78 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 65 



Thus the degree of approximation to integers seems to be about the 

 same as in the case of the atomic weights of the elements. Further, 

 in arriving at the number 9 (8.78) for the curious paramagnetic salt 

 K 2 HgI 4 , Weiss makes corrections for the diamagnetism of the three 

 constituent elements, a thing which is apparently not done in other 

 cases. 



It seems, therefore, that whatever may be the significance of 

 the integral values for the metals Fe, Co, Ni (even here the value 

 for Co is poor), the larger numbers obtained for the various hydrated 

 and complex salts shown above cannot have any simple theoretical 

 meaning — certainly none in so far as they may profess to represent 

 definite numbers of natural unit magnets. Almost any mechanistic 

 interpretation of Weiss' magneton involves the fallacy that elemen- 

 tary magnetic units can be additive in their effect on the magnetism 

 . of atoms and molecules in the same way as elementary electric units 

 can be. This is no more true than that the moments of bar magnets, 

 in an assemblage of such, are in general additive. Recently H. S. 

 Allen (Phil. Mag., May, 191 5) has discussed, in connection with 

 Weiss' magneton, a magnetic atom model in which he surmounts 

 this difficulty by ascribing the different magneton numbers to the 

 presence of different numbers of electrons in a rotating ring, and to 

 different angular velocities of this ring and of the central positive 

 charge, which is also supposed to rotate. But the insuperable objec- 

 tions to hypotheses 6f rotating rings of electrons have already been 

 explained (§2) ; and besides, the arbitrary nature of the assump- 

 tions which this model requires compares very unfavorably with the 

 simplicity of Langevin's scheme or with the " automatic " way in 

 which the model atoms of the present theory show a qualitative 

 agreement with the most diverse facts of magnetism. 



The futility of trying to express the magnetic properties of most 

 atoms as simple functions of their magneton constitutions has been 

 amply demonstrated in §§19-22. Apart from the paramagnetism 

 expected in the isolated H atom, the only case in which, in the present 

 state of the theory, we can make an absolute prediction of even the 

 sign of the magnetism, is when the atom or molecule contains no free 

 magnetons and only groups of eight. The atoms of He, Ne, A, 

 Kr, Xe fulfill this condition, and for two of them we know the 

 values (T) : 



He (y) -38.8, 



A (37) -212.8 ( = 3X7o.9) = (3X38.8)+96. 4 ). 



Unlike paramagnetic moments, diamagnetic moments must always 



