154 



MEMOIliS OF THE NATIONAL ACADEMY OF SCIENCES. 



(r = 450 10', ;)/ = 6.0), &c., it was concluded that the deviation of rays whose wave-leugth is 1.767 

 is 450 10' ; and each point in this detennination being obtained from the mean of five observations 

 the result is partly free from irregularities caused by changes in the state of the sky, and minute 

 instrumental variations from extraneous causes, which here become of great relative importance, 

 owing to the feeble heat measured. 



Subsequent determinations, like the preceding, gave for the deviation of the same ray 45'^ 06' 

 and 450 07', and from a consideration of all, the deviation adopteil was (instead of 45° 21', as given 

 by Bbiot's formula) 45° 08', corresponding to a refractive index of 1.5.149. 



By means of measurements like the one described above, the deviations of various obscure 

 rays of known wave-lengths were determined. The indices of refraction were then computed by 

 the usual formula 



where 



sin ^ a 

 a = 620 34' 43". 



The results are contained in the following table, where, however, only the results of successful 

 days are given, most of the observations having been lost through changes of the sky during the 

 course of one determination.* 



Table II. — Uxperimental determination of d or n as a function of \ [Hilger prism). 



Date of observation. 



April 1, 188-2 



April 9 



June 27 



Juue 13-27 . 



Jnly 14 



June 7 



In. 010 J- 0. 0053 

 1/i. 200 J- 0. 00(39 

 1//. (i58 -J- 0. 0091 

 \^i. 767 ± 0. 0094 

 -in. 090 ± 0. 0104 

 2//. 356 ± 0. 0110 



1. 5654 

 1.5625 

 1. 5562 

 1. 5549 

 1. 5511 

 1.5478 



We observe that where measures are taken in the prismatic spectrum alone, we can generally 

 use with advantage a bolometer of as small an aperture as oneflfth of a millimeter, but that here 

 it is advisable to open it to 2""", owing to the relative expansion of the spectrum and to the very 

 feeble heat. 



Owing to the difiSculties arising from the almost infinitesimal amount of heat in question, 

 numerous subsidiary observations are requisite for a single determination, which it therefore takes 

 long to make, each final value resting npon between 20 and 100 readings. If it should possibly 

 appear to the reader that in the three months ot consecutive labor which were given to this part 

 of the work, more than six points might have been determined in the curve, he is asked to 

 remember that what is here difficult has till now been impossible. 



Plotting the points given by the data in Table II, and drawing a smooth curve through them, 

 we obtain the curve of "observation," showing n as a iuuction of A in the lower curve of Plate 2 

 and d as a function of A in the curve of Fig. 3, M-here the points obtained by observation are dis- 

 tinguished by small circles. 



There would be no gain in accuracy at this stage in attempting to work from a formula repre- 

 senting the eciuation of the curve obtained, as the graphical construction is fully as trustworthy 



'AH these observations for discovering the relation betvreen n and 1 can be conducted with at leasf as much 

 advautugo by a powerful and constunt electric li'^ht as by suuliyht. The latter only, however, was at the observer's 

 actual command. 



