Serviss — Internal Temperature Gradient of Metals. 457 



steady conditions were not realized. I attempted to obviate 

 this difficulty by making the bottom of both vessels conical at 

 an angle of 45°, hoping that the weight of the ice would cause 

 it to settle and replace that which melted, and so establish a 

 steady condition. But I found almost immediately that owing 

 to the comparatively narrow annular space between the two 

 cans and the regelation of the ice, this device alone was insuf- 

 ficient. Still it is an improvement over a flat-bottomed 

 thermostat, for the ice is kept in contact with the whole of the 

 conical bottom all the time, which materially aids in maintain- 

 ing constant conditions. 



Throughout the work, ice was added at regular intervals, 

 usually every four or five hours, sufficient to keep it near a fixed 

 level. At the same time, the drain water was weighed. 

 .During the use of the first arrangement, these weights were 

 irregular, indicating a honeycombing of the ice and regelation. 

 When the ice was crowded clown, considerable changes were 

 noticed in the temperature, similar to those observed by Dr. 

 Thwiug. About 12 kgms. of ice were used in 24 hours, the 

 amount depending somewhat on the room temperature. 



"Next a spuclger was built out of an iron casting and six 

 J-inch iron rods projecting vertically downward in a circle mid- 

 way between the thermostat and the vat. A fan motor and 

 reducing gear raised and lowered this, with some freedom of 

 rotation, every seventeen seconds. A very good quality of 

 natural ice was used throughout this work. During these tests 

 it was ground into pieces weighing 10 grams or less. For a 

 time this seemed to be successful. But the pressure of the 

 prongs raised the melting point sufficiently to disturb the 

 gradient and regelation followed. The ice from the upper 

 third of the vat was removed, and the remainder was found 

 frozen into nearly a solid mass, in which the melting point was 

 variable under the changing pressure of the spudger. The 

 external junctions showed sudden and considerable changes in 

 temperature within a few minutes. Consequently this scheme 

 was abandoned. 



The third attempt was successful. Larger lumps of ice 

 were used, averaging 35 or 40 grams in weight. These were 

 cut from the 100-lb. pieces with a sharp chisel, and the corners 

 rounded by washing in running tap water, which prevented 

 relegation, and the consequent freezing into a single, solid 

 mass. But the melting point still varied a little, as indicated 

 by the fluctuations in the gradient. It is a well known fact 

 that the temperature of a mixture of ice and water rises con- 

 siderably above zero, depending on the amount of water that 

 has accumulated. By stirring the ice and collecting the drip 

 at short intervals, it was found that water was held between 

 the lumps of ice by capillary action. \ As the stirring was con- 



