462 Serviss — Internal Temperature Gradient of Metals. 



sitive as in the other runs. The ice was kept covered with a 

 blanket, and crowded down from time to time. A steady con- 

 dition was not reached, owing to the manner of managing the 

 ice-bath. The gradient was variable, even reversing at times, 

 after the ice had melted out, at times rising to values fully as 

 large as those observed by Dr. Thwing ; but even this imj)erfect 

 series gave no indication of a steady gradient, i. e. there was 

 no positive effect toward which it returned after being dis- 

 turbed. 



The secoud run (May 8-20). The spudger had meanwhile 

 been installed, the sensitiveness of the galvanometer increased 

 to its final value (4"lXl0~ n amp /mm from zero at a scale 

 distance of 4*2™), Owing to trouble with the spudger and a 

 final breakdown, this run gave no additional evidence on the 

 problem. 



The third run (June 29- August 1). The use of the spudger 

 was at first continued, but after a few days was replaced by 

 stirring the ice at frequent intervals. In this run the work 

 was continued day and night, much of the observing being 

 done at night when the room temperature was less variable 

 and the building was free from mechanical jars. The inter- 

 missions were brief, and during them the apparatus was left 

 under the care of some other person, who stirred the ice occa- 

 sionally to prevent too great an accumulation of water and 

 who added more ice if necessary. Two samples of lead and one 

 of iron were very carefully tested and sufficiently steady con- 

 ditions were maintained for three or four hours at a time to 

 show unmistakably that no mean gradient as great as o- 0000001 

 existed. 



It would be interesting to compare the cooling curves with 

 those computed, but as the labor of maintaining steady condi- 

 tions was so great I did not attempt to make observations until 

 I knew the cylinders had cooled practically to zero throughout. 

 The specimens were placed in the thermostat at about 20°, 

 and it seemed to require fully 24 hours to bring the junctions 

 of the thermopile to sensibly the same temperature. This is 

 considerably longer than is required for the cylinders to cool 

 to the same gradient, the difference indicating a lag between 

 the junctions and the cylinders, due to the insulation of the 

 asphaltum, but this is no evidence that they did not ultimately 

 assume the temperature of the adjacent metal. The pressure 

 at the bottom of the thermostat, due to the weight of 25 cm of 

 ice, is about 1/50 of an atmosphere, which will cause a lower- 

 ing of the melting temperature of o, 00015. Possibly, when 

 the vat is jarred by the stirring, convection currents may be 

 set up, but they will not affect the temperature of the copper 

 wire projecting beyond the cylinders, for the additional mica 



