Serviss — Internal Temperature Gradient of Metals. 461 



(6) The specimens were suspended in the air bath so as to 

 expose them at all surfaces equally, instead of piling them on 

 a block of stearine and a paper box. 



(7) Finer wires and thinner insulation were used in the ther- 

 mopile, bringing the cylinders closer together. 



(8) The lead wires from the thermopile to the outside con- 

 nections were much longer, insuring against conduction of 

 heat to the end junctions. 



Observations and Results. 



I shall now give a brief history of the experiments. As 

 Dr. Thwing found the largest effect in lead, it was selected; 

 iron was taken for a second metal, chiefly because of the ease 

 with which it can be worked. Four cylinders of each metal 

 12x5 cm and two 12x9 cm , were cast early in January; all of 

 each metal were of the same pouring. Two of the smaller size 

 were also cast from a different sample of lead. One pair of 

 the smaller cylinders of each metal was turned at once and 

 laid aside while the rest of the apparatus was being con- 

 structed ; owing to delays and trouble with the ice bath, the 

 final observations were not made until July. 



When I outlined this investigation, I provided for certain 

 variations and check experiments. I planned to use cylinders 

 from the same pouring with fresh surfaces and with aged sur- 

 faces, hence the extra small cylinders and the considerable 

 allowance for turning ; also as a check experiment, to use 

 cylinders twice as thick ; but for this, I should have made my 

 thermostat 5 to 8 cm shorter. As my results have been entirely 

 negative, these other forms of the experiment have not been 

 made. The work on the radio-activity of metals shows that 

 it may be much lessened by scraping the surfaces, and if a 

 specimen long exposed to the air does not give a positive 

 result, a fresh one could not be expected to do so. The varia- 

 tion with thicker cylinders was also given up for the same 

 reason. 



Throughout the work, all the ice and drip-water was weighed 

 to detect honeycombing of the ice. The ice was kept at nearly 

 the same level by the addition of a sufficient quantity every 

 four or five hours to fill the vat to a certain mark. The ther- 

 mopile was tested before and after each pair of specimens. 

 There was a very slight e.m.f. detected, clue probably to uncon- 

 trolled external conditions. This was less than the slight 

 continuous movement of the zero. 



The first run (April 5-13) was made on lead. In this set of 

 observations the current from the junctions was sent directly 

 through the galvanometer, which was only about 1/20 as sen- 



