478 Mr. G. W. Walker on the 



greater than the value of K — 1. Aoain, since fl — Et^\ dimi- 



P 



nishes as &)^ increases we require that should diminish 



with increasing temperature. We also require that -~^ 



should be less than one. The first portion of the curve has 

 thus comparatively limited power of explaining the actual 

 tacts, and the conclusion that 2^'"^ should be less than 0)'^ is 

 highly improbable. 



Turning now to the portion of the curve beyond A, we see 

 that refraction increasing with the frequency may be ex- 

 plained. There the values of _ are greater than unity^ and 



the function M — E^-J is negative. 



To explain a refractive index greater than unity we re- 



7. 2 



quire that k^ p should be greater than '?^ \\. — 'E]~-A in the 



spectrum. Now this can be secured even when -^ is 



u 



greater than kip by taking 4 sufficiently large. Thus we 



may have ^- as the important term in K— 1, but kip will 



be the important term in the visible spectrum. 

 Writing the formula 



we see that the variation with temperature will be greater 

 or less than that indicated by Gladstone and Dale's law, 

 according as the coefficient of ki p diminishes or increases 



2 / 2 



with temperature. That is according as ir^/l — E?^) dimi- 



nishes or increases in numerical value as 4;, increases. Re- 

 ferring to the curves we see that the minimum point marked 

 B on the curve LAl — 'E^—^] occurs for a value of 4 



ft)^\ ft)"/ CO 



greater than that for the point A on the curve (1 — Ec^j. 

 Hence, for values of 4 between those corresponding to the 



