yyieon/ of Bef faction in Gases. 485 



It is clear that the observations at l8°-4 and 19° cannot 

 both be correct, and the same applies to the observations 

 at 95°-3 and 108''-4. Excluding the first and last, the agree- 

 ment is about 1 per cent. 



Hence we get 



K^l^ I'^^P f . '0135P 1 r i + -00040(9 \ 



(1 + •00380713 I "^(1 + •0038076/ 1 6 J 



or 



j^_ •000784P_r -0135?^ IJl-g.lT^ol 



(1 + -0038 076 r^(l + -0038 076J U"^^^^^J • 

 Hence we deduce 

 , . -000784? f. -0135? 1/^ . „6>o/ ?^M 



^•-^-(l^W3W)76l^-^(r +-0038076 }i^ + ^-^V(^~^&)}- 



For the dispersion Lorenz gives ^^~ ^ = 1-0038 between Li 



J.2 yUp' — 1 



red and Na, /. Sr, = 1'297. 

 P 

 Thus we have 



1-9.17^ 



1-9'17'^x 1-297 



^ = 1-0048, 



from which we obtain 



9-17-2 =-016, 



or ^ =-0017 for the :>^a line and 0° C. 



P' 



Using this value, we deduce from Baedecker's measure- 

 ments the theoretical value 



yu, = 1-000391 for the Na line at 76 cms. and 0° C. 



My experiments on refraction gave us 



/^ = l-000379. 



As in the case of SOq, the agreement is very satisfactory. 

 The values of K — 1 are about 10 times the value of yu,^— 1 in 

 the visible spectrum. 



Atmospheric Air. 



In this case no experiments on the variation of K with 

 temperature have been made, and the only datum we have is 



K = 1-000558 at 76 cms. and 16° C. (Boltzmann). 



