permitted by (6) only those are occupied by an appreciable number 



of the systems which nearly coincide with the levels of p given 



by (7). 



B. The transitions which take place have almost without excep- 



h 

 tion the magnitude — (and not a multiple of it) (See (13)). This is 



again in good agreement with the fact that for /= qd the Fourier 

 expansion of the x (or y) component contains only the fundamental 

 and no higher harmonics so that for this case the Principle of 

 Correspondence allows only the transitions (see (7)) for which 

 m, — m, =: ± 1. 



7. A question must now be mentioned the precise explanation of 

 which would be of value. For the discussion of thermal equilibrium 

 in our complex we must know the ''weights" (the a priori proba- 

 bility) to be ascribed to each p level. For ƒ =t oo it would appear 

 that the same weight should be given to every stop of (6) — in- 

 dependently of the value of ƒ and independently of the density 

 with which the levels follow each other. On the other hand for 

 f=zco only the levels given by (7) are to have a weight (the same 

 for all). A closer examination of this case will probably make it 

 necessary to consider the fact that we are concerned here with a 

 double limit viz. lim t:^ oo (the lapse of an intinitel}^ long time for 

 the establishment of thermal equilibrium) and /m /=od; our dis- 

 satisfaction is really based on an unconscious demand that the result 

 should be independent of the order in which the two limits are 

 approached. 



The junior author of the paper (G. Breit) is Fellow of the National 

 Research council, United States of America. 



The University, Leiden. 



