Volume 81, Issue 11-12

Scanning tunneling microscope (STM) images and scanning tunneling spectroscopy (STS) spectra of hematite (α-Fe 2 O 3 ) surfaces were calculated using ab-initio methods, not only to interpret experimentally collected STM data, but also to gain insight into atomic level changes in electronic structure that are associated with heterogeneous surface reactions. The electronic structure and wave functions inside the studied crystal were obtained as a periodic solution of the SchrOdinger equation by using the program Crystal92. STM images and STS spectra were calculated by applying a technique similar to the Tersoff and Hamann (1985) method. Experimental STM images of the upper valence band of hematite (001) surfaces, cleaved in air, show a periodic array of bright spots that differs slightly from the O-O separation in the bulk. However, our calculations show that these spots are located at the Fe positions of the surface Fe atoms and above the Fe atoms between the first and second hexagonally close-packed O layers. The calculated STS spectra for tip positions above the three nonequivalent Fe positions show significant differences, in particular because the contribution of O 2p-like and Fe 3d-like states changes with the distance between the tip and the respective Fe atom underneath. Hematite crystals that were used to obtain STM images experimentally in previous studies were cleaved in air, and the presence of adsorbed H 2 O and O 2 was considered in this study. Calculations that optimize the surface atomic arrangement with respect to total energy of the slab indicate that Hp and O 2 adsorbed to the surface have binding energies too low to withstand the dragging force and the electric potential applied during the scanning process. In addition, only calculations of STM images of fresh hematite surfaces exactly mimic the periodicity of high electronic density spots, as observed in experiments. STS spectra calculated for equivalent Fe positions on terraces and near steps show the increased electron density of the top of the valence band for step sites, which is experimentally observed as higher intensities of bright spots at steps. These calculations show that the local electronic structure of surfaces can be very different from bulk electronic properties and that conclusions drawn from cluster calculations representing the bulk can be misleading. In addition, this theoretical approach helps to explain the increased reactivity at specific sites on hematite, such as steps and kinks, in terms of the electronic surface structure of this mineral.

High-resolution 25 Mg NMR data are reported for a single crystal of pure forsterite (Mg 2 SiO 4 ) at temperatures to about 1400 °C, which allow rates of exchange of Mg cations among specific sites in the structure to be obtained. Results are consistent with estimates of hopping frequencies derived from macroscopic diffusivity data, suggesting that this approach holds potential for tracing microscopic diffusion pathways in even very refractory minerals.

X-ray absorption fine-structure (XAFS) spectra at the Fe K edge in Bamble enstatite (Mg 0.88 Fe 0.12 )SiO 3 were analyzed in conjunction with theoretical XAFS spectra to determine the bonding configuration of Fe in this structure. The structural analysis involved deter­mination of the Fe distribution between the octahedral M1 and M2 sites, and Fe-O bond lengths in the M2 site, into which Fe strongly partitions. Our analysis yielded bond lengths for Fe in the M2 site of 1.97(2), 1.98(2), 2.09(2), 2.16(2), 2.43(2), and 2.65(10) Å, in agreement with bond lengths determined from X-ray and neutron-diffraction analysis of the two orthopyroxene end-members. The average Fe-O bond length in the M2 site is 2.22(2) Å, longer than that of the Mg end-member (2.151 Å) but approximately the same as that of the Fe end-member (2.228 Å) of the orthopyroxene solid-solution series. Octa­hedral distortion of the M2 site may be greater than that of either the Fe or Mg end- member. The presence of a minor amount of Fe 3+ was inferred by our analysis of the M1 site and was also suggested by our bond-valence calculations, which yielded a charge of 2.07 for Fe in the M2 site and a charge of 2.78 for Fe in the M1 site. Simple calculations using our data and those of other studies show that the average Fe-O bond length in the M2 site is constant along the Fe-Mg join in the orthopyroxene solid-solution series.

The effects of compositional changes and nonsymmetry-breaking Al-Si ordering on the displacive phase transition in anorthoclase have been studied using X-ray powder diffrac­tion. The results were analyzed using a model based on Landau theory. The observed transition temperature, T ∗ c , was found to be independent of orthoclase content in the range 0 < X Or < 0.02. For samples with X Or > 0.02, Tf decreases linearly with increasing X Or . This plateau behavior is explained by the finite volume of the strain field around each K + ion: Chemical mixing behavior is observed only for average K -K + distances of <20 Å. Increasing Al-Si order between the T1 and T2 sites causes T c to increase; this is unlike T1o-T1m ordering, which prevents the transition to monoclinic symmetry altogether. The range of T ∗ c values reported in the literature is shown to be consistent with different degrees of T1-T2 order.

The Al-Si configuration in ordered and disordered lead feldspar (PbAl 2 Si 2 O 8 ) has been investigated by single-crystal X-ray diffraction. Single crystals synthesized from melt and cooled from T = 1280 to 1000 °C in 1 h and then to T= 700 °C in 15 h show a completely disordered Al-Si configuration with no b>-type superstructure reflections (C2/m, a = 8.428, b = 13.054, c = 7.174 Å, p = 115.32°; R = 5.8%). Subsequent structure refinement of a crystal hydrothermally annealed at T = 500 °C and P H₂O = 2 kbar for 216 h shows an ordered Al-Si configuration (I2/c, a = 8.388, b = 13.067, c = 14.327 Å , p = 115.19°; R = 4.7%; Q od ≈ 0.9; 950 b reflections with F o ≥ 4σ). In the ordered lead feldspar the Pb polyhedron is significantly distorted in comparison with the Sr polyhedron in strontium feldspar. The coordination of the Pb site is reduced from sevenfold, as in strontium feldspar, to a rather irregular sixfold configuration. Such distortion could be related to the lone-pair effect in Pb 2+ . In the disordered lead feldspar the Fourier map shows an anomalous elec­tron-density distribution around the Pb site and consequently a split-site model for Pb was assumed in the refinement (Pb-Pb′ split = 0.557 Å).

Procrystal models for the electron-density distributions of low albite and microcline were constructed by placing spherically averaged Clementi-Roetti atomic wave functions at experimentally observed atomic positions. The topography of the model electron-density distribution was analyzed to locate (3, -1) critical points between the Na or K positions and the positions of the surrounding framework O atoms. Because the number of (3,-1) critical points associated with a given atom corresponds with its coordination number, the analysis indicated that Na is fivefold and K is sevenfold coordinated. In particular, the results indicate that the Oco atom in low albite is not coordinated with Na, whereas in microcline it is coordinated with K. This suggests that in low albite the Oco atom is underbonded and thus a possible hydrophilic site, whereas in microcline the bonding requirements of Oco are satisfied. Details of the local bonding explain the effects of pressure, H, and H 2 0 on ordering of Al and Si, as well as compressibility systematics.

The crystal chemistry and structure of the layered lead oxychloride parkinsonite, ideally Pb 7 MoO 9 Cl 2 , have been reevaluated in the light of TEM observations of natural and syn­thetic samples and the strong analogy with the Dirt and DO 22 ordering patterns of Ni-Mo alloys. D 1 a-like superstructures occur in synthetic parkinsonite and V-bearing parkinsonite because of Pb-Mo and Pb-V ordering, with ordering stoichiometries Pb 9 Mo (10-site scheme) and Pb 18 V 2 (20-site scheme). Natural parkinsonite orders on the Pb 7 Mo (8-site) or Pb 14 Mo 2 (16-site) scheme. This ordering involves the formation of an incommensurate state that may be associated with a I pseudohexagonal motif similar to the DO 22 superstruc­ture in Ni-Mo alloys. It is our contention that X-ray methods usually miss important information about the superstructures in these minerals because of very high X-ray ab­sorption. As a result, single-crystal structure refinements can lead to erroneous identifica­tion of the tetragonal or pseudotetragonal I subcell as the true unit cell. Transmission electron microscopy allows examination of extremely thin crystallites and observation of weak superstructure reflections. Crystal-chemical inferences about the structure and order­ing behavior in these minerals are supported by EXAFS and XANES studies of a synthetic parkinsonite ordered on the 9:1 scheme, Pb 9 MoO 11 Cl 2 , indicating that Mo 6+ may well be present in fivefold square-pyramidal coordination.

Transmission electron microscopy (TEM) investigation of natural kalsilite, (K 1.77 ☐ 0.09 Na 0.14 ) Σ₂ (Si 2.09 Al 1.79 Fe 0.12 ) Σ₄ O 8 , from San Venanzo, Italy, shows that it contains a pervasive domain structure related to a superstructure. The probable space group for the superstructure is P2 1 . The relationship between the unit-cell parameters of the P6 3 subcell and those of the P2 1 supercell is a super ≅ b supcr ≅ V3 sub ; C super = c sub , and the a super axis is rotated 30° about the c* axis with respect to the a sub axis. Nonperiodic distribution of elon­gated twin domains causes streaking of superstructure reflections normal to the c* axis. The average structure of this kalsilite sample has P6, symmetry. On the basis of previous structure refinements, this average structure has three O1 positions slightly displaced from the threefold axis, and each O1 position has a site occupancy of ⅓. High-temperature annealing studies indicated that there is a displacive phase transition from the P2 1 super­structure to the P6 3 substructure. The twin domains are related by threefold twin operations. The phase-transition temperature for kalsilite may range from 500 to 600 °C and is de­pendent on the crystal composition (e.g., the proportions of Na, Ca, and vacancies). Kalsilite crystals prepared from nepheline by Na-K exchange are structurally complex. Domain structures in Na-bearing kalsilite crystals with more than 2.5 mol% NaSiAlO 4 are similar to those in the natural kalsilite sample but exhibit additional twin boundaries par­allel to the c axis. Na-poor and Na-ffee kalsilite crystals are composed of (0001) domains with P6 3 and P31c symmetries, respectively, and merohedral twins between the neighbor­ing P6 3 or P31 c domains. The merohedral twins in the synthetic crystals formed during transformation of the nepheline structure to the kalsilite structure. The intergrown domains with P31c symmetry formed during the phase transition on cooling of the Na-poor kalsilite. There is no simple symmetry hierarchy among the known structures in the proposed phase diagram.

Selected-area electron diffraction, dark-field images, and high-resolution transmission electron microscope images show that there are pervasive domains in hillebrandite from Mexico. The domains are in thin layers that are almost parallel to (010). They range from one to several unit cells thick across the b direction, with an average thickness of about 10 nm. The actual repetition period (7.28 Å) along the a crystallographic direction is that of the wollastonite-like chain in the hillebrandite structure. Neighboring superstructure domains are related by stacking faults, with stacking vector of ± ½ a sup . Nonperiodic ar­rangements of the superstructure domains give an average structure of hillebrandite with a 3.64 Å repetition period along the a axis and one-half occupancies of Si atoms in the tetrahedral sites in the two Si-O chains.

Transmission electron microscopy (TEM) was used to study the structural changes that occur in aluminate sodalite, Ca 8 [Al 12 O 24 ](CrO 4 ) 2 , when it is heated with the electron beam. Initially, a twinned orthorhombic crystal was heated, and it transformed to tetragonal and then to cubic symmetry. The cubic phase contains two-dimensional periodic antiphase boundaries (APBs) along the {110} planes. The APBs are about 54 Å apart. Structural modulations also occurred during the transformations. The transition from tetragonal to orthorhombic symmetry is also associated with random transformation twins on a {110} plane; this plane is both the composition and twin plane.

Microcrystalline opal was investigated using low-dose transmission electron microscopy (TEM) methods to identify microstructural characteristics and possible phase-transforma­tion mechanisms that accommodate silica diagenesis. High-resolution TEM (HRTEM) re­vealed that microcrystalline opal in opal-CT chert (>90 wt% silica) and opal-CT porcelanite (50-90 wt% silica) from the Miocene Monterey Formation of California displays various amounts of structural disorder and coherent and incoherent lamellar intergrowths. Species of microfibrous opal identified by HRTEM in early-formed opal-CT chert include length-slow opal-C and unidimensionally disordered length-slow opal-CT (“lussatite”)- These fibers often display a microstructure characterized by an aperiodic distribution of highly strained domains that separate ordered domains located at discrete positions along the direction of the fiber axes. Microfibrous opal occurs as several types of fiber-aggre­gation forms. TEM revealed that the siliceous matrix in later-formed opal-CT porcelanite consists of equidimensional, nanometer-size opal-CT crystallites and lussatite fibers. Pseu­do-orthorhombic tridymite (PO-2) was identified by HRTEM in one sample of opal-CT porcelanite. Burial diagenesis of chert and porcelanite results in the precipitation of opal-C and the epitaxial growth of opal-C domains on opal-CT substrates. Diagenetic maturation of lussatite was identified by TEM in banded opal-CT-quartz chert to occur as a result of solid-state ordering. The primary diagenetic silica phase transformations between non­crystalline opal, microcrystalline opal, and quartz occur predominantly by a series of dis­solution-precipitation reactions. However, TEM showed that in banded opal-CT-quartz chert, the epitaxial growth of quartz on microfibrous opal enhances the rate of silica diagenesis.

A systematic TEM investigation of interstratified chlorite-biotite crystals showed that the crystals are composed of domains of periodically interstratified chlorite-biotite, nonperiodically interstratified chlorite-biotite, biotite, and chlorite. The interstratified chloritebiotite Occurs as a vein filling and was apparently crystallized from a hydrothermal solution. The complex structure of the interstratified chlorite-biotite presumably results from a nonlinear growth phenomenon occurring under a nonequilibrium state. A simple nonlinear dynamics model derived from Duffing's equation was constructed with an additional chemical potential that accounts for the variation of structural configuration of tetrahedral sheets or 2: 1 layers in chlorite and biotite, a simple periodic fluctuation of hydrothermal fluid composition, and a simple damping force for two-dimensional lattice misfit on (001) resulting from the intergrowth of different types of layers with different structural configurations and other dissipation effects. Solutions to the equations of the model show that periodic interstratification, nonperiodic interstratification, and domains of the two end-member components (biotite, chlorite) can be formed during crystallization under various conditions. The nonperiodic sequences of biotite and chlorite layers along the c axis in the interstratified crystals produced by this model are chaotic rather than random. The calculations suggest that both periodic and nonperiodic interstratifications can result from periodic external force, e.g., compositional fluctuation of the fluid.

A periodic ab initio Hartree-Fock LCAO study was performed on the 1:1 sheet silicate lizardite, Mg 3 Si 2 O 5 (OH) 4 , which has P31m symmetry. A total of 258 atomic orbitals were described using double-zeta-quality basis sets augmented with polarization d (Si, Mg, O) and p (H) functions. Density of states and electron charge-density maps were calculated to investigate the electronic properties. The majority of the valence states are composed of O and Si atomic orbitals with little contribution from H atoms. Calculations showed that although there are about 0.5|e| in Si d and about 0.1|e| in Mg d orbitals, the population of O d orbitals is negligible. The maps of charge density show that interlayer hydrogen bonds fix adjacent 1:1 layers. Positions of the main O peaks in projected density of states evaluated for both three-dimensional (3D) and two-dimensional (2D) calculations were influenced by layer-to-layer interactions, especially hydrogen bonds.

Thermodynamics of order-disorder in minerals may be approached by treating the mineral as a solid solution between an independent set of end-members with which the range of possible states of ordering of the phase can be represented. Thus, a mineral of fixed composition (a one-component phase), requiring s independent order parameters to represent the state of order in the mineral, involves an independent set of s + 1 end-members. This approach is applied by means of symmetric formalism, with the entropy part of the Gibbs energy taken to be the ideal configurational entropy of mixing using a mixing-onsites formulation, and the enthalpy part taken to be that of a regular solution between the s + 1 end-members. Symmetric formalism is shown to be formally identical to the generalized Bragg-Williams or point approximation, and its treatment of convergent and nonconvergent cation ordering is compared with that of the Landau theory. Its flexibility in describing a wide range of order-disorder behavior is illustrated with applications to sillimanite, spinel, albite, and potassium feldspar, the latter two involving order-parameter coupling.

The thermodynamics of order-disorder in mineral solid solutions are handled with symmetric formalism, whereby the mineral is treated as a solid solution between an independent set of end-members with which the range of composition and states of ordering of the phase can be represented. An n-component mineral, requiring s independent order parameters to represent the state of order in the mineral, involves an independent set of n + s end-members. Symmetric formalism involves ideal mixing-on-sites with regular-solution activity coefficients. It is applied to omphacite, orthopyroxene, ferromagnesian clinoamphibole, and alkali feldspar. The model for omphacite, with a single order parameter, successfully produces the topology of paired miscibility gaps with tricritical points at their apices and with a critical curve connecting them. Ferromagnesian orthopyroxene is shown to behave effectively as an ideal solution at all geologically relevant temperatures. Cummingtonite- grunerite solid solutions are slightly positively nonideal in either a two-site or a three-site model. Na-K alkali feldspars with order-parameter coupling involving tetrahedral site occupancies can show the essential topologic relationships in this system, with only one independent binary interaction energy. The power of symmetric formalism comes from the simplicity of its representation of the thermodynamics of minerals and its flexibility with few adjustable parameters.

In unequilibrated eucrites, pyroxene is a sensitive recorder of variables that affect their crystallization histories. Therefore, major, minor, and trace element systematics of pyrox­ene were evaluated from several unequilibrated eucrites and an equilibrated eucrite, Juvinas, by EMP and SIMS techniques. The Cr, Al, and Ti zoning trends in pyroxene reflect its igneous crystallization history, including the sequence of crystallization of pyroxene and plagioclase. The primary substitutional couples in unequilibrated pyroxene among Cr, Al, and Ti are [6] Cr 3+ - [4] Al 3+ , [6] Al 3+ - [4] Al 3+ , and [6] Ti 4+ -2 [4] Al 3+ , with the [6] iTi 4+ -2 [4] Al 3+ couple dominating over [6] IA 3+ - [4] Al 3+ following the onset of plagioclase crystallization. Reequili­bration erases the primary igneous zoning trends of Cr, Al, and Ti, and the dominant substitutional couples in homogenized pyroxene are [16] Ti 4+ -2 [4] Al 3+ and [6] Cr 3+ - [4] Al 3+ in equal proportions. Trace element compositions in unequilibrated eucrites are variable be­cause of igneous zoning, but averaged analyses show core-rim differences that reflect their Ca-rich rims relative to Ca-poor cores. The pyroxene REE abundances in the unequilibrated samples are as much as ten times chondrite in their rims, whereas homogenized Juvinas pyroxenes have abundances between the core and rim values of pyroxene from unequilbrated eucrites. The ratio of REEs (homogenized pyroxene/bulk composition) for Juvinas is significantly higher than the ratio of REEs (pyroxene cores/bulk composition) for the unequilibrated eucrites. This is especially true for the HREEs. The higher observed REE concentrations in the equilibrated pyroxene result from the intracrystalline exchange between Ca-poor cores and Ca-rich rims and also from intercrystalline exchange with other phases, including plagioclase, phosphates, and mesostasis. Use of the equilibrated pyroxene REE concentra­tions and igneous D values leads to an incorrect estimate of the parental melt composition. The estimated chondrite-normalized melt REE concentrations are too high by a factor of two to four. These studies suggest that great care must be used to take into account sub­solidus elemental redistributions when attempting to estimate parental melt compositions from measured mineral trace element concentrations.

Although studies have shown that igneous cumulates can form by in situ crystallization without requiring crystal settling, it has not been demonstrated that crystal-size distributions (CSDs) are consistent with such a process. Plots of crystal-size fractions per unit volume vs. crystal size for chromite grains from the Stillwater complex show a log-linear distribution with negative slope at larger sizes and a concave-down distribution at smaller sizes. The log-linear portion of these trends is similar to previously reported trends for other silicate and oxide phases in crystallizing magmas. However, the lack of smaller grain sizes indicates that the original population was altered during a postnucleation crystalaging period that resulted in the loss of the smaller size fractions, an interpretation consistent with textural and volume-balance evidence. The CSD trends imply that Cr was not concentrated by settling of chromite but was brought to the site of the growing chromite grains. Similarities between the Stillwater data and CSD trends for garnet in metapelites suggest that such trends are a characteristic feature of any geologic system undergoing crystal aging after an initial period of nucleation and crystal growth.

Regional structural interpretation, thermobarometry, and Gibbs-method calculations were combined to determine an approximate P- T path for kyanite-grade metapelitic rocks of the Baltimore Gneiss terrane. Garnet growth along the estimated P-T path was modeled using the computer program DiffGibbs. Models were evaluated by comparing calculated profiles of garnet components with measured profiles of a large garnet rich in plagioclase inclusions in a kyanite-bearing sample from the area. Modifications in the P-T path con­strained by changes in garnet components along critical portions of the profile resulted in a very satisfactory match between measured and model-garnet profiles. In the most suc­cessful model, garnet growth began at 550 °C and 6500 bars in the assemblage chlorite + biotite + garnet + muscovite + plagioclase + quartz with increasing temperature and with decreasing and then constant pressure. At 570 °C and 5724 bars, staurolite was added to the assemblage and some garnet was consumed. In the model, garnet growth resumed after chlorite was consumed along a path having a substantial increase in pressure. The region near the rim with low measured grossular content was reproduced by cessation of garnet growth as temperature increased at constant pressure to a maximum of 615 °C at 6474 bars. After kyanite was added to the assemblage, grossular increased to the constant rim value only if temperature decreased and pressure increased. Cooling and decompression affected garnet zoning only very near the rim. Measured compositions of the rims of plagioclase inclusions in the garnet are consistent with plagioclase compositions predicted by the model. This study demonstrates that complex garnet zoning can be modeled successfully, even when the reaction history includes episodes of garnet consumption, over a path that is consistent with geologic evidence. Details of garnet zoning provide surprisingly tight con­straints on the P-T path.

This paperreports observations of etched internal {0001}, {101̅ 0}, {112̅0}, {101̅1}, and {112̅1} surfaces of Durango apatite. Three types of surfaces are distinguished: P type, S type, and T type. After initially acquiring a certain roughness, P-type surfaces grow smoother with etching time. S-type surfaces remain smooth for long etching times. T-type surfaces develop a progressively coarser texture. Polishing scratches are flat bottomed in P-type, sharp in S-type, and indistinct in T-type surfaces. Fission tracks are funnel shaped in P-type surfaces but simple channels in S- and T-type surfaces. Several lattice defects other than fission tracks are revealed in P-type surfaces but not in S- and T-type surfaces. S-type surfaces can retain the imprints of fission tracks after they have been etched out. The etching velocity perpendicular to the surface increases in the order of P, S, T. Track density is approximately the same in all surfaces near the limit of track revelation but diverges with prolonged etching. It decreases in P-type surfaces, remains fairly constant in T-type surfaces, and increases in S-type surfaces. These trends are partly explained by the interplay between the track geometries and the track-determination criteria. The classic ν b -ν t model fails to explain these track geometries or their dependence on surface orientation. An alternative model is proposed, based on the theories of crystal growth and dissolution. This model implies that the efficiency of fission-track counts cannot be calculated on the basis of etching velocities. It further predicts that confined track lengths do not increase indefinitely with etching time, and that the lengths of surface tracks decrease.

The crystal structure of bottinoite, Ni(H 2 O) 6 [Sb(OH) 6 ] 2 , was determined from a twinned crystal. This study revealed that the strong 3̅1m pseudosymmetry shown by all the natural and synthetic crystals examined results from {101̅0} twinning. The structure is trigonal (space group P3) with a = 16.045(4), c = 9.784(2) Å (natural bottinoite) and a = 16.060(3), c = 9.792(1) Å (synthetic analog). The structure consists of a sequence of pairs of layers parallel to (0001) and stacked along the c axis. Each layer consists of isolated octahedra, linked together by hydrogen bonds, which also connect adjacent layers. Ni 2+ cations are ordered in two of the ten independent octahedral sites in one layer, whereas the second layer of the pair is made up of only Sb 5+ octahedra. The average octahedral bond lengths are 2.06 and 1.98 Å for NiO 6 and SbO 6 , respectively. The ten independent octahedra show four orientations mutually related by pseudosymmetry operations. This particular feature is taken into account in the formulation of a hydrogen-bonding model and a twinning mechanism. Structural and geometrical relationships with the brucite-like M 2+ (OH) 2 structures are also discussed.

Crystal structures were determined by single-crystal X-ray methods for the {011}, {610}, and {010} growth sectors of brewsterite from Strontian, Scotland. Refinements R w = 4.3-6.6%) showed that all three growth sectors are triclinic and that their structures differ slightly.

A novel type of high-temperature, high-pressure cell has been designed for near-infrared and optical spectroscopy of hydrous silicate melts at pressures up to 3 kbar and temperatures up to 800 °c. It consists of an externally heated cell with a disk-shaped sample chamber (4 mm in diameter and 0.2 mm in height) that is compressed between a tungsten carbide piston and a cylindrical sapphire window. A platinum ring surrounding the sample prevents loss of melt or water during the experiment. The new spectroscopic cell was used for direct measurement of H 2 O species in aluminosilicate melts by near-infrared spectroscopy. Other possible applications of the cell include the study of silicate melts doped with transition metal ions in the UV-VIS-NIR and the observation of the crystallization and dissolution of minerals in hydrous silicate melts under in situ conditions.

An imaging-plate detector interfaced to a large-volume high-pressure device allows quantitative in situ powder X-ray diffraction measurements at the National Synchrotron Light Source (NSLS). High-quality diffraction patterns were recorded from an (Ni 1 Mg) 2 SiO 4 olivine solid-solution sample at 4 GPa and 800 °C as a function of time using 3 min exposures at 40, 63, and 109 min. Refinement of the crystal structure indicated an increase in the ordering of Ni 2+ and Mg 2+ cations over the M1 and M2 sites at high pressure. The unit-cell volume was found to decrease with increasing ordering. Kinetic phenomena associated with the cation redistribution were observed on the time scale of tens of minutes. Extrapolation based on an exponential law of ordering relaxation toward equilibrium gave a distribution coefficient (K D ) of 10.7(1) at 4 GPa and 800 °C; the starting sample, which was annealed at 800 °C and room pressure, gave K D = 8.3(1).

The dissolved H 2 O concentrations of experimentally produced, Fe-free, nominally an­hydrous minerals (NAMs) have been measured using, for the first time, 1 H magic-angle­spinning NMR spectroscopy. At least two types of hydrous species are present in all the samples studied, leading to peaks of different widths: (1) a broad resonance probably due to clustered OH such as that in static H 2 O molecules within the structure, clusters of OH or hydrogarnet substitution, and (2) narrower peaks presumably due to nonclustered OH associated with point defects. The solubility of H 2 O in clinopyroxene close to the Di-CaTs join is 0.26-0.41 wt% H 2 O at 1000-1100 °C and 1.5 GPa. The solubility of H 2 O in enstatite and forsterite is 0.087 and 0.18 wt%, respectively, under similar conditions. It is possible that the broad components result from submicroscopic hydrous mineral or glass impurities or static H 2 O molecules at grain boundaries; if only the narrow peaks are considered to be due to hydrous species within the structure of the NAM, the solubilities are 0.12-0.24 wt% for the clinopyroxene, 0.024 wt% for enstatite, and 0.056 wt% for forsterite.

A pressure-induced phase transition was observed in a sample of synthetic titanite (CaTiOSiO 4 ). The high-pressure structure was refined in space group A2/a [a = 6.8912(3), b = 8.6234(3), c = 6.4065(3) Å , β = 113.057(4)°] using powder diffraction data collected with synchrotron radiation on material in a diamond-anvil cell at 69.5 kbar. The observed P2 1 a ↔ A2/aa symmetry change is identical to the well-described phase change at 496 K. However, the mechanism of the phase transition is proposed to be different. The high- temperature phase transition is assumed to result from a loss of long-range order between Ti off-center dipoles, while locally preserving the characteristic off-center displacement of Ti atoms. In contrast, the high-pressure phase transition is a response to increasing over­bonding of the Ti atoms, which forces individual Ti atoms to shift to the center of the coordination octahedron. Thus, the A2/a high-pressure phase, unlike the A2/a high-tem­perature phase, has both local and long-range A2/a symmetry.

In situ, high-temperature Raman spectroscopy of Na 2 O·9SiO 2 glasses and melts with and without 2 mol% P 2 O 5 has been conducted to 1491 K (near the liquidus temperature). In the temperature range from 298 to ~1000 K (below the glass-transition temperature range), orthophosphate (PO 4 3- ) complexes appear to be the dominant phosphate species present. As temperature increases above ~1000 K, the orthophosphate complex is partially replaced by pyrophosphate complexes (P 2 O 4- 7 ) ) so that at ~1200 K >70% of the P exists in the latter form. The proportion of pyrophosphate species does not change significantly as the tem­perature further increases. This change in phosphate speciation is accompanied by an increase in the Q 3 /Q 4 silicate species abundance ratio. At the highest temperatures there is also evidence for Q 2 species in the melts. The transformation from orthophosphate to pyrophosphate implies depolymerization of the silicate network because the P/O ratio in the P 2 O) complex is smaller than in PO 4- 4 . This depolymerization is consistent with the change in Q 3 /Q 4 ratio in the silicate. Therefore, use of P speciation data from glasses would lead to overestimation of the degree of po­lymerization of P-bearing silicate melts.