250 PROCEEDINGS OF THE AMERICAN ACADEMY. 



in^ was next turned about a verticil axis until the direct ima^e of the 

 slit fell upon the plate. No lines below group (', wen- revealed by this 



process, • was tin- position of the lines greatly changed. It must be 



noted, however, that the turning of the grating will somewhat distort the 

 tra and change the value of /u to cm. 

 The second grating to be investigated was <»f 21 ft. radius, l 1.438 

 lines to the inch. It was arranged on Rowland's mounting. The in- 



tensity of the m tra given by this grating was verv inferior to that 



obtained with the 6 ft. instrument. Tin' first Bpectrum was selected to 

 work with, the magnesium -park was used as ;i source. It was found 

 necessary to give an exposure of an hour, and to use a slit ."_' nun. in 

 width. A good deal of care was taken to have the exposure of the 

 proper length : if too short, the lines themselves would not appear ; if too 

 long, the- lines would be obscured by fog. With this 21 ft. grating, a 

 verv little fog would quite shield any very fainl line. After some trouble, 

 the characteristic reproductions "f group 2790 were found. They now 

 irred, however, at positions corresponding to wave lengths of 

 1728 /x, 1079/1. The widths <>f the groups were 4.8mm., 2.8mm., 

 1.8 mm. Thus, though these phantom lines have different apparent wave 

 lengths when measured on different gratings, their dispersion seemed 

 proportional to their wave length. It must be noted that with grating 

 No. 2 only the lines called A,. (',, and A 4 , C 4 are visible, the other 

 fainter lines, as well as the llu ed effect in the background, being totally 



absent. This is well accounted for by the feebleness of the spectrum 



obtained from grating No. 2. The apparent wave length of these groups 



obtained by the same method as that previously employed. Owing 

 to the great distance between the head group (2790) C and its reproduc- 

 tions, a cathetometer was used. The results are correct, however, to 

 better than - A in units. The widths of the groups were measured 



on the dividing engine and are correct to within .03 mm. The value of 

 1 cm. of plate length was obtained from measurements between two sun 

 lint 



I feel certain that groups 2790 /x, (', and (', a- obtained with grating 

 No. 2 are not in 1:1': .'! ratio. 



The next grating Investigated was one of 21,000 line, to the inch and 



_'l ft radius. Here the illumination was even more feeble than with 



ting No. 2. Great trouble wi erienced from fog. The lines in 



magnesium, C] and C 4 , COuld not be found. The last grating examined 



tnilar to No. I. It had a radiuB of 183 cm., 14,438 lines to the 

 in<h; it was prepared by Brashear; i; was ruled by Rowland's engine at 



