MR13B-1255
Compositional effects on aluminum and silicon coordination changes in high pressure aluminosilicate glasses: NMR results
Understanding the structure of high pressure aluminosilicate glasses can better constrain the macroscopic properties and igneous processes in the upper mantle. As melts densify, bond angles and distances change and the average coordination number (CN) for the cations increases. Both Al and Si can increase CN, and are affected by the type and amount of network modifying cation (i.e. Ca, Mg, Na, K, etc…) present. This study uses nuclear magnetic resonance (NMR) spectroscopy to investigate several aspects of the compositional controls on the structure of aluminosilicate glasses quenched from liquids at high pressure. 29Si MAS NMR spectra of 6GPa Na2Si4O9 glasses show the presence of 2 % [5]Si and 2 % [6]Si. A similar glass with 1.2 wt % Al2O3 and the same nominal NBO/T has very similar Si coordinations, even though the Al has an average CN of 5.4. Thus, although Al increases CN more readily than does Si, the addition of Al does not inhibit the formation of high coordination Si. Extending the previously investigated range of modifier cation field strength, 27Al MAS NMR spectra of LaAlSi3O9 glasses show a substantial increase in average Al CN, from about 4.21 to 5.07 from ambient P to 8GPa, while Li3AlSi3O9 glasses have consistently lower average Al CN, increasing from 4.02 to 4.48 from ambient P to 6 GPa. This agrees with previously noted trends with modifier cation field strength. However, at 6 and 8 GPa the average Al CN is less than expected in the La-containing glasses, suggesting that the La3+ environment changes as well. 17O 3QMAS NMR spectra of the Li-containing glasses are consistent with previously proposed mechanisms for the generation of [5],[6]Al at the expense of non-bridging oxygen, but the La-containing glasses have nearly unchanged oxygen environments with increasing pressure. Finally, we present 27Al MAS NMR spectra of Na+Ca-bearing aluminosilicate glasses with up to 5 wt % FeO quenched from melts at 5 GPa and conclude that the substitution of small amounts of Fe2+ for Ca2+ does not significantly affect the average Al CN.
MR13B-1256
Composition Effects on Aluminoborosilicate Glasses Structure
Quantitative determinations of the atomic-scale structure of glasses, and the effects of composition on it, are critical to the development of physically accurate models of glass and melt properties. This study is an attempt to understand the effects of composition on network connectivity of alkali and alkaline earth aluminoborosilicate glasses, used widely in technology. It is also an examination of the extent of disorder of modifier cations around non-bridging oxygens (NBO) and an evaluation of the changes in boron and aluminum speciation as a function of the ratio of modifier cations (Na/Ca). The latter outcome may also help to understand how different cation field strengths effect aluminum coordination numbers in high pressure environments. We have used B-11 and Al-27 MAS NMR to investigate the formation of different boron and aluminum species. B-coordination changes from dominantly trigonal to dominantly tetrahedral as Na/(Na+Ca) increases, but the fraction of [4]B species is not an exact linear function of composition. When multiple modifier cations are present (Na and Ca), the fraction of [4]B is lower than what would be expected from linear combinations of the appropriate end-members. [4]Al is predominant among the Al species. The small amount of [5]Al (a few percent) increases with decreasing Na/(Na+Ca). The conversion of [3]B to [4]B is also expected to convert NBO to bridging oxygens. The NBO fraction can thus be calculated directly from the compositions and the amounts of four-coordinated B and Al. The estimated NBO in these glasses thus decreases with increasing Na/(Na+Ca). We have also used Na-23 MAS NMR to investigate the Na environments. The calculated mean isotropic chemical shifts become slightly higher with increasing Na/(Na+Ca), which suggests that Na-O distances increase when more Ca substitutes for Na. Therefore, Na is in a more bridging oxygen rich environment with an increasing Ca component, with a higher proportion of NBOs surrounding Ca. Quantification of spectra from the present study shows a decrease in B- and an increase Al-coordination with increasing average field strength of the modifier cation, meanwhile the fraction of NBO is increasing. These data also suggests the Na/Ca distribution is not random, and therefore that the entropy of mixing is reduced.
MR13B-1257
Numerical Simulations of Falling Sphere Viscometry Experiments.
The falling sphere technique based on Stokes' law is widely used to determine the viscosities of geologically relevant melts at high pressures. Stokes' law is valid when a rigid sphere falls slowly and steadily through a stationary and infinite Newtonian medium of uniform properties. High-pressure falling sphere experiments however, usually involve dropping a dense, refractory sphere through a liquid contained by a cylindrical capsule of finite size. The sphere velocity is influenced by the walls (Faxen correction) and ends of the capsule, and possible convective motion of the fluid. Efforts are made to minimize thermal gradients in laboratory experiments, but small temperature differences within the capsule can lead to convection complicating interpretation. We utilize GALE (Moresi et al., 2003;), a finite element particle-in-cell code, to examine these factors in numerical models of conditions similar to those of high-pressure experiments. Our modeling considers a three- dimensional box or cylinder containing a cluster of particles that represent the dense sphere in laboratory experiments surrounded by low viscosity particles representing the melt. GALE includes buoyancy forces, heat flow, and viscosity variations so our model can be used to assess the effects of the capsule's walls and ends, and the consequences of thermal gradients on the sphere's velocity and trajectory. Comparisons between our numerical simulations and real-time falling sphere experiments involving lower viscosity molten komatiite are made to assess the validity of Stokes' law with the standard Faxen correction included, and formulations considering end effects. The modeling also permits an evaluation of the uncertainties in recovering accurate liquid viscosities from Stokes' law when a dense sphere falls through a convecting low viscosity melt. It also allows us to assess acceleration to a terminal velocity that can provide constraints on melt viscosity in experiments in which the terminal velocity was not reached.
MR13B-1258
In situ buoyancy test for the density measurement of basaltic liquid at high pressure and high temperature
In situ buoyancy test for the density measurement of basaltic liquid at high pressure and high temperature Akio Suzuki, Eiji Ohtani, Hidenori Terasaki, Tatsuya Sakamaki, Keisuke Nishida, Kenichi Funakoshi We have been carried out density measurement of silicate and metallic liquid at high pressure and high temperature by using the buoyancy test. In this method, the density of liquid was bounded by that of the sinking or floating buoyancy marker. The uncertainty of the determined density of liquid is caused by 1) the uncertainty of the equation of state of the buoyancy marker 2) the uncertainty of the pressure of experiment. In the pervious experiments, the pressures were estimated on the basis of the pressure calibration curve, which had been established from the separate experiments. Therefore, the uncertainty of pressure is about 10% and makes it difficult to obtain the precise equation of state of liquid. This serious problem can be solved if the pressure of the experimental condition is determined precisely. In the present study we carried out buoyancy test to determine the density of basaltic liquid at high pressure and high temperature. The starting material has a composition of lunar high-Ti basalt, which was synthesized by using a gas-mixing furnace. Experiments were performed at BL04B1 beamline at SPring-8, Japan. The sinking and flotation of the buoyancy markers were observed from the X-ray radiography and X-ray diffraction study. The pressure was determined by using an equation of state of MgO. Sinking of diamond was observed at 3.88+-0.03 GPa and 1800K, and 4.19+-0.05 GPa and 1780K. Flotation of diamonds was observed at 4.74+-0.08 GPa and 1810K, and 5.08+-0.07 GPa and 1870K. A compression curve determined in the present study is consistent with Circone and Agee (1996).
MR13B-1259
Diffusivity of Diopside Liquid at High Pressure
With the evidence of deep xenoliths and the seismic observation of ultra lower velocity zone (ULVZ), silicate liquids are thought to extend to the lower mantle and even to the core-mantle boundary. Knowledge of their physical properties at high pressures is essential to understand deep magma oceans. Diffusivity controls rates of chemical exchange, and is connected with the mobility (viscosity) of melts, and their response to elastic waves. In this study, we investigated the diffusivity of diopside liquid across the entire mantle pressure-temperature regime by first principles molecular dynamics (FPMD) simulations, which have been successfully used to study structure, equation of state, and thermodynamic properties of silicate liquids. Our simulations are based on density functional theory (DFT) in the local density approximation (LDA) and the ultra-soft plane-wave pseudopotential method. The calculations were performed with the Vienna ab initio simulation package (VASP). All simulations in this study were performed for an 80-atom cubic unit cell. For each calculation, the total run is 8 ps. Larger systems (160 atoms) were also performed and produced results within the statistical uncertainty of our simulations. We extract the self-diffusion coefficients via the mean-squared displacement and the Einstein relation. The calculated total diffusion coefficients are fit to the Arrhenius relation, D = D0 exp[-(E*+PV*)/kT]. This relation, with constant activation energy and volume does not capture all relevant behavior as a maximum diffusivity is observed at relatively lower pressure. Differences in the self-diffusion coefficients of different atoms are subtle.
MR13B-1260
Configurational Contribution to the Compression of Silicate Liquids
Melting of silicates in the deep interior of planets has likely played an important role in their evolution. The segregation of silicate melts is controlled by the density contrast between melts and residual solids. Stolper et al. (1981) hypothesized that melts might be denser than their residues at high pressures due to the much larger compressibility of liquids than that of solids. Although the large compressibility of liquids is supported by ultrasonic measurements and static compression experiments using sink/float method, the compression mechanism of liquids is still not well understood. In general, equation of state of a material can be obtained by taking the volume derivative of the Helmholtz free energy. For a solid, free energy has two contributions, the potential energy of a static lattice, and the vibrational free energy (thermal contribution). Thus bulk modulus (the second derivative of free energy) of a solid at T=0 K is determined by the potential energy, while the vibrational part gives the temperature dependence. However, under compression atoms in a liquid can undergo structural rearrangement in addition to the uniform shortening of interatomic distances. Therefore another term, the configurational contribution, must be included in the liquid free energy. It is this contribution that leads to the different compression mechanism for liquids. In this work, we analyzed elastic properties determined by Brillouin spectroscopy, ultrasonic velocity measurements, and static compression experiments on solids, glasses, super- cooled liquids, and liquids for several silicate compositions including CaMgSi2O6, CaAl2Si2O8, Fe2SiO4, etc. We find that in the bulk modulus-density log-log plot (the slope of this plot is the Grüneisen-Anderson parameter), data for solids, glasses, and super-cooled liquids fall on the same straight line, while data for relaxed liquids fall on a distinctly different line. This means that the compression mechanisms for super-cooled liquids, glasses, and solids are essentially the same. Their differences in bulk modulus are mainly due to the volume difference (or bond length difference). However, unlike super-cooled liquids and glasses, configurational rearrangement plays an important role for the relaxed compression of liquids, which makes the bulk modulus of liquids further smaller. The percentages of configurational contribution to the total compressibility for different silicate liquids are then compared with the configurational entropy of the liquids to shed some light on the theoretical model of liquid compression.
MR13B-1261
Glass or Solution Structures can not be Validated by Radial Distribution Analysis
The manufacture of crystalline phases from first principle calculations has been hampered by uncertainties in describing the nucleation and growth process. For aluminosilicates the principal working hypothesis has been that differences or similarities in molecular moieties between glass or melt and crystalline precipitate as inferred from radial distribution analysis, originally designated the Prins function (Zernike & Prins, 1927, Bernal, 1964) affect the activation energy for nucleation. Thus anorthite composition glass, a fragile system, crystallizes easily presumably due to the presence of 4-membered rings as inferred from radial distribution analysis, whereas albite composition glass, a robust system, not at all due to the presence of 6-membered rings. Unease exists about the precise information content of such analysis as for instance manifested by the long standing discord about the presence of boroxyl rings vis a vis linear boron oxide chains, both of which being compatible with the radial distribution of B2O3 glass. To settle the issue concerning the information content of a radial distribution for amorphous solids or liquids once and for all, we have calculated this function for all known 14 crystalline single and mixed alkali disilicates (R2O-2SiO2). These silicates contain the, by stoichiometry preordained, local environment Q3, i.e. tetrahedra with 3 bridging and one non-bridging oxygen, as is characteristic for sheet silicates. Some of these compounds form sheet silicates with 6-membered rings, others contain 4, 4-8, 4-6-8; 4-8-12, or 14-membered rings occurring either as sheets, tubes, or 3-D networks. We calculated total and partial pair distributions eliminating the alkalis in the latter to enhance details of the silica moieties. Our results show that no ring system assignment could be made to any of the calculated PDF's, nor that the large variety of ring systems in the different phases could be distinguished from one another. Validating Molecular Dynamic calculations using as criterion the quality of fit with observed radial distributions for glasses or aqueous solutions, is therefore incompatible with the information content available in such analysis. Thus structural assessments of extended moieties in melts and aqueous solutions remain elusive.
MR13B-1262
Thermal diffusivity of pyroxene, feldspar, and silica melts, glasses, and single-crystals at high temperature
Thermal diffusivity (D) due to phonon transport (the lattice component) was measured using laser-flash analysis from oriented single-crystals and of glasses above the glass transition, which proxy as melts. Compositions include SiO2, CaMgSi2O6, LiAlSi2O6, NaAlSi3O8, and CaAl2Si2O8. KAlSi3O8 was studied previously. Viscosity measurements of the supercooled liquids, in the range 106.8 to 1012.3 Pas, confirm near-Arrhenian behavior. For all compositions and for crystal and glass, D decreases with T, approaching a constant generally near 1000 K: Dsat, which is larger in the crystal than in the glass. A rapid decrease in D as T is increased further (ca 1400 K for orthoclase) is consistent with crossing the glass transition, verified from our viscosity data on these systems. The amount of the decrease depends on the chemical composition and similar to the relative decrease observed in heat capacity. Orthoclase values for Dsat are 0.65± 0.3 mm2/s for bulk crystal and 0.53+/-0.03 mm2/s for the glass. Constant D = 0.475+/-.01 mm2/s represents melt. Thermal conductivity (klat) of orthoclase glass, calculated using previous results for heat capacity (CP) and our density data, increases with T due to CP strongly increasing with T, reaching a plateau near 1.45 W/m-K for melt, but is always below klat of the crystal. Similar results were obtained from the other systems studied. Melting of silica, pyroxene, and feldspars impedes heat transport, providing positive thermal feedback that may promote further melting in the continental crust. The consistency of the behavior for these different compositions and structures suggests that our results are universal, holding for oceanic lithosphere as well. Melts, due to being disordered, are poor transporters of heat via vibrations. However, d(ln klat)/dP depends inversely on bulk modulus, suggesting that at some high pressure, the thermal conductivity of the melt and corresponding crystal become equal so that retention of heat by melts may not occur deep inside the Earth.