MAGNETIC PERMEABILITY OF NICKEL AND COBALT 71 



been so much neglected. It may have been that a simple method of 

 experiment was not known; but if so, I believe that my method will be 

 found both accurate and simple, though it may be modified to suit the 

 circumstances. Professor Maxwell has suggested to me that it would 

 be better to use rods of great length than rings, because that in a ring 

 we can never determine its actual magnetization, but must always con- 

 tent ourselves with measuring the change on reversing or breaking the 

 current. This is an important remark, because it has been found by 

 MM. Marianini and Jamin, and was noticed independently by myself 

 in some unpublished experiments of 1870, that a bar of steel which has 

 lain for some time magnetized in one direction will afterwards be more 

 easily magnetized in that direction than in the other. This fact could 

 not have been discovered from a ring; and indeed if a ring got a one- 

 sided magnetism in any way we might never know it, and yet it might 

 affect our results, as indeed we have already seen in the case of the 

 magnetic curve. But at the same time I think that greater errors 

 would result from using long bars. I have tried one of iron 3 feet 

 long and inch diameter; and the effect of the length was still appar- 

 ent, although the ratio of length to diameter was 144. To get exact 

 results it would probably have to be several times this for the given 

 specimen of iron, and would of course have to be greater for a piece 

 of iron having greater permeability. This rod must be turned and 

 must be homogeneous throughout conditions which it would be very 

 difficult to fulfil, and which would be impossible in the case of nickel 

 and cobalt. We might indeed use ellipsoids of very elongated form; 

 and this would probably be the best of all, as the mathematical theory 

 of this case is complete, and it is one of the few where the magnetization 

 is uniform, and which consequently will still hold, although the permea- 

 bility may vary with the amount of magnetization. This form will, of 

 course, satisfy Professor Maxwell's objection. 



The method of the ring introduces a small error which has never 

 yet been considered, and which will affect Dr. Stoletow's results as well 

 as mine. The number of lines of induction passing across the circular 

 section of a ring-magnet we have seen to be 



/+ J ~Jp y* 



Jn a, x 



in which a is the mean radius of the ring, E the radius of the section, 

 n' the number of coils in the helix, and i the intensity of the current. 

 Xow in integrating this before, I assumed that ft was a constant 

 throughout the section of the ring: now we have found that 11 is a 



