40 PROCEEDINGS OF THE AMERICAN ACADEMY 



when, at any desired part of the stroke, it was closed for a short 

 time by the revolving crank of the engine. With this apparatus 

 there should be a current, and therefore an effect ujDon the gal- 

 vanometer, at closure of the circuit, so long as the two inner ends of 

 the antimony and bismuth bars, which press against the steel, are 

 at a different temperature from the outer ends, which are placed 

 side by side in a pot of melted paraffin or hot oil. But when, on 

 the other hand, there is no effect on the galvanometer at closure, 

 one may, if proper precautions have been taken, conclude that the 

 temperature at that surface of the steel which is touched by the 

 antimony and bismuth, to determine which is the object of the 

 experiment, is the same as that of the heated liquid surrounding 

 the ends a and h, which a thermometer at once makes known. 



This represents one stage of the investigation, but as I gradually 

 became convinced that considerable fluctuations of temperature 

 really occurred at measurable depths of the cylinder wall, I saw 

 that a better considered thermopile than the one at first used was 

 needed. I must discard the steel, the thermal conductivity of 

 which might differ considerably from that of the cast-iron compos- 

 ing the cylinder wall, and must replace the antimony and bismuth 

 by something not very different from cast-iron in thermal conduc- 

 tivity, for it is evident that the temperature existing at any mo- 

 ment at the surface of contact of the iron and the metal outside it 

 may depend greatly upon the conductivity of this second metal. 

 Moreover, this surface of contact should be of considerable breadth, 

 and the contact very perfect; in short, everything should be done 

 to make the surface of contact resemble in all its conditions, as 

 nearly as might be consistently with the object of the experiment, 

 the surface of contact of two strata of iron within the actual cylin- 

 der wall. Accordingly, I made the plug and the cover, which now 

 was fastened by four small steel screws to the inner end of the 

 plug (see Figure 3), from the same bar of cast-iron, — not ordinary 

 cast-iron, which would be too brittle for the thin cover, but cast 

 gun-iron of density 7.18, which I found by experiment to have a 

 thermal conductivity not very different from that of ordinary 

 Southern cast-iron of density 7.06. In place of the antimony 

 and bismuth bars 1 now used a single cylinder of cast nickel of 

 density 8.08, the thermal conductivity of which I have found by 

 experiment to be verj' nearly the same as that of the cast gun-iron 

 at 115° or 120° Centigrade. The specific heat of nickel is about 

 the same as that of iron. 



