We have already seen that the creation and recombination of electron-hole pairs in or near a PN junction makes it possible to detect or generate light with considerable ease, and even to obtain amplification by the transistor effect. Before semiconductors, the electrical effects of light were classified as the photoemissive, photovoltaic and photoconductive effects. The photoconductive effect is seen in CdS cells, the photovoltaic effect in "solar cells" that produce a voltage when illuminated, and the photoemissive effect in the devices we will study here. This classification is not a very good or inclusive one--it omits the electron-hole effects that are of so much importance today, for example.

The photoemissive effect, more commonly called the photoelectric effect (though not the only one, as we have just noted), is the emission of free electrons from metal surfaces illuminated by light. It was discovered by Heinrich Hertz in 1887, and was very important to the understanding of quantum effects. For this reason, it is interesting to study the effect in the phototubes that were used for light detection before semiconductors.

The valence electrons in metals are free to move about, but are confined to the volume of the metal by the potential well produced by the positive ion cores. This potential well is shown in the figure at the right, in the right-hand plot of electron energy versus distance, x. The potential well is a very deep one, so deep that it is amazing that electrons can be coaxed out under any circumstances. The key to the situation is shown in the left-hand plot of the number of electrons in the available states. There is not an infinite number of states, but a quite finite number, few at low energies and more and more at higher energies, the number of available electron states increasing as the square root of the energy. Electrons are curious in that only a pair of electrons at most can occupy any state, and this pair must have their spins opposite. The valence electrons fill up the states starting at the lowest energies, and continue upwards. At T = 0 K, all the states are filled up to some energy E called the Fermi level, shown in the figure. The electrons at the top have quite high energies, but still can't go anywhere. It takes an extra energy Φ, the work function, to allow them to depart. At finite temperatures, some of the electrons at the top are kicked up to higher levels, and this empties some states below the Fermi level, as shown by the dotted line. If the metal is made hot enough, some of the electrons may acquire more than the work function of extra energy, and leave. This, of course, is thermionic emission, as used in vacuum tubes.

Light may be absorbed by one of the electrons at the surface of the metal. If the frequency of the light is f, the energy transfers are in multiples of hf, where h is Planck's constant. This is the "quantum energy" of the light, which leads to the concept of photons, packets of energy of the amount hf. This is a bit oversimplified, but will do here. It was photoelectricity that first demonstrated and clarified many of these matters. If an electron absorbs an energy hf from the light, and requires energy Φ to escape, it is left with a kinetic energy E = hf - Φ, a relation called Einstein's equation. Einstein's Nobel Prize came on the basis of the photoelectric effect, and on work he had done in statistical mechanics, not for relativity.

The Einstein equation shows that there is minimum frequency of light that can emit a photoelectron, given by 0 = hf - Φ or f = Φ/h. In terms of wavelength, the limit is λ = 1240/Φ nm, where Φ is in electron-volts. Light of lower frequency, or longer wavelength cannot produce photoelectrons, no matter how intense it is. These were the original puzzles of photoelectricity, unexplainable classically. The number of photoelectrons, or photocurrent, is, however, proportional to the intensity of the light. Photoemission is a surface process, and so is affected by the nature of the surface, a fact which led to great experimental difficulties in its study. Most light incident on a metal is reflected, and most of the rest is absorbed as heat, making photoemission a rather rare process. The quantum efficiency of a photocathode is the number of photoelectrons per incident photon. The quantum efficiency of a photocathode falls off at short as well as at long wavelengths. At short wavelengths, the photons penetrate so deeply before they are absorbed that there is little chance of an electron's escaping.

To be sensitive at the peak sensitivity of the eye, 550 nm, the work function of the photocathode must be less than 2.2 eV, and to be sensitive over the whole visual range of 400-700 nm, 1.8 eV. Only the alkali metals Na, K, Cs and Rb have such low work functions, and are the basis of most photocathodes, generally as alloys rather than as pure metals. Na and K are difficult to work with, so most use Cs and Rb. There was a search for photocathode materials with low work function, and at the same time a large quantum efficiency, which is the average number of photoelectrons emitted per photon of the incident light. Some of the most useful materials are the S-1 Ag-CsO-Cs cathode, sensitive out to 1200 nm with a peak near 800 nm, the S-3 Ag-Rb cathode, sensitive to 900 nm with a peak at 430 nm, and the S-4 Sb3Cs cathode, the most sensitive, with a peak sensitivity near 400 nm and a maximum wavelength of 700 nm, sketched at the left. The S-1 photocathode appears silvery and somewhat pink, the S-4 grey. There were some 16 different cathode responses, most only used on a small number of devices. Cathodes like the S1 and S4 are not really metals, but complicated semiconductors that can have low work functions. The quantum efficiency of a photocathode is the average number of photoelectrons per quantum absorbed. For an S-4 cathode, it can be 0.1 at the maximum sensitivity. For the S-1 cathode, the efficiency is much lower, perhaps 0.004.

All photodetectors exhibit dark current, which is a small current in the absence of light. With photocathodes of low work function, particularly the S-1 cathode, thermionic emission at room temperature may be sufficient to cause noticeable dark current. Cosmic rays and natural radioactivity are other sources, in addition to leakage conductance. Dark current can be reduced by cooling, so most very sensitive detectors are cooled, often to liquid nitrogen temperatures.


A phototube consists of a photocathode of large area, in the shape of a cylinder, with an anode wire on the axis, in a glass envelope. The anode is made positive with respect to the cathode to collect the electrons emitted. In a vacuum phototube, the envelope is highly exhausted so that the electrons do not suffer collisions, and the photocurrent saturates at a low voltage. In a gas phototube, Ne or A at low pressure is introduced to amplify the photocurrent. The photocurrent initially saturates as in a vacuum phototube, but above 20 V it begins to increase, and may be multiplied by 7 at 80 V or so. At still higher voltages, a glow discharge begins, which does the tube no good, and is not useful. The mechanism of multiplication is not simple. An accelerated electron may, clearly, create positive ions and negative electrons on collision with neutral gas atoms, which add to the current. At a somewhat lower voltage, the electrons produce metastable ions which drift about until they collide with the cathode, when they cause the emission of extra electrons. The amplifiying effects all require time, so gas phototubes have severe frequency limitations. Vacuum phototubes have no such limitations, and are very prompt in their response.

The first commercial use of phototubes was in the reproduction of sound from motion picture films. The sound was recorded as a variable density sound track that was naturally synchronized with the pictures. Though the film moves in jerks in projection, it must move at constant speed in the sound pickup. This was an interesting technical problem, whose solution was ingenious and practical. Another use of phototubes was in alarms, automatic door openers and production line counting, where the breaking of a light beam was detected.

In the 1950's, RCA alone manufactured some 137 different types of photosensitive devices and cathode-ray tubes, from the simple gas and vacuum phototubes that are the subject of this page, through photomultipliers, camera tubes, image-converters, and storage tubes, to cathode-ray tubes for oscillographs, kinescopes and flying-spot cathode ray tubes. There were even monoscopes, which, when scanned produced a static video picture. The 2F21 gave an Indian-head pattern for testing transmitters and receivers. RCA even made a CdS photoresistive cell, the 6957, with an octal base. The sensitivity was classified as S-15 (maximum at 585 nm), the maximum power dissipation was 0.5W, the maximum voltage 250V. with 50V applied, the sensitivity was 1.64 A/lm, or 4 mA/ft-cd, with a dark current of 20 μA. See Optoelectronics for more information on CdS.

The basing for phototubes is shown at the right. These connections apply to nearly all phototubes with octal and small 4-pin bases, including the 1P40/930 gas and the 1P39/929 vacuum phototubes that are typical. If there is a cap, it should be obvious whether it is connected to the anode or cathode. The corresponding pin is then unused. The arrow shows the direction of incidence of light. The photocathodes for these tubes have a projected area 5/8" wide x 13/16" high, centered 1-5/8" above the seat of the base. The 1P39/929 has an S-4 cathode with maximum sensitivity at 400 nm, and is designed for photometry. The maximum anode voltage is 250V, the maximum photocurrent 5 μA, the maximum ambient temperature 75°C. The absolute sensitivity is 0.045 μA/μW at maximum, or 45 μA/lm for 2870K tungsten light. The dark current is 12.5 nA. The 1P40/930 has an S-1 cathode with maximum sensitivity at 800 nm, and is designed for sound reproduction and relays. The maximum anode voltage is 90V, the maximum photocurrent 3 μA (6 μA for anode voltages less than 70V), the maximum ambient temperature 100°C. The absolute sensitivity is 0.012 μA/μW at maximum, or 135 μA/lm at DC, 111 μA/lm at 5 kHz, and 101 μA/lm at 10 kHz. The maximum gas amplification factor is 10, and the dark current 100 nA. The voltages and specifications are quite typical of all gas and vacuum phototubes.

The glass used for phototube envelopes is probably the usual lime (crown) glass used for other vacuum tubes. Such glass is transparent from about 300 nm to 3000 nm, which includes most of the wavelengths to which the photocathodes are sensitive. The published data does not indicate whether it includes the effects of the glass or not. "Uviol" and similar glasses are transparent out to 250 nm or so, while quartz transmits in the wide range of 185-4500 nm. There is no advantage in using these glasses for phototubes, but they may have application to other optoelectronic devices. It should also be remembered that many substances considered transparent, like water or ice, are so only for the narrow band of visible wavelengths, and are quite opaque to ultraviolet or infrared radiation. Reflection by the glass envelope is on the order of 10%, half at each interface, allowing for oblique incidence. Light is not deflected by the parallel glass surfaces nor focused on the photocathode.

A special kind of phototube that is extremely sensitive because of amplification is the photomultiplier. An electron emitted from the cathode is accelerated and collides with a dynode, an electrode that produces secondary electrons. This continues for 9 or more stages, producing a shower of electrons that is collected by the anode. Under the proper conditions, the absorption of a single quantum can be detected. A typical small photomultiplier was the 931-A, with 9 dynodes and an S-4 cathode. Normally, there was 100V between successive dynodes, making a total of 1000V from cathode to anode. The average amplification at each dynode was 4.5, making the gain 800,000 overall. The maximum anode current was 1 mA. In the 1950's, RCA manufactured 19 types of photomultipliers. They found their greatest scientific use in astronomy and in scintillation counters, but were also used in such mundane applications as detecting auto headlights for an automatic dimmer. Photomultipliers must not be exposed to high light levels with voltages applied, nor to elevated temperatures. They often must "cool down" in the dark until their noise levels are acceptable.


The type 929 or 1P39 phototube has an S-4 photocathode and is a vacuum tube. The type 930 or 1P40 has an S-1 photocathode and is gas filled. These two make good examples, and are the ones I used, but there are many others. These tubes are rather rare and expensive, but are interesting to study. I found a type 1P40, similar to the 930, at a good price. All these have octal bases, with the cathode at pin 8 and the anode at pin 4. The octal key should face to the rear (light coming from the direction between pins 4 and 5). The test circuit, shown at the right, is very simple. Everything important can be seen with a supply voltage of 60 V maximum. The voltage at the top of the 100k (or 1M) load resistor gives the current, which will be a few microamperes, and is easily measured with the DMM.

It is convenient to make an enclosure for the phototube, using an aluminum box and an octal relay socket. I used a 2 x 3 x 5 box, with two terminals on top. A piece of brass tubing 5/8" ID and about 2" long, epoxied in a hole in the enclosure, directs the light. The inside of the tube and box should be painted matte black. The purpose of the enclosure is to block scattered and unwanted light while making connections easy.

As a light source, I used 25-, 40- and 60-watt incandescent bulbs. It would also be possible to use LED's and get some idea of the variation of the response with wavelength. The S-1 cathode is well matched to incandescent or LED light. The 60W bulb was advertised to give 840 lumens. If this were equally distributed in angle, the illumination at 45 cm would be 330 lux (lumens per square meter). A GE light meter said 42 ft-cd, or 452 lux, which is probably more accurate. The projected area of the photocathode is about 15 x 20 mm, so the total illumination is 0.136 lumen. The 929 gave a photocurrent of 6.6 μA with this illumination, so the efficiency of the S-4 cathode for tungsten light would be 6.6/0.136 = 48 μA/lm. A reference quoted 45 μA/lm, so the experiment is probably not far off. It would be easy to compare bulbs of different wattage. When you make measurements, look out for reflected light, perhaps coming from your clothes as you bend in to read the meters. One can use the phototubes to make photometric measurements of many kinds. There is no heater power, and no high voltage is necessary, so they are easy to use.

Measure the photocurrent as a function of the anode to cathode potential, and note the difference between vacuum and gas phototubes. Also measure the dark current as a function of anode to cathode voltage. An LED can be flashed on and off at varying rates to study the frequency response of the photocurrent. One can also try to modulate an LED so that an audio signal can be carried by the light and picked up by the phototube, making a light-beam telephone.

Lumens measure perceived visual intensity or brightness, and are related to the physical power, watts, in the light by the spectral sensitivity of the eye, which peaks at 550 nm, where there are 640 lumens per watt. A point source of intensity I = 1 candela (candle) emits 4π lumens equally in all directions. The 60W bulb, then, has an intensity of 67 cd. An illumination of one lux is one lumen per square meter, a ft-candle one lumen per square foot. Accurate measurements on phototubes with monochromatic light require a rather expensive monochromator, but some rough experiments could be done with a 60° prism and a few lenses, improvising a spectroscope, perhaps using the sun and a hole in the blind as Newton did.


J. Millman and S. Seely, Electronics (New York: McGraw-Hill, 1951), Chapter 15.

Radio Corporation of America, Photosensitive Devices and Cathode-Ray Tubes (Harrison, NJ: RCA, no date)

Return to Electronics Index

Composed by J. B. Calvert
Created 29 August 2001
Last revised 16 January 2002