332 



Mr Filon, On the variation with the wave-length 



In one case, namely S 57, the observations require a parabolic 

 curve, without a shadow of doubt. 



The observations, however, fall into three groups, which are 

 best fitted by three different parabolas. One of these accounts 

 fairly well for the three sets A., B. and E. of Table VII. If we 

 take it as giving the best determination, then the stress-optical 

 coefficient for this glass is given by the formula 



C=- 1-070 +316 



\ - 5400 

 1000 



•1375 



\-5400y 

 1000 / ' 



As this glass has been examined by F. Pockels (Annalen der 

 Physik, April, 1902) it was of interest to compare his values with 

 mine. I therefore reduced his results so as to bring them into a 

 form comparable with mine. 



He determined the coefficient C for three wave-lengths only, 

 namely those for lithium, sodium and thallium light. On exami- 

 nation of his reductions, I find that his assumed wave-lengths 

 (although they are not explicitly stated in the paper referred to) 

 were probably \ 6730 for lithium and \ 5360 for thallium. 



From these I obtained the following 



Glass S 57. 



Values of G. 



Thus, the absolute values found by Pockels for the stress 

 optical coefficient are nearly twice my own. It may be that 

 I have omitted a factor \ in the reduction of Pockels' observations, 

 although, on further examination, I am unable to find any such 

 omission and, in fact, the reduction, for this particular glass, has 

 been practically done by Pockels himself, for in one place he gives 



the values of his (8 X — B z ) -~ - — for S 57, and this, multiplied by 



the wave-length of sodium light, gives my C. The two sets of 

 values agree fairly nearly, as they should. 



On the other hand, I am quite satisfied that no error of this 

 kind has entered into the reduction of my own observations. 



The only possible conclusion, so far as I can see, is that the 



