LIGHT MEASUREMENTS USING PHOTOELECTRIC CELLS*

M. M. Gurevich, Leningrad

Introduction

The large number of articles devoted to the question of light measurements by means of photoelectric cells, both in previous years and in recent years, shows quite clearly that photoelectric cells have long been used for these purposes and in the most varied cases. Even if we at once set aside all measurements of invisible radiation, to which physicists very readily extend the concept of “light,” and confine ourselves to the considerably narrower region of the visible spectrum, we still encounter a very great variety of cases in which photoelectric cells are used. All these cases may be divided into three large categories. To the first category we assign spectrophotometric installations¹, sometimes made in the form of complex recording instruments², in which two monochromatic beams of the same wavelength are compared. This last circumstance very greatly simplifies the fundamental aspect of the question, since the matter reduces to a quantitative comparison of qualitatively identical and, moreover, elementarily simple compositions of radiation. The spectral properties of the photoelectric cell itself thereby prove to be of little significance.

They remain of little significance also for the second category of cases in which photoelectric cells are used, distinguished by the fact that it is not monochromatic beams that have to be compared, but complex beams with identical spectral compositions. The latter case occurs, for example, in studying the spatial distribution of the light from some source. Just as in the first case, the spectral properties of the photoelectric cell play no role, and likewise the whole interest of installations of this kind lies in another area—for example, in the automation of the entire process³,⁴. However, neither spectrophotometry nor the measurement of the light distribution of sources is the fundamental problem of photoelectric measurements.

* Report read before the Lighting Engineering Commission of the Technical Physics Group of the Department of Technical Sciences of the Academy of Sciences, 4/VI 1936.

The principal task in those cases which we assign to the third division consists in comparing light beams with different spectral compositions. As an example it would have been sufficient to mention the photometry of an ordinary gas-filled lamp, in which it has to be compared with a vacuum standard, were we not faced with the far more difficult problems of photometry of gas-discharge tubes. Here the spectral sensitivity of the photoelement receiving the light immediately comes to the fore; it must correspond to the spectral properties of the eye, something which at present cannot be considered achieved with a sufficient degree of accuracy. But since the photometry of differently colored sources is a problem of serious technical importance and, moreover, one that has no satisfactory solution by visual methods, the reason becomes quite clear that in recent years has given rise to a large number of works which we assign to the third division of our classification and which are devoted to photoelectric methods of comparing light beams with different spectral compositions.

There is neither the possibility nor the necessity of dwelling on the description of all cases of application of photoelements. With a greater or lesser degree of accuracy they are used for measuring luminous intensity, luminous fluxes, illuminances, determining coefficients of reflection and transmission, for measuring color, in combination with a microscope, telescope, etc.

The insufficiently high degree of accuracy of all these measurements is evident if only from the fact that the standardization of photometric quantities is still carried out exclusively visually. At the same time there is no doubt that continuously improving methods of photoelectric measurements will in the near future not only catch up with, but also surpass, visual methodology, which has stagnated in its possibilities for improvement. In view of this state of affairs, in the first part of the report we intended to give an outline of such a possibility, analyzing in this example the special features of light measurements with the aid of photoelements, and in the second part to dwell on the work of the photometric laboratory of the State Optical Institute in the field of the use of photoelements.

I. The possibility of photoelectric standardization of luminous quantities

Before speaking about the method of photoelectric measurement of a standard, let us pause briefly on the defects of visual methods of photometry. Without developing in detail the repeatedly noted shortcomings of visual methods, shortcomings connected with the psycho-physiological character of the basic process of establishing photometric equality, which depends on the state of the organism at a given moment, let us note one circumstance which, as it seems to us, has remained unappreciated until now. We have in mind

differences in the spectral properties of the human eye, established by a large number of investigators and perhaps best expressed by Coblenz and Emerson,^5 who plotted 125 sensitivity curves on one graph (Fig. 1).

In addition to this, it should be borne in mind that the sensitivity curve of each observer is not constant in time.

Fig. 1. Curves of the spectral sensitivity of the eyes of 125 observers

Let us choose two curves \(v_1\) and \(v_2\) (Fig. 2) from among those obtained by Coblenz and Emerson and see what results in photometry carried out by two corresponding observers. Suppose two fields are being compared—one emitting radiation corresponding to a color temperature* of \(2400^\circ\) K, and the other correspondingly to \(2800^\circ\) K. Suppose one observer establishes for himself a certain position of photometric equilibrium with great accuracy.

If there are no additional or interfering causes, then the picture obtained by the first observer will not appear to the second to be in equilibrium. The difference in the spectral curves will lead him to establish another position of photometric equilibrium, differing from the first by approximately 4%. In this case the first observer (\(v_1\)), in comparison with the second (\(v_2\)), will underestimate the radiation corresponding to a temperature of \(2400^\circ\) K, or, what is the same thing, will overestimate the radiation corresponding to a temperature of \(2800^\circ\) K.

An even considerably greater discrepancy between our observers will be obtained if, for example, we turn to the measurement of mercury gas-discharge lamps and compare them with radiation corresponding to a color temperature of \(2400^\circ\) K. We can pro—

* The color temperature of radiation is the temperature of an absolutely black body at which it gives radiation of the same color as the given source. The color temperature of the radiation of many incandescent metals is higher than their true temperature.

to carry out the comparison with the aid of the best flicker photometers*, which eliminate the color difference between the fields, and nevertheless, because of the difference in spectral sensitivities, we shall inevitably arrive at different estimates. In this case the discrepancy will amount to about 20%, and again the first observer \((v_1)\) will underestimate the radiation corresponding to a temperature of \(2400^\circ\mathrm{K}\), in comparison with the second observer \((v_2)\).

Fig. 2. Curves of the spectral sensitivity of two observers with normal color vision

We see, therefore, that the existing differences in the spectral properties of the eyes of individual observers make the precise measurement of luminous quantities on a visual basis impossible in principle.

The difficulties of heterochromatic photometry, which had become apparent quite long ago, gave rise in the literature to a number of proposals for a transition to purely physical methods of measuring luminous quantities. Among these proposals we shall note the relatively recently appearing articles by Ornstein \(^{6,7}\), who proposes abolishing special photometric (lighting-engineering) standards and units and using general physical CGS units.

Considering it necessary to preserve the generally accepted photometric units and at the same time to avoid the difficulties of heterochromatic photo-

* A flicker photometer is a photometer intended for comparing heterochromatic radiations. Its action is based on the rapid transfer of the radiations from one part of the field of vision to another, whereby the difference in colors disappears.

calibration, we raise the question of photoelectric methods of standard photometry.

Let us dwell in a few words on the reasons that at present do not allow the situation with the light standard to be regarded as satisfactory.

As the principal reason one should point to the instability of any radiation process associated with high temperature. Any material subjected to heating above \(2000^\circ\) K wears out relatively quickly; every setup changes, and does so the faster, the higher the temperature. Therefore light standards, which must be the most stable, have the lowest temperature; economically, lamps with a considerably higher temperature prove more advantageous and therefore are widely used.

Carbon standards of luminous intensity have a color temperature of \(2080^\circ\) K; tungsten standards, \(2360^\circ\) K, whereas lamps used for illumination have a color temperature of \(2700\)—\(3000^\circ\) K, depending on the power they consume. This difference in temperatures causes such a difference in colors that it creates serious inconveniences and produces large errors in visual measurements.

Moreover, the method now adopted of preserving the standard in the form of a group of lamps is, of course, not reliable and does not ensure the invariability of the standard. The desire to improve the preservation of the standard led to the idea of using an absolutely black body. At present the Americans\(^8\) have achieved good results along this path, realizing an absolutely black body at the solidification temperature of pure platinum and visually comparing its light with a carbon standard. There is no doubt, however, that the good results obtained in the USA are connected with the fact that the solidification temperature of pure platinum (\(2046^\circ\) K) happened by chance to be close to the temperature of the carbon standard (\(2080^\circ\) K), and that, in this way, the light beams compared had almost exactly the same composition.

However, even after overcoming the serious difficulties of obtaining a stable radiation regime of an absolutely black body at a temperature of \(2046^\circ\) K, we would still face the same obstacles in the visual comparison of this radiation with the radiations given by other light sources—difficulties whose cause lies in the physical differences existing in the spectral properties of the eyes of individual observers.

On the other hand, the photoelectric method, free of any connection with psychophysiology, will make it possible to eliminate errors associated with the uncertainty of the eye’s sensitivity curve.

The principal condition that would make it possible to carry out such work consists in bringing the spectral sensitivity of the photoelement into sufficiently close agreement with the established mean sensitivity curve of the human eye. Such adjustment is in itself a difficult task, but its solution has had

would make sense even in the case where it allowed only a transition from the primary standard to secondary standards and did not make it possible to carry out exact photometry of gas-luminous sources. At the same time it would prove possible to lower the temperature of the primary standard, which would make it more easily realizable and would increase its stability in operation.

It would be possible, for example, to pass from the melting temperature of platinum, which for visual measurements gives a light that is still too yellow, to the considerably more accessible temperature of \(1500^\circ\mathrm{K}\), at which an absolutely black body has a brightness still quite sufficient for photoelectric measurements.

The brightness of a black body at \(1500^\circ\mathrm{K}\) is about \(0.8\ \mathrm{sb}\), and if an illumination of \(20\ \mathrm{lx}\) is produced on a photocell of area \(10\ \mathrm{cm}^2\), then, with a photocell sensitivity of the order of \(300\ \mu\mathrm{A}/\mathrm{lm}\)*, we obtain a photocurrent of about \(6\ \mu\mathrm{A}\), the measurement of which with sufficient accuracy presents no difficulty.

Much more serious concern is caused not by the smallness of the energy to be measured, but by the simultaneous incidence upon the photocell of a relatively very powerful flux of infrared rays. Already in the region of wavelengths up to \(1.2\ \mu\) there is contained a power exceeding the measured one by hundreds of times and therefore representing a tremendous danger for the measurement, even in the presence of only a very slight sensitivity of the photocell. It should be noted that photocells most sensitive to visible radiation—in particular selenium photocells—still possess appreciable sensitivity in this part of the spectrum. Therefore, first of all, one must somehow get rid of this, albeit slight, sensitivity. Anticipating somewhat, let us say that at present, even if no ready solution of this problem exists, there are paths that promise to lead to the intended goal.

An even more considerable power, lying in the longer-wavelength region, already appears less dangerous for two reasons. First, the sensitivity of the photocell here falls practically to zero; and second, even if dangerous residual sensitivity were present, a ten-centimeter layer of pure water, which transmits the entire visible region excellently, constitutes an insurmountable barrier for this kind of ray\(^9\) (transparency not higher than \(0.0005\%\)).

But the absence of sensitivity in the infrared region is evidently insufficient for the photocell to be able to serve for light measurements with adequate accuracy. The curve of its spectral sensitivity must differ little from the curve adopted as the sensitivity curve of the eye**. Evidently, here it is difficult

* \(1\ \mu\mathrm{A} \approx 10^{-6}\lambda\). \(\mathrm{lm}\) is the abbreviated notation for the unit of luminous flux—the lumen.
* See any course in photometry, for example Sh. Fabry, General Introduction to Photometry, ONTI, 1934, or S. O. Maisel, Light and Vision*, GTTI, 1932.

expect absolute agreement, and it is hardly necessary. When comparing sources with a continuous spectrum, we can, obviously, impose different requirements on the accuracy of the agreement of the curves in different parts of the visible spectrum.

Taking into account the large amount of energy concentrated in the long-wave part of the spectrum, we shall obviously have to evaluate errors differently in the blue-violet and in the red regions.

The final process of photoelectric standardization appears to us in the following form. The aperture of an absolutely black body heated to \(1500^\circ\mathrm{K}\) illuminates the surface of the photoelement. If necessary, a cuvette with water is inserted in the path; it may be useful in order to avoid heating the photoelement. It is not difficult to obtain the spectral absorption of water with the required degree of accuracy.

Having measured the resulting photocurrent and knowing the illumination produced on the photoelement, we obtain a calibration of the photoelement for illumination. Then, replacing the black body by a secondary tungsten or carbon standard, we measure the illumination produced by it and thereby determine the luminous intensity of the standard.

In order to calculate the illumination produced on the photoelement by an absolutely black body, apart from the spectral properties of the optics, we need only the conventionally established curve of the sensitivity of the eye and the mechanical equivalent of light*. The latter is needed only in order to express illumination in practical units. And if, after carrying out the process described, we base the calculation on some arbitrarily chosen value for the mechanical equivalent, then by doing so we shall establish a new system of photometric units, based on radiation of an absolutely black body, connected by an exact relation with the absolute CGS system of units and arbitrarily close to those practical units which we are accustomed to use.

The incandescent lamps which now serve as standards must, in this case, be preserved, but instead of maintaining the primary group standard, they will have the role of secondary standards, objectively calibrated against the primary standard—an absolutely black body.

In concluding the presentation of considerations on the possibility of photoelectric measurements of the luminous standard, we note that, despite the great difficulties that undoubtedly stand in this path, the problem is of sufficient fundamental interest that it is worth expending effort on its solution.

* The mechanical equivalent of light is the ratio between the luminous and radiant fluxes for radiation with wavelength \(0.555\,\mu\), where the eye has maximum sensitivity. The mechanical equivalent of light is equal to \(0.00161\ \mathrm{W/lm}\).

II. Work of the Photometric Laboratory of the State Optical Institute in the Field of Applying Photoelements to Light Measurements

The first attempts to apply photoelements to light measurements were made in the photometric laboratory of the GOI quite a long time ago.

The most substantial of the earlier works were published in our press at the time [10, 11]. I consider it necessary only to recall that one of the first installations carried out was an apparatus for determining the spectral sensitivity of photoelements [12], assembled with the aid of a simple glass monochromator and a ribbon lamp with a known color temperature. This apparatus is still operating in our laboratory.

Fig. 3. Spectral sensitivity of a selenium photoelement and of the average eye

At first we used photoelements with an external photoelectric effect, potassium and cesium ones, manufactured in various laboratories. But, beginning approximately in 1933, our attention was drawn more and more toward a new type of photoelement, the so-called barrier-layer photoelements, or photoelements with a blocking layer, in particular—selenium photoelements. At present we are interested only in these photoelements and in the possibilities of using them for light measurements.

Let us briefly consider the properties of selenium photoelements with a blocking layer that make them so interesting from the photometric point of view. First of all, let us note the spectral

sensitivity of the photocell (Fig. 3) greater than the sensitivity of any other, approaching the sensitivity curve of the average eye and overlapping it on the right and on the left. The latter circumstance makes it possible, without a large loss of light, to use light filters that bring the properties of the photocell still closer to those of the eye.

Next, one should point to the high integral sensitivity of these photocells, which has one of the highest values among all photocells known to us. For our photocells it ranges from 300 to 500 μA/lm (at a color temperature of 2800° K), which in most cases makes it possible to dispense with amplification of the photocurrents.

Furthermore, the enormous practical convenience of these photocells consists in the fact that barrier-layer photocells do not require the application of any external electromotive force for their operation. In the absence of amplification this reduces the photometric system to the combination of a photocell and a galvanometer. The combination of all these properties makes the selenium photocell extremely interesting for the photometrist.

In a few words, the process of manufacturing selenium photocells is as follows. A thin layer of selenium—30–80 μ thick—is deposited in vacuum onto a specially prepared iron plate. The resulting selenium, in a non-conducting state, is converted into a semiconducting modification by being kept for several hours at a temperature close to 200°C. After this the selenium is covered with an extremely thin (semi-transparent) layer of gold, which serves as one of the electrodes. The transparent electrode is applied by cathode sputtering.

This seemingly very simple manufacturing process leads to the production of photocells with a very considerable individual variability of properties. Finding the cause of one or another deviation proves to be an extremely difficult problem and, in most cases, one that has not yet been solved. As a result, the entire process still contains a significant element of crude empiricism, and the necessary result is not fully assured.

As a result, a rather significant portion of the photocells manufactured proves not entirely satisfactory. It should nevertheless be noted that recently this portion has greatly decreased, undoubtedly in connection with certain regularities we have found.

First of all, one should note a property of selenium photocells, discovered by us quite unexpectedly, of losing their sensitivity when the photocell is placed in vacuum. This process was observed repeatedly and sometimes was completed within 2–3 hours. A photocell removed from under the bell jar did not recover its sensitivity; its color changed, and the surface of the selenium began to conduct electric current considerably better.

Since photocells covered with a thin layer of shellac retained their sensitivity in vacuum, this led to the supposition that the blocking layer of the photocell is a layer of gas which is released from it in vacuum and is retained by the shellac film[^13]. Without going into details, we shall note only that at present we have abandoned this supposition and have come to the conclusion that the loss of sensitivity is connected with the destruction of the blocking layer under the action of mercury vapors present under the evacuated bell jar.

In addition to the very substantial practical consequences with which the property discovered is connected, it enabled us to convince ourselves that the cause determining the photosensitivity of the photocell does indeed lie in a very thin layer on the surface of the semiconducting selenium. This is confirmed by the fact that if, from a photocell that has completely lost its sensitivity, one scrapes off from above a very thin layer, then such a plate can afterward again be converted into a photocell by applying a semitransparent gold electrode. It follows from this that the harmful influence of mercury, which destroyed the photosensitivity of the photocell, did not extend to any deeply lying layers of selenium.

Fig. 4. Equivalent circuit of a selenium photocell of the valve type.

The presence of mercury absorbed by the upper layer was established spectroscopically in photocells taken out of the “mercury vacuum.”

Before turning to the consideration of the stability and temperature coefficient of selenium photocells, let us examine the generally accepted equivalent circuit of a photocell, shown in Fig. 4. Here 1 and 2 are the semitransparent gold layer and the surface layer of semiconducting selenium, forming, as it were, the plates of a certain capacitor. Under the action of light penetrating through the gold layer, photoelectrons are torn out of the selenium layer in an amount proportional to the radiant energy incident on the photocell. The accumulation of electrons is hindered by the potential difference created on the capacitor plates and by the reverse transition of electrons into the semiconductor, for which the blocking layer represents a certain resistance, denoted in Fig. 4 by \(r\). The letters \(r_1\) and \(r_2\) denote the resistances of the semitransparent gold layer and of the semiconducting selenium layer. \(W\) is the external resistance across which the photocell is closed.

The electron flux \(I\) torn out by the light is divided into two parts: \(i_{\phi}\), the photocurrent in the external circuit, and \(i_r\), the reverse leakage current through the resistance \(r\).

In this case

\[ I = i + i_{\phi}. \]

We shall call the sum \(r + r_1 + r_2 = R\) the resistance of the photocell. Then the relation between the total current \(I\) and the photocurrent \(i_{\phi}\) will be determined by the formula:

\[ I = i_{\phi}\frac{R+W}{r}. \]

If \(W\) is sufficiently small and may be neglected in comparison with the resistance \(R\), then the photocurrent in the external circuit will be denoted by \(i_0\) and called the short-circuit current. Then

\[ i_0 = I\frac{r}{R}. \]

The other limiting case of operation of the photocell corresponds to \(W=\infty\)—the case of an open external circuit, or no-load operation. We shall denote the open-circuit voltage by \(E_0\). Since in the latter case \(i_{\phi}=0\), it follows that \(I=i\) and \(E_0=Ir\). It is not difficult to see that the resistance of the photocell \(R\) can be obtained from this by dividing the open-circuit electromotive force by the short-circuit photocurrent:

\[ R=\frac{E_0}{i_0}. \]

The principal characteristic of a photocell is its integral sensitivity \(A\), which represents the ratio of the electron current \(I\) to the total luminous flux \(F\) incident on the photocell,

\[ A=\frac{I}{F}, \quad \text{or} \quad I=AF. \]

In addition, as we see,

\[ E_0=AFr \]

and

\[ i_0=AF\frac{r}{r_1+r_2+r}. \]

In Fig. 5 are shown the dependences of \(E_0\) and \(i_0\) on illuminance for two photocells made by us. The solid lines refer to one photocell, and the dotted lines to the other.

In order that the photocell be sufficiently stable, i.e., change little with time, it is necessary that the quantities,

determining the open-circuit voltage and the short-circuit current would retain their value. For most photocells exposed for a long period of time (months) to the free action of atmospheric air, a considerable decrease in \(E_0\) and a slight decrease in \(i_0\) are observed (Fig. 6, solid lines).

Fig. 5. Dependences of the open-circuit e.m.f. \((E_0)\) and the short-circuit current \((i_0)\) on illumination for two photocells manufactured at GOI.

From the formulas for \(E_0\) and \(i_0\) it may be concluded that such changes can be caused by a decrease in the leakage resistance \(r\) at constant photocell sensitivity \(A\).

A good effect on the stability of the photocell is produced by coating the photocell with a thin and transparent protective layer of lacquer (we used zapon lacquer), especially in the case of photocells for which the application of the gold electrode was carried out in air. These photocells show the greatest degree of constancy over sufficiently long intervals of time (several months), especially with respect to the short-circuit current (Fig. 6, dotted line).

In barrier-type photocells, a considerable temperature dependence of the photoeffect is often observed; under the influence of a changing temperature, at unchanged illumination, both the short-circuit photocurrent \(i_0\) and the open-circuit electromotive force \(E_0\) change. Moreover, for different specimens this dependence proves to be quantitatively different.

As is seen from Fig. 7 (solid curves), at low temperatures of the order of \(-100^\circ\)C there is a fairly rapid increase of \(i_0\) and \(E_0\)

Fig. 6. Change in the e.m.f. of the open circuit (\(E_0\)) and the short-circuit current (\(i_0\)) with time

with increasing temperature. This increase gradually slows down, then stops (which occurs at different temperatures

Fig. 7. Dependence of the e.m.f. of the open circuit (\(E_0\)) and the short-circuit current (\(i_0\)) on temperature

for \(i_0\) and \(E_0\)) and changes into the decline observed at room temperature for most photoelements.

Differences in the properties of individual photocells forced us to seek such cases in which the temperature dependence reaches a minimum or even, if possible, is reduced to zero. It turns out that such cases are observed. In Fig. 7 (dashed line) there is shown the temperature behavior of one of the photocells we made. As can be seen, within the limits from \(-100^\circ\text{C}\) to \(+100^\circ\text{C}\) the short-circuit photocurrent remains practically constant with changing electromotive force. Here the best of the results we obtained is presented, but there is a series of photocells which, even if not so successful, are in any case quite suitable for practical purposes.

A more attentive consideration of the differences in the properties of individual photocells leads to the assumption that the temperature coefficient of a selenium photocell is determined by the ratio between the sum of the resistances \(r_1 + r_2\)—of the gold electrode and the resistance of the semiconducting selenium—and the resistance \(r\) of the blocking layer of the photocell.

The total electronic current \(I\), caused by light falling on the photocell, is divided under short circuit into two parts: \(i\) and \(i_0\). The ratio of these two parts depends on the ratio of the resistances \(r\) and \(r_1 + r_2\). We have already seen that

\[ i_0 = I \frac{1}{1 + \frac{r_1 + r_2}{r}} . \]

Assuming that \(I\) does not depend on temperature, we shall have constancy in two cases: either when \(r_1 + r_2\) changes proportionally to \(r\), which is difficult to expect in view of the completely different nature of these resistances, or when the ratio

\[ \frac{r_1 + r_2}{r} \]

is so small in comparison with unity that even a noticeable change in it will not affect the magnitude of \(i_0\).

Concluding with this account of the properties of selenium photocells with a blocking layer, let us now turn to those experiments which we carried out with a view to applying these photocells to light measurements.

The rather good agreement existing between the spectral sensitivity of the selenium photocell and that of the average human eye immediately prompts an attempt to replace the eye by a photocell, at least for the simplest sources with a continuous spectrum. In this case, what is essential is the influence that the difference between the sensitivity curves will have on the results of the measurements.

To clarify the question, we proceeded as follows.

A single tungsten lamp was placed in an Ulbricht sphere,* and, by changing its supply conditions, radiations were obtained corresponding to various color temperatures between the extreme limits of 2300°K and 2800°K.

Table 1 gives the results of measurements made visually and with the aid of two photocells. The value of the luminous fluxes for the lowest temperature was taken as common.

TABLE 1

As is evident from the table, one of the photocells gave results fairly close to the visual ones, whereas the other regularly increased its errors as the temperature of the source rose, systematically underestimating the values of the luminous fluxes. For the highest temperature the discrepancies reached 20%, which, of course, cannot be considered at all satisfactory.

The reason for such deviations is the exaggerated, in comparison with the eye, sensitivity of photocells in the red and infrared regions of the spectrum; moreover, this sensitivity differs for different photocells (Fig. 3). Therefore it was quite natural that the question arose before us of some measures for dealing with this, apart from the selection of random photocells possessing the required properties. These measures were all the more necessary because, in addition to purely technical aims, we also have in mind more accurate measurements.

The obvious method of using a specially selected light filter proved insufficient. The point is that, in order to fit the photocell curve in the visible part of the spectrum, we arrive at a green filter. But we were unable to obtain a green filter that would not transmit in the infrared part of the spectrum. Therefore, in combination with photocells having appreciable infrared sensitivity, a green filter may even worsen the results. An attempt to apply the method proposed

* An Ulbricht sphere is an instrument intended for measuring luminous fluxes and consisting of a hollow sphere whose walls are covered with an even layer of white matte paint, into which the source to be measured is placed.

Dresler’s method[^15], consisting in covering different parts of the photoelement with different filters and eliminating fluctuations in the properties of an individual photoelement by individually adjusting the parts of its surface covered by one or another filter, also ended unsuccessfully because the corresponding filters were not available to us. We then conceived the idea of overcoming the difficulty of obtaining filters that are highly transparent in the visible part of the spectrum and do not transmit infrared rays in the following way.

We took a filter possessing the opposite properties, namely: opaque in the visible part of the spectrum and transmitting the far red and infrared part of the spectrum, and used it to cover a second photoelement placed next to the main one. Illuminating both photoelements simultaneously, we pass the currents arising in them through two windings of one galvanometer, so that their actions are subtracted. By suitably selecting the sensitivities of the windings, we obtained results that were considerably more favorable. They are given in Table 2, which shows that under the previous conditions

TABLE 2

this system gives discrepancies with visual measurements of only 3%, i.e. the discrepancies between the visual and objective measurements already lie within the limits of error (Table 2).

Finally, quite recently, by changing the process of manufacturing the photoelements, we succeeded in obtaining a sensitivity curve of the photoelement itself (solid line in Fig. 8) that is considerably closer to the eye curve, at least in the red part of the spectrum. As for its blue-violet part, the values obtained from the curve are undoubtedly too large here because of the presence of a relatively large amount of scattered light.

Among the instruments based on the use of the selenium photoelement, the objective luxmeter undoubtedly occupies first place. Being a combination of a photoelement and a sufficiently sensitive pointer galvanometer equipped with a scale graduated directly in lux, the objective luxmeter is a very convenient and simple instrument to use, pro-

[[unclear: beginning of word]] which has gained enormous distribution in the West. It is enough to look through the advertisements in lighting-engineering journals to be convinced that the objective luxmeter will soon be almost as commonly used an instrument as the thermometer.

The requirements imposed on the photoelement of a luxmeter are not as great as the requirements for a laboratory instrument. Here a fairly considerable departure in spectral properties from the properties of the eye is permissible for two reasons: first, an accuracy

Fig. 8. Spectral sensitivity of selenium–sernistic photoelements

of 10–15% is quite sufficient for most technical purposes, and, secondly, correction coefficients can be introduced for different types of radiation, a table of which can be supplied with the instrument. Some foreign firms follow this latter course. In this case these coefficients may deviate quite strongly from unity.

An additional requirement which the objective luxmeter must satisfy is the proportionality of the instrument readings to the cosine of the angle of incidence of a parallel beam of rays. An analogous requirement is also imposed on the test plate of a visual luxmeter, and it is never fulfilled exactly. If the mounting of the photoelement does not cast strong shadows on the light-sensitive surface, then photoelements satisfy this requirement fairly well. If for some reason it is desirable to approximate the exact law to a greater degree, then this can be accom-

late, by creating above the photocell a special hood that intensifies the action of rays falling at a very large angle onto the surface[^16].

The sensitivity of a photocell itself, equal to from \(3 \cdot 10^{-8}\) to \(5 \cdot 10^{-8}\ \mathrm{A/lx\ cm^2}\), is quite sufficient for measuring illuminations, with a pointer instrument of sensitivity \(10^{-7}\ \mathrm{A}\), beginning with \(0.1\ \mathrm{lx}\) for a surface of \(25\ \mathrm{cm^2}\). If it is not required to measure such small illuminations, then the sensitivity of the instrument may be reduced. The need for objective luxmeters in our Union is undoubtedly great, and, strictly speaking, there are no fundamental obstacles to beginning the corresponding production. We make photocells (besides the Optical Institute, they are manufactured by the TsRL and the Electrotechnical Institute, and also by a number of laboratories), and galvanometers too, but there is still as yet no production of objective luxmeters.

A very small number of luxmeters was issued by the Optical Institute, and with respect to some of them we have information that they are working properly.

Besides the objective luxmeter, the photometric sector of the GOI is also engaged in other applications of selenium photocells. In particular, we are developing compensation circuits for connecting photocells, which make it possible to carry out relative measurements without strictly controlling the constancy of the voltage on the lamp.

At present we have worked out the conditions for the functioning of the dc circuit shown in Fig. 9, where \(\Phi_1\) and \(\Phi_2\) are two photocells connected in opposition to one another through a galvanometer \(G\). Both photocells are illuminated by one light source, and when the incandescence of the lamp changes the readings of the galvanometer should not change. To overcome the differences between the photocells that interfere with the operation of the circuit, we have successfully applied the method of shunting the photocells by external resistances, small in comparison with the resistance of the photocell. With a suitable selection of photocells and resistances, the circuit acquires sufficient stability, without losing too much of its sensitivity.

Fig. 9. Compensation circuit (dc) for connecting two selenium photocells

It seems to us that circuits of this kind may be successfully applied to create a series of new types of objective instruments for various light measurements, for example, for measuring transmission and reflection coefficients. An installation assembled according to this principle for measuring transparency with a galvanometer of sensitivity \(10^{-9}\ \mathrm{A}\) gave a zero displacement within \(6\ \mathrm{mm}\) when the voltage on the lamp was changed by \(50\%\) and the devia-

…a decrease of 200 mµ when ordinary glass is placed in one of the beams*.

References Cited

1. K. S. Gibson, Bureau of Stand. Bull. No. 349, 1919.
2. A. Hardy, J. O. S. A., 25, 305, 1935.
3. W. F. Little and H. J. Eckweiler, Trans. Ill. Eng. Soc., 26, 810, 1931.
4. C. L. Dows and G. R. Baumgartner, Trans. Ill. Eng. Soc., 30, 476, 1935.
5. W. W. Coblentz and W. B. Emerson, Bureau of Stand. Bull., 14, 1918–1919.
6. L. S. Ornstein, Rev. Optique, 12, 390, 1933.
7. L. S. Ornstein, Rev. Optique, 12, 385, 1933.
8. H. T. Wensel, Wm. F. Roeser, L. E. Barrow and F. R. Caldwell, Bureau of Stand. Journ. Res., 6, 1103, No. 6, 1931.
9. E. Aschkinass, Ann. Phys. u. Chem. N. F., 55, 401, 1895.
10. E. Putseiko, Journal of Technical Physics, 4, 1163–1169, 1934.
11. E. Putseiko, Lighting Engineering, No. 5, 16–18, 1934.
12. M. Gurevich and E. Putseiko, Journal of Technical Physics, 4, 748–767, 1934.
13. S. Freibert, Journal of Technical Physics, 5, issue 10, 1935.
14. E. Putseiko, Temperature Coefficient of a Selenium Photoelectric Cell, Journal of Technical Physics, 7, 10, 1937.
15. A. Dresler, E. T. Z. 54, 476, 1933.
16. V. B. Weinberg, Proceedings of the All-Union Electrotechnical Association, 4, 28–38, 1935.

* All the results listed above concerning selenium photoelectric cells were obtained by the following staff members of the Photometric Laboratory of the State Optical Institute: group leader E. K. Putseiko, postgraduate student of the State Optical Institute S. I. Freibert, and staff members N. B. Berdnikov, B. I. Kamenetsky, L. N. Meyer, and V. I. Rutkovsky.