spectral energy characteristics of the mercury vapor lamp

Journalof the

Optical Society of Americaand

Review of Scientific InstrumentsVol. 10 JANUARY, 1925 Number 1

SPECTRAL ENERGY CHARACTERISTICS OF THEMERCURY VAPOR LAMP

By George R. Harrison* and George Shannon Forbes

ABSTRACT

By means of a large aperture quartz spectroradiometer the spectral energy distributionof a special variable-length mercury lamp was measured between 14,000 A and 2300 A underwidely varying conditions. The effects of current, voltage and ventilation on the radiationat fifteen important maxima were measured, including three maxima in the infrared, four inthe visible, and eight in the ultraviolet. The composition of each maximum is given, andcurves are plotted showing the degree of resolution used; the variation in intensity of eachmaximum with voltage gradient at constant current; the variation of intensity of the strongultraviolet maximum near 3660 A with voltage for five current values from two to fouramperes; the stationary characteristic curves of a high pressure arc; two typical cases ofspectral energy. distribution at constant power; and the variation with current of the intensityof each maximum at constant power.

It was found that for constant current the energy in each maximum except those in theinfrared increased linearly with the voltage after a certain minimum (near 8 volts for 2.5amperes) was reached. The pressure was measured for each stationary state used, the extremevalues being 20 mm of mercury and two atmospheres. Where power input was constant theenergy radiated in each maximum increased rapidly with decreasing current. It was foundthat the most satisfactory condition for running an arc at pressures under two atmospheres,at least, was to have the pressure as high as possible, so that for a given power input thevoltage was high and the current low. This condition gives the greatest efficiency and theleast change in energy distribution with total energy variation. The relation of pressure tothe other variables and the effects at higher pressures are being investigated further; also anabsolute determination is being made of the energy available in each wave-length for photo-chemical purposes.

Although a number of papers on the quartz mercury vapor lamphave appeared since it was first designed some twenty years ago,surprisingly little information is available concerning the variation inits spectral energy distribution to be expected with changing condi-tions. Many workers have measured the relative intensities of the

* National Research Fellow.

HARRISON AND FORBES [J.O.S.A. & R.S.I., 10

various lines in the visible and ultraviolet, but usually for only one ortwo power rates in a given lamp, and for conditions occurring only intheir particular researches. It is, of course, not to be expected that aspectral energy curve for one arc will apply exactly to another, butsufficient data can be secured to determine, for example, what condi-tions should be chosen so that any variation of total energy radiatedwould produce the least variation in energy distribution.

Desiring a steady source of visible and ultraviolet light of intensitysuitable for photochemical and photographic investigations, one whoseenergy distribution could be exactly determined in absolute units, wehave studied the behavior of the quartz mercury lamp under widelyvarying conditions of use. While isolation of certain spectral regionsby filters is sufficient for most commercial uses, the separation obtainedin this way is not great enough for the photochemist. The presentpaper deals with the energy distribution obtained with a high aperturequartz prism spectroradiometer, and although the length of the spec-trum from 11000 to 2100A was only three inches, sufficient resolutionwas obtained for the purpose desired, with small sacrifice of intensity.

The most complete investigation of the energy distribution of radia-tion from the quartz mercury lamp was made by KUch and Retschinsky'in 1906. They measured the total intensity of the visible, and theintensities of the individual visible lines, by means of a flicker pho-tometer, and of the total ultraviolet radiation with a photoelectric cell,while a photographic method was used for the continuous background.They varied the power input into the lamp over a wide range, but madeno effort to separate the variables, current and voltage. In otherpapers they give results of experiments determining the temperaturesin various parts of the arc,2 and the absorption by the luminescentmercury vapor of its own light.' Pfluiger4 has repeated and verifiedportions of their work, using a high aperture quartz-rocksalt spec-trometer and a Rubens thermopile. He resolved the lines sufficiently toshow that members of the same series increased in intensity together,which KUch and Retschinsky had predicted from their results. Athan-asiu5 also has recently verified this.

1 KUch and Retschinsky, Ann. der Phys., 20, p. 563; 1906.

2 Ibid, 22, p. 595; 1907.

3 Ibid, 22, p. 852; 1907.

4 Pffilger, Ann. der Phys., 26, p. 789; 1908.

6 Athanasiu, Compt. Rend., 178, p. 2071; 1924.

2

MERCURY ARC CHARACTERISTICS

Ladenburg,6 Fabry and Buisson, Wintherj Souder,9 Coblentz,Long and Kahler,' 0 Coblentz and Kahler,'1 and Koppius 2 have meas-ured the energy distributions in certain lamps for one or two powerinputs.

Buttolph 3 ,A has summed up the most important known character-istics of the mercury arc, traced the development of the quartz lamp,and discussed its theory and applications. He has collected a bibliog-raphy of papers dealing with the construction, characteristics, and usesof mercury arcs.

It is evident from a study of the results of these workers that astatement of the power put into a mercury lamp is not sufficient todetermine the energy distribution of the radiation emitted by it. Aswith time and intensity in the photographic plate, a "reciprocity law"does not hold between current and voltage in the arc, since not onlytheir product is important, but their ratio as well. In no published work,so far as we have been able to find, have the effects of current andvoltage been separated definitely, by measuring the variation in energydistribution at constant current and variable voltage, and at variablecurrent and constant voltage.

The purpose of the present work was to study the energy distributionfrom a typical lamp under widely varying conditions, to separate theeffects of current and voltage, and to find what conditions, at highefficiency and high total intensity, were such that a small variation intotal intensity would produce the least variation in energy distribution.

To get a clear picture of the relation of the variables concerned in theoperation of the lamp, some discussion of its stationary characteristiccurves is necessary. Fig. 1 shows several such characteristic curvesmarked T1 , T2, and T3. The stationary characteristic curve shows thecurrent and voltage relations in the arc when a steady state has beenreached. That is, for a given ventilation of the arc we get a curve T1

when the line voltage is kept constant and various resistances RI, R 2,etc. are placed in series with it. If, however, the arc is allowed freer

Ladenburg, Phys. Zeits., 5, p. 524; 1904.' Fabry and Buisson, Compt. Rend., 153, p. 93; 1911.8 Winther, Z. Electrochem., 20, p. 109; 1914.9 Souder, Phys. Rev., 8, p. 683; 1916.

Coblentz, Long and Kahler, Bur. Stds. Sci. Paper No. 330; 1918.Coblentz and Kahler, Bur. Stds. Sci. Papers, 378, 233; 1920.

12 Koppius, Phys. Rev., 18, p. 443; 1921.1Buttolph, Cooper-Hewitt Bull., 105.14 Buttolph, Gen. Elec. Rev., 23, 741 and 858 and 909; 1920.

Jan., 19251 3


ventilation so that its heat can escape more rapidly, it will move alongthe stationary characteristic curve T when its series resistance ischanged. The steady state of the lamp is attained in a period varyingfrom three minutes to an hour or more, according to conditions.

Arc length const

1 upplyVotage conzt

Arc Current IFIG. 1. Yolt-ampere diagram for a quartz mercury vapor lamp of constant length, wit/i

constant supply voltage.Curves T1. . . .T3 are stationary characteristic curves, applying only after the lamp has

come to a steady state. To pass from one of these to another a change must be made in theventilation of any lamp.

It is evident that if we wish to study the lamp at constant currentwhile varying its voltage, along line C1, for instance, we must vary theventilation of the lamp to do so. It is also evident that it is much moredifficult to keep the voltage constant while varying the current, asalong line VI, since no part of any characteristic curve is so nearly

4

Jan., 1925] MERCURY ARC CHARACTERISTICS 5

parallel to a constant voltage line, as to a constant current line. Undercertain conditions the characteristic curves bend back, so that anincrease in applied voltage produces a decrease in current; the lamp,when it gets too hot, goes out.

The usual type of mercury arc is built with constant length, and thecurves in Fig. 1 assume a constant arc length. The arc used in thepresent work was designed to have variable length, for several reasons;all results are reduced to unit length however, after a correction of12 volts had been made for the electrode drop.

The arc used is shown in Fig. 2. The legs of the inverted quartz Udip into two pyrex reservoirs, and are sealed to them with de Khotinskycement, the upper part of each reservoir serving as a water cooler andthe lower part containing the mercury for the arc. Iron wire electrodesare sealed into the sides of the reservoirs, and dip almost to the bottomof the mercury pools. A quartz trap is sealed to the top of the arc, sothat the last trace of air in the tube can be exhausted by boiling themercury up into the trap.

In the original model of the lamp the lower legs and the trap tubewere of large bore, but the arc showed a great tendency to oscillate,the levels of the mercury moving up and down several centimeterscontinuously, while the arc voltage varied twenty per cent or more.In the final form these legs were made of quartz capillary of one milli-meter internal diameter below the water coolers, which damped theoscillations down to a few hundredths of a millimeter and kept thevoltage quite constant. A finer thread of mercury would not havecarried the maximum current of five amperes.

The bends in the upper part of the quartz tube were made withthickened walls so that under the greater heat caused by deflected ionsthe tube would not fuse. The two reservoirs and the trap were con-nected through rubber pressure tubing by a system of stopcocks tovacuum, so arranged that any one of them could be separately ex-Lausted either through an inch of thermometer capillary or through aube of large bore. In this way exact control of all mercury levelsvas possible. Arrangements vere also made to apply air pressure tohe reservoirs, so that pressures up to two atmospheres could be reached.Ihe pressures in the reservoirs could be read at any time from man-meters connected with them.

In starting the arc, both reservoirs and the body of the lamp wereexhausted to a fraction of a millimeter, and then the mercury wasallowed to rise slowly in the arms of the tube. The outside of the

HARRISON AND FORBES [J.O.S.A. & R.S.I. 10

lamp, wherever it could be reached, was heated with a Bunsen flameand the mercury boiled vigorously. When the mercury had risenso as to partly fill the trap the rubber tube connecting this to the pumpwas closed with a pinchcock as near the trap as possible. The current

Tto Fhmp

Tap

.M 2

W Water outSpectrometer

Sl1it

Water inWaterout

ToMonometers

Waterin -l ectrode

A ' 4 | '-To Manometr

+E1ectrode -R, QUARTZ MERCURY ARC

arranged for varyingpressure and length.

FIG. 2. Te special design of quartz mercury vapor lamp used in this work, arranged forvariation of pressure, arc length, voltage gradient, current, series resistance, and rate of cooling.

was then turned on, the upper electrode being always negative. Thus,the greatest flickering occurred at the surface farthest from the pec=trometer slit. The two reservoirs were then slowly exhausted until the

~6

Jan., 1925] MERCURY ARC CHARACTERISTICS 7

mercury broke at the top of the U, and the arc length then increasedas pressure was built up inside of the lamp. The height of the mercurycolumns was finally adjusted by varying the resistance in series with thelamp, the rate of cooling, and the pressure in the two reservoirs.Although the upper reservoir gradually filled at the expense of the lower,the excess could be returned without extinguishing the arc. A coppertube provided with small holes was bent to the shape of the quartztube and currents of air from this were directed against the lamp overthe whole luminous column, when necessary.

A lamp of this type possesses several important advantages. Whenthe quartz becomes devitrified, after say 1500 hours of use, a shortcylinder can be removed where the light is taken from the arc, and afresh one put in its place. The lamp thus has an indefinite life. Thenthe arc length is variable, which is a desirable feature for certain work.The most important advantage of this type of arc, however, is that achange can be made from one stationary state to another on a differentcharacteristic curve much more rapidly than in the ordinary arc. Itwas found possible to build up the pressure in the arc very rapidly, whenfirst struck, by lengthening it to 25 cm, putting a large current throughit, and then shortening the column thus building up the voltage drop.Where some twenty runs through the whole spectrum were taken in aday under various conditions it was very important to be able toarrive at the desired new state quickly.

The whole arc was surrounded by a large box with holes which couldbe opened or shut to provide different degrees of ventilation, the topof the box being open to a large pipe leading to a hood. A window ofheavy glass was placed in one side of the box so that the arc could beobserved from the outside, and a cathetometer was arranged to measure,while the arc was burning, the heights of all the mercury columns in-volved. All controls were outside of the box, so that it was possibleto make runs continuously for a whole day, under widely varying con-ditions, without its removal.

In series with the arc were placed a fairly high inductance to damp outoscillations, and a resistance which would carry five amperes, variablefrom 200 ohms downward in small steps. The arc was fed by a 550 voltD.C. generator run on an induction motor; the terminal voltage wasextremely constant. In no case, in the present work, was the voltagedrop across the arc greater than 300 volts, which corresponded to about17 volts per cm gradient. A precision ammeter was connected in serieswith the lamp, and a voltmeter across its terminals.

HARRISON AND FORBES [J.O.S.A. & R.S.I., O

The spectroradiometer is shown in Fig. 3. The lamp, having a tubenine millimeters in internal diameter, was placed practically in contactwith the slit, 4 cm long and variable in width from 1 mm to 0.05 mm.The slit was built upon the base of the instrument as a separate unitfrom the collimating lens, so that ample room for a shutter and severalabsorption cells was provided between them. In order to eliminate asmuch as possible the effect of stray infrared radiation the shutter wasmade of "electric smoke" glass, kindly loaned us by Mr. D. C. Stock-barger of the Massachusetts Institute of Technology, whom we wish to

Ghu fe r Table

Arco X

Gpace fr-\,,/Absorption Cells \

Screen-

LARGE APERTURE QUARTZ

5PEOTRO- RADIOMETER

FIG. 3. The spectroradiometer Used to mneasure the itensity distribution in the spectrum ofthe quartz mercury vapor lamp.

thank also for the loan of a resistance for the arc and for technicalinformation regarding the Cooper-Hewitt mercury arc. This glasstransmits the short-wave infrared, and is perfectly opaque to the visibleand the ultraviolet, so that when null readings were taken in theseregions, the shorter infrared penetrated the shutter, and thus cancelledout. Very little stray radiation was found, however.

The Cornu prism, made by Hilger, had a face nine cm long and six cmhigh. The two quartz lenses were six cm in diameter and twenty cm infocal length for D-light. The apparatus was collimated for about3500A, and the prism was set at minimum deviation for this wave-length. The focal plane had a very small curvature.

The screw carrying the thermopile mounting was placed parallel tothe focal plane, and with a motion of four inches the spectrum could becovered from 14,OOOA to 2000A. The thermopile mounting was free to

8


rotate about a vertical axis on its carriage, and a rod, passing throughtwo guides in the top of the mounting and swivelled at its far end overthe optical center of the prism, kept the hot junctions of the thermo-pile normal to the incident light. The prism and lenses remainedfixed.

The thermopile was constructed of bismuth and silver, following thedirections of Coblentz with certain minor variations. It containedfourteen junctions, and was used with a Leeds and Northrup high-sensi-tivity d'Arsonval galvanometer of 12 ohms resistance, 6 seconds period,and sensitivity 14.1 mm per microvolt. The junctions were 1 mm wide,but in most cases were covered with a slit .4 mm wide, and the deflec-tions for single maxima were in many cases greater than 180 cm, whichwas the limit of the scale. This was 3.5 meters from the galvanometer,and was read with a telescope, a one-fourth inch mirror being used. Theeffective length of the thermopile slit was 18 mm, so that the entirelength of the collimating slit was not used. The thermopile-galvan-ometer combination was calibrated in terms of a standard lamp fromthe Bureau of Standards, so that all measurements could be reducedto absolute energy values if desired. The present report deals only withrelative values, however.

In Fig. 4 are shown two typical runs of the thermopile through thespectrum. Although the screw could be turned by hand, and theradiometer slit set on any desired wave-length, it was also arrangedto be turned by a small synchronous motor properly geared down.The whole spectrum could be covered in any desired time from twominutes to fourteen, but in the present work the time was uniformlyeight minutes, giving the highest speed consistent with negligiblegalvanometer lag. One observer sat at the galvanometer telescopeand recorded the readings at all maxima, or at regular intervals,occasionally closing the shutter and taking a null reading; he also readthe voltmeter just after each maximum was passed. The other observermeasured the pressures registered for both sides of the arc. He alsoread the ammeter at stated intervals, and the heights of the arc mercurycolumns with the cathetometer. The bottom of the U in the trap,marked T in Fig. 2, was taken as a bench mark, and the other heightsM1 and M2, bounding the arc, and RI and R2, were then measured fromthis. From these readings, those of the external manometers, and thebarometric height the arc pressure was computed on each side, and theresults averaged. A full-size chart of the lamp was drawn to scale, andthe arc length was taken from a table computed from this. This length

Jan., 1925] 9


was measured at the center of the tube, where the thread of light isconcentrated when the pressure is high.

Fig. 4 shows the degree of resolution used, which, although not high,is sufficient for most photochemical purposes, and is much greater thanany obtainable with the usual filters. Greater resolution would have

o1 w vvave -enqTn \-.- II00',A 5000 4000 3600 2500

FIG. 4 Two energy distribution curves taken for approximately tihe sane power inputs,showing the maxima measured, the degree of resolution obtained, ad te increase in intensityobtainable by using high voltages and small currents for a given power.

been obtainable by sacrificing light intensity. Usually only the maxi-mum and null values were read, but the curves of Fig. 4 were taken withreadings made every four seconds.

It should be noted that the curves are for practically identical powerinputs into the lamp, the difference being that the upper was for lowcurrent, high voltage and pressure, and the lower for high current,low voltage and pressure.

10

Jan., 19251 MERCURY ARC CHARACTERISTICS 11

Table 1 gives the approximate values of the principal lines makingup each maximum. The first column contains the arbitrary number

TABLE 1

Maximum Lines Series (Fowler) ApproximateRel. Intensity

1, 2 Selective quartz radiation?3 General radiation

11288 A ls-2p, 210140 lP-2S 5

4 6234 2S-4P 16152 _ 2

5 5819 - 35804 s-4ph 35790 1P-2D 35769 1P-2d' 65461 lpi- ls 5

7 4359 1P2- ls 54348 1P-3D 5

8 4078 Ip 2 -2S 34047 lp3-ls 3

9 3663 lpi-2d" 43655 lp 1-2d' 63650 lpi-2d 10

10 33523342 lp,-2s

11 3132 1p2 -2d" 43126 lp 2 -2d' 5

12 3026 lp 1-3d" 13024 lp-3d' 23022 lpl-3d 2

13 2804 etc. lpi-4d 1-4d' 0-4d" 0

14 2700 _ 2

15 2537 1S-1p 2 2

16 2483 etc. 1p2 -4d" etc. 1

_ _ . __


given to the maximum, the second the wave-lengths of the lines includedin it, the third their series notation according to Fowler's Report onSeries in Line Spectra, the last a rough estimate of the relative inten-sities of the various members of each group. Since the heights of thevarious maxima were measured, and not their areas, the effects observedmay be considered as being due to the strongest line in each case, withsome effect due to the weaker lines. Since, however, the large ultra-violet maximum, No. 9, near 3660, is composed of lines belonging toone series only, members of a single diffuse triplet, all of these shouldincrease together, and it may be considered as a single line. Its behavioris indistinguishable from that of the other maxima, those in the infraredexcepted.

In comparing the relative intensities of the va ious maxima it shouldbe remembered that no correction has been made for reflection orabsorption in the optical train, or for the effective width of the slitwhich is greater at longer wave-lengths than at shorter. This correctionwould have to be made in studying the true spectral energy distributionfrom the lamp; we are here interested only in measuring the varyingintensities of single lines.

In Fig. 5 are given curves showing how the various maxima increasein intensity as the voltage gradient in the arc is increased at constantcurrent. In order to keep the current constant the degree of coolingof the lamp had to be varied in most cases. But under ordinary ventila-tion, with no forced draft, the lamp tended to draw about 2.5 amperesover a large voltage range. It is evident that the intensity of anymaximum goes up linearly with the voltage gradient after a particularlimiting voltage, varying with the individual lines, and of course withthe current, has been reached. It seems possible that the voltage-energycurves may bend over again after a certain pressure has been reached,and this point is now being investigated with a very high pressure arc.Maximum 3 (10, 140A) is the only one which shows a slope not pro-portional to its relative height, and the increase is evidently due togeneral radiation from the hot quartz of the lamp. Table 2 gives theratios of the values of the maxima at various voltages for constantcurrent of 2.5 amperes. The slopes of the straight portions of thecurves are also given, and under the heading "intercept" is listed thevoltage gradient at which the straight portion of each curve would cutthe voltage axis if it continued at the same slope. It is evident thatthe changing spectral energy distribution at low voltages is due largelyto changing slope of the energy-voltage curves. At high voltage, how-

12


ever, the change is determined mainly by the diff erent intercepts whichthe straight portions of the curve would have if extrapolated to thevoltage axis. These intercepts are grouped within the region 4-7 volts,and undoubtedly have an important physical meaning, which shouldthrow light on the mechanism of the mercury arc if interpreted. Before

1PG. 5. Curves showing the variation of energy with voltage gradient at constant current forten of the most important maxima.

definite interpretation can be made, however, the maxima must beseparated into their individual lines.

Fig. 5 shows how rapidly intensity increases with voltage. Takingthe strong ultraviolet line near 3660, for example, by increasing the

Jan., 1925] 13


voltage three and one-half fold, from 4 v/cm to 14 v/cm, the energyof the line is increased from 9 to 154 units, or 17.1 times. Between 7and 14 volts, where the energy-voltage curve is linear a 4.5-fold incre-ment is produced due to the fact that the intercept is positive.

The degree of variation of individual points from the plotted curvemust not be taken as an indication of the accuracy attainable. Thevariation is due to the fact that those points lying farthest from thecurve correspond to states in the arc which had not become quitestationary. It was possible to test such points by plotting pressureagainst voltage for them; it was found in general that the greatest

TABLE 2

Ratio to Maximum 5 (Y- G line)Maximum Slope Intercept at

volts I. _ _ __ _ _ _ _ _

5v 7v 9v lv 14v

3 370 6.4 .347 .337 .359 .375 .4025 590 5.6 1.000 ----7 4 0 ° 4.5 .725 .738 .655 .600 .5628 1 270 4.25 .375 .474 .407 .369 .3459 590 5.15 .909 1.100 1.120 1.080 1.050

10 400 4.45 .587 .705 .638 .545 .54512 270 5.5 .173 .279 .303 .332 .32914 150 5.0 .013 .147 .168 .166 .16515 160 5.1 .013 .170 .180 .171 .165

errors occurred when a large pressure change had just taken place.Without the variable length feature, however, it would have been muchmore difficult to produce the desired change of stationary state in theshort time allowed between runs.

Fig. 6 gives a family of energy-voltage curves for different currentstrengths for maximum 9, the strong ultraviolet triplet near 3660.This was chosen because of its simple structure and its intensity. Datafor curves similar to these were taken for all maxima, but this set isrepresentative. Voltage-energy curves were taken for each currentstrength separately, and current-energy curves at constant voltage werethen made, these being used to tie the former set of curves together.The points taken from the double network of readings were checkedtogether, and the final curves were plotted from a graphical average.The dotted curves in Fig. 6 show the variation in radiation of maximum9 for two power inputs, with variable voltage. If a voltage-current

14


"(reciprocity law' held these dotted curves would be horizontal andstraight.

Fig. 7 shows the variation in energy radiated in each maximum whenthe power input per cm is kept constant, but the current and voltage

° 1 4 5 6 7 8 9 10 Il 12 13 14

FIG. 6. A faiily of curves showing the variation of the intensity of niaxinuti No. 9 (3660 A)with voltage gradient, for different current values.

The dotted curves connect points for the same power input.

are changed. The current has been plotted against galvanometerdeflections for ten of the principal maxima, each of which is marked

Jan., 1925] 15

16 HARRISON AND FORBES [J.O.S.A. & R.S.I., 10

with the most important wave-length in its group. The curves ingeneral have similar shapes and may be compared directly, all beingtaken under the same conditions. The point at 2.5 amperes lies belowpractically every curve, and although no error in calculation can be

'Sc

14

131

12(

l0(

7(

6C

AC

30

20

t0

n11.5 2.0 2.5 3.0 3.5 4.0 4.

FIG. 7. Curves showing the variation of energy in the most imtportant maxima with current,at a constant power input of 25 watts per cim.

found, apparently either the voltage or arc length was read incorrectly,since other results confirm the general shape of the curves as plotted.Lines 3 and 11 are somewhat anomalous; the first is the infrared groupnear 10140, and since the maxima were not corrected for background,

C

)Power = 25watts per Cm

5lit 0mmArc 2 eries 8

<,5780 etc.

E13 3663\

U)

C- 5

U S

020

u ~~ = I= _ _~~136

6

Amperes Arc Current


the increasing temperature radiation from the quartz is evidently thecause of the increase in slope. No explanation can be found for theanomalous curvature of the graph of maximum 11, and this point isbeing investigated further.

The figure shows at once the great gain in efficiency to be expectedfrom using high voltages and low currents for a given power input; inother words, within the limits investigated, (0-2 atmospheres) thehigher the pressure the better. Whether this holds to very high valuesis being investigated in the especially constructed high pressure arcmentioned before. It is evident that at 25 watts per cm and 4.5 amps.,only one-half as much energy will go into the yellow-green maximum asif the current were 1.5 amps. and the voltage three times as great.Thus the arc, for high efficiency, should be cooled as little as possible.

Since the preceding curves all indicate that the highest efficiencies,within the limits investigated, are to be reached by building up pressureand voltage gradient as high as possible by keeping the rate of coolingat a minimum, it is important to see how far a further pressure increasewill result in improved operation. The exact value of the best currentwould undoubtedly vary with the cross-section of the arc tube, due notso much to the change in current density as to the changed temperaturegradient from walls to center. We may conclude from the resultsalready obtained, however, that the average arc will show the leastchange in energy distribution for a given variation in total intensityif the pressure is as high as possible. Since this condition is also theone for greatest intensity of all maxima, and for greatest efficiency, thebest operating conditions seem to be those with as high a pressureand voltage drop as is consistent with the life of the lamp and itsmechanical structure.

We are continuing the study of the effect of pressure on the othervariables, and hope to be able to reach pressures up to four or fiveatmospheres with an arc of different construction. After finally deter-mining the best form and rating of arc to be used in photochemcialwork we intend studying the degree of constancy in energy distributionto be expected from it, and the number of ergs per second per squaremm obtainable in each wave-length band chosen for use.

In conclusion, we wish to thank the Cyrus M. Warren Fund for agrant covering the cost of building the special arc used.

JEFFERSON PHYSICAL LABORATORY (G. R. H.) AND

CHEMICAL LABORATORY OF HARVARD COLLEGE (G. S. F.)

HARVARD UNIVERSITY.OCTOBER 21, 1924.

Jan., 19251 17

spectral energy characteristics of the mercury vapor lamp

Documents