of Electro-Magnetism by Heat. 63 
In determining the effects of the common galvanic apparatus, 
an almost insuperable difficulty occurs from the loss of power, 
arising from the gradual saturation of the acid, and oxidation of 
the plates. In the electro-magnetism excited by heat, provided 
that the extremities of the bars are kept at an uniform temperature, 
(which there is no difficulty in doing) the power remains unaltered. 
There is therefore, by this process, no impediment in determining 
the relative conducting powers of different metals, or the effects 
of different dimensions of the same metal. 
Some experiments, I had previously made, led me to imagine 
that the conducting powers of metals were materially affected, not 
only by the diameters of the wires, but by their lengths; the law 
of which it would be desirable to ascertain. For this purpose, the 
circuit from a bar of antimony kept at a steady temperature, was 
made through a connecting copper wire 32 feet in length, the devi- 
ation was 7°; 16 feet of the same wire gave 10°; 8 and 4 feet gave 
154° and 20°. Allowing for the unaveidable inaccuracies of an ex- 
periment, in which the divisions of the compass were estimated by 
the eye, it seems that the conducting power of the wire diminishes 
in a much lower ratio than its length increases. A slight change 
in the deviations making them 6°, 11°, 16°, 21°, would form an 
arithmetical series corresponding to the diminutions of length tn 
geometrical progression. 
The diameter of the wire used in this experiment was + inch; 
when 8 feet of wire of t was used, the deviation, which had been 
151°, was reduced to 64°, and with platina wire of ;4 it became 
certainly not more than 3°. When the minute quantity of electri- 
city developed in this experiment is considered, it might have been 
expected that the platina, and much more the copper wire of 4 inch, 
would have conveyed it without loss; yet I found that, even the 
larger wire of +, was not sufficient, but that the deviation was still 
augmented by employing wire of 4. This seemed the limit, for 
