LYMAN. — FALSE SPECTRA FROM THE ROWLAND GRATING. 243 



trace of the lines could then be found. Since the light causing these 

 lines is absorbed by glass, and is not absorbed by air, it would seem to 

 have a wave length between 3500 /x and 2000 /x, according to the accepted 

 absorbing power of these two media. 



It was now observed that the groups at 2790 /x, 1873^, and 924 jx 

 were, even to casual inspection, strikingly characteristic and identical in 

 general appearance. Further, on measuring their dispersions, they were 

 found to be proportional to the displacements of the grorps from the 

 slit. Therefore, these groups formed either a remarkable series repro- 

 duction or a heretofore unobserved diffraction phenomenon. Thus the 

 issue was definitely raised whether these lines were real first spectrum 

 lines, produced by light of wave lengths 924 /x, 1873 /x, and 2790 /x,, and 

 arranged in al, 2, 3 series, or whether they were all produced by light 

 of one wave length and owed their position to some diffraction phenom- 

 enon connected with the grating. 



To distinguish between these alternatives, recourse was had to the 

 relative refrangibility of these lines through quartz. To this end a 

 small angle quartz prism was placed between slit and grating and a 

 short distance from the slit. The effect of this prism was to produce 

 virtual images of the slit whose displacements were nearly proportional 

 to the refrangibility of the light. This device resulted in the displace- 

 ment of each line in the grating spectra by an amount proportional to 

 the refrangibility of the light producing that line. 



If the lines at 1873 /x and 924 /x were caused by light of the same 

 wave length, they would suffer equal displacements. If they were 

 caused by light of different wave lengths, they would be displaced by 

 different amounts. 



The experiment was tried, and it was found that the groups at 2790 p, 

 1873 /x, and 924 /x all suffered about equal displacements. From 2790 /x 

 down to the region of 1873 /x, the displacement of the groups increased 

 as the wave length decreased. There was then a discontinuity in the 

 rate of increase of the displacement before 1873 /x and a second discon- 

 tinuity between 1873 /x and 924//. 



The most obvious conclusion from these data is that the lines at 1873 /x 

 and 924 fx are not due to light of wave lengths corresponding to the 

 positions of the lines in the spectrum. They seem to be caused by 

 light having a wave length near 2800 /x ; they seem to be reproductions 

 of some part of the true spectrum, — curious phantom liues due to some 

 property or imperfection of the grating. An examination of the lines 

 below 1900^ taken with aluminum terminals fully carried out this 



