CO2–rock–brine interactions in Lower Tuscaloosa Formation at Cranfield CO2 sequestration site, Mississippi, U.S.A.

Chemical Geology 291 (2012) 269–277

Contents lists available at SciVerse ScienceDirect

Chemical Geology journal homepage: www.elsevier.com/locate/chemgeo

Research paper

CO2–rock–brine interactions in Lower Tuscaloosa Formation at Cranfield CO2 sequestration site, Mississippi, U.S.A. Jiemin Lu a,⁎, Yousif K. Kharaka b, James J. Thordsen b, Juske Horita c, Athanasios Karamalidis d, e, Craig Griffith d, J. Alexandra Hakala d, Gil Ambats b, David R. Cole g, Tommy J. Phelps c, Michael A. Manning f, Paul J. Cook h, Susan D. Hovorka a a

Bureau of Economic Geology, Jackson School of Geosciences, The University of Texas at Austin, TX, USA U.S. Geological Survey, Menlo Park, CA, USA c Department of Geosciences, Texas Tech University, Lubbock, TX, USA d National Energy Technology Laboratory, Pittsburgh, PA, USA e Carnegie Mellon University, Pittsburgh, PA, USA f U.S. Geological Survey, Jackson, MS, USA g School of Earth Sciences, The Ohio State University, Columbus, OH, USA h Lawrence Berkeley National Laboratory, Berkeley, CA, USA b

a r t i c l e

i n f o

Article history: Received 25 May 2011 Received in revised form 20 September 2011 Accepted 27 October 2011 Available online 29 October 2011 Editor: B. Sherwood Lollar Keywords: CO2 storage Rock–water–CO2 reaction Brine chemistry Carbon isotopes Tuscaloosa Formation Autoclave experiment

a b s t r a c t A highly integrated geochemical program was conducted at the Cranfield CO2-enhanced oil recovery (EOR) and sequestration site, Mississippi, U.S.A.. The program included extensive field geochemical monitoring, a detailed petrographic study, and an autoclave experiment under in situ reservoir conditions. Results show that mineral reactions in the Lower Tuscaloosa reservoir were minor during CO2 injection. Brine chemistry remained largely unchanged, which contrasts with significant changes observed in other field tests. Field fluid sampling and laboratory experiments show consistently slow reactions. Carbon isotopic composition and CO2 content in the gas phase reveal simple two-end-member mixing between injected and original formation gas. We conclude that the reservoir rock, which is composed mainly of minerals with low reactivity (average quartz 79.4%, chlorite 11.8%, kaolinite 3.1%, illite 1.3%, concretionary calcite and dolomite 1.5%, and feldspar 0.2%), is relatively unreactive to CO2. The significance of low reactivity is both positive, in that the reservoir is not impacted, and negative, in that mineral trapping is insignificant. Published by Elsevier B.V.

1. Introduction Geologic sequestration (geologic storage) of CO2 is a potential approach to reduction of CO2 emissions in a fossil-fuel-based economy. Among possible storage reservoirs for CO2, saline reservoirs have the largest potential (DOE-NETL, 2010). Depleted oil and gas reservoirs are excellent candidates for geologic storage of CO2 because they have proven capacity for holding hydrocarbon fluids at geologic timescales. Another benefit of these reservoirs is that they have been extensively explored and have a variety of datasets available for the planning and management of CO2 storage. After injection, CO2 will dissolve into the formation water—a dissolution that is rapid (a matter of minutes to hours) and that quickly lowers the pH and changes the distribution of many dissolved species (Kharaka and Cole, 2011). CO2 solubility in brine is quantified by models

⁎ Corresponding author. E-mail address: [email protected] (J. Lu). 0009-2541/$ – see front matter. Published by Elsevier B.V. doi:10.1016/j.chemgeo.2011.10.020

on the basis of experimental data (e.g., Duan and Sun, 2003; Portier and Rochelle, 2005), and dissolution of CO2 will initiate a variety of geochemical reactions with brine and rock-forming minerals that are more complicated and less well constrained. Some of the reactions could be beneficial by helping to chemically trap the CO2 as dissolved species and by forming new carbonate minerals (Bachu et al., 2007); other reactions may be damaging to the reservoir and seal integrity by dissolving rock-forming minerals. These mineral reactions could alter rock structure, petrophysical properties, and chemical composition of the reservoir and cap rock. Injectivity may be affected in the near term, whereas longer-term effects may include reservoir and seal integrity and mineral-trapping capacity. Brine chemistry will most likely be altered by a decrease of pH upon CO2 injection. Rock–brine–CO2 reactions take place mostly in the form of mineral dissolution/ precipitation and adsorption/desorption at the mineral surface. Previous field tests have shown dissolution of carbonate minerals and metal release during CO2 injection. At the Weyburn CO2-EOR site, for example, CO2-driven dissolution of calcite increased concentrations of Ca and HCO3− in brine and changed the carbon isotopic

270

J. Lu et al. / Chemical Geology 291 (2012) 269–277

ratio of CO2 gas. The distribution pattern of brine chemistry in the reservoir was altered across the field during CO2 injection (Emberley et al., 2005). Following CO2 breakthrough at the Frio pilot project, pronounced increases in the concentrations of HCO3− (100–3000 mg/L), Fe (30–1100 mg/L), and other metals (Ca, Mn, etc.) were observed as pH dropped (Kharaka, et al., 2006, 2009; Xu et al., 2010). The Nagaoka pilot test in Japan has also shown elevated concentrations of HCO3−, Ca, Mg, Fe, and Mn in the formation water after injection of 10,400 tons of CO2 (Mito et al., 2008). Mineral-trapping capacity of a storage formation is another important geochemistry-related aspect of CO2 storage projects. Mineral trapping is considered a desirable mechanism because CO2 is permanently stored in solid phases, such as calcite, dolomite, magnesite, siderite, or dawsonite (e.g., Gunter et al., 1993, 1997; Johnson and Nitao, 2002; Xu et al., 2003, 2005; Knauss et al., 2005; Palandri and Kharaka, 2005; White et al., 2005; Zerai et al., 2006). Mineral-trapping capacity depends primarily on fluid composition and mineral composition, as numerical modeling has shown (e.g., Xu et al., 2005; Zhang et al., 2009). Assessment of mineral-trapping capacity requires modeling that is based on detailed characterization of reservoir mineralogy, petrography, and brine chemistry. Field observations will provide valuable calibration to modeling. An understanding of the overall impact of the geochemical interactions between CO2 and rock/brine is important in satisfying regulatory requirements and ensuring public confidence. However, these processes will depend on mineralogy, brine chemistry, and conditions

of individual storage reservoirs. Geochemical assessment and monitoring must account for mineral dissolution, precipitation, and pHchange aspects of the systems. In this paper, we present results of a highly integrated geochemical research program, including extensive field geochemical monitoring, detailed petrographic study, and a batch experiment under reservoir conditions. The series of studies were conducted during a large-scale CO2-enhanced oil recovery (EOR) and sequestration project at Cranfield field, Mississippi, U.S.A. The goal of this study was to assess the role of reservoir mineralogy and petrography in controlling geochemical processes during CO2 injection, thereby furthering design of monitoring programs for future CO2 sequestration projects. 2. Geology and field operations Cranfield field is a salt-cored, simple domal structure located ~20 km east of Natchez in Adams and Franklin County, southwest Mississippi, U.S.A. (Fig. 1). The reservoir (CO2 injection zone) is in the Upper Cretaceous Lower Tuscaloosa Formation at depths of ~3000 m (10,000 ft). The reservoir unit in the basal Lower Tuscaloosa Formation, locally called Sand “D-E,” is composed of a 15- to 25-m-thick (45- to 80-ft) package of porous and permeable fluvial sandstones with conglomerates. Core examination shows that the sandstones and conglomerates consist of vertically stacked, fluvial point-bar and channel deposits, similar to those described by Stancliffe and Adams (1986). The cores appear to be light green owing to abundant chlorite.

Fig. 1. Structural contour map of the top of the Lower Tuscaloosa Formation reservoir (Sand D-E). Contours in true vertical depth subsea (TVDSS) with interval of 3.048 m (10 ft). Sample wells named and nearby injection wells marked.

J. Lu et al. / Chemical Geology 291 (2012) 269–277

The oil and gas field, discovered in 1943, was produced to depletion in 1966 (Mississippi Oil and Gas Board, 1966). Water injection was tested only briefly on the west side of the field in 1958–59 but was considered unsuccessful and was abandoned (Mississippi Oil and Gas Board, 1966). Fluid perturbation by extensive waterflood by diverse fluids common in EOR was avoided in the study area, and strong aquifer drive returned the reservoir to hydrostatic pressure prior to CO2 injection in July 2008. Reservoir temperature was ~125 °C, and reservoir pressure was 32 MPa at a depth of 3040 m prior to CO2 flood. This situation was optimal for geochemical monitoring because, unlike most EOR fields in which a prolonged period of pressure maintenance by water injection before CO2 injection complicates reservoir chemistry, CO2 injection at Cranfield was preceded by >40 years of reequilibration. Thus, the initial pressure buildup and geochemical disruption caused by CO2 injection at Cranfield are expected to be analogous to those of CO2 sequestration in saline formations. The reservoir was brought under CO2-EOR flood by Denbury Onshore LLC in July 2008. Mass injected CO2 (not including recycled CO2) reached 2.5 million tons in early 2011. CO2 from natural geologic accumulation at Jackson Dome (near Jackson, Mississippi) is transported 160 km via pipeline to the field and has been injected continuously (without water injection) to support EOR. The CO2 is in supercritical state at reservoir conditions. CO2 was injected through six injectors in the north part of the field in 2008. A CO2 flood later extended to the east side, and the number of injectors increased to 24 in early 2011. Owing to considerable heterogeneity of the fluvial system, CO2 migration pathways are complex. Arrival time of the injectate has varied widely at different production wells and observation wells. 3. Sampling and methods 3.1. Brine and gas sampling and analyses CO2 injection started in July 2008. One brine sample prior to injection was taken from Injector 29–10 (Fig. 1) by the field operator in October 2007. Following initiation of CO2 flooding, three sampling campaigns were conducted in March of 2009, December of 2009, and April of 2010, covering 21 months from the start of injection. Water and associated gas samples were obtained from the wellhead at 10 production wells in the north part of the field and two dedicated observation wells completed below the oil–water contact on the east flank of the field (Fig. 1). Seven production wells were sampled in March 2009, and eight production wells (including the five sampled in March) were sampled in December 2009. Although another four wells were sampled in April 2010, results of these samples will be reported in a later publication. On December 1, 2009, CO2 injection started through Injector 31F-1 at a rate of ~250 metric tons of CO2 per day into the downdip water leg below the oil–water contact, away from other field activities (Fig. 1). Injection rate increased to ~490 metric tons per day on December 20, 2009. A series of geochemical and geophysical monitoring activities were conducted at two observation wells, 31F-2 and 31F-3, which are 68 and 112 m east of the injector, respectively (Fig. 1). The observation wells were constructed of fiberglass casing and coated tubing, which reduce dissolution of steel as a contaminant—a problem encountered in earlier tests (Kharaka et al., 2006, 2009). Installed U-tube samplers (Freifeld et al., 2005) enabled continuous fluid sampling from the perforation. A downhole Kuster sampler was deployed when the U-tube samplers were blocked temporarily by solids between December 5 and 12 (2009). In total, 11 brine and 830 gas samples were collected between December 1, 2009, and January 1, 2010. From mid- to late December, U-tube samplers collected only gas because high-mobility CO2 gradually filled the well bore and displaced brine in the perforated intervals. Continuous sampling enabled documentation of geochemical processes as the CO2 front passed the monitoring wells.

271

The brine and gas samples collected were subjected to onsite filtration and preservation. Electrical conductivity (EC), pH, and alkalinity were measured at atmosphere pressure and room temperature in the field immediately after the samples were taken. The pH measurements have an uncertainty of ±0.2 pH units. Note that, owing to degassing, pH values measured at atmosphere pressure are not the actual values under reservoir conditions. Subsurface pH values are computed by using field-measured pH and raising the pCO2 to 300 bars at 120 °C using SOLMINEQ. Previous field studies have demonstrated that such measurements are able to document influence of injected CO2 and major chemical changes (e.g., Kharaka et al., 2006). Water samples for cations, anions, and trace metals were filtered, acidified, and preserved for laboratory determinations following the protocol described in Kharaka and Hanor (2007). Uncertainties in the determination of major cations and anions and, therefore, total dissolved solids (TDS) are ±1–3%; for trace metals uncertainties are ±5%, but as much as 20% when values are close to the detection limit. Gases (CO2, CH4, C2–C6) were analyzed using a Carle AGC 400 Gas Chromatograph that is equipped with both thermal-conductivity (TCD) and flame ionization (FID) detectors. The resulting component peak areas are quantified (given raw percent values) by comparing them to previously run standards. The raw percentage values for each sample are recorded and normalized. Preparation of samples for measurement of the δ13C of CO2 was performed by first injecting a sufficient amount of the sample gas into a gas chromatograph (GC) for CO2 separation. After GC separation, the CO2 was passed via a helium (He) carrier gas into a liquid nitrogen trap. After removal of the He carrier gas by evacuation, a dry-ice/isopropylalcohol bath was substituted for the liquid nitrogen bath and the CO2 was transferred cryogenically into a cold finger. After yield was calculated using an electronic manometer connected to the cold finger, the CO2 was transferred into 1/4 in. Pyrex tubing immersed in liquid nitrogen, and sealed for mass spectrometric analysis (Finnigan MAT 252). 3.2. Transmitted light microscopy and scanning electron microscopy (SEM) As part of the geochemical study, a total of 63 thin sections from 4 cores (29–12, 28–1, 31F-2, 31F-3) from the reservoir interval were examined using a light microscope. Samples were examined in both transmitted light (plane and polarized) and in bright-field reflected light. Another 27 thin sections were coated with carbon and examined using a field-emission SEM and an FEI Nova™ NanoSEM 430 at The University of Texas at Austin (UT). The SEM was aided by energy dispersive spectrometry (EDS) for qualitative analysis of elemental composition. 3.3. X-ray diffraction (XRD) mineralogy Thirty-seven core samples from the injection zone of Well 31F-2 were quantitatively analyzed using random-powder X-ray diffraction (XRD), which was prepared by wet grinding and spray drying (Hillier, 1999). Core samples were disintegrated using a TEMA ball mill prior to wet grinding using a McCrone Micronizing Mill. Slurry samples were sprayed through the heated chamber of a spray drier and dried. XRD was conducted on a Bruker AXS D8 diffractometer at UT, and quantitative analysis was performed using Topas 3, which is PC software based on the Rietveld method (Bish, 1994). Sample-preparation methods and analytical parameters can be found in Lu et al. (2011). 3.4. Autoclave CO2–rock–brine experiment CO2–rock–brine interactions under reservoir temperature and pressure were simulated in a laboratory-scale experiment using Dickson-type, flexible-gold-bag, rocking autoclaves. The rocking autoclave consisted of an electrical furnace capable of rotating 180° at 4 rpm, a stainless-steel pressure vessel, and a flexible gold bag (250 mL)

272

J. Lu et al. / Chemical Geology 291 (2012) 269–277

capped with a commercially pure titanium head fitted with a capillary-lined stainless-steel sampling tube. The sampling tube was equipped with a meter valve that allowed the experiment to be sampled under temperature and pressure. At the beginning of the experiment, the gold bag was filled, under an N2 atmosphere, with 193.2 g of reservoir brine (taken from Well 31F-2 using a U-tube, Fig. 1) and 10 g of loosely crumbled core material (3190.9 m, Well 31F-2) (see Table 1 for mineral composition), sealed with the Ti head. Then the capped-gold-bag assemblage was placed in the pressure vessel, which was filled with deionized water. The confined deionized water served as the pressure medium, and pressure was controlled by an HPLC syringe pump. The pressure vessel was pressure tested with He gas before being heated and agitated, then it was placed in the rocking furnace and brought to temperature (120 °C) and pressure (32.4 MPa). The brine–rock sample was reacted at pressure and temperature for 75 days prior to supercritical CO2 injection. Fluid samples were taken and analyzed to ensure that brine chemistry stabilized and equilibrium between rock and brine was achieved. Then CO2 was injected into the gold bag on days 75 and 76. In total, 25.35 g of CO2 was added. Given the solubility under experimental conditions (37 g/kg H2O), ~7 g of CO2 would dissolve into the brine (Duan and Sun, 2003). The rest of the CO2 would remain in head space, and the brine was saturated with CO2 for the rest of the experiment. Three water samples were taken before CO2 was introduced, and eight afterward. Four samples were taken within 12 h of injection to capture initial changes in brine chemistry resulting from addition of CO2. The water samples were analyzed for Table 1 Mineralogy of reservoir sandstone of Well CFU31F-2 determined by XRD. Well location: latitude: 31.563336; longitude: −91.140564. Depth (m)

Quartz Kaolinite Chlorite Illite Albite Calcite Dolomite Anatase % % % % % % % %

3178.1 3178.9 3179.8 3180.1 3181.1 3181.7 3182.0 3183.0 3183.8 3184.4 3184.8 3185.2 3185.8 3187.0 3187.6 3188.3 3188.8 3189.3 3189.6 3190.0 3190.1 3190.8 3190.9a 3191.2 3191.5 3193.2 3193.6 3193.9 3194.2 3194.9 3195.6 3195.9 3196.1 3196.8 3197.7 3199.7 3200.3 Average

77.9 73.7 66.9 77.0 79.4 78.6 74.7 82.0 81.5 85.8 83.9 83.5 83.2 79.3 79.3 80.7 82.3 81.3 82.5 78.8 83.0 81.0 80.7 75.6 77.9 55.2 84.6 81.4 85.5 85.1 85.7 80.8 82.0 81.9 83.6 71.2 70.5 79.4

a

5.1 4.3 5.1 6.0 3.9 4.0 4.6 3.3 3.7 3.3 3.4 3.4 3.3 4.2 3.3 3.1 2.1 2.6 2.7 2.7 2.4 1.5 1.7 2.5 2.8 1.2 3.0 2.2 2.5 2.2 2.0 2.1 2.2 2.2 1.7 4.8 4.7 3.1

11.0 13.8 18.4 18.8 11.0 11.1 14.4 9.7 10.1 7.3 8.4 9.0 9.1 11.2 13.0 12.6 12.2 12.0 10.7 14.0 10.7 13.4 14.1 17.1 15.3 4.7 8.8 12.8 8.7 10.1 9.2 14.0 13.1 12.5 10.5 10.6 13.1 11.8

1.3 1.9 2.8 2.9 2.0 2.1 2.3 1.8 1.8 1.2 1.3 1.6 1.1 1.2 0.9 1.2 1.0 1.0 1.2 1.3 1.3 1.2 0.8 1.0 0.9 0.6 0.8 0.4 0.4 0.3 0.5 0.3 0.1 1.2 1.4 1.9 1.5 1.3

0.5 2.5 2.9 1.0 0.9 1.1

36.8

0.5 0.6

0.5

0.2

Sample used in laboratory autoclave experiment.

1.1

9.8 6.5 0.4

4.2 3.7 4.0 4.3 2.8 3.1 4.1 3.2 2.9 2.5 3.0 2.5 3.3 4.1 3.4 2.5 2.5 3.1 2.9 3.2 2.7 2.9 2.6 3.7 3.2 1.5 2.8 3.3 2.4 2.3 2.0 2.9 2.7 2.2 2.3 1.8 3.7 3.0

major cations by ICP-OES, for trace cations by ICP-MS, for anions by ion chromatography, and for dissolved CO2 by CO2 coulometer. The reacted rock sample was examined using SEM and compared with the unreacted rock samples. 4. Results 4.1. Mineralogy and petrography Bulk XRD mineral compositions of core samples are shown in Table 1. Abundance of authigenic mineral phases and rock fragment was estimated by thin-section microscopy and point count. The Lower Tuscaloosa reservoir sandstones are quartz arenite, composed mainly of quartz (79.4%), chlorite (chamosite) (11.8%), kaolinite (3.1%), and illite (1.3%) (Table 1), which have a lower dissolution rate and reactivity than those of K-feldspar and carbonate minerals (Palandri and Kharaka, 2004). The reservoir rock contains small amounts of more active minerals, such as calcite (1.1%), dolomite (0.4%), and albite (0.2%). According to point count, the sandstones contain an average 4.5% of rock fragments that are mostly metamorphic and igneous. Many rock fragments have been dissolved, leaving secondary pores (average 2.3%), and some have been partly replaced by chlorite. Feldspar minerals are rare in thin sections and undetectable by XRD for most samples. Albite was identified by XRD in five samples in the upper part of the reservoir, with an abundance of as much as 2.9% (Table 1). Most chlorite occurs as fibrous coats surrounding detrital grains (Fig. 2A, B), although some occurs as pore-lining and pore-filling phases. Authigenic kaolinite most commonly occurs as pore-filling vermicular and booklet-stacking patterns. Quartz cement (up to 1% according to point count) is present in the upper part of the reservoir. Calcite and dolomite concretions (some of which are iron rich, according to SEM/EDS analysis) exist in conglomerates in the lower part of the reservoir. In total, four concretions with irregular edges were identified in two cores from the monitoring wells (31F-2 and 31F-3). Under microscope, most calcite and dolomite cements (up to 40%) are poikilotopic and occur as pore-filling and replacing habits. Outside of concretions, carbonate minerals occur only in trace amounts. They are commonly surrounded by grain coating and pore-lining chlorite (Fig. 2C). 4.2. Gas chemistry The Middle Tuscaloosa Formation water was saturated with CH4 (~60 mM) before CO2 injection. During the sampling period, CH4 content in the gas samples ranged from 4.3% to 89.9% and CO2 from 1.9% to 94.8%. (Table S1, online supplementary data). The gas samples with low CO2 content (b5%) had low C isotopic ratios (−8.0 to −10.3‰) (Fig. 3). Those containing high CO2 abundance (>50%) showed higher δ13C values, similar to those of Jackson Dome CO2 (δ13C=−2.6‰) (Table S1). The cross-plot of CO2 content and δ13C shows that data points fall largely on the mixing trend between injectate and original formation gas (Eq. 1), when Jackson Dome gas (CO2%=99%; δ13C=−2.6‰) and Sample CRF2-339 (CO2%=4%; δ13C=−10.5‰ ) are used as end members (Fig. 3).

Y¼

δ13 C f  C i  C f −δ13 C i  Ci  C f 100 δ13 C i  C i −δ13 C f  C f  ; þ C i −C f C i −C f X ð1Þ

where Y is the C isotopic ratio of the gas samples, X is the CO2 content (%) of the samples, δ13Cfis the C isotopic ratio of original formation gas (−10.5‰), δ13Ci is the C isotopic ratio of the injectate (−2.6‰), Cf is the CO2 content in formation gas (4%), and Ci is the CO2 content in injectate (99%).

J. Lu et al. / Chemical Geology 291 (2012) 269–277

273

Fig. 2. (A) Light optical microscopic image showing chlorite coating surrounding detrital grains, rock fragments, and secondary pores; 3190.3 m, Well 31F-3; (B) backscattered SEM image showing fibrous chlorite coating surrounding quartz grains; 3190.8 m, Well 31F-2; (C) backscattered SEM image showing rare-occurring calcite (cc) grain confined by chlorite rims, limiting its contact with pore water; 3190.8 m, Well 31F-2.

4.3. Brine chemistry

~500 mg/L as CO2 content increased from 3% to 75%. The pH shows a decrease of ~0.8, followed by a rebound of 0.5. Other elements do not show significant changes with time.

The Lower Tuscaloosa reservoir brine is an Na–Ca–Cl-type water, with relatively uniform concentrations of major cations (Fig. 4) and salinity (~150,000 mg/L TDS). The brine contains relatively high concentrations of Mg (890–1180 mg/L) and Sr (480–760 mg/L) but low values of SO42 − (25–55 mg/L) (Table S2). Overall, TDS do not vary significantly across the field (Figs. 4, 5). TDS of the earliest water sample (155 g/L) taken before CO2 injection were slightly higher than the average (149.6 ±4.5 g/L, N=26) of those taken in 2009. For these five wells (CFU 28–2, 29–1, 29–3, 29–6, and 29–9) sampled both in March and December 2009, average TDS increased slightly from 148.1±6.4 g/L to 150.2±6.8 g/L (N=5). However, TDS of all the samples were not correlated with CO2 contents in the associated gas phase (Fig. 5). In particular, metals such as Ca, Mg, Mn, and Sr, which are generally associated with carbonates, appear to be constant, with varying CO2 content. For example, Ca concentrations vary between 7600 and 14,000 mg/L but do not appear to increase as CO2 content increases (Fig. 5). Only alkalinity and iron show increasing trends as CO2 content increases (Fig. 5). HCO3− increased from ~ 300 mg/L to ~600 mg/L and Fe from ~ 90 mg/L to ~ 160 mg/L when CO2 increased from 2% to 95%. Continuous sampling from Observation Well 31F-2 documented geochemical signals of CO2 arrival from Injector 31F-1. CO2 content in associated gas started to increase 9 days after the initiation of injection on December 1, 2009 (Fig. 6) (CO2 front arrived at Well 31F-2 earlier, when U-tube sample was blocked and no brine and gas samples were taken). Note that three samples between December 9 and 12 were collected by the Kuster wireline sampler while the U-tube was temporarily plugged. The different sampling method may have caused some of the variations in results. HCO3− increased from ~400 mg/L at the start to

Water analyses from the experiment show that major metals (Ca, Mg, Na, Al, Mn, Sr, Ba) kept relatively constant upon introduction of CO2 (Fig. 7). The sample of day 76.3 shows higher concentrations of Ca (16,947 mg/L), Mg (1409 mg/L), Sr (1100 mg/L), and Ba (70 mg/L) than the rest of the samples (average Ca, 11,952 mg/L; Mg, 993 mg/L; Sr, 747 mg/L; Ba, 58 mg/L). The elevated concentrations were not maintained in subsequent samples. Concentrations of Ni, Mo, and U fluctuated within the range of one order of magnitude (Fig. 7). Similar to what was found in field observations, Fe shows an overall increase from b1 mg/L before CO2 was added to 68 mg/L at the end of the experiment. Other metals, Zn, Cu, Pb, Sn, and Sb, appear to follow the pH trend. Low pH decreased their contents immediately after CO2 was introduced, and high pH gradually increased their concentrations toward the end of experiment. The pH, which is also measured at atmosphere pressure, dropped from 5.7 to 4.8 after CO2 was added; then it rebounded to 5.2 toward the end of the experiment. The pH trend matches that of the field observation well at Well 31F-2. The reacted rock sample was retrieved by filtering remaining fluids from the gold bag. The retrieved material was disintegrated, and its physical occurrence was altered considerably from its original state. Salts (mainly CaCl2 and NaCl) were precipitated when the reacted material was dried in an oven prior to SEM examination. Such physical changes make comparison between the original and the reacted material inconclusive.

Fig. 3. CO2 concentration vs. stable C isotopic ratio of gas samples. Data points align with mixing line of injectate (CO2% = 99%; δ13C = − 2.6‰, triangle) and original formation gas (CO2% = 4%; δ13C = −10.5‰). Mixing trend defined by Eq. 2.

Fig. 4. Concentrations of major cations and anions of 27 brine samples. Most cations show similar concentrations; alkalinity and Fe display larger variations. Sample from Well 29–10 taken before CO2 injection shown as green diamonds. Taken by the field operator, it was not filtered, which may be why its Fe is relatively high.

4.4. Autoclave CO2–rock–brine experiment

274

J. Lu et al. / Chemical Geology 291 (2012) 269–277

Fig. 6. Brine and gas composition change in Monitoring Well 31F-2 with time. CO2 injection at Injector 31F-1 started on December 1, 2009. CO2 arrival marked by increase of CO2 content in gas phase, increase of HCO3−, and decrease in pH. As injection continued, gas displaced brine in the well, ending brine sampling.

Fig. 5. Major cations, Cl, HCO3−, TDS, and pH of brine samples plotted against CO2 content in associated gas samples; pH measured at surface. HCO3− shows noticeable increase as CO2 content increases, indicating limited carbonate dissolution. Fe also shows modest increase from ~90 to ~170 mg/L, which may be caused by dissolution of iron oxyhydroxides.

5. Discussion The multiple datasets of this study, including gas composition, C isotopes, brine chemistry, and autoclave CO2–rock–brine reaction, are consistent with one another. Field observations and the laboratory experiment show a largely unaffected brine and rock system with possible minor carbonate dissolution during CO2 injection. Before its being affected by injected CO2, the Cranfield reservoir brine was saturated with CH4, and CO2 content in associated gas is ~1–5%. The gas samples with low CO2 content (b5%) and low C isotopic ratios (−8.0 to −10.3‰) represent preinjection formation gas. CO2 content jumped to 95% after the injectate reached some of the sampling wells. The samples with high CO2 content show higher C isotopic ratios, indicating the influence of Jackson Dome CO2 (δ13C ~−2.63‰) (Table S1). Data points of CO2 content and C isotopic composition in gas samples fell close to a simple two-end-member mixing line, suggesting no significant additional carbon source reaction (Fig. 3). Similar mixing between baseline and injected CO2 was found at the Cardium CO2 monitoring site, where gas carbon isotopes were monitored at production wells (Johnson et al., 2011). The carbon isotopic ratio of CO2 can be affected by mineral reactions. For example, during Weyburn CO2-EOR operations, δ 13C of CO2 gas became higher than baseline and injected CO2, suggesting that additional carbon with higher δ 13C was derived from dissolution of carbonates in the carbonate reservoir (Emberley et al., 2005). At

Cranfield, calcite in the reservoir rock shows variable δ 13C between − 0.30‰ and − 12.99‰ with an average of − 6.25‰ based on 6 samples (Table S3). Dissolution of the calcite would release CO2 gas with much lower δ 13C (average −8.64‰) than the injected CO2 (−2.63‰) according to fractionation equation between HCO3− and CO2(gas) of Deines et al. (1974) (most carbon from calcite dissolution transforms to HCO3−). If a significant amount of calcite had been dissolved, the data points in Figure 3 would have fallen below the mixing trend. The samples align well with the mixing trend between formation gas and injected gas, suggesting that carbonate dissolution was limited. Brine chemistry, including Electric Conductivity (EC), major cations, and most anions, remained relatively unchanged, whereas CO2 abundance increased in some wells to the level of the injectate. Between March and December 2009, average TDS increased slightly from 148 to 150 g/L. However, concentrations of most cations are not correlated with CO2 content (Fig. 5), and no evidence shows that high CO2 concentrations increased cation concentrations in brine. High CO2 concentrations signal the arrival of injectate and the start of potential CO2-induced reactions. The constant cation concentrations, therefore, suggest limited mineral reactions. Alkalinity increased from ~300 mg/L to ~600 mg/L as HCO3− when CO2 rose above 90% (Fig. 5). Carbonate mineral dissolution is the most probable HCO3− source (Eq. 2), although C isotopes and related cation concentrations do not show evidence of dissolution. The 300-mg/L increase in bicarbonate is equivalent to dissolution of the 0.002% rock mass of calcite (average porosity of reservoir rock ~25%; rock density ~2.68 g/cm3). CaCO3 þ CO2 þ H2 O→Ca

2þ

−

þ 2HCO3

2þ

CaMgðCO3 Þ2 þ 2CO2 þ 2H2 O→Ca

ð2Þ

2þ

þ Mg

−

þ 4HCO3

ð3Þ

Amount of calcite dissolution is estimated to mobilize 98 mg/L of Ca, which is insignificant when compared with the high Ca content of this brine (7600–14,000 mg/L). Assuming dolomite dissolution (Eq. 3), 49 mg/L of Ca and 30 mg/L of Mg could have been released, respectively. However, the initially high concentrations of Ca and

J. Lu et al. / Chemical Geology 291 (2012) 269–277

Fig. 7. Cation and anion concentrations during CO2–rock–brine autoclave experiment. CO2 added before sample of day 75.8 (marked by arrow). Only small variations in major cations and anions can be seen. Fluctuations in trace cations may be related to surface desorption/adsorption. Note: time scale distorted to better show dense data points immediately following CO2 introduction.

Mg (7600–14,000 mg/L, 770–1260 mg/L) exceed estimated release by many times, and, consequently, any carbonate dissolution is masked. Fe increases from ~90 to ~170 mg/L and may be related to dissolution of iron oxyhydroxides as grain coating, which was proposed as an iron source at the Frio CO2 injection site (Kharaka et al., 2009). Dissolution of chlorite is unlikely because Mg and other components in chlorite did not increase. The small amount of carbonate dissolution contrasts sharply with that of the Frio test, in which ~30 times the bicarbonate increase (100–3000 mg/L) occurred upon the arrival of CO2. At Frio, dissolution of carbonate minerals and iron oxyhydroxides is thought to be the main source of elevated alkalinity (corrosion of well tubing and casing is another possible source) (Kharaka et al., 2006, 2009). The difference between the two sites is rooted in the difference between mineral composition and rock texture. The Frio injection interval contains abundant orthoclase and plagioclase feldspar (~20%) and a small amount of calcite as reactants. Only minor clay coatings exist on mineral grains, which do not limit CO2–mineral reactions. At Weyburn, the Midale carbonate aquifers contain abundant dolomite and calcite to react with

275

injected CO2, and observation shows significant increases in concentrations of related cations and anions (Emberley et al., 2005). In contrast, Cranfield reservoir rocks contain only a trace amount of calcite and dolomite and rare feldspar in high-permeability zones. Abundant fibrous, pore-lining, and grain-coating chlorite would greatly reduce the contact area between reactive minerals and CO2-saturated brine (Fig. 2C), keeping reactions to a minimum. Continuous sampling at Well 31F-2 shows that alkalinity increased slightly from 400 to 500 mg/L with time, suggesting some dissolution of carbonate minerals. The pH rebounded after an initial decrease, probably indicating some degree of mineral buffering. However, metals, except Fe, do not exhibit correlated changes. Again, this phenomenon suggests that carbonate dissolution is limited. The potential quantities of mobilized metals from carbonates (Ca, 16–33 mg/L; Mg 0–10 mg/L) are small and could be masked by high and variable background values (Ca, 7600–14,000 mg/L; Mg, 770–1260 mg/L). Overall, this set of data also reveals a largely unreactive geochemical system with limited dissolution of carbonates at the arrival of injected CO2. In the batch reaction experiment, most cations and anions remained largely unchanged by the introduction of CO2 (Fig. 7). Concentrations of Ca, Mg, Sr, and Ba increased slightly at day 76.3, but they decreased to earlier levels. The fluctuations may have been caused by temporary disruptions in temperature or pressure in the reaction vessel, which may have temporarily encouraged carbonate dissolution. The spike in K concentration on day 76.1 (1184 mg/L compared with the average of the rest of the samples, 449 mg/L) is suspicious. No correlated changes show up in Al or Si trends; therefore, the spike is unlikely related to dissolution of potassium-bearing feldspars or clay minerals. Fe concentrations fluctuated but showed an overall increasing trend, indicating a possible contribution from iron oxyhydroxides. Similar to the field observation of Well CFU 31F-2, pH rebounded following an initial decrease after CO2 (Fig. 7). The pH rebound was most likely caused by mineral buffering by mineral dissolution. However, as in field results, major cations remained constant, suggesting that only limited mineral dissolution may have occurred in response to CO2. Overall, the reaction experiment reproduced field observations, further confirming that the specific combination of mineral assemblage and formation water of the Lower Tuscaloosa reservoir is not very reactive in contact with CO2 in the short term. Reservoir sandstones are composed mainly of minerals that are kinetically limited in their reaction with CO2 (e.g., quartz, chlorite, kaolinite, and illite). Small amounts of albite are present only in the upper part of the reservoir. K-feldspar is almost absent. Most rock fragments were already dissolved or severely altered early in diagenesis, producing secondary pores (Fig. 2A). During geologic time, reactive components in the reservoir may have already been consumed by reactions aided by CO2 produced from organic matter or converted from CH4. For example, degradation of volcanic rock fragments could be the source of the abundant chlorite in the Lower Tuscaloosa Formation (e.g., Stancliffe and Adams, 1986). Therefore, the present low reactivity of the reservoir system would be the result of reactions of naturally occurring CO2 at geologic timescales. Porelining and grain-coating chlorite may have contributed to the isolation of the rare reactive minerals from reservoir fluids. Carbonate cements, mainly present in concretions, may have limited mineral–brine contact because porosity and permeability are reduced in concretions, according to thin-section observation. The limited-contact area may have prevented large-scale carbonate dissolution. Rock properties are unlikely to be altered during CO2 injection in a reservoir with a composition similar to that of the Lower Tuscaloosa owing to the limited reactivity of minerals. A small amount of calcite and/or dolomite (~0.002%) may be dissolved in the reservoir, most dissolution probably occurring at concretion margins because porosity inside the concretions is low and carbonate minerals may not be exposed to brine. Overall increases in permeability should be limited because chlorite, the most important mineral affecting permeability, appears to be largely unreactive toward CO2.

276

J. Lu et al. / Chemical Geology 291 (2012) 269–277

The mineral buffering process is restricted by a lack of buffering minerals such as feldspar and carbonate minerals in the reservoir sandstone. Consequently, the capacity of mineral trapping is limited. Mineral trapping of CO2 will not be important in the reservoir because the trace amount of available plagioclase and carbonate restricts the supply of Ca and Na. Previous modeling studies suggest that chlorite could be a significant source of Mg and Fe, which could lead to precipitation of ankerite and siderite (Xu et al., 2005; Zhang et al., 2009). Potential CO2 mineral trapping depends on the abundance of chlorite and its reaction-rate constant. Chlorite dissolution rate is generally low (log k = −12.52 log mol m− 2 s− 1 for neutral mechanism, Palandri and Kharaka, 2004). Dissolution is unlikely to be significant in the short term because field monitoring and the laboratory experiment showed no evidence of dissolution. On a timescale of thousands of years or longer, however, a slightly higher actual dissolution rate could increase the capacity of mineral trapping. Slow CO2–rock–brine reactions leave dissolution of CO2 as the main geochemical process at Cranfield. Injected CO2 will dissolve into brine and produce carbonic acid, H2CO3, which dissociates into bicarbonate and carbonate ions, HCO3− and CO32 −: −

þ

H2 CO3 HCO3 þ H

−

2−

þ

HCO3 CO3 þ H :

ð4Þ

largely unaffected. Only a small amount of carbonate, equivalent to 0.002% of rock affected by injectate, may have dissolved. The limited geochemical reactions are the result of nonreactive reservoir rocks, which are composed mainly of quartz, chlorite, kaolinite, and illite. Reactive minerals such as carbonates, plagioclase, and Kfeldspar are rare in the reservoir. Abundant chlorite occurs as mineral coats, further separating reactive minerals from reservoir fluids. Limited contact between fast-reacting minerals and CO2-saturated brine prevented reactions from taking place on a large scale. Supplementary materials related to this article can be foundonline at doi:10.1016/j.chemgeo.2011.10.020.

Acknowledgments The study was funded by the Department of Energy and managed by the National Energy Technology Laboratory (Bruce Brown, DOE project manager) through the Southeast Regional Carbon Sequestration Partnership (SECARB) (managed by the Southern State Energy Board). We thank Denbury Onshore LLC for hosting the project and providing tremendous assistance during the study. We also acknowledge assistance from T. A. Meckel in making the field map. The article benefited immensely from thorough and constructive reviews by two anonymous reviewers. Publication authorized by the Director, Bureau of Economic Geology.

ð5Þ

Because the capacity of mineral trapping is low, solubility trapping must be the major chemical trapping mechanism. Under reservoir conditions, CO2 solubility is ~ 37 g/kg H2O (Duan and Sun, 2003). The unique combination of mineral assemblage and chlorite occurrence leads to special geochemical responses during CO2 injection, contrasting with rapid mineral reactions and cation releases at the Weyburn, Frio, and Nagaoka sites and previous reaction experiments (Emberley et al., 2005; Kaszuba et al., 2005; Kharaka et al., 2006, 2009; Mito et al., 2008; Ketzer et al., 2009). The contrast highlights the importance of site-specific mineralogy, petrography, and brine chemistry. In the design of a geochemical monitoring program for future CO2 sequestration projects, reservoir rock and formation water should be characterized in advance to enable prediction of CO2– rock–brine interactions and development of a monitoring strategy. For nonreactive reservoirs, potential risks related to mineral reactions are low, and monitoring should focus on the trapping mechanism and potential structural-leakage pathways or abandoned wells; for reactive reservoir systems, storage efficiency may be improved because CO2 would be consumed and converted to solid phases by mineral reactions. However, thorough studies need to be conducted on potential impacts of mineral interactions on reservoir mechanical integrity, porosity and permeability (injectivity), rock framework velocity, etc. Further studies on the Middle Tuscaloosa Formation, the primary confining unit for the injection zone, need to be conducted to evaluate reactivity of the mudstone formation in contact with CO2-saturated brine and supercritical CO2. Studies of mudrock–CO2 interactions based on and calibrated to field observations are lacking. A similar field-based program coupled with laboratory experiments would improve our understanding of long-term caprock performance. 6. Conclusions Multiple phases of geochemical monitoring were conducted during CO2 injection at the Cranfield CO2-EOR and sequestration site. A laboratory autoclave experiment was conducted using reservoir rock and brine samples to investigate CO2–rock–brine reactions under reservoir conditions. Field data and experimental results agree well and show that interactions of CO2 with rock-forming minerals are limited, in contrast to those of field observations at other sites. Brine chemistry is

References Bachu, S., Bonijoly, D., Bradshaw, J., Burruss, R., Holloway, S., Christensen, N., Mathiassen, O., 2007. CO2 storage capacity estimation: methodology and gaps. International Green House Gas Control 1, 430–443. Bish, D.L., 1994. Quantitative X-ray diffraction analysis of soil. In: Amonette, J.E., Zelazny, L.W. (Eds.), Quantitative Methods in Soil Mineralogy. Soil Science Society of America, Wisconsin, pp. 267–295. Deines, P., Langmuir, D., Harmon, R.S., 1974. Stable carbon isotope rations and the existence of a gas phase in the evolution of carbonate ground water. Geochimica et Cosmochimica Acta 38, 1147–1164. Department of Energy (DOE), National Energy Technology Laboratory (NETL), 2010. Carbon Sequestration Atlas of the United States and Canada—Atlas, third ed. http://www.netl.doe.gov/technologies/carbon_seq/refshelf/atlasIII/. Duan, Z., Sun, R., 2003. An improved model calculating CO2 solubility in pure water and aqueous NaCl solutions from 273 to 5333 K and from 0 to 2000 bar. Chemical Geology 193, 257–271. Emberley, S., Hutcheon, I., Shevalier, M., Durocher, K., Mayer, B., Gunter, W.D., Perkins, E.H., 2005. Monitoring of fluid–rock interaction and CO2 storage through produced fluid sampling at the Weyburn CO2-injection enhanced oil recovery site, Saskatchewan Canada. Applied Geochemistry 20, 1131–1157. Freifeld, B.M., Trautz, R.C., Kharaka, Y.K., Phelps, T.J., Myer, L.R., Hovorka, S.D., Collins, D.J., 2005. The U-tube: a novel system for acquiring borehole fluid samples from a deep geologic CO2 sequestration experiment. Journal of Geophysical Research 110, B10203. doi:10.1029/2005JB003735. Gunter, W.D., Perkins, E.H., McCann, T.J., 1993. Aquifer disposal of CO2-rich greenhouse gases: reaction design for added capacity. Energy Conversion and Management 34, 941–948. Gunter, W.D., Wiwchar, B., Perkins, E.H., 1997. Aquifer disposal of CO2-rich greenhouse gases: extension of the time scale of experiment for CO2-sequestering reactions by geochemical modelling. Mineralogy and Petrology 59, 121–140. Hillier, S., 1999. Use of an air brush to spray dry samples for X-ray powder diffraction. Clay Minerals 34, 127–135. Johnson, J.W., Nitao, J.J., 2002. Enhanced caprock integrity and self-sealing of the immiscible plume through mineral trapping during prograde and retrograde CO2 sequestration in saline aquifers. AAPG Bulletin 86, 161–162. Johnson, G., Mayer, B., Shevalier, M., Nightingale, M., Hutcheon, I., 2011. Tracing the movement of CO2 injected into a mature oilfield using carbon isotope abundance ratios: the example of the Pembina Cardium CO2 Monitoring project. International Journal of Greenhouse Gas Control 5, 933–941. Kaszuba, J.P., Janecky, D.R., Snow, M.G., 2005. Experimental evaluation of mixed fluid reactions between supercritical carbon dioxide and NaCl brine: relevance to the integrity of a geologic carbon repository. Chemical Geology 217, 277–293. Ketzer, J.M., Iglesias, R., Einloft, S., Dullius, J., Ligabue, R., de Lima, V., 2009. Water–rock–CO2 interactions in saline aquifers aimed for carbon dioxide storage: experimental and numerical modeling studies of the Rio Bonito Formation (Permian), southern Brazil. Applied Geochemistry 24, 760–767. Kharaka, Y.K., Cole, D.R., 2011. Geochemistry of geologic sequestration of carbon dioxide. In: Harmon, R.S., Parker, A. (Eds.), Frontiers in Geochemistry: Contribution of Geochemistry to the Study of the Earth. Blackwell Publishing Inc, pp. 133–174. doi:10.1002/ 9781444329957.

J. Lu et al. / Chemical Geology 291 (2012) 269–277 Kharaka, Y.K., Hanor, J.S., 2007. Deep fluids in the continents I: sedimentary basins. In: Drever, J.I. (Ed.), Surface and Ground Water, Weathering and Soils: Treatise on Geochemistry, v. 5. Elsevier Science, pp. 1–48. Kharaka, Y.K., Cole, D.R., Hovorka, S.D., Gunter, W.D., Knauss, K.G., Freifeld, B.M., 2006. Gas–water–rock interactions in Frio Formation following CO2 injection: implications for the storage of greenhouse gases in sedimentary basins. Geology 34 (7), 577–580. Kharaka, Y.K., Thordsen, J.J., Hovorka, S.D., Nance, H.S., Cole, D.R., Phelps, T.J., Knauss, K.G., 2009. Potential environmental issues of CO2 storage in deep saline aquifers: geochemical results from the Frio-I Brine Pilot test, Texas, U.S.A. Applied Geochemistry 24, 1106–1112. Knauss, K.G., Johnson, J.W., Steefel, C.I., 2005. Evaluation of the impact of CO2, co-contaminant gas, aqueous fluid, and reservoir rock interactions on the geologic sequestration of CO2. Chemical Geology 217, 339–350. Lu, J., Milliken, K., Reed, M.R., Hovorka, S., 2011. Diagenesis and sealing capacity of the middle Tuscaloosa mudstone at the Cranfield carbon dioxide injection site, Mississippi. Environmental Geosciences 18, 1–19. Mississippi Oil and Gas Board, 1966. Mississippi Oil and Gas Board (MOGB), Cranfield Field, Cranfield unit, basal Tuscaloosa reservoir, Adams and Franklin Counties, pp. 42–58. Mito, S., Xue, Z., Ohsumi, T., 2008. Case study of geochemical reactions at the Nagaoka CO2 injection site, Japan. International Journal of Greenhouse Gas Control 2, 309–318. Palandri, J.L., Kharaka, Y.K., 2004. A compilation of rate parameters of water–mineral interaction kinetics for application to geochemical modeling. U.S. Geological Survey, Open File Report 2004–1068. California, 64 pp.

277

Palandri, J.L., Kharaka, Y.K., 2005. Ferric iron-bearing sediments as a mineral trap for CO2 sequestration: iron reduction using sulfur-bearing waste gas. Chemical Geology 217, 351–364. Portier, S., Rochelle, C., 2005. Modelling CO2 solubility in pure water and NaCl-type waters from 0 to 300 °C and from 1 to 300 bar: application to the Utsira Formation at Sleipner. Chemical Geology 217, 187–199. Stancliffe, R.J., Adams, E.R., 1986. Lower Tuscaloosa fluvial channel styles at Liberty Field, Amite County, Mississippi. Gulf Coast Association of Geological TRANSACTIONS 36, 305–313. White, S.P., Allis, R.G., Moore, J., Chidsey, T., Morgan, C., Gwynn, W., Adams, M., 2005. Simulation of reactive transport of injected CO2 on the Colorado Plateau. Chemical Geology 217, 387–405. Xu, T., Apps, J.A., Pruess, K., 2003. Reactive geochemical transport simulation to study mineral trapping for CO2 disposal in deep arenaceous formations. Journal of Geophysical Research B: Solid Earth 108, 3–13. Xu, T., Apps, J.A., Pruess, K., 2005. Mineral sequestration of carbon dioxide in a sandstone– shale system. Chemical Geology 217, 295–318. Xu, T., Kharaka, Y.K., Doughty, C., Freifeld, B.M., Daley, T.M., 2010. Reactive transport modeling to study changes in water chemistry induced by CO2 injection at the Frio-I Brine Pilot. Chemical Geology 271, 153–164. Zerai, B., Saylor, B.Z., Matisoff, G., 2006. Computer simulation of CO2 trapped through mineral precipitation in the Rose Run Sandstone. Applied Geochemistry 21, 223–240. Zhang, W., Li, Y., Xu, T., Cheng, H., Zheng, Y., Xiong, P., 2009. Long-term variations of CO2 trapped in different mechanisms in deep saline formations: a case study of the Songliao Basin, China. Green House Gas Control Technology 3 (2), 161–180.