Manipulating hydrogenation pathways enables economically viable electrocatalytic aldehyde-to-alcohol valorization

Significance

Electrocatalytic reduction (ECR) of biomass-derived aldehyde such as furfural is a promising pathway for chemicals manufacturing. However, insufficient understanding of the intricate ECR mechanisms and the lack of mechanism-guided catalyst design result in unsatisfactory production rates. Herein, integrating theoretical and isotopic kinetic analyses, we introduce an amorphous ceria support to Cu clusters to specially construct a H* transfer channel, thus switching a proton-coupled electron transfer mechanism to a desirable hydrogen-atom transfer one. The resulting catalyst achieves the economically viable electrosynthesis of a single alcohol product. This work offers deep insights into the significance of manipulating hydrogenation pathways in electrocatalytic valorizing biofeedstocks toward economically viable or even industrial-scale productivity.

Abstract

Electrocatalytic reduction (ECR) of furfural represents a sustainable route for biomass valorization. Unfortunately, traditional Cu-catalyzed ECR suffers from diversified product distribution and industrial-incompatible production rates, mainly caused by the intricate mechanism−performance relationship. Here, we manipulate hydrogenation pathways on Cu by introducing ceria as an auxiliary component, which enables the mechanism switching from proton-coupled electron transfer to electrochemical hydrogen-atom transfer (HAT) and thus high-speed furfural-to-furfuryl alcohol electroconversion. Theoretical and kinetic analyses show that oxygen-vacancy-rich ceria delivers an efficient formation−diffusion−hydrogenation chain of H* by diminishing H* adsorption. Spectroscopic characterizations indicate that Cu/ceria interfacial perimeter enriches the local furfural, synergistically lowering the barrier of the rate-determining HAT step across the perimeter. Our Cu/ceria catalyst realizes high-rate HAT-dominated ECR for electrosynthesis of single-product furfuryl alcohol, achieving a high production rate of 19.1 ± 0.4 mol h⁻¹ m⁻² and a Faradaic efficiency of 97 ± 1% at an economically viable partial current density of over 0.1 A cm⁻². Our results demonstrate a highly efficient route for biofeedstock valorization with enhanced techno-economic feasibility.

The dwindling fossil resources coupled with the global Sustainable Development Goals have promoted the blossoming of biorefineries worldwide (1–3). Catalytic valorization of renewable biofeedstocks offers a sustainable approach to chemicals manufacturing. Thereinto, a prime exploration is the valorization of lignocellulose-derived aldehydes such as furfural (FUR) with a substantial economic market (over US$600 million in 2023) (4–6). Its dominating downstream product is furfuryl alcohol (FA) with 65% of the market and the highest revenue share of 86% (US$473 million in 2023) (7, 8). The industrial manufacturing of FA is achieved via thermocatalytic hydrogenation employing a toxic copper chromite catalyst with elevated temperatures (140 °C to 300 °C) and pressurized H₂, which is energy-intensive and poses an explosion risk (5). Thus, an economical and safe valorization technique is urgently needed.

Renewable energy-powered electrocatalytic reduction (ECR) using water as the hydrogen source provides a sustainable alternative due to its compatibility with ambient temperatures and pressures (Fig. 1A) (9, 10). Nevertheless, current FUR-to-FA conversion via the Cu-catalyzed ECR process suffers from industrial-incompatible production rates of below 30 mA cm⁻² in most cases (SI Appendix, Table S1) (7, 11). According to the a priori techno-economic analysis (Fig. 1B and SI Appendix, Fig. S1), the lowest economically viable current density for the FUR-to-FA conversion is ca. 75 mA cm⁻² [calculated with an optimistic Faradaic efficiency (FE_FA, 90%) and a conservative global average levelized cost of onshore wind-powered electricity (6 ~ 7 cents kW h⁻¹)] (12, 13), which is determined as the profitable threshold for the ECR of FUR. The existing dilemma mainly stems from the inadequate understanding of the intricate ECR mechanisms and the lack of mechanism-guided design of high-performance electrocatalysts.

Fig. 1.

Valorization process, techno-economic analysis, and ECR mechanisms of FUR-to-FA conversion. (A) Schematic illustrating of electrocatalytic biomass valorization for FA production. (B) Plant-gate levelized cost for the FUR-to-FA conversion as a function of electricity cost and operation current density. (C) Introduction of an auxiliary component to steer the H behaviors and convert the ECR from PCET-domination to HAT-domination. Gold, brown, pink, gray, cyan, and purple balls represent Cu⁰, Cu^δ+, O, H, H*, and Ce atoms, respectively.

The ECR pathway is widely recognized to have an impact on product selectivity and is also critical for the ECR rate though it is largely unexplored. Cu-catalyzed ECR of aldehydes may proceed following two competitive pathways: proton-coupled electron transfer (PCET) pathway and hydrogen-atom transfer (HAT) pathway via chemisorbed hydrogen (H*) from water (SI Appendix, Fig. S2) (14–19). In the PCET pathway, the randomly distributed proton source in the electric double layer leads to nonoriented protonation, normally resulting in mixed products that pose an impediment for product separation and extraction (SI Appendix, Fig. S3). Moreover, the tunneling barrier for electron transfer across the solvation layer causes slow kinetics. These two facts restrict the PCET-dominated ECR for the high-rate synthesis of single product, especially under large current densities (14). As for the HAT process, H* is confined on the electrode surface and transferred to adsorbed aldehyde carbonyls, theoretically enabling only carbonyl hydrogenation and thus high selectivity for single alcohol products (15, 20). HAT-dominated ECR commences with electrochemical Volmer reaction to generate H*, followed by nonelectrochemical coupling between H* and aldehyde adsorbates. Thus, the rate of HAT-dominated ECR directly depends on the H*-related electrokinetics (generation/transfer/addition) and the aldehyde adsorbate coverage (21).

Therefore, to achieve the high-rate ECR valorization for single product FA, it is expected to inhibit the undesired PCET and amplify the HAT contribution to the overall ECR. However, Cu with intermediate hydrogen overpotentials is not an ideal site for water activation. Thus, HAT is not expected to contribute appreciably on monocomponent Cu. In fact, previous studies suggested that the PCET process dominates on Cu even in neutral and alkaline electrolytes (21–25). To achieve the mechanism switching, we propose the introduction of an auxiliary component which should enable facile water dissociation and offer a balanced H* adsorption. In the ideal scenario, the auxiliary component can serve as a reservoir of H* that continuously spills over to realize the high-speed HAT for hydrogenation (Fig. 1C). There exist hints at this conception in previous studies, where metal oxides such as ceria facilitate the HAT in thermocatalytic hydrogenation (26–28). In addition to regulating the hydrogen kinetics, it is also essential to strengthen the aldehyde adsorption because Cu locates on the right branch of the carbonyl adsorption−activity volcano plot (SI Appendix, Fig. S4) (16, 29).

Based on the aforementioned analyses, here we leveraged the theoretical predictions followed by systematic catalysis experimental verifications to identify ceria as an ideal auxiliary component for realizing HAT-dominated ECR. We further found that oxygen vacancy (O_V)-rich ceria could deliver an efficient formation−diffusion−hydrogenation chain of H* and accelerate the rate-determining HAT step across the Cu/ceria interfacial perimeter by diminishing the H* adsorption. Local enrichment of FUR* was realized by preserving perimeter sites during the ECR. Following these mechanistic insights, our specially designed Cu/a-CeO_2-x catalyst (high-density Cu atomic clusters supported on amorphous ceria) achieved the economically viable electrocatalytic valorization for single product FA with a high rate of 19.1 ± 0.4 mol h⁻¹ m⁻² and a FE_FA of 97 ± 1% at a partial current density (j_FA) of over 0.1 A cm⁻² in the continuous-flow electrolyzer. Our design made the electrocatalytic FUR-to-FA valorization a feasible route for practical application.

Results

Density Functional Theory (DFT) Simulations of the HAT-Based ECR.

As a prototypical defective oxide, ceria enables the dissociative activation of water via the Ce−O Lewis acid−base pairs (30–32). The capability to accommodate H* through Ce−O−H* bonds and facilitate the H-spillover makes ceria a potential candidate for realizing fast HAT and eventually high-rate ECR (27, 33, 34). Furthermore, the controllable stoichiometry endows ceria with flexible structural tunability. These intriguing properties, therefore, motivated us to first investigate the HAT feasibility and the ECR pathway on Cu/ceria using DFT simulations (Fig. 2A).

Fig. 2.

Theoretical predictions of the HAT-based ECR on Cu/ceria. (A) Modeling of CeO₂(111), Cu/CeO₂(111), and Cu/CeO_2-x(111). (B) Calculated H* adsorption energy. (C) Energy profiles for initial states (ISs), transition states (TSs), and final states (FSs) of water-assisted HAT. (D) Projected density of states (PDOS) calculations for the Ce 4f and O 2p orbitals (Inset: differential charge density distributions between O_V and nearby Ce atoms in CeO_2-x). (E) Gibbs free energy profiles for the stepwise FUR hydrogenation in the perimeter regions of Cu/CeO₂ and Cu/CeO_2-x. F represents the furyl ring.

Based on our concept, H-involved behaviors were first studied on stoichiometric CeO₂(111) and reduced CeO_2-x(111) with O_V. Water dissociates on CeO₂ and CeO_2-x spontaneously with near-zero barriers (SI Appendix, Fig. S5). H* and OH* adsorb on top-surface O (O_top) and Ce atoms, respectively (SI Appendix, Fig. S6 and Table S2). Compared with pristine CeO₂, the presence of O_V weakens the H* adsorption on CeO_2-x (Fig. 2B). Subsequently, H* diffuses through dynamical cleavage/generation of O_top−H* bonds with water molecules as bridges (Fig. 2C and SI Appendix, Fig. S7). On CeO₂, molecular H₂O adsorbs on Ce⁴⁺ and facilitates the one-step HAT with a barrier of 0.12 eV. In contrast, water-assisted HAT on CeO_2-x undergoes the H₂O dissociative adsorption first to give an intermediate state (O_V−Ce³⁺−OH*) due to the high affinity between O_V-neighboring Ce³⁺ and H₂O (35). Therefore, the two-step HAT on CeO_2-x exhibits a lower barrier. Furthermore, HAT dominates on ceria instead of the undesired H−H coupling to H₂ (SI Appendix, Table S3).

PDOS and Bader charge analyses (Fig. 2D and SI Appendix, Fig. S8) reveal that excess electrons released from O_V are localized onto nearby Ce atoms, reducing them to Ce³⁺. Localized 4f electrons at Ce³⁺ enhance its capability of transferring electrons to adsorbed O, thereby facilitating the water dissociative adsorption on Ce³⁺. Also, within the Ce³⁺−O_top bonds, more electrons are pushed toward O_top, which reduces its capability of receiving electrons from H*. Therefore, H* adsorption on O_top is weakened (see more details in SI Appendix, Note S1) (36). Moreover, H* diffusion barriers maintain low at higher H* coverage (SI Appendix, Table S4). These findings suggest that ceria steers the H*-involved steps thermodynamically and kinetically, confirming the feasibility of the HAT pathway.

After loading the Cu, Cu transfers electrons to nearby O atoms, creating interfacial perimeter (Cu–[O_x]–Ce) for the FUR hydrogenation. FUR prefers to adsorb on the perimeter region (SI Appendix, Fig. S9). The total energy profiles (Fig. 2E and SI Appendix, Figs. S10–S12 and Tables S5–S7) demonstrate that the initial H-hopping and addition (F−CHO* + H* → F−CHOH*) serve as the rate-determining step (RDS), with kinetic barriers of 0.75 and 0.55 eV on Cu/CeO₂ and Cu/CeO_2-x, respectively. The accelerated RDS across the perimeter region is ascribed to the weakened H* adsorption on Cu/CeO_2-x. Note that under the real conditions, a higher density of O_V will be involved. Therefore, more excess electrons can be released to weaken the H* adsorption, which is expected to further facilitate the surface HAT and subsequent hydrogenation steps (34).

In short, ceria with O_V facilitates the H* kinetics (water dissociation and surface diffusion) and significantly accelerates the rate-determining HAT step across the Cu/ceria perimeter to reduce the ECR barriers.

Catalyst Synthesis and Characterizations.

As a proof-of-concept, such Cu/ceria catalysts were synthesized through a facile impregnation and annealing process. Ceria supports were obtained by the controlled pyrolysis of Ce–salen complex (SI Appendix, Figs. S13–S16). The crystallinity and defective degree of ceria were precisely modulated by varying the pyrolysis temperature. Cu was immobilized on crystalline and amorphous ceria supports (denoted as Cu/c-CeO₂ and Cu/a-CeO_2-x, respectively). Cu/a-CeO_2-x exhibited a typical amorphous feature with no X-ray diffraction (XRD) peaks observed (SI Appendix, Fig. S17), differing from the crystalline Cu/c-CeO₂ with sharp XRD peaks of cubic fluorite ceria. Cu or Cu oxide phases were not detected, indicating that Cu species was highly dispersed on ceria supports, with no prominent aggregation during the deposition. Cu mass loadings measured by inductively coupled plasma-atomic emission spectroscopy were 1.84 and 1.79 wt% in Cu/a-CeO_2-x and Cu/c-CeO₂, respectively.

Transmission electron microscope (TEM) and high-angle annular dark-field scanning TEM (HAADF-STEM) images showed that Cu/a-CeO_2-x possessed a dominantly plate-like stacked morphology (SI Appendix, Fig. S18). High-resolution TEM image showed that Cu/a-CeO_2-x was highly amorphous (Fig. 3A). In contrast, Cu/c-CeO₂ exhibited lattice fringes of the (111) facets of fluorite CeO₂ with a spacing of 0.31 nm (Fig. 3B). Corresponding selected area electron diffraction (SAED) patterns displayed a typical amorphous diffused halo for Cu/a-CeO_2-x, while clear diffraction rings for Cu/c-CeO₂. Diffraction rings or spots of Cu were not observed (SI Appendix, Fig. S19), in line with the absence of large Cu aggregation. Aberration-corrected HAADF-STEM image (Fig. 3C) proved that the surface of Cu/a-CeO_2-x consisted of continuous amorphous domains. Energy-dispersive X-ray elemental mapping (Fig. 3D) also proved the uniform distribution of Cu and Ce. Note that the bright spots in Fig. 3C correspond to Ce atoms. Due to the low Z-contrast of Cu compared to Ce and the thickness effect of ceria support, small-sized Cu species was hard to distinguish in the HAADF-STEM mode (37–39). The analysis of Cu state should be combined with the X-ray photoelectron and absorption spectroscopies (XPS and XAS, see below).

Fig. 3.

Structural and morphological characterizations. (A and B) High-resolution TEM images of Cu/a-CeO_2-x (A) and Cu/c-CeO₂ (B). Inset: corresponding SAED patterns. (C and D) Aberration-corrected HAADF-STEM image (C) and energy-dispersive X-ray elemental mapping (D) of Cu/a-CeO_2-x. (E–J) Electron paramagnetic resonance (EPR) (E), O 1s XPS (F), Ce 3d XPS (G), Cu K-edge X-ray absorption near-edge spectroscopy (XANES) (H), Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) (I), and Ce L₃-edge XANES (J) spectra of Cu/a-CeO_2-x and Cu/c-CeO₂.

The O_V was characterized by EPR spectroscopy (Fig. 3E and SI Appendix, Table S8). The prominent isotropic signal at g = 2.003 was attributed to the unpaired electrons in O_V, manifesting the abundant O defects in Cu/a-CeO_2-x (40). In contrast, Cu/c-CeO₂ was deficient in O_V, confirmed by its silent EPR signal.

The chemical environment and elemental valence were further investigated by XPS (SI Appendix, Fig. S20 and Table S9). Deconvoluted O 1s spectra (Fig. 3F) contained three peaks: lattice oxygen (O_L), oxygen species adjacent to O_V, and surface-bound oxygen (O_S) (32, 40). The O_V concentration of Cu/a-CeO_2-x was considerably higher than that of Cu/c-CeO₂ (integral area of 45% versus 22%), in line with the EPR results. The fraction of Ce³⁺, directly related to the O_V, exhibited similar tendency (Fig. 3G) (35, 41). The Ce³⁺ ratio of Cu/a-CeO_2-x was much higher than that of Cu/c-CeO₂ (43% versus 25%).

Furthermore, we employed Cu 2p XPS and Auger electron spectroscopy (AES) to investigate the metal−support interactions (MSI). The main peaks at 932.7 eV were assigned to Cu⁰/Cu⁺ (SI Appendix, Fig. S20C). The absence of characteristic Cu²⁺ satellite peaks in 940 to 950 eV indicated that Cu²⁺ was negligible in our catalysts (38, 41, 42). Compared with Cu/c-CeO₂, the Cu⁰/Cu⁺ peak of Cu/a-CeO_2-x shifted toward higher binding energy by ~0.4 eV, indicating a more pronounced charge transfer from Cu to adjacent Ce atoms. The strong electronic MSI of the a-CeO_2-x support can facilitate the Cu⁺ formation at the Cu−[O_x]−Ce interfacial perimeter (35, 43). Cu LMM AES revealed that Cu⁺ in Cu/a-CeO_2-x was much higher than that in Cu/c-CeO₂ (62% versus 36%) (SI Appendix, Fig. S20D) (35, 44). Combined with the indistinguishable morphology due to the small size and the absence of Cu²⁺ XPS signal, we deduced that Cu in Cu/a-CeO_2-x and Cu/c-CeO₂ might exist as atom-level clusters.

The Cu valence state was also investigated by K-edge XANES. Characteristic 1s → 4p transition peaks at 8,982.2 eV confirmed the presence of Cu⁺ (Fig. 3H) (29, 45). Pre-edge spectra of Cu/a-CeO_2-x and Cu/c-CeO₂ lay between those of bulk Cu₂O and Cu foil, and Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) spectra showed two peaks of Cu−O and Cu−Cu coordination (Fig. 3I), implying positively charged Cu^δ+ (0 < δ < 1) states (46). The pre-edge XANES spectrum of Cu/a-CeO_2-x was closer to Cu₂O than Cu/c-CeO₂, suggesting the more Cu⁺ content in Cu/a-CeO_2-x, consistent with the AES results. Ce L₃-edge spectra presented double white-line features and the adsorption edges lay between those of CeCl₃ and CeO₂, indicating a mixture of Ce³⁺/Ce⁴⁺ states (Fig. 3J). Ce³⁺ was more prominent in Cu/a-CeO_2-x after fitting the L₃-edge spectra (SI Appendix, Fig. S21 and Table S10) (32).

Taken together, the above results demonstrated the successful synthesis of Cu atomic clusters supported on ceria with diverse O_V degree and interfacial perimeter.

Mechanistic Investigations.

We then sought further mechanistic insights into the H-related kinetics and FUR adsorption behaviors. We measured the hydrogen/deuterium kinetic isotope effect (KIE_H/D) to investigate the H transfer in the ECR (21, 47, 48). Note that due to the different vibrational zero-point energies of H/D atoms, the participation of a H-involved elementary step in the RDS (in our case, HAT across the interfacial perimeter for the initial hydrogenation) will inevitably cause different overall ECR rates in H₂O and D₂O-based electrolytes, namely, a KIE_H/D value greater than 1 (SI Appendix, Note S2) (21). As shown in Fig. 4A, the ECR on Cu/c-CeO₂ was severely suppressed when the electrolyte changed from KOH + H₂O to KOD + D₂O, with KIE_H/D values varying from 2.4 to 2.9 in the ECR potential range (Fig. 4B and SI Appendix, Fig. S22). We also measured the KIE_H/D on pure Cu powder for comparison, which exhibited a larger KIE_H/D of over 3.9 (SI Appendix, Fig. S23). The mitigated KIE_H/D on Cu/c-CeO₂ suggested that c-CeO₂ support should convert the ECR from a PCET-dominated pathway to a relatively slow HAT-dominated one, that is, contribution of the HAT process to the overall ECR was highlighted with the auxiliary support. In contrast, Cu/a-CeO_2-x underwent much slighter current attenuation in deuterated electrolytes, exhibiting the characteristic of primary KIE_H/D (1.4 ~ 2.0), suggestive of the HAT-involved RDS and significantly accelerated HAT on Cu/a-CeO_2-x (47, 48), in line with the preceding theoretical predictions. To further verify the HAT mechanism on Cu/a-CeO_2-x, we analyzed the dependence of ECR rates for FA on evaluated FUR concentrations. For ECR via a HAT process, namely, H* generation first followed by FUR* hydrogenation, it follows a Langmuir−Hinshelwood pathway where H* and FUR* compete for adsorption sites (17, 24, 49). As shown in SI Appendix, Fig. S24, when increasing the FUR concentration, a negative reaction order (−0.70) was observed on Cu/a-CeO_2-x, indicative of the competitive adsorption between H* and FUR*. Therefore, the ECR on Cu/a-CeO_2-x is indeed a HAT process.

Fig. 4.

ECR mechanistic investigations. (A) Polarization curves in H₂O and D₂O electrolytes with 50 mM FUR. (B) KIE_H/D values of Cu/a-CeO_2-x and Cu/c-CeO₂. (C) Quasi in situ EPR spectra obtained after 8 min of ECR at −0.45 V_RHE with 150 mM DMPO. (D) Time-dependent FUR conversion with or without 200 mM tert-butanol. (E) ATR-SEIRAS spectra of FUR adsorption on our catalysts and bare Si. (F) Operando Cu K-edge XANES spectra of Cu/a-CeO_2-x during the ECR using chronoamperometry. (G) Calculated Cu^x species ratios extracted from linear combination fitting of the XANES spectra. (H) Corresponding FT-EXAFS spectra.

To demonstrate the discrepancy in HAT kinetics and detect the ECR intermediates, quasi in situ EPR experiments were performed using 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) as the radical trapping reagents. With electricity input, two sets of peaks appeared (Fig. 4C), including a nonet (hyperfine coupling constants, A_N = 16.2 G, A_βH = 22.2 G, intensity ratio of 1:1:2:1:2:1:2:1:1) assigned to DMPO−H adduct and a sextet (A_N = 15.6 G, A_βH = 22.3 G, intensity ratio of 1:1:1:1:1:1) assigned to DMPO−α-C adduct (SI Appendix, Fig. S25) (44, 50, 51). The captured H radicals were ascribed to the in situ generation and diffusion of H* from the surface. The captured α-C radicals (·CH(OH)C₄H₃O, ketyl radical) originated from the adsorbed FUR (F−CHO*) coupling with H*, proving that the ECR on our catalysts follows a HAT pathway (Langmuir−Hinshelwood mechanism).

Of note, the EPR signal intensity is proportional to the concentration of DMPO-adducts (50). Our trapping experiments followed strictly consistent operating conditions to ensure the feasibility of qualitative analysis, because the lifetimes of DMPO-adducts are minute-level. Cu/a-CeO_2-x produced significantly higher concentration of both DMPO−H and DMPO−C adducts. In situ formed H* has three destinations: dimerizing to H₂ [i.e. hydrogen evolution reaction (HER)], participating in the hydrogenation, or trapped by DMPO. Under alkaline conditions, HER was quite limited, confirmed by the poor HER performance of both Cu/a-CeO_2-x and Cu/c-CeO₂ (SI Appendix, Fig. S26). The stronger C signal of Cu/a-CeO_2-x indicated a faster H* transfer to the F−CHO*. Even so, Cu/a-CeO_2-x still presented a stronger H signal, confirming a higher steady-state H* coverage, which is conducive to the continuous H* supply for hydrogenation. Namely, Cu/a-CeO_2-x can achieve a highly efficient formation−diffusion chain of active H*.

The H* contribution was further identified by the radical quenching experiment with tert-butanol as the H* scavenger (Fig. 4D). The reduced FUR conversion in the presence of tert-butanol indicated that active H* supply played a rate-determining role in the ECR kinetics (8, 21).

We then used attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) to probe the FUR adsorption behavior, which was conducted in D₂O to avoid the overlap of H₂O peak with carbonyl peak (SI Appendix, Fig. S27). For the pure FUR, the strong peak at 1,670 cm⁻¹ corresponds to the carbonyl stretching vibrational mode (υ_C=O) of trans-FUR conformer (52, 53). The υ_C=O peak redshifted to 1,667, 1,662, and 1,655 cm⁻¹ on a-CeO_2-x, Cu/c-CeO₂, and Cu/a-CeO_2-x, respectively (Fig. 4E). Considering the 4 cm⁻¹ spectral resolution, the significant redshift on Cu/a-CeO_2-x manifested a stronger FUR activation adsorption than that on Cu/c-CeO₂ (15 cm⁻¹ versus 8 cm⁻¹), which can present a FUR enrichment and carbonyl bond-weakening effect, contributing to lowering the hydrogenation barrier (29). Due to the negligible adsorption on the a-CeO_2-x, we deduced that FUR adsorbs on the Cu⁺−[O_x]−Ce³⁺ perimeter sites, as implied by DFT simulations (SI Appendix, Fig. S9). Other peaks in Fig. 4E all remained at the same position, including cis-FUR C=O stretching, ring breathing, and aldehyde C−H rocking (SI Appendix, Fig. S28 and Table S11), which indicated that except for the adsorbed carbonyl group, the rest of FUR molecule is pushed away from the catalyst surface, leading to a tilted orientation. This repulsion effect is due to the overlap of the Cu 3d band with the aromatic furyl ring (54). Therefore, FUR adsorbs through the trans-FUR carbonyl O and adopts a η¹-(O)-aldehyde configuration, which benefits the selective hydrogenation of carbonyls and avoids side reactions related to the rings (42).

Operando XAS was then performed to monitor the Cu oxidation state evolution. As the potential decreased, the absorption edge of Cu/a-CeO_2-x moved slightly toward lower energy (Fig. 4F). In contrast, the absorption edge of Cu/c-CeO₂ almost overlapped with that of Cu foil (SI Appendix, Fig. S29). We used linear combination fitting to quantitatively analyze the Cu oxidation states (Fig. 4G and SI Appendix, Fig. S30) (46). 61% of the initial Cu⁺ was preserved in Cu/a-CeO_2-x under the ECR potentials while only 46% was preserved in Cu/c-CeO₂. The FT-EXAFS spectra coincided with the XANES results (Fig. 4H). The majority of Cu in Cu/c-CeO₂ evolved into Cu⁰, whereas Cu/a-CeO_2-x maintained a considerable ratio of Cu⁺/Cu⁰, indicating the stabilizing effect of the a-CeO_2-x support. The large reservation of Cu⁺ could be attributed to the strong electronic MSI in the Cu⁺−[O_x]−Ce³⁺ interface, in line with previous reports (55, 56). According to the DFT results (SI Appendix, Fig. S9), the Cu⁺−[O_x]−Ce³⁺ interface promoted the FUR adsorption. Thus, the stabilized Cu⁺ in Cu/a-CeO_2-x can provide more adsorption sites to enrich the local FUR* and lower the ECR barrier.

ECR Performance of Cu/Ceria Catalysts.

We first assessed the ECR performance in a standard three-electrode H-cell configuration in 1 M KOH with 50 mM FUR (SI Appendix, Fig. S31). Linear sweep voltammogram was recorded to establish the electrolysis windows (SI Appendix, Fig. S32). With the addition of FUR, Cu/a-CeO_2-x and Cu/c-CeO₂ showed enhanced current densities and prominent FUR reduction peaks from −0.3 to −0.6 V versus reversible hydrogen electrode (RHE) (57), confirming their intrinsic ECR activities. More importantly, Cu/a-CeO_2-x achieved nearly twice the current density of Cu/c-CeO₂ in the reduction peak region, indicating that the a-CeO_2-x support greatly accelerated the ECR kinetics. The FEs, selectivity, and yield rates of ECR products were then measured under steady-state constant potential electrolysis using gas chromatography-mass spectrometer (GC-MS) with the internal standard method. Total ion chromatograms showed that FA was the single product without hydrogenolysis, over-hydrogenation, or dimerization byproducts (SI Appendix, Figs. S33 and S34), which was also confirmed by ¹H NMR spectroscopy (SI Appendix, Figs. S35 and S36). A weak NMR signal of furoic acid was observed due to inevitable nonelectrochemical disproportionation (Cannizzaro reaction) of FUR (25). Cu/a-CeO_2-x reached a maximum FE_FA of 97 ± 2% at −0.4 V_RHE and maintained a high FE_FA plateau across a wide potential range, and nearly 100% FA selectivity was achieved on Cu/a-CeO_2-x (Fig. 5A). There was no significant difference in FE_FA and FA selectivity between Cu/a-CeO_2-x and Cu/c-CeO₂ (SI Appendix, Fig. S37). In a sharp contrast, Cu/a-CeO_2-x presented a much larger j_FA (65.7 ± 2 mA cm⁻²) and FA yield rate (8.9 mol h⁻¹ m⁻²) at −0.5 V_RHE, both 1.9-fold higher than those of Cu/c-CeO₂ (Fig. 5B and SI Appendix, Fig. S38). Given the similar Cu content and product selectivity, the discrepancy in the activity originates obviously from that in intrinsic ECR kinetics caused by the ceria supports. Note that Cannizzaro reaction also produces FA. We quantified the furoic acid after the chronoamperometry test and found that less than 5% of the obtained FA came from Cannizzaro reaction (SI Appendix, Table S12). It is believed that the detected FA was predominantly ECR product, especially under larger current densities. Therefore, we used apparent values of FE_FA, j_FA, FA yield rate, and FUR conversion to evaluate ECR activity in this work, which also conduced to our subsequent comparison with other works.

Fig. 5.

ECR performance. (A) FE_FA and FA selectivity over Cu/a-CeO_2-x in H-cells. All error bars represent SD with three independent measurements hereinafter. (B) FA yield rates in H-cells. (C) FUR-to-FA conversion at −0.4 V_RHE in H-cells. (D) FE_FA and FA yield rates over Cu/a-CeO_2-x at −0.4 V_RHE during 20 consecutive cycles in H-cells. (E) Schematic of the GDL-based flow cell. (F) Polarization curves (without iR-compensation, scan rate: 10 mV s⁻¹) in 1 M KOH with or without 50 mM FUR in flow cells. (G) j_FA and FA yield rates in flow cells. (H) Comparison of Cu/a-CeO_2-x with state-of-the-art electrocatalysts in terms of FE_FA and j_FA. Detailed data are provided in SI Appendix, Table S1.

Moreover, the FUR conversion was linearly related with the applied charge and eventually reached 97.6% after the theoretical charge required for the complete conversion was applied (Fig. 5C). The continuous steady conversion of FUR indicated the high ECR activity of Cu/a-CeO_2-x even under an ultralow FUR concentration. The ECR recyclability was then evaluated by running 20 consecutive electrolysis cycles (Fig. 5D). Impressively, the FE_FA maintained over 94% and the FA yield rates fluctuated slightly in each cycle but essentially remained stable, indicating the superior durability of Cu/a-CeO_2-x.

To verify the economic viability of our electrocatalytic FUR-to-FA conversion process, we assembled a continuous-flow electrolyzer with a narrow inter-electrode gap and gas-diffusion layer (GDL)-supported Cu/a-CeO_2-x as the working electrode (Fig. 5E and SI Appendix, Fig. S39). Analogously, Cu/a-CeO_2-x delivered a much larger FUR-to-FA conversion current, and maintained superior single product selectivity (FE_FA of 97 ± 1% at −0.5 V_RHE and nearly 100% selectivity) (Fig. 5F and SI Appendix, Fig. S40). The production rate reached 19.1 ± 0.4 mol h⁻¹ m⁻² with an average j_FA of 102 ± 2 mA cm⁻² at −0.5 V_RHE (Fig. 5G). Compared with state-of-the-art electrocatalysts (Fig. 5H and SI Appendix, Table S1), Cu/a-CeO_2-x clearly outperforms others in terms of j_FA and exceeds the profitable threshold as revealed by the cost analysis (SI Appendix, Fig. S41), indicating its great potential in economic viability and industrial application. In addition, the versatility of Cu/a-CeO_2-x was confirmed by extending the substrate scope to other bioderived aldehydes (SI Appendix, Fig. S42).

Discussion

In summary, we have successfully leveraged the mechanistic analysis, DFT simulations, and systematic kinetic experiments to identify that copper/ceria can realize high-speed HAT-dominated ECR for the electrosynthesis of single product FA. Ceria with abundant O_V achieves high-efficiency formation−diffusion−hydrogenation of H* by diminishing H* adsorption, as well as the preservation of Cu⁺−[O_x]−Ce³⁺ interfacial perimeter for enhancing the FUR adsorption, synergistically reducing the barrier of the RDS (F−CHO* + H* → F−CHOH*) (Fig. 6). Our Cu/a-CeO_2-x catalyst achieved the electrocatalytic FUR-to-FA valorization with a high FE_FA of 97 ± 1% and an economically viable j_FA of 102 ± 2 mA cm⁻² at −0.5 V_RHE in the flow electrolyzer. This work offers avenues to electrocatalytic valorizing biofeedstocks toward economically viable or even industrial-scale productivity.

Fig. 6.

Schematic plot of the ECR mechanism. Proposed mechanism and the role of ceria in the electrocatalytic FUR-to-FA valorization.

Materials and Methods

Materials.

o-vanillin (99%), 2,2-dimethylpropane-1,3-diamine (98%), Ce(NO₃)₃·6H₂O (99.95%), Cu(NO₃)₂·2.5H₂O (99.99%), FUR (99.5%), FA (98%), 2-furoic acid (98%), 2-methylfuran (98%), p-xylene (99%), D₂O (99.9 at% D), DMPO (97%), and tert-butanol (99.8%) were purchased from Aladdin. KOH (85%), NaOH (98%), ethyl acetate (99.8%), NaCl (99%), Ni foil (0.1 mm, 99.5%), Cu foil (0.05 mm, 99.8%), and IrO₂ (99%) were purchased from Alfa Aesar. Ethanol was purchased from Beijing Chemical Works in analytic grade. Nafion ionomer dispersion (5 wt%, DuPont), carbon paper (TGP-H-060, Toray), GDL (YLS-30T, Toray), and Nafion-212 membrane (DuPont) were obtained from commercial suppliers. Deionized water (18.2 MΩ cm, Millipore) was used for all experiments.

Characterizations.

XRD was conducted on a PANalytical Empyrean diffractometer with Cu Kα radiation (λ = 1.5416 Å) at 40 kV and 40 mA. TEM images were collected by JEM-2100F microscope (JEOL, Japan) with an energy-dispersive X-ray mapping detector (Oxford Instrument, UK) at an accelerating voltage of 200 kV. HAADF-STEM was performed on a JEM-ARM200CF instrument (JEOL, Japan) at 200 kV with double hexapole Cs correctors (CEOS, Heidelberg, Germany) and cold-field emission gun. XPS was performed on a Thermo Scientific ESCALAB 250Xi spectrometer with 300 W monochromatic Mg Kα radiation. XAS was measured at 1W1B beamline of Beijing Synchrotron Radiation Facility, China, and collected in fluorescence mode at Cu K-edge or Ce L₃-edge with a Lytle detector. Cu foil was used for K-edge energy calibration. EPR was performed on a Bruker EMXplus-9.5/12 spectrometer (Bruker, Germany). Thermogravimetric analysis was performed on a DTG-60 analyzer (Shimadzu, Japan) under a N₂ flow. Inductively coupled plasma atomic emission spectroscopy was performed on a Thermo iCAP-RQ atomic emission spectrometer. ¹H NMR spectra were recorded on a Bruker Avance III 400 HD spectrometer with water suppression.

Synthesis of Salen-Type Ligand (S-Ligand).

To a stirred 100 mL ethanol solution of o-vanillin (4.053 g) was added dropwise 30 mL ethanol solution of 2,2-dimethylpropane-1,3-diamine (1.474 g), which was stirred under reflux at 90 °C for 6 h. Ethanol was removed under reduced pressure to obtain S-ligand crystals, which were dried at 60 °C in vacuo.

Synthesis of Ce–Salen Complex.

Ten mL ethanol solution of Ce(NO₃)₃·6H₂O (2.605 g) was added dropwise into 100 mL ethanol solution of S-ligand (2.054 g), which was stirred under reflux at 60 °C for 6 h before cooling to room temperature. The crude product was filtered, washed with cold ethanol (50 mL × 3), and dried at 60 °C in vacuo to give Ce–salen complex.

Synthesis of Cu/Ceria Catalysts.

Ce–salen was heated at 600 °C for 1 h in a tube furnace with a ramp rate of 10 °C min⁻¹ under N₂ to obtain a-CeO_2-x. a-CeO_2-x (50 mg) was dispersed in NaOH solution (8 mL, 0.6 mM) and sonicated for 30 min, to which was added dropwise Cu(NO₃)₂·2.5H₂O solution (2.5 mL, 10 mg mL⁻¹) under stirring at 1,000 rpm. The resulting dispersion was stirred for 12 h. The solid was collected by centrifugation, washed with ethanol, dried at 50 °C in vacuo, and heated at 500 °C for 1 h with a ramp rate of 10 °C min⁻¹ under Ar to give Cu/a-CeO_2-x. Cu/c-CeO₂ was prepared with the same procedure except that pyrolysis temperature of Ce–salen was 900 °C.

Electrochemical Measurements.

Catalyst (5 mg) was dispersed in a mixture of ethanol (970 μL) and Nafion dispersion (5 wt%, 30 μL). The ink was sonicated for 30 min and drop-casted onto carbon paper (1 mg_cat cm⁻², 0.5 × 0.5 cm²). The electrodes were dried at 50 °C in vacuo before use.

Electrochemical measurements were conducted on an electrochemical station (CHI660E) at the scan rate of 10 mV s⁻¹ without iR-compensation in the H-cell. Anolyte (1 M KOH) and catholyte (1 M KOH+50 mM FUR) were separated by a Nafion-212 membrane. The three-electrode system included the as-prepared electrode as working electrode, Ag/AgCl (saturated KCl) as reference electrode, and Pt wire as counter electrode. Before the measurements, working electrode was scanned for 10 cyclic voltammetry laps from 0.2 to −1.2 V_RHE. The catholyte was stirred with Ar gas protection for 15 min before electrolysis and purged with Ar throughout the electrolysis. Chronoamperometry was conducted under stirring at 500 rpm for 1 h. pH of the electrolyte was monitored before (13.8) and after (13.7) electrolysis to make sure no obvious change was observed.

Electrodes after electrolysis were rinsed with water and dried in vacuo before remeasurements. Potentials were converted to the RHE reference scale using

$$E_{\text{RHE}}=E_{\mathrm{Ag}/\text{AgCl}}+0.059\times \text{pH}+0.197,$$

The flow electrolyzer consisted of catalyst-deposited GDL (2 mg_cat cm⁻²) as working electrode, IrO₂-deposited Ni foil (1 mg cm⁻²) as the anode, and Cu foils as current collectors. Reference electrode was located inside the cathode flow channel. Nafion-212 membrane was used to separate the anolyte (1 M KOH) and catholyte (1 M KOH+50 mM FUR). The flow channel was 3 mm thick and the window area was 1 × 1 cm². The flow rates of electrolytes were imposed by peristaltic pumps.

Product Analysis.

Products were quantitatively analyzed by a GC-MS (Agilent 8890) with a Waters Xevo TQ MS detector. Catholyte (3 mL) was collected after electrolysis, saturated with NaCl and extracted with ethyl acetate (3 mL) for successive two times by stirring with a vortex mixer and collecting the upper ethyl acetate phase. The extraction was strictly controlled each time to ensure that the results were not affected by the extraction operation. p-xylene served as an internal standard. FEs, yield rates, and selectivities were calculated as follows:

$${\text{FE}}_{\text{FA}}(\%)=\frac{n_{\text{FA}}\times z_{\text{FA}}\times F}{Q}\times 100\%,$$

$$\text{Yield} \text{rate} \text{of} \text{FA} (\text{mol} {\mathrm{h}}^{-1} {\mathrm{m}}^{-2}) =\frac{n_{\text{FA}}}{t\times S},$$

$$\text{Selectivity} \text{of} \text{FA} (\%) = \frac{c_{\text{FA}}}{c_{\text{FUR}}^0-c_{\text{FUR}}}\times 100\%,$$

$$\text{Conversion} \text{of} \text{FUR} (\%) = \frac{(c_{\text{FUR}}^0-c_{\text{FUR}})}{c_{\text{FUR}}^0}\times 100\%,$$

where n_FA is the mole amount of FA produced, z_FA is electron transfer number, F is Faraday constant, Q is the total charge, c⁰_FUR is the initial FUR concentration, c_FUR and c_FA represent the time-dependent concentrations; t is the electrolysis time, and S is the effective electrode area.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.