1904] Thermoelectric Power produced by Magnetisation. 415 



tion in any field up to 1600 (the strongest reached), was found to be 

 about equal numerically to the retraction in ten-millionths multiplied by 

 145 x 10 -5 . The curves for two pieces of an impure nickel also exhibit 

 general similarity of form, but since the dimensional ratios (ratio of 

 length to cross section) of the two pieces are very different, the curves 

 are not strictly comparable. The changes of thermoelectric force are, 

 like the changes of length, much greater for nickel than for iron. 

 Tensile stress produces, as in iron, corresponding variations in the two 

 classes of curves. The effect of tension upon the magnetic retraction 

 of nickel is, as I have shown in a former paper,* not so simple as in the 

 case of iron. In weak fields the contraction is diminished by tension ; 

 in fields of more than about 150 units the contraction is increased by 

 tensile stress up to a certain critical value of the stress depending 

 upon the strength of the field, and diminished by greater tension. 

 Thus it happens that the retraction curves for a wire loaded with two 

 different weights may cross each other. In one of my published 

 experiments the retraction curves for a nickel wire carrying loads of 

 420 and of 980 kilogrammes per sq. cm. crossed when H reached 220. 

 With nearly the same loads, the two curves of thermoelectric force also 

 crossed, though in a weaker field, H being only 150. Wire of the 

 same quality was used in both experiments, but in the first it was 

 hard, while in the second it was annealed, which may or may not 

 be the reason of the difference. In any case, it is interesting to find 

 this complex and unexpected phenomenon qualitatively reproduced by 

 the thermoelectric curves. 



The anomaly above referred to consists in the fact that the direction 

 of the thermoelectric force due to magnetisation is the same for nickel 

 as for iron, whereas length is affected oppositely in the two metals, 

 iron being extended, nickel contracted. 



Cobalt.— For cobalt I have not succeeded in finding any relation 

 between the thermoelectric and dimensional changes attending 

 magnetisation. The thermoelectric curve somewhat resembles that 

 for nickel, but it is much lower. The strain curve is opposite in 

 character to that of iron ; in weak fields the metal contracts, in strong 

 fields it is elongated. 



Nomenclature. 



When the direction of the thermoelectric current between two 

 metals, A and B, was from A to B through the hot junction, it was 

 formerly the custom to say that A was positive to B, and B negative 

 to A ; bismuth, for example, was said to be thermoelectrically positive 

 to antimony. Of late years this custom has to a large extent been 

 reversed, and the terminology is at present in an unsettled condition. 

 Following what I believe to be the best modern authorities, I shall in 



* ' Eoj. Soc. Proc.,' vol. 47, p. 474, 1890. 



2 G 2 



