Properties, Specific Resistance, and Hardness of Steel. 363 



we pass from the exterior to the interior, the circumstances 

 become more complicated. 



Furthermore, suppose the ends of a thick glass-hard cylinder 

 to be kept at temperatures T and t (T>t). In this case, 

 since each of the infinitesimally thin cylindrical shells has a 

 particular T. E. H. corresponding to its hardness, we are led 

 to infer that thermo-currents closing themselves in the inte- 

 rior of the cylinder are the result — the direction of these in 

 the outer (harder) parts being from t to T, in the inner from T 

 to t. In fig. 3 (vertical section) the hypothetical condition of 

 the cylinder in this case is indicated. As will be seen, its elec- 

 trical state corresponds with that of a rod circularly magnetized. 



For the purpose of studying this question experimentally, 

 a steel cylinder, 30millims. in diameter and 50 millims. long, 

 was turned and glasshardened. This was placed vertically 

 directly before the needle of a magnetometer (the deflection 

 of which could be read off with telescope and scale) in such a 

 manner that the position of equilibrium of the needle was left 

 unaltered. The relative position of needle and cylinder, in 

 other words, was such that the axis of the former, if prolonged, 

 would intersect the axis of the latter at its middle point. 

 Upon now cooling the upper end of the cylinder with a piece 

 of ice, or warming by projecting a jet of steam against it, very 

 decided deflections were observed toward the right or left re- 

 spectively, which increased with the difference of temperature 

 T— -t, and vanished as this difference became nil. 



As the cylinder was not magnetic, it is improbable that these 

 phenomena can be referred to a change in the state of mag- 

 netic distribution. With reference to the direction of the 

 currents, however, no simple results could be arrived at. 



e. In § VII. c, we ascribed the very high value of the 

 T. E. H. of a glass-hard steel rod to the large proportion of 

 chemically combined carbon contained therein. If this be 

 true, the thermoelectric difference between soft steel and soft 

 iron, in both of which combined carbon is either wholly absent 

 or exists only in traces, must be quite small. This inference 

 is supported by the data actually found for soft iron. Making 

 allowance for the difference of circumstances involved, the 

 result to be derived from experiments of Kohlrausch and 

 Ammann also agrees sufficiently herewith. 



On the other hand, Joule * has long since shown that ordi- 

 nary cast iron is thermoelectrically negative towards copper, 

 all the more, therefore, towards soft steel — a result which we 

 should be inclined to predict from the quantity of combined 

 carbon contained. 



* Joule, Phil. Mag. [4], vol. xv. pp. 538, 030, J857. 



