OR/14/048 Implications for repository performance

Rochelle, C A, Purser, G, and Milodowski, A E. 2014. Results of laboratory carbonation experiments on NRVB cement. British Geological Survey Internal Report, OR/14/048.

Information in the proceeding sections allowed identification and quantification of processes occurring during cement carbonation, and this will help inform predictive modelling of repository evolution. The buffer/backfill cement appears to have coped with carbonation well, generally remaining intact, with samples showing no evidence of overall shrinkage or swelling.

Carbonation was identifiable by a colour change of the cement samples (dark grey to light brown), and this progressed from the outside of the samples towards their centres. Carbonation was rapid, especially early in the experiments, with some samples showing at least partial carbonation through to the centres of the samples (12.5 mm minimum travel distance) within 40 days. Most of the carbonation observed was relatively uniform on greater than mm scales. However, a small number of (mainly diffusion) samples showed some evidence for carbonation along specific zones. This may have been aided in certain cases by heterogeneities within the samples. These appear to have been caused by segregation of cement grains during casting/setting. The coarser layers had their pore space filled with Ca(OH)₂ and this was more susceptible to carbonation than the finer-grained CSH phases. Similar features could form in a repository setting, and could provide preferential routes for CO₂ migration if they occurred on a large scale. Such cement heterogeneity could be minimised through the use of organic additives (superplasticisers) to enhance cement flow behaviour. However, Francis et al. (1997) and Young et al., (2013) note that the presence of such organics may enhance the mobility of some radionuclides. Without them however, some degree of grain segregation and hetergoenious carbonation may be a feature within a repository. It will be important therefore, to consider carefully the pros and cons of adding additives to repository buffer/backfill cement.

In terms of bulk samples, carbonation was not associated with a sample volume change that meant that overall NRVB density increased. In detail however, this density change was not uniform. At microscopic scales carbonation resulted in a patchwork of low-density domains composed mainly of silica gel enclosed by higher density zones of secondary carbonate. The low-density domains had higher porosity, and the higher density zones had lower porosity. These higher density zones may have formed as the carbonation front moved through the sample, with initial local microcracking that was subsequently infilled by carbonate precipitation. Potentially, the many thin ‘walls’ of carbonate precipitate might reduce sample permeability (see Purser et al., 2013; Rochelle et al., 2013). Reductions in permeability could be beneficial in terms of containment, as they would act to limit radionuclide migration. At a somewhat larger (millimetric) scale, there was some evidence that unconfined samples could develop stress cracking in the partly-carbonated cement close to the carbonation reaction fronts, though these features were very localised and could be filled by secondary carbonates as the reaction fronts moved through the sample. It is possible however, that there might be a dynamic zone of increased permeability that moves just ahead of the main carbonation front.

Carbonation resulted in a reaction zone several mm wide, within which were several reaction fronts delineating 4 main reaction zones. These were:

Zone 1: Unreacted cement

Zone 2: Partially carbonated cement

Zone 3: Fully carbonated cement

Zone 4: Leached cement (where slightly acidic CO₂-rich water had re-dissolved some of the secondary carbonate)

The visually most apparent reaction front occurred between Zones 2 and 3. Note that Zone 4 probably formed relatively early in the experiment when waters surrounding the NRVB samples were not fully saturated with carbonate minerals, and as such may not occur in a repository setting. It is apparent however, that multiple reaction zones are likely to form within a repository setting as a consequence of carbonation.

Carbonation resulted in the degradation of the high pH buffering capacity of the NRVB cement. Given that many metallic radionuclides have low solubility under alkaline conditions, but increased solubility as pH decreases, carbonation may potentially lead to increased metal corrosion rates and consequent increased potential for radionuclide migration. That said, if as seems likely, the repository contained an excess of cement that was more than sufficient to react with all the produced CO₂, overall pH buffering to alkaline conditions could still be effective for many parts of the repository. Whilst we did not measure the porewater pH in the fronts directly, their presence can be determined by the abrupt nature of the reactions fronts and removal of portlandite and CSH phases due to the migration of lower pH conditions associated with the CO₂.

In the experiments carbonation and uptake of CO₂ increased the weight of the cement by up to 8.5%. This has two potential benefits:

1. Reduction in potential for pressure increases due to gas production due to consumption of CO₂.
2. As some wastes contain ¹⁴C, it provides a mechanism for released ¹⁴CO₂ to be immobilised in secondary carbonate minerals.

One less expected observation concerning solute migration was the enhanced localised uptake of dissolved chloride (Cl^-) by the cement. ³⁶Cl presents particular issues within repositories, due to its long half-life, ease of uptake into biological systems, and relative mobility in many settings. Uptake of Cl^- was most clearly revealed by significant decreases in dissolved Cl^- in the experiments using more ‘evolved’ porewater compositions (i.e. porewaters reflecting interaction between saline groundwater and the cement) (see Figure 13).

Many of the experiments showed decreases in dissolved Cl^-, not just those pressurised with CO₂. The NRVB cement appeared to be immobilising Cl^-, especially at lower temperatures (in this study 20°C) where Cl^- uptake was especially favoured. Smaller, but still significant, amounts of additional Cl^- were taken up in equivalent experiments pressurised with CO₂. The additional stability provided by the presence of CO₂ possibly suggests that it may exist as part of a solid- solution series with Cl- and CO₃-endmembers.

Detailed mineralogical observations in the region around carbonation fronts revealed the presence of small amounts of two fine-grained, Cl-rich solid phases. One was a gel-like Cl-rich CSH, the other was only found in the 20°C experiments and occurred as fine-grained radial fibrous crystal aggregates of a calcium chloroaluminate phase, probably hydrocalumite (Ca₄Al₂O₆Cl₂.10H₂O) (Rochelle et al., 2013; Milodowski et al., 2013) (Figure 16).

Figure 16 BSEM photomicrograph of radial fibrous secondary calcium chloroaluminate phase. This formed within altered cement just behind the leading edge of the carbonation front.

Element mapping clearly showed that these Cl-rich phases were found on the largely unreacted cement-side of the reaction front (Figure 17). It would appear that they are only stable under higher pH conditions (i.e. where CSH phases were still present). It is thought that they have progressively broken down as the carbonation reactions progressively consume the CSH around them. The released Cl^- then diffused towards the remaining CSH and re-precipitated. This dissolution/precipitation process would continue until all of the CSH is reacted (in the case of a limited quantity of cement), at which point the Cl^- would be released back into solution.

Figure 17 High resolution images of a cement carbonation front. Note the abundance of Cl on the partly-carbonated side of the reaction front, and its near absence in the fully-carbonated cement.

The uptake of Cl^- in CO₂-free cement systems concurs with observations undertaken as part of previous experimental buffer/backfill cement studies (Glasser et al., 1998). Cl^- uptake in CO₂-rich systems has also been previously noted as part of borehole stability studies for the deep underground storage of CO₂. Rochelle et al. (2006, 2009) found increased Cl^- uptake in CO₂-rich experiments compared to CO₂-free experiments, and Carey et al. (2007) report a Cl-rich secondary phase in recovered samples of borehole cement that had been exposed to CO₂-rich fluids for 30 years.

The apparent likely occurrence of Cl-rich phases within cementitious repositories, raises the question of whether performance assessment calculations could include them (many current approaches assume conservative, i.e. non-reacting, behaviour of Cl^-). Inclusion of such phases would likely retard the overall predicted migration of ³⁶Cl, lowering eventual releases to the biosphere, and hence improving safety calculations. Equally, they may allow for higher initial loadings of ³⁶Cl in the waste. It would be useful to investigate these Cl-rich phases further.

Carbonation features and secondary phases observed in these experiments using a relatively porous/permeable cement, bear many similarities to those found in far lower porosity/permeability borehole cements used in CO₂-storage operations (e.g. Rochelle and Milodowski, 2013). There are also similarities to samples of naturally-occurring CSH phases which have been naturally-carbonated over prolonged timescales (Milodowski et al., 1989, 2009, 2011). A number of common carbonation processes may be operating in all these systems, and consideration of all these sources of information is needed to help provide an overall picture of cement carbonation over a range of temporal and spatial scales.