Testing geochemical models of bentonite pore water evolution against laboratory experimental data

Abstract The determination of a bentonite pore water composition and understanding its evolution with time underpins many radioactive waste disposal issues, such as buffer erosion, canister corrosion, and radionuclide solubility, sorption, and diffusion, inter alia . Previous modelling approaches have tended to ignore clay dissolution–precipitation reactions, a consequence of which is that montmorillonite is theoretically preserved indefinitely in the repository system. Here, we investigate the applicability of an alternative clay pore fluid evolution model, that incorporates clay dissolution–precipitation reactions as an integral component and test it against well-characterised laboratory experimental data, where key geochemical parameters, Eh and pH, have been measured directly in compacted bentonite. Simulations have been conducted using different computer codes (Geochemist’s Workbench, PHREEQC, and QPAC) to test the applicability of this model. Thermodynamic data for the Gibb’s free energy of formation of MX-80 smectite used in the calculations were estimated using two different methods (‘Polymer’ and ‘Vieillard’ Models). Simulations of ‘end-point’ pH measurements in batch bentonite–water slurry experiments showed different pH values according to the complexity of the system studied. The most complete system investigated revealed pH values were a strong function of partial pressure of carbon dioxide, with pH increasing with decreasing P CO 2 (with log  P CO 2 values ranging from −3.5 to −7.5 bars produced pH values ranging from 7.9 to 9.6). A second set of calculations investigated disequilibrium between clay and pore fluid in laboratory squeezing cell tests involving pure water (pH = 9.0) or a 1 M NaOH solution (pH = 12.1). Simulations carried out for 100 days (the same timescale as the experiments) showed that smectite remained far from equilibrium throughout, and that the lowering of pH due to smectite hydrolysis was trivial. However, extending the duration of the simulations to that required for clay–fluid equilibrium, necessitated timescales of 7 and 65 years for pure water and 1 M NaOH, respectively, but again produced relatively minor reduction in pH (in the order of 0.1–0.2 pH units). If the (equilibrium) precipitation of secondary minerals was included in the simulations, then not only was the clay–fluid equilibration period extended dramatically (from 7 to 360 years for pure water, and from 65 to 2600 years for 1 M NaOH), but concomitant changes in pH were significant, decreasing from 9.0 to 8.6 (pure water) and from 12.1 to 9.0 (1 M NaOH). Repetition of these latter calculations using an alternative method for Δ G f 0 smectite produced an increase in equilibration time for reaction with 1 M NaOH from 2600 to 5000 years, highlighting the potential effects of the uncertainty in thermodynamic data for smectite. A final set of calculations was carried out to investigate both the time- and space-dependent variations in pore fluid composition in laboratory in-diffusion experiments conducted for over 1200 days, initially with pure water and ‘spiked’ after 271 days with a Na–Ca–OH–Cl solution (pH = 11.7). Here, the sensitivity of the results to both variations in a number of parameters/conditions (porosity, reaction rate of secondary minerals, the degree of mixing of the external fluid reservoirs in the experiments, the effective diffusion coefficient) and the inclusion/exclusion of key processes (clay hydrolysis, secondary mineral precipitation, ion exchange, clay edge protonation–deprotonation reactions) was investigated. These calculations confirmed that smectite dissolution–precipitation reactions alone have an insignificant impact upon pH buffering over laboratory timescales and that the pH buffering observed is most likely controlled by clay protonation–deprotonation reactions, and kinetic secondary mineral (brucite + tobermorite) precipitation. Ion exchange reactions were found to have little effect on pH. Alternative data for the kinetic dissolution of smectite produced no observable differences, and the adoption of a reduced diffusion coefficient produced a poorer fit to experiment results. In conclusion, modelling predicts that the effects of smectite dissolution on the chemistry of bentonite pore waters would be essentially undetectable over experimental time scales, but when the model is combined with plausible constraints on the precipitation of secondary minerals, significant changes in solution chemistry and mineralogy are predicted to occur over time scales that are relevant to repository near-field evolution (hundreds to thousands of years). There are remaining fundamental uncertainties related to the variable chemistry of the smectite clays, the nature of porosity in highly compacted buffer materials, the reactive surface area of smectite, and the thermodynamic properties of these clay minerals.