Propagation of Light in a Gravitational Held. 591 



Imperial College, most suitable for the purpose, for various 

 orientations of the polarizer, the eye being focussed, through 

 the analyser, on the light emerging from the polarizer. From 

 this procedure there accrued two advantages, as compared 

 with the usual method of observing the illumination, by 

 means of an eyepiece, in a plane near the analyser. It was 

 possible to separate the actual light under observation from 

 stray unpolarized light not coming from the polarizer. Also, 

 this stray light seen, as it was, close to, but not overlapping 

 the polarized beam, and suffering no change of brightness 

 when the analyser was rotated, served as a standard of com- 

 parison for testing the completeness of extinction in relation 

 to the orientation of the light vector. No difference of 

 extinction at all could be detected. A small speck of stray 

 unpolarized light (referred to above) was fortuitously in 

 existence by the side of the polarizer. For all orientations 

 of the polarizer it was possible to extinguish the polarized 

 beam to such an extent that it was definitely less luminous 

 than the stray speck mentioned. 



Measurements were now made to establish an upper limit 

 for the ellipticity of the light. It was found that in order 

 to bring about a revival Of light equal in luminosity to the 

 standard speck, a rotation of the analyser through o, 012, or 

 about 0*0002 radian was necessary. Thus the ellipticity 

 produced by gravitation was certainly distinctly smaller 

 th;.n 2xl0~ 4 *: 



Substituting this result, i. e. k <2 . 10 -4 with L = 4000 cm.. 

 and X = 5 x 10" 5 cm., say, in equation (9), we have 



^~^<8.10- 13 . 

 c 



In order to realize the meaning of this result, imagine a 

 light wave travelling under similar conditions for one second. 

 Then if one component lags at all behind the other the 

 difference would be smaller than three hundredths of a 

 millimetre in the three hundred thousand kilometres covered 

 in that time. There are, perhaps, few pairs of physical 

 magnitudes whose equality could be stated with equally 

 high precision. 



We take the opportunity of expressing our grateful thanks 

 to Messrs. Adam Hilger for the loan of the Babinet com- 

 pensator and the analysing nicol used in this investigation. 



London, 

 21 January, 1920. 



* It is perhaps worthy of notice that this corresponded to a relative 

 change of brightness 4x 10~ y , which could be detected by the eye 



