636 



Lord Kelvin 



[May 21, 



rature and connecting them by wires of any kind witli brass or otber 

 terminals of a galvanometer of higb enough resistance not to sensibly 

 diminish the difference of potentials between A and B, provided all 

 the connections between metals of different quality except J and K 

 are kept at one and the same temperature (or pairrf of them, properly 

 chosen, kept at equal temperatures). 



§ 20. Suppose, now, instead of breaking a circuH of two metals at 

 a place in one of the metals, as A B in copper in Fig. 9, we break it 

 at one of the junctions between the two metals, as at C 0, 1' I, Fig. 10. 

 C D represents a movable slab of copper which (for § 22 below) may 

 be pushed in so as to be wholly opposite to I' I, or at pleasure drawn 

 out to any position, still resting on the copper below it as shown in 

 the diagram. Calling zero the uniform potential over the surfaces 

 C C D, the potential at I' I would be about + • 16 volt (according to 



tRON 



Fig. 10. 



Murray's results for emery-polished copper and iron surfaces) If the 

 temperature at J and throughout the system is uniform at about 15° G. 

 Keeping now the temperature of C C, 1' I exactly at 15°, let the tem- 

 perature of J be raised to 25°. The difference of potentials between 

 C'C a-jd I'l would be increased to •16148 volt, supposing '16000 to 

 have been exactly the difference of potentials when the temperature 

 of J was 15°. This difference of differences of potentials would be 

 just perceptible on the most delicate qua;lrant electrometer connected 

 as indicated in the diagram. Lastly, raise the temperature of C G 

 and I' I to exactly 25°, J being still kept at this temperature : the 

 spot of light of the electrometer will return exactly to its metallic 

 zero. But, would the Volta-difference of potentials between the 

 surfaces G' G, I' I remain unchanged, or would it return exactly 

 to its previous value of '16000, or would it come to some other 

 value ? We cannot answer this question without experiment. The 



