Supercritical CO2 uptake by nonswelling phyllosilicates.

Interactions between supercritical (sc) CO2 and minerals are important when CO2 is injected into geologic formations for storage and as working fluids for enhanced oil recovery, hydraulic fracturing, and geothermal energy extraction. It has previously been shown that at the elevated pressures and temperatures of the deep subsurface, scCO2 alters smectites (typical swelling phyllosilicates). However, less is known about the effects of scCO2 on nonswelling phyllosilicates (illite and muscovite), despite the fact that the latter are the dominant clay minerals in deep subsurface shales and mudstones. Our studies conducted by using single crystals, combining reaction (incubation with scCO2), visualization [atomic force microscopy (AFM)], and quantifications (AFM, X-ray photoelectron spectroscopy, X-ray diffraction, and off-gassing measurements) revealed unexpectedly high CO2 uptake that far exceeded its macroscopic surface area. Results from different methods collectively suggest that CO2 partially entered the muscovite interlayers, although the pathways remain to be determined. We hypothesize that preferential dissolution at weaker surface defects and frayed edges allows CO2 to enter the interlayers under elevated pressure and temperature, rather than by diffusing solely from edges deeply into interlayers. This unexpected uptake of CO2, can increase CO2 storage capacity by up to ∼30% relative to the capacity associated with residual trapping in a 0.2-porosity sandstone reservoir containing up to 18 mass % of illite/muscovite. This excess CO2 uptake constitutes a previously unrecognized potential trapping mechanism.

CO2 is increasingly used in many deep subsurface engineering processes, such as enhanced oil recovery, geological carbon storage in saline formations, depleted gas and oil fields (1, 2), waterless hydraulic fracturing, and enhanced geothermal systems (3, 4). Tight rocks (shales and mudstones) function as caprocks in geological CO2 sequestration and as host rocks in unconventional gas and oil extractions. Reactions with supercritical CO2 (scCO2) under elevated pressure and temperature conditions of the deep subsurface can significantly change physical and chemical properties of rocks and minerals. CO2 uptake can lead to rock volume expansion, increased rock stress, and geomechanical deformation (5–7). Wettability alteration resulting from scCO2 reactions has been reported by a number of researchers (8–10), although the measurements can contain large uncertainties (11). Dewetting and changes in capillary forces controlling saturation have been reported to result from long-term exposure to scCO2 (12–16), and strong CO2 adsorption can increase CO2 storage capacity and CO2 preferential adsorption enhance petroleum gas release (17, 18).

Wilson et al. (19) reviewed the mineralogy of the major unconventional hydrocarbon shale reservoirs in the United States and reported that the clay mineralogy of practically all of the shale reservoirs older than the Upper Cretaceous is dominated by illitic clay, as discrete illite and/or mixed-layer illite/smectite. There is no occurrence of smectite as a discrete phase in the shale formations older than the Upper Cretaceous, although it is not clear whether all of the illite is derived through illitization of a smectite precursor (19). The name “illite,” given by Grim, Bray, and Bradley in 1937 (20), is a general term for the mica-type mineral with 10-Å spacing that does not expand on treatment with glycol. Although illite belongs to the mica group, compared with the well-crystallized micas, illite has a lower degree of order, smaller particle size, lower potassium content, and variable cation population in its tetrahedral and octahedral structure (20). Smectite is a swelling or expandable clay with the interlayer spacing dependent on hydration state and counterion occupancy. Many recent studies reported that smectites have strong CO2 adsorption capacity and that CO2 can intercalate into smectite interlayers (21–29). Espinoza and Santamarina (30) compiled petrographic properties of shale caprocks from eight carbon storage sites and showed that illitic clay is the most prominent mineral (can be up to 70% of the caprock). In sandstone reservoirs, illite can constitute up to 18% of the rock mass (31).

In addition to being common in the deep subsurface sediment rocks, muscovite is widely used in material sciences research and in many engineering processes, because of its high chemical and physical stability, and as large and easily cleaved tetrahedral–octahedral–tetrahedral sheets. Upon cleaving, muscovite yields atomically smooth and large basal surfaces. Because of these properties, we chose muscovite as the model nonswelling 2:1 phyllosilicate mineral to study. Its formula KAl2(AlSi3O10)(F,OH)2 represents two ditrigonal tetrahedral (T) Si sheets (with ∼25% Al substitution) and an octahedral (O) Al sheet in between. Muscovite dissolution rates increase with acidity (important for water equilibrated with CO2) and alkalinity (32). Illite and muscovite are both dioctahedral 2:1 layered phyllosilicate minerals, and they have similar 00l reflection positions. They are differentiated by diffraction of random-ordered mounts (33). Illite has a small number (<5%) of smectite interlayers. Muscovite does not have any interstratified layers with smectite. Thus, using muscovite as a proxy of illite to evaluate its reactivity with scCO2 is appropriate, although probably not the other way around. It was reported that Cs+ ions can intercalate into frayed edges of illite interlayers, resulting in interlayer expansion (34). Hu et al. (35) and Hu and Jun (36) investigated biotite dissolution in CO2-saturated brine and observed surface cracking and formation of illite, goethite, and kaolinite. They speculated that CO2 intercalation into biotite interlayers upon exposure to scCO2 saturated brines. Although biotite is less common in deep subsurface sedimentary rocks due to its high reactivity and transformation to illite and kaolinite, it does share the common mica structure with muscovite. Molecular dynamics simulations of CO2 diffusion within muscovite interlayers predict that transport may be significant when the slit pore dimensions are >0.25 nm (37). Gases trapped within muscovite interlayers and released during cleaving in ultrahigh vacuum amounted to 0.1–1 molecules per nanometer squared of cleaved surface area (38). Wan et al. (11) observed blisters at muscovite surface upon reaction with water–scCO2, and they speculated that the blisters are possibly the result of scCO2 diffusion into the interlayer from the edges of the mica plate and gas expansion upon depressurization. That conjecture motivated further research presented here. It should be noted that the atomic force microscopy (AFM) images presented in figure 7 in Wan et al. (11) contained contaminants in addition to blisters (distinguished by the phase images that we realized later).

To investigate how muscovite responds to CO2 exposure under deep subsurface conditions, we incubated single muscovite plates (highest purity of commercially available natural crystals, freshly cleaved) first with four different fluids under otherwise identical conditions—wet-scCO2, dry-scCO2, wet-scN2, and scCO2-saturated brine—and then focused experiments on the most relevant and interesting fluid: water-saturated-scCO2. The experiments were conducted at 12 MPa and 90 °C in a high-pressure and -temperature vessel, over times ranging from 7 d up to 1 mo. The incubation experiment setup is shown in Fig. S1. Then, a series of measurements were conducted on the reacted and unreacted control samples. The morphology of basal surfaces was examined with AFM, and the interlayer spacing was measured by X-ray diffraction (XRD) (ex situ). The surface chemistry was examined by X-ray photoelectron spectroscopy (XPS). An off-gassing experiment from reacted muscovite plates was performed to measure CO2 release from reacted muscovite plates (the off-gassing experiment setup is shown in Fig. S2).

Results

Surface Morphology of Basal Surfaces.

The AFM analyses of the control sample showed atomically smooth basal surfaces with measured roughness equal to the instrument limit (Fig. 1 and Fig. S3 ). For the samples incubated with scCO2-saturated brine (sample was submerged into 10 mM NaCl solution that coexisted with scCO2), the basal surfaces were roughened, indicative of muscovite dissolution (Fig. S3 ). For the samples incubated with dry-scCO2 and wet-scN2, the images show sparsely distributed adventitious nanoparticles that are considered practically unavoidable on high-energy muscovite surfaces (39). Surprisingly, dome-shaped features were observed on the basal surfaces of samples incubated in wet-scCO2, but not under the other conditions. Fig. 1 shows two selected areas from different samples. The homogeneity of the phase images (Fig. 1 ) indicates that the surfaces are chemically uniform (indicating no contamination). The slight variations in phase are predominately associated with changes in amplitude (Fig. 1 ) due to topography, but the phase shift associated with the large feature in Fig. 1 indicates that it has a lower stiffness than unaltered mica. The topographic variations are not caused by deposition of foreign material or contamination onto the basal surface. The structures are distributed over the basal surface, with sizes ranging from tens to hundreds of nanometers across and up to 2–3 nm tall. We did not observe significant variations in feature size and population with incubation time (7 and 14 d). We also did not observe discernable variations in size or population at different distances from the edges of muscovite plates. We interpreted the domed features to be blisters formed by the expansion of CO2 intercalated into mica when the reactor vessel was depressurized from 12 MPa. Because the blistering occurred only on the samples incubated with wet-scCO2, it appears that the water film is required for CO2 entry. We also found numerous examples of features that appeared to be collapsed blisters (e.g., Fig. 1 ), suggesting escape of trapped CO2 from local surface defects after depressurization. It is worth noting that we do not expect blisters to form under subsurface conditions because depressurization and gas expansion do not occur.

Fig. 1.

AFM images of muscovite basal surface showing blistering after exposure to moist supercritical CO2. (A) A 3-µm-height image. (B–D) A 1-µm zoom-in of boxed region of A with section across a blister (B) including amplitude (C) and phase (D) data. (E) Height image of freshly cleaved mica as baseline instrument noise. (F) A 2-µm-height image. (G–I) A 350-nm zoom in of boxed region of F with section across a collapsed blister (G) including amplitude (H) and phase (I) data. All height images have an 8-nm color scale. In B and G, height profiles along dashed lines are shown in Inset graphs.

Interlayer d-Spacing Measurements.

The ex situ XRD measurements obtained on single muscovite plates (samples previously incubated under elevated pressure and temperature, then measured under 1-atmosphere pressure conditions) showed no alteration of the basal spacing; with measured values of d(002) spacing 9.97 ± 0.028 Å in both control and wet-scCO2–incubated plates, consistent with the standard value (40). The bulk muscovite interlayers did not swell in response to exposure to scCO2. Indeed, d-spacing alteration of bulk samples is not expected, giving the strong electrostatic binding between interlayer K+ and the negative charges localized in the immediately adjacent silicate layers (41).

Chemical Characterization of the Basal Surface.

XPS is a surface-sensitive spectroscopic technique. Approximately 95% of the detected photoelectrons originate from the surface down to the depth of 3λ, with λ being the electron inelastic mean free path. In muscovite, λ ∼ 0.86–0.88 nm for photoelectrons associated with the C 1s binding energy (283–290 eV) (42); thus, 3λ can include C photoelectrons emanating from the upper surface down through the two interlayers. The XPS analyses were conducted to compare freshly cleaved muscovite samples with samples exposed to scCO2-saturated brine, and to water-saturated scCO2. Survey scans showed peaks for expected muscovite elements: Al, Si, O, K, and, in a few cases, F. In addition, the only nonmuscovite element present was C. Analysis of the control sample yielded 150 ppm total C (total organic carbon/total inorganic carbon analyzer; Shimadzu TOC-VCPH), even though it was considered of highest purity. High-resolution XPS scans were conducted over the O 1s region to calibrate spectra to the O 1s peak at 531.4 eV for muscovite (43) and over the C 1s and K 2p energy regions (Fig. 2). Suitability of the calibrations to the O 1s peaks was demonstrated through good agreement of the K 2p peaks with literature values (43). All samples including the control displayed C 1s peaks of varying intensity in the organic/adventitious/graphitic region at ∼285–284 eV, and most also contained a peak in the range of 282–281 eV associated with carbidic C (43). Native and adventitious C of reduced-organic nature have previously been reported in XPS measurements of freshly air-cleaved muscovite (39, 44), and the highest-grade muscovite has been reported to contain graphitic carbon.

Fig. 2.

XPS high-resolution scans indicating interlayer CO2 and CO2 releasing as time in vacuum. The distinct C 1s peak at 290.5 eV (CO2) appeared only on samples incubated with wet-scCO2 and was absent from the control and CO2-saturated brine-incubated samples. Note the intensity of C 1s peak collected at a single location decreased with time (30–240 min for one sample and 40–220 min for another) under the 10−9 Torr vacuum. Spectra for the other conditions were collected at randomly selected locations on samples, generally within 0.5–3 h after removal from the reactor.

Unique to samples incubated in wet scCO2 was a distinct peak at ∼290.5 eV, adjacent to the K 2p3/2 peak. No evidence of resonance in the vicinity of 290.5 eV was observed for either the control sample or the muscovite incubated in CO2-saturated brine. It should be noted that the carbidic, graphitic, and organic C peaks all occurred at binding energies that were significantly lower than those associated with CO2 and carbonates. Note that physisorbed CO2 would only be expected to persist with (at most) monolayer coverage on basal surfaces after samples were returned to atmospheric pressure and that most of this CO2 should be desorbed during evacuation of the XPS sample chamber (10−9 torr) before collection of XPS data. Thus, it is reasonable to believe that the measured peak at ∼290.5 eV is interlayer trapped CO2. Previously, Bhattacharyya (44) detected very small amounts of C adsorption on muscovite (C 1s peak intensity ∼15% that of K 2p3/2) following low pressure (10−5 torr) exposure to CO2, and loss of CO2 with exposure to higher vacuum (44). Time-dependent losses of CO2 from our scCO2-exposed muscovite were evident from the decreases in the 290.5-eV peak (Fig. 2) through collection of spectra at the same spot for three to four times within up to 4 h under the ultrahigh vacuum. This suggested that CO2 diffused out of the near-surface interlayers during the course of the measurement in the vacuum chamber. We therefore designed a separate off-gassing experiment that allowed quantification of CO2 release from muscovite.

Off-Gassing of CO2 from Reacted Muscovite Samples.

To quantify adsorbed and surface region-associated CO2, off-gassing measurements were performed through comparing samples exposed to wet-scCO2 with control samples. Each sample consisted of six small muscovite plates with the total surface areas of 33.6 cm2 from basal surfaces and 1.5 cm2 from edges. For the scCO2-exposed samples, the mica plates were transferred to a N2 glove bag after removal from the depressurized reactor, to purge weakly sorbed CO2 off external surfaces of samples, then sealed in a glass serum bottle for later analyses of off-gassed CO2 (photographs of the sample vials are shown in Fig. S2). The results presented in Fig. 3 show the cumulative CO2 released from each of three or four replicate samples. The CO2 recovered from the control samples showed amounts similar to those obtained from the blanks (empty serum bottles), reflecting a low background CO2 off-gassed from the septum stoppers.

Fig. 3.

Comparison of CO2 release with bulk surface area, indicating CO2 interlayer uptake. (A) Cumulative CO2 recovered from blank (sealed empty vials), control, and scCO2-exposed samples of three or four replicates. (B) Blank-subtracted CO2 recovery per unit bulk surface area of muscovite (left vertical axis) and as equivalent numbers of CO2 monolayers (right vertical axis).

In contrast, scCO2-exposed muscovite samples yielded approximately seven times more CO2 relative to the controls and blanks. After subtracting the average of blank samples, the controls were found to release negligible CO2. Fig. 3 shows CO2 release per unit bulk surface area of the muscovite plates along the left vertical axis. Using an area of 0.16 nm2 per CO2 molecule (45), the cumulative CO2 released was also expressed in terms of equivalent monolayer thicknesses along the right vertical axis of Fig. 3. A condensed film of ∼15 molecular layers of CO2 on the exterior surfaces of muscovite clearly was not present after depressurization of the reactor and removal of the samples within the purged N2 glove bag. While large, this amount of released CO2 was insignificant compared with the storage capacity expected if intercalation was pervasive, because each ∼0.3-mm-thick muscovite plate contained ∼3 × 105 interlayers. We therefore conclude that the CO2 recovered in the off-gassing experiment must have originated from shallow depths in exposed, scCO2-reacted basal surfaces and along edges disturbed during cutting of plates.

Discussion and Implications

Although the AFM, XPS, and off-gassing measurements of control vs. wet-scCO2–reacted samples suggested possible CO2 intercalation into muscovite interlayers, the XRD measurements showed no interlayer expansion. A key question is whether CO2 molecules can diffuse into muscovite interlayers without expansion. The interlayer widths between centers of two opposite oxygen planes was ∼0.34 nm (44, 46), which, upon accounting for oxygen radii (0.121 nm) (47), left a minimum gap of only 0.10 nm (Fig. 4). The separations between O on individual Si tetrahedral provided wider channels, ∼0.2 nm, within interlayers. However, these channels were substantially smaller than the 0.33-nm kinetic diameter for CO2 (48). Even the smaller diameters proposed for the effective cross-section of CO2, ranging from 0.31 (45) to 0.232 nm (49), remained prohibitively large, and K+ occupying all siloxane cavities present in-plane obstacles (Fig. 4). While previous studies showed release of CO2 (48) and even larger molecules N2 (0.364–0.380 nm) and CO (0.369 nm) (38) from muscovite interlayers, and our AFM, XPS, and off-gassing analyses collectively showed the CO2 interlayer trapping, our comparisons of interlayer dimensions and molecular sizes suggest that these gases cannot intercalate the interlayers of an undisrupted crystal structure.

Fig. 4.

Comparisons between muscovite interlayer dimensions and assumed 0.33-nm kinetic diameter of CO2. (A) Edge view, showing 0.34-nm average distance between centers of two oxygen on opposing ditrigonal silica sheets result in minimum gaps of ∼0.10 nm. (B) Basal surface view, showing the occupancy of all ditrigonal cavities by K+ imposes additional 0.19-nm constrictions within the interlayer plane.

Based on these interlayer dimensions and molecular size considerations and the lack of increased blistering near the edges of the muscovite plates, as well as the muscovite solubility considerations in the scCO2-acidified water films described below, we hypothesize that CO2 enters mainly through local defects formed on basal surfaces rather than by diffusing from the edges deeply into interlayers. Our experiment results showed that only the reaction with wet-scCO2 caused blistering, while scN2, dry-scCO2, and scCO2-saturated brine did not. This suggests that a scCO2-acidified water film between bulk scCO2 and muscovite is the necessary condition for CO2 molecules to enter the interlayers. Previous studies have predicted and demonstrated the stability of nanometer-scale brine films on muscovite under scCO2 confinement (50, 51). Knauss and Wolery (52) reported that muscovite dissolution rates increased as pH decreased <0.1 MPa and 70 °C. Assuming a pH of 2.9 for the brine film (53) and a dissolution rate of 1.7 × 10−12 mol⋅m−2⋅s−1 (52) leads to a predicted average dissolution depth of 0.15 nm within 7 d of reaction. Nonuniform distributions of Al substitutions for Si may lead to localized sites on basal surfaces that are more susceptible to dissolution in the adsorbed scCO2-acidified water films, allowing CO2 to enter damaged locations under elevated pressure and temperature. Vacancies of K+ ions and the other crystal defects could provide the necessary spaces for CO2 entry. The volume fraction associated with these defects would be too small to affect XRD measurements of bulk samples. Although this hypothesis on the mechanism for CO2 intercalation needs further testing, our calculated additional CO2 storage capacity in reservoirs based on the measured CO2 uptake by muscovite shows the significance of this partitioning.

The potential significance of CO2 uptake by muscovite, and by extension onto illite, on enhancing reservoir storage capacity becomes evident upon comparison with storage attributable to capillary trapping, one of the main CO2 storage mechanisms (1, 2). The muscovite/illite-associated CO2 capacity can be approximated as the product of equivalent CO2 uptake per muscovite/illite surface area times the abundance of this surface area per bulk volume of reservoir rock. As shown in (Tables S1 and S2), CO2 partitioning onto muscovite/illite under representative reservoir conditions may amount to as much as 30% of the capacity expected from capillary trapping in sandstone (Fig. S4). Future measurements of CO2 excess sorption isotherms on muscovite could provide additional insights into partitioning over a broad range of reservoir pressures.

The results presented in this work are the beginning stages toward understanding CO2 uptake by nonswelling clays. The pathways for CO2 entering the interlayers are still hypotheses needing more testing, especially under in situ (high-pressure and -temperature) conditions. The rates and extent of the CO2 uptake, and dependence on varying geochemical conditions, need to be quantified to apply this knowledge. We hope this preliminary research will stimulate more creative studies on this topic, especially given the abundance of illite and muscovite, and their importance in many energy and environment issues.

Materials and Methods

Materials.

V-1–grade natural muscovite plates (catalog no. 71855-05; Electron Microscopy Sciences) were used. High-purity CO2 (99.99%; Praxair), ultrapure deionized water (resistivity ≥18 MΩ; Milli-Q), and American Chemical Society-grade chemicals were used. Muscovite plates (70 × 25 × 0.3 mm thick) were cut and then cleaved to ∼35 × 8 × 0.2 mm before placement in the reactor. The cleaved fresh surfaces were marked, and only the fresh surfaces were used for after-reaction AFM and XPS characterizations.

High P-T Muscovite–CO2 Incubation.

The incubations were conducted in a high-pressure–temperature Hastelloy vessel (Parr; model 4560), with the pressure controlled by an ISCO pump (model 500HP; Teledyne ISCO) and temperature-controlled by a Parr Temperature Control Module. The pressure (12 ± 0.01 MPa) and temperature (90 ± 0.1 °C) were constants through the experiments. Under these pressure and temperature conditions, water solubility in scCO2 was ∼1.5%, and CO2 in water was ∼1.8% (54). The reaction durations were 7, 14, and 30 d for different tests. Each experiment was repeated three times or more, with much attention given to avoiding potential mica surface contamination. For the reaction with CO2-saturated brine, both the inside and outside of the polytetrafluoroethylene (PTFE) bottle were filled with brine up to the holes of the PTFE bottle, and the remaining space was filled with scCO2. For the reaction with water-saturated-scCO2, 3 mL of deionized water was placed inside the bottom of the vessel, outside the PTFE container, to ensure water saturation in scCO2 without direct contact between muscovite plate and bulk water. For the reaction with dry-scCO2, no water added in the reactor and the pure CO2 was used as-is from the tank. The experiment setup is illustrated in Fig. S1.

AFM.

The nanometer-scale surface morphological features of the control and reacted samples were quantified by using the tapping mode on a Bruker Icon AFM. Height, amplitude, and phase images were collected simultaneously. Tap150-G (BudgetSensors) tips were used (drive frequencies were ∼150 kHz, and scan rates ranged between 0.3 and 1 Hz). For each sample, at least five areas over the basal surface were analyzed, and at each area, several images at different scales were recorded. The Nanoscope Analysis software (Version 1.40; Bruker) was used to analyze the images and calculate the area and line roughness.

XPS.

XPS (Physical Electronics PHI 5300) was used to determine the chemical composition of the basal surface for unreacted muscovite and samples reacted with CO2-saturated brine and water-saturated scCO2. The spectra were acquired with an Al Kα X-ray source operated at 1,486.6 eV and 350 W (15 kV–25 mA) in a 10−9 torr vacuum chamber. Measurements were taken with sample stage flat (take-off angle of 35.3° from the surface). Survey spectra were acquired with 1 eV per step and pass energy of 178.95 eV. High-resolution scans were acquired with 0.1 eV per step and 35.75 eV pass energy. The energy axis was calibrated by setting the O 1s peak to 531.4 eV on the high-resolution scan (43, 55). After calibration, the Al 2p, Si 2p3/2, and K 2p3/2 peaks all lined up within 0.4 eV of their reference values (74.1, 102.4, and 292.8 eV, respectively) for muscovite (43).

XRD.

For the ex situ XRD measurements, muscovite monocrystals (∼45 × 12.5 × 0.2 mm) were used for both control and reacted samples, with the latter incubated in water-saturated scCO2 for 14 d. The diffraction pattern was collected ∼2 h after the beginning of depressurizing the high-P reactor. XRD spectra were collected on a powder diffractometer PANalytica X’Pert pro (Co Kα radiation, 40 kV, 40 mA, X’Celerator detector). Since the samples were single crystals, they were mounted on spinner sample holder, with the basal plane as the sample surface. This procedure simulated a perfectly oriented polycrystal, allowing collecting of all the valid (00l) reflections of muscovite (and only those) in a powder-like diffraction profile, allowing a precise measurement of the basal d-spacing.

Off-Gassing of CO2 from Muscovite.

This experiment was designed to help determine whether or not CO2 entered muscovite interlayers by comparing samples exposed to wet-scCO2 with unexposed control samples. Each sample (four replicates for CO2-incubated and triplicate for controls) consisted of six small plates (35 × 8 × 0.3 mm). Each sample had an external surface area of 35 cm2, primarily from basal surfaces (33.6 cm2), with a smaller contribution from edges (1.5 cm2). The high P-T muscovite–CO2 incubation and sampling steps were conducted as described above (instead of one plate at a time, the 24 small plates were all in one PTFE bottle). After opening the Parr reactor, the PTFE bottle containing all of the mica plates was brought into a N2 glove bag and transferred to a glass dish. The N2 bag was then vacuumed and refilled two times to remove the surface-accumulated CO2 and free CO2. Then, six small plates were placed into a 25.6-mL glass serum bottle under the N2 atmosphere and sealed with a thick blue chlorobutyl septum stopper (Fig. S2). Sampling of the headspace in each bottle, including triplicate empty (blank) bottles, was done by initially collecting 60 mL of gas into a syringe ∼1 h after sealing. Two subsequent 20-mL gas samples were collected from each bottle at 4 and 15 d. The sampled gas was immediately injected into the N2 carrier gas stream flowing into a CO2 gas analyzer (Li-Cor LI-840A) (56). The cumulative CO2 recovered from the scCO2-exposed and control samples were compared after subtraction of measurements obtained on blank samples.