Engineering Peripheral Metal‐Oxide Catalysis: Interparticle Spacing in Cu/ZrO₂ Catalysts for Methanol Synthesis by CO₂ Hydrogenation

Abstract

The periphery surrounding oxide‐supported metal nanoparticles plays a crucial role in many catalytic reactions that exhibit strong metal‐oxide promotional effects. Engineering this catalytically active periphery, where kinetically relevant surface intermediates are efficiently turned over, offers a pathway to optimized performance, yet it remains challenging due to the need for precise control over nanospatial catalyst features. Herein, we address this subject for the relevant case of methanol synthesis by CO₂ hydrogenation on Cu/ZrO₂ catalysts. The methanol synthesis rate reaches a maximum at a surface‐to‐surface Cu interparticle distance of ca. 15 nm. Operando modulation–excitation diffuse reflectance infrared spectroscopy reveals that this optimal spacing maximizes the fraction of surface‐bound HCOO* intermediates, stabilized on coordinatively unsaturated Zr(IV) Lewis acid sites on the ZrO₂ support, which are dynamically involved in catalysis. This particle spacing represents a shift in the reaction's kinetic control regime and the apparent activation energy for methanol synthesis. Engineering Cu interparticle spacing to the optimal value results in exceptionally high metal‐specific methanol formation rates under industrially relevant reaction conditions. More broadly, our findings highlight that, beyond metal particle size, interparticle spacing is a key design parameter for catalyst systems featuring functional metal‐oxide interfaces.

The catalytically active periphery around oxide‐supported metal nanoparticles is key in reactions featuring marked metal‐oxide promotional effects. For Cu/ZrO₂ CO₂ hydrogenation catalysts, a Cu interparticle spacing of ∼15 nm optimizes methanol production rate and selectivity. This spacing maximizes the surface density of formate intermediates on peripheral ZrO₂ which are dynamically turned over, marking a shift in the reaction's kinetic control.

Introduction

In heterogeneous catalysis, interfacing chemically dissimilar catalytic functionalities, such as metal nanocrystals with oxide or chalcogenide support materials, often results in performances that deviate significantly from a mere additive combination of the individual components' behaviors.^{[ 1} , ² , ³ , ⁴ , ⁵ , ^{6 ]} Bifunctionality is often considered to involve either the development of a new type of interfacial active site or the surface migration (spillover) of reaction intermediates across material interfaces, transitioning between active surfaces that adhere to different scaling relationships, thereby breaking with those limitations imposed by single‐component catalysts.^{[ 7} , ^{8 ]}

The concerted site action likely occurs within the metal/support periphery, where the bifunctional mechanism operates. Engineering these active peripheral areas at the nanoscale is anticipated crucial for optimizing performance but remains challenging due to the need for precise control over nanospatial features of supported metal catalysts. While the effects of nanoparticle (NP) size have been widely studied, nanoparticle spacing has only recently attracted attention as a design parameter,^{[ 9 ]} often circumscribed to catalyst stability.^{[ 10} , ^{11 ]}

Methanol synthesis via the selective hydrogenation of CO₂ is a process of eminent interest in the context of storage and conversion of renewable energy.^{[ 12} , ¹³ , ^{14 ]} Additionally, it represents a major example for a heterogeneously catalyzed reaction wherein interfacing metallic (Cu) and oxidic (ZrO₂, ZnO, CeO₂, etc.) nanomaterials unlocks performance levels which remain out of reach for either of the materials separately. Earlier work with model inverse,^{[ 15} , ¹⁶ , ^{17 ]} as well as conventional oxide‐supported^{[ 10} , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ^{28 ]} metal catalysts, has shown that the concerted action of Cu nanoparticles and metal/lanthanide oxide promoter/support materials is critical to attain relevant methanol formation rates and selectivity. Moreover, fundamental kinetic parameters for methanol synthesis with oxide‐supported Cu nanoparticles have been correlated with physicochemical descriptors of coordinatively unsaturated Lewis acid centers on the oxide support surface, confirming the involvement of these sites in kinetically relevant reaction steps.^{[ 29} , ^{30 ]}

Previous research has shown that formate (HCOO*) species accumulate on the surface of the oxide catalyst component/support, becoming the predominant surface species under industrially relevant CO₂ hydrogenation conditions. Both experimental^{[ 31 ]} and computational^{[ 21} , ²⁶ , ^{32 ]} studies have identified formates as particularly stable reaction intermediates, and their further hydrogenative conversion most likely associated with a high degree of rate control for the overall reaction. Other studies proposed further methoxy hydrogenation to methanol as the slowest reaction step.^{[ 33 ]} Additionally, formate species have been suggested to participate in dynamic surface migration processes across Cu/oxide interfaces.^{[ 34} , ^{35 ]} On the working catalyst, only a fraction of these surface intermediates is expected to engage in the methanol synthesis mechanism. The remaining exhibit too low net renovation rates and thus act as mere surface spectators, dominating vibrational spectroscopy observations but not contributing effectively to catalysis turnover events.^{[ 33 ]}

Herein, we address the challenge of quantifying and engineering the catalysis‐relevant periphery (CRP) at the interface of metal and oxide catalyst components. We tackle this challenge by systematically controlling the average nanoparticle spacing in a series of Cu/ZrO₂ supported methanol synthesis catalysts with similar Cu nanoparticle sizes. Modulation–excitation diffuse–reflectance Fourier‐transform infrared spectroscopy (ME‐DRIFTS) is applied under process conditions to quantitatively elucidate the abundance and dynamics of surface reaction intermediates as a function of Cu interparticle distance, resolving them from stagnant, surface spectator species. This approach enables the rational identification of an optimal metal nanoparticle spacing of ∼15 nm, for which remarkably high methanol formation rates are attained, e.g., >19 mmol_MeOH mmol_Cu ⁻¹ h⁻¹ at 513 K.

Results and Discussion

A series of Cu/ZrO₂ supported catalysts was synthesized via impregnation of a solution of Cu(NO₃)₂ metal precursor on a monoclinic ZrO₂ (P2₁/c) support material (for details refer to Methods in the Supporting Information). The nominal surface‐specific Cu content (δ _Cu) was systematically varied in the sub‐monolayer range of 0.3–7.5 Cu_at nm⁻², while thermal nitrate decomposition conditions were adjusted to drive copper oxide redispersion on the surface of the zirconium oxide support material.^{[ 10} , ^{36 ]} The experimental copper surface contents, as determined by inductively‐coupled plasma optical emission spectroscopy (ICP‐OES), agreed with the nominal values, with deviations <1.1% (Table S1 in the Supporting Information). Powder X‐ray diffraction (XRD) analysis showed no diffraction signals attributable to CuO in the oxidic catalyst precursors, indicating a high dispersion of Cu(II) oxide species in the form of entities lacking long‐range atomic order (Figure S1). Similarly, no diffraction signals for Cu⁰ were detected following reductive activation in 15%H₂/N₂ at 523 K, confirming that copper existed as very small crystallites (<4 nm), which remained therefore XRD‐undetectable (Figure S2).

Neon low‐energy ion scattering (²⁰Ne⁺‐LEIS) was applied to assess the geometric copper coverage and dispersion on the ZrO₂ carrier following in situ catalyst reductive activation (Figure S3). LEIS specifically determines the chemical composition of the topmost atomic layer, therefore, relative signals for Cu and Zr elements provide a measurement for the fraction of the ZrO₂ surface which is masked by Cu nanoparticles. A Cu(111) single‐crystal and the bare, nanocrystalline ZrO₂ support material were applied as reference materials to calibrate the LEIS signals for surface Cu and Zr atoms, respectively. To realistically depict the multifaceted character of the supported metal nanocrystals, a correction factor was systematically applied to account for the difference in surface Cu atom density between the Cu(111) monocrystal (17.8 Cu_at nm⁻²) and the supported Cu nanoparticles, as derived from the corresponding first‐principles Wulff construction assuming cuboctahedral nanocrystal morphology (14.2 Cu_at nm⁻²) (Figure 1a). The monolayer coverage for Cu on ZrO₂ was approximated by the maximum surface‐specific content of Cu atoms which could be accommodated on a representative m‐ZrO₂ (‐111) surface as determined with periodic DFT modelling, i.e., 15.5 Cu_at nm⁻² (Figure 1b). As shown in Figure 1c, the normalized ²⁰Ne⁺‐LEIS copper signal [S _Cu/S _Cu ^Max],^{[ 37 ]} i.e., the extent to which the topmost surface of the ZrO₂ carrier is masked by Cu nanoparticles, was found to scale essentially linearly with the Cu surface content over the range of Cu loading studied, indicating that Cu dispersion remains fundamentally constant across the series of catalysts.

Figure 1.

Metal dispersion and interparticle spacing in Cu/ZrO₂ catalysts. a) Wulff construction of a 3 nm diameter, cuboctahedral Cu nanoparticle and facet relative surface areal contributions. b) Zenital view on a DFT‐optimized (PAW‐PBE‐D3) model for a 1 ͯ 1 ͯ 4 Cu/m‐ZrO₂(‐111) slab at the maximum Cu surface content which can be accommodated with stable 2D Cu dispersion, i.e., in the absence of 3D Cu clustering and ion shadowing effects (7 Cu_at/(ZrO₂ (‐111) cell)⁻¹ or 15.5 Cu_at nm⁻²). Atoms at the bottom two layers of the slab are depicted with a squared color pattern. c) Dependence of the normalized copper ²⁰Ne⁺‐LEIS signal (S _Cu/S _Cu ^Max) with the surface copper content (δ_Cu, Cu_at nm⁻²). d) Dependence of the average Cu nanoparticle diameter (d _Cu, nm), as derived from the analysis of Cu‐K EXAFS spectra after in situ catalyst reduction, and the corresponding average Cu surface‐to‐surface inter‐particle distance (D _NP‐NP, nm), with the Cu surface content for the series of Cu/ZrO₂ catalysts. The dashed lines are provided as a guide to the eye.

N₂O oxidative chemisorption results showed unrealistically high N₂O uptakes even at 303 K, suggestive of subsurface oxidation of Cu⁰ nanoparticles or additional oxidative titration of partially reduced sites on the ZrO₂ support (see Experimental Methods in the Supporting Information) and therefore proved unreliable to evaluate Cu dispersion for this series of highly dispersed catalysts. Therefore, extended X‐ray absorption fine‐structure spectroscopy (EXAFS) has been applied to determine average Cu nanoparticle sizes following in situ reduction activation of the Cu/ZrO₂ catalysts. The Cu‐K EXAFS spectra, as well as the results for spectra fitting, are provided in Figure S4 and Table S2 in the Supporting Information. The average Cu nanoparticle diameter was derived from the corresponding 1^st‐shell Cu–Cu coordination number (CN) according to a geometrical model for supported fcc Cu nanoparticles with an equatorially truncated cuboctahedra geometry (see details in the Experimental Methods section of the Supporting Information). A slight progressive increase in average Cu nanoparticle diameter (d _Cu) was detected, from 2.02 to 2.90 nm, on increasing δ _Cu from 0.36 to 8.44 Cu_at nm⁻² (Table S1). Yet, as shown in Figure 1d, the variation in d _Cu is within the experimental uncertainty, with the only exception of the catalyst with the lowest Cu loading (0.36 Cu_at nm⁻²), emphasizing an essentially uniform metal nanoparticle size across the range of surface Cu loadings examined. This result is in line with the fundamentally constant Cu dispersion inferred from the complementary ²⁰Ne⁺‐LEIS results.

The comparatively uniform Cu nanoparticle size ascertained across the series of Cu/ZrO₂ catalysts emerges as a result of atomic layer deposition‐like phenomena from the sublimation of pore‐confined Cu(NO₃)₂(anh) under the metal nitrate drying and decomposition procedure applied herein for catalyst synthesis.^{[ 36} , ³⁸ , ^{39 ]} It is inferred that nucleation, rather than growth, exerts kinetic control over the development of Cu⁰ nanoparticles from the initially dispersed Cu (II) oxidic species in the as‐calcined catalyst precursors. As observed earlier for other model planar supported metal catalysts synthesized through atomic layer deposition mechanisms in inert gas (N₂) plasma atmospheres followed by reduction with hydrogen,^{[ 40 ]} this is expected to provide a regular series of materials wherein the metal nanoparticle surface density scales systematically with the surface Cu content, whilst nanoparticle size varies only within a very narrow margin.

Further experimental evidence supporting the minimal variation in Cu nanoparticle size across catalysts with markedly different surface metal loadings–as well as insights into metal particle distribution, which cannot be resolved by averaging techniques such as EXAFS or ²⁰Ne⁺‐LEIS–was obtained through independent analyses using high‐resolution energy‐dispersive spectroscopy (EDS) in a double aberration‐corrected scanning transmission electron microscope (STEM). While direct high‐angle annular dark‐field (HAADF)‐STEM visualization of the Cu nanocrystals was hampered by their small size and their limited Z contrast relative to the ZrO₂ carrier (Z _Cu = 29 versus Z _av,ZrO2 = 19), using an aberration‐free sub‐nm electron probe in combination with four‐quadrant Super‐X EDS detection served to resolve the supported Cu nanoparticles (Figure S5). Average Cu NP diameters of ca. 3.2–3.5 nm were determined with EDS‐STEM regardless of the Cu surface content, in fair agreement with the Cu nanoparticle size determined by EXAFS. Cu nanoparticles appeared homogeneously distributed over ZrO₂, with spatial distributions matching closely those expected for a statistically random location across the oxide carrier material (Figure S6), hence on average uniformly spaced. Only for the case of the highest Cu loading of 7.5 Cu_at nm⁻² were few aggregates detected, with locally higher metal density, which nevertheless consisted of individual, smaller Cu nanoparticles in closer proximity (Figure S5).

The average surface‐to‐surface Cu interparticle distance (D _NP‐NP) was determined from the corresponding average Cu NP diameter and surface nanoparticle density. As shown in Figure 1d D _NP‐NP decreases from 22.4 to 5.7 nm on increasing the surface metal loading across the series of catalysts, which are hereafter denoted as Cu/ZrO₂(DX), where X stands for D _NP‐NP. Only in the case of the highest Cu loading catalyst, i.e., lowest interparticle distance, is D _NP‐NP expected to be locally shorter within the few events of Cu nanoparticle clustering detected by EDS‐STEM mapping.

The performance of the series of catalysts for methanol synthesis by CO₂ hydrogenation was assessed under relevant operation conditions (gas feed CO₂:H₂:Ar 23.5:71.5:5 (vol), T = 493–523 K, P = 2.5 MPa, WHSV = 580–180 g_CO2 g_Cu ⁻¹ h⁻¹). The CO₂ conversion per reactor pass was limited to <9%, thereby keeping approach to equilibrium terms <0.06 (see Experimental Methods in the Supporting Information). Methanol formation rates and selectivity were extrapolated to zero CO₂ conversion, to ensure differential conditions. As shown in Figure 2, the methanol formation rate (r _MeOH,0) at 513 K showed a pronounced volcano evolution with interparticle spacing, peaking at 19.2 ± 1.1 mmol_MeOH mmol_Cu ⁻¹ h⁻¹ for an average D _NP‐NP of ca. 15 nm (Table S3). Similar dependences were observed at other reaction temperatures in the studied range, as well as when the methanol formation rate was normalized per unit Cu‐ZrO₂ peripheral length (Figure S7). Methanol selectivity remained high and essentially constant in the range of 59%–68% on increasing the interparticle spacing up to ca. 15 nm (Figure 2). CO was the other major product, whereas the selectivity to methane and dimethyl ether (secondary methanol dehydration product) remained fairly low at <0.5% and <0.3%, respectively (Table S4). Remarkably, further enlarging the interparticle spacing beyond 15 nm led to a significant decrease in methanol selectivity with D _NP‐NP (Figure 2).

Figure 2.

Methanol synthesis rate and selectivity in CO₂ hydrogenation with the series of Cu/ZrO₂ catalysts as a function of the copper interparticle spacing (D _NP‐NP). Reaction conditions: gas feed CO₂:H₂:Ar 23.5:71.5:5 (vol), T = 513 K, P = 2.5 MPa, WHSV = 180–580 g_CO2 g_Cu ⁻¹ h⁻¹. Rate and selectivity data reported at differential CO₂ conversion conditions (see Methods section 1.3 in the Supporting Information for details). Error bars along the x‐axis represent the standard error of the mean Cu interparticle spacing, after propagation of the standard error of the mean Cu nanoparticle size as derived from the analysis of in situ EXAFS results. Error bars along the y‐axes represent the standard error of the mean, ca. 5.5% for r_MeOH,0, as determined from three independent test repetitions with selected catalysts. For methanol selectivity, the error bars (ca. 3.1%) are smaller than the symbols.

For catalysts Cu/ZrO₂(D22.4), Cu/ZrO₂(D15.2) and Cu/ZrO₂(D5.7), which span across the range of Cu interparticle spacing studied, in situ EXAFS was extended to also assess changes in d _Cu upon exposure to methanol synthesis conditions. The corresponding Cu‐K EXAFS spectra as well as the results for spectra fitting are provided in Figure S8 and Table S2 in the Supporting Information. Marginal increments in d _Cu of 0.3–0.4 nm were detected following catalysis, which are smaller than the experimental uncertainty for d _Cu and therefore statistically negligible. Besides, during catalyst testing, the initial test conditions (T = 493 K, P = 2.5 MPa and WHSV = 580 g_CO2 g_Cu ⁻¹ h⁻¹) were re‐evaluated, after each catalyst had been exposed to the full range of operating conditions in this study (T = 493–523 K, $X_{{\mathrm{CO}}_2}$ = 0.9%–8.1%). In all cases, methanol formation rates were reproduced within 3%–8% of those measured for the fresh, as‐reduced catalyst. Collectively, these results discard any significant contribution of neither Cu nanoparticle sintering nor catalyst deactivation on the trends discussed above for performance as a function of Cu nanoparticle spacing.

The nature and dynamics of surface reaction intermediates were studied under operando conditions using modulation–excitation diffuse‐reflectance infrared Fourier‐transform spectroscopy (ME‐DRIFTS).^{[ 41} , ⁴² , ^{43 ]} A periodic external stimulation was achieved through a step‐function on the H₂:CO₂ molar ratio in the gas feed, from 0 to 3.0, at a high gas space velocity (WHSV) of 2.1–49.0·10³ g_CO2 g_Cu ⁻¹ h⁻¹. This approach allowed for the specific study of adsorbates dynamically involved in the reaction, effectively discriminating them from a background of unresponsive (spectator) species. Phase‐sensitive detection (PSD) was applied to demodulate the periodic system response.

Following the attainment of a pseudo‐steady state under reaction conditions, DRIFT spectra in the υ(OCO) region corresponded to a convolution of different contributions, primarily surface formates (HCOO*), and additionally methoxy (H₃CO*) and carbonate (OCOO*) species (Figure S9 and Table S5). Bands at 1586–1595 cm⁻¹, alongside less intense but clear split doublet at ∼1374 and ∼1390 cm⁻¹ have been assigned to υ_as(CO₂ ⁻), δ(CH) and υ_s(CO₂ ⁻) vibrational modes in bidentate formate species, respectively. Mono‐ and bidentate surface carbonates were detected with bands in 1400–1430 cm⁻¹ and 1600–1750 cm⁻¹ (υ_as(CO₂ ⁻)), as well as in the 1240–1350 cm⁻¹ region (υ_s(OCO)).^{[ 44 ]} Surface methoxy species have been identified based on bands in the 1420–1500 cm⁻¹ spectral range, attributed to the characteristic δ _as(CH₃) vibrational mode.^{[ 29} , ^{33 ]}

To discriminate the vibrational fingerprints for adspecies stabilized onto the ZrO₂ oxide carrier and on the surface of the metallic Cu nanoparticles, a reference Cu/SiO₂ catalyst was synthesized (Figure S10) and tested under identical methanol synthesis conditions. The SiO₂ support in this catalyst exhibits a silanol‐terminated surface and lacks Lewis acid sites. As a result, the steady‐state DRIFT spectrum recorded under CO₂ hydrogenation conditions showed no bands ascribable to reaction intermediates stabilized on the oxide support. Inspection of the υ(CH) spectral range showed bands peaking at ∼2852 and ∼2942 cm⁻¹, corresponding to the υ(CH) stretching vibrational mode of HCOO* and the υ_as(CH₃) mode of H₃CO* adspecies, respectively, sitting on Cu⁰ surfaces (Figure S10c), in agreement with earlier studies on metallic Cu single crystals.^{[ 45} , ⁴⁶ , ^{47 ]} For the set of Cu/ZrO₂ catalysts, HCOO* and H₃CO* species were additionally, and primarily, detected adsorbed on the Zr(IV) Lewis centers exposed on the ZrO₂ support, with υ(CH) and υ_as(CH₃) vibrational modes peaking at ∼2871 and ∼2929 cm⁻¹, respectively.

Figure 3 shows the operando ME‐DRIFTS results in both time panel (a) and phase (b) domains for the specific case of Cu/ZrO₂(D15.2), as a showcase. The corresponding spectral collections for additional catalysts are provided in Figures S11 to S14 in the Supporting Information.

Figure 3.

Assessment of surface intermediates and dynamics by ME‐DRIFTS. a) Initial steady‐state DRIFT spectrum (upper graph) and time‐domain operando ME‐DRIFT spectra collection (contour plot); and b) phase‐domain operando spectra collection for Cu/ZrO₂(D15.2) as a representative showcase. See Figures S11 to S14 in the Supporting Information for data collected with additional Cu/ZrO₂ catalysts. c) Evolution of the ME‐DRIFTS in‐phase angle for different surface adspecies with the Cu interparticle spacing. d) Dependence of the degree of rate control for formate hydrogenation relative to the overall carbonate to methoxy conversion (RC_HCOO* ) and apparent activation energy for methanol synthesis (E _app) with the Cu interparticle spacing. The following fingerprint IR bands have been used to track different adspecies: carbonate on ZrO₂ (υ_as(CO₂ ⁻),1623 cm⁻¹), formate on ZrO₂ (bd‐υ_s(CO₂ ⁻),1371 cm⁻¹) and methoxy on ZrO₂ (δ _as(CH₃), 1461 cm⁻¹). Experimental ME‐DRIFTS conditions: T = 493 K, P = 0.5 MPa, WHSV = 2.1–49.0·10³ g_CO2 g_Cu ⁻¹ h⁻¹; H₂:CO₂ molar ratio modulated with a step function from 0 to 3.0, 12 ME cycles of 400 s duration. In panels (c) and (d), error bars along the y‐axes correspond to the standard error of the mean determined from three independent test repetitions with selected catalysts, while error bars along the x‐axis represent the standard error of the mean Cu interparticle spacing, after propagation of the standard error of the mean Cu nanoparticle size as derived from the analysis of in situ EXAFS results. Dashed lines are added to guide the eye.

A swift development of surface carbonates was detected at the onset of each modulation–excitation cycle, i.e., for H₂:CO₂→0. Formates emerged subsequently, following the addition of hydrogen to the feed. Finally, methoxy species were detected towards the end of the cycle. This cyclical sequence of surface speciation agrees well with the extended mechanistic proposal which considers a sequential hydrogenation of surface carbonates to formates and further to methoxy intermediates, along the methanol synthesis path. Supporting this inference, methanol production was detected at the outlet of the IR cell synchronously to the development of spectroscopy fingerprints for surface H₃CO* and the disappearance of the corresponding DRIFT signals for surface HCOO*, respectively, during the modulation–excitation cycles (Figure S15).

Further, kinetic information was retrieved by analysing the ME‐DRIFTS data in the phase domain (Figure 3c). The in‐phase angle (${\mathrm{\Phi }}^{\mathbf{PSD}}$) represents the phase shift between the modulation signal and the surface species' response, providing kinetic insight into the dynamics of the latter. Carbonates showed an in‐phase angle ${\mathrm{\Phi }}_{\mathbf{OCOO}\ast }^{\mathbf{PSD}}$= 300 ± 10°, which remained constant over the entire range of D _NP‐NP. Formate intermediates showed essentially the same in‐phase angle as carbonates for interparticle spacings of ca. 15 nm or shorter. However, at longer D _NP‐NP, ${\mathrm{\Phi }}_{\mathbf{HCOO}\ast }^{\mathbf{PSD}}$ decreased conspicuously with D _NP‐NP, down to 166° for a D _NP‐NP of 22.4 nm. The absence of a phase lag between surface carbonates and formates at short interparticle distances indicates a coupled kinetics, wherein the interconversion of these species, i.e., carbonate hydrogenation to formate, proceeds with fast intrinsic rates and it is therefore expected to be kinetically irrelevant for methanol synthesis. In contrast, the progressive development of a phase lag with increasing D _NP‐NP strongly suggests a sequential conversion, wherein the degree of kinetic control for the carbonate to formate conversion step increases beyond the threshold ca. 15 nm interparticle spacing.

Methoxy intermediates showed the lowest in‐phase angle of all three surface species, indicative of the slowest kinetic response to the periodic modulation. Moreover, the evolution of ${\mathrm{\Phi }}_{{\mathbf{H}}_3\mathbf{CO}\ast }^{\mathbf{PSD}}$ as a function of D _NP‐NP qualitatively resembled that of formates, plateauing up to D _NP‐NP∼15 nm before decreasing with further interparticle spacing. To assess the degree of rate control of the formate hydrogenation step, relative to the overall carbonate to methoxy conversion, the following parameter RC_HCOO* has been defined as:

$$R{C}_{HCOO\ast }=\frac{\mathrm{\Delta }{\mathrm{\Phi }}_{\mathbf{HCOO}\ast conversion}^{\mathbf{PSD}}}{\mathrm{\Delta }{\mathrm{\Phi }}_{\mathbf{Overall}conversion}^{\mathbf{PSD}}}=\frac{{\mathrm{\Phi }}_{\mathbf{HCOO}\ast }^{\mathbf{PSD}}-{\mathrm{\Phi }}_{\mathbf{H}\mathbf{3}\mathbf{CO}\ast }^{\mathbf{PSD}}}{{\mathrm{\Phi }}_{\mathbf{OCOO}\ast }^{\mathbf{PSD}}-{\mathrm{\Phi }}_{\mathbf{H}\mathbf{3}\mathbf{CO}\ast }^{\mathbf{PSD}}}$$ (eq. 1)

As shown in Figure 3d, RC_HCOO* took values > 0.95 for D _NP‐NP ≤ 15 nm. These results provide experimental evidence supporting the hypothesis that the reaction step of HCOO* hydrogenation to H₃CO* exhibits a comparatively high degree of rate control, in line with earlier proposals based on first‐principles and microkinetic analyses.^{[ 21} , ^{32 ]} However, RC_HCOO* decreased progressively to 0.4 upon further increasing interparticle spacing to 22.4 nm. Therefore, while the degree of rate control for formate hydrogenation to methoxy is inferred to be significant over the entire range of interparticle spacing examined, it appears to decrease, indicating that carbonate hydrogenation becomes also kinetically relevant, as the Cu nanoparticles are spaced beyond the 15 nm threshold. A change in the relative value for the degree of rate control for different elementary conversion steps, involving intermediates of different enthalpic contributions, is expected to entail a change in the overall apparent activation energy (E _app).^{[ 32 ]} Therefore, the fact that the evolution of RC_HCOO* with D _NP‐NP was mirrored by the evolution of E _app for methanol formation, which remained essentially constant at 53 ± 3 kJ mol⁻¹ for D _NP‐NP ≤15 nm, then surged progressively up to 69 ± 4 kJ mol⁻¹ with further increasing interparticle spacing (Figure 3d), provides support for a change in the rate control contributions at this specific average interparticle spacing.

Given that the spectroscopic fingerprints for HCOO* and H₃CO* intermediates on Cu and ZrO₂ surfaces, respectively, could be resolved in the υ(CH) spectral regime, their intrinsic dynamics were probed with ME‐DRIFTS. Carbonate species do not form on Cu, therefore, the difference in in‐phase angle for formate and methoxy species has been used for studying the kinetics of the formate‐to‐methoxy hydrogenation step. Compared to reaction intermediates on the ZrO₂ support, the IR bands associated to intermediates on the metallic Cu surface showed a lower ${\mathrm{\Phi }}_{\mathbf{HCOO}\ast }^{\mathbf{PSD}}-{\mathrm{\Phi }}_{{\mathbf{H}}_3\mathbf{CO}\ast }^{\mathbf{PSD}}$ in‐phase angle difference of ca. 18° which remained essentially invariable as a function of the Cu interparticle spacing in the studied range (Figure S16). This indicates a comparatively swifter dynamic response (faster reactivity) of intermediates on the Cu surface, in line with earlier findings that reaction intermediates on the metallic Cu surface are turned over faster.^{[ 48 ]}

Moreover, the fact that only the dynamics of intermediates detected on ZrO₂ shows a dependence on Cu particle spacing which matches that of the overall methanol formation rate strongly suggests that the latter intermediates have the largest contribution to the macroscopically determined methanol production. This observation is reinforced by the ca. 30‐fold lower methanol formation rate and ca. fourfold lower methanol selectivity (17%) attained with the Cu/SiO₂ catalyst, which lacks the ability to stabilize reaction intermediates on a Lewis acidic oxide component.

Therefore, further analysis concentrated on the ZrO₂‐stabilized reaction intermediates. In addition to the characteristic period and phase shift, the amplitude of the PSD response, which reflects relative changes in the surface concentration of the responsive species, was examined as a function of D _NP‐NP (Figure S17). As shown in Figure 4, surface formates exhibited maximum concentration changes per ME cycle when the Cu interparticle spacing was approximately 15 nm. Both shorter and, particularly, longer interparticle spacings resulted in reduced cyclic concentration changes. Consistent with the evolution of the methanol formation rate (vide supra, Figure 2), these results indicate that the optimal dynamic engagement of HCOO*‐ZrO₂ intermediates in the reaction occurs when the perimeters of neighbouring Cu nanoparticles are spaced about 15 nm apart.

Figure 4.

Amplitude of the phase‐sensitive detection (PSD) response (phase‐angle) per ME cycle in the studied range of the Cu interparticle spacing for formate species on ZrO₂ (bd‐υ_s(CO₂ ⁻),1371 cm⁻¹). Experimental ME‐DRIFTS conditions: T = 493 K, P = 0.5 MPa, WHSV = 2.1–49.0·10³ g_CO2 g_Cu ⁻¹ h⁻¹; H₂:CO₂ molar ratio modulated with a step function from 0 to 3.0, 12 ME cycles of 400 s duration.

The results above suggest the existence of a catalysis‐relevant periphery (CRP) around each supported Cu nanoparticle, as schematically sketched in Figure 5. The CRP may be defined as the area around the metal‐oxide interface where kinetically significant surface intermediates are dynamically turned over with a relevant frequency. For the herein studied Cu/ZrO₂ system, a surface‐to‐surface interparticle distance of ca. 15 nm maximizes performance, indicating that the width of the individual CRP is ca. 7.5 nm, or approximately nine m‐ZrO₂ surface unit cells, at the optimal Cu nanoparticle spacing.

Figure 5.

Schematic qualitative illustration showing how variations in the spacing between next‐nearest‐neighbor (NNN) Cu nanoparticles is proposed to influence the spatial arrangement of their catalytically relevant Cu‐ZrO₂ periphery (CRP) areas, with implications for the methanol production rate during CO₂ hydrogenation with Cu/ZrO₂ supported catalysts.

The above optimal interparticle spacing is expected to maximize the global area of the CRP, thereby optimizing peripheral surface site utilization. Closer nanoparticle proximity results in an overlap of individual CRP for next nearest Cu nanoparticles, thus a submaximal surface density of HCOO* intermediates stabilized on the functional surface of the ZrO₂ support with sufficiently fast turnover dynamics. For such short Cu interparticle distances, our results additionally anticipate a maximum methanol selectivity and minimum apparent activation energy for methanol synthesis (approx. 45–60 kJ mol⁻¹), corresponding to the case of formate hydrogenation holding a high degree of rate control. These have indeed been observed for technically relevant Cu‐ZnO‐Al₂O₃ (and Cu‐ZrO₂) catalysts, where high copper loadings (50–70  wt%) enforce particularly short Cu interparticle spacings.^{[ 49} , ⁵⁰ , ^{51 ]} On the contrary, particle spacings beyond the optimal value are inferred to result in increasingly higher surface coverages of stagnant (non‐responsive) spectator adsorbates on the ZrO₂ carrier, and thus an increasingly higher contribution of the rather unselective behavior of metallic sites on the surface of the Cu component to the overall catalyst performance. A comparison of the catalytic performance with those reported in state‐of‐the‐art studies highlights the critical role of optimizing Cu interparticle spacing in achieving exceptionally high specific methanol production rates (Figure S18).

Conclusion

The performance of Cu/ZrO₂ catalysts in methanol synthesis via CO₂ hydrogenation strongly depends on the nanoscale spacing between supported Cu nanoparticles. The Cu‐specific methanol formation rate follows a volcano trend, peaking at an interparticle distance of ca. 15 nm. Methanol selectivity is also maximized at and below this optimal interparticle spacing. Operando modulation–excitation DRIFT spectroscopy reveals that the surface fraction of formate (HCOO*) reaction intermediates stabilized on the Lewis acidic Zr(IV) sites of the oxide support which show turn over dynamics is maximized at the optimal Cu interparticle spacing. Particularly for more distant Cu particles, the contribution from barely reactive (spectator) formate species increases significantly. The optimal interparticle spacing marks a shift in kinetic control, reflected by changes in the apparent activation energy for methanol synthesis. The results suggest the existence of a catalysis‐relevant periphery (CRP) on the ZrO₂ support, around each Cu nanoparticle, where kinetically significant formate intermediates, stabilized on Lewis acidic centers, are dynamically turned over at a relevant frequency. For optimally spaced Cu nanoparticles, this CRP is approximately 8 nm wide (ca. nine ZrO₂ unit cells). The dynamics of formate intermediates at the CRP mirrors the spatial trends of the overall methanol formation rate, while those on the metallic Cu surface show conversion dynamics unaffected by particle spacing, supporting the fact that the former are a major contribution to the overall methanol synthesis rate. The results uncover the importance of spatial factors, such as interparticle spacing, for the rational design of advanced supported Cu catalysts for methanol synthesis, and prospectively a wider array of catalytic processes exhibiting marked promotional effects at functional metal‐oxide interfaces.

Conflict of Interests

The authors declare no conflict of interest.

Supporting information

Supporting Information

López‐Luque I., Hack J., Ródenas T., Henao W., Mundet B., Patil P., Pichler C. M., Marini C., Agostini G., Meier D. M., Prieto G., Angew. Chem. Int. Ed.. 2025, 64, e202420126. 10.1002/anie.202420126

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.