In situ carbonation of peridotite for CO₂ storage

Mantle peridotite is composed largely of the minerals olivine [(Mg,Fe)₂SiO₄] and pyroxene [(Ca,Mg,Fe)₂Si₂O₆], which react with H₂O and CO₂ near the Earth's surface to form hydrous silicates (serpentine), Fe-oxides (magnetite), and carbonates (calcite, magnesite, and dolomite). Such reactions may generally be formulated as:Evidence for natural, low-temperature hydration and carbonation of mantle peridotite can be found in springs and associated travertines in catchments composed of mantle peridotite (12–19), and in outcrops of altered peridotite with abundant carbonate veins (e.g., refs. 20–26). High alkalinity, stable isotope ratios, and formation of travertine and carbonate cemented conglomerates in springs (Fig. 1A) indicate ongoing serpentinization involving meteoric water at low temperature. In addition to travertine at springs, carbonate veins are also found within host peridotite (Fig. 1 B and C).

Vein and travertine formation are linked (e.g., refs. 15–19). Groundwater reacting with peridotite in near-surface, open systems forms water rich in Mg and HCO₃⁻, which we call Type 1 waters, according to Barnes and O'Neil (18). When these waters become isolated from the atmosphere, continued reaction with peridotite leads to precipitation of abundant magnesite and dolomite as veins; the resulting waters become progressively richer in Ca and OH⁻, and impoverished in dissolved carbon, approaching a pH of 12. When these Ca-OH⁻-rich, carbon-poor, Type 2 waters emerge near the surface, to mix with Mg-HCO₃⁻ waters or react with the atmosphere, they precipitate abundant calcite and dolomite in near-surface veins, carbonate cement in unconsolidated sediment, and travertine.

The rate of CO₂ uptake via weathering of peridotite is poorly known. We sampled solid carbonate forming from peridotite over a wide area in the Samail ophiolite [Fig. 2 and supporting information (SI) Table S1], including veins from ridges far from present day springs as well as currently forming travertine. Previous workers inferred that most veins far from present-day springs are 30–90 million years old, related to formation of oceanic crust, emplacement of the ophiolite, and Eocene extension (e.g., refs. 15, 21, 22, 27). However, all of our samples have ¹⁴C ages from 1,600 to 43,000 years, similar to the previously measured range of 840 to 36,000 years in the vicinity of a single, actively forming travertine in Oman (28). Samples of veins from ridges are mainly composed of dolomite and magnesite. In general, they are somewhat older than calcite-rich travertine and calcite-dolomite veins near active springs. However, the vein samples have an average age of ≈26,000 years, with a fairly “flat” age distribution (Fig. 3), and none are too old to date with ¹⁴C.

The observed volume of carbonate terraces and veins in the Samail ophiolite, together with their ages, can be used to estimate the rate of CO₂ uptake via formation of solid carbonate minerals in 2 ways. First, we can estimate the mass of veins directly, and divide this by their average age. Poupeau et al. (29) estimated an erosional denudation rate of ≈0.3 mm/yr for northern Oman. The ages of carbonate veins in peridotite suggest that veins form mainly in a thin weathering horizon that keeps pace with erosion; this horizon must generally be ≈15 m thick (erosion rate ≈0.0003 m/yr · maximum age of carbonate veins ≈50,000 years). Newly created road cuts in Oman peridotites reveal abundant, submillimeter carbonate veins on joint surfaces. We measured the vein abundance as ≈1 vol% in transects along road cuts (Table S2); 1% of the volume of a 15-m-thick weathering horizon in the Oman peridotite corresponds to ≈10¹² kg of CO₂, for an average CO₂ uptake of ≈4·10⁷ kg/yr.

We can independently estimate the mass of travertine formed at and near the surface by alkaline springs, and infer the associated mass of carbonate veins far from the surface that must be formed during recharge of these springs. Based on our relatively detailed, although incomplete, mapping in the southern third of the ophiolite, we estimate that there are ≈45 travertine terraces in the Samail ophiolite (Fig. 2) that are ≈1 m thick, with exposed areas ≈200,000 m² (Fig. S1), comprising a total of ≈10⁷ m³ of exposed travertine. Travertine extends beneath alluvium downslope from outcrop areas, and travertine deposits are underlain by a zone ≈10 m thick with ≈5% calcite-rich veins (Table S2), so that their total volume is probably ≈2.5 times the exposed volume. Near-surface deposits, similar in composition and age to the travertine terraces, occur as massive carbonate bands, veins, and cement in alluvial terraces and conglomerates with peridotite clasts (Fig. 4). The volume of carbonate cement derived from Ca-OH⁻ waters in peridotite sediments is hard to estimate, but is at least as large as the volume of travertine terraces. All of these factors, taken together, suggest that the volume of near-surface travertine and carbonate in peridotite sediments in Oman is ≈5.5·10⁷ m³ or more, corresponding to at least ≈10¹¹ kg of CO₂.

Spring waters and shallow groundwater in peridotite catchments fall into 2 compositional groups, as discussed above and illustrated in Fig. S2. We can estimate carbonation rates from water compositions, assuming (i) all carbon in type 1 waters is consumed to form solid carbonate during formation of type 2 waters, and (ii) the difference in Ca between type 2 and type 1 waters is precipitated as calcite when type 2 waters reach the surface. In California, type 1 waters have ≈0.2 mmol of Ca per liter, and up to 24 mmol of carbon per liter (19). Type 2 waters have essentially no carbon, and ≈1.5 mmol Ca per liter (Fig. S2). Thus, for every mole of calcite near the surface, up to ≈24/(1.5–0.2) or 18 mol of magnesite form in the subsurface.

There is a maximum of ≈8 mmol/L, carbon in Oman Type 1 waters, lower than in California, whereas Ca concentration is ≈0.8 mmol/L, higher than in California. There is essentially no carbon, and ≈1.6 mmol Ca in type 2 waters in Oman. It is not clear whether these values reflect lower carbon concentrations in Oman waters compared with those in California, or whether end-member type 1 waters in Oman have not yet been sampled. If, for every mole of near-surface calcite, 8/(1.6–0.8) or 10 mol of magnesite are precipitated as veins, this yields ≈10¹² kg of CO₂ in veins, consistent with the estimate derived from measured vein abundance and the inferred thickness of the veined horizon.

In summary, estimates of the volume of carbonate deposits formed during ongoing weathering of peridotite, and their average age of 26,000 years, indicate that ≈4·10⁷ kg of atmospheric CO₂ per year are consumed via mineral carbonation in the Samail ophiolite, or ≈2 tons/km³ of peridotite. This strikingly rapid rate is compared with CO₂ flux in rainwater and groundwater, and discussed further in the SI Text. Here, we emphasize that a factor of 100,000 increase in this rate could consume 4 billion tons of CO₂ per year, ≈10% of the annual increase in atmospheric CO₂ because of anthropogenic emissions, via carbonation of peridotite in Oman.

In this section, we propose and evaluate ways to increase CO₂ uptake in situ in tectonically exposed peridotite massifs. In the Samail ophiolite and other large massifs, an obvious approach is to increase the depth of the weathering horizon by a factor of 200, from ≈15 m to ≈3 km in the peridotite via drilling and hydraulic fracture (30). Additional fracture may be anticipated as a result of thermal expansion during heating (31), volume increase during hydration (32–34), and volume increase during carbonation. Carbonation of olivine (Eq. 2b) results in ≈44% increase in the solid volume, which can lead to enormous stresses that may be relieved by cracking and additional expansion (Fig. 1 B and C).

An additional increase in the carbonation rate, by a factor of > ≈10⁶, could be achieved by raising the temperature of the peridotite and injecting CO₂-rich fluids. There is an optimal temperature for peridotite carbonation. Heating from low temperature speeds the diffusive kinetics of hydration and carbonation. However, the chemical potential driving the reaction is reduced as the temperature approaches the equilibrium phase boundary for serpentine or carbonate mineral stability. The combined effect yields a maximum reaction rate at a temperature intermediate between surface conditions and the equilibrium phase boundary (Fig. S3). The reaction rate for serpentinization as a function of temperature has a maximum value at ≈260 °C over a range of pressure (35), whereas the rate of carbonation is optimized at, for example, 185 °C and 150 bars CO₂ pressure.* We fit data on rates of serpentinization of olivine with grain size 58–79 μm (35) and carbonation of olivine with grain size ≤75 μm* as a function of temperature and CO₂ partial pressure, yielding a serpentinization rate (Fig. S4)and a carbonation rate (Fig. S5 and Fig. S6).both in units of mass fraction per second. Heating and raising the partial pressure of CO₂ can increase the carbonation rate by a factor of >10⁶ (Fig. 5), and with the potential for increasing the thickness of the weathering horizon by fracturing, the overall increase could be a factor of ≈10⁹. Together with the estimated present-day CO₂ uptake at the end of the previous section, this corresponds to 2·10⁹ tons/km³ per year.

The change in temperature for a particular volume in a subsurface, porous aquifer can be approximated in 1 dimension aswhere T_in is the temperature of incoming water or aqueous fluid (°C or Kelvin), T is the current temperature in the volume, T_o is the far-field temperature, outside the volume, which is equal to the initial temperature in the volume, ρ_f and ρ_s are the densities of the fluid and solid, C_p^f and C_p^s are the heat capacities of the fluid and solid, φ is the porosity or volume fraction of fluid (nondimensional, 1% in all calculations shown here), w is the fluid flow velocity (m/s), d is the dimension or “size” of the volume (m, 1,000 m in all calculations shown here), κ is the thermal diffusivity (10⁻⁶ m²/s), Γ is the reaction rate, which is a function of temperature (units of 1/s), A is the fraction of the rock available for reaction in the volume (nondimensional), and ΔH is the enthalpy change due to reaction. The use of a reactive volume fraction term <1 accounts for the fact that most natural rocks do not have an effective grain size of ≈70 μm, unlike the experiments used to calibrate Eqs. 3 and 4. The volume available for reaction is product of a diffusion distance times the surface area of grains. The surface area is proportional to the radius squared, so the use of a reactive volume fraction of 0.01% corresponds to modeling of an effective “grain size” or fracture spacing of ≈7 mm, 100 times larger than in the experiments.

For these calculations, densities, heat capacities, and enthalpies were obtained from standard references reviewed and updated by Gottschalk (36) and similar data from the National Institute of Standards (NIST) Chemistry WebBook (NIST Standard Reference Database Number 69, June 2005 Release) at http://webbook.nist.gov/chemistry/. For our calculations, we fit simple functions to the temperature dependence of thermodynamic properties from 25 to 300 °C. For H₂O and CO₂, we used 300 bars pressure for the calculations discussed in this section. Above 300 °C, and far from this pressure, our calculations would be inaccurate.

Fig. 6 illustrates results from Eq. 5 in terms of temperature change, dT/dt, versus initial temperature, for fluid porosity of 0.01 (1%), fluid temperature of 25 °C at 300 bars pressure, with reactive volume fraction A of 0.01%. At high flow rates with cold (25 °C) fluid, the volume cannot be heated by exothermic reactions. At low flow rates, advective cooling is negligible, and temperature is controlled by exothermic heating and diffusive cooling. To optimize olivine carbonation rates, fluid flow should be adjusted to yield dT/dt = 0 at ≈185 °C. Heating due to hydration (serpentinization) is less effective than heating due to carbonation. This is partly because, for example, at1 bar and 25 °C, ΔH is ≈250 kJ/kg for serpentinization (Eq. 1), whereas carbonation (Eq. 2) evolves ≈760 kJ/kg, and partly because serpentinization is slower than carbonation for temperatures between 25 and 185 °C (Fig. 5).

One approach is to take maximum advantage of the exothermic heat output available from the carbonation reaction, by raising a rock volume to the optimal temperature for peridotite carbonation. To reach and maintain 185 °C, it is necessary to preheat the rock volume. This can be achieved via a variety of flow rates, fluid temperatures, and fluid compositions. Initial heating should probably be via high flow rates by using preheated fluids. Later, because large volumes of rock are to be held at 185 °C to optimize CO₂ uptake, output fluid can be used to heat other areas. This may happen spontaneously as hot fluid flows into colder, surrounding rock.

Because pumping rates for 25 °C fluid must remain low to maintain high temperature, dissolved CO₂ in surface water cannot be supplied rapidly enough to keep pace with the enhanced carbonation rates modeled here. Instead, injection of pure CO₂, or a CO₂-rich fluid mixture, is required to keep pace with the enhanced reaction rate. As seen in Fig. 6, dT/dt resulting from carbonation is zero at 185 °C when the flow rate of pure CO₂ injected at 300 bars and 25 °C is ≈0.040 m/s and the reactive volume fraction is 0.01%. At these conditions, our 1-dimensional model delivers ≈0.166 kg of CO₂ per s to a 1 × 1 × 1000 m³ rock volume, and consumes ≈0.127 kg of CO₂ per s to form solid magnesite. All olivine is consumed after ≈190 days, consuming ≈2,000 tons of CO₂ per 1,000 m³, or—scaling up—2·10⁹ tons of CO₂ per km³ at ≈4·10⁹ tons/yr. Note that this is an independent estimate of CO₂ uptake, which is consistent with the rate of ≈2·10⁹ tons/km³ per year derived at the end of the section entitled Enhancing Rates of Peridotite Carbonation in Situ.

Fig. 7 provides an example of a 3-step process, with drilling and hydrofracture of peridotite at depth, followed by injection of hot fluid to heat the newly fractured peridotite to 185 °C, followed in turn by injection of pure CO₂ at 25 °C to sequester carbon whereas exothermic carbonation maintains the system at 185 °C. Our simple calculations show that a factor of more than 1 million enhancement in the carbonation rate is achievable. Note that we have not incorporated the “cost-free” heating afforded by the geothermal gradient; if northern Oman lies along a typical continental geotherm of 10–20 °C/km (37), then the initial temperature at the bottom of a 3-km drill hole will be 55–85 °C, not the initial 25 °C used in our model. Indeed, Neal and Stanger (15) report that alkaline springs in Oman peridotites have temperatures up to 15 °C hotter than normal groundwater in the same locations, and infer that the alkaline waters have been heated during circulation at depths of 700 m or more.

Our calculations are done assuming that the reactive volume fraction is constant, whereas, in practice, reactive surfaces may become depleted. It may be necessary to reduce the fluid flow rate as this occurs, to avoid cooling the reacting volume. At some point, in particular, if reaction-driven cracking does not occur, it may be necessary to hydraulically fracture the system again to expose additional reactive surface area. More optimistically, temperature change and the large increases in solid volume due to mineral hydration and carbonation will cause cracking and increased permeability. In any case, eventually all accessible olivine in a given rock volume will be depleted. Before this occurs, fluid heated by reaction in 1 region can be pumped into an adjacent area to begin the process anew.

An alternative process could avoid prolonged pumping of fluid and use of purified CO₂. In Oman, New Caledonia, and Papua New Guinea, peridotite is present beneath a thin veneer of sediment offshore. Here, peridotite could be drilled and fractured, and a volume could be heated. Again, little heating would be required if, for example, the initial temperature at the bottom of a 5-km bore hole is 100 °C (Fig. 8). Then, controlled convection of near-surface water through the rock volume could sustain high temperature via exothermic hydration of olivine at a flow rate of ≈4·10⁻⁶ m/s (as seen in Fig. 3 Right). The carbonation rate would be limited by supply of dissolved CO₂ in convecting seawater—only ≈10⁴ tons of CO₂ per km³ of peridotite per year at a flow rate of 4·10⁻⁶ m/s—but the cost would be relatively low.

Because these proposed methods of in situ mineral carbonation use the chemical potential energy inherent in tectonic exposure of mantle peridotite at the Earth's surface, the optimal temperature for carbonation can be maintained in a rock volume at little expense. Further, rock volumes at depth are, inherently, at relatively high pressure and elevated temperature. Thus, compared with engineered, mineral carbonation “at the smokestack,” this method does not involve quarrying and transportation of peridotite, processing of solid reactants via grinding and heat treatment, or maintaining high temperature and pressure in a reaction vessel. Instead, the major energy investments in this method would be for drilling, hydraulic fracturing, pumping fluid, preheating fluid for the first heating step, and purification of CO₂. Also, unlike ex situ mineral carbonation, this method may require on-site CO₂ capture or transport of purified CO₂ to the in situ carbonation locality.

Clearly, more elaborate models combined with field tests will be required to evaluate and optimize this method. For example, it is difficult to predict the consequences of hydraulic fracturing of peridotite, plus cracking associated with heating, hydration, and carbonation, in terms of permeability and reactive volume fraction. Such processes are all-but-impossible to simulate in the laboratory. Large-scale field tests should be conducted, because the proposed method of enhanced natural CO₂ sequestration provides a promising potential alternative to storage of supercritical CO₂ fluid in underground pore space, and to engineered, ex situ mineral carbonation.

We thank everyone at the Geological Survey of Oman and the Directorate General of Minerals in the Ministry of Commerce and Industry, Sultanate of Oman, for their generosity, particularly, Hilal Al Azri, Ali Al Rajhi, and Salim Al Busaidi. We thank many friends, especially Karen Benedetto, Martin Collier, Brad Hacker, Karen Hanghøj, Greg Hirth, Sam Krevor, John Rudge, and Lisa Streit for help during the 2007 and 2008 Oman field seasons; Hacker, Hirth, Rudge, Wally Broecker, Mark Ghiorso, Al Hofmann, and Marc Spiegelman for scientific discussions and advice; Benedetto, Hanghøj, Streit, Bill Curry, Kathy Elder, Al Gagnon, Susan Handwork, Rindy Osterman, and Margaret Sulanowska for assistance with (ongoing) sample preparation and analysis; Mike Purdy and David Hirsh for moral and financial support; Dave Walker for editorial advice; and Greg Dipple and an anonymous reviewer for helpful suggestions. This work was supported by a Columbia Research Initiative in Science and Engineering Grant (to P.B.K. and J.M.).