PHYSICS AND TECHNOLOGY OF LUMINESCENT LAMPS

V. A. Fabrikant

“I do not know whether inventors could devise anything better than a candle that would burn without the aid of tongs.”

—Goethe

1. INTRODUCTION

In 1938, at the World’s Fair in New York, new light sources, called by Americans fluorescent, or, in abbreviated form—

Fig. 1. External appearance of a luminescent lamp.

F-lamps1, were used for lighting the Soviet and other pavilions. It is more correct to call these lamps luminescent, since they make use of the glow of phosphors. In their outward appearance, luminescent lamps differ sharply from incandescent lamps and consist of long, uniformly glowing tubes with four contacts in the form of pins at the ends (Fig. 1). The production of luminescent lamps, despite the war—or rather, even thanks to the war—began to grow extremely rapidly: in 1938, 200,000 units; in 1939, 1,600,000 units; in 1940, 7,100,000 units; in 1941, 19,000,000 units; in 1942, 30,000,000 units; and in 1943, 40,000,000 units.

A specialized plant was built for the manufacture of luminescent lamps, with a design capacity of 200,000 lamps per day. Dozens of firms took up the manufacture of accessories for luminescent lamps. In general, there is undoubtedly talk of the birth of a new major branch of industry and of a genuine revolution in the field of lighting technology.

In our Soviet Union, the first luminescent lamps of the modern type were built at the beginning of 1940 by staff members of the All-Union Electrotechnical Institute (VEI), F. A. Butaeva and V. I. Dolgopolov.

At present, the All-Union Electrotechnical Institute, the Moscow Electric-Lamp Plant, and the Institute of Physics of the Academy of Sciences of the USSR are carrying out a number of works connected with the deployment of production of luminescent lamps.

The purpose of the present survey is to give as complete an idea as possible of the properties of luminescent lamps and, above all, to analyze all the basic physical processes that determine these properties. We shall try to trace how the microcharacteristics of the individual elements of a luminescent lamp are connected with such macroscopic characteristics of lamps as luminous efficacy.

Of course, before speaking about luminescent lamps, it is necessary to recall the reasons that made it necessary for new light sources to appear in place of incandescent lamps. It is also necessary to consider in somewhat greater detail the question of the ideal light source. It is clear that there is and can be no universal ideal light source, but, on the other hand, “if you do not know to what harbor to steer, there can be no favorable wind” (Seneca).

2. SHORTCOMINGS OF INCANDESCENT LAMPS

First of all, let us dwell briefly on the reasons that make the incandescent lamp an irrational light source in need of replacement. The principal shortcoming of the incandescent lamp, as is well known, is its low efficiency. Table 1 gives data on the energy balance in 100 W incandescent lamps with a service life of 1000 hours². In Table 1, by visible radiation

Table 1

Energy balance in incandescent lamps

Energy distribution	Vacuum lamp—straight filament	Lamp with argon—ordinary coil	Bispiral lamp with argon	Lamp with ordinary coil—krypton-xenon
Visible radiation . .	7	10	12	13
Invisible radiation	86	68	74	76
Losses in supports	7	3	2	2
Losses in gas . .	0	19	12	9
Total . . . .	100	100	100	100

is understood as the region of the spectrum, approximately from 0.40 to 0.75 μ. Invisible radiation consists essentially almost entirely of infrared rays. Thus, the incandescent lamp is very economical as a source of infrared rays. This latter circumstance has led to the wide use of incandescent lamps for drying various objects.^3 However, it follows from Table 1 that, as a source of visible light, the incandescent lamp is considerably less economical. It is important to note here that the luminous efficiency of an incandescent lamp is incomparably lower than the 12% corresponding to the fraction of energy falling within the visible part of the spectrum. Let us recall that the luminous efficiency is the ratio of the luminous flux supplied by a light source to the maximum value of the luminous flux that can be obtained with the same expenditure of power.^2 In determining the luminous efficiency, a very essential role is played by the course of the curve of the spectral sensitivity of the eye within the visible spectrum. This curve has a sharp maximum for the wavelength 0.556 μ. The visibility coefficient for 0.556 μ is taken to be equal to unity. The greatest luminous flux is obtained, naturally, when all the energy is converted into radiation of this wavelength; in that case each watt of energy is converted into 621 lumens of luminous flux. If we compare with this figure the luminous efficacy of a 100-watt incandescent lamp, 15.5 lm/W, we obtain a luminous efficiency equal to only 2.5%!

3. OPTICAL CHARACTERISTICS OF AN “IDEAL” LIGHT SOURCE

From the standpoint of economy, the ideal source would be a source of monochromatic green light with wavelength 0.556 μ. However, the wide application of such a light source would encounter serious difficulties. The point is that what is important to us is not only the quantity of light supplied by the source, but also its quality, i.e. its spectral composition. Ordinary gas-discharge lamps, especially sodium lamps, have not found sufficiently wide application precisely because of the unsatisfactory spectral composition of the light they emit. In monochromatic light, all distinction between the colors of illuminated objects disappears. If the light is not monochromatic but consists of a small number of spectral lines, then a strong distortion of the visible colors of objects occurs. The human eye has adapted to the color relations that arise when objects are illuminated by daylight or sunlight. Experience also shows that the human eye is, to a certain extent, accustomed to the distortion of colors when objects are illuminated by the yellow-red light of incandescent lamps. However, in a number of cases color distortion under illumination by incandescent lamps is unacceptable. When exact distinction of colors is important, there arises the complex problem of artificial daylight (the textile industry, printing, museums, etc.). The standard of daylight adopted is source C with spectral-

with an energy distribution close to the energy distribution in the spectrum of an absolutely black body at \(6500^\circ\) K (Fig. 2)\(^4\). It must be emphasized that, in order to create artificial daylight, what is needed is a source close to source \(C\) in its spectral characteristic, and not merely in its color. Two sources that are close in color and represented in Maxwell’s color triangle by nearby points may differ very sharply in their spectral properties. For example, a source emitting two spectral lines corresponding to complementary colors (red, green), and a source emitting a continuous spectrum. When a white surface is illuminated by such sources, in both cases the surface will preserve its color, but as soon as we illuminate with them any multicolored objects (human faces, colored fabrics, pictures, etc.), sharp differences in the appearance of these objects will immediately arise. The poverty of the spectrum of the line source will appear in the falling out of some colors and the unnaturally high brightness of others. Therefore, for a correct evaluation of the color properties of a light source, it is necessary either to measure the distribution of energy in the spectrum, or, what is more practical, to evaluate the color rendering of a number of standard color samples when illuminated by the light source under investigation\(^5\).

Fig. 2. Energy distribution in the spectrum of source \(C\).

For a visual representation of distortions in color rendering, it is most convenient to use an equal-contrast diagram constructed according to Judd\(^6\). On this diagram, unlike the usual Maxwell triangle, one and the same distance between two points corresponds to one and the same difference in color, regardless of the region in which these points lie. Fig. 3 shows an equal-contrast diagram. Along the edges are plotted the monochromatic colors of the spectrum. The straight line connecting 700 and 400 \(m\mu\) corresponds to purples. The points \(C_1\)—white, \(C_2\)—yellow, \(C_7\)—red, \(C_8\)—crimson, \(C_{13}\)—blue, \(C_{15}\)—light blue, and \(C_{22}\)—green—are samples under illumination by the standard daylight source \(C\). The points \(H\) depict the colors of the same samples under illumination by an incandescent lamp, the points \(P\)—by a mercury lamp, and the points \(P + H\)—by a combination of an incandescent lamp and a mercury lamp. We see that under illumination by an incandescent lamp the white sample becomes orange (is displaced toward 610 \(m\mu\)); under illumination by a mercury lamp it becomes greenish-blue (is displaced toward 470 \(m\mu\)). The yellow sample \(C_2\), under the incandescent lamp, reddens, while under the mercury ...

green. No less significant are the changes in the color of the other samples.

In the ideal case, the difference in the color characteristics of the samples when illuminated by the standard and by the given light source should not exceed, for all samples, 1–2 thresholds (a threshold being the smallest difference in color detected by the eye).

If we now return to the question of the limiting luminous efficacy of a light source, already taking into account the formulated requirements for the quality of the light, then we shall obtain figures considerably lower than 621 lm/W. Let us imagine a light source that converts all the power supplied to it into radiation with wavelengths lying only within the visible part of the spectrum, and let the distribution of energy in this part of the spectrum be the same as that of an absolutely black body (Fig. 4). Then

Fig. 3. Equal-contrast chart according to Judd.

Fig. 4. Energy distribution in the spectrum of an “ideal” light source.

for such an idealized source one can calculate the luminous efficacy, which will be different for different values of the temperature of the absolutely black body, i.e., in essence, for different color temperatures of the light source.

Figure 5 shows the course of luminous efficacy with color temperature1. We see that the maximum luminous efficacy is approximately 220 lm/W, i.e., almost 3 times less than the maximum luminous efficacy for a monochromatic light source.

It is clear that the values of luminous efficacy plotted in Fig. 5 depend strongly on the setting of the limits of the visible spectrum (0.4–0.7; 0.4—

0.75; 0.4–0.8; 0.38–0.76, etc.). The curve in Fig. 5 corresponds to limits of 0.4–0.76 μ.

Fig. 5. Luminous efficacy as a function of the color temperature of an “ideal” light source.

The course of the curve in Fig. 5 shows that changes in color temperature from 2500 to 8000°K have little effect on luminous efficacy, i.e., we may vary the color properties of the idealized source quite strongly without changing its luminous efficacy.

For most practical applications it is reasonable to regard as ideal a source with the properties described above, which corresponds to a limiting luminous efficacy of the order of 200 lm/W. In those cases, however, where the monochromaticity of the light is an advantage (illumination of highways) or where color rendering is not essential (illumination of large workshops in the metalworking industry, outdoor lighting, etc.), the fundamental limit for the luminous efficacy of the source must be considerably higher.

In any case, the incandescent lamp is very far from the ideal light source both in efficiency and in its spectral properties.

4. SHORTCOMINGS OF GAS-DISCHARGE LAMPS

Until 1938, the development of new light sources proceeded mainly along the line of creating gas-discharge lamps using the visible glow of a gas discharge. Here a number of serious results were obtained in the sense of increasing efficiency and brightness, but with respect to color characteristics matters remained very unsatisfactory throughout. High luminous efficacy was achieved chiefly at the expense of the limiting selectivity of the light sources, i.e., owing to the line structure of their spectra (the limiting case being the monochromatic sodium lamp).

As we indicated above, the line character of the spectrum is extremely unfavorable from the point of view of color rendering. In the glow of a gas discharge a continuous spectrum can be obtained only at a very high degree of ionization of the gas, when the processes of recombination of free electrons with ions begin to play the main role. Moller[^8] made an interesting calculation showing that under such conditions a discharge in cesium vapor may prove to be a source of white light with a luminous efficacy of the order of 180 lm/W. However, the construction of a lamp with cesium vapor, operating for a long time at high temperature, presents enormous purely technological difficulties.

Cesium is extremely chemically active and rapidly destroys the walls of a discharge tube. Likewise, attempts to use molecular gases2 are still, in practice, of little interest. Another, more realistic but less effective path for obtaining a light source with the desired color properties consists in combining gas-discharge lamps (mercury lamps) with incandescent lamps. Installations of this type were fairly widely used in the USA for lighting schools and other public buildings. However, this solution to the problem is, of course, associated with a great loss in economy and, most importantly, still does not provide good color rendering. In the graph of Fig. 3 the points \(P+H\) depict the colors of samples under illumination by a combined light source. According to the laws of mixing, the points corresponding to the colors of test objects under combined illumination must lie on the straight lines connecting the points \(P\) and \(H\). Since these straight lines do not pass through the points \(C\), corresponding to the correct colors of the test objects, it is clear that no combination (in the sense of power ratio) of a mercury lamp and incandescent lamps can provide correct color rendering.

5. COMBINATION OF LUMINOPHORES WITH A GAS DISCHARGE

A far more fruitful path has proved to be that associated with the use of luminophores in gas-discharge lamps. As is known, the emission spectra of solid luminophores are fairly sharply bounded, but continuous, bands. Thus, luminophores possess spectral properties analogous to the properties of the ideal light source of which we spoke above. The emission of luminophores is concentrated in a definite spectral region, but within this region all wavelengths are present.

It must be said that the idea of using luminophores in combination with a gas discharge is of considerable age. As early as Becquerel observed the glow of luminophores in discharge tubes3, and more than 30 years ago Cooper-Hewitt conducted experiments with fluorescent reflectors intended to add red light to the green glow of a mercury lamp4. Later, over a number of years, successful experiments on the use of luminophores were carried out in France by Claude and his collaborators5 and by Randall in England6.

But, as we have already emphasized at the beginning, the practical application of luminescent lamps acquired its real scope only beginning in 1938. The principle of operation of gas-discharge lamps with luminophores is extremely simple. A thin layer of some luminophore, glowing under the action of the radiation of a gas discharge, is applied to the wall of the lamp bulb. In this case, the ultraviolet part of the discharge-emission spectrum is usually used to excite the luminophore. In the case where the luminophore is used only to correct the chromaticity of the lamp, the visible radiation of the discharge plays a very substantial role. For example, when correcting the chromaticity of high-pressure mercury lamps, zinc-cadmium sulfide, applied to the in-

the inner surface of the outer bulb gives a negligible fraction of the lamp’s total luminous flux. In this case the luminous efficacy is, in essence, determined entirely by the visible radiation of the discharge. It is clear that the presence of bright lines in the emission spectrum greatly spoils the color characteristics of the lamp. Therefore the use of phosphors to correct the color rendering of gas-discharge lamps having a high luminous efficacy of the discharge itself is a difficult problem. Up to now, only very limited results have been obtained in this direction. The matter is simpler in the case where the visible radiation of the discharge itself is very weak in comparison with the ultraviolet radiation. Here conditions can be created under which almost the entire luminous flux of the lamp will consist of the glow of the phosphor. The role of the gas discharge is reduced only to the generation of ultraviolet radiation that excites the phosphor. Lamps of this type are called fluorescent lamps.

Fig. 6. Section of a fluorescent lamp. The inner surface is coated with phosphor.

In fluorescent lamps, which have become widespread, an electric discharge is used in a mixture of mercury vapor with argon, and a mixture of luminescent silicates, tungstates, and borates is used (Fig. 6). We shall successively consider the individual elements of fluorescent lamps and their interaction with one another.

6. ENERGY BALANCE OF THE GAS DISCHARGE

The output of radiation in a gas discharge depends extremely strongly on such parameters as the gas pressure and the current intensity. It is essential here to distinguish between resonance radiation, emitted by gas atoms in transitions to the lowest energy (normal) level, and radiation emitted in transitions between higher levels. If we roughly divide all the levels of the atom into three “stories” (Fig. 7), then the most intense resonance lines correspond to transitions from the second story to the first, while the most intense nonresonance lines correspond to transitions from the third story to the second. Figure 8a shows, in a very schematic form, the energy balance in the positive column of a gas discharge at various pressures, calculated on the basis of extensive experimental material by Klarfeld7. Losses at the walls consist of energy given to the walls by ions and metastable atoms; heat losses in the volume are connected with the heating of the gas due to elastic collisions of electrons with atoms. Losses at the walls decrease with increasing pressure, owing to the shortening of the mean free path of the ions, which makes it more difficult for the ions to reach the walls.

The fraction of the energy expended on thermal losses in the volume of the gas has a sharp maximum in the region of intermediate pressures. The heat released in this process, owing to the thermal conductivity of the gas, is also transferred to the walls of the discharge tube.

It is interesting to note that in the region of high pressures \(\eta_b\) drops sharply. This is connected with the disappearance of the difference between the electron temperature and the gas temperature and with the establishment, along the axis of the discharge, of very high temperatures, of the order of \(7000^\circ\).[^15]

Thermal losses in the volume may be regarded as a peculiar process of heat transfer from the hot electron gas, “heated” by the electric field of the discharge, to the colder atomic gas, on which the electric field does not act directly.

Fig. 7. Simplified scheme of the energy levels of the mercury atom.

From the diagram in Fig. 8a it is seen that there are two pressure regions optimal for the generation of radiation in the discharge.

In the region of low pressures the discharge is especially effective as a source of resonance radiation; in the region of high pressures, on the contrary, the yield of nonresonance lines is especially large.

In what follows we shall be interested only in the low-pressure discharge. Let us briefly consider the mechanism of excitation of radiation in such a discharge.

7. MECHANISM OF EXCITATION OF RADIATION IN THE DISCHARGE

As is known, the scheme of the basic energy transformations in a low-pressure discharge is fairly simple. An electron acquires a reserve of kinetic energy in the electric field of the discharge, then collides with an atom and excites it, i.e., transfers the atom from a lower energy level to a higher one. After a certain very short interval of time has elapsed (the lifetime of the excited atom is \(10^{-7}\)—\(10^{-8}\) sec.), the excited atom returns back to a lower energy level, emitting in the process a quantum of radiant energy. However, this

the simple scheme is complicated by a number of secondary processes, on which we shall dwell somewhat later.

It is obvious that only electrons possessing an energy greater than the excitation potential, equal to the distance between the energy levels of the atom expressed in electron-volts, can take part in the excitation process. The distribution of electrons over energies has a Maxwellian character, with the electron temperature as a parameter. It is customary to express the electron temperature in electron-volts—1V corresponds to 7800° K[^16].

Fig. 8a. Energy balance in the positive column of a gas discharge.

$\eta_{el}$ — fraction of the power converted into resonant radiation; $\eta_t$ — nonresonant radiation; $\eta_p$ — heat losses in the volume of the gas; $\eta_s$ — losses at the walls.

Fig. 8b. Intensity of the 1850 and 2537 Å lines.

The maximum of the Maxwellian curve corresponds to an electron energy equal to one third of the electron temperature, expressed in electron-volts. In a discharge the maximum of the distribution curve almost always lies considerably to the left of the excitation potential, so that only a small fraction of all the electrons participates in the excitation of atoms. The lower the electron temperature, the smaller this quantity.

With increasing pressure, at constant discharge-current strength, the concentration of normal atoms and electrons increases (in connection with the decrease in drift velocity), and the electron temperature falls. As a result of the superposition of these mutually opposite effects, at a certain pressure a maximum is observed in the product of the concentration of normal atoms and the number of electrons possessing energy sufficient to excite these atoms. This must correspond to a maximum of the radiation of the discharge in the low-pressure region.

The absolute intensity of the radiation depends very strongly on the magnitude of the effective cross section of the atom for the process of excitation by electron impact. The effective cross section of the atom has different values for transitions between different levels and is a function of the energy of the exciting electron—the “excitation function.” Figure 9 gives the excitation functions for four levels of the mercury atom[^18]. All these levels belong to the “second”

“floor” of mercury-atom levels. From the \(6^1P_1\) level begins the line 1849 Å, and from the \(6^3P_1\) level—the line 2537 Å; the \(6^3P_2\) and \(6^3P_0\) levels are metastable.

The absolute value of the effective cross section of a mercury atom at the maximum of the excitation function of the \(6^1P_1\) level is equal to \(15\ \text{cm}^2/\text{cm}^3\) (the area of all targets at \(0^\circ\)C and \(1\) mm Hg—a quantity reciprocal to the mean free path), which amounts to approximately 15% of the gas-kinetic cross section of the mercury atom; for the \(6^3P_1\) level, correspondingly, \(4\ \text{cm}^2/\text{cm}^3\) \(^{19}\). It should be pointed out that the data available in the literature for the \(6^3P_1\) level are contradictory, but after the introduction of the corresponding corrections they agree fairly well with one another \(^{20}\). In atoms of inert gases the cross sections are much smaller; for example, in neon the effective excitation cross section amounts to fractions of a percent of the gas-kinetic cross section \(^{21}\). Conversely, in such atoms as sodium and potassium, the efficiency of collisions is still greater than in mercury. Roughly speaking, the probability of excitation decreases with increasing excitation potential.

Fig. 9. Excitation functions of the mercury atom.

In the absence of secondary processes, the effective cross section of the atom is the only atomic characteristic that influences the intensity of radiation of a gas discharge. The connection between the macroscopic characteristics of the discharge and the atomic constants can be established only on the condition that the electronic characteristics of the discharge are known, i.e., the electron concentration and the electron temperature.

For a discharge in sodium vapor there are the most complete data on the electronic and optical characteristics \(^{22}\). For this discharge, excellent agreement is obtained between the values of absolute intensities calculated from atomic constants and the experimental data \(^{23}\).

Unfortunately, the mercury discharge has been investigated much less thoroughly. In works devoted to absolute measurements of the intensity of resonance lines, not everything is satisfactory from the methodological point of view. There are no direct data on the intensity of the 1850 Å line. The maximum intensity of resonance radiation is obtained in the pressure region of the order of \(10^{-2}\) mm Hg, which is in qualitative agreement with theoretical calculations \(^{24}\). In Fig. 8b the course of the intensity of the 1850 and 2537 Å lines, calculated theoretically, is shown. A more exact comparison is difficult, since there is almost no information

on the concentrations of electrons and electron temperatures for the same discharge conditions under which the radiation measurements were made. For example, in the work of Found and Hennelly[^25] there are detailed data on the electronic characteristics, but the results of measurements of the 2537 Å line are of a relative nature.

8. DISCHARGE IN LUMINESCENT LAMPS AS A SOURCE OF ULTRAVIOLET RADIATION

Until now we have been speaking of the radiation of a discharge in pure mercury vapor. For a number of subsidiary reasons, which will be discussed below, luminescent lamps use a discharge in a mixture of mercury vapor with argon. The argon pressure is of the order of 3–4 mm Hg. The role of argon in the mechanism of discharge radiation has not yet been fully clarified. In any case, the presence of argon hinders the motion of electrons and ions, since it causes a decrease in the mean free path. The latter should lead to an increase in the electron concentration (at constant current) and to a lowering of the electron temperature.

When 4 mm of argon is added, according to the measurements of F. Butaeva,[^26] the following changes occur in the discharge: the electron concentration increases by an order of magnitude from \(9 \cdot 10^{10}\) to \(2.6 \cdot 10^{11}\), while the electron temperature decreases from 18,100 to 13,000°K. The temperature of the lamp walls was maintained constant and equal to 40°C, which corresponds to a mercury-vapor pressure of the order of \(10^{-2}\) mm Hg. The current strength was also constant—350 mA. The decrease in the electron temperature should change the ratio of the intensities of the 1850 and 2537 lines in favor of the latter.

Figure 15 shows the radiation output as a function of the wall temperature of a luminescent lamp.[^27] We see that 40°C (110°F) corresponds to the optimum conditions for generating resonance radiation.

Table 2 gives the data of Teele and Burns[^28] for the energy distribution in the spectrum of the gas discharge of lamps 25 mm in diameter.

Table 2

Output of spectral lines in luminescent lamps

Current, A	Wall temperature, °C	Fraction of power converted into line \(\lambda\): 2537	Fraction of power converted into line \(\lambda\): 3129	Fraction of power converted into line \(\lambda\): 3651	Fraction of power converted into line \(\lambda\): 4047	Fraction of power converted into line \(\lambda\): 4358	Fraction of power converted into line \(\lambda\): 5461	Fraction of power converted into line \(\lambda\): 5780	Sum	Fraction on line 2537	Lumens per watt
0.25	48	62	0.53	0.45	0.59	1.30	0.86	0.18	65.9	94	5.5
0.50	57	55	0.68	0.62	0.86	1.45	1.26	0.27	60.1	92	7.3
1.00	69	40	0.89	0.82	1.02	1.79	1.49	0.39	46.4	86	10.0

at three different current strengths. In Table 2, the power expended in the positive column of the discharge is taken as 100%. It follows from it that, under these conditions, more than 50% of the power supplied to the positive column is converted into the 2537 Å line, and that this line accounts for more than 90% of the radiation of the discharge (the latter is not exact, since the 1850 Å line and the infrared lines were not measured).

These data are in excellent agreement with the general picture of the energy balance of the discharge shown in Fig. 8 at the beginning of the present section. It should be noted, however, that because of the complexity of the energy levels of mercury, a quantitative theoretical explanation of such high values of the yield for the 2537 Å line as those given in Table 2 is far from a trivial problem.

Indeed, as is seen from Fig. 9, the probability of excitation of both metastable levels \(6^3P_0\) and \(6^3P_2\) is approximately twice as large as the probability of excitation of the 2537 Å line, which originates from the \(6^3P_1\) level. Consequently, if more than 50% of the power is expended on excitation of the 2537 Å line, then more than 100% of the power should be expended on excitation of the metastable levels, and considerably more than 150% of the supplied power on excitation of all four lower levels\(^ {29}\). In this respect the situation is quite different in the sodium discharge: there even values of the radiation yield close to 100% find a simple explanation, since there are no metastable states. The contradiction that arises in attempting to interpret theoretically the data of Table 2 can be resolved only by taking into account collisions of the second kind. In collisions of the second kind, metastable mercury atoms either give their energy back to electrons, or pass to the \(6^3P_1\) level, becoming radiating atoms.

It is possible that the latter process plays a significant role in the mechanism of excitation of the 2537 Å line. The presence of argon should considerably increase the number of collisions of the second kind experienced by metastable atoms.

In any case, the data presented indicate that the discharge in fluorescent lamps is a very economical source of short-wavelength ultraviolet radiation. On the contrary, as a source of visible radiation this discharge is disadvantageous. In the last column of Table 2 are given the values of the luminous efficacy of the discharge itself, and we see that these figures are lower even than the luminous efficacy of incandescent lamps.

As has already been pointed out above, the weak intensity of the visible lines is also an advantage of fluorescent lamps. Finally, from Table 2 it is clear how harmful it is to force the lamp operating conditions: the yield of resonance radiation drops sharply and the role of the visible lines increases. We shall return to this question in the section devoted to the choice of the technical parameters of fluorescent lamps.

9. PHOSPHORS USED IN LUMINESCENT LAMPS

In choosing phosphors for luminescent lamps one must take into account a whole series of circumstances. First of all, these phosphors must possess high sensitivity to short-wave ultraviolet. Next, these phosphors must give visible luminescence with quite definite spectral characteristics, in the sense of the position of the luminescence band in the spectrum and the distribution of energy within the band. Finally, phosphors for lamps must be vacuum-resistant, i.e., when introduced into an electrovacuum device such as a luminescent lamp, they must not decompose and must be sufficiently stable in the presence of the mercury discharge. The latter condition greatly complicates the selection of phosphors for lamps. It is often necessary to reject phosphors that are very advantageous in their optical characteristics, since they are not vacuum-resistant in the indicated sense of the word.

At present an assortment of phosphors used in luminescent lamps has already been established. Table 3 gives the principal characteristics of these phosphors according to the work of O’Dea and Cizel[^30].

Table 3

Principal phosphors used in luminescent lamps

Phosphor	Color	Excitation band	Sensitivity maximum	Emission band	Emission maximum
CaWO$_4$	Blue	2200—3000	2720	3800—7000	4400
MgWO$_4$	Blue-white	2200—3200	2850	3800—7200	4800
ZnSiO$_3$	Green	2200—3960	2537	4500—6200	5250
ZnBeSiO$_4$	Yellow-white	2200—3000	2537	4500—7200	5950
CdBO$_3$	Pink	2200—3600	2500	4000—7200	6150

It should be noted that the short-wave boundary of the excitation band, 2200 Å, is determined conventionally, since the apparatus used did not make it possible to advance farther toward the region of shorter waves.

As shown by Bizet’s measurements[^31], carried out with the aid of a vacuum spectrograph, the excitation band of the phosphors extends far beyond 1850 Å; moreover, a second, higher, sensitivity maximum is even observed in the region of 1850 Å (see Fig. 10). Unfortunately, Bizet’s data are somewhat contradictory. He observes a quantitative discrepancy between the results of measurement with a vacuum spectrograph and with a monochromator (for wavelengths \(> 2200\) Å).

F. Butaeva\(^ {32}\) carried out experiments which, it is true, were of a somewhat indirect character, but which pointed quite definitely to the great sensitivity of phosphors to the 1850 Å line. F. Butaeva placed

Fig. 10. Sensitivity of phosphors to the short ultraviolet.

phosphor specimens at a short distance from a low-pressure mercury quartz lamp and covered them with filters transmitting the 2537 Å line (from 33 to 50%) and cutting off the 1850 Å line. It turned out that when they were covered with such filters the brightness of the phosphor decreased by a greater factor than followed from the transmission coefficient for the 2537 Å line. For example, with a filter transmission of 33%, the brightness decreased for some phosphors to approximately 26%, and for others to 18% of the initial value. This indicated that in the excitation of phosphors the 1850 Å line, cut off by the filter, plays a substantial role. As was to be expected, all effects associated with the 1850 Å line disappeared at a sufficiently large distance between the lamp and the phosphor specimen. It is known that air strongly absorbs the 1850 Å line.

Fig. 11. Sensitivity of phosphors in the region of wavelengths greater than 2200 Å.

Figure 11 gives the curves of the spectral sensitivity of phosphors in the region of wavelengths greater than 2200 Å. We see that they all have sensitivity maxima lying close to 2537 Å and, consequently, should be well excited by this spectral line. The emission bands of the phosphors cover, as is seen from Table 3, the entire visible spectrum. In Table 4

Table 4

Energy distribution in luminescence bands.

Phosphor	Wavelength of energy maximum	Intensity relative to maximum	Intensity relative to maximum	Intensity relative to maximum	Intensity relative to maximum	Intensity relative to maximum
		λ 4500	5000	5500	6000	6500 Å
CaWO₄	4400	98	60	24	8	2
MgWO₄	4800	86	97	56	26	9
ZnSiO₃	5250	1	33	48	4	0
ZnBeSiO₃	5950	1	9	68	99	55

quantitative data on the distribution of energy in luminescence bands are given, according to Tetry and Berns.

In Fig. 12 the results of measurements by F. Butaeva of phosphors synthesized at the VEI by V. I. Dolgopolov are shown on the same scale[^34]. Naturally, the properties of phosphors are very strongly

Fig. 12. Distribution of energy in the spectral bands of phosphors.

MgN 23 — magnesium tungstate; Be 44 — zinc-beryllium silicate activated with manganese; Be 858 — the same, but with a different quantitative composition (more manganese, less beryllium); B 75 — cadmium borate.

affected by the technology of their preparation. Contamination of phosphors with heavy metals such as iron, copper, lead, etc., is very dangerous.

In Fig. 13 the luminescence yield is shown as a function of the percentage content of the impurity[^35]. We see that concentrations of the order of \(10^{-5}\)—\(10^{-8}\) are sufficient to cause a noticeable decrease in the yield. This circumstance compels us to impose extremely

high requirements on the purification of the starting materials for the synthesis of phosphors. The activator also plays a major role—a heavy metal that imparts to the phosphor its specific properties. Thus manganese serves as the activator in zinc silicate (willemite) and zinc–beryllium silicate. In general, manganese is a widespread activator of phosphors, and the properties of phosphors activated by manganese are evidently closely connected with the properties of the manganese ion \(^{35}\).

There exists a definite concentration of the activator corresponding to the maximum luminescence yield. In the case of silicates this concentration is of the order of \(10^{-6}\), but the exact value depends strongly on a number of circumstances (composition of the main substance, annealing regime, etc.). We cannot here go into the details of phosphor technology and shall therefore confine ourselves to the following brief remarks.

Fig. 13. Influence of impurities on the luminescence yield for cadmium borate activated with manganese.
Vertical axis: relative yield. Horizontal axis: percentage of impurity.

The theory of solid phosphors, based on concepts of the energy bands of a solid, is still in an embryonic state \(^{37}\). Even if this theory does provide a certain qualitative picture of the mechanism of excitation and emission of phosphors, in any case all questions connected in one way or another with the intensity of luminescence still remain unanswered. This is explained by the great complexity of the object. Therefore the synthesis of new phosphors proceeds by a purely empirical route, and a successful recipe is the secret of one firm or another.

10. THE PHOSPHOR LAYER IN FLUORESCENT LAMPS

The phosphor is applied to the inner walls of the lamp in the form of a layer of finely crystalline powder. The size of the individual crystallites is of the order of a micron. The size of the crystallites has a substantial effect on the luminescence yield. In the optimum case, approximately \(1\text{–}2\) mg of phosphor fall on \(1\ \text{cm}^{2}\) of the lamp surface, which corresponds to \(\sim 10^{9}\) crystallites. The crystallites in this case pile up on one another (pilling) \(^{38}\). The optical properties of lamps depend very strongly on the structure and thickness of the phosphor layer.

If the layer is too thin, the exciting ultraviolet is not completely absorbed; if the layer is too thick, the harmful absorption of the luminescence light in the layer itself already begins to have an effect. The latter is aggravated by the fact that the phosphor layer in lamps operates in transmission, i.e., the glow is observed from the side opposite to that being excited. Roughly speaking, a luminescent lamp may be regarded as a closed fitting made of milk glass. For such a fitting the efficiency is determined, as is well known,^2 by the following simple formula:

\[ \eta=\frac{p}{1-r}, \tag{1} \]

where \(p\) is the transmission coefficient of the layer, and \(r\) is the reflection coefficient of the layer.

In deriving formula (1), repeated reflections of light from the walls are taken into account. If \(p+r=1\), then \(\eta\) is always equal to unity, independently of the thickness of the layer. Conversely, even small deviations of \(p+r\) from unity already cause a noticeable decrease in efficiency. For example, if \(p\) of the layer is equal to \(0.3\), and \(r=0.65\), i.e. \(p+r=0.95\), and the absorption in a single passage through the layer is \(0.05\), then the efficiency determined by (1) already proves to be equal to \(0.86\), i.e. the light losses reach \(0.14\). This is explained by multiple reflections of light in such a closed cavity. It has been established experimentally that in ordinary luminescent lamps \(\eta\) is approximately equal to \(0.9\).

The question of the optimum characteristics of the phosphor layer leads to a rather interesting problem in the optics of scattering media.^39

The technology for obtaining homogeneous layers on the inner walls of long tubes is based on the use of suspensions of phosphors in various liquids. After they are applied, the adhesive substance is removed. There is an enormous number of patents describing various methods of applying phosphors. Here the capillary properties of the liquid medium and the adhesive ability of the phosphor itself play a large role. For this reason, small additions of surface-active substances prove very useful.

11. SELECTION OF THE DIMENSIONS OF LUMINESCENT LAMPS

We now turn to the description of technical specimens of luminescent lamps. As has already been indicated above, up to the present ultraviolet radiation of the positive column of the discharge has been used in luminescent lamps. Therefore luminescent lamps have the form of long tubes with a large distance between the electrodes. Other conditions being equal, the greater the length of the lamp, the higher its luminous efficacy, since then a larger fraction of the total power is expended in the positive column of the discharge, and the smaller the role played by losses at the electrodes (cathode and anode drops).

On Fig. 14 the luminous efficacy of a luminescent lamp is shown as a function of its length.

It is clear that, with increasing length, the power consumed by the lamp increases. The luminous efficacy reaches saturation at approximately a length

Fig. 14. Luminous efficacy of a luminescent lamp as a function of its length.

equal to 120 cm (Fig. 14)^40. The diameter of the lamp is determined by the necessity of obtaining a definite wall temperature, to which the optimal pressure of mercury vapor corresponds. This temperature depends on the power consumed by the lamp and on the dimensions

Fig. 15. Luminous efficacy as a function of the wall temperature of a luminescent lamp.

of the lamp. Too small a lamp diameter leads to an increase in temperature above the permissible limit and thereby to a reduction in the output of ultraviolet radiation, which, in turn, entails a reduction in the luminous efficacy of the lamp. In Fig. 15 is shown

luminous efficacy of the fluorescent lamp as a function of the temperature of the tube walls8.

Table 5 gives the dimensions of low-voltage fluorescent lamps. We see that a 40-watt lamp is a tube more than 1 m long and almost 4 centimeters in diameter. Such large

Table 5

Dimensions of fluorescent lamps

Wattage, watts	6	8	14	15	15	20	30	40	65	100
Length, centimeters	22.5	30	35	45	45	60	90	120	90	150
Diameter, centimeters	1.56	1.56	3.75	2.5	3.75	3.75	2.5	3.75	5.3	5.3

dimensions of fluorescent lamps are in many cases their advantage, but sometimes lead to difficulties (local lighting).

In addition to low-voltage lamps, high-voltage lamps have recently become increasingly widespread9. Their length reaches 250 cm, and their diameter is from 1.5 to 2.5 cm. Finally, fragmentary information has appeared in the American press about the production, in 1944, of spherical lamps.

12. ELECTRODES IN FLUORESCENT LAMPS

Alternating-current fluorescent lamps have two identical electrodes. Each electrode serves as the cathode for one half-period of the current and as the anode for the other half-period. In low-voltage lamps, electrodes in the form of tungsten double spirals coated with oxide paste are used (see Fig. 6). The oxide paste serves to reduce the work function of the electrons. Under normal operation the electrodes are heated by the discharge itself as a result of bombardment (“self-heating”). In order to limit the heating of the electrodes, before the war the Americans soldered onto them nickel “whiskers,” which took upon themselves part of the current during the anode period. During the war, for reasons of nickel economy, these whiskers were eliminated10. The sum of the cathode and anode drops lies within the range from 12 to 18 V11. As was already indicated in the preceding section, the presence of these drops noticeably lowers the luminous efficacy of short lamps. It is curious that the oxide cathode is not equipotential in this case, and a single bright spot of the cathode operates, gradually moving along the cathode over the service life of the lamp. When this spot passes from one end of the double spiral to the other, the lamp ceases to operate. Evidently, in this case the entire

surface of the cathode. As a result of sputtering of the electrodes, ring-shaped black deposits form at the ends of the tube*).

In high-voltage tubes, cold cathodes in the form of cylinders are used. These cylinders are usually made of iron, copper, or other metals treated on the surface with substances such as barium azide. The sum of the cathode and anode voltage drops reaches hundreds of volts, which makes it necessary to make very long and narrow lamps in order to increase the share of the power falling on the positive column of the discharge^42. The service life of cold cathodes is very long, on the order of 10,000 hours.

13. ELECTRICAL CHARACTERISTICS OF FLUORESCENT LAMPS

Like any gas-discharge device, a fluorescent lamp has rather complex electrical characteristics. First, it is necessary to distinguish two modes of lamp operation: the ignition mode and the normal burning mode. Second, a fluorescent lamp has a falling volt-ampere characteristic, i.e., as the current increases, the voltage drop across the lamp decreases.

All this greatly complicates the circuit for connecting a fluorescent lamp (low-voltage—120–220 V). Let us first present the electrical characteristics of lamps in the normal burning mode (Table 6). We see that the voltage drop across the lamps is

Table 6

Electrical characteristics of low-voltage fluorescent lamps

	6	8	14	15	15	20	30	40	45	100
Watts	6	8	14	15	15	20	30	40	45	100
Current, in amperes	0,15	0,18	0,37	0,30	0,33	0,35	0,34	0,41	1,35	1,45
Voltage drop across the lamp	45	54	41	56	48	62	103	108	50	72
Line voltage	110–125	110–125	105–125	110–125	110–125	110–125	220–225	220–250	110–125	220–260

not more than \(50\%\) of the line voltage. This is explained by the fact that the ignition potential of the gas discharge is much higher than the voltage drop across the discharge gap when the discharge is already burning^16. Therefore, with a smaller difference between the line voltage and the voltage drop across the burning lamp, reliable ignition and stable operation of the lamp can no longer be obtained. For when the lamp operates on alternating current, ignition of the discharge must occur every

*) The service life of the cathodes is approximately 3000 hours. Sputtering of the electrodes is reduced because of the presence of argon.

half-period. Figure 16 shows oscillograms of the current and voltage of a luminescent lamp12.

The lamp ignites only when the instantaneous value of the line voltage reaches a certain magnitude, equal to the ignition potential*). After this, the voltage drop across the lamp

Fig. 16. Oscillograms of the current, voltage, and luminous flux of a luminescent lamp. On the left is shown the vector voltage diagram for a circuit with a luminescent lamp.

decreases sharply and remains constant for a certain interval of time, almost until the next extinction of the lamp. The choice of current is determined by the desire, on the one hand, to obtain high efficiency and, on the other hand, by the need to have lamps of a certain power. Figure 17 shows a series of curves of luminous efficacy at different discharge-current strengths[^28]. From these curves it is evident that, as the current increases, the luminous efficacy decreases. The same follows from the data of Table 2. We shall see below that, for the same reason, 65- and 100-watt lamps have reduced luminous efficacy, despite their impressive dimensions. Currents that are too small correspond to low temperatures of the lamp walls and to insufficient power values.

Fig. 17. Luminous efficacy at different current strengths.

Somewhat different relations occur in high-voltage lamps. In them the voltage drop reaches 1–2 kV, and with a current strength of 60–100 mA a power of 15–20 W is already obtained. The narrow diameter of the lamps provides a sufficient wall temperature. It should be pointed out that obtaining currents above 100 mA when using cold cathodes is still, in general, a very difficult problem.

14. STARTING FLUORESCENT LAMPS

The circuit for switching on a low-voltage fluorescent lamp must ensure reliable starting of the lamp and then the normal mode of its operation. A low-voltage lamp is started with the aid of a very ingenious device called a glow-discharge relay[^16]. The point is that, in order to start the lamp, it is first necessary to heat the electrodes. With cold electrodes the starting potential of an ordinary low-voltage lamp reaches almost 1000 V. Conversely, preliminary heating of the electrodes produces strong electron emission, facilitating ignition of the discharge. But as soon as the discharge has been ignited, the heating of the electrodes must be switched off, since the electrodes then begin to be heated by the discharge itself. Both these operations—switching the heating on and off—are performed by the glow-discharge relay.

Fig. 18. Schematic circuit for switching on a fluorescent lamp.

Fig. 19. Glow-discharge relay.

In Fig. 18 the basic circuit for switching on a low-voltage fluorescent lamp is shown (the purpose of the choke will be explained below). Thus, a glow-discharge relay is connected in parallel with the lamp. A similar circuit was once used for sodium lamps, but with a switch instead of a relay. In Fig. 19 this relay is shown. The glow-discharge relay is a miniature discharge tube filled with an inert gas, with electrodes made of bimetallic plates. In the normal state there is a small gap between the plates (see Fig. 19, position a). When a sufficient voltage is applied between the plates, a glow discharge flashes up. The glow discharge causes heating of the electrodes, which bend as a result of this heating toward one another until they touch (see Fig. 19, position b). In this position the relay is, of course, short-circuited, and since the discharge has ceased, the electrodes begin to cool and, consequently, to straighten, moving apart from one another. If the applied voltage remains the same, the glow discharge flashes up again, and everything begins anew. Oscillatory behavior arises. The starting potential of the relay is selected to lie between the line voltage and the starting potential of the fluorescent lamp (with heated electrodes). Therefore, at the first moment after switching on, the discharge flashes up between the electrodes of the relay, but not in the lamp. Both electrodes of the lamp prove to be connected in series in the circuit, but only a very small current flows in this circuit, because of the large internal resistance of the relay.

However, after a very short interval of time (fractions of a second), the bending of the relay plates described above takes place. The resistance of the relay then falls almost to zero, and the current passing through both electrodes rises sharply, reaching several hundred milliamperes. This current is usually almost twice the normal discharge current of the lamp. As a result, the electrodes heat up to a very high temperature. However, this stage too lasts only fractions of a second, because the cooling of the relay electrodes causes them to straighten and the circuit to open. The situation is now quite different from what it was at the beginning. Owing to thermal inertia, the lamp electrodes retain their heat and therefore possess increased electron emission. Of the two parallel discharge gaps—the lamp and the relay—the lamp now has the lower ignition potential, and therefore the discharge flashes over in the lamp and not in the relay. If, for some reason, the lamp has not ignited, the lamp electrodes cool down and everything begins again from the start. The relay will not “lag behind” the lamp until ignition of the discharge in it has occurred. It must be said that this automatic action of the relay, which at first seemed a great advantage, quickly became a major drawback. As soon as any lamp failed (the electrodes became deactivated), a peculiar “signaling” began—the lamp began to “blink.”

In large workshops, where hundreds and thousands of lamps are installed at the same time, one or two lamps may always fail. With such “blinking” they stand out sharply to the eye, producing an unpleasant effect; moreover, during prolonged operation in the “blinking” mode the chokes become strongly overheated. As a result of long operation, new “non-blinking” relays have been produced in the USA. The introduction of these relays was accompanied by noisy advertising (they were called “revolutionizing”), but their construction has not yet been described*). The advantage of high-voltage lamps consists in the absence of special devices for ignition. The high voltage ensures reliable ignition of the discharge.

15. CIRCUIT FOR SWITCHING ON FLUORESCENT LAMPS

Let us turn to the regime of normal burning of the lamp. As has already been pointed out, the voltage drop across the lamp is then approximately half the line voltage. Hence a ballast resistance, absorbing this voltage difference, must be connected in series with the lamp. The ballast resistance, in addition, limits the discharge current of the lamp. Without a ballast resistance, because of the falling characteristic of the discharge[^16], the current in the lamp would begin, immediately after ignition, to grow catastrophically and would quickly destroy the electrodes.

*) There is only a description of an old-type “non-blinking” relay (see the book by Amick[^40]).

It is clear that making the ballast resistance purely ohmic is meaningless, since this would lead to the conversion of approximately half of the supplied power into Joule heat. Therefore, a choke is used as the ballast resistance: a small coil of copper wire wound on an iron core. The specific designs of chokes for luminescent lamps are very diverse and are manufactured by dozens of different firms. The presence of a choke facilitates ignition of the lamps, since, owing to its inductance, at the moment when the relay circuit is broken a strong overvoltage arises.

Fig. 20. Circuit for switching on two luminescent lamps. In the circuit of one of the lamps there is a capacitor, which creates a phase shift between the lamps and improves the power factor.

The improvement of chokes proceeds along the line of reducing power losses and decreasing dimensions. Losses in chokes range from 30 to 16% of the lamp power. The use of chokes causes a sharp deterioration of the power factor (cosine phi) of the installation. For the lamp itself, as follows from the oscillograms in Fig. 16, the power factor is approximately 0.95, whereas for a lamp with a choke it is only 0.55. Therefore it is necessary to further complicate the lamp switching circuit by introducing capacitance to compensate the phase shift. Switching devices for 2, 3, and 4 lamps simultaneously have become most widespread. Fig. 20 shows the circuit of such a device for two lamps. They are more compact and have smaller losses. Table 7 gives the characteristics of the most widespread types of switching devices.

Table 7

Switching devices for luminescent lamps

Type of lamps	Losses in W	Power factor	Note
For one 15-watt lamp . .	4.5	0.55	Choke
» 15- » . .	4.5	0.90–0.95	Choke with capacitor
» two 15-watt lamps . .	9.0	0.55–1.00
» one 20-watt lamp . .	4.5	0.90–0.95	Choke with capacitor
» two 20-watt lamps . .	9.0	0.95–1.00
» one 40-watt lamp . .	12	0.90–0.95	Choke with capacitor
» two 40-watt lamps . .	14.5	0.95–1.00
» four 40- » . .	31.0	0.95–1.00
» one 100-watt lamp .	35.0	0.95–1.00
» four 100-watt lamps	64.0	0.95–1.00

It is interesting to note that for more powerful lamps, starting devices for two and four lamps have considerably smaller losses (relative) than devices for single lamps. For example, for a 100-watt lamp, 35 and 16 W (6434).

The presence of four contacts in a low-voltage fluorescent lamp makes it necessary to use special sockets, and for each lamp 2 sockets are needed. Near one of the sockets, a glow-discharge relay is mounted in a special socket. For high-voltage lamps, special sockets are also used, but of a simpler design.

Fig. 21. Oscillograms of the luminous flux of fluorescent lamps.

a) one lamp, b) two lamps with phase shift, c) three lamps on three phases.

Starting devices for two or more lamps make it possible to eliminate one drawback of fluorescent lamps, since, owing to the insufficient inertia of the phosphors, they produce strong stroboscopic effects. In Fig. 21, on the same scale, curves are shown for the change with time of the luminous flux of incandescent and fluorescent lamps. Table 8 gives quantitative data for the modulation depth of the luminous flux during one period of alternating current[^47].

In Table 8, the oscillations of the luminous flux of a 200-watt incandescent lamp are conventionally taken as unity. Three fluorescent lamps connected to three different phases of three-phase current give, owing to the phase shift, a resultant curve with smaller oscillations than even a 40-watt incandescent lamp. Two lamps are connected according to the circuit of Fig. 20. One lamp is connected only with a choke, the other with a choke and a capacitance, which creates a phase shift of one lamp relative to the other, and both lamps together give a sufficiently smoothed curve. In this case the power factor is close to unity.

Table 8

Oscillations of the luminous flux of lamps during one period of alternating current

Lamp	Oscillations
Incandescent lamp 100	1
Three fluorescent lamps on three phases	3
Incandescent lamp 40	7
Two white fluorescent lamps	9
Two daylight fluorescent lamps	10
One white fluorescent lamp	19
One daylight fluorescent lamp	21

In some special cases (on railway transport), fluorescent lamps are used in direct-current circuits[^48]. Then, naturally, an ohmic resistance has to be introduced, the role of which is usually performed by incandescent lamps. Clearly,

that in these cases the economy of installations with fluorescent lamps proves to be reduced. There are fragmentary indications that fluorescent lamps without ballast resistance have been built in the United States, but there are no data on their economy and service life.

16. LUMINOUS CHARACTERISTICS OF FLUORESCENT LAMPS

Let us consider in more detail the luminous characteristics of fluorescent lamps. In Fig. 22 the entire energy balance of a 40-watt fluorescent lamp is presented in the form of a detailed diagram. From this diagram it follows that in a fluorescent lamp 18.5% of the supplied power is converted into visible light, i.e., the energy efficiency of a fluorescent lamp is approximately one and a half times higher than that of a 100-watt incandescent lamp (see Table 1). However, this increase in energy efficiency is not the only and, perhaps, not even the chief reason for the high economy of fluorescent lamps.

Fig. 22. Energy balance of a 40-watt fluorescent lamp.

In Table 9 the luminous characteristics of fluorescent lamps are given, which confirm this circumstance with complete clarity.

We see that the luminous efficacy of white 40-watt lamps reaches \(52.5\ \text{lm/W}\), which is 3.4 times greater than the luminous efficacy of a 100-watt incandescent lamp (\(15.5\ \text{lm/W}\)). The luminous efficacy of green lamps reaches \(75\ \text{lm/W}\). These figures show what enormous importance is possessed by the distribution of energy already within the limits of the visible spectrum. The principal reason for the high luminous efficiency of white fluorescent lamps is the advantageous distribution of energy in the visible part of their spectrum. In Figs. 23, 24, and 25 are shown the spectral curves of white, daylight, and soft-white fluorescent lamps. To obtain such spectral curves it is necessary to use mixtures

Fig. 23. Distribution of energy in the spectrum of a white lamp. The areas of the rectangles represent, on the corresponding scale, the energy of the lines of the mercury spectrum.

Table 9

Light Output of Fluorescent Lamps

Lamp color	6 W luminous flux, lm	6 W light output, lm/W	8 W luminous flux, lm	8 W light output, lm/W	14 W luminous flux, lm	14 W light output, lm/W	15 W luminous flux, lm	15 W light output, lm/W	20 W luminous flux, lm	20 W light output, lm/W	30 W luminous flux, lm	30 W light output, lm/W	40 W luminous flux, lm	40 W light output, lm/W	65 W luminous flux, lm	65 W light output, lm/W	100 W luminous flux, lm	100 W light output, lm/W
White	180	30	300	37.5	450	33	615	41	900	45	1450	49	2100	52.5	2100	32	4200	42
Daylight	155	26	250	31	370	26.5	495	33	730	36.5	1200	40	1700	42.5	1800	27	3350	33.5
Soft white	—	—	—	—	325	23	435	29	640	32	1050	35	1500	37	—	—	—	—
Blue	—	—	—	—	—	—	315	21	460	23	780	26	—	—	—	—	—	—
Green	—	—	—	—	—	—	900	60	1300	65	2250	75	—	—	—	—	—	—
Pink	—	—	—	—	—	—	300	20	440	22	750	25	—	—	—	—	—	—
Golden	—	—	—	—	—	—	375	25	540	27	930	31	—	—	—	—	—	—
Red	—	—	—	—	—	—	45	3	60	3	120	4	—	—	—	—	—	—

phosphors described in Section 9. For example, in a white lamp zinc-beryllium silicate is used as the principal phosphor, and small amounts of cadmium borate and magnesium tungstate have been added to it.^47 In a daylight lamp the percentage of phosphors with a blue emission is considerably higher than in white lamps. Unfortunately, the usual laws of mixing do not apply to mixtures of phosphors,^89 and therefore the selection of mixtures is carried out purely empirically.

Fig. 24. Energy distribution in the spectrum of a daylight lamp.

Fig. 25. Energy distribution in the spectrum of a soft-white lamp.

F. Butaeva^26 calculated the mean visibility coefficients for the spectral curves of white and daylight lamps. It turned out that for a white lamp this coefficient is equal to 0.50, and for a daylight lamp to 0.44; the luminous efficiency of the radiation is 310 and 245 lm/W. For a 100-watt incandescent lamp the same coefficient is only 0.21. This is explained by the fact that in white and daylight fluorescent lamps a much larger share of the energy is concentrated in the central part of the visible spectrum, corresponding to the maximum sensitivity of the human eye. The difference in the luminous outputs of white and daylight lamps is explained by their spectral characteristics. Soft-white lamps, which give a more pleasant light, have a considerably lower luminous output precisely because the central part of their spectral curve has been artificially reduced. Conversely, the colossal luminous output of the green lamp is partly explained by the fact that all the radiation of this lamp (containing ZnSiO₃) is concentrated in a narrow region close to the maximum sensitivity of the eye.

The choice of the forms of the spectral curves for fluorescent lamps was determined by colorimetric considerations. The daylight lamp was matched to source C (see Section 2). The white-light lamp approximately corresponds to a color temperature of 3500°K.

In Fig. 26 the colors of test objects under illumination by GEC and VEI fluorescent lamps are shown on a uniform-contrast diagram. The points \(C_1\), \(C_2\), etc., correspond to the colors of the test objects under illumination by the standard source \(C_5\).

In Table 10, \(x\) and \(y\) are the color coordinates, \(\rho\) is the reflection coefficient of the test objects. The most important quantities are \(\Delta S\), giving in terms of thresholds the altered colors of the test object. We see that \(\Delta S\), i.e. the distortions of color rendering under illumination by daylight fluorescent lamps, do not exceed one or two thresholds. The reflection coefficients are almost unchanged. Thus, daylight fluorescent lamps practically coincide in their colorimetric characteristics with source C and give genuinely true daylight, while white fluorescent lamps give a light incomparably whiter than incandescent lamps.

Fig. 26. Color rendering under illumination by daylight fluorescent lamps.

17. APPLICATION OF FLUORESCENT LAMPS

In practical application, the economy of installations is characterized by figures lower than those given in Table 9.

Table 10

Colorimetric characteristics of daylight fluorescent lamps

Test objects	Test objects	Source C	Source C	Source C	OES lamp	OES lamp	OES lamp	OES lamp	VEI lamp No. 5-6	VEI lamp No. 5-6	VEI lamp No. 5-6	VEI lamp No. 5-6
No.	Color	\(x_0\)	\(y_0\)	\(\rho_0\)	\(x\)	\(y\)	\(\Delta S\)	\(\rho\)	\(x\)	\(y\)	\(\Delta S\)	\(\rho\)
1	White	0.310	0.316	1.0	0.311	0.327	0.3	1.0	0.316	0.336	0.9	1.0
2	Yellow	0.458	0.461	0.58	0.460	0.449	0.7	0.62	0.476	0.488	0.7	0.61
7	Red	0.546	0.320	0.14	0.555	0.323	0.7	0.15	0.552	0.324	0.6	0.14
9	Crimson	0.382	0.208	0.14	0.378	0.195	1.0	0.15	0.391	0.214	0.6	0.14
13	Blue	0.200	0.144	0.12	0.206	0.150	0.7	0.11	0.206	0.141	0.7	0.11
15	Light blue	0.188	0.221	0.20	0.188	0.262	2.1	0.19	0.200	0.254	1.8	0.19
22	Green	0.201	0.514	0.21	0.222	0.485	1.3	0.21	0.234	0.508	1.7	0.21

This is explained by the fact that it is necessary to take into account the losses in the switching device and to determine the luminous efficiency as the ratio of the luminous flux to the entire power consumed by the installation. Then we obtain the data summarized in Table 11.

Table 11

Luminous efficiency of installations with luminescent lamps, in lm/W

Color	Power, W	15	20	40	100
White		32	37	44	36
Daylight		25	29	36	29
Soft white		22	26	31	—

Thus, despite the indicated correction, the luminous efficiency of an installation with 40-watt white luminescent lamps reaches 44 lm/W, i.e. almost three times exceeds the luminous efficiency of 100-watt incandescent lamps. The difference is so great that it becomes clear what a formidable competitor has appeared for the incandescent lamp. It must be said that, besides their high economy and color, luminescent lamps also possess a number of serious advantages in comparison with the incandescent lamp. For purposes of general lighting of large premises (factory shops, public buildings), a positive quality of luminescent lamps is their low surface brightness. Instead of the hundreds of stilbs of the filament of an incandescent lamp, a luminescent lamp has a brightness of from 0.15 to 0.75 stilb. Therefore luminescent lamps do not have a blinding effect on the eyes and can be used in open trough-shaped fittings[^49]. In the USA, lighting installations with luminescent lamps are widespread. Usually the lamps are formed into continuous luminous lines, creating uniform diffused illumination[^50].

Fig. 27. Decrease in luminous efficiency during the lamp service life.

Next, a serious advantage of luminescent lamps is their long service life—2500–3000 hours for low-voltage lamps and 10000 hours for high-voltage lamps.

Incidentally, the data on the luminous efficiency of lamps in Table 9 are given after the first 100 hours of burning. During these 100 hours the luminous efficiency drops rather strongly—by 10%; subsequently the luminous efficiency decreases very slowly. In Fig. 27 is shown...

luminous-output depreciation curve over the service life of a fluorescent lamp[^51].

Sometimes the question arises whether the use of fluorescent lamps entails an increased consumption of such scarce materials as copper and other metals. In the United States during the war this question became especially acute. At first glance it seems that the need to use ballasts must certainly lead to a greater consumption of copper than in the case of incandescent lamps. However, a calculation made by Keverly[^52] showed that this is not so. Table 12 gives the results of this interesting calculation. From

Table 12

Consumption of copper and steel in lighting by fluorescent lamps and incandescent lamps

Type of installation	Generator—distribution network: Copper (kg)	Generator—distribution network: Steel (kg)	Installation: Copper (kg)	Installation: Steel (kg)	Total consumption: Copper (kg)	Total consumption: Steel (kg)
Incandescent lamp (2,500 W) in a diffusing fixture	65	465	0	14	65	479
Fluorescent lamp	26	184	4	25	51	209
Total saving	—	—	—	—	14	270

Table 12 it is above all evident that the largest quantities of copper and steel are consumed at the power station itself, in the generators and in the networks distributing the electric power. Therefore a reduction in the consumption of electric power (at the same illumination) leads to such a large saving of these materials (through a reduction in generator capacity and a reduction in the cross-section of conductors) that this saving more than offsets the increase in the consumption of materials in the lighting installation itself. As a result, paradoxical as it may seem, there is even a certain overall saving of copper and steel. We cannot dwell here on questions of the cost of lighting by fluorescent lamps, but, despite their relatively high cost, they quickly pay for themselves and subsequently yield savings[^53]. In concluding the present section, we shall give some statistical data on the prevalence of the individual types of fluorescent lamps.

In terms of spectral composition, white lamps are the most widespread (65%); daylight lamps constitute 30% of all installed lamps, and the remaining colors—5%[^54]. This ratio is explained by the higher economy of white lamps and their more pleasant color properties. Daylight lamps are used mainly where accurate color rendering is important. Soft-white

lamps give a very pleasant light, but their wide application is limited by their relatively low luminous efficacy.

There are also characteristic data on the consumption of fluorescent lamps of various powers. Naturally, the most widespread are 40-watt lamps, accounting for approximately 70%, followed by 20-watt lamps—20%, while the remaining types account for 10%.

18. WAYS OF FURTHER IMPROVING FLUORESCENT LAMPS

Despite their enormous advantages in comparison with incandescent lamps, fluorescent lamps are, of course, not free from certain shortcomings. Moreover, the advantages themselves of fluorescent

Fig. 28. Luminous efficacy of a fluorescing lamp as a function of the temperature of the surrounding air. Blowing causes a lowering of the temperature of the walls. Conversely, enclosing the lamp in a sealed fixture raises the temperature of the walls.

lamps can be considerably strengthened in the course of their further improvement.

It may be said without much exaggeration that the fluorescent lamp of the 1943 type is related to the fluorescent lamp of the future as the carbon lamp is to the coiled gas-filled incandescent lamp. After all, we are witnessing the very first stage in the development of fluorescent lamps. It is therefore especially important to imagine, at least in general form, those fundamental paths along which it will be possible to achieve a significant improvement in the characteristics of fluorescent lamps. Naturally, much in this field is still unclear and highly debatable. The main shortcoming of ordinary fluorescent lamps is their sensitivity to external temperature. Figure 28 shows the luminous efficacy of a lamp as a function of

external temperature. Such a dependence is quite natural, since the external temperature strongly affects the temperature of the lamp walls. The optimum external temperature corresponds to 25° C.

When the external temperature is lowered by 1° C, the luminous flux of the lamp decreases by approximately 1.5%. At temperatures below 0° C the mercury freezes out, and the lamps ignite poorly. Attempts have been made to adapt ordinary fluorescent lamps for operation at low temperatures by lowering the argon pressure[^47]. However, these attempts lead to a noticeable decrease in luminous efficacy and service life. For the construction of fluorescent lamps operating at any temperatures, it would be expedient to abandon the use of mercury. Lamps filled only with inert gas would possess the required temperature properties, but experiments carried out with inert gases by a whole series of investigators (Jenkins, Fonda, and others[^55]) have so far given very disappointing results in terms of luminous efficacy. The main reason limiting the luminous efficacy of fluorescent lamps is considered to be the unfavorable ratio between the wavelength of the exciting radiation and the wavelengths of luminescence. As is known, the resonance radiation of inert gases lies in the range from 584 Å for helium to 1469 Å for xenon.

If one assumes that, in the best case, for each quantum of exciting radiation one quantum of luminescence radiation is obtained, then the energy yield of luminescence upon excitation by the Schumann region should lie approximately within the limits from 0.10 to 0.30 (taking the mean wavelength of luminescence to be 5000–6000 Å). The validity of the quantum law of luminescence yield, first stated and established by S. I. Vavilov, was verified experimentally for a whole series of liquid and solid luminophores[^56].

The impossibility of “exchanging” one “large” short-wave quantum for two or more “small” quanta greatly lowers the limiting value for the luminous efficacy even of ordinary fluorescent lamps using the mercury lines 1850 and 2537 Å. In them the energy yield of luminescence cannot, according to this law, appreciably exceed approximately 0.5. Thayer and Barnes[^28] carried out the corresponding calculations and measurements for 1.5-watt lamps 2.5 cm in diameter, coated with various luminophores. The maximum possible luminous efficacy \(E_M\) was calculated as follows:

\[ E_M = 621CR \times 0.45 \times 0.90 = 251CR\ \text{lm/watt}, \tag{2} \]

where 621 is the number of lumens per watt of radiant energy for 5560 Å, \(C\) is the visibility coefficient of the luminescence radiation, determined from the spectral curve, \(R\) is the quantum ratio—averaged-

... the ratio of the energy of the luminescence quantum to the energy of the exciting quantum; 0.45 is the fraction of the lamp power converted into the 2537 line; 0.90 is \(\eta\), the efficiency of the lamp’s scattering layer (see Section 10).

Table 13 gives the results of calculations and measurements by Thayer and Barnes. For some reason they did not carry out the same calculations for the practically interesting cases of the white and daylight lamps. Such calculations by formula (2) were made by F. Butaeva\(^{28}\) and gave the following results: for the white lamp, \(50\ \mathrm{lm/W}\); for the daylight lamp, \(55\ \mathrm{lm/W}\).

However, F. Butaeva’s experiments\(^{39}\) with a filter cutting off the 1850 Å line, carried out already inside a fluorescent lamp

Table 13

Determination of the quantum yield for phosphors by comparing calculations and measurements of 15-watt lamps

Phosphor	Visibility coefficient \(C\)	Quantum ratio \(R\)	Calculated luminous efficacy \(E_M\)	Best measured luminous efficacy*)	Quantum yield	Energy yield
\(\mathrm{CaWO_4}\)	0.21	0.58	30	21	0.70	0.41
\(\mathrm{MgWO_4}\)	0.37	0.53	50	35	0.70	0.37
\(\mathrm{ZnSiO_3}\)	0.78	0.48	95	70	0.74	0.36
\((\mathrm{Zn,Be})\mathrm{SiO_3}\)	0.55	0.43	60	32	0.53	0.23
\(\mathrm{CdSiO_3}\)	0.53	0.42	55	30	0.55	0.24
\(\mathrm{CdB_2O_5}\)	0.34	0.41	35	23	0.66	0.27

(a small spot on the wall, which was covered by the filter, was coated with phosphor; see Section 9), showed that, in addition to the 2537 Å line, the 1850 Å line also participates in exciting the phosphors to approximately the same extent. It follows from this that formula (2) for calculating \(E_M\) is incorrect, and the limiting luminous efficacies given in Table 13 are significantly underestimated. The values of the quantum yield, on the contrary, are overestimated by approximately a factor of two, since in the calculations only the 2537 Å line was taken into account, while what was essentially measured was the total brightness of the phosphor produced by both lines, 2537 and 1850 Å. Unfortunately, so long as the intensity of the 1850 Å line is unknown, it is impossible to calculate precisely the fundamental limit of luminous efficacy for fluorescent lamps with mercury vapor.

*) In the measurements, the radiation of the discharge itself, giving \(4\ \mathrm{lm/W}\), was excluded.

In any case, according to the law of quantum yield this luminous efficacy cannot exceed \(621 C \times R\), i.e. for white lamps \(128\ \mathrm{lm/W}\), and for daylight lamps \(134\ \mathrm{lm/W}\).

Such luminous efficacies would be obtained with 100% conversion of the power into the 2537 Å line and with a luminescence quantum yield equal to unity. These luminous efficacies are very high, but still considerably less than \(220\ \mathrm{lm/W}\) for an ideal light source (see section 3).

Let us recall that, in this case, the mean visibility coefficient of white lamps is almost one and a half times higher than that of the “ideal source” (see section 3), owing to the appreciable deviations of the spectral curve of the white lamp from the spectral curve of a temperature radiator.

The question naturally arises whether deviations from the “quant-for-quant” law may be possible, i.e. whether an “exchange” of large quanta for several small ones may be observed. With optical excitation of the luminescence of sodium vapor such an “exchange” of quanta, as is known, is observed\(^{60}\). When the line at 3302.34–94 is excited, sodium vapor emits three lines: 22057–84, 11388–11404 and 5889.96 and 5895.93. Upon excitation, sodium atoms are raised to the level \(4^{2}P_{3/2,\ 1/2}\), passing from this level to the level \(4^{2}S_{1/2}\) and emitting the line 22057–22084, then from \(4^{2}S_{1/2}\) to \(3^{2}P_{1/2,\ 3/2}\), emitting the line 11382–11404, and finally from the level \(3^{2}P_{1/2,\ 3/2}\) to the level \(3^{2}S_{1/2}\), emitting the line 5889.96–5895.93 (Fig. 29).

Fig. 29. Scheme of “exchange” of quanta in the fluorescence of sodium vapor.

It is possible that phosphors will be created (especially with the participation of rare earths) having analogous properties. This would at once sharply raise the fundamental limit of the luminous efficacy of luminescent lamps and would make it realistic to create mercury-free luminescent lamps with high luminous efficacy.

Finally, it is necessary to simplify the circuit for switching on the lamps. Despite all the ingenuity of the principle of relay operation, the complexity of the switching circuit is undoubtedly a drawback of luminescent lamps. Here the development of a lamp with cold electrodes seems promising. There seem to be hopes for the creation of electrodes with a large cold emission of electrons *).

*) Reports have appeared in American journals about a new switching device developed in the transformer laboratory of the General Electric Company. The new device ignites lamps without a relay, but in this case lamps of the old type quickly fail. Beginning May 1, 1944, mass production of 40-watt lamps of a new type, with normal service life when operated with the new switching device, was begun.

ADDENDUM 1. ULTRAVIOLET LUMINESCENT LAMPS

During the war, quite distinctive luminescent lamps were produced which emitted chiefly invisible ultraviolet rays. At first ultraviolet luminescent lamps were made in the form of long, narrow tubes for alternating current at a voltage of 120 V; later these lamps took the form of bulbs, like the one shown in Fig. 30, and were designed for direct-current supply at a voltage of 24–26 V[^61]. In addition to the design, the phosphor applied to the wall of the lamp also underwent a change. In the old type, evidently, calcium tungstate was used, whereas in the new lamps a special phosphor 360 B was employed, having an emission maximum at a wavelength of 3650 Å (Fig. 31). Ultraviolet lamps are used with a black Wood filter, which readily transmits long-wave ultraviolet in the region of 3650 Å and completely absorbs visible radiation. The combination of an ultraviolet phosphor with a Wood filter makes it possible to create an excellent radiation source for exciting phosphors applied to the scales of aircraft instruments.

Fig. 30. Ultraviolet luminescent direct-current lamp, 24 V.

Fig. 31. Distribution of energy in the spectrum of an ultraviolet luminescent lamp.

In this case, all glare from the instrument glasses is absent, since the exciting lamp gives practically no visible light. When ultraviolet luminescent lamps are used, two successive transformations of radiation already take place. First, in the lamp itself, the short-wave ultraviolet of the discharge is converted by the phosphor 360 B into long-wave ultraviolet; then this ultraviolet, falling on the luminescent inscriptions, is converted into visible radiation. As measurements show, despite this double transformation, the use of luminescent ultraviolet lamps is very effective.

ADDENDUM 2.

Natural light	Color temperature °K	Artificial sources
Exceptionally clear deep blue northern sky	28000	1 carbon + 1 day. flame lamp
Deep blue northern sky	26000
Blue filters \(5400^\circ K\)–\(30000^\circ K\)	24000	Blue filters \(5400^\circ K\)–\(30000^\circ K\)
Blue sky with thin white clouds	22000
Blue sky	20000	1 carbon + 2 day. flame lamp
Uniformly clouded sky	18000	1 carbon + 4 day. flame lamp
	16000	1 carbon + 8 day. flame lamp
	14000	Daytime flame lamp
	12000
	10000
	8000
Midday sun	6000	4 day. + 1 white flame lamp
3 hr 30 min in the afternoon	5600	3 day. + 1 white flame lamp
4 hr 30 min in the afternoon	5000	2 day. + 1 white flame lamp; Day photolamp
2 hr	4500	1 day. + 1 white flame lamp
1 hr 30 min	4000	500 W Mazda day. flame lamp incl.; Photoflood
40 min	3500	150 W Mazda day. flame lamp incl.; white flame lamp; \(C\) Ø photolamp
30 min	3000	Gas lamps
20 min	2500	Empty; ordinary incandescent lamps
Sunrise	2000	Candle flame

Left-side scale annotations: Light of the sky; Sun; Time after sunrise.

Fig. 32 (addendum 2).

LITERATURE

1. S. Hibben, J. E. S. Trans., 34, 855, 1939.
2. B. F. Fedorov, General Course of Lighting Engineering, 2nd ed., Moscow—Leningrad, 1944.
3. L. Ickis, H. Haynes, G. E. R., 42, 145, 1939; Metropolitan Vickers Gazette, 20, 342, 1943.
4. N. T. Fedorov, General Color Science, Moscow, 1939.
5. D. Nickerson, J. E. S., Trans., 34, 1233, 1939. E. S. Ratner and D. A. Shklower, Bull. VEI, 5, 35, 1941. A. Taylor, Ill. Eng., 35, 625, 1940.
6. D. Judd, JOSA, 22, 65, 1932.
7. A. P. Ivanov, Electric Light Sources, Part 1, Moscow—Leningrad, 1938.
8. P. Mohler, JOSA, 29, 152, 1939.
9. T. Kolian, JOSA, 28, 49, 1938.
10. A. Becquerel, Ann. de phys. et chim., 55, 93, 1859; La Lumière, Paris, 1867.
11. R. Woud, Physical Optics, Moscow—Leningrad, 1936.
12. G. Claude, C. R., Paris, 203, 1203, 1936, 207, 1206, 1938.
13. J. Randall, Journ. Roy. Soc. Arts, 135, 4398, 1937.
14. Electronic and Ionic Devices, GEI, Moscow—Leningrad, 1940, p. 217.
15. W. Elenbaas, Physica 1, 211, 678, 1934; 2, 45, 1935; De Ingenieur, 50, 83, 1935.
16. D. A. Rozhanskii, Physics of Gas Discharge, Moscow—Leningrad, 1937.
17. M. Druyvesteyn, F. Penning, Rev. Mod. Phys., 12, 1940.
18. W. Penney, Phys. Rev., 39, 467, 1932.
19. F. Arnot, G. Baines, Proc. Roy. Soc., 151, 256, 1935; P. Bricout, Journ. d. Phys., 9, 88, 1928.
20. V. Fabrikant, I. Tsir, DAN, 16, 279, 1937.
21. K. I. Panenkin, ZhETF, 9, 1607, 1939.
22. M. Druyvesteyn, N. Warmoltz, Phil. Mag., 17, 1, 1934.
23. V. Fabrikant, DAN, 15, 451, 1937.
24. V. Fabrikant, ZhETF, 8, 549, 1938; M. Townsend, Tr. AJEE, 61, 607, 1942.
25. C. Found, E. Heunelly, Report to the 34th Annual Meeting of the American Association of Illuminating Engineers, 1940.
26. F. Butaeva, DAN, 27, 654, 1940.
27. J. Marden, N. Beese, G. Meister, JOSA, 30, 184, 1940.
28. R. Thayer, B. Barnes, JOSA, 29, 131, 1939.
29. F. Butaeva and V. Fabrikant, Izv. Acad. Sci. USSR, Phys. Ser., 1944.
30. A. Oday, R. Cissel, J. E. S. Trans., 34, 1165, 1939.
31. N. Beese, JOSA, 29, 278, 1939; J. Marden, N. Beese, Ill. Eng., 36, 235, 1941.
32. F. Butaeva, Technical Report of VEI, 1944; ZhTF, 1945 (ready for press).
33. J. Marden, G. Meister, J. E. S. Trans., 34, 503, 1939.
34. V. I. Dolgopolov, Bull. VEI, No. 5, 29, 1941.
35. Kéfer, Dissertation 1940 (Holland).
36. C. Fonda, J. Phys. Chem., 43, 564, 1939.
37. F. Seitz, R. Johnson, Advances in Physical Sciences, 29, 89, 1940. N. Mott, R. Gurney, Electronic Process in Ionic Crystals, Oxford, 1940. H. Leverenz, Pt. J. R. E., 32, 256, 1933. V. Levshin, V. Antonov-Romanovskii and D. Blokhintsev, Izv. Acad. Sci. USSR, Phys. Ser., 1944 (in press).
38. J. Marden, G. Meister, Ill. Eng., 36, 1286, 1941.
39. F. Butaeva, V. Dolgopolov and V. Fabrikant, Izv. Acad. Sci. USSR, Phys. Ser., 1944 (in press).
40. C. Amick, Fluorescent Lighting Manual, McGraw Hill, N. Y., 1942.
41. E. Beggs, Lighting a Lamps, 42, No. 5, 18, 1942.
42. K. Channon, K. Barr, Ill. Eng., 37, 769, 1942; W. Phillips, Electrical

Wordl, Apr., 17, 62, 1943; A. Van, C. Cutler, Electrical World, May, 1, 38, 1943; J. Sabatini, Ill. Eng., 38, 171, 1943.
43. D. Caverly, Ill. Eng., 38, 133, 1943.
44. J. Marden, G. Meister, J. E. S. Trans., 34, 1095, 1939.
45. R. F. Hays, D. Gustin, J. E. S. Trans., 35, 939, 1940.
46. R. Hays, Electronics, 13, 14, 1940; El. Eng., 60, 226, 1941.
47. E. Beggs, Lighting a. Lamps, 42, No. 4, 14, 1942; R. Neumann, Electronic Engineering, 16, 202, 1943.
48. Railway, Electr. Eng., 31, 135, 1940.
49. W. Potter, W. Darley, Ill. Eng., 35, 759, 1940; F. Winkler, Ill. Eng., 37, 220, 1942; D. Severance, ibid., 38, 49, 1941; R. Daglas, D. Adams, Ill. Eng., 38, 141, 1943.
50. G. Taylor, Ill. Eng., 36, 1414, 1941; W. Brown, Mag. Light, 5, 20, 1941; K. Frank, Ill. Eng., 37, 217, 1942; M. Luckiesh, F. Moss, Ill. Eng., 37, 81, 1942 (general information); R. Ackerley, Elect. Eng., 63, 203, 1943 (English); E. Cleaver, India Rubber World, 102, 33, 1940 (rubber); R. Hays, El. Eng., 60, 539, 1941; Ill. Eng., 36, 570, 1941 (airfields); L. Long, Rayon, Textil Monthly, 22, 179, 1941; G. Taylor, Cotton, 105, 105, 1941; D. Caverly, Ill. Eng., 36, 489, 1939 (textiles); E. Derrah, G. E. R., 46, 593, 1943; Steel, 105, 42, 1940; Lighting Rev. Dec., 19, 1912 (aircraft instruments); R. Blair, Telephony, 120, 9, 1941; B. Durden, Telephon Eng., 45, 23, 1941 (telephone stations); El. World, 120, 77, 1943 (metro); Power, Plant. Eng., 42, 67, 1942 (F. D.); Architect. Record, Apr., 69, 1943 (shoe factory); Mag. Light, 1, 13, 1943 (underground factories); El. Times 102, 30, 1942 (picture galleries).
51. C. Amick, loc. cit.
52. D. Caverly, Ill. Eng., 38, 133, 1943.
53. B. Greenberg, Ill. Eng., 37, 165, 1942; Lighting a. Lamps, 41, 19, 1943.
54. P. S. Miller, Electrical Eng., 63, 126, 1944.
55. G. Fonda, H. Huthsteiner, JOSA, 32, 156, 1942.
56. S. I. Vavilov, ZS. Physik, 42, 311, 1927; V. Fabrikant, Sow. Phys., 3, 567, 1933; G. Harrison, P. Leighton, Phys. Rev., 38, 5, 1932; Solomin, DAN, 1941.
57. C. Christensen, G. Roliefson, Phys. Rev., 34, 1157, 1929.
58. Ill. Eng., 39, 235, 1944; Lighting a. Lamps, Apr. 1944.
59. E. Beggs, JOSA, 33, 61, 61, 1943; G. E. R. Jan., 1942.

Reviews

1. S. I. Vavilov, Izv. AN SSSR, physical series, 1941.
2. A. Morin, Amer. Techn. and Ind., 19, No. 1–2, 31, 1942.
3. P. I. Chernousov, Electricity, 1944.
4. C. Amick, Fluorescent Lighting Manual, McGraw Hill, N. Y., 1942.
5. P. S. Miller, Electrical Engineer, 63, 126, 1944.
6. L. Davis, H. Ruff a. W. Scott, J. J. E. (London), 89, 447, 1942 (English luminescent lamps).
7. R. Neumann, Electronic Engineering, 16, 154, 202, 1943 (historical survey).

1. F-lamps. ↩↩

7. 14. ↩

12. *) The ignition of luminescent lamps is greatly facilitated by the presence of argon. ↩