VISIBLE AND NEAR-VISIBLE RADIATION 193 



calculated difference from the value 1362, obtained by absolute measure- 



o 



ment of the Hnes in the region of 3130 A and shorter wave-lengths, of 

 18 per cent, and an observed difference of 15 per cent. Where correction 

 factors computed from the transmission curves are applied, our observa- 

 tions on standard-type mercury arcs have indicated an error of the order 

 of 2.5 to 3 per cent. If, however, a great intensity exists in the region of 

 cut-off, much more serious errors will be introduced. 



For instance, if one is concerned with the measurement of solar energy 

 in the region specified, much greater difficulties present themselves. 

 Apparent solar distribution is shown in arbitrary units indicated on the 

 right; the relative distribution of response, as measured by the inter- 

 national ultra-violet standard, is indicated by Curve S in the lower 

 section. Obviously, much the greater part of the energy lies in the region 

 of wave-lengths longer than 3200 A. Consequently, corrections of a 

 rather large order must be applied in order to secure an evaluation of 

 the extreme region. 



In the curve CD-BF, we have a typical case of a band selectivity 

 secured by the differential thermocouple method, combining corex D with 

 barium flint. In using such a method, however, precautions must be 

 taken to avoid a change in selectivity resulting from an increased opacity 

 of the glass due to ultra-violet exposure. 



PHOTOCELL 



By far the commonest selective detector in use is the photoelectric 

 cell. Commercially, a wide wave-length .range of sensitivity has been 

 desirable. The majority of such cells employ a sensitive surface com- 

 posed of cesium on cesium oxide on silver. Such a sensitive surface is 

 placedin an evacuated tube and a potential applied between an auxiliary 

 electrode and the sensitive surface. When illumination falls upon the 

 sensitive surface, electrons escape and produce a current between the 

 electrodes. This is measured either directly by a galvanometer or 

 indirectly by means of an amplification circuit. Such a detector is 

 capable of the greatest sensitivity of any known method. If the inten- 

 sities are extremely small, the individual electrons may be counted by the 

 method developed by Geiger and Miiller (58). 



Figure 31 shows the sensitivity curves of a group of commercially 

 available photocells. The full-hne curves 1 to 4 are representative of the 

 various types of cells of cesium on cesium oxide on silver, where the tube 

 is completely evacuated and the surface sensitized by a special process. 

 These present two regions of maximum sensitivity, one lying near 

 3600 A, and the other varying from 7000 to 8700 A. They all present 

 a minimum sensitivity between the two maxima in the region of 5000 A. 

 A large amount of technical investigation has been expended in order to 



