Serviss — Internal Temperature Gradient of Metals. 453 



long and the copper into lengths of 7*6 cm . Pairs of these 

 were bntted and brazed with an alcohol flame. Silver solder 

 was used with borax as a flux. Care was taken to use as 

 small a quantity of silver as practicable, but if too small a 

 bead is used, it flows entirely to one or the other of the wires. 

 All junctions which suffered an accident in making, were bent, 

 or happened to have an undesirably large excess of silver, or 

 which did not successfully flow the first time, were rejected. 

 After the experience of making 100 good junctions, I was able 

 to join these couples end to end in a similar manner, winding 

 them on the mica ring as completed, so that all the junctions 

 of the thermopile were as uniform as they well could be. 

 The outer diameter of the mica ring was l^'S 0111 and the 

 diameter of the hole was 2-4 cm . Dr. Thwing placed the outer 

 row of junctions of his thermopile outside the cylinders and 



the inner row between them. I so placed the constantan that 

 my outer row of junctions would come - 6 cm inside the con- 

 vex surface of the cylinders, and the inner row was, of course, 

 3 cm nearer the center. This change was made in order to 

 expose all the junctions in the same manner. This is a funda- 

 mental difference between my work and his, and it will be 

 discussed at length in connection with the results. 



Experiments in this laboratory have shown " that moderately 

 sharp bending (with a radius of - 5 cm ) is not very injurious in 

 the ordinary thermo-electric use of german-silver wire "* 

 but, of course, the bend about the ring of mica is necessarily 

 much sharper than this. Under these circumstances, it 

 seemed best, since the softer and more homogeneous copper 

 would suffer less than the harder alloy, to leave the constan- 

 tan straight; the junctions were, therefore, all on the same 

 *Proc. Amer. Acad., xli, p. 559, 1906. 



