274 DR. L. N. G. FILON ON THK 



dispersion between the lithium and thallium lines. The error will be greater when 

 we reach the violet end of the spectrum, but it will still be too small to affect the 

 observations. 



In experiments demanding great accuracy it might be desirable to polarize the 

 light before it passes through the slit. The accuracy possible under the present 

 circumstances did not seem to justify this additional complication in the apparatus. 



5. Effect of Finite Breadth of the Source. 



It may be shown that if the slit have a finite breadth '2e, the intensity of the light 

 that gets through is proportional to 



sin q0 







where 



= 27rK /A, g = o-e 



HO being the relative retardation corresponding to the middle of the slit. 

 In the actual case g = J-^Q approximately. 

 The minima are given approximately by 



= 2nrr (!-&). 



The proportional error in the wave-length of extinction is therefore BTSOO- which 

 is negligible. 



Also the minimum no longer corresponds theoretically to perfect darkness, but with 

 a slit between ^ millim. and 1 millim. wide the bands were very dark and quite 

 definite. 



6. Effect of Relative Rise and Fall of the Two Beams and of Elastic Yielding of 



the Bed-plate and Knife-edges. 



Owing to the elasticity of the glass, the middle parts of the beams will undergo a 

 vertical shift owing to flexure, and the bed-plate and apparatus as a whole will sink. 

 In consequence we have variations 8z,, Sz 2 , h depending on the applied load. 

 Thus the right-hand side of equation (8) is multiplied by a factor 



This may be allowed for by supposing M 2 (or W) multiplied by the same factor, 

 equation (8) remaining otherwise unaltered. 



The effect is then to add to the applied weight a correction 



W[8z I -8z 2 +o-(8z l -8/ t )]/[z 1 -2 2 +o-(z l -A)]. 



Now the relative rise and fall of the beams themselves is an elastic effect and may 

 be taken, in such a small correction, strictly proportional to the load. 



