486 Mr Filon, On the variation with the wave-length 



In our case a is always less than this value, being in fact of 

 the order 7r/180 and therefore small. But ^(o^k a/ 2 and tends 

 to a/ 2 if a be small. So that the intensity never actually vanishes, 

 but it falls to a very small minimum whenever 



-^- (pyo - qy') =^nir 



, \n 



or pyo-qy=^j- 



This therefore corresponds to a black band in the field of view. 

 If now y {n) , y {n+1) be the y 's of two consecutive black bands 



Cp(y ^-y ™)=\. 



This equation is of precisely the same form as in the ideal case 

 previously discussed, except that the coefficient p is slightly 

 different. 



Hence, in so far as relative values of G for different Vs are 

 concerned, y (n+1) — y {n) being the quantity observed, no error what- 

 ever will be introduced by the obliquity, or by the finite solid 

 angle of the pencils. 



As a matter of fact p is known, as all the quantities which 

 enter into its calculation can be easily measured : so the absolute 

 values can be deduced when required. 



5. It was now necessary to find the exact wave-length of the 

 light corresponding to any vertical line on the scale. 



To do this the arc lamp, which was used to give a continuous 

 spectrum, was removed and the spark-holder with the poles made 

 of the alloy mentioned above was placed behind the slit. 



The spark spectrum of the alloy then appeared on the scale. 



The positions of 25 of the most prominent lines of this spec- 

 trum, ranging from A.4000 to X6800, were then carefully read off 

 on the horizontal millimetre scale. The lines were identified and 

 their accurate wave-lengths obtained. 



The wave-lengths were then plotted on squared paper to the 

 scale divisions. A smooth curve was drawn as nearly as possible 

 through the 25 or so points obtained and from this the wave- 

 length corresponding to any given scale division could be read off 

 practically correct to the first three figures. 



Before each experiment the "zero" was determined by reading 

 off the position on the scale of a well-defined, easily recognizable, 

 line of the comparison spectrum. The wave-lengths for each 

 observation could then be deduced by means of the diagram. 



The spectrum was carefully re-mapped and a new diagram 

 plotted, at certain intervals, especially whenever the lenses, prism 

 or beams had been moved or shaken in any way. 



