Technical Report NTB 05-01

Grimsel Test SiteInvestigation Phase VHPF-Experiment:Modelling Report

Cement is a major component of the engineered barrier system in proposed underground repositories for low- and intermediate-level radioactive waste. The interaction between the hyperalkaline solutions derived from the degradation of cement and the rocks hosting such repositories may change the physical and chemical properties of the host rocks. The Hyperalkaline Plume in Fractured Rock (HPF) project at the Grimsel Test Site (Switzerland) addresses this issue in the context of a fractured crystalline host rock (granite).

The granite at Grimsel is characterized by the presence of ductile shear zones (Bossart & Mazurek, 1991), with thicknesses ranging up to at least meter to decameter scales. These shear zones include intensely deformed mylonitic bands with thicknesses up to several tens of centimeters. Brittle fractures developed in later stages of deformation, mainly within the mylonitic bands. These fractures, with thicknesses in the millimeter range, are at least partially filled with a porous fault gouge.

The HPF project, which is funded by ANDRA (France), JNC (now JAEA, Japan), NAGRA (Switzerland), SKB (Sweden); POSIVA (Finland) and the DOE (USA), included an underground field experiment (injection of a hyperalkaline solution in a hydraulic dipole setting, tracer transport experiments), small-scale laboratory experiments and structural and mineralogical characterization. In order to understand the interaction of the hyperalkaline solution with the fractured shear zone at the Grimsel Test Site, the laboratory and underground field experiments were analyzed by means of numerical modeling.

Initially, different sets of scoping calculations were performed to asses the spatial scale and magnitude of the mineralogical alteration associated with the injection of a hyperalkaline solution into a fracture at the Grimsel Test Site, and to aid in the design of the experiment. The contemplated duration of the experiment in the simulations ranged from 1 to 3 years. All concepts were based on one-dimensional reactive transport modeling of the interaction between an injected hyperalkaline solution and the mineralogy of the fault gouge filling the fracture. In addition, matrix diffusion into the wall rock was also considered in one of the conceptual models that were investigated. The two main conclusions from the entire set of scoping calculations were (1) that relatively large flow rates had to be implemented in the field in order to expect a significant amount of mineralogical alteration, and (2) that the precipitation of significant amounts of secondary minerals was only to be expected if the reactive surface areas of the primary minerals in the rock were at least close to the values of specific surface area measured with the BET technique. Other identified uncertainties were associated with the mineralogy of the potential secondary phases that could precipitate and with the magnitudes of mineral reaction rates.

After the experiments were started, one-dimensional flow and reactive transport models were applied to reproduce the breakthrough curves measured in the small-scale laboratory experiment (injection of a high-pH solution into a drill core containing a fracture). An important finding from this experiment was that the interaction between the hyperalkaline solution and the fault zone in Grimsel granite caused a significant reduction in the hydraulic conductivity of the rock core, even though the amount of mineral alteration was minor. The modeling results confirmed that the dissolution of primary minerals was kinetically controlled. The two modeling approaches (GIMRT and 3FLO) used rate constants based on published experimental results for the primary minerals and larger rate constants for the secondary minerals (simulating conditions close to local equilibrium for these secondary phases). In order to obtain a reasonable agreement between models and experimental results, reactive surface areas of the order of 105 m2/m3 rock had to be used for both GIMRT and 3FLO. These values are much smaller than those measured for the fault gouge filling the fracture, which are of the order of 10- 107 m2/m3 rock. However, the match between the simulations and the observations was improved by adding a small fraction of fine-grained mineral, which could explain the high initial peaks in Al and Si concentration. With the inclusion of this fine-grained fraction, the initial surface areas in the model were within the range of the measured specific surface areas of the fault gouge.

Tracer tests at the Grimsel Test Site were interpreted assuming both homogeneous (a) and heterogeneous (b) distributions of hydraulic conductivity of the shear zone.

a)  Although this approach was based on a relatively simple model (isotropic hydraulic properties identical for all dipole configurations), peak arrival times and the shape of the breakthrough curves were generally reproduced. A long-lasting tailing seems to be induced by a continuous release of low levels of tracer in the injection interval, possibly by diffusion from stagnant zones in the surrounding formation or from the test equipment. The modeling systematically overpredicted tracer concentrations in the effluent for all dipole configurations, as if the amount of tracer was a factor of 3 to 5 lower than injected. It is difficult to explain this trend by adding complexity to the flow field within the shear zone plane or by using different boundary conditions. It has been hypothesized that unexpected tracer losses may have resulted from the injection into features corresponding to intervals other than the ones used directly in the experiment, although this effect has not been confirmed by the presence of high-pH conditions in any of these other intervals.

b)  Dipole experiments were interpreted using different concepts to reconcile transport processes and radionuclide migration within a shear zone altered by high-pH fluid. Parameter fits were possible for a multiple fracture-matrix approach, as well as for a two-dimensional heterogeneous medium approach. Discrimination between these two approaches was not possible, although the extended dipole flow field geometry, the quite dissimilar breakthrough curves measured for experiments with different dipole geometries and the measured lateral spreading of the high-pH plume favored the heterogeneous porous medium approach. Using a heterogeneous porosity distribution, together with an empirical Kozeny-Carman equation that relates porosity and hydraulic conductivity, a heterogeneous flow field could be calculated for the shear zone. This flow field was used to predict the interaction of the hyperalkaline solution with the shear zone. Calculations were also performed to predict the spreading of Cs, Co and Eu radionuclide tracers within the shear zone altered by high-pH interaction. It has been shown that the calculated Cs, Co and Eu breakthroughs and their concentration distributions depend on the assumptions on sorption behavior. In addition, the observed decrease in hydraulic conductivity of the system, which is observed both in the field and in small-scale core infiltration experiments, and the related changes in the flow field, which are linked to mineral alteration, will strongly influence the migration of radionuclides.

Two-dimensional reactive transport modeling of the geochemical evolution of the system was also performed. Preliminary modeling with GIMRT and a first comparison with experimental observations showed an important retardation in Na and K experimental breakthrough compared to the model, which was partly due to the change in the flow field during the experiment and most probably due to chemical retardation (sorption) of both Na and K. Also, the fact that the concentrations at extraction at steady-state were the same as the concentrations at injection pointed to a channeling effect, which gradually prevented mixing of the injected solution with the surrounding groundwater. These results also indicated that there was a strong possibility that secondary mineral precipitation in the shear zone was only minor and therefore difficult to detect. Modeling with 3FLO suggested the following conclusions:

  • Na and K concentrations in the effluent could be reproduced if a sorption term were assumed. Otherwise, the initial flow and transport model was consistent with the observations.
  • The values of Na and K concentrations at late stages of the experiment are in fairly good agreement with the experimental observations, although the shapes of the breakthrough curves are not identical.
  • The trends in Ca, Al and Si concentration in the effluent were globally well simulated, although the magnitudes were not.
  • The observed trends and orders of magnitude of the pH breakthroughs at the observation wells were reproduced, although they occurred slightly too fast at the wells further away from injection (99.008 and 98.004).
  • The increase in the injection pressure was reproduced but the modeled values were too high, suggesting that the precipitation-induced permeability reduction was overestimated.
  • The evolution of the breakthrough curves corresponding to the dipole tests performed under high-pH conditions was not reproduced. The model does not predict the formation of any preferential pathway with time, which is necessary to reproduce the experimental results.

Overall, the major conclusion from the modeling of the laboratory and field experiments can be summarized as follows:

  • Injection of the high-pH solution modifies the hydraulic conductivity of the flow field, significantly altering tracer travel times and even the geometry of the flow field. The results of the field experiment point to a channeling effect, which severely limits mixing of the injected high-pH solution with background Grimsel groundwater at late stages of the experiment.
  • Relatively little pH buffering of the hyperalkaline plume by the Grimsel granite occurs, indicating that the use of kinetic formulation for the mineral dissolution was appropriate.
  • The fracture zone appears to be sufficiently heterogeneous and random that it is unlikely that the results from other tests can be predicted deterministically. However, the basic assumption is that the average or ensemble behavior, given the stochastic nature of the hydraulic conductivity, porosity and mineral distributions, can still be understood.

Adequate matching of major cation retardation (and, by implication, the retardation of radionuclides like Cs that sorbs according to an ion exchange mechanism) may require a more sophisticated and comprehensive ion exchange and/or surface complexation model.