1904.] Thermoelectric Power produced by Magnetisation. 421 



So far the presumption in favour of the reality of the compressive 

 stress appears to be fairly strong, but it is, I think, greatly strengthened 

 by the results obtained for nickel. In fig. 7 (L) and (k) are curves 

 of change of thermoelectric power and of change of length for this 

 metal. There is obviously a strong likeness between the two, though 

 they lie on opposite sides of the axis. Curve is derived from (k) 

 simply by inverting the latter and plotting the ordinates to a slightly 

 different scale. The inverted curve is seen to be throughout its whole 

 length almost a counterpart of the thermoelectric curve ; and this 

 although it has not been "corrected" for compressive stress. Why 

 then should the correction which is indispensable for iron be un- 

 necessary for nickel 1 The answer is that while for iron the calculated 

 correction for mechanical stress is relatively very considerable, for 

 | ickel it turns out to be very small, so small as to be negligible. The 

 changes of length indicated by the compression curve for iron, (d), 

 fig. 4, are generally much greater than those indicated by the curve (b), 

 to which the correction is to be applied ; and the two curves have very 

 different forms. On the other hand the compression curve for 

 nickel (in), fig. 7, is an almost straight line making a very small angle 

 with the axis of H, the changes of length which it indicates amounting 

 to not more than 2 or 3 per cent, of the changes exhibited by the 

 uncorrected curve for corresponding fields. Both curves, moreover,, 

 rise gradually from the origin to their highest points. Thus it happens 

 that the uncorrected and the corrected curves for nickel, if referred 

 respectively to scales of ordinates so chosen that the two curves may 

 be of the same height, are sensibly identical. Curve (n) may be 

 regarded either as the uncorrected or as the corrected curve, according 

 as the ordinates are referred to the scale given in the right-hand 

 margin or to a different scale in which each unit is very slightly 

 increased. 



Thus the absence of any need for the correction in the case of nickel 

 where a priori it ought not to be required, tends to show that the 

 success of its application in the case of iron is not a mere accident, and 

 that the compressive stress is consequently a vera causa. 



Apparatus and Methods of Experiment. 



Two methods of experiment were employed. In the one the 

 thermoelectric force of the magnetisable metal in conjunction with 

 copper was opposed by a similar couple in which the first metal was 

 unmagnetised ; in the other it was opposed by an electromotive force 

 derived from a battery. 



The arrangement adopted for Method I is shown diagrammatically in 



former communications to the Eojal Society. It is unfortunate in the present 

 connection 1 hat some of the experiments were not carried further. 



