58 PROCEEDINGS OF THE AMERICAN ACADEMY. 



band advancing down the spectrum. Finally by further increase the 

 red appears again in full intensity, being reflected unpolarized, and 

 mixing with the metallically reflected violet gives the brilliant reddish 

 purple. Very similar appearances can be observed with selenium, 

 though the colors are not as saturated and not nearly so brilliant. 



Having roughed out, so to speak, the dispersion in the ultra-violet by 

 photographing the spectrum of the reflected light at various angles of 

 incidence through a Nicol prism, an attempt was next made to secure 

 more reliable data by means of the method of crossed prisms. A prism 

 of small angle, held in a clamp, was mounted with its refracting edge in 

 a horizontal position, immediately behind the prism of the quartz spec- 

 trograph. A fine platinum wire immediately below the edge of the 

 prism, when heated by a current of suitable strength, kept the nitroso in 

 a fluid condition. A system of small screens was arranged so that the 

 light could either be made to pass through the nitroso prism, or through 

 a small clear space in the quartz plates immediately above it, thus pho- 

 tographing the deviated and uudeviated spectra one above the other. 

 The current strength was so adjusted as to keep the temperature of the 

 prism as nearly as possible at the melting point of the nitroso. It was 

 possible in this way to give exposures of an hour or more, and obtain a 

 photographic record of the dispersion curve from the orange down to the 

 extreme end of the ultra-violet. The cadmium spark was employed as 

 a source of light, and the slit of the spectrograph was diaphragmed down 

 to a length of less than a millimeter in order to obtain very narrow 

 spectra. As the conditions necessary for success were determined by 

 repeated experiments, prisms of larger angle were used, and the two 

 spectra, which overlapped in the first experiments in the region where 

 the refractive index had a low value, were completely separated. I 

 found that a great improvement resulted from carefully grinding the 

 edge of the quartz plate, which rests against the other plate, perfectly 

 straight with fine emery. It was not until this expedient was adopted 

 that the larger angles became possible. The deviated spectrum is quite 

 sharp except close to the edge of the absorption band, where absorption 

 jiroduces a broadening of the image by reducing the effective width of 

 the beam of light, as is always the case with strongly absorbing prisms. 

 It is even possible to follow the general trend of the dispersion curve 

 right through the baud of metallic absorption, though the broadening 

 resulting from diffraction is, in this region, so great that accurate meas- 

 urements were impossible. The continuity of the curve is shown, how- 

 ever, better than I have ever seen it in any photograph. To deduce 



