CONTEMPORARY ADVANCES IN PHYSICS 363 



variously close approximations to perfect freedom, the theories are 

 good enough to make it possible to bring about quantitative agreement 

 between theory and experiment, simply by choosing appropriate 

 values for the magnetic moments of these particles. The values so 

 determined nearly always lie between 10~^^ and 10"^'' C.G.S. units. 



Ferromagnetic substances are solids, and we need not be surprised 

 that the mutual influence of the atoms becomes so great as to make 

 the task of devising a theory much more difficult. Ewing, it is true, 

 did show that elementary magnets of a particular shape and crowded 

 close together would form systems displaying the peculiar features 

 (hysteresis, and a crooked magnetization-curve) of ferromagnetics. 

 Weiss did show that atomic magnets, subject to the agencies which 

 bring about thermal equilibrium and maintain it, and in addition to 

 a field proportional to the magnetization of the assemblage and 

 enormously great, would form systems displaying residual magnetism 

 below a certain temperature, and paramagnetic above. Dazzling as 

 these achievements are, the theories are not so good that they can be 

 brought into complete accord with the data, simply by choosing 

 appropriate values for the moments of the imagined elementary 

 magnets. 



Can we at least assign a value of the order familiar among para- 

 magnetics, 10~^^ for instance, to the magnetic moment of (say) the 

 iron atom — that is to say, the atoms of a piece of solid pure iron, 

 since iron is not in all conditions ferromagnetic — without definitely 

 contradicting any fact of experience? Probably we can. In fact, 

 the saturation-values of the magnetizations of iron, nickel, and cobalt 

 support this idea. If saturation signifies that all the atomic magnets 

 are parallel, then the magnetic moment of each must be the quotient 

 of /max. by the number of atoms in unit volume; at all events, the 

 magnetic moment of the atom cannot be less than the quotient, by 

 that number of atoms, of the highest value of / ever observed. Now 

 the highest values of / are observed at the lowest temperatures; 

 extrapolating from the data (shown in Figure 9) to zero absolute, 

 Weiss obtained values of the quotient which are indeed of the order 

 10-19 — eleven "magnetons" for iron and three for nickel, and probably 

 eight for cobalt. This concordance with the values of magnetic 

 moment to which we are accustomed among free atoms is evidently 

 important. However, as Ewing found, we cannot take the natural 

 next step of supposing that each atom is a long slender magnet having 

 its ends very close to the ends of the adjacent magnets; for then the 

 /-vs.-// curve of the assemblage would not agree with the initial 

 curves observed in practice. 

 24 



