560 Prof. A. A. Michelson on Metallic 



The quantities in the last column are proportional to k ; 

 but the actual values of Jc thus deduced from observations of 

 transmitted light are considerably less (about 1/3 of the 

 value given), possibly because of the unevenness of the film 

 which makes the measurement of the actual thickness (of 

 the order of one thousandth of a millimetre) uncertain. 



The agreement in the last tvvo columns of the tables, while 

 somewhat imperfect, is still enough to show that the results 

 are of the right order of magnitude- — and if it be remembered 

 that the properties of the specimens vary considerably with 

 the method of preparation, it is probable that the outstanding 

 differences may be thus accounted for. 



In any case the agreement is much better than it would 

 be, had the ellipticity been attributed to absorption alone. 



In the aniline colours the absorption varies enormously 

 with the colour, and we have all the gradations from metallic 

 reflexion to almost perfect transparence combined in a single 

 specimen. The measurement of the phase-change and the 

 amplitude- ratio for these substances show changes in the form 

 of the curves almost identical with those given in the 

 preceding figure. 



PI. IV. fig. 2 shows the curves obtained for fuchsine and 

 for diamond green. It may be noted that in both these 

 figures the " phase " curve is much more characteristic in its 

 changes than the "amplitude" curve. 



These specimens are prepared by dissolving the aniline 

 colour in hot alcohol, filtering hot, and covering a hot glass 

 surface with the solution. The alcohol evaporates rapidly, 

 leaving a mirror surface of thickness of the order of a 

 thousandth of a millimetre. 



A quite remarkable alteration occurs in the phase curves 

 when the solution is diluted. The film deposited is then 

 very much thinner than before (ftom one-tenth to one one- 

 hundredth of the former thickness) and, for some colours, 

 the thickness is so small that considerable light is reflected 

 from the surface of the glass. The resulting phase curve 

 may then be negative, as shown in PI. IV. fig. 3, for the 

 colours red, orange, and yellow*. 



Such a result has been predicted from theoretical con- 

 siderations f, but so far as I am aware, no attempt has been 

 made to show that this depends on the colour of the incident 



* The lower curves show more clearly how the maximum value of k 

 varies with the colour. 



t Drude, ' Theory of Optics,' p. 204. 



