168 Dr. E. H. Hall on the " Rotational Coefficient ' 5 



we can yet be sure that the bend is not due to faulty 

 observations ; for to make an error of 1 per cent, in the value 

 of 3 at this point would require an error of very many per 

 cent., say 20 or 30, in the value of a as determined by Prof. 

 Rowland. From the manner and rate at which \l was 

 changing at the points where his experiments ceased, it seems 

 almost certain that these lines would continue to bend, and for 

 a time to bend rapidly. Indeed the curve in which Prof. 

 Rowland has continued jju beyond the range of his experiments 

 would indicate that the curves (2) and (3), if continued a short 

 distance further, would turn downward and approach the base- 

 line. This, however, would mean that the magnetization 3 

 actually decreases after a certain point with increase of the 

 magnetizing-force. The possibility of this is spoken of by 

 Rowland*; but there seems to be no experimental evidence of 

 such an effect ; and if it does not exist, it appears altogether 

 probable that the lines (2) and (3) would become asymptotic to 

 horizontal lines lying considerably higher than any points 

 reached by the curves as here given. 



We can therefore say that, so far as actual experiments have 

 gone, there seems to be much tendiug to prove a very simple 

 and intimate relation in nickel between the transverse effect 

 and the "magnetization according to the German theory." 



It would, of course, be desirable to test for some more 

 minute agreement than has yet been traced between the curves 



for -=jr and 3 ; but such a testing would probably be difficult to 



make. An exact agreement could not be expected; for it 

 would probably be almost impossible to obtain exactly the 

 same quality and condition of metal in the very different 



E' 



shapes required for experiments on 3 and experiments on ^. 



There are, however, certain minute characteristics which 

 would belong to all curves for 3- Thus (2) and (3) should not 

 be straight at any point. They are lines of double curvature, the 

 steepest part of each being not far from 33 = 2000. The cur- 

 vature in this region, however, is very slight; and to detect a 



E' 

 corresponding curvature in the line for =^, if such exists, 



would be a matter of considerable difficulty, though not, 

 perhaps, impossible. 



Having gone thus far with nickel, we might, were it not for 

 the anomaly presented by the sign of the rotational coefficient 



* Phil. Mag. Nov. 1874, p. 322. 



