348 Messrs. Trowbridge and Sabine on Wave-lengths 



the pointer of the carriage on the scale beside the track indi- 

 cated that light of wave-lengths 4200 to 4800 in the first 

 spectrum and 2100 to 2400 in the second spectrum would fall 

 on the plate. The shutter was turned so as to expose only 

 the lower half of the plate and a photograph of the solar 

 spectrum from 4200 to 4800 taken. The shutter was again 

 turned, and the upper half of the plate given a long exposure 

 to the light of the spark. Both spectra were in focus. The 

 wave-lengths of the metal lines were then found directly, by 

 interpolation on the normal spectrum, from the solar lines 

 whose values were given in Rowland's Photographic Map 

 and table of wave-lengths *. The interpolation was made by 

 means of measurements on a divided engine. In order to 

 correct for any displacement due to the motion of the spark 

 from side to side, or to jarring arising from the great noise of 

 the spark, and also in order to sift out the lines belonging to 

 the first spectrum from those belonging to the second, the 

 girder was moved to the violet of the third, with its magnified 

 dispersion and different underlying spectra. The metal and 

 solar lines were taken side by side, and the interpolation for 

 the wave-lengths of the metal lines made as before. From 

 this the correction to be applied to the previous plate was 

 found, amounting in some cases to '2 of a tenth metre. The 

 correction thus found was applied to all of the lines on the 

 plate. The girder was now moved so that the sensitive plate 

 was in the extreme ultra-violet of the first spectrum, and the 

 plate exposed to the light from the spark. From this nega- 

 tive the values of the wave-lengths of the faint lines were 

 obtained by interpolation from the values of the stronger lines 

 as determined by the first plate. It also served as the final 

 test whether the lines on the first negative w r ere of the first 

 or second order. All of the lines more refrangible than line 

 2123*1 were in the case of copper found from this negative 

 and from line 2136*1 by direct measurement. 



Another method of distinguishing which lines on the first 

 negative belong to the second and which to the first spectrum, 

 is to place in front of the slit while taking the metal lines a 

 piece of plane glass. The second spectrum for this refrangi- 

 bility will be completely cut out, and only the metal lines of 

 the first remain, being in the visible violet. 



The only source of error was in the setting of the micro- 

 scope upon the broad or faint lines. The probable error of 

 this is about *1 tenth metre. For the few most refrangible 

 lines it may be greater. 



* American Journal of Science, March 1887 ; Phil. Mag. March 1887. 



