Residence times of nanoconfined CO2 in layered aluminosilicates

Nanoconfinement of CO2 in layered aluminosilicates contributes to the capture and release of this greenhouse gas in soils.


Introduction
Layered aluminosilicate minerals play important roles in technology and nature due to their tremendous solute and solvent exchange capacities. Those of the expandable type can readily confine gases between nanometer-thick interlamellar layers. Exchanges between these materials and CO 2 are under increasing scrutiny to follow the fate of this greenhouse gas in soils and for its deep geological storage. In particular, thawing of soils of the cryosphere is of great concern as mineralbound and newly bio-respired CO 2 are expected to be released into the atmosphere over the next few decades. [1][2][3] Along these lines, a 2018 report 4 by the Intergovernmental Panel on Climate Change highlights the devastating impact that a 1.5°C hike in global warming could have by the end of the century.
The positive feedback to a CO 2 -driven warming climate is of particular concern given that frozen soils contain about one-third of Earth's soil carbon, and that all soils contain about three times more carbon than the atmosphere. 5 Understanding how global warming will alter land-air exchanges requires knowledge of the ability of soil-building minerals to dynamically exchange CO 2 with the atmosphere as soils begin to thaw. This is central to our ability to evaluate possibilities for geological storage of CO 2 .
A primary focus on common soil-building swelling clay minerals is central because this important class of phyllosilicates entraps CO 2 between nanometre-thick layers of negatively charged aluminosilicate layers (Fig. 1). [6][7][8][9][10][11] CO 2 molecules are intercalated with their main axis parallel to the basal face (Fig. 1), either as dimers or as part of the hydration shell of interlayer cations (e.g. sodium). [12][13][14] Studies 9,15-20 under supercritical CO 2 conditions show that humidity facilitates uptake, and that the greatest loadings are achieved with interlayers populated with ∼1 monolayer (1W) of water. Less CO 2 can be stored at greater humidity as thicker water interlayers (>1W) suppress cation-CO 2 and CO 2 -CO 2 interactions. 12 Additionally, other studies 21,22 under environmentally relevant conditions show that very high CO 2 pressures and prolonged exposure times are required to intercalate CO 2 in dehydrated clays. Seasonal and climatic variations in moisture levelsfrom wet to dry and from warm to coldshould therefore exert important changes in the dynamic exchanges of CO 2 with soil-building clay minerals. This can have especially important implications on the loadings and isotopic signatures of soil inorganic carbon. 23,24 To address this issue, we measured the rates of CO 2 release from the interlayer of montmorillonite (MMT), a representative soil-building swellable clay mineral. The rates were calculated from time-resolved vibration spectra of CO 2 released by venting dry and hydrated MMT over a range of temperatures well covering those of the cryosphere up to the warmest regions of the planet. Molecular simulations provided the insight needed to explain molecular interactions controlling CO 2 mobility in the interlayer region. In this study we suggest that the dynamic breathing 25 of CO 2 by clays will be attenuated by warming soils.

Results and discussion
CO 2 was intercalated in the interlayer of MMT upon exposure to 20 000 ppm CO 2 , a representative amount in high CO 2 soils. 26,27 To uncover the intrinsic abilities of MMT to dynamically exchange CO 2 , intercalation was first carried out on MMT with the lowest levels of water that can be achieved after a 30 min period of outgassing in vacuo. We note that all final water loadings were no more than 26 mg H 2 O per g MMT (∼1.4 mmol H 2 O per g), which is equivalent to (i) ∼23% of a 1W (∼112 mg H 2 O per g MMT) 28 or (ii) ∼1.6 H 2 O per interlayer Na + if water is homogeneously distributed throughout the particles (cf. ESI † for details and Fig. S1-S5). This hydration level can therefore be regarded as one of the lowest that could be achieved by MMT in dry air at ambient temperatures. Additionally, it is responsible for maintaining a partially opened interlayer region needed to intercalate CO 2 . The expanded interlayer of our samples was confirmed via mixed layer modelling of X-ray diffraction data, as discussed in the ESI † (Fig. S5). CO 2 intercalated in Na-exchanged MMT was readily identified through a single asymmetric band of the stretching mode (ν 3 ) at 2343 cm −1 (Fig. 2a). The 2343 cm −1 band of CO 2 in this MMT was most clearly seen once free gaseous and sorbed CO 2 were evacuated from the clay ( Fig. 2a and S6 †). It was strongly comparable to that for supercritical CO 2 in clays 15,16 and in liquid 29 or solid 30,31 water, as well as in organic-based frameworks. 32 The confining effects of the MMT interlayer on CO 2 are also confirmed by the absence of rotational sidebands, which are typical of gaseous CO 2 (Fig. 2a). 33 The spectra provided no other evidence of changes in water content, (bi)carbonate formation or the formation of solid CO 2 ice 34 (Fig. S2 †).
Intercalated CO 2 was systematically removed by prolonged exposure to vacuum, a means to simulate the action of venting on soils by displacing pore gases outside clay mineral particles. No or little changes in ∼1.4 mmol H 2 O per g interlayer water took place during CO 2 release (Fig. S2 †). While complete removal took place well within ∼10 min at 60°C, it took a substantially longer time than the ∼20 min originally needed to intercalate CO 2 at lower temperatures. Consequently, this provided a first level of indication pointing to molecular forces responsible for holding CO 2 in the interlayer region.
The loss in area (a) of the 2343 cm −1 band with evacuation follows first-order reaction kinetics (lnĲa/a 0 ) = −kt, where a 0 is the band area prior to venting and k is the reaction rate), confirming that the loss of CO 2 from the interlayer region is predominantly a simple molecular desorption process (Fig. 2b). These rates revealed CO 2 residence times (τ = 1/k) ranging from τ = ∼6 min at 60°C to ∼1616 min at −50°C. Additionally, these rates were unchanged when CO 2 was vented by a flow of 101 kPa N 2 Ĳg) instead of vacuum (Fig. 3).
The temperature dependence of these rate constants obeys the Arrhenius relationship (ln k = ln A + E a /RT), with an activation energy of E a = 34 kJ mol −1 (Fig. 2c). This value is ∼2.3 times larger than the one estimated for site-hopping of CO 2 within the interlayer 13 but is most comparable to Fig. 2 (a) Asymmetric (ν 3 ) stretching region of CO 2 of MMT (∼0W) originally exposed to CO 2 Ĳg) at −30°C during the 200 min evacuation (<0.3 Pa). (b) Examples of loss of the normalized band area of the 2343 cm −1 band during evacuation of MMT originally exposed to 20 000 ppm CO 2 . Lines are generated from a first-order kinetic model. (c) Arrhenius plot of the first-order rate constants for the release of CO 2 from the interlayer region of MMT with ∼0W. The release rate of CO 2 resulting from evacuation with 101 kPa N 2 Ĳg) is shown through the pink data point at −30°C (see Fig. S4 † for data and analysis). Enhanced release rates with ∼1W and 2W at −10°C and −50°C, respectively, are shown in turquoise.   37 We note that the pre-exponential factor (A = 3.8 × 10 4 min −1 ) is seven to nine orders of magnitude lower than typical values for in vacuo desorption of CO 2 from open mineral surfaces. 36,37 We ascribe this difference to the effects of interlayer confinement, which is expected to decrease the frequency of desorption events.
In another set of experiments, CO 2 removal rates were tracked in MMT containing interlayer water at −10°C and −50°C (Fig. 4). The rates were generally 2-4 ln k larger than in ∼1.4 mmol H 2 O per g but were also strongly coupled to the stability of interlayer water ( Fig. 4b and c), which we estimate here from the O-H stretching region of water, as explained in Fig. 4 and S8. † The loss of water in 1W-bearing MMT at −10°C was rapid within the first ∼10 min (τ = 14 min) but then slowed considerably after ∼40 min of evacuation (τ = 1213 min) (Fig. S9 †). CO 2 evacuation rates followed accordingly, as highlighted by the instantaneous rates shown in Fig. 4a. In contrast, higher loadings of interlayer water were retained over longer time periods at −50°C (Fig. 4c).
High CO 2 removal rates were therefore maintained under extended periods of venting.
Theoretical calculations of CO 2 diffusion in the interlayer region of Na-exchanged MMT added further insight into the mechanisms affecting the removal of CO 2 during venting ( Fig. 5) (for basal spacing data, see Fig. S9-S11 †). First, we note that simulations confirmed previous findings 19,20 showing that oxygen atoms of CO 2 at low hydration levels predominantly lie in the middle of the interlayer, with one oxygen atom above Si tetrahedra and the other over ditrigonal cavities (Fig. 1). The carbon atom interacts predominantly with the basal oxygens sharing a Si-tetrahedral edge, or occasionally with a corner sharing tetrahedral oxygen when CO 2 oxygen atoms are positioned over two corner-sharing Si tetrahedra. Under completely dehydrated conditions (0W), CO 2 molecules remain in the middle of the interlayer nanopore, but with oxygen atoms coordinating to Na + , which is often in ditrigonal cavities of the basal plane adjacent to sites of isomorphic substitution. Simulations also suggest that large CO 2 loadings favour dimers and clusters with a slipped plane parallel coordination. 13 However, no such configurations could be unambiguously ascertained in the vibrational spectra (Fig. 2a) because of the strong overlap of the main 2343 cm −1 band with those of dimers. 38,39 Simulated CO 2 diffusion coefficients ( Fig. 5a and S10 †) were as small as 10% of bulk H 2 O values 40,41 at both low CO 2 and H 2 O loadings. Values were however greatly enhanced at larger CO 2 loadings, or by intercalation of greater populations of water (see full results in Tables S1-S5 †). We note that our 2W value at low CO 2 loadings falls in line with another simulated value 42 in 2W MMT.
Activation energies for diffusion (E a ), obtained by the Arrhenius relationship to computed diffusion coefficients, were also strongly loading dependent. Values were as large as ∼15-20 kJ mol −1 at the lowest CO 2 loadings, and down to ∼5 kJ mol −1 above 0.4 g CO 2 per g MMT (Fig. S12 †). Expressed on a d-spacing basis, the trend of Fig. 5 denotes a decrease in E a of ∼4-6 kJ mol −1 per nanometer of interlayer expansion, irrespective of water loading. Expansion of the interlayer region, which can result from gains in both CO 2 and water, Fig. 3 Evacuation of CO 2 upon flushing of 200 sccm N 2 Ĳg) to CO 2loaded MMT (∼0.1W) at −30°C. (a) ν 3 region of CO 2 during the ∼200 min evacuation period. In contrast to in vacuo removal, gaseous CO 2 lingers over ∼60 min in the reaction cell as MMT-bound CO 2 begins to be released to the atmosphere. The strong overlap between the spectra of free CO 2 Ĳg) and intercalated CO 2 required a multivariate curve resolution alternating least squares (MCR, version 2004) 35 analyses. (b) MCR spectral components for CO 2 Ĳg) + CO 2 ĲMMT) and CO 2 ĲMMT). (c) Concentration profiles of MCR-ALS components of (b). The values of CO 2 ĲMMT) were used to estimate the first-order rate constant for the evacuation of CO 2 from MMT by N 2 Ĳg).    13 and yet are no more than 50-60% of our experimentally derived value of 34 kJ mol −1 . This latter difference could point to more convoluted processes limiting the interlayer mobility of CO 2 in real materials. For instance, contributions from clay particle packing motifs, frayed particle edges, 43 internal defects or chemisorption could have contributed to these differences. Resolution of these and related phenomena should add further insight needed in predicting gaseous exchange involving natural soil-building materials.

Conclusions
Atmospheric exchanges of CO 2 with MMT are strongly affected by water content and temperature. MMT with low levels of resilient interlayer water (up to ∼1.4 mmol H 2 O per g) stored CO 2 with residence times as long as ∼1616 min at −50°C to ∼6 min at 60°C. The activation energy for the release of CO 2 is 34 kJ mol −1 , and thus at least twice that expected from theoretical simulations on idealised MMT.
Our results suggest that as even the driest soils thaw under a warming climate, we should expect a decrease in the residence times of CO 2 of ∼15 min for each 1°C rise in temperature. These residence times are even shorter under conditions of stable interlayer water, and especially below −10°C , where smaller release rates of water promote CO 2 release by maintaining a large interlayer spacing in MMT. Residence times will nonetheless continue to be longer than in the warmest regions of the planet. However, considering the vast quantities of carbon stored in cold terrestrial environments, these accelerated rates are likely to result in important shifts in land-air fluxes of CO 2 .
Consequently, this work suggests that exposure of CO 2bearing soil clay minerals to increasingly longer periods of thawing will accelerate fluxes of CO 2 to the atmosphere. Conversely, exposure to cold and dry periods will increase the residence time of CO 2 in soils. Given that soil warming of the planet's cold regions is expected to be faster in the winter than in the summer months, 44 wintertime venting of soil CO 2 will likely be an increasing contributor to terrestrial emissions of CO 2 to the atmosphere in the years to come.

Materials
The MMT was Wyoming clay powder (SWy-2) obtained from the Clay Mineral Repository. A portion of this clay was Na-exchanged, as detailed by Schaef et al. 45 The Brunauer-Emmett-Teller specific surface area (25.3 m 2 g −1 ) was calculated from 90-point adsorption/desorption N 2 gas isotherms (Micromeritics). X-ray photoelectron spectroscopy (Kratos Axis Ultra DLD electron spectrometer) was used to detect the surface compositions of Na-exchanged MMT (Na/Fe/Mg/Al/Si/ O = 0.1 : 0.1 : 0.1 : 1.0 : 2.3 : 7.8) and to confirm that the particles were free of impurities, other than adventitious forms of atmospheric carbon at the 285.0 eV line.

Fourier transform infrared spectroscopy (FTIR)
Aqueous suspensions of MMT of 25 g L −1 in deionized water (18.2 MΩ cm) were equilibrated for 30 min and then centrifuged at ∼2000g for 10 min. The top-to-middle portion of the centrifuged wet paste was sampled to avoid accessory minerals and then applied as a thin layer on a fine tungsten mesh (Unique wire weaving, 0.002″ mesh diameter). Then, it was dried under N 2 Ĳg) overnight, and the mass (7-8 mg) of the resulting dry MMT film was determined. Next, the dried sample-tungsten support assembly was inserted into an electrically heated or N 2 Ĳl)-cooled copper shaft connected to a K-type thermocouple used to monitor temperature. The shaft was thereafter inserted into a transmission optical gas reaction chamber (AABSPEC #2000-A) equipped with CaF 2 windows.
The sample was dried further in vacuo (<0.3 Pa, the detection limit of the capacitance manometer; MKS, Baratron) in the reaction chamber for 30 min at 25°C to remove atmospheric gases bound to the materials (Fig. S1 †). Next, the sample was either heated or cooled to reach the desired temperature in this study (−50°C, −30°C, −10°C, 10°C, 25°C, and 60°C; experiments at 95°C and 135°C were abandoned due to excessively high rates of CO 2 release). Then, the sample was equilibrated for another 30 min. During this time, residual water contents were monitored through the O-H stretching (∼3250 cm −1 ) and water bending (∼1630 cm −1 ) bands. The water content in the clay was quantified using the ratio of the intensities of the 1630 cm −1 (ν 3 from interlayer water) and the 1850 cm −1 combination band (Si-O stretch of MMT + OH deformation), 46,47 using the water vapor binding data from Yeşilbaş et al. 28 A Matlab (The MathWorks, Inc.) code using this strategy is given in Table  S6. † CO 2 was intercalated in the samples upon exposure to a flow of 200 square cubic centimetres per minute (sccm) of ∼20 000 ppm CO 2 . This gas mixture was prepared by mixing a 7 sccm flow of pure CO 2 Ĳg) with 193 sccm dry N 2 Ĳg) and controlled by mass-flow controllers (MKS, Baratron). A nondispersible infrared device (Li-7000, Licor, Inc.) was used to continuously monitor the partial pressures of CO 2 Ĳg) and to ensure that the gases were free of H 2 OĲg). Removal of intercalated CO 2 was then observed as a function of time through the asymmetric C-O stretching band (ν 3 ) at 2343 cm −1 upon evacuation of the reaction cell (<0.3 Pa). Repeated sessions of intercalation and evacuation showed that an exposure time of ∼20 min was sufficient to maximize the 2343 cm −1 band. One experiment at −30°C involved CO 2 removal by flushing with 200 sccm N 2 Ĳg).
All FTIR spectra of the samples were collected in transmission mode using a Bruker Vertex 70/v FTIR spectrometer equipped with a DLaTGS detector. The FTIR spectra were collected in the spectral range of 600-4000 cm −1 at 4 cm −1 resolution with a 10 Hz forward/reverse scanning rate. We used the Blackman-Harris three-term apodisation function with 16 cm −1 phase resolution and the Mertz phase correction algorithm. Each spectrum was obtained from 100 coadded spectra and collected over an 89 s period.

Chemometrics
Multivariate curve resolution alternating least squares (MCR, version 2004) 35 50 was used to model clay minerals, and a flexible EPM2 model 19 was used to treat CO 2 molecules. Na + was modelled with the pair potentials originally optimized for the SPC/E water model. 51 In this study, however, water was modelled with a flexible SPC water model. 52 The harmonic and anharmonic terms of this model have better dielectric, vibrational and diffusive properties compared to the SPC water model used in the original parameterization of CLAYFF. CO 2 or H 2 O removal from the interlayer region was mimicked through a custom-made simulation protocol allowing simulation systems to automatically generate sequentially reduced loadings. This was achieved by removing either CO 2 or H 2 O molecules between production runs. Systems of fixed water populations (0-2W) were simulated with varying CO 2 content (0-10 molecules per unit cell; 0.5 g CO 2 g clay −1 ). Systems denoting 1.5 CO 2 and 3.0 CO 2 contained a fixed amount of 1.5 and 3.0 CO 2 molecules per MMT unit cell, and were simulated as a function of water loading by varying between 0 and 20 water molecules per MMT unit cell (0.6 g water g clay −1 ). The initial configurations with maximum loadings were equilibrated by energy minimization, a 50 ps solute restrained simulation under NVT conditions, followed by 5 ns of volume optimization under fixed NPT conditions. The production runs were performed under fixed NPT conditions for 5 ns with a 0.5 fs time step and preceded by a short equilibration step of 100 ps between the evaporative steps. Selected configurations were also simulated for an additional 10 ns using a 1 fs time step under fixed NVT conditions and individual thermostating, in order to obtain diffusion coefficients at −10°C, 10°C, 25°C and 50°C. The simulation results were analysed by computing density maps, radial distribution functions, and diffusion coefficients from the mean square displacement.

Conflicts of interest
There are no conflicts to declare.