102 Prof. Wood on the Dispersion, Absorption, 



spark was reflected into the slit of the instrument at various 

 angles of incidence, and dark bands were found running 

 across the two polarized spectra furnished by the Rochon 

 prism. I spent some time endeavouring to extract data from 

 these bands, and finally came to the conclusion that either 

 they were not due to the selective polarization by reflexion, 

 or else that the dispersion-curve as determined for selenium 

 by me^ns of the interferometer was in error. On experi- 

 menting further, I found that these bands were due solely to 

 the rotatory dispersion of the light in the quartz collimating- 

 lens, some colours being rotated through such an angle as to 

 be quenched in one spectrum by the Rochon, and others in 

 such a degree as to be absent in the other. I mention this 

 defect in the instrument, as it may be of interest to others 

 working along similar lines. The proper design of the in- 

 strument should have called for a collimating-lens made of 

 two thin lenses, one of right-handed, the other of left-handed 

 quartz. I remedied the defect in my instrument by placing 

 a plate of left-handed quartz immediately behind the lens. 

 This plate had a thickness equal to the thickness of the lens 

 at its centre, and abolished the bands entirely, when the lens 

 was stopped down to a small area at the centre. 



I found, however, in working with the horizontal cell of 

 fluid nitroso, that better results were obtained by using a 

 Nicol prism in front of the slit of the spectrograph than with 

 the Rochon prism. The nicol was made transparent to ultra- 

 violet light by separating the two halves, cleaning off" the 

 balsam, and substituting glycerine. 



Evidence of the very low value of the refractive index on 

 the ultra-violet side of the absorption-band was obtained 

 before any polarization experiments were tried. The light of 

 a cadmium spark was reflected from the pool of liquid nitroso 

 at nearly normal incidence, and then thrown into the spectro- 

 graph by means of a quartz total-reflecting prism. A series 

 of spectra was taken, with times of exposure varying from 

 2 seconds to 3 minutes. By comparison of the difi^erent 

 spectra it was possible to form a rough estimate of the re- 

 flecting power in diiFerent parts of the spectrum. A similar 

 series was made with a flat reflector of magnalium, which is 

 remarkably constant in reflecting power throughout the entire 

 spectrum. It was found that the nitroso, in the region of its 

 absorption-band, reflected almost as strongly as magnalium, 

 while just on the ultra-violet side of the band there was a gap 

 in the spectrum, due to the very low reflecting power which 

 the substance has for these waves (Plate VIII. tig. 8). The 

 reflexion coefficient at this point was estimated at about 



