â44 



C. G. KXOTT AND S. KIMUIIA. 



The g-eiiei'al pliiii of (jx[)uriiiieiitiiig w.u oiiiiple eiKJUg-h. After 

 the tulje und coniaiiied wire h'id attained a ;iteidy coiidici(jii a.i regard« 

 temperature, a «eriei of reading.^ of the différent thermoelectrie and 

 thermouietrie enrrent« were taken a« rapidly a.; poy.-sible, with a .suf- 

 ficient nunilier of repetition« of the same to yield a good mean for 

 each individual circuit. Thi.s operation wa« carried through for a 

 series of ascending and descending value« of tension. The small 

 value of the current in the thermoelectric circuit as compared with 

 that in the thermometric required that only a small shunted portion 

 of the latter should be taken through the galvanometer. This neces- 

 sitated a somewhat complicated arrangement of resistances and com- 

 mutators, which hoAvever it is unnecessary to describe. 



In the earlier experiments the thermometric circuits were of 

 copper andiron, and the thermoelectric of iron and iron. J5y using 

 copper and iron, we expected to be able to get good measurements of 

 the true temperature values of the junctions ; and this because of the 

 existence of a neutral point at an easily attainable temperature. It 

 was found, however, that the uncertainties of reduction from the 

 parabolic temperature scale of experiment to the linear scale of 

 accepted use far (outweighed the advantag(3s of having an (_)b-ierved 

 neutral point as a guide. Accordingly a.fter many experiments had 

 been made the Copper Iron thermometric jiUiction was aljandoned in 

 favour of a Gierman- silver Ir()n junction. As is well known, the elec- 

 tromotive force of this pair of metals varies in ;in approximately linear 

 manner with temperature up to a dull red heat. The graphical com- 

 parison of the thermoelectric with the thermometric currents will not 

 in tliis case differ greatly in appearance from wliat would be the case if an 

 accurate aljsolute scale of temperature were used instead. Ultimately, 

 of course, the thermometric readings were reduced to the ordinary 

 temperature scale by calculation frcjm the results of direct ex[)eriment. 



