HALL. — THERMO-ELECTRIC HETEROGENEITY IN ALLOYS. 547 



8. By keeping the intermediate parts of the german-silver wires not 

 far below the temperature of the cooler junction in the calibration tests, 

 the change of sensitiveness observed on reversing the temperature inter- 

 val of the junctions, about 40°, was reduced from an average of about 

 3 per cent to an average of about 0.8 per cent. 



9. As a result of all the calibration work done on thermo-electric 

 couples of the kind described above it appears that, if we take 2100 X 

 10" 8 volt as the mean "thermo-electric height," electromotive force 

 with 1° difference of temperature of junctions, at 100° C., and if we 

 take 4.8 X 10 -8 as the rate of increase per degree rise of temperature, 

 we shall get values which will not at any point between 0° and 218° 

 differ as much as 1 per cent from the most probable value which we 

 have for that point. 



10. A piece of manganin wire subjected to the ice test described in 

 (7) behaved very much as the german-silver wire did, in magnitude of 

 local electromotive forces developed, in effect of annealing, and in non- 

 effect of stretching. 



11. A piece of constantin wire subjected to the ice test showed much 

 smaller local electromotive forces than did german-silver and manganin, 

 perhaps one fifth as great, and annealing, although it affected the dis- 

 tribution of the local peculiarities, did not sensibly increase their 

 magnitude. 



Annealing this constantin wire, which had been hardened by drawing 

 down from 0.051 cm. to 0.036 cm. diameter, changed its " thermo-elec- 

 tric height " with respect to copper about 1.5 per cent, the annealed wire 

 being nearer to copper thermo-electrically than the hard-drawn wire. 



12. As the thermo-electric height of manganin with respect to copper 

 is two or three times as great as that of german-silver with respect to 

 copper, at ordinary temperatures, as, further, annealing seems to make 

 less proportional difference in constantin than in copper, and as, finally, 

 the thermo-electric heterogeneity of constantin appears to be relatively 

 less than that of german-silver, it would seem that couples of constantin 

 with copper are to be preferred to couples of german-silver with copper 

 for ordinary use. 



It must be remembered, however, that the experiments here described 

 relate to one kind only of german-silver and to one specimen only of 

 constantin. 



