The heat capacities of 27 glasses have been determined from room temperatur
e to temperatures corresponding to supercooled liquid behavior. The investi
gated compositions are based on a haplogranite (near the 2-kbar pH(2)O mini
mum melt composition in the system NaAlSi3O8-KAlSi3O8-SiO2) to which alkali
and alkaline earth oxides (Li2O, Na2O, K2O, Rb2O, Cs2O, MgO, CaO, SrO, and
BaO) have been added individually. Where comparison is possible, our data
below the glass transition are consistent with predictive models previously
proposed in the literature. In addition, the partial molar heat capacity o
f Li2O in silicate glasses is determined from our data. Extrapolated glassy
and relaxed liquid enthalpies intersect at a temperature defined as the li
miting fictive temperature (T-f'). At this temperature, the glassy heat cap
acity of all the studied compositions is close to the theoretical upper lim
it of 3R per gram-atom, where R is the gas constant. For samples cooled at
5 K/min, liquid viscosity of all samples is 10(12.56 +/-0.43) Pa s at T-f'.
For other cooling rates, this result implies that log eta (T-f') = 11.5 -
log Q, where eta (T-f') is the viscosity at the limiting fictive temperatur
e and Q is the cooling rate (K/s). Liquid heat capacity is found to general
ly increase with addition of all oxides, although the details of the variat
ions are obscured by the fact that experimental uncertainties are of a simi
lar magnitude to variations in heat capacity caused by compositional change
. On the other hand, the "configurational heat capacity" (C-p(conf)(T-f')),
defined as the difference between the fully relaxed liquid heat capacity a
nd the glassy heat capacity at the limiting fictive temperature, shows much
less dispersion as a function of composition. Its variation is a nonlinear
function of composition, with little, if any, change for additions of oxid
e less than 10 mol%, but increasing values for greater additions of oxides.
By use of previously determined liquid expansivities, we calculate that vo
lume changes account for similar to 15% of the configurational heat capacit
y. We conclude that liquid heat capacity should be considered as the sum of
a vibrational contribution, of value close to 3R per gram-atom, and a conf
igurational contribution related to liquid structure, rather than trying to
define a single partial molar heat capacity for each liquid oxide componen
t. Copyright (C) 2001 Elsevier Science Ltd.