A computer simulation study of natural silicate melts. Part II: High pressure properties

The thermodynamic, structural and transport properties of natural silicate melts under pressure are investigated by molecular dynamics simulation with the help of a force field recently introduced by us [Guillot B. and Sator N. (2007) A computer simulation study of natural silicate melts. Part I: low pressure properties. Geochim. Cosmochim. Acta71, 1249-1265]. It is shown that the simulation reproduces accurately the bulk moduli of a large variety of silicate liquids as evaluated from ultrasonic studies. The equations of state (EOS) of the simulated melts are in good agreement with the density data on mid-ocean ridge basalt, komatiite, peridotite and fayalite as obtained either by sink/float experiments or by shock-wave compression. From the structural point of view it is shown that the population of ^[5]Al and ^[6]Al species increases rapidly upon initial compression (0-50 kbar) whereas for Si these highly coordinated species are found in a significant abundance (>5%) only above ∼50 kbar for ^[5]Si and ∼100-150 kbar for ^[6]Si. This increase of the coordination of network formers is not the only response of the melt structure to the densification: there is also a large redistribution of the T-O-T (T = Si, Al) bond angles with the pressure and noticeably upon initial compression in rhyolitic and basaltic liquids. Furthermore, a detailed analysis of the population of bridging oxygens (BO) and nonbridging oxygens (NBO) points out that the polymerization of the melt generally increases when the pressure increases, the magnitude of this polymerization enhancement being all the more important that the melt is depolymerized at low pressure. The role of triclusters (threefold coordinated oxygens to network former cations) is particularly emphasized in acidic and basaltic liquids. The pressure-induced redistribution of the oxygen atoms through the melt structure is also stressed. Finally, the simulation predicts a nonmonotonic behavior of the diffusivity of network former ions when the pressure increases, a feature with depends on the melt composition. This could have a counterpart in the electrical conductivity at sufficiently high temperature when the viscosity of the liquid is low.     

A computer simulation study of natural silicate melts. Part I: Low pressure properties

In implementing into a molecular dynamics simulation code a simple interionic potential developed to describe the nine component system K₂O-Na₂O-CaO-MgO-FeO-Fe₂O₃-Al₂O₃-TiO₂-SiO₂ (KNCMFATS), it has been possible to reproduce satisfactorily a number of thermodynamic, structural and transport properties of a representative set of natural silicate melts. An important conclusion reached in this study is the good transferability of the potential from felsic to ultramafic compositions although this transferability becomes less accurate with high silica contents (rhyolitic composition and beyond) and with very iron-rich silicates (e.g. fayalite). A key feature of the simulation is to make the link between macroscopic properties of the melt and its microscopic structure and dynamics. We thus obtain a relationship between the molar volume of the melt, the number of network modifiers and the oxygen coordination number. The simulation also allows one to quantify the coordination environment around the cations as function of the melt composition. Furthermore, the electrical conductivity of the high temperature liquid is investigated.     

Carbon dioxide and argon diffusion in silicate melts: Insights into the CO2 speciation in magmas

To investigate the influence of temperature and composition on the diffusivities of dissolved carbon dioxide and argon in silicate melts, diffusion experiments were performed at magmatic pressure and temperature conditions in (a) albite melts with excess Na₂O (0-8.6 wt%) and a constant Si/Al ratio of 3, and (b) albite₇₀quartz₃₀ to jadeite melts with decreasing SiO₂ content and a constant Na/Al ratio of 1. We obtained diffusion coefficients at 500 MPa and 1323-1673 K. In the fully polymerized system Ab₇₀Qz₃₀ - Jd, the change in composition only has a weak effect on bulk CO₂ diffusivity, but Ar diffusivity increases clearly with decreasing SiO₂ content. In the system Ab + Na₂O, bulk CO₂ and Ar diffusivity increase significantly with gradual depolymerisation. The relatively small change in composition on molar basis in the depolymerized system leads to a significantly larger change in diffusivities compared to the fully polymerized Ab₇₀Qz₃₀-Jd join. Within error, activation energies for bulk CO₂ and Ar diffusion in both systems are identical with decreasing silica content (Ab + Na₂O: 159 ± 25 kJ mol⁻¹ for bulk CO₂ and 130 ± 8 kJ mol⁻¹ for Ar; Ab₇₀Qz₃₀-Jd: 163 ± 16 kJ mol⁻¹ for bulk CO₂ and 148 ± 15 kJ mol⁻¹ for Ar) even though this results in depolymerisation in one system and not the other.Although there is a variation in CO₂ speciation with changing composition as observed in quenched glasses, it has previously established that this is not a true representation of the species present in the melt, with the ratio of molecular CO₂ to carbonate decreasing during quenching. Thus, diffusion coefficients for the individual CO₂ species cannot be directly derived by measuring molecular CO₂ and CO₃^2- concentration-distance profiles in the glasses. To obtain diffusivities of individual CO₂ species, we have made two assumptions that (1) inert Ar can be used as a proxy for molecular CO₂ diffusion characteristics as shown by our previous work and (2) the diffusivity of CO₃²⁻ can be calculated assuming it is identical to network forming components (Si⁴⁺ and Al³⁺). This is derived from viscosity data (Eyring eqn.) and suggests that CO₃²⁻ diffusion would be several orders of magnitude slower than molecular CO₂ diffusion.The systematics of measured bulk CO₂ diffusivity rates and comparison with the Ar proxy all suggest that the faster molecular CO₂ species is much more dominant in melts than measurements on resulting quenched glasses would suggest. This study has confirmed an observation of surprisingly consistent bulk CO₂ diffusivity across a range of natural compositions were Ar diffusivity significantly increases. This is consistent with an actual increase in molecular CO₂ mobility (similar to Ar) that is combined with an increase in the proportion of the slower carbonate in the melt.These results demonstrate that the CO₂ diffusion and speciation model provides an insight into the transport processes in the melt and is promising and an alternative tool to in situ speciation measurements at magmatic conditions, which at the moment are technically extremely difficult. We present the first high pressure high temperature in situ MIR spectra of a CO₂ bearing albitic glass/melt suggesting that molecular CO₂ is a stable species at high temperature, which is qualitatively consistent with the modelled CO₂ speciation data.     

Structure of silicate melts: Raman spectroscopic data and thermodynamic simulation results

High-temperature Raman spectroscopy was applied to study model silicate melts in the M₂O-SiO₂ system, where M = K, Na, or Li. Structural units of the melts and equilibrium constants of reactions between them are determined. Thermodynamic calculations were conducted for the dependence of the Q ⁿ distribution on the composition and temperature, and the results of thermodynamic simulations were demonstrated to be consistent with the experimental results.     

Transition metal ions in silicate melts—I. Manganese in sodium silicate melts

Optical absorption spectra obtained on glasses quenched from sodium silicate melts show Mn³⁺ to be the dominant species for melts heated in air and Mn²⁺ to be the dominant species for melts heated at Po₂ = 10^?17 bar. The absorption spectrum of Mn³⁺ consists of an intense band at 20,000cm^?1 with a 15,000cm^?1 satellite possibly arising from the Jahn-Teller effect. The independence of the spectrum from melt composition and the high band intensity is offered as evidence for a distinct Mn³⁺ complex in the melt. The spectrum of Mn²⁺ is weak and many expected bands are not observed. A two-band luminescence spectrum from Mn²⁺ has been tentatively interpreted as due to Mn²⁺ in interstitial sites in the network and Mn²⁺ coordinated by non-bridging oxygens.     

Infrared spectroscopic study of the silicate anionic structures of some magmatic silicate melts

Summary The relation between the species of silicate anions in a silicate melt and their infrared characteristic frequency is discussed. A simple relation approximated with a quadratic equation is established between the ratio of non-bridging oxygens to silicon atoms of a silicate anion and the characteristic frequency. Based on this relation, the silicate anions in some magmatic silicate melts were estimated. The results obtained in the present study agreed well qualitatively with those estimated by some researchers on the basis of other spectroscopic methods.The constituent silicate segments and the distribution of AlO₄ tetrahedra in fully polymerized melts, albite, jadeite and anorthite, have been investigated. The polymerization degree of the silicate segments in the network seemed to decrease with the increase of AlO₄ tetrahedron. It is also suggested, that AlO₄ in the albite melt distributed more randomly, while those in anorthite melt had a tendency to form an aluminous segment.

---

Infrarot-spektroskopische Untersuchungen der Silikatanionen-Strukturen einiger magmatischer Silikatschmelzen

Zusammenfassung Die Beziehungen zwischen den Spezies der Silikatanionen in den Silikatschmelzen und ihrer charakteristischen Infrarot-Frequenz wurde untersucht. Es gibt eine einfache Beziehung zwischen dem Verhältnis nicht brückenbildender Sauerstoffatome zu Si-Atomen der Silikatanionen und der charakteristischen Frequenz, die mit einer quadratischen Gleichung beschrieben werden kann. Auf Grund dieser Beziehung werden die Silikatanionen in einigen magmatischen Schmelzen abgeschätzt. Das Ergebnis dieser Untersuchung stimmt qualitativ mit den Werten überein, die von einigen Forschern durch andere spektroskopische Methoden erhalten wurden.Die Silikatsegmente und die Verteilung der AlO₄-Tetraeder in vollkommen polymerisierten Schmelzen, in Albit, Jadeit und Anorthit wurden untersucht. Der Polymerisationsgrad der Silikatsegmente in den Netzwerken scheint mit der Zunahme der AlO₄-Tetraeder geringer zu werden. Die Untersuchungläßt erkennen, daß AlO₄-Tetraeder in Albit unregelmäßig verteilt sind, und in Anorthit die Neigung haben, ein Al-Segment zu bilden.

---

---

With 6 Figures     

A review and analysis of silicate mineral dissolution experiments in natural silicate melts

We present a database and a graphical analysis of published experimental results for dissolution rates of olivine, quartz plagioclase, clinopyroxene, orthopyroxene, spinel, and garnet in basaltic and andesitic melts covering a range of experimental temperatures (1100–1500°C) and pressures (10⁵ Pa-3.0 GPa). The published datasets of Donaldson (1985, 1990) and Brearly and Scarfe (1986) are the most complete. Experimental dissolution rates from all datasets are recalculated and normalized to a constant oxygen basis to allow for direct comparison of dissolution rates between different minerals. Dissolution rates (ν) range from 5·10⁻¹⁰ oxygen equivalent moles (o.e.m.) cm⁻² s⁻¹ for olivine in a basaltic melt to 1.3·10⁻⁵ o.e.m. cm⁻² s⁻¹ for garnet in a basaltic melt. Values of ln ν are Arthenian for the experiments examined and activation energies range from 118 to 1800 kJ/o.e.m. for quartz and clinopyroxene, respectively.

The relationship between calculated A/RT for the dissolution reactions, where A is the thermodynamic potential affinity, and values of ν is linear for olivine, plagioclase, and quartz. We interpret this as strong evidence in support of using calculated A as a predictor of ν for, at least, superliquidus melt conditions.     

Carbon isotopic fractionation between CO2 vapour,silicate and carbonate melts: an experimental study to 30 kbar

The carbon isotopic fractionation between CO₂ vapour and sodamelilite (NaCaAlSi₂O₇) melt over a range of pressures and temperatures has been investigated using solid-media piston-cylinder high pressure apparatus. Ag₂C₂O₄ was the source of CO₂ and experimental oxygen fugacity was buffered at hematite-magnetite by the double capsule technique. The abundance and isotopic composition of carbon dissolved in sodamelilite (SM) glass were determined by stepped heating and the ¹³C of coexisting vapour was determined directly by capsule piercing. CO₂ solubility in SM displays a complex behavior with temperature. At pressures up to 10 kbars CO₂ dissolves in SM to form carbonate ion complexes and the solubility data suggest slight negative temperature dependence. Above 20 kbars CO₂ reacts with SM to form immiscible Na-rich silicate and Ca-rich carbonate melts and CO₂ solubility in Na-enriched silicate melt rises with increasing temperature above the liquidus. Measured values for carbon isotopic fractionation between CO₂ vapour and carbonate ions dissoived in sodamelilite melt at 1200°–1400° C and 5–30 kbars average 2.4±0.2, favouring¹³C enrichment in CO₂ vapour. The results are maxima and are independent of pressure and temperature. Similar values of 2 are obtained for the carbon isotopic fractionation between CO₂ vapour and carbonate melts at 1300°–1400° C and 20–30 kbars.     

Modeling diffusive dissolution in silicate melts

Here empirical models for calculating self-diffusion coefficients and diffusion matrices are combined with MELTS, a thermodynamic model for silicate minerals and melts, to estimate diffusive dissolution rates, interface melt compositions and melt diffusivities. Simulations of olivine dissolution experiments in basalt show that the overall model is capable of accurately reproducing diffusive dissolution rates, and the resulting diffusion profiles, over a range of pressures and temperatures. However, the overall model is less successful at reproducing olivine dissolution in andesite, diopside dissolution in either basalt or andesite, or anorthite dissolution in picrite. Yet, even for these systems the predicted dissolution rates are generally within about a factor of two of the measured ones. Comparisons between simulations and experiments suggest that errors in the self-diffusion and thermodynamic models are responsible for the differences, and show that dissolution experiments could be a powerful way of testing and calibrating these and similar models. The overall model will also be a useful tool for designing future experiments, and for identifying the parameters that control diffusive dissolution (and crystallization) in silicate melts under a wide range of conditions.     

高温硅酸盐熔体粘度与网络分数维值的相关性研究

对高温硅酸盐熔体粘度的估算一直是国际地学界热点问题之一,本文在研究了高温硅酸盐熔体网络分数维值的基础上,建立了估算熔体粘度的新模式 (简称 FD模式 ) ,阐明了熔体粘度值除了与温度成反比外,还与分子网络介观尺度的自相似比 ri 和分数维值 D相关,即与分子网络 (纳米尺度以上 )中的桥氧数 Ni成正比 (Ni∝ ) , 与单位硅氧四面体中的非桥氧数成反比.经对 4个硅酸盐系列高温熔体的粘度测定证实, FD模式的理论计算值与实测值吻合,且优于现今国际通用的 VTF模式.为探索岩浆迁移演化规律和完善新型低维材料的性能提供了新的理论依据.     

Activities of trace elements in silicate melts

A theoretical model has been developed which describes the amphoteric character of oxides in silicate melts. This has been used to account for the increased stability of the higher oxidation states of altervalent trace elements in silicate melts with increasing basicity and to derive a general expression for the estimation of trace element activities in silicate melts.     

The speciation of water in silicate melts

Previous models of water solubility in silicate melts generally assume essentially complete reaction of water molecules to hydroxyl groups. In this paper a new model is proposed that is based on the hypothesis that the observed concentrations of molecular water and hydroxyl groups in hydrous silicate glasses reflect those of the melts from which they were quenched. The new model relates the proportions of molecular water and hydroxyl groups in melts via the following reaction describing the homogeneous equilibrium between melt species: H₂O_molecular (melt) + oxygen (melt) = 2OH (melt). An equilibrium constant has been formulated for this reaction and species are assumed to mix ideally. Given an equilibrium constant for this reaction of 0.1–0.3, the proposed model can account for variations in the concentrations of molecular water and hydroxyl groups in melts as functions of the total dissolved water content that are similar to those observed in glasses. The solubility of molecular water in melt is described by the following reaction: H₂O (vapor) = H₂O_molecular (melt).These reactions describing the homogeneous and heterogeneous equilibria of hydrous silicate melts can account for the following observations: the linearity between f_H2O and the square of the mole fraction of dissolved water at low total water contents and deviations from linearity at high total water contents; the difference between the partial molar volume of water in melts at low total water contents and at high total water contents; the similarity between water contents of vapor-saturated melts of significantly different compositions at high pressures versus the dependence on melt composition of water solubility in silicate melts at low pressures; and the variations of viscosity, electrical conductivity, the diffusivity of “water,” the diffusivity of cesium, and phase relationships with the total dissolved water contents of melts.This model is thus consistent with available observations on hydrous melt systems and available data on the species concentrations of hydrous glasses and is easily tested, since measurements of the concentrations of molecular water and hydroxyl groups in silicate glasses quenched from melts equilibrated over a range of conditions and total dissolved water contents are readily obtainable.     

Isovalent trace element partitioning between minerals and melts: A computer simulation study

We present a new approach for the rationalisation of trace element partitioning between silicate melts and minerals, which is not based on the empirical, parameterised continuum models in common use. We calculate the energetics of ion substitution using atomistic simulation techniques, which include an explicit evaluation of the relaxation energy (strain energy) contribution to this process. Solution energies are estimated for isovalent impurities in CaO, diopside, orthoenstatite, and forsterite. These show a parabolic dependence on ionic radius, similar to the variation of mineral-melt partition coefficients with ionic radius. The success of the empirical models, which often include only the strain energy, appear to have been due to the partial cancellation of energy terms, and to the empirical fitting of the parameters included in these models. Our approach can be readily extended to aliovalent substitution.     

Growth of diamond in hydrous silicate melts

This study demonstrates that a hydrous, halide bearing silicate melt is a viable medium for diamond growth. Experiments were conducted in the MgO–SiO₂–H₂O–C ± KCl ± NaCl system, which was used as a model for harzburgitic mantle. In no case did we observe crystals that could be interpreted as spontaneously nucleated, but growth of diamond on seed crystals at 1,400–1,600°C and 7 GPa in experiments of 4 h duration was observed. The addition of KCl to the system produced crystallization of diamond at temperatures as low as 1,400°C. At higher temperatures, larger growth features were produced than those that seen in the KCl-free system at the same conditions. The NaCl-bearing system is different; in these experiments, the diamond seed crystals show evidence of possible dissolution and layer growth, albeit more subdued growth than in the KCl system. Therefore, NaCl may be an inhibitor of diamond growth in a hydrous silicate melt. Based on these results, hydrous silicate melts could play a role in formation of diamond in either deep subduction zones, or above slabs imbricated against a lithospheric ‘root’ in the sub-continental lithospheric mantle. The water and halide necessary for their formation could be transported into the mantle in hydrous phases such as serpentine in subducting lithospheric slabs. Dehydration of serpentine at >200 km depth would release hydrous, halide-bearing fluids into the overlying mantle wedge or lithospheric root, triggering melting at conditions similar to those of the formation of natural diamond.     

Rutile solubility and titanium coordination in silicate melts

The solubility of rutile has been determined in a series of compositions in the K₂O-Al₂O₃-SiO₂ system ($\text{K}{}^1\text{=}\text{K}{}_2\text{O}\text{(K}{}_2\text{O}$ + Al₂O₃) = 0.38–0.90), and the CaO-Al₂O₃-SiO₂ system ($\text{C}{}^1\text{=}\text{CaO}\text{(CaO + Al}{}_2\text{O}{}_3\text{)}\text{= 0.47–0.59}$). Isothermal results in the KAS system at 1325°C, 1400°C, and 1475°C show rutile solubility to be a strong function of the $\text{K}{}^1$ ratio. For example, at 1475°C the amount of TiO₂ required for rutile saturation varies from 9.5 wt% ($\text{K}{}^1\text{= 0.38}$) to 11.5 wt% ($\text{K}{}^1\text{= 0.48}$) to 41.2 wt% ($\text{K}{}^1\text{= 0.90}$). In the CAS system at 1475°C, rutile solubility is not a strong function of $\text{C}{}^1$. The amount of TiO₂ required for saturation varies from 14 wt% ($\text{C}{}^1\text{= 0.48}$) to 16.2 wt% ($\text{C}{}^1\text{= 0.59}$).The solubility changes in KAS melts are interpreted to be due to the formation of strong complexes between Ti and K⁺ in excess of that needed to charge balance Al³⁺. The suggested stoichiometry of this complex is K₂Ti₂O₅ or K₂Ti₃O₇. In CAS melts, the data suggest that Ca²⁺ in excess of A1³⁺ is not as effective at complexing with Ti as is K⁺. The greater solubility of rutile in CAS melts when $\text{C}{}^1$ is less than 0.54 compared to KAS melts of equal $\text{K}{}^1$ ratio results primarily from competition between Ti and Al for complexing cations (Ca vs. K).TiK_β x-ray emission spectra of KAS glasses ($\text{K}{}^1\text{= 0.43–0.60}$) with 7 mole% added TiO₂, rutile, and Ba₂TiO₄, demonstrate that the average Ti-O bond length in these glasses is equal to that of rutile rather than Ba₂TiO₄, implying that Ti in these compositions is 6-fold rather than 4-fold coordinated. Re-examination of published spectroscopic data in light of these results and the solubility data, suggests that the 6-fold coordination polyhedron of Ti is highly distorted, with at least one Ti-O bond grossly undersatisfied in terms of Pauling's rules.     

Experimental study of evaporation and isotopic mass fractionation of potassium in silicate melts

The behavior of Na and K during evaporation from chondrule composition melts was studied using a vacuum furnace. Though Na is the less volatile of the two as an element, it is lost more rapidly than K from silicate melts. Mass fractionation of K isotopes was measured by ion microprobe and Rayleigh fractionation is observed for vacuum evaporation (10⁻⁵ atm). With higher pressures of air, the K loss rate decreases but with increasing hydrogen pressure, K is lost more rapidly. δ⁴¹K decreases with higher pressures, because of back reaction between melt and K in the gas. With long heating duration, the release of light K condensed within the furnace leads to interaction with the K-depleted melt and a further reduction of δ⁴¹K. Natural chondrules differ in some ways from our experimental residues. Some (especially type IIA) have superchondritic Na and K, despite their assumed formation in nebular hydrogen, which enhances volatile loss, and chondrules do not show K isotopic fractionation. Type I chondrules in Semarkona (LL3.0) either plot on our evaporation trend, or are depleted in K but slightly enriched in Na, relative to K. In Bishunpur (LL3.1), type I chondrules are mostly K-depleted but moderately to strongly enriched in Na. In petrologic type 3.2 to 3.4 chondrites they are enriched in both K and Na, like type II chondrules. The alkali contents suggest type I chondrules experienced evaporation and subsequent metasomatism. Their normal δ⁴¹K values suggest closed-system evaporation of a chondritic precursor in a gas with relatively high K pressures due to vaporization of dust accompanying chondrule precursor aggregates. Type II chondrules are volatile-rich, as well as unfractionated in K isotopes. They probably formed in a gas with higher pK than in the case of type I chondrules, due to heating of a more dust-rich parcel of gas.     

Behaviour of Silicate Melts in Respect of Volume

The volumes per oxygen of some silicate melts have been calculated and then compared with those of silicate glasses.It is suggested that the volume of a silicate melt can be divided into two parts.One is contri buted by the silicon-oxygen network and the other by the “oxides”.Variation patterns of VPOs suggest that the volume of the Si-O network generally remains unchanged and the expansion of the melt is caused mainly by the locat expansion of the “oxides”.It is further proposed that the radius of O^2- shows little variation,in striking contrast to the radius of cations.The mechanism governing the expansion is discussed in detail.     

Dual speciation of nitrogen in silicate melts at high pressure and temperature: An experimental study

Solubility and speciation of nitrogen in silicate melts have been investigated between 1400 and 1700 °C and at pressures ranging from 10 to 30 kbar for six different binary alkali and alkaline-earth silicate liquids and a Ca-Mg-alumino silicate. Experiments were performed in a piston-cylinder apparatus. The nitrogen source is silver azide, which breaks down to Ag and molecular N₂ below 300 °C. At high pressure and temperature, the nitrogen content may be as high as 0.7 wt% depending on the melt composition, pressure, and temperature. It increases with T, P and the polymerization state of the liquid. Characterization by Raman spectroscopy and ¹⁵N solid state MAS NMR indicates that nitrogen is not only physically dissolved as N₂ within the melt structure like noble gases, but a fraction of nitrogen interacts strongly with the silicate network. The most likely nitrogen-bearing species that can account for Raman and NMR results is nitrosyl group. Solubility data follow an apparent Henry’s law behavior and are in good agreement with previous studies when the nitrosyl content is low. On the other hand, a significant departure from a Henry’s law behavior is observed for highly depolymerized melts, which contain more nitrosyl than polymerized melts. Possible solubility mechanisms are also discussed. Finally, a multi-variant empirical relation is given to predict the relative content of nitrosyl and molecular nitrogen as a function of P, T, and melt composition and structure. This complex speciation of nitrogen in melts under high pressure may have significant implication concerning crystal-melt partitioning of nitrogen as well as for potential elemental and isotopic fractionation of nitrogen in the deep Earth.