of Light by Potassium Vapour. 301 



that the value of a x is a little too large, and therefore also 

 the value of aja 2 . This error is not, however, likely to alter 

 the general decrease in the value of aja 2 from 1-8, since 

 the difference between the values for 3 and 4 is large. The 

 shape of the first curve of the series for wave-lengths near 

 the blue absorption line is largely a matter of conjecture, 

 and since the value of a 2 is determined principally by this 

 part of the curve it cannot be relied upon. The observed 

 and calculated values of n lie sufficiently near together on 

 the curve to justify the assumption that the value of ai/a 2 is 

 higher for curve 1 than for curve 2, but the numerical value 

 may be anywhere between 1000 and 2000. 



The values of a^a 2 for the series I. 1, 2, 3, previously 

 referred to, are given below. The results are not in them- 

 selves very reliable, but they seem to corroborate those of 

 the above series. 



Series I. 1, 2, 3. 



Temperature range increases from 1-3. 

 Wave-length range 4070-5400. 

 The actual temperatures at which the photographs were 

 taken were not in this case recorded. 





2e 2 = 



•27038 





Xe/\ 



= -•51240 





Zef= 



-11-534 





S./A 



= 28-128 





Xf = 



1268-94 





Xl/X 



»= 1-0609. 







"Een. 



2/k 





s»/x. 



1 





-11-800 



3766 





21-043 



2 





-11-484 



327-3 





20067 



3 





-14-583 



31907 





24-394 



Je. a, /«„. 



1 



- 6249 



-•10136 



- 7-663 



616-5 



2 



- 66 83 



— 12917 



- 9 946 



517-4 



3 



- 102-27 



-•22538 



-20-430 



453-76 



These values give fairly good agreement for observed and 

 calculated values of n for all three curves. It is seen from 

 the values of a x that the first two curves lie very close 

 together. The amount of dispersion in these photographs 



