and Dispersion in certain Metals. 367 



centim., occupying the same relative positions as the prisms 

 and windows of our metallic biprisms (see Plate XII. fig. 1). 

 The collimation of the two outer windows was adjusted as 

 usual, with the result that the inner pair of windows now gave 

 an apparent angle of *2", and an apparent deviation of *4". A 

 platinized glass plate treated in the same manner gave values 

 1"*3 and 1"*8 respectively; this had been previously labelled 

 " middling," and was less perfectly plane than any of those 

 carrying the biprisms finally experimented upon. The latter 

 plates were of course carefully selected and tested by means 

 of a Gauss's eyepiece. 



These observations prove the method of collimation to have 

 been sufficiently accurate for the object in view, as the re- 

 fracting angles of the prisms * lay between 15" and 25", the 

 deviation generally exceeding 20" (up to 200", compare § 11). 

 As regards the measurement of the angles and deviations, it 

 may be remarked that these cannot appreciably have been 

 interfered with through multiple inner reflexions. In fact 

 the prisms were in every case so thick that films of double 

 the thickness would have proved almost quite opaque f. 



§ 5. It is impossible to give a general estimate of the 

 accuracy reached ; each prism exhibiting its particular indi- 

 viduality in this respect* In order therefore to enable the 

 reader to form a judgment on our measurements, we have 

 given a few detailed sets of observations (Tables I. and III.) 

 besides the final values (Tables II. and IV.). 



Our prisms were always kept in exsiccators, and proved to 

 remain quite unaltered when thus treated. Thus we found 

 the refracting angle as well as the deviation of a cobalt prism 

 (Co II].), which had been left to itself for six months, not to 

 have changed appreciably during this interval. The same 

 was the case with an iron prism (Fe IIL)> the black coating 

 of which had moreover been removed and replaced by a new 

 one : this also proves the surfaces of the metal to have been 

 sufficiently plane, as the reflecting patches left open on blacken- 

 ing certainly had not exactly the same boundaries in both cases. 



We may remark that certain other experiments were occa- 

 sionally made in this laboratory by Mr. Shea, using our prisms 

 with a spectrometer by Schmidt and Haensch ; these always 

 quite confirmed our results. 



* As it is necessary to keep the thinner part of the prisms >3xl0-6 

 centim. and the thicker part ■<: 13 X 10- 6 centim., it follows that prisms 

 with an angle much exceeding the highest value given in the text cannot 

 be used. The mean breadth of the prisms may be estimated at - 15 centim. 



t Compare Wernicke, Pogg; Ann. clV. p. 88 (1875) ; also Rathenau, 

 Inaugural dissert. Berlin, 1889. 



2C2 



