344 
PHYSICS: W. P. WHITE 
tonite-diopside-enstatite, and anorthite-andesine-albite, show almost perfect 
concordance. The temperature distribution in the furnace, the most dangerous 
source of error, was studied with especial care, and the maximum irregularity 
detected corresponded to 0.3 per mille. The precautions against systematic 
error thus appear to be unusually numerous, and the conclusion seems war- 
ranted that most of the original results, for mean atomic heats over inter- 
vals, are correct to better than 3 per mille, of which less than 1 per mille is 
accidental. 
The method of obtaining from these interval atomic heats the true atomic 
heats at single temperatures was demonstrably such that original accidental 
TABLE 1 
Interval Atomic Heats of Silicates, Compared by Being Given as Differences from 
a Fictitious Standard Silicate 
SILICATE, MINUS STANDARD 
0°-100° 
0°-300° 
0°-500° 
0°-700° 
0°-900° 
0°-1100° 
0.023 
0.036 
0.039 
0.039 
0.038 
0.043 
MgSi0 3 Pyroxene 
0.036 
0.059 
0.084 
0.108 
0.109 
0.108 
0.109 
0.109 
0.107 
PW + Py 
2 
0.129 
0.115 
0.111 
P. Woll 
0.223 
0.170 
0.137 
0.113 
0.093 
0.082 
Woll 
0.149 
0.139 
.0.136 
0.012 
(008) 
0.013 
0.001 
0.006 
0.019 
-0.055 
-0.056 
-0.057 
-0.053 
Anor + Alb 
-0.048 
-0.048 
-0.056 
-0.052 
2 
Albite 
-0.107 
-0.109 
-0.108 
-0.113 
-0.109 
-0.107 
-0.096 
-0.114 
-0.112 
-0.112 
-0.116 
-0.126 
0.006 
0.003 
0.013 
0.017 
0.021 
0.063 
4.067 
4.558 
4.913 
5.143 
5.316 
5.453 
errors of 1 per mille might possibly produce single errors of 1% and would be 
certain to produce single errors of about 5 per mille in a few cases, but would 
not make any series of results wrong by much more than the original error. 
Results.— Results for seven materials are given in table 2 as true mean 
atomic heats. Of the substances included, silica glass is known to have an 
almost zero expansion coefficient, hence its atomic heat, though obtained at 
constant pressure, is the atomic heat at constant volume, or A v . Cristobalite, 
a crystalline form of silica, shows a still lower atomic heat, and also shows it 
up to a considerably higher temperature. Quartz, which also is lower, has a 
negative coefficient of expansion from 600° to 1000° and the very low value of 
A p for cristobalite may be partly or wholly a result of negative expansion. 
If the small thermal effect of the contraction of quartz is computed and taken 
