p. LECOMTE DU NOtJY 753 



it really amounts to 1.5 per cent approximately, by concentrating 

 the beam of light before and after passing through the prism. 



Finally, the curves were integrated graphically. Fig. 2 shows 

 the different aspects of the curves before and after corrections due 

 to absorption. (These curves are not to scale, in order to emphasize 

 the difference.) The results were: 



r.0.7 



3.6 



Jt7 rtO.S ^0. 



= 110.3, io.8= =4.8, Lo.T-= = 

 0.4 Jo.4 Jo.4 



The ratios are: 



Upper limit = 0.8, ^ = 0.0435 

 Upper limit = 0.7, ^ = 0.0316 



These results are in good agreement with those of Ingersoll, who, 

 with the upper limit 0.76, found figures between 0.036 and 0.046, 

 and in contradiction with those of Drude (20), who gives 12 per cent 

 as luminous efficiency of Nernst lamps, (instead of 4.35 per cent). 

 Lux gives 5.96 per cent, but takes no account of the fact that the 

 screening method he used allowed most of the radiation up to 1.2 ^ 

 to pass. Coblentz and Nichols found 0.033 for the efficiency of acet- 

 ylene flame. 



Hence, as we have established that our source radiated 87 watts 

 equatorially, the quantity radiated as luminous waves is equal to 

 4.35 per cent of 87 (limit 0.8 m) = 3.8 watts, and 3.16 per cent of 

 87 (limit 0.7 fx) = 2.75 watts. Let us take the quantity corresponding 

 to the limit 0.7 ix for example. These 2.75 watts are radiated at the 

 distance of 103 cm. from the source, (distance of the thermopile) 

 by square centimeter. The energy then becomes 0.0000206 watts. 

 The area of the sHt being 0.1 sq. cm., it only receives 0.00000206 

 watt seconds, or 20.6 X 10~^ watts = 20.6 ergs. This quantity 

 of energy is spread over the range delimited by the upper and lower 

 limits of integration (0.4 fj, to 0.7 fx) and the energy distribution 

 curve corrected for the absorption by mirrors, and by the prism, 

 that is, over a surface of 83 sq. cm., (obtained by graphical integration 

 of the curve, Fig. 3). This corresponds, in the scale chosen, to 

 0.248 ergs by unit of surface. Similarly, between the Hmits 0.4 n 



