Single-atom Cu catalyst in a zirconium MOF for highly selective hydrogenation of unsaturated aldehydes to unsaturated alcohols

Abstract

Achieving nearly 100% selectivity in the hydrogenation of α,β-unsaturated aldehydes to unsaturated alcohols, while maintaining complete conversion, remains a significant challenge in heterogeneous catalysis, especially with non-noble metals. Herein, we report an efficient single-atom Cu catalyst (Cu₁@MOF-808) constructed by covalently anchoring Cu atoms onto defective sites of Zr oxide clusters in MOF-808, where Cu species are chelated and stabilized via coordination with oxygen atoms from dangling –OH/–OH₂ groups. The unique configuration enables preferential adsorption of the C C bond at isolated Cu sites and the capture of H* by neighboring coordinated oxygen. This synergistic, spatially dispersed adsorption mechanism drives highly selective C O bond hydrogenation. The resultant Cu₁@MOF-808 catalyst achieves nearly 100% selectivity toward cinnamyl alcohol (COL) with full conversion of cinnamaldehyde (CAL). Mechanistic studies revealed that the high oxidation state of single-atom Cu sites, coupled with electron-rich oxygen coordination and Lewis acid sites, enables spatial separation between reactant adsorption and proton transfer, thereby effectively regulating the product desorption and optimizing the reaction pathway.

The precise spatial confinement of MOF and atomically dispersed Cu species, coupled with Lewis acid sites, enabled the single-atom Cu₁@MOF-808 catalyst to selectively promote C O bond hydrogenation while suppressing C C bond hydrogenation.

Introduction

Chemoselective hydrogenation of α,β-unsaturated aldehydes to unsaturated alcohols is of fundamental and practical significance in both scientific research and industry applications, as these compounds are valuable and versatile intermediates in the production of pharmaceuticals, cosmetics, resins, fragrances, and dyes.^1–5 Nevertheless, the intrinsic thermodynamic and kinetic favorability of C C hydrogenation over C O reduction poses a persistent selectivity challenge for heterogeneous catalysts (Scheme 1A).^6–12 Conventional supported metal catalysts (Pt, Au, Pd, Ni, Co, etc.) have been employed for this transformation,^13–23 but they often suffered from low selectivity due to the preferential adsorption and activation of the C C bond on extended metal surfaces.

Scheme 1. Representative catalytic systems for the hydrogenation of α,β-unsaturated aldehydes to unsaturated alcohols. (A) Monometal-supported catalysts. (B) Modified catalysts, including bimetallic catalysts, sterically hindered catalysts, encapsulated catalysts, and confined catalysts. (C) This work: single-atom Cu catalyst, Cu₁@MOF-808.

To circumvent this limitation, two classical strategies have been developed: modification of electronic properties by introducing second metals^24–33 and construction of spatial steric hindrance^34–38 or a confinement effect^6,39–41 (Scheme 1B). In addition, non-contact hydrogenation⁴² and an oscillation potential strategy⁴³ have also been employed to improve the hydrogenation selectivity. While these approaches have achieved some success, they are largely empirical and offer limited capacity for the precise, bottom-up design of the catalytic environment. More fundamentally, these strategies still operate within the paradigm of modifying metal-based active sites, rather than rationally engineering the entire catalytic microenvironment.

Single-atom catalysts (SACs), with their well-defined coordination structures and maximum atom utilization efficiency, provide an ideal model system for probing structure–selectivity relationships.^44,45 However, the local environment of the isolated metal centers is inherently constrained by the support's surface chemistry, making it difficult to independently tune the microenvironment without altering the metal's intrinsic electronic state. We hypothesized that metal–organic frameworks (MOFs) could offer a transformative solution to this problem. Unlike conventional supports, MOFs have chemically tunable pore surfaces, adjustable coordination geometries, and programmable functionality at the molecular level.^46–50 Such construction, which leverages the highly tunable nature of MOF matrices for atomic-level precision in regulating the coordination environment and electronic states of active metal species, may offer a pathway to enhance the selectivity of the targeted catalytic reactions toward desired products.

Herein, we demonstrate a MOF-based microenvironment engineering strategy to achieve highly chemoselective hydrogenation of α,β-unsaturated aldehydes. The isolated Cu atoms were covalently attached to the defect sites of the Zr clusters in MOF-808, forming abundant Zr–O–Cu–O–Zr structures in the frameworks, which facilitated the exposure of catalytic sites (Scheme 1C). The precise spatial confinement of MOF and atomically dispersed positively charged Cu species, as well as electron-rich coordination O atoms, coupled with the Lewis acid sites, enabled the single-atom Cu₁@MOF-808 catalyst to selectively promote C O bond hydrogenation while completely suppressing C C bond hydrogenation. Consequently, the catalyst achieved 99.7% selectivity to cinnamyl alcohol (COL) at full conversion of cinnamaldehyde (CAL), outperforming other non-precious metal-based catalysts and rivaling Pt-based systems. The Cu₁@MOF-808 catalyst showed no significant decay in terms of activity and selectivity over five consecutive cycling tests, indicating the robust nature of the catalyst. Moreover, it effectively catalyzed the hydrogenation of diverse α,β-unsaturated aldehydes to their corresponding unsaturated alcohols with satisfactory selectivity. This work demonstrates MOF-enabled microenvironment engineering as a general and rational design strategy for selectivity control in single-atom catalysis.

Results

Synthesis and characterization of catalysts

MOF-808 was synthesized by a modified solvothermal method, and it possesses a 6,3-connected 3D structure with spn topology, featuring a long-range ordered framework comprising both tetrahedral and adamantine-type cages (Fig. 1A and S1).⁵¹ Cu chelated MOF-808 (Cu₁@MOF-808) was prepared by a facile and scalable post-modification method. The activated MOF-808 powder was dispersed in DMF solution containing a certain amount of copper salt under constant heating, during which the Cu ions were coordinated to the oxygen atoms of dangling –OH or –OH₂ ligands on Zr₆ clusters of the MOF. The Cu species were precisely chelated and stabilized by the nodes of MOF-808 (Fig. 1B). Further activation under dynamic vacuum was performed to remove guest molecules and generate active Cu-anchored MOF-808. Finally, reduction treatment in a H₂/Ar mixed atmosphere was conducted, resulting in the creation of reduced Cu₁@MOF-808. The accurate Cu loading in Cu₁@MOF-808 was calculated to be 1.37 wt%, which was determined by inductively coupled plasma optical emission spectrometry (ICP-OES, Table S1).

Fig. 1. Scheme for the synthesis of Cu₁@MOF-808 and structural characterization. (A and B) Structures of MOF-808 and the illustration of Cu species metalated in MOF-808 for Cu₁@MOF-808. (C–F) HAADF-STEM images of reduced Cu₁@MOF-808, and (G–J) corresponding EDS elemental mapping images of C, O, Zr, and Cu in a single MOF-808 crystal. The scale bar is 500 nm in (C), 300 nm in (D), 10 nm in (E), and 350 nm in (F–J).

MOF-808 displayed an octahedral shape with an average particle size of approximately 700 nm (Fig. S2), as determined by scanning electron microscopy (SEM). The morphology of reduced Cu₁@MOF-808 crystals could be maintained after chelation of Cu species and reduction treatment (Fig. S3 and S4), without obvious Cu nanoparticles or clusters on the surface of MOF, implying the even distribution of Cu species. Moreover, high-angle annular dark-field scanning TEM (HAADF-STEM) was used to analyze the structures of reduced Cu₁@MOF-808. No apparent lattice fringes assigned to metallic Cu or CuO_x species were present, further suggesting the encapsulation of Cu atoms in the pores of MOF-808 (Fig. 1C–E). Additionally, the elemental mapping of Cu matched well with C, O, and Zr elements, suggesting the even distribution of Cu species throughout the entire single MOF crystal (Fig. 1F–J), as evidenced by energy dispersive spectroscopy (EDS).

Cu₁@MOF-808 showed similar sharp powder X-ray diffractions (PXRD) compared with the as-synthesized MOF-808 and the simulated one, indicating the high crystallinity of Cu₁@MOF-808 (Fig. 2A). Moreover, no characteristic reflections assigned to Cu species were observed, suggesting the high dispersion of Cu species over the MOF support. Cu₁@MOF-808 displayed high thermal stability up to 290 °C (Fig. S5), as evidenced by thermogravimetric analysis (TGA). The N₂ adsorption–desorption isotherms revealed a Brunauer–Emmett–Teller (BET) surface area of 1586 m² g⁻¹ for Cu₁@MOF-808, which was lower than that of MOF-808 (2158 m² g⁻¹) as predicted (Fig. S6a). In addition, the pore size distribution revealed a small reduction after Cu chelation from 1.53 nm to 1.26 nm (Fig. S6b). X-ray photoelectron spectroscopy (XPS) was performed to test the chemical valence state of Cu₁@MOF-808 (Fig. 2B and S7).The binding energies centered at 934.6 eV (Cu 2p_3/2) and 953.6 eV (Cu 2p_1/2), along with two distinct satellites, were visible for fresh Cu₁@MOF-808 (Fig. 2B). These characteristic satellites belonged to Cu²⁺ ions.^52,53 Meanwhile, Cu 2p_3/2 and Cu 2p_1/2 at binding energies of 932.8 and 952.8 eV were present in fresh Cu₁@MOF-808, corresponding to Cu¹⁺ or Cu⁰ species. However, the existence of Cu⁰ was ruled out by Auger electron spectroscopy (AES) of Cu LMM at a binding energy of 572.6 eV (Fig. 2C), indicating the existence of Cu¹⁺ in fresh Cu₁@MOF-808.^53–55 The Cu¹⁺ species might result from partial reduction of Cu²⁺ during the synthetic process due to the reducibility of the DMF solvent.⁵⁶ These results manifested the co-presence of both Cu¹⁺ and Cu²⁺ in fresh Cu₁@MOF-808. After reduction treatment, only Cu¹⁺ species were present in reduced Cu₁@MOF-808, which was confirmed by the combination of the disappearance of characteristic satellites of Cu²⁺ and the observation of a typical Cu¹⁺ signal in Cu LMM spectra (Fig. 2B and C).

Fig. 2. Electronic structure characterization of Cu₁@MOF-808. (A) PXRD patterns of reduced Cu₁@MOF-808, MOF-808, and the simulated one. (B) Cu 2p XPS and (C) Auger electron spectra of Cu LMM for fresh and reduced Cu₁@MOF-808. (D) XANES and (E) EXAFS spectra of reduced Cu₁@MOF-808, Cu foil, Cu₂O, and CuO at the Cu K-edge. (F) R- and (G) k-space of Fourier transform magnitudes of Cu K-edge EXAFS of reduced Cu₁@MOF-808. (H) Wavelet transform of the Cu K-edge EXAFS spectra of reduced Cu₁@MOF-808, Cu foil, Cu₂O, and CuO, respectively.

Furthermore, X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) were employed to identify the electronic structure and local coordination environment of Cu species in reduced Cu₁@MOF-808 (Fig. 2). The normalized XANES spectra of absorption edges for Cu₁@MOF-808 shifted to higher energy as compared with Cu foil (Fig. 2D and S8), revealing the positively charged state of Cu atoms. Moreover, the absorption edge of Cu species in Cu₁@MOF-808 was very close to that of the Cu₂O reference, implying that the average oxidation state of Cu atoms was around +1.^57,58 In the EXANES spectra of Cu₂O, CuO, and Cu foil, the peaks at ∼2.66, ∼2.42, and ∼2.23 Å were attributed to the Cu–O–Cu and Cu–Cu bonds (Fig. 2E and S9–S11).⁴⁸ Besides, the peaks at ∼1.47 and ∼1.53 Å were assigned to the Cu–O bond for Cu₂O and CuO,^58,59 respectively. In contrast, only a single peak located at approximately 1.50 Å was observed in Cu₁@MOF-808, which is attributed to Cu–O coordination (Fig. 2E). The absence of a Cu–Cu bond peak further confirmed the atomic dispersion of Cu species.⁵⁸ One prominent peak located at around 1.42 Å, a fingerprint peak attributed to Cu–O coordination,^58,59 was also found in the Fourier transformed EXAFS spectra in the R and K space in Cu₁@MOF-808 (Fig. 2F and G). Subsequently, the detailed atomic structures of Cu atoms were analyzed by fitting the EXAFS spectra (Table S2). The coordination number of the Cu–O bonding in the first coordination sphere was 2.15 ± 0.30 at a distance of 1.97 ± 0.01 Å, suggesting that one Cu atom was coordinated with two O atoms. To further elucidate the atomic dispersion of Cu species, the wavelet transform (WT) of the Cu K-edge EXAFS oscillation was analyzed (Fig. 2H). The WT contour plots of Cu₁@MOF-808 exhibited only one intensity maximum at approximately 5.2 Å⁻¹, which corresponds to the Cu–O coordination. In contrast to the reference samples (Cu foil, Cu₂O, and CuO), no intensity maxima associated with Cu–Cu or Cu–O–Cu contributions were observed.⁵⁹

Selective C O bond hydrogenation

The selective hydrogenation of α,β-unsaturated aldehydes to unsaturated alcohols was performed, and cinnamaldehyde (CAL) was used as the model chemical compound (Table 1). First, the hydrogenation of CAL was conducted in isopropanol solvent at 100 °C and 2 MPa H₂ pressure. Control experiments confirmed that no reaction occurred either in the absence of a catalyst or when using only MOF-808 as a catalyst (Table 1, entries 1 and 2). In marked contrast, the reduced Cu₁@MOF-808 catalyst could effectively catalyze the selective hydrogenation of CAL toward target cinnamyl alcohol (COL) with 99.6% selectivity at 71.6% conversion (Table 1, entry 3). When the reaction time was prolonged to 10 hours, the CAL conversion reached nearly 100% while preserving high selectivity to COL (99.7%) (Table 1, entry 4). For comparison, other catalysts (Pt₁@MOF-808, Pd₁@MOF-808, Cu₁@UO-66, and Cu/ZrO₂, which are carefully characterized, see the SI, Fig. S12–S22) were also systematically evaluated under identical reaction conditions. Pt₁@MOF-808 and Pd₁@MOF-808 displayed full CAL conversion, but no COL was obtained; instead, the over-hydrogenation product hydrocinnamyl alcohol (HCOL) was obtained (Table 1, entries 5 and 6). Moreover, neither Cu₁@UiO-66 nor Cu/ZrO₂ exhibited catalytic activity for this reaction (Table 1, entries 7 and 8). Notably, when the temperature was raised to 150 °C, Cu₁@UiO-66 gave 73.5% conversion of CAL with excellent selectivity toward COL (>99%, Table S3). In contrast, even at an elevated temperature of 200 °C, Cu/ZrO₂ showed ultralow activity (<3% conversion) and produced a mixture of COL, hydrocinnamaldehyde (HCAL), and HCOL with a selectivity of 51.4%, 34.8%, and 13.8% (Table S3), respectively.

Table 1. Selective hydrogenation of CAL to COL over various catalysts^a.

Entry	Catalyst	Conv. (%)	Selectivity (%)
			COL	HCAL	HCOL
1	Blank	n.d.c	n.d.	n.d.	n.d.
2	MOF-808	n.d.	n.d.	n.d.	n.d.
3	Cu1@MOF-808	71.6	99.6	n.d.	n.d.
4	Cu1@MOF-808b	99.9	99.7	n.d.	n.d.
5	Pt1@MOF-808	99.9	n.d.	n.d.	>99
6	Pd1@MOF-808	99.9	n.d.	n.d.	>99
7	Cu1@UiO-66	n.d.	n.d.	n.d.	n.d.
8	Cu/ZrO2	n.d.	n.d.	n.d.	n.d.

^aReaction conditions: CAL (1 mmol), catalyst (20 mg), solvent (IPA, 10 mL), reaction temperature (100 °C), H₂ pressure (2 MPa), reaction time (4 h), stirring speed (600 rpm).

^bReaction time was 10 h.

^cNot detected.

Furthermore, the effects of pressure and temperature on CAL hydrogenation over the Cu₁@MOF-808 catalyst were investigated. The catalytic performance showed minimal dependence on hydrogen pressure (Table S4), while the reaction temperature exerted a pronounced effect on both activity and selectivity (Table S5). At 80 °C, the reaction achieved high COL selectivity (>99%) with only 27.5% CAL conversion. Remarkably, when the temperature increased to 100 °C, the conversion increased to nearly 100% without compromising the product selectivity (Table S5). Upon further elevation of temperature, a gradual decline in COL selectivity was observed, accompanied by the formation of side products, ascribed to the occurrence of a condensation reaction at high temperature,⁶⁰ as confirmed by GC-MS (Fig. S26). The solvent effect was studied over the Cu₁@MOF-808 catalyst (Fig. S27). In comparison with nonpolar solvent (n-hexane) and aprotic polar solvent (tetrahydrofuran), protic solvent (isopropyl alcohol) was more beneficial for the CAL hydrogenation reaction. Under the optimized conditions, Cu₁@MOF-808 delivered complete CAL conversion, 99.7% COL selectivity, and a superior productivity of 23.2 h⁻¹, outperforming non-precious metals by an order of magnitude (Table S6) and rivaling precious metal-based catalysts (Table S7).

The evolution of reactant and product distributions over Cu₁@MOF-808 as a function of time was investigated. The CAL conversion increased gradually over time, reaching full conversion after 10 h. Throughout the reaction, COL remained the sole product, with no observable changes in selectivity even after prolonged reaction times (Fig. S28). These results demonstrated the excellent selectivity of Cu₁@MOF-808 toward C O bond hydrogenation, while effectively suppressing the hydrogenation of the C C bond. Moreover, the catalyst was reusable for at least five consecutive cycles without significant decay in activity and selectivity (Fig. S29), implying outstanding stability and durability. After cycling, PXRD confirmed the well-retained crystallinity and structure of MOF-808 (Fig. S30), and HAADF-STEM showed no observable Cu clusters or nanoparticles (Fig. S31). A negligible leaching of Cu species was detected, as evidenced by ICP-OES analysis (Table S8). XPS revealed new features assigned to Cu²⁺ ions (Fig. S32), implying partial oxidation of Cu¹⁺ to Cu²⁺ during cycling. All these results demonstrated the robustness of the Cu₁@MOF-808 catalyst for selective C O bond hydrogenation, an industrially desirable feature.

To further demonstrate the universality of Cu₁@MOF-808, a series of α,β-unsaturated aldehyde substrates were investigated (Table 2). As anticipated, cinnamyl aldehyde and its various phenyl derivatives, such as p-methylcinnamaldehyde, p-fluorcinnamaldehyde, and p-bromocinnamaldehyde, were all converted to their corresponding unsaturated alcohols and achieved high selectivity (97–99%, Table 2, entries 1–4). Furthermore, when the heterocyclic substrate 3-(2-furyl)acrolein was employed, the corresponding unsaturated alcohol was obtained with 92% selectivity (Table 2, entry 5). In particular, aldehydes containing a benzoyl group, for instance benzalacetophenone, could be transformed into the corresponding unsaturated alcohols (Table 2, entry 6). It is generally difficult to achieve linear selective synthesis of unsaturated alcohols from chain unsaturated aldehydes due to the fact that they lack conjugated functional groups on their C C bond, resulting in a negligible effect of spatial steric hindrance, which makes it easy for C C bond hydrogenation. In our catalytic system, the hydrogenation of the C C bond was restrained and the target linear unsaturated alcohols were achieved in high selectivity (Table 2, entries 7–9). All these results suggested that Cu₁@MOF-808 exhibited excellent tolerance to a variety of substrates.

Table 2. Selective hydrogenation of various α,β-unsaturated aldehydes or ketones over Cu₁@MOF-808^a.

Entry	Substrate	Product	Con. (%)	Sel. (%)	bCB
1			>99	>99	99
2			>99	98	98
3			>99	99	98
4			>99	97	96
5			>99	92	92
6			>99	89	89
7			>99	99	98
8			>99	95	95
9			>99	96	96

^aReaction conditions: substrate (1 mmol), catalyst (20 mg), solvent (IPA, 10 mL), pressure (H₂, 2 MPa), stirring speed (600 rpm), 100 °C and 10 h.

^bCB: carbon balance.

Reaction mechanism investigation

To gain more insight into the reaction mechanism, several control experiments were conducted. To rule out the possibility that the reaction pathway proceeds via a catalytic transfer hydrogenation (CTH) route, we conducted a comparative experiment under an inert atmosphere. The reaction did not proceed under an N₂ atmosphere (Table S9), confirming that the hydrogen originated from H₂. When HCAL was used as the substrate, HCOL was identified as the main product over both Cu₁@MOF-808 and Pt₁@MOF-808 catalysts (Fig. 3A and Table S10). However, when COL was employed as the reactant, Pt₁@MOF-808 hydrogenated the C C bond to form HCOL, whereas Cu₁@MOF-808 completely suppressed this reaction (Fig. 3A and Table S11). These results demonstrated that Cu₁@MOF-808 exhibited excellent hydrogenation selectivity for the C O bond while fully inhibiting C C bond hydrogenation.

Fig. 3. Mechanistic investigations. (A) Control experiments of selective hydrogenation of HCAL or COL over Pt₁@MOF-808 and Cu₁@MOF-808 catalysts. Reaction conditions: substrate (1 mmol), catalyst (20 mg), solvent (IPA, 10 mL), pressure (H₂, 2 MPa), stirring speed (600 rpm), 100 °C, and 3 h. (B) H₂-TPD spectra of Cu₁@MOF-808 and Pt₁@MOF-808 catalysts. (C) Py-IR spectra of Cu₁@MOF-808 and Pt₁@MOF-808 catalysts. (D) ATR-IR spectra of CAL adsorption on Cu₁@MOF-808 and Pt₁@MOF-808 catalysts.

Temperature programmed desorption (TPD) was performed to study the H₂ activation capacity of catalysts (Fig. 3B). The temperature for the TPD test increased from 50 to 250 °C at a rate of 5 °C min⁻¹ and was then held at 250 °C for 1 h. Three desorption peaks located at around 175, 230, and 250 °C were observed for Cu₁@MOF-808 and Cu₁@UiO-66, which may be attributed to the desorption of hydrogen spilled over onto the Zr-oxide cluster (Fig. S33).⁶¹ Note that the total H₂ desorption amounts of Cu₁@MOF-808 were higher than those of Cu₁@UiO-66, suggesting the higher hydrogenation ability of Cu₁@MOF-808, which was consistent with the high activity (Table 1). However, Pt₁@MOF-808 displayed much lower desorption peaks at about 122 °C and 198 °C, suggesting a weak interaction between the catalyst and H₂ molecules. This result demonstrated the higher activation ability for H₂ over Pt₁@MOF-808, which accounts for its highest activity. For comparison, Cu/ZrO₂ showed no noticeable H₂ desorption peaks, explaining its poor performance (Fig. S33).

Pyridine-IR spectroscopy was conducted to identify the acid type and the amount of catalysts. Generally, the bands located at around 1610/1450 and 1540 cm⁻¹ were assigned to the Lewis and Brønsted acid sites,^62,63 respectively. However, the bands in the range of 1400–1700 cm⁻¹ were difficult to distinguish from those of MOFs with strong peaks because of the benzene ring and carboxylate vibrations. As a result, we focused on the signal peaks in the interval of 1000 to 1100 cm⁻¹.⁶⁴ Three characteristic peaks at around 1069, 1043, and 1013 cm⁻¹ were observed (Fig. 3C), corresponding to the ν₁₈, ν₁₂, and ν₁ ring models of pyridine adsorbed on Zr clusters, respectively, which acted as Lewis acid sites.⁶⁵ The Lewis acid sites would facilitate the adsorption and activation of the carbonyl group, thus enhancing the reaction selectivity.^11,14 Cu₁@MOF-808 exhibited the highest Lewis acid amount of 140.7 µmol g⁻¹ at 200 °C, which was higher than those of Pt₁@MOF-808 (122.5 µmol g⁻¹), Cu₁@UiO-66 (79.8 µmol g⁻¹), and Cu/ZrO₂ (11.9 µmol g⁻¹) (Fig. S34 and Table S12). This result was generally consistent with the catalytic performance, demonstrating that Lewis acid sites played a key role in the selective hydrogenation of the C O bond.

Furthermore, attenuated total reflectance infrared (ATR-IR) spectra were recorded to evaluate the adsorption of CAL on the surfaces of different catalysts (Fig. 3D). The pure CAL displayed two characteristic vibration peaks at 1677 cm⁻¹ (C O) and 1625 cm⁻¹ (C C).⁶⁶ Upon adsorption on Cu₁@MOF-808 and Pt₁@MOF-808, both the C O and C C peaks exhibited a slight but consistent blueshift (1680 cm⁻¹vs. 1677 cm⁻¹; 1627 cm⁻¹vs. 1625 cm⁻¹, respectively). These results indicated that CAL adsorbed on Cu₁@MOF-808 and Pt₁@MOF-808 through both the C C and C O bonds. By correlating the distinct hydrogenation products over Cu₁@MOF-808 and Pt₁@MOF-808 with ATR-IR analysis, we proposed that the reaction pathway depends on factors beyond just CAL's adsorption configuration. To elucidate the origin of Cu₁@MOF-808's superior performance and underlying reaction mechanism, we conducted density functional theory (DFT) calculations.

We first established stable structural models of Cu and Pt atoms anchored on MOF-808 (Fig. 4A and S35). The intrinsic Lewis acidity of the MOF-808 node originated from its Zr₆–O clusters, creating an electron-withdrawing coordination environment for the anchored metal atoms. Atomic charge distribution analysis revealed distinct oxidation states induced by oxygen coordination, with Cu and Pt losing 0.59 and 0.26|e| to the coordinated oxygen atoms (Fig. 4B), respectively. The substantially greater electron transfer from Cu (0.59 vs. 0.26|e|) indicated that the Lewis-acid-driven metal–support interaction is markedly stronger for Cu sites, resulting in a higher-valence Cu center with depleted d-electron density. Guided by these electronic structure differences, we systemically evaluated competitive adsorption and hydrogen atom capture at these single-atom centers—key factors governing the selective hydrogenation mechanism of CAL. For both Cu and Pt sites, the adsorption of C C, C O, and H atoms was considered. The calculated adsorption energies followed the order C C > C O > H (Fig. 4C), suggesting that both C O and C C bonds are preferentially activated on Cu₁@MOF-808 and Pt₁@MOF-808. Bader charge analysis further confirmed that greater charge transfer promotes stronger adsorption (Fig. S36). Critically, the Lewis-acid-induced electron depletion at the Cu site led to weaker C C adsorption relative to Pt, because the diminished d-electron density at Cu reduced the back-donation from the metal d-orbitals to the π* antibonding orbitals of C C. This electronic effect is the microscopic origin of the experimentally observed suppression of C C hydrogenation on Cu₁@MOF-808.

Fig. 4. Theoretical calculations and possible reaction paths for hydrogenation. (A) Side view of the optimized TM₁@MOF-808 structure. (B) Atomic charge coloring chart of Cu₁@MOF-808 and Pt₁@MOF-808. (C) Initial competitive elementary reactions on the active site. (D) Capture of active H atoms at different sites. (E) Schematic diagram of CAL hydrogenation on the TM₁@MOF-808 catalyst. (F) Relative energy diagram of two paths for selective hydrogenation of CAL to COL, and (G) the structures of the intermediate. “*” denotes the hydrogenation site of the intermediate or the active site on the catalyst surface. The added H* in each step is represented by blue atoms.

Further analysis addressed the origin of H atoms in CAL hydrogenation. On Cu₁@MOF-808, H* preferentially adsorbs on the Lewis-basic coordinated oxygen (adjacent to the Lewis-acidic Zr center), benefiting from its high electron density (−1.00|e|). This creates a spatial separation between CAL adsorbed on Cu and H* on oxygen. On Pt₁@MOF-808, the lower oxidation state of Pt enables a co-adsorption configuration of C C and H*, with an adsorption energy as low as −3.05 eV (Fig. 4D and S37). These distinct behaviors reflect different underlying mechanisms: in Cu₁@MOF-808, the Lewis acidity of the node enforces substantial electron transfer between Cu and coordinated oxygen, creating a synergistic bifunctional environment where Cu captures CAL and the oxygen captures H in the initial stage. In contrast, the electron-rich Pt site in Pt₁@MOF-808 retains its metallic character and higher electron density, thus acting as an independent active center.

Based on experimental observations and product distribution analysis, two possible reaction pathways were proposed. Path 1 involves the preferential attack of H* on the C C bond, generating HCAL as an intermediate, which undergoes complete hydrogenation to yield HCOL. Path 2 proceeds via selective addition of H* to the C O group to form COL as the primary product, followed by hydrogenation to HCOL (Fig. 4E). The energy profiles and intermediate structures of CAL hydrogenation on the two catalytic models are presented in Fig. 4F and G. Building upon the established preferential activation of the C C bond, the C O bond and the H* captured by coordinated oxygen adopt a vertically aligned configuration over Cu₁@MOF-808 (inset in Fig. 4D), requiring only a 0.61 eV migration barrier. Subsequent attack on the C atom exhibited a comparable barrier of 0.63 eV, with a maximum adsorption energy change of 0.47 eV for these two elementary steps. In contrast, initial hydrogenation of the C C bond encountered a considerable barrier of 1.12 eV, making this pathway unfavorable. This selective low-barrier process benefits from steric effects induced by the coordination structure that activates H species, thereby facilitating target product formation. For Pt₁@MOF-808, the first elementary step of C O bond hydrogenation showed a strong adsorption energy (−3.36 eV); however, further hydrogenation was precluded by a high energy barrier of 1.96 eV, rendering this pathway unviable. We then evaluated the feasibility of Path 1 for C C bond hydrogenation. The second protonation step during HCAL formation was endothermic by 1.18 eV with a 1.54 eV barrier and is considerably more favorable than Path 2. Consequently, HCAL was thermodynamically driven to further hydrogenate to HCOL on Pt₁@MOF-808. These findings aligned with the established hydrogenation selectivity of noble metals, where the combined effect of C C adsorption and H* activation on Pt surfaces favors saturated alcohol production. Therefore, comparative simulations identify three superior features of Cu₁@MOF-808: (1) a unique electronic structure with strongly interacting high-valence Cu sites and electron-rich coordinated oxygen, both originating from the Lewis acidity of the MOF-808 node; (2) dispersed adsorption of reactants and H* that lowers the initial protonation barrier and optimizes the reaction pathway; and (3) moderate adsorption strength that facilitates product desorption.

Conclusions

In summary, we successfully engineered atomically dispersed Cu species stabilized in MOF defects through Cu–O coordination, which effectively prevented their aggregation and leaching. The resulting catalyst achieved exceptional performance in the selective hydrogenation of cinnamaldehyde to cinnamyl alcohol, demonstrating 99.7% selectivity at complete conversion. Combined with experiments and theoretical calculations, the study revealed that the precisely controlled geometric and electronic structures of these isolated Cu sites and Lewis acid sites in the MOF are the key factors for the high selectivity and efficiency. The low Cu electron density and high O electron density precisely regulated the adsorption of both reactants and H* and the desorption capability of the product, ultimately improving the selectivity. The present defect-engineering strategy for stabilizing non-precious single-atom catalysts in MOFs offers a versatile platform for designing high-performance catalysts for a wide range of selective hydrogenation reactions.

Author contributions

Y. W., M. D., and A. Z. conceived and designed the project. Y. W. performed the initial discovery and optimization of the test conditions. Y. L. performed the DFT calculations. Y. W., Y. L., M. D., and A. Z. wrote the manuscript and all authors contributed to the final manuscript and the interpretation of the results.

Conflicts of interest

The authors declare no competing financial interests.

Data availability

The supporting data have been included in the supplementary information (SI). Supplementary information: experimental section, synthetic procedures, and characterization data (including N₂ adsorption–desorption, HR-TEM, and XPS) for catalysts; catalytic performance of CAL hydrogenation. See DOI: https://doi.org/10.1039/d6sc04923k.

All data generated and analyzed during the study are available from the corresponding authors on reasonable request.