White — Specific Heats of Silicates and Platinum. 339 



distance from the water. The cover used here is in the form 

 of a floating cup, and therefore had always the same temper- 

 ature as the rest of the calorimeter surface. ( 4 ) An inno- 

 vation has been made in calorimeter practice by working with 

 temperature intervals much greater than usually employed (in 

 one case, 23°, which is more than ten times the ordinary rise ). 

 This method increases the accuracy by diminishing the relative 

 value of thermometric and other important errors. It requires 

 an allowance for the variation in the cooling rate over the wide 

 temperature intervals ( deviation from Newton's law ), but the 

 difficulties of this correction have proved absolutely insig- 

 nificant — far less than had been anticipated.* 



No work has yet been done with this calorimeter of suffi- 

 cient precision in other respects to fully test its accuracy. 

 From the agreement (0*1 of 1 per cent) obtained in determina- 

 tions of a heat quantity no more than one-sixth to one- 

 twentieth of that usually employed, the accuracy is seen to be 

 more than sufficient for all requirements of the present work. 



The Specific Heats. — Specific heat, like density and con- 

 ductivity, is a property varying with the temperature. Unlike 

 them, it is almost never determined directly for any particular 

 temperature. For the specific heat is, essentially, the heat 

 given out by a body in falling through a given temperature 

 interval divided by the interval. To give the true specific 

 heat at any temperature, this interval should be infinitesimal ; 

 in practice, it is necessarily finite and often very large. The 

 result obtained is the mean specific heat for the interval, from 

 which the different true values occurring within the interval 

 -may vary widely. If a single interval only is employed, the 

 relation between the mean and the true heat can not be deter- 

 mined ; hence, unfortunately, most published values are 

 of the mean heat only and give merely approximations to the 

 true heats. If data are available for several different intervals 

 all values of the true specific heat within them can generally 

 be obtained. Three computations were required in the present 

 work : (1) The mean heats were corrected down to zero, (2) 

 to even temperatures at the upper end, and (3) the true 

 specific heats were then derived from these corrected mean 

 heats. The first correction was performed as follows: Let 

 M be the observed mean heat, found between the tempera- 

 tures, t and o . Let m 1 be the mean heat from to 1? M 2 from 

 to # , and M 3 from to the even upper temperature ® 3 . 

 Equating total heats, M 2 2 = M (0..-0J + m x 0, 



whence M 2 = M + (m l - M ) -± (1) 



2 



*See Phys. Rev., xxviii, 462, 1909. 



