## Introduction

Owing to the significant climate change caused by the steady increase of CO2 in the atmosphere along with industrial activity1, the technology of CO2 capture and storage with respect to less-energy intensiveness, cost effectiveness, and environmental friendliness has been long-awaited. In power plants, electricity is generated by burning fossil fuels such as coal and natural gas, which could be a large point source yielding ~26% of global CO2 emission through the combustion process2. An approach to reduce CO2 emission is the efficient capture of CO2 and its storage in a stable manner before release to the atmosphere. There are three pathways for CO2 capture: pre-combustion capture, oxyfuel combustion, and post-combustion capture3. Pre-combustion capture is the decarbonation process before combustion, in which primary fuels are converted into a mixture of hydrogen and CO2 using gasification or reforming. In oxyfuel combustion, fuels are burned with an oxygen-enriched gas mixture instead of air, resulting in flue gases that mainly comprise CO2 and H2O. Post-combustion capture removes diluted CO2 from the flue gases produced by the combustion of fuel in the air, which could be applied to most conventional power plants owing to its adaptable operation.

Post-combustion technology for CO2 capture is based on the fundamental process of mainly chemical and physical absorption with absorbents of high CO2 solubility, and physical adsorption employing solid sorbents3. Chemical absorption is the CO2 capture process involving the reaction of CO2 with chemical solvents in aqueous solution, whereas CO2 is absorbed into solvents by applying pressure to promote physical absorption. In the standard environment of power plants, flue gases contain 12–14 vol.% of CO2 and are emitted under atmospheric conditions, therefore, they require treatment at elevated pressure to promote CO2 removal4. Physical adsorption utilizes the physisorption of CO2 molecules onto the surface of an adsorbent via the quadrupole interaction in addition to the size effect, thus requiring microporous materials with a high specific surface area5. At present, physical adsorption of CO2 is still in the early research stage together with the subject as to the thermal and chemical stability of the sorbed CO2, mechanical strength as well as production cost.

Another important factor of CO2 capture and storage is the long-term sequestration of CO2, where CO2 storage in geological reservoirs has been often considered6. In this approach, CO2 is injected into a deep geological formation to be physically confined below an impermeable or very low permeability caprock, such as a shale, allowing for a sequence of possible trapping mechanisms7. A fraction of injected CO2 is fixed as thermodynamically stable mineral carbonates, as e.g., CaCO3 or MgCO3 in the geological formation via the reaction with alkaline minerals there8. The formation of stable carbonates in the deep underground geology, known as in situ mineral carbonation9, could be suitable for long-term storage of CO210,11. On the contrary, ex situ mineral carbonation is the above-ground process involving the reaction of CO2 with alkaline earth metals extracted from basic rock12. Ex situ carbonation is generally conducted using acidic solutions at high temperature/pressure to accelerate the reaction between CO2 and materials, resulting in an energy-intensive process that generates a number of liquid wastes12,13.

Saponite, a silicate clay mineral abundantly and ubiquitously available in nature, is structured through stacks of 2D nanosheets with thicknesses of a few nm, which are the minimum structural unit. The 2D nanosheets have a variety of sizes and cannot be perfectly stacked, resulting in partial overlapping as schematically illustrated in Fig. 1a, which has been observed in images of field-emission type scanning electron microscopy14. This results in the formation of nanoscale open spaces (see Fig. 1b), which have been identified by positronium (Ps) annihilation spectroscopy together with molecular dynamics simulation as is detailed later15,16,17. Naturally, there exist local molecular sites in the interior of above open spaces such as nanosheet edges that are chemically active owing to the presence of unpaired electrons at ionically bound octahedron18. In the present work, CO2 adsorption in the open spaces originated from overlapped nanosheets in saponite clay minerals is explored by solid-state nuclear magnetic resonance (NMR) spectroscopy, open space analysis using positronium (Ps), and Fourier transform infrared (FT-IR) spectroscopy coupled with conventional chemical techniques. Besides the Na-type saponite, the Cs type is studied to find out an approach of environment-friendly recycling for low contaminated soil in Fukushima. The emergence mechanism of instantaneous mineral carbonation together with the physisorption of CO2 gas molecules, feasible in the absence of energy-consumption process as well as chemical solution enhancement, is highlighted as a future strategy of CO2 capture and storage.

## Results

### CO2 adsorption on 2D nanosheet surfaces

Chemical analysis by inductively coupled plasma (ICP) spectroscopy indicated that the Cs-type saponite contained ~0.04 mmol/g Cs (see ccesium in Table 1). According to a recently developed analytical method using the data of elution test, 133Cs magic-angle spinning (MAS) NMR, and radiocesium interception potential, the local molecular structures, as e.g., nanosheet surface, nanosheet edge, and oncoming hexagonal cavity, have been shown to act as Cs adsorption sites18,19. In this analysis, 69% of the loaded Cs are found to physisorb on the surface of the 2D nanosheet, which amounts to a concentration of ~0.03 mmol/g. The carbon, hydrogen, and nitrogen (CHN) elemental analysis revealed that the C contents, ccarbon, in the Na- and Cs-type saponite samples after CO2 loading are ~0.02 and ~0.18 wt.%, respectively (see Table 1). In light of the fact that the samples are isolated from the air during CO2 loading, the C contents detected in the CHN analysis are solely associated with CO2. Therefore, CO2 molecules are sorbed to the Na- and Cs-type samples at concentrations of ~0.02 and ~0.15 mmol/g, respectively (see cCO2 in Table 1). The concentration of CO2 higher than that of Cs on the nanosheet surface implies the adsorption of several CO2 molecules at the Cs cation sites, which will be discussed more in detail later.

Ps lifetime spectroscopy prior to CO2-loading reveals two kinds of open spaces for both the Na- and Cs-type saponite samples. The similar sizes of small and large open spaces with R1 ~ 3 Å and R2 ~ 9 Å are obtained for the Na- and Cs-type samples (see Table 2). Our former studies revealed that the above two open spaces commonly observed for both the Na- and Cs-type saponite samples are caused by overlapped nanosheets: the small and large open spaces are the consequence of one- and two-nanosheet insertion into the interlayer spaces15. The relative intensity of large open space I2 for the Cs-type sample is ~25%, much higher than that of the Na-type sample, though the intensities I1 of small open spaces at 5% are similar to each other. The high intensity I2 for the Cs-type sample indicating the large amount of 9 Å open space is resultant from insufficient self-assembly toward densification, which originates from interlayer Cs cations with low hydration degree16,17.

Figure 2 shows the 133Cs MAS NMR spectra obtained for the Cs-type saponite before (black curve) and after CO2 loading (red curve), and their difference spectrum (blue curve). The dominant signal at ~−134 ppm observed for the sample prior to CO2 loading correspond to the Cs+ interlayer cations physisorbed on the surfaces of the 2D nanosheets, whereas the additional broad signal at ~−15 ppm originates from Cs2O compounds at nanosheet edges18,19,20. The dominant peak arising from the Cs+ interlayer cations is slightly shifted and becomes broader upon CO2 loading, which can be seen in the difference spectrum as well. This demonstrates that the CO2 molecules adsorb at the Cs+ cation sites on the nanosheet surfaces at a concentration of ~0.03 mmol/g, as estimated above.

### Physisorption and chemisorption of CO2 molecules

It is of interest that both the physisorption and chemisorption for CO2 molecules occur at the alkali metal cations located on the surface of the 2D nanosheet. It is most probable that the CO2 physisorption on the nanosheet surface is caused by the quadrupole interaction between the alkali metal cations and CO2 molecules. Molecular orbital calculation, on the one hand, predicts the resultant carbonate species Cs2CO3 from CO2 chemisorption, in which the CO2 molecule stabilizes by bridging with two alkali cations on the nanosheet surface (Fig. 3b). The adsorption energy for the optimized structure calculated with the SCIGRESS program (Fujitsu Ltd. Japan) is −97.8 kcal/mol, which is similar to that of CO2 chemisorption onto the graphitic surface with two-bond conformation22. The concentrations of CO2 physisorption, cPhysCO2 and CO2 chemisorption, cChemiCO2 for the Na- and Cs-type samples obtained from the total amount of CO2 and the ratio of two peak intensities in 13C MAS NMR spectra are listed in Table 1. It is noted that ~13% of the loaded CO2 is instantaneously chemisorbed as the carbonate species in both samples without employing strong acid solution at ambient pressure and temperature. As the concentration of Cs+ cations physisorbed on the surface of the 2D nanosheet is ~0.03 mmol/g (see Table 1), the cesium carbonate formed with two Cs+ cations on the nanosheet surface has the concentration of ~0.02 mmol/g. This is consistent with the concentration of CO2 chemisorption, cChemiCO2 (see Table 1), signifying that Cs+ cations on the nanosheet surface dominantly take part in the CO2 chemisorption. On the other hand, the concentration of CO2 physisorption, cPhysCO2 is higher than that of Cs+ cations on the nanosheet surface. Presumably, a few CO2 molecules are physisorbed at Cs+ cations with the distance of several angstroms (see Fig. 3b).

## Discussion

According to a number of earlier works on zeolite materials with CO2 gas, the polarizing power of exchangeable alkali cations is one of the decisive factors for the capacity of CO2 gas adsorption23. The polarizing power of cations is inversely proportional to the ionic radius of alkali metal cations. The capacity of CO2 adsorption owing to the cation-quadrupole interaction thus increases as follows: Cs+<Rb+<K+<Na+<Li+23. This is in sharp contrast with the present observation that the concentration of both the physisorption and chemisorption increase for the Cs-type saponite. It is noted here that the fraction of open space with the size of ~9 Å for the Cs-type saponite probed by Ps is much higher than that of the Na-type one (see relative intensity I2 in Table 2). The open space formed by nanosheet overlap for the Cs-type saponite offers large enough surface area to accommodate CO2 molecules, thus being responsible for both the physisorption and chemisorption for CO2 molecules.

Here, we discuss why the chemisorption of CO2 gas molecules, i.e., ex situ mineral carbonation, instantaneously occurs in saponite samples without acid solution under the ambient condition. This unique adsorption nature is explained by the interior structure of the open nanospace characteristic for 2D materials. Saponite possess a 2:1 layered structure with 2D nanosheets consisting of tetrahedra and distorted octahedra. O and Si atoms are located at the vertices and central site of the tetrahedron, respectively. On the other hand, the O atoms or OH groups sit on the vertices of distorted octahedron, whereas a metallic Mg atom is located at the central site of octahedron (see the inset of Fig. 4). In the case of stevensite, the same family of saponite in aluminosilicate-type 2D materials, ca. 6.7% of the central Mg atoms in the octahedra are absent (see the inset of Fig. 4). An influence of Mg missing in the octahedron on the chemical bond is visible in the wavenumber region of 600–850 cm−1 in the FT-IR spectrum of stevensite. The FT-IR spectrum for the stevensite exhibits intense and small absorption peaks at the wavenumbers of around 650 and 775 cm−1, which can be ascribed to Si-O-Mg and Mg3OH bending vibrations, respectively24. The wavenumbers of the above bending vibration bands for the stevensite are lower than those of saponite, as the missing Mg atom causes lower frequencies in the both bending vibrations. It is noted here that the absorption peaks of Si-O-Mg and Mg3OH bending vibrations are red-shifted for CO2 loaded saponite as well implying atom missing in the octahedron (see Fig. 4). It is reasonably inferred that the above red-shift arise from the disappearance of O atoms from the vertices of the octahedra by the analogy of stevensite as illustrated in the inset of Fig. 4.

Based on the present findings of 133Cs and 13C MAS NMR and FT-IR spectroscopy, we can draw the scenario of instantaneous ex situ mineral carbonation as schematically illustrated in Fig. 5. In saponite, the O atoms are weakly bound to the octahedron by ionic bonding, in contrast to the strong semi-covalent bonding of O and Si atoms in the tetrahedron. CO2 gas molecules thus easily pick up the O atoms from the octahedra at the nanosheet edges in the interior of the above open spaces to be activated as the carbonate ions of CO32− type (see right-hand side of Fig. 5). The CO32− ions are then chemisorbed at the alkali metal cations on the surface of 2D nanosheets (see left-hand side of Fig. 5). The softness of the octahedra is also anticipated from the fact that the decomposition of octahedral sheets by mechanochemical milling proceeds prior to tetrahedral sheets25.

## Methods

### Materials

Synthetic Na-type saponite Na0.66[Mg5.34Li0.66]Si8O20[OH]4 produced by Kunimine Industries Co. Ltd., Japan, was employed in the present work. Cs loading was conducted by impregnating the Na-type saponite with a 1 M CsCl solution for ion exchanging with Cs. Aqueous solution was completely eliminated from the sample, which is referred to as Cs-type saponite. The chemical element in the Cs-type sample was examined by ICP spectroscopy (ICPS-8100, Shimadu). The sample was first treated at 423 K for 8 h under a vacuum condition of ~10–5 Torr in order to eliminate physisorbed H2O molecules. The dehydrated sample was successively replaced with 1-mbar CO2 atmosphere for CO2 loading without any contact to air and then kept there for 30 min, which was then analyzed for CHN using an elemental analyzer and 133Cs MAS NMR. In parallel to that, the dehydrated sample was exposed to a 13C-enriched CO2 (13CO2) atmosphere at 50-mbar for 30 min, which was subjected to 13C MAS NMR.

The sizes of open nanospaces and their fractions were investigated by Ps annihilation lifetime spectroscopy employing digital oscilloscope (WaveSurfer 10, Teledyne LeCroy). The positron source (22Na), sealed in a thin foil of Kapton, was mounted in a sample-source-sample sandwich configuration. The 1.27 MeV positron birth γ ray from a 22Na source and one of the 511 keV γ rays emitted as a result of positron annihilation in the samples are detected by BaF2 scintillators with 1'' diameter × 1'' thickness coupled with photomultiplier tubes (H3378-51, HAMAMATSU). Positron lifetime spectra were numerically analyzed. A fraction of energetic positrons injected into samples forms the bound state with an electron, Ps. Singlet para-Ps (p-Ps) with the spins of the positron and electron antiparallel and triplet ortho-Ps (o-Ps) with parallel spins are formed at a ratio of 1:3. Hence, three states of positrons: p-Ps, o-Ps, and free positrons exist in samples. The annihilation of p-Ps results in the emission of two γ-ray photons of 511 keV with a lifetime ~125 ps. Free positrons are trapped by negatively charged parts, such as polar elements, and annihilated into two photons with a lifetime ~450 ps. The positron in o-Ps undergoes two-photon annihilation with one of the bound electrons with a lifetime of a few ns after localization in nanospaces. The last process is known as o-Ps pick-off annihilation and provides information on the free volume size R through its lifetime τo–Ps based on the Tao-Eldrup model27,28.

$$\tau _{o - {\rm{Ps}}} = 0.5\left[ {1 - \frac{R}{{R_0}} + \frac{1}{{2\pi }}{\mathrm{sin}}\left( {\frac{{2\pi R}}{{R_0}}} \right)} \right]^{ - 1}$$
(1)

where R0 = R + ∆R, and ∆R = 0.166 nm is the thickness of homogeneous electron layer in which the positron in o-Ps annihilates. On the one hand, the relative intensity of τo–Ps is assumed to be correlated with the amount of open nanospaces.

### Solid-state nuclear magnetic resonance

The 133Cs and 13C MAS NMR experiments were performed using a Bruker Avance III 400WB spectrometer at the resonance frequencies of 52.5 MHz and 100.6 MHz, respectively. 133Cs MAS NMR spectra were recorded with single-pulse excitation of 2.0 µs and the repetition time of 3 s. A sample spinning rate of 22 kHz was used and 24,000 scans were accumulated. The chemical shifts were referenced to a 1.0 M solution of CsCl. 13C MAS NMR measurements were performed via exciting the 13C spins with single-pulses of 2.0 µs and with a repetition time of 20 s thus avoiding relaxation effects by T1. The sample spinning rate was 8 kHz and 320 scans were collected for each spectrum. The chemical shifts of the 13C nuclei in the adsorbed organic species were determined with respect to tetramethylsilane as the external reference with an accuracy ±1 ppm.

### FT-IR spectroscopy

Attenuated total reflection (ATR) FT-IR spectra were measured using a Nicolet iS5 FT-IR spectrometer (Thermo Fisher Scientific Inc.) using the ATR device with a diamond crystal plate. All the FT-IR spectra were measured at room temperature with the resolution of 4 cm−1. The measurements were repeated 100 times and the final spectra were obtained by averaging them. OMNIC 8.2 software was used to display absorbance spectra by converting ATR data, where the absorption band in the range of wavenumbers from 600 to 850 cm−1 is focused.