Inorganic Materials, Vol.36, No. 6, 2000, pp. 529-543. Translatedfrom Neorganicheskie Materialy, Vol.36, No. 6, 2000, pp. 647--662.

Original Russian Text Copyright 9 2000 by Zaitsev, Shelkova, Mogutnov.

Thermodynamics of Na20-SiO2 Melts

A. I. Zaitsev, N. E. Shelkova, and B. M. Mogutnov
Kurdyumov Institute of Physical Metallurgy and Functional Materials,
Bardin Central Research Institutefor the Iron and Steel Industry,
Vtoraya Baumanskaya ul. 9/23, Moscow, 107005 Russia
ReceivedApril 15, 1998; in final form, March 29, 1999

Abstract--The thermodynamic properties of Na20-SiO 2 solid (942-1285 K) and liquid (1103-1719 K,

19.5-61.8 mol % Na20) silicates were studied by Knudsen cell mass spectrometry, To determine the activities
of the constituent oxides, these were reduced to volatile suboxides directly in effusion cells. Mass spectra of the
+ +
saturated vapor over Na20-SiO2 showed the presence of the Na+, Na2O+, NaO +, 02 , TaO § TaO 2 , NbO+,
NbO~, MoO+, MoO 2 , MoO 3 , and NiO + ions resulting from the ionization of the Na, Na20, NaO, NaO 2, 02,
TaO, TaO 2, NbO, NbO 2, MoO, MoO 2, MoO 3, and NiO molecules. The activities calculated by two different
procedures were found to coincide within the experimental error. The enthalpies and Gibbs energies of forma-
tion of sodium silicates were shown to be extremely low. The formation of solid ortho- and metasilicates is
accompanied by a decrease in entropy, in contrast to the other sodium silicates. Sodium orthosilicate has the
lowest enthalpy and Gibbs energy. A thermodynamic model for Na20-SiO 2 melts is proposed which relies on
associated solution theory and takes into account silica polymerization. The model describes the composition
and temperature dependences of the activities of the constituent oxides in the melt with an accuracy no worse
than the experimental error (2-3%). The model, in combination with the thermodynamic functions of formation
of all the intermediate solid phases, was used to calculate phase equilibria in the Na20-SiO 2 system. The results
agree well with the experimental data obtained by physicochemical methods.

INTRODUCTION to volatile suboxides directly in effusion cells:

Sodium silicates enter into the composition of many nNa20(cr) + R(cr) = RO~(g) + 2nNa(g), (1)
minerals and are therefore of immense importance in nSiO2(cr) + R(cr) = ROn(g) + nSiO(g), (2)
chemistry, geochemistry, and petrology. They are
where R is the reducing agent (cell material: Nb, Ta,
widely used in the production of glass and ceramics, in
Mo, or Ni). In a few experiments, powders of reducing
civil engineering, etc. The unique ability of silica to agents were added to the mixtures to be studied. Ther-
form polymeric molecules of arbitrary dimensions and modynamic calculations and measurements show that,
orientation engenders considerable theoretical interest in the Na20-SiO2 system, only reaction (1) takes place.
in sodium silicates. Since the activity of the reductant was in all cases unity,
In earlier studies [1-5], we developed an approach the equilibrium constant of this reaction is given by
which allowed us to significantly extend the composi- p ( R O n ) [ p ( N a ) ] 2n
tion and temperature ranges of Knudsen cell mass spec- Kp(1) =
trometry and provided the theoretical basis for model- [a(Na20)]" (3)
ing the thermodynamic behavior of silicate melts. The 0 0 2n
= p (ROn)[p (Na)] ,
objective of this work was to obtain reliable data on the
thermodynamic properties of phases in the Na20-SiO 2 where the pressures labeled 0 refer to mixtures in which
system and to apply the thermodynamic model of liq- the activity of Na20 is unity. Therefore, studying mix-
uid silicates [1, 5, 6] to Na20-SiO 2 melts, characterized tures with high and low Na20 contents, one can deter-
by strong interaction between the constituent oxides. mine Na20 activity as

[a(Na20)]n = p(ROn)[P( Na)]2n

0 0 2n
EXPERIMENTAL p (ROn)[p (Na)]
(4)
The thermodynamic properties of the melts were /(RO:)[I(Na+)] 2n
studied by Knudsen cell mass spectrometry. As in ear-
lier studies [1-5], the constituent oxides were reduced I~176 2n"

0020-1685/00/3606-0529525.00 9 2000 MAIK "Nauka/Interperiodica"

530 ZAITSEV et al.
Here, I and I0 are the corresponding ion intensities--
either the sums of the ion currents produced by ioniza-
tion of particular molecules or those parts of the total
ion currents corresponding to molecular ions. In both
cases, in going from partial pressures to ion intensities,
the total or partial ionization cross sections of the corre-
sponding molecules canceled. This notably improves the
accuracy in the activity of the constituent oxides [1, 4].
Because of the high volatility of Na20, mixtures with a
high Na20 content could not be studied at high tem-
peratures, and a(Na20) values were calculated by equa-
tion (3) using reference data [7] for determining Kp
in (1).
Samples for this investigation were prepared from
extra-pure-grade SiO2 dried in vacuum and Na20
obtained by thermal dissociation of extra-pure-grade
Na2CO3. Some of the samples were synthesized
directly in effusion cells, the others were prepared by
reacting the constituent oxides in closed Ni crucibles
under a vacuum of 10-4 Pa or better. Chemical and
x-ray diffraction analyses of solidified samples showed
that the dissolution of Ta, Nb, Mo, and Ni oxides in the
samples was insignificant and the oxygen stoichiome-
try of the constituent oxides remained virtually
unchanged. In other words, the representative point of
the sample composition remained on the Na20-SiO2
pseudobinary join of the Na-O-Si system. The x-ray
patterns contained only lines corresponding to equilib-
rium phase relations in the Na20-SiO2 system.
In our experiments, we used double effusion cells.
The sample was placed in one compartment and the ref-
erence standard in the other. As reference materials, we
used Ag (99.99% purity), Ni (99.98%), and Cu
(99.999%). To verify that a nearly equilibrium state was
reached in the effusion cell, we varied the effusion area,
all other dimensions being the same, and used cells
from different materials. Under the conditions of our
measurements, changes in sample composition were
negligible. The procedure of high-temperature mass
spectrometric measurements was described in detail
elsewhere [8, 9].
Mass spectra of the saturated vapor over Na20-SiO2
showed the presence of the Na§ Na2O§ NaO § 02,
TaO +, TaO;, NbO +, NbO 2 , MoO +, MoO 2 , MOO;,
and NiO + ions resulting from the ionization of the Na,
Na20, NaO, NaO 2, 02, TaO, TaO2, NbO, NbO2, MoO,
MOO2, MOO3, and NiO molecules. The strongest sig-
nals were those from Na*, TaO 2 , NbO 2 , and MoO 2 .
O~ ions were detected only in experiments with Ni and
Mo cells. The peaks from the species produced by ion-
ization of Mo, Nb, and Ta oxides were identified using
earlier results [2-5]. The origin of the Na*, Na2O+,
NaO § and 02 ions can be understood only if the mech-
anism of Na20 vaporization is known. Na20 was
reported to vaporize congruently [10, 11] by the reaction
2NazO(cr ) - 4Na(g) + O2(g). (5)
Therefore, Na and 0 2 would be expected to be the
major components of the saturated vapor over Na20,
with p(Na) = 4p(O2). The concentrations of Na20,
NaO, Na202, and NaO 2 molecules were assumed to be
much lower, by three and more orders of magnitude
[ l 0, 11 ]. However, according to a recent detailed study
[12], NaO 2 molecules are quite stable in the gas phase
and their concentration in the vapor over Na20 is rela-
tively high. This implies that Na20 may vaporize not
only by reaction (5) but also according to the equation
2Na20(cr ) = 3Na(g) + NaO2(g) (6)
with p(Na) = 3p(NaO2). Thus, the ion current I(Na§
may comprise two contributions, one due to the ioniza-
tion of Na atoms and the other to the ionization of NaO2
molecules. The formation of O2 ions is due mainly to
the dissociative ionization of NaO 2 molecules, which is
unlikely to yield NaO + ions. The absence of the 02 sig-
nal in the mass spectrum of the saturated vapor over
Na20-SiO2 in Nb and Ta cells testifies to low concen-
trations of NaO2 and 02. Therefore, Na atoms are the
only source of Na + ions. In contrast, the presence of the
02 signal in experiments with Ni and Mo cells indi-
cates that there is a second source of Na + ions---NaO2
molecules---clearly, if there is a second mechanism of
vaporization. To check this assumption, we carried out
measurements in Nb, Ta, Mo, and Ni cells for pure
Na20, Na20-SiO2 melts of a particular composition,
and mixed-phase samples with compositions falling in
the same two-phase region. The Na partial pressures
measured with Nb and Ta cells differed markedly, by a
factor of 1.6-1.7, whereas the activities calculated by
equations (3) and (4) coincided within the experimental
error (1-3%). The Na20 activities determined with Mo
and Ni cells were substantially higher, obviously
because the second mechanism--dissociative ioniza-
tion of NaO2 molecules--came into play, contributing
to the Na + signal. To evaluate this contribution, we car-
ried out experiments in which the residual oxygen in
the spectrometer chamber was frozen out using a liq-
uid-nitrogen trap situated in the source region or by
cooling the jacket of the Knudsen unit with liquid nitro-
gen. This enabled us to eliminate the corresponding
systematic error in the O 2 signal and to determine the
INORGANIC MATERIALS Vol. 36 No. 6 2000
 THERMODYNAMICS OF Na20-SiO 2 MELTS 531

intensity ratio of the Na § and 02 ion currents produced 241

.1
by the ionization of NaO2 molecules. Indeed, we have +2
D3
l ( N a § NaO2) = bl(02, NaO2) = bl(O2),
+ (7) 240

l ( N a § Na) = l ( N a § - l ( N a § NaO2)
(8)
= l ( N a § - bl(O~, NaO2), ~
,..r.
239

where b is a constant of the mass spectrum, and l(Na § 9

and I(O2 ) are the measured signal intensities. The rate Z 238
of reaction (1) and, hence, the Na and NaO 2 partial
pressures depend on the cell material, which can be I
used to separately assess the contributions to the Na § 237 o
signal. Constant b was calculated using the equilibrium
constant of reaction (6) expressed through ion currents,

{/'(Na § - br(02) }/{ l"(Na § - br'(02) } 236 i i i J a I

(9) 960 980 1000 1020 1040 1060 1080

, + + 1/3
= {r(o2)/r(o2)} , T,K
where the primed and doubly primed ion intensities Fig. 1. Partial Gibbs energy of Na20 as a function of tem-
refer to the results obtained for the same composition in perature for the N3S8 + SiO2 and N3S8 + NS2 two-phase
Mo and Ni cells, respectively. Constant b was found to regions: {1) x(SiO2) = 0.805, Nb cell, diameter of the effu-
be independent of sample composition and temperature sion hole D = 0.203 mm; (2) x(SiO2) = 0.733, Ta cell, D =
and equal to 1.72 at an ionizing-electron energy of 0.314 mm; (3) x(SiO2) = 0.709, Nb cell, D = 0.247 mm;
(4) x(SiO2) = 0.671, Ta cell, D = 0.356 ram.
50 eV. Using this value to correct the Na § signal inten-
sities measured in experiments with Mo and Ni cells,
we obtained Na20 activities which coincided, within 216
the experimental error, with those determined in exper- ,1
iments with Nb and Ta cells. Subsequently, the results r
obtained with Mo and Ni cells were used in activity cal- 215 a3
culations only if I(O 2 ) could be measured reliably and,
hence, the contributions to l(Na +) could be determined
separately. Using the equilibrium constant of reaction (6), ~214
we were also able to calculate a(Na20) from the exper-
9
imental data for mixtures with high and low Na20 con-
tents. In all cases, the Na20 activities determined under z 213
different conditions coincided within the experimental r~
error (1-3%). i

212 * * ~ * ~ ~
RESULTS AND DISCUSSION
Solid silicates. According to earlier reports [ 13-18],
the Na20-SiO 2 system contains five compounds: N2S, 211 J I I -
NaS 2, NS, NS2, and N3Ss (hereafter, N = Na20 and S = 950 1000 1050 1100 1150
SiO2). For each of the two-phase regions, we studied T,K
a few compositions. In addition, we studied the stoichi-
ometric compounds NS and N2S. Our data on the par- Fig. 2. Partial Gibbs energy of Na20 as a function of tem-
tial Gibbs energies of Na20 in the two-phase regions of perature for the NS + NS2 two-phase region: (1) x(SiO2) =
the Na20-SiO2 system are summarized in Figs. 1-3. As 0.625, Nb cell, D = 0.322 mm; (2) x(SiO2) = 0.573, Ta cell,
D = 0.208 mm; (3) x(SiO2) = 0.524, Mo cell, D = 0.351 mm.
the standard state of Na20 we took the ~-form. To
reduce our results to the standard state, we used the ref-
erence data given in the IVTANTERMO database of positions lying within a particular two-phase region
thermodynamic properties [7]. As apparent from coincide within the experimental error. This indicates
Figs. 1-3, the AG(Na20) values determined under dif- that the procedure used to calculate Na20 activity is
ferent conditions (effusion area, cell material) for com- quite correct and that the effusion system was in equi-

INORGANIC MATERIALS Vol. 36 No. 6 2000

532 ZAITSEV et al.

145 ever, in later, probably more detailed studies, its for-

140
~ 135
A
9 130
Z mental error below 1068 K and differ at higher temper-
~ 125
I
+5
120 ~ 7
• mined from the temperature dependences of the chem-
115 I I I I
900 1000 1100 1200 1300
T,K
Fig. 3. Partial Gibbs energy of Na20 as a functionof tem-
peraturefor the N3S2 + NS, N3S2 + N2S,and N2S + NS two-
phase regions: (1) x(SiO2) = 0.477, Nb cell, D = 0.311 mm;
(2) x(SiO2) = 0.430, Ta cell, D = 0.203 mm; (3) x(SiO2) =
0.450, Mo cell, D = 0.207 mm; (4) x(SiO2) = 0.405, Mo cell,
D = 0.351 mm; (5) x(SiO2) = 0.382, Nb cell, D = 0.209 mm;
(6) x(SiO2) = 0.382, Ta cell, D = 0.314 mm; (7) x(SiO2) =
0.360, Mo cell, D = 0.248 mm; (8) x(SiO2) = 0.349, Ni ceil,
D = 0.212 mm.
librium during measurements. Therefore, the thermo-
dynamic data obtained are accurate and reliable.
The curves obtained for the NS 2 + N3S a and
NaSa + S regions intersect (Fig. 1). The intersection
point corresponds to the peritectoid decomposition of
N3S a into sodium disilicate and quartz. The temperature
of this transformation was calculated to be 973 K,
which coincides with the value obtained by William-
som and Glasser [17] in physicochemical studies. Note
that there is no general agreement regarding the compo-
sition of the higher sodium silicate. Budnikov and
Matveev [ 19] argued that its composition is NS3; how-
mula was determined to be N3S a [17]. Our results are
consistent with the latter composition, since the chem-
ical potential of Na20 at x(SiO2) = 0.733 falls in the
range characteristic of the equilibrium between the
higher sodium silicate and quartz rather than NS2.
An unusual behavior is exhibited by the chemical
potential of Na20 in the N2S + N3S2 and N3S 2 + NS
two-phase fields (Fig. 3). The AG(Na20) values for
these two-phase regions coincide within the experi-
atures. This indicates that N3S2, stable at lower temper-
atures, decomposes peritectoidally into a mixture of
sodium ortho- and metasilicates at 1068 K.
The partial enthalpy and entropy of Na20, deter-
ical potential by least squares fitting, are listed in
Table 1, where errors are given as 95% confidence
intervals. The thermodynamic functions of formation
of intermediate phases in the Na20-SiO2 system were
calculated for stoichiometric compositions by the
Gibbs-Duhem equation:
AF{x'(SiO2)} = [x'(SiO2)lx"(Si02)]
xAF{x"(SiO2)} + { 1 -x'(SiO2)/x"(Si02)} (10)
x AF{x'(Si02), x"(SiO2) },
where AF{x'(SiO2)} and F{x"(SiO2)} are the total
molar functions of phases (') and C) with silica contents
x'(SiO2) and x"(SiO2), respectively; C) designates the
neighboring phase on the silica-rich side; and
AF{x'(SiO2), x"(SiO2) } is the partial molar function of
Na20 in a phase mixture of compounds with silica con-
tents x'(SiO2) and x"(SiO2).
The calculation results (Fig. 4, Table 2) show that
the entropies of formation of sodium ortho- and meta-
silicates are negative, whereas the formation of the
other silicates is accompanied by an entropy gain.
Sodium orthosilicate has the lowest enthalpy and Gibbs
energy. However, since the entropy of formation of N2S
is notably larger in magnitude than AjS(NS), the Gibbs
energies of ortho- and metasilicates approach one

Table 1. Partial thermodynamic functions of Na20 in phase Table 2. Thermodynamic functions of formation of sodium
mixtures of the Na20-SiO 2 system; standard state, []-Na20 silicates from quartz and ~-Na20

Phase A T, K -AH(Na20)' AS(Na20)' Phase AT, K -AfH, J/mol AtS, J/(mol K)

mixture J/mol J/(mol K)
N3Ss 974-1071 58300 + 755 6.70 + 0.70
NaSa + SiO2 974-1071 213800 + 2800 24.55 + 2.70
NS 2 + N3Ss 973-1078 242600 + 2160 -5.00 + 2.10 NS2 973-1078 73660 + 715 5.70 + 0.70
NS + NS 2 953-1110 240500 + 1040 -26.30 + 1.00 NS 953-1110 115400 + 600 -2.30 + 0.60
N3S2 + NS 1068-1285 90800 + 1680 39.90 + 1.40
N3S2 1068-1285 110500 + 580 6.15+0.55
N2S + N3S2 1072-1271 210200 + 2270 -72.20 + 1.90
N2S + NS 942-1063 150520 + 2330 -16.10 + 2.30 N2S 942-1271 127100 + 620 -6.90 + 0.60

INORGANIC MATERIALS Vol. 36 No. 6 2000

 THERMODYNAMICS OF Na20-SiO 2 MELTS 533
another with increasing temperature. The Gibbs ener- 140
gies of formation of sodium silicates from the neigh-
boring phases, AfG", provide a more conclusive assess- -.o-1
120
ment of their relative stabilities (Fig. 4). As might be
expected, two compounds, NS and N2S, stand out in the O
Na20-SiO2 system, N2S having the highest Gibbs
energy.
In this work, the high-temperature thermodynamic b 80
functions of sodium silicates were determined for the I
first time. The high-temperature functions reported in ~ 60
the literature are derived from heat capacity measure-
ments and low-temperature enthalpies of formation. To I
compare the data reported for N2S, NS, and NS2 with 40
the present results, we calculated standard thermody-
namic functions using the data given in Table 2. We also 20
used heat capacity data [20-22] and the thermody-
namic functions of SiO2 [23] and Na20 [7]. The results,
together with references data [22], are presented in
Table 3. Note that the uncertainties indicated in Table 2 0 0.2 0.4 0.6 0.8 1.0
are associated only with random errors. In evaluating x(SiO2)
the errors in the standard thermodynamic functions, we
took into account uncertainties in both the measured Fig. 4. Thermodynamicfunctions of formationof sodium
silicates from I~-Na20 and quartz: (1) AfG; (2) AfH;
parameters (temperature, ion currents, effusion area, (3) AfG', 1070K.
ionization cross sections of gaseous species) and the
thermodynamic functions used in the calculations. It
can be seen from Table 3 that the standard thermody- (Table 3). The estimates obtained for N3S 2 differ some-
namic functions of N2S, NS, and NS2 calculated in this what from experimental data, particularly in the case of
work and those recommended in [22] coincide within enthalpy. The discrepancy arises most likely from the
the experimental error. Of particular importance is the insufficient accuracy in thermodynamic functions
agreement between the entropies of formation and found by solving the reverse problem of chemical ther-
absolute entropies, because the values given in [22] are modynamics [18].
derived from low-temperature heat capacity data. The
thermodynamic functions of N3S2 and N3S8 were deter- Melts. Our measurements yielded a set of 550
mined here for the first time. The reported estimates a(Na20) activities in Na20-SiO 2 melts for various
[ 18] were obtained by optimizing the available experi- compositions and temperatures. SiO 2 activities were
mental data on the thermodynamic functions and phase calculated by integrating the Gibbs-Duhem equation
equilibria in the Na20-SiO2 system. The estimated with the use of the (z function: tz(Na20) =
thermodynamic functions of N3S8 are in close agree- ln{~(Na20) }/[ 1 -x(Na20)] 2. The position of the line of
ment with the present results. This appears quite natural SiO2-saturated melts, necessary for integration, was
considering that N3S8 is only stable in a narrow temper- determined from the coordinates of the points at which
ature range, and that its thermodynamic properties the composition dependences of Na20 activity level
were deduced from the thermodynamic functions of off, corresponding to a melt + solid SiO 2 two-phase
sodium disilicate [22], which agree with our data equilibrium. Subsequently, the positions of these points

Table 3. Standard thermodynamic functions of sodium silicates (this work and [22])

S~ (298.15 K), -AfH, kJ/mol AfS, J/(mol K) AfH, kJ/mol

Silicate Souse
J/(mol K) from oxides from elements
NS2 This work 165.4 + 4.0 227.4 + 5 7.3 + 3.0 2464.2
[22] 164.055 230.54 6.03 2467.3
NS This work 114.8+3.0 228.1 + 5 -1.7 + 2.0 1554.0
[22] 113.805 230.120 -2.76 1556.03
N2S This work 194.1 + 5.0 363.0 + 8 2.4 + 4.0 2104.0
[22] 195.811 359.8 4.15 2100.8

INORGANIC MATERIALS Vol. 36 No. 6 2000

534 ZAITSEV et al.

-ot(Na20) were refined in phase-diagram calculations (see

60 below). In all cases, the coordinates determined by the
two procedures differed insignificantly. For example,
55 according to the 1473-K isotherm, the saturation of the
melt with SiO 2 was attained at x(SiO2) = 0.805; phase-
50 diagram calculations yielded x(SiO2) = 0.808. Integra-
tion was carried out by the Simpson method. A typical
45 composition dependence of o~(Na20) is shown in Fig. 5.
Table 4 compares the measured and calculated a(Na20 )
40 and a(SiO2) data for some compositions.
In Fig. 6, we compare our data on the Na20 activity
35 in Na20-SiO 2 melts with those reported in the litera-
ture. In earlier studies, Na20 activities were determined
30 by various methods of chemical thermodynamics: emf
measurements in reversible concentration cells [24-
25 i i i i I 30], mixed-phase equilibrium studies [31-35], Knud-
0.4 0.5 0.6 0.9 0.7 0.8
sen cell mass spectrometry [36-38], and even ionic-
x(SiO2) molecular equilibrium studies [39]. Kohsaka et al. [26]
Fig. 5. Composition dependence of ot(Na20) and Ravaine et al. [28] mesured Na20 activities with
lnl~Na20)}/[l -x(Na20)] 2 at 1473 K. respect to melts with x(SiO2) = 0.6 and 0.67. To reduce
their data to the standard state (pure liquid Na20), we
used the Na20 activities found in this work for the cor-
responding SiO2 concentrations. Both the activities and
their temperature variations obtained in this work agree
-4
well with the data obtained in emf measurements [24,
26-30], mixed-phase equilibrium studies [34], and
studies of ionic-molecular equilibria [39] (Fig. 6,
Table 5). Somewhat higher a(Na20) values were
obtained by Knudsen cell mass spectrometry [36, 37]
-6 9 .e ~ ; 9 9
and in transfer measurements of the Na vapor pressure

Z
i9 over Na20-SiO 2 melts in equilibrium with graphite at a
fixed CO pressure [33]. It seems likely that these dis-
crepancies are related to the complex mechanism of
Na20 vaporization. Since the results reported in [12]
O were obtained later than those in [36, 37], Shul'ts et al.
~--8 [36] and Chastel et al. [37] took into account only reac-
tion (5) and attributed the entire Na + signal intensity to
the ionization of Na atoms. Reactions similar to (6)
oo might also contribute to p(Na) over Na20-SiO2 melts
0 x 1 ell
m in equilibrium with graphite in vapor pressure measure-
u ~ ~ 9 2 e 12
r 03 v13 ments by the transfer method [33]. Under the experi-
-10 , | 4 o 14
m D 9: . 5 =15
mental conditions used in [33], the formation of silicon
r 6 m16 carbide was quite possible. Dissolving in the silicate
r tl 7 t>17 melt, SiC could raise the Na20 activity. As can be seen
a 9 8 lXI8
9 9 VI9 from Fig. 6, Na20-SiO2 melts are characterized by
V o 10 ,20 extremely low Na20 activities. With increasing temper-
-12 I I I ature, a(Na20 ) rises rapidly, which corresponds to large
0.1 0.2 0.3 0.4 0.5 x(Na20) magnitudes of the partial enthalpy of Na20. This
Fig. 6. Na20 activityin Na20-SiO2 melts with respect to behavior was observed in all studies except in those
pure liquid Na20. This work: (/) 1573 K, (I/) 1473 K, reported by Pearce [31, 32], who found the opposite
(1II) i 373 K; earlierdata: (IV)1573K [33], (1/) 1673K [33], tendency. Note in this context that his a(Na20) data dif-
(I) 1450K [261,(2) 1373K [30],(3) 1573K [30],(4) 1773K fer markedly from those obtained in the other studies.
[30], (5) 1293 K [36], (6) 1423 K [36], (7) 1373 K [37],
(8) 1323 K [29l, (9) 1573 K [34], (10) 1473 K [34], The isotherms of SiO2 activity (Fig. 7) intersect at
(I1) 1373 K [34], (12) 1473 K [27], (13) 1473 K [25], x(Na20) - 0.3, indicating that the partial enthalpy of
(14) 1173 K [28], (15) 1273 K [28], (16) 1373 K [28],
(17) 1273K [31,32], (18) 1373 K [31,32], (19) 1473K [31, SiO 2 changes from positive at high silica concentra-
32], (20) 1473 K [24]. tions to negative at high Na20 concentrations. This

INORGANIC MATERIALS Vol. 36 No. 6 2000

 THERMODYNAMICS OF Na20-SiO 2 MELTS 535

Table 4. Comparison of measured and calculated SiO2 and Na20 activities in Na20-SiO 2 melts; standard states, pure liquid
SiO2 and Na20

a(SiO 2) a(Na20)
x(SiO2) T, K
experiment calculation experiment calculation

0.805 1473 0.860 0.854 7.19 x 10-1~ 7.13 x 10-1~

0.805 1573 0.842 0.845 2.61 x 10-9 2.65 x 10-9
0.753 1273 0,735 0.729 5.29 x 10-11 5.35 x 10-11
0.753 1473 0.714 0.712 1.35 x 10-9 1.34 x 10-9
0.753 1673 0.697 0.702 2.61 x 104 2.65 x 104
0.709 1273 0,532 0.536 1.20 x 10-1~ 1.22 x 10-1~
0.709 1473 0.541 0.536 2.94 x 10-9 2.89 x 10-9
0.709 1673 0.536 0.537 3.24x 104 3.24 x 10-8
0.671 1173 0.342 0.337 4.67 x 10-11 4.72 x 10-ll
0.671 1373 0.355 0.355 1.63 x 10-9 1.66 x 10-9
0.671 1573 0.365 0.370 2.41 x 104 2.38 x 104
0.625 1173 0.151 0.152 2.01 x 10-l~ 2.05 x 10-1~
0.625 1373 0.178 0.176 6.10 x 10-9 6.04 x 10-9
0.625 1573 0.199 0.196 7.63 x 104 7.65 x 10-8
0.573 1273 0.0565 0.0574 6.10 x 10-9 6.13 x 10-9
0.573 1473 0.0703 0.0698 1.02 x 10-7 1.01 • 10-7
0.573 1573 0.0757 0.0756 3.20 x 10-7 3.16 x 10-7
0.524 1373 0.0162 0.0160 1.45 x 10-7 1.46 x 10-7
0.524 1473 0.0177 0.0180 5.24x 10-7 5.17 x 10-7
0.524 1573 0.0204 0.0201 1.53 x 10--6 1.56 x 10-6
0.477 1373 1.75 x 10-3 1.77 x 10-3 1.32 x 10-6 1.31 x 10-6
0.477 1473 2.19 x 10-3 2.16 x 10-3 4.27 x 10-6 4.33 x 10-6
0.430 1373 3.12x 10-4 3.07 x 10-4 5.71 x 10-6 5.67 x 10-6
0.430 1473 3.80 • 10-4 3.83 x 10-4 1.83 x 10-5 1.83 x 10-5
0.405 1423 1.38 x l0 -4 1.40 x 10-4 2.01 x 10-5 1.98 x 10-5
0.405 1473 1.58 x 10--4 1.56 x 10-4 1.56 x 10-5 3.44x 10-5
0.382 1383 4.76 x 10-5 4.73 x 10-5 4.73 x 10-5 2.43 x 10-5
0.382 1423 5.15 x 10-5 5.20 x 10-5 5.20 x 10-5 3.71 x 10-5
0.349 1373 3.77 x 10-6 3.80 x 10-6 3.80 x 10-5 8.62 x 10-5

behavior can be understood in terms of the effect of
basic oxides on the structure of molten silica. Upon the
introduction of Na20 into high-silica solutions, the gain
in enthalpy due to the formation of bonds between
Na20 and SiO2 is lower than the energy spent for the
breaking of polymeric structures in silica. In basic
melts, most of the polymeric structures are disrupted
and the gain in enthalpy due to the formation of bonds
between Na20 and SiO2 prevails. This composition
dependence of a(SiO 2) correlates well with the general
trends established for silicate melts [ 1-6]. For example,
the intersection of the a(SiO2) isotherms lies at
x(CaO) = 0.6 for CaO-SiO2 melts [4] and at x(MnO) =
0.675 for MnO-SiO2 melts [1, 4], reflecting the
changes in basicity and interaction with SiO 2 in the
series Na20, CaO, and MnO.
The present and earlier data on the SiO2 activity in
Na20-SiO2 melts (Fig. 7) were obtained by integrating
the Gibbs-Duhem equation. It can be seen that our
a(SiO2) data agree well with those derived from emf
measurements [25, 30] and mixed-phase equilibrium
studies [34]. The slight discrepancy between our
a(SiO2) values and the results reported in [24, 36] can
be explained by the difference in the lower limits of

INORGANIC MATERIALS Vol. 36 No. 6 2000

536 ZAITSEV et al.

0 integration, that is, in the points on the line of silica-sat-

9 urated melts. It will be shown below that the line of sil-
~
9 Q ica-saturated melts calculated in this work from activity
0
data is in excellent agreement with the results obtained
by physicochemical methods [13, 14].
The composition dependences of the total thermo-
-1 u 0 dynamic functions of formation of Na20-SiO2 melts
are displayed in Fig. 8 in comparison with earlier data.
The curves for the Gibbs energy and enthalpy of forma-
tion are seen to be asymmetric: the extrema are shifted
O~
9 m to Na20-rich compositions. The formation of high-sil-
A
ica melts is accompanied by an entropy gain, whereas
~-2 the entropy of formation of low-silica melts is larger in
magnitude and negative. These findings correlate well
with the structural changes occurring in Na20-SiO2
melts with increasing Na20 concentration. Our data on
the composition dependences of the Gibbs energy,
enthalpy, and entropy of formation of Na20-SiO2 melts
-3 are in perfect agreement with earlier results obtained by
e2 9
emf measurements [24] and in mixed-phase equilib-
03 rium studies [33]. Knudsen cell mass spectrometric
e4 9
05 measurements [36] yield smaller magnitudes of Gibbs
06 / energy and larger magnitudes of entropy. The origin of
~7 this discrepancy is probably the same as in the case of
, , , , , , , , Na20 activities. The Gibbs energies of Na20-SiO2
0.1 0.2 0.3 0.4 0.5 0.6 x(Na20) melts derived from emf measurements with a solid
Fig, 7. SiO2 activity in Na20-SiO2 melts with respect to electrolyte [29] are in excellent agreement with the
pure liquid silica. This work: (/) 1573 K, (I/) 1473 K, present results, whereas the enthalpies of formation are
(Ill) 1373 K; earlierdata: (IV) 1473 K [34], (V) 1573K 133], substantially smaller in magnitude. Similar distinctions
(V/) 1673 K [33], (1) 1073 K [24], (2) 1473 K [24], are observed in the case of the partial thermodynamic
(3) 1473K [25], (4) 1573 K [30], (5) 1293 K [36], functions of Na20 deduced from mixed-phase equilib-
(6) 1423 K 136], (7) 1323 K [29].
rium studies [34]: the partial Gibbs energies agree with
our data, whereas the partial enthaipies and entropies
160 differ markedly from those obtained in this work
.1
*2 (Table 6). Moreover, the composition variations of par-
tial entropy are also different: According to our data,
x 4 IV~"
9 5 / the partial entropy of Na20 changes from positive to
120 o 6 ./ ~, negative with increasing x(Na20), in accordance with
"6 the reported structural changes in the melts. By con-
trast, according to Tsukihashi and Sano [34], AS(Na20)
varies in the opposite way. The possible origin of this
~'~ 80 discrepancy is that, when measurements are made in a
<1 narrow temperature range, both the emf and mixed-
I
t5 'd, phase equilibrium studies ensure a relatively high accu-
racy in Gibbs energy, which is derived directly from the
I
40 o,;, o o ~\\ measured quantities, whereas the accuracies in
enthalpy and entropy, determined by the temperature
0
\ variation of Gibbs energy, are substantially lower.

I I I I I I I I I

0.2 0.4 0.6 0.8x(Na2O) THERMODYNAMIC MODELING

It appears interesting to develop a thermodynamic
Fig. 8. Composition dependencesof the total thermody- model taking into account the detailed structure of
namic functions of formation of Na20-SiO2 melts. This
work: (IV)-AfH, 1473 K; (V) - AfG, 1473 K. Earlierdata: Na20-SiO2 melts. Such a model would be of practical
(l) -AfH, 1300 K [31]; (l) -AfG, 1293 K 131]; (2) -AfG, importance--in analyzing phase equilibria and reac-
1423 K [31]; (ll)-AfG, 1573 K [28]; (Ill) -AfG, 1673 K tions involving Na20-SiO2 melts and predicting the
[28l; (3) -Aft7, 1473 K [18]; (4) -AfH, 1473 K [18]; properties of related materials under various condi-
(5) -AfG, 1323 K [24]; (6) -AfH, 1323 K [24]. tions-and of theoretical interest--in gaining further

INORGANIC MATERIALS Voi. 36 No. 6 2000

 THERMODYNAMICS OF Na20-SiO 2 MELTS 537

Table 5. Comparison of the Na20 activities in Na20-SiO 2 based on ionic concepts of melt structure failed to ade-
melts determined in this work with earlier data [39]; standard quately account for the composition and temperature
state, pure liquid Na20 dependences of Na20 and SiO 2 activities.'
-log[a(Na20)] In our previous works [1, 2, 4, 5, 6], relying on the
x(Na20 ) T, K associated solution model, we developed a thermody-
this work [39]
namic theory of silica polymerization, which offered
0.103" 1430 9.394 9.823 the first quantitative approach to treating this effect in
1473 9.158 9.770 inorganic systems of different nature. This theory takes
into account the formation of polymeric molecules with
0.254 1430 9.094 9.572 arbitrary dimensions and configurations and leans on
1473 8.827 9.515 the only assumption that the equilibrium constant of
1515 8.580 8.981 attachment of Si-O groups is independent on the size
and configuration of the polymeric molecule. Detailed
1557 8.348 8.790
studies of the CaO-SiO 2, MnO-SiO2, and CaF2-SiO 2-
0.303 1344 9.261 9.090 CaO systems [1, 2, 4] demonstrate that the physico-
1387 8.966 8.859 chemical properties of simple silicate melts are gov-
1430 8.689 8.59 erned by the competition between SiO 2 polymerization
and the formation of MO. SiO 2 and 2MO. SiO 2 com-
1473 8.429 8.286
plexes. The presence of these complexes in MnO-SiO 2
1515 8.188 8.065 melts is evidenced by small-angle x-ray scattering data
0.402 1255 8.674 8.143 [41, 42]. Calculations show that SiO2 polymerization is
1344 8.062 7.756 characterized by a large enthalpy, which is greater in
1387 7.794 7.596 magnitude than the enthalpy of formation of MnO 9
SiO2 complexes and is comparable to that of CaO 9
1430 7.543 7.374 SiO2. This indicates that the energy of SiO2-SiO2
1473 7.305 7.392 bonds in polymeric structures is higher than that of
0.500 1255 7.080** 6.558 MnO-SiO 2 bonds. Thus, the formation of basic oxide-
1300 6.784** 6.236 SiO2 melts is due in large measure to the gain in entropy
1344 6.512"* 5.841 associated with the breakdown of polymeric Si-O
structures. As shown above, the formation of high-sil-
1387 6.264 5.641 ica Na20-SiO2 melts also results in entropy gain.
1430 6.030 5.576
When applying the model in question to the
1473 5.810 5.429 Na20-SiO 2 system, account should be taken of the
* Meit-tridymiteequilibrium. extremely strong interaction between the constituent
** Determinedby extrapolatingmeltpropertiesto the stabilityfield oxides in sodium silicates (Tables 1--4). As shown previ-
of sodiummetasilicate. ously [5, 6], the introduction of alumina strengthens
interactions in CaF2-CaO-SiO 2 and CaO-SiO 2 melts,
resulting in the attachment of Si-O polymeric struc-
insight into the structure of matter and the nature of the tures to associates. Given that the interactions in Na20-
chemical bond. Previous attempts to construct a thermo- SiO 2 melts are even stronger than those in the CaO-
dynamic model for liquid sodium silicates with consid- A1203-SIO2 and CaF2-CaO-Al203-SiO 2 systems, it is
eration for silica polymerization were unsuccessful. The reasonable to assume that NS associates in Na20-SiO 2
models for Na20-SiO 2 melts proposed in [30, 33, 40] melts may combine with polymeric Si-O groups to

Table 6. Comparison of the 1473-K partial thermodynamic functions of Na20 in Na20-SiO 2 melts determined in this work
with earlier data 34]; standard state, pure liquid Na20
-AH(Na20), kJ/tool AS(Na20), kJ/(mol K)
x(Na20)
this work [34] this work [34]
0.4 223.2• 279.5• 11.2• 38.2•
0.45 216.5• 225.9• -18.9• ~4.3•
0.5 206.3• 183.8• ~8.8• -19.7~0.6
0.55 197.9• 141.7• ~9.2• -1.1•

INORGANIC MATERIALS Vol. 36 No. 6 2000

538 ZAITSEV et al.
form NSj (2 < j < 8) complexes. Besides, of all NSj
groups in Na20-SiO 2 melts, the NS2 complex must
play a special part since solid sodium disilicate melts
congruently. In view of this, NS2 must be regarded as a
distinct constituent of associated solutions.
Based on the above, earlier findings [1, 2, 4-6], and
the stability of intermediate phases, it seems reasonable
to assume that Na20-SiO2 melts contain N2S, NS, and
NS 2 associates and Sk (2 _<k < ~) and NSj (3 _<j < ~)
polymeric molecules. Further, we take, as in previous
works [ l, 2, 4-6], that the equilibrium constants of the
attachment of Si-O groups are independent of the size
and configuration of polymeric complexes:
Kp = K(Sk) = K(NSj). (11)
The major formulas of the model in question were
derived in [1, 2, 4-6]. Here, it will suffice to mention
that, in the associated solution model, the activity of a
component is calculated as the product of the molar
fraction of monomers and the activity coefficient,
which can be determined from the excess Gibbs energy
by a standard procedure,
AfGe = NT' l... [n(N)]i[n(S)]j (12)
.~. -t.l n(i+j_l) ' It follows from parameters (13) and earlier results
I, J
where n(N) and n(S) are the numbers of moles of the
starting components, n = n(N) + n(S) is the total num-
ber of moles of the starting components, and Li, j a r e
some functions of temperature.
As in previous studies, the parameters of the associ-
ation and polymerization reactions and the number of
terms and the parameters in the expression for the
excess Gibbs energy were determined by a least
squares fitting procedure, minimizing the sum of the
squares of the differences in the calculated and mea-
sured component activities, [(aex p - acalc)[aexp]2, over the
entire set of experimental data. Although the thermody-
namic parameters of SiO2 polymerization were deter-
mined earlier [l, 2, 4-6], they were considered
unknown at the first stage of calculations. As a result,
we found that the enthalpy and entropy of SiO2 self-
association, AHp -- -92370 J/tool and ASp =
-27.1 J/(mol K), differ insignificantly from the values
found earlier. In view of this, in subsequent calcula-
tions, we used the values of AHp and ASp obtained in [l,
2, 4-6]. This did not lead to notable changes in other
model parameters. The calculations showed that an
adequate description of the thermodynamic properties
in the melt system under consideration could only be
achieved with the fitting parameters
AHp = -92575 J/mol,
ASp = -27.2 J/(mol K);
AfH(NS) = -322830 J/mol,
AfS(NS) = -32.5 J/(mol K);
AfH(N2S) = -520580 J/mol, (13)
AfS(N2S) = -81.3 J/(mol K);
AfH(NS2) = -433220 J/mol,
AfS(NS2) = -55.5 J/(mol K);
Lll = 24920J, L22 = - 1 1 9 0 4 0 J .
With these parameters, the composition and tempera-
ture dependences of component activities can be
described to within the experimental accuracy, as illus-
trated in Table 4. The discrepancy between the mea-
sured and calculated activities is insignificant in all
cases and does not exceed 2-3%.
It should be emphasized that the thermodynamic
parameters of silica polymerization in Na20-SiO2
coincide with those found earlier in the silicate and alu-
minosilicate systems CaO-SiO2 [4], MnO-SiO2 [1, 4],
CaO-A1203-SiO2 [5], CaF2-CaO-SiO 2 [2], and C a F 2-
CaO-AI203-SiO 2 [5]. In this aspect, our approach dif-
fers from the models based on the ionic concepts of
oxide melt structure [43-47], in which the parameters
of one and the same process vary from system to
system.
[1, 2, 4-6] that the formation of complexes in Na20-
SiO2 melts is accompanied by substantially larger gains
in enthalpy and Gibbs energy than that in the systems
CaO-SiO2 and, especially, MnO-SiO2. This observa-
tion is consistent with the known scale of oxide basic-
ity. However, in spite of this distinction, the three sys-
tems in question are closely similar in the character of
the interaction between components, which highlights
once again that the model under consideration is of
general nature. The stability of the equimolar complex
with respect to the orthosilicate group is higher in
CaO-SiO2 melts than it is in MnO-SiO2 [1,4]. In going
to soda-silica melts, the difference becomes even
larger. Thus, the rise in the stability of orthosilicate
associates in silicate melts with increasing basicity of
the second component appears to be a general trend. It
follows from (13) and our earlier data [1, 2, 4-6] that
the parameters describing the excess Gibbs energy of
formation of associated Na20-SiO2, CaO-SiO2, and
MnO-SiO2 solutions differ markedly, even though
these solutions are described by AfG of the same form.
This suggests that the interactions between the compo-
nents in these systems are of the same nature--acid-
base interactions of the basic oxides Na20, CaO, and
MnO with the acid oxide SIO2.
In addition to the above-mentioned studies [30, 33,
40], thermodynamic modeling of Na20-SiO2 melts
was carried out in [18, 24]. Shakhmatkin and Shul'ts
[24] treated the composition dependence of Na20
activity in terms of two approaches. In one of them,
they used the ideas of the associated solution model.
They considered two concurrent equilibria in the melt,
INORGANICMATERIALS Vol. 36 No. 6 2000
 THERMODYNAMICS OF Na20-SiO 2 MELTS 539

Na20 + SiO2 = Na20. SiO2 and Na20 + 2SIO2 = Na20.
2SIO2, and assumed that all of the Na20 is incorporated
in complexes. As a result, they obtained quantitative
agreement with experiment. The other approach was
based on analysis of different states of oxygen, O ~ +
02- = 20-, and ensured only qualitative agreement with
experiment. Wu et al. [18] optimized the thermody-
namic and phase-diagram data for the Na20-SiO 2 sys-
tem by fitting the temperature and composition depen-
dences of thermodynamic functions to a modified
"structureless" quasi-chemical model [48], which rests
on the assumptions that cations occupy only cation
sites and that the composition dependences of the con-
tributions from ordering processes to the enthalpy and
nonconfigurational entropy can be described by poly-
nomials. Their results were in reasonable agreement
with the experimental data on the composition and tern-
perature dependences of Na20 activity. However,
although much attention was given to phase-diagram
data in determining the model parameters, the calcu-
lated lines of phase equilibria were in poor agreement
with experiment, especially at high SiO2 concentra-
tions. According to Wu et al. [18], the uncertainty in the
calculated temperatures of phase equilibria was +10 K
on the whole and :1:50 K for x(SiO2) > 0.75. Moreover,
the metastable miscibility gap calculated in [18] proved
notably narrower than follows from experimental data,
even though the best-fit polynomials included terms up
to the seventh order.
PHASE DIAGRAM
The above melt model and the thermodynamic func-
tions of formation of all the intermediate solid phases

T,K
1996

1900

1800 N2S N382 NS NS 2 N3S8

; ;
1700

1600

1500

405 1397(1397)
1400

1300 tridymite + L
t2 ) 0.442

1200
- [ + 1148(1148) L 14__3
i

1100 ~-Na20+ N2S Z 0.630(0.630)l.~=~=.X

1quartz+L~
'l'

0.749(0.742)
023
N3S 8 +
1000
N2S + NS [ NS + NS 2
),-Na20 + N2S
NS 2 + quartz ~[
-t,,,,J

Na20 0.2 0.4 0.6 0.8 SiO2

mole fraction
Fig. 9. Na20-SiO2 phase diagramcalculatedin this work(solid lines).The open circlesshowthe data obtainedby Kracek[13].
Earlierdata on the positionsof invariantpoints [13-17] are givenin parentheses.

INORGANIC MATERIALS Vol. 36 No. 6 2000

540 ZAITSEV et al.
T,K
1150
~ % ,1
o2
., + + +3
1100 o4
94- + O 5
1050 ~6
1000
950
900
850
800 I x-ray diffraction study [17], the stoichiometry of this
0 0.05 0.10 0.15 0.20
x(Na20)
Fig. 10. Metastableliquidmiscibilitygap in the Na20-SiO2
system; the curve represents calculation results, and the
points showexperimentaldata: (I) [49], (2) [50], (3) [51],
(4) 153l,(5) [52], (6) [54].
were used to calculate phase equilibria in the Na20-
SiO2 system. In calculations, relying on fundamental
thermodynamic laws, we solved, using a standard pro-
cedure, the system of equations obtained by equating
the chemical potentials of components in coexisting
phases. We considered all possible phase relations in
the Na20-SiO2 system. The phase assembly resulting
in a minimum in the Gibbs energy was considered to be
an equilibrium state of the system; otherwise, phase
relations were considered metastable and were not
included in the phase diagram. The Gibbs energies of
the phase transformations in Na20 and SiO2 were taken
from [7, 23]. The calculated Na20-SiO 2 phase diagram
is displayed in Fig. 9 in comparison with the available
experimental data on the line of SiO2-saturated melts
[ 13, 14] and invariant points [ 13-17].
Earlier, the phase equilibria in the Na20-SiO 2 sys-
tem were studied by Kracek [13, 14], who analyzed
heating curves of quenched materials. According to his
results, NS and NS2 melt congruently at 1363 and
1148 K, respectively. The positions of the eutectic
points are x(SiO2) = 0.630, 1120 K and x(SiO2) = 0.742,
1057 K. These data are in excellent agreement with
those reported in [15], where NS and NS 2 were found
to melt congruently at 1363 and 1148 K, respectively,
and the eutectic points were located at x(SiO2) = 0.615,
1114 K and x(SiO2) = 0.742, 1057 K. According to
Kracek [13], N2S decomposes peritectically at 1392 K
and forms a eutectic with NS at x(SiO2) = 0.438, with
the melting point at 1296 K. According to recent data,
the melting point of Na20 is 1405 K [7], and no stable
compounds exist in the composition range between
Na20 and N2S. In view of this, incongruent melting of
sodium orthosilicate appears unlikely, as also sug-
gested by the data in Table 2 and Fig. 4, according to
which sodium orthosilicate is the most stable com-
pound in the system under consideration and, hence,
must melt congruently. This inference is supported by
the experimental data obtained by d'Ans and Loeffler
[15] and Loeffler [16], who reported N2S to melt con-
gruently at 1358 K. They also identified a new com-
pound, N3S2 (Tm = 1397 K), which was found to melt
congruently and form eutectics with N2S (x(SiO2) =
0.361, 1275 K) and NS (x(SiOz) = 0.455, 1289 K).
According to Budnikov and Matveev [19], the system
also contains a silica-rich compound of tentative sto-
ichiometry NS 3. In a later, and probably more careful,
compound was shown to be N3S8. It melts incongru-
ently at 1082 K, yielding quartz and a liquid phase and
decomposes at 973 K by a peritectoid reaction into
quartz and NS2. These findings on the melting behavior
of N3S8are in conflict with earlier data [ 13, 15], accord-
ing to which the eutectic point on the silica-rich side
lies beyond the N3S8 stoichiometry and, therefore, the
peritectic decomposition of N3S8 into quartz and liquid
is unlikely. The present results reconcile this conflict by
showing that the incongruent melting of N3S8 yields
sodium disilicate and liquid. It seems likely that the
controversy in question stems from the narrow temper-
ature stability range of N3S8, which complicates exper-
imental studies of its melting behavior.
There is also ample evidence [49-55] that, in the
subsolidus region of the Na20-SiO 2 phase diagram,
there is a liquid miscibility gap. The boundary of the
immiscibility region calculated in this work is shown in
Fig. 10, together with the experimental data obtained
by determining the clarification and maximum opales-
cence temperatures and by small-angle x-ray scattering
[49-54]. The calculation results are seen to agree rea-
sonably well with experiment. It is of interest to note
that Hailer et al. [49] achieved good agreement with
experimental data on the metastable liquid miscibility
gap in the Na20-SiO2 system using a usual type of the
regular solution model and considering Na20 9 3SIO2
and (SiO2)8 to be immiscible liquids, i.e., resorting to
ideas of the associated solution model. Phase-diagram
data [1, 4, 56] show that the temperature and composi-
tion ranges of liquid immiscibility broaden in the
sequence Na20-SiO2, CaO-SiO2, MnO-SiO2, and
FeO-SiO 2. In soda-silica glasses, liquid immiscibility
is metastable. According to Kracek et al. [57], the
KEO-SiO2 system contains no liquid miscibility gap at
all. Thus, the presence and extent of the miscibility gap
in silicate melts are related directly to silica polymer-
INORGANICMATERIALS Vol. 36 No. 6 2000
 THERMODYNAMICS OF Na20--SiO2 MELTS
ization and the basicity of the second component,
which determines the strength of its interaction with
SiO2. These relations can be described qualitatively
and quantitatively using the model presented in this
work and the proper characteristics of complexation.
It follows from the above observations and Figs. 9
and 10 that our phase-diagram calculations are in excel-
lent agreement with available physicochemical data for
the Na20-SiO2 system. Note once again the agreement
with the positions of phase boundaries derived from the
activity isotherms of the melt. The major advantages of
the thermodynamic approach used in this work is that
good agreement with experimental phase-diagram data
is achieved by using only Gibbs energies of melting of
the constituent oxides, without resorting to additional
data, and that the detailed melt structure and its changes
with melt composition are taken into account.
CONCLUSION
The thermodynamic properties of all the solid and
liquid phases in the Na20-SiO2 system were deter-
mined by Knudsen cell mass spectrometry.
The formation of liquid and solid sodium silicates is
characterized by extremely low enthalpies and Gibbs
energies, which reach extrema in sodium orthosilicate.
A thermodynamic model for liquid sodium silicates
is proposed which relies on associated solution theory
and considers the formation of Si-O groups with arbi-
trary dimensions and configurations. The model
describes the composition and temperature depen-
dences of the thermodynamic functions of Na20--SiO2
melts with an accuracy no worse than the experimental
error.
The phase equilibria in the Na20-SiO2 system were
assessed using thermodynamic calculations. The
results agree well with the experimental data obtained
by physicochemical methods.
ACKNOWLEDGMENTS
We are grateful to A.D. Litvina and I.A. Lipgart for
their assistance in performing the experimental work
and preparing the manuscript.
This work was supported by the Russian Foundation
for Basic Research, grant no. 99-03-32616.
REFERENCES
1. Zaitsev, A.I. and Mogutnov, B.M., Thermodynamic
Properties and Phase Equilibria in the MnO--SiO2 Sys-
tem, Z Mater. Chem., 1995, vol. 5, pp. 1063-1073.
2. Zaitsev, A.I., Litvina, A.D., and Mogutnov, B.M., Ther-
modynamic Properties of CaF2-SiO2-CaO Melts,
Neorg. Mater., 1997, vol. 33, no. 1, pp. 76-86 [lnorg.
Mater. (Engl. Transl.), vol. 33, no. 1, pp. 68-77].
INORGANIC MATERIALS Vol. 36 No. 6 2000
541
3. Zaitsev, A.I., Litvina, A.D., and Mogutnov, B.M., Ther-
modynamic Properties and Phase Equilibria in the CaF2-
SiO2-AI203--CaO System: I. Experimental Study of
CaF2-SiO2-AI2Oa--CaO Melts, Neorg. Mater., 1997,
vol. 33, no. 12, pp. 1489-1498 [lnorg. Mater. (Engl.
Transl.), vol. 33, no. 12, pp. 1265-1273].
4. Zaitsev, A.I. and Mogutnov, B.M., Thermodynamics of
CaO-SiO2 and MnO-SiO2 Melts: I. Experimental Study,
Neorg. Mater., 1997, vol. 33, no. 7, pp. 839-847 [lnorg.
Mater. (Engl. Transl.), vol. 33, no. 7, pp. 707-714].
5. Zaitsev, A.I., Litvina, A.D., Lyakishev,N.P., and Mogut-
nov, B.M., Thermodynamics of CaO-AI203-SiO 2 and
CaF2-CaO-AI203-SiO 2 Melts, J. Chem. Soc., Faraday
Trans., 1997, vol. 93, no. 17, pp. 3089-3098.
6. Zaitsev, A.I. and Mogutnov, B.M., A General Approach
to Thermodynamics of High-Temperature Liquid Solu-
tions, High Temp. Mater. Sci., 1995, vol. 34, nos. 1-3,
pp. 155-171.
7. Gurvich, L.V., IVTANTERMO: An Automated Data-
base of the Thermodynamic Properties of Substances,
Vestn. Akad. Nauk SSSR, 1983, no. 3, pp. 54-65.
8. Zaitsev, A.I., Korolyov, N.V., and Mogutnov, B.M., The
Vapor Pressures and the Heats of Sublimation of CaF2
and SrF2, High Temp. Sci., 1990, vol. 28, pp. 341-350.
9. Zaitsev, A.I., Korolev,N.V., and Mogutnov, B.M., Vapor
Pressures of CaF2 and SrF2, Teploflz. Vys. Temp., 1989,
vol. 27, no. 3, pp. 465-471.
10. Hildenbrand, D.L. and Murad, E., J. Chem. Phys., 1970,
vol. 53, p. 3403.
11. Lamoreaux, R.H. and Hildenbrand, D.L., J. Phys. Chem.
Ref. Data, 1984, vol. 13, p. 151.
12. Steinberg, M. and Schofield, K., A Reevaluation of the
Vaporization Behavior of Sodium Oxide and the Bond
Strengths of NaO and Na20: Implications for the Mass
Spectrometric Analyses of Alkali/Oxygen Systems,
J. Chem. Phys., 1991, vol. 94, no. 5, pp. 3901-3907.
13. Kracek, EC., The System Sodium Oxide-Silica, J. Phys.
Chem., 1930, vol. 34, pp. 1583-1598.
14. Kracek, EC., Phase Equilibrium Relations in the System
Na2SiO3-Li2SiO3-SiO2, J. Am. Chem. Soc., 1939,
vol. 61, pp. 2863-2877.
15. D'Ans, J. and Loeffler, J., Untersuchungen im System
Na20-SiO2-ZrO 2, Z. Anorg. Allg. Chem., 1930,
vol. 191, pp. 1-34.
16. Loeffler, J., Uber den alkalischen Teil des System Na20-
SiO2, Glastech. Ber., 1969, vol. 42, no. 3, pp. 92-96.
17. Williamsom, J. and Glasser, EE, Phase Relations in the
System NaESi2Os-SiO2, Science (Washington, D.C.,
1883-), 1965, vol. 148, no. 6, pp. 1589-1591.
18. Wu, E, Eriksson, G., and Pelton, A.D., Optimization of
the Thermodynamic Properties and Phase Diagrams of
the Na20-SiO2 and K20-SiO2 Systems, Z Am. Ceram.
Soc., 1993, vol. 76, no. 3, pp. 2059-2064.
19. Budnikov, EE and Matveev, M.A., Synthesis and Prop-
erties of Crystalline Na20 9 3SiO2, Dokl. Akad. Nauk
SSSR, 1956, vol. 107, pp. 547-550.
542 ZAITSEV et al.
20. Berman, R.G. and Brown, H.T., The Heat Capacity of
Minerals in the System Na20-K20--CaO-MgO-FeO -
Fe203-A1203-SiO2-TiO2-H20-CO: Representation, Esti-
mation, and High Temperature Extrapolation, Contrib.
Mineral Petrol., 1985, voi. 89, pp. 168-183.
21. Berman, R.G. and Brown, H.T., Erratum Heat Capacity
of Minerals in the System Na20-K20-CaO-MgO-
FeO-Fe203-A1203-SiO2-TiO2-H20--CO: Representa-
tion, Estimation, and High Temperature Extrapolation,
Contrib. Mineral. PetroL, 1986, vol. 94, p. 262.
22. Knacke, O., Kubaschewski, O., and Hesselmann, K.,
Thermochemical Properties of Inorganic Substances,
Berlin: Springer, 1991, 2nd ed.
23. Hillert, M., Sundman, B., and Wang, X., Assessment of
the CaO-SiO2 System, MetalL Trans. B, 1990, vol. 21,
pp. 303-312.
24. Shakhmatkin, B.A. and Shul'ts, M.M., Thermodynamic
Properties of Na20-SiO 2 Glass-Forming Melts in the
Range 800-1200~ Fiz. Khim. Stekla, 1980, vol. 6,
no. 2, pp. 129-135.
25. Frohberg, M.G., Caung, E., and Kapoor, M.L., Measure-
ment of the Activity of the Oxygen Ions in the Liquid
Systems Na20-SiO 2 and K20-SiO 2, Arch. Eisenhutten-
wes., 1973, vol. 44, no. 8, pp. 585-588.
26. Kohsaka, S., Sato, S., and Yokokawa, T.E., Measure-
ments of Molten Oxide Mixtures: III. Sodium Oxide +
Silicon Dioxide, J. Chem. Thermodyn., 1979, vol. 11,
pp. 547-551.
27. Yamaguchi, S., Imai, A., and Goto, K.S., Activity Mea-
surement of Na20 in Na20-SiO 2 Melts Using the Beta-
Alumina as the Solid Electrolyte, Scand. J. MetalL,
1982, vol. 11, pp. 263-264.
28. Ravaine, D., Azandegbe, E., and Souquet, J.L., Mesures
potentiom6triques de chaines 616ctrochimiques com-
prenant des silicates fondus: Interpr6tation des resultats
par un mod/~le statistique, Silic. Ind., 1975, vol. 12,
pp. 333-340.
29. Neudorf, D.A. and Elliott, J.E, Thermodynamic Proper-
ties of Na20-SiO2-CaO Melts at 1100 to 1200~ Met-
all. Trans. B, 1980, vol. 1l, pp. 607-614.
30. Goto, K.S., Yamaguchi, S., and Nagata, K., The Chemi-
cal Activity of Component Oxides in Na20-Based Slags,
Proc. of 2nd Int. Symp. on Metallurgical Slags and
Fluxes, Lake Tahoe, 1984, pp. 467-481.
31. Pearce, M.L., Calculation of Oxygen Ion Activities in
Sodium Silicate and Sodium Borate Melts, J. Am.
Ceram. Soc., 1965, vol. 48, pp. 611--613.
32. Pearce, M.L., Solubility of Carbon Dioxide and Varia-
tion of Oxygen Ion Activity in Soda-Silica Melts, J. Am.
Ceram. Soc., 1964, vol. 47, no. 7, pp. 342-347.
33. Rego, D.N., Sigworth, G.K., and Philbrook, W.O., Ther-
modynamic Study of Na20-SiO 2 Melts at 1300 and
1400~ Metail. Trans. B, 1985, vol. 16, pp. 313-323.
34. Tsukihashi, E and Sano, N., Measurement of the Activ-
ity of Na20 in Na20-SiO 2 Melts by Chemical Equilibra-
tion Method, Tetsu to Hagane, 1985, vol. 71, no. 7,
pp. 815-822.
35. Holmquist, S., Oxygen Ion Activity and the Solubility of
Sulfur Trioxide in Sodium Silicate Melts, Z Am. Ceram.
Soc., 1966, vol. 49, no. 9, pp. 467-473.
36. Shul'ts, M.M., Stolyarova, V.L., and Ivanov, G.G., Mass
Spectrometric Study of the Thermodynamic Properties
of Na20-SiO 2 Melts and Glasses, Fiz. Khim. Stekla,
1987, vol. 13, no. 2, pp. 168-172.
37. Chastel, R., Bergman, C., Rogez, J., and Mathieu, J.C.,
Excess Thermodynamic Functions in Ternary Na20-
K20-SiO2 Melts by Knudsen Cell Mass Spectrometry,
Chem. GeoL, 1987, vol. 62, pp. 19-29.
38. Piacente, V. and Matousek, J., Mass Spectrometric
Determination of Sodium Partial Pressure over the
System Na20-2SiO 2, Silicates, 1973, vol. 4,
pp. 269-280.
39. Rudnyi, E.B., Vovk, O.M., Sidorov, L.N., et aL, Activity
of the Alkali Oxide in Soda-Silica Melts by the Ionic-
Molecular Equilibrium Method, Fiz. Khim. Stekla, 1988,
vol. 14, no. 2, pp. 218-225.
40. Yamaguchi, S. and Goto, K.S., Activity Measurement on
Na20 in Na20-SiOz-P205 Melts at 900-1400~ Using
Beta-Alumina Electrolyte, Scand. J. Metall, 1984,
vol. 13, pp. 129-136.
41. Sokol'skii, V.E., Kazimirov, V.E, and Galinich, V.I.,
X-ray Diffraction Study of MnO-SiO 2 Melts, lzv. Akad.
Nauk SSSR, Neorg. Mater., 1983, vol. 19, no. 4,
pp. 629-633.
42. Sokol'skii, V.E., Galinich, V.I., Kazamirov, V.E, et al.,
Structure of MnO-TiO2-SiO 2 and MnO-ZrO2-SiO2
Melts, Rasplavy, 1987, vol. 1, no. 6, pp. 34 40.
43. Masson, C.R., An Approach to the Problem of Ionic Dis-
tribution in Liquid Silicates, Proc. R. Soc. London, A,
1965, vol. 287, no. 1409, pp. 201-221.
44. Masson, C.R., Smith, I.B., and Whiteway, S.G., Activi-
ties and Ionic Distributions in Liquid Silicates: Applica-
tion of Polymer Theory, Can. J. Chem., 1970, vol. 48,
p. 1456.
45. Gaskell, D.R., Thermodynamic Models of Liquid Sili-
cates, Can. MetalL Q, 1981, vol. 20, no. 1, pp. 3-19.
46. Esin, O.A., The Nature of Molten Metallurgical Slags,
Zh. Vses. Khim. O-va. im. D. I. Mendeleeva, 1974,
vol. 16, no. 5, pp. 504-514.
47. Novikov, V.K., Evolution of the Polymer Model of Sili-
cate Melts, Rasplavy, 1987, vol. 1, no. 6, pp. 21-33.
48. Peiton, A.D. and Blander, M., ThermodynamicAnalysis
of Ordered Solutions by Modified Quasichemical
Approach--Application to Silicate Slags, MetalL
Trans. B, 1986, vol. 17, pp. 807-815.
49. Hailer, W., Blackburn, D.H., and Simmons, J.H., Misci-
bility Gaps in Alkali-Silicate Binaries--Data and Ther-
modynamic Interpretation, J. Am. Ceram. Soc., 1974,
vol. 57, no. 3, pp. 120-126.
50. Moriya, Y., Warfington, D.H., and Douglas, R.W., Meta-
stable Liquid Immiscibility in Some Binary and Ternary
Alkali Silicate Glasses, Phys. Chem. Glasses, 1967,
vol. 8, no. 1, pp. 19-25.
51. Andreev, N.S., Goranov, D.A., Porai-Koshits, E.A., and
Sokolov, Yu.G., Chemically Nonuniform Structure of
Soda-Silica and Lithia-Silica Glasses, in Stekloobraz-
INORGANIC MATERIALS Vol. 36 No. 6 2000
 THERMODYNAMICS OF Na20-SiO 2 MELTS
noe sostoyanie. Vyp. 1: Katalizirovannaya kristalli-
zatsiya stekla (Glassy State. Issue 1: Catalyzed Glass
Crystallization), Leningrad, 1963, pp. 46-53.
52. Andreev, N.S. and Aver'yanov, V.I., Structural Studies of
Soda-Silica Glasses in the Range of Metastable Segre-
gation, Trudy IV vsesoyuznogo soveshchaniya po stek-
loobraznomu sostoyaniyu (Proc. IV All-Union Conf. on
Glassy State), Moscow, 1965, pp. 94-97.
53. Hammel, J.J., Proc. VII Int. Congress on Glass, Charle-
roi, 1966, vol. 1, paper 36.
INORGANIC MATERIALS Vol. 36 No. 6 2000
543
54. Porai-Koshits, E.A. and Averjanov, V.I., Primary and
Secondary Phase Separation of Sodium Silicate Glasses,
J. Non-Cryst. Solids, 1968, vol. 1, no. 1, pp. 29-38.
55. Charles, R.J., Origin of Immiscibility in Silicate Solu-
tions, Phys. Chem. Glasses, 1969, vol. 10, no. 5,
pp. 169-178.
56. Schlackenatlas, Dtisseldorf: Stahleisen, 1981, p. 35.
57. Kracek, EC., Bowen, N.L., and Morrey, G.W., Equilib-
rium Relations and Factors Influencing Their Determi-
nation in the System K2SiO3-SiO2, Z Phys. Chem.,
1937, vol. 41, pp. 1183-1193.