Direct observation of carbon dioxide adsorption and binding at the air/aqueous interface

Abstract

Carbon dioxide removal (CDR) involves reducing carbon dioxide (CO₂) concentrations. Developing new technologies and enhancing existing ones for extracting and converting CO₂ are ongoing areas of research. In all these technologies, the movement of CO₂ molecules through an interface is a common process. At liquid surfaces, the nanometer-thick interfacial region is expected to play a fundamental role in enhancing or hindering the process. The interface can have significantly different conditions, such as pH, ion concentration, and ion speciation, compared with the bulk. Despite this, our knowledge of the molecular-scale details of CO₂ capture and conversion at liquid interfaces is limited. Here, we report direct observation of CO₂ surface adsorption and conversion to bicarbonate at the air/aqueous interface of potassium orthovanadate solutions using vibrational sum frequency generation spectroscopy. We show that orthovanadate ions enhance the hydrated CO₂ population at the interface, indicated by a strong peak at 2,336 cm⁻¹. DFT calculations suggest that CO₂ molecules are bent with respect to their original linear structure, demonstrating the initiation of CO₂ to $HCO3−$ conversion. With increasing orthovanadate concentration and/or time of exposure, the CO₂ peak disappears, and (bi)carbonate peaks appear. The characterization of the bulk solutions as well as the precipitated products suggests that the observed interfacial species are transient, different from the final products. This study provides a better understanding of CO₂ transport into aqueous media, not only for CDR technologies but also for environmental and atmospheric chemistry in general.

Introduction

Developing new carbon dioxide removal (CDR) technologies is an active area of research (1, 2). Carbon dioxide (CO₂) is selectively extracted from the atmosphere via different approaches, such as membranes, sorbents, and electrochemical and thermal swing (1, 3, 4). Some common sorbents include alkali/alkaline earth (hydr)oxides, mineral oxides, metal organic frameworks, and amine-based sorbents. Liquid sorbents are especially important as they are scalable and compatible with most industrial processes. The interfacial reactions during CO₂ capture and conversion are keys to the success of liquid sorbents. The nanometer-thick region at the liquid interface has significantly different properties, including different ion concentrations and ion-pairing behavior, pH gradients, and hydrophobic self-assembly (5, 6). Therefore, interfacial behavior cannot simply be deduced from bulk properties. Understanding the fundamental physical and chemical processes at these interfaces is crucial to develop and improve CDR technologies.

The interfacial behavior of carbonate, bicarbonate, and other related carbon capture products has been studied and debated extensively. Surface-sensitive sum frequency generation (SFG) spectroscopy studies of carbonate and bicarbonate solutions inferred the ion behavior via secondary effects on the interfacial water (7). Ambient pressure X-ray photoelectron spectroscopy and resonantly enhanced deep-UV second harmonic generation spectroscopy studies showed that carbonate anions are more abundant than the bicarbonate ions at the interface, in contrast to the widely used models, which suggest that doubly charged anions would be repelled from the interface due to the image charge interactions (8, 9).

Interfacial structure during CO₂ capture and conversion has been less explored. Combined X-ray reflectivity and neutron reflectivity studies investigated liquid/air interface during CO₂ capture in ionic liquid-propanol solutions and suggested that CO₂ is adsorbed with anions beneath the cation-rich top surface (10). Richmond and coworkers used SFG spectroscopy to study the reaction of highly concentrated monoethanolamine (MEA) solution and CO₂ at the air/liquid interface. They observed significant spectral differences at the air/MEA interface upon reaction with CO₂ (11, 12). Broadening of the methylene bending and stretching modes of MEA was observed, indicative of the presence of charged species at the interface, driving reorientation of the MEA and surface-active moieties. More importantly, the loss of the N–H bending mode and the presence of the amide II band (N–H deformation) upon reaction with CO₂ were signatures of the formation of amide-containing species (carbamic acid) upon the reaction with CO₂ (11–13). Premadasa et al. used SFG spectroscopy and MD simulations to study the CO₂ capture at the air/aqueous amino acid interface (leucine, valine, and phenylalanine) under highly alkaline conditions (pH 12), at which the anionic species predominate (14). The spectral variations of the CH stretches were used as molecular reporters on the structural and orientational changes promoted by CO₂ capture. In a recent study, the same team demonstrated that added bicarbonate, a possible direct air capture (DAC) product concentrated in the subsurface region, induced concentration and charge gradients at the interface (6). Neither CO₂ nor (bi)carbonate species were directly observed at the air/liquid interface in these studies.

Recently, we reported the direct observation of (bi)carbonate species at the air/liquid interface of niobium polyoxometalate aqueous solutions under enhanced CO₂ atmosphere. SFG studies showed a vibrational signature of bidentate bicarbonate (1,412 cm⁻¹) upon exposure to CO₂, whose intensity was counter-cation dependent (15). No interfacial signature of hydrated CO₂ was observed in that study.

Here, we provide a direct observation of the adsorption and binding of CO₂ and the subsequent formation of (bi)carbonate species at the air/liquid interface of orthovanadate solutions using SFG spectroscopy. Prior, inorganic oxoanions were shown to enhance the hydration and absorption of CO₂ into aqueous solutions (16–18). Among all the studied oxoanions, orthovanadate has the highest catalytic efficiency for CO₂ hydration, providing the ideal conditions to observe interfacial CO₂ (17, 18).

Solutions for SFG measurements were prepared by dissolving orthovanadate and orthomolybdate salts in deionized water. Solutions in a small Teflon dish were placed inside a custom-made, sealed chamber for controlled atmosphere (Fig. 1a). Multiple spectral regions were scanned for signatures of carbonate and CO₂, in addition to the OH region of water (2,800–3,800 cm⁻¹). In a typical SFG measurement, a fixed visible (532 nm) and tunable infrared (IR) (1,000–4,000 cm⁻¹) picosecond pulsed lasers is overlapped spatially and temporally at the air/liquid interface. The generation of a nonlinear SFG signal is forbidden in the centrosymmetric medium and therefore can only be generated at the interface where the symmetry is broken. The SFG signal is proportional to the square of the second-order nonlinear susceptibility (χ⁽²⁾) (see Materials and methods). Therefore, the SFG signal intensity is related to both number density and the orientational ordering of the oscillators. All SFG data were normalized against nonresonant SFG signal collected from a gold surface to correct for the variations in the laser power and IR absorption due to the atmospheric CO₂. Bulk solution conditions were investigated by Fourier transform IR spectroscopy (FTIR) measurements. The final products were characterized by FTIR, powder X-ray diffraction (PXRD), and thermogravimetric analysis–mass spectrometry (TGA–MS).

Results

Direct observation of interfacial CO₂

Figure 1b shows the SFG spectra collected at the air/100 mM aqueous K₃VO₄ interface in the absence and presence of excess CO₂ from 2,200 to 2,400 cm⁻¹. In the presence of ambient CO₂, the orthovanadate SFG spectrum shows two spectral dips at 2,340 and 2,360 cm⁻¹ (Fig. S1). The peak positions of these two features are consistent with previously reported CO₂(g) features at 2,336 and 2,362 cm⁻¹ as the R branch and P branch of CO₂(g) rotational–vibrational spectrum (19, 20). Upon introducing CO₂ into the sample cell, the CO₂(g) R branch and P branch disappear and a narrow, sharp peak at 2,336 cm⁻¹ grows (Fig. 1b, red spectrum). We assign this peak to the ʋ₃ asymmetric stretch of CO₂, which is at 2,350 cm⁻¹ in vacuum (19). The emergence of this peak is a direct indication of CO₂ hydration and adsorption at the air/water interface in CO₂-saturated atmosphere (19, 21). This is unexpected because CO₂ is a centrosymmetric molecule, and therefore, it cannot have a vibrational mode that is both Raman and IR active (a necessary condition for a vibrational mode to be SFG active). We investigated the validity and the implications of this observation in depth.

We first investigate the kinetics and the stability of CO₂ adsorption. When 100 mM orthovanadate solution is exposed to a 40-mmol/dm³ CO₂ atmosphere (pCO₂ ∼1 atm), the peak is observed as early as 10 min, and its intensity increases for 60 min and reaches a plateau (Fig. 1d, inset). After 60 min, we removed the lid of the sample cell completely, exposing it to the ambient atmosphere. The peak intensity dropped very slowly to 80% of its maximum within 60 min, indicative of a long-lived population, possibly kinetically trapped in this metastable state. We also simultaneously probed the spectral regions for any CO₂ capture products, such as carbonates and bicarbonates, but did not see any peaks within the first 2 h.

Figure 2a–c shows the increase of the surface-adsorbed CO₂ with the increasing concentration of CO₂ in the sample chamber. Each data point is measured with a fresh solution (100 mM $VO43−$⁠) with variable CO₂ atmospheric concentration for 30 min. Assuming that the overall orientational ordering of the interfacial CO₂ is similar for all concentrations, we can calculate the relative coverage of the adsorbed CO₂ from the amplitude of the SFG signal. Then, it is possible to fit the adsorption isotherm by the Langmuir model:

where q_e, q_max, K_L, and C_e represent the adsorbed CO₂ amount (mmol/dm²), maximum CO₂ capacity (mmol/dm²), Langmuir constant (dm³/mmol), and CO₂ equilibrium concentration (mmol/dm³), respectively.

Fitting the experimental data (SFG amplitude) to Eq. 1 shows a good agreement with the Langmuir adsorption isotherm (Fig. 2b). Plotting the SFG amplitude to the linearized Langmuir equation (Eq. 2) yielded a straight line with a correlation coefficient, R² of 0.9989 and a Langmuir constant 0.080 dm³/mmol (Fig. 2c).

Given that vanadium speciation is highly dependent on vanadium concentration, which would ultimately control its catalytic activity for CO₂ hydration, we also conducted SFG measurements with varying orthovanadate concentrations. Figure 2d shows the amplitude of the CO₂ asymmetric stretch normalized to the maximum SFG amplitude for SFG spectra collected at different vanadium concentrations (500 mM, 1 M, and 1.5 M) while increasing CO₂ concentration. CO₂ adsorption was found to increase with CO₂ concentration and saturate only at 100 mM orthovanadate. At higher vanadium concentrations, CO₂ hydration at the interface increased at low CO₂ levels, then dropped thereafter, especially with higher orthovanadate concentrations. These results indicate that once the interfacial (bi)carbonate formation starts the hydrated CO₂ species can more easily transform, similar to a nucleation event. Indeed, we were not able to observe any interfacial CO₂ at 2 M bulk orthovanadate concentration, suggesting that at this concentration even the ambient CO₂ is enough to start (bi)carbonate formation in a few minutes.

Direct observation of interfacial carbon capture products

As we will discuss in detail below, these results suggest that we are observing a metastable transient species of orthovanadate–CO₂ at the interface. To observe the final carbonate products, we exposed our samples to a saturated CO₂ atmosphere for ∼15 h and investigated the possible formation of interfacial (bi)carbonate species. SFG spectra collected after longer exposure of 1 and 2 M orthovanadate solutions to CO₂ show the growth of two new SFG features at ∼1,412 and 1,640 cm⁻¹ (Fig. 3a, green and red spectra). The CO₂ asymmetric stretch is not observed simultaneously with these two features. We attribute these two peaks to the symmetric and asymmetric stretching modes of bidentate bicarbonate formed by CO₂ binding, respectively (22–26). These spectra are very similar to the bicarbonate species we observed at the air/aqueous interface of [Nb₆O₁₉]⁸⁻ solutions. Also, in an SFG study of carboxylic acid films at the air/liquid interface, Sthoer et al. (24) observed intensity increase of the 1,416 cm⁻¹ COO⁻ symmetric stretch and a drop in the C=O stretch of arachidic acid with increasing Y³⁺ and La³⁺ ion concentrations due to the increased metal binding of the carboxylic acid monolayer. The presence of these two modes indicates CO₂ binding and conversion to (bi)carbonate upon reaction with orthovanadate (scheme in Fig. 3b). At 2 M, the bicarbonate SFG features are more intense and broader (Fig. 3a, red spectrum), suggesting increased carbonate formation and heterogeneity of the air/liquid interface at higher orthovanadate concentration. We only see the spectral signatures of CO₂ hydration and binding as bicarbonate in the ssp polarization combination, suggesting that the CO₂ molecules are parallel to the air/water interface (no spectral signatures were observed in ppp or sps polarization combinations).

To compare the carbon capture activity of the orthovanadate anion with other transition metal oxoanions, we also collected SFG spectra on the air/aqueous molybdate (Na₂MoO₄) interface of 1 and 2 M solutions after reaction with CO₂ for ∼15 h (K₂MoO₄ did not dissolve completely at equivalent concentration as K₃VO₄, hence not used for comparison). No spectral evidence of adsorption of CO₂ or binding as carbonates was observed below 2 M. At 2 M, we see evidence of CO₂ capture and binding as (bi)carbonate species (Fig. 3c), indicated by multiple (bi)carbonate features (Table S1).

Interfacial water behavior

The formed (bi)carbonate species at the air/liquid interface affects the orientation and structure of interfacial water (Fig. 3d) (7, 13, 27). A typical air/water SFG spectrum consists of two different water populations: hydrogen-bonded and nonhydrogen-bonded OH populations. The hydrogen-bonded OH oscillators consist of three broad peaks, 3,100, 3,200, and 3,400 cm⁻¹, from strongly to weakly hydrogen-bonded OH groups, respectively. The nonhydrogen-bonded (free) OH groups, dangling out from the surface, vibrate at ∼3,700 cm⁻¹ (see gray spectrum in Fig. 3d) (28–30). SFG spectra collected at the air/aqueous orthovanadate interface are spectrally similar to the neat air/water spectrum, exhibiting three distinct features (Fig. 3d, black and blue spectra): two slightly red-shifted and less intense broad bands at 3,150 and 3,350 cm⁻¹ due to strongly and weakly hydrogen-bonded water, and a dangling OH peak at ∼3,700 cm⁻¹. The 2 M spectrum shows a weaker 3,150 band and a stronger free-OH peak compared with 1 M, highlighting the effect of the salt concentration on the hydrogen-bonding network at the air/water interface. Upon reaction with CO₂, the intensities of all the interfacial OH features significantly decrease (Fig. 3d, green and red spectra). The 3,350 cm⁻¹ peak is barely observable at 2 M $VO43−$⁠. We attribute these significant spectral differences after reaction with CO₂ to the perturbation of the interfacial water populations by the newly formed (bi)carbonate/metavanadate species mentioned prior. The “merging” of hydrogen-bonded broad water band with the free-OH peak at 3,700 cm⁻¹ is intriguing. At free air/water interface, these water populations have opposite orientations, creating a dip between their representative peaks, due to the destructive interference. The absence of this feature suggests that the hydrogen-bonded water population is also up-oriented like the free OH. This could be enhanced by the presence of anionic surface-active species.

Bulk carbon capture products

Bulk FTIR measurements were conducted on the same samples investigated by SFG. FTIR measurements revealed no noticeable spectral differences after short-time exposure of 100 mM potassium orthovanadate to CO₂ (Fig. S5). However, longer exposure to CO₂ gives rise to new (bi)carbonate and polynuclear metavanadate species (see Figs. 4a and S8), consistent with our proposed mechanism of CO₂ binding (Fig. 3b) (31, 32). In contrast, no new products were formed from the reaction of sodium orthomolybdate with CO₂ (Fig. 4b), suggesting that the formed surface/interface active (bi)carbonate species block the interface, hindering the migration of the formed species into the bulk. Characterization of the 2-day reaction solid product of 2 M orthovanadate solution using FTIR, TGA–MS, and PXRD reveals the formation of potassium bicarbonate and potassium metavanadate species (Figs. 4c, d, S6, and Table S2).

Discussion

The direct observation of the spectral signature of interfacial CO₂ is highly unexpected and opens new avenues to investigate CO₂ reactivity at liquid interfaces. The fact that the ʋ₃ asymmetric stretch of CO₂ is SFG active (both IR and Raman active) at the orthovanadate solution interface provides key insights into the interfacial interactions. In the gas phase, this mode is only IR active with a frequency of 2,350 cm⁻¹ (19). Our DFT calculations indicate that as the CO₂ molecule bends, due to its interactions with water and/or orthovanadate, the Raman activity of this mode increases, and its frequency begins to redshift. Previous studies suggested bent CO₂–oxoanion structures as intermediate during the transformation from CO₂ to carbonate (16, 18). Our DFT calculations further reveal that protonated orthovanadate, i.e. [HVO₄]²⁻, can bend CO₂ molecules to an O–C–O angle of 175°, sufficient for Raman activity (Fig. 1e). DFT calculations tend to overestimate the absolute value of the vibrational frequencies; however, the 10–15 cm⁻¹ shift in the frequency is a more reliable parameter and similar to the experimental observation (see Fig. S7) (33). The experimental frequency of 2,336 cm⁻¹ is also red-shifted compared with the hydrated CO₂ vibrational signature (19), suggesting that CO₂ adsorbed at the air/liquid interface is involved in relatively strong interactions, explaining its persistence under ambient atmosphere (inset of Fig. 1d). Ab initio molecular dynamics simulations, better reflecting the complex interfacial environment, will be performed in the future.

A comparable direct observation of interfacial CO₂ was reported at a platinum electrode—ionic liquid interface during CO₂ reduction, where the CO₂ stretching frequency was at 2,348 cm⁻¹ (21). It was suggested that CO₂ formed an intermediate species with the cation of the ionic liquid (1-ethyl-3-methylimidazolium), making it SFG active. Although our system differs, being at the air/water interface, and not electrochemically driven, with (bi)carbonate as its final product, this prior study provides supportive evidence to the possible formation of stable intermediate species detectable by SFG.

Orthovanadate hydrates CO₂ more strongly than molybdate due to its differentiating properties. $HVO42−$ is the most dominant and reactive vanadium (V) species under all solution conditions of pH and vanadium concentration (Fig. S9a and b), whereas our speciation calculations show that $MoO42−$ is the most dominant molybdate species under our solution conditions ([Molybdate] = 100 mM–2.0 M, and pH = 9.0–12.0; Fig. S9c and d). $HVO42−$ has a basic –OH group that directly binds to CO₂ (Fig. 1c). The presence of this hydroxyl group in addition to the two negative charges localized on two oxygen atoms contribute to the high catalytic activity of orthovanadate in hydrating CO₂ (see mechanism in Fig. 3b) (34, 35). Furthermore, for an inorganic oxyanion to accelerate CO₂ absorption, it must have a high pK_a value (17). The active V (V) species in CO₂ hydration, $HVO42−$ and $HV2O73−$ have pK_a values of 13.4 and 9.74, respectively (36). Therefore, $HVO42−$ and $HV2O73$ have very high catalytic activity for CO₂ hydration.

What are the implications of enhanced interfacial CO₂ concentration, and why was it not observed in other systems? It is possible that in commonly studied amine systems, the intermediate CO₂ capture products are mainly carbamates (11, 14), or the (bi)carbonate products are not surface/interface active or simply not SFG active. The bending of the CO₂ molecule from its linear geometry requires an asymmetric interaction with the orthovanadate. Nevertheless, recent SFG studies with amine systems suggest that increasing the residence time of CO₂ at the interface is beneficial for DAC, as it enhances the likelihood of reactive transport of CO₂ and binding as (bi)carbonate species (6, 14, 37). Indeed, the comparison of orthomolybdate and orthovanadate supports this view; orthomolybdate solutions show neither interfacial CO₂ nor carbonate products in the bulk following prolonged exposure. Finally, we note that this is a brief report of the direct observation of hydrated CO₂ at the air/liquid interface, but further investigations are needed to gain a deeper molecular insight into interfacial CO₂ hydration and adsorption at environmentally relevant interfaces.

The observation of interfacial (bi)carbonate species after longer exposure times and at higher orthovanadate concentrations (1–2 M) supports the hypothesis that interfacial CO₂ enhancement correlates with reactive transport of CO₂ into the bulk. Although numerous studies in the last two decades have utilized SFG spectroscopy on bulk carbonate and bicarbonate solutions, none reported direct spectral signatures of bicarbonate species at aqueous interfaces (7, 13, 27). Only organic carbonates, such as propylene carbonate, were observed at the liquid/air interface with a spectral signature around 1,750 cm⁻¹ (38, 39). It is likely that the (bi)carbonate species observed in this study are visible to SFG due to their formation from highly ordered CO₂–orthovanadate complexes at the interface, resulting in orientational ordering. The dominant 1,412 and 1,640 cm⁻¹ peaks with orthovanadate correspond to bidentate chelating bicarbonate symmetric and asymmetric stretches, which structurally resembles the intermediate orthovanadate–CO₂ observed in earlier stages. The mechanism of bidentate bicarbonate formation upon prolonged reaction of orthovanadate with CO₂ involves CO₂ binding to the basic –OH group of the $HVO42−$ species in a carbonic anhydrase intermediate mechanism (34), eventually dissociating into bicarbonate and metavanadate species (Fig. 3b).

We observe an increase in the intensity of the 1,412 cm⁻¹ SFG signal with higher orthovanadate concentrations (Fig. 3a). Concurrently, bulk IR bicarbonate peaks also increase in intensity (Fig. 4a). This suggests that the surface species are relatively stable and are also transported into the bulk, creating ideal conditions for DAC. Conversely, with orthomolybdate solutions, (bi)carbonate species are observed only at 2 M concentration, with multiple peaks in the SFG spectra indicating more heterogeneous orientational ordering (Fig. 3c and Table S1). These (bi)carbonate products are limited to the interface with no significant presence in the bulk (Fig. 4b). A similar observation was made with [Nb₆O₁₉]⁸⁻ solutions in our recent study. In that case, the solutions with K, Rb, and Cs counter ions showed mainly 1,412 cm⁻¹ peak at the surface and a significant amount of carbon capture products in the bulk. However, solutions with tetramethylammonium (TMA) cation showed multiple SFG (bi)carbonate features but no significant carbon capture in the bulk. In that case, we attributed the lack of interface to bulk transport to the hydrophobic, and hence, surface-active nature of TMA. In the molybdate case, the reason may be the lower charge of molybdate (2− vs. 3− for orthovanadate) or a few number of free oxo-ligands (2 vs. 3 for orthovanadate), and hence increased hydrophobicity.

Before concluding, we rule out the possibility that the observed CO₂ signal is an artifact of the SFG setup. Given that the IR power is strongly absorbed by ambient CO₂, one might argue that the region between R and P branches appears as a peak. However, if this were the case, then the peak should be positioned at 2,350 cm⁻¹, precisely between the R and P branches (19). Furthermore, if the peak were an artifact of the laser setup, it would appear under all solution conditions. Nonetheless, as demonstrated, the peak responds specifically to the solution and reaction conditions and only appears at the surface of orthovanadate solutions. Moreover, the peak persists even after removing the CO₂-enhanced atmosphere (Fig. 1d), indicating that it is not affected by the bulk atmosphere above the surface and is indeed a genuine surface signal. All control experiments and additional details are provided in the Supplementary material.

Conclusion

In this study, we have reported the first direct observation of hydrated CO₂ and subsequent (bi)carbonate product formation at the air/water interface using SFG spectroscopy. The increased interfacial CO₂ concentration is advantageous for enhanced reactive capture of CO₂. However, if the products are not effectively transported into the bulk, they may obstruct the interface and hinder the process. As our focus was on the direct observation of interfacial products, we worked with smooth, stable air/liquid interfaces. It is reasonable to consider that under more realistic conditions, with agitation and stirring of the liquid, some of these effects might differ. Additionally, it is possible to control the interface with surfactants as proposed in earlier studies (40, 41). The direct observation of CO₂ and captured carbonates facilitates a rational comparison and contrast of these effects. We believe that this work will pave new paths in interfacial investigations. Lastly, this research contributes another example to the potential uses of oxoanions in CO₂ extraction (15, 42–44).

Materials and methods

SFG spectrometer

SFG experiments were conducted with an EKSPLA spectrometer, which has been described in the earlier publications (45, 46). Briefly, 1,064 nm laser pulse from a mode-locked Nd:YAG laser (EKSPLA, 29 ps, 50 Hz) is split, and frequency is doubled to 532 nm inside a harmonic unit. One of the 532 nm beams is directly used to probe the sample while the remaining beams are parametrically combined to generate a tunable IR beam. The beams are overlapped on the air/water interface in space and time in a reflection geometry; the visible and IR excitation angles, with respect to the sample normal, are θ_vis = 60° and θ_IR = 55°, respectively. The 532 nm polarization is adjusted with a λ/2 waveplate, and the SFG signal is selected using a Glan polarizer. SFG spectra were collected at the air/liquid interface of aqueous solutions. After the sample, the SFG signal is directed to a monochromator and collected with a photomultiplier tube (Hamamatsu, R7899). Each spectrum was collected with a 4-cm⁻¹ increment over the ranges 1,200–1,800, 2,200–2,400, and 3,000−3,800 cm⁻¹ average of 50 laser shots per point. SFG spectra were collected in ssp polarization combination (s-SFG, s-VIS, and p-IR) and normalized to gold under the same experimental conditions.

SFG measurements at the air/liquid interface of CO₂ capture experiments

SFG spectra were collected at the air/liquid interface of aqueous solutions in a 1.0-inch diameter polytetrafluoroethylene (PTFE) dish. The PTFE dish was placed in a sealed homemade chamber designed specifically for the SFG measurements with two KBr windows for the input and output laser beams. Short-time SFG measurements were conducted in the sample cell during the exposure of the liquid samples to CO₂ dry ice. For prolonged exposure experiments, aqueous solutions were placed in the same PTFE dish and exposed to a CO₂-rich environment in a vacuum desiccator (∼3–4 lbs dry ice) for ∼15 h; then SFG spectra were collected on the samples before and after reaction with CO₂. Control measurements with varying pH and solution conditions are provided in Figs. S1–S4.

Bulk and solid characterization of CO₂ capture products

Attenuated total reflectance (ATR)-FTIR data were collected with a Nicolet Nexus 870 FTIR spectrometer with an ATR accessory with the same aqueous solutions used for SFG measurements. PXRD for all structures was collected on the solid products of CO₂ capture experiments at 150 K, on a Rigaku Oxford Diffraction Synergy-S equipped with a PhototJet-S Cu source (λ = 1.54178 Å) or Mo source (λ = 0.71073) and a HyPix-6000HE photon counting detector. TGA was performed on TA Instruments SDT Q600, and MS was collected by Hiden Gas Analyzer HPR-20 QIC EGA. About 10 mg of the sample was transferred into alumina crucibles for measurements under Argon gas flow (the purge flow is 100 mL/min) up to 900 °C at the heating rate of 10 °C per minute. Approximately 90 min of Ar gas chamber flush (100 mL/min) was operated before the TGA–MS experiment to minimize the background noise.

Computational methods

Density functional theory calculations were carried out in the Gaussian 16 Rev. A.03 software package (47) with the CAM-B3LYP functional (48) and the default ultrafine integration grid. Carbon, oxygen, and hydrogen were modeled with the 6-31+G* basis set, while LANL2DZ (49) was used for vanadium. The isolated CO₂ calculations for the bond angle scan were performed in vacuo without empirical dispersion. The [HVO₄]²⁻–CO₂ system was embedded in an implicit water solvation environment through the polarizable continuum model (50). To more accurately account for the weak, noncovalent interaction between [HVO₄]²⁻ and CO₂, the GD3 empirical dispersion correction (51) was included.

Acknowledgments

The authors thank Thomas Persinger for the SFG sample cell design. Part of this work was conducted at ANL, operated by UChicago Argonne, LLC for the United States Department of Energy.

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

This work was supported by the US Department of Energy, Basic Energy Sciences, grant DE-SC0022278.

Author Contributions

M.R.: conceptualization, data curation, investigation, methodology, writing—original draft, review, and editing. Z.M.: investigation, writing—review and editing. J.S.H.: formal analysis, investigation, writing—review and editing. T.J.Z.: supervision, funding acquisition. M.N.: conceptualization, supervision, funding acquisition, writing—review and editing. A.U.: conceptualization, supervision, funding acquisition, methodology, writing—review and editing.

Data Availability

All Cartesian structures and input files have been deposited on this public GitHub repository: https://github.com/tjz21/CO2_SFG_DFT. All experimental data are included in the manuscript and Supplementary material.

References

Author notes