Cement-Bentonite Interaction Experiments

The goal of the completed PhD project (Dolder, 2015, Dolder, 2014) was to characterize and quantify the cement/bentonite skin effects spatially and temporally in laboratory experiments. A newly developed mobile X-ray transparent core infiltration device was used, which allows performing X-ray computed tomography (CT) periodically without interrupting a running experiment. CT scans allowed tracking the reaction plume and the core volume/diameter in the running experiments. The experiments attested an effective buffering capacity for bentonite and sand/bentonite, a progressing coupled hydraulic-chemical sealing process and also the preservation of the physical integrity of the interface region in this setup with a total pressure boundary condition on the core sample. No complete pore-clogging was observed but the hydraulic conductivity got rather strongly reduced in 3 experiments, explained by clogging of the intergranular porosity (macroporosity). Such a drop in hydraulic conductivity may retard the saturation time of the buffer in a nuclear waste repository.

Evolution of the reaction plume
Evolution of the reaction plume based on CT images and analysed by image processing: (a) sketch of the analysed section; (b) zone of reaction consisting of an increased X-ray attenuation (time in days after infiltration of artificial Portland cement pore water) (Dolder, 2015)

The Mt. Terri CI (cement clay interaction) experiment contains bentonite in contact with different cements. Intact interfaces have been sampled by RWI, and are currently in the focus of mainly Japanese research (CRIEPI).

Cement-Opalinus Clay Interaction Experiment

Extensive investigations on concrete-Opalinus Clay (OPA) interfaces recovered after 2.2 y of interaction from the Cement-Clay Interaction (CI) experiment at the Mont Terri Underground Laboratory (St. Ursanne, Switzerland) revealed significant alteration in the different cements within a couple of millimetres, but minor changes in OPA (Jenni et al., 2014).

Owing to the especially developed drilling technique for coring undisturbed interfaces (Jenni et al., 2014), a larger number of high-quality samples could be retrieved from the same interfaces after 4.85 y of interaction, allowing for more extensive investigations. These include

  • standard optical and scanning electron microscopy with energy dispersive spectroscopy, resulting in element maps and semi-quantitative chemical data (SEM EDX)
  • synchrotron micro x-ray diffraction spectroscopy (µXRD), resulting in spatially resolved phase information (Dähn et al., 2014)
  • synchrotron x-ray absorption near edge structure (XANES), resulting in spatially resolved phase and elemental coordination information
  • µRaman and infrared spectroscopy, both supporting spatially resolved phase identification and phase structure information
  • autoradiography, resulting in spatially resolved porosity information
  • determination of cation exchange occupancy and aqueous leaching of the clay

The thin interaction layers in both cement and OPA, the multi-mineral concrete aggregates that disturb investigations on the cement in between, and the presence of poorly crystalline and/or solid-solution phases, necessitate this comprehensive suite of spatially resolved expensive techniques.

RWI studies concentrate on three concrete-OPA interfaces, whose concretes contain 1) Portland cement (OPC), 2) ESDRED cement especially designed for repository applications (40% of cement substituted with silica fume), and 3) low alkali cement (LAC, containing slag and nanosilica).

Sampling of interfaces
Schematic of optimised stabilisation technique for sampling of interfaces: access borehole from top right through Opalinus Clay, followed by six anchor boreholes with glued-in anchors (olive), and subsequent overcoring of three interfaces through anchors (Jenni et al., 2014))

In general, the chemical zonation and the thicknesses of the zones did not change from 2.2 y to 4.85 y of interaction, but strongly depend on the type of cement. The striking Mg enrichment close to OPA occurs only in the low-pH cements LAC and ESDRED, where the OPA is also enriched close to the interface. Only a fraction of this Mg enrichment can be attributed to the observed increase of Mg on the clay exchanger (at the expense of Ca and Na), but µXRD indicates the presence of a poorly crystalline magnesium silicate hydrate (MSH). The same phase is present in the Mg enriched cement, as shown by µXRD (Figure) and XANES, both spectra are compared with measurements on pure synthesised MSH. Structure and thermodynamic property investigations are ongoing, first results indicate similarities to talc. Further characteristics of the interaction zone are a decalcification in all three cements close to OPA, and a sulphur depletion, followed by a layer enriched in S further away. Thicknesses of observed carbonation layers mainly depend on the water/binder ratio of the cement. Decalcification decreases, carbonation increases porosity in the cements.

Sulphur SEM EDX map
Sulphur SEM EDX map of Portland cement concrete (top) in contact with Opalinus Clay (bottom). High sulphur concentrations in white, interface in red, different layers in the cement and clay indicated in green and blue, respectively

Data from 2.2 y and 4.85 y samples both suggest diffusion of CO2/HCO3-, SO42-, and Mg species, from OPA into cement. Possible pH decrease in the cement next to the interface leads to instability of sulphur-bearing cement hydrates, the S in solution diffuses towards higher pH (away from the interface), where ettringite can form and incorporate S. In turn, the cement releases Ca, which is partially found in short calcite veins extending into the OPA.

Reactive Transport Modelling Across Cement-Opalinus Clay Interfaces

Reactive transport modelling helps to understand mechanisms, which interact and leave behind complex traces in interface samples. Modelling cement – clay rock interaction is extremely demanding:

  • Clay rocks contain a pore fraction influenced by the negatively charged surfaces of clay minerals (Donnan porosity). This pore fraction, typically around half of total pore volume (Opalinus Clay), has a solute composition different from the charge-balanced intergranular porosity. Porosity clogging is expected to occur mainly within intergranular porosity, whereas solute transport continues in the Donnan porosity, although transport properties herein are substantially different. Conventional model approaches treat the Donnan porosity as immobile cations undergoing cation exchange reactions.
  • Several studies indicate that interactions start right after concrete emplacement, and slow down relatively soon thereafter. The entire cement hydration at the interface, interacting with the clay rock, has to be modelled. Therefore,
  • the model needs to cover porosity evolutions ranging from 0.7 in a fresh cement paste, down to zero at a clogged interface. Furthermore, immense spatial gradients occur in porosity, pH, and solute concentrations.
  • Many cement hydrates form process – relevant solid solutions, of which one exclusively observed at cement – clay interfaces has only recently thermodynamically characterised and implemented in a numerical simulation (MSH).

Two multicomponent reactive transport codes meet these demands: CrunchFlowMC (Carl Steefel, clay-specific extensions developed in collaboration with RWI, Alt-Epping et al, 2014), and an in-house developed extension of FLOWTRAN. The set-up of the Mt. Terri CI experiment, a concrete – Opalinus Clay interface, was simulated from right after casting the concrete into a drill hole. The model results reproduce chemical zonations measured by SEM EDX after 3 and 5 y, and meet mineralogical gradients observed by various methods (Jenni et al., 2014). The numerical results allow for isolating the major interacting mechanisms:

  • Decalcification of cement, Ca increase on clay exchanger
  • S diffusion from clay into cement, and further away from interface until pH is again high enough for precipitation of S containing hydrates
  • Mg diffusion from clay to the interface, precipitation of mainly MSH
  • C diffusion from clay into cement, followed by carbonation of cement hydrates

All these mechanisms influence porosity at different locations. The model outcome can give an estimation of porosity evolution with time, and emphasise the change in transport properties.

Ca distribution
Ca distribution in the cement matrix and in the clay after 5 years of interaction: model prediction (orange line), EDX spot measurements in the cement matrix (left), bulk EDX measurements in the clay (right), compared with a EDX Ca map (Jenni, in prep.)