Photoelectric Devices

S. Yu. Lukyanov and A. A. Razdel, Leningrad

Photoelectric devices, long used in physical research, are finding ever wider application in various fields of technology.

The operation of these devices may be based on:

a) the external photoelectric effect,
b) the internal photoelectric effect,
c) the photoelectric effect of the blocking layer (Sperrschichtphotoeffekt).

Devices based on the last two phenomena are used relatively little in technology; this is partly connected with the insufficiently complete study of these phenomena. In addition, the technical specimens available have certain organic shortcomings that limit the range of their application. These shortcomings include, chiefly, the impossibility of using them at high frequencies. Even at ordinary audio frequencies (of the order of 1000 Hz) a noticeable inertia is observed. Moreover, the small internal resistance greatly complicates the subsequent amplification of the photocurrents obtained in photocells of these types. Therefore, in what follows we shall dwell on photocells and other devices based on the external photoelectric effect.

From the design standpoint, and depending on the requirements imposed, photoelectric devices based on the external photoelectric effect may be divided into four groups:

1) photocells proper,
2) photorelays,
3) television devices,
4) photoelectric devices with secondary emission (photodynators).

There is no possibility of enumerating here all the possible applications of photocells in science and technology. It may be said that the photocell, which not so long ago was used in scientific laboratories, has followed the same path as the electron tube and has become an equally widespread technical device. It is sufficient to point to sound cinema, certain fields of telemechanics, and objective photometry. In television, despite the appearance in recent years of special devices, photocells continue to play a major role in all systems of mechanical television.

For certain purposes, where it is desirable to obtain direct control of large currents by light and where inertial properties do not play a major role, so-called photorelays may be used. In principle the device is a combination of a photocell with an amplifying tube. Many investigators have tried to construct such a device, designed as a special electron tube containing a controlling photocathode. However, these devices have not become widespread because of the difficulty of manufacturing them with uniform and predetermined parameters. Therefore, unless special compactness is required, preference is given to special circuits made of photocells and amplifying tubes, which provide large amplifications and large currents with a minimum number of circuit elements.

At present, mechanical television is developing only in the direction of improving the structural design—chiefly of the optical part of the television apparatus—using, in essence, ordinary photocells for converting fluctuations of light intensity into electrical pulses. Since the possibilities for improving the quality of the transmitted image are limited by the amount of light used, it is, of course, photocells with high yield and inertia-free operation up to very high frequencies (of the order of \(10^6\) Hz) that are used for modern mechanical television. Alongside mechanical television, in recent times so-called cathode television has developed strongly, using the photoelectric effect on new principles. The two directions in cathode television are defined by the work of Zworykin \(^1\) and Farnsworth \(^2\). In both cases the television devices are special apparatuses without the participation of any mechanical parts. Both of these systems compete with one another, far surpassing ordinary mechanical devices in convenience of operation and in sensitivity.

One more, quite recently introduced and very fruitful, application of the photoeffect consists in the use of secondary electron emission from photoactive surfaces. Secondary emission makes it possible to obtain large and inertia-free amplification of photocurrents. It can therefore be applied as “distortion-free amplification” in cathode television, which opens up still greater prospects for the latter.

I. Photocells

The ejection of electrons by light quanta occurs only beginning with a certain wavelength, which for most metals lies in the far ultraviolet region. This wavelength, called the “red limit” of the photoeffect, lies in the visible part of the spectrum only for the alkali metals. Therefore it is precisely the alkali metals Li, Na, K, Rb, and Cs that are used as photoelectric surfaces in technical photocells. The alkali metals, generally having a low work function of electrons, [[unclear: text cut off at bottom of page]]

corresponds to the position of the red limit in the visible part of the spectrum), are characterized, in addition, by the presence of a certain maximum of photoemission on the spectral-distribution curve (Fig. 1). The presence of this “selective” maximum leads to a general increase in the integral photoemission in the visible region. Referring the integral photoemission to a unit of incident light energy, even for these metals we obtain extremely small values, not exceeding fractions of a microampere per lumen,* which is insufficient for technical use. Therefore all the efforts of experimenters were directed toward increasing the overall sensitivity of photoelectric surfaces. It turned out that an increase in sensitivity of this kind is obtained by treating the surface of alkali metals with various gases or with vapors of certain substances. At the same time the work function decreases (the red limit shifts toward longer wavelengths), and sometimes a new selective maximum appears (also lying in the longer-wavelength part of the spectrum). Such treatment sharply increases the sensitivity of the photocell, bringing it, for some surfaces, up to 20–40 μA/Lm.

Fig. 1. Spectral characteristics of pure alkali metals.

Along with increasing photoelectron emission by sensitizing the surface, another method of increasing it is also used, consisting in filling the finished photocell with an inert gas. With a sufficient potential difference between the cathode and anode of the photocell, the electrons torn out by light, accelerated by the field, ionize gas molecules and increase the number of charged particles moving between the electrodes. As a result, the current through the photocell increases by approximately 10–12 times.

The first technical types of photocells were potassium photocells sensitized with hydrogen. Their manufacture is carried out as follows. After the usual evacuation and heating of the bulb, a thick layer of potassium is deposited on the walls, with a transparent window left for the passage of light. Hydrogen is admitted into the photocell under a pressure of the order of tenths of a mm Hg, and a glow discharge is passed between the cathode and anode. In this process the surface of the potassium changes color, which serves as a criterion for the amount of hydrogen absorbed. At the same time the sensitivity of the photosensitive layer increases, reaching a maximum at

* Lumen (Lm) is a practical unit of luminous flux. One international candle gives a total flux of \(4\pi\) lumens.

in a certain amount of absorbed hydrogen. To increase the sensitivity, after treatment of the surface the photocell is usually filled with gas, at which point its manufacture ends. This type of sensitization proved imperfect: with time the absorbed hydrogen can be released from the surface, as a result of which the sensitivity of the photocell falls.

The integral sensitivity of hydrogen–potassium photocells does not exceed several microamperes per lumen (for vacuum specimens); the selective maximum is localized in the violet part of the spectrum (4460 Å).

The next stage of development was sulfur–potassium photocells. In contrast to the preceding ones, the potassium layer here is deposited on a magnesium underlayer previously evaporated onto the wall, serving for better contact with the lead-in through the glass. The potassium surface is then treated with sulfur vapors until a certain optimal sensitivity is obtained. In mass production here, as in hydrogen–potassium photocells, the degree of readiness of the photocells is judged by the change in color. A well-treated surface has a golden-yellow color. The sensitivity is likewise of the order of several microamperes per lumen and, after filling with gas, reaches 30–40 microamperes in good specimens.

Fig. 2. Spectral characteristic of a sulfur–potassium photocell

Sulfur–potassium photocells proved more stable with time and became widely used in technology. Their selective maximum lies within the range from 4300 to 4500 Å, depending on the degree of treatment of the surface with sulfur. The red limit reaches 7000 Å (Fig. 2).

In the USSR, two designs of photocells of this type are chiefly in use. Since evaporation of magnesium onto the walls of the bulb is performed by heating a piece of magnesium in a tungsten spiral, in one design the anode is made in the form of a nickel ring that fastens the ends of such a spiral (Figs. 3a and 3b). In addition to the indicated types, for special purposes (for example, a Bildtelegraph) photocells of special designs are manufactured, which in principle have the same properties and are adapted only for the most convenient and complete use of light in the given optical system.

As we see, the sensitivity of potassium photocells with both methods of treatment is still not too high. The development of television imposed high demands for further increase—

…increase in the sensitivity of photocells. It was natural to turn to the next alkali metals in the periodic system, which have an even smaller work function in comparison with potassium. The best in this respect among the elements known at present is cesium, but pure cesium has a sensitivity lower than treated potassium. From a number of experimental and theoretical works³˒⁴ it became clear that the smallest work function and an intense selective maximum should be possessed by complex surfaces consisting of alternating layers of substances with small and large work functions. Direct treatment of the cesium surface with sulfur, hydrogen, and various other substances did not, however, give the desired sharp increase in sensitivity.

Fig. 3. Two designs of the anode of a photocell.

The works of Koller⁵ and Campbell⁶ showed that in this case it is necessary to proceed by another, more complicated route. Cesium photocells received their technical form in the works of Zworykin⁷ and Prescott and Kelly⁸. It turned out that the treatment of a cathode containing cesium should advisably be carried out not from the surface of a thick layer of metallic cesium, but, on the contrary, cesium should be deposited from vapor onto a metallic backing previously prepared by suitable treatment.

As the backing in cesium photocells, silver is used, either deposited on glass chemically or by evaporation, or in the form of a silver plate. The silver layer is oxidized by a discharge in oxygen at a pressure of about 1 mm Hg to a depth of about 100 atomic layers. The depth of the oxidized layer can be determined both by direct measurement of the amount of oxygen absorbed and by the accompanying change in the color of the silver surface. The various stages of oxidation at which the oxidation process is terminated range from a blue to a yellow-green shade. After evacuation of the oxygen, metallic cesium, obtained by reduction from a salt, is deposited on the surface of the oxidized silver, partly reacting with it and forming cesium oxide. In this form the photocell does not yet possess the required sensitivity. Further activation of the surface is carried out by heating the photocell in a furnace at a temperature of about \(200^\circ\mathrm{C}\) under continuous pumping. During this process the photoemission gradually increases, reaching a considerable value. Good specimens of manufactured photocells have from 20 to 40 \(\mu\mathrm{A}/\mathrm{Lm}\). Filling with gas brings the integral sensitivity up to 300 \(\mu\mathrm{A}/\mathrm{Lm}\).

The spectral characteristic of cesium photocells differs sharply from the characteristic of potassium photocells. In Fig. 4 one can see two characteristic maxima: one lying in the violet part of the spectrum, the other—in the near infrared region.

The red limit reaches 12,000 Å. The data given by different authors differ somewhat in this case, especially with respect to the red limit.

As we have already mentioned, cesium photoelements may be of two types: the first type, made by the preliminary deposition of silver on glass, has the same construction as the potassium one described above (a spherical bulb and an anode of various forms); the second type is made in the form of a semicylindrical silver plate (cathode), in front of which there is a nickel anode in the form of a rod or a grid (Fig. 5). The whole construction is enclosed in a cylindrical bulb. Photoelements of such construction are manufactured in the USSR, for example TsLPS⁹, and the spherical type with photocathodes on a silver layer deposited on glass—by VEI.

From the spectral characteristics of the various photocathodes it follows which photoelements should preferably be used, given one or another spectral distribution of the light source employed in a particular case. Cesium photoelements, by virtue of their high integral sensitivity, are convenient for the entire region of the visible spectrum, as well as for the near infrared region. Potassium photoelements are convenient for the blue-violet part of the spectrum.

Fig. 4. Spectral characteristic of a cesium photoelement.

For the ultraviolet part of the spectrum one may use both the types of photoelements considered and photoelements with cathodes of pure metals, of which cadmium is the most convenient; of course, the bulbs of such photoelements must be made of quartz (or have quartz windows) or of special kinds of glass transparent to such rays.

Fig. 5. Construction of a cesium photoelement with a plate cathode.

As we have already mentioned, to increase the sensitivity of vacuum photoelements they are filled with some inert gas. Usually neon or argon serves for this purpose. The gases must be absolutely pure, free of impurities that can change the properties of the cathode. Not to mention oxygen or organic contaminations (vapors of sealing compound, etc.), even chemically rather inactive nitrogen, being adsorbed on the cathode surface, sharply worsens its photoelectric properties. Special purity of the gas is necessary for the stability of the photoelement over time.

Typical current-voltage characteristics of vacuum photoelements are shown in Fig. 6. Various vacuum photoelements—

except for the magnitude of the saturation current, which depends on the sensitivity of the cathode, they also differ in the anode voltage at which this saturation occurs. The corresponding anode voltage depends chiefly on the design. For spherical photocells with a central anode (the most common type

Fig. 6. Volt-ampere characteristics for vacuum photocells.
$A$—semicylindrical cathode, rod anode; $B$—semicylindrical cathode, grid anode; $C$—central cathode.

of photocells), saturation occurs at rather large anode voltages, of the order of 50–100 V. This happens because the electrons from the cathode surface fly out in all possible directions, and some of them, not reaching the anode, return to the cathode.

With the reverse arrangement, i.e., with a central cathode and a spherical anode surrounding it, all electrons emitted from the center of the sphere inevitably reach the surrounding surface even at the smallest anode voltages. Technically such a construction is disadvantageous because of the small cathode surface, but if it is necessary to have saturation at a small anode voltage, this can be done in another way. By making the cathode in the form of a plate and placing a wire grid serving as the anode at a sufficiently short distance from the cathode, we obtain a saturation current already at 5–10 V.

Fig. 7. Volt-ampere characteristic of a gas-filled photocell. $P$—operating point.

A completely different picture is observed for gas-filled photocells (Fig. 7). At small anode voltages, before ionization begins, they behave like vacuum cells. Then the current begins to increase out of proportion to the applied voltage. Next there follows a certain, almost linear, section, which, with a further increase of voltage, passes

Photoelectric Devices

into the vertical straight line corresponding to a glow discharge. The current here is limited only by the ballast resistance in the photocell circuit, and in this region there is no longer any dependence on illumination.

The ignition potential of the self-sustained discharge depends on the nature and pressure of the gas filling the photocell, on its design, and on the number of initial electrons. Therefore the ignition potential in the dark will in general be somewhat higher than when the photocell is illuminated. Even at the very low intensities of light flux ordinarily used (\(\sim 0.05\ \mathrm{Lm}\)) this difference reaches 50 V. The curve of the dependence of the ignition potential on illumination, taken for a cesium photocell (of the same design as in Fig. 5), is presented in Fig. 8.

It is clear from this what voltages should be applied to the photocell in order to use it in the most advantageous way. In fact, from Fig. 7 it is clear that the operating point for a gas-filled photocell must be chosen on the steep portion of the characteristic, where the amplification is maximal. In reality, the possibilities in this direction are limited by the proximity of the discharge. Therefore it is difficult to speak of the “sensitivity” of a gas-filled photocell—it will depend on many factors, and this makes its value considerably less definite than for vacuum photocells. If for vacuum photocells sensitivity is understood as the photocurrent, recalculated per unit of light flux of a definite spectral composition and measured in the saturation region, then for a gas-filled photocell the corresponding photocurrent must be measured at the operating point of the characteristic, whose position is determined, as we have seen, more or less arbitrarily. The desire to obtain the highest possible sensitivity compels one to approach as closely as possible the discharge potential, which is especially tempting in view of the steep increase of sensitivity in voltage intervals close to the discharge. The high sensitivity, reaching in such measurements 500–600 \(\mu\mathrm{A}/\mathrm{Lm}\), is, however, of very relative value, since operation in the immediate vicinity of the discharge is accompanied by an increase of inertia; in addition, fatigue phenomena appear. The entire operation of the photocell becomes unstable and characteristic of the transition regime from a non-self-sustained to a self-sustained discharge in the gas. In view of

Fig. 8. Dependence of the ignition potential on illumination.

one should not choose the operating point too close to the ignition potential of the discharge in light; thus, for example, for photoelements with an ignition potential of 250–300 V, at the illuminations usually used (0.1–0.01 Lm), the operating point must be 30–40 V below the discharge potential.

The integral sensitivity of a photoelement thus depends on a number of factors: its geometry, the pressure and kind of gas, and, chiefly (for gas-filled photoelements), on the applied voltage. But, in addition, as we have already indicated when speaking of the spectral characteristics of photoelements, the sensitivity depends to a large extent also on the spectral composition of the rays incident on the surface of the cathode.

Fig. 9. Dependence of photocurrent on illumination.

Even with illumination by so-called “white light” one may obtain quite different values of sensitivity in microamperes per lumen (which is the usual technical measure of sensitivity) with small changes in the incandescent temperature of the light source. Indeed, when the temperature is increased, the maximum of the energy emitted by the light source shifts into the violet part of the spectrum according to Wien’s law. Therefore, it would be physically correct to measure the sensitivity of photoelements in coulombs per unit (calorie or erg) of incident monochromatic energy. Since this would require rather complicated optical apparatus, in practice lumens are nevertheless used as units of luminous flux, but with an indication of the absolute color temperature of the source at which the sensitivity of the photoelement is measured. In the technical use of photoelements with different cathodes, this circumstance plays an important role.

All measurements of sensitivity are, of course, made on the assumption of strict proportionality between the illumination of the cathode and the photocurrent. In technical photoelements this simple law of photoelectricity ceases to be absolutely strict, and in the region of high illuminations deviations from direct proportionality become appreciable. These deviations are due to a number of causes: the appearance of charges on the walls of the bulb of the photoelement, a change in the conditions of reflection of light, and a change in the structure of the cathode itself. From the curve in Fig. 9 it is clear, however, that over the entire broad range of illuminations used in practice (from 0.001 Lm to several lumens), i.e., when the illumination is changed by 10,000 times, the proportionality of the two quantities is fully preserved.

So far we have confined ourselves to considering the properties of photoele-

ment only in the static regime. Meanwhile, in the majority of technical applications photoelements operate at a high frequency of light pulses. Under these conditions gas-filled photoelements exhibit certain new properties—the so-called inertia. The physical nature of the phenomenon consists in the fact that the electrons knocked out by an individual light pulse do not immediately ionize the gas filling the photoelement; the ions formed also require a certain time to reach the corresponding electrodes, since they move considerably more slowly than electrons. Then, at a high frequency of light pulses, the next pulse may arrive before the ions created by the preceding pulse have reached the electrodes. As a result, both the rise of the current and its disappearance will occur with a delay relative to the frequency of the light pulses. This phenomenon is illustrated by the curve in Fig. 10. The total quantity of electricity per unit time, i.e., the current passing through the photoelement, of course remains unchanged at any frequency for a given light intensity. However, the amplitude of the current passing through the photoelement (the depth of modulation), with increasing frequency of the light pulses, decreases, and consequently the amplitude of the voltage across the resistance at the amplifier input decreases. In practice this means that high frequencies are slightly cut off, which introduces distortion into the transmission. From the curve in Fig. 11 it is evident that the inertia acquires a noticeable magnitude beginning approximately at 10,000 Hz. Thus, for sound-cinema purposes gas-filled photoelements are sufficiently suitable, since the range of sound frequencies reproduced by audio-frequency amplifiers and loudspeakers usually does not exceed this limit. Conversely, for television purposes, where the range of frequencies used is extremely wide and, for satisfactory reproduction of the image, requires frequencies above 500,000 Hz, gas-filled photoelements are plainly unsuitable. Indeed, in television exclusively vacuum photoelements are used.

Fig. 10. Curve of the photocurrent for a gas-filled photoelement at a high frequency of light pulses.

Fig. 11. Frequency characteristic of a gas-filled photoelement.

Summarizing all the above, it may be said that gas-fill

...photoelements having high output are limited in their applications. Therefore, further improvement of vacuum photoelements, free from the indicated shortcomings of gas-filled photoelements, is desirable in order to increase their sensitivity. The most perfect of the photoelements developed up to the present time are cesium photoelements. Without going into theoretical details, it may be said only that further technical progress in the manufacture of photoelements can proceed in two directions:

1. By treatment of the cathode of the photoelement directed toward better utilization of the light incident on the photoelement (increasing the roughness of the surface^11, composite layers^12).
2. By using certain phenomena not directly connected with the photoelectric effect (electron reflection, the dynatron effect). These questions will be considered later (§ 4).

II. Photorelays

As we have already indicated, attempts were made to produce a device which, depending on the light, would give, without a complicated amplifying circuit, large currents capable of controlling electromagnetic relays. One solution of this problem is the construction^13 shown in Fig. 12. The principle of the device consists in the use of a “free grid.” At the center of the device there is an incandescent filament possessing a large electron emission. On both sides of it are two electrodes, of which one serves as the anode, while the other is insulated from the remaining parts of the device. A photosensitive layer is deposited on the wall of the bulb, to which, with respect to the filament, a large negative potential is applied. Under the action of light the photocathode emits electrons, which are accelerated in the direction of the anode. Some of them may fall on the free electrode \(G\) and charge it to a certain negative potential. In this case electrode \(G\) will play the role of a negatively charged grid, causing the anode current from the incandescent filament to decrease; within certain limits the anode current will depend on the illumination of the photolayer.

Fig. 12. Schematic of the arrangement of a photorelay with a hot cathode.

After the action of light ceases, the potential of electrode \(G\) again reaches a potential close to zero (taking the potential of the incandescent filament as zero), as a result of which the anode current again increases to its former value in the dark. This occurs through the discharge of electrode \(G\) by positive gas ions present in the device (at a very low pressure) when it is darkened.

The quantitative theory of the phenomenon, as well as experimental verification, show that the charge and discharge of the controlling electrode require...

...require an appreciable time (inertia); this time depends on the illumination, the geometry of the device, and the pressure of the gas filling the photorelay. For the device described, the inertia at the illuminations usually used reaches hundredths of a second. On the other hand, the currents obtained amount to tens of milliamperes per lumen. The use of such devices is, of course, impossible for audio frequencies, but they are very convenient for all sorts of control and regulating devices. A whole series of various kinds of photorelays is briefly described in the book by Simon and Zuurman[^14].

Photorelays nevertheless have not become widespread, owing to the difficulty of their manufacture. Special circuits are used for the same purposes, some of which we shall present here.

The simplest of them is the Rosenberg circuit[^15] (Fig. 13), which in essence operates on the same principle as the hot-cathode photorelay described above. Here a combination of a photocell and an amplifying tube is used. The anode of the photocell is connected directly to the grid of the tube and plays the role of the isolated electrode of the photorelay. When the photocell is illuminated, it becomes a conductor and applies a negative bias from the auxiliary battery to the grid of the tube. As a result the tube is “blocked,” i.e., the anode current falls. When the photocell is darkened, the grid of the tube is isolated and is charged by a positive grid current. Thus a necessary condition for operation of the circuit is the presence in the tube of a positive (ionic) grid current. Despite its simplicity, this circuit is extremely sensitive and has an enormous amplification factor. In Geffcken’s book[^15] there are many other, more complicated circuits, the detailed discussion of which we cannot dwell on here.

Fig. 13. Rosenberg circuit.

It is not the purpose of the present article to give an exhaustive review of all possible applications of amplifier circuits with photocells; at present such “phototelemechanics” has become enormously widespread, especially in the USA. At the recently held “Century of Progress” exhibition in Chicago (at the end of 1934), a large number of such devices were demonstrated. Photocells opened doors when people approached; illuminated display windows, previously dark, for passers-by; served as reliable guardians of exhibited valuables, etc. Moreover, the signal for the opening of the exhibition and the sudden illumination of it were switched on by the light of a star that at a certain time was passing through the meridian of Chicago.

It must be said that all these remarkable applications of photocells are not only a demonstration and advertisement of technical achievements, but are rapidly penetrating into all branches of industry.

III. Television Devices

The first stage in the development of television was so-called picture telegraphy, i.e., the comparatively slow transmission of still images, chiefly printed or handwritten text. The principle of the transmitting and receiving system of a picture-telegraph apparatus is clear from Fig. 14. A beam, focused to a point, falls on a cylinder moving along a screw thread. The cylinder is wrapped with a sheet of paper bearing the image (text) to be transmitted. The diffusely reflected light falls on a photocell. The photocurrent is determined by the coefficient of reflection of light from the various points of the image. In this way we obtain current pulses that vary according to the dark and light areas of the image; these pulses are amplified and enter the transmission line.

Fig. 14. Picture-telegraph system, \(S\)—transmitting device, \(E\)—receiving device.

The receiving device is shown on the right-hand side of Fig. 14. The incoming current pulses are amplified and sent to a light modulator, which is usually a Kerr condenser or a point gas lamp. The light from the modulated source falls on photosensitive paper, blackening it in complete correspondence with the transmitted image. To avoid distortions, the receiving cylinder must move entirely synchronously with the transmitting one; this is accomplished by a special synchronizing device. The modulated light source must, of course, possess sufficient photochemical activity.

Fig. 15. Television system with a Nipkow disk. \(S\)—Nipkow disk, \(F\)—image of the “picture” on the disk, \(G\)—light source, \(C\) and \(O\)—optical system, \(B\)—transmitted “picture,” \(L\)—focusing lens, \(Z\)—photocell.

Television proper, or the transmission of moving images,

PHOTOELECTRIC DEVICES

is currently carried out in several different ways. The oldest of these is “mechanical” television, which has not yet lost its technical significance even now, especially in connection with certain very recent improvements. Let us briefly describe the most widespread system of mechanical television, with a Nipkow disk. Light passing through the image is scanned by the Nipkow disk and focused on the photoelement (Fig. 15). For each revolution of the Nipkow disk, each aperture successively runs along the corresponding “line” of the image. The number of apertures determines the number of lines in the transmitted image. The fluctuations of light and shadow in the image are converted in the photoelement into electrical impulses. The more detail we wish to transmit in the image, the more apertures there must be on the Nipkow disk, and the smaller the area of each aperture must be. More detailed scanning is associated with a decrease in the intensity of the light falling on the photoelement. This does not present particular difficulties for stationary images, since a stronger light source can be used. But already in the transmission of a motion picture, and still more in the transmission of naturally illuminated objects or landscapes, we encounter ever greater difficulties because of the limited sensitivity of the photoelement at such low light intensities, since amplification of very small photocurrents presents great, and sometimes even insurmountable, difficulties. A somewhat better use of light is provided by the application of the so-called mirror screw, which in essence gives the same scanning as the Nipkow disk; but even in this case the light from naturally illuminated objects is insufficient for detailed scanning, i.e., for good image quality. Thus mechanical television has reached an impasse, from which a way out has been indicated only very recently in several different directions.

One of the ideas leaves the very principle of mechanical television unchanged and has proceeded along the path of improving the photographic part of the apparatus. Instead of directly transmitting a poorly illuminated scene, in the “intermediate-film” method (Zwischenfilm) the object is first photographed on motion-picture film, which is developed in time in a special apparatus and then passed through the television transmitting apparatus. Both parts of the apparatus (the photographic and the television parts) are mounted together; the time between photographing and transmission, owing to the considerable improvement of photochemical processes, is so short (less than a minute) that this method makes it possible to transmit almost immediately the image of any object illuminated sufficiently for filming with an ordinary motion-picture camera. The difficulties with light are thus overcome here, but the bulkiness of the apparatus of the mechanical television transmitter is further increased by the introduction of fairly complex photochemical devices.

A fundamentally quite different approach to the problem of transmission

images we find in so-called “cathode television.” Two systems of cathode television have become widespread—Zworykin’s^21 and Farnsworth’s^20. Zworykin’s system has already been described in this journal. Let us note that the conversion of an invisible image into a visible one, characteristic as one of the applications of Zworykin’s system, has also been described by another author^16; the corresponding device was called a light converter.

Let us now proceed to a description of the television system according to Farnsworth’s method. Farnsworth’s device, in its original form, has no advantage over mechanical television with respect to sensitivity. This device, which he called a “dissector” (literally, “cutter”), is shown in Fig. 16. The photosensitive cathode is deposited on glass as a semitransparent layer, so that the image can be projected onto the photocathode from the outside. In this way an electronic image is obtained on the cathode, i.e., a photocurrent of different intensity flows from differently illuminated regions of the cathode. If a plane anode is placed in the path of the emitted electrons, then we obtain some average current, independent of the distribution of illumination over the surface of the photocathode.

Fig. 16. Farnsworth dissector.

However, at each point of the anode there arrives an electron current of the same intensity as is obtained at the corresponding point of the electronic image on the photocathode. The electronic image will in this case be transferred to the anode without distortion only in the case of a homogeneous electrostatic field between the electrodes, which is achieved by applying a metallic layer with high resistance to the side walls of the device (in addition, magnetic focusing is used). At the center of the anode there is a small aperture (about 0.4 mm in diameter), through which the electrons, accelerated by a rather strong field (600 V is applied between the cathode and the anode), reach a second anode placed behind the aperture. In contrast to Zworykin’s iconoscope, where the electronic image was stationary and its scanning was carried out by a narrow electron beam moving over it, here the electronic image itself moves, and through the aperture there pass in succession, and strike the second anode, separate portions of the electronic image.

Scanning is carried out by two mutually perpendicular magnetic fields.

Despite the measures adopted to improve the homogeneity of the field, distortions are obtained at the edges of the image in Farnsworth’s dissector.

Thus, at the present time there exist the following television systems: mechanical television, which has become widely dissem-

distribution in Germany, especially thanks to its improvement by the intermediate-film method, and cathode television, which is developing successfully in America along both of the indicated lines. It is difficult to give preference to any one of these methods, since the colossal increase in sensitivity in Farnsworth’s method, achieved by the use of the electron multiplier (of which we shall speak in more detail below), can also be used in mechanical television.

The receiving apparatus for television, of course, does not depend on the transmitter system employed. Formerly the same Nipkow disk with a neon lamp or Kerr element was used as the television receiver. This device has a number of shortcomings; the chief of them is, of course, the low light intensity in the received image, the small dimensions of the image itself, and also the distortions inherent in the neon lamp itself. Now the television receiver of this type has been almost completely abandoned, and the cathode television receiver has been adopted, which in principle is a cathode oscillograph improved for television purposes. In the USSR, at the Svetlana plant (Leningrad), such cathode receiving tubes (“kinescopes”) have been developed.

IV. Electron Multiplier

It has been known for quite a long time^17 that electrons striking the surface of a metal with a certain velocity are diffusely reflected from this surface. The reflection coefficient, i.e., the ratio of the number of scattered electrons to the number of incident ones, depends very strongly on the velocity of the incident electrons. For certain values of the velocity (i.e., certain values of the applied potential difference), characteristic for the given surface, the reflection coefficient becomes greater than unity. For most pure metals the reflection coefficient becomes greater than unity when the electrons attain velocities corresponding to 30–40 V. In Fig. 17 is shown the general character of the course of the reflection coefficient \(R=\dfrac{I}{I_0}\)

Fig. 17. Dependence of the reflection coefficient on the velocity of incident electrons for a pure metal.

depending on the velocity of the incident electrons[^18]. At small velocities, at first \(R\) rapidly falls, reaching a minimum value, after which it begins to increase more smoothly, becomes greater than unity, and passes through a diffuse maximum.

The initial fall of the reflection coefficient can be explained by the fact that at small velocities a large part of the electrons is reflected; as the velocities increase, the electrons penetrate more deeply into the thickness of the metal, and the reflection falls. But in parallel with this, new electrons begin to be torn out from the depth of the metal. Their number must increase with an increase in the energy of the incident electrons, which explains the new rise of the curve. The velocity of the electrons torn out (and consequently their energy), as a detailed investigation has shown, is much less than the velocity of the incident electrons; therefore their number may exceed the number of incident electrons, i.e., the reflection coefficient becomes greater than unity. Finally, at very large velocities of the incident electrons they can penetrate into the metal so deeply that escape to the outside becomes difficult for the electrons torn out, and we observe a gradual decline of the reflection curve.

The picture described in general outline1 applies to any surfaces; however, the magnitude and position of the maximum on the curve depend to a high degree on the properties of the surface. Experimental data show that the maximum \(R\) increases as the work function of the surface decreases. Indeed, the surface of ordinary cesium cathodes, having the smallest known work function, gives a large coefficient, reaching \(R = 8—10\).

In technology one has encountered a large reflection coefficient \(R > 1\) in work with cathode tubes; this phenomenon, called the “dynatron effect,” disturbed the normal operating regime of the tube, and therefore it was combated in every possible way. The few attempts to use it proved not very successful.

At the present time this phenomenon has been approached from an entirely different side, and new ideas for using the effect have proved extraordinarily fruitful and promising.

Let a beam of electrons fall on a strongly reflecting surface. If the reflecting surface has a reflection coefficient \(R\), and the current in the primary beam is \(I_0\), then a current \(RI_0\) will flow to the collecting electrode (collector), which has a positive potential relative to the reflecting plate. Consequently, for \(R > 1\) we shall obtain a peculiar amplification of the primary current. The device described here in principle bears the name “electron multiplier.” This type of amplification differs fundamentally from the ordinary tube amplifier circuit, where each stage contains a combination of capacitances and resistances. Instead of voltage amplification at each stage in an electron multipli-

in the body we obtain a direct increase in the quantity of electricity passing through the collector circuit. For an even greater increase of current in the electron multiplier, the electron beam can be made to be reflected several times, which will correspond to a multistage tube amplifier.

A simple device based on amplification of the photocurrent with a single reflection was built by Iams and Salzberg2. Since their aim was to construct a photocell with a large output, the design had to satisfy certain technical requirements. The anticathode (the secondary-emission cathode) and the collector should, as far as possible, screen the photocathode less from the incident light; the collector must have a high permeability for photoelectrons, but at the same time, at not excessively high potentials, collect well the electrons reflected from the anticathode. In addition, the interelectrode capacitances must be sufficiently small (a general requirement for devices operating at high frequencies). In the “photodynatron” of Iams and Salzberg the cesium photocathode has the form of a semi-cylindrical plate, edge-on to which is placed a flat anticathode (also with a cesiated surface), surrounded by a collector in the form of a flat sparse grid (Fig. 18). With an initial photocathode sensitivity of 27 μA/Lm, this photodynatron gives about 150 μA/Lm, which corresponds to approximately sixfold amplification. A positive potential of 200 V is applied to the anticathode, and 250 V to the collector relative to the cathode. The reflection coefficient, and consequently the amplification, can be increased by applying larger accelerating voltages to the anticathode and collector, but this increase is insignificant.

Fig. 18. Photodynatron of the Iams and Salzberg design.

A similar device was developed at the VEI with a somewhat modified design. The cesium cathode and anticathode are deposited opposite one another on the walls of a spherical bulb. The collector (a sparse grid) is located right next to the anticathode. Light falls on the cathode through the transparent side surface of the bulb (Fig. 19). The reflection coefficient here is approximately 6, and the output is close to 160 μA/Lm.

As we have already said, it is more advantageous to use multiple reflection. Let \(n_0\) be the initial number of electrons emitted by the cathode, and \(R\) the reflection coefficient of the surface of the anticathode. Then after the first reflection the number of electrons will be \(n_0R\), after the second \(n_0R^2\), and so on. After \(k\) reflections the number of electrons will increase to

\[ n = n_0 R^k, \]

which, even for small \(k\), gives enormous amplification. At the Institute of Telemechanics (Leningrad) a number of devices of various designs based on this principle have been developed. We shall describe one of these

devices (Fig. 20). Several electrodes (up to 10) are sealed into a cylindrical tube, each of which is a photocathode.

The first of them, under illumination, emits photoelectrons, which are “turned” by a perpendicular magnetic field in the direction of the neighboring electrode. Each electrode has a positive potential with respect to the preceding one; the magnetic and electric fields are chosen so that the electrons emitted from one electrode strike the next with a velocity sufficient to knock out secondary electrons. The last electrode has the greatest positive potential and serves as the collector. The currents obtained amount to several milliamperes at illuminations of the order of several hundredths of a lumen. The amplification given by the device is approximately \(10^5\). By improving the reflecting surfaces, still greater amplifications can be obtained.

Fig. 19. Photodynatron of VEI design.

Farnsworth, who has long been engaged in the reflection of electrons, at the end of last year published the results obtained and their application to television. He developed multistage electron multipliers of two types—static and dynamic. The static device is similar in its principle to the device of the Institute of Telemechanics described above. Farnsworth’s static multiplier is shown in Fig. 21. A thin semitransparent film of platinum or nickel is deposited on the wall of the tube by evaporation or spraying. The resistance between the two ends of this layer ranges from 0.5 to \(2\ \mathrm{M}\Omega\). Over the nickel film a layer of silver is deposited, from which, by suitable treatment, a cesium photocathode is obtained. Along the axis of the tube a thin wire is stretched, serving as the anode. To one end of it is attached a metallic collector in the form of a funnel. To both ends of the semiconducting photosensitive layer a considerable potential difference (over 500 V) is applied, so that a potential gradient is created along the tube. The anode (collector) has the greatest positive potential. Photoelectrons emitted from the upper end of the tube are accelerated toward the anode, at the same time moving along it not as a result of the potential gradient. They strike the opposite wall with a certain velocity of the order of 100 V or more, depending on the point of impact. At these velocities they knock out secondary electrons, which in turn are accelerated in the same way; reaching the opposite wall, they

Fig. 20. Electron multiplier of the Institute of Telemechanics. The dotted line indicates a magnetic field perpendicular to the plane of the drawing.

again knock out an increased number of electrons. Thus each transverse passage gives an increase in the number of electrons. The amplification obtained in this way depends on the mean potential difference traversed before each impact (the reflection coefficient depends on the velocity of the electrons) and on the number of impacts, according to the formula given above. Because of the appearance of space charge, the final electron current is directly proportional to the number of initial electrons. In this device the amplification obtained was \(10^4\)-fold and even greater, depending on the quality of the tube and the applied potentials. The advantage of Farnsworth’s static device is the possibility of control without a magnetic field.

Fig. 21. Static Farnsworth multiplier.

In the described types of electron multipliers, amplification was achieved by multiple reflection from a series of surfaces; it is possible to limit oneself to only two surfaces, but to make the electron beam be reflected many times from each of them. This principle is used in Farnsworth’s dynamic (radio-frequency) multiplier. In Fig. 22 the construction of this device is shown in its main features. It consists of a cylindrical tube, on both sides of which there are flat cesium photocathodes. Between these photosensitive plates is placed a cylindrical anode (collector).

Fig. 22. Dynamic Farnsworth multiplier.

The collector may occupy the entire space between the plates or be only a central ring. A solenoid, supplied with direct current, is placed on the tube in order to create a pro-

...a longitudinal focusing magnetic field between the cathodes. A high-frequency voltage of the order of \(5 \cdot 10^7\) Hz or higher is applied to the cathodes. A positive potential from an auxiliary battery is applied to the anode; the negative pole (the midpoint of the high-frequency generator) is grounded. An electron (for example a photoelectron), emitted by one of the cathodes, is accelerated toward the opposite one under the action of the anode voltage. It does not reach the anode because of the focusing action of the magnetic field, and reaches the opposite electrode if the dimensions of the tube, the voltage, and the period of the high-frequency field are such that the electron manages to traverse the distance between the plates in a time equal to a half-period of the applied high frequency. During this time the high-frequency field does not change its direction and continuously accelerates the electron. The time of flight is determined by the distance between the plates and by the velocity acquired from the anode voltage, plus the integral effect of the potential between the cathodes during the time of flight. If the high-frequency voltage is sufficiently large, secondary electrons are torn out, which then, under the action of the field that has changed sign, are accelerated in the opposite direction in order, in turn, to tear electrons out of the first plate. Subsequently the process is repeated. Thus, with each impact, the number of electrons is multiplied in accordance with formula (1):

\[ n = n_0 R^k, \]

where \(k\) still denotes the number of reflections. It should be noted that, without a high-frequency voltage on the cathodes, the anode potential determines only the mean velocity of the electrons passing through the tube, but does not affect the impact velocity, since the acceleration received by an electron leaving one of the cathodes, as it approaches the anode, is completely neutralized by the deceleration as it approaches the other cathode. In order to obtain a reflection coefficient greater than unity, it is necessary to apply to the cathodes a high-frequency voltage of not less than 50 V. Since the time of flight at a frequency of \(5 \cdot 10^7\) must be of the order of \(10^{-8}\) sec, the distance between the cathodes, as approximate calculations show, must be of the order of several centimeters. According to Farnsworth’s data, this distance was \(6\) cm; by selecting the anode voltage, it was possible to achieve exact coincidence of the electron time of flight with the half-period of the high-frequency field. This value of the voltage was determined from the sharp increase of the current at the collector.

Indeed, although the arrival of a single electron at the collector is almost impossible owing to the presence of the focusing magnetic field, with the given form and arrangement of the latter a certain fraction of all the electrons will nevertheless be collected on it. This fraction will depend on the positions of the cathodes emitting secondary electrons (i.e., on whether the electrons strike closer to the center or to the periphery of the plates), and on the transverse component of the electric field in the tube, which will be determined

under the influence of the space charge. The greater the focusing field, the greater the values that the oscillatory current between the plates can attain. The increase of the current between the plates, however, is limited by the increase in the density of the space charge. Indeed, ultimately a state will be reached in which the number of newly emitted secondary electrons becomes equal to the number collected at the collector—the “suction” current (the current to the collector) will become constant (saturation). Under conditions of an equilibrium regime, when as many electrons arrive at the collector per unit time as are produced upon reflection, according to Farnsworth one can obtain a current of up to 0.5 A. One might think that, in order to obtain such a suction current, a very large number of reflections would be necessary. However, even if one assumes that the initial current is negligibly small, amounting, for example, to \(10^{-18}\) A, and that the reflection coefficient is \(R = 6\), then, using equation (1), we obtain:

\[ 0.5 = 10^{-18} 6^k, \]

whence

\[ k = \frac{17 \cdot \lg 5}{\lg 6} = 16. \]

Thus, under these conditions, only 16 reflections are sufficient. It is easy to see that, already with 105 reflections, a single initial electron gives \(10^{82}\) electrons, and all the electrons in the entire universe (according to Einstein–de Sitter their number is \(10^{79}\)) would suffice only to maintain a current a thousand times weaker.

In reality, although in this type of multiplier enormous amplification factors can be achieved at small initial currents, in practice, at the illuminations used, amplifications of only several thousand times have so far been obtained, owing to the limiting action of the space charge. This amplification is, of course, not limiting and can be increased by an appropriate selection of electric and magnetic fields or by the use of several multiplication stages.

Farnsworth used a dynamic multiplier in his television device. Fig. 23 shows a device consisting of a combination of a television tube with a multiplier. On the glass of the tube a semitransparent cesium photocathode \(AB\) is deposited. From the outside, the optical image is focused onto it. Opposite the cathode is placed an auxiliary anode with an aperture in the center. The anode and cathode are connected by a semiconducting film of metal, which creates a uniform field gradient and removes charges from the walls. After the auxiliary anode there is placed a dynamic multiplier, and in the cathode of the multiplier facing the auxiliary anode there is likewise an aperture. The distance between the cathodes is about 6 cm. The collector is a ring located midway between the cathodes. The entire amplifying tube is enclosed in a metallic sheath. Between the cathode on which pa-

gives the image, and a high voltage of the order of 600 V is applied to the auxiliary anode.

The auxiliary anode has a small positive potential relative to the first cathode of the multiplier, in order to eliminate secondary electrons that arise near the aperture and can cause distortions. Between the cathodes, through a variable capacitor, a high-frequency voltage is applied, so that the high-frequency potential can also be varied. The collector has a positive potential of the order of 90 V relative to the cathodes of the multiplier. The amplified pulses in the electron multiplier are fed to a tube amplifier from a resistance in the collector circuit.

Fig. 23. Farnsworth television apparatus with dynamic multiplier.

In Farnsworth’s apparatus magnetic focusing and scanning are used. It is extremely difficult to completely eliminate the mutual influence of the magnetic fields of the focusing and scanning coils; therefore the images transmitted by Farnsworth’s apparatus have characteristic distortions. This, however, is not a fundamental shortcoming of the system, and it may be hoped that in the future these technical defects will be reduced to a minimum. The sensitivity of Farnsworth’s system with an electron multiplier is not inferior to the sensitivity of Zworykin’s iconoscope and is entirely suitable for television under natural illumination.

Farnsworth also developed a special receiving device, which is an ordinary television receiver with a cathode oscillograph, but likewise with magnetic focusing. In addition to television, the electron multiplier can be used for many other radio-engineering purposes.

LITERATURE

1. Zworykin, J. I. E. E. 73, 437, 1933.
2. Farnsworth, U. S. Patent N I. 773, 980.
3. Fowler, Proc. Roy. Soc. A 128, 123, 1930
4. Olpin, Phys. Rev. 36, 251, 1930.
5. Koller, Phys. Rev. 36, 1699, 1930.

1. Campbell, Phil. Mag. 12, 173, 1931.
2. Sewig, Z. Physik 76, 91, 1932.
3. Prescott & Kelly, Bell Syst. Techn. Journ. 11, 334, 1932.
4. Kiselev and Luk’yanov, Zhurnal tekhn. fiziki IV, 1560, 1934.
5. Usikov, Bull. Ts. L. P. S. 2, 38, 1934.
6. Kluge, Phys. Z. 22, 844, 1933.
7. Asao a. Susuki, Proc. Inst. of Radio Eng. 19, 655, 1931.
8. Videts, Zhurnal tekhn. fiziki 4, 1174, 1934.
9. Simon u. Suhrmann, “Lichtelektrische Zellen und ihre Anwendung,” Berlin, 1932.
10. Geffken, Richter, Winkelmann, “Lichtelektrische Zelle als Technisches Steuerorgan,” 143.
11. Holst, de Boer, Teves a. Veenemans, Physica 4, 297, 1934.
12. Bayer, Phys. Z. 10, 168, 1909; Hull, Phys. Rev. 7, 1, 1916.
13. Lukirsky, “Izvestiya G. F. T. I.” 1, 107, 1922.
14. Jams & Salzberg, Proc. Inst. of Radio Eng. 23, 55, 1935.
15. Farnsworth, Journ. Frankl. Inst. 218, 411, 1934.
16. Zvorykin, Uspekhi fiz. nauk 14, 778, 1934.

1. Both here and below we use the term “reflection” for any values of \(R\), less than or greater than unity, although physically \(R > 1\) means the tearing out of electrons. ↩

2. 19. ↩