1904.] Thermoelectric Power produced by Magnetisation. 417 



to magnetisation had apparently become zero. His paper contains no 

 record of an actual reversal of the force in stronger fields, though the 

 probability of such reversal is suggested. My experiments, which 

 were made with several different specimens of iron, show no approach 

 to zero in fields nearly five times as strong as that mentioned. As 

 regards the behaviour of steel, Houllevigue's results are at variance 

 both with those of W. Thomson and with my own. 



Mr. E. Khoads* has given an account of two thermoelectric 

 experiments, one of which was made with iron, the other with nickel, 

 his work having been undertaken "in looking for some property of 

 iron that would vary Math the magnetisation in the same way as the 

 length." After referring to the experiments of Chassagny, and to 

 those of Houllevigue " who found that the diminution [of thermo- 

 electric force] continues in higher fields, and that a reversal actually 

 takes place," he remarks that "this suggests change of length, and 

 made me wish to work out the cyclic curve for comparison with it." 

 The cyclic curve which he obtained exhibits hysteresis, and bears a 

 general resemblance to one given for change of length, though the 

 latter is not carried beyond H = 90. The thermoelectric curve for 

 iron is represented as crossing the axis of H at H = 400, continuing to 

 fall in an almost straight line up to H = 500, the strongest field 

 applied. Rhoads considers that it agrees with the curve for change of 

 length when the latter is corrected for " Maxwell's stress " (which is 

 given as B 2 /47r) being apparently unaware of the fact demonstrated by 

 More, and afterwards by •Klingenberg, that the corrected curve never 

 crosses the horizontal axis. The thermoelectric curve for nickel is 

 shown in Rhoads's diagram as lying, like the curve for change of length, 

 below the horizontal axis (in agreement with Thomson's experiment), 

 while for iron, both the thermoelectric and the corrected length curves 

 are above the axis. " The two properties," he says, "turn out to be 

 related in the opposite sense in the two metals." As regards form, the 

 nickel curves for change of thermoelectric power and change of length 

 are of the same character, even though, as is remarked, they are not 

 strictly comparable, being made with very different specimens. No 

 mention is made of Maxwell's stress in relation to nickel. 



Mechanical Stress clue to Magnetisation. 



Before describing the experiments it is desirable to discuss briefly 

 the controverted question of compressive stress which has such an 

 important bearing upon the results. In a paperf on the changes of 



* 'Physical Review,' vol. 15, p. 321. This paper was published about 2 months 

 later than my article above referred to. My experiments were begun before I had 

 beard of Mr. Rhoads's work. 



f 'Phil. Trans.,' vol. 179, p. 216, 1888. 



