Contrib Mineral Petrol (2005) 149: 276287 DOI 10.

1007/s00410-004-0648-1

O R I GI N A L P A P E R

Antonio Acosta-Vigil David London George B. Morgan VI

Contrasting interactions of sodium and potassium with H2O in haplogranitic liquids and glasses at 200 MPa from hydrationdiffusion experiments
Received: 2 August 2004 / Accepted: 8 December 2004 / Published online: 3 February 2005 Springer-Verlag 2005

Abstract This study examines hydrationdiusion in the metaluminous haplogranite system at 200 MPa H2O and 800300C. At 800C hydration is accompanied by melting and uphill diusion of sodium from anhydrous glass toward the region of hydration and melting, whereas potassium diuses away from the hydration front and into anhydrous glass. Silicon and aluminum are simply diluted upon hydration. There is no change in molecular Al/(Na + K) throughout the entire hydration-diusion aureole and, therefore, (1) there is no loss of alkalis to the vapor, and (2) K migrates to replace Na in order to maintain local charge balance required by IVAl. Alkali diusion occurs over a viscosity contrast from 104.1 Pa s in hydrous liquid to %1011.81013.5 Pa s in anhydrous glass. From these results, we interpret that: (1) Na is structurally or energetically favored over K as a charge-balancing cation for IVAl in hydrous granitic liquids, whereas the opposite behavior has been observed for anhydrous melts, and (2) the diusion of alkalis through silicate melts is largely independent of viscosity. Results from 600C are similar to those at 800C, but hydration at 300C involves a loss of Na and concomitant increase in molar Al/(Na + K) in the hydration zone due to hydrogen-alkali exchange

between uid and glass. Hydration behavior at 400C is transitional between those at 300C and 600C, suggesting that the change in hydration mechanism occurs near the glass transition.

Introduction
The component H2O exerts profound eects on phase relationships, viscosity, density, and diusivities in silicate liquids (Tuttle and Bowen 1958; Lange 1994; Watson 1994). Though interactions of H2O with melt components in the haplogranite system (AbOrQtz) have been thoroughly investigated, the interpretation of results remains a matter of debate (Burnham 1975; Kohn et al. 1989; McMillan 1994; Kohn 2000; Schmidt et al. 2000; Zeng et al. 2000; Schmidt et al. 2001). In most models, the alkalis (Na and K) are treated as equivalent during the solvation of H2O in melt, though dierences in water solubility between albite- and orthoclase-rich compositions have been noted (423% relative, Holtz et al. 1992a; Behrens 1995; Holtz et al. 1995; Romano et al. 1996). Diusion experiments in silicate melts can provide valuable information about the interactions among chemical constituents because the concentration gradients that reveal these interactions occur in melt at magmatic temperatures (e.g. Wolf and London 1994; Acosta-Vigil et al. 2002). This is in contrast to spectroscopic measurements on glasses, for which the structure and associations of some components in melt may change upon quenching to glass, and may vary with rate of quench (e.g. the speciation of H2O: Dingwell and Webb 1990; Nowak and Behrens 1995; Romano et al. 1995; Shen and Keppler 1995; Zavelsky et al. 1998). In some instances, the information derived from diusion experiments in melts has found verication in spectroscopic studies of similar glasses. This is the case for the speciation of phosphorus in granitic melts, for which Wolf and London (1994) found evidence of an Al-P associated species that was also indicated by spectro-

Editorial Responsibility: T.L. Grove A. Acosta-Vigil D. London (&) G. B. Morgan VI School of Geology and Geophysics, University of Oklahoma, 100 East Boyd Street Room 810 SEC, Norman, OK 73019, USA E-mail: dlondon@ou.edu Tel.: +1-405-3253253 Fax: +1-405-3253140 Present address: A. Acosta-Vigil Dipartimento di Mineralogia e Petrologia, Consiglio Nazionale delle Ricerche, Instituto di Geoscienze e Georisorse, ` di Padova, Corso Garibaldi 37, Universita 35137, Padova, Italy

277

scopic work (Mysen et al. 1981, 1999; Gan and Hess 1992; Toplis and Schaller 1998; Cody et al. 2001; Mysen and Cody 2001). This study concentrates on hydration and H2O diffusion in cylinders of glass with a composition corresponding to the metaluminous haplogranite minimum at 200 MPa H2O. Most experiments were conducted at 800C with the goal of investigating the interaction between H2O and other components in granitic melt that help to delineate the systematics of diusion by which the chemical potential gradients in H2O are erased. Hydration-diusion experiments at 300, 400, and 600C were conducted with the aim of examining the hydration reaction as a function of temperature to elucidate differences in hydration and diusion mechanisms between glass and melt.

Materials and methods

Starting materials and experimental methods The starting material (Table 1) is a synthetic anhydrous haplogranite glass (Corning Lab Services, New York, USA) with the eutectic composition (Ab38.20Or27.94 Qtz33.60Crn0.25) (mineral symbols after Kretz 1983) of the haplogranite system at 200 MPa H2O (Tuttle and Bowen 1958). Glass cylinders measuring 3.5 mm in diameter and 46 mm in length were loaded into gold capsules (I.D. of $3.6 mm) with enough deionized-ultra ltered water ($10 wt%) to permit saturation of
Table 1 Electron microprobe analyses of starting and fully hydrated haplogranitic glasses Material Number of analyses Oxides (wt%) SiO2 Al2O3 CaO Na2O K2 O Total H2O by dierence Molar ratios ASI Na/K Al/Si Na/Si Norms (wt%) Ab Or Qtz Crn Starting glass 50a 77.95 (0.33) 12.78 (0.15) 0.02 (0.02) 4.51 (0.10) 4.73 (0.11) 99.99 (0.45) 1.016 1.45 0.19 0.11 38.20 (0.78) 27.94 (0.60) 33.60 (0.77) 0.25 (0.20) Hydrated (Acasi 279) glass 116b 73.06 (0.31) 11.78 (0.13) 0.01 (0.01) 4.12 (0.11) 4.43 (0.11) 93.40 (0.39) 6.60 1.016 1.41 0.19 0.11 37.35 (0.90) 28.02 (0.64) 34.38 (0.87) 0.25 (0.25)

the entire silicate mass. Capsules were sealed by DC argon plasma arc welding while keeping the capsule frozen to prevent volatilization of added water, placed overnight in an oven at $130C and then re-weighed to ensure no leakage. Experiments were conducted in subhorizontal NIMONIC 105 cold-seal pressure vessels with water as the pressure medium. Two sets of experiments were performed (Table 2): one in which glasses were taken to a target temperature and pressure of 800C and 200 MPa for dierence lengths of time, and a second set in which glasses were taken to temperatures of 300, 400, and 600C at the same pressure (200 MPa) for a single experimental time. Variations in the target temperature and pressure during the experiments were %3C and 1 MPa, respectively. Temperature was monitored with an internal Chromel-Alumel thermocouple, and pressure was monitored with a factory-calibrated Heise bourdon tube gauge; uncertainties in temperature and pressure are <10C and <10 MPa, respectively. The samples were rst pressurized cold, and then temperature was raised to the target value at a rate of $40C/min. Run times ranged from 0 h to 479 h (Table 2), with the 0-h experiment being quenched immediately upon reaching 800C. Experiments were quenched isobarically at a rate of $75C/min using a jet of air and water. After quench, capsules were weighed, punctured, placed in a desiccator overnight, and then weighed to check for the loss of free water. Products were mounted in Buehler Transoptic thermal plastic, ground to the center of the cylinders, and polished to a nal grit size of 0.3 lm for microprobe analysis. Analytical methods Starting and experimental glasses were analyzed with a Cameca SX-50 electron microprobe at the University of Oklahoma, using an accelerating voltage of 20 kV, a beam current of 2 nA, and a 20 lm spot. Matrix reduction used the PAP correction algorithm (Pouchou and Pichoir 1985). Sodium, potassium and aluminum were concurrently analyzed rst to minimize alkali volatilization and attendant changes in elemental ratios. Counting times were 30 s on peak for all elements, yielding calculated 3-r minimum detection limits of 0.02 wt% for Na2O, K2O and Al2O3, and $0.05 wt% for SiO2. Based on counting statistics, analytical uncertainties relative to their reported concentrations in glass are in the range of $0.51.0% for SiO2 and Al2O3, and $1.53.0% for Na2O and K2O. Using these methods, the maximum loss of sodium in water-saturated glass ($2.02.5% loss relative to the mean value, based on change in stoichiometry from the starting composition as determined by inductively coupled plasma) and grow-in of aluminum and silicon intensities during analysis are negligible and comparable to or less than the analytical uncertainties, hence no corrections to the data were performed (Morgan

Number in parentheses indicate 1-sigma standard deviations Mineral symbols are taken from Kretz (1983) a Transect along longitudinal centerline of glass cylinder, 100 microns spacing b Transect along longitudinal centerline of glass cylinder, 50 microns spacing

278 Table 2 Starting materials and durations of the conducted experiments Run no. Acasi Acasi Acasi Acasi Acasi Acasi Acasi Acasi Acasi Acasi
a

Starting materials H2O H2O H2O H2O H2O H2O H2O H2O H2O H2O + + + + + + + + + + Metaluminous Metaluminous Metaluminous Metaluminous Metaluminous Metaluminous Metaluminous Metaluminous Metaluminous Metaluminous haplogranite haplogranite haplogranite haplogranite haplogranite haplogranite haplogranite haplogranite haplogranite haplogranite glass glass glass glass glass glass glass glass glass glass core core core core core core core core core core

Temperature (C) 300 400 600 800 800 800 800 800 800 800

H2Oa (wt%) 10.54 17.71 17.58 10.11 10.64 9.94 10.47 10.80 11.62 10.62

Duration (min) 28,720 1,375 1,388 0 211 368 540 740 1,454 7,268

276 283 284 270 271 272 274 275 277 279

All experiments conducted at 200 MPa Proportion of water sealed inside the capsule with respect to the haplogranite glass

and London 1996). H2O concentrations in glass were calculated by dierence of the electron microprobe analysis totals from 100%. With the current analytical methods, accuracy for H2O by dierence is generally better than 10% relative for perfectly at and wellpolished samples with H2O concentrations in the range 210 wt% (Morgan and London 1996; Acosta-Vigil et al. 2003). The starting anhydrous glass often contains tiny (0.55 lm) air-lled vesicles that tend to disappear as hydration and melting proceed. Hence, small surface imperfections (pits) arising from these vesicles (that may be hidden by the carbon coat) account for most of statistical variation in nearly to completely anhydrous glasses (H2O 1.0 wt%) in the present experiments. The maximum uncertainty for the reported ASI values is 0.035 (ASI = molar Al2O3/[CaO + Na2O + K2O]), calculated by the propagation of errors. To ensure (1) statistical accuracy at critical points in the diusion prole, and (2) that diusion perpendicular to the cylinder axis did not interfere with diusion parallel to the cylinder axis, we conducted several analytical transverses throughout the experimental glasses: two transverses near the center and parallel to the cylinder axis, and one transverse at the middle and perpendicular to the cylinder axis.

Results
800C and variable experimental time Melting of the glass cores accompanied the inward diffusion of H2O. Concomitantly, Na diused uphill from anhydrous glass into and toward the hydrated liquid, whereas K diused uphill away from the H2O diusion front and into the anhydrous portion of the charges (Figs. 1, 2, 3). These results demonstrate that Na and K were coupled with H2O during diusion in the sense that their migration occurred in response to the inux of H2O. Note that in the 3-h experiment, for instance, the diusive exchange of Na and K occurred over 1.5 mm in the silicate cylinder, through the hydration front

(%1.0 mm) and %0.5 mm into the anhydrous glass (Figs. 1, 2). Throughout the entire charges at all experimental durations, the molar ratios Al/[Na + K] (ASI) and Al/ Si remained constant and equal to those in the starting glass (Fig. 1). When normalized to an anhydrous basis (Fig. 2), concentration proles for silica and alumina are at and similar to those in the starting glass within analytical uncertainty. In most of the central anhydrous region of the charges, the proles for sodium and potassium remain at and equal to the starting values but develop a minimum in Na and a maximum in K well above analytical uncertainties within 250500 microns of the hydration front(s), and then monotonically increase (Na) and decrease (K) from the hydration front(s) toward the hydrated edges of the glass cylinder (Fig. 2). These observations indicate that SiO2 and Al2O3 were simply diluted during the incorporation and migration of H2O, that there is no loss of alkalis to the vapor, and that the molar uxes of sodium and potassium were equivalent but opposite in direction to maintain the local charge balance required by IVAl. Based on the immobility of SiO2 during the diusion of H2O, we calculated the molar decrease in K (DK, equivalent to moles of Na gained) in the region of hydrated glass, and observed that the molar ratio DH/DK is approximately constant and equal to $100 to 200. These observations suggest that H2O diuses along a direction in the oxide composition space in which mol. H/Na $100200 and mol. Na=K. Between 12 h and 24 h the hydration fronts meet at the center of the charges (Figs. 1, 2), and after 24 h the direction of alkali migration reverses. Sodium ions spatially associated with water (in whatever form) near the edges of the core return toward the core interior as the H2O concentration gradientthe driving force for the segregation of Na and Kdecreases with the progressive hydration of the melt (Figs. 1, 2). Concomitant with the hydration of the granitic liquid to a uniform saturation value of $6.6 wt% H2O at 200 MPa (Table 1), alkali concentration gradients erase in about 48 h and do not change in experiments of up to 120 h duration.

279

Fig. 1 Composition (oxide concentrations and molar ratios) of the 800C experimental glasses along microprobe transverses parallel to the cylinder axis, as a function of experimental time and distance to the H2O-core interface. Each prole represents the mean values of two transverses near the center of the cylinder axis. H2O proles calculated by dierence of totals from 100%. In this and following gures, dashed vertical lines mark the position of the H2O diusion front, and error bars represent maximum analytical uncertainties of the analyses (analytical uncertainties are equal or smaller than the symbols when not shown)

300600C and single experimental time As a comparison to hydration-melting at 800C, we conducted three more hydration experiments at 300, 400, and 600C (Table 2). For a granitic melt of eutectic composition containing 6 wt% H2O, the glass transition temperature is calculated at $375C based on the viscosity of silicate melt at the glass transition from the data of Dingwell (1998). These three experiments, therefore, span this transition temperature. The results (Figs. 4, 5) at 600C are similar to those at 800C: Na diuses from the anhydrous glass toward the region of hydration, K migrates into the anhydrous glass in the opposite direction, and there is no net transfer of alkalis to the aqueous vapor (ASI remains constant). The results at 400C, close to the calculated glass transition temperature for this composition, begin to show a change in the alkali migration patterns. The

ASI remains essentially constant through the charge, so there is no noticeable loss of alkalis to vapor. The Na/K ratio increases in the direction of increasing hydration, as in the experiments at higher temperature, but alkali migration across the hydration front, where H2O content goes to zero, is not so evident, and the element proles mostly reect dilution of the glass components by H2O. The result at 300C reveals a change in the behavior of alkalis upon hydration: there is a slight loss of Na (Na/K falls) across the hydration aureole, and there is a corresponding increase in ASI. Thus, at 300C, hydration of the glass is accompanied by loss of Na to the vapor phase. Potassium, however, is enriched at the interface of the hydration front in the glass, as in all other experiments at higher temperatures.

Discussion
Mechanism of hydration: liquid versus glass An experimental H2O diusion study by Zhang et al. (1991) reported no alkali concentration gradients, but their methodology entailed diusion of H2O out of hydrous glasses at 1 atm pressure, which imposed substantially dierent conditions than those employed in this study. Perhaps the most important dierence is the

280

Fig. 2 Composition of the 800C experimental glasses along microprobe transverses parallel to the cylinder axis, normalized to an anhydrous basis. H2O prole added for reference. Dashed horizontal lines show the composition of the starting anhydrous glass

H2O concentration. Their glasses contained 1.8 wt% H2O, most of them <1.0 wt%, and the associated redistribution of alkalis would be negligible and would probably go unnoticed within the analytical scatter of the electron microprobe analyses. A single diusion prole from Karsten (1982) involving hydration of rhyolitic glass cores at 750C, 70 MPa showed the same alkali fractionation observed here. Karsten (1983), however, suggested that the penetration of H2O into glass/melt occurred in response to the exchange of alkalis (dominantly Na) from glass or melt with H+ or H3O+ from the aqueous vapor phase, because this hydration mechanism was well established from previous studies involving subsolidus hydration of rhyolitic glasses (e.g. Friedman and Smith 1958; Friedman et al. 1966; Jezek and Noble 1978). Our results at 800C, however, demonstrate that the ASI remains constant at the initial value for the glass across the entirety of all experimental charges; hence, there is no measurable loss of alkalis from the silicate mass. The alkali gradients observed in the 800C experiments, therefore, cannot be driven by ion exchange with aqueous uid. The experi-

ment at 600C shows results comparable to the 800C runs, and we suspect that the prole cited by Karsten (1982) also entails alkali migration in melt without extensive ion exchange with the vapor phase, though the data presented in her abstract do not permit a calculation of ASI to conrm this hypothesis. The experiment at 400C shows behavior intermediate between the high temperature and the low temperature hydration experiments, suggesting that a change in the mechanism of hydration occurs near the glass transition temperature. The experiment at 300C replicates the results of Friedman and Smith (1958), Friedman et al. (1966), and Jezek and Noble (1978), in which hydration of the glass is accommodated by alkali loss, particularly of Na, to the aqueous vapor. We conclude that the mechanism driving alkali migration by hydration-melting of granitic glass is dierent from hydration of glass below the glass transition temperature, and that the change in hydration mechanism occurs across the glass transition temperature. At 300C potassium still diuses uphill throughout the entire hydration volume (Figs. 3, 4) and, therefore, kinetics do not seem to be responsible for the change in hydration mechanism. We can only speculate that the change in hydration mechanism might be related to a change in the partitioning behavior of alkalis between granitic glass and aqueous uid with decreasing temperature, analogous to the pronounced isobaric increase in the partitioning of Na into saline aqueous vapor

281

Fig. 3 Composition of the 800C experimental glasses along microprobe transverses perpendicular to the cylinder axis near the middle of the cylinder, as a function of experimental time and distance to the interface

equilibrated with alkali feldspars from 450C to 350C (Lagache and Weisbrod 1977). Interactions between alkalis and other oxide components in hydrous granitic melt The interaction of H2O with silicate melt components has a long history of inquiry and contrasting interpretations. In a quasi-crystalline albite melt model, Burnham (1975) proposed a solvation mechanism in which the rst mole of dissolved H2O dissociated to form two OH groups, one of which served to break a bridging oxygen bond between Si-O-Si, and another that displaced Na from its charge-balancing position on IVAl to

an interstitial site. From spectroscopic measurements, Kohn et al. (1989, 1998; Kohn 2000) observed a strong interaction between Na and OH or H2O, but they found no evidence for disruption of AlOSi or SiOSi bonds (i.e. no AlOH or SiOH formed from bridging oxygen atoms). Ab initio models, however, led Sykes and Kubicki (1993; Kubicki and Sykes 2004) to conclude that TOT bond-breaking was probable. Other nuances of hydration models lie between these two extremes (e.g. see review by McMillan 1994; Schmidt et al. 2000), but the crux of the dierences concerns the extent to which solvation of H2O breaks bridging oxygen linkages in the TOT structure of melt. The potential signicance of dierent charge-balancing cations, Na+ or K+, which appear to be involved in the hydration process, has not been examined in the context of these melt structural studies, and only Schmidt et al. (2001) have presented recently some data (see below). The current experiments demonstrate that the smaller in CN=6: ShanNa+ (eective ionic radius of 1.16 A

282 Fig. 4 Oxide concentrations and molar ratios in the 300, 400, and 600C experimental glasses along microprobe transverses parallel to the cylinder axis, as a function of distance to the H2O-core interface. Each prole represents the mean values of two transverses near the center of the cylinder axis

non 1976) is preferred over K+ (eective ionic radius of in CN=9: Shannon 1976) in the environment 1.72 A created by the hydration melting of metaluminous granitic glass (NBO/T=0: Mysen 1988). In anhydrous aluminosilicate liquids, M+OAl bonds become more stable with decreasing charge density among Group IA cations (Shannon 1976; Roy and Navrotsky 1984). This behavior stems from the inability of smaller cations to eectively shield highly charged T cations like Si (e.g. Pauling 1960), and from the tendency of network-modifying cations with higher eld strength to destabilize melt or glass structures by drawing electron density away from T-O bonds (e.g. Shannon 1976; Roy and Navrotsky 1984). For instance, in a study of partitioning between immiscible anhydrous melts (Watson 1976), only Cs+, the Group IA cation of lowest eld strength, partitioned into the higher-silica melt. In our experiments, dilution of silica accompanies hydration-melting of the glass. It might be construed, therefore, that the

migration of K into the anhydrous portions of the hightemperature glasses reported here reects these tendencies in alkali ion stabilization, and that the entire pattern of diusion behavior observed here is driven by the movement of K rather than of Na. However, two lines of evidence indicate otherwise. (1) A parallelism can be establish between these and previous diusion experiments in which chemical gradients in Al in hydrous haplogranite melts induced the molar Na/Al ratio to adjust quickly to a constant value everywhere in the melt through the diusion of Na, whereas K diused concomitantly in such a way as to maintain or bring the ASI of melt to an equilibrium value (Acosta-Vigil et al 2002, 2004, and unpublished work in review). This constancy of the Na/Al ratio demonstrates a special structural association of Na and Al in hydrous alkali aluminosilicate melts. (2) As in the experiments reported here, Na also migrates toward higher concentrations of B, P, and F where gradients in these ligands exist in hydrous

283 Fig. 5 Composition of the 300, 400, and 600C experimental glasses along microprobe transverses parallel to the cylinder axis, normalized to an anhydrous basis. H2O prole added for reference

haplogranite melts, and liquid lines of descent produced by crystallization culminate in melts highly enriched in Na (over K) and all the above ligands (London 1999, 2004a, b; London et al. 2004). The liquid line of descent in uxed granitic melts clearly demonstrates that these liquids are stabilized by Na, whereas K becomes insoluble and precipitates as alkali feldspar and mica (London et al. 1989; 2004). Furthermore, the fractionation of Na and K observed here is consistent with the relationships found in bulk equilibrium studies on granitic compositions, in which the solubility of H2O is greater in Na-rich (albitic) than in K-rich (orthoclase-rich) liquids (Holtz et al. 1992a; Behrens 1995; Holtz et al. 1995; Romano et al. 1996). Together with other diusion and bulk equilibrium studies (Dingwell et al. 1984; Holtz et al. 1992b; Holtz et al. 1995; Linnen et al. 1996; Romano et al. 1996; Mungall et al. 1998; Acosta-Vigil et al. 2002, 2003, and unpublished work in review), these

results demonstrate a greater compatibility and interaction of H2O with Na2O and Al2O3 than with K2O and SiO2 in granitic liquids, although the underlying cause of these interactions remains uncertain. Regarding interaction between H2O and alkalis, NMR spectroscopy on albitic, orthoclase-rich, and haplogranitic glasses shows a progressive variation in the 23Na spectra with increasing water concentration (Kohn et al. 1989, 1998; Zeng et al. 2000; Schmidt et al. 2000, 2001). Kohn et al. (1989, 1998) interpreted the changes in the 23Na spectra as being due to the presence of more hydrated Na complexes, whereas Schmidt et al. (2001) suggested that the changes are consistent with the formation of a hydration shell around Na. Zeng et al. (2000) argued that the variations do not constitute clear evidence for the presence of hydrated Na complexes. None of these studies report any NMR spectra for potassium, although Schmidt et al. (2001)

284

found that spectra for 1H have no measurable dependence on composition (from albitic to orthoclase-rich to haplogranitic glasses). We suggest that the preference of IVAl for Na over K in hydrous granitic liquids expresses a steric eect that maximizes distances of separation, and hence minimizes repulsion, between neighboring M+ (H, Na, or K) cations by the incorporation of the smaller of the alkali ions when protons are added to the melt. This should apply whether or not H+ displaces Na+ or K+ from Al, though we cannot evaluate this substitution with our current experiments. Acosta-Vigil et al. (2002), however, proposed that H+ contributes to the charge balance on excess Al in hydrous peraluminous granitic melts; this would allow Al to remain four-coordinate (T) without the need to invoke tetrahedral clustering (Lacy 1963) or a coordination change to VIAl, for which there is little supporting evidence from spectroscopic measurements or numerical simulations (e.g. Stebbins et al. 2001; Kubicki and Toplis 2002). This is based on results from diusion and bulk equilibrium studies (Dingwell et al. 1984, 1997; Holtz et al. 1992b; Linnen et al. 1996; Behrens and Jantos 2001; Acosta-Vigil et al. 2002, 2003) which strongly suggest an interaction between excess aluminum and H2O in melt. In any case, the constancy of ASI across the entire glass prole in our diusion experiments above 400C reveals that if H displaces Na or K from coordination with Al upon glass/melt hydration, then the displaced alkali ion is conserved locally, i.e. in the immediate vicinity of the hydration reaction. If the dissolution of H2O involves the creation of a hydration shell of molecular water around alkali cations, as proposed by Schmidt et al. (2001), then the volumetric constraints on that hydration sphere may again favor the smaller alkali cation. Eective binary diusion coecient for total H2O The calculated concentration proles for H2O in haplogranitic glass are controlled by diusion in the glass/liquid rather than by convection in the liquid. This is indicated by three lines of evidence. (1) The advance of compositional proles into glass is proportional to the square root of time (Fig. 6); i.e. the concentration proles overlap when plotted against distance normalized to the square root of time (see Crank 1975). (2) The concentration proles obtained are regular and similar inward from both the ends and sides of the glass cylinders (compare Figs. 1, 2, 3 with Shaw 2000). (3) The high viscosity of the glass/liquid (see below) precludes convection on the time scale of these experiments. We have used the data from the 3-h to 12-h experiments to calculate an eective binary diusion coecient (EBDC, Cooper 1968) by the Boltzmann-Matano method (e.g. Crank 1975) for total H2O without dierentiating between OH groups and
Fig. 6 Al2O3, K2O and H2O concentrations versus distance to the H2O-core interface normalized to the square root of experimental time, in glasses of the 3-, 6-, 12-, and 24-h hydration experiments at 800C

molecular H2O (e.g. Stolper 1982). Corrections for the eects of the run-up step on the diusion proles (isobaric heating from 25C to 800C at 200 MPa in $27 min, 0-h run, Fig. 1) followed procedure (2) of Zhang and Behrens (2000). These methods yield EBDC $1.71011 to 2.61011 m2 s1 for melts with 4 wt% H2O, and EBDC $3.01011 to 4.81011 m2 s1 for melts with 5 wt% H2O. These EBDC compare well with previously determined H2O diusivities for granite melts under similar PTX conditions: 11010.4 to 11010.7 m2 s1 (Nowak and Behrens 1997) and 1.21011 to 1.91011 m2 s1 (Zhang and Behrens 2000).

285

Alkalis and diusion in the H2O-saturated haplogranite system Together with recent diusion studies, this paper shows that alkalis are important for the diusion of other melt components in the H2O-saturated haplogranite system at 800C and 200 MPa. With the exception of silica, alkali migration is clearly related to the movement of other oxide components, and sodium and potassium diuse at least as rapidly as the fastest-moving component in the system. In experiments involving the dissolution of corundum (Acosta-Vigil et al. 2002), the creation and diusion of a peraluminous component (diusivity of $11014 m2 s1) controlled the rates of alkali migration through the melt. In the present study, H2O is the rate-limiting component with a diusivity $21011 m2 s1, and hence the diusion of the alkalis occurs at this same rate. This value is essentially identical to the diusivities of the alkalis when their migration is driven by their own chemical potential gradients in otherwise homogeneous hydrous melts ($11011 m2 s1 : Freda and Baker 1998). Diusion of alkalis and viscosity The migration of Na and K along chemical potential gradients in these experiments is largely independent of liquid viscosity. In the 3-h experiment at 800C, the exchange of Na and K via diusion extended up to $1.5 mm into domains of the silicate core that remained anhydrous (Figs. 1, 2). The calculated viscosity (Schulze et al. 1995) of this nominally anhydrous composition (wt% H2O varied from 0.01 to 0.001) at 800C is 1011.8 to 1013.5 Pa s, which spans the structural transition of liquid to glass (Dingwell 1998). The alkali diusion proles are continuous up to the outer edge of the hydration aureole, where the liquid is saturated in H2O for this P and T, and for which the calculated viscosity (Schulze et al. 1995) is 104.1 Pa s. Mungall (2002) similarly concluded that the diusion of alkalis was largely independent of melt viscosity (at a specied temperature), though his interpretation arose from multiple data sets rather than single hydration-diusion experiments of the type employed in this study. Moreover, the same mechanisms and directions in the diusion of alkalis operate to comparatively low temperatures, at least to temperatures near the glass transition, and consequently across much higher gradients in viscosity.

librium studies deal with chemically homogeneous glass/ melt at any given experimental time. Diusion studies, on the other hand, produce melts with noticeable concentration gradients at a given time and, therefore, have the potential to magnify the eects of the interactions among components. Such behavior helps to explain the large fractionation in the Na/K molar ratio in melt due to varying H2O concentrations (45% relative, this study) as compared to the moderate increase in H2O concentration arising from changes in the Na/K ratio in melts at equilibrium (423% relative, Holtz et al. 1992a, 1995; Behrens 1995; Romano et al. 1996). This indicates the utility of diusion studies in deciphering the interactions among melt components that may be extensively erased once the melts achieve chemical homogeneity or are quenched to glass. The current hydration-diusion experiments extend our recent series of diusion studies (Acosta-Vigil et al. 2002, and unpublished work in review) that have revealed dierences in behavior between Na and K in hydrous granitic liquids, but consistencies of the same alkali ions among diering experimental programs. Both Na and K diuse in response to the creation of Al and H chemical gradients in hydrous haplogranite melts. Sodium diuses to maintain a constant Al/Na molar ratio throughout the entire melt reservoir when gradients in Al are present, or toward H-rich environments when gradients in H are present. Potassium diuses in the direction needed to bring the ASI of melt to the equilibrium value and to maintain local charge balance on IVAl (Acosta-Vigil et al. 2002; unpublished work in review; and this study). The diusion of alkalis can operate over the dimensions of the entire melt volume (i.e. the diusion caused by chemical potential gradients is independent of distance, Acosta-Vigil et al. 2002; London et al. 2004; Acosta-Vigil et al. unpublished work in review) and across very large viscosity gradients (this study). The diusion of Na and K along chemical potential gradients induced by the addition of other components (e.g. excess Al, H2O) or by removal of Na or K via the crystallization of feldspars represents another viable mechanism for the fractionation of Na and K in granitic systems (London 2004c; London et al. 2004).
Acknowledgements Support for this research was provided by National Science Foundation grants EAR-990165 and EAR-0124179. The Electron Microprobe Laboratory at the University of Oklahoma was created with US DOE grant DE-FG22-87FE1146 and upgraded with NSF grant EAR-9404658 and support from the University of Oklahoma Oce of Research Administration. Comments on the manuscript by James Kubicki and thoughtful reviews by Simon Kohn and Yang Liu improved the paper and are very much appreciated.

Concluding remarks
Diusion studies such as the present work represent an important tool for examining the interactions of components in silicate liquids. A basic dierence between diusion studies versus bulk equilibrium and spectroscopic studies is the degree of chemical homogeneity in the glass/melt products. Spectroscopic and bulk equi-

References
Acosta-Vigil A, London D, Dewers TA, Morgan GB VI (2002) Dissolution of corundum and andalusite in H2O-saturated haplogranitic melts at 800C and 200 MPa: constraints on diusivities and the generation of peraluminous melts. J Petrol 43:18851908

286 Acosta-Vigil A, London D, Morgan GB VI, Dewers TA (2003) Solubility of excess alumina in hydrous granitic melts in equilibrium with peraluminous minerals at 700800C and 200 MPa, and applications of the aluminum saturation index. Contrib Mineral Petrol 146:100119 Acosta-Vigil A, London D, Morgan GB VI (2004) Composition and distribution of melt during partial melting of a leucogranite at 200 MPa H2O. Geol Soc Am Abstr Prog 36:45 Behrens H (1995) Determination of water solubilities in high-viscosity melts: an experimental study on NaAlSi3O8 and KAlSi3O8 melts. Eur J Mineral 7:905920 Behrens H, Jantos N (2001) The eect of anhydrous composition on water solubility in granitic melts. Am Mineral 86:1420 Burnham CW (1975) Water and magmas; a mixing model. Geochim Cosmochim Acta 39:10771084 Cody GD, Mysen BO, Saghi-Szabo G, Tossell JA (2001) Silicatephosphate interactions in silicate glasses and melts; I. A multinuclear (27Al, 29Si, 31P) MAS NMR and ab initio chemical shielding (31P) study of phosphorous speciation in silicate glasses. Geochim Cosmochim Acta 65:23952411 Cooper AR Jr (1968) The use and limitations of the concept of an eective binary diusion coecient for multicomponent diusion. In: Wachtman JB Jr, Franklin AD (eds) Mass transport in oxides. National Bureau of Standards, Special Publication 196:7984 Crank J (1975) The mathematics of diusion. Clarendon Press, Oxford Dingwell DB (1998) The glass transition in hydrous granitic melts. Phys Earth Planet Int 107:18 Dingwell DB, Webb SL (1990) Relaxation in silicate melts. Eur J Mineral 2:427449 Dingwell DB, Harris DM, Scarfe CM (1984) The solubility of H2O in melts in the system SiO2-Al2O3-Na2O-K2O at 1 to 2 kbars. J Geol 92:387395 Dingwell DB, Holtz F, Behrens H (1997) The solubility of H2O in peralkaline and peraluminous granitic melts. Am Mineral 82:434437 Freda C, Baker DR (1998) Na-K interdiusion in alkali feldspar melts. Geochim Cosmochim Acta 62:29973007 Friedman I, Smith RL (1958) The deuterium content of water in some volcanic glasses. Geochim Cosmochim Acta 15:218228 Friedman I, Smith RL, Long WD (1966) Hydration of natural glass and formation of perlite. Geol Soc Amer Bull 77:323328 Gan H, Hess PC (1992) Phosphate speciation in potassium aluminosilicate glasses. Am Mineral 77:495506 Holtz F, Behrens H, Dingwell DB, Taylor RP (1992a) Water solubility in aluminosilicate melts of haplogranite composition at 2 kbar. Chem Geol 96:289302 Holtz F, Johannes W, Pichavant M (1992b) Peraluminous granites: the eect of alumina on melt composition and coexisting minerals. Trans R Soc Edinb: Earth Sci 83:409416 Holtz F, Behrens H, Dingwell DB, Johannes W (1995) H2O solubility in haplogranitic melts: compositional, pressure, and temperature dependence. Am Mineral 80:94108 Jezek PA, Noble DC (1978) Natural hydration and ion exchange of obsidian: an electron microprobe study. Am Mineral 63:266 273 Karsten JL (1982) Potassium and sodium mobility during water diusion in obsidian melts. Eos Trans Am Geophys Union 63:1137 Karsten JL (1983) Mechanisms of water diusion in silicate glasses and melts. Eos Trans Am Geophys Union 64:339 Kohn SC (2000) The dissolution mechanisms of water in silicate melts; a synthesis of recent data. Mineral Mag 64: 389408 Kohn SC, Dupree R, Smith ME (1989) A multinuclear magnetic resonance study of the structure of hydrous albite glasses. Geochim Cosmochim Acta 53:29252935 Kohn SC, Smith ME, Dirken PJ, van Eck ERH, Kentgens APM, Dupree R (1998) Sodium environments in dry and hydrous albite glasses: improved 1H solid-state NMR data and their implications for water dissolution mechanisms. Geochim Cosmochim Acta 62:7987 Kretz R (1983) Symbols for rock-forming minerals. Am Mineral 68:277279 Kubicki JD, Sykes DG (2004) Ab initio calculation of 1H, 17O, 27 Al, and 29Si NMR parameters, vibrational frequencies and bonding energetics in hydrous silica and Na-aluminosilicate glasses. Geochim Cosmochim Acta 68:39093918 Kubicki JD, Toplis MJ (2002) Molecular orbital calculations on aluminosilicate tricluster molecules; implications for the structure of aluminosilicate glasses. Am Mineral 87:668678 Lacy ED (1963) Aluminum in glasses and melts. Phys Chem Glasses 4:234238 Lagache M, Weisbrod A (1977) The system: two alkali feldspars KCl-NaCl-H2O at moderate to high temperatures and low pressures. Contrib Mineral Petrol 62:77101 Lange RA (1994) The eect of H2O, CO2, and F on the density and viscosity of silicate melts. In: Carroll MR, Holloway JR (eds) Volatiles in magmas. Min Soc Am Rev Mineral 30:331369 Linnen RL, Pichavant M, Holtz F (1996) The combined eects of fO2 and melt composition on SnO2 solubility and tin diusivity in haplogranitic melts. Geochim Cosmochim Acta 60:4965 4976 London D (1999) Melt boundary layers and the growth of pegmatitic textures. Can Mineral 37:826827 London D (2004a) Geochemistry of alkalis and alkaline earths in ore-forming granites, pegmatites, and rhyolites. In: Linnen RL, Sampson I (eds) Rare-element geochemistry of ore deposits. Min Assoc Can Short Course Handbook (in press) London D (2004b) Granitic pegmatites: an assessment of current concepts and directions for the future. In: Ra mo OT, Kosunen PJ, Lauri LS, Karhu JA (eds) Granitic systemsstate of the art and future avenues, an issue honoring the retirement of Prof. Ilmari Haapala, University of Helsinki. Lithos (in press) London D (2004c) Vertical zonation of alkalis in mineralized granites and pegmatites. Geol Min Assoc Can Abstr 29:197 London D, Morgan GB, Herving RL (1989) Vapor-undersaturated experiments with Macusani glass + H20 at 200 MPa, and the internal dierentiation of granitic pegmatites. Contrib Mineral Petrol 102:117 London D, Morgan GB VI, Acosta-Vigil A. (2004) Alkali fractionation and feldspar zonation in granitic pegmatites. Geol Soc Am Abstr Prog 36:46 McMillan PF (1994) Water solubility and speciation models. In: Carroll MR, Holloway JR (eds) Volatiles in magmas. Min Soc Am Rev Mineral 30:131156 Morgan GB VI, London D (1996) Optimizing the electron microprobe analysis of hydrous alkali aluminosilicate glasses. Am Mineral 81:11761185 Mungall JE (2002) Empirical models relating viscosity and trace diusion in magmatic silicate liquids. Geochim Cosmchim Acta 66:125143 Mungall JE, Romano C, Dingwell DB (1998) Multicomponent diusion in the molten system K2O-Na2O-Al2O3-SiO2-H2O. Am Mineral 83:685699 Mysen BO (1988) Structure and properties of silicate melts. Elsevier, Amsterdam Mysen BO, Cody GD (2001) Silicate-phosphate interactions in silicate glasses and melts; II. Quantitative, high-temperature structure of P-bearing alkali aluminosilicate melts. Geochim Cosmochim Acta 65:24132431 Mysen BO, Ryerson FJ, Virgo D (1981) The structural role of phosphorus in silicate melts. Am Mineral 66:106117 Mysen BO, Holtz F, Pichavant M, Beny JM, Montel JM (1999) The eect of temperature and bulk composition on the solution mechanism of phosphorus in peraluminous haplogranitic magmas. Am Mineral 84:13361345 Nowak M, Behrens H (1995) The speciation of water in haplogranitic glasses and melts determined by in situ near-infrared spectroscopy. Geochim Cosmochim Acta 59:34453450 Nowak M, Behrens H (1997) An experimental investigation on diusion of water in haplogranitic melts. Contrib Mineral Petrol 126:365376

287 Pauling L (1960) The nature of the chemical bond, 3rd edn. Cornell University Press, Ithaca Pouchou JL, Pichoir F (1985) PAP /(qz) correction procedure for improved quantitative microanalysis. In: Armstrong JT (ed) Microbeam Analysis. San Francisco Press, San Francisco, pp 104106 Romano C, Dingwell DB, Behrens H (1995) The temperature dependence of the speciation of water in NaAlSi3O8-KAlSi3O8 melts: an application of ctive temperatures derived from synthetic uid inclusions. Contrib Mineral Petrol 122:1 10 Romano C, Dingwell DB, Behrens H, Dol D (1996) Compositional dependence of H2O solubility along the joins NaAlSi3O8KAlSi3O8, NaAlSi3O8-LiAlSi3O8, and KAlSi3O8-LiAlSi3O8. Am Mineral 81:452461 Roy BN, Navrotsky A (1984) Thermochemistry of charge-coupled substitutions in silicate glasses: The systems M1/n n+ -Al2O3SiO2 (M=Li, Na, K, Rb, Cs, Mg, Ca, Sr, Ba, Pb). J Am Ceramic Soc 89:606610 Schmidt BC, Riemer T, Kohn SC, Behrens H, Dupree R (2000) Dierent water solubility mechanisms in hydrous glasses along the Qz-Ab join: evidence from NMR spectroscopy. Geochim Cosmochim Acta 64:513526 Schmidt BC, Riemer T, Kohn SC, Holtz F, Dupree R (2001) Structural implications of water dissolution in haplogranitic glasses from NMR spectroscopy: inuence of total water content and mixed alkali eect. Geochim Cosmochim Acta 65:29492964 Schulze F, Behrens H, Holtz F, Roux J, Johannes W (1995) Inuence of water on the viscosity of a haplogranitic melt. Am Mineral 81:11551165 Shannon RD (1976) Systematic studies of interatomic distances in oxides. In: Strens RGJ (ed) The physics and chemistry of minerals and rocks. Wiley, London, pp 403431 Shaw CSJ (2000) The eect of experiment geometry on the mechanism and rate of dissolution of quartz in basanite at 0.5 GPa and 1350C. Contrib Mineral Petrol 139:509525 Shen A, Keppler H (1995) Infrared spectroscopy of hydrous silicate melts to 1000C and 10 kbar: direct observation of H2O speciation in a diamond-anvil cell. Am Mineral 80:13351338 Stebbins JF, Oglesby JV, Kroeker S (2001) Oxygen triclusters in crystalline CaAl4O7 (grossite) and in calcium aluminosilicate glasses: 17O NMR. Am Mineral 86:13071311 Stolper E (1982) The speciation of water in silicate melts. Geochim Cosmochim Acta 46:26092620 Sykes D, Kubicki JD (1993) A model for H2O solubility mechanism in albite melts from infrared spectroscopy and molecular orbital dynamics. Geochim Cosmochim Acta 57:10391052 Toplis MJ, Schaller T (1998) A 31P MAS NMR study of glasses in the system xNa2O - (1-x)Al2O3- 2SiO2- yP2O5. J Non Crystalline Solids 224:5768 Tuttle OF, Bowen NL (1958) Origin of granite in the light of experimental studies in the system NaAlSi3O8-KAlSi3O8-SiO2H2O. Geological Society of America Memoir 74, 153 pp Watson EB (1976) Two-liquid partition coecients: experimental data and geochemical implications. Contrib Mineral Petrol 56:119134 Watson EB (1994) Diusion in volatile-bearing magmas. In: Carroll MR, Holloway JR (eds) Volatiles in magmas. Min Soc Am Rev Mineral 30:371411 Wolf MB, London D (1994) Apatite dissolution into peraluminous haplogranite melts: an experimental study of solubilities and mechanisms. Geochim Cosmochim Acta 58:41274145 Zavelsky VO, Bezmen NI, Zharikov VA (1998) Water in albite glasses: OH-groups, isolated molecules, and clusters. J Non Crystalline Solids 224:225231 Zeng Q, Nekvasil H, Grey CP (2000) In support of a depolymerization model for water in sodium aluminosilicate glasses: information from NMR spectroscopy. Geochim Cosmochim Acta 64:883896 Zhang Y, Behrens H (2000) H2O diusion in rhyolitic melts and glasses. Chem Geol 169:243262 Zhang Y, Stolper EM, Wassenburg GJ (1991) Diusion of water in rhyolitic glasses. Geochim Cosmochim Acta 55:441456