Introduction

Electrochemical catalysis has the potential to revolutionize organic synthesis by creating high-value products using electricity. By utilizing voltage and current, it is possible to bypass the need for high temperature, pressure, and stoichiometric amounts of redox agents, directly altering the functional groups’ structure and oxidation state1,2,3,4,5. In recent years, the increasing adoption of renewable energy has made green electricity more accessible and affordable6, leading to growing interest in coupling electrolysis technologies with novel synthesis approaches. The combination of electrocatalysis and bio-derived compounds shows great promise for improving both the technological and economic aspects of organic and green chemical synthesis, paving the way for a more sustainable future for green chemistry7.

Carbonyl groups received significant attention due to their wide existence in electrochemical biomass and CO2 valorization approaches8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26. The carbonyl group can be electrochemically oxidized to carboxylates or reduced to oxygenates and alkanes20,23,27. Previously, we showed that formaldehyde could be reduced to alcohols on metals22. However, these electro-organic reactions are catalyst-dependent and involve complicated proton-coupled and -decoupled reaction steps, making detailed mechanistic understanding at the atomic scale elusive.

Figure 1 depicts generalized steric configurations of how a carbonyl group can be adsorbed on the catalyst surface. At negative potentials, adding a proton-electron pair or hydrogen to the C=O group, the initial reaction intermediate adsorption can occur through oxygen binding (Fig. 1a) or carbon binding (Fig. 1b), each leading to various channels and products. Following Fig. 1a (oxygen-down form), Eley-Rideal hydrogenation on the unbound carbon atom can split the C-O bond, producing alkanes, such as formaldehyde reduction on single-site Fe–N–C candidates26. Alternatively, Langmuir-Hinshelwood hydrogenation on the “oxygen-down” intermediate may yield alcohols on metals22,23,24. Additionally, in the case of the initial horizontal two-site adsorption type as proposed in ketone reduction on Pt facets, strong binding energy can dissociate the C=O bond and fully protonate the intermediate to hydrocarbons. However, this adsorption type is not considered on these M–N–C catalysts. In contrast, weak adsorption leads to oxygenate formation19,20. These reduction mechanisms contain multiple possibilities and remain unaddressed.

Fig. 1: Speculated reaction cascade of carbonyl reduction on the catalyst surface.
figure 1

a Adsorption in oxygen-down form and b adsorption in carbon-down form. R1 and R2 represent the alkyl group or H.

Full size image

Herein, we investigated the electrochemical reduction of carbonyl groups on single-site M–N–C catalysts (M: Fe, Co, and Ni) using formaldehyde, acetaldehyde, and acetone as model reactants. We strive to establish correlations between the nature of the metal center, adsorption geometry (oxygen-down vs carbon-down), and the products selectivity (hydrocarbons vs oxygenates). As the scaling relations link oxygen-down (\(\Delta {E}_{{CH}3O* }=\Delta {E}_{{CH}3{CH}2O* }=\Delta {E}_{{CH}3{CHO}* {CH}3}\)) and carbon-down (\(\Delta {E}_{* {CH}2{OH}}=\Delta {E}_{{CH}3* {CHOH}}=\Delta {E}_{{CH}3* {COHCH}3}\)) adsorption configurations according to our Density Functional Theory calculation, our findings highlight the critical role played by the metal center’s oxyphilicity and the resulting adsorption geometry in controlling product selectivity. Specifically, Fe–N–C exhibited oxyphilic behavior, leading to oxygen-down adsorption for carbonyl groups and enhancing the selectivity for hydrocarbons formation during aldehyde group reduction. Conversely, carbophilic Co–N–C exhibited the ability to adsorb the carbon site of carbonyl groups in acetaldehyde and acetone, yielding ethanol and 2-propanol, respectively.

Results

Density functional theory predicted adsorption geometry

In this work, we study the electrocatalytic reduction of carbonyl compounds using a set of single-site M–N–C as model catalysts. The M–N–C catalysts feature isolated single metal atoms with nitrogen coordination, forcing sole atom adsorption (displayed in Fig. 1a, b). We selected formaldehyde, acetaldehyde, and acetone as reactants (C1–C3 carbonyls) for the electrochemical reduction reaction22. By varying the metal center of the active M–Nx motifs, the binding strength to the reactive intermediates (C vs. O) can be tuned, resulting in distinct reaction paths and corresponding selectivity. Our framework utilizes the adsorption geometry of the carbonyl group (oxygen-down vs. carbon-down) and the binding strength as the selectivity-determining indicator for the production of alcohols or alkanes.

Density Functional Theory was used to determine the binding energies of the initial adsorbed carbonyl intermediates after a proton–electron transfer step, in both carbon-down and oxygen-down forms. Figure 2 illustrates the binding strengths of these two adsorption configurations for various M–Nx site’s structure (red) and metal (111) surfaces (black), featuring the carbonyl groups ranging from C1 to C3: formaldehyde (Fig. 2a), acetaldehyde (Fig. 2b), and acetone (Fig. 2c). The diagonal in all panels represents balanced adsorption energy line of the oxygen-down and carbon-down forms. Regions below the diagonal line favor oxygen-down adsorption configuration, while the upper regions prefer carbon-down adsorption (schematic illustrations are detailed in all panels). An important observation is that the binding energy patterns in a, b, and c are similar due to scaling relations between oxygen intermediates: (CH3O*, CH3CH2O*, and CH3CHO*CH3), which all have one oxygen bond to the catalyst surface and between carbon intermediates: (*CH2OH, CH3*CHOH and CH3*COHCH3), which all have one carbon bond to the surface. The simplistic picture is general, with one oxygen versus one carbon bond as represented for OH* vs. *CH3 shown in Fig. S1. The origin of the fundamental scaling relation between the oxygen bonds and carbon bonds can be shown for both M–N–C catalyst and for metal catalysts as depicted in Fig. S2.

Fig. 2: Carbonyl reduction intermediates.
figure 2

a Formaldehyde (CH2O), b acetaldehyde (CH3CHO), and c acetone (CH3COCH3) with the oxygen-bound intermediates on the y-axis and carbon-bound intermediates on the x-axis. While M–N–C’s catalyst (red cycle) falls close around the diagonal, metal catalysts (gray cubic) are more scattered, with some having stronger oxygen binding.

Full size image

Focusing on Fig. 2, we notice the theoretical perspective that most metallic catalysts exhibit a preference for binding with oxygen-down over carbon-down (Fe, Co, Ru, Cu, and Ag), except for precious metals like Pt, Au, Rh, and Pd, which are close to the diagonal. In contrast, most M–N–C candidates exhibit similar binding strengths for both adsorptions, positioned close to the diagonal line. Our benchmark study of M–N–C s showed that Co–N–C is located in the carbon-down region, while Fe–N–C is slightly in the oxygen-down region for all projected carbonyl compounds, e.g., the difference in formaldehyde carbon binding is 0.16 eV (Co–N–C: −0.76 eV, Fe–N–C: −0.6 eV) and oxygen binding is 0.3 eV (Co–N–C: −0.51 eV, Fe–N–C: −0.81 eV) for Co–N–C and Fe–N–C. This difference could potentially provide a different reaction path on our two selected catalysts. In line with previous work on oxygen reduction reaction, Fe–N–C allows the four-electron reduction of oxygen to water, whereas the weaker oxyphilic Co–N–C site is more selective toward the two-electron reduction toward H2O228. This suggests Fe–N–C’s capability to bind oxygen (O*) at the active site, whereas this is partly limited to Co–N–C. On Ni–N–C, both carbon-down and oxygen-down exhibit similar but weak binding energies for intermediate adsorption.

Electrochemical reduction of carbonyl groups on M–N–C catalysts

We conducted electrochemical reduction experiments on Fe–N–C, Co–N–C, and Ni–N–C catalysts derived from ZIF-8. Those candidates have been shown to possess single metal atoms as active sites, and their exposed surface area and metal concentration remain in a similar range (Supplementary Information Table S1, Fig. S3 and Note)29. A reference candidate without our studied metal centers in this work is named N–C. All reactants selected in this set study are CH2O, CH3CHO, and CH3COCH3, with an initial concentration of 5 mM.

We first analyzed the behavior of formaldehyde reduction by performing linear sweep amperometry (LSV) from −0.2 to −0.8 VRHE before the bulk electrolysis and measuring the geometric current density as a function of scanning potential (Fig. S4). The LSV plots of N–C, Fe–N–C, and Ni–N–C catalysts showed only minor differences with the presence of CH2O. Interestingly, the Co-based catalyst seems poisoned by CH2O (Fig. S4a).

Later, we held the cathode potential stationary for 75 min for product analysis (details are given in “Method” section and Supplementary Information). The hydrogen evolution reaction was the prominent process, leaving only <10% faradaic efficiency for CH2O conversion. Moreover, within 75 min of electrolysis, other liquid products, such as CH3OH, were not detectable. This latter observation is different from CH2O reduction on metal catalysts22.

Figure 3a, b shows the reaction rate and faradaic efficiency of CH4 formation, highlighting the role of the nature of the metal centers for the reactivity of CH2O reduction. Fe–N–C yielded CH4 with a maximum efficiency of 6% at −0.5 VRHE. The reaction rate increased with more cathodic potentials while the faradaic efficiency decreased. On the contrary, Co–N–C was entirely inactive for CH2O to CH4 conversion. The LSV profile evidenced that CH2O poisoned Co–Nx sites and thus deactivated the catalyst (Fig. S4). The reference N–C catalyst showed a very low, yet, in a certain potential range, finite CH4 formation. We hypothesize that this minor CH4 reactivity can be attributed to the presence of distinct nitrogen species or residual Zn metal atoms.

Fig. 3: Reactivity of electrochemical CH2O reduction.
figure 3

a CH4 production rate and b CH4 faradaic efficiency as a function of the applied iR-free electrode potential on N–C (orange), Fe–N–C (red), Co–N–C (gray), and Ni–N–C (blue) catalysts. Data are averages over 75 min electrolysis obtained at 15 min, 45 min, and 75 min of each electrolysis. Line to guide the eye. Conditions: 0.05 M K3PO4 + 0.05 M H3PO4 neutral solution. Catalyst loading: 0.75 mg cm−2 on glassy carbon. The polarization curve of CH2O reduction is shown in Fig. S4 and pH dependence of CH2O reduction is presented in Fig. S5.

Full size image

We then investigated the reduction of CH3CHO on these single-site catalysts and found that C2H6 and C2H5OH achieve roughly 3% faradaic efficiency (Fig. 4), whereas the rest of the current majorly contribute to hydrogen evolution. The Ni–N–C catalyst was inactive, yielding only a low production rate (<0.1 nmol cm−2 s−1) at ~−0.8 VRHE. In addition, the Fe–N–C catalyst exhibited a higher selectivity for C2H6 formation, with the onset at −0.5 VRHE and a peak faradaic efficiency of 3% at −0.6 VRHE. This aligns with the observed preference for oxygen-down binding and protonation on the unabsorbed site, as seen in the CH2O reduction to CH4. On the other hand, the Co–N–C catalyst produces C2H5OH with a reaction rate of 1.5 nmol cm−2 s−1 and a faradaic efficiency of 3% at −0.7 VRHE. Notably, this differs from experimental observation for CH2O reduction in that only hydrocarbons are observed on Fe–N–C and no product from Co–N–C, which could be plausibly attributed to the Keto-Enol tautomerism of acetaldehyde molecule.

Fig. 4: Reactivity of electrochemical CH3CHO reduction.
figure 4

a C2H6 production rate, b C2H6 faradaic efficiency, c C2H5OH production rate, and d C2H5OH faradaic efficiency as a function of the applied iR-free electrode potential on N–C (orange), Fe–N–C (red), Co–N–C (gray), and Ni–N–C (blue) catalysts. Data are averages over 75 min electrolysis obtained at 15 min, 45 min, and 75 min of each electrolysis. Line to guide the eye. Conditions: 0.05 M K3PO4 + 0.05 M H3PO4 neutral solution. Catalyst loading: 0.75 mg cm−2 on glassy carbon. The polarization curve of CH3CHO reduction is shown in Fig. S6 and pH dependence study is presented in Fig. S7.

Full size image

In our final investigation, we focused on the electrochemical reduction of CH3COCH3 on M–N–C candidates (Fig. 5). Only the Co–N–C catalyst showed a measurable electrochemical reduction of CH3COCH3 to 2-propanol. The catalytic conversion started at around −0.5 VRHE and increased with more negative potentials, reaching a maximum production rate of 1.5 nmol cm−2 s−1. However, the faradaic efficiency for this reaction remained low at around 0.2% in our studied potential range.

Fig. 5: Reactivity of electrochemical CH3COCH3 reduction.
figure 5

a 2-propanol production rate and b 2-propanol faradaic efficiency as a function of the applied iR-free electrode potential on N–C (orange), Fe–N–C (red), Co–N–C (gray), and Ni–N–C (blue) catalysts. Data are averages over 75 min electrolysis obtained at 15 min, 45 min, and 75 min of each electrolysis. Line to guide the eye. Conditions: 0.05 M K3PO4 + 0.05 M H3PO4 neutral solution. Catalyst loading: 0.75 mg cm−2 on glassy carbon. The polarization curve of CH3COCH3 reduction is shown in Fig. S8.

Full size image

Discussion

In summary, our study employed a combination of theoretical and experimental approaches to screening electrochemical carbonyl reduction on single-site M–N–C catalysts, highlighting the crucial role of the adsorption geometry of the carbonyl group in determining product selectivity. Unlike the metal surfaces (oxygen-down adsorption selectively leads to oxygenates formation, summarized in Table 1), the oxyphilic Fe-functionalized active site preferentially binds to the oxygen site of aldehyde groups (Fig. 1a, oxygen-down adsorption) and exhibits significant faradaic efficiencies towards CH4 and C2H6 (8% and 3%, respectively). Notably, during the acetaldehyde reduction, C2H5OH arises as a by-product at more negative potentials, plausibly through a specific reaction mechanism after the initial Keto-Enol tautomerism. For acetone, the steric structure and I-effect from the two methyl groups pose a hindrance to the overall reaction reactivity. As a result, through predicted oxygen-down adsorption, the Fe–N–C catalyst yields neither hydrocarbons nor oxygenates during the ketone group reduction. On the contrary, the carbophilic (less oxyphilic) Co–N–C, according to our prediction, prefers carbon binding, selectively converting acetaldehyde and acetone to oxygenates (or being poisoned by formaldehyde). This finding demonstrates a ketone-to-oxygenate path through the carbon-down adsorption geometry. Overall, our insights into carbon-based molecule conversions on single-site catalysts could contribute to a better fundamental understanding of electrochemical CO2 and biomass valorization reactions and could pave the way for potential large-scale electrochemical carbonyl hydrogenation processes.

Table 1 Summary of product selectivity of electrochemical carbonyl reaction on our studied single site M–N–C catalysts in combination with DFT predicted initial adsorption geometry.
Full size table

Methods

The catalyst synthesis is identical to our previous work29 and the protocol is detailed in the Supplementary Information. For catalyst ink, 15 mg catalyst powder was first mixed with 50 μL Nafion solution (5 wt% solved in ethanol solution, SigmaAldrich), 150 μL isopropanol, and 800 μL DI-water, and later sonicated using SoniferHorn for 15 min. 50 μL catalyst ink was later drop casted on 1 cm2 glassy carbon plate and dried to our working electrode, for a catalyst loading of 0.75 mg cm−2.

All electrochemical assessments were conducted in a custom-made two-compartment and three-electrode H-cell, using EC-Lab SP-200 Potentiostat. The cathode chamber and anode chamber were separated using a Nafion 117 membrane. Pt mesh was deployed as the counter anode. A leak-free Ag/AgCl was used as the reference, located in the cathode chamber, close to our working electrode.

5 mM carbonyl compounds, namely, the CH2O, CH3CHO, and CH3COCH3, were added in N2-purged (30 sccm) KH2PO4/K2HPO4 (0.1 M anion concentration, Sigma-Aldrich) neutral buffer electrolyte as reactants. For measurements, the impedance between reference electrode and working electrode was measured with Potential Electrochemical Impedance Spectroscopy (PEIS) at −1.0 V vs Reference. The measurements were controlled using constant potential with 50% automatic IR correction, whereas the rest were manually corrected. Gas products were detected by an online Gas Chromatograph and the residual electrolyte were analyzed using HPLC and liquid Gas Chromatograph for liquid products. Quantification details are given in the Supplementary Information and Supplementary Eqs. (1)–(4).

For Density Functional Density calculation, the M–N–C model was created in ASE30 by a 3 × 5 unit graphene cell with a functionalized M-N4 site by removing carbon atoms. The outmost carbon atoms were fixed in position and periodic boundaries were applied. Further, the metal (111) model was built by a standard 3 × 3 × 4 slab including a vacuum region and the two lower layers fixed. The electronic calculations were carried out with the GPAW software31 with the projector augmented wave method, spin polarization and the revised Perdew–Burke–Ernzerhof functional32. We applied a 0.18 grid spacing together with a (2 × 2 × 1) k-point sampling for M-N-Cs and (3 × 3 × 1) k-point sampling for the metals and all the structures were relaxed to a force below 0.1 eV/Å.