Article pubs.acs.org/est

Mineralization of Basalts in the CO2−H2O−SO2−O2 System Herbert T. Schaef,* Jake A. Horner, Antoinette T. Owen, Chris J. Thompson, John S. Loring, and Bernard P. McGrail Pacific Northwest National Laboratory, Richland, Washington 99352, United States S Supporting Information *

ABSTRACT: Sequestering carbon dioxide (CO2) containing minor amounts of co-contaminants in geologic formations was investigated in the laboratory through the use of high pressure static experiments. Five different basalt samples were immersed in water equilibrated with supercritical CO2 containing 1 wt % sulfur dioxide (SO2) and 1 wt % oxygen (O2) at reservoir conditions (∼100 bar, 90 °C) for 48 and 98 days. Gypsum (CaSO4) was a common precipitate, occurred early as elongated blades with striations, and served as substrates for other mineral products. In addition to gypsum, bimodal pulses of water released during dehydroxylation were key indicators, along with X-ray diffraction, for verifying the presence of jarosite−alunite group minerals. Welldeveloped pseudocubic jarosite crystals formed surface coatings, and in some instances, mixtures of natrojarosite and natroalunite aggregated into spherically shaped structures measuring 100 μm in diameter. Reaction products were also characterized using infrared spectroscopy, which indicated OH and Fe−O stretching modes. The presences of jarosite−alunite group minerals were found in the lower wavenumber region from 700 to 400 cm−1. A strong preferential incorporation of Fe(III) into natrojarosite was attributed to the oxidation potential of O2. Evidence of CO2 was detected during thermal decomposition of precipitates, suggesting the onset of mineral carbonation.

■

objective of this current work was to characterize the fluid− mineral reaction products occurring through exposing natural basalt formations to CO2 containing minor concentrations of SO2 and O2, contaminants expected to be present in CO2 captured from post- and oxy-combustion systems.18

INTRODUCTION Capturing CO2 emitted from fossil fuel combustion for longterm geologic storage is a strategy aimed to slow increases in atmospheric CO2 concentration. Geologies considered for storage are primarily deep saline formations, but unconventional reservoirs such as continental flood basalts and depleted shale gas reservoirs can expand subsurface storage options if proven viable, especially in regions where conventional storage may be constrained.1−6 Of these options, basalts provide the unique opportunity to convert injected CO2 to a solid mineral,1,7 ensuring safe and permanent storage. However, the critical carbonation reactions necessary to achieve this conversion are strongly dependent upon a complex set of chemical reactions occurring between the mineral phases in the basalt, CO2 and other contaminants present in the gas stream, and water. Although carbonation reactions in pure CO2 have received considerable study, the impacts of contaminants in the CO2 stream are far less understood. Every CO2 stream captured from a fossil fuel source will contain contaminants such as H2S, SOx, NOx, and O2, as further purification processes to obtain high-purity CO2 are not economically practical.8,9 In addition, subsurface management of these contaminants instead of implementing costly gas cleanup systems could reduce the overall cost of CO2 capture and storage.1,8−17 In our previous studies on basalts, we demonstrated that reactions occurring in the CO2−H2S−H2O system produced coatings of pyrite and divalent metal carbonates (calcite and dolomite)1,15 that temporarily halted carbonation. The © 2014 American Chemical Society

■

MATERIALS AND METHODS A unique set of basalt samples were selected from North America, India, and Africa to examine the potential for longterm storage of CO2. Samples representing the Columbia River Basalts (CRB) and the Central Atlantic Mafic Provence (CAMP) are from cores collected from wells penetrating deep subsurface basalts in Eastern Washington State and near Charleston, SC, respectively. Collected from a roadside outcrop near Branford, CT, the Newark Basin (NB) basalt is similar to the Holyoke basalt common to the Northeast United States. The Deccan basalt (DECCAN) was collected from an outcrop near Ghod in Western Ghats, India and is part of the Khak Dongar flow. Finally, the Karoo basalt (KAROO), obtained from a roadside outcrop in South Africa, was collected from a portion of the Karoo Formation. The 2.0−0.42 mm sized basalt chips utilized in this study were previously described by Schaef Received: Revised: Accepted: Published: 5298

November 7, 2013 February 7, 2014 March 26, 2014 March 26, 2014 dx.doi.org/10.1021/es404964j | Environ. Sci. Technol. 2014, 48, 5298−5305

Environmental Science & Technology

Article

et al.,1 including specific location, chemical composition (X-ray fluorescence), and mineralogy (petrographic modal analysis). On the basis of petrographic modal analysis, these basalts are mineralogically similar and contain plagioclase feldspar (35−49 wt %), interstitial glass (21−45 wt %), and pyroxene (18−31 wt %).1 Concentrations of Fe(II) and Fe(III) for each basalt indicate a narrow range of iron concentrations (8.5−11.2 wt %) but an enrichment of Fe(III) for CRB and Deccan in comparison to the Karoo.15 Furthermore, during processing, weathered products on the surface of outcrop samples were avoided to minimize potential impacts to reactivity. High-pressure static testing was conducted for 48 and 98 day intervals using 25 mL Parr reactors constructed of titanium. Approximately 0.5−5.0 g of crushed basalt was added to each reactor along with 0.25−2.5 mL of deionized water (maintaining a solid to water ratio of 2.0); limited availability of basalt samples necessitated restrictions in the amounts used. Reactors were then placed into an oven, heated to 90 °C, and subsequently pressurized with gas phase CO2 containing a mixture of 1 wt % each of SO2 and O2 to a supercritical phase target pressure of ∼100 bar. Between 7.1 and 9.3 g of gas was added to each reactor (Table 1). On the basis of these

depressurization, basalt grains were removed from the vessels while hot and either rinsed with deionized water (48 day) or allowed to air-dry on filter paper without rinsing (98 day). Grains containing visible surface precipitates were hand selected for further analysis using optical microscopy and characterized following an array of methods. Scanning electron microscopy/energy dispersive spectrometry (SEM/EDXS) was used to image individual mineral phases and to estimate chemical compositions for each phase observed. X-ray diffraction (XRD) was used to identify the mineral phases. Thermogravimetric analysis coupled with mass spectrometry (TRA-MS) and infrared (IR) spectroscopy were used to discriminate between phases that were not apparent from XRD or otherwise validate initial interpretations. A detailed description of each characterization method is provided in the Supporting Information.

■

EXPERIMENTAL RESULTS Under natural lighting, reacted grains appeared encrusted in a white to yellowish precipitate with the original underlying darkgray basalt color persisting. Solids were loosely attached to the basalt surface and easily removed; pure granules of fine-grained yellow material were present in the bottom of each glass vial after limited handling. Noticeably less precipitate was associated with the shorter duration testing, which we partially contributed to the rinsing step. Tendency for particle aggregation was most pronounced in the 98 day sampling, in which large clusters (2− 4 mm) formed, primarily in the CAMP and KAROO. Wellformed gypsum blades were abundant in all CAMP, KAROO, and DECCAN, typically forming flowered clusters and elongated bundles. In contrast, fibrous gypsum crystals formed knots of fine needles separate from the basalt grains and reached sizes near 1 mm in the CRB material. Products from NB (98 day) contained a green gel-like phase that appeared “wet” in addition to the yellow precipitate. Unique opaque, domed-shaped structures on the basalt surface were also visible. Overall, precipitates from NB and, to a lesser extent, the KAROO (98 day) appeared “wet” after air-drying, a sharp contrast to the shorter 48 day precipitates. Additionally, reaction vessels containing the NB and KAROO contained a yellowish sticky coating in the bottom that was difficult to remove. Table 2 lists the original masses and reacted masses obtained from both sets of tests. A negative mass change for the 48 day experiment is contributed to the sample rinsing step. The rinsing step was omitted in the 98 day testing, and the mass of reaction products increased (5−25%). X-ray Diffraction. Characterization of the precipitates by XRD indicated little difference (in most cases) between the crystalline phases formed during 48 and 98 day testing;

Table 1. Experimental Parameters for High-Pressure Static Tests Exposing Basalt Chips Submerged in Water Equilibrated to CO2 Containing 1 wt % Each of SO2 and O2 ID

basalt (g)

water (g)

STD

0

2.5

CR DECCAN CAMP KAROO NB

5.0 4.0 2.0 1.0 0.5

CR DECCAN CAMP KAROO NB

5.0 4.0 2.0 1.0 0.5

initial pressure (bar)

CO2/ SO2/O2 mass (g)

Standard Gas Mixture 107 7.9 48 Day Test (Rinsed) 2.5 100 8.3 2.0 86 8.8 1.0 110 9.2 0.5 83 8.8 0.25 103 8.0 98 Day Test (Unrinsed) 2.5 106 7.6 2.0 110 7.8 1.0 103 8.0 0.5 114 9.3 0.25 88 7.1

final mass (g)

% mass change

1.30 1.37 1.44 1.37 1.25

4.70 3.96 1.72 0.75 0.31

−6 −10 −14 −25 −38

1.19 1.22 1.25 1.45 1.11

5.25 4.31 2.25 1.25 0.61

+5 +8 +13 +25 +22

SO2 (mmol) 1.23

measurements, 1.1−1.5 mmol of SO2 was initially present. The reactor vessels were maintained at pressure (83−114 bar) and temperature (90 °C) for 48 or 98 days. In situ gas samples were collected from the 98 day sampling directly from the reactors by the procedure described by Glezakou et al.10 Following

Table 2. Quantitative Estimates (wt %) of Crystalline Phases Identified from Surface Precipitates Removed from Basalt Chips Reacted with Water Equilibrated with scCO2 Containing 1 wt % SO2 and 1 wt % O2 (88−114 bar) for 98 Days (90 °C)

a

phase ID

formula

DECCAN

CAMP

KAROO

CR

NBa

gypsum jarosite natrojarosite natroalunite schwertmannite hexahydrite butlerite

CaSO4·2H2O (K,H3O)Fe3(SO4)2(OH)6 NaFe3(SO4)2(OH)6 NaAl3(SO4)2(OH)6 Fe16O16(SO4)3(OH)10·10 H2O MgSO4·H2O Fe3+(OH)SO4·2H2O

51

26

12

34 45

Xb

31 17

34 22

66

19

23

X X 21 X

XRD quantification was not performed on the NB precipitates. bIdentified by SEM/EDXS. 5299

dx.doi.org/10.1021/es404964j | Environ. Sci. Technol. 2014, 48, 5298−5305

Environmental Science & Technology

Article

[Fe16O16(SO4)3(OH)10·10 H2O] (PDF 47-1775). However, another Fe3+ containing phase, butlerite [Fe3+(OH)SO4·2H2O] (PDF 25-0409) could not be ruled out, on the basis of the presence of a single weak reflection at 35.76° 2θ; overlapping reflections with natrojarosite (17.72° 2θ) also added to the uncertainty. Thermal Decomposition. Precipitates extracted from each of the basalts were characterized by TGA−MS to validate the presence of hydrated components and sulfur bearing phases and to confirm the occurrence of carbonation. As an example, the thermal decomposition (mass loss) of the CAMP precipitate is presented in Figure 2 as a function of

therefore, the following discussion will focus primarily on the longer duration tests. Powder diffraction patterns collected from each precipitate were compared to powder diffraction files (PDFs) for gypsum [CaSO4·2H2O] (33-0311), natroalunite [NaAl 3 (SO 4 ) 2 (OH) 6 ] (14-0130), natrojarosite [NaFe3(SO4)2(OH)6] (36-0425), and hexahydrate [MgSO4· 6H2O] (01-0354). In all cases with hexahydrate, overlapping reflections from other phases prevented definitive identification by XRD. Figure 1 is the graphical illustration of background

Figure 2. Thermal gravimetric analysis of precipitates collected from the CAMP 98 day experiment.

Figure 1. Graphical representation of XRD tracing obtained from precipitates formed after exposing basalts (DECCAN, CAMP, CRB, KAROO, and NB) to water equilibrated with scCO2 containing 1 wt % SO2 and 1 wt % O2 at ∼100 bar and 90 °C for 98 days (values in parentheses indicate a scaling factor).

temperature; the remaining TGA−MS graphs (DECCAN, CRB, NB, and KAROO) are shown in Figure SI1 (Supporting Information). Releases of H2O, CO2, and SO2 are represented by changes in ion current for atomic masses 18 (H2O), 44 (CO2), and 64 (SO2), respectively. DECCAN lost 10.6 wt % water in a single event at 105 °C, whereas pulses of water evolved from the CAMP (12.6 wt %), KAROO (17.9 wt %), and CRB (12.2 wt %) starting at 50 °C and ending near 116 °C. No water was observed being released from the NB before 300 °C. Early releases of water occurring with CRB (50 °C), KAROO (60 °C), and CAMP (60 and 102 °C) are in good agreement with water losses from hydrated sulfates. 19 Hexahydrate, identified by XRD in CAMP, CRB, and KAROO, undergoes a first-stage mass loss (41.7 wt %) attributed to the loss of water starting at 50 °C.19,20 In combination with the XRD characterization, we interpret a mass loss peaking between 105 and 116 °C due to the loss of water from gypsum.21,22 Water losses for the DECCAN over this temperature interval translate into ∼45 wt % gypsum, similar to estimates by XRD (52 wt %). Multiple water releases occurring with the CRB, KAROO, and CAMP prevent quantitative estimates of gypsum from the TGA−MS data. At ∼300 °C, mass loss due to the discharge of H2O appeared as a single event in the NB (5.8 wt %), but was delayed to ∼390 °C in KAROO (10.5 wt %) and CRB (4.5 wt %). In contrast, a bimodal release of water occurred in the CAMP (22.0 wt %) and DECCAN (6.1 wt %) at ∼390° and 480 °C. Some of the water release at 350 °C seems to be associated with the release of CO2 (CAMP, DECCAN, CRB, and KAROO), evident through a broad subtle increase in atomic mass 44 occurring at the same temperature; no evidence of CO2 release was seen for NB. Dehydroxylation of the jarosite structure occurs at ∼405

subtracted stacked XRD tracings and stick patterns of PDF phases plotted as a function of degrees 2θ (10°−60° 2θ). Reflections corresponding to gypsum constitute the dominant phase in the DECCAN (∼51 wt %). All remaining reflections were assigned to natroalunite (31 wt %) and natrojarosite (17 wt %). In comparison, CAMP contained less gypsum (26 wt %) and natroalunite (22 wt %) but more natrojarosite (34 wt %). Identification of hexahydrite, based on the primary peak (20.16° 2θ) appearing as a shoulder on gypsum (021) reflection (20.72° 2θ), contributed ∼19 wt %. Along with gypsum (34 wt %), CRB contained reflections most consistent with jarosite [(K,H3O)Fe3(SO4)2(OH)6] (45 wt %) and hexahydrite (21 wt %). Although reflections produced by jarosite (PDF 36-0427) and natrojarosite (PDF 36-0425) are similar, in the case of the CRB, a better match was obtained with the former on the basis of the less intense reflections positioned at 14.87 and 15.84. Precipitates from KAROO contained hexahydrate (23 wt %) along with natrojarosite (66 wt %) and a minor amount of gypsum (12 wt %). Patterns generated by the NB precipitate were of the poorest quality and thus were not quantitatively analyzed. The pattern, shown in Figure 1b, represents a combination of two types of precipitates: (1) a yellowish colored fine grained material and (2) a green-blue gel-like coating encapsulating crystalline material. Reflections assigned to natrojarosite were present in the tracing and constitute the only common phase compared to the other basalts tested. Peaks positioned at 18.24°, 26.32°, and 35.00° 2θ aligned well with the Fe3+ phase schwertmannite 5300

dx.doi.org/10.1021/es404964j | Environ. Sci. Technol. 2014, 48, 5298−5305

Environmental Science & Technology

Article

°C and appears consistent with these TGA−MS results.22,23 Moreover, the release of SO2 in three distinct stages, most notable in the KAROO (520, 676, and 804 °C) and CAMP (492, 663, and 817 °C) are typical for synthetic natrojarosite.22 Precipitate Morphologies. Reacted grains were examined under a light microscope and individually selected for SEM analysis. Representative examples of particle morphologies most common to each basalt are presented in Figure 3.

in the DECCAN and CAMP, and the presence of jarosite was restricted to CRB. These phases are subgroups of the alunite supergroup of minerals and share similar crystal habits, making morphology-based identification difficult. Classification is typically based on cation substitution occurring between Fe3+ and Al3+ for B into the ideal formula AB3(SO4)2(OH)6. The B site is occupied by Al3+ for alunites and Fe3+ for jarosites with alunites containing more Al than Fe. In the CAMP experiments (48 day), precipitation of these minerals utilized gypsum crystals as nucleation sites (Figure 3a); similar trends were observed with the KAROO (48 day). However, in most instances, well-developed surface coatings formed directly on the basalts. Common coatings consisted of distinct individual grains (Figure 3b), thick mats of triangular particles (Figure 3c), pseudocubic shaped crystals either as dense surface coatings (Figure 3d) or aggregating into large (100 and 200 μm) spherical structures (Figure 3e), and as delicate thin platelets (Figure 3f inset). Unique morphologies such as the spheres shown in Figure 3e were first observed in the shorter duration testing (KAROO) but were only just beginning to form in the longer 98 day CAMP test. With the CRB, two distinct crystal habits appeared: (1) individual clusters (∼10 μm) consisting of laminated layers, and (2) similar sized solid pseudohexagonal grains (Figure 3b). During grain selection, we encountered a pure white precipitate in the CAMP (98 day). Subsequent characterization by SEM revealed well-developed blocky pseudocubic crystals uniformly sized (1 μm; Figure 3d). The structural formula derived from EDX was Na0.830Al2.876Fe3+0.176(SO4)1.99(OH)6, a nearly pure end member natroalunite. However, chemistries most representative of the dominant phase in the CAMP are best illustrated by the following structural formula: [K0.026Na0.684Al2.121Fe3+0.802(SO4)2.048(OH)6]. Similar chemical compositions, determined by EDX, were observed for the natrojarosites in DECCAN (Figure 3c), KAROO (Figure 3e), and NB (Figure 3f inset), with a range of Fe3+ substitutions (0.63 and 2.01) occurring arbitrarily. Still, the most notable chemical difference occurred in the CRB where jarosite particles (Figure 3b) produced the structural formula nearly void of Al3+ [K0.616Na0.181Al0.211 Fe3+2.451(SO4)2.033(OH)6]. Along with blades of gypsum and well-crystallized surface coatings of natrojarosite, the presence of botryoidal clusters, either isolated or as aggregates, and amorphous-like coatings dominated the NB precipitates (Figure 3e). Typically measuring ∼100 μm in diameter, the dome-shaped features appeared to have formed in stages with multiple structures positioned adjacent or partially overlapping. Elemental analysis by EDX indicated Mg, S, and Fe were the primary components. Often tiny precipitates (5−10 μm) composed of Al−Mg−S−Fe were attached to the surface. A rather nondescript phase void of any visible structure was also present and chemically dominated by Al and S. FTIR Spectroscopy. Diffuse reflectance infrared spectra collected from each precipitate are shown in Figure SI2 (Supporting Information) over the wavenumber range 4000− 400 cm−1. Samples from CAMP, DECCAN, CRB, and KAROO each contained evidence for gypsum, based on strong bands near 1120 and 1145 cm−1 (ν3 asymmetric stretching bands of SO42−), another strong band at 671 cm−1 (ν4 asymmetric bending mode of SO42−), medium-intensity bands at 1622 and 1684 cm−1 (ν2 water bending modes), and strong, broad bands at approximately 3405 and 3550 cm−1 (O−H stretching modes).24 The spectrum of NB also has features

Figure 3. Microphotographs of surface precipitates associated with basalts exposed to CO2 containing 1 wt % SO2 and 1 wt % O2 at 90 °C and 90 bar: (a) CAMP (48 day), (b) CRB (98 day), (c) DECCAN (98 day), (d) CAMP (98 day), (e) KAROO (48 day), and (f) NB (98 day).

Gypsum crystals maintained elongated blade-shaped morphologies with striations approaching 1 mm in length and occurred as discrete particles or delicate surface clusters. In some instances (CRB), individual grains were narrower and more needle-like (10−25 μm) compared to those associated with the CAMP and KAROO, which obtained widths of 50−100 μm. Gypsum grains in the short duration testing (48 day) served as substrates for precipitation of natrojarosite and natroalunite in the CAMP samples (Figure 3a); similar observations were observed with the KAROO. This is in contrast to the clean gypsum surfaces observed with the NB, DECCAN, and CRB following 98 days of testing (Figure 3b). In the case of the NB, gypsum appeared as short laminated blocks ∼10 μm thick and