PHYSICS: W. P. WHITE 
343 
THE GENERAL CHARACTER OF SPECIFIC HEATS AT HIGH 
TEMPERATURES 
By Walter P. White 
Geophysical Laboratory, Carnegie Institution of Washington 
Communicated by A. L. Day, September 14, 1918 
• 
The wide variations presented by atomic heats at low temperatures have 
been brought under a single generalization, usually expressed in terms of the 
quantum hypothesis, in the light of which all follow a common law. This law 
still demands that at sufficiently high temperatures all atomic heats at constant 
volume should have the same constant value, approximately 5.96. The ap- 
plicability of this theory, however, has been doubted; the supposition has been 
made that the atomic heats continue to increase with temperature, and this 
has been proved to be true of some metals. The present paper gives data for 
some non-metallic substances which show the probable existence of such an 
effect in every case. The substances in question are silica and various sili- 
cates. Their specific heats were determined by dropping relatively large 
charges (50 to 80 grams) from electric furnaces into calorimeters. Tempera- 
tures up to 1400° were used in three cases, and up to 1100° at least in most 
of the others. 
The accuracy of the results determines the value of the evidence, and there- 
fore needs attention. 1 (1) As to the 'accidental,' Visible,' or continually 
varying errors, the discrepancy between duplicate measurements averaged 
under 0.5 per mille. Moreover, this small amount of discrepancy was largely 
, accounted for by the irregularity of heat loss in dropping the hissing hot body 
into water, while this loss, except for its irregularities, was completely can- 
celed out by the procedure used. (2) The unvarying, or systematic, errors 
were checked by using two each of practically every piece of apparatus, 
such as furnaces and calorimeters. Comparisons nearly always showed 
discrepancies under 1 per mille. 
The temperature scale was established by means of one of the original Day 
and Sosman thermoelements, confirmed by numerous concordant metal melt- 
ing points. 
(3) The slowly or occasionally varying errors, the most insidious and detri- 
mental, were checked by repeating some determinations at intervals of several 
months, with uniformly satisfactory results, and also by general comparisons 
as in table 1 . Here about half the results are compared by giving their dif- 
ferences from a standard fictitious silicate which is approximately the mean of 
the whole. Of forty-five results above 0° to 100° only three, sporadic ones, 
those italicized, show discrepancies from a smooth curve exceeding, and then 
only slightly, the allowed accidental discrepancy of 1 per mille, and, what is per- 
haps more important, the different comparable series of values, pseudo-wollas- 
