Key Drivers of Activity and Selectivity in Cu‐Based Catalysts for Methanol Synthesis From CO₂: Insights From Atomically Dispersed Promoters

ABSTRACT

The synergistic Cu–metal oxide (Cu–MO_x) interface is critical for selective CO₂ hydrogenation to methanol, yet its mechanistic function, the central, long‐debated feature of industrial Cu/ZnO/Al₂O₃, remains ambiguous. Using operando transient DRIFTS‐SSITKA, we elucidate the roles of M⁺ sites (Zn²⁺, Ga³⁺, and In³⁺) on model Cu─M/SiO₂ catalysts prepared by surface organometallic chemistry (SOMC). X‐ray absorption spectroscopy reveals that these promoters restructure from alloys after reduction to cationic species at the interface under reaction conditions. We find the promoter's electronic effect on formate bond strengths provides quantitative descriptors for both activity and selectivity. All catalysts follow a common formate spillover mechanism, with methoxy hydrogenation/desorption as the rate‐limiting step. The intrinsic CH₃OH formation rate follows a Sabatier‐type volcano with C─O bond strength, while selectivity correlates linearly with C─H bond strength (a proxy for the formate decomposition barrier). Cu─Ga/SiO₂ shows the fastest spillover, suppressing CO formation and yielding the highest selectivity; Cu─Zn/SiO₂ has optimal binding for the highest activity; In³⁺ binds formate too strongly, creating a kinetic trap on the strong‐binding side of the volcano. These findings bring mechanistic clarity to the debated Cu─Zn(O) synergy, showing selectivity is governed by a balance of electronic stabilization and spillover dynamics.

Mechanistic studies on model Cu─M/SiO₂ catalysts (M = Zn, Ga, and In) reveal that promoters enable a formate spillover pathway from Cu to M^{n +} (n = 2 or 3) sites. The promoter's electronic effect on formate bond strengths serves as a quantitative descriptor for activity and selectivity, balancing spillover dynamics with intermediate stabilization to suppress CO formation and drive methanol synthesis.

1. Introduction

One promising strategy to mitigate greenhouse gas emissions and combat global warming is the conversion of CO₂ into valuable products using green (e.g., produced through water electrolysis sourced by renewable energies) or blue H₂. Among such chemicals, methanol is positioned as the most versatile chemical feedstock and energy carrier toward a fossil‐fuel–free economy, known as the “methanol economy” [1].

Methanol synthesis from syngas (CO and H₂) in the presence of a small amount of CO₂ is a commercially established technology, catalyzed by Cu/ZnO/Al₂O₃ [2]. While the Cu/ZnO/Al₂O₃ catalyst is widely used in industrial methanol synthesis due to its balance of activity, stability, and selectivity, the precise nature of its active sites and the reaction mechanism remain subjects of ongoing debate, particularly in the context of CO₂ hydrogenation [3]. The Cu/ZnO/Al₂O₃ catalyst is dynamic and can change its structure [4]. During reductive catalyst activation, Cu─Zn alloy is formed via migration of ZnO_x onto Cu⁰ [5]. Under CO₂ hydrogenation conditions, CO₂ oxidizes Zn, leading to Cu⁰ particles that are in contact with ZnO_x [6]. Such structural changes are proposed to increase the activity by generating adsorption sites for oxygen‐containing intermediates (e.g., formate). Further migration of Zn forms metastable ZnO_x overlayer on the Cu surface [7]. This critical ZnO─Cu interface, which facilitates methanol formation via a formate pathway, is also reported to form in situ from the surface oxidation of Cu─Zn alloys under reaction conditions [8]. This interdependent structural relationship, which dictates catalytic performance, is established by two pivotal events: the reduction of CuO and the subsequent ripening and re‐oxidation of ZnO [9].

In elucidating active sites, reaction intermediates, and reaction mechanisms during CO₂ hydrogenation, two gaps, namely pressure and material gaps, are widely recognized, representing obstacles and complicating the establishment of structure–activity relationships [10]. The pressure gap toward industrial conditions is being addressed by clarifying pressure‐dependent active catalyst structures [11]. However, the “material gap” remains a significant impediment for mechanistic studies. Industrial Cu/ZnO/Al₂O₃ catalysts have an inherent complexity and active site heterogeneity, which requires the use of a model catalyst to deconvolute the crucial Cu─ZnO interaction and gain precise molecular insights [12]. This analytical challenge is compounded by the high loading of bulk Cu within the industrial Cu/ZnO/Al₂O₃, which strongly absorbs infrared (IR) radiation and hinders mechanistic studies using IR spectroscopy. In this regard, model Cu‐based catalysts can be prepared using surface organometallic chemistry (SOMC), which allows precise control of active sites, composition, interfacial sites, and metal‐support interactions [13]. This method allows for the creation of isolated metal sites on support, for example, SiO₂, while removing interference from the bulk properties of metal oxides. Importantly, these isolated M⁺ sites are distinct from bulk oxides such as ZnO, In₂O₃, or ZrO₂, which can hydrogenate CO₂ to methanol in their own right via cooperative surface chemistry (e.g., oxygen vacancies, extended acid–base ensembles); here, the isolated cations instead serve primarily to modify the Cu─M interface. For example, the Cu─Zn/SiO₂ catalysts generated via SOMC with Zn²⁺ single sites provide insights into Cu─Zn alloy and Zn redox behaviors without the contribution of bulk ZnO; extrapolation of the information can provide useful insights into Cu/ZnO/Al₂O₃ systems [14].

During CO₂ hydrogenation over Cu─ZnO‐based catalysts, there is a general consensus that formate is the key surface intermediate that is further hydrogenated to methanol [8]. However, formate intermediates were not observed ex situ with IR and/or solid‐state nuclear magnetic resonance (NMR) analyses of several related SOMC‐derived catalysts, including Cu─Zn/SiO_2, Cu─Ga/SiO₂, Pd─Ga/SiO₂ [14, 15, 16, 17], and the Cu─In/SiO₂ system investigated herein (Supporting Information, Figure S9), whereas their presence was confirmed over Cu─Ti/SiO₂ and Cu─Zr/SiO₂ [18, 19]. In contrast, operando diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) confirms the presence of formate intermediates under reaction conditions for Pd─Ga/SiO₂ [16]. These findings underline the importance of operando techniques. Furthermore, they highlight the general lack of information regarding the state and reactivity of formate species, in particular, how they dictate the product selectivity.

Since these SOMC‐made Cu─Zn/SiO₂ catalysts also show activity comparable to industrial systems, they provide an ideal platform for gaining mechanistic insights into the Cu─MO_x interface's role in CO₂‐to‐methanol conversion. This approach also enables the systematic study of other promoter effects, allowing a direct comparison of different single‐site promoters by ensuring uniform Cu particle size, metal dispersion, and total surface area [13].

In this study, we shed light on reactive surface species by operando DRIFTS to elucidate the nature of the active and selective sites for CO₂ hydrogenation to methanol over the Cu‐based catalysts with/without single‐site promoters (Cu/SiO₂ and Cu─M/SiO₂, M = Zn, Ga, and In) prepared via SOMC [14, 15, 16, 17]. This approach enables the study of the roles of each promoter (M) close to the Cu nanoparticle (e.g., the Cu─Zn/SiO₂ structure is illustrated in Figure 1).

FIGURE 1.

Structure of Cu─Zn/SiO₂ prepared via SOMC reported elsewhere [14].

Transient methods in combination with the use of isotopologues, such as steady‐state isotopic transient kinetic analysis (SSITKA), clarified that the structure and thus stability of surface formate species are affected by the nature of the interfacial sites. This explains how the reaction pathway is altered toward desired methanol or undesired carbon monoxide by an effective promoter, clarifying a widely debated yet important mechanistic aspect.

2. Results and Discussion

2.1. What Is the Impact of Single‐Site Promoters on Catalytic Activity and Selectivity?

The series of catalysts, consisting of Cu nanoparticles supported on SiO₂ decorated with isolated promoter sites (Cu─M/SiO₂, where M = Zn, Ga, Zr, or In) or on bare SiO₂ (Cu/SiO₂), see Figures S5 and S6 for synthesis and characterization details and Table S4 for the Cu loadings, Cu particle sizes (measured from transmission electron microscopy, TEM), metal dispersions (N₂O chemisorption), and BET surface areas. All catalysts were evaluated for catalytic performance under identical CO₂ hydrogenation conditions as shown in Figure 2. Noteworthily, the isolated promoter sites are not catalytically active on their own: under identical reaction conditions, M/SiO₂ supports (M = Zn, Ga, Zr, and In) exhibit CO₂ hydrogenation conversions below detection limits. This is consistent with previous reports for Zn/SiO₂ [14], Ga/SiO₂ [15], and Zr/SiO₂ together with related group IV/V single sites [20], and is expected, because the isolated M cations dispersed on the otherwise innocent SiO₂ support cannot readily dissociate both CO₂ and H₂, a step that requires the Cu nanoparticles (vide infra).

FIGURE 2.

Catalytic activity of Cu/SiO₂, Cu─Zr/SiO₂, Cu─Zn/SiO₂, Cu─Ga/SiO₂, and Cu─In/SiO₂: (a) overall CH₃OH selectivity and CO₂ conversion and (b) Intrinsic formation rates for CH₃OH and CO. Pretreatment was performed at 300°C under 20 NmL min⁻¹ of H₂ for 2 h. Reaction conditions are ca. 250 mg catalyst mixed with 5 g of SiC, 230°C, 25 bar, H₂/CO₂ = 3, total flow rate 6–100 NmL min⁻¹, TOS = 180 min.

The overall methanol selectivity shows a strong dependence on the single‐site promoters. As shown in Figure 2a, all promoted catalysts exhibit significantly higher methanol selectivity (>70%) compared to Cu/SiO₂ (48%) at low conversions (<1%). Notably, Cu─Ga/SiO₂ maintains exceptionally high selectivity (89%–92%) even as conversion approaches 3–4%, outperforming Cu─Zn/SiO₂ (71%–80%), Cu─In/SiO₂ (75%–82%), and Cu─Zr/SiO₂ (36%–69%), all of which show a more pronounced decline with conversion. Cu/SiO₂ exhibits the poorest selectivity (below 40%) at higher conversions, clearly highlighting the role of promoters. Notably, CH₄ and DME were below detection limits, and the only carbon‐containing by‐product is CO. Its selectivity ranges from ∼8%–11% over Cu─Ga/SiO₂ to >50% over Cu/SiO₂ at comparable conversions; the complete dataset of CO₂ conversions, CH₃OH and CO selectivity is compiled in Table S2, and the corresponding intrinsic formation rates in Table S3.

Interestingly, a different trend emerges when examining the intrinsic formation rates (Figure 2b). Because quantifying the dynamically changing, dual‐site active atoms is inaccurate for these alloys (vide infra), these rates are normalized by copper mass (µmol s⁻¹ g_Cu ⁻¹) to ensure a fair comparison across the catalysts. Cu─Zn/SiO₂ displays the highest intrinsic activity toward methanol surpassing Cu─Ga/SiO₂ and Cu─Zr/SiO₂. Both are significantly more active than the unpromoted Cu/SiO₂. In contrast, Cu─In/SiO₂ exhibits lower intrinsic methanol formation rates than even the unpromoted catalyst under these conditions. The selectivity trends are further clarified by the intrinsic CO formation rates (Figure 2b). Cu/SiO₂, Cu─Zn/SiO₂, and Cu─Zr/SiO₂ exhibit similar, relatively high CO rates. In stark contrast, both Cu─Ga/SiO₂ and Cu─In/SiO₂ demonstrate significant suppression of the CO formation pathway. These catalytic results highlight distinct roles for each promoter compared to Cu/SiO₂: Zn²⁺ and Zr⁴⁺ primarily boost the intrinsic rate of methanol formation without suppressing CO; Ga³⁺ strongly suppresses CO while maintaining high methanol activity, leading to superior overall selectivity; In³⁺ also suppresses CO but at the cost of significantly reduced methanol activity. Understanding the origins of these pronounced differences in activity and selectivity requires a detailed investigation of the surface intermediates and reaction pathways, which will be explored in the following sections.

2.2. What Is the Structural Nature of the Promoter under Reaction Conditions?

The promotional role of isolated, Lewis acid sites, such as Zr⁴⁺ sites at the Cu interface in Cu─Zr/SiO₂ prepared via SOMC remains in their oxidation state after reduction and under reaction conditions, giving this well‐established baseline for a nonreducible promoter [19]. The present study focuses primarily on comparing the effects of Zn²⁺, Ga³⁺, and In³⁺ promoters, which exhibit more dynamic structural and redox behavior.

Catalysts prepared with Zn²⁺ or Ga³⁺ initially form CuZn or CuGa alloy nanoparticles upon reduction, alongside residual Zn²⁺ or Ga³⁺ sites. Crucially, in situ X‐ray absorption spectroscopy (XAS) showed that these alloys undergo de‐alloying under CO₂ hydrogenation conditions [14, 15, 16, 17]. For Cu─Ga/SiO₂, this resulted in complete re‐oxidation to form Cu⁰ nanoparticles interfaced with Ga³⁺ sites [15], while for Cu─Zn/SiO₂, the de‐alloying was partial, leading to a dynamic mixture of Cu⁰, isolated Zn²⁺ sites, and some remaining reduced Zn⁰ species under reaction conditions [14]. XAS analysis of the Cu─In/SiO₂ catalyst reveals analogous behavior. In situ Cu K‐edge near edge spectra (XANES; Figure 3a, see Figure S1 for experimental procedure) indicated that Cu was present primarily as metallic Cu⁰ after pretreatment with H₂ and remained so under CO₂ hydrogenation conditions, similar to observations for related Cu─M/SiO₂ systems. In addition, ex situ In K‐edge XANES spectra (Figure 3b) of the reduced catalyst showed a significant shift in the absorption edge energy relative to the In₂O₃ reference, consistent with the formation of reduced In species. Fine structure (EXAFS) analysis confirms the formation of a Cu─In_x alloy: a simultaneous fit of the Cu and In K‐edges of the reduced catalyst reveals a Cu─In/In─Cu scattering path at 2.69 Å, with a contraction of the Cu─Cu distance to 2.475 Å relative to fcc Cu (2.55 Å), consistent with substitutional incorporation of In into the Cu nanoparticles (Figure S7 and Table S5). The In K‐edge fit further shows that In is present in two co‐existing environments: an oxygen‐coordinated In³⁺ environment comparable to the In/SiO₂ precursor and a metallic In environment from which roughly one‐third of the In is estimated to be alloyed in the as‐prepared catalyst, the remainder persisting as isolated In³⁺ sites (see Supporting Information, Section 2.2.3). Upon exposure to reaction conditions, the edge energy of the ex situ reacted sample shifts back to values indicative of In³⁺, suggesting that the Cu─In/SiO₂ system also undergoes re‐oxidation and de‐alloying (Figure S8).

FIGURE 3.

Operando XAS spectra of the Cu─In/SiO₂ catalyst. (a) In situ Cu K‐edge XANES spectra recorded after air exposure (purple), after reduction with H₂ (green), and under reaction with CO₂ + H₂ (yellow). The spectrum of a Cu foil is included as a reference (dashed grey). (b) Ex situ In K‐edge first derivative XANES spectra recorded after air exposure (purple), after reduction with H₂ (green), and after reaction with CO₂ + H₂ (yellow). The spectrum of In₂O₃ is included as a reference (dashed grey). Pretreatment was performed at 300°C under 10 NmL min⁻¹ of 20% H₂ for 1 h. Reaction conditions are ca. 10 mg catalyst, 230°C, 5 bar, H₂/CO₂ = 3, total flow rate 10 NmL min⁻¹, TOS = 60 min.

Therefore, Zn²⁺, Ga³⁺, and In³⁺ promoters, prepared via SOMC, all share characteristic dynamic restructuring involving dopant redox, differentiating them from isolated Zr⁴⁺ promoters. The alloyed fraction of the promoter in the as‐prepared (reduced) state is comparable across the three systems (≈25%–50% for In, Ga, and Zn; Table S6), while the extent of re‐oxidation under reaction conditions differs: it is total for Ga³⁺ and In³⁺, but only partial for Zn²⁺ [21]. This in situ generation of interfacial M⁺ sites via de‐alloying represents a fundamentally different scenario compared to the static Zr⁴⁺ sites. This dynamic nature is thought to be the key to their distinct activity and selectivity, as it potentially creates a high density of active interfacial sites in proximity to the Cu nanoparticles. However, the atmosphere‐dependent generation of interfacial M⁺ sites means their density under reaction conditions is not directly accessible by ex situ characterization (e.g., N₂O chemisorption). Consequently, catalytic rates are compared on a per‐Cu‐mass basis rather than as site‐based turnover frequencies (TOFs) (vide supra). Importantly, this re‐oxidized, cationic state is the one present during formate binding: under the same reaction atmosphere, κ² ‐HCOO* forms on both Cu and the promoter (vide infra). It is therefore the oxidized, Lewis‐acidic M⁺ site, not a reduced or alloyed form, that stabilizes the formate intermediate.

2.3. Surface Intermediates in CO₂ Hydrogenation

The surface intermediates formed during the CO₂ hydrogenation over Cu/SiO₂, Cu─In/SiO₂, Cu─Ga/SiO₂, and Cu─Zn/SiO₂ catalysts were investigated by operando DRIFTS (Figure 4a), after the pretreatment in H₂ (followed by purging with He, see Figure S2 for experimental procedure). The temporal evolution of surface species was monitored after switching to the reaction mixture of H₂/CO₂. The assignments of the IR bands are listed in Table S1. Upon passing the reaction mixture, bidentate formate on Cu (κ ²‐HCOO* (Cu)) at 2856 and 2931 cm⁻¹ was the first and most common surface species detected over all catalysts. Moreover, κ² ‐HCOO* (Cu) appeared simultaneously with bidentate formate on Zn²⁺ (κ ²‐HCOO* (Zn²⁺)) over Cu─Zn/SiO₂ at 2896 cm⁻¹. The time‐resolved operando DRIFT spectra of Cu/SiO₂, Cu─In/SiO₂, and Cu─Ga/SiO₂ catalysts are similar (Figure S10). Interestingly, while κ ²‐HCOO* was clearly observed on Zn²⁺, its presence on Ga³⁺ or In³⁺ could not be confirmed firmly by this approach alone (without isotopic and transient experiments). After 90 min, methoxy on SiO₂ CH₃O*(SiO₂) at 2858 cm⁻¹ and surface‐bound methanol (CH₃OH*) at 2981 cm⁻¹ built up slowly on the catalysts (Figure 4c). The delay in mass spectrometry response of CH₃OH compared to that of H₂O (Figure 4b) hinted at a stronger interaction of CH₃OH with SiO₂ than that of H₂O, for example, by the adsorption of CH₃OH product over Si─OH transforming into Si─O─CH₃ (i.e., methoxy or CH₃O*(SiO₂)). It should be noted that carbonyl species (CO*) were not detected over any of the catalysts studied.

FIGURE 4.

(a) Time‐resolved operando DRIFT spectra of the surface species formed during the CO₂ hydrogenation reaction over pre‐reduced Cu─Zn/SiO₂ catalysts and structures of relevant surface species. (b) Corresponding normalized ion current signal obtained from mass spectrometry. (c) Operando DRIFT spectra of the surface species formed during the CO₂ hydrogenation reaction over Cu/SiO₂, Cu─In/SiO₂, Cu─Ga/SiO₂, and Cu─Zn/SiO₂ catalysts after reaching steady state (2 h). Pretreatment was performed at 300°C under 10 NmL min⁻¹ of H₂ for 2 h. Reaction conditions are 10 mg catalyst, 230°C, 10 bar, H₂/CO₂ = 3, total flow rate of 10 NmL min⁻¹.

The detection of κ ²‐HCOO* (Cu), particularly in the presence of promoters such as Zn²⁺, Ga³⁺, or In³⁺, was initially unexpected since the κ ²‐HCOO* (Cu) was unobservable and represented by only κ ²‐HCOO* (Zn²⁺) in our previous study where the Cu‐Zn/SiO₂ was prepared via SOMC and atomic layer deposition (ALD) [17]. In the current study, the observation of both κ ²‐HCOO* (Cu) and κ ²‐HCOO* (Zn²⁺) as kinetically distinguishable species (vide infra) help understand the interplay between these formate species—on Cu and promoter sites—may critically influence CH₃OH selectivity as discussed later.

2.4. Relationship between Formate and Methoxy Surface Species and Methanol Formation Rate

Transient isotopic labeling experiments are powerful methodologies that can provide insights into active species with minimal perturbation to the steady state [22]. As shown in Figure 5 after switching from H₂+CO₂ to D₂+CO₂ and vice versa (see Figure S3 for experimental procedure), κ² ‐HCOO* (Cu) at 2931 and 2850 cm⁻¹ and κ² ‐HCOO* (Zn²⁺) at 2896 cm⁻¹ exchanged instantaneously to κ² ‐DCOO*(Cu) at 2167 cm⁻¹ and κ² ‐DCOO*(Zn²⁺) at 2184 cm⁻¹, while CH₃O*(SiO₂, 2958 and 2858 cm⁻¹) and CD₃O*(SiO₂, 2080 cm⁻¹) exchanged slowly. This exchange simplifies the spectra, as C─H combination bands (e.g., ν(C─O) + δ(C─H) at 2931 cm⁻¹) shift out of this region, leaving only the fundamental ν(C─D) stretches at 2167 cm⁻¹. For methoxy, the symmetric ν _s(C─H) stretch at 2858 cm⁻¹ shifts down to the observed ν _s(C─D) band at 2080 cm⁻¹, while the asymmetric ν _as(C─H) stretch at 2958 cm⁻¹ is expected to shift to ∼2160 cm⁻¹, where it overlaps with the intense deuterated formate bands (see Supporting Information for calculations of C─D stretching frequencies).

FIGURE 5.

Transient responses of surface species and gas products during the steady‐state isotopic switching from CO₂ + D₂ to CO₂ + H₂ over Cu─Zn/SiO₂ catalysts. (a) Time‐resolved DRIFT spectra of ν(C─H) region and ν(C─D) region. (b) Component's spectra obtained by MCR applied on the time‐resolved DRIFT spectra. (c) Concentration profiles of the spectra of the corresponding components obtained by MCR. (d) Corresponding normalized ion current signal of isotope‐labeled products. Reaction conditions: 10 mg catalyst, H₂(or D₂)/CO₂ = 3, T = 230°C, P = 10 bar, F_total = 10 NmL min⁻¹.

The separation of kinetically unique spectral features of the surface species during the SSITKA experiment was obtained using multivariate curve resolution (MCR), enabling blind‐source separation (i.e., without references) [23, 24]. As shown in Figure 5b, MCR analysis of the time‐resolved spectra (Figure 5a) yielded deconvoluted spectra of kinetically distinct species (Figure 5b) and their concentration profiles (Figure 5c). The normalized gas concentration profiles determined by mass spectrometry during the transient study are presented in Figure 5d. The kinetics of formation and hydrogenation of the κ² ‐HCOO* (Zn²⁺) were indistinguishable from κ² ‐HCOO* (Cu) with the time resolution of this study (10 s). The fast decay responses of surface κ² ‐HCOO* (Cu and Zn²⁺) and the concomitant increase of κ² ‐DCOO* species indicate that kinetic isotope effects (KIE) are negligible. This may also imply that these formates are easily hydrogenated to CH₃OH and are not spectators in CO₂ hydrogenation. The isotopic exchange of κ² ‐HCOO*/κ ²‐DCOO* species was almost instantaneous (Figure 5c, red and green lines), suggesting that H₂/D₂ dissociative adsorption and formate formation were fast, and not the rate‐limiting step in methanol formation [25].

On the other hand, prominent KIE was observed for methanol formation (Figure 5d, green line (CH₃OH, upon switching to H₂+CO₂) vs. yellow line (CD₃OD, upon switching to D₂+CO₂)), showing significant delay in the formation of CD₃OD compared to that of CH₃OH. This suggests the presence of KIE for formate hydrogenation to CH₃OH and/or for CH₃O* activation by H addition/activation toward formation/desorption of CH₃OH.

The formation profiles of the transiently produced methanol species (CH₃OD vs. CD₃OH, Figure 5d) after the isotopologue switch are convenient reaction‐rate indicators of formate hydrogenation (C─H/C─D formation) or activating methoxy species (O─H/O─D formation) that dominate before the switch. To determine the rate‐limiting step, we analyzed the products formed immediately after the isotopic switch from CO₂​+D₂​ to CO₂​+H₂​ (Figures 5d and S11a). The formation of CD₃​​OH arises from the hydrogenation of pre‐existing deuterated surface species (like CD₃​O*(SiO₂​)), while the formation of CH₃​​OD requires the full hydrogenation of newly adsorbed carbon with incoming H₂. Crucially, the formation of both CD₃​​OH and CH₃​​OD were observed to be very similar. If C─H bond formation was the rate‐determining step, the rate for CH₃​​OD would be substantially lower than for CD₃​​OH. The comparable rates therefore indicate that the KIE for carbon hydrogenation is small and that these steps are not rate‐limiting. This suggests that a different step, such as the methanol formation or desorption, controls the overall reaction rate. The reverse experiment, switching from CO₂​+H₂​​ to CO₂​+D₂​​, yielded analogous results, supporting this conclusion (Figures 5d and S11b). More details on the transient responses and steady‐state distribution of isotopologues are summarized in Figure S13).

Having established that the initial carbon hydrogenation steps are rapid, the rate‐limiting step can be identified from the spectroscopic data. A slow isotopic exchange response for the surface methoxy species (CH₃​O*(SiO₂​), or CD₃​O*(SiO₂​)), correlates directly with the delayed formation of gas‐phase methanol (CH_3​OH or CD₃OD) observed by mass spectrometry (Figures 5c and 5d). This strong correlation, along with a significant kinetic isotope effect for this process (KIE = 2.78, Figure S12), led to the conclusion that the final hydrogenation and/or desorption of the methoxy species is the rate‐limiting step for CO₂​ hydrogenation over Cu‐Zn/SiO₂. Further analysis of the MCR‐resolved responses reveals two surface methoxy populations: a rapidly exchanging reactive pool, likely in proximity to the Cu/Zn²⁺ interface, that governs both the observed KIE and the rate‐limiting behavior, as well as a slowly‐exchanging spectator pool tentatively assigned to remote SiO₂ sites (Figure S14). This insight refines our previous understanding of this system. While we previously considered the CH_3​O*(SiO₂​) species to be a stable, non‐participatory spectator [17], this work reveals its crucial role as a key intermediate participating directly in the rate‐determining step of the catalytic cycle, an insight made possible by isotopic transient analysis.

Notably, SSITKA experiments on Cu/SiO₂, Cu─Ga/SiO₂, and Cu─In/SiO₂ showed similar responses of surface species and isotopic methanol products, suggesting a common reaction mechanism via formate hydrogenation for all catalysts (Figures S15–S17). However, the intriguing details lay in the kinetically resolved spectra after MCR analysis. Similar to Cu─Zn/SiO₂, the kinetics of κ² ‐HCOO* (Ga³⁺) are indistinguishable from κ² ‐HCOO* (Cu) (i.e., they appear in the same MCR‐resolved spectrum since they are not kinetically discernible, Figure S16) and suggest a similar role of Ga³⁺ to Zn²⁺.

On the other hand, κ² ‐HCOO*(In³⁺) was not observed after MCR analysis (Figure S17). While confirming a common mechanism, this MCR analysis is unable to resolve the ambiguity for the Cu─In/SiO₂ system: it cannot be distinguished whether the κ² ‐HCOO*(In³⁺) formate is unobservable because it is consumed rapidly or because it is not formed in the first place. However, the electronic properties of each promoter directly influence the bond characteristics of the formate intermediates, which will be discussed in the next section.

2.5. Effect of Neighboring Promoter on Formate Species and Methanol Formation Over Cu‐Based Catalysts

The application of MCR also allowed for the precise isolation of component spectra, providing insights into the effects of single‐site promoters near Cu nanoparticles. To avoid interference with methoxy bands, which can complicate the spectra, the ν(C─D) stretching frequencies were specifically analyzed. This approach provided clear insights into the bond characteristics of formates on Cu─Zn/SiO₂, revealing distinct vibrational frequencies at 2184 and 2167 cm⁻¹ for κ² ‐DCOO* (Zn²⁺) and κ² ‐DCOO* (Cu) (Figure 5b), and provides direct spectroscopic evidence for the dual role of the Zn²⁺ promoter: it not only provides a distinct Lewis acid site for formate stabilization but also electronically influences the formate bound to the adjacent Cu surface.

For the other promoters (Ga³⁺ and In³⁺), distinct bands for promoter‐bound formate were not resolved in the ν(C–D) region, possibly due to spectral overlap with the κ² ‐DCOO* (Cu) band or a low steady‐state concentration. Nonetheless, the electronic influence of all promoters on the overall formate population is evident. As summarized in Figure 6, the introduction of any single‐site promoter (Zn²⁺, Ga³⁺, or In³⁺) leads to observable shifts in the vibrational frequencies of both the C─D and C─O bonds of the detected formate species compared to Cu/SiO₂. Since higher vibrational frequencies correspond to a shorter and stronger bond, these shifts indicate that all promoters exert a similar electronic effect, which weakens the C─O bond while strengthening the C─H bond of the surface formate.

FIGURE 6.

Catalytic performance of Cu/SiO₂, Cu─Zn/SiO₂, Cu─Ga/SiO₂, and Cu─In/SiO₂ catalysts as a function of formate intermediate bond strengths obtained from the isotopic transient experiment. (a) The intrinsic CH₃OH formation rate versus the ν(C─O) stretching frequency, which serves as a proxy for the formate‐promoter binding strength and strength of the Lewis acid site. (b) The intrinsic CH₃OH selectivity versus the ν(C─D) stretching frequency of the κ² ‐DCOO* species. The κ² ‐DCOO* species wavenumber for Cu─Zn/SiO₂ (2173 cm^{− 1}) represents the band center of the overlapping κ² ‐DCOO* (Cu) and κ² ‐DCOO*(Zn²⁺) components (see Figure 5b (bottom)), which was determined by Gaussian fitting.

This electronic influence on the formate bonds provides the key to understanding the catalytic performance. The ν(C─O) stretching frequency, which acts as a proxy for the formate binding strength (and Lewis acid strength), exhibits a distinct volcano‐shaped relationship with the intrinsic CH₃OH formation rate (Figure 6a). This trend illustrates the Sabatier principle. At high ν(C─O) frequencies (weak binding), unpromoted Cu/SiO₂ binds formate too weakly for efficient activation, resulting in a low rate. Conversely, at low ν(C─O) frequencies (strong binding), the strong Lewis acidity of In³⁺ causes the formate to over‐stabilize. This creates a kinetic trap that prevents subsequent hydrogenation, thus suppressing the overall reaction rate. At the optimum, Cu─Zn/SiO₂ and Cu─Ga/SiO₂ possess an intermediate binding strength that balances formate activation with its subsequent hydrogenation, leading to the highest intrinsic activities.

These findings are consistent with previous reports on related SOMC‐derived catalysts. For example, the enhanced CH₃OH formation rate on Cu─M/SiO₂ (M = Ti, Zr, Hf, Nb, and Ta) was related to the Lewis acidity of the dopant and attributed to the stabilization of HCOO* and CH₃O* intermediates at the Cu periphery [20]. It has been demonstrated that isolated Zr⁴⁺ [19] and Ti⁴⁺ [26] sites adjacent to Cu nanoparticles—as opposed to bulk oxide properties [19] —are sufficient to enhance CH₃OH selectivity by stabilizing these formate and methoxy intermediates [26]. Similarly, Lewis acid sites (M = Y, Sc, Zr, and Ta) over Cu─M/γ‐Al₂O₃ were shown to simultaneously weaken the C─O bond and strengthen the C─H bond of κ² ‐HCOO* via their electron‐withdrawing character, an effect that destabilizes the κ² ‐HCOO* intermediate, increases the electrophilicity (positive charge) of the central carbon, makes it more susceptible to nucleophilic attack by a hydride (H⁻) from the Cu, and eventually reduces the apparent activation energy for CH₃OH formation [27]. However, these studies, which observed a linear increase in reactivity with Lewis's acidity, effectively mapped only the weak‐binding side of the volcano plot. Indeed, over‐stabilization of HCOO* on strongly Lewis acidic γ‐Al₂O₃ has been shown to be detrimental, opening a side‐reaction pathway via methyl formate (HCOOCH₃) that decomposes to promote CO selectivity [28]. Notably, this methyl formate decarbonylation is specific to the strongly Lewis acidic γ‐Al₂O₃ surface [28]; on SiO₂‐ and ZnO‐based supports the ester is instead preferentially hydrogenated back to methanol [29], so this CO‐forming route is not expected to be significant for our Cu‐M/SiO₂ catalysts, where CO formation is instead governed by direct κ² ‐HCOO* decomposition on Cu.

While the C─O bond strength dictates the overall activity according to the Sabatier principle (Figure 6a), the C─D bond strength serves as a powerful and direct descriptor for CH₃OH selectivity. As shown in Figure 6b, the intrinsic CH₃OH selectivity correlates directly and linearly (R ²> 0.99) with the ν(C─D) stretching frequency of the κ² ‐DCOO* intermediate.

This provides a clear mechanistic link: κ² ‐HCOO* is also the key intermediate for the CO (RWGS) pathway on Cu, and its subsequent decomposition to CO involves both C─O and C─H bond cleavage (HCOO* → CO* + OH*) [30]. Based on DFT calculations, the HCOO* based pathway for CO formation (RWGS) on Cu is kinetically favorable proceeding via HCOOH* → cis‐COOH* → CO* pathway and hydrogenation of HCOOH* to H₂COOH* is a crucial route to CH₃OH [31]. Therefore, strengthening the C─H bond via promoter introduction (Figure 6b) could, in principle, suppress κ² ‐HCOO* decomposition pathway while facilitating CH₃OH formation. This conclusion is strongly supported by HCOOH‐TPSR which proves that the higher stability of HCOO* can inhibit its decomposition into CO and directly correlates with higher methanol selectivity [32].

2.6. Role of Promoter in Driving Methanol Formation—Relation of Formate Spillover and Methanol Selectivity

The nature of the active site of Cu─ZnO‐based catalyst remains highly debated, and it was proposed that CO₂ hydrogenation to CH₃OH occurs via zinc formate Zn(HCOO)₂ intermediates originating from hydrogen dissociation over Cu, H‐spillover to ZnO, and subsequent CO₂ activation over ZnO [33]. While ZnO can generate formate directly via this route, it is associated with bulk or clustered ZnO; in our catalyst the working Zn speciation is dominated by isolated interfacial Zn²⁺ (in situ XAS), which is itself inactive for CO₂ hydrogenation and remains fully oxidized in the absence of Cu [14], making independent CO₂ activation at Zn²⁺ an unlikely formate source and pointing instead to formate generation on Cu followed by spillover (vide infra, Figure 7) [34]. Notably, operando studies of industrial‐type catalysts indicate that the catalytically relevant zinc under reaction conditions is not bulk ZnO but is instead oxidized, dispersed interfacial Zn. This includes a reactive zinc formate intermediate and, under high‐conversion conditions, low‐nuclearity ZnO_x species resembling silica‐supported Zn, consistent with the single‐site Zn²⁺ in this study [33, 35].

FIGURE 7.

Transient responses of surface species and gas products during switching from He + CO₂ to H₂ + He over Cu─Zn/SiO₂ catalysts. (a) Component spectra obtained by MCR applied on the time‐resolved DRIFT spectra. (b) Concentration profiles of the spectra of the corresponding components obtained by MCR‐ALS. Reaction conditions. (c) Corresponding normalized ion current signal of CH₃OH products. Reaction conditions: 10 mg catalyst, He/CO₂ = H₂/He = 3, T = 230°C, P = 20 bar, F_total = 10 NmL min⁻¹.

Even using a well‐defined Cu─Zn/SiO₂ prepared via SOMC, it was also challenging to pinpoint the origin of the high CH₃OH selectivity due to the presence of Lewis acidic Zn sites and/or reduced Zn sites (alloy) [14]. While some recent operando studies have concluded that formates bound directly to the metallic Cu⁰ surface are the main reaction intermediates [36], other works highlight that Lewis acidic sites at the perimeter can play an essential role as an active reservoir for κ² ‐HCOO*, which can spillover from the metal particle and be stabilized [34, 37, 38]. The selectivity to CH₃OH over CO should be related to the spillover rate toward M⁺ sites (Zn²⁺, Ga^3+, and In³⁺) as well as the stability of HCOO* over M⁺ sites, which can be elucidated by transient experiments.

To directly probe this proposed formate spillover pathway, a transient experiment separating the CO₂ and H₂ activation steps was employed (see Figure S4 for experimental procedure). By alternatingly passing CO₂ or H₂, this method provides further insights into CO₂ activation and subsequent hydrogenation. The limited supply of reactants (especially H₂) allows the capturing of the fast‐forming species (e.g., CO* or κ² ‐HCOO* over Zn²⁺, Ga³⁺, In³⁺), which were readily hydrogenated during steady‐state CO₂ hydrogenation.

The evolution of the surface species during the CO₂ and H₂ phases over the Cu─Zn/SiO₂, is presented in Figure 7. Unlike the previous steady‐state experiment, this alternating‐feed method allowed the formation and consumption of κ² ‐HCOO* (Cu) and κ² ‐HCOO* (Zn²⁺) to be kinetically distinguished using MCR‐ALS (Figure 7a).

The resulting concentration profiles (Figure 7b) reveal a clear temporal sequence of events at the beginning of the H₂ phase. First, κ² ‐HCOO* (Cu) forms and is consumed rapidly. Second, after a slight delay, κ² ‐HCOO* (Zn²⁺) forms and then begins to decay, but at a much slower rate than the κ² ‐HCOO* (Cu). Finally, the decay of this κ² ‐HCOO* (Zn²⁺) directly correlates with the formation of CH₃O*(SiO₂) and the detection of gas‐phase CH₃OH (Figure 7c). This sequential transfer of the κ² ‐HCOO* intermediate from Cu to Zn, which then leads to the final product, provides direct experimental evidence for the proposed formate spillover pathway from Cu to adjacent Zn²⁺ sites, where the intermediate is stabilized for hydrogenation to CH₃OH.

For Cu‐Ga/SiO₂, the MCR analysis showed kinetically indistinguishable spectra comprising κ² ‐HCOO* (Cu), κ² ‐HCOO* (Ga³⁺), CH₃O*(Ga³⁺) and adsorbed CH₃OH (Figure S20), all of which were consumed rapidly under H₂. This suggests that the spillover of κ² ‐HCOO* from Cu to Ga³⁺ and its subsequent hydrogenation are both fast, such that the intermediates cannot be resolved at this time scale (faster than that over Cu‐Zn/SiO₂). In contrast, for Cu‐In/SiO₂, while MCR analysis also shows evidence of spillover of kinetically indistinguishable spectra comprising κ² ‐HCOO* (In³⁺) and CH₃O*(In³⁺) (Figure S19), they were not consumed during the H₂ phase and instead increased and reached a stable plateau. This result suggests that κ² ‐HCOO* once spilled over from Cu to In³⁺ was less favorable to further hydrogenate to produce CH₃OH, which is due to overly strong binding of κ² ‐HCOO* on In³⁺ and explains the lower intrinsic CH₃OH formation rate.

This transient experiment also provided crucial insights into the origin of CO formation. During the CO₂ phase, Cu/SiO₂, Cu─In/SiO₂ and Cu─Ga/SiO₂ all produced both κ² ‐HCOO* and CO* (on Cu⁰) as detectable surface species (Figures S18–S20). An analysis of their interplay revealed that the κ² ‐HCOO* (Cu) signal decreased as the CO* signal increased. This provides strong evidence that κ² ‐HCOO* (Cu) acts as an intermediate for CO* formation via decomposition (the RWGS pathway) as previously discussed. CO* was then rapidly converted upon the introduction of the H₂ phase (the reason we could not observe it under steady state). In sharp contrast, CO* remained undetectable over Cu─Ga/SiO₂ (Figure S20). This finding clearly demonstrates why Cu─Ga/SiO₂ is so selective: the formate spillover and subsequent hydrogenation are so fast that this pathway completely outcompetes the parallel decomposition pathway on Cu, preventing CO formation.

While this study used a CO₂/H₂ feed, in technical CO/CO₂/H₂ syngas CO₂ remains the immediate carbon source for methanol, with co‐fed CO acting mainly through the water‐gas shift to regenerate CO₂ and scavenge product water [3]. The formate spillover pathway and the C─O/C─H bond‐strength descriptors established here are therefore expected to extend to CO‐blended feeds, with the CO/CO₂ ratio additionally tuning the oxidation state of the dynamic Cu─M⁺ interface.

In summary, while Cu─Zn/SiO₂ is the most active catalyst for methanol production, Cu─Ga/SiO₂ achieves superior methanol selectivity by more effectively inhibiting the undesired CO side‐reaction via fast formate spillover. Conversely, Cu─In/SiO₂ exhibits high selectivity but low activity due to the over‐stabilization of the formate after spillover, which inhibits further hydrogenation. As summarized schematically in Figure 8, these findings reveal how a promoter dictates product selectivity through control of formate spillover and stabilization. Ultimately, they provide a foundation for rational catalyst design: vibrational descriptors of the formate intermediate can guide the computational screening of promoters that balance intermediate stability against decomposition, while synthetic strategies that maximize the density of active interfacial sites can favor spillover over decomposition, together offering a route to further suppressing the unselective CO by‐product.

FIGURE 8.

Proposed dual‐pathway mechanism for CO₂ hydrogenation on Cu─M/SiO₂ catalysts, detailing the molecular steps that control product selectivity. The unselective pathway (left) involves the decomposition of the formate intermediate on the Cu surface to produce CO. The selective pathway (right) is enabled by the crucial spillover of formate from Cu to an adjacent promoter site (M⁺). This protects the intermediate and initiates a reaction sequence that proceeds via a stable methoxy species on the SiO₂ support, whose final hydrogenation to release methanol is the overall rate‐limiting step.

3. Conclusion

The Cu─M/SiO₂ catalysts (M = Zn, Ga, and In) were prepared by surface organometallic chemistry to position isolated cationic M sites adjacent to Cu nanoparticles in the form of silicates. Although these promoters can initially form alloys upon reduction, in situ and ex situ XAS show they dynamically restructure into isolated interfacial cationic M sites under reaction conditions. Operando DRIFTS identified Cu‐bound formate as a common intermediate across all catalysts, with the cationic M sites enabling formate spillover from Cu to the promoter (Figure 8). Relative to unpromoted Cu/SiO₂, where formate is prone to decomposition into CO, the cationic M sites stabilize the intermediate, suppressing decomposition and lowering CO formation.

This stabilization stems from the promoter's electronic effect on the adsorbed formate: it weakens the C─O bond and strengthens the C─H bond. Each bond strength then acts as a quantitative descriptor of a different performance metric: the C─O bond strength sets activity through a Sabatier‐type volcano, while the C─H bond strength varies linearly with methanol selectivity. Cu─Zn/SiO₂ shows the highest activity, consistent with its optimal formate–Zn²⁺ binding on the Sabatier volcano, whereas the faster spillover to Ga³⁺ more effectively suppresses CO formation and gives the highest methanol selectivity. In contrast, In³⁺ binds formate too strongly after spillover, creating a kinetic trap that lowers activity.

All catalysts proceed through the same formate‐hydrogenation mechanism. Isotopic transient experiments showed negligible KIE for H₂ dissociation, formate formation, and formate hydrogenation, but a significant KIE for methoxy hydrogenation/desorption to methanol, identifying this as the rate‐limiting step. These insights clarify the long‐debated Cu─ZnO synergy underlying industrial Cu/ZnO/Al₂O₃ catalysts and provide a foundation for rationally designing selective catalysts by tuning the interfacial environment and promoter acidity where formates form and are stabilized.

Author Contributions

Nat Phongprueksathat: conceptualization, methodology, investigation, writing – original draft, data curation, software, formal analysis, validation, and visualization. Gina Noh: methodology, writing – review and editing, and investigation. Scott R. Docherty: investigation, writing – review and editing, and methodology. Damian Vico van Berkel: investigation, methodology, software, formal analysis, writing – review and editing, and validation. Yannik Stiefel: investigation, writing – review and editing, and methodology. Christophe Copéret: funding acquisition, writing – review and editing, project administration, resources, supervision, and methodology. Atsushi Urakawa: project administration, resources, supervision, methodology, conceptualization, funding acquisition, and writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: Experimental procedures, commercial chemicals, catalyst preparation procedures, instrument specifications, and characterization techniques are covered in greater detail in the Supporting Information. The authors have also cited additional references within the Supporting Information [14, 15, 16, 17, 19, 23, 24, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49].

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.