TABLE OF STANDARD WAVE-LENGTHS OF THE SPECTRAL LINES 563 



between the radius from the centre of the photographic plate and the 

 line drawn from that point to the centre of the grating. When prop- 

 erly adjusted, f will be zero. Also, we make 2 r = R, to obtain perfect 

 focus throughout. So that 



/ = ^ (sin 4- sin -g-j . 



nN\ 2 / 



Calling ^ O the wave-length at the centre of the plate, we have ap- 

 proximately 



* 



The first quantity, - , is the value of / ^ , assuming the spectrum to 



^ ?l jLJ 



be normal. The last term is the required correction expressed in terms 

 of the provisional wave-length. The correction in actual practice has 

 been made from a plot of the correction on a large scale, and never 

 amounted to more than a few hundredths of a division of Angstrom, even 

 for the longest plate. 



In two or three plates the camera was displaced, so that 7- had a value. 

 In such cases no attempt was made to measure f , but the plates were 

 only used for local interpolation by drawing a curve through certain 

 points used as substandards. 



These substandards Mere principally used for working up the last 

 set of photographic plates containing the solar spectrum and the metal 

 spectra of the same or higher orders, or both. Some of them contained 

 three metallic spectra. 



Thus the region 3900 in the solar spectrum has been obtained from 

 both wave-lengths 5200 and 5850. The mean of these gave values of 

 the substandards for working up the plates taken at this point, and 

 containing also metallic lines at 2700. 



Again, the boron lines 2496 and 2497 have been obtained from the 

 regions 4800, 3200 and 3600. The mean values give substandards for 

 working up the metallic spectra of that region. Also the near coinci- 

 dence in the values of the wave-lengths of these lines indicate the rela- 

 tive accuracy of the regions 2496, 3200, 3600, and 4800. 



The use of these substandards is as follows: The photographic plates, 

 mostly 19 inches long, were measured mostly on a machine giving wave- 

 lengths direct. The differences of the results from the substandards 

 were then plotted on a paper having the curve of correction for length 

 upon it in such a way that the final marks should theoretically be a 

 straight line. This was actually the case in all but a few plates, in 



