Introduction

Combating the worsening impacts of climate change requires urgent implementation of effective decarbonization strategies that can operate at commercial scales1. Indeed, investments in carbon capture technologies, including direct air capture and point source capture, have risen exponentially in the past few years2. However, the success of these technologies at commercial scale will rely largely on efficient implementation of sustainable carbon storage strategies3,4,5. Geologic carbon sequestration, via rapid mineralization in Fe-Ca-Mg rich subsurface formations such as flood basalts, permanently stores carbon dioxide as stable carbonates and ensures little to no leakage risk over near-term to geologic timescales, thus requiring minimal site monitoring efforts and significantly decreasing the cost of implementation5,6,7,8,9,10. Given the widespread abundance of such rock formations around the world, they represent an enormous untapped resource to tackle climate change. For instance, the Columbia River Basalt Group, in the United States, has a CO2 storage potential greater than 100 gigatons6 – the equivalent of ~20 years of CO2 emissions in the United States, given present rates11.

Two field-scale pilot studies have demonstrated successful rapid CO2 mineralization in basalt reservoirs, one in the Pacific Northwest of the United States (Wallula, WA) and another in Iceland (Carbfix, hf.)7,12,13. At the Wallula Carbon Storage Demonstration, ~977 metric tons of dry supercritical CO2 (scCO2) was injected in the Indian Ridge and Ortley members of the Grand Ronde Basalt formation. Core sections retrieved from the injection site two years later revealed carbon mineralization in the form of spherical nodules within basaltic vugs12,14,15. Bulk compositional analyses of these nodules, using micro X-ray fluorescence (μ-XRF) and scanning electron microscopy based energy dispersive spectroscopy (SEM-EDS), are consistent with Ca-Mn-Fe carbonates largely devoid of Mg, while X-ray diffraction of crushed nodules identified ankerite, Ca(Fe,Mn)(CO3)212,14,16. However, the nodule is zoned, with a Ca/Mn-rich interior and an Fe-rich exterior rim, indicating an evolving mineralogy during carbonation, potentially consisting of multiple phases16. Since the accuracy of geochemical models depends heavily on reaction endpoints, it is critical to characterize all phases, as well as their relative timing, during carbon sequestration in reactive reservoirs.

To understand the mineralogical evolution during carbonation, the compositions and microstructures of carbonate nodules extracted from the Wallula field demonstration site were characterized from the micrometer- to the atomic-scale. The results reveal striking mineralogic variations in lattice structure, composition, and morphology from early- to late-stage carbonate growth. Unexpectedly, we discovered a Mg-absent cation-ordered ankerite phase that formed at near-ambient temperatures (~40 °C) within 2 years post injection. This contrasts with established wisdom that ordering of double carbonates in low-temperature geologic settings requires thousands of years17. This newly discovered ordered ankerite phase, not yet synthesized in the laboratory nor found in nature18, could control the thermodynamics of carbonation during CO2 sequestration in basalt reservoirs.

Results

Carbonate nodules (Fig. 1A) were extracted from core sections retrieved from 857.1 m (2812 ft) depth below ground surface (bgs) at the Wallula injection site (Supplementary Fig. 1), mounted in epoxy, and cross-sectioned for SEM-EDS analysis. No textural/morphological variations were detected across a nodule (Fig. 1B and Supplementary Fig. 2); however, SEM-EDS (Supplementary Fig. 3) and µ-XRF (Supplementary Figs. 4 and 5) revealed compositional zonation and that Mg was mostly absent except for trace quantities near the periphery. These results are consistent with prior studies speculating that the core region has a manganoan ankerite-like composition and the rim region resembles a calcian siderite14,16. Such compositional zonation is common in the medium-to-large (200 µm to 1 mm) carbonate nodules; consequently, one representative nodule was selected for further characterization15,16.

Fig. 1: Morphological and chemical analysis of a carbonate nodule extracted from a depth of 857.1 m (2812 ft) bgs.
figure 1

A Schematic illustration of CO2 injection showing a photograph of a recovered core section from Zone 2. Inset: Optical microscope image revealing small-to-medium nodules within pore spaces. B SEM secondary electron image of the cross-sectional surface of an extracted carbonate nodule. The highlighted areas (‘core’ and ‘rim’) represent the regions analyzed by FIB-STEM. C EBSD grain map of the nodule where different colors represent specific crystal orientations. D SEM-EDS line scan, measuring the composition along the white arrow shown in (B).

Full size image

Electron-backscatter diffraction (EBSD) reveals that the core of this nodule is comprised of two (~150 µm) distinct rhombohedral-like regions that appear to be aggregates (Fig. 1C). The rim, in contrast, contains randomly oriented grains that are much smaller (<10 µm) in size, likely pointing to differences in nucleation rates between early- and late-stage crystallization. Scanning transmission electron microscopy (STEM) was used to probe the atomic structure, lattice ordering, defect architecture, and chemical composition corresponding to the core and rim regions. Thin lamellae (~50 nm) were prepared by conventional focused ion-beam milling (FIB) techniques in SEM, followed by argon ion nano-milling (see ‘Methods’ section and Supplementary Fig. 6).

We first examined the lamella corresponding to the large crystallographically aligned regions in the core region of the carbonate nodule by transmission electron microscopy (TEM), to find this is in fact composed of smaller crystallites at the nanoscale (Fig. 2A, B). The only elements detected by STEM-EDS were Ca, Mn, Fe, C, and O, consistent with prior studies16 and SEM-EDS analyses (Fig. 1D), suggesting an Mg-absent Ca/Mn-rich ankerite composition Ca1.0[Ca0.05Fe0.81Mn0.14](CO3)2 (Table 1). However, STEM-EDS mapping revealed two compositionally distinct phases within the core that were previously undetected (Fig. 2C and Supplementary Fig. S7). Constituting more than 65% of the total analysis area, the major phase is a Mn-bearing ankerite with the composition Ca1.0[Mn0.15Fe0.85](CO3)2 (‘Mn-ankerite’ in Fig. 2C). The minor phase is richer in Ca, with a composition resembling a calcium-rich ankerite, Ca1.0[Ca0.14Mn0.13Fe0.73](CO3)2 (‘Ca-ankerite’ in Fig. 2C).

Fig. 2: TEM analysis of the Ca/Mn-rich core region of the carbonate nodule.
figure 2

A HAADF-STEM image showing microstructure of the core region. B Higher-resolution HAADF-STEM image of the region highlighted in (A). C EDS elemental map highlighting compositional variations in (A). Compositions corresponding to the highlighted boxes are in Table 1. D Atomic-resolution HAADF-STEM image of the major phase. (inset) digitally magnified HAADF image showing lattice spacings. E FFT of the HAADF-STEM image showing ordering reflections.

Full size image
Table 1 Composition of mineral phases in the carbonate nodule determined using STEM-EDS
Full size table

Atomic-resolution high-angle annular dark-field (HAADF)-STEM images of the Mn-ankerite phase, along the [100] direction of ankerite, reveal a highly crystalline lattice (Fig. 2D and Supplementary Fig. 8). A fast Fourier transform (FFT) of the HAADF-STEM image (Fig. 2E and Supplementary Figure 9) shows ordering reflections (0kl, l = odd), indicating that the centrosymmetric nature of the calcite structure is broken via ordering of the cation lattice into an alternating ABAB stacking pattern, where A and B represent large and small cations, respectively. Consequently, the structure is best described as having rhombohedral symmetry with lattice parameters a = b = 4.96 Å, c = 16.67 Å, and the space group R\({\bar{3} }\). Disordered CaFe(CO3)2 (space group R\({\bar{3} }\)c), in contrast, has a smaller unit cell in agreement with studies showing cation ordering leads to an expansion of the unit cell18. This Mg-absent ordered ankerite phase has not been observed in nature or in the laboratory18,19.

Unlike the core, a layered microstructure is observable in a low-magnification STEM image of the ‘rim’ region (Fig. 3A), likely representing crystal growth fronts. Such textures were observed in early TEM studies on rhombohedral carbonates, such as smithsonite20. STEM-EDS mapping revealed compositional variation perpendicular to the growth direction; that is, the major phase is Fe-rich (Fe0.6Ca0.37Mn0.02Mg0.01CO3 – ‘Ca-siderite’ in Fig. 3B), while the minor phase has nearly equimolar Fe and Ca (Fe0.51Ca0.47Mn0.01Mg0.01CO3 – ‘Ca-Fe-ankerite’ in Fig. 3B). The average composition of the rim region determined from low-magnification STEM-EDS closely agrees with SEM-EDS (Fig. 1D) and prior XRF studies16. Again, Mg is only a trace component (<1 at%), with slightly higher amounts in the Fe-rich major phase compared to the endmember ankerite-like minor phase (Table 1).

Fig. 3: TEM analysis of FIB lift-out from Fe-rich outer rim of carbonate nodule.
figure 3

A HAADF-STEM image showing layered microstructure. Arrows represent direction of carbonate growth fronts. B STEM-EDS elemental map showing compositional variability across the sample. Compositions of the two highlighted regions are shown in Table 1. C Atomic-resolution HAADF-STEM image of Fe-rich major phase. D FFT of the HAADF-STEM image showing the diffraction pattern with no ordering reflections. E iDPC-STEM image of the Fe-rich phase showing the carbonate groups in addition to the metal lattice. Brown: Ca/Fe, purple: C, red: O. F Atomic-resolution HAADF-STEM image of minor phase labeled ‘Ca-Fe-ankerite’, and G FFT of the HAADF-STEM image showing subtle ordering reflections (arrowed).

Full size image

Atomic-resolution HAADF-STEM (Fig. 3C) and the accompanying FFT (Fig. 3D and Supplementary Fig. 9) indicate the major Fe-rich phase has a highly crystalline lattice that, in contrast to the major phase in the core region, lacks cation ordering. Based on the relative cation composition (Fe>Ca), this phase is best described as a Ca-rich siderite with a space group of R\({\bar{3} }\)c and lattice parameters a = b = 4.83 Å, c = 16.05 Å (vs a = b = 4.69 Å, c = 15.38 Å for pure siderite)21. In addition to visualizing the metal lattice, the carbonate groups were resolved using integrated differential phase contrast STEM imaging (iDPC-STEM) (Fig. 3E)22. Unlike HAADF-STEM, where the contrast is proportional to Z2, the contrast in iDPC-STEM mode is roughly proportional to Z, which enables imaging of light atoms including C, O, and N23. Though used in material science applications, this technique has only been sparingly used for geochemical and mineralogical studies24,25. Aside from the novelty of this being only the second time that carbonate groups have been imaged by TEM, direct observation of carbonate groups provides further confidence in our structural determination of this phase26.

Electron diffraction of the interfacial region between successive growth fronts shows a large intersection angle of about 11.7° (Supplementary Fig. 10C). These few-micron wide misoriented growth fronts cover the entirety of the outer rim and are consistent with the SEM-EBSD map (vide infra) (Fig. 1C) that reveals randomly oriented grains at the cross-sectional surface. Such stacking faults likely accommodate the lattice strain within the calcian siderite structure due to an excess of the larger Ca2+ ion (1.0 Å) compared to Fe2+ (0.78 Å). At the nanoscale, this internal strain yields cracks/voids preferentially along the [01\({\bar{4} }\)] direction, leading to subtle dislocations in the crystal structure with tilt angles of about ~1° (Supplementary Fig. 10D, E).

An atomic-resolution HAADF-STEM image of the minor phase, Ca0.94[Fe1.02Mn0.02Mg0.02](CO3)2, represented by the cyan regions in Fig. 3B, shows a highly crystalline lattice (Fig. 3F). An FFT of the HAADF-STEM image (Fig. 3G and Supplementary Fig. 9) shows subtle ordering reflections, indicating that the minor phase is at least partially ordered with lattice parameters (a = b = 4.93 Å, c = 16.59 Å) that are close to those of the ordered manganoan ankerite phase in the core region but appreciably larger than those for the adjacent disordered Fe-rich major phase. The expansion of the lattice is due to the presence of a greater amount of the larger cation, Ca2+, and possibly cation ordering. While kinematical effects in diffraction could lead to apparent ordering due to twinning in an R\({\bar{3} }\)c-type lattice27,28, this scenario is unlikely in the present study (specifically along the [100] axis) as no such twinning is observed in the atomic-resolution HAADF-STEM images from which the FFTs are derived (Supplementary Fig. 8A, C).

Discussion

This study identifies four distinct carbonate phases in the zoned carbonate nodule extracted from interflow Zone 2 (Supplementary Fig. 1) of the Indian Ridge Member of the Grand Ronde Basalt formation. Cation lattice ordering is observed for phases where the ratio between large and small cations is approximately unity (Table 1). Given that very similar compositional profiles are widespread in the Wallula demonstration project, we pose the question whether ordered ankerites could be a common endpoint phase during carbon sequestration in basalt reservoirs. How sensitive these phases are to variations in basalt chemistry and injection conditions remains an open question.

The inability to synthesize an ordered endmember ankerite in the laboratory or find one in nature has been a longstanding enigma among mineralogists given that several ordered Ca-based double carbonates, such as CaM2+(CO3)2 (M = Mg, Mn, Zn), exist naturally18,19. However, their formation in geologic settings is usually associated with elevated temperatures, and it is speculated that it could take thousands of years at ambient conditions. Indeed, they have not been synthesized in the laboratory unless moderate-to-high temperatures are employed (i.e., ≥~80 °C)17,29,30. So far, attempts to synthesize an ordered Mg-absent ankerite in the laboratory, even under hydrothermal conditions, have also been unsuccessful19,31. To this end, several favorable factors likely contributed to the nucleation and growth of these complex carbonate phases at the Wallula demonstration site. With respect to the compositional zonation, fluid sampling studies at the pilot site showed elevated concentrations of carbonate-forming cations, such as Ca, Fe, Mn, and Mg14. It is well known that water exchange rates for divalent cations follow the order Ca>Mn>Fe>Mg, which tracks the compositional zonation from the core to rim of the carbonate nodules32. Consequently, we hypothesize that in an aqueous-dominated scCO2 environment, such as at Wallula, competitive cation dehydration rates play a rate-limiting role in determining the compositional zonation of the nodules. With respect to the low concentrations of Mg in the nodules, even near the rims, enthalpy of hydration values indicates that a hydrated Mg2+ complex is thermodynamically more stable than analogous Ca2+ and Mn2+ complexes33. At the Wallula site, the temperature at 857.1 m bgs was ~40 °C12, which likely inhibited the formation of Mg-based carbonates34,35. Indeed, it is plausible that Mg-based carbonate phases would precipitate when CO2 is sequestered in higher geothermal gradient terrains, for example in mid-ocean ridge basalts, or during injection at much greater depths (>2 km) in continental flood basalts.

Collectively, the results present a more complete picture of complex carbon mineralization processes in a basalt reservoir. The injection of scCO2 results in the dissolution of host rock components such as glass, pyroxenes, feldspar, and olivine. These dissolution processes are known to be temperature and pH dependent, and fluid sampling studies show the release of carbonate -forming cations, Ca2+, Mg2+, and Fe2+ into the pore fluid8,14,36,37,38. In addition, Mn2+ is released via the dissolution of oxyhydroxide/oxide phases from pore-lining alterations in basalt that formed due to long-term groundwater/basalt interaction16. This creates localized regions with high concentrations of Mn2+ near the pore lining, thus kinetically favoring its early incorporation in carbonate phases while inhibiting the nucleation of calcite39. The length of the Mn-rich region in the SEM-EDS line scan correlates well with the lateral size of the large crystallites observed in the EBSD grain map (Fig. 1C, D), suggesting that the incorporation of Mn2+ in the ankerite lattice could be playing a key role in stabilizing a cation-ordered lattice structure. This hypothesis can be tested in future long-term core-scale batch reactor studies.

As observed by EBSD (Fig. 1C), the formation of the two ankerite aggregates within the core of the nodule suggests two near-simultaneous nucleation events. It follows that, over time, as the availability of Mn2+ decreases, the structure shifts from the largely ordered manganoan ankerite to the disordered calcian siderite. Preferential precipitation of siderite from acidic CO2-brine containing a mixture of Ca2+, Mg2+, and Fe2+ has been documented in prior studies35, where magnesite was only observed at high temperatures and high Mg2+ supersaturation with respect to other competing cations (e.g., aMg2+/aCa2+ > 234). An ordered near endmember ankerite (~ CaFe(CO3)2) is also present in the rim region, intergrown with the major Ca-rich siderite phase. While this minor phase has the appearance of exsolution lamellae, the low ambient temperatures render this unlikely40. It is difficult to pinpoint a single cause for the stabilization of ordered near endmember ankerite due to the inherent environmental complexity in the subsurface. However, unlike dolomite, where high temperatures or dissolution-reprecipitation processes over thousands of years lead to cation ordering17, the sequestration conditions (and timescales) at the injection site render this scenario unlikely for ankerite. We therefore speculate that the low ambient temperature played a primary role in the stabilization of this ordered ankerite phase. Furthermore, this agrees with prior studies hypothesizing that an ordered endmember ankerite could only exist, if at all, at low-temperatures18,31. Whether additional factors, such as possible growth on the adjoining Ca-siderite or simply the longer-than-usual reaction times (i.e., ≤2 years) compared to laboratory studies, helped promote the growth of these phases is a subject of further investigation41

This study provides unique insight into the formation of anthropogenic carbonates in deep basalt reservoirs as part of a pilot demonstration of geologic carbon sequestration. We accurately identified the various mineral phases, microstructures, defects, and textures within the sequestration outcomes using atomic-resolution microscopy. In turn, this enabled pinpointing of mineralization endpoints, such as ordered Mn-ankerite, Ca-siderite, and endmember ordered ankerite. Notably, both the ordered Mn-ankerite and ordered ankerite have not been documented in nature or synthesized successfully in the laboratory. These insights are critical to enable better predictive geochemical modeling for carbon sequestration at the field scale, refinement of capacity estimates, and enhanced techno-economic analysis. Coupling this knowledge to the widespread occurrence of continental flood basalts brings the community closer to using this resource for commercial-scale carbon sequestration.

Methods

Preparation of polished cross-section of carbonate nodules

Several core-sections were retrieved from the Wallula Basalt Carbon Storage demonstration site two years post-CO2 injection12. Site selection and geological provenance has been reported earlier42. From the core-section corresponding to 857.1 m bgs (2812 ft), carbonate nodules were manually extracted. These were then epoxied to cavities in Al holders such that half of the nodule was exposed above the cavity of the aluminum holder. This allowed the Al holder to serve as a mask during cross-section polishing of the nodule, which was performed on a JEOL IB-09010CP cross section polisher (JEOL USA, Inc.). Then ion milling was performed at an acceleration voltage of 6 kV using Ar+-ions. The sample was mounted 90° to the ion beam and was continuously rotated during the milling process. The sample was milled until the surface of the nodule was in plane with the top of the Al holder.

Electron-backscatter diffraction (EBSD) in SEM

The polished sample was sputter-coated with 2.5 nm of Ir to provide a conductive surface and avoid charging. Samples were analyzed using a JSM-7001F field-emission gun scanning electron microscope equipped with a Bruker Xflash 6ǀ60 EDS and Bruker e-Flash electron backscatter diffraction detector (EBSD) to collect elemental maps and EBSD patterns. To determine the grain structure of the carbonate nodules EBSD analysis was performed using beam conditions of 30 kV and 50–70 nA, EBSD camera resolutions were 320 × 240 and 400 × 300 pixels, and collection times 14.9 and 13.3 fps, for maps collected at magnifications of 330X and 4300X, respectively.

Scanning electron microscopy and energy-dispersive spectroscopy

Scanning electron Microscopy for TEM lamella preparation was performed on a Helios NanoLab 610 DualBeam system. The sample was coated with 2 nm Au coating and surface morphology was analyzed using a secondary electron detector or a solid state backscatter electron detector using an acceleration voltage between 3 and 5 kV and a current between 0.17 and 0.34 nA. Energy dispersive spectroscopy and mapping analysis was performed using an accelerating voltage of 15 kV and a current between 4 and 6 nA. Oxford INCA software was used to collect and analyze point scans and compositional mapping data.

Focused ion-beam milling preparation of TEM lamella

Focused-ion beam (FIB) milling in SEM was performed to prepare two thin TEM lamellae for structural and compositional analysis at the nanoscale. Supplementary Fig. 6 shows the progression of steps to prepare TEM lamellae. A microscopic area of interest (~2 × 30 µm) was first coated by Pt using the gas-injection system in the SEM (Supplementary Fig. 6A), protecting the area from subsequent ion damage during milling. The section was isolated by trenching the sides using a Ga-ion source (Supplementary Fig. 6C). The section was removed from the bulk nodule (Supplementary Fig. 6D) and transferred to Cu TEM grid (Supplementary Fig. 6E). The resulting lift-out (shown in cross-section in Supplementary Fig. 6F) was then further thinned to <100 nm by Ga-ion milling to be electron transparent, as needed for TEM analyses (Supplementary Fig. 6G).

Argon ion nano-milling

The FIB lamellas were ion milled with Fischione 1040 Nanomill to remove surface damage imparted by FIB preparation. The ion milling was performed consecutively at 900 eV and 500 eV for 10 minutes at each side of the lamella using Argon ions. The milling angle was 10 degrees.

Transmission electron microscopy

TEM analysis was performed with aberration corrected Thermo-Fisher Themis Z scanning/transmission electron microscope (S/TEM) operated at 300 kV. The observations were performed in conventional TEM and Scanning mode using a HAADF and segmented DF detectors. The probe convergence angle was 25 mrad. HAADF images were collected using multi-frame integration. The acquisition and basic image processing were performed with Thermo-Fisher’s Velox software. TEM and diffraction acquisition was performed with CMOS based CETA II. Elemental analysis and mapping was performed with X-ray energy dispersive spectroscopy (EDS) using Thermo-Fisher Super-X Silicon Drift Detector (SDD).