and Dispersion in certain Metals. 371 



at any angle to the left ( + ) or right ( — ) of the telescope's 

 axis. After a few preliminary trials we made regular measure- 

 ments for values of i=0°, 30 6 , 40°, 50°, 55°, 60°, 65°, because 

 with these intervals the increments of the deviation were most 

 uniform. In certain cases we could still observe at 70° inclina- 

 tion ; the deviations obtained in this case were considerable, e.g. 

 with a cobalt prism we found a = 200", i. e. over three minutes. 

 However, the projection of the prism on the objective lens 

 now becomes too narrow ; consequently the image of the slit 

 is so considerably drawn out by diffraction that measurement 

 becomes uncertain and inaccurate notwithstanding the strong 

 deviation. 



For these observations the source of light was a Linnemann's 

 zirconia burner seen through red glass. For the reasons 

 stated above (§ 5) a complete set of observations is given in 

 Table I. It will be seen that each observer determined one 

 deviation for + and one for — inclination (each by 20 read- 

 ings) ; these are seen to agree on the whole. The angle /3 was 

 measured three times by each of us, the deviation with normal 

 transmission four times ; this was done on account of the pre- 

 ponderating influence of these values on the final results. 

 The mean values for the two observers were used in the subse- 

 quent calculation. 



We have carried out this part of our work with but one 

 prism of each metal, while using three for the measurements 

 of dispersion. In the latter the results prove quite inde- 

 pendent of the refracting angle (§ 18), and it may therefore 

 be inferred that such is also the case with oblique transmission. 

 It was not so much our object here to find numerical constants 

 as it was in the second part of our work, where a great 

 number of observations had to be collected in order to obtain 

 sufficient accuracy*. 



§ 12. The results for the three metals are given in Table II. 

 From the observed values of a (first horizontal line in each 

 section of the table) values of n were first calculated by 

 equation (3) (third line); this was done only in order to see 

 whether Snellius's law might apply. Now it is evident that 

 for nickel the values of n at last begin to increase with i ; 

 such is also the case for cobalt, though to a less degree. For 

 iron the deviations from Snellius's law appear to lie within 

 the errors of observation. However, we feel bound to assume 

 that the latter metal differs from the two others only quan- 

 titatively, not qualitatively. 



* Compare the check by an independent observer mentioned in the 

 last paragraph of § .5, which was applied here. 



