118 Mr. G. F. Fitzgerald, 



cally polarized. Now, the amount of elliptic polarization, we may 

 expect, can be calculated in the following manner: — I have esti- 

 mated by a series of unfortunately rather rough experiments that 

 when a red hot ball is plunged into cold water to a depth of half 

 a centimetre, the thickness of the Crookes' Layer formed, is about 

 a quarter of a millimetre. This is, of course, only a rough ap- 

 proximation, but it will give us results which determine with 

 what order of quantity we are dealing. Hence, the excess of 

 pressure in the vertical over that in the horizontal direction that 

 I measured was half a gramme per sq. centimetre, which is about 

 the '0005 of the atmospheric pressure. Now, I will make an as- 

 sumption which is, however, only partially true, but as one object 

 of the experiment is to determine to what extent it is so it is legiti- 

 mate provisionally, and it is that we may treat the strain in the 

 gas as due entirely to a difference of density in different directions. 

 That we may do so to some extent, at least, is manifest, for, ac- 

 cording to theory, the number of molecules moving in the direction 

 of the strain is greater than the number moving in other di- 

 rections. To what extent this is true could only be determined 

 either by elaborate theoretical investigations into the state of the 

 gas or else by experiments such as I am proposing. Assuming 

 then the strain to be wholly due to a difference of density we 

 can proceed as follows : — The law connecting the refractive in- 



— 1 — ' — 1 

 dices of a gas at different densities is ~r~ ~ ,, ' * so that in air 



when /* = 1*0002940, and in the case we are considering where 



-^-=1-0005 we have p! = 1'0002941. As there are 100,000 



vibrations of light per five centimetres in vacuo there will be 

 100029*40 when the density is /*, and 100029*41 when it is /, 

 and consequently a difference of phase of *01 of a wave length 

 will be introduced per five centimetres or a twentieth of a 

 wave length in 25 centimetres. Now, as the intensity of light 

 in the analyzer depends upon the square of the sine of the 

 difference of phase, this will give the intensity as the square 

 of the sine of 9°, which is *02 or -^-th. Hence I conclude 

 that if a ray of plane polarized light be transmitted through 25 

 centimetres of a Crookes's Layer between two surfaces, one of 



