Oxygen-17 Nuclear Magnetic Resonance Study of the Structure of

Aug 20, 2008 - School of Earth and Environmental Sciences, Seoul National University, Seoul, ... Geophysical Laboratory, Carnegie Institution of Washi...
0 downloads 0 Views 891KB Size
Download PDF
11756

J. Phys. Chem. B 2008, 112, 11756–11761

Oxygen-17 Nuclear Magnetic Resonance Study of the Structure of Mixed Cation Calcium-Sodium Silicate Glasses at High Pressure: Implications for Molecular Link to Element Partitioning between Silicate Liquids and Crystals Sung Keun Lee,*,† George D. Cody,‡ Yingwei Fei,‡ and Bjorn O. Mysen‡ School of Earth and EnVironmental Sciences, Seoul National UniVersity, Seoul, 151-742 Korea, and Geophysical Laboratory, Carnegie Institution of Washington, 5251 Broad Branch Road, NW, Washington D.C. 20015 ReceiVed: May 20, 2008; ReVised Manuscript ReceiVed: July 5, 2008

The structure of silicate glasses and the corresponding liquids at high pressure and their structure-property relations remain difficult questions in modern physical chemistry, geochemistry, and condensed matter physics. Here we report high- resolution solid-state O-17 3QMAS NMR spectra for mixed cation Ca-Na silicate glasses quenched from melts at high pressure up to 8 GPa. The spectra provide the experimental evidence for the varying pressure-dependence in two different types of nonbridging oxygen (NBO) environments (i.e., Na-O-Si and mixed {Ca,Na}-O-Si) in the single glass composition. The percentage of NBO drops significantly with increasing pressure and is a complex function of melt composition, including cation field strength of network modifying cations. A decrease in NBO fraction with pressure is negatively correlated with the element partitioning coefficient between crystals and liquids at high pressure. Introduction Although it is well appreciated that the atomic and nanoscale structure of silicate melts and glasses at high pressure controls their macroscopic properties, detailed knowledge about their atomic and molecular structures, such as cation ordering and the nature of changes in the degree of melt polymerization at high pressures in general remains rather poorly constrained. At near ambient pressures, alkali cations (e.g., Ca2+, Na+, and Mg2+) in silicate melts operate as network modifiers, reducing the number of bridging oxygens (BO’s, e.g., [4]SisOs[4]Si and [4]SisOs[4]Al) through the formation nonbridging oxygen moieties (NBO’s, e.g., NasOs[4]Si, oxygen coordinated by [4]Si and several alkali cations).1 The degree of polymerization is often defined by the fraction of NBO and is one of the dominant controls on the thermodynamic (e.g., energy, entropy, and meltcrystal partitioning) and transport properties (e.g., diffusivity and viscosity) of melts and glasses at 1 a.m.2-4 The structure and the degree of polymerization of silicate melts and glasses quenched at high pressure are expected to be very different from those formed ambient pressure.5-10 The effect of pressure on silicate melt structure in general has been extensively studied using various experimental techniques, which revealed gradual pressure-induced coordination transformations of framework cations (e.g., Si and Al) at the expense of NBOs and changes in topological disorder including bond length and angle distributions.8,11-13 The free volume and ring size of silicate glasses and melts are also expected to vary with pressure. Detailed discussion of the pressure-induced coordination transformation in silicates and aluminosilicate glasses can also be found in recent reviews.8,14,15 * To whom correspondence should be addressed. Prof. Sung Keun Lee, Assistant Professor, School of Earth and Environmental Sciences, Seoul National University, Seoul, Korea, 151-742. E-mail: [email protected]. Phone: 822-880-6729. Fax: (822) 871-3269. Web: http://plaza.snu.ac.kr/ ∼sungklee. † School of Earth and Environmental Sciences, Seoul National University. ‡ Geophysical Laboratory, Carnegie Institution of Washington.

One of the important properties relevant to silicate melt structure at high pressure includes the element partition coefficients (Dmcrystal-melt)16-18 between liquids and crystals in equilibrium at high pressure, which has significant geophysical implications for chemical differentiation in the Earth’s interior. It has been reported that Dmcrystal-melt, can vary up to 2 orders of magnitude as a function of melt composition, hence the percentage of NBO at constant pressure.19,20 For instance, recent partitioning experiments revealed that the partitioning of potassium-40 between silicate melts and FeS (molten metals) changes by nearly an order of magnitude with variation in the percentage of NBO at pressures of up to 2 GPa.21 The various other chemical variables affecting the crystal-melt partitioning coefficient include activity coefficient of element (or metal oxide), oxygen fugacity, oxidation states, and the charge of the elements (e.g., see ref 17 for recent review of the effect of melt composition and structure on the element partitioning at low pressure regime). Previous predictive modeling efforts have focused on understanding the role of the crystalline phase,22 where lattice strain energy (e.g., Lattice strain model), charge, and net relaxation have all been shown to be valuable predictors of behavior. The role of melt structure at high pressure, however, has not been well-known due to lack of experimental constraints of atomic structures of silicate glasses at the elevated pressure ranges. In order to have better insights into the atomic structure of amorphous oxide at high pressure and its effect on macroscopic thermodynamic properties of the silicate melts including the elemental partitioning at high-pressure, we explore a structure of a mixed cation Ca-Na-silicate glasses [CNS, (Na2O)0.75(CaO)0.25. 3SiO2] at high pressures up to 8 GPa. The CNS glass can be regarded as one of the archetypal glass formers with significant fundamental physical interests and have diverse applications in the glass industry and advanced materials. The composition is also similar to natural basaltic magmas with the NBO/T (nonbridging oxygen per silica tetrahedron) value of 0.67 at 1 atm. Although glass of this composition has been

10.1021/jp804458e CCC: $40.75  2008 American Chemical Society Published on Web 08/20/2008

Structure of Calcium-Sodium Silicate Glasses

J. Phys. Chem. B, Vol. 112, No. 37, 2008 11757

manufactured since the Roman Empire and has been studied extensively with various experimental and theoretical methods, the nature of polymerization and the Ca-Na mixing in the CNS glass at 1 atm was only recently determined with a direct measurement of their oxygen environments using O-17 NMR (see ref 23 and the references therein). Recent developments in high-resolution solid-state 17O NMR provide exquisite detail on the distribution and quantification of oxygen clusters (hence, NBO and BO environments) at high pressures.13,24,25 In particular, the 2D O-17 3Q (triple quantum) magic-angle spinning (MAS) NMR techniques separate, along one dimension, the incompletely averaged and overlapping quadrupolar (MAS dimension) spectra responsible for broadening from the purely isotropic dimension,26-28 where individual oxygen clusters are clearly revealed both at ambient and high pressures.13,23,24,29-32 The study on the structure of mixed cation silicate glasses with two types of network modifying cations at high pressure allows us to explore the pressure-induced structural changes in several distinct atomic configurations due to Na-Ca mixing in glass network: As there are a number of chemically different NBO environments in Ca-Na mixed silicate glasses, the unknown pressure-dependence of NBO fraction with composition remains to be explored. Here, we explore the structure of Na-Ca mixed silicate glasses at high pressure using 2D O-17 3QMAS NMR and present previously unknown structural details of pressureinduced changes in their nonbridging oxygen environments. We then discuss the effect of melt structure at high pressure on the element partitioning between melts and other phases using these new experimental NMR data with emphasis on the oxygen cluster environment at high pressure. 2. Experimental Section Sample Preparation. CNS glass [(Na2O0.75CaO0.25).3SiO2] was synthesized from Na2CO3, CaCO3, and 40% 17O-enriched SiO2 with 0.2 wt.% of Gd2O3 added to reduce the spin-lattice relaxation time. The 40% enriched O-17 enriched SiO2 was synthesized by hydrolyzing SiCl4 with O-17 enriched water.25,32 The starting CNS glass was synthesized at 1 atm by fusing these mixtures for an hour at 1773 K in an Ar environment (in a single batch) in the Pt crucible and was loaded in a multi-anvil apparatus with a 18/11 (octahedron edge length/ truncated edge length of the anvils) assembly. The CNS glass was then fused at 2100 K for about 10 min, and the CNS melt was quenched to glasses at 6 and 8 GPa. The cooling rate is estimated to be about 500 C/s in the first few seconds.3 NMR Spectroscopy. O-17 3QMAS NMR spectra were collected on a CMX Infinity 300 spectrometer (7.1 T) at a Larmor frequency of 40.7 MHz for 17O with a 2.5 mm ZrO2 rotor in a Chemagnetic double resonance MAS probe. 17O 3QMAS NMR spectra were collected using a shifted-echo pulse sequence [3.3 us-τ (t1delay)-1.3 us-τ(delay)-20 us].26,31 The recycle delay for 17O MAS and 3QMAS NMR is 1 s and a magic-angle sample spinning speed of 19 kHz was employed. All spectra were referenced to tap water. Radio frequency (rf) field strength for the first hard pulse is about 120 kHz. Scaled dwell time in t1 dimension is 51.67 µs and about 70 FIDs are necessary to have processed spectra free from truncations without linear predictions. The t2 dimension is apodized with an appropriate Gaussian weighting function of about 200 Hz. A detailed description of the NMR methods is given elsewhere.26 We note that the newly made sample in the current study was analyzed at 7.1 T, while the previous result for CNS glass at 1 atm was collected at higher static field of 9.4 T.23 An identical

Figure 1. O-17 3QMAS NMR spectra for [(Na2O0.75CaO0.25).3SiO2] (CNS) glasses quenched at different pressures as labeled. Contour lines are drawn at 5% intervals from 13% to 93% of relative intensity, with added lines at the 4% and 6.5% and 9% levels to better show lowintensity peaks.

processing method with the previous study besides axis reversal was used for raw time domain 2D NMR data.23 Results and Discussion O-17 NMR Results. Figure 1 shows the two-dimensional, contoured, O-17 3QMAS NMR spectra for calcium-sodium trisilicate glasses (CNS) quenched at 1 atm, 6, and 8 GPa, respectively. The peak at ∼-25 ppm in the isotropic dimension for the 1 atm quenched melt corresponds to the NBO coordinated by [4]Si and several Na (Na-NBO, i.e., NasOs[4]Si). A mixed cation NBO peak [{Ca, Na}sOs[4]Si, oxygen coordinated by both Na, and Ca as well as [4]Si] is observed at ∼-42 ppm in the isotropic dimension; This peak is well-resolved from

11758 J. Phys. Chem. B, Vol. 112, No. 37, 2008 the pure Na-NBO peak allowing for unambiguous assessment of the effect of pressure on both types of NBOs. The intense peak at ∼-56 ppm in the isotropic dimension corresponds to bridging oxygen, [4]SisOs[4]Si cluster. These NBO and BO peak assignments are consistent with the previous NMR study at 9.4 T.23 A peak at ∼-75 ppm, in the isotropic dimension, develops only at high pressures and is assigned to [4] SisOs[5, 6]Si.7,25 The increase in the abundance of [4]SisOs [5, 6]Si compensates for the reduction of all oxygen species, a trend observed previously in quenched sodium silicate melts.7,13 Figure 2A presents the projections along the isotropic dimension (Figure 1) of the O-17 3QMAS NMR spectra for the CNS quenched melts at different pressures. A systematic shift in the peak positions of all of the oxygen cluster types to lower frequencies is observed with increasing pressure, reflecting a progressive increase in SisO bond length and closer proximity of network modifying cations to the NBOs. Significantly, although the full width at half-maximum (fwhm) of the [4]Si-O[4]Si peak also increases with pressure reflecting a widening of the bond length distribution function with pressure, this does not obscure the ability to resolve the primary speciation. As is shown in Figure 2B, these isotropic projection spectra are fit well with four Gaussian peaks resolving clearly the wellresolved oxygen clusters. The previous reports showed that the peak broadening in the oxygen sites with pressure in 2D O-17 NMR spectra is mostly due to changes in isotropic chemical shifts rather than due to variations in quadrupolar coupling constants.13,33 It was shown that topological disorder and thus topological entropy due to bond angle and length distribution in oxide glasses often increases with increasing pressure, partly due to the increase in polyhedral distortion with pressure (see13 for detailed information). We also note that the presence of minor fraction of NasOs[5]Si in binary Na-silicate glasses was reported at high pressure,3 implying that there could a minor fraction of {Ca,Na}sOs[5]Si or NasOs[5]Si in the NBO peaks of CNS glasses at high pressure. While it is difficult to quantify their fractions, the robust fitting results with two Gaussian peaks representing {Ca,Na}sOs[4]Si or NasOs[4]Si imply that the fraction of NBO with highly coordinated Si is unlikely to be significant (Figure 2B). While it is rather difficult to collect Si29 MAS NMR spectra for the samples studied here, due to low signal/noise ratio at relatively low magnetic field (7.1 T), future Si-29 MAS NMR (at 9.4 T or higher magnetic field) will be potentially useful to identify five- and six-coordinated Si in the glasses. The variation in the percentage of various oxygen clusters in CNS quenched melts with pressure is presented in Figure 3. Because the NMR signal intensity in the 3QMAS NMR experiment is controlled by efficiencies in triple quantum excitation and single quantum reconversion that are mostly dependent on magnitude of the interactions between nuclear quadrupolar moment and electric field gradient,27 the observed NMR signal intensity is scaled to take into account the magnitude of the quadrupolar interactions characteristic for each oxygen cluster type by numerical simulation.34 This approach has been successfully applied to quantify oxygen cluster populations in oxide glasses: the calculated (and thus scaled) fractions were consistent with the experimental results.13,31,33,35 These scaled concentrations of Na-NBO, mixed-NBO, and [4]SisOs[4]Si clusters for CNS glasses at 1 atm are 12.6, 14.1, and 73.3%, respectively. Note that the percentage of Na-NBO to total NBO is 47%, slightly greater than a previous assessment, ∼43%,23 where scaling had not been employed. Significant cation ordering is evident by presence of extensive Ca-Na pairs

Keun et al.

Figure 2. (A) Total isotropic projection of O-17 3QMAS NMR spectra for {CaOsNa2O}s3SiO2 glass quenched from melts at different pressures as labeled. (B) Fitting results for {CaOsNa2O}s3SiO2 glass using 4 Gaussians representing Na-NBO, Mixed-NBO, [4]SisOs[4]Si, and [5, 6]SisOs[4]Si as labeled.

around NBO [∼53-57% of {Ca,Na}sOs[4]Si]. This is consistent with the earlier O-17 NMR study for Ca-Na glasses with varying Ca/Na ratio at 1 atm (9.4T).23

Structure of Calcium-Sodium Silicate Glasses

J. Phys. Chem. B, Vol. 112, No. 37, 2008 11759

Figure 3. Variation of the oxygen cluster population with pressure in {CaOsNa2O}s3SiO2. Closed diamonds, squares, circles, and diamonds refer to NasOs[4]Si, [4]SisOs[4]Si, [5,6]SisOs[4]Si, and CasOs[4]Si, respectively. Open circles are a fraction of total NBO with pressure. The thick curves show the trend lines connecting experimental data.

With increasing pressure, the percentage of Na-NBO decreases down from 12.6% (1 atm), to 9.5% (6 GPa), and to 7.3% (at 8 GPa). The percentage of {Ca,Na}sOs[4]Si also decreases, albeit less dramatically, from 14.1% (1 atm), to 13.6% (6 GPa) to 9.4% (at 8 GPa). Note that whereas the fraction of Na-NBO decreases continuously from 1 atm up to 8 GPa, reduction in the mixed cation NBO fraction is only significant (Figure 4A) moving from 6 to 8 GPa. Such differential densification among NBOs with pressure is unexpected and no theoretical prediction has been made so far that would have anticipated this result. Along with the reduction in NBOs, the percentage of the [4]SisOs[4]Si BO fraction also decreases with increasing pressure, down from 73.3% at 1 atm, to 68.1% at 6 GPa and to 64% at 8 GPa. Compensating the decreases in NBO and [4]SisOs[4]Si, the BO cluster, [4]SisOs[5,6]Si, increases from 0 at 1 atm to 6.8% and 15.8% at 6 and 8 GPa, respectively. The following schemes, therefore, account for the observed changes in oxygen clusters in the CNS glasses quenched from melts with increased pressure from 1 atm to 8 GPa: [4]

SisOs[4]Si + NasOs[4]Si ) [4]

[4]

SisOs[5]SisOs[4]Si + Na+ * (1)

SisOs[4]Si + {Ca, Na}sOs[4]Si ) [4]

SisOs[5,6]SisOs[4]Si + Ca2+ * + Na+ * (2)

where Na+* and Ca2+* now charge balance the under-bonded BOs associated with the high coordinated silica species ([5]Si, [6]Si). Pressure-induced coordination transformation is thus accompanied by the reduction of NBO as explicitly shown in eqs 1 and 2. Detailed information of pressure-induced coordination transformation in Si as well as Al can be found elsewhere.4,10,36 Pressure-induced shifts to the right in eqs 1 and 2 increase the extent of melt polymerization in the glasses with pressure and constitute the primary densification mechanisms.3 These results are significant, due to the fact that the pressuredependence on the oxygen site distributions [e.g., (∂XNBO/∂P)T] will affect the dynamic properties (viscosity) of melts in the Earth’s interior, as discussed in detail in the previous study.4,10,36

Figure 4. (A) Variation of NBO cluster with pressure in CNS glasses with pressure. (B) Populations of NBO fractions in binary silicates (Na2O:SiO2 ) 1:3, NS3 and Na2O:SiO2 ) 1:4, NS4), ternary mixed cation silicates, and aluminosilicate glasses (Na2O:Al2O3:SiO2 ) 1.5: 0.5:6, NAS) as labeled.3,7,39

The associated changes in coordination environment of framework units such as Si and Al [e.g., (∂[5, 6]Si/∂P)T] must be expected to affect the trace element portioning.9 Effect of Composition on the Pressure-Induced Changes in the Degree of Polymerization, (DXNBO/DP)T. While NBO/T is a convenient and popular way to describe the polymerization

11760 J. Phys. Chem. B, Vol. 112, No. 37, 2008 states of melts at 1 atm, the NBO fraction decreases and the coordination transformation is prevalent with increasing pressure. The NBO fraction is thus more sensible choice of the degree of polymerization in the melts at high pressure, as it can be directly probed by 17O NMR.3,12 In order to show that these results are both general and compositionally complex, the percentage of alkali NBO for different model compositions of silicate glasses as a function of pressure are presented in Figure 4B. Note that while all of the glasses studied exhibit a general trend of decreasing NBO concentration with pressure, the details of this relationship (∂XNBO/∂P)T is strongly dependent on the composition of melts [e.g., the degree of melt polymerizations total (Na + Ca)/(Si + Al) ratio, Si/Al ratio, and types of network modifying cations, Na/Ca.] at 1 atm. It has been suggested that this pressure derivative, (∂XNBO/∂P)T, may explain the nonlinear pressure dependence of transport properties (e.g., diffusivity and viscosity).3,10,37,38 Comparing these new CNS glass data with a sodium trisilicate (NS3) glass with nearly the same degree of depolymerization at ambient pressure7 reveals that the addition of Ca dramatically increases |(∂XNBO/∂P)T|, suggesting strong compositional dependence (the degree of polymerizations, types of framework and nonframework cations). While further experimental data are necessary and it is difficult to generalize the effect of network modifying cation on the variation of NBO with pressure, the nonframework cation with larger field strength (charge/ionic radii, e.g., Ca2+) can apparently stabilize the highly coordinated Si or Al more effectively. Thus, the formation of SisOs[5,6]Al is more favorable in Ca-Na silicate glasses than in Na-silicate glasses, leading to a rapid decrease in the NBO fraction in the former with pressure. The difference between sodium trisilicate glasses (NS3) and sodium tretrasilicate glasses (NS4) has been previously discussed:39 with increasing Na content in binary silicate glasses, there is an increase in steric hindrance for the coordination transformation. A more drastic change in NBO fraction with pressure can thus be observed in the sodium silicate glasses with lower Na2O content.25 The degree of polymerization increases by adding Al in sodium binary silicate melts. The presence of Al in the silicate melts at high pressure adds another complexity in the densification mechanisms. Recent Al-27 NMR study at high field MAS NMR clearly showed the effect of cation field strength on pressureinduced changes in Al coordination environments.36 Implication for Element Partitioning between Silicate Melts and Crystals. The specific role that pressure and XNBO play in partitioning of elements between melts and crystals at high pressure is far from well understood, even though it is clear that melt composition and structure play important roles in controlling partitioning, as suggested from recent experimental studies.17-19,40-42 As the partitioning coefficient at high pressure can be dependent on many factors discussed above, it is challenging to provide complete and quantitative atomistic model incorporating melt contribution due to the numerous variables required to be understood and the fact that these cannot yet be constrained by experimental data on melt structure at high pressure: full atomic configurations and thus short to medium range order around element of interest in melts are not well-known. In the current discussion of crystal-melt partitioning at high pressure, we thus tried to focus on the effect of available melt structure at high pressure obtained from the current O-17 NMR experiment [i.e., the pressure-induced decrease in nonbridging oxygen fraction, (∂XNBO/∂P)T] on crystal-melt partitioning. The crystal-melt partitioning coefficient of element Dmelt-crystal of a given element m can be expressed as follows:

Keun et al.

ln

crystal-melt Dm ∝-

xs,crystal xs,melt [Gm - Gm (XNBO)] kT

(3)

where k and T are the Boltzman constant and temperature (K), respectively. Gxs,crystal and Gxs,melt are the excess lattice energy m m to incorporate element m in a given lattice or interstitial sites in the crystal and available sites for element, m, in melts, respectively. The first term is related to the interaction potentials (electron-electron interactions) between element m and lattice site. Earlier pioneering modeling efforts found it convenient to divide this excess energy into a steric (lattice strain) and charge potential accompanied by lattice relaxation after substitution.16 Gxs,melt is a less satisfactorily defined, complex function of the m atomic configurations around m and their mutual interactions as a function of composition (including XNBO), temperature, and pressure. Gxs,melt includes the sum of the pair potentials [u(r), m mostly charge-charge interaction] for a given atomic configurations g(r), the radial distribution function around m] and, thus, Gxs,melt may be expressed as 4π∫∞0 u(r)g(r) r2dr, capturing m complex atomic environments (topology as well as composition) around element m. Considering the pressure dependence of Gxs,melt and Gxs,crystal , an expression for the pressure dependence m m of crystal-melt partitioning may, therefore, be derived, e.g.,

(

crystal-melt ∂ln Dm ∂P

-

)

T

∝

xs,crystal [(∂Gm ⁄ ∂P)T - (∂Gmxs,crystal ⁄ ∂XNBO)T(∂XNBO ⁄ ∂P)T] kT (4)

where the first term on the right side of the proportionality encompasses all aspects of the pressure dependence on the crystallographic configurations and lattice structures in relation to inclusion of elements m; this term can be derived theoretically. The remaining contributions due to melt structure are divided into a XNBO dependence of excess energy in the melts (∂Gxs,melt /∂XNBO)T and effect of pressure on XNBO, (∂XNBO /∂P)T, m highlighting the critical importance of knowing NBO percentage with pressure. Gxs,melt is likely to decrease with increasing XNBO as network m rigidity of melts decreases with NBO: the energy penalty to include m in the melt thus decreases with XNBO and the sign of this term can be negative. Its sign can also depend on the network disorder in the melts, types of network modifying cations, and the interaction among network modifying cations, and the element of interest, suggesting complex and nonlinear nature of Gxs,melt with NBO fraction as manifested in complex m NBO dependence of Dcrystal-melt .17 m From the current O-17 NMR data, it is clear that the sign of (∂XNBO /∂P)T term is negative. Assuming a likely behavior of xs,melt crystal-melt negative (∂Gm /∂XNBO)T value, Dm is expected to increase due primarily to a decrease in XNBO with increasing pressure, suggesting that a knowledge of (∂XNBO/∂P)T would be important if the effect of pressure on partition coefficients is to be determined. Currently, (∂XNBO/∂P)T can only be determined experimentally, addressing the importance of melt structure at high pressure in describing the variation of partitioning coefficient with pressure. We note again that the above discussion was only for a qualitative suggestion on the link between experimental atomic structure at high pressure [(∂XNBO /∂P)T] and macroscopic thermodynamic property. It is important to note that in order to measure (∂XNBO/∂P)T using NMR one is restricted to studying glasses quenched from melts. The structure of glass studied here represents the atomic

Structure of Calcium-Sodium Silicate Glasses configuration of a supercooled liquid frozen at the glass transition temperature at high pressure, where the structural features at high pressure are quenchable. The question naturally arises as to how does the structure frozen at the glass transition temperature differs from that at realistic melt temperatures. This is a long standing question that has been addressed by studying glasses quenched at different cooling rates, thereby establishing a range of so-called “fictive temperatures”. Contrasting changes in molecular structure across these glass transition temperatures enables one to confidently explore the effect of temperature on parameters such as XNBO. What is known is that for alkali silicate quenched melts (at ambient P) XNBO does not change significantly with increasing glass transition temperature.43 Thus, it is likely that (∂XNBO/∂P)T measured in these quenched melts are closely representative of the melts at elevated temperatures. We also note that ongoing computational modeling utilizing molecular dynamics simulation or ab initio molecular orbital calculations would provide complementary interpretation of the atomistic origins of crystal-melt partitioning at high pressure. Acknowledgment. S.K.L. was supported by Korea Science & Engineering Foundation through the National Research Laboratory Program (2007-000-20120-0). We thank Prof. P. Grandinetti for RMN software for 2D NMR data processing. We gratefully acknowledge the support of the W.M. Keck foundation and the NSF for the NMR facility at the Geophysical Laboratory. References and Notes (1) Mysen, B. O.; Richet, P. Silicate Glasses and Melts, Properties and Structure; Elsevier: Amsterdam, 2005. (2) Kushiro, I. J. Geophys. Res. 1976, 81, 6347. (3) Lee, S. K.; Cody, G. D.; Fei, Y.; Mysen, B. O. Geochim. Cosmochim. Acta 2004, 68, 4189. (4) Poe, B. T.; McMillan, P. F.; Rubie, D. C.; Chakraborty, S.; Yarger, J. L.; Diefenbacher, J. Science 1997, 276, 1245. (5) Hemley, R. J.; Mao, H. K.; Bell, P. M.; Mysen, B. O. Phys. ReV. Lett. 1986, 57, 747. (6) Lee, S. K.; Eng, P. J.; Mao, H. K.; Meng, Y.; Newville, M.; Hu, M. Y.; Shu, J. F. Nat. Mater. 2005, 4, 851. (7) Lee, S. K.; Fei, Y.; Cody, G. D.; Mysen, B. O. Geophys. Res. Lett. 2003, 30, 1845. (8) Wolf, G. H.; McMillan, P. F. Pressure Effects on Silicate Melt Structure and Properties. In Structure, Dynamics, and Properties of Silicate Melts; Stebbins, J. F., McMillan, P. F., Dingwell, D. B., Eds.; Mineralogical Society of America: Washington, D.C., 1995; Vol. 32; pp 505. (9) Xue, X.; Stebbins, J. F.; Kanzaki, M.; Tronnes, R. G. Science 1989, 245, 962. (10) Yarger, J. L.; Smith, K. H.; Nieman, R. A.; Diefenbacher, J.; Wolf, G. H.; Poe, B. T.; McMillan, P. F. Science 1995, 270, 1964.

J. Phys. Chem. B, Vol. 112, No. 37, 2008 11761 (11) Allwardt, J. R.; Stebbins, J. F.; Schmidt, B. C.; Frost, D. J.; Withers, A. C.; Hirschmann, M. M. Am. Mineral. 2005, 90, 1218. (12) Lee, S. K. Geochim. Cosmochim. Acta 2005, 69, 3695. (13) Lee, S. K. J. Phys. Chem. B. 2004, 108, 5889. (14) Lee, S. K.; Cody, G. D.; Fei, Y.; Mysen, B. O. The Application of 17 O and 27 Al Solid-State (3QMAS) NMR to Structures of Non-Crystalline Silicate at High Pressure. In AdVances in High-Pressure Techniques for Geophysical Applications; Chen, J., Wang, Y., Duffy, T. S., Shen, G., Dobrzhinetskaya, L. F., Eds.; Elsevier: Amsterdam, 2005; pp 241. (15) Allwardt, J. R., Stebbins, J. F.; Schmidt, B. C.; Frost, D. J., The Effect of Composition, Compression, And Decompression on the Structure of Highpressure Aluminosilicate Glasses: An Investigation Utilizing 17 O and 27 Al NMR. In AdVances in High-Pressure Techniques for Geophysical Applications; Chen, J., Wang, Y., Duffy, T. S., Shen, G., Dobrzhinetskaya, L. F., Eds.; Elsevier, Amsterdam,2005; pp 211. (16) Blundy, J.; Wood, B. Nature 1994, 372, 452. (17) Mysen, B. O. Chem. Geol. 2004, 213, 1. (18) Walter, M. J. Science 1995, 68, 1186. (19) Kohn, S. C.; Schofield, P. F. Chem. Geol. 1994, 117, 73. (20) Toplis, M. J.; Corgne, A. Contrib. Mineral. Petrol. 2002, 144, 22. (21) Murthy, V. R.; van Westrenen, W.; Fei, Y. Nature 2003, 423, 163. (22) Blundy, J. D.; Wood, B. J. Nature 1994, 452. (23) Lee, S. K.; Stebbins, J. F. J. Phys. Chem. B. 2003, 107, 3141. (24) Allwardt, J. R.; Schmidt, B. C.; Stebbins, J. F. Chem. Geol. 2004, 213, 137. (25) Xue, X.; Stebbins, J. F.; Kanzaki, M. Am. Mineral. 1994, 79, 31. (26) Baltisberger, J. H.; Xu, Z.; Stebbins, J. F.; Wang, S.; Pines, A. J. Am. Chem. Soc. 1996, 118, 7209. (27) Frydman, I.; Harwood, J. S. J. Am. Chem. Soc. 1995, 117, 5367. (28) Amoureux, J.-P.; Bauer, F.; Ernst, H.; Fernandez, C.; Freude, D.; Michel, D.; Pingel, U.-T. Chem. Phys. Lett. 1998, 285, 10. (29) Dirken, P. J.; Kohn, S. C.; Smith, M. E.; Vaneck, E. R. H. Chem. Phys. Lett. 1997, 266, 568. (30) Stebbins, J. F.; Xu, Z. Nature 1997, 390, 60. (31) Lee, S. K.; Stebbins, J. F. J. Phys. Chem. B. 2000, 104, 4091. (32) Xu, Z.; Maekawa, H.; Oglesby, J. V.; Stebbins, J. F. J. Am. Chem. Soc. 1998, 120, 9894. (33) Lee, S. K.; Stebbins, J. F. Geochim. Cosmochim. Acta 2006, 70, 4275. (34) Bak, M.; Rasmussen, J. T.; Nielsen, N. C. J. Magn. Reson. 2000, 147, 296. (35) Lee, S. K.; Lee, B. H. J. Phys. Chem. B. 2006, 110, 16408. (36) Allwardt, J. R.; Stebbins, J. F.; Terasaki, H.; Du, L. S.; Frost, D. J.; Withers, A. C.; Hirschmann, M. M.; Suzuki, A.; Ohtani, E. Am. Mineral. 2007, 92, 1093. (37) Angell, C. A.; Cheeseman, P. A.; Tamaddon, S. Science 1982, 218, 885. (38) Diefenbacher, J.; McMillan, P. F.; Wolf, G. H. J. Phys. Chem. B. 1998, 102, 3003. (39) Lee, S. K.; Cody, G. D.; Fei, Y.; Mysen, B. O. Chem. Geol. 2006, 229, 162. (40) Gaetani, G. A. Contrb. Mineral. Petrol. 2004, 147, 511. (41) Jaeger, W. L.; Drake, M. J. Geochim. Cosmochim. Acta 2000, 64, 3887. (42) Watson, E. B. Contrib. Mineral. Petrol. 1976, 56, 119. (43) Stebbins, J. F.; Farnan, I.; Xue, X. Y. Chem. Geol. 1992, 96, 371.

JP804458E