216 Prof. J. A. Ewing on the Molecular 



it is clear that an increase of permeability is to be expected 

 from the theory ; expansion with rise of temperature involves 

 a separation of the molecular centres, and therefore a reduc- 

 tion of stability. As regards the almost sudden loss of 

 susceptibility which occurs at a high temperature, it may do 

 no harm to hazard a rather wild conjecture. We may suppose 

 the molecular magnets to be swinging more or less, the 

 violence of the swings increasing as the temperature rises, 

 until finally it develops into rotation. Should this happen, 

 all trace of polarity would of course disappear. The con- 

 jecture that the molecular magnets oscillate more and more 

 as the temperature rises, is at least supported by the fact 

 (carefully investigated by Hopkinson * in iron and nickel ; 

 data for cobalt also have lately been supplied by du Bois f ) 

 that under strong magnetic forces rise of temperature reduces 

 magnetism ; for with strong forces the molecular magnets 

 are already ranged so that their mean direction is nearly 

 parallel to *£j: hence the earlier effect of heat (to diminish 

 stability and facilitate alignment) does not tell, and the in- 

 creased swinging simply results in reducing the mean value 

 for each molecule of its moment resolved parallel to the 

 magnetizing force. 



Before referring to effects of stress we may consider shortly 

 the stability of a pair or line of magnets, treating each as a 

 pair' of poles subject to the law of inverse squares. Take first 

 a single pair of equal magnets with centres at C and C x (fig. 7). 

 The poles P P' would lie in the line CC, but for the imposed 

 force <£), which produces a deflexion CO'P' or C / OP = #. 



Let a be the angle which &> makes with the line of centres, 

 m the pole-strength, and r the half length of the magnetic 

 axis of each magnet. The deflecting moment is 



2$mr sin (a— #), 

 and the restoring moment is 



JTp/2 J 



ON being drawn normal to PP'. The restoring moment at 

 first increases with 0, but passes a maximum at a value of 6 

 which depends on the relation of r to the distance between the 

 centres. The condition of equilibrium is 



2£™rsin(a-6 , )=-=- ; 



* Phil. Trans. 1889, A, p. 443 j Koy. Soc. Proc, June 1888. 

 t Phil. Mag. April 1890. 



