Structure of Aluminosilicate Melts

Abstract

In peralkaline and meta-aluminous melts, essentially all Al³⁺ (>95%) occupy tetrahedral coordination, whereas for peraluminous melts, complex mixtures of aluminum triclusters with 4-fold coordinated Al³⁺ and Al³⁺ in 5- and 6-fold coordination with oxygen describe the structure. Aluminum in tetrahedral coordination requires electrical charge-balance. With alkali metals (M⁺) in this role, the proportions are M⁺=Al³⁺. The overall structure is dominated by three-dimensionally interconnected tetrahedra to form 6-membered rings of tetrahedra. The Al/(Al+Si) of these tetrahedra are simple positive functions of the bulk melt Al/(Al+Si). When tetrahedrally-coordinated Al³⁺ is charge-balanced by divalent cations, the M²⁺ cation charge-balances 2Al³⁺ tetrahedrally coordinated cations. This structure is dominated by SiO₄, (Si,Al)O₄, and AlO₄ entities.

In peraluminous melts, where there is insufficient proportion of M²⁺ and M²⁺ cations for charge-balance, aluminum exists in triclusters with Al³⁺ in tetrahedral coordination. In peralkaline aluminosilicate melts, there coexist discrete structural units with different degree of silicate polymerization. These units are termed Qⁿ-species where the superscript, n, is the number of bridging oxygen in individual units. Equilibria among these units are of the type, 2Qⁿ = Qⁿ⁺¹ + Qⁿ⁻¹. In these melts, Al³⁺ is distributed among these units. The Al³⁺ in peralkaline aluminosilicate melts strong preference Q⁴ units. This preference is, however, temperature-dependent as reflected in changes in the ΔH of the Qⁿ-species reaction.

1. Introduction

Solution of Al₂O₃ in silicate melts and glasses results in significant changes in their properties such as, for example, chemical durability, glass hardness, melt and glass viscosity, thermal expansion, and compressibility. Alumina addition also enhances resistance to devitrification as well as greatly reduced compositional range of liquid immiscibility. These and other property changes reflect how solution of Al³⁺ in silicate melt and glass causes significant structural changes. The details of the structural behavior Al³⁺ in silicate melts and glasses with implication for some important properties are the focus of this review.

A few major features of properties will be summarized first in order to illuminate the importance of Al₂O₃. This brief summary will be followed by a more detailed review of aluminosilicate melt structure and, finally, a few examples of how aluminosilicate melt properties may be understood in terms of the aluminosilicate melt structure.

2. Terminology

Aluminosilicate compositions are conveniently described within the system Mⁿ⁺O_n/2–Al₂O₃–SiO₂, where Mⁿ⁺O_n/2 is an alkali or an alkaline earth oxide (Fig. 1). The Mⁿ⁺O_n/2–Al₂O₃–SiO₂ system is divided into peralkaline, meta-aluminous, and peraluminous. Compositions with 1/nMⁿ⁺_n/2>Al³⁺ are peralkaline, those with 1/nMⁿ⁺_n/2=Al³⁺ are meta-aluminous, and those with 1/nMⁿ⁺_n/2<Al³⁺ are termed peraluminous. The properties and structure of melts and glasses in these three categories typically are significantly different. Those differences, which will be described below, reflect how the structural role of Al³⁺ in aluminosilicate melts varies depending on its proportion relative to metal cations, Mⁿ⁺_n/2, and on the electronic properties of this cation. In this review, we will describe their electronic properties in terms of their ionization potential, Z/r², where Z is formal electrical charge and r is ionic radius.

Fig. 1.

Compositional range of peralkaline, peraluminous, and meta-aluminous compositions in the system Mⁿ⁺O_n/2–Al₂O₃–SiO₂ where M is an alkali and alkaline earth oxide.

3. Properties

3.1. Liquid Immiscibility

In all Mⁿ⁺O_n/2–Al₂O₃–SiO₂ systems, there is immiscibility along the SiO₂–Al₂O₃ join (Fig. 2). This immiscibility gap extends somewhat into the Mⁿ⁺O_n/2–Al₂O₃–SiO₂ field. The extent of this expansion depends on the electronic properties of the Mⁿ⁺ cation. Unmixing near the SiO₂–Al₂O₃ join in the Mⁿ⁺O_n/2–Al₂O₃–SiO₂ system is marginal for Mⁿ⁺=Sr²⁺, small for both Ba²⁺ and Ca²⁺, but significant for Mg²⁺.^1,2,3,4) There is also immiscibility along the Mⁿ⁺O_n/2– SiO₂ join when the M-cation is an alkaline earth. There is no miscibility gap above the liquidus surface in alkali silicate melts.^2,3) By adding Al₂O₃ to these systems, the immiscibility gap shrinks in the same order as for the immiscibility gap near and at the SiO₂–Al₂O₃ gap (Fig. 2).

Fig. 2.

Extent of liquid immiscibility along the metal oxide-silica and alumina-silica joins in the systems Na₂O–Al₂O₃–SiO₂ and MgO–Al₂O₃–SiO₂ modified after.¹⁾

The immiscibility trends illustrated with the two examples in Fig. 2 reflect the fact that when Al³⁺ substitutes for Si⁴⁺ in tetrahedral coordination in the melt structure, the silicate framework polymerizes because of the metal cations that associate with Al³⁺ for an effective charge-compensation. Other things being equal, the influence of such cations on immiscibility along Mⁿ⁺O_n/2– SiO₂ join is the least effective in the magnesium system (Fig. 2). This leads to the suggestion that Mg²⁺ is the cation that associates the least strongly with Al³⁺.

3.2. Thermodynamics

Mixing Behavior. The enthalpy of solution, H_s, has been determined for a number of melts/glasses in Mⁿ⁺_n/2O–Al_nO_2n–SiO₂ systems^5,6) (Fig. 3). In those studies, the thermodynamic data were obtained by solution calorimetry at 700°C. Whether those data were obtained for a glass or supercooled liquid depended whether or not the experimental temperature, 700°C, was above or below the glass transition temperature for these compositions. Ideally, in order to compare all as liquids (or supercooled liquids), an adjustment near 5 kJ/mol would be needed.^7,8) This adjustment is, however, so small compared with the heat of solution values, that the adjustment is not made for the results in Fig. 3.

Fig. 3.

Energetics of the Al,Si substitution along various SiO₂–Mⁿ⁺Al_nO_n joins determined at 700°C. A. Enthalpy of solution, ΔH_s, as a function of composition. B. Enthalpy of solution, ΔH_s, as a function of ionization potential, Z/r², were Z is formal electric charge, and r is ionic radius (modified after⁶⁾).

The enthalpy-composition relationships in Fig. 3 illustrate how H_s-values succeed each other in the order of ionization potential of the metal cation. One of the most striking features of these data is the negative enthalpy of mixing between SiO₂ and MAlO₂ (M: alkali metal), suggesting strong affinity between these components. Moreover, as the Al/(Al+Si) increases so does the heat of solution (becomes more negative.)⁶⁾ Interestingly, whereas such simple correlations between H_s and Al/(Al+Si) is clear for any group of metal cations with the same formal electrical charge, there is a separate trend for series of differently charged metal cations (Fig. 3(B)).

There is an H_s-minimum near 50 mol% SiO₂,⁹⁾ the magnitude of which varies strongly as a function of the electronic properties of the M-cation. The minimum is about +10 kJ/mol for Mg- and less than −20 kJ/mol for K-aluminosilicates. The Mg meta-aluminate join also is the only meta-aluminosilicate join for which the enthalpy-composition relationship in Fig. 3(A) shows an initial maximum where, therefore, the SiO₂ deviates positively from ideal mixing.¹⁰⁾ It is notable that Mg-aluminosilicates unmix at very high SiO₂ content along the meta-aluminous join.⁴⁾

Heat Capacity. Heat capacity can be divided into vibrational and configurational (C_p^vib and C_p^conf, respectively). The heat capacity of glasses is vibrational, whereas for melts, the total heat capacity is the sum of vibrational and configurational heat capacity. Configurational heat capacity is introduced as a glass is transformed to supercooled melt at the glass transition. The configurational heat capacity of melts often is not very sensitive to temperature, but can vary when the melt structure is dependent on temperature.^7,11) Such features have been reported for aluminosilicate, ferrisilicate, and titanosilicate melts, for example (for review of such experimental data, see¹²⁾).

Heat capacity is sensitive to glass and melt structure. The vibrational density of states in glass and melt depends on oxygen coordination of every cation^13,14) which, therefore, governs C_p^vib. Heat capacity data for peralkaline compositions suggest major changes in short-range order around alkali cations when associated with aluminum. This association is temperature-dependent thus leading to temperature-dependent configurational heat capacity, C_p^conf, of alkali aluminosilicate melt.¹¹⁾ The oxygen environment of alkaline earth cations, on the other hand, does not change significantly regardless of details of their structural roles, which would suggest that the C_p^conf of alkaline earth aluminosilicate melts is not sensitive to temperature.

The evolution of configurational heat capacity with metal oxide/alumina ratio for both peralkaline and peraluminous compositions indicates that there must be major changes in short-range order around alkali cations caused by association with tetrahedral Al³⁺ (Fig. 4; see also^15,16)). For alkali aluminosilicate melts, there is a C_p^conf-increase on both sides of the meta-aluminate silicate join (Fig. 4). The C_p^conf-value at this minimum becomes less negative with increasing electronegativity of the alkali metal (see insert in Fig. 4). For alkaline earth aluminosilicates, the configurational heat capacity increases to a maximum value near the meta-aluminate join. This heat capacity maximum increases with increasing Z/r² of the metal cation. Notably, the configurational heat capacity becomes less sensitive to the type of metal cation the more electronegative the metal cation. This change reflects greater similarity of the M–O and Al–O bond strength with increasing Z/r² of the M-cation.

Fig. 4.

Variations in configuration heat capacity at the glass transition temperature, C_P(T_g), as a function of M_xO/(_xO+Al₂O₃) at 67 mol% SiO₂. Insert shows configurational heat capacity as a function of ionization potential of charge-balancing cation at M_xO/(_xO+Al₂O₃)=0.45–0.5 (modified from^15,16)).

Notably, the configurational heat capacity of aluminosilicate melts is essentially a linear function of Al/(Al+Si) for the Mg- and Ca-meta-aluminosilicate melts, whereas this is not so for the SiO₂–NaAlO₂ melts.¹⁷⁾ This difference likely is because for the alkaline earth aluminosilicates, two Al³⁺ cations are charge-compensated for each metal cation, whereas only one Al³⁺ is charge-compensated by the monovalent Na⁺ cation. There is, therefore, less flexibility and probably longer-range order in the alkaline earth aluminosilicate glasses and melts.

Viscosity. The viscosity of aluminosilicate melts varies from several to many orders of magnitude depending on their chemical composition.^{17,18,19,20,21)} One compositional variable is the Al/(Al+Si). A second is metal/alumina ratio of the melt (Fig. 5). How the Al/(Al+Si) affects melt viscosity depends on metal oxide/(silica+alumina) of the melt.

Fig. 5.

Activation energy of viscous flow, assuming Ahrrenian viscosity behavior, of melts along the joins NaAlO₂–SiO₂ and Na₂Si₂O₅–Na₂(NaAl) ₂O₅ modified after.^21,23) The activation energy is derived from melts at temperatures above the glass transition. For the compositions shown in this Figure, the viscosity is essentially Ahrrenian at temperatures above the glass transition.

For example, along the SiO₂–NaAlO₂ join, the melt viscosity and activation energy of viscous flow decreases systematically with increasing Al/(Al+Si) (Fig. 5). For these melts, the principal control is the smaller Al–O bond energy compared with that of the Si–O.²²⁾ However, in depolymerized and, therefore, peralkaline Na₂O–Al₂O₃–SiO₂ melts such as those along the Na₂Si₂O₅–Na₂(NaAl)₂O₅ join, for example, the activation energy of viscous flow actually decreases and then increases slightly with increasing bulk melt Al/(Al+Si) (Fig. 5, lower curve). In this case, the behavior is linked to the Al³⁺ distribution among structural units in the melt.^23,24)

Melt viscosity also is sensitive to the Al³⁺/(Al³⁺ + Mⁿ⁺_n/2) abundance ratio (Fig. 6). In the example in Fig. 6 from the Na₂O–AlO₃–SiO₂ system,^21,25,26) melt viscosity (and activation energy of viscous flow) increases with increasing Al/(Al+Na) at constant SiO₂ in peralkaline melts and viscosity decreases with further decrease of Al/(Al+Na). In the per-aluminous region, increasing Al/(Al+Na) leads to formation of Al-bearing structural entities with much lower bond strength than the Al–O bonds in the peralkaline melts and, therefore, decreasing activation energy of viscous flow (and any other transport property). The viscosity maximum is less pronounced for alkaline earth aluminosilicate melts. This difference of alkali aluminosilicate melt viscosity can be ascribed to decreased strength of the association of the metal cation with aluminum for alkaline earths.

Fig. 6.

Viscosity of aluminosilicate melts along the system Na₂O–Al₂O₃–SiO₂ at 75 and 67 mol% SiO₂ as a function of Al/(Al+Na) at 1700°C. Melts with Al/(Al+Na)<0.5 er peralkaline, those with this ratio at 0.5 are meta-aluminous and melts with Al/(Al+Na)>0.5 are peraluminous (modified after²⁵⁾).

Molar Volume. Molar volume of aluminosilicate melts and partial molar volume of Al₂O₃ therein also are variables that are sensitive to the structural role of Al³⁺. Whereas for peraluminous and meta-aluminous melts, the molar volume is an additive function of the partial molar volume of the oxide components,^27,28) for melts along the SiO₂–Al₂O₃ join, linear additivity of volume no longer is the case²⁹⁾ (see also Fig. 7). The departure from linearity in Fig. 7 results from the molar volume of liquid Al₂O₃ (about 33 cm³/mol above 2000°C³⁰⁾) is lower than the 38–39 cm³/mol derived for the partial molar volume of Al₂O₃ in peralkaline melts.

Fig. 7.

Molar volume of binary SiO₂–Al₂O₃ melts at the temperatures indicated. Data from²⁴⁾ for binary compositions,^30,110) for pure Al₂O₃ and¹¹³⁾ for SiO₂. The dashed curve showing linearly additive is 1800°C model values of.¹¹¹⁾

Volume measurements are available for binary SiO₂–Al₂O₃ melts.²⁹⁾ These volume data join smoothly with the data for pure Al₂O₃ and SiO₂. A definite nonlinearity is observed beyond about 60 mol% Al₂O₃ at 1900–2000°C although the thermal expansion coefficients of liquid Al₂O₃ and of the most Al₂O₃-rich melt are not well constrained.

The density of binary Al₂O₃–SiO₂ melts, when approached from ternary metal aluminosilicate systems, depends on the electronic properties of the M-cation. For the Na₂O–Al₂O₃–SiO₂ system, additivity of molar volume Al/(Al+Na) breaks down as soon as the meta-aluminous join is crossed (Fig. 8). The extent of deviation from linearity increases with decreasing SiO₂ content and, therefore, with increasing Al₂O₃ concentration. In the MgO–Al₂O₃–SiO₂ system, there is less deviation from linearity than in the sodium aluminosilicate system (see insert in Fig. 8). The MgO–Al₂O₃–SiO₂ data extrapolate to a volume of 31 cm³/mol for pure Al₂O₃. This value is essentially the same as the molar volume of liquid Al₂O₃ on its liquidus (2073°C³⁰⁾).

Fig. 8.

Molar volume of Na-alumosilicate melts at 1800°C at constant SiO₂ content as indicated on diagram (data from²⁵⁾). Insert shows magnesium aluminosilicates at 50 mol% SiO₂ (data from^109,113)).

4. Glass versus Melt; Glass Transition

Most of the properties discussed above were determined on molten silicate and aluminosilicate. These data offer suggestions into how melt structure may cause property changes. It is, however, important to recognize that a majority of the experimental structural data employed to rationalize melt properties have been on obtained on glasses formed upon temperature quenching from the melt. Given the dynamic nature of most melt properties as well as the dynamic, but different, nature of glass structure, it is necessary to address the structure changes as a melt is transformed to a glass.

The temperature at which a glass is transformed to a glass is termed the glass transition temperature, T_g. At and below T_g the material is not relaxed to its equilibrium state, whereas above the glass transition the material has reached its relaxed, equilibrium state. However, in contrast to the temperature of an equilibrium transition, the glass transition temperature is not a fixed temperature, but can vary with experimental conditions. The T_g is higher for higher cooling rates or shorter experimental time scales. Such variations of T_g can be several hundred degrees as illustrated in Fig. 9 for the glass transition of CaAl₂Si₂O₈. In this latter example, the time scale of viscometry, ultrasonic, and Brillouin scattering experiments are on the order of the order of 10², 10⁻⁶ and 10⁻¹⁰ s, respectively. As a result, the glass transition temperature determined with these three different techniques varies by nearly 1200 degrees.

Fig. 9.

Time-dependence of the glass transition for CaAl₂Si₂O₈ (from data of¹¹⁵⁾) showing the glass transition temperature, determined with the different methods, as a function of property time. The transition curve separates the fully relaxed liquid from the unrelaxed glass. It is plotted against experimental timescale to show the conditions under which the glass transition takes place for the three kinds of experiments indicated.

Glass transitions can be detected with thermodynamic, kinetic, and volumetric measurements. Here, I will only show a thermodynamic example. Configurational properties are particularly sensitive. For glasses, configurational heat capacity undergoes very rapid changes at the glass transition (Fig. 10). The magnitude of the configurational heat capacity change across the glass transition is;   

$$C_{\mathrm{p}}{}^{\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{f}}=C_{\mathrm{p}}{}^{\mathrm{l}\mathrm{i}\mathrm{q}}–C_{\mathrm{p}}{}^{\mathrm{g}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}}({\mathrm{T}}_{\mathrm{g}}),$$(1)

Fig. 10.

Heat capacity of a variety of silicates glasses and melts as a function of temperature below and above the glass transition temperature. (See¹¹⁶⁾ for data sources).

Its value is sensitive to composition and tends to increase the higher the metal oxide/silica ratio of the material.³¹⁾

5. Structure

Aluminosilicate melt structure has been examined both via numerical simulation methods and with direct structural determination using methods such as x-ray and vibrational and NMR spectroscopy, for example.

5.1. Al₂O₃

Results from numerical simulation of melt structure indicate more than one coordination state of Al³⁺ in molten Al₂O₃ (Fig. 11). The average Al–O distance is in 1.71–1.79 Å range, which is near that of Al³⁺ in tetrahedral coordination.^32,33) In comparison, the Al–O distance in crystal structures with 6-fold coordination of Al³⁺ such as in jadeite, the Al–O distance is about 1.9 Å.

Fig. 11.

Structure of molten Al₂O₃. A. Al–O–Al angle distribution calculated at 2300°C (modified after⁴⁸⁾). B. Probability distribution of oxygen coordination number surrounding Al³⁺ in Al₂O₃ melt at 2500 K (modified after³⁹⁾).

The Al–O–Al bond-angle distribution in molten Al₂O₃ is asymmetric and does, in fact, show at least two distinct distributions (Fig. 11(A)). Those results have led to different interpretations. For example,³⁴⁾ proposed that it reflects coexisting 4-, 5-, and 6-fold coordination of Al³⁺, with 5-fold coordination dominating. The 5- and 4-fold coordinated Al³⁺ exist in about equal proportions in molten Al₂O₃ as has been suggested.³⁵⁾ In another study, it was concluded that the average oxygen coordination around Al³⁺ is slightly higher than 4 (4.25) with about 85% of the Al³⁺ in 4-fold coordination.³³⁾ This conclusion also is in accord with,³²⁾ who found that Al³⁺ for the most part is in 4-fold coordination, based on the good agreement between the calculated and experimental melting temperature of Al₂O₃.³⁶⁾

Results from ²⁷Al NMR spectroscopy of molten Al₂O₃ to temperatures near 2450°C have been interpreted to indicate an Al³⁺ coordination number greater than 4, but for the most part less than 6.^37,38) From a neutron diffraction study,³⁹⁾ it also was found that 4-fold Al³⁺⁻to dominates (~60%), with significant contributions from both higher and lower coordination numbers (Fig. 11(B)). The conclusion that molten Al₂O₃ comprises mostly 4-fold coordinated Al³⁺ agrees with the interpretation of an Al-O bond distance of about 1.76Å derived from the x-ray radial distribution function of Al₂O₃ near and above its 2070°C melting temperature.³⁶⁾ These experimental results for the most part also accord with the structural interpretation of numerical simulations.^40,41,42)

5.2. Al₂O₃–SiO₂

Simulated structure of Al₂O₃–SiO₂ melt and glass reveals similarities of the Al³⁺ structural role with that in the Al₂O₃ endmember. However, differences also exist. The differences become more pronounced as the composition becomes more SiO₂-rich. These structural differences include oxygen triclusters, which have been reported to be common in simulated structure of SiO₂–Al₂O₃ melts and glasses.^30,40,41) Moreover, the AlO₄-tetrahedra coexisting with SiO₄ tetrahedra are different in that edge-sharing is common for AlO₄, tetraheddra, which has not been reported for SiO₄ tetrahedra.

In Al₂O₃–SiO₂ melt structure, Al–O polyhedra with different oxygen coordination exist. For example. From computed Si–O–Si and Al–O–Al angle distributions in Al₂O₃–SiO₂ melt, there are multiple average angles for a composition such as Al₂O₃·2SiO₂.⁴²⁾

The average oxygen coordination number from the angle distribution computations of Al₂O₃–SiO₂ melt ranges between about 4.5 and 5.^40,41,42) These coordination numbers also vary with temperature. The 3-fold and 5-fold coordinated Al³⁺ becomes less abundant with increasing temperature, whereas 4-fold coordinated Al³⁺ abundance increases⁴¹⁾ (Fig. 12). This temperature-dependent evolution of coordination numbers seems independent of the Si/Al ratio, although the Si/Al ratio will affect the relative abundances of these different structural forms of Al³⁺.^40,41)

Fig. 12.

Aluminum coordination numbers in Al₂O₃–SiO₂ melt at temperatures in excess of 2000 K with 13 mol% SiO₂ from numerical simulation (modified from⁴¹⁾).

In experimental determination of Al₂O₃–SiO₂ melt and glass structure, it has been suggested that as much as 6 wt% Al₂O₃ may dissolve in molten SiO₂ without significant disruption of the SiO₂ structure.⁴³⁾ In another experimental study,⁴⁴⁾ as much as 59 wt% Al₂O₃ could be dissolved in SiO₂ without disruption of its structure. Despite problems with crystal nucleation during quenching of these melts, it was found that the structure of silica-rich (>70 mol% SiO2) glasses along the SiO₂–Al₂O₃ join resembles that of pure SiO2. This conclusion differs somewhat, though, from the structural interpretation of multinuclear NMR spectra of Al₂O₃–SiO₂ glasses with Al₂O₃ contents between 0.4 and 12 wt%.⁴⁵⁾ Here, 4-, 5-, and 6-fold coordinated Al³⁺ was reported. In yet another another study⁴⁴⁾ with alumina contents in excess of 9 wt%, it was suggested that a significant portion of the Al³⁺ occurs in some form of aluminate clusters. comprising ^[4]A, ^[5]Al, and ^[6]Al.

From ²⁷Al MAS NMR spectra of SiO₂–Al₂O₃ glass temperature-quenched from melt, it also has been concluded that multiple oxygen coordination states.⁴⁶⁾ They proposed, in accord with the x-ray data,⁴⁷⁾ that these clusters contain 4-, 5-, and 6-fold coordinated Al³⁺ (Fig. 13). From these latter data, it can be concluded that even with several tens of mol% SiO₂ added to Al₂O₃, the Al³⁺ structural role may not differ greatly from its structural role in pure Al₂O₃ melt. Such structural models accord with those from numerical simulation.⁴⁷⁾

Fig. 13.

The Al coordination numbers from ²⁷Al MAS NMR spectrum of SiO₂–Al₂O₃ glass with 30 mol Al₂O₃ modified after.⁵⁰⁾

Problems associated quenching Al₂O₃–SiO₂ melts to glass without nucleation of crystals were avoided in another experimental study conducted at temperatures above that of melting.⁴⁸⁾ Here, high-temperature ²⁷Al NMR spectroscopy was coupled with molecular dynamics simulation of SiO₂–Al₂O melts to temperatures above 2000°C. The ²⁷Al isotropic shift is broadly correlated with the Al₂O₃/SiO₂ abundance ratio. This evolution was interpretaed to suggest changes in average Al³⁺ coordination number, which, in turn, was interpreted to to reflect several different aluminate complexes in the SiO₂–Al₂O₃ melt.This observation is at least qualitatively in accord with other interpretation of Raman and NMR spectra of SiO₂–Al₂O₃ glasses^44,45,49,50)

5.3. Charged-Balanced Al³⁺

At low pressure, the ionic radius of Al³⁺ (r = 0.47 Å) is not greatly different from to the 0.34 Å ionic radius of Si⁴⁺. However, the formal charge of Al is 3+, which means that it can occupy tetrahedral sites of the silicate framework only by associating with a neighboring charge-compensating cation so that a 4+ formal charge can be obtained. An alkali (univalent) cation, as illustrated by albite (NaAlSi₃O₈), for example, can play this role as can an alkaline earth (divalent) cation, which compensates for two different Al³⁺, as illustrated in crystals by anorthite (CaAl₂Si₂O₈).

Alkali metals and alkaline earths in aluminosilicate melts and glasses may, therefore, occur both as cations that charge-balance tetrahedrally coordinated Al³⁺ and as network modifiers that form bonds with oxygens at the end of SiO₄ polymers. The dual structural roles of alkali metals and alkaline earths has been documented by experiment. The Na-NBO (NBO: nonbridging oxygen; bonds attached to oxygen at the end of aluminosilicate polymers) bond is shorter than Na-BO (BO: bridging oxygen; bonds that link neighboring aluminosilicate tetrahedra) distances and that Ca-NBO distances are shorter than for Ca-BO distances.^51,52,53,54) For Na⁺, for example, this bond distance difference results in more electronic deshielding of the Na nucleus as the proportion of Na-NBO bonds increases. That suggestion is in accord with ²³Na NMR data as well as other structural data for binary Na₂O–SiO₂ glasses and ternary meta-aluminosilicate glasses along the alkali and alkaline earth aluminosilicate joins join^55,56) (Fig. 14). The Na-NBO bond distance might be as much as 10% shorter for Na-NBO than for Na–O when Na⁺ serves to charge-balance tetrahedrally coordinated Al³⁺. The exact relationship to shielding of the Na nucleus also depends, however, on Na/Si and Al/(Al+Si).⁵⁵⁾

Fig. 14.

²³Na MAS NMR spectra of glasses of Na₂Si₂O₅ and NaAlSi₆O₁₄ composition (modified after⁵⁶⁾).

5.4. Al³⁺ Substitution for Si⁴⁺

The average T–O bond length in aluminosilicate melts, glasses, and crystals increases with increasing Al/(Al+Si). This increase results from the generally slightly longer Al–O bond length compared with Si–O bonds in the oxygen tetrahedra.^22,57)

For melts and glasses on the SiO₂–NaAlO₂ join, there is near random substitution of Al³⁺ for Si⁴⁺ in the structure so that the average (Si,Al)–O bond length increases monotonously with increasing Al/(Al+Si).^58,59,60) However, with divalent cations for Al³⁺ charge-balance, there is more structural disorder.⁵⁹⁾ This difference is evident, for example, in the much greater line width in ²⁹Si MAS NMR spectra of Ca-aluminosilicate glasses compared with the spectra of Na-aluminosilicate glasses (Fig. 15). Although this line width in general also depends on Al/(Al+Si), even at constant Al/(Al+Si), it increases as the charge-balancing cation becomes more electronegative. These relationships are consistent with increasing S⇔Al disorder.^59,60)

Fig. 15.

Full width at half height of the ²⁹Si peak in MAS NMR spectra of glasses along the joins NaAlO₂–SiO₂ and CaAl₂O₄–SiO₂ as a function of their bulk Al/(Al+Si) modified after.⁶⁰⁾ Insert shows examples of the actual NMR spectra.

The electronic properties of the charge-balancing metal cations affect, therefore, the aluminosilicate structure. These properties also affect physical and chemical properties of the materials. This, in turn, reflects the variations in the bond strength and bond angle distribution depending on how Al³⁺ is charge-compensated.^59,60,61,62) A measure of this effect can be seen in the perturbation of the T–O bond as a function of the nature of the charge-balancing cation. This perturbation, Δ(TO)_aver, has been expressed as a deviation from ideal Si–O and Al–O bond lengths as a function of X=Al/(Al+Si);⁶¹⁾   

$$\mathrm{\Delta }{\left(\mathrm{T}\mathrm{O}\right)}_{\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{r}}=\mathrm{X}\cdot \mathrm{\Delta }(\mathrm{A}\mathrm{l}\mathrm{O})+\text{(}1-\mathrm{X}\text{)}\cdot \mathrm{\Delta }(\mathrm{S}\mathrm{i}\mathrm{O}),$$(2)

where Δ(AlO) and Δ(SiO) represent the difference of Si–O and Al–O bond lengths from ideal values, 1.712 and 1.581 Å, respectively

The perturbation of the T–O bond lenths affects the stability of meta-aluminosilicate melts and glasses. For example, the stabilization enthalpy, ΔH_stab, relative to the enthalpies of solution of SiO₂ and aluminosilicate, ΔH_soln, has been expressed as:⁶²⁾   

$${\mathrm{H}}_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{b}}=\left[\mathrm{\Delta }{\mathrm{H}}_{\mathrm{s}\mathrm{o}\mathrm{l}\mathrm{n}}({\mathrm{M}}_{\mathrm{X}/\mathrm{n}}\mathrm{A}{\mathrm{l}}_{\mathrm{X}}\mathrm{S}{\mathrm{i}}_{1-\mathrm{x}}{\mathrm{O}}_2)-\mathrm{\Delta }{\mathrm{H}}_{\mathrm{s}\mathrm{o}\mathrm{l}\mathrm{n}}(\mathrm{S}\mathrm{i}{\mathrm{O}}_2)\right]/\mathrm{X}.$$(3)

The perturbation of the T–O bond length is negatively correlated with the stabilization enthapy (Fig. 16). In other words, the more perturbed (weakened) the T–O bond by various charge-compensating cations, the less stable is the aluminosilicate complex. To some extent, this effect also is affected by the Al/(Al+Si) and the enthalpy of solution (see also Fig. 3), but the electronic properties of the charge-balancing cation dominate the stabilization enthalpy.

Fig. 16.

Stabilization enthalpy, ΔH_stab, as a function of perturbation of the T–O bond, Δ(TO)_aver for charge-compensated Al³⁺ of aluminosilicate glasses. The numbers in perentheses indicate the Al/(Al+Si) ratio of the material (modified after⁶¹⁾).

5.5. Structure of Meta-Aluminosilicate

Different Al,Si ordering resulting from the different electronic properties of the charge-balancing cation is also seen in the distribution of (Si,Al)–O–(Si,Al) angles.⁶³⁾ The wider the angle range and the more angle maxima, the more disordered is the aluminosilicate structure (Fig. 17). The angle maxima range between about 120° and 180°. However, both the maxima and the distribution around this maximum depend on the metal cation properties.

Fig. 17.

Intertetrahedral (Si,Al)–O–(Si,Al) angle distribution from simulated structure in melts of Mg_0.5AlSiO₄ and NaAlSiO₄ composition modified after.⁶⁴⁾

The structure of melts and glasses along the SiO₂–NaAlO₂ join resembles that of the glass and melt structure of pure SiO₂.^64,65,66) The structure of molten SiO₂ is likely made up of a small number of 3-dimensionally interconnected rings of SiO₄ tetrahedra.^67,68) It has been proposed⁶⁷⁾ that melts and glasses along the SiO₂–NaAlO₂ join comprises coexisting, 3-dimensionally interconnected structures whose intertetrahedral angles differ by 5°–10°. This angle difference is in response to slightly different Al/(Al+Si) in the two structural units (Fig. 18(A)).

Fig. 18.

Structure of melts and glasses along met-aluminosilicate joins. A. Data along the join NaAlO₂–SiO₂. Numbers along the (Si,Al)-unit(I) are the Al/(Al+Si) ratios. B. Data along the join CaAl₂O₄–SiO₂ (Figures modified from⁶⁸⁾).

From this analysis, one cannot determine whether the two ring structures differ in the number of (Si,Al)O₄ tetrahedra. The angle difference could also be due to differing extent of puckering (see⁶⁹⁾ for discussion of energy minimization modeling and its relationship to intertetrahedral angle variations). A variation of intertetrahedral angle of 5°–10° could also result from differences in Al/(Al+Si) as increasing Al³⁺ in tetrahedral coordination induces a decrease of intertetrahedral angle, at least in crystalline aluminosilicates.²²⁾

The Al, Si ordering in anorthite crystals resembles that of anorthite glass to 5Å distance or perhaps even slightly more⁵⁹⁾ (Fig. 19). This conclusion is also in agreement with the interpretation of wide-angle x-ray study of the same glass. Here,⁷⁰⁾ it was suggested that the x-ray spectrum of other glasses on the SiO₂–CaAl₂O₄ join could be described as a mixture of anorthite-like and SiO₂-like structures. This structural model resembles that of⁷¹⁾ from their interpretation of Raman spectra of CaAl₂Si₂O₈ melt at high temperature. It follows, therefore, that distinct structural entities coexist in SiO₂–CaAl₂O₄ glasses.^67,70) These units have been referred to as an Al-unit (without Si) together with Al_0.5Si_0.5- and Al-free SiO₂-like structure (Fig. 18(B)). Their abundance varies with Al/(Al+Si) (Fig. 18(B)), but the Al/(Al+Si) of the individual units does not change with bulk melt (and glass) Al/(Al+Si). Analogous entities may exist in SiO₂–MgAl₂O glasses.⁶⁷⁾

Fig. 19.

X-ray radial distribution functions of glassy and molten NaAlSi₃O₈ and CaAl₂Si₂O₈ (data from^59,116)).

Ideally, one might expect that melts and glasses along meta-aluminosilicate joins are fully polymerized (all oxygen atoms serve to link neighboring aluminosilicate tetrahedra). Spectroscopic data, relying for the most part on NMR spectroscopy with ²⁹Si and ²⁷Al as the principal nucleii used, indicate, however, that this assumption is not entirely correct.^72,73,74) There is, in fact, evidence of a small percentage of 5-fold coordinated Si⁴⁺. Their exact proportion depends on the nature of the charge-balancing cation.^72.73) Notably, at least in the SiO₂–CaAl₂O₄ glasses, those with the highest Al-content have the greatest abundance of nonbridging oxygens (~5%).⁷⁴⁾

There has been a small number of experimental studies aimed at characterization of meta-aluminosilicate melts at temperatures above that of their melting.^71,75,76) The x-ray radial distribution functions of molten NaAlSi₃O₈ at 1200°C and CaAl₂Si₂O₈ at 1600°C⁷⁵⁾ resemble those of the glasses of the same composition at ambient temperature (Fig. 19) although, at least for NaAlSi₃O₈, the first (Si, Al)–O radial distance near 1.6–1.7 Å (See arrow in Fig. 19) might be slightly shorter in the glass than in the melt. This effect might be what is expected as the structure of a melt opens up with increasing temperature.⁷⁹⁾ Larger angles at higher temperature could cause a shortening of the (Si, Al)–O bridging bond distance.^22,79) Such a structural effect cannot be seen, however, in the x-ray radial-distribution functions of CaAl₂Si₂O₈ glass and melt (Fig. 19). An interpretation of high-temperature Raman spectra of CaAl₂Si₂O₈ melt was that temperature does not affect intertetrahedral angles and, thus, presumably, (Si, Al)–O bond lengths.⁷¹⁾

The (Al,Si)-ordering in an aluminosilicate network also depends on temperature.^79,80) There is, for example, a small temperature effect on the proportion of two coexisting, 3-dimentionally connected structural units in Na-metaaluminosilicate melts (Fig. 20). The abundance of the unit with the highest Al/(Al+Si) (denoted unit “1” in Fig. 20) decreases slightly with increasing temperature.⁷⁸⁾ The distribution of Al³⁺ among the coexisting, fully polymerized (Si,Al)O₄ tetrahedra in SiO₂–NaAlO melts appears, therefore, to depend on temperature. This temperature effect differs, therefore, from the effect on SiO–CaAl₂O₄ melts where there is no evidence for tempeature-dependent changes in the Al³⁺ distribution among structural units.

Fig. 20.

Abundance evolution of the two structural entities in NaAlO₂–SiO₂ melt as a function of temperature (modified after⁷⁹⁾).

As discussed above, aluminosilicate with 1/nMⁿ⁺_n/2>Al³⁺ are termed peralkaline. In melts and glasses of such compositions, the excess M-cations are network-modifiers. Network-modifying cations link the silicate or aluminosilicate network units via bonding with nonbridging oxygen (Fig. 21).

Fig. 21.

Schematic representation of the silicate structure that involves nonbridging and bridging oxygen.

Melts and glasses with network-modifying cations are depolymerized. The term, nonbridging oxygen per tetrahedrally coordinated cations, NBO/T, is commonly used as a measure of the extent which a silicate and aluminosilicate is depolymerized. In this notation, T-cations are the tetrahedrally coordinated cations. In aluminosilicate melts, T=Si+Al as long as there is no excess Al³⁺ over that which can be charge-balanced with M-cations.

5.6. Structural Principles of Depolymerized Silicate Glasses and Melts

Before addressing peralkaline aluminosilicate melt and glass structure, it is necessary to describe briefly the rules that govern the structure of depolymerized silicate melt and glass. With the aid of a the so-called “fingerprint technique” where vibrational spectra of crystalline materials and their glasses were compared, it was noted early on that the main anionic entities of CaMgSi₂O₆ glass resemble those of diopside (CaMgSi₂O₆) crystals because the main Raman peaks in both spectra are at approximately the same frequency.⁸¹⁾ A comparison of the Raman frequencies in the spectra of CaMgSi₂O₆ melt and diopside recorded at the diopside liquidus temperature (1392°C⁸²⁾) has revealed even greater similarity between spectra of crystalline diopside and CaMgSi₂O₆ melt at this high temperature.⁸³⁾

There are systematic relationships between silicate glass and melt structure and chemical composition^84,85) much like that of the original observations by Etchepare.⁸¹⁾ Those structural concepts were expanded^86,87) by investigating glasses ranging from pure SiO₂ to nearly as depolymerized compositions as orthosilicate (consisting of isolated SiO₄ tetrahedra only). In these glasses, not only do SiO₃- and Si₂O₅-like structures exist, but there are additional types of units such as SiO₂, Si₂O₇, and SiO₄ groups. These unites subsequently have been described in terms of Qⁿ units where the superscript, n, refers to the number of bridging oxygen in the unit. The number of nonbridging oxygen then simply is 4-n.

From the Raman and NMR data of metal oxide – silica glasses and melts,^{76,87,88,89,90)} it follows that there does indeed exist a small number of discrete silicate structures (n=0, 1, 2, 3, and 4). For example, in a comprehensive ²⁹Si MAS NMR study of structure of alkali silicate glasses, individual peaks remain at nearly constant frequency, but changed their relative intensity as a function of the alkali/silicon ratio (Fig. 22). These frequencies are similar to those observed in the ²⁹Si NMR spectra of crystalline alkali silicates of equivalent degree of polymerization.⁹⁰⁾ The chemical shift of bands assigned to individual Qⁿ-species (Fig. 22) varied slightly because there is a small deshielding effect even for the individual Qⁿ-species. This deshielding of ²⁹Si nucleus, in turn, depends on Si–O bond length, Si–O–Si bridging bond angles, and on changes in 2nd − and 3rd-nearest neighbor environments.^{90,91,92,93,94,95)}

Fig. 22.

A. ²⁹Si MAS NMR spectra of glasses in the SiO₂–Na₂O system with the Na₂O content (mol%) indicated on the individual spectra (modified after⁷⁶⁾). B. Schematic representation of the different Qⁿ-species in silicate glasses and melts. (Online version in color.)

The ²⁹Si MAS NMR data (Fig. 22) provide strong support, therefore, for the concepts that a small number of Qⁿ-species exist in silicate glasses. Such glass data have been extended to the behavior of melts at high temperature.⁹⁷⁾

A simple equation of the form;   

$$2{\mathrm{Q}}^{\mathrm{n}}={\mathrm{Q}}^{\mathrm{n}-1}+{\mathrm{Q}}^{\mathrm{n}+1},$$(4)

is now used to describe the equilibria among the Qⁿ-species in silicate (and aluminosilicate) melts.⁸⁷⁾

The Qⁿ-species abundance data from ²⁹Si MAS NMR spectra, shown for K₂O–SiO₂ and Na₂O glasses in Fig. 23, are consistent with the idea that the principal structural variations in metal oxide silicate glasses are the abundance of the individual structural units. In pure SiO₂ glass, only Q⁴ species are detected. By adding metal oxides to SiO₂ melt, Q⁴ species begin to transform into Q³ whose abundance reaches a maximum near the disilicate stoichiometry (Si₂O₅). With a further addition of metal oxide, a maximum in Q² abundance is observed for composition near the metasilicate stoichiometry (SiO₃). The degree of polymerization of the individual structural units does not evolve as a function of overall melt and glass polymerization. Only the proportion of the units varies.

Fig. 23.

Abundance of Qⁿ-species in Na₂O–SiO₂ and K₂O–SiO₂ glasses as a function of their composition (data from⁹⁰⁾).

The NMR data in Fig. 23 also show that Qⁿ-species abundance depends on the type of metal cation. For a fixed M/Si-ratio, for example, the abundance of Q³ species in a binary metal oxide silicate glass decreases with increasing ionization potential, Z/r², of the alkali metal, whereas those of Q² and Q⁴ increase. It follows that reaction (3) shifts to the right with increasing ionization potential of the M-cation.^96,97,98)

5.7. Temperature and Qⁿ-Speciation

The Qⁿ-speciation in the glass does not vary with temperature. However, the onset structural relaxation at the glass transition appears as a distinct change in the slope of the temperature versus Qⁿ abundance relationship⁹⁶⁾ (see also Fig. 24). At and below the glass transition temperature, T_g, the Qⁿ-species abundance does not vary with temperature. However, at temperatures higher than T_g, the abundance varies systematically with temperature. In this temperature regime, the equilibrium constant for equilibrium (4) is a near linear function of 1/T (K⁻¹)^{99,100,101,102,103)} (see also Fig. 24(B)).

Fig. 24.

Evolution of Q⁴, Q³, and Q² species as a function of temperature below and above the glass transition temperature as indicated from Na₂Si₂O₅ composition glass and melt (modified after¹⁰³⁾).

5.8. Peralkaline Aluminosilicate Glasses and Melts

The structural behavior of Al³⁺ in meta-aluminosilicate glass and melt structures with NBO/T~0 illustrates the principles governing substitution of Si⁴⁺ by Al³⁺ in tetrahedral coordination. However, in peralkaline glasses and melts, which are depolymerized (NBO/T>0) and where, therefore, multiple Qⁿ-species coexist, there is further structural complexity affecting Al,Si substitution. Here, essentially all the Al³⁺ enters Q⁴ species with little or no Al³⁺ in the coexisting, more depolymerized structural units such as Q³ and Q^{2 24)} (Fig. 25). Therefore, when Al/(Al+Si) increases in systems in which the nominal NBO/T (T = Si+Al) is constant, the abundance of Q² and Q⁴ structural units increases, whereas Q³ units become less abundant (Fig. 26). A consequence of this is that an equilibrium such as in Eq. (4) shifts to the right with increasing bulk melt Al/(Al+Si) at constant temperature and NBO/T.

Fig. 25.

Al/(Al+Si) of coexisting Qⁿ-species in glasses along the join Na₂Si₃O₇–Na₂(NaAl)₃O₇ as a function of bulk melt Al/(Al+Si) (modified from²⁴⁾).

Fig. 26.

Concentration of Q⁴, Q³, and Q² species in glasses along the join Na₂Si₃O₇–Na₂(NaAl)₃O₇ as a function of bulk melt Al/(Al+Si). The Qⁿ_species abundance calculated with Si⁴⁺ and Al³⁺ incorporated in the individual Qⁿ-species (modified from²⁴⁾).

The temperature-dependence of Qⁿ-abundance in peralkaline aluminosilicate melts qualitatively resembles that observed for the Al-free alkali silicate endmembers.^101,102,103) The concentration of Q², Q³, and Q⁴ units is insensitive to temperature over several hundred degrees below the temperture of the glass transition (Fig. 27). Above the glass transition, the abundances of Q⁴ and Q² species increase and that of Q³ decreases.²⁴⁾

Fig. 27.

Evolution of Qⁿ species in aluminosilicate melts with Al/(Al+Si) as indicated as a function of temperature below and above the glass transition (modified from¹⁰³⁾).

The Qⁿ-species abundance evolution with temperature in peralkaline aluminosilicate melts (above the glass transition) is somewhat more sensitive to temperature than in silicate melts without Al³⁺. Therefore, the enthalpy for the disproportion reaction, Eq. (4), depends on the Al/(Al+Si) (Fig. 28). It also depends on the electronic properties of the alkali metal as it is higher the more electronegative the alkali metal that serves to charge-compensate Al³⁺ (Fig. 29).

Fig. 28.

Enthalpy of reaction (4) calculated from the abundance of Q⁴, Q³, and Q² structural units at temperatures above the glass transition range for tetrasilicate melts (modified after¹⁰³⁾). for melts along the join Na₂Si₄O₉–Na₂(NaAl)₄O₉ as a function of bulk melt Al/(Al+Si).

Fig. 29.

Enthalpy of reaction (4) for melts along the joins K₂Si₄O₉–K₂(KAl)₄O₉, Na₂Si₄O₉–Na₂(NaAl)₄O₉, and Li₂Si₄O₉–Li₂(LiAl)₄O₉ at bulk melt Al/(Al+Si) = 0.2 as a function of ionization potential, Z/r², for cation in 6-fold coordination with oxygen (modified from¹⁰³⁾).

6. Properties and Structure

Properties of aluminosilicate glasses and melts are systematic functions of their composition and, therefore, their structure. The main structural factors are the Al³⁺ charge compensation, the electronic properties of the charge-balancing cation (or cations), and the degree of silicate polymerization as characterized by the Qⁿ distribution and abundance.

6.1. Liquidus Relations and Mixing Behavior of Aluminosilicate Melts

In the compositional range of aluminosilicate systems where tridymite is on the liquidus, from the Van’t Hoff relation, this liquidus surface can be used to calculate the activity of SiO₂ and activity coefficient of SiO₂, ${\gamma }_{Si{O}_2}^{melt}$. From these calculations, the trajectory of ${\gamma }_{Si{O}_2}^{melt}$ depends on whether we consider meta-aluminosilicate compositions with nearly all Q⁴ species, or peralkaline, depolymerized aluminosilicate melts where Q⁴-spcies coexist with Q³ and Q² species (Fig. 28). In the latter system, ${\gamma }_{Si{O}_2}^{melt}$ varies significantly with bulk melt Al/(Al+Si) because Al³⁺ substitutes preferentially for Si⁴⁺ in Q⁴ units. In contrast, in melts along the SiO–NaAlO₂ join (meta-aluminosilicate), only Q⁴ structural units are present and the Si,Al substitution is essentially that of an ideal mixture. This is, therefore, reflected in the activity coefficient remaining equal to 1 as the Al/(Al+Si) is increased (Fig. 30).

Fig. 30.

Activity coefficient of SiO₂ in melts in the Na₂O–Al₂O₃–SiO₂ system, γSiO₂^melt, for compositions along the meta-aluminosilicate join, SiO₂–NaAlO₂ (solid line), and along the peralkaline composition join, Na₂Si₄O₉–Na₂(NaAl)₄O₉ (dashed line) from the temperature-composition trajectories of silica polymorphs on the liquidus. (modified after¹²⁾).

The non-random Al³⁺ distribution among structural units in depolymerized peralkaline aluminosilicate melts results in changes in liquidus phase relations along composition joins with constant NBO/T, but where Al/(Al+Si) varies. The join Na₂Si₂O₅–Na₂(NaAl)₂O₅ (bulk melt NBO/T=1) is an example (Fig. 31). Here, increasing Al/(Al+Si) results in disappearance of Na₂Si₂O₅ as the liquidus phase. This results in an increase in the activity of both Q⁴ and Q² structural units in the melts. As a result, the liquidus volume of the NaSiO₃ phase (with Q² structure) expands.

Fig. 31.

Liquidus phase relations, at ambient pressure along the join Na₂Si₂O₅–Na₂(NaAl)₂O₅ (Liquidus relations after^2_). Chemical formulae denote crystalline phase on the liquidus in the composition range indicated.

6.2. Viscosity

The viscosity of Al₂O₃ melt is non-Ahrrenian.^{19,104,105,106)} The curvature in the viscosity versus 1/T relationship (Fig. 32(A)) is consistent with multiple AlO_n-species in Al₂O₃ melt, the abundance of which depends on temperature. An example such a temperature-controlled structural evolution is the temperature-dependent average coordination number for Al³⁺ in Al₂O₃ above its liquidus temperature.^{37,38,40,41,42)} In contrast, the Ahrrenian nature of the viscosity of liquid SiO₂ is consistent with little of no structural change of SiO₂ melt with temperature.

Fig. 32.

Viscosity of Al₂O₃ (A) and SiO₂ (B) melt as a function of temperature. Al₂O₃ data redrawn after.^105,109) Data sources for SiO₂ melt indicated on the diagram.

Substitution of charge-balanced Al³⁺ for Si⁴⁺ weakens the (Si,Al)–O bridging oxygen bonds, which means that any property that depends on (Si,Al)–O bond strength reflects this substitution. This can be seen in decreasing viscosity and activation energy of viscous flow of meta-aluminosilicate melts with increasing Al/(Al+Si)^{21,25,107,108)} (Fig. 5). The magnitude of this viscosity change varies, however, with the electronic properties of the metal cation that serves to charge-compensate for Al³⁺.^{21,25,106,107,108)} This effect, in turn, reflects in part the fact that the strength of bridging Al–O bonds is negatively correlated with the ionization potential of the charge-compensating metal cation.

Melt viscosity can also be correlated with its degree of polymerization, NBO/T. The well-known maximum in melt viscosity at or near the meta-aluminosilicate composition in aluminosilicate melt (Fig. 33) is because of the relationship between viscosity and melt NBO/T.¹⁰⁹⁾ This maximum to shifts, however, slightly to the peraluminous side of the join (Fig. 33; see also²¹⁾). Most recent structural models for peraluminous aluminosilicate melts have invoked Al-triclusters, the formation of which involves release of charge-compensating Na⁺ cation, and, therefore, the shift of the viscosity maximum to the peralumnious side of the SiO₂–NaAlO₂ join.¹⁰⁹⁾ The greater importance of triclusters in SiO₂–CaAl₂O₄ and SiO₂–MgAl₂O₄ melts.^73,74) compared with SiO₂–NaAlO₂ melts may be at least partly responsible for the diminished importance of viscosity maxima in melts in the CaO–Al₂O₃–SiO₂ and MgO–Al₂O₃–SiO₂ systems.

Fig. 33.

Compositional position of the viscosity maximum in Na₂O–Al₂O₃–SiO₂ melt at 1600°C as a function of composition (redrawn after²¹).

7. Concluding Remarks

Solution of Al₂O₃ in silicate melts and glasses affects properties such as chemical durability, glass hardness, melt and glass viscosity, thermal expansion, and compressibility. These and other properties reflect how solution of Al³⁺ in silicate melt and glass causes significant structural changes.

In peralkaline and meta-aluminous melts, essentially all Al³⁺ (>95%) occupy tetrahedral coordination. For peraluminous melts, on the other hand, complex mixtures of aluminum triclusters with 4-fold coordinated Al³⁺ and structural entities with Al³⁺ in 5- and 6-fold coordination with oxygen describe the structure. With alkali metals (M⁺) for electrical charge-balance of Al³⁺ in tetrahedral coordination, the proportions are M⁺=Al³⁺. The overall structure is dominated by three-dimensionally interconnected tetrahedra to form 6-membered rings of tetrahedra. The Al/(Al+Si) of these tetrahedra are simple positive functions of the bulk melt Al/(Al+Si). When tetrahedrally-coordinated Al³⁺ is charge-balanced by divalent cations (M²⁺), the M²⁺ cation charge-balances 2Al³⁺ tetrahedrally coordinated cations. This structure is dominated by SiO₄, (Si,Al)O₄, and AlO₄ entities.

In peralkaline aluminosilicate melts, there coexist discrete structural units with different degree of silicate polymerization. These units are termed Qⁿ-species where the superscript, n, is the number of bridging oxygen in individual units. Equilibria among these units are of the type, 2Qⁿ = Qⁿ⁺¹ + Qⁿ⁻¹. In these melts, Al³⁺ is distributed among these units. The Al³⁺ in peralkaline aluminosilicate melts strong preference Q⁴ units. This preference is, however, temperature-dependent as reflected in changes in the ΔH of the Qⁿ-species reaction. This structural behavior has consequences for transport properties of peralkaline aluminosilicate melts. It will also affect liquidus phase relations and mineral/melt element partitioning.

References

• 1)   F. Y.  Galakhov,  V. I.  Aver’yanov,  V. T.  Vavilonova and  M. P.  Areshev: Sov. J. Glass Phys. Chem., 11 (1985), 712.
• 2)   E. F.  Osborn and  A.  Muan: Phase Equilibrium Diagrams of Oxide Systems, Plate 4. The System Na₂O-Al₂O₃-SiO₂, American Ceramic Society, Columbus, OH, (1960), Plate 4.
• 3)   E. F.  Osborn and  A.  Muan: Phase Equilibrium Diagrams of Oxide Systems, Plate 3. The System MgO-Al₂O₃-SiO₂, American Ceramic Society, Columbus, OH, (1960), Plate 3.
• 4)   E. F.  Osborn and  A.  Muan: Phase Equilibrium Diagrams of Oxide Systems, Plate 2. The System CaO-Al₂O₃-SiO₂, American Ceramic Society, Columbus, OH, (1960), Plate 2.
• 5)   A.  Navrotsky,  G.  Peraudeau,  P.  McMillan and  J. P.  Coutures: Geochim. Cosmochim. Acta, 46 (1982), 2039.
• 6)   B. N.  Roy and  A.  Navrotsky: J. Am. Ceram. Soc., 67 (1984), 606.
• 7)   P.  Richet and  Y.  Bottinga: Geochim. Cosmochim. Acta, 48 (1984), 453.
• 8)   G. L.  Hovis,  M. J.  Toplis and  P.  Richet: Chem. Geol., 213 (2004), 173.
• 9)   A.  Navrotsky: Physics and Chemistry of Earth Materials, Cambridge University Press, Cambridge, UK, (1994), 417.
• 10)   V. L.  Stolyarova,  S. I.  Shornikov and  M. M.  Shultz: High Temp. Mater. Sci., 36 (1996), 15.
• 11)   J. A.  Tangeman and  R. A.  Lange: Phys. Chem. Miner., 26 (1998), 83.
• 12)   B. O.  Mysen and  P.  Richet: Silicate Glasses and Melts, 2nd ed., Elsevier, Amsterdam, (2019), 708.
• 13)   G. R.  Robinson and  J. L.  Haas: Am. Mineral., 68 (1983), 541.
• 14)   P.  Richet,  P. A.  Robie and  B. S.  Hemingway: Geochim. Cosmochim. Acta, 57 (1993), 2751.
• 15)   S. L.  Webb: Chem. Geol., 256 (2008), 92.
• 16)   S. L.  Webb: Eur. J. Mineral., 23 (2011), 487.
• 17)   P.  Richet and  D. R.  Neuville: Thermodynamic Data; Systematics and Estimation, ed. by S. K. Saxena, Springer, New York, (1992), 132.
• 18)  J. O’M. Bockris, J. D. Mackenzie and J. A. Kitchner: Trans. Faraday Soc., 51 (1955), 1734.
• 19)   G.  Urbain,  Y.  Bottinga and  P.  Richet: Geochim. Cosmochim. Acta, 46 (1982), 1061.
• 20)   P.  Richet: Geochim. Cosmochim. Acta, 48 (1984), 471.
• 21)   M. J.  Toplis,  D. B.  Dingwell and  T.  Lenci: Geochim. Cosmochim. Acta, 61 (1997), 2605.
• 22)   G. E.  Brown,  G. V.  Gibbs and  P. H.  Ribbe: Am. Mineral., 54 (1969), 1044.
• 23)   D. B.  Dingwell: Geochim. Cosmochim. Acta, 50 (1986), 1261.
• 24)   B. O.  Mysen,  A.  Lucier and  G. D.  Cody: Am. Mineral., 88 (2003), 1668.
• 25)   E. F.  Riebling: J. Chem. Phys., 44 (1966), 2857.
• 26)   S. L.  Webb,  E.  Muller and  H.  Büttner: Am. Mineral., 89 (2004), 812.
• 27)   Y.  Bottinga,  D. F.  Weill and  P.  Richet: Geochim. Cosmochim. Acta, 46 (1982), 909.
• 28)   S.  Takahashi,  D. R.  Neuville and  H.  Takebe: J. Non-Cryst. Solids, 411 (2015), 5.
• 29)   I. A.  Aksay,  J. A.  Pask and  R. F.  Davis: J. Am. Ceram. Soc., 62 (1979), 332.
• 30)   E. E.  Shpil’rain,  K. A.  Yakimovich and  A. F.  Tsitsarkin: High Temp. High Press., 5 (1973), 191.
• 31)   P.  Richet and  Y.  Bottinga: Rev. Geophys. Space Phys., 24 (1986), 1.
• 32)   R.  Ahuja,  A. B.  Belonoshko and  B.  Johansson: Phys. Rev. E, 57 (1998), 1673.
• 33)   G.  Gutiérrez and  B.  Johansson: Phys. Rev. B, 65 (2002), 104202.
• 34)   M.  Hemmati,  M.  Wilson and  P. A.  Madden: J. Phys. Chem. B, 103 (1999), 4023.
• 35)   D. K.  Belashchenko and  A. D.  Fedko: Inorg. Mater., 28 (1992), 85.
• 36)   S.  Ansell,  S.  Krishnan,  J. K. R.  Weber,  J. J.  Felten,  P. C.  Nordine,  M. A.  Beno,  D. L.  Price and  M.-L.  Saboungi: Phys. Rev. Lett., 78 (1997), 464.
• 37)   P.  Florian,  D.  Massiot,  B.  Poe,  I.  Farnan and  J.-P.  Coutures: Solid State Nucl. Magn. Reson., 5 (1995), 233.
• 38)   C.  Bessada,  V.  Lacassagne,  D.  Massiot,  P.  Florian,  J.-P.  Coutures,  E.  Robert and  B.  Gilbert: Z. Naturforsch. A, 54 (1999), 162.
• 39)   C.  Landron,  L.  Hennet,  T. E.  Jenkins,  G. N.  Greaves,  J.-P.  Coutures and  A. K.  Soper: Phys. Rev. Lett., 86 (2001), 4839.
• 40)   A.  Winkler,  J.  Horbach,  W.  Kob and  K.  Binder: J. Chem. Phys., 120 (2004), 384.
• 41)   P.  Pfleiderer,  J.  Horbach and  K.  Binder: Chem. Geol., 229 (2006), 186.
• 42)   V. V.  Hoang,  N. N.  Linh and  N. H.  Hung: Eur. Phys. J. Appl. Phys., 37 (2007), 111.
• 43)   D.  Kato: J. Appl. Phys., 47 (1976), 5344.
• 44)   P.  McMillan and  B.  Piriou: J. Non-Cryst. Solids, 53 (1982), 279.
• 45)   S.  Sen and  R. E.  Youngman: J. Phys. Chem. B, 108 (2004), 7557.
• 46)   M.  Okuno,  N.  Zotov,  M.  Schmucker and  H.  Schneider: J. Non-Cryst. Solids, 351 (2005), 1032.
• 47)   M. C.  Wilding,  C. J.  Benmore and  J. K. R.  Weber: J. Phys. Chem. B, 114 (2010), 5742.
• 48)   B. T.  Poe,  P. F.  McMillan,  B.  Cote,  D.  Massiot and  J.-P.  Coutures: J. Phys. Chem., 96 (1992), 8220.
• 49)   S. H.  Risbud and  J. A.  Pask: J. Am. Ceram. Soc., 60 (1977), 418.
• 50)   P. K.  Sato,  P. F.  McMillan,  P.  Dennison and  R.  Dupree: J. Phys. Chem., 95 (1991), 4483.
• 51)   T.  Uchino and  T.  Yoko: J. Phys. Chem. B, 102 (1998), 8372.
• 52)   A. N.  Cormack and  J.  Du: J. Non-Cryst. Solids, 293–295 (2001), 283.
• 53)   S.  Ispas,  M.  Benoit,  P.  Jund and  R.  Jullien: J. Non-Cryst. Solids, 307–310 (2002), 946.
• 54)   L.  Cormier,  D.  Ghaleb,  D. R.  Neuville,  J.-M.  Delaye and  G.  Calas: J. Non-Cryst. Solids, 332 (2003), 255.
• 55)   S. K.  Lee and  J. F.  Stebbins: Geochim. Cosmochim. Acta, 67 (2003), 1699.
• 56)   L.  Cormier and  D. R.  Neuville: Chem. Geol., 213 (2004), 103.
• 57)   K. L.  Geisinger,  G. V.  Gibbs and  A.  Navrotsky: Phys. Chem. Miner., 11 (1985), 266.
• 58)   M.  Taylor and  G. E.  Brown: Geochim. Cosmochim. Acta, 43 (1979), 61.
• 59)   S. K.  Lee and  J. F.  Stebbins: Am. Mineral., 84 (1999), 937.
• 60)   S. K.  Lee and  J. F.  Stebbins: Geochim. Cosmochim. Acta, 70 (2006), 4275.
• 61)   A.  Navrotsky,  K. L.  Geisinger,  P.  McMillan and  G. V.  Gibbs: Phys. Chem. Miner., 11 (1985), 284.
• 62)   S. K.  Lee and  J. F.  Stebbins: Geochim. Cosmochim. Acta, 66 (2002), 303.
• 63)   C. A.  Scamehorn and  C. A.  Angell: Geochim. Cosmochim. Acta, 55 (1991), 721.
• 64)   L. L.  Hench and  J. K.  West: Annu. Rev. Mater. Sci., 25 (1995), 37.
• 65)   E.  Bourova and  P.  Richet: Geophys. Res. Lett., 25 (1998), 2333.
• 66)   T.  Charpentier,  P.  Kroll and  F.  Mauri: J. Phys. Chem. C, 113 (2009), 7917.
• 67)   F. A.  Seifert,  B. O.  Mysen and  D.  Virgo: Am. Mineral., 67 (1982), 696.
• 68)   D. M.  Zirl and  S. H.  Garofalini: J. Am. Ceram. Soc., 73 (1990), 2848.
• 69)   J. D.  Kubicki and  D.  Sykes: Am. Mineral., 78 (1993), 253.
• 70)   B.  Himmel,  J.  Weigelt,  T.  Gerber and  M.  Nofz: J. Non-Cryst. Solids, 136 (1991), 27.
• 71)   I.  Daniel,  P.  Gillet,  P. F.  McMillan and  P.  Richet: Mineral. Mag., 59 (1995), 25.
• 72)   D. R.  Neuville,  L.  Cormier and  D.  Massiot: Geochim. Cosmochim. Acta, 68 (2004), 5071.
• 73)   D. R.  Neuville,  L.  Cormier,  V.  Montouillout,  P.  Florian,  F.  Millot,  J. C.  Rifflet and  D.  Massiot: Am. Mineral., 93 (2008), 1721.
• 74)   L. M.  Thompson and  J. F.  Stebbins: Am. Mineral., 96 (2011), 841.
• 75)   H.  Maekawa,  T.  Maekawa,  K.  Kawamura and  T.  Yokokawa: J. Non-Cryst. Solids, 127 (1991), 53.
• 76)   S. B.  Liu,  A.  Pines,  M.  Brandriss and  J. F.  Stebbins: Phys. Chem. Miner., 15 (1987), 155.
• 77)   P. F.  McMillan,  B. T.  Poe,  P.  Gillet and  B.  Reynard: Geochim. Cosmochim. Acta, 58 (1994), 3653.
• 78)   D. R.  Neuville and  B. O.  Mysen: Geochim. Cosmochim. Acta, 60 (1996), 1727.
• 79)   G. V.  Gibbs,  E. P.  Meagher,  M. D.  Newton and  D. K.  Swanson: Structure and Bonding in Crystals, Chapter 9, ed. by M. O’Keefe and A. Navrotsky, Academic Press, New York, (1981), 195.
• 80)   J.  Etchepare: Amorphous Materials, ed. by R. W. Douglas and E. Ellis, Wiley-Interscience, New York, (1972), 337.
• 81)   J. F.  Stebbins,  E. V.  Dubinsky,  K.  Kanehashi and  K. E.  Kelsey: Geochim. Cosmochim. Acta, 72 (2008), 910.
• 82)   I.  Kushiro: Am. J. Sci., 267–A (1969), 269.
• 83)   P.  Richet,  J.  Ingrin,  B. O.  Mysen,  P.  Courtial and  P.  Gillet: Earth Planet. Sci. Lett., 121 (1994), 589.
• 84)   S. A.  Brawer and  W. B.  White: J. Chem. Phys., 63 (1975), 2421.
• 85)   S. A.  Brawer and  W. B.  White: J. Non-Cryst. Solids, 23 (1977), 261.
• 86)   B. O.  Mysen,  D.  Virgo and  F. A.  Seifert: Rev. Geophys. Space Phys., 20 (1982), 353.
• 87)   J. F.  Stebbins: Nature, 330 (1987), 465.
• 88)   W.-A.  Buckermann,  W.  Muller-Warmuth and  G. H.  Frischat: Glasstechnische Bericht, 65 (1992), 18.
• 89)   P.  McMillan: Am. Mineral., 69 (1984), 622.
• 90)   G.  Engelhardt and  D.  Michel: High-Resolution Solid-State NMR of Silicates and Zeolites, Wiley, New York, (1987), 485.
• 91)   J. B.  Murdoch,  J. F.  Stebbins and  I. S. E.  Carmichael: Am. Mineral., 70 (1985), 332.
• 92)   R.  Oestrike,  W.-H.  Yang,  R. J.  Kirkpatrick,  R.  Hervig,  A.  Navrotsky and  B.  Montez: Geochim. Cosmochim. Acta, 51 (1987), 2199.
• 93)   E.  Schneider,  J. F.  Stebbins and  A.  Pines: J. Non-Cryst. Solids, 89 (1987), 371.
• 94)   X.  Xue,  J. F.  Stebbins and  M.  Kanzaki: Am. Mineral., 79 (1994), 31.
• 95)   F.  Mauri,  A.  Pasquarello,  B. G.  Pfrommer,  Y.-G.  Yoon and  S. G.  Louie: Phys. Rev. B, 62 (2000), R4786.
• 96)   B. O.  Mysen: Contrib. Mineral. Petrol., 127 (1997), 104.
• 97)   C. C.  Lin,  S. F.  Chen,  L. G.  Liu and  C. C.  Li: Mater. Chem. Phys., 123 (2010), 569.
• 98)   J. F.  Stebbins: J. Non-Cryst. Solids, 106 (1988), 359.
• 99)   M. E.  Brandriss and  J. F.  Stebbins: Geochim. Cosmochim. Acta, 52 (1988), 2659.
• 100)   P. F.  McMillan,  G. H.  Wolf and  B. T.  Poe: Chem. Geol., 96 (1992), 351.
• 101)   B. O.  Mysen and  J. D.  Frantz: Am. Mineral., 78 (1993), 699.
• 102)   B. O.  Mysen: Geochim. Cosmochim. Acta, 63 (1999), 95.
• 103)   B. O.  Mysen and  J. D.  Frantz: Contrib. Mineral. Petrol., 117 (1994), 1.
• 104)   G.  Urbain: Rev. Int. Hautes Temp. Refract., 19 (1982), 55.
• 105)   G.  Urbain,  Y.  Bottinga and  P.  Richet: Geochim. Cosmochim. Acta, 46 (1982), 1061.
• 106)   D.  Langstaff,  M.  Gunn,  G. N.  Greaves,  A.  Marsing and  F.  Kargl: Rev. Sci. Instrum., 84 (2013), 124901.
• 107)   S.  Sukenaga,  N.  Saito,  K.  Kawakami and  K.  Nakashima: ISIJ Int., 46 (2006), 352.
• 108)   E. F.  Rlebling: Can. J. Chem., 42 (1964), 2811.
• 109)   B. O.  Mysen and  M. J.  Toplis: Am. Mineral., 92 (2007), 933.
• 110)   B. S.  Mitin and  Y. A.  Nagibin: Russ. J. Phys. Chem., 44 (1970), 741.
• 111)   Y.  Bottinga and  D. F.  Weill: Am. J. Sci., 269 (1970), 169.
• 112)   Y.  Bottinga and  P.  Richet: Geochim. Cosmochim. Acta, 59 (1995), 2725.
• 113)   J. W.  Tomlinson,  M. S. R.  Heynes and  J. O. M.  Bockris: Trans. Faraday Soc., 54 (1958), 1822.
• 114)   V.  Askarpour,  M. H.  Manghnani and  P.  Richet: J. Geophys. Res., 98 (1993), 17683.
• 115)   Y.  Marumo and  M.  Okuno: Materials Science of the Earth’s Interior, ed. by I. Sunagawa, Terra Scientific, Tokyo, (1984), 25.
• 116)   P.  Richet and  Y.  Bottinga: Structure, Dynamics and Properties of Silicate Melts, ed. by J. F. Stebbins et al., Mineralogical Society of America, Washington, D.C., (1995), 67.