800 18.4 175 Rotating 15,800 =8x8 350,000 22.2 800 Deuterium Lamp 40 1.0 (Nominal irradiance at 250 nm at (diam.)

154 Radiant Energy and Sources

Table F-V Typical Characteristics of a Number of Useful Crystal Laser Systems Host Dopant Wavelength Mode and Highest of Laser Temperature of . Operation (K)

Al2O3 0.05% 0.6934 CW,P 350 Cr3 + 0.6929 P 300

Al2O3 0.5% 0.7009 P 77 Cr3 + 0.7041 P 77 0.7670 P 300

MgF2 1% 1.6220 P 77 Ni2 +

MgF2 1% 1.7500 P 77 Co2 + 1.8030 P 77 ZnF2 1% 2.6113 P 77 Co2+ CaWO4 1% 1.0580 CW 300 Nd3 + 0.9145 P 77 1.3392 P 300 CaF2 1% 1.0460 P 77 Nd3 +

CaMoO4 1.8% 1.0610 CW 300 Nd3 + 3 Y3Al5O12 Nd + 1.0648 CW 360 P 440

LaF3 1% 1.0633 P 300 Nd3 + LaF3 1% 0.5985 P 77 Pr3 +

CaWO4 0.5% 1.0468 P 77 Pr3 +

5% 0.6113 P 220 Y2033 Eu3 + 3 CaF2 Ho + 0.5512 P 77 CaWO4 0.5% 2.0460 P 77 Ho3 + 3 Y3Al5O12 Ho + 2.0975 CW 77 P 300 CaWO4 1% 1.6120 P 77 Er3 + 3 Ca(NbO3)2 Er + 1.6100 P 77 3 Y3Al5O12 Er + 1.6602 P 77 3 CaWO4 Tm + 1.9110 P 77 3 Y3Al5O12 Tm + 2.0132 CW 77 P 300 Tm3+ 1.9340 CW 77 Yb3+ 1.0296 P 77

CW = Continuous P = Pulse

155 Photomultiplier Handbook

Table F-V Typical Characteristics of a Number of Useful Crystal Laser Systems (Cont’d) Host Dopant Wavelength Mode and Highest of Laser Temperature of Operation (K) CaF2 0.05 % 2.6130 P 300 U3+ CW 77 3 SrF2 U + 2.4070 P 90 CaF2 0.01 % 0.7083 P 20 Sm2+ SrF2 0.01% 0.6969 P 4.2 Sm2+ CaF2 0.01% 2.3588 CW 77 Dy2 + P 145 CaF2 0.01% 1.1160 P 27 Tm2+ CW 4.2

CW = Continuous P = Pulse

Table F-VI Comparison of Characteristics of Continuous Crystalline Lasers Material Sensitizer Optical Pump Wave- Power Operating Active System Length Eff.(%) (Watts) Temp.(K)

- W 2.36 - 0.06 1.2 77 - Hg 0.69 0.1 1.0 300 - W 1.06 0.2 2 300 1.06 0.6 15 300 Nd3+Y Al 0 - Plasma Arc 1.06 0.2 200 300 3 3 5 12 Nd +Y Al 0 Cr3+ Na Doped Hg 1.06 0.4 0.5 300 3 3 5 12 Nd +Y 3Al 5 012 Hg 10 300

Table F-VII Typical Characteristics for Two Liquid Lasers Liquid Principal Pulse Energy(J) Wavelength (Pulse Width)

Eu(O-ClBTFA)4 DMA* 0.61175 0.1 J 3 Nd + *:SeOC 3 l2** 1.056 +dimethylammonium salt of tetrakis -ortho-chloro-benzoyltrifluoracetonate. **trivalent in selenium oxychloride.

156 Radiant Energy and Sources

Table F-VIII Typical Characteristics of a Number of Useful Gas Lasers Gas Principal Output Power Wavelengths Pulsed or Typical Maximum Continuous Ne 0.3324 - 10 mW Pulsed (ionized) 10 mW CW Ne 0.5401 - 1 kW Pulsed (unionized) He-Ne 0.5944-0.6143 - Pulsed (unionized) 0.6328 1 mW 150 mW 1.1523 l-5 mW 25 mW CW 3.3913 <1 mW 10 mW CW Xe 0.4603-0.6271 5 mW - Pulsed (ionized) 0.5419-0.6271 10 mW 1 W CW 0.4965 - 0.5971 1 mW 1 W Pulsed Xe 2.026 1 mW 10 mW CW (unionized) 3.507 - 1 mW CW 5.575 0.5 mW 5 mW CW 9.007 0.5 mW 5 mW CW A 0.4880 0.5 W 5W CW (ionized) 0.5145 0.5 W 5 W CW 0.4545 to 1.5 W 40 W CW 0.5287 0.33 - 200 kW Pulsed N2 (ionized - 100 mW CW Kr 0.3507 - 300 mW CW (ionized) 0.5208-0.6871 - 3 W CW 0.5682 - 100 mW CW 10.552 CO2 (molecular 10.572 50 W 3.2 kW CW excitation) 10.592

CF3I 1.315 - 10 kW Pulsed 27.9 - 1.2 W Pulsed H2O (molecular 118 - 1 mW Pulsed excitation) 118 - CW CN 337 - 50 mW Pulsed (molecular excitation)

157 Photomultiplier Handbook

Table F-IX Typical Characteristics of p-n Junction Light-Emitting Diodes crystal Wavelength Laser Action

PbSe 8.5 PbTe 6.5 Yes InSb 5.2 Yes PbS 4.3 Yes 3.15 Yes 0.85-3.15 Yes

In PxAs1-x) 0.91-3.15 Yes GaSb 1.6 No InP 0.91 Yes GaAs 0.90 Yes 0.80-0.90 Yes 0.55-0.90 Yes CdTe 0.855 No (ZnxCd l-x)Te 0.59-0.83 No CdTe-ZnTe 0.56-0.66 No BP 0.64 No Cu2Se-ZnSe 0.40-0.63 No Zn(SexTe1-x) 0.627 No ZnTe 0.62 No GaP 0.565 No GaP 0.68 No SiC 0.456 ?

LIGHT SOURCES FOR TESTING it is relatively simple, stable, and inexpen- Monochromatic sources of many wave- sive, and maintains its calibration. The lengths may be produced by narrow-band tungsten lamp emits a broad band of energy filters or monochromators. Narrow-band with relatively smooth transitions from one filters are more practical for production end of the spectrum to the other. Its prin- testing, but, at best, such tests are time- cipal disadvantage as a general source is its consuming and subject to error. Monochro- lack of ultraviolet output and relatively low matic sources are not used in general- blue output. purpose testing because most applications Sources such as arcs and glow discharges involve broader-band light sources; a are difficult to calibrate and show serious monochromatic test might grossly misrepre- time variations. sent the situation because of spectral- Filters are frequently used to narrow the response variations. spectral range for specific purposes; how- A broad-band source is probably more ever, they sometimes contribute to errors useful as a single test because it tends to in- because of significant transmission outside tegrate out irregularities in the spectral- the band of interest. Filters are also subject response characteristic and more nearly rep- to change in transmission with time and are resents the typical application. The tungsten very difficult to reproduce with identical lamp has been used for many years because characteristics.

158 Radiant Energy and Sources

In many applications it is appropriate to REFERENCES test photomultipliers in the same manner in 121. J.C. DeVos, “A new determination of which they are to be utilized in the final ap- the emissivity of tungsten ribbon,” Physica, plication. For example, photomultipliers to Vol. 20, pp 107-131, (1954). be used in scintillation counting may be 122. “Color temperature, luminous efficacy tested by means of an NaI:Tl crystal and a and the international practical temperature 137Cs source or a simulated NaI light source scale of 1968,”National Bureau of Stan- utilizing a tungsten lamp whose light passes dards Technical News Bulletin, Vol. 54, No. through a one-half stock-thickness Corning 9, pp 206-7, Sept. 1970. CS5-58 filter (5113 glass). Interference-type 123. R.W. Engstrom and A.L. Morehead, filters are becoming increasingly important “Standard test-lamp temperature for in isolating specific wavelengths for testing photosensitive devices-relationship of ab- photomultiplier tubes for laser applications. solute and luminous sensitivities,” RCA When a photomultiplier is manufactured Rev., Vol. 28, pp 419-423 (1967). for a variety of purposes, including scientific 124. E.R. Cohen, CODATA Bulletin 11, applications, it would be highly desirable if Dec. 1973. See also D.G. Fink and H.W. sensitivity were specified by a complete spec- Beaty, Standard Handbook for Electrical tral response in terms of quantum efficiency Engineers, 11th Edition, McGraw-Hill, or radiant sensitivity. This information 1978. could then be utilized with the known spec- 125. E.G. Zubler and F.A. Mosby, “An tral emission of any source to compute the iodine incandescent lamp with virtually 100 response of the photomultiplier to that per cent lumen maintenance,” Illuminating source. Complete spectral sensitivity data, Engineering, Vol. 54, No. 12, pp 734-740, however, are rarely provided because it is Dec. 1959. unnecessary for most practical situations 126. R. Stair, W.E. Schneider and J.K. and would considerably increase device Jackson, “A new standard of spectral irra- costs. diance,”Appl. Opt., Vol. 2, p 1151, (1963).

159 Photomultiplier Handbook

Appendix G - Statistical Theory of Noise in Photomultiplier Tubes

Generating Functions In this treatment of noise and signal-to- noise ratio in photomultiplier operation, generating functions 127 will be used to devel- The variance is defined as op the statistics of the different processes and their combinations. This approach has been selected because it provides a general, straightforward, and relatively simple method of solving problems of this nature. Those not acquainted with the use of generating functions may find it worthwhile to investigate some of the sources in reference127. A short summary of generating functions and their use is provided in this section. Others may use this Appendix for the summarized expressions relating to photomultiplier statistics. A generating function may be defined as

Substituting G-3 and G-8 in G-6 establishes the identity of G-6 and G-7, and thus of G-6 and G-5.

Additive Events Now, suppose there are two independent events whose scores are to be added. An ex- ample is the roll of two dice, or in the case of a photomultiplier, the photocathode current generated thermally and photoelectrically. Assume two corresponding generating func- tions: QA(s) and QB(s). It may be shown that the generating function for the sum is and

160 Statistical Theory of Noise in Photomultiplier Tubes

PHOTON NOISE

The case of subtraction is just a modifica- tion of the addition. In this case G-13 and, as before, G-13a Cascade Events Consider a pair of independent sequential events such as the emission of photoelec- . G-22 trons and the multiplication that occurs at the first dynode of a photomultiplier. Each photoelectron is multiplied by a secondary emission factor that has an average value and a variance. Assume two generating func- tions: A(s) for the primary event, and B(s) for the secondary or multiplying event. It may then be shown that the generating func- tion for the cascaded event is G-25 G-14 Thus;the signal-to-noise ratio of the photon Q AB(s) = A[B(s)] There are a few cases in which the fluctua- tion in the photon flux has an additional term. 128 In broad-band thermal sources, however, Eq. G-24 yields the proper expres- sion for the noise in the input photon flux.

In the case of the cascaded event relating PHOTOEMISSION NOISE to the photomultiplier, each electron from the preceding stage is acted on independently The physics of photoemission and by the secondary emission gain. In contrast, photocathodes is discussed in Chapter 2 on consider the case of multiplication of two in- Photomultiplier Design. In the following dependent scores: i.e., a die is rolled and the discussion it is assumed that the time be- score noted, then rolled again and scores tween the absorption of a photon and the subsequent emission of an electron, when multiplied. The order of the events is not 12 significant. The generating function for one emission occurs, is short (about 10- event may be second). In addition, it is assumed that all the statistical processes of absorption, elec- G-17 tron transport within the photocathode, and photoemission can be described and characterized by one number, the quantum G-18 function of the photon wavelength.

161 Photomultiplier Handbook

A simple model of photoemission will aid in understanding the noise contributions of G-30 the photocathode. For each photon that strikes the photocathode, an electron is Thus, for quantum efficiencies less than unity, the statistics of the photoemission reflection effects at the various cathode in- process result in a finite signal-to-noise ratio for the photoelectron number even though The chance that no electron is emitted is there is no assumed noise in the photon stream. In practice, the signal-to-noise ratio of the input photon flux is never infinite; the flux always contains some noise. In most sources, the signal-to-noise ratio of the photon flux is given by Eq. G-25.

The photoelectron signal-to-noise ratio is G-33 A quantum efficiency of 40 per cent reduces the signal-to-noise ratio of the photoelectron flux to about 63 per cent of that of the photon flux. It is important to realize that the degrada- being less than unity is irreversible in that no amount of noise-free amplification can im- prove the photoelectron signal-to-noise ratio. In some applications, multiphoton pulses form the input signal. In these applications, integral numbers of photoelectrons are emitted from the photocathode within a time

162 Statistical Theory of Noise In Photomultiplier Tubes that is short with respect to the resolution ber of ways such an output can occur among time of the photomultiplier. Examples of the m-photon inputs. this type of input can be found in radioactive tracer scintillations, such as those observed in tritium and carbon spectroscopy.

92cs-32483 G-l - Photoelectron output pulse spec- trum resulting from a flux of I-photon input

A

Fig. G-2 - Photoelectron output pulses resulting from a flux of 4-photon input pulses or

is no photoemission only 13 per cent of the time; most of the photoemission is divided between 1- and (2-, 3-, of 4-) photoelectron Eq. G-39 is just the coefficient of sr in the ex- pulses in the ratio of 13 to 20. In sharp con- occur 96 per cent of the time. Of the times the chance of (m-r) failures to photoemit when photoelectrons are emitted, single- and the binomial coeffi- electron pulses occur 70 times more often than any others; 3- and 4-photoelectron pulses almost never occur. It should be clear that in multiphoton-pulse spectroscopy it is

163 Photomultiplier Handbook

The signal-to-noise ratio expected in the For a given primary energy, it is possible photoelectron pulse distribution generated to obtain any number of secondaries ns from from an m-photon input depends upon the zero to a maximum ns(max). The maximum quantum efficiency. From Eqs. G-34 and number is given by the quotient of the G-35, the relation primary energy EP and the energy required to produce a hole-electron pair within the G-40 G-42 Over many repeated measurements using primary electrons of the same energy, a trun- THERMAL EMISSION ADDED TO cated distribution is obtained for nS. A PHOTOEMISSION NOISE model that describes the observed distribu- Another source of photocathode noise is tions from most practical dynodes follows. the thermionic emission of single electrons. The observed number distributions for The strength of the emission varies with secondary electrons vary among the dif- photocathode type. In the bialkali cathode, ferent types of dynodes used commercially. for example, thermionic emission is virtually Nearly all the distributions fall within the absent; on the other hand, Ag-O-Cs class limited by a Poisson distribution130 at photocathodes exhibit relatively large dark one extreme and an exponential distribution currents as a result of thermionic emission. at the other.131 To describe this wide variety These currents can be eliminated to some ex- of distributions the Polya, or compound tent by cooling the tube. Poisson, distribution is employed.132 Randomly emitted thermionic electrons Through the adjustment of one parameter, add a term to the fluctuation in the the distribution runs from purely Poisson to photoelectron current proportional to their exponential. Therefore, this one distribution number. Because of their independence with can be used to describe and to interpret the respect to any usual signal current, the term bulk of the observed secondary-emission adds to that of the signal noise in statistics. quadrature. (See Eq. G-l1.) That is, The distribution has the following form:

STATISTICS RELATED TO where P(n,b) is the probability of observing SECONDARY EMISSION distribution, and b is the parameter con- Although the amplification (multiplica- trolling the shape of the distribution. With tion) of the photoelectron current in the dynode chain of a photomultiplier is often b = 0, the distribution is Poisson, as follows: referred to as noise-free, careful examina- tion of the statistics of the gain mechanisms G-44 involved shows that this statement is not en- 1 distribution tirely correct. Approximate noise-free opera- tion can be attained, however, with the use of proper electronoptics and newly G-45 developed dynode materials. In the fol- Fig. G.3 shows a family of distributions for lowing discussion the statistical gain pro- various values of b. cesses in the individual dynodes are ex- The generating function for P(n,b) is amined and then combined to yield the given by statistical properties of the entire multiplier chain. Much of the work presented was ac- complished at a very early stage in photomultiplier history. 129 G-47

164 Statistical Theory of Noise In Photomultiplier Tubes

Fig. G-4 - A comparison of the SNR as a Polya distributions. For Poisson statistics (b = 0), the SNR increases as the square root

square root of b - for large mean gains.

Poisson statistics significantly reduce the signal-to-noise ratio at moderately high gains of 10 to 20. It can be anticipated that departures from Poisson statistics degrade the single-electron pulse-height resolution. The Polya distribution has an interesting interpretation with respect to secondary- emission statistics. 132 For non-zero values of b, the distributions described by Eq. G-43 can be shown to be composed of a number of different Poisson processes, each with a different mean value. The mean values are, in turn, distributed according to the Laplace distribution. When b equals 0, the distribu- Fig. G-3 - Single-particle output distribution tion of the mean values collapses to a delta for a dynode displaying Polya statistics. A value of b = gives an exponential distribu- Poisson distribution. For b equals 1, the tion; b = 0 gives a Poisson distribution; b = 0.2 distribution of the mean values is exponen- is intermediate between the two extreme tial. The physical interpretation for a dynode values. displaying non-Poisson statistics is that physical non-uniformities on the dynode sur- Fig. G-4 shows a log-log plot of the signal- face cause each element of the surface to have a different mean value for emission. parameter. The signal-to-noise ratio im- Although each small element exhibits Poisson statistics with respect to emission, distribution, but approaches unity with large the total emission from the entire dynode is non-Poisson because it comprises a distribu- any non-zero value of b, the signal-to-noise tion of Poisson distributions. It is possible that the basic emission process from a given in Fig. G-4, even small departures from dynode is not a Poisson process. However,

165 high-gain GaP dynodes exhibit nearly Statistics for a Series of Dynodes 130 Poisson statistics and at present it is Assume that one primary electron im- believed that departures from this norm are pinging on the first dynode releases, on the caused, to a large extent, by dynode non- uniformities. The output of the first dynode striking the The departure from Poisson statistics af- second dynode produces an average gain at fects the single-particle pulse-height resolu- tion. Fig. G-5 shows the output-pulse distribution from a single dynode for a events, the average gain and its variance may number of multiple-particle inputs. The be related to the individual dynode statistics resolution is clearly degraded in passing as follows: from a Poisson distribution to an exponen- G-50 tial one. With the exponential distribution shown in Fig. G-5, it would be difficult to and distinguish among one-, three-, and five- 2 G-51 particle input pulses, whereas the problem virtually disappears for a Poisson distribu- emission and variance for the second dynode for a single-input electron. Continuing in this manner, the gain and fluctuation from the third stage are given by G-52 and 2 G-53 Eq. G-53 can be rearranged to read

Eq. G-55 states the expected results: that the total average gain for a series of k dynodes is the product of the secondary- emission yields of the individual dynodes in the series. Eq. G-56 shows that the relative contribution of any state to the total fluctua- 0 tion decreases with the proximity of the dynodes to the output end of the chain. The 92CS-32478 first stage contributes most to the total Fig. G-5 - A comparison of multiple-particle variance. The higher the first-stage gain, the distributions from a single dynode having less each subsequent stage contributes to the Poisson and exponential distribution. Clearly, total variance. This property is an important the particle resolution characteristics of the feature of the high-gain GaP first-dynode exponential distribution are much poorer. photomultipliers.

166 Statistical Theory of Noise in Photomultiplier Tubes

The signal-to-noise ratio for the multiplier evolves toward a steady-state distribution chain is given by after four or five stages, and exhibits little change thereafter. Fig. G-6 shows some single-electron distribution132 computed by use of the Polya statistics for each stage in the chain as explained above. As b ap- proaches 1, the distribution becomes more sharply peaked.

For large first-stage gains, the multiplier signal-to-noise ratio is high. Most of the noise contribution is from the first stage. If, in addition to a large gain, the first stage ex- hibits Poisson statistics, as explained above, the signal-to-noise ratio becomes

G-58

0 I 2 3 4 The noise added to the input signal is very PULSE HEIGHT small. It is in this sense that the multipli- cation chain is said to provide noise-free 92CS-32479 gain. Fig. G-6 - Computed singleelectron distribu- tion for a range of values of parameter b. Multiple-Particle Inputs Parameter b is defined in Appendix B. The output-pulse distribution for multiple-particle inputs is obtained from the generating function for the multiplier chain. Computer values of multiple-particle out- puts for a two-stage structure are shown in By an extension of Eq. G-14 for the 131 generating function for a cascaded event, the Fig. G-7. The two curves relate to two dif- generating functions for a chain of k dy- ferent structures that have the same dynode nodes may be obtained: as a first stage. The solid-line curve shows the output when the first stage is followed by G-59

The probability Pk(n) of observing n elec- nearly pure Poisson statistics (b = 0.01). The trons (for a one-electron input) at the output output peaks are sharply defined, and pulses of a k-stage chain is derived from Eq. G-4 as up to a ten-electron input pulse are clearly follows: resolvable. The dashed line describes the final output distribution when the first stage

The generating function for a k-stage multiplier chain for multiple-particle inputs is given by extension of Eq. G-9 as follows: m Qk(s,m)=[Qk(s)] G-61 The probability of observing n output elec- trons from m input particles is therefore given by

NUMBER OF SECONDARY ELECTRONS IN A PULSE FROM DYNODE SURFACE

92CS-32480 When identical dynodes are used, the output Fig. G-7 - Theoretical pulse-height distribu- distribution for single-electron input pulses tion,

167 Photomultiplier Handbook

having an exponential output distribution. Individual peaks are no longer discernible; the large variance associated with the ex- ponential statistics of the second stage eliminates all the structure in the output of the first stage. BURLE has developed a high-gain gallium phosphide dynode which, when used as the first stage in a conventional copper beryllium multiplier chain, greatly increases the pulse-height resolution of the photomultiplier. The high-gain first stage in a photomultiplier having multiple photoelec- 92CS-32482 tron events originating from the photo- Fig. G-Q - A comparison of the tritium scin- cathode is similar to the case illustrated in tillation pulse-height spectra obtained using Fig. G-7 for a multiplier where the high-gain a conventional photomultiplier having ail second dynode amplifies the multiple pulses CuBe dynodes and a photomultiplier having a originating from the first dynode which in GaP first stage. Note: The first photoelectron turn are initiated by single electrons. Typi- peak of the 8850 spectra includes dark noise from the photomultiplier, chemiluminescence cal gains for the gallium phosphide dyn- and phosphorescence from the vial and odes are 30 to 45, and their statistics are cocktail, as well as 3H disintegration. nearly Poisson. Fig. G-8 shows the multiple- particle pulse-height distribution for the tube, and Fig. G-9 shows the pulse-height FLUCTUATIONS IN THE TUBE curves for a tritium scintillation input for a AS A WHOLE conventional tube using the standard In the previous sections the noise con- copper-beryllium first dynode along with tributions from the photocathode and the that for a tube with a gallium phosphide first multiplier chain were considered. These dynode. The increased resolution of the results can be combined to obtain the signal- gallium phosphide dynode is clearly shown. to-noise ratio for the photomultiplier as a Note: The first photoelectron peak of the whole. 8850 spectra includes dark noise from The average number of photoelectrons the photomultiplier, chemilumi- nescence and phosphorescence from G-63 the vial and cocktail, as well as H3 disintegration. The variance is given by 2 G-64

flux of photons displays Poisson statistics. If

process, Eq. G-l6 must be used to obtain 2

RESOLUTION = 40% Using these expressions to describe the in- put to the photomultiplier chain, the average number of electrons collected at the anode can be stated as follows: G-65 0 I 2 3 4 5 6 PULSE HEIGHT-PHOTOELECTRON EQUIVALENTS

92cs-32481 average gain of a k-stage multiplier. The variance for the output electron stream is Fig. G-8 - Typical photoelectron pulse-height given by spectrum for a photomultiplier having a GaP first dynode.

168 Statistical Theory of Noise in Photomultiplier Tubes

variance in the average gain of a k-stage total anode fluctuation then becomes multiplier chain. Eq. G-66 can be rearranged as follows: G-67 For large dynode gains, For equal-gain stages described by Poisson G-76 statistics in the multiplier chain, Eq. G-56 becomes Even for large dynode gains, exponential

G-68 drop in SNR, is accompanied by a severe loss in single- and multiple-electron pulse-height resolution, as shown in Fig. G-7. To resolve single-photoelectron pulses, the multiplier chain must exhibit both high gain and good G-69 (i.e., Poisson) statistics. A significant improvement in SNR, results when the photocathode quantum efficiency In this case

G-70 0.7 would improve SNRa by a factor of 1.4. would have an SNR, equal to the square root much less than 1 and hence that G-7 1 the ideal SNRa by a factor of 0.59. Con- siderable improvement can be expected with The signal-to-noise ratio at the anode is the development of photocathode materials given by of increased sensitivity. The application of these equations to pre- G-72 sent photomultipliers indicates that the available photocathode quantum efficiency For high-gain dynodes exhibiting Poisson is the principal degrading influence on statistics, therefore, SNRa is essentially that of the photoelectrons, SNRp.e., as given in the multiple-photon input-pulse resolution Eq. G-33. In a photomultiplier in which the dynode greater than 6) do not significantly degrade gain is not high but still exhibits Poisson the input signal-to-noise ratio of photoelec- statistics, SNRa is given by trons provided the dynode statistics are nearly Poisson. G-73 OTHER SOURCES OF NOISE IN PHOTOMULTIPLIER TUBES Within a photomultiplier there are sources decreased by a factor of 0.87 from its value of noise that are not associated directly with the processes of photoelectric conversion changes the degradation factor to 0.94. Fur- and electron multiplication. These sources can, in general, be separated into two very much. groups: (1) those that are not correlated In the case of fully exponential dynode with and (2) those that are correlated with statistics, the variance for each dynode in a the signal pulse. chain of identical dynodes is given by Eq. Non-Correlated Noise Sources G-48; i.e., The materials used to fabricate the in- G-74 ternal structure and the glass envelope of a

169 Photomultiplier Handbook photomultiplier may contain amounts of first dynode. The positive ion thus created certain radioactive elements that decay and travels backward to the cathode where it give off gamma rays or other high-energy may release one or more electrons from the particles. If one of these emitted particles photocathode. Because there is a time delay strikes the photocathode or one of the first in the ion-emitted electron pulse equal to the few dynodes, it will produce an anode dark time of flight of the ion to the photocathode, pulse. The size of the anode pulse may be the resulting pulse, usually referred to as an equivalent to one or more photoelectrons afterpulse, occurs after the true signal pulse emitted at the photocathode. These pulses at the anode. Afterpulses are caused mainly are randomly emitted. In tube manufacture by hydrogen ions and their occurrence can this type of emission is minimized through be minimized in tube processing. the careful selection of materials. Primary electrons produce photons as well Analytically, the fluctuations resulting as secondary electrons within the dynodes of from non-correlated random sources add in the multiplier chain. Despite the low effi- quadrature to those of the signal. Because ciency of this process, some of the emitted they are random, the dark-pulse variances photons may eventually reach the photo- - cathode and release additional electrons. A time delay is observed corresponding to the transit time for the regenerated electron pulse to reach the point of origin of the light. Depending upon the type of dynode The total anode variances are given by: multiplier cage, this time may be of the order of 20 nanoseconds. Most of the photons comprising the light feedback originate in the region of the last few dynodes or of the anode. If the voltage across the tube is increased, the dark-pulse rate also increases and usually are the average single- produces some observable light near the photoelectron and the average single-dark- anode region, a fraction of which is fed back emitted electron numbers observed in a time to the photocathode. The result of this positive feedback. is that, at a certain anode signal-to-noise ratio. Again, in those voltage,the photomultiplier becomes instances where the incoming radiation unstable and allows the output dark-pulse signal comprises more than one photon, rate to increase to an intolerably high level. coincidence techniques can be employed to The voltage at which this increase occurs is reduce the effect of randomly emitted elec- generally above the recommended maximum trons originating at the cathode. operating voltage. Electrons may originate at dynodes well Not every signal pulse initiates an after- along in the multiplier chain. At the anode, pulse. Therefore,coincidence techniques, these electrons appear as fractional- using more than one photomultiplier, can be photoelectron pulses. Such pulses also result employed in some instances to eliminate this from interstage skipping, generally near the source of noise as well as the uncorrelated beginning of the chain. With good statistics sources discussed above. in the chain, the fractional pulses may be If correlated noise sources cannot be discriminated against because the single- eliminated by time discrimination or other electron peak in the pulse-height spectrum means, an analytical treatment for the total stands out sharply. variance would follow the form given by Eq. Correlated Noise Sources G-12. A gas atom or molecule within the photomultiplier may be ionized by a NOISE AND THE BANDWIDTH photoelectron pulse, This ionization may OF THE OBSERVATION occur at the first-dynode surface or in the At high counting rates, noise calculations vacuum between the photocathode and the are performed with the average and variance

170 Statistical Theory of Noise in Photomultiplier Tubes of the photoelectron current rather than with individual photoelectron pulses. The expres- sions which have been developed for signal- to-noise ratio by consideration of the where i is the photocathode emission current in amperes and e is the charge of the elec- variance from average may be converted to tron. If the signal current is considered as i, expressions of signal-to-noise ratio involving the noise in the bandwidth B for the currents and bandwidth, B, by considering photocurrent is the familiar shot noise the reciprocal nature of the observation time formula (2eiB)1/2. and the bandwidth. The equation for the photon noise squared Noise equivalent bandwidth B may be is given by defined133 as follows: G-86 G-79 The photocurrent noise squared is given by transfer response of the circuit, and Am is Both these equations involve the photon "circuit” in this case counts pulses in a time “current” I,. If the photons are not randomly emitted, Laplace transform of the impulse response. Eq. G-86 must be modified. In the case in The impulse response is a rectangular pulse which the variance in the photon current is case,

For a randomly emitted photon, the noise and Am is found to be equal to 7. The noise current squared at the anode is given by equivalent bandwidth is then readily found to be G-81 where the multiplier chain is assumed to be It is of interest to compare this relation with composed of k high-gain dynodes exhibiting the equivalent noise bandwidth for exponen- Poisson statistics, each with an average gain tial impulse response as in an RC circuit; in this case At the anode, the input resistance and ca- pacitance of a preamplifier generate a noise B=1/4RC G-82 current squared given by

From Eq. G-25, the SNR for a random photon flux is given by G-83 where I, is the average photon arrival rate and 7 is the time interval of the count. When where g = l/R is the shunt conductance in the anode lead, C is the shunt capacitance, B is the bandwidth, k is Boltzmann’s constant,

T is the absolute temperature, and Rn is the equivalent noise resistance of the preamplifi- In the case of the photocathode electron cur- er input. The total noise current squared rent, Photomultiplier Handbook

statistics of the measurement are improved by increasing the count time. The present discussion recommends the optimum time division between the dark and the light counts.

G-93 G-94 The variances for the light and dark mea- surements are given by The value of I, shown in Eq. G-92 is the lower limit for the average photon current G-95 and makes the squared anode-current fluc- tuation greater than the squared noise cur- G-96 rent in the photomultiplier preamplifier in- The final “signal” which is a measure of I, put by a factor of ten. The resulting value of is given by the average photon current corresponds to a photoelectric current of about 10-14 am- pere, or about 6 x 104 photoelectrons per G-97 second. In cases in which the dark current is effectively higher than 6 x 104 photoelec- Assume that the times of the counts are ac- trons per second, the photomultiplier sen- curately determined so that the respective sitivity limit is set by the dark current and not by the preamplifier noise. It is the noise- use of Eq. G-21 for multiplication (by the free gain of the multiplier chain which in- reciprocal of the count times), the variance creases the rms photocurrent shot noise by a of each term in G-97 is then operation. See also Fig. 65 in Chapter 4, G-98 Photomultiplier Characteristics.

PULSE COUNTING STATISTICS* G-99 In utilizing a photomultiplier in a “photon” counting mode, individual anode For the variance of a difference, using Eq. pulses initiated by electrons from the G-11, photocathode are counted. Because the count at very low light levels originates from G-100 both photoemission and thermionic emis- sion, two separate counts are required: one in the light and one in the dark. In deter- The signal-to-noise ratio in the determina- mining the photoelectron count rate, the tion of I, is then given by

*Measured signal-to-noise ratios are similar for pulse- counting or for current -measurement techniques.134 Baum135 has shown that in some cases, pulse counting G-101 can have an advantage of the equivalent of a factor of 1.2 in quantum efficiency.

172 Statistical Theory of Noise in Photomultiplier Tubes

Let the average number of photons exiting to be minimized with respect to the division from the scintillator onto the photocathode per photoelectrically converted gamma-ray nd from G-93 and G-94: average quantum efficiency of the photo-

emission process is as before (Eq. G-28): G-102 G-104 sidering the other quantities as invariant. Now, using G-15 and G-16 for a cascaded event, the average number of photoelectrons imum SNR condition: per pulse is G-105

G-103 and the variance in the numberofphotoelec- trons per pulse is The implication of Eq. G-103 is as follows: For example, if the signal count is much less than the dark count, about equal times should be spent for both readings. If It is instructive to consider the signal-to- the signal count and dark count are about noise ratio at each stage of the photo- multiplier. For the photocurrent in the pulse, using G-105 and G-106:

PULSE-HEIGHT RESOLUTION IN SCINTILLATION COUNTING In scintillation counting, for the case of a For the current pulse leaving the first K2CsSb photocathode and a NaI:Tl scin- tillator, the typical photoemission yield is dynode, assuming a secondary emission approximately 8 photoelectrons per keV of gamma-ray energy absorbed in the crystal- statistics), again using G-15 and G-16: or 125 eV per photoelectron. The peak of the emission spectrum for NaI:Tl is approx- imately at 415 nm which corresponds to a Variance in this number = photon energy of 3 eV. The quantum effi- ciency of the bialkali photocathode for the blue spectrum of the scintillator is approx- imately 25 per cent. If the crystal were 100 SNR (pulse out of first dynode) = per cent efficient in converting gamma-ray energy to light energy, one would expect a photoelectron for every 12 eV. Thus, the G-108 conversion efficiency of the crystal is about 10 per cent. The statistics of the pulse-height distribu- tion, therefore, involve both the crystal and Similarly, for the pulse out of the second the photomultiplier processes. On the other dynode: hand, if the conversion efficiency of the crystal were 100 per cent, the photoelectric conversion of the gamma-ray would yield Variance in this number = essentially a constant number of photons per conversion and the crystal would then not contribute to the statistical process. Photomultiplier Handbook

SNR (pulse out of second dynode) = SUMMARY The more important expressions relating to signal and noise discussed in this Appen- dix are summarized below. Beginning at the input to the photomultiplier, the signal is followed through the tube, and the variance and the SNR associated with the signal are noted. If a tube having k stages of equal secon- Photon flux: Conditions: random emis- sion, Poisson statistics. Terminology: av-

G-l10

In pulse-height resolution measurements it is common to refer to the Full-Width-Half- Maximum (FWHM) which may be related to the SR as follows: FWHM =2.355/SNR Dark emission from a photocathode: Ter- minology: average number of thermionic

G-111

The first term in the brackets represents the relative variance of the photomultiplier con- Combined dark and photoemission from tribution and the second term is the relative the photocathode: Terminology: average variance of the crystal contribution. the number of photons per pulse. In this case,

Photoemission determined from total emission: Conditions: dark emission de- Actually, the crystal statistics are worse than the above assumption. This question is so that the variance in the measurement is discussed in more detail in Chapter 4, negligible; dark emission to be subtracted Photomultiplier Characteristics in the sec- from total emission to determine photoemis- tion on “Pulse Counting.” sion .

174 Statistical Theory of Noise In Photomultiplier Tubes

Photomultiplier tube: Conditions: Each stage identical with exponential statistics.

(Also, see section above on Pulse Counting gain dynodes assumed, high photon- Statistics for cases in which the dark emis- counting rates, output circuit noise contribu- sion count is not determined over a long time tions included. Terminology: charge on the interval.) electron, e; bandwidth, B; Boltzmann’s con- Multiplier chain with single electron input: stant, k; temperature, T, (degrees Kelvin); equivalent noise resistance of the input of the preamplifier that processes the anode

signal, Rn; shunt conductance in the anode 2 lead, g; shunt capacitance in the anode lead, C.

Photomultiplier tube: Conditions: high- gain dynodes assumed, gain and output cur- rent sufficient so that circuit noise com- ponents can be neglected. Terminology: photocathode emission current, i.

Photomultiplier tube: Terminology: av- erage number of electrons collected at the

Photomultiplier tube and scintillator: Conditions: scintillations resulting from gamma-ray photoelectric excitation in the crystal; equal gain per stage assumed. Ter- minology: full width half maximum, FWHM; photocathode quantum efficiency,

average number of photons incident on the Photomultiplier tube: Conditions: Each variance in the number of photons in each stage is identical and has Poisson statistics.

FWHM = 2.355/SNR

The first term inside the brackets is the relative variance associated with the photo- multiplier; the second term is the relative variance associated with the scintillator. Photomultiplier Handbook

REFERENCES first dynode photomultipliers,” Appl. Phys. 127. T. Jorgensen,“On Probability Gener- Lett., Vol. 13, p 356 (1968). ating Functions,”Am. J. Phys., Vol. 16, p 131. L.A. Dietz, L.R. Hanrahan and A.B. 285 (1948); E. Breitenberger , “Scintillation Hance,“Single-electron response of a Spectrometer Statistics,” Prog. Nuc. Phys., porous KCl transmission dynode and appli- Vol. 4, (Ed. O.R. Frisch, Pergamon Press, cation of Polya statistics to particle counting London, p. 56 (1955); William Feller, An In- in an electron multiplier,” Rev. Sci. Inst., troduction to Probability Theory and Its Ap- Vol. 38, p 176 (1967). plications, Vol. 1, John Wiley and Sons, 132. J.R. Prescott, Nuc. Instr. Methods, Third Edition (1968). “A statistical model for photomultiplier 128. R.H. Brown and R.Q. Twiss, “Inter- single-electron statistics,” Vol. 39, p 173 ferometry of the intensity fluctuations in (1966). light,”Proc. Royal Soc. London, Vol. 133. M. Schwartz, Information Transmis- 243A, p 291 (1958). sion, Modulation, and Noise, McGraw-Hill, 129. W. Shockley and J.R. Pierce, “A p. 207 (1959). theory of noise for electron multipliers,” 134. F. Robben,“Noise in the measure- Proc. IRE, Vol. 26, p 321 (1938); V.K. ment of light with photomultipliers,” Appl. Zworkykin, G.A. Morton, L. Malter, “The Opt., Vol. 10, No. 4, pp 776-796, (1971). secondary emission multiplier-a new elec- 135. W.A. Baum,“The detection and tron device,”Proc. IRE, Vol. 24, p 351 measurement of faint astronomical sources,” (1936). Astronomical Techniques, Edited by W.A. 130. G.A. Morton, H.M. Smith, and H.R. Hiltner, The University of Chicago Press, Krall, “Pulse height resolution of high gain (1962).

176 Index

Index

Absorptance ...... 18,125 Darkcurrent ...... 4,8,16,126 Absorption coefficient ...... 18 Dark current and noise ...... 54 Acceptor...... 125 Dark current and temperature ...... 8 Acceptor levels ...... 14 Dark current, increase following exposure After pulse ...... 57, 125 to fluorescent lamps ...... 43,58 Aging ...... 58, 75 Dark current, sources of ...... 19,53 Air pressure and photomultiplier operation ... . 76 Dark noise pulse spectrum ...... 59 Angle of incidence and photoemission .... . 39,92 Dark noise reduction with cooling ...... 50,54,58 Angle of polarization and photoemission .. . 39,40 Degaussing ...... 46 Angstrom unit ...... 125 Delayline ...... 12 6 Anode ...... 3,34,125 Delta function ...... 59,63,126 Anode current maximum ...... 50 Densitometry ...... 105 Anode pulse current maximum ...... 47 Detectivity ...... 56,90, 126 Anticoincidence circuit...... 125 Deuterium lamp ...... 151 Applications list...... Appendix A Differential cooling ...... 74 Argon arc lamp...... 153 Discriminator, pulse height ...... 126 Astronomy, application to ...... 5,112 Discriminator setting ...... 68 Avalanche photodiodes ...... 8 Donor ...... 126 Dynamic compression of output signal ...... 87 B ackground counts ...... 12 5 Dynode...... Band bending ...... 14 Dynode, Cs-Sb ...... 96 Bandwidth and noise spectrum . .59,125,170,171 Dynode, Cu-Be ...... 98 Basing ...... 88 Dynode, GaP:Cs ...... 68,83 Beta radiation, effect on photomultiplier ...... 44 Dynode glow...... 56 Black-body radiation ...... 148 Dynode voltage control ...... 83 Bleeder...... 125 Box-andgrid dynode structure ...... 28 EADCI (Equivalent Anode Dark Current Input)...... 55, 127 Cage...... 125 E2B ...... Calibration, standard lamp...... 92 Einstein ...... 10 Candela ...... 125,139 Electron affinity ...... 13,127 Capacitors, by pass...... 84 Electron multiplier...... 127 Carbon-arc lamp ...... 153 Electron optics ...... 26-35 CAT scanner ...... See CT-scanner Electron resolution (See also Pulse-height Cerencov radiation ...... 19,57,92,97,126 resolution, single electron) ...... 127 Channel multiplier ...... 32,126 Electron volt ...... 2O, 127 Channel number ...... 126 Electrostatic focus ...... 4 Circular cage ...... 4,27 Energy diagram ...... 10, 13,14,94 Coincidence counter., ...... 72 Energy distribution, photoelectrons ...... 12 Collection efficiency ...... 27 ENI (Equivalent Noise Input) ...... 8,9,55,127 Collection uniformity ...... 41 Environmental effects ...... 74 Color temperature ...... 126,139 Environment, high humidity ...... 76 Color-temperature standard...... 149 ...... Environment, pressure ...... 76 Compton effect .93 ERMA ...... 15,16,127 Conduction band ...... 126 ...... Connections, terminal ...... 88 Escape depth 13,18 Constant-fraction triggering ...... 99 Cooling photomultipliers ...... 74 Fall time...... 62,127 Count ...... 12 6 Fatigue ...... 50,51,127 Counting efficiency ...... 126 Fatigue and dynode materials ...... 51 Count-rate stability ...... 52,96 Feed back...... 4 Crossed-field photomultiplier ...... 33 Fermi-Dirac energy distribution function . . 11,13 Crosstalk...... 12 6 Fermilevel ...... 10,11,13,127 Cross talk in liquid scintillation counting ...... 73 Fiber-optic communication systems...... 8 Cryostats for photomultipliers...... 54 Field emission ...... 60 CT-scanner ...... 6,7,102,125 Filters, narrow bandpass color ...... 92,158 Curie...... 12 6 Filters, neutral density ...... 91 Current-voltage characteristic ...... 36 Fluorescent lamp ...... 154

177 Photomultiplier Handbook

Fluorometry...... 106 Flying-spot scanner...... 115,127 Focussing electrode...... 127 Focussing-electrode voltage...... 42 Foot candle...... 127 Foot Lambert...... 127 Forbidden band...... 13,127 FWHM (Full Width at Half Maximum) . .63,64,127

Gain...... 3,9,127 Gain control...... 83 Gain, maximum...... 75 Gain, variation with temperature...... 49 Gain, variation with voltage...... 44 Gamma radiation and glass discoloration.....43 Gamma-ray camera...... 6,100,128 Generating functions...... 160 Glass, browning of...... 19,77 Glass charging effects...... 56 Glass, properties of...... 19 Glass transmission...... 19 Ground potential...... 75

Headlight dimmer...... 6 Helium penetration...... 57,76,90 Hertz (Hz)...... 128 History of photomultiplier development...... 3-6 Hofstadter...... 5 “Hysteresis” (cyclic instability)...... 47,52,127

Illuminance...... 128,140 Illuminated area, photocathode...... 92 Insulator charging...... 47,52 IPA’s (Integrated Photodetection Assemblies) .80 Irradiance...... 126

...... 13 2 Johnson noise...... 61

Lambert’s cosine law...... 128,141 Laplace’s equation...... 27 Laser range finding...... 8,113 Lasers ...... 155-157 Lasers, Nd:YAG...... 8 Leading edge timing...... 9 6 Leakage, ohmic...... 53 Life expectancy...... 3 Light chopper...... 92 Light emitting diodes, (LED’s)...... 8,158 Light level for photomultiplier use...... 89,90 Light level reduction methods...... 90,91 Light pipe...... 95,128 Light shielding...... 75 Light sources for testing...... 158 Linearity...... 4 7 Linearity and photocathode resistivity...... 47 Linearity and voltage divider current...... 82 Linearity measurement...... 49 Linearity, pulse measurement of...... 86 Liquid scintillation counting....69,72,87,97,128 fluoride...... 19 Lumen(lm)...... 128 Luminance...... 128.14 0 Luminous efficacy...... 128 Luminous Intensity...... 128,129 Lux...... 128

178 Index

179 Photomultiplier Handbook

180