12 COBLENTZ: CONSTANTS OF SPECTRAL RADIATION 



/ 



first term (1 —e~''''^^') log e, etc.) is used. As in equation (2) an 

 approximate value of Co = 14,500 is used in applying the second 

 term correction. For wave-lengths up to 1 /z this correction term 

 is small being only 2.1 and 4.3 for temperatures intervals {T^ - Ti) 

 of 363° and 623°, respectively, when using Tr = 1450°C. How- 

 ever these corrections increase very rapidly with wave-length 

 beyond 1 /x so that at 2 ix, with the same temperature intervals 

 just mentioned, the second term corrections to the values of Cz 

 amount to 168 and 227 respectively. 



As already stated, it was deemed of greater importance to 

 obtain experimental data than to spend the time discussing the 

 bearing of the data at hand upon existing theories. From the 

 data now at hand, this procedure seems amply justified. More 

 than 180 isothermal energy curves have been obtained, and by 

 actual count 75 to 80 per cent of the rehable sets of these curves 

 are found to fit the Planck equation, within the experimental 

 errors of observations. The numerical values of the constajtits 

 are smaller than the older determinations of Paschen, and of 

 Lummer and Pringsheim (and, for that matter the earlier values 

 of the present data, obtained by a different system of computa- 

 tion, and not including all the correction factors for reflection). 

 However, as will be shown in the complete paper, the data of 

 previous observers are in agreement with present values, when 

 computed on the same basis. ^ 



The data now available were obtained with different fluorite 

 prisms, water cooled shutters, air and vacuum bolometers, and 

 thermocouples. The radiators were platinum-wound porcelain 



1 For example, Paschen's data, if computed by the present methods would give a 

 value of XmT = 2912 and C2 = 14,460. The data of Lummer and Pringsheim are 

 wrong owing to an error in their calibration cujve which amounts to 0.02/x for the 

 region of the spectrum up to 2.5^. This would decrease many of their values of 

 Xmoj by almost 1 per cent, and reduces their mean value to \„iT = 2930. Their 

 energy curves did not fit the Wien equation and since in the present research, 

 radiators of their design were used under conditions which were similar to theirs, 

 it is possible to recalculate their data by the present methods of computation. 

 This gives a mean value of XmT = 2911 and Cs = 14,450. If we exclude their 

 last value of \m.T= 2814, which is evidently not comparable with the rest, their 

 mean value of XmT = 2924 and d = 14,500. Considered as a whole, a fair esti- 

 mate of the older data is C2 = 14,460 to 14,500. 



