Economical Light Sources

D. Zernov, Moscow

The problem of obtaining a highly economical light source has in recent years raised the question of using, for this purpose, the luminescence phenomena of rarefied gases and metal vapors when an electric current is passed through them. It is known that the maximum luminous efficiency of so-called temperature radiators, i.e., lamps in which an incandescent body is used to obtain light, can under the most favorable conditions reach only 14% of the power expended on incandescence.* This luminous efficiency can be obtained at a temperature of the incandescent body of about 6000°, at which the position of the maximum intensity of the spectrum emitted by the incandescent body coincides with the maximum of the sensitivity curve of the human eye. The luminous efficiency of practical types of incandescent lamps with a tungsten filament proves to be still lower and amounts to only 2–3% of the power expended. This is explained, as is known, by the fact that the spectrum emitted by an incandescent body, in addition to the visible region, always contains an intense infrared and ultraviolet part not perceived by the eye.

Since the spectra of luminous gases consist of separate lines, one may expect that with a suitable choice of gas, i.e., in the case where the most intense lines of its spectrum fall in the visible region, the conversion of electrical energy into light will occur with a considerably higher efficiency. If all the energy expended on radiation were emitted in the form of a line with a wavelength corresponding to the maximum of the eye’s sensitivity curve \((\lambda = 0.55\ \mu)\), then the luminous flux thereby obtained would amount to 620 lumens for each watt of power expended (the mechanical equivalent of light for \(\lambda = 0.55\ \mu\)).

This value of luminous efficiency is thus the limiting

* The calculation of the percentage of energy emitted in the form of light is made as follows: the curve of the energy distribution over the spectrum (for the given source) is constructed, and its ordinates are multiplied by the corresponding ordinates of the sensitivity curve of the eye. On the basis of the data thus obtained a new curve is constructed, the area of which, referred to the magnitude of the power expended on radiation, gives the required percentage of energy emitted in the form of light.

for all light sources, including gas-discharge tubes. Of interest is the question of how closely the luminous efficacy of light sources based on the phenomenon of gas discharge can approach this limiting value, and what causes affect the magnitude of the luminous efficacy obtained from them in practice. First of all it must be noted that the values of the luminous efficacy of gas-discharge tubes obtained in practice will, of course, always be below the indicated limiting value, since, first, in the spectra of all gases there are lines belonging to regions not perceived or only weakly perceived by the eye, and, second, the phenomenon of electrical discharge in gases is always associated with losses of energy due to the release of heat. The indicated maximum value of luminous efficacy can be regarded only as a limit that may be approached either by selecting a gas whose most intense lines fall in a region close to the maximum of the eye-sensitivity curve, or by selecting a regime in which the lines falling in the visible region become most intense, and also by choosing conditions under which the losses due to heat release constitute, as far as possible, a smaller part of the power expended.

The first designs of light sources using an electric discharge in gases, the so-called Geissler tubes with cold electrodes, could not find application for lighting purposes for the following reasons:

a) These tubes required a high operating voltage, since a large part of the applied voltage, owing to the appearance of a positive space charge, fell in a comparatively thin layer at the surface of the cathode (the so-called cathode fall), and this fall usually amounted to several hundred volts.

b) An increase in the brightness of the lamp by increasing the strength of the discharge current was limited by destruction of the electrodes under the action of ionic bombardment (positive ions in the layer of the cathode fall acquire very considerable velocities).

c) The economy, owing to large power losses at the electrodes, was insignificant (for example, for neon glow lamps the luminous efficacy is only 0.8 lumen/watt).

Only after heated oxide electrodes, possessing considerable electron emission and making it possible, by destroying the positive space charge, to reduce the cathode fall, and at the same time the total operating voltage, to such an extent that the tubes could burn from the ordinary mains voltage while carrying considerable discharge currents, began to be used in gas-discharge tubes, did the problem of using an electric discharge in gases for lighting acquire a real basis.

At the present time the greatest use has been gained by gas-discharge tubes containing neon, mercury vapor, and sodium vapor. As for the first, owing to the insufficiently favorable distribution of energy in the spectrum of neon, the economy of modern-

the neon lamps does not exceed the efficiency of ordinary incandescent lamps. The reason for this is that a large part of the neon lines that fall within the visible region lies at the red end of the spectrum, so that even without taking into account losses due to heating and radiation in the infrared and ultraviolet parts, the theoretical limit of luminous efficacy is only about 25% of the power expended. Owing to their relatively low efficiency, the field of application of neon lamps is limited to advertising and architectural lighting and to signaling purposes.

Tubes with mercury vapor have proved to be considerably more advantageous; their luminous efficacy reaches 35–40 lumens/watt. This high efficiency, however, can be obtained only at high current densities and considerable mercury-vapor pressures. With a decrease in current density and mercury-vapor pressure, the efficiency of mercury lamps rapidly falls.

Fig. 1. Sodium lamp of the Osram firm.

This is explained by the fact that at low mercury-vapor pressures and insignificant current density, the positive column of the mercury discharge emits chiefly the resonance lines of mercury, which lie in the ultraviolet part of the spectrum. Only at considerable current densities and vapor pressures, as a result of the action of reabsorption and stepwise excitation,* do the lines lying in the visible region begin to play the predominant role, which entails an increase in the luminous efficacy of the lamp. Thus the most advantageous tubes are those designed for high power and, consequently, for high luminous intensity. Less powerful tubes, which could be used for domestic lighting, turn out to be insufficiently economical. This circumstance makes mercury gas-discharge tubes suitable chiefly for lighting high rooms and squares, thereby limiting to a certain extent the field of their application.

At present, one of the most economical sources of light is gas-discharge tubes with sodium vapor. These tubes (like all gas-discharge lamps with metallic vapors), in addition to

* For more details, see the discussion of the discharge in sodium vapor.

of metallic sodium, contain some inert gas (the so-called main gas), in most modern lamps neon, which at the moment of ignition serves as a conductor of current. After several minutes of burning, under the action of the heat developed in the discharge, the sodium begins to evaporate, and its spectrum gradually displaces the spectrum of the main gas. The maximum luminous efficiency attained up to the present time by such lamps is 40—50 lumens/watt.

The sodium lamps now being manufactured for lighting purposes are divided into two principal types: lamps in the form of a tube and lamps in the form of a broad cylindrical bulb.

The lamp of the first type, developed by the Osram firm in Germany, is shown in Fig. 1. The discharge tube is sealed into a vacuum jacket, which serves as thermal insulation. The light output of such a lamp is about 300 candles at

Fig. 2. Circuit diagram for switching on a sodium lamp with two heated electrodes. \(T\)—filament transformers; \(D\)—choke.

Fig. 3. Construction of a sodium lamp with one heated electrode (Philips firm).

a total power consumption of 70 W. The luminous efficiency of the whole installation is about 40 lumens/watt. Fig. 2 shows the switching circuit, containing transformers serving to supply the heating of the electrodes, and a choke connected in the alternating-current circuit (220 V) in series with the lamp, absorbing part of the voltage supplied to the lamp.

Fig. 3 shows the second type of lamp, developed by the Philips lamp works in Holland. The lamp contains an oxide cathode placed in the center of the bulb, and two anodes located on both sides of the cathode. The voltage drop on the tube at a discharge current of 6 A is only 12—15 V; therefore several lamps are usually connected in series in the circuit (Fig. 4). The supply is by direct current. When operating on alternating current the lamp is connected according to a rectifier circuit so that each anode operates during one half-period. Thermal insulation

is achieved by placing the lamp inside a glass Dewar vessel. The lamps are made of glass resistant to the action of sodium vapor (ordinary grades of glass quickly darken in the presence of hot sodium vapor).

Both types of sodium lamps have been used successfully for lighting experimental sections of highways, building facades, etc. In comparison with mercury lamps, sodium lamps have the advantage that with their aid greater economy can be obtained at relatively low powers and small luminous intensities. The monochromaticity of the radiation given by sodium lamps, which makes them unsuitable where it is necessary to distinguish the coloration of objects, is, on the other hand, their advantage where this is not required, since in monochromatic light, owing to the absence of chromatic aberration of the eye, visual acuity is considerably increased. This property extends the field of application of sodium lamps to the lighting of workshops in which small parts are produced and precision mechanical work is carried out, and also permits their use for illuminating objects observed when working with a microscope.

Fig. 4. Circuit for switching on a sodium lamp with one heated electrode.
\(K\) — cathode; \(A_1, A_2\) — anode; \(T\) — filament transformer; \(a\) — negative pole of the anode-voltage source; \(b\) — positive pole; \(1,2\) — wires supplying anode voltage; \(3,4\) — network feeding the filament transformers.

In addition, sodium lamps are used when working with polarization instruments (in the sugar industry), for calibrating optical instruments, for lighting photographic laboratories when working with special grades of photographic plates insensitive to yellow light, and in a number of other cases of laboratory practice.

One of the important applications of the sodium lamp is its use as a modulated light source in television installations. A lamp intended for this purpose, developed by the firm Osram, is shown in Fig. 5. During operation the lamp is placed in an electric furnace that maintains the necessary

for vaporizing the sodium. * In the front part of the furnace there is a window, covered with frosted glass, which serves as the frame surface. Owing to its considerable brightness (126 candles with a frame surface of \(3 \times 4\) cm, a current of 0.1 A, and a voltage of about 100 V), this lamp made it possible to receive images on a television set with a Nipkow disk designed for a large number of elements (10,800).

The study of the electric discharge in sodium vapor is of great interest, since the most intense resonance lines of sodium (\(\lambda = 0.589\,\mu\)) are situated very close to the maximum of the eye’s sensitivity curve (for \(\lambda = 0.589\,\mu\), the sensitivity of the eye is 76.5% of the maximum), and, consequently, the sodium lamp offers great possibilities with respect to obtaining an economical light source. Since the theoretical value of the luminous efficacy of the sodium discharge is 520 lumens/watt (assuming that all the energy is emitted in the form of resonance lines), it is of interest to clarify the possibility of further increasing the efficiency, achieved up to the present, of practical samples of sodium lamps, and to establish the path that must be followed to attain this goal. The solution of this question requires, above all, a detailed consideration and study of the elementary processes taking place in the discharge. Below is given a survey of the literature material concerning the dependence of the luminous efficacy of the sodium discharge on various parameters characterizing the discharge.

Fig. 5. Sodium lamp for television (Osram firm)

Characteristics of Sodium Lamps

The fundamental question that must be dealt with in determining the conditions for obtaining the optimum luminous efficacy of sodium lamps,

* Since in the present case the current density is insufficient for self-heating of the lamp.

is the study of the dependence of luminous efficacy on the density of the discharge current, the temperature of the tube, the pressure, and the nature of the main gas.

A characteristic feature of the luminous characteristics of all types of sodium lamps, which relate the luminous efficacy to the density of the discharge current, is an optimum of luminous efficacy occurring over a certain range of current density, accompanied by a similar optimum in luminous intensity. In Fig. 6 a curve is given, taken from the work of Fonda and Young,¹ showing the dependence of luminous flux and luminous efficacy on the intensity of the discharge current. This curve was obtained for a sodium lamp operating on alternating current, consisting of a cylindrical bulb 6.3 cm in diameter, provided at the ends with electrodes 15 cm apart, and containing

Fig. 6. Luminous intensity and luminous efficacy of a sodium lamp as a function of current intensity.

Fig. 7. Dependence of the efficiency of the resonant radiation of sodium on the intensity of the discharge current for various pressures of sodium vapor.

neon as the main gas at a pressure of 1.5 mm. An analogous course of this dependence was obtained at one time by the author.² At low current densities the temperature of the tube is insufficient, the spectrum of the main gas participates in the glow, and the luminous efficacy is insignificant. With increasing current density the luminous efficacy rises up to a certain limit, after which it begins to fall. Since in practical specimens, with an increase in current density, there is a simultaneous increase in the temperature of the tube and, consequently, in the density of sodium vapor, it is of interest to study the dependence of luminous efficacy on each of these two parameters separately. For this purpose the tube under investigation was usually placed inside an electric furnace, by means of which the temperature was regulated. Fig. 7 shows the curves, presented in the article by Krefft, Reter, and Rompe,³ for the dependence of the efficiency of the resonant lines of sodium on the density of the discharge current at various values of the sodium-vapor pressure. As can be seen from the curves, the efficiency of the resonant lines in all cases decreases with increasing current density.

As for the dependence of luminous efficacy on the temperature of the tube, then, according to indications available in the literature, as well as to direct measurements made by the author jointly with I. M. Ta-

studies in the Light-and-Vacuum Engineering Department of the VEI,* with an increase in temperature the luminous efficacy of the sodium discharge rises to a certain limiting value, after which, with a further increase in temperature, it begins to fall. As regards the question of the influence of the kind and pressure of the main gas on the maximum value of the luminous efficacy of practical samples of sodium lamps, at the present time we still do not possess a sufficient amount of experimental material.

The data on the dependence of luminous efficacy on the pressure and kind of the main gas refer only to conditions of constant current strength and temperature (for a tube heated by an electric furnace), and it may be considered established that under these conditions the luminous efficacy increases with decreasing pressure and with the transition from lighter to heavier gases. However, on the basis of these data one cannot judge the optimum filling of practical samples of sodium lamps.

Energy balance in the positive column

The character of the dependences described above is explained partly by the redistribution of energy over the spectrum with a change in the discharge conditions, and partly by a change in the relative share of the total radiation in the general energy balance.

The radiation of sodium consists mainly of the resonance doublet, \((\lambda = 0.589—6\,\mu)\), and of three infrared lines with wavelengths \(0.820\,\mu\) \((2p—3d)\), \(1.14\,\mu\) \((2p—2s)\), and \(2.20\,\mu\) \((2s—3p)\). Lines corresponding to higher members of the series are weak and, under ordinary excitation conditions, together amount to only a few percent of the resonance radiation. As has already been noted, the conditions for obtaining the optimum luminous efficacy of the sodium discharge correspond to conditions under which the greater part of the supplied power is radiated in the form of the resonance \(D\)-lines. However, the relative intensity of the \(D\)-lines in the sodium spectrum depends to a considerable degree on the discharge conditions, namely—their relative intensity falls both with an increase in the density of the discharge current and with an increase in the density of sodium vapor. The table below, taken from the work of Druyvesteyn,^4 gives the relative—

TABLE 1

* Unpublished.

relative intensity of the resonance lines and of three infrared sodium lines, expressed as percentages relative to the total radiation, for different values of the discharge-current strength (the vapor pressure in the tube was kept constant).

At still higher current densities the relative intensity of the infrared lines increases to 50% of the total radiation. With increasing temperature the relative intensity of the infrared lines also increases (in the temperature interval investigated by Droivestein, from 273 to 287°C, the relative intensity of the infrared lines increased by 3%). The change in the relative intensity of the resonance lines in the spectra of the alkali metals with changes in the discharge-current density and the tube temperature (vapor density) was investigated in detail by Krefft.^5 He established that both with increasing temperature and

Fig. 8. Curves of the energy distribution in the discharge spectrum in cesium vapor at vapor pressures: a) \(p = 4\) mm and b) \(p = 20\) mm.

with increasing discharge-current density, for all alkali metals there is observed a decrease in the relative intensity of the resonance lines, owing to an increase in the relative intensity of the lines of the I and II subordinate and Bergmann series, a redistribution of energy within the series themselves (an increase in the relative intensity of the higher members of the series), the appearance of lines corresponding to the so-called “forbidden” transitions \(1s — md\), and of recombination spectra at the series limits. As an illustration one may cite the intensity-distribution curves obtained by Krefft for the spectrum of cesium at vapor pressures \(p = 4\) and \(p = 20\) mm Hg (Fig. 8). On passing from the lower pressure to the higher one, the curves show an increase in the relative intensity of the lines of the I and II subordinate and Bergmann series, situated near the resonance doublet \((\lambda = 0.8521\) and \(0.8943\,\mu)\), which at a pressure of 20 mm proves to be strongly self-reversed.

Krefft’s work gives no quantitative data for the sodium spectrum; however, one may expect that the situation here is exactly the same.

The explanation of the phenomena described is as follows. As the temperature rises, the number of sodium atoms in the normal (unexcited) state increases. The presence of these atoms causes strong absorption of the quanta of resonance radiation, so that each quantum, before leaving the tube, under-

moves from one atom to another, and the duration of its stay inside the tube increases. This is equivalent to increasing the lifetime of the atom in the excited state and, consequently, leads to an increase in the probability of stepwise excitation of higher levels and to an increase in the probability of quenching of excited atoms through collisions of the second kind, as a result of which the relative intensity of the resonance lines decreases and the lines themselves become self-reversed. An increase in current density,

Fig. 9. Diagram of the energy balance in the positive column of the discharge.

associated with an increase in the electron density in the discharge, also leads to an increase in the probability of stepwise excitation. In addition, a considerable concentration of ionized sodium atoms, occurring at high current density, promotes their recombination with electrons in the volume of the gas, which is associated with the appearance of continuous spectra at the series limits; and the presence of strong ionic fields leads to the appearance of forbidden series.

The phenomena described partly explain the behavior of the characteristics of sodium lamps, giving the dependence of luminous output on the density of the discharge current and on the temperature of the tube. However, alongside these phenomena it is also necessary to take into account the change in the overall energy balance with changes in the discharge conditions.

The overall energy balance in the positive column of the discharge may be represented by the diagram shown in Fig. 9. The most significant

considerable component parts in the overall energy balance, apart from radiation, are:

a) Losses at the walls due to the kinetic energy carried by ions and electrons reaching the wall, and the energy of their recombination.

b) Losses in the gas volume in elastic collisions of electrons with gas atoms. Among the secondary processes leading to losses in the gas volume are: collisions of positive ions with gas atoms, recombination in the volume, and impacts of the second kind. However, the losses due to these processes prove to be considerably smaller than the losses at the walls and the losses due to elastic collisions of electrons, and, in a first approximation, may be disregarded in the overall energy balance.

Fig. 10. Dependence of the output of the total radiation of a sodium discharge on the discharge-current strength for sodium vapor without admixture of an inert gas.

Measurements of the energy balance in the positive column of a discharge in pure sodium vapor and in mixtures of sodium with inert gases (Ne, He, Ar) were carried out by Druyvesteyn. He measured the total radiation (resonance lines + infrared part), the losses at the walls, and the losses in the gas volume as functions of the discharge-current density, and also of the pressure and type of the principal gas under conditions of constant sodium-vapor density. Measurements of the energy balance in pure sodium vapor (without admixture of inert gas) showed that the power expended in a certain section of the positive column was always, within a few percent, equal to the sum: radiation plus losses at the walls. Deviations occurred only in cases where the tube contained impurities of non-noble gases, even a small quantity of which very sharply affects the increase of volume losses. In some cases the sum—radiation plus losses at the walls—proved even greater than the power expended in the given section, which may be attri-

…caused by the action of fast electrons penetrating from the cathode into the region of the tube under investigation. On the basis of the data presented by Druyvesteyn, a curve was constructed giving the dependence of the specific weight of the total radiation (in percent relative to the power expended) on the strength of the discharge current (Fig. 10). The curve shows that, with increasing current strength, the specific weight of the total radiation falls; in the present case (i.e., in the absence of an inert gas) this decrease occurs entirely at the expense of the growth in the specific weight of losses at the walls, which at a current of 6 A reach 70% of the power expended. Measurements of the energy balance, made by Druyvesteyn for a mixture of sodium vapor with inert gases, showed that at low pressures of the inert gas (of the order of several millimeters Hg) the power expended still goes entirely into radiation and losses at the walls; however, at more considerable pressures volume losses begin to make themselves felt, the specific weight of which increases as the pressure rises (Fig. 11). Of the three gases studied (He, Ne, Ar), this effect appears most sharply in helium, where the volume losses constitute a substantial fraction of the total energy balance already at a pressure of the order of several millimeters Hg.

Fig. 11. Dependence of the specific weight of volume losses on the pressure of the inert gas.

In argon and neon the increase in the specific weight of losses in the volume takes place somewhat more slowly, but at high pressures also reaches considerable values (at a neon pressure of about 80 mm and a current of 0.2 A, losses in the volume reach 60% of the power expended). As the theoretical calculation made by Druyvesteyn shows, these losses may be attributed almost entirely to the energy given up by electrons in elastic collisions with atoms of the inert gas. Simultaneously with the growth of the specific weight of losses in the gas volume, there is a decrease in the specific weight of the total radiation, which falls both with increasing pressure and with the transition from heavier to lighter gases (Fig. 12). It should be noted that, in absolute magnitude, the losses in helium, at equal current density and equal pressure, considerably exceed the losses in neon and argon. The comparatively high radiation output for helium is explained by the fact that the absolute magnitude of the radiation in helium also considerably exceeds the absolute magnitude of the radiation in neon and argon; this is a consequence of the higher electron temperature and, consequently, of the greater probability of excitation of the resonance lines of sodium (the excitation function of the sodium D-lines,

as is known, has a maximum at electron velocities of about 7 V).

An increase in the density of the discharge current in the presence of an admixture of an inert gas, as in the case of pure sodium vapor, leads to a decrease in the specific weight of the total radiation; at low pressures this decrease occurs mainly at the expense of an increase in the specific weight of losses at the walls, and at more appreciable pressures—at the expense of an increase in the specific weight of losses in the volume of the gas.

Conclusions

The results presented above from studies of the spectral composition of luminous sodium vapor and of the energy balance in the positive column of the discharge make it possible to draw certain conclusions applicable to the question of the optimum conditions for the operation of practical specimens of sodium lamps.

Fig. 12. Dependence of the yield of total radiation on the pressure of the inert gas.

As follows from the foregoing, an increase in current density leads to a lowering of the specific weight of the resonance lines of sodium in the overall energy balance. Since, on the other hand, in practical lamps maintenance of the required vapor density is effected through heat losses in the discharge itself, in order to obtain high luminous efficacy it is necessary to have the possibility of obtaining the required tube temperature at the lowest possible density of the discharge current. Hence there follows first of all the requirement of the best possible thermal insulation of the tubes, which in the newest designs of sodium lamps is achieved, in addition to placing the tube in a vacuum jacket, by arranging intermediate glass partitions which absorb the thermal energy radiated by the tube and radiate part of it back in the direction of the tube.⁶

As regards the choice of the pressure and kind of the main gas, here it is necessary to take into account that in practical specimens at high—

... at higher pressures of the inert gas, or when using a gas with a small atomic weight, the losses necessary for maintaining a sufficient density of sodium vapor can be obtained at a lower current density; consequently, under conditions of constant current density and sodium-vapor density, the decrease in luminous efficacy will be compensated by its increase due to the reduction in current density. In any case, this question requires further experimental development.

References

1. Fonda and Yong, Journ. Optic. Soc. Am. 24, No. 2, 1934.
2. D. V. Zernov, Journal of Technical Physics III, issue 8, 1266, 1933.
3. Kreft, Reger and Rompe, Z. techn. Phys. 14, 242, 1933.
4. Druyvesteyn, Phil. Mag. 17, I, No. 110, 1934.
5. Kreft, Z. Physik. 77, 752, 1932.
6. R. Rompe, Techn. Wiss. Abhandl. aus d. Osram-Konzern, B. III, S. 44.