96 Wheeler — Kej>eri 'mental In restitution, on the Reflection 



spherical form when placed on it, unless it has heen recently 

 cleaned in the manner described. It is not necessary to start 

 with especially clean mercury, — it cleans itself in a very short 

 time. Nor is it necessary to test the cleanliness of the surface 

 except after each fresh filling of the reservoir. A good surface 

 gets contaminated in the air in a brief time, but retains its 

 cleanliness long enough to permit the necessary observations to 

 be taken comfortably ; while a fresh surface can be obtained 

 easily and quickly. 



Thus it is reasonably certain that the reflecting surfaces used 

 were free from surface contamination. Of course it is not so 

 certain, a priori, that the film of gas was removed when jthe 

 mirror was submerged. It is natural to expect, however, that 

 the rapid overflowing of the mercury would tend to carry away 

 any such film and leave a more perfect contact between the two 

 media. And from the final result of these experiments it would 

 seem that this expectation is fully justified. During the course 

 of a set of observations under a liquid it was of course neces- 

 sary to close the overflow tube (D), consequently the cylinder 

 (C) was made large enough to hold all of the mercury which it 

 was necessary to allow to overflow in the course of such a set. 

 The liquid thus displaced from the box (E) was made to over- 

 flow at one corner and was caught in a suitable vessel. 



In order to avoid any deviation of the beam of light by 

 refraction at the free surface of the liquid, glass tubes termi- 

 nated by plane-parallel glass ends were carried in suitable brass 

 holders by both the collimator and telescope arms ; when the 

 mirror was in position the ends of these tubes were just sub- 

 merged in the liquid in (E). The length of the box (E) was 

 made as great as the space between the tube-holders would 

 allow, in order to obtain as large an angle of incidence as pos- 

 sible. The space thus available was, however, too small to 

 permit the use of angles of incidence greater than 60°, which, 

 though lessening the precision attainable in this kind of exper- 

 iment, still yielded ample accuracy for the ends in view. 



The apparatus for measuring the phase difference and azi- 

 muth of restored polarization was of a familiar type and needs 

 but brief description. The spectrometer* was one by the 

 Societe Genevoise, arranged so that the table could be swung 

 to a vertical position to accommodate a horizontal mirror. The 

 collimator arm carried the polarizing nicol behind the lens, the 

 Babinet compensator was mounted before the objective of the 

 telescope, while the analyzing nicol was carried at the other 

 end of the telescope tube. The nicol and compensator came 

 with the spectrometer and were of good quality. By means of 



* This was obtained by the aid of a grant from the Baehe fund of the 

 National Academy of Science. 



