Physico-Chemical Properties of Silicate Melts

Abstract The shear viscosity, density, thermal expansivity and
specific heat capacity are important factors controlling the
morphology, rheology, and texture of volcanic flows and deposits.
These physical properties of silicate melts largely depend on
chemical composition, water content, crystal content, bubble
content and stress applied to the melt. Recently, it has been
recognized that the applied stress plays an important role in the
so called “glass transition” area of silicate melts. This kinetic
boundary between brittle and ductile behavior affects the eruptive
style. Thorough knowledge of the physical processes that occur at
this brittle/ductile transition can affect the decision making of
governments during volcanic crises and help to reduce and/or avoid
loss of life and assets. Scientific knowledge from this research
can be directly applied to the geomaterial industry. In addition,
natural magmatic rocks are the major raw material in the production
of microfibres and continuous fibres. Compared to normal glass
fibres, rock fibres have a remarkable high temperature endurance,
acid and alkali resistance and anti-heat impact. Rock products can
be used as substitutes for metal and timber. They are likely to
become more widely used in the near future. Further use for natural
magmatic rocks include crushed stone, concrete aggregate, railroad
ballast, production of high quality textile fibres, floor tiles,
acid-resistant equipment for heavy industrial use, rockwool, basalt
pipers, basalt reinforcement bars, basalt fibre roofing felt
(ruberoid), basalt laminate (used as a protective coating),
heat-insulating basalt fibre materials and glass wool (fibre
glass). Since Bottinga and Weill (1970) first suggested that the
density of melts in two or three component systems could be used to
determine partial molar volumes of oxide components in silicate
liquids, several models based upon this approach have been proposed
in the Earth sciences literature. Considering that knowledge the
densities of 8 Zn-bearing silicate melts have been determined, in
equilibrium with air, in the temperature range of 1363 to 1850 K.
The compositional joins investigated [sodium disilicate (NS2)- ZnO;
anorthite-diopside 1 atm eutectic (AnDi)-ZnO; and
diopside-petedunnite] were chosen based on the pre-existing
experimental density data set, on their petrological relevance, and
in order to provide a test for significant compositionally induced
variations in the structural role of ZnO. The ZnO concentrations
investigated range up to 25 mol% for sodium disilicate, 20 mol% for
the anorthite-diopside 1 atm eutectic, and 25 mol% for petedunnite.
Molar volumes and expansivities have been derived for all melts.
The molar volumes of the liquids decrease with increasing ZnO
content. The partial molar volume of ZnO derived from the
volumetric measurements for each binary system is the same within
error. A multicomponent fit to the volumetric data for all
compositions yields a value of 13.59(0.55) cm3/mol at 1500 K. I
find, no volumetric evidence for compositionally induced
coordination number variations for ZnO in alkali-bearing vs.
alkali-free silicate melts nor for Al-free vs. Al-bearing silicate
melts. The partial molar volume of ZnO determined here may be
incorporated into existing multicomponent models for the prediction
of silicate melt volume. High temperature density determinations on
ZnO-bearing silicate melts indicate that a single value for the
partial molar volume of ZnO is sufficient to describe the
volumetric properties of this component in silicate melts. The
presence of alkalies and Al does not appear to influence the
partial molar volume of ZnO within the temperature range
investigated here. There is no volumetric evidence across this
temperature range presented for composition to influence the
coordination polyhedron of ZnO in silicate melts. The next physical
property to be studied was thermal expansivity. Ten compositions
from within the anorthite-wollastonite-gehlenite (An-Wo-Geh)
compatibility triangle were investigated. Due to the lack of
information about the thermal expansivities at supercooled liquid
temperatures this study focused on the measurement of thermal
expansivity using a combination of calorimetric and dilatometric
methods. The volumes at room temperature were derived from
densities measured using the Archimedean buoyancy method. For each
sample density was measured at 298 K using glass that had a
cooling-heating history of 10-10 K min-1. The thermal expansion
coefficient of the glass from 298 K to the glass transition
interval was measured by a dilatometer and the heat capacity was
measured using a differential scanning calorimeter from 298 to 1135
K. The thermal expansion coefficient and the heat flow were
determined at a heating rate of 10 K min-1 on glasses that were
previously cooled at 10 K min-1. Supercooled liquid density, molar
volume and molar thermal expansivities were indirectly determined
by combining differential scanning calorimetric and dilatometric
measurements assuming that the kinetics of enthalpy and shear
relaxation are equivalent. The data obtained on the supercooled
liquids were compared to high-temperature predictions from the
models of Lange and Carmichael (1987), Courtial and Dingwell (1995)
and Lange (1997). The best linear fit combines the supercooled
liquid data presented in this study and the high temperature data
calculated using the Courtial and Dingwell (1995) model. This
dilatometric/calorimetric method of determining supercooled liquid
molar thermal expansivity greatly increases the temperature range
accessible for thermal expansion. It represents a substantial
increase in precision and understanding of the thermodynamics of
calcium aluminosilicate melts. This enhanced precision demonstrates
clearly the temperature independence of the melt expansions in the
An-Wo-Geh system. This contrasts strongly with observations for
neighboring system such as Anorthite-Diopside and raises the
question of the compositional/structural origins of the temperature
dependence of thermal expansivity in multicomponent silicate melts.
In addition, the partial molar volumes and the thermal
expansivities of 10 samples from within the An-Wo-Geh compatibility
triangle have been determined. They have been incorporated into
existing multicomponent models in order to predict silicate melt
volume. The resulting supercooled liquid volumes near glass
transition temperatures (1135 - 1200 K) and at superliquidus
temperature were combined to yield temperature independent thermal
expansivities over the entire temperature range. In light of
results presented in this study, together with the published data,
it seems that binary and ternary systems have temperature
independent thermal expansivities from the supercooled liquid to
the superliquidus temperature at 1 atmosphere. By combining the
high temperature densitometry data (i.e., above liquidus) from the
literature with volume and expansivity data obtained at Tsc, a wide
temperature range is covered. There is no volumetric evidence
across this temperature range for temperature independent thermal
expansivities in the An-Wo-Geh compatibility triangle. Furthemore,
the thermal expansivities of three multicomponent glasses and
liquids have been obtained over a large temperature interval (298 -
1803 K) which involved combining the results of low and high
temperature measurements. The sample compositions investigated were
derived from three natural lavas; Vesuvius 1631 eruption, Etna 1992
eruption and an Oligocene-Miocene lava flow from Slapany in the
Bohemian massif. The original rocks are tephri-phonolite,
trachybasalt and basanite, respectively. This is the first time
this calorimetric/dilatometric method has ever been applied to
natural magmatic melts. The low temperature volumes were derived
from measurements of the glass density of each sample after cooling
at 5 K.min-1 at 298 K, followed by measurements of the glass
thermal expansion coefficient from 298 K to the samples´ respective
glass transition interval. Supercooled liquid volumes and molar
thermal expansivities were determined by combining scanning
calorimetric and dilatometric measurements, assuming that the
kinetics of enthalpy and shear relaxation are equivalent (Webb,
1992). High temperature densities were measured using Pt double bob
Archimedean densitometry. In addition, the oxidation state of iron
was analyzed using a wet chemistry method. Small amounts of samples
were taken from the liquids using a “dip” technique at regular
temperature steps during high temperature densitometry. The
measured high temperature densities have been compared with
predicted densities across the same temperature interval calculated
using the multicomponent density models of Lange and Carmichael
(1987) and Lange (1997). The resulting data for liquid volumes near
glass transition temperatures (993 - 1010 K) and at super-liquidus
temperatures (1512 - 1803 K) are combined to yield temperature
dependant thermal expansivities over the entire supercooled and
stable liquid range. These results confirm the observation of
Knoche et al. (1992a); Knoche et al. (1992b); Toplis and Richet
(2000); Liu and Lange (2001); Gottsmann and Dingwell (2002) of the
temperature dependence of thermal expansivity. The molar volumes
indicate, in general, a significant negative temperature dependence
of the expansivity. The thermal molar expansivity of the glasses
increase from SiO2-poor (basalt-basanite composition) to relatively
SiO2-rich melts (tephri-phonolite composition). The thermal molar
expansivity at supercooled liquid temperatures increases in the
same manner as the glasses. In contrast, the thermal molar
expansivity of the superliquidus liquid decrease from SiO2-poor to
relatively SiO2-rich melts. Non-linear dependency of molar volume
has been observed for all studied samples above the glass
transition area. Molar volume from just above the glass transition
area to about 1873 K can be predicted by a non-linear logarithmic
curve. This study examined the expansivities and molar volumes of
relatively basic compositions. Extending such a study to more
SiO2-rich, but still geologically relevant, compositions remains a
challenge, because the high viscosities of such melts preclude the
use of immersion techniques. This problem can be solved using a
high temperature densitometry where the volume is measured on
levitated sample. I would like to urge studies of this sort in the
future. Results from such studies should provide important
information regarding a number of geological processes, which occur
in such extremely high viscous liquids. A new viscosity measurement
for melts spanning a wide range of anhydrous compositions
including: rhyolite, trachyte, moldavite, andesite, latite,
pantellerite, basalt and basanite are discussed in the last
chapters. Micropenetration and concentric cylinder viscometry
measurements cover a viscosity range of 10-1 to 1012 Pas and a
temperature range from 973 to 1923 K. These new measurements,
combined with other published data, provide a high-quality database
comprising ~800 experimental data on 44 well -characterized melt
compositions. This database is used to recalibrate the model
proposed by Giordano and Dingwell [Giordano, D., Dingwell, D. B.,
2003a. Non-Arrhenian multicomponent melt viscosity: a model. Earth
Planet. Sci. Lett. 208, 337–349] for predicting the viscosity of
natural silicate melts. The recalibration shows that: a) the
viscosity (η)–temperature relationship of natural silicate liquids
is very well represented by the VFT equation [log η=A+B/ (T−C)]
over the full range of viscosity considered here, b) the use of a
constant high-T limiting value of melt viscosity (e.g., A) is fully
consistent with the experimental data. There are 3 different
compositional suites (peralkaline, metaluminous and peraluminous)
that exhibit different patterns in viscosity, the viscosity of
metaluminous liquids is well described by a simple mathematical
expression involving the compositional parameter (SM) but the
compositional dependence of viscosity for peralkaline and
peraluminous melts is not fully controlled by SM. For these extreme
compositions we refitted the model using a temperature-dependent
parameter based on the excess of alkalies relative to alumina
(e.g., AE/SM). The recalibrated model reproduces the entire
database to within 5% relative error. On the basis of this extended
database the T-variation of the viscous response of strong and
fragile liquids within a wide range of compositions shows three
clearly contrasting compositional suites (peralkaline, metaluminous
and peraluminous). As a result, I present an extended model to
calculate the viscosity of silicate melts over a wide range of
temperatures and compositions. This model constitutes a significant
improvement with respect to the Giordano and Dingwell (2003a) study
in that: 1) The number of experimental determinations over which
the model is calibrated is larger. 2) The range of investigated
compositions is larger. 3) The investigated temperature range is
larger. 4) The assumption is made that at infinite temperature, the
viscosity of silicate melts converges to a common, but unknown
value of the pre-exponential factor (A=−4.07, Equation (7.1)). In
particular the compositional range involves a large number of
viscosity determinations for peralkaline and peraluminous
compositions in a temperature interval between 949 and 2653 K.
Furthermore, it is shown that the assumption of a common value of
the pre-exponential parameter A produces an equally good
representation of the experimental data as that produced by each
melt having its own specific A-value. This optimization also
induces a strong coupling between data sets that stabilizes the
range of solutions and allows the different rheological behaviour
of extreme compositions (peralkaline and peraluminous vs.
metaluminous) to be discriminated. It was demonstrated that,
although the parameter SM (Giordano and Dingwell, 2003a) can be
used to model compositional controls on the viscosities of
metaluminous liquids, it does not capture the viscosity of
peralkaline and peraluminous liquids. The differences in the
rheological behaviour of these extreme compositions reflect
important differences in the structural configuration of
metaluminous, peralkaline and peraluminous melts. Subsequently, a
second regression of the experimental data was performed involving
a second compositional parameter (AE) that accounts for the excess
of alkali oxides over the alumina. Incorporating this
temperature-dependent compositional parameter (i.e., AE) into the
SM-based model (Equation 7.7) appears to account for the anomalous
rheological behaviour of peralkaline and peraluminous liquids. The
resulting model reproduces the entire experimental database to
within an average RMSE of 0.45 log units. The model presented here
is recommended for the estimation of the viscosity of anhydrous
multicomponent silicate melts of volcanic interest.