384 PROCEEDINGS OF THE AMERICAN ACADEMY. 



through the sodium tube, after which it was brought to a focus by a 

 second mirror on the slit of a quartz spectrograph. The tube was heated 

 by a row of small Bunsen burners, and as soon as a steady state was 

 reached, which could be told by observing the curved branches of the 

 spectrum with an eye-piece, the photographic plate was inserted and ex- 

 posed. Owing to the low reflecting power of the silver the ultra-violet 

 region was somewhat under-exposed. The curvature of the spectrum in 

 the immediate vicinity of the first two ultra-violet lines of the principal 

 series was very pronounced, however. The plates obtained with the 

 small quartz spectrograph are too small to reproduce well, especially as 

 the density is not very great in the most interesting region, and it has 

 seemed better on this account to make drawings from them. The general 

 appearance of the photographed spectrum is shown in Figure 2, which 

 has been enlarged about two diameters. It is apparent that the effect of 

 these lines on the refractivity of the medium is negligible except in their 

 immediate vicinity. 



Plates obtained in the manner described were useless for purposes of 

 measurement, as no record was made of the position of the spectrum on 

 the plate before its deviation by the sodium prism. 



As the displacement is very slight in the remote ultra-violet, the two 

 records would overlap in this region, even if the spectrum was very 

 narrow. To overcome this difficulty the following method was devised. 

 A slide was fitted to the plate holder, perforated with a number of verti- 

 cal slits, each about 2 mms. wide, the widths of the opaque bars and 

 slits being equal. This slide was left in the plate holder during the ex- 

 posures, and screened off the spectrum at regular intervals. In order to 

 obtain greater deviations a large tube of Jena glass 80 cms. long was 

 used. It was heated in a combustion fuinace, the burners of which were 

 turned down low, the best results being obtained when the tops of the 

 flames only partially enveloped the tube. 



As it was desired to extend the observations as far down into the ultra- 

 violet as possible the spark was substituted for the arc, and to avoid loss 

 of light by reflection one of the mirrors was dispensed with. 



The spark was furnished by a large induction coil operated with an 

 alternating current of 110 volts, a Leyden jar of large size having been 

 joined in parallel with the terminals. Cadmium electrodes were used, 

 the spark gap being placed close to the horizontal slit, which was mounted 

 at the centre of curvature of a concave reflector of magnalium of one 

 meter radius. The convergent beam from the reflector passed through 

 the sodium dispersion tube and came to a focus on the slit of the spec- 



