2S2 Prof. Barkla and Miss White on the 



wave-length of a characteristic radiation on the shorter 

 wave-length side of the absorbing substance, f(X) becomes 

 X 3 simply. This is certainly very near to the truth when X 

 is much less than \k the wave-length of the K radiation of 

 the absorbing substance. Aluminium is probably the best 

 substance with which to test this relation, as the deter- 

 minations of the absorption coefficient are probably more 

 accurate than in any other substance. We have, therefore, 

 plotted values of log (fi/p — "2) in aluminium against logX 

 over a long range of wave-lengths (*39 to 2'28 X 10~ 8 cm.) 

 greater than Xj for aluminium, using Barkla and Sadler's 

 and our values for p,/p and Siegbahn/s for X. The relation 

 is not exactly a linear one and shows that if we write 

 (fi/p — *2) = 0A, W , n is not a constant, but varies even over 

 this range from about 2*7 to 3. (No importance can be 

 attached to the exact value *2 as it is usually very small 

 in comparison with p,/p, and is only introduced here as 

 indicating a physical fact of which account must be taken 

 over other ranges.) 



We have shown, too, that the absorptions in other sub- 

 stances are proportional to those in aluminium for radiations 

 of wave-length considerably shorter than those of the nearest 

 spectral line. Thus, if it were true for aluminium it would 

 be true approximately for other substances. 



It should be remembered, however, that such a simple 

 relation as is obtained by giving n a constant value has not 

 been found to hold accurately for wave-lengths anywhere 

 really near to that of a K spectral line on its shorter wave- 

 length side. In fact, if we can regard the absorption 

 experiments as having given results approximating to those 

 which would have been obtained with perfectly homogeneous 

 radiations, the relation cannot hold even to a first approxi- 

 mation when the wave-length is only slightly shorter than 

 that of a K spectral line say. But further experiments are 

 needed. 



Hull and Bice, from their experiments on the absorptions 

 in aluminium, copper, and lead, showed that fi/p — '12 = k\ z 

 with a fair degree of accuracy and concluded that the 

 scattering coefficient (<r/p) = '12. 



It is probable, however, that a slight variation of (o-jp) 

 with X makes the scattering appear somewhat less than it 

 actually is — at any rate with the longest waves used. It is 

 highly improbable that the " fluorescence absorption " in 

 this region is so strictly proportional to the cube of the 

 wave-length as to enable us to use this as a method of 

 determining a/p with any accuracy. 



