Solubility of carbon dioxide in albitic melt

Document Sample
Solubility of carbon dioxide in albitic melt Powered By Docstoc
					                                  American Mineralogist, Volume 72, pages 107 1-1085, 1987

                               Solubility of carbon dioxide in albitic melt
              Eowl.nn Slor.pnn" Grn-lr,o FINE.TTrrouls                 JoHusoN,2 Slrr,v NBwvrall
                                                                                             California 91125, U.S.A.
                                                  California Institute of Technology,Pasadena,
     Division of Geological and Planetary Sciences,

                Infrared spectroscopyhas been used to measurethe concentrations of molecular CO,
             and carbonate in albitic (NaAlSirOr) elasses   quenchedfrom melts equilibrated with CO,
             vapor at high pressures  (15-30 kbar) and temperatures  (145{u--1625 At constanttem-
             perature, the concentrations of carbonate and molecular CO, as well as the CO] /CO,
             ratio increasewith increasingpressureunder vapor-saturatedconditions. The secondde-
             rivative of these speciesconcentrations with respect to pressureunder vapor-saturated
             conditions at constant temperature is positive over the range of conditions studied. At
             constant pressureunder vapor-saturated conditions, the solubility of molecular CO, de-
             creases  with increasingtemperature,but the concentration ofcarbonate increases.  The net
             efect is that total CO, solubility is nearly independent of temperature. According to our
             results, CO, solubility and speciation changegradually over the range of conditions that
             we have studied and do not indicate major or abrupt structural changesin albitic melts
             in this P-Z range.
                Our results can be described thermodynamically in terms of two reactions. The first,
             COr(vapor) : COr,molecular(melt),describesthe heterogeneous     equilibrium betweenmelt
             and vapor. The second,CO,,molecular(melt) + O'?-(melt): COI-(melt), models the ho-
             mogeneousequilibrium betweenmelt species.       Volume and enthalpy changes thesetwo
             reactionshave been constrainedby our solubility and speciationdata. We emphasizethat
             the solubilities of volatile components that dissolve in melts in several different forms
             must be treated by such coupled heterogeneous       and homogeneousequilibria and that
             spectroscopic   methods provide direct insights into them.

                    INrnooucrroN                             in a melt in equilibrium with CO, vapor) is the sum of
   The solubility of CO, in both natural and synthetic       the concentrationsof the various forms of dissolved COr.
silicate melts has been studied extensivelyin recentyears.   Full understandingof the effectsof pressure,  temperature,
There have been several motivations for these studies.       and silicate composition on the solubility of CO, in melts
First, solubility measurementsgive upper limits on the       must therefore take account both of the heterogeneous
amounts of CO, that can dissolve in magmas at various        equilibrium between melt and vapor and of the homo-
                                                             geneousequilibria between various melt species.Mea-
temperaturesand pressures.    Such data are essentialifwe
are ever to understand the degassing   behaviors of mag-     surementsof the concentrationsof the different forms of
mas as they ascend(e.g.,Moore, 1979; Speraand Berg-          dissolved CO, in vapor-saturatedmelts thus have the po-
man, 1980; Mathez, 1984). Moreover, information on           tential to provide greater insights into the thermody-
the pressureand temperature dependenceof CO, solu-           namics of solubility and into the microscopic mecha-
bility and spectroscopicstudies of quenched COr-satu-        nisms that contribute to observedvariations in solubility
rated melts may provide insights into the mechanismsof       than do sirnple measurements bulk dissolved CO, con-
CO, dissolution(e.9.,Mysen, 1976;Rai et al., 1983)and        tents that give no direct information on speciation.
into the structures and properties of silicate melts (e.g.,     In this paper, we report new measurementsof the sol-
Mysen, 1976, 1977).                                          ubility of CO, in albitic melt at pressuresof 15-30 kbar
   It is known that CO, dissolves in silicate melts in at    and temperaturesof 1450-1625'C. Bulk CO, concentra-
least two microscopic forms: molecules of CO, and car-       tions were determined by summing the concentrationsof
bonate ion complexes(Mysen, 1976; Fine and Stolper,          dissolved carbonateions and moleculesof CO, in glasses
                                                             quenched from melts equilibrated with COr-rich vapor
 1985,1986).CO, solubility (i.e.,the concentration CO,
                                                             at high temperaturesand pressures.The concentrations
  t Presentaddress:                                          of the individual C-bearing speciesin the glasseswere
                    Research and Development Division, Corn-
ing GlassWorks, Corning,New York 14831,U.S.A.
                                                             determined by infrared spectroscopy, which the inten-
  2Present address: Department of Geology and Geophysics, sities ofabsorptions due to the presenceofthese species
University of California, Berkeley,California 94720, U.S.A.  were measured.The albitic composition was chosenbe-
0003-{04xl87/ | | t2-107ts02.00                            r07l
t0'72                              STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT

         TneLe1. Average of five microprobeanalyses                                                 wovelenglh(/lm)
                 of the albiticglass used in this study

                               Analyzed                                              T J_ 3 6
                                                                                     scoled 5opm thickness
              Naro               10.38            11.82
              Alr03              20 43            19.44                    o
              sio,               69.66            68.74
                Total           100.47           100.00                   a
           Nofe. Microprobe conditions were acclerating volt-             o    v.w

         age, 15 kV; sample current, 5 nA on brass; beam size,
         40-50 pm. Also given is an analysis of stoichiometric

cause it is a traditional and simple starting point for mod-
eling real magmatic systems, because previous measure-                     r.
                                                                        Fig. Spectrum                      (rJ-36)
                                                                                      #,-"i:ili,J'fr,l.rru,,    scared
ments of solubility and speciation are available in the
                                                                      to a thickness of 50 pm. The six absorption bands discussedin
literature for comparison (Brey, 1976; Mysen, 1976;My-                the text are indicated by arrows. The lower spectrumis of a COr-
sen et al., 1976; Mysen and Virgo, 1980), and because we              freealbite glass(ABC- l2: 1400'C, 20 kbar, 0.19 wto/o
                                                                                                                          HrO) scaled
have previously calibrated the infrared technique for                 to the same thickness.
quantitative determination       of the concentrations of
C-bearing species in albitic glass (Fine and Stolper, I 985).         es were prepared for us by A. L. Boettcher of the University of
                                                                      California. These samples were prepared in a l-in. (2.54-cm)
               ExpBnrnrnNTAL         TECHNTeuEs                       piston-cylinder apparatususing dried crystalline albite plus sil-
                                                                      ver oxalate as starting materials.
   The synthesis of COr-bearing glass samples is discussedin             A few albitic andjadeitic glasses were synthesized pressures
detail elsewhere (Fine and Stopler,1985, 1986),but is reiterated      of 100 to 1000 bars at l2OO'C in an internally heated Ar pres-
briefly here. Starting material of approximately albite (NaAl-        sure vesselin J. R. Holloway's laboratory at the Arizona State
SirOr) composition was synthesized grinding Johnson-Matth-
                                       by                             University. Thesesampleswere uniformly chargedwith bubbles
ey SpecpureNarCOr, AlrOr, and SiO, in an agatemortar for 6            up to 20 pm in size.
h, followed by melting at 1580 "C for 12 h in air at I atm. This         The quenched glasseswere sectioned with a diamond saw,
decarbonated   glasswas ground under ethanol for 6 h and dried        ground into plates30-250 pm in thickness,and polished on both
at 850 "C in air for 2 d to remove adsorbed water and hydro-          sidesin a slurry of AlrO, and HrO. Sample thicknesswas mea-
carbon residue. An electron-microprobe analysis of the decar-         suredwith a digital dial indicator. The polished glassplateswere
bonatedglassis given in Table 1 Although microprobe analyses          placed over metal apertures50-1000 pm in diameter. The sam-
of soda-rich glasses  are difficult becauseof Na mobility under       ples were then examined microscopically to determine whether
the electron beam, the glass appears to be deficient in NarO          bubbles were present in the region exposedby the aperture. In
relative to stoichiometric albite, probably owing to Na volatili-     most cases, spectrawere obtained on bubble-freeor nearly bub-
zation during decarbonation and drying. The compostion that           ble-free regions. In some cases,however, bubble-rich regions
we studied is consequentlyslightly peraluminous (3.40lo     norma-    were purposelyexamined so that we could examine the spectro-
tive corundum) and enriched in silica relative to stoichiometric      scopic signatureof CO. contained in bubbles.
albite.                                                                 Transmissioninfrared spectrawere obtained on the regionsof
   Powdered silicate starting material and silver oxalate (AgrCrOo)   the samplesexposedby the metal aperture using both a Perkin-
were weighedinto Pt capsules     that were then sealedby arc-weld-    Elmer 180 infrared spectrophotometer and a Nicolet Instru-
ing. Amounts of each were chosen using published solubility           ments 60SX Fourier transform infrared spectrometer (r'rrn).
data to produce vapor-saturatedmelts with a minimum of excess         These machines yielded quantitatively similar results, but the
vapor. These Pt capsuleswere in turn loaded into larger Pt cap-       rrn is quicker and has the advantageofsmaller potential beam
sules containingsintered   hematiteand run in a 0.5-in. (1.27-cm)     sizes(lessthan tens of micrometers).Infrared spectraon the rrrn
piston-cylinder apparatususing an NaCl and Pyrex pressureme-          were obtained using a HgCdTe, detector,KBr beamsplitter,glo-
dium at a variety ofpressures (15-30 kbar) and temperatures           bar source, a mirror velocity of 1.57 cmls, and lO24 to 8192
(1450-1625 "C) using the proceduresdescribedin Fine and Stol-         scans.
per (1985).
   During run conditions, the silver oxalate presumably disso-
                                                                      Ixrnrnno       spEcTRoscopy     oF CO,-BEARTNG cLAssES
ciatesto Ag metal and COr. Upon quenching,glasses        containing
disseminatedAg and dissolvedCO. are formed. No quenchcrys-            Band assignments
talline carbonates were observed.The glasses often yellow or
                                                are                      Typical spectra of a COr-bearing and a COr-free albitic
bluish-orange,probably owing to finely disseminatedAg (Sew-
                                                                      glass are shown in Figure l. Six absorption bands are
ard, 1980).All ofthe glasses    contain bubbles from 5 to 100 pm
                                                                      present in the COr-bearing glass that are not observed in
in diameter that are assumedto contain gaseous liquid COr;
the bubbles tend to occur in streaks and clusters distributed         the COr-free glass.
throughout the sample. We have interpreted the presenceof these          2352 cm r. This sharp, intense band is due to the z,
bubbles as an indication of vapor-saturated conditions during         antisymmetric stretching mode of '2CO, molecules dis-
the experiments.                                                      solved in the glass (Fine and Stolper, 1985). The band is
   A few vapor-saturatedand undersaturatedCOr-bearingglass-           slightly asymmetrical; it can be approximated as the sum
                             STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT                                  1073

of two nearly Gaussianbands, one at about 2350 cm-' to the stretching of OH goups and water moleculesdis-
(FWHM : full width at half maximum : 19                 and solved in the glass(Stolper, 1982a).It appearsto be dif-
a secondband at about 2370 cm ' (FWHM : "--'1 cm-r) ficult to completely exclude water from glassessynthe-
with about 150/o the intensity of the first band. The sized at elevated pressuresand temperatures in solid
position and shape of the band in bubble-free glasses media piston-cylinder apparatuses (Fine and Stolper,
changelittle betweenroom temperatureand liquid-nitro-        1985). This is particularly true in CO,-bearing experi-
gen temperature. This contrastswith the spectraof bub- ments, which systematicallyhave higher water contents
ble-rich samples,which at room temperaturehave an ab- than glasses         synthesizedwithout COr.
sorption maximum at about 2358-2366 cm        ', sometimes      f600-f700 cm t and 1375 cm'. These bands are at-
with a shouldet at 2340-2350 cm-', and often showing a tributed to CO, dissolved in the form of distorted CO3-
conspicuousdip in the baselineof the spectrum centered ionic complexes         (Fine and Stolper, 1985).The shapes   of
at about 2310 cm-'. At liquid-nitrogentemperature,       the thesebands changelittle betweenroom and liquid nitro-
band at -2360 cm ' is split into three sharp peaks at gentemperatures. band between1600and 1700cm-'
2337,2352, and 2366 cm-' plus a sharp peak at 2281 is clearly a composite; we have found that it can be ap-
cm-r, and the anomalous baseline in this region of the proximated as a sum of two nearly Gaussiancomponents,
spectrumis no longer present.We presumethat this spec- one at 1670-1680 cm ' (FWHM : 70-80 cm-') and
                                                                                                                '). The
trum is characteristicof crystalline CO, within the bub- anotherat 1600-1610cm-' (FWHM : 70-80 cm
bles at low temperatures.                                    relative intensities of thesetwo componentsare variable,
    2287 cm'. This weak band on the sloping tail of the but are usually within about 200/o each other for the
band at 2352 cm' is attributed to the z, antisymmetric albitic glassesreported on in this paper. The same two
stretch of dissolved moleculesof "CO, (Fine and Stolper, components are present in CO'-bearing jadeitic glasses,
 1985).No attempthasbeenmade to model its shape,         but but the componentat 1600-1610cm 'is typically about
its FWHM is approximately 20 cm-i and changeslittle 50-800/o           larger than the one at 1670-1680cm '. This is
between room temperature and liquid-nitrogen temper- apparent in the spectra presented in Fine and Stolper
ature, though the background is flatter under these con- (1985). In both albitic and jadeitic glasses, band at
ditions because   ofa narrowingofthe band at 2352 cm'         1375 cm ' can be approximatedby a single Gaussian
in this vicinity.                                            component(FWHM : -'70 cm'). Following our earlier
    For bubble-rich samplesat room temperature,this band interpretation (Fine and Stolper, 1985, and references
is usually a poorly defined shoulder on the '2CO2   band at therein), thesebands, including the component at 1670-
 2352 cm', and it is often difficult to seebecause the 1680 cm-', are interpreted as splittings of the z, stretching
 distortion ofthe baselineofvery bubble-richglasses      de- vibrations of distorted sodium carbonate ionic com-
 scribed above. It is sometimesresolvableinto two      weak plexes.We do not have specificassignmentsor interpre-
bandsin bubble-richsamples:       one at 2310 cmrand a tations of the observed spectral components in terms of
 secondat 2275 cm I. Thesetwo bandsmay be the '3CO, the local structural details ofthe dissolvedcarbonateion-
 equivalents the '2CO,bands observedat -2360 cm-' ic complexes responsible for them, but similar spectral
 ard -2345 cm-t in bubble-rich samples.In bubble-rich features observed in nitrate glasseshave been analyzed
 samples at liquid-nitrogen temperatures, however, the in somedetail (Furukawaet al., 1978).
 broad bands in this region are replacedby a distinct band      At elevatedwater contents,we would expectto observe
 at 2281 cm '. In samplescontaining significant CO, both an absorptionat 1630 cm-' due to molecular HrO. We
 in bubbles and in the enclosingglass,the spectrum again have not observed such a component in the spectra of
 shows a broad shoulder in this region at room tempera- the glasses       included in this study, nor have we observed
 ture, which resolvesinto bands at about 2290 and 2284 a correlation betweentotal water content and the ratio of
 cm I at liquid-nitrogen temperature. The band at higher the integrated intensities of the 1600-1700-cm-' and
 wave numbers is assigned dissolvedmoleculesof '3CO, 1375-cm ' bands that we would expect to accompany
 and the one at lower wave numbers to frozen '3CO, in such a component. On the basis of the total water con-
 the bubbles.                                                 tents ofour COr-saturatedglasses, observedrelation-
    3710 cm-1. This weak band is due to the      presenceof ship betweentotal water and molecular HrO contents of
 moleculesof CO,, as indicated by its strength in bubble- albitic glasses,and the molar absorptivity of the 1630
 rich samplessynthesizedat low pressuresand the excel- cm I band in albitic glasses A. Silver and E. Stolper,
 lent (r : 0.98) correlationof its intensity with the '3CO, in prep.),the intensity of the 1630-cm molecularHrO
 band at 2287 cm-t in bubble-free    glasses.We tentatively band    would be expectedin all casesto be less than l0o/o
 assignthis band to a combination of the v, ar;.d modes of the intensity of the carbonatebandsat 1600-1700cm-'.
 of r2CO,molecules(Nakamoto, 1978; Fine and Stolper,
  1985).In bubble-free  specimens, band shapechanges Band intensities
 very little betweenroom temperatureand liquid-nitrogen          Quantitative measurementsof band intensities were
  temperature; in bubble-rich specimens,the band sharp- made of most of the absorption bands on eachof the glass
  ens considerablyat liquid-nitrogen temperature.             samples.These measurementswere made on each spec-
    3550 cm-l. This broad. asymmetric band is attributed trum after the spectrum of a COr-free albite glass,scaled
t0'14                                STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT

TreLe 2. Molar absorptivities and integrated molar absorptivi-              aluminosilicate bands in the mid-infrared. If the samples
         ties for relevant infrared absorption bands in albitic             are thicker than 200 pm or so, the backgroundcan be so
                                                                            intensethat the carbonateabsorptions are offscale. Thin
   Band                                   ee'                               samples(<100 pm) can also presentproblems because
   (cm')           Species           (           (,)         interferencefringes are sometimespresent in the spectra
   3710        molecularCO,          1 3 . 9+ 1 . 1       485 + 40          of such specimensand theselimit the accuracywith which
   35500       0H, H,O                 70+2             35000+500           absorption intensities can be measured,particularly for
   2350        molecularlrCO,        945 + 45           25200 + 12OO
   2287        molecular'3CO,        1 1 . 7+ 1 . 0 8                       weak absorptions.
   1610c       carbonate              199 + 17          27300 + 2300
   1375        carbonate             235+20             16300+1400          Determination of extinction coefficients
   Note. Values determined as described in the text. Errors for €2s50and        In order to determine the concentration in the glassof
(j37sevaluated as described in Fine and Stolper (1985). All others based
    propagationof theseerrors with the standarderrors obtainedby regres-
                                                                             the speciesresponsible for a particular absorption, we
sion of band intensitiesagainst the 2350-cm 1 or 1375-cm 1 band inten-       must know the band intensity, the sample thickness,the
sities;i.e.,the best-fitratioof intensities the 2350-and 371o-cm-1
                                          of                       bands     glass density (estimated in this study from the data of
                                            :       :
is 68 with a standard error of 4, so €3710 er"us/68 13.9, and the error
on €3710 13 9 x [(4/68), + (451945)21112.
                                                                             Kushiro, 1978,on COr-freealbitic glasses     quenched   from
   Speciesconcentrations(as weight percentof CO, that would be released      melts at temperaturesand pressuressimilar to those of
from the sample if all oI the specieswere converted to CO, and removed)      our study), and the extinction coefficient (or molar ab-
can be calculated follows:c: (absorbance 44 ol)/(densityx thick-
ness x e), where density is in g/L and thickness is in cm. lf integrated     sorptivity) ofthe band in question. The extinction coef-
absorbance is used, € is replaced with e.. For water contents, 44.01 is      ficient is the constant ofproportionality between the in-
replaced 18.02
           by                                                                tensity of the absorption and the number of absorbersper
  o From Silverand Stolper(in prep.)
  BThis value has been refined since our preliminaryreport of this work      unit area in the path of the infrared beam. It must be
(Johnson et al., 1985). The concentrationslisted in Table 3 are based on     determined empirically by determining the intensity of
the extinction coefficientsgiven here and supercededpreviouslyreported
vatues.                                                                      the absorption band in samplesin which the concentra-
  c This refers to the maximum peak height or total integratedintensityof    tion of the absorber is known. Molar absorptivities and
the group of bandsbetween1600 and 1700 cm 1.                                 integrated molar absorptivities have been previously de-
                                                                             termined for the molecular ''CO, band at 2352 cm I and
                                                                             for the carbonate   bandsat 1600-1700and,1375cm-' by
 to the thickness of the COr-bearing glass,had been nu-                      Fine and Stolper(1985) using a seriesof syntheticCOr-
 merically subtractedfrom it. This resulted in a spectrum                    bearingglasses   near the jadeite-silicajoin. Details ofthe
 with a relatively flat background from which the peak                       procedure used to determine these coefficientsand of its
 heightsand areasofthe bandsat 3550, 2350, 1600-1700,                        uncertaintiesare given in that paper. The values ofthese
 and 1375cm-' could be readily determined.The band at                        constants used in this paper are listed in Table 2; the
 37 l0 cm-' sits on the high-energy of the band at 3 550
                                     tail                                   integrated molar absorptivities for the carbonate bands,
cm '; its intensity and integrated intensity were deter-                    which were determinedas in Fine and Stolper(1985)bv
mined after the 3550 cm I band had been approximately                       the best-fit ratio of the integrated band intensity to the
nulled out by numerical subtraction of the spectrum of a                    peak height for each band, are slightly different from those
water-bearing,  COr-freealbitic glass.The 2287-cm ' band                    given in Fine and Stolper (1985) since we have enlarged
due to '3CO, sits on the low-energytail of the 'rCO, band                   the data set used in determining this ratio by including
at 2352 cm '. Unfortunately, we do not have spectraof                       the data obtained in this study. The best-fit value ofthe
 ''COr-bearing glasses  that contain no r3COr,so the shape                  ratio ofthe intensitiesofthe 1375-and 1600-1700-cm '
of the low-energy tail of the '2CO2band on which the                        carbonate bands has also been updated by the data ob-
 ''CO, band sits is not unambiguously known. Conse-                         tained in this study.
quently backgroundsfor this band were drawn by hand.                            The molar absorptivity for the 3710-cm ' band listed
    Although intensities of all six of the absorption bands                 in Table 2 was determined by regressing intensity of
could, in principle, be measuredfrom each spectrum, in                      this band with that of the 2352-cm-' band for sevenun-
practice, there are several limitations. The first is that                  dersaturated   albitic and jadeitic COr-bearingglasses  from
absorbances   greaterthan 1.5-2.0 cannot be reliably mea-                   the study of Fine and Stolper (1985).Unfortunately, when
sured on our instruments. Hence, for very intense ab-                       the 2352-cm-' band is on scale,the 3710-cm ' band is
sorptions,intensitiescannot be determinedprecisely.This                     very weak, and this ratio could be more tightly con-
is usuallythe casefor the '2CO,band at 2352 cm-r; sam-                      strained. The integrated molar absorptivity of the 3710-
ples would have to be unrealistically thin (i.e., less than                 cm ' band and the molar absorptivity of the 2287-cm-l
a few tens of micrometers) in order for this absorption to                  '3CO, band were determined by regressing integrated
be on scale for the concentrations of molecular CO, in                      intensity of the 3710-cm-1band and the peak height of
most of the vapor-saturatedglasses      that we synthesized.                the 2287-cm' band againstthe intensity of the 3710-
Consequently,    the bands at 31 l0 and 2287 cm-r were                      cm-r band. Note that the extinction coefficientofthe 2287-
usednearly exclusively in this study for determining mo-                    cm-r band is used to give the total dissolved molecular
lecular CO, concentrations.In the caseof the carbonate                      COr, not just the dissolvedmolecular'3COr,      and thus can
bands, these sit on the high-energytail of the prominent                    only be used when the t3C/t2Cratio is normal. We also
                              STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT                                       1075

note that the ratio of the band at 2352 cm-' due to mo-                              o olbite
lecular '2CO, to the band at 2287 cm ' due to molecular                              tr jod€iie
''CO, is, accordingto our preferred value, 80 + 5, which                             A euleclrc
is similar to their abundanceratio of 89; though there is                 o
no a priori reasonthat theseratios should be identical, it                o                           o
is encouragingthat they are similar.                                       N

Accuracy and precision
    Figure 2 compares the "actual" concentration of COz                   ;e                      (a)(o)

 (basedon the amount of CO, included in the syntheses                      ;
 of these standard glasses)  with the concentration of CO,
 obtained by summing the concentrations of dissolved                             o         o.5        ,i
 carbonate and molecular CO, in the set of glassesused                                wf. 7o CO. looded
by Fine and Stolper (1985) to determinethe molar ab-
                                                                                       of                    of
                                                                  Fig. 2. Comparison total CO, contents undersaturated
 sorptivitiesof the bands at 2352, 1600-1700,and 1375           sodiumaluminosilicate  glassessynthesizedby   Fine and Stolper
cm-t. The closecorrespondence       betweenthesetwo values      (1985) based (l) the amount
                                                                             on                loaded into thecapsule (2)
demonstrates the feasibility of determining total dis-          infraredspectra usingthe extinctioncoefficients givenin Table
 solved CO, concentrations in glassesalong the jadeite-         2. This figure been
                                                                              has     updated from the work of FineandStol-
 silica join using infrared spectroscopy.                       per (1985)by re-analyzing samples with the rrrn and in partic-
    In a few cases, Fine and Stolper (1985) noted significant                    the
                                                                ular by analyzing central  portionsof the samples    whenpos-
deviations between the amounts of CO, loaded into ex-           sible. threesymbols parentheses samples appear
                                                                      The              in            show           that
periments and the concentration of dissolved CO, deter-                                                   it
                                                                to havelostsignificant duringsynthesis; is Iikelythatthese
                                                                samples  camefrom nearthe capsule  wall. These   threesamples
mined spectroscopicallyin quenched glassesrecovered
                                                                werenot usedin the determination extinction     coefrcients.
after the experiments. The measuredconcentration was
in every caselower than the amount loaded into the ex-
periment. Using the finer spatial resolution possiblewith       ual measurements),in estimated density (+2o/o), and
the Nicolet FrrR,we have exploredpossiblecauses theseof         sample thickness (+3 pm) and varies from sample to
deviations. Figure 3 demonstratesthat a typical under-          sample. However, as discussedabove, a number of fac-
saturated COr-bearing glass of the sort used in the cali-       tors (interferencefringes in thin samples,ambiguities in
bration procedure is zoned, with higher dissolved CO,           backgroundsubtractions,nonlinearity of absorbances    for
and lower HrO contents in the core of the sample than           very intense absorptions) may degradethe precision for
in the rim next to the enclosingPt capsule.Though other         any individual measurement.On the basis of repeated
explanations are possible, we suggestthat this is due to                        of
                                                                measurements identical spots on several samplesand
diffusion of H into the Pt capsule followed by reaction         multiple measurementsof individual speciesconcentra-
of the H with dissolved CO, to form some HrO and                tions on a single spot (e.g.,we could make six potential
reduced C species.    The reduced C specieseither then re-      measurements molecular CO, concentration from any
mains in situ or diffusesas atomic C out of the Pt capsule      spectrum: peak heights and integrated intensities of the
as suggested Watson et al. (1982). This can account
               by                                               bands at 3710, 2352,2287 cm-t), typical uncertainties
for the systematicallyhigher dissolved water contents of        (lo) for the samplesreported on in this study are on the
glasses   synthesized with CO, over those without COr, the                         or
                                                                order of 0. I wto/o better for dissolved molecular COr,
positive correlation between run length and dissolved                      for
                                                                0.02 wto/o CO, dissolved as carbonate,and 0.01 wto/o
water content and its negativecorrelation with dissolved        for dissolvedwater. The poorer precisionof the molecular
CO, content that has been observed for undersaturated           CO, determinations is due to the low intensities of the
glasses(seeTable 3 for the results of experiments con-          3710- and 2287-cm' bandsusedin nearlyevery sample
ducted by A. L. Boettcher),and the fact that undersatu-         reported on in this study to determine molecular CO,
rated glassesoccasionally contain significantly less CO,        concentrations.If thicker samples,much thinner samples
than was loaded into the capsule prior to synthesis.As          (bringing the 2352-cm ' band on scale),or '3C-enriched
shown in Figure 3, however, the concentration measured          CO, were used (increasingthe intensity of the 2287 -cm I
in the center of the capsule by infrared spectroscopyis         band), the precision of the molecular CO, concentrations
essentiallyidentical to the amount loaded into the cap-         could be improved considerably,and we recommendthat
sule, giving us added confidencein the accuracy of our          efforts be made in future studies to do one or more of
primary calibration.                                            thesethings.
   The precision of the infrared determinations of species
concentrationsis difficult to assess general.Under op-
                                       in                                                RBsur-rs
timum circumstances,the precision will be on the order            The conditions of the experiments and the measured
of a few percent basedon potential uncertaintiesin mea-         HrO, molecular COr, carbonate,and total dissolved CO,
sured absorbances     (typically 0.005 for an absorbanceof      contents of the quenched glassesare listed in Table 3.
0. I with the rrrn basedon the reproducibility of individ-      The pressureand temperature dependence the molec-
1076                                    STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT

    3.      of          results
TeeLe Summary experimental
                                             Dura-        wt%
                 Starting         f           tion        CO,                 H,O"                   MolecularCO2B            CO6 ec               Total COro
    Run          materiala       fc)         (min)       loaded              (wt7.)                      (wt%)                 (wt%)                 (wt%)

                                                                              15 kbar
TJ41              Ab(c)          1450           65         1.9        0 . 6 11 2 , 1 2 , 0 . 0 3 )
                                                                            (                            (3,
                                                                                                     0.59 s, 0.04)      0 . 1 8 (9,34,0.02)        O.77(0.04)
ALB2965           xtal           1450          120         5                (2,
                                                                      0.14 2, 0.03)                  0.70H 1, -)
                                                                                                          (2,           0 . 1 6 (1 2, 0.01)
                                                                                                                                  ,                   (0.01)
ALB2998           xtal           1450          180         1          0.33 1, -)
                                                                            (1,                      0.60" , 1,-)
TJ42              Ab(c)          1525           60         2.4        o.72  (4,4,O.O1l                   (3,
                                                                                                     0.55 4, 0.03)      0.21 (3,9, 0.02)           0.76(0.04)
                                                                              20 kbar
TJ37              Ab(c)          1450           60         2.3        0.65 , 1,-)
                                                                            (1                       112(1,1,-l         0.27 ( 1 , 2 , 0 . 0 1 )   1.39
TJ48              Ab(c)          1450           60                          (1
                                                                      0.14 , 2, 0.01)                    (3,
                                                                                                     0.99 5, 0.09)      0.27 (3,9, 0.01)           1.26(0.10)
ALB2995G          xtal           1450           60         1                (3,
                                                                      0.20 3, 0.07)                      (3,
                                                                                                     0.72 5, 0.03)      0.20 (3,8, 0 01)           0.92(0.03)
ALB2982n          xtal           1450          180         1                (3,
                                                                      0.34 3, 0.01)                      (1,
                                                                                                     0.58 3, 0.09)      0 . 1 8(1, 2, 0.01)        0.76(0.10)
ALB299OG          xtal           1450          360         1          0.49 1,-)
                                                                            (1,                      0.39 , 1,-)
                                                                                                         (1             0 . 1 4(1, 3, 0.01)        0.s3
TJ4               Ab(A)          1525          100         2.4        0.42  (s,5,0.01)                   (5,
                                                                                                     0.95 13,0.06)      0.28 (3,10,0.01)           1.23(0.06)
TJ5               Ab(A)          1625           60         2.9              (3,
                                                                      0.28 3, 0.00)                      (3,
                                                                                                     0.80 s, 0.03)      0.31 (3,10,0.02)           1.11(0.04)
                                                                              25 kbar
TJ12              Ab(A)          1450             60       2.0              (2,
                                                                      0.34 2, 0.01)                      (4,
                                                                                                     1-26 10,0.07)      0.39 (3,10, 0.01)          1.65(0.07)
TJl 5             Ab{A)          1525             65       2.2              (4,
                                                                      0.44 4,0.O2)                       (2,
                                                                                                     1.O8 4,0.12)       o.42 (4,12,0.011           1.50(0.12)
TJ22              Ab(c)          1625             60       2.0        0.42 1,-)
                                                                            (1,                      1.130 , 3, 0.05)    _E                         -E

TJ36              Ab(c)          1625             60       2.2              (6,
                                                                      0.56 6, 0.03)                  1.0s(6,14,0.13)    0.47 (6,22,
                                                                                                                                  0.02)            1.52(0.13)
                                                                              30 kbar
TJ32              Ab(c)          1450             60       2.6        0 . 6 7 1 ,1 ,- )
                                                                            (                            (1
                                                                                                     1.44 , 3, 0.04)    6 . 5 91 i ,1 ,_ )
                                                                                                                               1                   1.94
TJ33              Ab(c)          1450             65       2.7        0 . 6 9 1 ,1 ,- )
                                                                            (                            (1
                                                                                                     1.35 , 3, 0.07)    0 . 4 6( 1 ,1 ,- )         1.81
TJ43              Ab(c)          1450             60       3.4        o.54  (4,4,0.02)                           1)
                                                                                                     1.42(4,8, 0.1      0.s4 (2,6,0.03)            1. 9 6( 0 . 1 )
TJ4O              Ab(c)          1525             90       3.0        0.76  (2,2,0.07)               1 28(2,4,0.O5)            (2,
                                                                                                                        0.72" 6, 0.03)             2.00 (0.06)
T t20             Ab(c)          1625             70       3.0        0.67  (3,3,0.04)                   (2,
                                                                                                     1.21 3, 0.05)      0.63 (4,9,0 06)            1.84(0.08)
  ^ Two separate batches of albitic glass starting material
                                                             [Ab(A) and Ab(C)]were used. The ALB sampleswere preparedfrom crystalline(xtal) albite.
   s c (n., nr,o): c is the average concentrationin weight percent; 4 is the number of separatespots or fragments for which spectra were obtained; n2
is the number of measurementsof concentrationdeterminedfor this sample. n2 > q since several measurements speciesconcentrationcan usually
be obtained from one spectrum. o is the error of the reported concentration(the mean of the n, separate measuresof concentration)given by 10 of
the distributionof the n, measurements.
   c This is the amount of CO, dissolvedas carbonate.
   Dc (o): c is the sum of the amount of CO, dissolved as molecularCO, and as carbonate. o is the error in this concentrationbased on the 10 errors
in the two soecies' concentrations.
   EConcentrationnot determined.
   FNot known, but the sample contains excess vapor in bubbles.
   GThese three sampleswere loadedwith the same amount of CO, (1 wto/")and held at run conditionsfor differentlengthsof time. Note the decreasing
CO" and increasingH,O content with time. ALB2982 and 2990 do not have bubbles. Bubbles(=10 rrm) are present but uncommonin ALB2995.
   HThese values are based on spectra of less than optimal quality.

                          .    CO2 , lolol                                           ular CO, content, carbonatecontent, and total dissolved
                          I    CO2 os corbonole
                          a    moleculor COz
                                                                                     CO2 content are displayed in Figures 4 and 5.
                          o    H20, lotol                                              The effects of pressure.Molecular CO, and carbonate
                                                       o mounl ot
                                                                                     concentrations,and consequentlytotal CO2 content, in-
                o.4                                    COz looded                    creasewith increasingpressureat constant temperature.
                                                       inlo copsule
                                                                                     The ratio of carbonate to molecular CO, also increases
            ;e "'"                                                                   under vapor-saturated conditions with increasing pres-
            E o.z                                                                    sure at constant temperature.
                                                                                       The effects of temperature.The concentration of mo-
                                                                                     lecular CO, under vapor-saturated conditions decreases
                                                                                     with increasingtemperature at constant pressure,but the
                               posilion                                              concentration of carbonate increasesslightly. These two
                                                                                     effectscounterbalanceeach other, leading to a total CO,
                                                                                     solubility that is nearly independent of temperature un-
                                                                                     der isobaric conditions. The ratio of carbonateto molec-
                                                                                     ular CO, increases  with increasingtemperatureunder va-
                                                                                     por-saturatedconditions at constant pressure.


  Fig. 3. Concentrationprofiles for dissolved COI-, molecular
COr, total COr, and water in sample ABC-58 (Fine and Stolper,                         a piston-cylinder run. Each analyzedspot is I 00 pm in diameter.
1985). A map showing the approximate location of each mea-                            Note that the total CO, concentration at the capsule'scenter is
surementis also shown. The glassusedis a horizontal section of                        very similar to the amount loaded into the capsule.
                                STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT                                         ro77

                                                        o l5 Kbor e 20Kbor
                                                        r 25Kbor o 3OKbor

                                                                              m o l e c u l o rC O ,



                                                                                30 K bor


                                   r 600

                         T("C)                                                                         b

                                              l o fo I C O 2



                                                               w t . 7 oC O z
  Fig. 4. (a) Molecular COr, (b) CO, dissolved as carbonate,and (c) total CO, concentrationsvs. temperature for COr-saturated
albitic melts. Error bars are 1o for the distribution of measurementsfor each sample; when only one measurementwas available
or the error is smaller than the symbol, no error bar is shown. SeeTable 3 for the errors on eachmeasurement.
                                                                                                            Lines are calculated
values based on the fits to Equations 4 and 7.

  The effects of water. Two pairs of experiments equili-               We note that the water contents of all of our run prod-
brated at the same temperatures and pressures(TJ-37                 ucts are disturbingly high. We have discussed possible
and TJ-48; TJ-41 and ALB2965) contain different                     sourcesof this water in these nominally anhydrous runs
amounts of dissolvedwater, yet their dissolvedmolecular             at length previously (Fine and Stolper, 1985, and above).
CO, and carbonate contents and hence their total dis-               We reiterate that we believe that it is unavoidable in pis-
solved CO, contents are, within error, the same. This                                                  of
                                                                    ton-cylinder experimentsbecause a combination of ad-
suggeststhat at leastat low water contents,CO, solubility           sorbed water on the starting materials (particularly the
and speciationare not strongly dependenton water con-               silver oxalate)and diffusion ofhydrogen into the capsules
tent.                                                               during run conditions. We doubt that previous studiesof
I078                               STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT

                ^ ' t 4 5 0 " c ! 1 5 2 5 " C. , | 6 2 5 ' C   it was in the undersaturatedexperiments,but the excess
                                                               CO, present in bubbles throughout the vapor-saturated
                                                               experiments probably approximately buffered the activ-
                                                               ity of CO, in these experiments.
                                                                  Achievementof equilibrium. The experimentswere not
                                                               reversed,so equilibrium has not been proved. However,
                                                               our run lengthswere l-1.5 h, greatlyin excess     ofthe l0
                                                               min or so that Mysen (1976) demonstrated to be suffi-
                                                               cient to reach a time-independent result, which he in-
                                                               ferred to be indicative of the attainment of equilibrium.
                                                               The two vapor-saturated experiments included in this
                                                               study that were conductedby A. L. Boettcher (4L82965
                                                               and 2998) were held at pressureand temperature for 2
                                                               and 3 h, respectively,and are similar to our l-h experi-
                                            c0rDonote          ment (TJ-41) held at the samerun conditions.
                                                                  Molecular COr: dissolved or in bubbles? We are fre-
                                                               quently asked if we can be certain that the molecules of
    P( kb o r )                                                CO, that we detect spectroscopically dissolved in the
                                                               glassrather than presentas bubbles or inclusions. Spectra
                                                               were obtained only on regions ofglass that, on the basis
                                                               of examination with an optical microscope, are bubble-
                                                               free or contain no more than a handful of bubbles. With
                                                               the help of Ian Hutcheon, we have demonstratedwith a
                                                               scanning-electronmicroscope that regions that appear
                                                               bubble-free with an optical microscope are free of bub-
                                                               bles down to a scaleof a few hundred lngstr0ms. In ad-
                                                               dition, we have demonstrated, as described above, that
                                                               the infrared spectraof CO, bubble-rich samplesdiffer sig-
                                                               nificantly from those of optically bubble-free samplesat
                                                               room temperature, and that the spectra of bubble-free
                                                               samplesshow negligible changesbetween room temper-
                                                               ature and liquid-nitrogen temperature whereas bubble-
                                                               rich specimensshow dramatic changesover this temper-
                                                               ature range,presumably owing to the freezingof the CO'
                  o                   1                        in the bubbles. Unfortunately, these observations only
                                w l . "/"CQ22
                                      "/"CQ                    rule out the presenceof CO, in relatively large (100 A?)
   Fig. 5. (a) MolecularCO,, (b) CO, dissolved carbonate,
                                                       as      bubbles.We cannot prove that molecular CO, is not pres-
and (c) total CO, concentration pressure COr-saturated ent in smaller clustersthat are not readily visible and do
                                    vs.            for
albitic melts.Error barsasdiscussed Fig. 4. Curves cal- not behave like a bulk, separatephase. However, since
                                           for             are
culated  values    based the fits to Equations and7.
                       on                          4           dissolved CO, molecules would probably be located in
                                                               "holes" in the melt structure and possibly in clusters,at
CO, solubility using analogousmethods of glass prepa- this level the distinction between bubbles and dissolved
ration have been successfulin establishing anhydrous CO, may be a semantic one. Nevertheless,the simple
conditions either.                                             patterns in our results [e.g.,regularitiesin the concentra-
   Sample heterogeneity.For most samples,spectrawere tion of molecular CO, with pressureand temperatureand
obtained on several diferent bubble-free regions of the in the ratio of molecular CO, to carbonatewith pressure,
glass,allowing us to assess     sample homogeneity.Table 3 temperature,      and melt chemistry (Fine and Stolper, 1985)l
lists the number of spots analyzed for each specieson strongly suggest          that the concentrationsof molecular CO,
each sample. The standard deviation of the distribution that we observeare controlled by thermodynamics, rath-
of analyses typically similar to their precision, indicat- er than by the vagaries of some mechanism of physical
ing that the samplesare usually homogeneous the level entrapment.
that we are capable of determining. There are a few ex-
ceptions,but only in a few cases the standarddeviation
                                         is                                 TrrnnvroovlqAMlc TR-EATMENT
greater than about l0o/oof the value of the mean. Note            Our data can be used to constrain a thermodynamic
that the homogeneityof the vapor-saturatedsamplescon- description of the solubility of CO, in silicate melts, in
trasts with the marked zoning of undersaturatedexperi- much the same way that Speraand Bergman (1980) did
ments (seeFig. 3). CO, may have been lost continuously for previous measurementsof CO, solubility in silicate
by reduction in the vapor-saturatedexperimentsmuch as melts. However, in that case, only bulk solubility was
                                STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT                                      t079

considered,whereasour data permit a separation of the            (seeFine and Stolper, 1985).This choice of activity-com-
heterogeneousequilibrium between melt and vapor and              position relationshipis analogous the one we have used
the homogeneous equilibrium between C-bearing species            for water-bearingmelts (Stolper, 1982b1,  Silver and Stol-
in the melt.                                                     per, 1985).It is unambiguoussince it is independentof
                                                                 the formula unit chosenfor the C-free silicate,but it does
Yapor-melt equilibrium                                           not take into account the difference in size between the
  Let us first consider the heterogeneous
                                        equilibrium be-          various species,the facts that there are probably many
tween CO, in the vapor and molecular (mol.) CO, in the           distinguishablesorts of oxygensin the melt and that they
melt. This may be describedby the reaction                       probably mix as polymeric groups rather than as individ-
                                                                 uals, and that CO, may mix on sites different from the
              COr(vapor): COr,mol.(melt).                 (l)
                                                                 other species. However, the particular choice of actiwity-
The conditionof equilibriumis                                    composition relationship will probably make little ditrer-
                                                                 encein the determination of the thermodynamic param-
                      trff,:   pF3';._".                  (2)    eters since the concentrations of C-bearing speciesare
This can be readily modified into the following relation         Iow.
that allows the activity of molecular CO, in melt satu-            Given this choice of activity-composition relationship,
rated with CO, vapor at any P and T to be calculated             Equation 3 becomes
provided that it is known at some referencepressure(Po)
and temperature (?"0):                                                xF3'l-,n:XF6'j...'
                                                                          e,        1t".,;##3
      aF8j.-'T): aFSj.-",
                    <e",fi1ffffi                                                             .-
                                                                                        , -"Y l1 4 s ; [ " ( P - & )
                                                                                        "               RT

                         '*01 @*#:ln)                                                         -         - I ll
                                                                                                  af/o(P")l I
                                                                                                     n Lz tlJ' ror
                                  ry[+-^.]],                     where trfgr'[", is now the partial molar volume of molec-
                                                                 ular CO, in the melt and f10(COr,mol.,melt) is now the
                                                                 partial molar enthalpy of molecular CO, in the melt. If
where a[[rj,.", (P, 7) and a[irj,*" (Po, TJ are the activities
of molecular CO, in the vapor-saturatedmelt at P and T           we know X.or.-", for a vapor-saturatedmelt at a reference
                                                                 pressure and temperature and have values for tr?3r"'1,",
and at Po and,To,respectively,relative to a standardstate
of pure CO, molecules at the pressureand temperature             and for AHoat the referencepressure,Equation 4 can be
ofinterest (Fine and Stolper, 1985);f.o,(P, T) andf"o,(Po,       used to calculateX36';.-.,in vapor-saturatedalbitic melt
T) are the fugacitiesof CO, in the vapor at P and T and          at any other P and T.
at Po and To; Vo.Elli""is the molar volume of molecular            Using our data on the concentrationsof molecular CO,
CO, in the melt in its standard state and has been taken         and carbonateas a function of P and I and the modified
to be independent of P and T in deriving Equation 3;             Redlich-Kwong equation of state of CO, vapor (Hollo-
Mo(P) is 110(CO,,mol.,melt)- ,F10(CO,,vapor) po and              way, 1977), we have obtained values for the parameters
                                                                 in Equation 4 by least squares.These values are given in
f, where flo(COr,vapor) is the enthalpy per mole of CO,
vapor at a pressuresufficiently low for the vapor phase          Table 4. The reference temperature and pressure were
to be perfect at Zo and .I10(COr,mol.,melt) the molar
                                                is               taken as 1450 "C and 20 kbar. The partial molar volume
enthalpy of molecular CO, in the melt in its standard            of molecular CO, given by this regressionis 28.6 cm3/
state at Po and Zo.Note that in deriving Equation 3, AHo         mol. This is somewhatlower than the value of 33-35 cm3/
has been assumed to be independent of temperature,               mol determined by Spera and Bergman (1980) for the
but it is a function of pressure !AH"(P) : AH\(P) +              partial molar volume of CO, in albitic melt basedon the
4ff[", (P - P.)1.                                                data Mysen et al. (1976). It is also lower than the molar
  Assuming that the melt can be treated as an ideal mix-         volume of the CO, vapor under these conditions (34-40
ture of CO, molecules,carbonategroups, and oxygen at-            cm3/mol), but similar to the D parameter for CO, mole-
oms (Fine and Stolper, 1985),we have                             cules in the Redlich-Kwong equation of state (29.7 cm3/
              4331,-" n: 43:,^",(P,
                                                                 Molecular COr-carbonateequilibrium
               aE31-(P, re&-(P,7)
                      D:                                            Let us now consider the homogeneous equilibrium be-
                 a\it:(P, T) : XStt:(P,T),                       tween CO, molecules,carbonategroups, and the silicate
                                                                 framework in albitic melt. We will describe this via the
where the mole fraction of each speciesis simply the             following reaction:
number of moles of that speciesin the melt divided by
the total number of moles of all three speciesin the melt              CO,,mol.(melt)+ Ots(melt): CO3-(melt).          (5)
1080                            STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT

        TreLe 4.   Best-fit thermodynamicparameters           sumed to be independent of temperature in deriving
                   for Equations4 and 7                       Equation 7, but it is a function ofpressure.
                                                                 Using our data on the concentrationsof molecular CO,
                Cor(vapor) : CO,,molecular(melt)
                                                              and carbonateas a function ofPand 7, we have obtained
                       Po:20kbar                              values for the parametersin Equation 7 by least squares'
                       ro: 1450"C
             XH." (Po, : 0.0067 0.0002
                                 +                            Thesevalues are given in Table 4. The referencetemper-
              t4trh (e D:28.6 t 0.5cm3/mol                    ature and pressurewere taken as 1450 "C and 20 kbar.
                  aH" (4) : 9200+ 1600cal/mol                 The negativevolume changeof this reaction and the pos-
                                                              itive enthalpy change are reflections of the increasesin
           Co,,molecular(melt)+ O'?(melt): CO3 (melt)
                                                              the carbonate to molecular CO, ratio observed with in-
                       Po:20 kbal                             creasingpressureand temperature.
                       %: 1450.C
               lQ (P", To): 0.27 + O.01
                                                              Calculation of CO' solubility and speciationin
               AVP e D: -3I + 0.8cm3/mol
                 AF/P(P.): 13500 + 3000 cal/mol
                                                              albitic melts
                                                                 The parameters given in Table 4 can be used to cal-
                                                              culate CO, solubility and speciation in vapor-saturated
                                                              albitic melts as a function of pressureand temperature'
We recognizethis as a gross oversimplification since it       First, Equation 4 is used to calculateXco2,^or a given
does not specify the structural changesthat take place in     set of values of P, T, andf.or. Then, Equation 7 is used
the silicate framework of the melt as this reaction pro-      to calculate Kc at this P and T. Given X55';,-"r + X35? +
ceeds.In the absence    ofany concreteinformation on the       }ff5t:: 1, and the value of Xp5rj,-' derived from Equation
details of this reaction, we will describethe homogeneous     4, the calculated value of K. can be readily used to de-
equilibria with this generalized equation plus the as-        termine XBi\ and Xg:':.
sumption of ideal mixing of the CO, molecules,carbon-             Weight pi:rcentages molecular CO, and of CO, dis-
ate groups, and oxygen atoms described above. Though          solved as carbonate as well as total dissolved CO, con-
this treatment, in which basically all of the oxygens in      tents were calculated for the pressure and temperature
the melt will be considered to be indistinguishable and       range covered by our experiments using the parameters
equally available for reaction with molecules of COr, is      given in Table 4 and compared to our measuredvalues
an oversimplification, the thermodynamic parameters           in Figures 4 and 5. Not surprisingly, since thesewere the
derived from it are not expected to depend strongly on        data used to constrain the thermodynamic parameters
the details becausethe concentrations of the C-bearing        given in Table 4, the calculations are a good fit to the
speciesare low (Fine and Stolper, 1985); that is, similar     data; the mean of the absolute value of the deviation
derivative parameters(e.g., AV, A1{) would be obtained        betweenmeasuredand calculatedmolecular CO, concen-
regardless the details of the reactions being modeled.
            of                                                trations is about 0.07 wto/o  and for carbonate it is about
  The condition for equilibrium for Reaction 5 is             0.03 wt0/0.  These deviations are comparable to our pre-
                          +      :
                   p35'j,*., p.8tl p33?                 (6)   cision at a lo level. The maximum deviations (0.22 wto/o
                                                              for molecular CO' and 0.06 wto/ocarbonate [except for
This can be rearranged give
                     to                                       TJ-40, which was excluded from the fitting procedurel)
                                                              are at the 2o to 3o level.
      T):   r.r*nl -L4(!: P")
  K,(P. K.(P.,                                                    The calculated total CO, solubility at I bar and 1450
                                                              "C is about 0.5 ppm by weight, about 900/o which is
                                 - an9(ruf t.ll ,1\           dissolvedas molecules COr.of
                                     R L7-;r;ll''"        Calculation of the Cor-saturated solidus of albite
where                                                       Using the proceduresand data given in Silver and Stol-
                                    vmerl                 per (1985)plus the parameters   listed in Table 4, we have
                              -    "co(- '            (8) calculatedthe effect of CO, on the melting of crystalline
            &:;#sgl-                     XBir'
                 tr.o,,*o,.4[5I XF$']...r.
                                                          albite. The calculated freezing-point depression(i.e., the
                                                                    in              betweenthe dry solidus and the
Po(bars)and Zo(K) are a referencepressureand temper- difference temperature
                                                          beginning of melting of albite in the presenceof CO, va-
ature; P and I are some other pressureand temperature;
                                                          por) increasesfrom 4 "C at 5 kbar, to 13 'C at 15 kbar,
A7'Pis the volume changeof the componentsof Reaction
                                                                      25 kbar. These values are consistentwith the
5 in their standard states,which is equal to the volume to 25 "C at
                       given our assumption of ideal mix- data reported by EggJer  and Kadik (1979) and Boettcher
changeof the reaction
ing and was taken to be independent ofP and Z in de-      et al.(1987).
riving Equation 7; AH?(P,)is the changein enthalpy of
the components of Reaction 5 in their standard statesat           Covrp.tnrsoN wITH pREvlous sruDrEs
Po,which is equal to the enthalpy changeof the reaction      There have been several previous studies of CO, sol-
at Po given the assumption of ideal mixing. AIIp was as- ubility in albitic melts. Mysen et al. (1976) used both
                                STOLPER ET AL.: SOLUBILITY OF CO. IN ALBITIC MELT                                           108I

                                                            1 Ok b o r                            1 5k b o r



                                                            2O kbor                              25 kbor


                                15 0 0


                                           30 kbor                               w l % C O 2t o t o l
                                                                                 Mysen t ol (1976)
                                                                             o   p - t r o c kc o u n t r n g
                                                                             o    gos chromologrophy
                                                                                B r e y( 1 9 7 6 )
                                r500                                         tr qoschromotogrophy

                                                                                 M y s e n n d V i r q o( 1 9 8 O )
                                                                             ^   B - t r o c kc o u n t i n q
                                         o1234                               I     i nfroredspeclroscopy
                                                 wl. % CO, lolol
  Fig. 6. Total CO, solubility vs. temperature for various pressures from this work and the literature. Error bars from this work
as describedfor Fig. 4. Error bars from Mysen et al. (1976) and Mysen and Virgo (1980) are as reported by them in their data
tables.Lines are calculatedvalues based on the fits to Equations 4 and 7.

B-track radiography and gaschromatographyto measure                                                    As
                                                                      Total solubility measurements. shown in Figure 6,
CO, solubility in albitic melts at 5-30 kbar and 1350-             there is a conspicuous lack of correspondencebetween
 1700 'C. In a companion paper, Mysen (1976) reported              the severalreports of total CO, solubility in albitic melts.
molecular CO, and carbonate concentrations based on                In general,our values are less than those of Mysen et al.
infraredspectroscopy.  Brey (1976)reporteda singlemea-             (1976) and Mysen and Virgo (1980), but the deviations
surementat 1450'C and 30 kbar using gas chromatog-                 are not systematic.  For example,at 1450'C and 30 kbar,
raphy and also reported some infrared spectroscopicre-             there is reasonableagreementbetween the four reported
sults. Mysen and Virgo (1980) used B-track radiography             values: l. 6 wto/o                              (Brey, 197
                                                                                     (Mysenet aI., 1976), 2. I wto/o         6),
to measure   CO, solubility at l0-30 kbar and 1450-1750            2.06 wto/o (Mysenand Virgo, 1980),and I .8-2.0 wto/o   based
'C; they used Raman spectroscopyto measure
                                                  relative         on three experimentsfrom this study. Similarly, at 1525-
molecular CO, and carbonateconcentrations.                          1550 'C and 30 kbar, our value of 2.0 wto/o bracketed
   All previous measurements total CO, solubility un-
                                of                                 by the valuesof 1.8 wto/o reportedby Mysen etal. (1976)
der nominally anhydrousconditions, which should be di-             and 2.31wto/o   reportedby Mysen and Virgo (1980).How-
rectly comparable to our results, are shown in Figure 6            ever, under most other conditions, our values tend to be
along with our data and our best-fit solubility function           lower than those of Mysen et al. (1976) and Mysen and
basedon the parametersgiven in Table 4. Measurements               Virgo ( 1980),evenexcludingthe gaschromatographicval-
of molecular CO, and carbonateconcentrationsfrom this              ues that Mysen et al. (1976) rejected as probably being
study are compared with those of Mysen (1976) in Figure            too high becauseoftrapped bubbles.
7. In this section,we discussthese different data sets.              We have considered several possible explanations for
1082                            STOLPER ET AL.: SOLUEILITY OF CO, IN ALBITIC MELT

                                                          1350 "C

                                                                 | 450 "C
                                                                    . 45 2 q o a

                       ;e             !,"!"!,
                                      \l /,


                                        lO Kbor

                                o                     0.5              1.o
                                                    wt."/" CO2 os corbonole
   Fig. 7. Molecular CO, vs. carbonateconcentrationsfor COr-saturatedalbitic melts from this work and Mysen (1976). Curves
are calculatedisotherms and isobars basedon the fits to Equations 4 and'l , Error bars as describedfor Fig. 4. Note the wide scatter
in the results of Mysen (1976) compared to those in this work.

the differencesbetweenour results and those of Mysen et               how this would lead to lower values under some condi-
al. (1976)and Mysen and Virgo (1980)that would reflect                tions but not under others. (3) If there were problems
problems with our experiments or analyses:(l) The fact                with our calibration, deviations would be expectedto be
that we obtain similar results whether we analyze our                 systematic.(4) A final possibility is that under some con-
experiments or those conducted by A. L. Boettcher sug-                ditions there are forms of C other than molecular CO,
geststhat differencesare not due to some peculiarity of               and carbonate dissolved in our glassesthat we fail to
our solid-media apparatusor of our starting material. $./e            detect by infrared spectroscopy.  We note that if the mys-
note that run number ALB2998 (15 kbar, 1450 "C,7o/o                   tery species were a reducedform ofC, this could explain
CO, loaded) contains widespreadbubbles up to 30 pm in                 the zoning and in particular the apparent C-deficiency
diameter, suggesting-in agreementwith our results but                 observed near the rims of undersaturated experiments
conflicting with those of Mysen and his co-workers-a                  (Fig. 2). However, if a mystery speciesis the reason for
solubility of lessthan I wto/o CO, under theseconditions.             the observed deviations, its concentration does not ap-
The presence uncommon, small (=10-pm diameter)
                of                                                    pear in vapor-saturatedsamplesto be correlatedwith dis-
bubbles in 4L2995 (20 kbar, 1450 "C, loloCO, loaded)                  solved water content (as would be expected if it were
would suggest   solubilities even lower than those we have            related to an influx of H) and would have to be highest
reported for 20 kbar, but rare, small bubbles may not                 at low pressures  and at high temperatures.
necessarily reflect vapor saturation (Burnham and Jahns,                 There is nothing obviously wrong with either our mea-
 1962). (2) Our experiments were run much longer than                 surementsor those of Mysen et al. (1976) and Mysen and
those of Mysen and his co-workers, but it is hard to see              Virgo (1980) using B-track radiography, so the cause of
                              STOLPER ET AL.: SOLUBILITY OF CO, IN ALBITIC MELT                                     l 083

the deviations betweenthesetwo data setscannot be as-          we reiterate that conclusion here. Our results do not sup-
signed.We note, however, that contrary to the statement        port the earlier suggestions  (Mysen, 1976 Mysen et al.,
in Mysen and Virgo (1980), their data are not in very           1916) that there are abrupt changesin melt polymeriza-
good agreement with the earlier measurementsby the             tion over the l5-30-kbar range that lead to dramatic
sametechnique     first reportedby Mysen et al. (1976).This    changesin CO, speciation and solubility. Our results in-
is apparent in Figure 6, where experiments conducted           dicate that there is a subtle, gradual increasein the ratio
under identical conditions have CO, concentrationsmea-         of carbonate to molecular CO, over this pressurerange
sured by Mysen and his co-workers differing by more            and that most of the increasein total CO, solubility is
than 0.5 wto/o  and experiments conducted under similar        due to the increasing concentration of molecular CO,
conditions (i.e., differing in temperature by 50 "C) differ    driven by the difference in volume between CO, in the
by almost 0.9 wto/0. l0 kbar, the earlier solubilities are
                      At                                       vapor and CO, moleculesin the melt.
the high ones; at 30 kbar it is the reverse.There clearly         The effectsof water on CO, solubility. Previous studies
is some lack of reproducibility in the CO. solubilities re-    have indicated that CO, solubility in albitic melt in-
ported by this group. Sinceour solubility values typically     creasesby about 0.8 wto/owhen the water content in-
deviate from theirs by 0.5 wt0/oor less, perhaps the de-       creases from zero to about 8-9 wt0/o  (Eggler,1973;Mysen,
viations between our data sets reflect the fact that the        1976; Egglerand Kadik, I 979). This is thought to be due
precision of the B-track measurements poorer than has
                                          is                   to increasedconcentrationsofdissolved carbonate com-
been reported.                                                 plexes under hydrous conditions. Our data show no de-
   In summary, we are reasonablyconfident of our results,      tectableenhancementof molecular CO, solubility or car-
and there appearsto be a lack of reproducibility in the        bonate concentration under vapor-saturated conditions
 B-trackresultsreported by Mysen et al. (1976) and Mysen       as the dissolved water content increasesfrom about 0.1
and Virgo (1980). Nevertheless,we cannot track down            to 0.7 wto/0. the increaseis CO, solubility were linear
with certainty the causeof the discrepancies    betweenour     with water content, the CO, content would be expected
results and those in the literature and conclude that this     to increaseby about 0.05 wo/o as the water content in-
will only be resolved through the use of an independent        creases  from 0.1 to 0.7 wto/o; this could probably be de-
technique for measuringdissolved CO, contents.                 tected ifthe increasewere entirely in the carbonate con-
   Temperature dependenceof solubility. As shown in            centration. More work will be needed to determine the
Figures 4 and 6, our results show a negligible to slightly     exact nature of the relationship betweendissolved water
negative dependence total CO, solubility on tempera-
                         of                                    and CO. contents of vapor-saturatedmelts at low water
ture. A similar result was inferred by Brey (1976) for         contents.
albitic melt. Similar results have been found in investi-
gations of CO, solubility in several other silicate melt                             DrscussroN
compositions(Pearce,       1964;Faile and Roy, 1966; Shar-        Given the unexplaineddiscrepancies   betweenour mea-
 ma, 1979;Rai et al., I 983). Mysen et al. (1976) found a      surements of CO, solubility and those of Mysen et al.
similar result at about l0 kbar for albitic melt, but found    (1976) and Mysen and Virgo (1980),we are reluctantto
a strong positive temperature dependenceof CO, solu-           overinterpret the details of our results. There are, how-
bility at higher pressures  (Fig. 6). We have found no sim-    ever, two important points that we believe will stand in-
ple explanation for the discrepancybetween our results         dependentof the resolution of these discrepancies.
and those of Mysen et al. (1976) and Mysen and Virgo              The first is that understandingofthe total solubility of
(1980) and again suggestthat an independent technique          a volatile component such as CO, dependscritically on
be used to resolve it, while pointing out that the results     an understandingof the speciation of this component in
of these two studies by Mysen and his co-workers using         the melt. Becausethe total solubility of CO, is the sum
identical techniquesare not entirely consistentgiven their     of the CO, dissolved as molecules of COr, as carbonate,
reported error bars.                                           and perhapsin other forms as well, modeling of the vari-
   Molecular CO, vs. carbonateconcentrations.We have           ations in solubility of CO, must consider the effects of
found that pressureand temperature have an effect on           temperature,pressufe,and melt composition both on the
the ratio of dissolved molecular CO, to carbonate (i.e.,       equilibria betweenvapor and molecular CO, and between
Af{ and A,V! are nonzero in Eq. 7). However, as shown          the various dissolvedspecies.  For example,in albitic melt
in Figure 7, the variations in this ratio with pressureand     the solubility of molecular CO, decreases  with increasing
temperature are gradual and small compared to the re-          temperature, the concentration of carbonate in equilib-
sultsof Mysen (1976).We had previously concluded(Fine          rium with the dissolved molecular CO, increaseswith
and Stolper,1985),on the basisofdata from glasses       with   increasingtemperature,and the sum-the total solubility
lower total CO, concentrationsthan those studied by My-        of COr-displays almost no temperature dependence.
 sen (1976),that the large rangein CO'?lCOl- ratios that       However, in other melt compositions, the competition
he reported and the nongradual changesthat he found in         betweenthe temperaturedependencies ReactionsI and
 this ratio with pressureand temperaturewere artifacts of      5 may balance out differently. For example, in molten
 his use of the imprecise KBr pellet technique for his in-     SiOr, which probably dissolves CO, nearly entirely as
 frared determinations of carbonate concentrations, and        molecularCO, (Fine and Stolper, 1985),there will be no
I 084                          STOLPER ET AL.: SOLUBILITY OF CO. IN ALBITIC MELT

 carbonatewith its concentration showing a positive tem-         behavior. The secondderivative of the activity of a mo-
 perature dependence,so a negative temperature depen-            lecular species(in vapor-saturated melt) with respect to
 dence of solubility would be expected. In contrast, for         pressureis given by
jadeitic and nephelinitic melts, in which carbonatedom-
 inates over molecular COr, the positive temperature de-                     d2a f r dw lv"-v^\'1
 pendence of dissolved carbonate concentration could
                                                                             ap,:oln-nrE \-nr / I'
 overwhelm the negative temperature dependenceof the
 concentration of the minor species,molecular COr, with          where tr/" the molar volume of the vapor and Z- is the
 the net result being a positive temperaturedependence      of   molar volume of the molecular speciesin the melt in its
 the bulk CO, solubility.                                        standardstate(which is taken as a fictive form ofthe pure
    Treatments of the solubilities of other volatile com-        molecular species P and O. If the partial molar volume
 ponents such as water and SO, that can dissolve both as         of the speciesin the melt is zero and the gasis ideal, the
 molecular species   and as species  formed by reaction with     secondderivative is zero, and the activity of the species
 the silicate framework (e.g., hydroxyl groups, sulfate          is proportional to pressure.If the parlial molar volume
 groups) will require similar considerations both of het-        is positive and the gasis ideal, the secondderivative will
 erogeneous   equilibia betweenvapor and molecular species       be negative at pressuresless than 2RT/V^. At elevated
 in the melt and of homogeneous       equilibria betweenmelt     pressures  where the gasis nonideal, though the first term
 species. most cases, temperaturedependence the
                          the                          of        in Equation 9 is always negative,it becomessmall and if
 solubility of the molecular speciesis expectedto be neg-        the partial molar volume of the speciesin the melt is
 ative, reflecting the usually positive enthalpy of vapori-      small relative to the molar volume of the vapor, the sec-
zation of such species(i.e., things usually boil when heat       ond term, which is always positive, overwhelms the first
is added to them). For very large molecular species,    how-     and the secondderivative is positive. Basedon our data,
ever, the temperature dependencemay well be positive,            the partial molar volume of molecular CO, (28.6 cm1/
reflectingthe energeticcostsofforcing such molecules(or          mol) is small relative to the volume of the vapor in the
atoms, in the caseof the heavy rare gases)into holes in          l5-30-kbar range;consequently,the secondderivative is
 the melt structure into which they do not quite fit. In         positive in this pressurerange.The point is that the shapes
addition, the differencein enthalpy betweenthesemolec-           ofsolubility curves and other phaseboundaries can only
ular speciesin the gas phase and in the melt, even if            with difficulty be usedto infer microscopic behaviors and
positive at low pressure,will tend to decrease pressure
                                                   as            that, in particular, the rapid increase in CO, solubility
increases   from 1 bar, so that the temperaturedependence        observed between 15 and 30 kbar does not necessarily
tends to become less negative, and may even become               signal anomalous changesin melt structure or polymeri-
positive, as pressureincreases.Indeed, the temperature           zation over this pressurerange.
dependence water solubility may switch from negative
to positive with increasingpressure(e.g.,Kennedy et al.,                                 Surrtulny
 1962;Khitarov and Kadik,1973), and this could reflecr               l. Infrared spectroscopy    has been used to measuredis-
such an effect. In any case,the temperature dependence           solved molecular COr, carbonate, and water concentra-
of water solubility reflectsa competition betweenthe sol-        tions of albitic glasses    quenchedfrom melts equilibrated
ubility of molecular H,O and the homogeneous        equilibria   with COr-rich vapor at l5-30 kbar and 1450-1625'C.
betweenmolecular water and hydroxyl groups,which are                 2. At constant temperature,the molecular CO, and the
probably favored by increasing temperature (Stolper et           carbonateconcentrationsof vapor-saturatedalbitic melts
al., 1983), and thus is probably similar to the casewe           increasewith increasing pressure.For each speciesand
have describedfor CO, solubility.                                their sum (i.e., the total dissolved COr), the second de-
   The secondpoint that we want to make has to do with           rivative ofconcentration with respectto pressureis pos-
the pressuredependence CO, solubility. Note in Figure
                            of                                   itive over the l5-30-kbar range. The ratio of molecular
5 that the solubility of CO, doesnot increaselinearly with       CO, to carbonatedissolvedin albitic melts decreases     with
pressure,as is often supposedto be its normal behavior           increasingpressureat constant temperature.
(e.g.,Harris, 198l; Des Marais and Moore, 1984).In fact,             3. At constant pressure,the concentrationof molecular
accordingto our results, the rate ofincrease in solubility       CO, in vapor-saturated albitic melts decreases      with in-
increases   with increasingpressureover the pressurerange        creasingtemperature,but the concentration ofcarbonate
that we have investigated.Although a small part of this          increases.The sum of the concentration of these two
shapeis due to the increasingratio of carbonate to mo-           species   (i.e., the total dissolved CO, concentration)is ap-
lecular CO, with increasingpressure(i.e., to the negative        proximately independent of temperature at constant
AV! of Reaction 5), this shape is also observed for the          pressureunder the conditions that we have investigated.
molecular CO, solubility.                                            4. Thermodynamic analysis of the solubility data that
   We would be mistaken if we interpreted this as unusual        we have presentedrequiresthe considerationoftwo equi-
behavior reflecting changesin melt structure over this           libria. The first, CO,(vapor) : CO,,mol.(melt) (Reaction
pressure   range(e.g.,Mysen, 1976).In fact, note that the         l), describesheterogeneous       equilibrium between vapor
calculatedmolecular CO, solubility basedon Equation 4            and melt. The second, CO,,mol.(melt) + O'z(melt):
and the parameters given in Table 4 also display this            COI-(melt) (Reaction 5), describes      homogeneous  equilib-
                                     STOLPER ET AL.: SOLUBILITY OF CO' IN ALBITIC MELT                                                              5

rium betweenmelt species.   Our data on the variations in                        ities of molecular speciesin supercritical fluids of geologc interest.
the concentrations of dissolved molecular CO, and car-                                                                        69,
                                                                                 Contributions Mineralogyand Petrology, 315-318.
                                                                              Furukawa, T., Brawer, S.A., and White, W.B (1978) Raman spectro-
bonate under vapor-saturatedconditions can be well de-                           scopicstudy ofnitrate glasses.  Joumal ofChemical Physics,69, 2639-
scribed by Reactions 1 and 5 if aI 20 kbar and 1450 "C,                          2651
the partial molar volume of molecular CO, in albitic melt                     Harris, D.M. (1981) The concentration CO, in submarinetholeiitic
is 28.6 cm3/mol, the differencein enthalpy between dis-                          basalts. Joumal ofGeology, 89, 689-701.
                                                                              Holloway, J.R. (1977) Fugacity and activity of molecular speciesin su-
solved molecular CO, and pure CO, gas at low pressure                            percritical fluids. In D. Fraser, Ed., Thermodynamics in geology,p.
is 9.2 kcal/mol, the volume change Reaction5 is -3.9
                                    of                                            161-181.D. Reidel,Boston,Massachusetts.
cm3/mol,and the enthalpy changeof Reaction 5 is 13.5                          Johnson, T , Fine, G.J., and Stolper, E M. (1985) Solubility of carbon
kcal/mol.                                                                        dioxide in molten albite.EOS,66, I130.
                                                                              Kennedy, C., Wasserburg,
                                                                                          G                   G.J.,Heard,H.L., and Newton,R C. (1962)
   5. There are significant, nonsystematic discrepancies
                                                                                 The upper three phaseregion in the systemSiOr-HrO. American Jour-
between CO, solubility in albitic melt and its pressure                          nal of Science.  260. 501-521
and temperature dependenceas measuredby us and as                             Khitarov, N.I., and Kadik, A.A (1973) Water and carbon dioxide in
measured   usingp-track radiographyby Mysen et al. (1976)                        magmatic melts and peculiaritiesof the melting process.  Contributions
and Mysen and Virgo (1980).Thesewill have to be re-                              to Mineralogyand Petrology,     41,205-215.
                                                                              Kushiro, I. (1978) Viscosity and structural changes albite (NaAlSirOJ
solved by future work. The reports by Mysen (1976) of
                                                                                 melt at high pressuresEarth and PlanetarySciencelrtters, 41,87-90.
substantial variations in the ratio of molecular CO, to                       Mathez, E.A. (1984) Influenceofdegassingon oxidation statesofbasaltic
carbonatein COr-saturatedalbitic glasses the pressure
                                           in                                    magmas.     Nature, 3 10, 371-375.
and temperature range investigated in this study can,                         Moore, J.G. (1979) Vesicularity and CO, in mid-ocean ridge basalt Na-
however, be discounted. This ratio varies gradually over                         ture. 282, 250-253.
                                                                              Mysen, B.O. (1976) The role of volatiles in silicate melts: Solubility of
this range of conditions and only by relatively small                            carbon dioxide and water in feldspar,pyroxeneand feldspathoidmelts
amounts.Our data do not suggest   major or abrupt changes                        to 30 kb and 1625'C AmericanJoumal ofScience,276,969-996.
in melt structure or polymerization over this range of                        -(1977)          The solubility of HrO and CO, under predicted magma
conditions.                                                                      genesis  conditions and somepetrologicaland geophysical   implications.
                                                                                  ReviewsofGeophysics       and SpacePhysics,15, 351-361.
                     AcrNowr,oocMENTs                                         Mysen, B.O., and Virgo, D. (1980)The solubility behavior of CO, in melts
                                                                                  on the join NaAlSiror-caAlsi,o6-co, at high pressures    and tempera-
  We thank GeorgeRossmanfor his cooperationand Paula Rosenerfor                   tures:A Raman spectroscopic    study. American Mineralogist, 65, I 166-
her help with the rrn. The interest and assistanceofA. Boettcher,J. T             lL75.
Cheney, and J R Holloway are appreciated.Ian Hutcheon generously              Mysen, B.O., Eggler,    D.H., Seitz,M.G., and Holloway,J.R. (1976)Car-
helpedus with seu examination ofour glasses. This work was beganunder             bon dioxide in silicate melts and crystals:Part I. Solubility measure-
an UndergraduateSummer Internship (to T.J.) in the Division of Geo-               ments.AmericanJournalof Science,276,       455-479'
logical and Planetary Sciences,Caltech, and was the basis of an honors        Nakamoto, K (1978) Infrared and Raman spectraofinorganic and co-
thesissubmitted to Amherst College.The research  was supportedby NSF              ordination compounds (third edition). Wiley, New York.
Grants EAR-8 2 12765 and EAR- 84 l7 434, CaltechDivision of Geological        Pearce,M L ( I 964) Solubility of carbon dioxide and variation of oxygen
and PlanetarySciences  Contribution Number 4274                                   ion activity in soda-silicamelts. American Ceramic Society Journal,
                                                                              Rai. C.S.,Sharma,     S.K., Meunow,D.W., Matson,D.W, and Byers,C D.
                      RprnnrNcns crrnn                                            (1983) Temperature dependenceof CO, solubility in high pressure
Boettcher, A.L., Luth, R.W., and White, B.S. (1987) Carbon in silicate            quenchedglasses diopside composition. Geochimicaet Cosmochim-
  liquids: The systemsNaAlSi,O.-COr, CaAlrSirO8-CO,,     and KAlSirOr-            icaActa,47,953-958.
  CO,. Contributions to Mineralogy and Petrology,in press.                    Seward,T.P., ilI. (1980) Coloration and optical anisotropy in silver-con-
Brey, G. ( 1976)CO, solubility and solubility mechanismsin silicatemelts          taining glasses.Joumal of Non-Crystalline Solids, 40, 499-513.
  at high pressures.Contributions to Mineralogy and Petrology,51,215-         Sharma,S.K. (1979) Structureand solubility of carbon dioxide in silicate
  22r.                                                                            gJasses diopside and sodium melilite compositionsat high pressures
Burnham, C W , and Jahns, R.H. (1962) A method for determining the                from Raman spectroscopicdata. CarnegreInstitution of Washington
   solubility of water in silicate melts. American Journal of Science,260,        Year Book 78, 532-537.
   72t-745.                                                                   Silver, L.A., and Stolper,E. (1985) A thermodynamic model for hydrous
Des Marais,D.J., and Moore,J.G. (1984)Carbonand its isotopes mid-  in             silicatemelts.JournalofGeology, 93, l6l-178
   oceanicbasaltic glasses.     Eanh and Planetary ScienceLetters, 69, 43-    Spera,F.J., and Bergman,S C (1980) Carbon dioxide in igneouspetro-
   57.                                                                            genesis: Aspectsofthe dissolution ofCO, in silicate liquids. Contri-
Eggler,D H (1973) Role of CO, in melting processes the mantle Car-
                                                         in                       butions to Mineralogy and Petrology,74,55-66.
   negieInstitutionof Washington      Year Book 72,457467.                     Stolper, E. (1982a) Water in silicate glasses: An infrared spectroscopic
Eggler,D.H., and Kadik, A.A. (1979) The systemNaAlSi3Os-H2O-CO,         to        study. Contributions to Mineralogy and Petrology,81, l-17.
   20 kbar pressure:I. Compositional and thermodynamic relations of           -(1982b)          The speciationof water in silicate melts. Geochimica et
   liquids and vapors coexistingwith albite. American Mineralogist, 64,           CosmochimicaActa. 46, 2609-2620.
    1036-1048.                                                                 Stolper,8.,Silver,L.A , and Aines,R.D. (1983)The effects     ofquenching
Faile,S.P.,and Roy, D.M. (1966)Solubilities Ar, N,, CO' and He in
                                                  of                              rate and temperatureon the speciationofwater in silicateglasses  (abs.).
   glasses pressuresto l0 kbars. American Ceramic Society Journal,
             at                                                                    EOS.64. 339.
    49, 638-643.                                                               Watson, E.B., Sneeringer,     M.A., and Ross,A. (1982)Diffusion of dis-
Fine, G., and Stolper, E. (1985) The speciationofcarbon dioxide in so-             solved carbonate in magmas: Experimental results and applications'
   dium aluminosilicateglasses.     Contributions to Mineralogy and Petrol-        Earth and Planetary ScienceLetters, 61, 346-358
   o g y , 9 l ,1 0 5 - 1 2 1 .
-(1986)             Dissolved carbon dioxide in basaltic glasses:
   tions and speciation.Earth and PlanetaryScience     Letters,76,263-278.
Flowers,G.C. (1979)CorrectionofHolloway's (1977)adaptationofthe                                 Noveunen 25, 1985
                                                                              MrNuscnrrr REcEIvED
   modified Redlich-Kwong equation of statefor calculation of the fugac-      MeNuscnrsr AccEPtEDJvLv 2,1987