by taking several preliminary swings, so as to give the thermopile chance to settle into a steady state, the rate of shift of zero was kept pretty steadily and the corrections were easily applied. It was also necessary to photometer the arcs in the actual condition in which they were under test. To this end the apparatus was set up as shown in Figure 4. Here A is the arc lamp, B the thermopile, C the galvanometer, D the telescope and scale, E an adjustable rotating sector disc just in front of the arc, F the quartz cells in their sliding screen in front of the thermopile window, G a silvered plate glass mirror which could be quickly interposed in the beam between the arc and

the thermopile so as to deflect the rays into the portable photometer H, set up on the other side of the photometer room. The coefficient of reflection of the mirror had previously been many times determined as the mirror had been in use for photometric work. The photometer was ready for use at any time simply by closing the switch on the standard lamp. When in course of a series of thermopile measurements it was desired to test the c. p. of the lamp the disc was started, the mirror swung into place and readings were then taken on the portable photometer.

The carbon arc was first attacked and it proved to be a difficult subject for investigation. The particular lamp used was of the enclosed type, having the globe fitted with a short side tube and a quartz window so as to keep the arc as steady as possible without losing the ultra violet. To the same end it was found desirable to adjust a magnet behind the arc so as to keep it burning on the side of the carbons next the thermopile instead of wandering round and round the carbons in the usual manner.

The arc thus operated gave a prodigious amount of ultra violet radiation, showing a continuous spectrum far down into the ultra violet and the three enormously intensive carbon bands usually ascribed to cyanogen, one of them in the extreme violet and the other two near wave lengths 380 μμ and 360 μμ respectively. Reduced to the standard distance the deflection due to the ultra violet cut off by the Euphos glass amounted to 74 cm., being 30% of the whole energy which passed through the quartz cell. It has, of course, been long known that the naked electric arc gives off very powerful ultra violet radiations and its effect in the production of ophthalmia electrica has been known for more than half a century, but in this case the extent of the ultra violet activity was somewhat unexpected.

It was undoubtedly considerably enhanced by the intensive cyanogen bands as regards that portion of the radiation lying near the visible spectrum, but on the other hand the extreme ultra violet, wave length 300 μμ and less, is unquestionably stronger in the case of an open arc than in the enclosed arc on account of the very intense continuous spectrum emitted from the crater, which is much lessened when the are is enclosed. No separation between these parts of the ultra violet was attempted with the lamp under consideration since its unsteadiness was a constant source of annoyance and the ordinary variations of independent readings from the mean amounted to 5 or 6. It was sufficiently evident, however, that a powerful enclosed arc in a globe which permits all the radiations to pass is an enormously powerful source of ultra violet light. The carbon arc, however, is rapidly passing out of general use so that attention was next directed to the luminous arc.

Magnetite Arcs. - The magnetite arc is one of the commonest and most generally useful outdoor illuminants. It gives a very intense nearly white light due almost wholly to the arc stream itself. The spectrum of this, the active electrode being composed almost wholly of the oxides of iron and titanium, is immensely complicated, containing thousands of bright lines so closely packed as almost to obtain

the effect of a continuous spectrum. The actual character of the spectrum photographed with a fairly wide slit, is shown in Plate 2, d. Here, with the quartz arc spectrum for reference at a is shown the radiation from the magnetite are through a quartz window and below it the spectrum of the same arc taken through its ordinary globe. A quartz window was used merely to insure steadiness of the light, which would have been lost by taking off the globe. A glance shows that this spectrum is exceedingly rich in powerful lines all through the ultra violet clear down to wave length 230 μμ. The glass globe cuts off the spectrum quite sharply near wave length 300 μu, as in Plate 2, e, but from this region to the visible spectrum lies an almost continuous mass of strong lines, very intense in the region where the quartz mercury arc is conspicuously weak, say from the group at wave length 313 μ to the group near wave length 365 μμ.

For radiometric measurements the magnetite arc, which was operated at 6.6 amperes and about 80 volts, proved much more steady than the carbon arc, showing more small and quick fluctuations, but fewer of the large and relatively slow variations which interfered most with the readings. As a consequence the deflections obtained agreed more closely, the average variations of a single setting running between 3 and 4. For the magnetite arc through the quartz window the cut-off of Euphos glass amounted to 29 cm., 28% of the total deflection. Through the ordinary glass globe the deflection was reduced to 22.4 cm., 22.5% of the total deflection. The difference between these results shows that while there is a large amount of energy of short wave length produced by the magnetite arc, most of the ultra violet energy is of wave length greater than 300 μμ. As compared with the quartz mercury arc used without its globe the magnetite arc gave relatively about 60% less energy of wave length below 300 μμ and about 40% more energy in the wave lengths above 300 μ. The candle power in the horizontal direction as measured by the method just described amounted to 760 in the run with the quartz window, and 700 in the run with the ordinary globe.

The Nernst Lamp. — Finally a series of readings was taken on the Nernst lamp. The lamp investigated was of the single glower type for 220 volts, taking 91 watts and giving a downward c. p. of 68. As the spectrum of this source runs to less than wave length 300 μμ and reaches that vicinity with somewhat material strength an attempt was at first made to run the Nernst glower without a globe. It proved so difficult to get steady deflections under these conditions, on account of the effect of air currents, that this measurement was

abandoned and the readings taken with the globe on, which proved reasonably easy, the precision being comparable with that obtained with the ordinary incandescent lamps. But even then the lamp proved very sensitive to small changes of voltage and only by very careful regulation of the current could consecutive series of readings be held in reasonably close agreement.

In the average the deflection due to the ultra violet in the Nernst lamp with its globe was 1.81 cm. and the percentage of energy thus cut off was 5.2. This completed the radiometric investigation of ordinary illuminants. Two others which it seemed desirable to investigate, that is the ordinary flame arc, and the arc between iron electrodes as used by Finsen were studied on the spectrograph, since their fluctuations were of a character to make their study by means of a galvanometer of so long period as that used in this investigation quite impracticable. The peculiarities of these sources will be referred to in discussion of the general results.

Sun Light. Finally it seemed advisable to take some comparative readings on sunlight as a source of ultra violet radiations, particularly with reference to the amount of ultra violet energy with respect to the intensity of the light. Of course the solar radiation in absolute amount has been investigated with great thoroughness, but the ultra violet has received less attention than the rest of the spectrum. In general the sun radiates energy substantially like an incandescent black body at about 6000 degrees C. except in so far as its energy, particularly in the ultra violet, is cut off by the absorption of its own and the terrestrial atmosphere. It behaves then, like an enormously hot incandescent body shining through a medium that cuts off all the ultra violet of less wave length than about 295 μu and greatly diminishes the shorter radiations even into the violet of the visible specOne would expect therefore to find relatively little total ultra violet per unit of illumination so far as the direct light of the sun is concerned. On the other hand as Schuster 14 and others have shown, much of this cutting off of the ultra violet is due to scattering of the short waves by the molecules of the atmosphere and small bodies suspended in it. In other words, the violet and ultra violet are not wholly lost, but appear in radiation from the blue sky.

Of the energy thus radiated from the sky the maximum lies almost in the edge of the ultra violet. The arrangement of the apparatus for experiments on sunlight is shown in Figure 5. Through the

14 Nature, XXXI, p. 97.

courtesy of the Director, this part of the work was done in the Rogers Laboratory of Physics where the conditions for getting natural light were good. In Figure 5, A is a porte lumiére receiving the light from the sun and forming by means of the iris diaphragm B, stopped to 3 mm. diameter, an image of the sun on the thermopile front at C, before which was placed the usual quartz cell D. The thermopile was connected with the galvanometer F, read by the telescope and scale G. By the use of the diaphragm, forming a species of "pin hole" image on the face of the thermopile, at a distance of 3 meters,

the light and energy were cut down so as to be readable with comparative ease.

To measure the intensity of the illumination a Simmance-Abady flicker photometer H was set up close alongside the thermopile so that the solar image could be quickly moved so as to fall squarely on the photometer disc. On the other side of the photometer at I was an 80 watt tantalum lamp which was previously calibrated, in terms of the current flowing through it, against a standardized Gem lamp. From the source of supply the current was taken to this lamp through an adjustable rheostat J and a mil-amperemeter K. In measuring the light-intensity of the beam which was allowed to fall on the thermopile, it was simply shifted from the face of the thermopile to the face of the photometer and by means of the rheostat J