Localized mass transport channels for electro-upgrade of dilute CO₂ toward high-yield C₂₊ products

Abstract

Electrocatalytic upgrade of CO₂ offers a promising approach for recycling of global CO₂ emissions, facilitating the achievement of carbon neutrality. Nevertheless, direct utilization of practical dilute CO₂ is urgently important yet rather difficult, which is hindered by the balance of reaction kinetics and mass transport of CO₂ to the catalytic sites. Herein, we propose coordinating the local environment and active catalyst by constructing covalent organic frameworks (COF) on single-atomic In-doped Cu₂O (In₁@Cu₂O) for a high tolerance of CO₂ inlet concentrations (15% to 100%). The optimized amounts of COF functionalized by the trifluoromethyl group act as the local CO₂/CO diffusion channels via steric confinement effects and C···F electronic effects. Besides, the formation of key intermediates for C₂₊ products is greatly facilitated by the promoted COOH adsorption. Hence, a total current of 81.7 A is realized in a 4 × 100 cm² electrolyzer stack with over 770 mmol/h C₂₊ products at an inlet of dilute CO₂. Such a electrode architecture sheds light on the dilute CO₂ electrolysis at the potential industrial scale.

Introduction

Recycling CO₂ emissions at ambient conditions by electrocatalytic conversion with renewable electricity grants an elegant carbon-neutral route^1,2,3. Such electrochemical CO₂ reduction reactions (CO₂RR) have shown techno-economic viability to synthesize valuable chemicals and feedstocks⁴. Nevertheless, the imbalance of reaction kinetics and mass transport with an increasing conversion rate and industrial dilute CO₂ limits the large-scale applications of this technique. So far, most of the research on CO₂RR is focused on pure CO₂ gas to benchmark catalyst or device performances^5,6,7. However, the general CO₂ concentration is only around 70–95% after capture and separation by the membrane plants from flue gas (15–20%)^8,9,10. Hence, the direct utilization of impure or dilute CO₂ inlets is highly significant.

The intrinsic challenge for continuous electrolysis of dilute CO₂ is that the solubility or partial pressure of CO₂ becomes too low to be fed sufficiently to the catalytic sites^11,12. Therefore, a concerted manner of concentrating CO₂ at a local level and efficiently converting CO₂ is urgently needed for further practical developments. Particularly, it is crucial to manage both localized reaction environments and catalytic sites for enrichments of local reactants or intermediates and improvements of conversion rates, respectively. Researchers have been investigating the use of covalent organic frameworks as sorption materials for capturing CO₂ because of a high affinity for CO₂¹³. Meanwhile, COF-based materials have also been extensively studied for CO₂RR, either using them as potential catalysts with tuned active sites or introducing them into catalytic layers with functionalized groups¹⁴. Attributed to the unique porous configuration and conjugated electronic structure, COF materials can be employed for hydroxide and gaseous feedstock adsorption¹⁵ or hydrophilic–hydrophobic nanochannels¹⁶ for regulating the local environment.

Herein, we coordinate reaction kinetic promotions and mass transport regulations by constructing localized mass transport channels for the scale-up of the electrolyzer stack with industrial-level electrosynthesis of C₂₊ products from dilute CO₂. As shown in Fig. 1a, through constructing mass transport channels with a COF layer on the single atomic In doped Cu₂O (In₁@Cu₂O), CO₂ concentrating and converting processes are coupled together. Hence, the dilute CO₂ (70–90%) and even simulated flue gas (15% CO₂) are efficiently converted into C₂₊ products at ambient conditions. In detail, a COF functionalized by the trifluoromethyl group (TfCOF) acts as localized channels for localized CO₂/CO mass transport. Besides, active In₁@Cu₂O is found to strengthen the adsorption of *COOH for the production of CO key intermediate toward C₂₊ products¹⁷. Correspondingly, such TfCOF-In₁@Cu₂O electrode converts dilute CO₂ with a Faradaic efficiency for C₂₊ products (FE_C2+) of 83.5% at E_cell of 3.4 V, presenting only a 3.4% decrease in FE_C2+ compared with the performance testing under 100% pure CO₂ inlet. Such high selectivity is well maintained for over 96 h with a current density of over 700 mA/cm². Additionally, a 4 × 100 cm² electrolyzer stack is assembled to achieve over 770 mmol/h C₂₊ products with a total current of 81.7 A with a dilute CO₂ inlet.

Fig. 1: Concept, fabrication, and characterizations of TfCOF-In₁@Cu₂O electrode.

a Schematic of relationships among reaction kinetics, mass transport, and industrial-level electrosynthesis. The atomic schematic is a magnified view of the highlighted region in the green boxes of the catalytic layer to highlight the reactive interface between TfCOF and In₁@Cu₂O. TfCOF layer on single atomic In₁@Cu₂O surface acting as localized channels for CO and CO₂ diffusion, which synergistically balances the reaction kinetics promotions and facilitates the scale-up of electrolyzer stacks. b Synthetic scheme of TfCOF by an ion-exchange process. Tp is 1,3,5-triformyl phloroglucinol and BTBD is 2,2′ Bis(trifluoromethyl)benzidine. c The schematic structure of TfCOF with the pore size of 18 Å. d WAXS patterns of TfCOF and the carbon paper background. e TEM, HRTEM, STEM, and element mapping images of synthesized TfCOF before the activation step. f HAADF-STEM image of TfCOF-In₁@Cu₂O. The yellow circles highlight representative single In atoms, which are brighter compared with Cu atoms. g Three-dimensional visualization of region ${{\rm{\alpha }}}$ in (e) to depict single atomic In atoms. h Element mapping images of TfCOF-In₁@Cu₂O.

Results

Electrode synthesis and in-situ activation

The TfCOF is proposed to tune the mass transport at the local environment near the catalytic sites, which is synthesized from 1,3,5-triformyl phloroglucinol (Tp) and 2,2′-Bis(trifluoromethyl)benzidine (BTBD) with the formation of Enol-imine form and further Keto-enamine form (Fig. 1b). The structures of TfCOF are illustrated as shown in Figs. 1c, S1, and S2a. The pore size is around 18 Å, which is characterized by the wide-angle X-ray scattering (WAXS, Fig. 1d), the nitrogen adsorption/desorption analysis (Fig. S2b, c), and the small-angle X-ray diffraction (XRD) patterns (Fig. S2d). The high crystallinity and the compositions of TfCOF are also observed by the crystalline lattice in transmission electron microscopy (TEM) images and elemental mapping images (Fig. 1e), respectively.

On the other hand, the active site is also delicately designed to promote the reaction kinetics of CO₂-to-C₂₊ products. As reported¹⁸, In doping of Cu-In alloy highly suppresses the hydrogen evolution and regulates the electronic structure of Cu for selective synthesis of CO intermediate. Therefore, single atomic In is further introduced to modify the surface Cu₂O site. The synthesis process of TfCOF-In₁@Cu₂O and the structure evolutions are presented in Fig. S3. An epoxide gelation method was firstly employed to prepare In-Cu gel, which consists of Cu₂(OH)₃Cl and In(OH)₃. This In-Cu gel was aged in the oven for 24 h, and then it was mixed with TfCOF to prepare catalyst ink, which was drop-casting on the carbon paper to form the electrode (Fig. S4). Under CO₂RR conditions, the synthesized electrode was subjected to in-situ electrochemical activation at a constant potential for 30 min, and then TfCOF-In₁@Cu₂O electrode was eventually obtained.

It is revealed by scanning electron microscopy (SEM) images that abundant interlaced channels are formed in the as-prepared TfCOF-In₁@Cu₂O by cross-linking of nanoparticles (Figs. S5 and S6). To further investigate this material, the high-resolution TEM (HRTEM) is examined as depicted in Figs. S7 and S8. It is found that such nanoparticles are quite fine and highly dispersed. Moreover, as shown in the high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images, isolated bright spots are identified as representing atomically dispersed In atoms in the Cu₂O matrix (Figs. 1f and S9). The lattice fringe with a d-spacing of 0.245 nm corresponds to the Cu₂O (111) plane, which is increased to 0.250 nm with the substitution of single atomic In over the Cu₂O matrix. This structural distortion directly evidences In incorporation into the bulk lattice^19,20. As shown in Fig. 1g, h, the 3D visualization and element mapping images further depict the distribution of single atomic In atoms.

Synchrotron X-ray absorption spectroscopy (XAS) characterizations are further performed to reveal the local electronic structures and coordination environment of Cu and In in TfCOF-In₁@Cu₂O catalyst^21,22. The Cu K-edge of X-ray absorption near-edge structure (XANES) for TfCOF-In₁@Cu₂O is depicted in Fig. 2a, as well as the spectra of copper foil, Cu₂O, and CuO references. The near-edge absorption peak of TfCOF-In₁@Cu₂O is located between those of Cu foil and Cu₂O, indicating the chemical valence of copper species is between 0 and +1, and close to +1^23,24. As for the catalyst modification, it is well known that monovalent Cu(I)-based catalysts are favorable for C–C bond formation^25,26,27, while single-atom alloys are reported to steer the electronic features of host material with the introduction of foreign atoms²⁰.

Fig. 2: Structural characterizations and evolutions of TfCOF-In₁@Cu₂O with in-situ activation under CO₂RR.

a XANES spectra of Cu K-edge. EXAFS spectra of b Cu K-edge and c In K-edge, respectively. d Operando WAXS patterns detected during the activation process. Operando WAXS spectra of e In(OH)₃ leaching and f Cu₂O generation during the activation step. g 3D images of TOF-SIMS elemental depth distribution of Cu and In elements before and after in-situ activation step. The bottom images are the surface distributions of single atomic In formation. Quasi in-situ NAP-XPS spectra of h Cu LMM and i In 3d spectra during the activation process.

With the analyses of the extended X-ray absorption fine structure (EXAFS) for Cu K-edge (Fig. 2b), the peak in R space at 1.46 Å is ascribed to the contribution of Cu-O bonds from Cu₂O^28,29. The broad peak at 2.35 Å is ascribed to the co-existence of Cu-In and Cu-Cu paths^24,30. Meanwhile, XAS curves of In K-edge for TfCOF-In₁@Cu₂O, indium foil, and In₂O₃ references are shown in Figs. 2c and S10. Two peaks at 1.64 Å and 2.47 Å are attributed to In-O and In-Cu paths, respectively³¹. Notably, there is no In-In coordination in TfCOF-In₁@Cu₂O, revealing that indium is atomically dispersed on Cu₂O.

Single atomic In is formed during the in-situ activation step, and such evolution of crystal structures and 3D elemental distribution of TfCOF-In₁@Cu₂O is captured by operando synchrotron 2D-WAXS measurements as shown in Fig. 2d–g. Before the in-situ activation (0 min), the crystal patterns of Cu₂(OH)₃Cl (JCPDS#50-1559) and In(OH)₃ (JCPDS#16-0161) are presented. Notably, these peaks gradually disappear and the existence of Cu₂O (JCPDS#78-2076) can be detected as the activation proceeds^24,32. It is worth noting that no In-In peaks are observed for the TfCOF-In₁@Cu₂O material, which further confirmed that there are no In particles formed. These results are further clearly verified by the integrated XRD curves in Fig. 2e, f. The In(OH)₃ (2 theta = 5.04) leaches out rapidly, while Cu₂(OH)₃Cl (2 theta = 3.65, Fig. S11) is converted to Cu₂O (2 theta = 8.27).

Time-of-flight secondary-ion mass spectroscopy (TOF-SIMS) with vacuum interconnection was also performed to analyze the elemental distribution in different depths of the electrode³³. After the activation in the glove box, the electrodes were transferred to the test station for TOF-SIMS measurements immediately. The 3D rendering of Cu and In species in the electrode before and after the activation step is depicted in Fig. 2g. In elements are observed to be distributed along different depths of the electrode before the activation step. Then the content of In is greatly reduced after the activation step (Fig. S12), which is consistent with the 2D-WAXS results. These results reveal that a large amount of In atoms leach out during the in-situ activation process. The distribution of other elements for TfCOF-In₁@Cu₂O catalysts is shown in Figs. S13 and S14. Notably, the uniform distribution of K element in Fig. S11b indicates that the TfCOF-In₁@Cu₂O catalyst holds porous structures for electrolyte penetrations.

Additionally, quasi in-situ near-ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) is performed to further elucidate the evolution of electronic structure and chemical states^34,35. As shown in Fig. S15, the binding energy of all Cu 2p_3/2 spectra for TfCOF-In₁@Cu₂O catalyst is around 932.7 eV, attributing to Cu⁰ or Cu⁺ peaks^24,36. The Cu LMM Auger spectra further clarify the major Cu⁺ species in the material after the activation step (Fig. 2h). Moreover, according to In 3d spectra in Fig. 2i, the In³⁺ peaks of In-Cu gel are located at 452.5 eV and 444.9 eV, which show a positive shift of about 0.23 eV after the activation step. The result indicates that In atom has a higher valence state than In³⁺, probably due to re-oxidations after the in-situ activation process.

Localized mass transport channels for dilute CO₂ electrolysis

We employ TfCOF-In₁@Cu₂O for regulating both catalytic sites and localized diffusion channels to simultaneously promote the reaction kinetics and regulate local mass transport for efficient dilute CO₂ electrolysis (Fig. 3a). In detail, In₁@Cu₂O plays the role of accelerating the production of CO key species for C–C coupling via enhancing the adsorption of *COOH intermediate as illustrated in the following Density Functional Theory (DFT) simulations. On the other hand, the TfCOF layer acts as the CO and CO₂ diffusion channels and makes it possible for them to contact with electrons from the catalyst and protons from the electrolyte, as illustrated by combined Multiphysics and Molecular Dynamic simulations. It is worth mentioning that the TfCOF layer also avoids the rapid departure of CO species to the electrolyte.

Fig. 3: Construction of localized mass transport channels for dilute CO₂ electrolysis.

a Schematic diagram of the specialized roles for In₁@Cu₂O and TfCOF to promote reaction kinetics and regulate mass transport, respectively. Inset in the middle revealing the steric effect for CO₂ by calculating CO₂ adsorption distribution. b FEs of different products for Cu₂O and TfCOF-In₁@Cu₂O with dilute 80% CO₂ inlets. The cell voltage of E_cell is reported without iR correction. The error bars represent the standard deviation from the measurement of three independent electrodes. c FFs toward total carbon products (FE_Ctot) and d C₂₊ products (FE_C2+) for TfCOF-In₁@Cu₂O in a flow cell with different concentrations of CO₂ inlets. The simulated flue gas is 15% CO₂ and 85% N₂. e Calculated 2D distributions of CO concentration around electrodes with and without TfCOF layer with Multiphysics simulations. f Calculated coverages of adsorbed *CO and CO₂ on the catalytic surface. The error bars represent the standard deviation from the measurement of three independent electrodes. Source data are provided as a Source Data file.

With the target of extending the commercial utilization of CO₂RR systems, we employ a customized flow cell with a CO₂-saturated aqueous inlet to deliver the carbon sources³. The architecture of neutral electrolyte flow-through electrodes not only eliminates the degradation from carbonate salt precipitations and electrolyte leakage but also highly reduces carbon losses (<5%)³⁷ with a cation exchange membrane to avoid the crossover of bicarbonate and carbonate anions³. The catholytes are bubbled by a series of pure and impure CO₂ gases: 100%, 90%, 80%, and 70% CO₂. These typical concentrations are selected to simulate the CO₂ waste gas after the membrane plants from fuel gases^9,10. As shown in Fig. S16, FE toward total carbon-containing products (FE_Ctot) shows a sharp decrease from 82.6% to 49.4% at E_cell of 3.4 V on pure Cu₂O when changing the inlet from pure CO₂ to 80% CO₂. It is attributed to that low concentrations of CO₂ inlet reduce the partial pressure of CO₂ (P_CO2) and further decrease the solubility of CO₂³⁸, which highly reduces the selectivity of CO₂RR on pure Cu₂O.

Dilute CO₂ is concentrated according to the electrostatic interaction by the modifications of TfCOF in the electrode^13,14,39. As shown in Fig. S17a, Faradaic efficiency toward total carbon-containing products (FE_Ctot) is promoted to 86.9% using TfCOF-Cu₂O and 80% CO₂ inlet. We further performed the measurements of electrochemically active surface area (ECSA, Fig. S18) and ECSA-normalized current densities to exclude the porosity effect (Fig. S19). It can be found that which TfCOF-Cu₂O electrode still maintains substantially higher C₂₊ partial current density (5.6 mA cm⁻² at 3.4 V) compared to that of Cu₂O electrode (2.7 mA cm⁻² at 3.4 V). This indicates that the enhanced performance is dominated by the interaction between the TfCOF and the CO₂ reactant. Nevertheless, the selectivity of C₂₊ products is still less than 75%. The possible reason is the production and concentration of CO intermediate adsorbed on the catalysts are too low to proceed for the further C–C coupling on pure Cu₂O for impure CO₂ electrolysis. Meanwhile, it is found that In₁@Cu₂O has a higher ability to produce CO than Cu₂O (Fig. S20), which confirms the critical role of single atomic In doping for the facilitating of reaction kinetics.

As expected, the selectivity of C₂₊ products on TfCOF-In₁@Cu₂O demonstrates strong tolerance on a wide CO₂ concentration window, which shows high FE_Ctot and FE_C2+ in the flow cell as shown in Fig. 3b. A maximum value of 86.3 ± 2.4% at E_cell of 3.4 V is delivered for pure CO₂ electrolysis. The obtained FE_Ctot and FE_C2+ using 90% CO₂ inlet are almost the same as those using pure inlet at a wide potential window between 2.6 V and 3.6 V (Fig. S18a). Meanwhile, FE_C2+ reaches 83.5% and FE_Ctot is 95.5% at E_cell of 3.4 V with 80% CO₂ inlet, showing comparable performance with 100% pure CO₂ and the performance damping is less than 5%. The high selectivity and activity are attributed to the balance of effective capture of CO₂ by TfCOF and faster reaction kinetic by In₁@Cu₂O.

FE_Ctot drops when the inlet changes to 70% CO₂, which is around 85.8% ± 2.2% at E_cell of 3.4 V (Fig. 3c). The 11.7% reduction in FE_Ctot compared with pure inlet is likely due to the limited CO₂ supply for the reactions⁴⁰, which is more evident under larger cell voltages. With the inlet of simulated flue gas of 15% CO₂, TfCOF-In₁@Cu₂O still presents a high conversion ability of CO₂, and the FE_Ctot is over 60% at E_cell between 2.8 V and 3.6 V (Figs. 3d and S18b).

To reveal the effect of the TfCOF layer on the localized concentrations and coverages of CO and CO₂, Multiphysics simulations coupled with MD calculations are employed with considerations of the mass transport, buffer reactions, and electrochemical reactions in the near-surface region of the catalyst (Fig. 3e, f)^41,42. The diffusion coefficients of CO and CO₂ in the catholyte with diffusion channels are calculated from Mean Squared Displacement functions from MD calculations (Fig. S21), which are the inlet parameters used for Multiphysics simulations.

It can be seen that the local concentrations of CO and CO₂ around the catalyst are much higher after the modification by TfCOF (Figs. 3f and S22). The coverage of adsorbed CO (*CO) varies from 0.06 to 0.28 after adopting TfCOF layer, which shows 4.7 times promotion. This high coverage of *CO provides sufficient *CO for further reduction, which accounts for the higher FE_C2+ on TfCOF-In₁@Cu₂O than that on In₁@Cu₂O. The addition of such TfCOF also facilitates the coverage of *CO₂, swinging from 0.22 to 0.31, which provides more carbon sources for CO₂ electrolysis on the catalytic sites⁴³. The simulated results reveal that the existence of TfCOF layer could highly increase the coverage and localized concentration of CO and CO₂^44,45. The enrichment of these species around active sites can realize the further reduction reactions toward valuable C₂₊ products.

Besides Multiphysics numerical simulation, the calculated CO₂ distribution probability using MD simulations is also shown in the middle inset of Fig. 3a. The red dots in the pores of TfCOF indicate the distribution possibilities of CO₂ accumulation region. High possibilities occur at the corners of nanopores in TfCOF, which reflect the special steric confinement effect on CO₂ by TfCOF⁴⁶. TfCOF holds a large amount of CO₂ with ample adsorption sites (Table S1), ascribing to the high electrostatic interaction of CO₂. Such TfCOF additive layer further breaks the solubility limitations from the low pCO₂ and promotes the local concentration of CO₂ for the reactions, which further improves the tolerance of lower inlet concentrations of CO₂.

High yield and stability with impure CO₂ inlets

The synthesized TfCOF-In₁@Cu₂O not only presents a high selectivity for C₂₊ products but also exhibits improved current densities and yields. As shown in Fig. 4a, the total current density (J) for TfCOF-In₁@Cu₂O exceeds 1400 mA/cm² at E_cell of 3.8 V with the inlet of pure CO₂ in the flow cell. FE_Ctot is also over 87% at such a high current, which suggests a pronounced catalytic activity and CO₂ utilization. With the inlets of dilute 90% and 80% CO₂, the total current densities remain high at E_cell of 3.8 V, which are 1355 mA/cm² and 1245 mA/cm², respectively. TfCOF-In₁@Cu₂O also presents a total current density of 200 mA/cm² at E_cell of 3.4 V for the direct conversion of simulated flue gas with 15% CO₂. Meanwhile, the maximum selectivity of C₂H₄ reaches 72.9 ± 2.2% at E_cell of 3.4 V, with FE_C2+ of 83.5% and FE_Ctot of 95.5% when using the 80% CO₂ inlet. Likewise, we observe that TfCOF-In₁@Cu₂O electrode delivers a superior partial current density toward C₂₊ products (J_C2+) of 837 mA/cm² at 3.4 V with pure CO₂ inlet, while it remains 768 mA/cm² with 80% impure CO₂ inlet (Fig. 4b), showing only 9% decrease in the production rate with 20% decrease of CO₂ concentration at such a high current density.

Fig. 4: Continuous upgrade of dilute CO₂ by TfCOF-In₁@Cu₂O in the flow cell.

a Total current densities (J) of TfCOF-In₁@Cu₂O with different CO₂ concentrations. b Partial current densities toward C₂₊ products under different potentials and CO₂ concentrations. c Yield of CO, CH₄, C₂H₅OH, and C₂H₄ with the inlet of 80% CO₂. The yield for C₂H₆ is not shown in the figures due to the low selectivity (FE < 3%). d Total current density and partial current density for C₂₊ and C₁ products with different adding amounts of TfCOF. e Stability measurements of TfCOF-In₁@Cu₂O at E_cell of 3.4 V with inlet of 80% CO₂. f Photograph of the electrolyzer stack. g Total current (J_Ttot) and partial current for C₂₊ (J_TC2+) and C₁ (J_TC1) products with stack potentials in the electrolyzer stack with inlet of 80% and 100% CO₂. The error bars represent the standard deviation from the measurement of three independent electrodes. Source data are provided as a Source Data file.

As depicted in Fig. 4c, we realize a total yield for carbon products of 3.45 mmol/h/cm² at E_cell of 3.6 V with the impure inlet of 80% CO₂, and the C₂H₄ yield is 2.36 mmol/h/cm². It is worth mentioning that the separations of C₂H₄ from CO and CH₄ are comparatively mature industrial processes, which makes the TfCOF-In₁@Cu₂O catalyst a promising material for CO₂ electrolysis. The efficient conversion of dilute CO₂ is ascribed to the simultaneous steric confinement effect and electronic effects of TfCOF on CO₂ and CO as discussed previously. Specifically, dilute CO₂ molecules are concentrated by TfCOF with ample adsorption sites⁴⁷, and the adsorption energies are revealed by the DFT simulations (Figs. S23 and S24).

Different amounts of TfCOF additives are also conducted with weight ratios of 0.1, 0.2, and 0.3 relative to the In-Cu gel (Figs. S25 and S26). The amount of TfCOF is critical to ensure the high yield of the C₂₊ products and the optimized ratio of TfCOF is around 0.2. The morphologies of different additions of TfCOF remain similar to the fine particles after the in-situ activations (Figs. S27 and S28). The J_C2+ of 698 mA/cm² with 80% dilute CO₂ inlet is realized on the catalyst with TfCOF(0.3), showing around 10% reduction compared with that of TfCOF-In₁@Cu₂O. On the other hand, TfCOF(0.1)-In₁@Cu₂O shows 18.5% decrease of J_C2+ compared with that of TfCOF-In₁@Cu₂O (Figs. 4d and S29). This could be due to electron conductivity loss at higher TfCOF loadings and reduced CO/CO₂ concentrating effects at lower loadings.

The stability of TfCOF-In₁@Cu₂O is also measured with 80% CO₂ inlet. A current density of 760 mA/cm² is well maintained for 96 h at E_cell of 3.4 V with FE_Ctot over 95% and FE_C2+ over 80% during continuous operation (Fig. 4e). The high yield of C₂H₄ is around 2.1 mmol/h/cm², which is about 200 mmol after a four-day electrochemical CO₂ upgrade with 1 cm² electrode. It can be found that the morphology (Fig. S30) and electronic structures (Fig. S31) of TfCOF-In₁@Cu₂O are well-maintained during the 96-h stability testing (Fig. S32). Such high yield and stability performance with 80% CO₂ inlet surpasses lots of reports with pure CO₂ inlet (Table S2).

The single-pass conversion efficiency consistently surpasses 25% starting from an applied voltage of 3.0 V when using simulated flue gas with 15% CO₂ feed concentration (Fig. S33), demonstrating promising prospects for industrial application. As the partial pressure of feed CO₂ decreases to 70% and 15%, the total carbon conversion rate exhibits a moderate decline. Consequently, under a feed condition of 15% CO₂, the energy conversion efficiency drops to approximately 20% at 3.2 V with a current density of 200 mA/cm². Both the single-pass conversion efficiency and energy efficiency surpass the reported values under the same conditions (the same current density and CO₂ inlet rate), demonstrating significant competitiveness^3,48,49.

With the target of potential industrial applications, an electrolyzer stack is further assembled as shown in Fig. 4f, which consistents of four 10 × 10 cm² modules connected in series^3,50,51. The maximum production rate of C₂₊ products is over 770 mmol/h with an inlet of 80% CO₂, corresponding to a total current of 81.7 A (Fig. 4g). With the utilization of pure CO₂ inlet, the maximum total current of 95.9 A is achieved with the C₂₊ production rate of around 800 mmol/h. Therefore, such a dual strategy of mass transport regulations and reaction kinetics promotions is scalable for potential industrial-level applications.

Underling mechanism

To evaluate the catalytic site of a single atomic In dopant combined with monovalent Cu(I) on CO formation, we firstly set up four surface models by the periodic DFT computations: Cu (111), Cu₂O(111), In₁@Cu (111) and In₁@Cu₂O (111) as shown in Figs. S34–S37, respectively. The atomic coordinates of the optimized computational models are shown in Supplementary Data 1 file. The reaction energies are depicted in Fig. 5a for the pathways of CO production^52,53,54,55. Among the four surface models, In₁@Cu₂O (111) shows the lowest Gibbs free energy for COOH* formation, indicating the promoting effect from single atomic In doping in Cu₂O. The reaction energy for COOH formation and CO desorption steps are all lower on In₁@Cu₂O (111) facets compared with those on pure Cu₂O (111). This indicates that the observed performance improvements stem from the relatively stronger COOH adsorption and weaker CO adsorption by In-induced electronic modulation of Cu₂O surface.

Fig. 5: Mechanism analysis and calculations.

a Relative Gibbs free energy (ΔG) profiles of the CO₂-to-CO reaction on Cu(111), Cu₂O (111), In₁@Cu (111), and In₁@Cu₂O (111) facets. b Crystal orbital Hamiltonian population (COHP) analysis between COOH with In₁@Cu₂O(111) substrate and Cu(111) substrate. The dashed line in pink and green color indicates the weighted average location of the antibonding level. c Schematic diagram of COHP analysis. DOS is density of states. d C⋯F electronic effect for CO, and the molecular electrostatic potential map of CO interacted with TfCOF. Inset: representative snapshots showing the pure TfCOF and close interactions of CO-TfCOF, respectively. Cyan, red, blue, white, and pink sticks: C, O, N, H, and F atoms. e Radial distribution function (RDF) and coordination numbers (CN) of CO and CO₂ with the distance from TfCOF. Inset is the representative snapshot showing close interactions of TfCOF-CO. An electric field of 5 V/nm was added in the Z-direction. f CO and CO₂ distributions in the lattice implemented in the MD simulation considering reactive surface (yellow), K⁺ (blue balls), H₂O (red dots), HCO₃⁻ (short-stick structures, red, white, cyan: O, H, C) and TfCOF (long-stick structures, blue, white, and pink sticks: N, H, and F). The element color of (e) is the same as (f). g The density distribution of CO along the Z-direction of the modeling lattice. h, i In-situ FTIR spectrum of TfCOF-In₁@Cu₂O for CO₂RR in different wavenumber regions. Source data are provided as a Source Data file.

To better understand the intrinsic mechanism for the enhanced activity for CO formation on In₁@Cu₂O, the projected density of states of the C atom in the key COOH* adsorbate and corresponding surface Cu atoms in Cu(111), In₁@Cu (111), and In₁@Cu₂O (111) were calculated (Fig. S38). The electron density and the wave function are decomposed to reveal the atomic orbital contributions^56,57. The overlaps between the C-2p and Cu-3d states on In₁@Cu₂O are the largest among all the models, which shows the strong interaction between the C and Cu atoms⁵⁸.

The crystal orbital Hamiltonian population (COHP) analysis was further performed. The weighted average locations of antibonding levels were also identified in this work. We found that a higher antibonding orbital level ((d-π*)*) relative to the Fermi level was delivered for In₁@Cu₂O (111) than Cu substrate after the interaction of the frontier orbitals with COOH. Additionally, the calculated -ICOHP values reveal significantly stronger bonding between COOH and the In₁@Cu₂O (111) substrate (–ICOHP = 2.156, Fig. 5b) compared to that between COOH and Cu surface (–ICOHP = 1.943, Fig. 5b).

We also calculated the C–C coupling step to produce *OCCO intermediate from two *CO (Figs. S39 and S40)⁴⁸. We found that the reaction energy for *OCCO formation is lower on In₁@Cu₂O (111) facets compared with that on the pure Cu (111) surface. This indicates that the observed performance improvements stem from both the relatively stronger COOH adsorption and *OCCO intermediate stabilization by In₁@Cu₂O (111) facets.

Hence, such a strategic combination of foreign In atomic In atoms doped in oxidized monovalent Cu(I) leads to the highest selectivity towards CO by holding specific electronics and features that are rather different from their counterpart metals. The proposed mechanism is also depicted by the schematic of Fig. 5c. Incorporating single atomic In into Cu₂O results in a larger energy split of hybrid orbitals when incorporated with the *COOH^34,59. A larger –ICOHP results in stronger interactions between substrate and molecules.

TfCOF possesses unique 3D porous structures, which can availably regulate the mass transport and enrichments of CO and CO₂ near the catalyst surface. The electrostatic potential of the F atom becomes larger after interacting with CO (Fig. 5d), which is attributed to the electron density transfer from F atom of TfCOF to C atom in CO. Hence, the negative regions in the molecular electrostatic potential maps extend to the oxygen atom of CO. The distance of C and F atoms is found to be 2.25 Å (Figs. 5e and S41). Such C···F interactions make the TfCOF become CO diffusion channels near the catalytic surface. As revealed, the strong electrostatic interaction between CO and TfCOF is further confirmed by the radial distribution function (RDF) and coordination numbers (CN) around TfCOF.

The distributions of CO and CO₂ are shown in the lattice for the MD simulation (Figs. 5e, f and S42). The random TfCOF molecules are gathering around 5 nm away from the catalytic surface. This assembled structure is likely causing the existence of multiple layers of CO and CO₂ between the TfCOF layer and catalytic surface due to different electrostatic interactions, which results in the analogous normal density distribution of it along the Z-direction (Fig. 5f). The top layers of CO (5–8 nm) are enriched by TfCOF layer ascribed to the stronger van der Waals’ interactions, while the bottom layers of CO with moderate interactions are close to the catalytic sites (around 1 nm in the simulation lattice) and can be released from TfCOF to the catalytic surface for further reduction. This is the proposed mechanism for the simultaneous capturing and releasing of CO and CO₂ molecules by TfCOF with the C$\cdots$F electronic effect and steric effect, respectively.

Such mass transport regulation effects are further depicted by the experimental confirmations with in-situ electrochemical attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy to monitor the intermediates of In₁@Cu₂O/COF at various potentials (Fig. 5h, j)^60,61,62. The important intermediates for the generation of C₂₊ products were observed, including the CO* (at 2065 cm⁻¹)⁶³, and CO bound to the copper surface (at 1967, 1919 cm⁻¹)⁶⁴. The intensity of these peaks increased with the cathode potentials, indicating the high coverage of *CO species. The enhancement in *CO adsorption promoted the subsequent C−C coupling. Moreover, the characteristic peak at 1541, 1423, and 1396 cm⁻¹ corresponds to COOH* and *COO⁻ (two O coordinated), respectively^17,65,66. The *H₂O peak is observed at 1649 cm⁻¹ (Fig. 5i)⁶³. The peaks at 1743, 1554, and 1519 cm⁻¹ are attributed to CO₃²⁻ species⁶⁷. The pronounced peak at 2344 cm⁻¹ is assigned to the CO₂* (Fig. S43), demonstrating the adsorption of CO₂ on TfCOF-In₁@Cu₂O. These findings further demonstrate the effective regulations of CO and CO₂ species near the catalytic sites by the TfCOF.

In short summary, the C₂₊ product formation is enhanced by the combined effect of In doping and TfCOF. Besides, the high tolerance of CO₂ concentration during CO₂RR is delivered by the TfCOF-mediated confinement effects. In detail, (1) the role of In doping: stronger COOH adsorption and *OCCO intermediate stabilization are obtained by In-induced electronic modulation of Cu₂O surface (Fig. 5a–c); (2) the role of TfCOF: the optimized amounts of COF functionalized by the trifluoromethyl group act as the local CO₂/CO diffusion channels via steric confinement effects and C···F electronic effects (Fig. 5d–g).

Discussion

Efficient electrolysis of dilute CO₂ (70–90% and 15%) toward C₂₊ products is realized through employing combined TfCOF and single-atom active sites of In₁@Cu₂O. Supported by the multi-scale simulation and experimental characterizations, TfCOF is proved to be the local diffusion channels of CO₂/CO and optimized amounts of TfCOF provide the proper concentrating of these molecules. In₁@Cu₂O plays the role of promoting the production of CO key intermediate for C–C coupling. Hence, the maximum FE_C2+ of 83.5% and FE_Ctot of 95.5% are obtained at E_cell of 3.4 V when using the 80% CO₂ inlet. The high C₂H₄ yield of 2.1 mmol/h/cm² is delivered for four days with an impure CO₂ inlet. Such a dual strategy is further scalable to the electrolyzer stack with a high yield of C₂₊ products of over 770 mmol/h. The primary contributions of this work are: (1) a high tolerance of CO₂ inlet concentrations (15%–100%) is realized; (2) steric confinement effects and C···F electronic effects of the Tf Group are implemented; (3) CO₂ concentrating and converting processes are coupled together.

Methods

Chemicals

Copper(II) chloride dihydrate (CuCl₂·2H₂O, metal basis >99.99%), indium nitrate hydrate (In(NO₃)₃·5H₂O, metal basis >99.99%), ethanol (>99.7%), propylene oxide (>99.5%), and isopropanol (> 99.7%) were purchased from Aladdin. Nafion solution (5 wt.%) was purchased from Macklin. Nafion 211 cation-exchange membranes were obtained from Dupont. All of the chemicals were used without further purification. The ultrapure water was obtained from the Milli-Q system.

Synthetic procedures

Preparation of TfCOF: 1 mmol 1,3,5-triformyl phloroglucinol (Tp, 210 mg) and 1 mmol 2,2′-Bis(trifluoromethyl)benzidine (BTBD, 324 mg) were added to a Pyrex tube containing 10 mL mixed solvent (dioxane/mesitylene = 1:1, volume ratio) with 1 mL acetic acid (6 M), which was then sonicated for 20 min. The mixture was flash-frozen under liquid nitrogen. After that, the mixture was sealed to the oven and dried at 120 °C for 3 days. The resultant precipitate was collected by filtration and washed with acetone, dichloromethane, and tetrahydrofuran, respectively. The precipitates were then heated at 120 °C for 24 h in a vacuum oven to get the material, which was named TfCOF.

Preparation of In-Cu gel: The CuCl₂·2H₂O (3.0 mmol) and In(NO₃)₃·5H₂O (0.2 mmol) were dissolved in 2 ml isopropanol. Under consistent bath sonication, 2 ml propylene oxide was added to the resulting solution and mixed for 20 s. Then, 0.2 ml H₂O was added and sonicated continuously until a homogeneous grey-green gel was formed. The above grey-green gel was aged for 24 h and eventually dried at 60 °C, lasting 24 h in a vacuum oven.

Preparation of TfCOF-In₁@Cu₂O electrode: First, the above In-Cu gel was formulated into ink solution in the usual way. In a typically prepared procedure of the working electrode, 5 mg of the dried In-Cu gel was mixed with TfCOF (In-Cu gel: TfCOF = 5:1, weight ratio), 1 ml water-isopropanol solution with a volume ratio of 7:3 and 80 μl Nafion solution (5 wt%) to prepare an ink solution. Then, the pristine electrode was obtained by loading the ink solution onto Sigracet 29 BC carbon paper of 1.0 × 1.0 cm and dried in an infrared drying oven. Finally, under CO₂RR conditions, the TfCOF-In₁@Cu₂O electrode was in-situ electrochemically activated in a CO₂-saturated 0.5 M KHCO₃ electrolyte at a constant potential of −3.14 V vs. RHE for 30 min (E(RHE) = E(SCE) + 0.235 + 0.0591 × pH). TfCOF-In₁@Cu₂O(0.1) and TfCOF-In₁@Cu₂O(0.3) electrodes were synthesized as described above, but with different additions of TfCOF (weight ratios of In-Cu gel:TfCOF = 1:0.124, In-Cu gel: TfCOF = 1:0.372, respectively).

Preparation of In₁Cu₂O/Cu electrode and Cu nanoparticles (Cu NPs): In₁Cu₂O/Cu electrode and Cu nanoparticles were synthesized in a similar way to that used for the TfCOF-In₁@Cu₂O catalyst, except that TfCOF powder and In(NO₃)₃·5H₂O were not added.

Electrochemical measurements

Electrochemical CO₂ reduction was conducted in a customized flow cell with a Nafion 211 membrane to separate the two compartments, preventing the reoxidation of CO₂RR products at the anode^3,40. Different inlets of pure and impure CO₂ were continuously bubbled into the catholyte bottle to saturate 0.5 M KHCO₃ solution (pH value is 7.2 ± 0.1). N₂ is used to balance and simulate the impure CO₂ from fuel gas^9,10. The flow rate of gas was controlled at 60 ml/min by a mass flow meter. The anolyte was also 0.5 M KHCO₃ solution. All electrochemical experiments were tested using a Biologic VMP-300 Potentiostat with a current booster. The cell voltage of E_cell is reported without iR correction. The current densities were normalized to the geometric surface area in this work. The partial current densities were equal to the total current densities multiplied by the Faradaic efficiency. Three independent measurements were performed under ambient pressure and room temperature (25 °C).

The electrolyzer stack was assembled by four module units with an electrode area of 10 × 10 cm². Two 1 L electrolyte bottles were employed for 0.5 M KHCO₃ solutions as the anolyte and catholyte. The inlet was separated into four modules in parallel, and the outlets were collected in the bottom flow channels.

Products analysis

The gaseous products were quantified by the online gas chromatograph (GC, GC-2014 Shimadzu) with a packed molecular sieve column. The online GC is equipped with a flame ionization detector (FID 1) for CO and CH₄, a flame ionization detector (FID 2) for C₂H₄, C₂H₆, and C₂H₂, as well as a thermal conductivity detector for H₂. The flow rates of the gaseous outlet are quantified by an additional flow-meter between the catholyte bottle and online GC. Liquid products in the catholyte were analyzed on a nuclear magnetic resonance spectroscopy (400 MHz ¹H liquid NMR spectrometer, Bruker Avance NEO) with the water suppression method. 1 ml of product-containing electrolyte was collected from the cathodic compartment after electrolysis. It was mixed with an internal standard of dimethyl formamide (Sigma-Aldrich, > 99.9%) and D₂O (Sigma-Aldrich, 99.9%), and then sealed for NMR measurements.

The Faradaic efficiency (FE) can be calculated as the following equation:

$${{\rm{FE}}}=\frac{{znF}}{Q}$$

(1)

where, $z$ is the number of electrons exchanged in the reaction, $n$ is the number of moles of product formed, $F$ is the Faraday constant, and $Q$ is the total amount of charges passed through the CO₂RR process. The partial current density was then obtained by multiplying the Faradaic efficiency with the average current density.

Characterizations

The XRD was performed to disclose the crystal structure of the as-prepared materials by Bruker/D8 ADVANCE X-ray diffractometer. Scanning electron microscope (SEM, ZEISS Sigma 500) and transmission electron microscopy (TEM, FEI Talos F200X) were applied to observe the morphologies and microstructures of materials and electrodes. The FEI Talos F200X TEM combines high-resolution STEM and TEM imaging with energy-dispersive X-ray spectroscopy (EDS) signal detection and 3D chemical characterization with compositional mapping. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) characterization was carried out on a Titan Cubed Themis G2 300 thermal field emission microscope with a high-tension voltage range of 60–300 kV.

XANES and EXAFS spectra of Cu K-edge and In K-edge were acquired on HXMA and BioXAS beamlines at Canadian Light Source (CLS), respectively. The results were analyzed via Athena and Artemis software. Solid-state ¹³C NMR experiments were recorded on an AVANCE NEO 600 MHz spectrometer (Bruker BioSpin Corp.) at the resonance frequency of 150.72 MHz for 13 C using the cross-polarization (CP), magic-angle spinning (MAS), and a high-power ¹H decoupling. The powder samples were placed in a pencil-type zirconia rotor of 4.0 mm o.d. The spectra were obtained at a spinning speed of 10 kHz (4.2 μs 90° pulses), a 2 ms CP pulse, and a recycle delay of 3 s. The C signal of tetramethylsilane at 0 ppm was used as the reference of 13 C chemical shift.

Time-of-flight secondary ion mass spectroscopy (TOF-SIMS) was employed to study the elemental composition of various catalysts. All depth profiles were acquired in non-interlaced mode, namely, sequential sputtering. A 50 × 80 µm² area centred on the analysis area was sputtered using a Cs⁺ ion beam (1 KeV ion energy, ~40 nA measured sample current).

Details of the operando, in-situ, and quasi in-situ measurements

The operando synchrotron two-dimensional wide-angle X-ray scattering (2D-WAXS) images were performed on the High Energy Wiggler Beamline in the Brockhouse Diffraction Sector (BXDS-WHE) at CLS, Canada.

Thermo Fisher Nicolet IS50 spectrometer was employed for in-situ electrochemical attenuated total reflection Fourier-transform infrared (ATR-FTIR) experiments. The Si prism with an incidence of a 60° angle was assembled into the bottom of the three-electrode in-situ electrochemical cell.

Quasi in-situ characterizations were conducted to explore the information on elemental composition and oxidation states, which were collected by a near-ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) located at the Vacuum Interconnected Nanotech Workstation (Nano-X), Suzhou Institute of Nano-Tech, and Nano-Bionics, Chinese Academy of Sciences.

DFT calculations

We used the Vienna Ab-initio Simulation Package⁶⁸ with the projector augmented wave method⁶⁹ for all DFT calculations. The Perdew-Burke-Ernzerhof (PBE) functional⁷⁰ combined with DFT-D3 correction⁷¹ was employed to treat the exchange-functional for describing the weak interactions between atoms. During the calculations, the plane-wave basis had a cut-off energy of 420 eV. The Brillouin zone integration was conducted with a 2 × 2 × 1 Monkhorst-Pack k-points for optimizing the geometry and lattice size. We performed structural relaxation until the electronic structure self-consistency loop reached 1 × 10⁻⁴ eV. Partial occupancies were permitted for the Kohn-Sham orbitals by using the Gaussian smearing method and a width of 0.1 eV. The energy change in a geometry optimization was considered convergent when it was less than 0.01 eV Å⁻¹. The implicit dipole and solvent corrections^72,73 were also taken into account in this work. For each elementary step, Gibbs reaction free energy ΔG is the difference of free energies between the initial and final states and is defined as follows:

$$\Delta {{\rm{G}}}=\Delta {{\rm{E}}}+\Delta {{\rm{ZPE}}}-{{\rm{T}}}\Delta {{\rm{S}}}$$

(2)

where the reaction energies of reactants calculated by DFT correspond to ΔE, ΔZPE, and ΔS are the zero-point energy and entropy changes due to the reaction. The definitions of adsorption energy (E_ads) for the adsorbate are depicted in the reported literature^44,72,74.

Molecular dynamics simulations

The all-atom MD simulations were performed using the Gromacs-4.6.7 software package⁷⁵, with a general AMBER force field⁷⁶ featuring RESP charges⁷⁷. The system was in a relaxed liquid configuration at 298 K, and the equilibrium MD simulation ran for a total of 10 ns NPT. We used the relaxed system as a starting configuration and performed energy minimization with a composite protocol of steepest descent, using 100 kJ/mol·nm termination gradients, prior to system relaxation MD. To maintain the equilibrium temperature at 298 K, the Nose´-Hoover thermostat⁷⁸ was employed, while periodic boundary conditions were enforced on all three dimensions. The Particle Mesh-Ewald method^79,80 was utilized to calculate long-range electrostatics with a relative tolerance of 1 × 10⁻⁶, and a 1 nm cut-off distance was applied to real-space Ewald interactions and van der Waals interactions. The bond lengths of hydrogen atoms were constrained using the LINCS algorithm⁸¹, and we used a leap-frog algorithm⁸² with a time step of 1 fs. Meanwhile, an electric field of 5 V/nm was added in the Z-direction.