Direct Infrared Discrimination of End-On and Bridged CO₂ Adsorption via ¹⁶OC¹⁸O Linkage Isomerism

Abstract

Identification of the adsorption geometry of molecular CO₂ on solid surfaces is essential for the rational design of effective adsorbents and catalysts for CO₂ capture and utilization, yet remains a persistent experimental challenge. In this work, we show that mixed-oxygen CO₂ isotopologues containing ¹⁶O and ¹⁸O, optionally combined with ¹³C substitution, provide a direct infrared spectroscopic criterion for determining the adsorption geometry of molecular, near-linear CO₂ at low coverage. The approach exploits isotopic linkage isomerism (ILI) of the intrinsically asymmetric ¹⁶OC¹⁸O isotopologue. End-on (η¹-O) adsorption necessarily generates two linkage isomers (coordination through either ¹⁸O or ¹⁶O) producing two ν₃ bands of equal intensity. The magnitude of the splitting correlates with adsorption strength. In contrast, bridged (μ-O,O′) adsorption involving two equivalent bonds suppresses ILI and yields a single ν₃ band. The general applicability of this diagnostic criterion is demonstrated for CO₂ adsorbed on alkali- and alkaline-earth-exchanged zeolites, Brønsted- and Lewis-acidic zeolites, metal–organic frameworks containing open metal sites or hydroxyl groups, and nonporous metal oxides. Pure end-on adsorption is observed on LiX and CaX, on acid sites in H-ZSM-5, on μ₃–OH groups in UiO-66, and on TiO₂, whereas Na-LTA and MIL-53 exhibit predominantly bridged adsorption. Mixed adsorption regimes are identified in NaX, NaY, and Ni-MOF-74. Isotopic shift factors between different CO₂ isotopologues further aid band assignment in complex spectra. This experimentally simple approach provides a robust and broadly applicable spectroscopic tool for resolving CO₂ binding geometries across a wide range of materials.

1. Introduction

Global warming and climate change place carbon capture among the major challenges facing modern society. Consequently, the adsorption of carbon dioxide is of great relevance for many scientific and technological fields, including gas capture, separation, and catalytic conversion of CO₂. , In these applications, not only the adsorption strength but also the adsorption geometry and the local coordination environment of CO₂ play a crucial role, as they govern adsorption capacity, selectivity, surface mobility, and the propensity of CO₂ to undergo subsequent chemical transformations.

Despite its apparent simplicity, CO₂ exhibits a complex adsorption behavior on a variety of materials, including metal oxides, zeolites, metal–organic frameworks (MOFs), and carbon-based adsorbents. Adsorption can be reactive, involving the formation of carbonate-like surface species, or nonreactive, where the molecular identity of CO₂ is retained. , Nonreactive adsorption is of particular importance for CO₂ capture applications, as it allows facile and energy-efficient regeneration of the adsorbent. In this context, porous materials are especially attractive because their high specific surface area is essential for achieving high adsorption capacities.

Among porous adsorbents, zeolites and MOFs have been extensively investigated owing to their well-defined structures and tunable adsorption properties. Zeolites combine crystalline order, uniform microporosity, and the presence of extra-framework cations, which provide adsorption sites with systematically varying size, charge, and coordination environment. − MOFs offer complementary flexibility in pore topology and chemical functionality. , In both classes of materials, CO₂ adsorption is largely nonreactive and dominated by electrostatic interactions, ,, in contrast to chemisorption on basic metal oxides. , This makes such systems especially suitable for investigating intrinsic molecular CO₂ adsorption geometries without complications arising from chemical activation or extensive formation of carbonate-like species.

Several adsorption geometries have been proposed for the initial stages of CO₂ on solid surfaces (Scheme S1). Besides weakly bound molecular, nearly linear species, strongly chemisorbed bent configurations may form, particularly on metal surfaces and highly basic sites, where the molecule undergoes substantial activation. Such bent species are typically characterized by pronounced structural perturbation and distinct spectroscopic signatures (bands in the 1780–1550 and 1300–1100 cm^–1 regions) and are therefore not considered in the present study, which focuses on molecularly adsorbed, near-linear CO₂.

For linearly adsorbed CO₂ at low coverage, two principal adsorption geometries are commonly proposed (Scheme a,b). In the first, CO₂ binds in an end-on (η¹-O) configuration, where one oxygen atom interacts with a single adsorption center, resulting in inequivalent oxygen atoms within the molecule. This adsorption mode is often considered for nonporous materials − and is frequently assumed for porous adsorbents as well. ,−

1. Schematic Representation of the Main Adsorption Geometries of Near-Linear CO₂ on Cationic Sites: (a) End-On (η¹-O) Coordination (Colinear or Tilted) and (b) Bridged (μ-O,O or μ-O,O′) Coordination Involving Simultaneous Interaction of Both Oxygen Atoms with Two Neighboring Cations; At Higher Coverages, Geminal Species (c) or More Complex Adsorption Motifs (d) May Be Formed.

In the second configuration, CO₂ adopts a bridged (μ-O,O′) geometry, interacting simultaneously with two neighboring adsorption centers, which tends to preserve, or only weakly perturb, the equivalence of the two oxygen atoms. Such a geometry is more often proposed for porous materials ,− but also for flat surfaces. It requires an appropriate distance between adjacent adsorption sites. At higher coverages, the situation becomes more complex, as geminal species ,,− ,,, or more extended adsorption motifs , may be formed (Scheme c,d). In all these configurations, a weak secondary interaction between the carbon atom and a suitably positioned basic site may also occur. Because such contacts do not significantly perturb the near-linear geometry of CO₂ or its spectral response, they are not considered as distinct cases in this work.

Because in the bridged configuration CO₂ is bound to the surface through both oxygen atoms, this adsorption form is considered thermodynamically more stable than the end-on configuration. ,,,, On this basis, bridged adsorption has been proposed for a variety of zeolites. A key prerequisite for the formation of such species is the presence of adsorption centers at a suitable distance to accommodate the linear CO₂ molecule, a condition that depends strongly on framework topology as well as on the size and location of the cations. On the basis of DFT studies, the optimum cation–cation distances for dual sites interaction were estimated to be 6.5, 7.3, and 8.3 Å for Li, Na, and K cations, respectively.

Despite the potentially profound implications of adsorption geometry for uptake and capacity, direct experimental differentiation between end-on and bridged CO₂ adsorption remains limited. Strong evidence for bridged adsorption has been reported only in few cases, through neutron powder diffraction and synchrotron X-ray diffraction combined with Rietveld refinement and Fourier difference mapping. − In most instances, however, conclusions are based on density functional theory (DFT) calculations − , or on indirect experimental observations, such as correlations with cation density. ,,,

Beyond their structural differences, the adsorption geometries of CO₂ are expected to have important consequences for adsorption capacity. It is well established that this capacity can be significantly enhanced when a single cation is able to coordinate more than one CO₂ molecule, forming so-called geminal complexes. ,,− ,,, Such species have been invoked to rationalize high CO₂ loadings observed in several low-silica zeolites and are generally assumed to form sequentially, starting from monoligand complexes. Geminal coordination is commonly associated with an initial end-on adsorption geometry. In contrast, when CO₂ is adsorbed in a bridged configuration, the interaction involves two (or even three) adsorption centers, which may hinder further coordination. Consequently, although bridged species may be thermodynamically more stable, they are likely to limit the maximum number of CO₂ molecules that can be accommodated per adsorption site.

Owing to its linear geometry, significant quadrupole moment, and lack of a permanent dipole moment, molecularly adsorbed CO₂ is especially sensitive to the nature of the adsorption site and to the local electrostatic field. As a consequence, variations in adsorption geometry can lead to pronounced changes in vibrational spectra, adsorption energetics, and reactivity. However, establishing the adsorption geometry of molecular CO₂ in a direct and unambiguous manner remains a major experimental challenge.

A variety of experimental techniques have been applied to investigate CO₂ adsorption on solid surfaces. Among them, infrared spectroscopy is one of the most widely used methods for probing adsorption strength, site heterogeneity, and coverage effects. ,,,, As a linear triatomic molecule, CO₂ has four fundamental vibrational modes which give rise to three bands: ν₁ (symmetric stretching, 1388 cm^–1; involved in the Fermi dyad with 2ν₂ at 1388 and 1285 cm^–1), ν₂ (doubly degenerate bending, 667 cm^–1), and ν₃ (antisymmetric stretching, 2349 cm^–1). Of these, ν₂ and ν₃ are infrared-active; however, ν₃ is more convenient for analysis because ν₂ is often obscured by absorption from the sample itself.

The ν₃(CO₂) band in the gas phase is located at ca. 2349 cm^–1. For porous materials, however, a more practical reference frequency is 2339 cm^–1, corresponding to CO₂ weakly adsorbed on silicalite. Upon coordination of CO₂ to a single cationic site (Scheme a), the molecule becomes polarized, which typically results in a blue shift of the ν₃ stretching frequency. ,, Numerous systematic studies on monovalent cations in zeolites ,,,, as well as on divalent cations in MOFs have demonstrated that the magnitude of this shift correlates with the polarizing power of the cation.

However, additional interactions complicate this picture. The “free” oxygen atom of CO₂ may interact with framework oxygen atoms, leading to a secondary red shift of the ν₃ mode. In some cases, weak interaction through the carbon atom can also occur, producing an additional blue shift. Moreover, the presence of combination and hot bands further complicates the spectral region. As a result, the absolute position of the ν₃ band reflects the contribution of multiple interaction effects rather than a unique structural parameter. These effects demonstrate that even for a formally simple end-on coordination geometry, the ν₃ frequency alone does not provide a direct or unambiguous description of the adsorption complex.

For CO₂ adsorbed in a bridging configuration (Scheme b), it has been reported that the increase in ν₃(CO₂) is more pronounced than for end-on species. However, this view is idealized. DFT calculations indicate that the difference between end-on and bridged species may be negligible, and in some cases relatively low ν₃ frequencies have been assigned to bridging configurations. ,, Thus, both experimental observations and theoretical analyses indicate that the absolute ν₃ position cannot serve as a reliable and general criterion for distinguishing CO₂ adsorption geometries. Despite the extensive spectroscopic literature, the coordination mode of molecularly adsorbed CO₂ remains insufficiently clarified. In this context, a direct and experimentally accessible spectroscopic criterion capable of differentiating between end-on and bridged adsorption geometries would be highly desirable.

Recently, we reported that end-on Ca²⁺···¹⁶O–¹³C–¹⁸O and Ca²⁺···¹⁸O–¹³C–¹⁶O species formed after ¹⁶O¹³C¹⁸O adsorption on CaX zeolite can be spectroscopically distinguished through splitting of the ν₃ band due to isotopic linkage isomerism (ILI). The ν₃(¹⁶O¹³C¹⁸O) modes of the two isotopomers differed by 7 cm^–1, providing strong evidence for end-on coordination. However, it remains unclear whether this effect is equally pronounced for other adsorption sites and material classes. In the case of dual-cation adsorption sites, no ILI is expected, as both oxygen atoms of CO₂ should be equivalently bonded. Experimental evidence for such behavior, however, is still lacking.

In the present work, we systematically investigate the ILI of ¹⁶O¹³C¹⁸O upon adsorption on a series of representative materials, including zeolites, metal–organic frameworks, and metal oxides, where both end-on and bridged adsorption configurations have been proposed. By correlating the presence or absence of ILI with the adsorption geometry, we demonstrate that this phenomenon provides a direct, unambiguous, and experimentally accessible spectroscopic criterion for distinguishing between end-on and bridged molecular CO₂ species: splitting of the ν₃(¹³C¹⁶O¹⁸O) band into two equal-intensity components is a universal spectroscopic signature of end-on adsorption.

More generally, our results establish ILI as a broadly applicable infrared spectroscopic approach for probing adsorption geometry on solid surfaces, offering a simple and robust complement to indirect spectral assignments and theoretical modeling.

2. Experimental Section

2.1. Materials

A series of materials, representative of zeolites, metal–organic frameworks, and metal oxides, were investigated. All samples were commercial materials and their basic characteristics are summarized in Table S1 of the Supporting Information (SI).

2.2. Methods

The main experimental technique in this study was FTIR spectroscopy. The FTIR spectra were recorded with a Thermo Scientific Nicolet 6700 FTIR spectrometer equipped with an MCT-A detector and accumulating 64–128 scans at a spectral resolution typically of 2 cm^–1. Resolutions of 0.5, 1.0, and 4.0 cm^–1 were also used for specific purposes. The accuracy was 0.02 cm^–1. Self-supporting pellets (ca. 10 mg cm^–2) were prepared from the sample powder and treated directly in a purpose-made IR cell allowing measurement at ambient and at lower temperatures down to 100 K. The IR cell was connected through a dosage valve to a vacuum-adsorption apparatus with a residual pressure below 10^–4 Pa.

The treatment of the spectra (background and gas-phase corrections, obtaining second derivatives, deconvolution) was performed by the OMNIC spectroscopy software (version 9).

2.2.1. CO Adsorption

The activated pellets were cooled to liquid-nitrogen temperature under dynamic vacuum. Five mbar of CO (Merck, 99.5%) were then introduced into the cell, and the sample was allowed to equilibrate until saturation. The pressure was subsequently reduced stepwise while IR spectra were recorded to monitor desorption.

2.2.2. CO₂ Adsorption

Adsorption of CO₂ (Linde, 99.9993%) was performed by successive addition of small doses (≥50 μmol g^–1) up to an equilibrium pressure of a few mbar. The pressure was then gradually reduced by (i) dilution and (ii) short evacuation steps, with spectra recorded after each step. No hysteresis in the ν₃ region was observed when comparing adsorption and desorption sequences.

2.2.3. Adsorption of ¹³C^16/18O₂ Isotopic Mixtures

Adsorption of mixed isotopologues was performed analogously. The isotopic mixtures were prepared by mixing appropriate amounts of ¹³C¹⁶O₂ (Sigma-Aldrich, isotopic purity >99 at. % ¹³C and >97 at. % ¹⁶O) and ¹³C¹⁸O₂ (CortecNet, nominal isotopic purity 99% ¹³C and 95% ¹⁸O) in a glass bulb. A slight difference in isotopic composition was observed between adsorption and desorption sequences. The first small doses yielded spectra enriched in ¹⁶O-containing species, gradually approaching the nominal gas-phase composition. This behavior is attributed to ¹⁶O/¹⁸O exchange with the sample and/or the apparatus walls. During desorption, the isotopic composition remained stable.

Prior to adsorption experiments, the zeolite samples and TiO₂ were activated in situ by heating in O₂ at 400 °C followed by evacuation at the same temperature. The MOF samples were activated under dynamic vacuum at 200 °C.

With UiO-66 adsorption was performed at −30 °C and the temperature was maintained using a NORHOF LN2 Microdosing System #915.

The samples were also characterized using several complementary techniques (TEM, EDX, XRD, DSC, DFT, UV–Vis) and details are provided in the Supporting Information.

3. Results and Discussion

For each material, adsorption of unlabeled CO₂ was first examined to obtain reference spectra and to assess adsorption strength, site heterogeneity, and coverage effects. These experiments are described in detail in the SI. Briefly, small successive doses of CO₂ were adsorbed up to equilibrium pressures of a few mbar, followed by spectral acquisition during stepwise evacuation to lower coverages. The analysis is focused on the ν₃(CO₂) region, which provides the most sensitive and informative probe of molecular adsorption geometry; information on other CO₂-related vibrational bands is therefore reported in the SI. In several cases, carbonate-like species were formed, predominantly at the initial stages of adsorption. These reactive species are described in the SI and are not discussed further here, as the present work focuses on molecularly adsorbed CO₂.

Subsequently, adsorption experiments were performed following the same experimental protocol using nearly equimolar mixtures of ¹³C¹⁶O₂ and ¹³C¹⁸O₂. Because rapid isotopic exchange occurs, the ¹⁸O atoms become statistically distributed, such that the resulting isotopic mixture contains ¹³C¹⁶O₂:¹⁶O¹³C¹⁸O:¹³C¹⁸O₂ in an approximate 1:2:1 ratio. The use of ¹³C-labeled molecules eliminates interference from atmospheric CO₂ and enables unambiguous identification of adsorbed species.

Although for the asymmetric ¹⁶O¹³C¹⁸O isotopologue the loss of inversion symmetry removes the strict distinction between symmetric and antisymmetric stretching modes, the vibrational modes are conventionally labeled according to their correlation with those of symmetric CO₂, and the stretching mode evolving from ν₃ is denoted as such. The recorded IR spectra exhibit three well-separated envelopes in the ν₃ region corresponding to the three CO₂ isotopologues, with the intermediate ¹⁶O¹³C¹⁸O envelope being approximately twice as intense as the outer ones, in agreement with the statistical distribution. Control experiments performed with ¹²C-based isotopic mixtures showed analogous behavior of the ν₃ band.

3.1. Zeolites

The first materials investigated were cation-exchanged FAU- and LTA-type low-silica zeolites, which are among the most promising zeolitic adsorbents for CO₂ capture owing to their high adsorption capacities. − , In dehydrated FAU-type zeolites, accessible extra-framework cations may occupy S_II and S_III positions in the supercage. The S_II positions are preferable while the S_III sites are occupied only at high cation loadings, resulting in a bimodal distribution of monovalent cations in fully exchanged low-silica materials. In contrast, divalent cations such as Ca²⁺ are present at approximately half the concentration of monovalent cations and are therefore expected to occupy almost exclusively S_II positions. In NaY, the lower aluminum content also restricts the number of cation sites, and accessible Na⁺ cations are considered to occupy predominantly S_II positions. ,,,,, On the basis of cation–cation distances and site availability, theoretical studies have suggested that bridged CO₂ species are unlikely in LiX (due to the small ionic radius of Li⁺, which does not provide a suitable intercation distance) , and in CaX (due to the absence of cations in S_III positions usually invoked as necessary for bridged adsorption). In contrast, the formation of bridged CO₂ species has been proposed for NaX. ,

In Na-LTA, three main extra-framework Na⁺ site families are commonly distinguished: S1 sites near the 6-membered rings linking the α- and β-cages, S2 sites in/near the 8-membered-ring windows of the α-cage, and S3 sites associated with 4-membered rings in the β-cage, generally regarded as inaccessible to guest molecules. The high cation density in this zeolite allows for short Na–Na separations across adjacent ring apertures. This suitable geometric configuration has led to the opinion that nearby Na⁺ cations may act as dual sites capable of adsorbing CO₂ in a bridging form. ,,, A solid experimental evidence for dual sites was also provided. This makes Na-LTA a particularly suitable system for evaluating the existence of bridged CO₂ species. ,,,

3.1.1. LiX

In the ν₃(CO₂) region, adsorption of unlabeled CO₂ on LiX leads to the appearance and synchronous growth of two signals: an intense band at 2364 cm^–1 and a weaker feature at 2384 cm^–1 (Figure S1). The former is assigned to CO₂ coordinated to Li⁺ cationic sites, while the latter corresponds to a high-frequency satellite, most plausibly associated with a combination band, as often observed for adsorbed CO₂. At saturation coverage, both bands become very intense and exceed the detection limit. Under these conditions, a weak band at 2294.5 cm^–1, originating from the natural abundance of ¹³CO₂, becomes clearly visible. Its position shows only a very small coverage-dependent shift, indicating that no measurable fraction of geminal CO₂ species is formed under these conditions. This observation is consistent with previous reports, , which show that coordination of a second CO₂ molecule to a single cation typically results in a discrete red shift of the ν₃ band by 4–10 cm^–1.

The spectra recorded after adsorption of a ¹³C^16/18O₂ isotopic mixture on LiX are shown in Figure A. Three main bands appear in the 2300–2260 cm^–1 region, with maxima at 2297, ca. 2280, and 2261 cm^–1. These bands are unambiguously assigned to the ν₃ modes of adsorbed ¹³C¹⁶O₂, ¹⁶O¹³C¹⁸O, and ¹³C¹⁸O₂, respectively. A high-frequency satellite is clearly detected only for ¹³C¹⁶O₂, appearing at 2316 cm^–1; for the other two isotopologues it is masked due to lower intensity and spectral overlap.

1.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 40% ¹⁸O) on LiX (a–e). B. Second derivative (b″) and computer deconvolution of spectrum (b).

Notably, the band centered near 2280 cm^–1, attributed to adsorbed ¹⁶O¹³C¹⁸O, is significantly broader than the corresponding bands of the symmetric isotopologues and exhibits a distinctly composite shape. Second-derivative analysis clearly reveals two components with maxima at 2282.5 and 2277.5 cm^–1, which are also well resolved after spectral deconvolution (Figure B). Based on DFT calculations in our previous work, the higher-frequency component at 2282.5 cm^–1 is assigned to Li⁺···¹⁸O–¹³C–¹⁶O species, whereas the lower-frequency component at 2277.5 cm^–1 is attributed to Li⁺···¹⁶O–¹³C–¹⁸O species.

Because the two linkage isomers are energetically essentially equivalent, they are expected to form in equal concentrations, irrespective of the ¹⁶O/¹⁸O ratio in the gas phase. Consistently, the spectra presented in Figure show that the two components exhibit practically identical intensities. This observation provides clear and direct evidence for ILI of adsorbed CO₂ and is fully consistent with an end-on (η¹-O) adsorption geometry.

Figure also demonstrates that having all three ¹³CO₂ isotopologues adsorbed under exactly the same conditions allows for a direct comparison of their spectra, which greatly simplifies spectral analysis.

3.1.2. CaX

Similar results were observed with CaX zeolite. CO₂ adsorption on this sample has already been reported and the spectra are not shown here. At low coverage, a band at 2363 cm^–1 accompanied by a weak satellite at 2386 cm^–1 (assigned to a combination band) is formed. The principal band is attributed to CO₂ coordinated to Ca²⁺ sites. With increasing coverage, geminal CO₂ complexes are formed and are characterized by a split ν₃(CO₂) band at 2372 and 2351 cm^–1. Therefore, in the present analysis we focus on the low-coverage regime, while the effect of geminal species is considered separately below.

Dosing of different ¹³C^16/18O₂ mixtures leads to the appearance of four bands at 2297, 2283, 2276, and 2261 cm^–1 (Figure A). The two side bands are assigned to adsorbed ¹³C¹⁶O₂ (2297 cm^–1) and ¹³C¹⁸O₂ (2261 cm^–1), while the two middle bands at 2283 and 2276 cm^–1 arise from the ν₃ modes of the ¹⁶O¹³C¹⁸O linkage isomers. The splitting between the two middle bands (7 cm^–1) is larger than that observed for LiX (5 cm^–1), consistent with the stronger interaction of CO₂ with divalent Ca²⁺ cations. At first glance, this may appear to contradict the similar ν₃ frequencies of adsorbed CO₂ in the two systems. However, as well documented for CO adsorption, the dependence of stretching frequencies on the polarizing power of the cation follows different trends for mono- and divalent cations and correlates with adsorption strength only within each cation class. In this respect, CaX provides a particularly instructive example, as the two linkage-isomer bands are well separated and can be directly identified without numerical analysis.

2.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 45% ¹⁸O) on CaX (a–j). B. FTIR spectra of ¹³C¹⁸O₂ (k) and ¹³C¹⁶O₂ (l) adsorbed on CaX. Both gases contain contaminants of ¹³C¹⁶O¹⁸O. Second derivatives of the spectra (k″, l″) are shown on the top part.

In this case again, the two linkage-isomer bands exhibit essentially identical intensities, as expected for energetically equivalent isomers. Only when ¹³C¹⁶O₂ strongly predominates does the band at 2283 cm^–1 seem slightly more intense and composite, due to superposition with a weak band of dicarbonyl species appearing at a very similar frequency (Figure B, spectrum l).

In summary, the spectra of ¹⁶O¹³C¹⁸O adsorbed on LiX and CaX zeolites reveal ILI. This provides direct spectroscopic evidence that, under the investigated conditions, molecular CO₂ is adsorbed on Li⁺ and Ca²⁺ sites in an end-on configuration and no indication of bridging species was found. For CaX this is consistent with the occupation of S_II positions only, whereas for LiX this is due to the lack of a suitable cation–cation distance, which prevents the formation of μ-O,O′ adsorption species.

3.1.3. Na-LTA

The next experiments were performed with Na-LTA zeolite, since several literature reports indicate that CO₂ adsorption on this material may involve, at least in part, bridging configurations. − ,,, This makes Na-LTA a particularly suitable system for testing the capability of mixed-oxygen isotopologues to discriminate between end-on and bridged adsorption geometries.

Reference spectra of ¹²C¹⁶O₂ adsorbed on Na-LTA are shown in Figure S2. At low coverage, they are dominated by a single ν₃(CO₂) band at 2361 cm^–1 increasing with CO₂ dosage. A weaker hot-band satellite at ca. 2350 cm¹ develops synchronously. No evidence of geminal species was found.

FTIR spectra recorded after adsorption of a ¹³C^16/18O₂ isotopic mixture on the sample are shown in Figure . Dosing of the mixture leads to the simultaneous growth of three bands at 2294, 2277, and 2258 cm^–1, which are unambiguously assigned to the ν₃ modes of adsorbed ¹³C¹⁶O₂, ¹³C¹⁶O¹⁸O, and ¹³C¹⁸O₂, respectively. In agreement with ¹²C¹⁶O₂ adsorption results, second-derivative spectra reveal weak low-frequency satellites for each band at wavenumbers lower by ca. 10 cm^–1.

3.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 45% ¹⁸O) on Na-LTA (a–e). B. Second derivative (d″) and computer deconvolution of spectrum (d).

In contrast to the behavior observed for LiX and CaX, only a single main band is detected for adsorbed ¹³C¹⁶O¹⁸O in this case. The absence of band splitting indicates the lack of ILI, providing direct spectroscopic evidence that CO₂ is predominantly adsorbed in a bridging configuration.

Closer inspection of the spectra shows that the full width at half-maximum (fwhm) of the middle band at 2277 cm^–1 is slightly larger (by less than 1 cm^–1) than the bands at 2294 and 2258 cm^–1. This small broadening suggests that the two Na⁺···OCO interactions are not perfectly equivalent, although their difference is insufficient to give rise to resolvable ILI. This conclusion is consistent with previous experimental and theoretical studies reporting the formation of bridging CO₂ complexes in which the molecule interacts through its oxygen atoms with two , or three , Na⁺ cations simultaneously. Such multicenter coordination leads to the formation of nearly, though not perfectly, equivalent CO bonds in the adsorbed molecule, in agreement with the absence of resolvable ILI in the ν₃ region.

The ν₃(¹³C¹⁶O₂)/ν₃(¹⁶O¹³C¹⁸O) isotopic shift factor in this case is 1.0077 ± 0.0001. This empirical value was reproduced in subsequent experiments on systems lacking resolvable linkage-isomer splitting (see Table S2) and can therefore serve as an operational reference value for cases where the two C–O bonds are effectively equivalent. In contrast, systems exhibiting ILI yield two distinct apparent shift ratios, reflecting the two ν₃ components of ¹⁶O¹³C¹⁸O; for example, values of 1.0057 and 1.0088 were obtained for CaX. For convenience, Table S2 summarizes all observed isotopic shift factors.

The results described above provide a clear and operational guideline for interpreting the ν₃ region of the spectra of adsorbed ¹³C¹⁶O₂/¹³C¹⁸O₂ mixtures. At low coverage, where CO₂ interacts either with a single adsorption center (end-on geometry) or with two adjacent centers (bridged geometry), only these two fundamental configurations need to be considered. Under such conditions, splitting of the intermediate ν₃ band into two components of equal intensity is diagnostic of ILI and therefore of end-on coordination. In contrast, the presence of a single, nonsplit intermediate band indicates bridged adsorption involving two equivalent bonds. Deviations from this pattern indicate mixed adsorption geometries or more complex structures, which, however, emerge at high coverages.

3.1.4. NaY

To clarify the CO₂ adsorption behavior of Na–FAU zeolites, NaY and NaX were studied. We started with NaY since in this material the accessible Na⁺ cations are generally considered to occupy exclusively S_II positions. ,,, For the present sample, this conclusion was supported by low-temperature CO adsorption experiments, which revealed only a single family of Na⁺ sites forming monocarbonyls that convert into dicarbonyls at higher coverage.

Previous studies of CO₂ adsorption on the NaY sample reported the synchronous evolution of a strong band at 2354 cm^–1 and a weaker feature at 2344 cm^–1, the latter attributed to a hot band. A very weak combination band at 2364 cm^–1 was also registered. Formation of geminal adducts was also observed but only at low temperature or high equilibrium CO₂ pressure. These findings are confirmed in the present work by stepwise CO₂ dosing experiments (Figure S3). Consequently, at low coverage, formation of end-on adducts, similar to that found for LiX and CaX, may be expected.

Adsorption of a ¹³C^16/18O₂ isotopic mixture on NaY leads to the parallel growth of three main bands at 2288, ca. 2270, and 2252 cm^–1 (Figure A). The band at 2270 cm^–1 is noticeably broader and exhibits a composite character. In addition, three weaker features appear at approximately 8 cm^–1 lower wavenumbers, manifesting as shoulders in the absorption spectra and becoming resolved in the second-derivative spectra (Figure B). Second-derivative analysis further reveals that the intermediate band centered near 2270 cm^–1 includes two distinct components at 2272.4 and 2268.9 cm^–1, providing clear evidence for ILI.

4.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 45% ¹⁸O) on NaY (a–i). B. Second derivatives of the spectra presented in Panel A (a″–i″).

However, a notable feature of the spectra is the systematically higher relative intensity of the component around 2269 cm^–1, a tendency that becomes slightly more pronounced with increasing coverage. Because this effect is already observed at very low coverages, it cannot be associated with the formation of geminal species. As already noted, the deviation from the ideal 1:1 intensity regime indicates the coexistence of both end-on and bridged adsorption forms.

The results can be rationalized by assuming that the ν₃(¹⁶O¹³C¹⁸O) band of the bridging species overlaps with the band around 2269 cm^–1 due to the Na⁺-¹⁶O¹³C¹⁸O linkage isomer. Indeed, using the above derived ν₃(¹³C¹⁶O₂)/ν₃(¹⁶O¹³C¹⁸O) reference isotopic shift factor (1.0077), the expected position of the band for a bridging species is calculated to be approximately 2270.5 cm^–1.

The two contributions to the 2269 cm^–1 band are too close in frequency to be resolved in the spectra or in their second derivatives, even at a spectral resolution of 1 cm^–1. However, Fourier self-deconvolution (Figure S4) suggests three components of the band, which is consistent with the coexistence of two linkage isomers and bridging species.

A plausible explanation for the observed behavior is that CO₂ adsorption induces partial migration of Na⁺ cations from S_II to S_III positions, as predicted in theoretical studies. Such redistribution would enable in situ formation of dual adsorption sites, a process depending on CO₂ coverage. This scenario accounts for the nearly parallel growth of all components within the 2270 cm^–1 region. Although this interpretation remains tentative, it is consistent with the experimental observations and with the known flexibility of extra-framework cation distributions in FAU-type zeolites. The present results therefore imply that CO₂ adsorption in NaY does not exclusively yield end-on species, as often assumed on the basis of the initial S_II cation distribution. ,,,,

In summary, ILI is characteristic of a substantial fraction of CO₂ adsorbed on NaY, demonstrating end-on coordination, while the spectral asymmetry indicates a concurrent contribution from bridged adsorption geometries.

3.1.5. NaX

Another zeolite for which a high population of bridged adsorption sites has been suggested is NaX. , In this case, however, the spectra are more complex, reflecting the coexistence of at least three adsorption forms with comparable stability. Adsorption of ¹²C¹⁶O₂ (Figure S5) reveals a main ν₃(CO₂) band composed of two closely spaced components at 2356–2355 and 2352–2351 cm^–1, together with a weak additional feature at approximately 2363 cm^–1.

The spectra recorded after adsorption of a ¹³C^16/18O₂ isotopic mixture are largely consistent with the absence of ILI (Figure A). The overall contour of the intermediate ¹⁶O¹³C¹⁸O band closely resembles those of the symmetric isotopologues and consists of two main components at 2271 and 2267 cm^–1, accompanied by a weaker feature at 2277 cm^–1. No clearly resolved doublet of equal intensity is observed within the intermediate envelope (Figure B).

5.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 40% ¹⁸O) on NaX (a–h). B. Second derivative (g″) and computer deconvolution of spectrum (g).

A more careful inspection of the spectra, however, reveals subtle but systematic deviations from this idealized picture. In particular, in the ν₃(¹⁶O¹³C¹⁸O) envelope the lowest frequency component (2267 cm^–1) is the most intense, which is in contrast to the situation in the side envelopes (Figure B, pink-colored bands). The overall pattern resembles that observed for NaY, although here the single-geometry scenario is predominant. This suggests the presence of a small fraction of end-on CO₂ species giving rise to ILI superimposed on a predominantly bridged adsorption regime.

Owing to the relatively low intensity of the end-on contribution and to spectral overlap, the exact positions of the corresponding linkage-isomer components cannot be determined directly. Nevertheless, by analogy with the NaY case and taking into account the characteristic splitting observed there, these contributions are expected to lie close to the main bands at 2271 and 2267 cm^–1.

Overall, the results point to predominantly bridged CO₂ adsorption on NaX at low coverage, accompanied by a smaller contribution from end-on species. This conclusion is in good agreement with the work of Yang et al., who proposed the coexistence of several bridged CO₂ configurations in NaX involving cations in S_II–S_III or S_III–S_III positions, as well as linear end-on complexes associated with cations located in S_II positions, for which linkage isomerism is expected. In contrast to NaY, where S_III positions are unoccupied prior to CO₂ adsorption, NaX contains a significant fraction of Na⁺ cations in S_III sites already in the activated material. This preexisting S_III population allows the formation of bridging complexes involving two S_III cations, which leads to a greater structural diversity of adsorption motifs than in NaY.

3.1.6. H-ZSM-5

Zeolites frequently contain Brønsted acidic hydroxyl groups as well as extra-framework aluminum (EFAL) species, and it is therefore of particular interest to elucidate the adsorption geometry of CO₂ on such chemically heterogeneous materials. To probe the interaction of CO₂ with such sites, CO₂ adsorption was investigated on H-ZSM-5.

Preliminary characterization of the material by low-temperature CO adsorption indicated that the dominant adsorption sites are acidic hydroxyl groups. Only at very low CO coverages were two types of Al³⁺–CO carbonyl species detected (Figure S6A).

At first sight, the FTIR spectra of adsorbed CO₂ appear inconsistent with this simple picture, as they are relatively complex and consist of several overlapping bands with comparable intensities (Figure S6B). At low coverage, a band at 2362 cm^–1 develops while the OH stretching region remains unaffected, indicating that this band arises from CO₂ adsorbed on Al³⁺ sites associated with EFAL species. Two weaker bands at 2371 and 2352 cm^–1 are also observed. The higher-frequency feature is associated with a combination band satellite. While the band at 2352 cm^–1 may also include a contribution of a hot band, it develops independently at higher coverage and is therefore attributed to a second type of CO₂ complex with Al³⁺ cations. Upon further CO₂ loading, an additional band at 2347 cm^–1 appears and becomes dominant in the spectra. The rise of this band is accompanied by a parallel erosion of the OH stretching band at 3615 cm^–1 and the simultaneous development of a new band at 3450 cm^–1 (Δν­(OH) ≈ −165 cm^–1). On this basis, and in agreement with literature data, the 2347 cm^–1 band is assigned to CO₂ interacting with Brønsted acidic OH groups.

The spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture are shown in Figure A. At low coverage, four main bands of comparable intensity develop in parallel at 2296, 2282, 2274, and 2260 cm^–1. Notably, the intensity ratio of the bands at 2282 and 2260 cm^–1 remains close to unity. Comparison with the spectra obtained for ¹²C¹⁶O₂ adsorption indicates that these bands arise from CO₂ complexes formed with the more acidic Al³⁺ EFAL sites. The presence of two well-resolved components for the ¹⁶O¹³C¹⁸O isotopologue provides clear evidence for pronounced ILI, demonstrating unambiguously that CO₂ is adsorbed in an end-on configuration on these Lewis acid sites. The spectral split in this case (8 cm^–1) is larger than in the systems described above, consistent with the high electrophilicity of the trivalent Al³⁺ cations.

6.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 40% ¹⁸O) on H–ZSM-5 (a–k). B. FTIR spectra of the same isotopic mixture at high (l) and intermediate high coverage (m) and the difference between them (n) with the second derivative (n″).

Analysis of the spectra at higher coverage is complicated by the superposition of multiple contributions from Lewis and Brønsted acid sites. However, owing to the substantial difference in adsorption enthalpy between CO₂ complexes formed with Al³⁺ sites and those involving Brønsted OH groups, it was possible to isolate the spectral signatures of the OH–CO₂ interaction by subtracting spectra recorded at high and intermediate-high coverages (Figure B). The difference spectrum is dominated by a single type of interaction, attributed to CO₂ adsorbed on the bridging OH groups of the zeolite (Brønsted acid sites). Three main bands are observed at 2281, ca. 2264, and 2245 cm^–1, corresponding to the three isotopologues. Second-derivative analysis reveals that the intermediate band consists of two components at 2265 and 2262 cm^–1, evidencing ILI also for OH-bound CO₂. The smaller splitting between the two components compared to Al³⁺ sites is consistent with the lower adsorption enthalpy of CO₂ on Brønsted acid sites.

Thus, the results demonstrate that CO₂ is adsorbed predominantly in an end-on configuration on both Lewis (Al³⁺) and Brønsted (OH) acid sites in H-ZSM-5. No spectroscopic indication of bridged CO₂ species was found.

3.2. Metal–Organic Frameworks

Metal–organic frameworks are promising candidates for CO₂ capture. To evaluate whether the ILI diagnostic extends beyond zeolitic materials, we examined a framework with open metal sites (Ni-CPO-27) and two hydroxyl-containing MOFs (UiO-66 and MIL-53­(Al)). Similarly to zeolites, MOFs are characterized by well-defined structures containing crystallographically equivalent hydroxyl groups and/or open metal sites, which typically give rise to well-resolved IR bands of adsorbed species. This structural uniformity makes these materials particularly suitable for evaluating the sensitivity of the ILI criterion.

3.2.1. Ni-MOF-74

MOF-74 materials are particularly suitable for this purpose because they contain open metal sites capable of directly coordinating CO₂. , In the present work, a Ni-MOF-74 sample (also called Ni-CPO-27) was examined.

Adsorption of increasing doses of CO₂ on Ni-MOF-74 (Figure S7) leads to a monotonic increase in intensity of a single ν₃(CO₂) band centered at 2340 cm^–1. Second-derivative analysis, however, reveals that this band consists of a dominant component at 2340.5 cm^–1 accompanied by a weak shoulder at 2335.5 cm^–1, indicating some heterogeneity of the adsorption sites.

Upon adsorption of a ¹³C^16/18O₂ isotopic mixture, three main bands appear at 2274.1, ca. 2256, and 2238.5 cm^–1, with intensity ratios reflecting the isotopic composition of the gas phase (Figure ). The intermediate band associated with ¹⁶O¹³C¹⁸O is clearly composite. Second-derivatives of the spectra reveal at least two components at ca. 2260 and 2253 cm^–1. Unlike the cases of purely end-on adsorption discussed above, these two components do not exhibit a 1:1 intensity ratio, with the lower-frequency component being more intense. Deconvolution of the full spectral envelope into five bands further indicates the presence of an additional weak component at ca. 2256 cm^–1. The two outer components (ca. 2260 and 2253 cm^–1) appear with nearly equal intensities.

7.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 50% ¹⁸O) on Ni-MOF-74 (a–j). B. Computer deconvolution of spectrum of adsorbed ¹³C^16/18O₂ isotopic mixture (k″).

These observations can be interpreted as evidence for ILI of CO₂ adsorbed on the dominant fraction of Ni²⁺ sites in Ni-MOF-74. For these species, the ν­(¹³C¹⁶O₂)/ν­(¹⁶O¹³C¹⁸O) isotopic shift factors are 1.0062 and 1.0092, corresponding to a splitting of 6.8 cm^–1, which is consistent with strong CO₂–metal interactions. In contrast, the isotopic shift factor associated with the minor component (1.0080) is close to the value expected for bridged CO₂ species. Taken together, these results suggest that CO₂ is predominantly adsorbed in an end-on configuration on the Ni²⁺ cations, while a small fraction of bridged species is also formed.

Although further theoretical work is required to elucidate the origin of these bridged species, their formation is plausibly associated with structural defects or the presence of quasi-bridged CO₂ trapped in the MOF pores.

3.2.2. UiO-66

To further extend the applicability of the method to hydroxyl-containing metal–organic frameworks, CO₂ adsorption was investigated on UiO-66. UiO-66 is known to frequently contain missing-linker defects, which may generate coordinatively unsaturated Zr sites and additional μ₃–OH groups capable of interacting with adsorbates. However, previous characterization of the present sample by low-temperature CO adsorption revealed the absence of detectable open Zr⁴⁺ sites and indicated a uniform distribution of μ₃–OH groups. In this material, CO₂ interacts predominantly with μ₃–OH groups of the Zr₆ clusters, and the interaction strength is relatively weak. , It was therefore of particular interest to examine whether such weak adsorption is sufficient to generate detectable ILI.

Adsorption of ¹²CO₂ leads to the development of a sharp band at 2339 cm^–1 (fwhm ca. 2.5 cm^–1) with a shoulder at 2334 cm^–1 (Figure S8). Simultaneously, a portion of the ν­(OH) band at 3675 cm^–1 shifts to 3642 cm^–1, clearly indicating interaction between CO₂ and the μ₃–OH groups. The band at 2339 cm^–1 is assigned to OH–CO₂ adducts. , Its narrow width suggests a highly uniform adsorption environment consistent with well-defined and crystallographically equivalent hydroxyl groups within the framework. In contrast, the broader low-frequency component is unlikely to be associated with OH–CO₂ interaction and may originate from weakly bound species stabilized by dispersive interactions or by local structural heterogeneity associated with defects.

To enhance CO₂ adsorption, experiments were performed at −30 °C which lead to a moderate increase in the relative intensity of the shoulder at 2334 cm^–1. Upon adsorption of a ¹³C^16/18O₂ isotopic mixture, seven bands appear in the ν₃ region at 2275.1, 2270.4, 2258.8, 2256.5, 2252.8, 2239.0, and 2234.0 cm^–1 (Figure ).

8.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 45% ¹⁸O) on UiO-66 (a–l) at −30 °C. B. Computer deconvolution of spectrum (f) from Panel A.

The lower-frequency shoulder manifests as a broad band with an ill-defined maximum. Its large width results in weak and broadened second-derivative features; however, this characteristic broad contour facilitates its reliable identification in all three isotopic envelopes. The ν₃(¹³C¹⁶O₂)/ν₃(¹⁶O¹³C¹⁸O) isotopic shift factor associated with this component is 1.0078, very close to the reference value (1.0077) established for systems lacking resolvable ILI. This observation is consistent with the proposal of Morris et al. that the low-frequency shoulder corresponds to CO₂ molecules stabilized primarily through dispersive interactions. Such species may be regarded as weakly bound, quasi-bridging forms in which the two CO bonds remain effectively equivalent. As in the case of Na-LTA zeolite, the ν₃(¹⁶O¹³C¹⁸O) band associated with this feature is approximately 0.5 cm^–1 broader than the corresponding bands of the symmetric isotopologues, suggesting only a minor perturbation of the two CO bonds.

Second derivative analysis also reveals a clearly resolved doublet in the ν₃(¹⁶O¹³C¹⁸O) envelope, evidencing pronounced ILI with the OH–CO₂ adducts. This conclusion is supported by the ν₃(¹³C¹⁶O₂)/ν₃(¹⁶O¹³C¹⁸O) isotopic shift factors of 1.0072 and 1.0082 for the two isomers. These results demonstrate that even relatively weak OH–CO₂ interactions are sufficient to produce detectable (>2 cm^–1) linkage-isomer splitting. UiO-66 thus demonstrates that the ILI approach remains experimentally viable even for weaker OH–CO₂ interactions than those observed for Brønsted acid sites in H-ZSM-5.

3.2.3. MIL-53­(Al)

The third MOF investigated, MIL-53­(Al), also contains hydroxyl groups capable of interacting with CO₂. This system is particularly intriguing because each pore contains two OH groups. According to previous studies, CO₂ adsorption at low coverage may lead to formation of species bridging the two hydroxyl groups, thereby stabilizing the narrow-pore form of the material.

Adsorption of ¹²CO₂ on MIL-53­(Al) leads to the development of a sharp band at 2337 cm^–1 (fwhm ca. 2.5 cm^–1) with a weak shoulder at 2326 cm^–1, the latter attributed to a hot band (spectra not shown). As in the case of UiO-66, the very narrow band profiles indicate a remarkably small distribution of adsorption geometries and interaction strengths, consistent with the presence of structurally well-defined and nearly equivalent adsorption sites.

Upon adsorption of a ¹³C^16/18O₂ isotopic mixture (Figure ), three sharp bands with coverage-independent maxima are observed at 2271.7, 2254.4, and 2236.0 cm^–1. The ν₃(¹³C¹⁶O₂)/ν₃(¹⁶O¹³C¹⁸O) isotopic shift factor of 1.0077 coincides with the reference value derived in this work from Na-LTA, where exclusively bridged CO₂ species are formed. In addition, weak satellite bands appear at ca. 2262, 2244, and 2225 cm^–1. Importantly, no splitting of the intermediate ν₃(¹⁶O¹³C¹⁸O) band is detected, nor is any significant band broadening observed.

9.

FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 45% ¹⁸O) on MIL-53­(Al) (a–f) and the second derivatives of spectrum (f).

The sharpness and symmetry of the intermediate band provide compelling evidence for the absence of detectable ILI and indicate a high equivalence of the two CO bonds. These findings are fully consistent with the formation of bridged CO₂ species during the initial stages of adsorption, as previously proposed. Notably, the exceptional band sharpness in this system offers a particularly clear illustration of the spectroscopic signature associated with genuine bridging adsorption.

3.3. Metal Oxides

3.3.1. TiO₂

The final sample investigated represents a different class of materials, namely nonmicroporous metal oxides. TiO₂ was chosen because CO₂ adsorption on this oxide is largely nonreactive, making it a suitable benchmark system for testing the robustness of the isotopic linkage-isomerism approach outside porous frameworks.

The FTIR spectra of CO₂ adsorption on TiO₂ are relatively complex (Figure S9). At low coverage, a band at 2363 cm^–1 predominantly develops and shifts toward 2359 cm^–1 with increasing coverage. Upon further CO₂ uptake, an additional band at 2352 cm^–1 appears and progressively becomes dominant, while a weak high-frequency satellite at 2383 cm^–1 develops in parallel and is attributed to a combination band. Under the present experimental conditions, the formation of carbonate-like species was negligible, as evidenced by the very low intensity of bands in the 1700–1200 cm^–1 region.

The spectra recorded after successive adsorption of a ¹³C^16/18O₂ isotopic mixture are shown in Figure A. Owing to the strong superposition of multiple bands, detailed analysis was performed primarily on the basis of second-derivative spectra (Figure B). In the ν₃ regions of ¹³C¹⁶O₂ and ¹³C¹⁸O₂, two dominant bands are observed for each isotopologue (∼2292 and ∼2286 cm^–1 for ¹³C¹⁶O₂; ∼2256 and ∼2250 cm^–1 for ¹³C¹⁸O₂). In contrast, the ¹⁶O¹³C¹⁸O envelope consists of three well-resolved bands at ∼2279, ∼2271, and ∼2267.5 cm^–1.

10.

A. FTIR spectra recorded after successive adsorption of small doses of a ¹³C^16/18O₂ isotopic mixture (ca. 40% ¹⁸O) on TiO₂ (a–i). B. Second derivatives of selected spectra from Panel A.

The second-derivative analysis clearly shows that at high coverage four bands predominate in the spectra: at 2286.1, 2270.7, 2267.5, and 2250.3 cm^–1. The two intermediate bands fall within the ¹⁶O¹³C¹⁸O envelope and provide clear evidence of ILI. This observation indicates that on the weaker adsorption sites CO₂ is adsorbed in an end-on configuration.

At first glance, based on a superficial inspection of spectrum (i), one might be tempted to conclude that the remaining three bands (2292.3, 2278.9, and 2255.9 cm^–1) reflect the absence of ILI and therefore correspond to bridged CO₂ species. However, closer analysis demonstrates that this interpretation is incorrect. First, the ν­(¹³C¹⁶O₂)/ν­(¹⁶O¹³C¹⁸O) isotopic shift factor for the high-frequency band is 1.0059, which differs markedly from the characteristic value established for genuinely bridged CO₂ species. Second, the three bands have comparable intensity, but the middle band should be ca. two times more intense for bridging species. Third, at lower coverage (spectrum h), where the contribution of the lower-frequency adsorption form is small, two bands at 2278.9 and 2270.5 cm^–1 dominate the ¹⁶O¹³C¹⁸O region. This clearly shows that ILI is also associated with the high-frequency CO₂ adsorption form. With increasing coverage, the relative intensities of the lower-frequency bands at 2270.7 and 2267.5 cm^–1 approach equality, reflecting the growing contribution of a second end-on adsorption form.

Thus, the ν₃ spectral split for the stronger CO₂ adsorption form appears to exceed 8 cm^–1, slightly larger than the split observed for Al³⁺ sites in H-ZSM-5, consistent with the high electrophilicity of the Ti⁴⁺ cations. For the second, weaker family of sites, the split is smaller, ca. 3 cm^–1, similar to that observed for CO₂ adsorbed on zeolite Brønsted acid sites. In our case, however, the spectra of the titania hydroxyl groups remain unaffected by CO₂, indicating that this weaker adsorption form is also associated with Ti⁴⁺ sites. This again demonstrates the dependence of the split magnitude on the adsorption strength.

Taken together, these results demonstrate that on the TiO₂ sample CO₂ is adsorbed predominantly in an end-on configuration on two distinct families of Ti⁴⁺ sites, in agreement with expectations for this oxide.

3.4. CO₂ Linkage Isomerism with Complex Structures

In the present work, the analysis is focused primarily on the low-coverage regime, where end-on and bridging species dominates and isotopic splitting provides a reliable probe of adsorption geometry. At higher CO₂ coverages, adsorption becomes increasingly complex due to lateral interactions, site crowding, and the formation of multiligand species (Scheme c,d). Under such conditions, additional spectral features may appear and the interpretation of ILI is less straightforward. If we consider only symmetric geminal species, seven different configurations are expected and detailed analysis may require support from DFT calculations to predict their spectral signatures.

The onset of geminal species formation in CaX (Figure S10) illustrates how additional band splitting may emerge as coverage increases. A comprehensive treatment of these more complex regimes is beyond the scope of the present study and will require dedicated future investigations.

3.5. Other CO₂ Vibrational Modes

Although the present analysis focuses on the ν₃(CO₂) region, other CO₂-related vibrational modes were also examined when possible. These bands are generally weak at low coverage and are often superimposed on framework vibrations, carbonate-related absorptions, and OH stretching modes, which limits their diagnostic value. Analysis of our results indicates that the only additional mode with potential for probing ILI is the ν₁ mode, observed around 1335–1330 cm^–1 for ¹⁶O¹³C¹⁸O. Although IR-silent in the free molecule, this mode becomes weakly activated upon adsorption due to symmetry reduction, but it remains of low intensity. The splitting associated with ILI is approximately two times smaller than that observed for the ν₃ band (Figure S11). Nevertheless, ν₁ may provide complementary information in cases where the ν₃ region becomes congested, particularly at higher CO₂ coverage.

3.6. Factors Affecting Spectral Signatures of the Linkage Isomerism

Comparison of the different materials investigated reveals that the magnitude of isotopic splitting correlates qualitatively with the strength of the CO₂–surface interaction. Stronger adsorption leads to larger perturbations of the C–O bonds and to more pronounced splitting of the ν₃(¹⁶O¹³C¹⁸O) band.

This trend can be exploited not only to distinguish between end-on and bridged adsorption geometries, but also to differentiate within a given geometry class. For end-on species, variations in the splitting magnitude reflect differences in site acidity (polarizing power) and can therefore help distinguish nonequivalent end-on adsorption centers, as illustrated for TiO₂. For bridged species, deviations from the reference “equivalently bonded” behavior may provide insight into asymmetric dual sites composed of two nonequivalent adsorption centers, such as two different cations.

The detectability of ILI depends on the spectral resolution and signal-to-noise ratio. In most cases, a resolution of 2 cm^–1 was sufficient to resolve the relevant features, while using a resolution of 4 cm^–1 one risk obscuring small splittings. This is demonstrated in Figure S12A which also shows that further increasing the resolution to 1 and 0.5 cm^–1 mainly increased spectral noise without improving interpretability. However, for systems exhibiting very small splitting, such as CO₂ adsorbed on OH groups, higher resolution becomes essential (Figure S12B). These observations highlight the importance of adapting experimental conditions to the specific adsorption system under investigation.

4. Conclusions

The present study demonstrates that adsorption of mixed-oxygen CO₂ isotopologues containing ¹⁶O and ¹⁸O (here combined with ¹³C substitution) provides a powerful, conceptually simple, and experimentally accessible spectroscopic approach for probing the geometry of molecularly adsorbed CO₂, particularly in the low-coverage regime. The method exploits the intrinsic asymmetry of mixed-oxygen isotopologues and converts it into a direct vibrational fingerprint of the adsorption configuration.

When ¹⁶O¹³C¹⁸O (or ¹⁶O¹²C¹⁸O) is adsorbed in an end-on (η¹-O) configuration, two linkage isomers are necessarily formed, corresponding to coordination through either the ¹⁶O or the ¹⁸O atom. Because the adsorption enthalpies of these isomers are essentially identical, they are present in equal concentrations, giving rise to two ν₃ bands of comparable intensity. The splitting between these bands typically falls in the range of 2–9 cm^–1 and increases with increasing adsorption strength, reflecting the growing perturbation of the molecular vibration by the adsorption site.

In contrast, when ¹⁶O¹³C¹⁸O is adsorbed in a bridged (μ-O,O′) configuration involving two equivalent bonds, ILI is absent and only a single ν₃ band is observed for the mixed-oxygen isotopologue. Minor deviations from perfect bond equivalence manifest themselves only as weak band broadening, without resolvable splitting. These contrasting spectroscopic signatures provide a direct criterion for distinguishing between end-on and bridged adsorption geometries without reliance on crystallographic structure determination.

Beyond ILI splitting itself, isotopic shift factors between the different CO₂ isotopologues constitute an additional and highly valuable diagnostic tool, particularly for the assignment of bands in complex spectra. The isotopic shifts are largely insensitive to coverage and site population and therefore allow reliable tracking of corresponding vibrational modes even when multiple adsorption forms coexist or when bands are partially superimposed. This aspect is essential for the correct interpretation of systems exhibiting mixed adsorption regimes or strong site heterogeneity.

Since pure C¹⁶O¹⁸O or ¹³C¹⁶O¹⁸O cannot be prepared, adsorption experiments inherently involve well-defined statistical mixtures of C¹⁶O₂, C¹⁶O¹⁸O, and C¹⁸O₂. Importantly, this does not represent a limitation of the approach but rather a key advantage. The simultaneous presence of three isotopologues enables direct internal comparison under strictly identical adsorption conditions and allows spectral overlap to be minimized by adjusting the isotopic composition of the gas phase.

The general validity of the approach is demonstrated across a wide range of materials, including H-forms, alkali- and alkaline-earth-exchanged zeolites, metal–organic frameworks with open metal sites or containing OH groups, and nonporous metal oxides. Pure end-on adsorption is observed on Li- and Ca-exchanged zeolites, on Lewis and Brønsted acid sites in H-ZSM-5, OH groups in UiO-66, and on TiO₂, whereas Na-LTA and MIL-53­(Al) represent clear cases of predominantly bridged CO₂ adsorption at low coverage. Intermediate systems such as NaX, NaY, and Ni-MOF-74 exhibit mixed regimes, in which end-on and bridged species coexist and can be disentangled spectroscopically.

In particular, the results from this study resolve long-standing ambiguities by demonstrating the presence of a fraction of bridged CO₂ species in NaY and confirm the genuinely bridging nature of CO₂ adsorption in MIL-53­(Al), consistent with the well-known CO₂-induced pore contraction (“breathing”) of this framework.

Certain practical limitations should nevertheless be acknowledged. Because the ν₃(CO₂) region lies within a relatively narrow spectral window, systems exhibiting extensive site heterogeneity or multiple coexisting adsorption forms can generate strongly overlapping bands, which complicates analysis. In such cases, high signal-to-noise ratios and complete adsorption series are essential for reliable interpretation. Very weakly adsorbing systems may also yield bands of insufficient intensity for quantitative evaluation. These constraints are primarily practical rather than conceptual and do not diminish the validity of the ILI criterion.

Overall, the use of mixed-oxygen CO₂ isotopologues establishes a direct, model-independent spectroscopic criterion for identifying adsorption geometry that is experimentally straightforward, robust against site heterogeneity, and applicable under ambient conditions. The conceptual framework is not restricted to surface chemistry and may be extended to other systems in which symmetry breaking of isotopically substituted molecules encodes structural information in their vibrational response.

5. Perspectives

The present results suggest that several widely accepted views on CO₂ adsorption geometriesparticularly the frequent assumption of end-on coordination on cationic sitesshould be reexamined. Application of ILI analysis to previously studied systems may enable reassignment of spectra and refinement of proposed adsorption geometries, leading to a more accurate understanding of structure–adsorption relationships. In particular, the approach may prove valuable for investigating asymmetric dual sites composed of two different adsorption centers, which are increasingly relevant in photocatalysis and cooperative activation processes but were not addressed in the present work. Systematic investigations at higher CO₂ coverage may further establish spectroscopic criteria for distinguishing geminal species from other complex adsorption motifs.

Although bent CO₂ species are generally spectroscopically distinguishable without recourse to ILI, the present approach may nevertheless provide valuable additional insight into their bonding. In particular, ILI is expected for side-on CO₂ species, whereas it should be absent for C-bonded configurations, thereby offering a complementary means for detailed structural characterization of these activated forms. Importantly, adsorption of mixed-oxygen isotopologues may also assist in distinguishing bent CO₂ species from carbonate-like surface species, which are frequently difficult to differentiate spectroscopically. Whereas bent molecular CO₂ is expected to generate at most two ν₃ components for ¹⁶O–C–¹⁸O, carbonate-like species give rise to more complex isotopic splitting patterns. The approach thus may provide an additional diagnostic dimension for characterizing activated CO₂-derived surface species, with potential applicability beyond surface chemistry

Because the ILI approach is experimentally simple and requires no specialized instrumentation beyond standard FTIR spectroscopy, it has the potential to evolve into a routine characterization tool for CO₂ adsorbents, catalysts, sensors, and related functional materials. The present results demonstrate that the relative populations of single-site and dual-site adsorption motifs may differ substantially even in materials of identical chemical composition or framework topology, emphasizing the need for geometry-specific characterization rather than purely compositional descriptors.

The present work represents a first systematic exploration of ILI as a spectroscopic probe of adsorption geometry. As experimental data accumulate across a broader range of materials and adsorption sites, new empirical relationships and generalizations are expected to emerge, further refining the interpretative power of the method.

Integration with density functional theory (DFT) calculations offers a particularly promising direction. Theoretical modeling can provide quantitative predictions of isotopic shifts, splitting magnitudes, and spectral signatures for increasingly complex adsorption motifs, thereby extending the applicability of the approach to systems where experimental interpretation alone becomes challenging. At the same time, although high-level DFT calculations can establish general interaction trends and idealized geometries, their direct transferability to specific real systems is often limited by uncertainties in cation distribution, defect density, and local structural heterogeneity. The combination of ILI experiments with theory therefore offers a complementary strategy, in which spectroscopy provides direct geometric validation under realistic conditions.

Finally, while this study has focused on CO₂ as a prototypical linear AB₂ molecule, the underlying concept of ILI is not restricted to this class. Extension to other suitable probe moleculesincluding symmetric A₂ species (e.g., N₂) and more complex molecular adsorbates may open new avenues for probing adsorption geometry and site heterogeneity in a wide variety of surfaces and porous materials.

More generally, ILI introduces a new dimension to infrared spectroscopy by transforming isotopic substitution from a tool for mode assignment into a direct probe of adsorption geometry.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsphyschemau.6c00044.

• Details on the samples and characterization techniques; a scheme of adsorption geometries of adsorbed CO₂, FTIR spectra of ¹²C¹⁶O₂ and ¹³C^16/18O₂ isotopic mixtures adsorbed on the samples (PDF)

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.