high pressures is essentiar ror undersranu,.', ''jffi i#ti ... · persed ruby grains were added for pressure calibration. The infrared absorption spectra were measured over the total

American Mineralogist, Volume 77, pages 258-269' 1992

Raman and infrared study of pressure-induced structural changes in MgSiOroCaMgSirO6, and CaSiO, glasses

J. D. Kunrcxr*Geophysical Laboratory and Center for High-Pressure Research, Carnegie Institution of Washington,

Washington, DC 20015, U.S.A.

R. J. Hrvlr-BvGeophysical Laboratory and center fo:,l;:i;3:

Tat?.."r1:"_egie rnstitution of washington,

and Department of Earth and Planetary Sciences, ih" John, Hopkins University, Baltimore, Maryland 21218, U.S.A.

A. M. HorrvrnrsrpnDepartment of Geology, University of California, Davis, California 95616, U.S.A.

Ansrucr

Raman and midinfrared spectra of MgSiOr, CaMgSirOu, and CaSiO, glasses have beenmeasured to 45 GPa in a diamond cell. The principal pressure-induced changes in spectraobserved are reversible, so that charucteization of the high-density states of these mate-rials requires in situ measurements at high pressure. The dominant change in all threeglasses is a marked shift of the midfrequency Raman bands (at 625-650 cm-') to higherfrequencies with increasing pressure. This indicates that the main compression mechanismis associated with a decrease in the average intertetrahedral Si-O-Si angles, as found in

other silicate materials. The intensities of high-frequency Raman and infrared bands at900-1200 cm ' decrease with pressure relative to those oflower frequency bands. This isattributed to distortions of SiOo tetrahedra. However, the Raman and infrared spectraprovide no direct evidence for a distinct t4rsi to t6rsi coordination change at room temper-ature in these glasses over the pressure range of this study. We present evidence thatpressure-induced structural changes in glasses, including changes in cation coordination,can be affected by sample preparation and trace impurities.

INrnooucrroN ing in the Si-O-Si chains, which can maintain a bonding

Information on the structure of molten silicate topologv based on si in tetrahedral coordination to high

high pressures is essentiar ror undersranu,.', ''jffi i"#ti':jil:T#J"ilg5l$:*l?J,11;l?:"1tli:HH:ing the phvsical and chemical properties of deep-seated ;i;;,;i;J5.airrutio.r. The relative contriburions of rhesemelts within the Earth. It is now recognized,h*.::t^.j_11- ;;d;;;;;pression mechanisms are expected to con-

:'#: :j"tffiJ;ttfi:T:'}|ffi'ffljili.ffi1,'i"ffi'n ""'iii" i""'ties achieved in morten slicates in rhe Earth's

they are inequilibrium 1e.g., nlgaen et al., 1984, 1988; -ln^'j^- -^

Heinz and Jeanloz, 1987; Knittle and Jeanloz, r9t#1. iiir .^11T:1t:"ttering and infrared absorption spectroscopv

has important consequences for the formatior" ̂ ; -^^- have been particularly useful for obtaining structural in-

chemical mantle reservoirs and planetary u,n"."t"itt"ill. l:tll,tl:: il silicate glasses (e'g'' Hibben' 1935)' leading

The mechanism by which silicate melts can ach'^-':;;-; to insights into the properties of melts at higher temper-

densiries is incompretery undersrood. tt i, o",Ji"J:1il*"1 ffJj,t tjl',""T"TJfffi*"ff#"::t?;:',ffi::XTJtff::

the silicate liquids and glasses of these "ornooti:1:T.Y:: ;;;":;; structure of these glasses (Sweet and white,

dergo pressure-induced changes in cation coord-ination i;;;;;. and White, 1977; Mysen et al., l9g0; Mc_

analogous to those documented for crystallint *",t::*t rtiriLrlllaaa, l9g9). one of the advantages of Raman

high pressure (waff, 1975; Matsui and Kawamul ?^1?' #ft;;; techniques relative ro orher merhods (wong

These materials may also be compressed bv angle bend-

""i erl"ii, rg76; cusack, lggT) is the possibility of using

them in high-pressure studies with the diamond cell (Ja-

*presentaddress: DivisionofGeologicalandplanetarySci- 11113"1-]-983; Hemleyetal' ' 1987;Mao' 1989)'Asa

ences, California Insritute of Technology, Pasadena,'dii";; result, structural changes in silicate glasses under deep

9125, U.S.A. Earth conditions can be examined without recourse to

oo03-004x/92l0304-0258$02.00 258

pressure quenching, which may not yield structural statesthat are representative of the high-pressure form of theglass.

Vibrational spectroscopic measurements on silicateglasses at high pressure in the diamond cell documentthat significant changes in local structure occur above l0GPa (e.g., Hemley et al., 1986; Williams and Jeanloz,1988). The results indicate that large distortions of SiOotetrahedra, and possibly an increase in Si coordination,occur at high pressure and room temperature in silicateglasses. Furthermore, the principal spectral changes in-duced by pressure are reversible and thus require in situhigh-pressure measurements to be observed (Hemley etal., 1986; Williams and Jeanloz, 1988). There is a needto examine this behavior for a variety of chemical com-positions relevant to the lower mantle. These high-pres-sure, vibrational spectra provide information on possibleshort- and intermediate-range order in calcium magne-sium metasilicate glasses under very high compressionand may be used to infer the structure of silicate melts inthe mantle.

In the present study, in situ Raman and midinfraredspectroscopic measurements of MgSiO, (enstatite),CaMgSirOu (diopside), and CaSiO, (wollastonite) glasseswere made to lower mantle pressures. MgSiO, perovskiteis thought to be the dominant end-member phase in thelower mantle (Liu, 1975; Yagi et al., 1978), and CaSiO,perovskite may be a secondary constituent (Liu and Ring-wood, 1975; Mao et al., 1989). An investigation of thevitreous state of these compositions provides a startingpoint for understanding the structure of mantle melts.Substitution of Ca for Mg is also used to clarify the roleof the network-modifying cations of the structure in theglasses. Because Raman and infrared spectroscopies de-tect different vibrational modes in the material, the useof both provides complementary information on the vi-brations in the glass that can be related to the high-pres-sure structure. These measurements are also comparedwith results from molecular dynamics simulations ofcompression in MgSiO, glass and melt.

ExpnnrrrnurAt, METHoDS

Glasses were prepared in our laboratory from spectro-scopic-grade oxide and carbonate powders mixed in anagate mortar and pestle under ethanol for 15 min. Thesamples were then decarbonated at 1000'C for over 4 h.Sintered samples were melted in a Pt-wound furnace atleast 25 "C above the melting point for I h in a Pt capsulewith 5-mm diameter, quenched in HrO, ground, and re-melted. Each sample was analyzed optically to ensure thatthe glasses were free of crystallites. Compositions weremeasured by electron microprobe to confirm the stoichi-ometries and were within +0.2 wto/o of the ideal com-positions. A small amount of grinding in an agate mortarand pestle was necessary to ensure that grain sizes oflessthan l0 rlm were available for the diamond cell. We alsomeasured Raman spectra of two MgSiO3 glasses preparedindependently ar 1620'C (by H. S. Yoder) and 1650 "C

259

(by J. Bass) and quenched in HrO at 0.1 MPa (l bar).Another MgSiO. glass, quenched at 0.2 GPa and 1650 "Cin a piston-cylinder apparatus (by F. R. Boyd and J. En-gland) at the Geophysical Laboratory, was used in theRaman experiments. All samples used for the high-pres-sure experiments were free of crystallites, as determinedby optical microscopy and Raman spectroscopy.

Samples were compressed in megabar-type Mao-Belldiamond-anvil cells using T-301 stainless steel gasketspreindented to 10-20 GPa (Mao, 1989). No pressure-transmitting media were used in most experiments in or-der to achieve the strongest possible signal. This was im-portant because of the very low Raman scattering crosssection of these glasses (particularly MgSiO.). Selectivemeasurements were performed on samples compressedin Ar pressure-transmitting media to examine effects ofnonhydrostatic stress (e.g., Hemley et al., 1986, 1987).Pressure measurements were made with the standard rubymethod (Mao et al., 1986). Precision was within +0. IGPa on any given ruby grain in the sample. Six separatepressure measurements were made for each spectrum todetermine the pressure variation in the sample. With rubydispersed throughout the sample, pressures were mea-sured on a grid of points over the entire gasket openingand averaged to obtain the quoted sample pressure. Stan-dard deviations in the pressure measurements across thesample were approximately l0o/o of the average pressureat the maximum loads.

Raman spectra were measured with a micro-opticalsystem designed for diamond-cell samples with weak Ra-man scattering cross sections (Hemley et al., 1986, I 987).The spectrometer consisted of a triple spectrograph (SpexTriplemate) and intensified diode-array detector (optical-multichannel analyzer, Princeton Instruments). A win-dow of 55 nm typically centered at 539.5 nm was used.Raman spectra were excited with the 514.5-nm line of anAr* laser (Spectra Physics model 165) at a power of 5-50 mW focused to a diameter of t3-10 pm. An aper-ture giving a sampling area of approximately 5 x l0 pmwas used to discriminate against background signal fromthe diamond anvils (Hemley et al., 1987). Spectra weretypically signal averaged from Vz to I h. Resolution of thespectra was approximately l0 cm-'with a grating of 600grooves/mm. Overlapping bands in the several sets ofspectra were curve fitted with Voight profiles using least-squares techniques.

Samples for the infrared measurements were preparedby crushing the sample between the diamond anvils intoa platelet of l-2 pm thickness (Hofmeister et al., 1989).A layer of CsI was placed on top of the sample, and dis-persed ruby grains were added for pressure calibration.The infrared absorption spectra were measured over thetotal sample area (i.e., diameter -200 1rm), whereas in-dividual Raman measurements were made on small por-tions of the sample (i.e., < l0 pm). Infrared spectra weremeasured on a Nicolet 7199 FT-IR spectrometer with aliquid Nr-cooled HgCdTe detector and a custom-de-signed beam condenser (Hofmeister et al., 1989). Three

KUBICKI ET AL.: SILICATE GLASSES AT HIGH PRESSURE

260 KUBICKI ET AL.: SILICATE GLASSES AT HIGH PRESSURE

MgSiOg

660I

990

I ross

CaMgSirOu

360660I

990

880 lroso

CaSiO, tfl98t)

875t ,

200 400 600 800 1000 1200

ITavenumber (cm-l)

Fig. 1. Raman spectra of MgSiOr, CaMgSirOu, and CaSiO,glasses at 0.1 MPa (l bar, 300 K). Spectra were measured ondiamond-cell sized samples with the micro-Raman spectrometerprior to compression.

sets of 4000 scans were averaged for each spectrum, whichhad a resolution of 2 cm-'. Background spectra were re-

corded using an empty diamond cell.

Rnsur,rsAmbient pressure spectra

Ambient pressure (0.1 MPa) Raman spectra of thesamples are presented in Figure l. Peak positions of thelow- and midfrequency bands agree with previously mea-sured spectra to within +2 cm' (Mysen et al., 1982;McMillan, 1984b). No temperature effects related tosample preparation were apparent in the 0. I MPa spectra(within the present experimental resolution). However,the high-pressure response of an MgSiO, sample preparedin a piston-cylinder apparatus at 0.2 GPa differed mark-edly from that of samples quenched at ambient pressure,as will be discussed below. The spectra of CaSiO, andCaMgSirOu have broad bands at 350-370 cm-'. The in-tensities of these bands increase with Ca content. In allthree compositions, a band of intermediate relative in-tensity is observed between 600 and 700 cm-' at ambientpressure. This peak has an asymmetric line shape, par-ticularly in the spectrum of CaSiO. and is broadest in theCaMgSirOu composition. A band from 800 to 1200 cm '

is observed in all three spectra, with shoulders flankingthe most intense peak at approximately 900 and 1100

MgSiO3 1070

* 800

CaMgSi2Ot

500 700 900 1100 1300

lYavenumber (cm-l)

Fig. 2. Mid-infrared absorption spectra of MgSiO', CaMg-SirOu, and CaSiO, glasses measured at low pressures (<2 GPa)in a diamond-anvil cell under nominal load (300 K). The aster-isk denotes a peak due to ruby used for pressure calibration.

cm r. The shoulders are more highly resolved from themain band for CaSiO, in comparison with the other twocompositions. Mysen et al. (1982) and McMillan (1984b)have shown that this high-frequency band may be fittedwith contributions from five peaks at 870, 9 I 5, 970, I 060,and I 150 cm ' .

Midinfrared spectra of the glasses at low pressure areshown in Figure 2. The spectra are similar and consist ofthree broad bands at 400-600, 600-800, and 800-1250cm '. All three bands shift to higher frequency with in-creasing Mg content of the glass. This shift is most pro-nounced in the lowest frequency band, which varied from480 cm-' in CaSiO, to 585 cm I in MgSiO, glass. Thespectrum of the glasses at low pressure is similar to thosereported previously (Williams and Jeanloz, 1988).

High-pressure spectra

In situ, high-pressure Raman spectra are presented inFigure 3. The Raman scattering cross sections of theseglasses are low, but spectral peaks are well above thebackground signal from the diamond cell to pressuresabove 30 GPa. The high-pressure spectra reveal largepressure effects on several bands. The largest changes ob-served occur in the 645-650 cm-r bands (0.1-MPa fre-quencies). The positions of these bands shift linearly withpressure up to at least -30 GPa. Above this pressure, the

Q)OtrdIti

IA!

0)

+)d

o)t!P.

>\{-)otrq)

+)tr

KUBICKI ET AL.: SILICATE GLASSES AT HIGH PRESSURE 261

>'+)

@

0)+d

22 GPa 't- \r"1J

\ l ! \\4w

0.1 MPa

a

200 400 600 800 1000 1200

l?'avenumber ("*-t)

35 GPa

89 GPa

?3 GPA

18 cPa

12 GPa

0.1 MPa

b

200 400 600 800 1000 1200

Yfavenumber ("*-t)

,[0 GPa

26 GPa

16 GPa

10 GPa

0.1 MPa

200 400 600 800 1000 1200

IVavenumber ("*-t)

peaks merge with the high-frequency bands and cannotbe traced to higher pressure. The low-frequency Ramanbands observed in CaSiO. and CaMgSirO6 (350 and 370cm r) have weaker, positive pressure shifts. In addition,a decrease in the broad scattering intensity at low fre-quency (<600 cm-') is observed initially with increasingpressure. Weak shifts in the positions of the low-frequen-cy bands and relative intensities are observed above 20GPa (particularly in CaMgSirO. and CaSiOr). The pres-sure dependence ofthe high-frequency bands at 850-1 100cm I are slight but weakly negative in MgSiO, andCaMgSirOu. This effect partially arises from the increasein relative intensity of the lower frequency (860-885 cm-t)components. This is borne out by preliminary curve fit-ting ofthe high-frequency bands as a function ofpressure.Notably, the structure of the high-frequency envelopechanges with pressure, which indicates that polymeriza-tion ofthe SiOo tetrahedra is altered even at low pressures(and 300 K). This is particularly marked in CaSiO, glass.

High-pressure Raman spectra were similar for glassesobtained from different sources (within experimental un-certainty). An important exception was the MgSiO, glasssample synthesized 0.2 GPa. The evolution of the spectrameasured for this sample with increasing pressure is shownin Figure 4. The spectra were similar to those measuredfor other samples up to approximately 25 GPa. At higherpressures, however, a new band at approximately 400cm-r was observed (Kubicki and Hemley, 1988). Thispeak appeared reproducibly in this pressure range in threeseparate experiments with this glass sample (two experi-ments with no pressure-transmitting medium and onewith an Ar medium). With increasing pressure, the inten-sity of the new peak increased and the bandwidth de-creased. Upon decompression, the peak diminished inintensity but remained detectable to l8 GPa. Similar re-sults were obtained both with and without an Ar medi-um, indicating that shear stresses have only a small effecton the high-pressure spectral changes. Spectra measuredup to 45 GPa on the other samples of MgSiO, glass didnot show this feature (see below).

Midinfrared spectra of the three glasses are shown inFigure 5. Changes in the spectra are continuous from 0.1MPa to 25 GPa. Low-frequency infrared band shifts aresmall (0.8-1.2 cm'/GPa) in all three samples. All three

ts

Fig. 3. High-pressure Raman spectra of (a) MgSiO, (glass A)(b) CaMgSi,O", and (c) CaSiO, glasses from 0.1 MPa to 35-45GPa measured in the diamond cell (300 K). Measurements ofpressure from different ruby grains embedded in the samplesindicate that rhe variation in pressure across each sample is lessthan 100/o ofthe average pressure. However, the pressure varia-tion across the portion ofthe sample probed by the laser for eachmeasurement is estimated to be of order lolo (neglecting possibleuniaxial stresses, which are unknown). The asterisks denote ex-traneous Raman or fluorescence bands. The rise in intensity above1200 cm ' is due to the strong Tr, band ofthe diamond anvil(peak position of 1333 cm-rat 0.1 MPa).

>'a

h

vtca)

+

262

200 400 600 800 1000 1200'Wavenumber (cm-t)

Fig. 4. Raman spectra of MgSiO, glass prepared by Boyd andEngland under 0.2 GPa pressure in a piston-cylinder (glass B).The high-pressure peak near 400 cm-' is identified by arrows inthe two highest pressure spectra. The spectral changes with pres-sure are reversible but exhibited hysteresis on decompression.The peak positions and relative intensities of the spectra below25 GPa are similar to those observed in the other MgSiO, glasssamples (Fig. 3a). A larger amount of low-frequency scatter wasobserved at 0.1 MPa as a result of the lower optical quality ofthe starting sample, but the sample appeared homogeneous byoptical microscopy.

bands also decrease in absorbance with pressure. A sec-ond low-frequency mode (-580 cm ') in MgSiO, andCaMgSirOu glasses shifts at 4.41+ 6.1; cm-'/GPa. Mid-frequency infrared modes in MgSiO, and CaMgSirOu (800and 730 cm ', respectively) decrease in frequency withpressure. In CaSiO. glass, the frequency of the 710-cm-'band is nearly independent of pressure. A gradual de-crease in absorbance between 900 and 1250 cm-r occursin each glass, and the region becomes flat with respect tothe background above 35 GPa. The high-frequency bandbroadens and decreases in absolute and relative intensitywith increasing pressure. In each case, the frequenciesshift linearly with pressure (within experimental uncer-tainty). The high-pressure infrared spectra of CaMgSirOuglass are comparable to that reported by Williams andJeanloz (1988), although there appears to be some differ-ences that are likely associated with diferences in themethod of sample preparation and loading (cf. Hofmei-ster et al., 1989).


hrJogo+tr

DrscussroN

One ofthe key observations ofthe present study is thelarge difference between spectra measured at ambient andhigh pressures (e.g., above -10 GPa), particularly in theRaman measurements. The changes induced by pressuregreatly exceed the differences due to Mg-Ca substitutionand differences in spectra measured at 0. I MPa beforeand after pressure cycling. In general, the bands becomeweaker with pressure, partly as a result of sample thin-ning. However, the low-frequency infrared bands are lessaffected and, in some cases, become stronger with pres-

sure. In contrast, the higher frequency bands markedlydecrease in intensity with pressure. The intensity of thebroad, low-frequency (<500 cm r) Raman features ini-

tially decreases with increasing pressure. Generally, thevibrational bands have positive frequency shifts withpressure, though midfrequency infrared bands are an ex-ception. The compositional effect on the magnitude ofthese shifts appears slight.

The results indicate that significant changes in structureoccur in these glasses under pressures at room tempera-ture. Although a quantitative interpretation of spectralchanges requires calculation ofthe optical response oftheglasses from accurate structural models, the observed fre-quency shifts can be understood at a qualitative level interms of possible changes in bond lengths, bond angles,and cation coordination with pressure. We begin such ananalysis with an examination of vibrational band assign-ments and the available information on the structure ofthe glasses under ambient conditions'

Vibrational band assignments

The low-, mid-, and high-frequency bands have beenassociated with Ca and Mg coordination polyhedra, Si-O-Si tetrahedral linkages, and SiOo tetrahedra, respec-tively. Etchepare (1972) has associated peaks in the fre-quency range of 300-400 cm ' with Ca and Mg in octa-hedral coordination based on comparisons between glass

and crystal spectra. Eckersley et al. (1988a, 1988b) reportevidence from EXAFS and neutron-scattering experi-ments that Ca has well-defined octahedral coordinationin CaSiO, glass. Analysis of X-ray diffraction results byYin et al. ( I 983) indicates that each Mg ion is surroundedby four O ions at 2.08 A and two more at 2.50 A (see

also Waseda and Toguri, 1977).Hanada et al' (1988) alsoconclude from X-ray emission spectra that Mg is in four-fold coordination in amorphous MgSiO, films. As dis-cussed below, molecular dynamics (MD) simulations(Matsui et al., 1981; Kubicki and Lasaga, 1987, l99l)predict that Mg coordination is ill-defined in MgSiO, glass

and ranges from tetrahedral to octahedral. The Ramanintensity of the low-frequency peak also decreases fromCaSiO. to MgSiO3 in these glasses. These observationsmay be taken as evidence for a decreasing concentrationof network-modifying cations in octahedral coordinationwith increasing Mg content (e.g., Navrotsky et al., 1985).

Midfrequency Raman bands (600-700 cm-') of a va-

263

riety of metasilicate glasses have been assigned to vibra-tional modes characteristic of Si-O-Si intertetrahedrallinkages (Furukawa et al., 198 l; Matson et al., 1983; Mc-Millan, 1984a). Specifically, these studies indicate that amode with contributions from bridging O symmetricstretching and Si-O-Si angle bending gives rise to the Ra-man intensity at this frequency. The high-frequency bandhas been deconvoluted into five peaks associated withsymmetric stretches of Si-O bonds in different polymer-ized species of Si tetrahedra [i.e., Qo-Qo, where the su-perscript denotes the number of bridging O atoms in theSiOo tetrahedra (McMillan, 1984b)1. By this analysis, in-tensities ofthe individual contributions to this band arerelated to the polymerization state of the sample.

Infrared bands in the low-frequency region (below 600cm ') have been attributed to combinations of O-Si-Obending modes with Si in tetrahedral coordination(Moenke, 1974). A mode of asymmetric Si-O-Si anglebending and bridging O asymmetric stretching vibrationscauses the midfrequency absorption. High-frequencybands in the midinfrared have been assigned to asym-metric Si-O stretching modes in SiOu tetrahedra (e.g.,Velde and Couty, 1987; Dowty,1987).

Principal compression mechanisms

The largest effect of pressure is observed for the mid-frequency Raman bands at 600-700 cm '. The pressureshift for these bands is generally linear, with large posilivedv,/dP of 5-6 cm-rlGPa. This frequency region has beenassigned to modes associated with bridging O stretches,with some contributions from Si-O-Si angle bending (Fu-rukawa et al., 198 l; Matson et al., 1983; McMillan et al.,1989). Furukawa et al. (1981) studied the relationshipbetween the intertetrahedral angle and vibrational fre-quency for a mixed mode of bridging O stretching andSi-O-Si angle bending in model SirOu chain units. Theirresults showed an inverse relationship between vibration-al frequencies and Si-O-Si angles. Decreasing the Si-O-Siangle from 180" to 105'increased the calculated vibra-tional frequency from 490 to 650 cm-r. Molecular orbitalcalculations on HuSirO, (Lasaga and Gibbs, 1987, 1988)show that the curvature of the potential well rises steeplyfor an intertetrahedral angle below 140". This increase incurvature causes an increase in the associated vibrationalfrequency. Hence, the observed behavior in the glassspectra indicates that the average intertetrahedral anglesdecrease appreciably with pressure.

The inverse relationship between the intertetrahedral

(-

Fig. 5. High-pressure midinfrared absorption spectra of (a)MgSiOr, (b) CaMgSi,Ou, and (c) CaSiO, glasses up to atleast24GPa measured in the diamond-anvil cell (300 K). In each casethe standard deviation in pressure is approximately l0o/o ofthetotal average pressure. Peaks marked with asterisks are fromruby grains used for pressure calibration.


o

L

o

d

o

L

o

0)

oo

500 ?00 900 1100 1300

Wavenumber (cm-r)

500 700 900 1100 1300

Wavenumber (cm-l)

o

6

k

o

q)

q,

o

500 700 900 1100 1300

fravenumber (cm-l)

264

TABLE 1, Si and Mg ion coordination as a function of densityfor an MgSiO3 glass at 300 K from the moleculardynamics simulations*

pres-Density sure*"(g/cm") (GPa) Cation

Cation coordination number (%)i


2.54 -6.0 Si4*Mg*

2.75 0.0 Sil*Mg'.

3.00 9.0 sio*Mgz*

3.40 29.0 Si4*Mg*

3.88 60.0 Si1+Mg*

4.25 93.0 Sio*Mgz'

99.0 1.066.3 28.6

100.0 0.067.4 30.296.0 4.027.5 42.191.2 8.05.7 34.0

72.8 26.60.0 12.2

34.5 51.10 .0 1 .1

0.0 0.0 0.05.1 0.0 0.00.0 0.0 0.04.7 0.5 0.00.0 0.0 0.0

27.7 1 .9 0.00.0 0.0 0.0

44.5 14.5 1.30.6 0.0 0.0

42.8 38.1 6.914.4 0.0 0.019.3 42.9 36.6

. Kubicki and Lasaga (1991)." Densities at a given pressure are calculated from a second-order Birch-

Murnaghan equation of state for MgSiO. glass with lG : 88 GPa (lG : 4assumed) and an experimental ambient pressure density of 2.75 glcms(Yoder, 1976; J. Bass, unpublished data).

f Coordination numbers defined with maximum bond distances of 2.0and 2.8 A for Si-O and Mg-O bonds, respectively.

angles and the frequency of the symmetric Si-O-Sistretching mode in SiO, polymorphs was pointed out bySharma et al. (1981). Matson et al. (1983) observed sys-tematic shifts in frequencies of midfrequency peaks inglasses ofother compositions and connected the changesin the average intertetrahedral angle. They noted that theaverage Si-O-Si angle in crystalline lithium disilicate sheetsis 137" and for a sodium disilicate it is 149.5'. A corre-sponding increase in the frequency ofthe 500-cm-' peakalso occurs in the spectra of glasses of the same compo-sitions. In addition, Sharma et al. (1981) and Matson etal. (1983) noted that the frequency of the Si-O-Si sym-metric stretching mode in silicates with four-memberedSi rings occurs above 500 cm-', whereas in those withsix-membered rings, the frequency is 480 cm r; this isconsistent with the inverse correlation between the av-erage Si-O-Si angle and ring size.

The compression of many polymerized silicate min-erals has been found to take place mainly by the closureof Si-O-Si angles (Levien et al., 1980; Hemley, 1987 Ha-zen et al., 1989). Quantitative methods for determiningaverage Si-O-Si angles (e.9., by X-ray techniques) havenot yet been developed for glasses at high pressures.Hence, it is useful to compare vibrational spectra of theglasses with those of crystalline silicates where bond an-gles have been determined by X-ray diffraction. The stud-ies of crystalline silicates suggest that the position of themidfrequency peak is a function of mean Si-O-Si angle.The most probable explanation for the shift in the mid-frequency peaks observed in this study is that the Si-O-Si angles decrease with pressure and this increases thefrequency ofthe associated vibrational mode.

Comparison of the high-pressure vibrational spectrawith the results of recent molecular dynamics (MD) sim-ulations provides additional insight into the mechanism

of compression. MD simulations carried out with poten-

tials derived from molecular orbital theory (Lasaga andGibbs, 1987) and electron-gas calculations (Cohen andGordon, 1976) have shown that the closure of the Si-O-Si angle is linear with pressure up to a volume compres-sion (V/V) of -0.50, where Zo is the volume at 0. I MPa(Kubicki and Lasaga, 1987). MD calculations predict thatthe average Si-O-Si angle decreases from 147'lo l32 fora density increase from 2.75 to 3.40 g/cm3, which corre-sponds to V/Vo: 0.8. The reduction of the Si-O-Si angleat 30 GPa from the MD simulations is l5'or -0.5'lGPa'

According to the calculations of Furukawa et al. (1981),a reduction of the Si-O-Si angle by 15' should shift theband by 60 cm-r. This is in qualitative agreement withthe observed shift. From the experimentally determinedbulk modulus of MgSiO, glass we estimate V/Vo: g.g ut30 GPa (Table l).

Another important observation is the decrease in dif-fuse Raman scattering intensity at low frequencies ( < 500cm ') with initial increase in pressure (< l0 GPa; Fig. 3).Similar changes occur in SiO, glass, where the efect iseven more pronounced (Hemley et al., 1986). The changein the SiO, spectra is likely due to a decrease in the in-tertetrahedral angle distribution, which results in an in-crease in intermediate-range order in the glass over thispressure range (Hemley et al., 1986). Similar effects arelikely in the present metasilicate glasses. The weak low-frequency bands in the Raman spectra of CaSiO, andCaMgSirOu glasses have positive shifts with pressure.These bands have been assigned to motions of Ca in oc-tahedral coordination. Hence, the pressure shifts in thesebands indicate a mode of compression involving a de-crease in the average Ca-O bond length. No evidence wasobserved of a change in Ca cation coordination. This be-havior is consistent with the high compressibility of Ca-O bonds in crystalline materials (Hazen and Finger, 1982).We suggest that the 360-cm-' band appears in spectra ofglasses with Ca in regular octahedral coordination (Eck-ersley et al., 1988a, 1988b) and not in low-pressureMgSiO, glass spectra because the coordination state ofMg is irregular and is closer to tetrahedral than octahedralcoordination (Yin et a1., 1983). The efect ofpressure maybe to increase the regularity ofthe coordination polyhe-dra of the Mg (see below).

Finally, high-frequency bands in both the Raman andinfrared spectra decrease in relative and absolute inten-sity with pressure in each of the glasses studied. However,the Raman bands in this frequency range have apprecia-ble intensity up to 45 GPa, which indicates that SiOotetrahedra are present in these glasses at least to this pres-sure. The observed intensity loss may be related to dis-tortions in the tetrahedra (Hemley et al., 1986) and sam-ple thinning with pressure. Within the high-frequencyband, an increase in relative intensity of the 900-cm-'shoulder is observed. This implies that the population of

Q' species (one bridging O atom) associated with thisband may increase with pressure. Xue et al. (1989) andDickinson et al. (1990) present evidence that pressure

favors the disproportionation reaction 2Q' = Q' + Qo inalkali silicates. The analogous reaction, 2Q'z = Q' + Q3,may occur under pressure in the present metasilicateglasses. However, Wolf et al. (1990) do not find evidencefor this reaction occurring on compression. There mayalso be contribution to the 900-cm ' band from a smallconcentration of tsrSi (McMillan et al., 1989; Kubicki andLasaga, l99l). The evidence for a discrete coordinationchange of Si is examined in detail below.

Reversibility of spectral changes

The gross spectral changes that occurred under pressureare fully reversible upon decompression. There are, how-ever, minor differences between the ambient pressurespectra measured before and after compression. Decom-pressed sample peak positions are within + l0 cm ' (gen-erally much closer) of the original positions at 0. I MPa.For example, after pressure cycling to 32 GPa, the posi-tion of the broad midfrequency peak in the Raman spec-trum of MgSiO, glass changes from 651 to 647 cm-t, andthe bandwidth (full-width at half-maximum, FWHM) de-creases slightly from 134 to I l3 cm-r. Another change isthe increase in intensity ofthe shoulder at 885 cm I rel-ative to the 650-cm-' band in this glass following decom-pression. These small differences in peak positions andwidths suggest that there may be a small degree of den-sification (irreversible compaction) of these glasses uponroom-temperature compression on the time scales (hoursto days) and over the pressure range (30-45 GPa) oftheseexperiments. However, the spectral changes are small incomparison with the large scale changes observed in thespectra measured in situ at high pressure.

Walrafen and Krishnan (1981) studied the Ramanspectra of SiO, glass compacted at 9 GPa and 23 'C andfound minor shifts in the high-frequency bands. Theseauthors noted a 5-13-cm ' frequency increase and l8-cm-rnarrowing of the 440-cm-t band in the compactedglass. Seifert et al. (1983) measured Raman spectra ofSiO, glass annealed at 900 'C up to 2 GPa and foundchanges of l0 cm-' samples densified up to 7010. McMillanet al. (1984) observed much larger spectral shifts in SiO,glass densified at 3.95 GPa and 530 "C, as the sample wascompressed approximately 60lo above the normal density.Raman spectra obtained by McMillan et al. (1984) hadshifts of 50 and 40 cm-r in the two high-frequency peaks,and the band near 430 cm ' in the normal glass increasedto 470 cm I with an accompanying 500/o decrease in peakwidth. Hemley et al. (1986) produced a pressure-quenchedSiO, glass at ambient temperature with a Raman spec-trum markedly different from the 0. l-MPa spectra. Thestrongest band shifted from 440 to 520 cm-', and thehigh-frequency bands shifted from 800 and I 060 cm ' to815 and 1000 cm-', respectively. These shifts are of sim-ilar magnitude to those induced with high-temperatureannealing at pressure, but the maximum pressure in thelow-temperature compression was over 30 GPa com-pared with 4 GPa in the McMillan et al. (1984) study. Inview ofthe evidence that significant frequency shifts ac-

265

company densification (Seifert et al., 1983; McMillan etal., 1984), it is unlikely that a sizable increase in density(i.e., > 5olo) could be accomplished in these glasses with-out inducing structural changes observable with vibra-tional spectroscopy. We suggest that the small differencesbetween the spectra of the starting material and decom-pressed samples indicate that the densification of theseglasses is small in these room-temperature experiments(e.g., in comparison to that observed for SiOr). This needsto be examined by direct measurements of the degree ofdensification, which will likely require larger samples thanthose studied here.

Evidence for structural transformations

As discussed above, changes in the low-frequency Ra-man spectra ofthe glasses are observed at pressures above-20 GPa. This behavior was most pronounced in spectraof the magnesium silicate glass sample quenched in apiston-cylinder apparatus above 25 GPa (glass B). A rel-atively sharp peak near 400 cm ' appeared, which in-creased in relative intensity and decreased in bandwidthwith further increase in pressure. The changes were re-versible but accompanied by hysteresis. Upon decom-pression, the peak diminished in intensity but remaineddetectable to 18 GPa. The peak was observed in threeseparate experiments on this sample, including an exper-iment carried out with Ar as a pressure-transmitting me-dium. Unfortunately, this sample had been consumed be-fore this difference with subsequently prepared sampleswas detected; infrared measurements were not made onthis glass sample.

We attribute the changes in the low-frequency Ramanspectra at high pressure to a structural transition in theglass. It is unlikely that the transition is associated withcrystallization (e.g., Hemley et al., 1989a) because of thereversible changes in the spectra (i.e., disappearance ofthe peak on decompression). In addition, no other crys-tal-like peaks are observed in the high-pressure state, andthe band is broad relative to those measured for crystal-line phases (see below). We also consider the possibility

of disproportionation of MgSiOr. There is some evidencefor this reaction in MgSiO, glass samples that have beenmelted in situ at 30 GPa with a Nd-YAG laser (Kubicki

and Hemley, 1989). Raman spectra of such samples,however, are distinct from those observed here. Further,the reversibility ofthe spectra on decreasing pressure sug-gests that no disproportionation of MgSiO, (e.g', to MgOand SiOr) has taken place. Such a reaction would requirechemical diffusion, which is slow at room temperaturecompared with the time scale of these experiments'

The difference in the high-pressure spectra may be af-fected by differences in sample preparation (especially thequench pressure). There is evidence that differences inquench rates can affect glass structure in SiO, (Galeenerand Geissberger, 1983; Grimsditch, 1986) and CaMg-SirOu (Brandriss and Stebbins, 1988), and variations inthermodynamic properties have been observed in calci-um magnesium metasilicate glasses with different ther-



200 400 600 800 1000 1200

Tfavenumber (cm-l)

Fig. 6. Comparison of high-pressure Raman spectra of twomagnesium silicate glasses, enstatite, and MgSiO, perovskite.The spectra ofglass A and B were measured at 35 and 32 GPa,respectively. The spectra of orthoenstatite and MgSiOj perov-skite were measured at 30 GPa and22 GPa (Kubicki and Hem-ley, unpublished data; Hemley et al., 1989b). The frequency shiftwith pressure for the band at 430 cm ' in the perovskite phaseis 2.4 (-r 0.1) cm '/GPa (Hemley et al., 1989b), so its frequencywould increase by approximately 24 cm I from 22 to 32 GPa.The intensities are relative; the Raman scattering cross sectionsfor orthoenstatite and perovskite spectra are approximately anorder of magnitude and a factor of 2 greater than those of theglass spectra, respectively. The sloping base line in the glass Bspectrum arises from residual diamond-anvil fluorescence.

mal histories (Richet and Bottinga, 1986). Although theambient pressure spectra of the MgSiO3 samples wereindistinguishable, we suggest that small differences at 0. IMPa are amplified by the effects of pressure. The behav-ior could also arise from impurities (e.g., HrO), annealing,and storage conditions. Our Raman measurementsshowed no sign of HrO, but the presence of very smallamounts of HrO (bound as OH) in the material could beresponsible for this behavior. The kinetics of high-pres-sure transformations in glasses are strongly dependent ontrace impurities (e.g., Fratello et al., l98l). Trace amountsof impurities such as HrO are known to afect stronglythe plastic flow of silica glass (e.g., Donnadieu, 1988).

Evidence of the difference in flow properties betweenthe two types of samples was visible. The samples thatdid not develop a new p€ak appeared heterogeneous above

10 GPa; the boundaries between glass shards were visibleup to the 45 GPa. The glass quenched at pressure, how-ever, became homogenous at approximately 5 GPa. Wesuggest that this difference arises from the lower yieldstrength (and perhaps viscosity) of the latter glass sampleat high pressure. Lower yield strengths may be associatedwith a structural transition. In this regard, the behaviormay be similar to the decrease in yield strength of SiO'glass with pressure reported by Meade and Jeanloz (1988).These changes could result from the destabilization anddistortion of the SiOo tetrahedra, if not from completechanges in Si coordination (Meade and Jeanloz, 1988;Williams and Jeanloz, 1988).

These observations can be explained by a pressure-in-duced transition between two structural states of the glass.Evidence for structural (or polymorphic) transitions hasbeen reported in other amorphous materials. Mishima etal. (1985) documented the transition with first-ordercharacter between high- and low-density amorphous ice.The hysteresis observed in the glass B samples suggests atransition with first-order character. The transition mayalso be similar to the changes in structure of SiO, glassobserved in situ at high pressure beginning at 9 GPa(Hemley et al., 1986; see also Grimsditch, 1984). How-ever, the transition in SiO, appears to be associated witha continuum ofhigh-density states, and the high-pressurestate can be partially quenched or recovered at room tem-perature (irreversible compaction), in contrast to the be-havior of glass B.

Cation coordination changes

It is useful to examine the evidence for discrete pres-sure-induced changes in cation coordination in both theSi and alkaline-earth ions. Here we consider both thechanges in the high-frequency (tetrahedral Si-O stretchingvibrations) as well as the low-frequency bands. Williamsand Jeanloz (1988) measured the midinfrared spectrumof CaMgSirOu glass and observed a decrease in absor-bance at 1000-1200 cm I and a concomitant increase at600-900 cm I at pressures above 25 GPa; they attributedthese changes to a transition from tetrahedral to octahe-dral Si. Similar changes are observed in infrared spectraof the three glasses studied here. However, the intensitiesof the high-frequency Raman bands are appreciable upto 45 GPa, and any increase in absorption at 600-900cm ' is very weak. These results indicate that a significantconcentration of SiOo groups persists to these pressures,although the tetrahedra may become increasingly dis-torted with compression of the glass. A small increase inthe concentration of more highly coordinated Si couldoccur, but no peaks were observed in the infrared spectrathat may be associated with well-defined t61si.

Raman spectra measured above 20 GPa for crystallineenstatite, MgSiO, perovskite, and the two magnesium sil-icate glasses are shown in Figure 6. The glass spectra aredistinct from that of enstatite, which is metastable at thesepressures. The Raman spectrum of enstatite shifts con-tinuously with pressures up to at least 30 GP4 which

>\{J

wtro)

{J-

Enstatite

Glass B

Perovskite

Glass A


indicates that the topology of the structure persists overthis range at room temperature with no evidence for ma-j or structural transitions including amorphization (Hem-ley, 1987; Hemley et al., 1988). However, small changesin splittings among some of the bands are observed, whichare interpreted as evidence ofdistortions ofthe low-pres-sure orthoenstatite structure (Kubicki and Hemley, un-published data). The stable phase of crystalline MgSiO.above 24 GPa is orthorhombic perovskite, which has oc-tahedrally coordinated Si (e.g., Yagi et al., 1978; Fei etal., 1989). This phase boundary is close to the pressureat which the 400-cm ' band emerges in the Raman spec-tra of glass B described above. The high-pressure Ramanspectrum of MgSiO, perovskite has a relatively strongpeak in this frequency range. Although such a correlationcould suggest that the peak may be characteristic of t6rSi,

lattice dynamics calculations on MgSiO. perovskite haveshown that Raman-active vibrations in this frequency re-gion (200-400 cm') have contributions from transla-tional modes of Mg in the distorted dodecahedral site ofthe perovskite structure (Hemley et al., 1989b). Hence,the new band in the MgSiO3 glass sample may be asso-ciated with Mg ion motions in a high-coordination en-vrronment.

Molecular dynamics simulations based on ionic pair-wise interatomic potentials were performed to explorefurther possible structural changes in MgSiO. glass underpressure. The techniques and details concerning the po-tentials used are described elsewhere (Kubicki and Lasa-ga, l99l). The Si and Mg coordinations as a function ofdensity predicted from the MD calculations are listed inTable l. The MD simulations indicate that the coordi-nation of Mg changes from predominantly tetrahedral atambient pressure to octahedral at 30 GPa, which corre-sponds to a volume compression of approximately 20o/o.The presence of tatMg at low pressures is supported bythe X-ray diffraction study of Yin et al. (1983) and X-rayemission spectra of Hanada et al. (1988). These resultsindicate that the peak may signal the onset of a lalMg toturMg (or higher) coordination change. If so, the peak maybe related to the low-frequency peaks in the CaSiO, andCaMgSirOu glass spectra due to Ca in octahedral coordi-nation. The proposed increase in Mg coordination withpressure could be coupled to an increase in average Sicoordination. The I4rSi to t6lSi transition may not be ap-parent in these spectra because the characterislic vibra-tional modes are weak (e.g., Hemley et al., 1989b), orbecause of the low concentration of the highly coordi-nated species, or both. Again, we emphasize that whethersuch a structural transition occurs at room temperaturemay be dependent on trace impurities in the glass andthe sample history. It is possible that structural transi-tions in the amorphous material may be kinetically hin-dered in the compressed glass at room temperature butmay become facile at higher temperatures (e.g., above theglass transition), perhaps at lower pressures.

In general, the high-pressure behavior of these glassesat room temperature is thus consistent with a wide range

of stability of tetrahedrally coordinated Si. However, itis possible to admit a small concentration of tulSi, perhapsalong a continuum of coordination states (e.g., Stolperand Ahrens. 1987). Such a coordination continuum couldimply the existence of I5lSi as an intermediate species. TheRaman and infrared signatures for this species are not yetknown, but McMillan et al. (1989) have suggested thatRaman intensity at approximately 900 cm-r in spectra ofquenched, high-pressure melts may be due to fivefold co-ordinated Si. This argument is consistent with the in-crease in Raman intensity with pressure at 900 cm ' inour spectra. Molecular orbital calculations on the

tS(OH)sl I molecule predict 1[31 tst$i is a stable specieswith Raman-active Si-OH stretches at768,825, and 951cm ' (Kubicki, l99l). Evidence for the formation of t'tSi

in glasses quenched at high pressure has been obtainedfrom recent NMR experiments by Stebbins and McMil-lan (1989). The presence of t5tSi in compressed MgSiO.glass at 300 K is also suggested by the MD simulations(Table l; see also Kubicki and Lasaga, l99l). MD sim-ulations of the high-temperature melts show that a sig-nificant population of I6rSi does not form in the meltsuntil very high compressions of V/Vo < 0.6 are reached(Matsui and Kawamura, 1980; Matsui et al., l98l; Ku-bicki and Lasaga. l99l).

CoNcr,usroNs

High-pressure Raman and infrared measurementsdemonstrate that the structures of metasilicate glasses arealtered dramatically at pressures above l0 GPa, as ob-served previously for SiO, glass. By far, the predominantchanges are reversible; hence, it is necessary to study theseglasses in situ in order to obtain information on high-pressure structures and properties. Structural studies ofglasses quenched from these pressures give at best an in-complete picture of the effect of compression on the short-and intermediate-range structure of these materials.Comparisons with previous measurements and existingtheoretical studies are used to show that the principalcompression mechanism at room temperature is the clo-sure of the Si-O-Si angle, at least up to 30 GPa. The lossof diffuse Raman scattering intensity at low frequenciesalso suggests there is an increase in intermediate rangeorder in these glasses under pressure, as in SiOr glass.Increases in cation coordination are inferred from changesin the low-frequency Raman spectra at higher pressures(above -20 GPa). However, the glasses retain spectralfeatures diagnostic of t4lsi up to 45 GPa, although thetetrahedra may be considerably distorted at high com-pressions. We also present evidence that sample prepa-ration, history, and experimental conditions can have asignificant effect on the structural state of the glass spectraunder pressure. It remains to be seen whether the struc-tural changes inferred from the present room temperaturespectra occur at comparable pressures at higher temper-atures (e.g., in high-density melts). These changes mayoccur at lower pressures as a result offaster kinetics as-sociated with higher O ditrusivity, which may also be


enhanced by the compression. Indeed, the variability inhigh-pressure behavior of glasses with nominally similarcompositions may be associated with differences in ki-netics for alternative compression mechanisms (i.e., cat-ion coordination) even at room temperature. Direct mea-surements at high pressure and temperature are requiredto examine these questions.

AcKNowLEDGMENTS

We thank H.K. Mao, T.C. Hoering, and L.W. Finger for valuable as-sistance, and F.R. Boyd, H.S Yoder, and J. Bass for supplying glass sam-ples and for communicating results prior to publication. We also thankC. Meade, R.E. Cohen, A.C. Lasaga, and B.O. Mysen for useful discus-sions, and P. McMillan, L.E. McNeil, C.T. Prewitr, D.E. Townsend, andL. Stixrude for helpful comments on the manuscript. This work was sup-ported by the National Science Foundation (grant nos. EAR-8418413,EAR-8708192, and EAR-8920239.

RrrnnrNcns crrEDBrandriss, M.E., and Stebbins, J.F. (1988) Effects of temperature on the

structures of silicate liquids: NMR results. Geochimica et Cosmochimi-ca Acta. 52.2659-2669.

Brawer, S.A., and White, W.B. (1977) Raman spectroscopic investigationofthe structure ofsilicate glasses. II. The soda-atkaline earth-aluminatemary and quaternary glasses. Joumal of Non-Crystalline Solids, 23,26r-278.

Cohen, A.J., and Gordon, R.G. (1976) Modified electron-gas srudy of thestability, elastic properties, and high-pressure behavior of MgO andCaO crystals. Physical Review B, 14,4593-4605.

Cusack, N.E. (1987) The physics of structurally disordered matter, 4O2 p.Adam Hilger, Philadelphia.

Dickinson, J.E., Scarfe, C.M., and McMillan, P.F. (1990) Physical prop-erties and structure of KrSinO" melt quenched from pressures up to 2.4GPa. Journal of Geophysical Research, 95, I 5675- I 568 1.

Donnadieu, P. (l 988) Influence ofimpurities on plastic flow ofsilica glass.Joumal of Non-Crystalline Solids, 99, 113-l17.

Dowty, E. (1987) Vibrational inreractions oftetrahedra in silicate glassesand crystals: I. Calculations on ideal silicate-aluminate-germanatestructural units. Physics and Chemistry of Minerals, 14, 80-93.

Eckersley, M.C., Gaskell, P.H., Barnes, A.C., and Chieux, P. (1988a)Structural ordering in a calcium silicate glass. Nature, 335, 525-527 .

-(1988b) The environment of Ca ions in silicate glasses. Journal ofNon-Crystalline Solids, 106, 132-136.

Etchepare, J. (1972) Study by Raman spectroscopy ofcrystalline and glassydiopside. In R.W. perglas and B. Ellis, Eds., Amorphous materials,papers presented to the International conference on the Physics ofNon-crystalline Solids, p. 337-346. Wiley-Interscience, New York.

Fei, Y., Saxena, S.K., and Navrotsky, A. (l 989) Internally consistent ther-modynamic data and equilibrium phase relations for compounds in thesystem MgO-SiO, at high pressure and high temperature. Journal ofGeophysical Research, 9 5, 6915-6928.

Fratello, V.J., Hays, J.F., and Turnbull, D. (1981) Dependence of growthrate ofquartz in fused silica on pressure and impurity content. JournalofApplied Physics, 5 l, 47 18-4728.

Furukawa, T., Fox, K.E., and White, W.B. (1981) Raman spectroscopicinvestigation of the structure of silicate glasses. III. Raman intensitiesand structural units in sodium silicate glasses. Journal of ChemicalPhysics, 75, 3226-3217.

Galeener, F.L., and Geissberger, A.E. (1983) Yibrational dynamics in rosi-

substituted vitreous SiO,. Physical Review B, 27, 6199-6204.Grimsditch, M. (1984) Polyrnorphism in amorphous SiOr. Physical Re-

view Irtters. 52, 2379-2381.-(1986) Annealing and relaxation in the high-pressure phase of

amorphous SiOr. Physical Review B, 34,4372-4373.Hanada, T., Soga, N., and Tachibana, T. (1988) Coordination state of

magnesium ions in rf sputtered amorphous fikns in the system MgO-SiO,. Joumal of Non-Crystalline Solids, 1O5,39-44.

Hazen, R.M., and Finger, L.W. (1982) Comparative crystal chemistry,231 p. Wiley-Interscience, New York.

Hazen, R.M., Finger, L.W., Hemley, R.J., and Mao, H.K. (1989) Hieh-pressure crystal chernistry and amorphization of a-quartz. Solid StateCommunications. 72. 507-51 I .

Heinz, D.L., and Jeanloz, R. (1987) Measurement of the melting curveof Mgo,Feo ,SiO, at lower mantle conditions and its geophysical impli-cations. Journal of Geophysical Research, 92, | 1437 -l 1444.

Hemley, R.J. (1987) Pressure dependence ofRaman spectra ofSiO, poly-morphs: c-quartz, coesite, and stishovite. In M.H. Manghnani and Y.Syono, Eds., High-pressure research in mineral physics, p.347-359.Terra Scientific, Tokyo, and American Geophysical Union, Washing-ton, DC.

Hemley, R.J., and Kubicki, J.D. (1991) Deep mantle melting. Nature,f49,283-284.

Hemley, R.J., Mao, H.K., Bell, P.M., and Mysen, B.O. (1986) Ramanspectroscopy of SiO, glass at high pressure. Physical Review l€tters,57,747-7 50.

Hemley, R.J., Bell, P.M., and Mao, H.K. (1987) Laser techniques in high-pressure geophysics. Science, 237, 605-612.

Hemley, R.J., Jephcoat, A.P., Mao, H.K., Ming, L.C., and Manghnani,M.H. (1988) Pressure-induced amorphization of crystalline silica. Na-fire,334, 52-54.

Hemley, R.J., Chen, L.C., and Mao, H.K. (1989a) New transformationsbetween crystalline and amorphous ice. Nature, 388, 638-640.

Hemley, R.J., Cohen, R.E., Yeganeh-Haeri, A., Mao, H.K., Weidner, D.J.,and Ito, E. (1989b) Raman spectroscopy and lattice dynamics of MgSiOr-perovskite at high pressure. In A. Navrotsky and D.J. Weidner, Eds.,Perovskite: A structure of great interest to geophysics and materialsscience, p. 35-44. American Geophysical Union, Washington, DC.

Hibben, J.H. (1935) Raman spectra in inorganic chemistry, ChemicalReviews. 13.345-4'18.

Hofmeister, A.M., Xu, J., Mao, H.K., Bell, P.M., and Hoering, T.C. (1989)Thermodynamics of Fe-Mg olivines at mantle pressures: Mid- and far-infrared spectroscopy at high pressue. American Mineralogist, 74,281-306.

Jayaraman, A. (1983) Diamond-anvil cell and high-pressure physical in-vestigations. Reviews ofModem Physics, 55, 65-108.

Knittle, E., and Jeanloz, R. (1989) Melting curve of (Mg,Fe)SiO, perov-skite to 96 GPa: Evidence for a structural transition in lower mantlemelts. Geophysical Research ktters, 16, 421-424.

Kubicki, J.D. (1991) Molecular orbital calculations of transitions betweencoordination states of Si(OH)" molecules. Eos, 72, 144.

Kubicki, J.D., and Hemley, R.J. (1988) In situ high-pressure Raman spec-tra of silicate glasses. Annual Report of Director of the GeophysicalLaboratory, I 987-1988, 53-56.

- ( I 989) Spectroscopic evidence for a new high-pressure magnesiumsilicate phase. Annual Report of Director of the Geophysical Labora-tory, 1988-1989, 9 l -94.

Kubicki, J.D., and Lasaga, A.C. (1987) Molecular dlnamics simulationsof MgrSiOo and MgSiO, melts: Structural and diffrsivity changes withpressure. Eos, 68, 436.

-(1991) Molecular dynamics simulation of pressure and tempera-ture effects on MgSiO, and Mg,SiOo melts and glasses. Physics andChemistry of Minerals, 17, 661-673.

Iasaga, A.C., and Gibbs, G.V. (1987) Applications of quantum mechan-ical potential surfaces to rnineral physics calculations. Physics andChemistry of Minerals, 14, lO7-117.

- (l 988) Quantum mechanical potential surfaces and calculations onminerals and molecular clusters. Physics and Chemistry of Minerals,16,29-41.

Levien, L., Prewitt, C.T., and Weidner, D.J. (1980) Structure and elasticproperties of quartz at pressure. American Mineralogist, 65, 920-930.

Liu, L.-G. (1975) Post-oxide phases offorsterite and enstatite. Geophys-ical Research Irtters. 2. 417-419.

Liu, L.-G., and Ringwood, A.E. (1975) Synthesis of a p€rovskite-typepolymorph ofCaSiO,. Earth and Planetary Science lrtters, 28,209-2t t .

Mao, H.K. (1989) Static compression of simple molecular systems in themegabarrange, In A. Polian, P. Loubeyre, and N. Boccara, Eds., Simple


molecular systems at very high densities, p.221-236. Plenum Press,New York.

Mao, H.K., Xu, J., and Bell, P.M (1986) Calibration of the ruby pressuregauge to 800 kbar under quasi-hydrostatic conditions. Journal ofGeo-physical Research, 91, 467 3-467 6.

Mao, H.K., Chen, L.C., Hemley, R.J., Jephcoat, A.P., Wu, Y., and Bas-sett, W.A. (1989) Stability and equation of state of CaSiOr-perovskiteto I 34 GPa. Joumal of Geophysical Research, 94, 17889-17E94.

Matson, D.W., Sharma, S.K., and Philpotts, J.A. (1983) The structure ofhigh-silica alkali-silicate glasses. A Raman spectroscopic investigation.Journal of Non-Crystalline Solids, 58, 323-352.

Matsui, Y., and Kawamura, K. (1980) Instantaneous structure of an MgSiO.melt simulated by molecular dynamics. Nature, 285,64E-649.

Matsui, Y., Kawamura, K., and Syono, Y. (1981) Molecular dynamicscalculations applied to silicate systems: Molten and vitreous MgSiO,and MgrSiOo. In S. Akimoto and M.H. Manghnani, Eds., High pressureresearch in geophysics, advances in earth and planetary science, vol.12, p. 511-524. Reidel, Boston.

McMillan, P. (1984a) A Raman spectroscopic study of glasses in the sys-tem CaO-MgO-SiOr. American Mineralogist, 69, 645-659.

-(1984b) Structural studies of silicate glasses and melts-applica-tions and limitations of Raman spectroscopy. American Mineralogist,69,622-644.

- (l 989) Raman spectroscopy in rnineralogy and geochemistry. An-nual Reviews in Earth and Planetary Sciences, 17, 255-283.

McMillan, P., Piriou, 8., and Couty, R. (1984) A Raman study of pres-sure-densified vitreous silica. Joumal of Chemical Physics, 81, 4234-4236

McMillan, P., Xue, X., Kanzaki, M., and Stebbins, J.F. (1989) Structuralchanges in quenched alkali silicate liquids at high pressure: Ramanspectroscopy. Eos. 70, 1374.

Meade, C., and Jeanloz, R. (1988) Effect ofa coordination change on thestrength of amorphous SiOr. Science, 24 l, 107 2-lO7 3.

Mishima, O., Calvert, L.D., and Whalley, E. (1985) An apparenfly first-order transition bgtween two amorphous phases ofice induced by pres-sure. Nature. 314. 76-78.

Moenke, H.H.W. (1974) Silica, the three-dimensional silicates, borosili-cates, and b€ryllium silicales. In V.C. Farmer, Ed., The infrared spectraof minerals, monograph no. 4, p. 383-423. Mineralogical Society ofhndon, London.

Mysen, 8.O., Virgo, D., and Scarfe, C.M. (1980) Relations between theanionic structure and viscosity of silicate melts-a Raman spectro-scopic study. American Mineralogist, 65, 690-710.

Mysen, B.O., Virgo, D., and Seifert, F.A. (1982) The structure of silicatemelts: Implications for chemical and physical properties of naturalmagma. Reviews ofGeophysics and Space Physics, 20,353-383.

Nayrotsky, A., Geisinger, K.L., McMillan, P., and Gibbs, G.Y. (1985)The tetrahedral framework in glasses and melts-Inferences from mo-lecular orbital calculations and implications for structure, thermody-namics, and physical properties. Physics and Chemistry of Minerals,1t,284-298.

Richet, P., and Bottinga, Y. (1986) Thermochemical properties ofsilicateglasses and liquids: A review. Reviews ofGeophysics, 24, l-25.

Rigden, S.M., Ahrens, T.J., and Stolper, E.M. (1984) Densities of liquidsilicates at high pressures. Science, 226, lO7 l-1074.

-(1988) Shock cornpression of molten silicate: Results for a modelbasaltic composition. Journal of Geophysical Research, 93, 367-3E2.

Seifert, F.A., Mysen, B.O., and Vireo, D. (1983) Raman study ofdensifiedvitreous silica. Physics and Chemistry ofGlasses, 24, l4l-145.

Sharma, S.K., Mammone, J.F., and Nicol, M.F. (1981) Raman investi-gations ofring configurations in vitreous silica. Nature, 292, 140-141.

Stebbins, J.F., and McMillan, P.F. (1989) Five- and six-coordinated Si inK.SioO, glass quenched from 1.9 GPa and 1200 qC. American Miner-alogist, 74, 965-968.

Stolper, E M., and Ahrens, T.J. (1987) On the nature ofpressure-inducedcoordination changes in silicate melts and glasses. Geophysical Re-search lrtters. 14. 123l-1233.

Sweet, J.R., and White, W.B. (1969) Study of sodium silicate glasses andliquids by reflectance spectroscopy. Physics and Chemistry ofclasses,r0,246-251

Velde, 8., and Couty, R. (1987) High-pressure infrared spectra of somesilicate glasses. Chemical Geology, 62, 35-41.

Waff, H.S. (1975) Pressure-induced coordination changes in magmaticliquids. Geophysical Research Irtters, 2, I 93- I 96.

Walrafen, G.E., and Krishnan, P.N. (1981) Raman spectrum of pressure

compacted fused silica. Journal ofChemical Physics, 74,5328-5330.Waseda, Y., and Toguri, J.M. (1977) The structure of molten binary sil-

icate systems CaO-SiO, and MgO-SiOr. Metallurgical Transactions, 88,563-568.

Williams, Q., and Jeanloz, R. (1988) Spectroscopic evidence for pressure-

induced coordination changes in silicate glasses and melts. Science,239,902-905.

Woll G.H., Durben, D.J., and McMillan, P.F. (1990) High-pressure Ra-man spectroscopic study of sodium tetrasilicate (NarSi.OJ glass. Jour-nal of Chemical Physics, 93, 2280-2288.

Wong, J., and Angell, C.A. (1976) Glass structure by spectroscopy, E64p. Marcel Dekker, Inc., New York.

Xue, X., Stebbins, J.F., Kanzaki, M., and Tronnes, R.G. (19E9) Siliconcoordination and speciation changes in a silicate liquid at high pres-

sures. Science. 245. 962-964.Yagi, T., Mao, H.K., and Bell, P.M. (1978) Structure and crystal chem-

istry of perovskite-type MgSiOr. Physics and Chemistry of Minerals,3,97-rr0.

Yin, C.D., Okuno, M., Morikawa, H., and Marumo, F. (1983) Structureanalysis of MgSiOr glass. Journal of Non-Crystalline Solids, 55, 131-t 4 l .

Yoder, H.S. (1976) Generation of basaltic magma, 265 p. National Acad-emy of Sciences, Washington, DC.

M.lNuscrusr REcEIVED Seprplranen 17, 1990MeNuscnrp'r AccEprED NoveMsen 5, l99l

Documents

high pressures is essentiar ror undersranu,.', ''jffi i#ti ... · persed ruby grains were added for pressure calibration. The infrared absorption spectra were measured over the total