420 



Dr. S. Bid well. On the Changes of 



[Apr. 11, 



writers insist upon the necessity of correcting curves of magnetic strain 

 for mechanical stress. Brackett attributes the recognition of this 

 necessity to Bowland, who, he says, defended his position as follows : — 

 The compressive force which will tend to close up a very thin air gap 

 in a divided magnet must exist in any magnet, for according to all our 

 ideas of matter there is no real difference in the case where the air gap 

 exists and where it does not, because we must still consider the gaps 

 between the molecules. 



I venture to think that the results of my thermoelectric experiments 

 with iron and nickel afford strong evidence of the reality of the 

 mechanical stress in question. Before any comparison is possible between 

 the two phenomena of change of length and change of thermoelectric 

 power, considered as due to the molecular effects of magnetisation, it is 

 clear that the effect of any extraneous mechanical action tending to 

 alter the length of the metal must be eliminated. The stress under 

 discussion, if it exists, is for our present purpose no less an extraneous 

 one than if it were produced by loading the metal with a weight. 

 Conjecturing that there might be a relation between the thermoelectric 

 and the strain effects of magnetisation, I plotted curves for the two 

 phenomena side by side (see curves (D) and (b), fig. 4) ; but although 

 these curves bear a superficial resemblance to each other, it is clear that 

 the phenomena to which they relate are by no means in correspondence. 

 In particular, one curve cuts the axis at an early stage, indicating a 

 reversal of sign, while the other appears to become asymptotic. Curve 

 (d) which is drawn to the same scale as curve (b) indicates, in ten- 

 millionths of length, the correction to be made in respect of the 

 hypothetical compression due to mechanical stress ; its ordinates are 

 proportional to P/M, P being the greatest weight which (as found 

 experimentally) could be lifted when the iron was magnetised by the 

 corresponding field, M being Young's modulus. Adding together curves 

 (b) and (cP), we get the "corrected" curve (E), which, as will be seen, 

 coincides with the thermoelectric curve (D) as nearly as could be 

 expected under the conditions of the experiments. 



Still more striking is the case of the wire under tension, curves (F) 

 and (/) fig. 5. Here there is not even a superficial resemblance; the 

 curve for change of length is an almost straight line inclined to the 

 horizontal axis and lying entirely below it, while the thermoelectric 

 curve begins above the axis and crosses it at H = 450. By making 

 the same addition for stress as in the former case, curve (cl) being again 

 employed for the purpose, we obtain a corrected curve of length, 

 indicated by the crosses on curve (F), which, when the ordinates are 

 plotted to a suitable scale, closely follows the thermoelectric curve (F), 

 and would, if it were continued a little farther,* evidently meet the 

 axis at just the same point. 



* All the curves for change of length referred to in this paper hare appeared in 



