Springer-Verlag 1997 Contrib Mineral Petrol (1997) 126: 377 – 385 Harald Behrens · Marcus Nowak The mechanisms of water diffusion in polymerized silicate melts Received: 26 March 1996 / Accepted: 23 August 1996 Abstract Diffusion of water was experimentally investigated for melts of albitic (Ab) and quartz-orthoclasic (Qz29Or71, in wt %) compositions with water contents in the range of 0 to 8.5 wt % at temperatures of 1100 to 1200 °C and at pressures of 1.0 and 5.0 kbar. Apparent chemical diffusion coefficients of water (Dwater) were determined from concentration-distance profiles measured by FTIR microspectroscopy. Under the same P-T condition and water content the diffusivity of water in albitic, quartz-orthoclasic and haplogranitic (Qz28Ab38 Or34, Nowak and Behrens, this issue) melts is identical within experimental error. Comparison to data published in literature indicates that anhydrous composition only has little influence on the mobility of water in polymerized melts but that the degree of polymerization has a large effect. For instance, Dwater is almost identical for haplogranitic and rhyolitic melts with 0.5–3.5 wt % water at 850 °C but it is two orders of magnitude higher in basaltic than in haplogranitic melts with 0.2–0.5 wt % water at 1300 °C. Based on the new water diffusivity data, recently published in situ near-infrared spectroscopic data (Nowak 1995; Nowak and Behrens 1995), and viscosity data (Schulze et al. 1996) for hydrous haplogranitic melts current models for water diffusion in silicate melts are critically reviewed. The NIR spectroscopy has indicated isolated OH groups, pairs of OH groups and H2O molecules as hydrous species in polymerized silicate melts. A significant contribution of isolated OH groups to the transport of water is excluded for water contents above 10 ppm by comparison of viscosity and water diffusion data and by inspection of concentration profiles from trace water diffusion. Spec- H. Behrens (&) · M. Nowak1 Institut für Mineralogie, Universität Hannover, Welfengarten 1, D-30167 Hannover, Germany 1 Present address: Bayerisches Geoinstitut, Universität Bayreuth, D-95440 Bayreuth, Germany Editorial responsibility: J. Hoefs troscopic measurements have indicated that the interconversion of H2O molecules and OH pairs is relatively fast in silicate glasses and melts even at low temperature and it is inferred that this reaction is an active step for migration of water. However, direct jumps of H2O molecules from one cavity within the silicate network to another one can not be excluded. Thus, we favour a model in which water migrates by the interconversion reaction and, possibly, small sequences of direct jumps of H2O molecules. In this model, immobilization of water results from dissociation of the OH pairs. Assuming that the frequency of the interconversion reaction is faster than that of diffusive jumps, OH pairs and water molecules can be treated as a single diffusing species having an effective diffusion coefficient DOHpair;H2 O . The shape of curves of Dwater versus water content implies that DOHpair;H2 O increases with water content. The change from linear to exponential dependence of Dwater between 2 and 3 wt % water is attributed to the influence of the dissociation reaction at low water content and to the modification of the melt structure by incorporation of OH groups. Introduction Diffusion of water in silicate glasses and melts becomes increasingly important in geosciences, e.g. for understanding and modelling of processes during volcanic eruptions. In the past twenty years several experimental and theoretical attempts were made to quantify the transport of water and to clarify the mechanisms of water diffusion. A comprehensive review of the experimental techniques and basic results is given by Watson (1994), and models for diffusion of water are summarized by Zhang et al. (1991a). Zhang et al. (1991a) have demonstrated that combination of water speciation data and water diffusion data can give new insights to the mechanisms of water diffusion. As shown recently, water species of the melt are not quenchable and, thus, its concentrations cannot be 378 determined by measurements on glasses at room temperature but only by direct measurements on the melt at high temperature and pressure (Nowak 1995; Nowak and Behrens 1995; Shen and Keppler 1995). For the haplogranite composition used to study water diffusion (Nowak and Behrens, this issue) quantitative measurements of water species in the melt are available from in situ near-infrared spectroscopy at temperatures up to 800 °C and pressure up to 3 kbar (Nowak 1995; Nowak and Behrens 1995). Furthermore, the influence of water on melt viscosity was investigated for the same haplogranitic composition by Schulze et al. (1996). Thus, in the present paper we firstly can combine water diffusivity data, water speciation data and viscosity data determined for the same melt composition to elucidate the mechanisms of water diffusion. The possible application of the results to other melt compositions (e.g. rhyolitic) is tested by studying the effect of anhydrous composition on water diffusivity in the haplogranite system and by comparison to literature data. Models for water diffusion in polymerized melts are critically reviewed on the basis of the new data and a preferred model for water diffusion is presented. Experimental methods Experiments were performed using the diffusion couple technique as described in detail by Nowak and Behrens (this issue). Starting materials were dry and hydrous glasses of Ab (albite) and Qz(quartz)29Or(orthoclase)71 (in wt %) anhydrous compositions. Dry glasses were synthesized by Schott company in Mainz, Germany (Ab) and by F. Holtz in Orléans, France (Qz29Or71). Electron microprobe analyses are given in Table 1. The same Ab composition [denoted as glass (I) in previous studies] was used for investigation of water solubility in melts (Behrens 1995) and of water speciation in glasses (Behrens et al. 1996). Hydrous glasses were prepared in internally heated pressure vessels (IHPV) using Ar as pressure medium. In the diffusion experiments two cylinders with different water contents were encapsulated in a platinum tube with the contact area marked by some grains of platinum powder. Runs were performed in a horizontally working IHPV. After the runs NIR (near-infrared) spectra were measured along the diffusion direction of water using an IR microscope. Concentration-distance profiles were calculated from the spectra using for Ab the linear molar extinction coefficients and the density/water content relation published by Behrens et al. (1996). For Qz29Or71 the values were determined experimentally using five hydrous glasses with known water content ( = Table 1 Electron microprobe analyses of dry glasses normalized to 100 wt % Na2O K2O Al2O3 SiO2 Ab Qz29Or71 11.96 – 18.83 69.19 – 17.76 19.23 63.02 1.69 l · mol)1 · cm)1 for the band at 4500 cm)1 due to OH groups and = 1.89 l · mol)1 · cm)1 for the band at 5200 cm)1 due to H2O molecules). Densities of hydrous Qz29Or71 glasses were calculated with a precision better than 2 % relative using the density equation for Qz28Ab38Or34 glasses given by Nowak and Behrens (this issue). Effective diffusion coefficients of water (Dwater) were determined from total water versus distance profiles using a modified Boltzmann-Matano analysis. The maximum uncertainty of the obtained diffusion coefficients is estimated to be 25 % relative for both Ab and Qz29Or71 samples. Results Experimental parameters and diffusion coefficients for selected water contents are given in Table 2. Because of the small temperature range in which experiments could be performed without crystallization no attempt was made to determine the temperature dependence and, hence, the activation energy of water diffusion for these melts. The Dwater determined for albitic and quartz-orthoclasic compositions agree within experimental error with corresponding values for haplogranitic compositions (Fig. 1). An influence of the type of alkali, Na or K, on water diffusion was not found. As observed for haplogranitic compositions, the dependence of Dwater on total water content (Cwater) changes from linear to exponential between 2 and 3 wt % water. Experiments with Ab melts at 1 and 5 kbar show the same trend as observed for haplogranitic melts. At higher pressure the diffusivity of water seems to be slightly lower than at low pressure at least for high water contents (Fig. 2). However, the small pressure range and the number of experiments can not give unequivocal proof of a pressure effect. Table 2 Parameters and results of water diffusion experiments with Ab and Qz29Or71 melts. (Cwater water contents of both parts of the starting diffusion couple, tn nominal run duration, te effective run duration considering heating and cooling time, MI Matano-interface) No. Cwater [wt %] T [ °C] P tn [kbar] [min] te [min] Shift of Dwater [10–7cm2/s] MI [mm] 0.5 wt % 1 wt % 1.5 wt % 2 wt % 3 wt % 4 wt % 5 wt % 6 wt % AbD07 AbD06 AbD05 AbD10 AbD08 8.43/1.28 4.02/0.05 4.16/0.02 6.28/0.04 6.25/0.03 1100 1150 1150 1150 1200 5.0 1.0 5.0 5.0 5.0 30 30 30 60 30 33.7 33.7 33.7 63.7 33.7 0.156 0.026 0.081 0.172 0.168 – 0.57 0.37 0.76 0.61 – 1.19 0.61 1.12 1.07 – 1.74 0.89 1.60 1.46 0.95 2.47 1.26 2.01 1.88 1.57 4.56 3.05 3.36 3.00 2.38 – – 5.82 5.22 3.73 – – 12.63 11.20 6.43 – – – – Qz29Or71D01 6.89/0.26 1200 5.0 60 66.9 0.214 0.73 1.31 1.75 2.20 3.31 5.39 9.24 17.52 379 Fig. 1 Comparison of water diffusivity in haplogranitic (Qz28 Ab38Or34, labelled AOQ), albitic (Ab) and quartz-orthoclasic (Qz71 Or29) melts. Data for AOQ are from Nowak and Behrens (this issue). Calculations for AOQ use Eq. (9) of Nowak and Behrens (this issue) Fig. 2 Effect of pressure on water diffusivity in Ab melts The effect of anhydrous composition on water diffusion The experimental results on Ab, Qz29Or71 and Qz28Ab38 Or34 compositions suggest that water diffusivity in haplogranitic melts is independent of anhydrous composition at least for feldspar-rich compositions. Furthermore, the good agreement of data for haplogranitic (Nowak and Behrens, this issue) and for rhyolitic melts (Shaw 1974; Karsten et al. 1982; Lapham et al. 1984) implies that the results for the haplogranite system can directly be applied to other polymerized silicate melt systems (Fig. 3). Differences of the shape of Dwater versus. Cwater of the various data sets are attributed to the Fig. 3 Comparison of water diffusion data for haplogranitic and rhyolitic melts. The water diffusivity in rhyolitic melt with low water content given by Jambon et al. (1992) is related to the average of initial and equilibrium water content of the sample used in the dehydration experiments. All other data refer to the water contents for which the diffusion coefficients are calculated. Preliminary data from hydration of obsidian by Jambon (1979) are in good agreement to the data of Shaw (1974) but are not included in the plot because they can not be related directly to the water content. Data for AOQ are from Nowak and Behrens (this issue) precision of the analytical techniques and to the evaluation procedure. It is worth noting that the curve of Shaw (1974) was not based on concentration profiling but on time dependent sorption of water in cylinders. Nevertheless, this curve gives a good approximation to the data obtained from diffusion profiles. Reinvestigation of the diffusion samples of Shaw (1974) by ion probe profiling and iterative determination of diffusion coefficients by Delaney and Karsten (1981) basically confirm the results of Shaw (1974). The only significant difference is that Shaw (1974) chose a linear function and Delaney and Karsten (1981) an exponential one to describe the dependence of diffusivity on water content. A direct comparison to results of Zhang et al. (1991a) for rhyolitic glasses is not possible because these authors evaluated individual diffusion coefficients of hydrous species (OH and H2O) from profiles produced by dehydration experiments below the glass transition temperature. We have determined graphically chemical diffusion coefficients of water from a concentration-distance profile (Run KS-D3, their Fig. 3) given by Zhang et al. (1991a). The obtained value of Dwater is for 0.5 wt % water 0.7 log units below the value calculated for haplogranitic composition using Eq. (7) of Nowak and Behrens (this issue) (Fig. 4). As shown by Cockram et al. (1969) and Haider and Roberts (1970) for several silicate and aluminosilicate glasses with low water contents the activation energy for water diffusion typically is higher below than above the glass transition (note the change in 380 mechanisms do not fundamentally change from polymerized (e.g. rhyolitic) to weakly depolymerized (e.g. andesitic) melt structures. However, in strongly depolymerized (e.g. basaltic) melts the diffusivity of water is considerably higher than in polymerized melts and a different diffusion mechanism seems to be operative. In basaltic melts with 0.2 wt % water Dwater is two orders of magnitude higher at 1300–1500 °C and 10 kbar (Zhang and Stolper 1991) than extrapolated values for haplogranitic melts (Nowak and Behrens, this issue). Furthermore, it is worth noting that trace water diffusion in simple silicate melts also strongly increases with depolymerization of the melt (e.g. Scholze and Mulfinger 1959; Moulson and Roberts 1961; Haider and Roberts 1970). For instance, at 1200–1300 °C the diffusivity is three orders of magnitude higher for Na2Si3O7 melts (Scholze and Mulfinger 1959) than for SiO2 supercooled melts (Moulson and Roberts 1961) (see Fig. 4). Fig. 4 Water diffusion in selected melt compositions with low water contents. Wt % in the plot are estimated averages of initial and equilibrium water contents for Na2O · 3SiO2 (Scholze and Mulfinger 1959), SiO2 (Moulson and Roberts 1961), Ab (Haider and Roberts 1970) and rhyolite A (Jambon et al. 1992). For other compositions (basalt, Zhang and Stolper 1991; haplogranite, Nowak and Behrens, this issue; rhyolite B, recalculated from Zhang et al. 1991a) wt % refer to the water content for which the diffusion coefficients are determined. Data for the haplogranite composition (AOQ) with 0.2 and 0.02 wt % water were extrapolated from experimental data using the linear relationship of diffusivity on water content. Note that diffusivities are higher in depolymerized melts (Na2O · 3SiO2, basalt) than in polymerized melts (haplogranite, rhyolite, SiO2, Ab) for given water contents slope for Ab at 690 °C in Fig. 4). Therefore, the deviation is attributed mainly to the extrapolation of our data to low temperature, because Zhang et al. (1991a) performed the experiments below the glass transition. For comparison to trace water diffusion (the term ‘‘trace water diffusion’’ is used for water contents up to several hundred ppm by weight which are typically for sorption and desorption experiments with glasses and melts at ambient pressure) in other polymerized melts we have linearly extrapolated the data of haplogranitic melts down to very low water contents of 0.02 wt %. The extrapolated values for the haplogranitic composition are in good agreement with experimental data for rhyolite (Jambon et al. 1992, revised data from Jambon 1979), Ab (Haider and Roberts 1970) and SiO2 (Moulson and Roberts 1961) considering that only average diffusion coefficients were determined in the latter studies (Fig. 4). The insensibility of Dwater to anhydrous composition implies that the bond strength of tetrahedral cations (Al or Si) to oxygen and the distribution of sizes of tetrahedral rings in the melt structure have little influence on water diffusion and that the diffusion mechanisms are the same for various polymerized melts. Stanton et al. (1988) noted that the diffusivity of water is similar in melts ranging from rhyolitic to andesitic compositions. This implies that the diffusion Mechanisms of water diffusion in polymerized melts At least five different diffusion mechanisms may contribute to the transport of water in silicate melts: (1) coupling of fluxes of H+ and OH); (2) parallel diffusion of OH) and alkali cations or interdiffusion of H+ (or H3O+) and alkali cations; (3) diffusion of OH groups bound to tetrahedral cations; (4) direct jumps of water molecules from one site into a neighbouring site without interaction with the silicate framework; (5) reaction of water molecules with bridging oxygens by formation of OH group pairs. After recombination water molecules may be expelled on different sites adjacent to the reformed oxygen bridge. These five possible mechanisms are discussed in the following sections. 1. Mechanisms involving diffusion of H+ and OH) Coupled motion of H+ and OH) has been inferred as a predominant mechanism for diffusion of water in silicate glasses and aluminosilicate glasses with alkali/Al » 1 (Scholze and Mulfinger 1959; Haider and Roberts 1970). The depolymerized glasses have high concentrations of singly bonded oxygens which can react with diffusing protons to form OH groups. This leads to a local charge imbalance which facilitates the attack of a diffusing OH) on one of the associated (Al, Si)-O-(Al, Si) bonds (Haider and Roberts 1970). Because the transport of the chemical component water always is determined by the slowest step of the diffusion and reaction process this mechanism implies that diffusion is controlled by movement of OH) or breaking of (Al, Si)-O-(Al, Si) bonds whereas protons can migrate rapidly within the silicate structure. Based on high temperature IR spectroscopy Keppler and Bagdassarov (1993) suggested a highly mobile type of proton at 1300 °C in a rhyolitic melt containing about 381 300 ppm water. For haplogranitc melts at 800 °C we found identical diffusion coefficients of protons (measured by H/D interdiffusion) and water (measured by chemical diffusion) (Nowak and Behrens, this issue). This implies that self diffusion of protons in haplogranitic melts is limited by the diffusion of water under these conditions. Thus, if protons really have high mobility their movement must locally be restricted or in other words rapid jumps of the protons occur only within cavities of the structure and not between cavities. It is emphasized that the field strength of a bare proton prohibits it existing in a condensed phase (Ernsberger 1980) and, therefore, a rapid motion of protons only is expected in the presence of singly bonded oxygens. Consequently, it is inferred that a mechanism involving H+ and OH) is less important for polymerized (e.g. rhyolitic) melts but could be more important for basaltic melts. 2. Mechanisms involving alkali diffusion Interdiffusion of alkali and protons or H3O+ is known to be a fundamental process during reaction of water with glasses (see, e.g. Doremus 1979; Scholze 1988, 1989). This process as well as the coupled motion of alkali and OH) should produce gradients of alkali along the diffusion profile. Such gradients are not observed for haplogranitic (Nowak and Behrens, this issue) and rhyolitic (e.g. Lapham et al. 1984; Zhang et al. 1991a) melts and thus we conclude that these mechanisms do not significantly contribute to the transport of water in polymerized silicate melts under the conditions studied. Mechanisms involving alkali diffusion may be important for magma/fluid interaction if alkalis are dissolved in the fluid or for magma mixing if water diffusion is coupled to chemical exchange fluxes. 3. Mechanisms involving OH diffusion Experimental data on the diffusivity of OH groups bonded to tetrahedral cations (DOH) are not available. As a raw approximation we have estimated DOH from viscosity measurements using the Eyring equation: D kT gk 1 where k is the Boltzmann constant, T the absolute temperature, g the viscosity and k the diffusive jump distance, Fig. 5 Scheme of diffusion of OH groups in silica glass for which the average distance of the tetrahedral cations (3.6 Å) in haplogranitic melts is used. The approximate validity of the Eyring equation has been demonstrated for oxygen diffusion in a polymerized (jadeite) melt (Shimizu and Kushiro 1984). Watson and Baker (1991) noted for several melt compositions that the Eyring equation can predict the diffusivities of non-alkalies during interdiffusion in the range 10)11–10)7 cm2/s within a factor of 3. The application of the Eyring equation is probably limited to polymerized melts, because data for more depolymerized melts such as diopside are not well reproduced by Eq. (1) (Shimizu and Kushiro 1984). This implies that the breaking of Si-O and Al-O bonds is a fundamental feature of viscous flow, and diffusion of oxygen and network formers in polymerized melts as noted by Rubie et al. (1993). Diffusion of isolated OH groups in polymerized melts is assumed to be similar to that of bridging oxygens. A possible mechanism relating both viscous flow and OH diffusion is the exchange of bonds between adjacent tetrahedra. This process is illustrated in Fig. 5 for hydrous silica glass. The movement of an OH group results in formation of new ring connections and, thus, in local displacement of the tetrahedra in the melt. The OH diffusivities estimated by the Eyring relation from viscosity data for AOQ (albite-orthoclase-quartz) melts at 800 °C given by Schulze et al. (1996) are 4 orders of magnitude below the measured water diffusivities (Fig. 6). This strongly supports that OH groups do not contribute to the transport of water under the conditions of our experiments. From the diffusivity and viscosity data for haplogranitic melts we can exclude a significant contribution of OH to the effective diffusion of water down to 1 wt % water (DOH = 3.6 · 10)11 cm2/s; Dwater = 1.26 · 10)7 cm2/s, both for 1 wt % water at 1200 °C and 5 kbar). The good agreement of extrapolated diffusivities for haplogranitic melts with experimental data for other polymerized melts (Fig. 4) implies that the diffusion mechanism does not change at least for water contents down to several hundred ppm. Trace water diffusion studies show that this is probably also the case for much lower water contents. Concentrationdistance profiles from sorption of tritiated water in silica glass (Drury et al. 1962; Roberts and Roberts 1964; Burn and Roberts 1970) and aluminosilicate glass (Haider and Roberts 1970) slightly deviate from error function shape indicating that in the range of water contents covered by the experiments Dwater is slightly increasing with water content. This conclusion is sup- 382 Fig. 6 Comparison of water diffusion (Nowak and Behrens, this issue) and chemical diffusion in hydrous haplogranitic melts. Chemical diffusivities of haplogranite melts are calculated from viscosity data of Schulze et al. (1996) using the Eyring equation. Dotted lines are extrapolations from the experimental range. Note the same slope of both curves ported by higher diffusion coefficients determined for removal of water from silica glass compared to entry of water (Moulson and Roberts 1961). Doremus (1995) noted that the diffusion profiles in silica glass measured by Roberts and coworkers can be fitted well by a diffusion coefficient varying linearly with water content. Combining the trace water diffusion data with our results it is inferred that a linear relationship between Dwater and Cwater exists over a large range of water contents from about 10 ppm to ca. 3 wt % in polymerized melts. We conclude that a contribution of isolated OH groups to the transport of the chemical component water only is expected for water contents below 10 ppm. This is in contrast to the assumption of Chekhmir et al. (1988) that OH is the dominant diffusing species below 0.4 wt % water. 4, 5. Mechanism involving diffusion of H2O molecules Diffusion of H2O molecules has been inferred as the fundamental process for the transport of water in silicate glasses and melts by several authors (e.g. Drury et al. 1962; Roberts and Roberts 1966; Doremus 1969, 1995; Tomozawa 1985; Wasserburg 1988; Zhang et al. 1991 a,b). The basic difference of the discussed models is the role of the reaction of H2O molecules with bridging oxygen. Mechanism 4 denoted as the diffusion-reaction mechanism by Doremus (1969, 1995) assumes that H2O molecules move through the silicate network by direct jumps from one cavity to another one without reaction with oxygen of the silicate network. The H2O molecules that are occasionally trapped as hydroxyl groups are considered as immobile. This model is favoured for in- stance by Wasserburg (1988) and Zhang et al. (1991a,b). In contrast, mechanism 5 further denoted as the interconversion-diffusion mechanism assumes that the reaction of H2O molecules with bridging oxygen is an active step during movement of water. After re-formation of the oxygen bridge the H2O molecule can be expelled on a site different from the starting site and, thus, an effective jump of the H2O molecule occurs. Such a mechanism is preferred for instance by Roberts and Roberts (1966) and Tomozawa (1985). The diffusion-reaction mechanism was examined from a theoretical point of view by Wasserburg (1988), Zhang et al. (1991b) and Doremus (1995). At temperatures above the glass transition structural relaxation of the melt is assumed to be fast enough to establish local equilibrium of water species during diffusion. Assuming that the reaction of H2O molecules with bridging oxygens results in an immobilization it follows that for low water contents the effective diffusion coefficient is proportional to the concentration of water. As noted above, such a dependence was experimentally found for trace water diffusion in silica glass and for diffusion of water up to 3 wt % in haplogranitic and quartz-feldspathic melts. At temperatures far below the glass transition structural relaxation is too slow to achieve equilibrium of hydrous species. A model for diffusion of water under such non-equilibrium condition was developed and successfully tested for silica glass by Doremus (1995). Zhang et al. (1991a) analysed dehydration experiments on rhyolite glasses by considering the diffusion of both H2O molecules and OH groups and the reaction between them. From fits of water diffusion profiles by a two species model Zhang et al. (1991a) concluded that DOH is negligible for the transport of water. In the context of their model Dwater simply is determined by the diffusivity of water molecules DH2 O and by the ratio of H2O/OH. The authors concluded that DH2 O is independent of total water concentration in the range 0.2 to 1.7 wt %. However, a basic assumption for the evaluation of the experiments by Zhang et al. (1991a) is that the water speciation is not modified during quench which may not strictly be valid according to recent in situ near-infrared spectroscopic measurements (Nowak 1995; Nowak and Behrens 1995; Shen and Keppler 1995). In order to test the applicability of the model of Zhang et al. (1991a,b) for our diffusion conditions we calculated Dwater for 850 °C and compared the calculated data with our experimental data (Fig. 7). For the same temperature Zhang et al. (1991a) found a good fit for their model to the data from diffusion couple experiments of Lapham et al. (1984). Assuming local equilibrium of OH and H2O effective diffusion coefficients of water can be calculated from the model using Dwater DH2 O dXH2 O dXwater 2 where XH2 O and Xwater are the mole fractions of water molecules and total water calculated on a one-oxygen basis. Assuming an ideal mixture of H2O, OH and O and 383 dence of Dwater on total water that DH2 O depends on water content and, hence, on the structure of the melt. It is inferred that the model of Zhang et al. (1991a) may be useful to analyse water diffusion below the glass transition but can not be extrapolated to higher temperatures and water concentrations. Previous models for water diffusion in silicate glasses and melts only consider one type of OH group and one type of H2O molecule (e.g. Chekhmir et al. 1988; Zhang et al. 1991a,b; Doremus 1995). However, near-infrared spectroscopy indicates at least two different types of OH groups in silicate glasses (e.g. Langer and Flörke 1979; Zhang et al. 1995; Nowak and Behrens 1995). For a more detailed view on the mechanism of water diffusion it is useful to specify the reaction between water molecule and bridging oxygen and to distinguish OH singletons and OH pairs. In the first step of the reaction a pair of OH groups is formed: Fig. 7 Comparison of experimental water diffusion data for haplogranitic melts (run AOQD002 of Nowak and Behrens, this issue) with calculations using the model of Zhang et al. (1991a). Note the different shape of experimental and calculated curves a concentration independent equilibrium constant K1 for the reaction (H2O + O = 2 OH) XH2 O can be calculated from Eq. (5.2) of Silver and Stolper (1985) XH2 O Xwater ÿ K1 K1 ÿ 4 " 2 1 1 ÿ 2 4 ÿ K1 ÿ 4 ÿ Xwater K1 ÿ 2 Xwater 1 1=2 # 3 Combining the derivative of Eq. (3) with Eq. (2) gives 2 Dwater ÿ1 6 DH2 O 41 ÿ h 1 4 ÿ K1 ÿ 4 K1 2 ÿ ÿ Xwater Xwater 3 1 ÿ 2 Xwater 7 1i1=2 5 4 In the calculation we use DH2 O = 7.47 · 10)8 cm2/s extrapolated from data for rhyolite melts of Zhang et al. (1991a) and K1 = 0.986 extrapolated from in situ NIR spectroscopic measurements on haplogranitic melts of Nowak and Behrens (1995). The Dwater values calculated by the model of Zhang et al. (1991a) are significantly lower than experimental ones, especially at high water content (Fig. 7). This can simply be explained by lower contents of molecular H2O in the melts at 850 °C than in glasses investigated at room temperature because the DH2 O values of Zhang et al. (1991a) are based on NIR measurements at room temperature. However, it is obvious that the shapes of the experimental and calculated curves are different. The second derivative is positive for the experimental data but it is negative for the calculated curve. This fundamental difference can be explained neither by possible different DH2 O for rhyolitic or haplogranitic melts (compare Fig. 4) nor by the use of in situ measured speciation data instead of speciation data of quenched glasses. Thus, assuming OH and H2O to be the only hydrous species we conclude from the depen- H2 O O OH 1 1 1 HO 5 The notation [OH···HO] implies an interaction of both OH groups and a thermodynamic stability of the associate. In the second step the pair can dissociate to form OH singletons: OH 1 1 1 HO 2 OH 6 The dissociation of OH pairs is very slow because it is controlled by diffusion of OH singletons. The time scale of this reaction is several minutes or more at 400–600 °C for rhyolitic glasses with 0.5–2.3 wt % water as inferred by kinetic experiments of Zhang et al. (1995). The time sale of interconversion of water species is much smaller than that of dissociation of OH pairs as indicated by several experimental observations. The NIR and MIR heating stage experiments on hydrous silicate glasses of various anhydrous compositions performed at 1 atmosphere (Nowak 1995; Schmidt et al. 1995, 1996) show that the water speciation can not be frozen even with cooling rates of 300 K/s. Rapid and reversible changes in water speciation are already observed at temperatures of several hundred degrees below the glass transition. A rapid interconversion of water species also is inferred by NMR experiments on D2O– containing rhyolite by Eckert et al. (1987). They suggest that protons might be exchanging between H2O molecules and OH groups on a time scale faster than 10)6 s above 100 °C. Condensation of SiOH groups and hydrolysis of Si-O-Si bonds are well known to take place during corrosion of glass at temperatures below 100 °C (Scholze 1989). Direct evidence for a condensation reaction at 55 °C during corrosion of potassium-calciumsilicate glass is given by 29Si-MAS-NMR (Böhm et al. 1995). These examples demonstrate that interconversion of water species can be highly dynamic even at very low temperature. On the other hand, no exchange of oxygen between diffusing H218O and the network was observed in short time sorption experiments with silica glass at temperatures between 100 and 200 °C (Helmich and Rauch 1993). However, this is not necessarily in con- 384 tradiction with the assumption of rapid interconversion reaction. An exchange of oxygen only can be expected if the formed OH groups are equivalent for the condensation reaction. Due to the rigidity of the network in the glass at low temperature this is probably not the case and the same oxygen preferentially re-forms the bridge between the tetrahedral cations. Because of the high interconversion rate of OH pairs and H2O molecules we infer that reaction (5) is a basic step during movement of water within polymerized silicate melts and glasses. However, a contribution of direct jumps of water molecules without interaction to the framework can not completely be ruled out. A distinction of both diffusion mechanisms by considering the activation volume for water diffusion in haplogranitic melts is not possible. The positive activation volume can be explained by an expansion of aluminosilicate rings if H2O diffuses as a rigid molecule as well as by an enlargement of the structure during reaction of H2O molecules and (Al, Si)-O-(Al, Si) bonds, e.g. by formation of transient 5-coordinated tetrahedral cations. A critical argument of Doremus (1969) against a diffusion mechanism which involves breaking of silicon oxygen bonds is that the activation energy should be at least as high as the bond energy of about 380 kJ/mol. However, the activation energy corresponds to the difference in energy of the stable states and the transition state. It can be considerably lower than the bond energy if, for instance, the process is catalysed by protons. In conclusion we favour an interconversion-diffusion model in which water mainly migrates by the interconversion reaction but small sequences of direct jumps are possible. In this model the mobile species are water molecules as well as OH pairs. With the assumption that the interconversion of both species is faster than the diffusive jump or, in other words, a water molecule reacts several times with the same oxygen bridge before it attacks another one, in first approximation both species can be considered as a single moving species (H2O,OH pair). Thus, effective diffusion of water in the melt can be described as the sum of contributions from (H2O,OHpair) and OH singletons Dwater DH2 O; OH pair dXH2 O; OH pair dXwater DOH single dXOH single dXwater 7 Quantitative modelling of water diffusion by Eq. (7) is complicated because the concentrations of OH pairs and OH singletons can not be determined directly from spectroscopic measurements. From our experimental results we conclude that the first term of Eq. (7) dominates for water contents > 100 ppm. The second term is expected to be dominant only at very low water contents (« 10 ppm). The convex shape of Dwater versus Cwater implies that DH2 O;OHpair increases with increasing water content as discussed before for the model of Zhang et al. (1991a). The change from linear to exponential dependence of Dwater on Cwater between 2 and 3 wt % observed for Ab, Qz29Or71 and Qz28Ab38Or34 compositions can result from the dissociation reaction (6) and from modification of the melt structure by incorporation of OH groups. The OH pairs becomes immobile by splitting into two OH singletons. With decreasing water content the mean distance between OH singletons becomes larger and, therefore, longer times are required for bringing OH singletons together to form a mobile OH pair. From this point of view, the change in the dependence of Dwater on Cwater is related to a critical distance of OH groups for which the dissociation reaction does not significantly affect the transport of water. This critical distance is estimated to be 6 Å assuming 3 wt % of water dissolved as randomly distributed OH groups. However, the diffusion-reaction problem is too complicated for a quantitative modelling of the diffusion data. As shown by Zhang et al. (1995) for the interconversion reaction a solution of this problem only is possible for relatively simple approaches, for instance, ignoring different subspecies of OH and H2O and assuming concentration independent mobilities of the species. Such approaches are not reasonable for the water diffusion process. The viscosity of haplogranitic melts changes by several orders of magnitude for water contents ranging from 0 to 1 wt % (Dingwell et al. 1996; Schulze et al. 1996) which indicates strong variation of the melt structure and of the mobility of OH singletons in this concentration range. Thus, the time in which water is immobile is not only determined by the distance of OH singletons but also by the melt structure. Moreover, it is emphasized that the same slope (0.25 log units per wt % water) is observed for water diffusion and viscous flow in haplogranitic melts at high water contents although the mechanisms of both processes are different. This implies that the exponential dependence of Dwater on water content is at least in part the result of modification of the melt structure by incorporation of OH groups. Concluding remarks On the basis of the water diffusivity data, in situ determination of hydrous species by NIR spectroscopy and viscosity data we conclude that OH pairs and water molecules are the hydrous species which control the transport of water in polymerized melts. It is inferred that the reversible reaction H2O + O = OH···HO is a basic step for movement of water but small sequences of direct jumps of H2O molecules also may contribute to migration of water. The proposed diffusion mechanism implies that only a part of the OH groups (OH singletons) is immobile on the time scale of water diffusion but the other (OH pairs) is directly involved in the transport of water. From viscosity data and trace water diffusion data it is estimated that OH singletons only may influence the transport of water at water contents « 10 ppm. The change in water concentration dependence of Dwater at about 3 wt % is attributed to the kinetics of dissociation and re-formation of OH pairs and to changes of the melt structure. 385 Acknowledgment We thank K. Roselieb and M. Carroll for stimulating comments on the manuscript. This work was supported by the SFB173 of the DFG. References Behrens H (1995) Determination of water solubilities in highviscosity melts: an experimental study on NaAlSi3O8 and KAlSi3O8 melts. Eur J Mineral 7: 905–920 Behrens H, Romano C, Nowak M, Holtz F, Dingwell DB (1996) Near-infrared spectroscopic determination of water species in glasses of the system MAlSi3O8 (M = Li, Na, K): an interlaboratory study. Chem Geol 128: 41–63 Böhm T, Leissner J, Chudek JA (1995) Investigation of glass corrosion in acid solutions with 29Si-MAS-NMR. Glastech Ber Sci Technol 68: 400–403 Burn I, Roberts JP (1970) Influence of hydroxyl content on the diffusion of water in silica glass. Phys Chem Glasses 11: 106–114 Chekhmir AS, Epel’baum MB, Simakin AG (1988) Water transport in magmas. Geochem Int 26: 125–127 Cockram DR, Haider Z, Roberts GJ (1969) The diffusion of ‘‘water’’ in soda-lime glass within and near the transformation range. Phys Chem Glasses 10: 18–22 Delaney JR, Karsten JL (1981) Ion microprobe studies of water in silicate melts: concentration-dependent diffusion in obsidian. Earth Planet Sci Lett 52: 191–202 Dingwell DB, Romano C, Hess K-U (1996) The effect of water on the viscosity of a haplogranitic melt under P-T-X conditions relevant for silicic volcanism. Contrib Mineral Petrol 124: 19–28 Doremus RH (1969) The diffusion of water in fused silica. In: Mitchell CW (ed) Reactivity of solids. Proc 6th Int Symp. Wiley, NewYork, pp 667–671 Doremus RH (1979) Chemical durability of glass. Treatise Mater Sci Techn 17: 41–69 Doremus RH (1995) Diffusion of water in silica glass. J Mater Res 10: 2379–2389 Drury T, Roberts GJ, Roberts JP (1962) Diffusion of ‘‘water’’ in silica glass. In: Adv Glass Technol Pap 6th Int Congr Glass, Plenum Press, New York, pp 249–255 Eckert H, Yesinowski JP, Stolper EM, Stanton TR, Holloway J (1987) The state of water in rhyolite glasses: a deuterium NMR study. J Non-Cryst Solids 93: 93–114 Ernsberger FM (1980) The role of molecular water in the diffusive transport of protons in glasses. Phys Chem Glasses 21: 146–149 Haider Z, Roberts GJ (1970) The diffusion of ‘‘water’’ into simple silicate and aluminosilicate glasses at temperatures near the transformation range. Glass Technol 6: 158–163 Helmich M, Rauch F (1993) On the mechanism of diffusion of water in silica glass. Glastech Ber 66: 195–200 Jambon A (1979) Diffusion of water in a granitic melt: an experimental study. Carnegie Inst Washington Yearb 78: 352–355 Jambon A, Zhang Y, Stolper EM (1992) Experimental dehydration of natural obsidian and estimation of DH2 O at low water contents. Geochim Cosmochim Acta 56: 2931–2935 Karsten JL, Holloway JR, Delaney JR (1982) Ion microprobe studies of water in silicate melts: temperature-dependent water diffusion in obsidian. Earth Planet Sci Lett 59: 420–428 Keppler H, Bagdassarov NS (1993) High temperature FTIR spectra of H2O in rhyolite melt to 1300 °C. Am Mineral 78: 1324–1327 Langer K, Flörke OW (1974) Near-infrared absorption spectra (4000–9000 cm)1) of opals and the role of ‘‘water’’ in these SiO2*nH2O minerals. Fortschr Mineral 52: 17–51 Lapham KE, Holloway JR, Delaney JR (1984) Diffusion of H2O and D2O in obsidian at elevated temperatures and pressures. J Non-Cryst Solids 67: 179–191 Moulson AJ, Roberts JP (1961) Water in silica glass. Trans Faraday Soc 57: 1208–1216 Nowak M (1995) Der Einbau von Wasser in haplogranitische Gläser und Schmelzen. PhD thesis, Univ Hannover, Germany 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: 3445–3450 Nowak M, Behrens H (1996) An experimental investigation on diffusion of water in haplogranitic melts. Contrib Mineral Petrol (this issue) Roberts GJ, Roberts JP (1964) Influence of thermal history on the solublity and diffusion of ‘‘water’’ in silica glass. Phys Chem Glasses 5: 26–31 Roberts GJ, Roberts JP (1966) An oxygen tracer investigation of the diffusion of ‘‘water’’ in silica glass. Phys Chem Glasses 7: 82–89 Rubie DC, Ross CR II, Carroll MR, Elphick SC (1993) Oxygen self-diffusion in Na2Si4O9 liquid up to 10 GPa and estimation of high-pressure melt viscosities. Am Mineral 78: 574–582 Schmidt MO, Behrens H, Nowak M (1995) MIR heating stage experiments on incorporation and dynamic of water species in hydrous silicate glasses. Eur J Mineral Beih 7: 212 Schmidt MO, Behrens H, Nowak M (1996) Temperature dependence of water speciation in silicate and aluminosilicate glasses (abstract). Terra Abstr 8: 58–59 Scholze H (1988) Glass-water interactions. J Non-Cryst Solids 102: 1–10 Scholze H (1989) The chemistry of glass surfaces. In: Proc 15th Int Congr Glass, Leningrad, pp 302–326 Scholze H, Mulfinger HO (1959) Der Einbau des Wassers in Gläsern. V. Die Diffusion des Wassers in Gläsern bei hohen Temperaturen. Glastech Ber 32: 381–385 Schulze F, Behrens H, Holtz F, Roux J, Johannes W (1996) The influence of water on the viscosity of a haplogranitic melt. Am Mineral 81: 1155–1165 Shaw HR (1974) Diffusion of H2O in granitic liquids. I. Experimental data. II. Mass transfer in magma chambers. In: Hofmann AW, et al (eds) Geochemical transport and kinetics. Carnegie Inst Washington Publ 634, pp 139–170 Shen A, Keppler H (1995) Infrared spectroscopy of hydrous silicate melts to 1000 °C and 10 kbar: direct observation of water speciation in a diamond anvil cell. Am Mineral 80: 1335–1338. Shimizu N, Kushiro I (1984) Diffusivity of oxygen in jadeite and diopside melts at high pressures. Geochim Cosmochim Acta 48: 1295–1303 Silver L, Stolper EM (1985) A thermodynamic model for hydrous silicate melts. J Geol 93: 161–178 Stanton TR, Hervig RL, Holloway JR (1988) Compositional effect on water diffusivity in silicate melts. EOS Trans Am Geophys Union 69: 511 Tomozawa M (1985) Concentration dependence of the diffusion coefficient of water in SiO2 glass. J Am Ceram Soc 68: C251– 252 Wasserburg GJ (1988) Diffusion of water in silicate melts. J Geol 96: 363–367 Watson EB (1994) Diffusion in volatile-bearing magmas. (Reviews in mineralogy, 30 Volatiles in Magmas, eds. Carroll MR, Holloway JR) Mineral Soc Am, Washington, DC, pp 371–411 Watson EB, Baker DR (1991) Chemical diffusion in magmas: an overview of experimental results and geochemical applications. Adv Phys Geochem 9: 120–151 Zhang Y, Stolper EM (1991) Water diffusion in a basaltic melt. Nature 351: 306–309 Zhang Y, Stolper EM, Wasserburg GJ (1991a) Diffusion of water in rhyolitic glasses. Geochim Cosmochim Acta 55: 441–456 Zhang Y, Stolper EM, Wasserburg GJ (1991b) Diffusion of a multi - species component and its role in oxygen and water transport in silicates. Earth Planet Sci Lett 103: 228–240 Zhang Y, Stolper EM, Ihinger PD (1995) Kinetics of the reaction H2O + O = 2OH in rhyolitic and albitic glasses: preliminary results. Am Mineral 80: 593–612
© Copyright 2026 Paperzz