1890.] Molecular Theory of Induced Magnetism. 347 



that there is stability for small displacements, but different positions 

 of the group may be stable in different degrees, and if members of 

 the group be turned through a sufficiently great angle, they become 

 unstable, and fall into a new position of stability, bringing about 

 a partial reconstruction of the lines that characterise the group. 

 Special interest attaches to squai'e patterns, from the fact that iron 

 and nickel (probably cobalt also) crystallise in the cubip system. 

 In a square pattern of many members, we find, in general, lines 

 running parallel with all sides of the square when the group settles 

 without directive force after a disturbance. 



Let the group, or collection of groups, be subjected to an external 

 magnetic force, , gradually increasing from zero. The first effect 

 is to produce a stable deflection of all members except those which 

 lie exactly along or opposite to the direction of . This results in 

 giving a small resultant moment to the group (assuming that there 

 was none to begiu with), which increases at a uniform or very nearly 

 uniform rate as increases. This corresponds to the first stage in 

 the magnetisation of iron or other magnetic metal (a, fig. 3). The 

 initial susceptibility is a small finite quantity, and is sensibly uniform 

 for very small values of <>. 



Suppose that, without going beyond this stage, we remove ; the 

 molecular magnets, not having been deflected beyond the limit of 

 stability, simply return to their initial places, and there is no residual 

 magnetism. This, again, agrees with the fact that no residual 

 magnetism is produced by very feeble magnetising forces. Up to 

 this point, there has been no magnetic hysteresis. But let the value 

 of < be increased until any part of the group becomes unstable, and 

 hysteresis immediately comes into play. At the same time, there 

 begins to be a marked augmentation of susceptibility that is to say, 

 a marked increase in the rate at which aggregate resultant moment is 

 acquired. It is not difficult to arrange groups in which the state of 

 instability occurs with one and the same value of > throughout the 

 group. But, in general, we shall have different elementary magnets, 

 or different lines of them, reaching instability with different values 

 of 4?- The range of , however, which suffices to bring about 

 instability throughout the whole, or nearly the whole, of the members 

 in most groups is not large ; we, therefore, find in the action of the 

 model a close analogy to the second stage (6, fig. 3) of the process of 

 magnetisation, in which the magnetism rises more or less suddenly, 

 as well as to the first stage (a). 



During the second stage (6), the magnetic elements fall for the 

 most part into lines which agree more or less exactly with the 

 direction of <. If, at the end of this stage, we remove , we find 

 that a very large proportion of the aggregate moment which the 

 group has acquired remains ; in other words, there is a great deal of 



2 B 2 



