Carbon dioxide in silicate melts: A molecular dynamics simulation study

Abstract

We have performed a series of molecular dynamics simulations aimed at the evaluation of the solubility of CO 2 in silicate melts of natural composition (from felsic to ultramafic). In making in contact within the simulation cell a supercritical CO 2 phase with a silicate melt of a given composition, we have been able to evaluate (i) the solubility of CO 2 in the P– T range 1473–2273 K and 20–150 kbar, (ii) the density change experienced by the CO 2-bearing melt, (iii) the respective concentrations of CO 2 and CO 3 2 - species in the melt, (iv) the lifetime and the diffusivity of these species and (v) the structure of the melt around the carbonate groups. The main results are the following: (1) The solubility of CO 2 increases markedly with the pressure in the three investigated melts (a rhyolite, a mid-ocean ridge basalt and a kimberlite) from about ∼2 wt% CO 2 at 20 kbar to ∼25 wt% at 100 kbar and 2273 K. The solubility is found to be weakly dependent on the melt composition (as far as the present compositions are concerned) and it is only at very high pressure (above ∼100 kbar) that a clear hierarchy between solubilities occurs (rhyolite < MORB < kimberlite). Furthermore at a given pressure the calculated solubility is negatively correlated with the temperature. (2) In CO 2-saturated melts, the proportion of carbonate ions ( CO 3 2 - ) is positively correlated with the pressure at isothermal condition and is negatively correlated with the temperature at isobaric condition (and vice versa for molecular CO 2). Furthermore, at fixed ( P, T) conditions the proportion of carbonate ions is higher in CO 2-undersaturated melts than in the CO 2-saturated melt. Although the proportion of molecular CO 2 decreases when the degree of depolymerization of the melt increases, it is still significant in CO 2-saturated basic and ultrabasic compositions at high temperatures. This finding is at variance with experimental data on CO 2-bearing glasses which show no evidence of molecular CO 2 as soon as the degree of depolymerization of the melt is high (e.g. basalt). These conflicting results can be reconciled with each other by noticing that a simple low temperature extrapolation of the simulation data predicts that the proportion of molecular CO 2 in basaltic melts might be negligible in the glass at room temperature. (3) The carbonate ions are found to be transient species in the liquid phase, with a lifetime increasing exponentially with the inverse of the temperature. Contrarily to a usual assumption, the diffusivity of carbonate ions into the liquid silicate is not vanishingly small with respect to that of CO 2 molecules: in MORB they differ from each other by a factor of ∼6 at 1473 K and only a factor of ∼2 at 2273 K. Although the bulk diffusivity of CO 2 is governed primarily by the diffusivity of CO 2 molecules, the carbonate ions contribute significantly to the diffusivity of CO 2 in depolymerized melts. (4) Concerning the structure of the CO 2-bearing silicate melt, the carbonate ions are found to be preferentially associated with NBO’s of the melt, with an affinity for NBOs which exceeds that for BOs by almost one order of magnitude. This result explains why the concentration in carbonate ions is positively correlated with the degree of depolymerization of the melt and diminishes drastically in fully polymerized melts where the number of NBO’s is close to zero. Furthermore, the network modifier cations are not randomly distributed in the close vicinity of carbonate groups but exhibit a preferential ordering which depends at once on the nature of the cation and on the melt composition. However at the high temperatures investigated here, there is no evidence of long lived complexes between carbonate groups and metal cations.