SPECTRUM ANALYSIS. 737 



These facts are of course easily explained by the inter- 

 ference of waves. It would be impossible to have a mope 

 convenient or suitable scale for measuring the spectra of 

 coloured solids and liquids. If we use a micrometer in 

 the eyepiece, an alteration in the width of the slit modifies 

 the readings, and the least movement of the apparatus 

 may lead to error, whereas this scale is not open to either 

 objection. Besides this, the unequal dispersion of the 

 spectrum makes the blue end too broad, so that a given 

 width, as measured with a micrometer in the eyepiece, 

 is not of the same optical value as the same width in 

 the red. The divisions in the interference- spectrum 

 bear, on the contrary, the same relation to the length 

 of the waves of light in all parts of the spectrum, and 

 no want of adjustment in the instrument alters their 

 position. As will be seen from the diagram (fig. 355), 

 the unequal dispersion makes the distance between the 

 bands in the blue about twice as great as in the 

 red. The perfection of a spectrum would be one in 

 which they were all at equal intervals ; but possibly 

 no such uniform dispersion could be produced. By 

 having a direct-vision prism, composed of one of flint- 

 glass of 60, and two of crown-glass of suitable angle, we 

 can place it over the eyepiece, and can diminish the 

 dispersion at the blue end, or increase that at the red end, 

 by turning it in one position or the other, and thus see 

 either end to the greatest advantage. 



" Since the number of divisions depends on the thick- 

 ness of the interference-plate, it became necessary to decide 

 what number should be adopted. Ten it was thought 

 would be most suitable ; but, on trying, it appeared to b& 

 too few for practical work. Twenty is too many, since it 

 then becomes extremely difficult to count them. Twelve 

 is as many as can be easily counted ; it is a number easily 

 remembered, gives sufficient accuracy, and has a variety of 

 other advantages. With twelve divisions the sodium-line 

 p comes very accurately at 3^ ; and thus, by adjusting the 

 plate so that a bright sodium-light is hid in the centre of 

 the band, when the Nicol's prisms are crossed, it is accu- 

 jately at 3|, when they are arranged parallel, so as to give 

 a wider field. The general character of the scale will 



SB 



