Interactions between CO2 and clay minerals have attracted interest in the scientific community in recent years, partly because geological structures are being investigated as storage sites for anthropogenic CO2. The cap-rock formations which act as flow barriers and seals in this context are known to contain high proportions of clay minerals1 and the long-term integrity of these formations is a prerequisite for avoiding CO2 losses to the atmosphere2,3. However, the physical parameters affecting the interactions between CO2 and clay minerals under reservoir conditions are still not well understood4.

Clay minerals are materials based on two-dimensional stacks of inorganic layers5. In some clay minerals (smectites), non-equivalent substitutions of atoms generate a negative charge on each layer surface which is balanced by exchangeable interlayer cations. These cations are responsible for the differences in the physico-chemical behavior of smectites such as water adsorption and retention, plasticity, swelling etc6,7. Smectite clay mineral particles typically consist of approximately hundred layers. Smectites have the ability to intercalate additional molecules into the interlayer space, thereby changing the repetition distance along the layer normal (z-direction), a process which is known as swelling8,9. Intercalation of water can also occur, since H2O is a polar molecule and this has been extensively studied with a wide range of techniques, such as neutron8,10,11 and X-ray scattering9,12,13, NMR spectroscopy14,15,16 tracer experiments17 or numerical modeling14,18.

Experiments19,20,21,22,23,24,25,26,27,28 and simulations18,29,30,31,32 have also shown that CO2 intercalates in some smectite clays, both in supercritical and in gaseous/liquid form. We have recently demonstrated that CO2 is able to intercalate in Na-fluorohectorite (NaFh) smectite clay mineral at conditions close to ambient (−20°C, 5 bar)22. In that work we also showed that under the same conditions neither H2O vapor nor N2 gas intercalates. These are not the typical conditions found in geological storage sites, but the conditions are relevant if clays are considered as a potential material for the capture or sequestration of CO2 and it is also of interest to study CO2 capture and retention under these conditions for the purpose of understanding the underlying molecular mechanisms. Several porous materials are currently being assessed for the purpose of CO2 capture and retention33,34,35. In this context, clay-containing materials could have a distinct advantage in that they are both cheap and ubiquitous31 and also because they generally provide a very large accessible effective surface area that arises from nanolayered stacked structures embedded in a mesoporous powder matrix.

Fluorohectorites (Fh) are synthetic smectites which have been used as a representative and clean model system of natural smectite clays36. Synthetic clays have the advantage that they possess a more homogeneous charge distribution and also contain significantly fewer impurities (e.g. carbonates, (hydr)oxides, silica and organic matter) than their natural counterparts7.

In the present work we studied the intercalation of CO2 in Li-fluorohectorite (LiFh), Na-fluorohectorite (NaFh) and Ni-fluorohectorite (NiFh). The only differences between these samples are the interlayer charge compensating cations used. Intercalation experiments under different temperature and pressure conditions were conducted. In order to investigate the potential of fluorohectorite clays for CO2 storage and capture, we also quantified the CO2 adsorption (wt%).


CO2 Intercalation

The layered nature of smectite clays gives rise to well-defined (00l) diffraction peaks and the angular position of these peaks is a direct measure of the interlayer repetition distance (the d-spacing). Figure 1 displays how the intensity of the (001) diffraction peak grows with time as the samples are exposed to CO2 at −20°C and 20 bar. All the samples were pre-dried before the measurements (see Methods section below). Intercalation of CO2 in the interlayer space manifests itself as the growth of an intercalation peak at a lower scattering angle (higher d-spacing) than the peak of dry, non-intercalated clay. As the intercalation progresses, the intensity of the CO2-intercalation peaks increase whereas the scattering from non-intercalated part of the sample decreases and eventually vanishes. LiFh and NaFh show similar intercalation behaviors in the X-ray diffractogram (XRD). The (001) peaks develop to d-spacings of 1.196 nm and 1.240 nm, for LiFh and NaFh respectively. This could correspond to a monolayer of intercalated CO2, in analogy to what occurs for H2O. For the NiFh sample we observe a similar intercalation state with d-spacing of 1.219 nm and in addition the development of another state with a larger d-spacing of 1.311 nm. To our knowledge, this is the first time such a complex CO2 intercalation state has been observed in a clay mineral, although other authors have found evidence of multiple intercalated layer type9,37. One may note that the secondary low-angle peak is at a d-spacing of ca. 1.3 nm, which is distinct from the ≈1.25 and ≈1.55 nm spacings of the one (1WL) and two water layer (2WL) smectite states as reported by Ferrage et al38,39. Other XRD studies have generally observed only blurred peaks in this region and have interpreted them as mixtures of peaks with the 1WL and 2WL spacings.

Figure 1
figure 1

The scattering intensity (arb. units) over a range of d-spacing values at different times (hours), for LiFh, NaFh and NiFh, respectively.

In all cases the CO2 adsorption conditions were −20°C and 20 bar. The red line () shows the time at saturation. Experiment performed at NTNU.

Figure 2 shows the comparison of (001) peak intensity vs. time for LiFh, NaFh and NiFh. In this Figure we plot the NiFh (001) peak intensities of d-spacing ≈ 1.21 nm, d-spacing ≈ 1.31 nm and the sum of them. We observe that the intercalation rate is significantly higher for NiFh and LiFh than for NaFh. This is similar to the case of water intercalation, where cations have been found to determine the stable states at varying relative humidity18,40,41,42,43 as well as the way that clay minerals exfoliate in aqueous dispersion6,44. Fripiat et al.27 suggested that the access of CO2 molecules to the interlayer space of montmorillonite clay is dependent on the size of the interlayer cation. Giesting et al.23 studied CO2-intercalation behavior of K-and Ca-montmorillonite, performing repeated measurements under the same conditions and also reported a significant dependence of dynamics on the cations. The dynamics of the CO2 intercalation can also be followed by observing the disappearance of the scattering intensity of the dehydrated peak, shown in Figure S1 in supporting information, which represents the d-spacing in the portion of the sample with no water or CO2 intercalated. In Figure S2 of the supporting information we show the d-spacing dynamics of the CO2 intercalated peak for each sample. Although the procedure used involves pre-drying of the samples it is possible that there is a minor amount of remaining H2O within the sample. This amount must in any case be very small since there is no detectable 1WL peak in the XRD patterns. Any residual H2O could affect the kinetics of the CO2 intercalation.

Figure 2
figure 2

Intensity of the intercalation peak, normalized to the maximum intensity (NI), as a function of time for different cations at 20 bar and −20°C.

The NiFh diffractogram has two contributions: Peak 1 is the peak with the maximum intensity with d-spacing ≈ 1.21 nm while peak 2 is the peak with d-spacing ≈ 1.31 nm.

A general equation for describing sorption kinetics is45:

where n is the amount of adsorbed molecules on a surface and is a rate constant that depends on temperature and pressure. The equation for the function depends on the type of adsorption mechanism. For a first order adsorption process, i.e. a process where the adsorbed molecules statistically occupy a single adsorption site46, . In the present case, is proportional to the normalized X-ray intensity (NI), where normalization is performed with respect to the intensity observed at the longest times, where the adsorption capacity of the material is reached. By integrating equation (1), we obtain

which is an exponential growth function towards saturation at 1. Applying the natural logarithm, we have a linear equation with the slope proportional to the rate constant :

The results of plotting the observed normalized intensity in this manner are shown in Figure 3. It is evident that equation (3) describes the data for NiFh and LiFh reasonably well, before the full adsorption capacity is reached, i.e. before all the sites have been occupied. However, for NaFh we observe deviation for −ln(1 − NI) < 1.6, which we relate to the Na+ ion providing a stronger layer adherence than the other two ions investigated here and thus the swelling is more difficult to achieve for the case of Na+, since the clay layers are closer together in the dehydrated state of NaFh, as shown in Figure 1 (d-spacing ≈ 0.97 nm). The difference in d-spacing between the dehydrated and the CO2 intercalated peaks is almost two times higher for NaFh compared to LiFh and NiFh (Figure 1). The horizontal line in Figure 3 represents the threshold between two regimes for the case of NaFh: Clay expansion accompanied by partial adsorption of CO2 (below the line) and adsorption of CO2 into the expanded interlayer of the clay mineral (above the line). This is in agreement with Figure 1, where it is noticeable that for the case of NaFh there is small shift with time of the monolayer CO2 Bragg peak even after expansion, indicating that more CO2 is adsorbed into the interlayer.

Figure 3
figure 3

The linearized intensity as function of time.

The horizontal line represents the threshold between two regimes for NaFh: Clay swelling accompanied by some intercalation of CO2 (below the horizontal line), which is described by the exponential function and intercalation of CO2 into the swollen interlayer of the clay mineral (above the horizontal line) described by equations (2) and (3). The initial swelling part of the dynamics is much faster for NiFh and LiFh as compared to NaFh, which is also confirmed in Figure S1 in the supporting information.

After the swelling process is finished, which means that the clay does not expand significantly anymore, the intercalation process is cation independent since the slopes of the curves shown in Figure 3 are the same for all samples, i.e. hours−1. This corresponds to a time constant τ of approx. 20 hours. A single exponential growth function (shown in Figure 2), with τ1 ≈ 51 hours, related to the clay swelling, describes well the data in the initial phase for NaFh, i.e. for −ln(1 − NI) < 1.6 in Figure 3.

The dependence of the intercalation intensity at various pressures is shown in Figure 4 for LiFh. The results show that the intercalation rate increases with pressure and this observation is in agreement with a previous study of NaFh22. We also measured intercalation at various temperatures (at a constant pressure of 20 bar) in the limited temperature range −5, −10 and −20°C and the data suggests (inset of Figure 4) faster adsorption at lower temperature. We have not measured the T-dependence of τ here, but for simplicity, we may assume this to follow an inverse Arrhenius behavior47.

Figure 4
figure 4

Left panel: The intensity of the (001) CO2 intercalation peak in LiFh as a function of time for different pressures (at constant temperature of −20°C). The intercalation rate increases with the pressure. Right panel: The quadratic dependence of the linear fit slopes Co/τ on CO2 pressure. Experiment performed at Maxlab.

The intensity of the CO2 intercalation peak follows a linear behavior for small times, in agreement with equation (2), which for short times t τ becomes NIt/τ. In Figure 4 we have not normalized the measured intensities to their saturation at long times, because we did not follow the experiments until saturation, thus the slopes (S) of the straight lines at short times in Figure 4 (left panel) can be expressed as where Co is a constant. As suggested above, one can assume that the adsorption dynamics after swelling is governed by an average inverse Arrhenius like trapping time:

where, is an average attempt frequency, E is an average energy barrier, kB is Boltzmann constant and T is the absolute temperature. Further we can assume that the attempt frequency increases with pressure, P, i.e. the higher the pressure, the more attempts are made by the CO2 molecules to cross the adsorption trapping energy barrier E. In the right panel of Figure 4, we have tested this assumption and fitted an empiric parabolic function to the pressure dependence of . Our temperature measurements are in such a limited range in absolute temperature that they do not allow us to test inverse Arrhenius trapping time assumption or any other temperature model for τ, but the desorption data for LiFh (shown below) indicates that the sorption trapping barrier is of the order of magnitude of . This gives where P is in units of bar and T in units of absolute temperature (K). From Figure 3 we estimated τ ≈ 20 hours for P and T equal to 20 bar and 253 K, respectively. This enables us to estimate τo ≈ 5 · 107 hours and thus the adsorption time would be τ ≈ 14 min for P and T equal to 200 bar and 250 K respectively, if we assume that trapping mechanism for liquid and gas CO2 are the same.

CO2 retention under ambient conditions

After exposing the three types of clay mineral samples to CO2 pressure for a sufficiently long time, the CO2 pressure was released and the cell was continuously flushed with N2, at atmospheric pressure, while increasing the temperature in steps of 5°C. It is known that the dry N2 does not intercalate into Na-fluorohectorite22. A plot of peak intensity versus time at different increasing temperatures is shown in Figure 5.

Figure 5
figure 5

Normalized Intensity (NI) of the (001) peak of LiFh (left) and NaFh (right) as a function of time for increasing temperatures.

Experiments performed at NTNU.

It has previously been concluded that the interlayer CO2 may cause an irreversible adsorption in clay, i.e. even if a clay sample is not exposed to the CO2 gas, CO2 molecules remain in the interlayer space48,49. This means that once intercalated with CO2 the clay mineral will retain these molecules. However, a temperature change can affect the CO2 retention50,51 and this makes the process of intercalation and release truly reversible. We found that at a certain threshold temperature, the intensity decreases until the contribution to the scattered intensity from the clay mineral with intercalated CO2 is negligible. Simultaneously, the peak corresponding to the dehydrated LiFh and NaFh reappears (data not shown). The threshold temperature, at which the CO2 is desorbed from the interlayer space of the clays, is highly dependent on the type of interlayer cation used. For LiFh, this temperature is about 35°C, whereas for NaFh it is about −15°C (Figure 5). This is consistent with the difference in size between the smaller Li+ cation versus the larger Na+ cation. Li+ has a more concentrated charge distribution than Na+ and can thus polarize the CO2 molecule more, forming a stronger bond to it. Loring et al also give a description of the CO2 intercalation mechanism20. In the case of NiFh the release, like the intercalation, has more complex features, as shown in Figure 6.

Figure 6
figure 6

(001) peak intensity of NiFh as a function of the d-spacing for different temperatures in N2 at atmospheric pressure.

As a reference the initial dehydrated peak of NiFh is included in the graph (circles, which is the same as the 0h curve in Figure 1). Experiments performed at NTNU.

Comparing the NiFh spectrum in Figure 1 with Figure 6 one can see that upon heating, the second CO2 peak merges with the first CO2 peak for NiFh. With increasing temperature, the intensity of the peak at the highest d-spacing value (about 1.31 nm) decreases and at 45°C it completely disappears while the lowest d-spacing value peak shifts to lower values and eventually contains all the (001) scattering. It appears that the final intercalation state is different from the original dehydrated state. This could suggest the formation of a complex CO2-Ni2+ structure within the interlayer space of the NiFh clay mineral, not present in the case of LiFh and NaFh. It is known that water intercalation experiments with NiFh can form a structure called Brucite (Ni[OH]2). Such a structure is formed in the cation exchange process from LiFh to NiFh60. It is possible that a Brucite-CO2 interaction could have an effect on the behavior. In addition this could occur due to the partially occupied d-orbitals of the Ni2+ ions, which allow multiple coordination geometries with CO2. These geometries can be possibly achieved by interactions of Ni d-orbitals with free oxygen orbitals present in polarized CO2 molecules.

Pressure composition Temperature Experiments

Figure 7 shows the excess CO2 adsorption isotherm of LiFh, obtained with the pcT-setup described in the Methods section. The excess adsorption is the amount of fluid taken up by the sample. The adsorption measurements were performed at room temperature and a pressure range from 1 bar up to 45 bar. The initial part of the isotherm (0 to ca. 9 bar) represents diffusion of CO2 into the mesoporous and interlayer network52 of the clay powder. Above approximately 9 bar it is likely that the swelling process of the clay has nearly finished and this will result in increased intercalation kinetics. With further increase in CO2 pressure, the excess of CO2 is seen to rise up to around 11 wt. % at a pressure of about 38 bar. At higher pressures, the apparent amount of adsorbed CO2 starts to decrease, likely due to the formation of an adsorbed layer with higher density and comparable to the volume of the clay mineral, associated with approaching the critical pressure for CO253,54.

Figure 7
figure 7

Excess CO2 adsorption isotherm of Li-fluorohectorite at room temperature and pressure range 0 to 45 bar.

The initial part of the isotherm (0 to ca. 9 bar) represents diffusion of CO2 into the mesoporous and interlayer network52 of the clay powder. Experiment performed at IFE.


The uptake of CO2 per weight of clay mineral can be inferred if it is assumed that the number of CO2 molecules coordinating exchangeable cations is similar to the number of H2O molecules within the interlayer space for the corresponding H2O-clay system (in the monohydrated state). This is approximately 2.4 molecules for each interlayer cation16,15. In case of LiFh and NaFh this would result in:

which is the same order of magnitude as measured by pcT (Figure 7), in this pressure range. This gives an amount of 3.2 mmol of CO2/g of LiFh. For other relevant CO2 capturing materials this number varies from 6.00 mmol of CO2/g, for e.g. metal organic frameworks (MOFs)55, to 5.00 mmol of CO2/g for Zeolites56. Both numbers are higher than the one we find for the clay mineral fluorohectorite. However, if we compare the adsorbed amount of CO2 per volume of the material, rather than per adsorbent mass, considering that the densities of zeolites (~2.2 g/cm3) and MOFs (~2.0 g/cm3) are lower than that of the clay minerals (~2.8 g/cm3). We find that a clay mineral, even with the cations considered here, is able to capture nearly the same mass of CO2 per volume (0.23 ton of CO2 per m3 of sample) as compared to the “best” zeolites (0.29 ton of CO2 per m3 of sample), or MOFs (0.32 ton of CO2 per m3 of sample). These numbers were calculated assuming 60% of packing density for all the materials. The commonly used benchmark Zeolite 13X captures 0.14 ton of CO2 per m3 of sample.


X-ray diffraction measurements were primarily performed on an in-house (NTNU, Trondheim) Bruker NanoSTAR X-ray scattering instrument, attached to a Xenox stationary electron impact source with a copper anode, producing Kα-radiation. The scattered intensity was recorded by a two-dimensional multiwire grid Xe gas detector (HI-STAR, Bruker). The beam diameter of the setup is 400 μm and the detectable range of momentum transfer q is (2.5 < q < 7.5) nm−1 (q is defined here as q = 4π sinθ/λ, where θ is the scattering angle and λ the wavelength of the X-rays). The relation between q and d-spacing is d-spacing = 2π/q nm which means that the d-spacing interval is between 2.5 and 0.84 nm. Complementary X-ray scattering data were collected at the I911-4 beamline of MAX IV laboratory using a 2D CCD detector (165 mm diameter, from Marresearch, GmbH) and 0.91 Å wavelength.

The samples were mounted in a custom-made sample holder22 which allows temperature control in the range of −30°C to 45°C and pressures in the range from ambient to 20 bar. To allow X-rays to pass through the sample, the cell has Kapton windows on both sides of the sample volume. Internal channels connect gas from valves on the surface of the cell to the sample volume and the gas pressure is controlled by standard reduction valves. The sample cell is depicted in Figure S3 in the supporting information.

In both X-ray setups, two-dimensional diffractograms were recorded and then azimuthally averaged to produce plots of intensity versus scattering vector, I vs q. Data reduction consisted in subtracting a background and normalizing the intensity profiles to the peak produced by the Kapton windows (see Hemmen et al.22 for details). The intensity, position and width of the intercalation peaks were found by fitting the peaks to Pseudo-Voigt profiles57,58.

The LiFh clay mineral used in the experiments was purchased from Corning Inc., with nominal chemical formula: Mx(Mg6-xLix)Si8O20F4 per unit cell59, where M is the interlayer cation (Li+, Na+ and Ni2+) and x the amount which balances the charge of clay mineral layers (x = 1.2 for monovalent ions and x = 0.6 for divalent ions). Each sample consisted of 7 ± 1 mg of clay powder packed in the available space in the sample chamber. Typically such packed clay powder samples have a mesoporosity of about 40%52, which in the present case enables access of the employed gases to the layered nanoporous structures that make up the individual powder grains (here the terms mesopore and nanopore follow the IUPAC definition).

At ambient conditions (23° and 40% of relative humidity), these clay mineral samples are in the monohydrated state9,60. Since the uptake of the CO2 molecules may be affected by the initial H2O concentration61, we investigated dehydrated native samples. For dehydration, the samples were heated in an oven at near 150°C for more than 10 hours and in a N2 flushed atmosphere. To remove residual humidity from the cell after loading the clay and to ensure that the sample remained dry, the cell was flushed with N2-gas. An X-ray scan was also recorded at ambient temperature and pressure while flushing with N2 to confirm that the sample remained dehydrated before starting the CO2 intercalation experiments. The sample was subsequently cooled to −20°C before the gas was changed to CO2. The gas outlet of the cell was closed and the pressure increased.

The CO2 used for experiments has a purity of 99.999% (Yara Praxair, grade 5). The N2 gas has a purity of 99.9999% (Yara Praxair, grade 6). To obtain a satisfactory signal-to-noise ratio, we varied the acquisition times from 30 to 60 minutes, depending on CO2 pressure, due to differences in X-ray absorption.

Pressure-composition-Temperature (pcT) isotherms were measured in a calibrated in-house built (IFE, Kjeller) volumetric Sieverts-type apparatus in order to obtain information on CO2 adsorption. Approximately 300 mg of Li-fluorohectorite was inserted in a sample holder and was degassed at 115°C under dynamic vacuum (<10−6 mbar) overnight, to remove residual humidity. Adsorption isotherms were acquired in the 0 – 45 bar range, with 3 bar step between each aliquot measurement at room temperature. The CO2 adsorption data were baseline corrected by the adsorption data collected from N2.