HALL AND AYRES. — HEAT CONDUCTION IN IRON. 301 



per square centimeter in water-gram units " equals 0.000252. Assum- 

 ing this rate of emission or absorption for each of the 50 sq. cm. of the 

 curved surface of the disk, we should get for the passage of lieat per 

 second 50 X 0.000252 X 5 = 0.0G3 units. The heat passii.g per second 

 through the disk from face to face was usually about one hundred times 

 as much as this. If we as&ume, for the moment, that the passage of heat 

 through the curved surface of the disk is equal to 1 % of that which Hows 

 from face to face, we may thereupon reason as follows. When the warmer 

 stream flows above the disk, the disk takes in heat from the jacket, and 

 the total amount passing out through the lower face of the disk exceeds 

 by 1% the amount flowing in at the upper face. The inflow at the 

 curved surface distorts the isothermals and lines of flow within the disk 

 in such a way that, with a given difference of temperature between the 

 faces, the flow from fixce to face, which is equal to the inflow at the upper 

 face, is less than it would be if the flow within the disk were adiabatic, 

 that is, if there were no inflow at the curved surface. On the other 

 hand, the outflow at tlie lower face is greater than the adiabatic flow 

 from face to face would be. We may conclude that, under the conditions 

 assumed, the actual inflow at the upper face is about 0.5% less than the 

 adiabatic flow would be with like temperatures at the faces, and that the 

 outflow at the lower face is about 0.5% greater than the adiabatic flow 

 would be. Our method of calculating ^assumes adiabatic flow, while 

 our observations give us the inflow at the upper surface. Accordingly, 

 under the conditions here assumed, the value obtained for K would be 

 about 0.5% too small. A similar error, in the same direction, would be 

 made with the colder stream flowing above the disk, so that we could 

 not eliminate it by combining two sets of observations. In fact, however, 

 the assumption that the rate of transmission at each square centimeter 

 of the curved surface of the disk, thickly wrapped with cotton, is the same 

 as that found by McFarlane for a bare blackened surface of copper, 

 gives a very large overestimate of the possible error from this source. 



Turning now to the surfaces between the top of the disk and the parts 

 Ji, Jo, Figure 2, we find the area of these surfaces to be about 300 sq. 

 cm. No systematic observations of the temperature of the jacket were 

 kept during the experiments of which this paper gives an account, but 

 from previous observations it appears probable that the mean tempera- 

 ture of the jacket was about 1° C. lower than the temperature of the 

 parts which we are now considering, when the warmer stream ran 

 above, and 0°.5 or less higher than the temperature of these parts when 

 the colder stream ran above. Assuming the difference of temperature 



