Thirty years ago, the Columbia River Basalt (CRB) in the western USA (mainly Washington State; Fig. 1) was a candidate for a high-level nuclear waste repository and possibly will be again, as a novel CO2 sequestration project has generated new interest in this attractive target for underground storage of natural gas, supercritical carbon dioxide and solid nuclear waste. On April 14, 2011, the US government decreed to stop funding the construction of the repository for spent nuclear fuel and other types of high-level nuclear waste in the Yucca Mountain, Nevada. Some days before that, Washington State granted a permit to inject 1,000 t of carbon dioxide into the CRB at Wallula, only a few kilometres from the Hanford Site, the former alternative to the Yucca Mountain repository. This comes at a time when the approach to nuclear waste disposal has undergone fundamental changes. The geologists no longer postulate that there exists a geological formation that can isolate the nuclear waste for a million years. Only a technical barrier in concert with the host rock is considered to be capable of completing this task. The waste container should be chemically compatible with the host rock. In the ideal case, the waste container is in thermodynamic equilibrium with the host-rock’s groundwater regime. The essay examines how this can be achieved.

Fig. 1
figure 1

Extent of the Columbia River Basalt (CRB) and location of the Hanford, Wallula and Yucca Mountain sites (USA). WA Washington, OR Oregon, ID Idaho, NV Nevada

Looking back, the CRB at Hanford was the only serious competitor left when eventually Yucca Mountain was designated by the Nuclear Waste Policy Act (US Department of Energy 2004) in 1987 to be the deep geological repository for high-level nuclear waste in the USA. As a consequence, the work on a deep repository in the CRB at Hanford was suspended; however, the CRB with its low-salinity groundwater regime offers what Yucca Mountain certainly has not: a geochemical environment in which metallic waste containers can survive without serious damage by corrosion.

The Yucca Mountain repository is unique because it is conceived for waste storage above the water table (Farmer et al. 2003). In other parts of the world, potential repositories are located below the water table (Schwartz 1996, 2008). The unsaturated zone precludes the protection of the metallic waste container by bentonite because the clay minerals would dehydrate and lose their isolating properties; thus, an entirely metallic barrier system has to be used. Evaporation of water from the metal barrier’s surface has been found to produce salinities up to 4 mol/L Cl in the water remaining on the barrier’s surface under the strongly oxidising conditions in the arid climate zone. Any kind of metallic barrier system is bound to fail in such an environment, long before the radioactive interior becomes harmless. In this respect, the Yucca Mountain project is similar to the salt-dome project at Gorleben, Germany (Schwartz 2012a), started in 1977 and ended in 2013 (German Federal Government 2013).

Worldwide, there are only two approved high-level waste repositories under construction—both the Forsmark project in Sweden (Schwartz 2012b) and the Olkiluoto project, Finland, are in a granite-dominated low-salinity environment and envisage the storage of copper-shielded waste containers. Compared with basalt, granite has lower concentrations of sulphur and higher Fe-III/Fe-II ratios, which is favourable when it comes to corrosion resistance of the copper shield, and low concentrations of copper (unfavourable with respect to resistance). Within a subaerially erupted basalt sequence (e.g. CRB), the top of an individual basalt flow has lower concentrations of sulphur and higher Fe-III/Fe-II ratios than the interior of the flow. Degassing at the top of the flow reduces sulphur concentrations and the reaction with atmospheric oxygen increases the iron-III/iron-II ratio.

The high-porosity tops of basalt flows are attractive targets for underground storage projects, not only for nuclear waste but also for natural gas (Reidel et al. 2002) and carbon dioxide (McGrail et al. 2011, 2014; McGrail and Schaef 2015; Matter et al. 2016). The reasons for the attractiveness, of course, are different. The top of a subaerially extruded sequence not only has low sulphur concentrations and high Fe-III/Fe-II ratios but also a high proportion of glassy phases containing calcium, magnesium and iron silicates that quickly react with wet CO2 to form stable carbonate minerals and amorphous SiO2; however, this is only one side of the story. Carbon-dioxide injection is like a large-scale hydrology experiment at its extreme. The huge volume of injected fluid causes high-speed groundwater movements, which are conveniently monitored at the surface by geophysical methods. The prediction of groundwater movement during the envisaged life of a nuclear waste repository, on the other hand, is one of the major tasks of a security analysis. Consider that a carbon-dioxide injection project in the CRB can be seen as a hydrology data collection event with a built-in time accelerator, which is convenient for projections concerning groundwater and nuclear waste disposal. The case of the deep CRB groundwater is especially interesting because it is very old: >30,000 years or even >100,000 years according to 14C or helium data, respectively (Reidel et al. 2002). Chemically, the old groundwaters are characterised by high pH as well as high sodium and fluoride concentrations, whereas calcium and magnesium concentrations are low. Both the Hanford Reference Repository and the Wallula CO2 injection zone (Table 1) are in typical deep CRB groundwater (Reidel et al. 2002; McGrail et al. 2009; Lavalleur 2012), where formation temperature ranges from 37 to 54 °C.

Table 1 Simplified stratigraphy of the CRB Group (Reidel et al. 2002) and depth ranges of the Reference Repository (Rockwell Hanford Operations 1982) and Wallula CO2 injection zone (McGrail et al. 2011)

Among the various candidate materials for nuclear waste packages, copper has unique oxidation characteristics. The conversion from native metal (Cu0) to metal oxide (Cu2O) occurs in a mildly oxidising to mildly reducing environment (positive Eh for pH <7.7 at 25 °C; Fig. 2), whereas the conversion of steel, titanium or nickel-chromium alloy is not only possible under oxidising conditions but also under strongly reducing conditions, i.e., below the stability field of liquid water. The example of naturally occurring native copper in the Earth’s crust must be taken into account in the disposal site design. The largest native copper deposits are located on the Keweenaw Peninsula of Michigan (mid-west USA), where they have been mined to a depth of 2.2 km (Butler and Burbank 1929; Broderick et al. 1946; White 1968). These deposits formed 1 billion years ago at temperatures of 100–200 °C. The mineralising fluids were slightly less reducing but had much higher copper concentrations than commonly present-day groundwater.

Fig. 2
figure 2

Eh-pH diagram for the system Cu-(Fe)-S-Cl-O-H at 25 °C and 0.1 MPa. The thermodynamic data are calculated with the SUPCRT92 program (Johnson et al. 1992) except for CuCl0 (Xiao et al. 1998) and FeSO4(c) (Hemingway et al. 2002). Two types of Cu0-Cu2S boundaries are shown: the black lines refer to the Fe-free system with ΣS of 10−2 and 10−5; the red interrupted line shows the limit of the Cu0 stability field for ΣS = 10−5 in the presence of hematite and FeSO4(c). Note that the position of the Cu0-CuCl0 boundaries (for ΣCu/Cl of 10–3.5 and 10−5) slightly varies with total salinity; this is not explicitly shown in the diagram but the thickness of the boundary lines is designed in such a way that it is wider than required for the salinity range of 0.005–0.05 mol/L Cl

In low-salinity groundwater (0.0005–0.05 mol/L Cl), CuCl0 is the dominant aqueous copper species. Accordingly, the thermodynamically relevant parameter is the ΣCu/Cl activity ratio, which refers to the sum of activities of all dissolved copper species divided by the Cl activity. Groundwater in the Deccan Basalt, India, has ΣCu/Cl ratios in the range from 10–3.5 to 10−5 (Muralidhar and Raju 1990). Similar conditions are achievable in the bentonite buffer around a copper container for non-basalt repositories if the bentonite contains artificial admixtures of fine-grained native copper. Thus thermodynamic equilibrium between the container and the external fluid on the high-Eh side of the native-copper stability field is implemented in the repository-container system as an analogue to the natural counterparts.

The low-Eh and low-pH side of the native-copper stability field is the copper-sulphide phase boundary (CuS and CuS2). The position of this boundary depends on the total activity of aqueous sulphur species (ΣS), which is ≤1.5 × 10−3 in CRB groundwater (Grande Ronde Basalt Formation; Reidel et al. 2002). Provided the bentonite backfill contains no sulphur-bearing minerals, the corrosion of the copper container is constrained by the diffusion of aqueous sulphide from the igneous rock aquifer through the bentonite to the container surface. This situation is quite different from corrosion of other package materials such as steel, titanium or nickel-chromium alloy, which all corrode by reaction with omnipresent water. It is not possible to control corrosion of these materials by a diffusion barrier around the container.

In the case of a sulphur-free buffer, general corrosion of the copper container can be determined by a simple calculation. The experimentally derived diffusion coefficient of aqueous sulphide in bentonite is approximately 7 × 10−8 cm2/s (King and Stroes-Gascoyne 1995; King et al. 2002). The CRB groundwater has an average sulphide concentration of 1.1 × 10−4 mol/L. Assuming a 35-cm-thick bentonite buffer, 2 × 10−16 mol per second and per square centimetre are transported through the buffer to the container surface. This is equivalent to a general corrosion rate of 10−7 cm/a, when Cu0 is transformed to Cu2S; thus, the corrosion depth is 0.1 cm within 1 million years and a service time for a 5–10-cm-thick copper shell beyond 1 million years is a realistic possibility.

The situation can be further improved if hematite is added to the bentonite buffer. Such a measure would expand the stability field of native copper to lower Eh in the presence of FeSO4(c) or pyrite by 0.07 V or 0.05 V, respectively. The average CRB groundwater with Eh = −0.3 V and pH = 9.4 would be in the stability field of native copper, and corrosion would be nil. The scenario replicates the natural process: the precipitation of native copper is linked to the dissolution of hematite in the Keweenaw deposits.

Paradoxically, the old CRB project (1968–1987) lives up to modern scientific standards. The more recent Yucca Mountain project (1978–2011) with its extremely high corrosion rates does not even come near those.