14 Lord Rayleigh on Reflexion from Liquid Surfaces 



to suppose that a difference of one minute in a can be detected 

 on repetition. From this we should get, by (14), as the least 

 observable value of k, 



k= J tan 30° x tan V = -00009. 



Jamin's arrangements do not appear to have allowed of the 

 determination of values of k less than *001. 



The first systematic experiments upon cleaned water- 

 surfaces showed that the ellipticity, if real, was pretty close 

 to the limit of observation. At this stage I expected to find 

 the marked ellipticity of greasy water gradually diminishing 

 to zero as the purifying process was carried further, but 

 remaining always of the same sign, so long as it could be 

 observed at all. This anticipation was not completely verified. 

 The larger differences of a, found with ordinary water upon 

 which camphor fragments were fully active, amounting say 

 to 40', rapidly diminished under the skimming process, so that 

 the final difference on the purest surfaces seemed just to 

 escape direct observation. It frequently happened that no 

 displacement of the dark spot relatively to the cross wires 

 could be detected on the reversal of /3. But when, in order 

 to the highest accuracy, many sets of alternate readings were 

 taken, the difference would come out sometimes in one direc- 

 tion and sometimes in the other. A small difference of 2', or 

 more, on the side of the contaminated water was easily 

 accepted as due to incomplete cleaning, but I was for a time 

 sceptical as to the significance of similar small differences in 

 the opposite direction. That these differences were not errors 

 of observation was soon apparent ; but I thought that they 

 might be of instrumental origin, due perhaps to some mal- 

 adjustment. 



The outstanding question was so small that it might per- 

 haps have been dismissed, but I was unwilling to stop without 

 a determined attempt to get to the bottom of it. The minute 

 reversal of ellipticity stood its ground in spite of repeated 

 remoun tings of the apparatus ; but I feared that it might 

 possibly be due to some, though I was unable to conjecture 

 what, defect in the optical parts themselves. When, however, 

 the nicols at first used were replaced by beautiful prisms 

 made by Steeg and Reuter and the effect still remained, it 

 had to be accepted as genuine, and the conclusion was forced 

 upon me that with some water-surfaces, and those presumably 

 the cleanest, there is a minute ellipticity in the opposite direc- 

 tion to that of ordinary water, and such that the difference of 

 settings for « amounted to about 2'. This corresponds to 

 k= +'0002. It will be understood that this is a very minute 



