V51D-01 INVITED 08:00h
Melt Structure and Properties: Progress and Prognoses
Recent advances in quantitative determinations of silicate glass structure are beginning to place important constraints on models of the physical and chemical properties of melts, but much remains to be done before such models can become entirely based on structure. For example, models of free energy and major component activities generally assume (lacking better constraints) that network species (e.g. Al, Si, "Qn" groups) and network modifiers (e.g. Na, Ca, K, Mg) each mix randomly. However, recent spectroscopic studies demonstrate strong ordering in the network (significant if incomplete Al avoidance; preference of Al for "Q4" groups, etc.) and between modifier cations differing greatly in field strength (e.g. K+, Mg2+). Solution models thus may need substantial revision, unless new studies of temperature effects on such ordering indicate approach to randomness at magmatic temperatures. Such studies are ongoing, facilitated by recent developments in hyper-quenching technology. On the other hand, discovery of such low-T ordering provides a likely (and long-suspected) mechanism for at least some of the configurational entropy in multicomponent melts, which in turn is a critical part of models of viscosity: at least in systems with high Al/Si ratios, for example, increase in Al/Si disorder with increasing T will a major part of this term. In many systems, however, the structural origins of Sconf remain mysterious. For both major and minor components, considerable information now exists to constrain models both of activity and of transport processes, but progress awaits development of useful model forms. Examples include new information on the local coordination of anions such as fluoride and chloride, which again show strong chemical ordering. Similarly, we now know a great deal about the concentration of minor "defect" species, that have been considered to be important to viscosity and diffusion (e.g. AlO5, SiO5, "excess" NBO). However, going from speculative mechanism to real prediction is a challenge that probably awaits improved ab-initio based molecular dynamics simulations. Recent results on Al, Si, and O coordination in glasses quenched at high pressure are beginning to provide solid constraints for models of density increase and of P effects on viscosity. These studies are especially promising when they address experimentally issues of thermal and decompression history on properties and structure of recovered glass samples.
V51D-02 08:15h
Molecular Dynamics Simulation of the Structural and Physical Properties of Silicate Melts in the CaO-MgO-Al$_{2}$O$_{3}$-SiO$_{2}$ System
Molecular dynamics (MD) simulation is used to calculate the structural and physical properties of both crystals and silicate melts in the CaO-MgO-Al$_{2}$O$_{3}$-SiO$_{2}$ (CMAS) system. A key component of numerical computations is the reliability and applicability of interatomic potentials used for simulation. In a previous publication (Matsui, 1996) we computed the properties of various silicate melts using the MD simulation with a rather simple rigid-ion model. An improved potential model incorporating many-body forces is developed in this study, in which each oxygen ion is allowed to deform isotropically under the effects of other ions in the system concerned, as reported by Matsui (1998; called breathing-shell model, BSM). Required energy parameters, including the net charges, repulsive radii, van der Waals coefficients of ions, and the oxygen breathing parameters, were derived empirically to reproduce not only the observed structures and bulk moduli of crystals in the CMAS system, but also the measured volumes of silicate melts in the CMAS system. The MD simulation with BSM is found to be quite successful in reproducing well these measured properties of both crystals and melts. The MD simulation is further used to predict the temperature-pressure-volume equations of state and structural details of silicate melts in the CMAS system over wide temperature and pressure ranges.
V51D-03 08:30h
A microscopic view on mass transport in silicate melts with neutron scattering
We investigate the interplay of structure and mass transport in alkali silicates and sodium aluminosilicate melts at temperatures up to 1600\,K with neutron scattering. Binary alkali silicates show evidence for the existence of alkali diffusion channels in the static structure. These channels percolate in the Si-O network and serve as preferential alkali ion conducting pathways. The non-homogeneous distribution of the alkali ions in the disrupted tetrahedral Si-O network on length scales of 6-8\,\AA\ strongly influences the structural relaxation, i.e. the viscosity, of the network. By systematically replacing Na$_2$O by Al$_2$O$_3$, albite and jadeite as well as their 75\,% per-alkaline compositions exhibit a significant decrease in the Na mobility and an increase in the viscosity. This goes along with a disruption of the channel structure found in binary alkali silicates with the addition of Al$_2$O$_3$. Neutron scattering on water bearing silica, sodium trisilicate and albite glasses reveals, that although dissolved water has a drastic impact on the melt viscosity, the structure is barely affected by the presence of OH groups. Vibrational spectra, that are dominated by the scattering of the hydrogen atoms, strongly depend on the anhydrous silicate composition, whereas the vibrational densities of states do not exhibit systematic changes in the concentration range of the dissolved water between 1 to 5 wt%.
V51D-04 INVITED 08:45h
Non-Newtonian and Viscoelastic Properties of Lava Flows
Lava flow models require an in-depth knowledge of the rheological properties of lava. Previous measurements have shown that, at typical eruption temperatures, lavas are non-Newtonian. The reasons for this include the formation and destruction of crystal networks and bubble deformation during shear. The effects of bubbles are investigated experimentally in this contribution using analogue fluids with bubble concentrations $<$20%. The shear-thinning behaviour of bubbly liquids noted by previous workers is shown to be dependent on the previous shearing history of the fluid. This thixotropic behaviour, which was investigated using a rotational vane viscometer, is caused by delayed bubble deformation and recovery when subjected to changes in shear stress. A rotational vane viscometer and torsional deformation apparatus were used to investigate the rheological properties of bubbly liquids and foams in order to determine a viscoelastic transition. These experiments have shown that the foams tested are viscoelastic power law fluids with a yield strength. Non-Newtonian properties and yield strength of foams are shown to be a probable cause of accelerating flow fragmentation in tube flow experiments on expanding foams. The flow of a bubbly fluid through a narrowing conduit may cause a pulsating regime of a flow due to periodic slip and slip-free boundary conditions near the walls of a conduit. Slip boundary conditions can lead to instability in viscoelastic shear flow causing short wavelength fluctuations at high shear rates. This mechanism may also take place during explosive volcanic eruptions. The frequency and amplitude of oscillation shear affect the structure of lavas which are thixotropic non-Newtonian liquids. The frequency dependent structure of lavas can be identified via frequency hysteresis and time-evolution of internal friction and viscosity. The rheological properties of basaltic lavas from Etna, Hawai'i and Vesuvius have been investigated at temperatures between $\sim500$ and 1150$\deg$C using a small strain oscillatory shear. The viscoelastic response of the lavas was analysed using small forced sinusoidal torques ($<10^{-3}$ N m) at frequencies between 0.002 and 20Hz. A purely viscous regime was only approached during experiments with Hawai'i samples. These experiments indicated that between $\sim$1070 and 1130$\deg$C, strain rate independent viscosities ($>10^9$ Pa s) could be measured at strain rates $<$ $\sim10^{-2}$ to $10^{-1} s^{-1}$. At 800$\deg$C, temporal variations in complex shear modulus and internal friction suggest that, over durations of up to 120h, structural adjustments were occurring within some of the samples. This time-varying behaviour of lava samples may be attributed to the slow closing (healing) of micro-cracks and pore space resulting in the apparent stiffening of lava samples under annealing. Thus, those parts of lava flows that undergoing slow cooling have more elastic properties. Regions which cool faster possess smaller shear moduli and higher internal friction due to thermal micro-cracking and less cohesion between crystal grains and bulk glassy matrix.
V51D-05 09:00h
Models for silicate melt viscosity
The prediction of viscosity in silicate liquids, over the range of temperatures and compositions encountered in nature, remains one of the most challenging and elusive goals in Earth Sciences. Recent work has demonstrated that there are now sufficient experimental measurements of melt viscosity to create new viscosity models to replace previous Arrhenian models [1],[2] and extend the compositional range of more recent non-Arrhenian models [3]. Most recently, [4] have developed an empirical strategy for accurately predicting viscosities over a very wide range of anhydrous silicate melt compositions (e.g., rhyolite to basanite). Future models that improve upon this work, will probably extend the composition range of the model to consider, at least, H$_{2}$O and other volatile components and may utilize a compositional basis that reflects melt structure. In preparation for the next generation model, we explore the attributes of the three most common equations that could be used to model the non-Arrhenian viscosity of multicomponent silicate melts. The equations for the non-Arrhenian temperature dependence of viscosity ($\eta$) include: a) Vogel-Fulcher-Tammann (VFT): log $\eta$ = A + B/(T - C) b) Adam and Gibbs (AG): log $\eta$ = A + B/[T log (T/C)], and c) Avramov (Av): log $\eta$ = A + [B/T]$^{\alpha}$ We use an experimental database of approximately 900 high-quality viscosity measurements on silicate melts to test the ability of each equation to capture the experimental data. These equations have different merits [5]. VFT is purely empirical in nature. The AG model has a quasi-theoretical basis that links macroscopic transport properties directly to thermodynamic properties via the configurational entropy. Lastly, the model proposed by Avramov adopts a form designed to relate the fit parameter ($\alpha$) to the fragility of the melt. [1] Shaw, H.R., 1972. Am J Science, 272, 438-475. [2] Bottinga Y. and Weill, D., 1972. Am J Science, 272, 438-475. [3] Hess, K.U. and Dingwell, D.B, 1996, Am Min, 81, 1297-1300. [4] D. Giordano & D.B. Dingwell, 2003. EPSL. 208, 337 (and related corrige EPSL 221, 449) [5] J.K. Russell, D. Giordano & D.B. Dingwell, 2003. Am Min 88, 1390
http://www.giordano.pi.it
V51D-06 09:15h
Viscosity of Anhydrous and Hydrous Basalt Melts at High Pressures
We performed in situ falling-sphere experiments to determine the viscosity of anhydrous and hydrous basaltic (48 wt% SiO$_{2}$) melts from 1.5 to 5.3 GPa between 1600 and 1840 K, using the T-25 MA8 multianvil apparatus at the GSECARS 13-ID-D beamline at the Advanced Photon Source, Argonne National Lab. These falling-sphere experiments included monitoring the simultaneous settling of Pt and Mo spheres, an approach that provides redundant viscosity measurements for individual experiments and offers the opportunity to recover simultaneously melt density. Our results show that the viscosity of anhydrous basalt melt decreases with pressure up to 5.3 GPa, with an activation volume for viscous flow of -8.1 to -9.5 cm$^{3}$/mol between 2 and 5.3 GPa. The addition of a few wt. % water reduces melt viscosity by roughly 0.5 log units; however, there is no resolvable influence on activation volume. This negative pressure dependence is consistent with previous results for basaltic melts up to 3 GPa [1, 2], while the activation volume at low pressure is indistinguishable from the activation volume for O self-diffusion in the same bulk composition [3]. Application of the Eyring equation using O self-diffusion data for basaltic melt [3] predicts anhydrous melt viscosities that are 30-90% of the values determined in this study. This result is in stark contrast with our recent results for dacitic melt (68 wt% SiO$_{2}$) melt [4], in which the Eyring equation overestimates viscosity by as much as 40% at pressures $<$ 5 GPa. The limited utility of the Eyring equation for naturally-occurring silicate melts illustrates the difficulties in relating O self-diffusion to viscous flow in polymerized liquids. Adam-Gibbs theory [5] provides a means for addressing structural controls on these transport properties. The negative pressure dependence for anhydrous and hydrous basalt viscosity suggests that the extraction of partial melts from mantle source regions will be enhanced with pressure to 5.3 GPa. Future work will extend these observations to higher pressures in an effort to constrain melt transport properties approaching conditions for the top of the transition zone where [6] have postulated the accumulation of a dense hydrous melt layer as a result of dehydration melting of transition zone material. [1] Fujii, T, Kushiro, I (1977) CIW Yrbk., 76, 419-424; [2] Ando, R, et al. (2003) SPring-8 report No. 11, 45; [3] Lesher et al. (1996) GCA, 60, 405-413; [4] Tinker et al. (2004) Am. Min., in press; [5] Adam, G, Gibbs, JH (1965) J. Chem. Phys., 43, 139-146; [6] Bercovici, D, Karato, S-I (2003) Nature, 425, 39-44
V51D-07 09:30h
The Influence of Melt Composition and Dissolved Water on the Melt-Vapor Surface Tension of Dacite and Rhyolite Magma
Melt-vapor surface tension ($\sigma$) is a poorly known quantity, yet it plays a critical role in many volcanic processes. In this study, we provide new data for $\sigma$ of hydrous natural dacite and rhyolite melt at 200 MPa, 800-1055$\deg$C, and 4.8-7.7 wt% H$_{2}$O using results from high temperature and pressure decompression experiments (this study; Mangan and Sisson, 2000; Mourtada-Bonnefoi and Laporte, 2002; 2004) and classical nucleation theory. The solutions give values for $\sigma$ that vary inversely with dissolved H$_{2}$O by 0.025 ($\pm$ 0.009) J/m$^{2}$ per wt% H$_{2}$O. Combining our results with data from hydrous haplogranite and rhyodacite (Epel'baum et al., 1973; Bagdassarov et al., 2000) reveals that melt-vapor surface tension also varies inversely with the concentration of mafic melt components (e.g., CaO, FeO$_{total}$, MgO). At constant H$_{2}$O content, $\sigma$ increases by 0.70 ($\pm$ 0.53) J/m$^{2}$ per tenth increment increase in the melt felsic index (FI = Na$_{2}$O+K$_{2}$O/Na$_{2}$O+K$_{2}$O+CaO) from FI $\sim$ 0.75 (dacite) to FI $\sim$ 1.0 (haplogranite). To understand these results, we consider the thermodynamic definition of $\sigma$, i.e., the work per unit surface area to create an interface between two phases. It can be expressed as: $\sigma$dA = $\omega_{dissociation}$ - $\omega_{interaction}$. The $\omega_{dissociation}$ term is the work ${\it done}$ to dissociate molecules from the bulk melt for incorporation in a diffuse, lower density interfacial zone only a few molecular layers thick. In contrast, the work of interaction, $\omega_{interaction}$, is work made ${\it available}$ as a result of across-interface attractions between gas and melt molecules. Increasing the concentration of dissolved H$_{2}$O and other network modifiers (e.g., Ca, Fe, and Mg) influence the balance of work terms as (1) the $\omega_{dissociation}$ declines due to lowered cohesion of the disrupted melt structure, and (2) the $\omega_{interaction}$ increases because enhanced molecular diversity supports more numerous and varied interfacial zone attractions. The net effect of decreased $\omega_{dissociation}$ and increased $\omega_{interaction}$ is a decrease in $\sigma$. Additional, systematic data on the variation of $\sigma$ in natural hydrous melts are needed before the volcanic implications can be fully explored. Nevertheless, it is evident from the data in hand that surface tension, much like melt density and viscosity, can and should be treated as a variable. Cooling, crystallization, and vapor exsolution impart a time-dependency to $\sigma$ that must be accounted for in modeling volcanic processes.