ON ODU KNOWLEDGE Off SPECTRUM ANALYSIS. 



325 



ment — the number, position, and refracting angle of the prisms, the 

 dispersive power of the material of which they are made, of variations in 

 the temperature, and of all other disturbing causes. It is clear that in 

 such a method each line can be mapped only by means of its colour, that 

 is to say, by the length of the wave of light hj which it is produced ; 

 and a spectrum so represented must be such a one as is produced by dif- 

 fraction, and not by dispersion. Dispersion-spectra obtained by the use 

 of prisms of different materials vary greatly in the relative breadth of 

 the colours, so that in mapping a spectrum it, is by no means sufficient 

 to give the positions of only two or three lines as points of reference. 

 Many otherwise valuable observations of spectra are entirely useless from 

 the insufficient number of reference lines observed. 



Three spectroscopes (each with a single prism and reflected scale), 

 constructed by Duboscq and intended to be exactly alike, differed as shown 

 in the following table. The numbers show the difficulty of constructing 

 two instruments with exactly similar scales : — 



Lines observed Spectroscope > 



In a diffraction'Spectram. the position of the lines is dependent solely 

 on their colour, and is precisely the same by whatever method the spec- 

 trum is obtained. 



The following table shows the relative positions occupied by the 

 Fraunhofer lines B D E F G in dispersion-spectra, produced by prisms of 

 60° of crown glass, of flint glass, and of carbon disulphide, with which are 

 compared the positions of the same lines in a spectrum produced by 

 diffraction. The interval between B and G is in each case divided into 

 1,000 equal parts. 



It will be noticed that the blue end of the spectrum is more compressed 

 in the diffraction-spectrum than in any of the dispersion-spectra, and the 

 red end is correspondingly lengthened out. 



In order that the results obtained by different observers may be com- 

 parable, either the spectra must be obtained directly by the method of 

 diffraction, or the results obtained with the prism must be reduced to ivave- 

 lengths. 



The admirable deterniinations of the wave-lengths of the chief solar 

 lines which we owe to Angstrom, will of course form the basis of the 

 reduction to wave-lengths, or when more convenient the measurements 

 based upon them of the bright lines of metallic spectra made by Thalen. 

 In the choice of reference-lines regard will of course be had to the accuracy 



' Spectres Lumincxix, p. 4. 



