Interface engineering of single-molecular heterojunction catalysts for CO₂ electroreduction in strong acid medium

Abstract

Electrochemical carbon dioxide reduction reaction (CO₂RR) under strongly acidic conditions enables high CO₂ utilization. However, especially in proton exchange membrane (PEM) electrode assembly reactors, achieving selective CO₂RR in such environments remains challenging due to uncontrolled interfacial water diffusion at high current densities. Here, we develop a nickel-based heterogeneous molecular electrocatalyst (NiPc-NH₂/CNT-SHP) featuring amino (-NH₂) functional groups and grafted long-chain hydrophobic molecules. Under acidic conditions, -NH₂ is in situ protonated to form amino cations (-NH₃⁺). The positively charged -NH₃⁺ groups and hydrophobic molecules effectively disrupt the protonated water (H₃O⁺)-rich network, inhibiting the invasion of H₃O⁺ and thereby suppressing the hydrogen evolution reaction, while enhancing selectivity for acidic CO₂RR. The catalyst achieves nearly 100% Faradaic efficiency for CO at current densities from 50 to 400 mA cm⁻², with approximately 76% CO₂ utilization efficiency in a flow cell, and sustains over 80% selectivity for more than 200 h in a self-designed PEM–porous solid electrolyte reactor. These findings highlight interfacial water management as a key design principle for efficient acidic CO₂ electroreduction.

Introduction

Converting CO₂ into high-value-added chemicals and fuels through the electrocatalytic carbon dioxide reduction reaction (CO₂RR) is a promising approach for producing valuable fuels and addressing carbon emissions^1,2. CO₂RR systems predominantly employ alkaline or neutral electrolytes due to the interfacial alkaline environment, which can promote the conversion of CO₂ molecules into *COOH (adsorbed carboxyl intermediate) and suppress the competing hydrogen evolution reaction (HER)^3,4,5. However, the alkaline solution exhibits a low CO₂ utilization efficiency, with theoretically 75% of the input CO₂ reacting with hydroxide (OH⁻) to form carbonate (CO₃²⁻), which then enters the anode through the anion exchange membrane (AEM), resulting in CO₂ loss^5,6,7,8. Additionally, the continuous influx of CO₂ lowers the electrolyte pH, disrupting the catalyst interface microenvironment and compromising both membrane integrity and catalyst stability⁹.

In response to this dilemma, increasing attention has shifted toward electrocatalytic CO₂ reduction in acidic media, which offers significant advantages for efficient CO₂ conversion. In acidic systems (pH <4), CO₃²⁻ or HCO₃⁻ (bicarbonate) formed at the electrode interface can rapidly reconvert to CO₂ upon diffusing into the bulk electrolyte, thereby minimizing the loss of CO₂ and reduction products^{9,10,11,12,13}. However, the high-concentration acidic environment poses significant challenges to catalysts and the efficiency of electrocatalytic CO₂ reduction, especially in acidic PEM reactors. Most metal catalysts and metal oxides are primarily susceptible to corrosion in acidic solutions, leading to catalyst deactivation^13,14,15,16. Additionally, the high concentration of protons (H⁺) intensifies HER competition^11,17,18, resulting in poor Faradaic efficiency (FE) of CO₂RR at high current densities. Therefore, strategies that enhance CO₂RR activity while suppressing HER remain critical for practical applications.

Recently, numerous studies have demonstrated that alkali metal cations are attracted to the negatively charged cathode field, resulting in their accumulation in the cathode capacitor layer and the formation of cation shielding. The formed OH⁻ though HER or CO₂RR will be preserved by cation shielding. Therefore, the local cathode environments, initially rich in a protonated water (H₃O⁺) network, can be transformed into an alkaline environment^19,20, effectively mitigating the competing HER in an acidic medium and thus improving the poor Faradaic efficiency of CO₂RR^21,22,23. Additionally, CO₃²⁻ or HCO₃⁻ species formed at the cathode surface diffuse into the electrolyte, where they react with acid to produce CO₂ molecules that subsequently return to the catalyst interface, maintaining a high carbon utilization efficiency. However, due to the strong Coulomb attraction to H₃O⁺ at more negative potentials²⁴, the catalyst interface makes it challenging to resist the invasion of harmful water species. A significant amount of H₃O⁺ is rapidly introduced into the electric double layer, which disrupts the alkaline environment of the double layer on the cathodic catalyst, thereby facilitating HER. Therefore, the high FE of acidic CO₂RR is limited to a small current window, as HER increasingly dominates at high current densities, limiting practical applications.

Herein, we develop a superhydrophobic heterogeneous molecular catalyst (NiPc-NH₂/CNT-SHP) featuring amino (-NH₂) functional groups and long-chain hydrophobic molecules (octadecylamine) grafted on multi-walled carbon nanotubes (CNT). Under acidic conditions, -NH₂ undergoes in-situ protonation to form amino cations (-NH₃⁺), which could repel H₃O⁺ via Coulombic repulsion, effectively suppressing HER activity. Simultaneously, the long-chain hydrophobic molecules modulate the interfacial water network, optimizing the ratio of free water to hinder H₂O reduction and facilitate CO₂RR under strongly acidic conditions (pH = 1). Therefore, the NiPc-NH₂/CNT-SHP achieves a high CO Faraday efficiency (FE_CO) of nearly 100% across a broad current density window of 50 to 400 mA cm^–2, with a carbon monoxide (CO) partial current density of 426 ± 14 mA cm^–2, and high single-pass CO₂ conversion efficiency in the flow cell. Thanks to the stable hydrophobic and amino groups, a proton-exchange membrane (PEM)-based porous solid electrolyte (PEM-PSE) reactor was developed to enable continuous, selective (> 80%), and stable (> 200 h) CO production using NiPc-NH₂/CNT-SHP as the cathode for practical applications.

Results

Catalyst design and preparation

To achieve high selective CO₂RR in the PEM reactor, the first step is to strategically design the catalytic site to suppress the competitive reaction of HER under high currents. We select a Ni-based molecular catalyst due to its well-defined Ni–N₄ coordination center, which imposes a high energy barrier for HER^25,26. Additionally, to mitigate H₃O⁺ intrusion at the catalyst interface, we aim to establish a built-in positive electric field that repels H₃O⁺ via Coulombic repulsion. Given that the protonation of amino (-NH₂) groups into amino cations (-NH₃⁺) under acidic conditions is a thermodynamically spontaneous process, we are inspired to employ amino-functionalized Ni phthalocyanine (NiPc-NH₂) as the catalytic unit. Under acidic conditions, -NH₂ spontaneously converts into -NH₃⁺, which is expected to effectively slow down the invasion of H₃O⁺ into the catalyst’s surface during acidic CO₂RR.

However, introducing a positively charged interface inherently increases affinity for H₂O (excluding H₃O⁺)²⁷, thereby promoting undesired H₂O reduction. To counteract this effect, we further engineer a hydrophobic interface by incorporating octadecylamine as a long-chain hydrophobic modifier, which modulates the interfacial water network and reduces H₂O reduction activity. Integrating these design strategies, (1) Coulombic repulsion via interfacial charge modulation and (2) hydrophobic surface engineering, are expected to achieve highly selective acidic CO₂RR at high current in the PEM reactor.

Synthesis and characterization of the NiPc-NH₂/CNT-SHP catalyst

Following this design strategy, we synthesized the superhydrophobic NiPc-NH₂/CNT-SHP catalyst. First, NiPc-NH₂, the molecular catalytic unit, was synthesized by hydrothermal and Soxhlet extraction (see Supplementary Methods, Supplementary Figs. S1, 2)^28,29). Commercial carboxyl-modified carbon nanotubes (CNT-COOH) were converted to acyl chloride-functionalized CNTs (CNT-COCl) through an acyl chlorination reaction under anhydrous and inert conditions. Subsequently, NiPc-NH₂ and octadecylamine were grafted onto CNT-COCl via nucleophilic substitution, forming the NiPc-NH₂/CNT-SHP catalyst (Supplementary Fig. S3, and Fig. 1a). For comparison, catalysts without hydrophobic modification (NiPc-NH₂/CNT) and without amino functionalization (NiPc@CNT) were also synthesized (see Supplementary methods).

Fig. 1: Synthesis and characterization of the NiPc-NH₂/CNT-SHP catalyst.

a Mechanism diagram of the NiPc-NH₂/CNT-SHP catalyst. b AC HAADF-STEM image of NiPc-NH₂/CNT-SHP, with the inset showing (c) contact angles of water on NiPc-NH₂/CNT-SHP. d High-resolution XPS N 1 s spectra of NiPc-NH₂, and NiPc-NH₂/CNT-SHP. e Experimental Ni K-edge XANES spectra and first derivative curves and (f) Fourier transformed magnitudes of the experimental Ni K-edge EXAFS signals of NiPc-NH₂/CNT-SHP, NiPc@CNT, and Ni foil. Source data are provided as a Source Data file.

To characterize the structural and morphological features of NiPc-NH₂/CNT-SHP, we conducted X-ray diffraction (XRD, Supplementary Fig. S4), infrared spectroscopy (Supplementary Figs. S5, S6), and ultraviolet–visible spectroscopy (UV-Vis, Supplementary Fig. S7), confirming the successful grafting of unordered NiPc-NH₂ and octadecylamine onto CNT. Field-emission scanning electron microscopy (SEM, Supplementary Fig. S8) and transmission electron microscopy (TEM, Supplementary Fig. S7) images revealed that NiPc-NH₂/CNT-SHP retains the characteristic one-dimensional morphology of CNT without the formation of aggregated NiPc-NH₂ crystals (Supplementary Fig. S9), indicating uniform Ni molecular site distribution. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM) further showed the presence of atomically dispersed Ni sites, as evidenced by the bright spots observed in Fig. 1b (yellow). Meanwhile, the water contact angle measurements found that the hydrophobicity of the interface changed, NiPc-NH₂/CNT-SHP exhibited a contact angle of ~151° (Fig. 1c), significantly higher than pristine CNT ( ~ 22°, Supplementary Fig. S9b) and NiPc-NH₂/CNT ( ~ 109°, Supplementary Fig. S9c), confirming that octadecylamine has successfully introduced a superhydrophobic interface. Elemental mapping analysis further demonstrated the uniform distribution of Ni catalytic sites (Supplementary Fig. S10) across the CNT. Moreover, inductively coupled plasma mass spectrometry (ICP-MS) determined the Ni content in NiPc-NH₂/CNT and NiPc-NH₂/CNT-SHP as 0.75 wt% and 0.73 wt%, respectively.

We performed X-ray photoelectron spectroscopy (XPS) to further elucidate the chemical states of elements in NiPc-NH₂/CNT-SHP (Fig. 1d, and Supplementary Figs. S11–13). Compared to NiPc-NH₂ (Fig. 1d) and NiPc/CNT (Supplementary Fig. S14), NiPc-NH₂/CNT (Supplementary Fig. S12) and NiPc-NH₂/CNT-SHP exhibited an additional peak at 401.70 eV, attributed to the –NH–(C = O)– species³⁰, confirming the successful grafting of both NiPc-NH₂ and hydrophobic functional groups onto CNT. The absence of shifts in Ni binding energy between NiPc-NH₂/CNT and NiPc-NH₂/CNT-SHP relative to NiPc-NH₂ (Supplementary Fig. S13) suggests that the central coordination structure remains intact after chemical grafting. To determine the geometric structure of the Ni sites in NiPc-NH₂/CNT-SHP, we conducted synchrotron X-ray absorption spectroscopy measurements (XAS). The X-ray absorption near-edge structure (XANES, Fig. 1e) revealed distinct spectral features, indicating differences in Ni coordination compared to the Ni foil and suggesting a higher Ni oxidation state. Furthermore, a weak pre-edge peak (peak A) and a pronounced 1 s → 4pz transition (peak B) were observed, indicative of high D_4h centro-symmetry³¹. The extended X-ray absorption fine structure (EXAFS) Fourier transform (FT) spectrum of NiPc-NH₂/CNT-SHP displayed a primary peak at 1.47 Å (Fig. 1f), corresponding to Ni–N scattering in the first coordination shell, and well below the Ni–Ni bond distance (2.17 Å), illustrating that the absence of Ni clusters or nanoparticle aggregation. Figure 1e, f show that the XANES and EXAFS spectra of NiPc@CNT closely resemble those of NiPc-NH₂/CNT-SHP, indicating a similar local coordination environment around the Ni center. Furthermore, only a single intensity maximum at 5.9 Å⁻¹ was identified in NiPc-NH₂/CNT-SHP (Supplementary Fig. S15a), corresponding to Ni–N coordination. This characteristic, also observed in NiPc@CNT (6.2 Å⁻¹, Supplementary Fig. S15c), clearly distinguishes both samples from Ni foil (Supplementary Fig. S15b). These structural characterizations confirm the successful synthesis of NiPc-NH₂/CNT-SHP with hydrophobic groups and NiPc-NH₂ while preserving the well-defined Ni–N₄ structure sites, in line with our design strategy.

Electrochemical acidic CO₂RR performance in flow cell

Given that the structural characterization of the catalyst aligns with our initial design, we evaluated the electrochemical acid CO₂RR performance in the flow cell (Supplementary method). In a strong-acid environment (0.5 M K₂SO₄ + H₂SO₄ (pH = 1)), online gas chromatography analysis identified CO and hydrogen (H₂) as the only gas-phase products, while liquid-phase products were undetectable by ¹H Nuclear Magnetic Resonance (NMR) spectroscopy (Supplementary Fig. S17). The NiPc-NH₂/CNT-SHP, NiPc-NH₂/CNT, and NiPc@CNT catalysts exhibited current densities of 35 mA cm^–2 at –1.04, –1.04, and –1.13 V versus the reversible hydrogen electrode (vs. RHE), respectively, with 80% iR compensation (Fig. 2a), consistent with recent reports³². Furthermore, NiPc-NH₂/CNT-SHP achieved a current density of 500 mA cm^–2 at –1.60 V vs. RHE, which is lower than that of NiPc-NH₂/CNT (–1.81 V vs. RHE) and NiPc@CNT (–1.88 V vs. RHE), highlighting the structural advantages of NiPc-NH₂/CNT-SHP in optimizing catalytic activity.

Fig. 2: Electrochemical acidic CO₂RR performance in flow cell.

a j_total (80% iR compensation), b FE_CO, c j_CO (80% iR compensation), d CO production rate (80% iR compensation), e TOF_CO, f j_(CO)ECSA (80% iR compensation) of NiPc-NH₂/CNT-SHP, NiPc-NH₂/CNT and NiPc@CNT in 0.5 M K₂SO₄ + H₂SO₄ (pH = 1), Resistance: 4.2 ± 0.42 Ω. g Total current density and FE_CO of NiPc-NH₂/CNT-SHP in acidic CO₂RR test (80% iR compensation). h Acid CO₂RR utilization efficiency of NiPc-NH₂/CNT-SHP with different CO₂ flow rates. i Comparison of acidic CO₂RR performance of NiPc-NH₂/CNT-SHP in flow cell with other reported catalysts. The error bars represent the standard deviation of three independent measurements. Source data are provided as a Source Data file.

We further analyzed the relationship between current and FE_CO (Fig. 2b). NiPc-NH₂/CNT-SHP exhibited high CO selectivity (FE_CO > 98%) over a wide current density range (50–400 mA cm^–2), maintaining a FE_CO of 85.26% at 500 mA cm^–2, outperforming both NiPc-NH₂/CNT and NiPc@CNT. Interestingly, the FE_CO of NiPc@CNT declined significantly beyond 250 mA cm^–2, reaching 49.50% at 300 mA cm^–2 (Fig. 2b). However, when the test condition is alkaline (1 M KOH), NiPc@CNT maintained a high FE_CO of 95.84% at 350 mA cm⁻² (Supplementary Fig. S18b), comparable to NiPc-NH₂/CNT, indicating that the in situ generated -NH₃⁺ effectively mitigates interfacial acidity, thereby enhancing acidic CO₂RR selectivity at high current densities.

The introduction of interfacial hydrophobic molecules also optimizes the performance of Ni sites under acidic conditions (Fig. 2c and Supplementary Fig. S19). As seen in Fig. 2c, NiPc-NH₂/CNT-SHP exhibited a maximal j_CO of 426 mA cm^–2 at –1.60 V vs. RHE (Fig. 2c), with the production rate of 7.97 mmol cm^–2 h^–1 (Fig. 2d), which is better than that of NiPc-NH₂/CNT (385 mA cm^–2, 6.78 mmol cm^–2 h^–1), and NiPc@CNT (198 mA cm^–2, 4.61 mmol cm^–2 h^–1). Furthermore, NiPc-NH₂/CNT-SHP demonstrated the highest CO turnover frequency (TOF_CO) is 17.76 s^–1 (63936 h^–1, Fig. 2e) compared to NiPc-NH₂/CNT (14.70 s^–1) and NiPc@CNT (6.20 s^–1). The double-layer capacitance (C_dl) of NiPc-NH₂/CNT-SHP (2.68 mF cm^–2) was comparable to NiPc-NH₂/CNT (2.63 mF cm^–2) and NiPc@CNT (2.23 mF cm^–2) (Supplementary Fig. S20), enabling the CO partial current densities of NiPc-NH₂/CNT-SHP normalized by the electrochemical active surface area (ECSA) also to be higher than NiPc-NH₂/CNT and NiPc@CNT (Fig. 2f). Despite the stronger acidic environment (pH = 0.4), NiPc-NH₂/CNT-SHP is still able to achieve the maximal j_CO of 561 mA cm^-2 (Supplementary Fig. S21c), which is still higher to that of NiPc@CNT (234 mA cm^-2, Supplementary Fig. S22c). Notably, the absence of a superhydrophobic interface in NiPc-NH₂/CNT resulted in small liquid accumulations at the back of the electrode after prolonged operation (Supplementary Fig. S24), with the overflowing liquid exhibiting an alkaline pH (Supplementary Fig. S25), consistent with previous reports⁹. In addition, NiPc-NH₂/CNT-SHP also achieves the highest activity of CO₂RR in alkaline and neutral solution, showing the maximal j_CO of 390 mA cm^–2 at –0.70 V vs. RHE (1 M KOH, Supplementary Figs. S18a–c) and 414 mA cm^–2 at –1.13 V vs. RHE (1 M KHCO₃, Supplementary Figs. S18d–f), indicating the good activity of NiPc-NH₂/CNT-SHP.

Given the performance of NiPc-NH₂/CNT-SHP, we then focus on a long-term test under high current density. Due to the absence of hydrophobic functionalization, NiPc-NH₂/CNT exhibited poor stability, with continuous operation limited to less than 2 h at 100 mA cm^–2 (Supplementary Figs. S26, 27), during which FE_CO rapidly declined after 2 h. By contrast, NiPc-NH₂/CNT-SHP maintained stable operation at 150 mA cm⁻² for over 48 h, with a consistent potential of around–1.37 V vs. RHE and FE_CO exceeding 96% (Fig. 2g), along with good stability in neutral and alkaline conditions (Supplementary Figs. S28, S29).

Furthermore, NiPc-NH₂/CNT-SHP significantly improved CO₂ utilization efficiency at high current densities. At 200 mA cm⁻², FE_CO reached ~98.41% as the CO₂ flow rate decreased from 25 to 2 mL min^–1, achieving a high single-pass carbon efficiency (SPCE) of ~76% at a flow rate of 2 mL min^–1 (Fig. 2h). This performance under acidic conditions is further summarized in Supplementary Table S2 and Fig. 2i, where NiPc-NH₂/CNT-SHP demonstrates competitive FE_CO compared to previous reports in acidic environment^{19,21,24,32,33,34}.

Intrinsic activity analysis for CO₂RR performance

In order to verify the spontaneous formation and stability of amino cations under acidic conditions during electrochemical CO₂ reduction, we performed in situ XPS measurements in an acidic electrolyte (CO₂-saturated 0.5 M K₂SO₄ + H₂SO₄ (pH = 1)). Initially, the as-prepared NiPc-NH₂/CNT-SHP electrode showed no signal attributable to protonated amine species (Supplementary Fig. S30). However, upon the introduction of acidic electrolyte, a distinct peak emerged at approximately 402.1 eV under open-circuit potential, which can be assigned to the –NH₃⁺ species^35,36,37. This observation aligns well with our initial hypothesis. Furthermore, when applying potentials ranging from 0 to –1.6 V vs. RHE, the –NH₃⁺ species remained clearly visible, indicating that the amino cation is stably maintained in the strongly acidic environment and plays a role in modulating the interfacial environment during electrochemical CO₂ reduction.

Subsequently, in view of the above differences in selectivity, we performed in situ attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) in a CO₂-saturated 0.5 M K₂SO₄ + H₂SO₄ solution (pH = 1) to obtain detailed molecular vibration information at the catalyst interface³⁸. The peak at 1298 cm^–1 is attributed to the OH-deformation of the *COOH intermediate and interfacial bicarbonate species^39,40, while a strong peak at 1381 cm^–1 is contributed to the C − O stretching of *COOH intermediate and interface carbonate^41,42, where the *COOH intermediate is crucial for the electrochemical reduction of CO₂ to CO. The formation of (bi)carbonate originates from the alkaline environment at the catalyst interface³⁸. Furthermore, another important peek at 2064 cm^–1 belongs to the linearly bonded *CO (*CO_L) on the Ni atom of NiPc-NH₂/CNT-SHP^43,44, and its signal intensity enhances with the tested potential, in line with that of the *COOH intermediate (Supplementary Fig. S31), indicating that *CO_L and *COOH are essential intermediates for CO₂ electrochemical reduction to CO. Similar spectral features were observed for NiPc@CNT. However, NiPc@CNT displays a lower *CO_L wavenumber (~1948 cm^–1) compared with NiPc-NH₂/CNT-SHP, indicating stronger CO adsorption^39,43,44, which is likely due to the modification of the Ni electronic structure by different peripheral functional groups on the molecular catalyst⁴⁵, leading to variations in CO adsorption strength at different sites. On the other hand, NiPc@CNT exhibited a noticeable shift in H₂O-related peak positions with the tested potential in acidic conditions (Fig. 3b), whereas this was not observed under neutral conditions (Supplementary Fig. S32).

Fig. 3: In-situ infrared spectroscopy of interface species.

a In-situ infrared spectroscopy of NiPc-NH₂/CNT-SHP (without iR compensation). b In-situ infrared spectroscopy of NiPc@CNT (without iR compensation). ATR-SEIRAS of interfacial water on NiPc-NH₂/CNT-SHP (c) and NiPc@CNT (d) at −0.80 V in CO₂-saturated 0.5 M K₂SO₄ + H₂SO₄ (pH = 1). e Potential-dependent area ratio of protonated water / (H-bonded water + free water) on NiPc@CNT (cyan) and NiPc-NH₂/CNT-SHP (blue) in CO₂-saturated 0.5 M K₂SO₄ + H₂SO₄ (pH = 1, without iR compensation). f Potential-dependent area ratio of free water / H-bonded water on NiPc@CNT (cyan) and NiPc-NH₂/CNT-SHP (blue) in CO₂-saturated 0.5 M K₂SO₄ + H₂SO₄ (pH = 1, without iR compensation). The error bars represent the standard deviation of three independent measurements. Source data are provided as a Source Data file.

To elucidate this phenomenon, we analyzed the bending vibration of H–O–H (1500–1750 cm⁻¹), which was deconvoluted into three peaks: ~1620 cm⁻¹ (free water)^46,47, ~1680 cm⁻¹ (H-bonded water)^48,49, and ~1715 cm⁻¹ (protonated water, H₃O⁺) (Fig. 3c, d, and Supplementary Figs. S33, S34)⁵⁰. Compared to NiPc-NH₂/CNT-SHP (Fig. 3c), NiPc@CNT exhibited a significantly higher proportion of H₃O⁺ and a reduced fraction of free water (Fig. 3d). The area ratio of protonated water to the sum of free and H-bonded water was consistently lower for NiPc-NH₂/CNT-SHP than for NiPc@CNT and increased with applied potential (Fig. 3e). The cation-dependent variations in the H–O–H stretching region (3000–3800 cm⁻¹) are consistent with those observed in the bending region (1500–1750 cm⁻¹), as shown in Supplementary Fig. S35. At –1.6 V vs. RHE, NiPc@CNT displayed an area ratio approximately 2.3 times higher than NiPc-NH₂/CNT-SHP, indicating greater H₃O⁺ infiltration into the catalyst interface, thereby reducing FE_CO (Fig. 2b) and weakening *COOH signal intensity.

Moreover, NiPc-NH₂/CNT-SHP exhibited a significantly higher ratio of free water to H-bonded water compared to NiPc@CNT, reaching ~9.8 times at –1.6 V vs. RHE (Fig. 3f). In contrast, NiPc@CNT showed a decreasing trend with applied potential, likely due to H₃O⁺ competition (Fig. 3e). Conversely, NiPc-NH₂/CNT-SHP maintained a stable ratio value of ~1.0, suggesting that the long-chain hydrophobic groups effectively manage the interfacial water, disrupt the water network, and promote free water infiltration into the electric double layer, facilitating CO₂RR³⁵. This enhanced interfacial water management accelerates the kinetics of adsorbed CO₂ conversion³⁹, corroborating the superior catalytic activity of NiPc-NH₂/CNT-SHP compared to NiPc@CNT (Fig. 2c).

To validate the above findings, we further measured the polarization curves of a rotating disk electrode (RDE) coated with NiPc-NH₂/CNT-SHP (Fig. 4a), NiPc-NH₂/CNT, and NiPc@CNT (Supplementary Fig. S36) at different rotating speeds in an Ar-saturated 0.5 M K₂SO₄ + H₂SO₄ (pH = 1). The onset potentials for HER were similar across all catalysts. A current density plateau, indicative of HER from diffusion-limited hydronium reduction, was observed on all electrodes (Fig. 4a, and Supplementary Fig. S37). However, NiPc-NH₂/CNT-SHP exhibited a lower plateau current density than NiPc@CNT, likely due to attenuated H₃O⁺ mass transport to the cathode surface (Fig. 3d)²⁴. The linear fitting of the plateau current density (j_plateau) using the Levich equation (Supplementary Figs. S38, 39) reveals an approximately 40% decrease in the H₃O⁺ diffusion coefficient (D_H3O⁺) for NiPc-NH₂/CNT-SHP compared to NiPc@CNT. The kinetic current densities were subsequently calculated using the Koutecký–Levich equation, showing a ~ 2.5-fold decrease in HER activity from ~990 mA cm^-2 for NiPc@CNT to ~400 mA cm^-2 for NiPc-NH₂/CNT-SHP (Supplementary Fig. S38). In addition, linear fitting of the total current density (j_plateau) against ω^1/2 indicates that NiPc-NH₂/CNT-SHP exhibits slower interfacial water reduction kinetics than both NiPc@CNT and NiPc-NH₂/CNT. Collectively, the RDE results demonstrate that the presence of hydrophobic domains and –NH₃⁺ functionalities effectively suppress both hydronium and water reduction, thereby promoting selective CO₂ electroreduction in acidic media, in line with the trends observed in the in situ ATR-SEIRAS spectra.

Fig. 4: Mechanism analysis of interfacial water influence.

a LSV curves of NiPc-NH₂/CNT-SHP in Ar-saturated electrolyte of 0.5 M K₂SO₄ + H₂SO₄ (pH = 1) under varied rotating speed (without iR compensation). b A comparison of NiPc-NH₂/CNT-SHP, NiPc-NH₂/CNT, and NiPc/CNT at 1000 rpm (without iR compensation). c A typical model of NiPc-NH₂/-SHP for MD simulations, red: O, white: H, blue: N, black: C, yellow: S, light blue: Ni. d Radial distribution functions (g(r)) of H₃O⁺ surrounding the Ni center for NiPc, NiPc-NH₃⁺ and NiPc-NH₃⁺/-SHP. e g(r) of H₂O surrounding the Ni center for NiPc, NiPc-NH₃⁺ and NiPc-NH₃⁺/-SHP. f Schematic presentation of water species distribution on the surface of NiPc@CNT, NiPc-NH₂/CNT, and NiPc-NH₂/CNT-SHP.

Furthermore, for NiPc-NH₂/CNT-SHP at 1000 rpm, a significant increase in current density for H₂O reduction was observed at –1.50 V vs. RHE, whereas this process commenced at –1.38 V and –1.30 V vs. RHE for NiPc-NH₂/CNT and NiPc@CNT, respectively. This shift illustrates the suppressed H₂O reduction in NiPc-NH₂/CNT-SHP (Supplementary Fig. S39), supporting its high FE_CO under high current density conditions, which can be attributed to the designed hydrophobic interfacial environment. The hydrophobic structure also demonstrated high stability (Supplementary Fig. S40).

To simulate the microenvironment surrounding our catalysts, we conducted molecular dynamics (MD) simulations (see Methods for details). We examined the influence of different interfacial environments and varying concentrations of charged species (namely K⁺, H⁺) on the transport of H₂O. In the present case, we paid particular attention to the distribution of H₂O and H₃O⁺ relative to the Ni catalyst center (Fig. 4c, and Supplementary Figs. S33, 34). We propose that in situ-generated NH₃⁺ at the catalyst periphery would repel other cationic species, while the long-chain hydrophobic molecules would hinder interfacial water adsorption. Such effects might lead to reduced accumulation of H₃O⁺ and H₂O near the catalytic center, thereby alleviating competing HER.

To assess the distribution of H₂O species around the solvated catalysts, we computed radial distribution functions, g(r), to analyze H₃O⁺ proximity to the Ni catalytic center (Supplementary Figs. S42, 43, Fig. 4d). The NiPc-NH₃⁺ and NiPc-NH₃⁺/-SHP (octadecylamine) lines showed that H₃O⁺ tends to reside beyond 7.7 Å from the Ni center compared with that of NiPc (5.7 Å), which could be attributed to coulombic repulsion with the positively NH₃⁺ present on the molecules. It leads to the electrolytic H₃O⁺ far away from the Ni center, which is consistent with the result of ATR-SEIRAS (Fig. 3e). Meanwhile, NiPc-NH₃⁺ is primarily distributed in a higher 8.5–10.0 Å range than that of NiPc (7.5–9.0 Å), but NiPc-NH₃⁺/-SHP showed the highest peak position of 9.5–11.0 Å. These data suggest that the combination of long-chain hydrophobic molecules and NH₃⁺ groups further inhibit the approach of protonated water, thereby supporting the enhanced CO₂RR selectivity of NiPc-NH₂/CNT-SHP. Meanwhile, further density functional theory (DFT) calculations suggest that H₃O⁺ could promote a Volmer-type reaction at the imine nitrogen site, leading to reduced catalyst stability and activity (Supplementary Figs. S44, 45), thereby highlighting the importance of interfacial exclusion of protonated water species.

Furthermore, we also analyzed the distribution of H₂O near the central Ni in NiPc, NiPc-NH₃⁺, and NiPc-NH₃⁺/-SHP (Fig. 4e), which shows a peak at ~2.3 Å, representing the first solvation shells. It is worth noting that the strength of the first solvation shell of NiPc-NH₃⁺/-SHP is lower than that of NiPc-NH₃⁺ and NiPc, suggesting that less H₂O exists at the interface of NiPc-NH₃⁺/-SHP and the distribution of interfacial water was affected by the long-chain hydrophobic groups, consistent with the result of ATR-SEIRAS (Supplementary Fig. S47). NiPc–NH₃⁺ narrows the water distribution range around NiPc from 2.0–3.2 Å to 2.0–2.8 Å, yielding a higher intensity. This implies that -NH₃⁺ not only influence H₃O⁺, but also modulate the distribution of H₂O (excluding H₃O⁺) at the interface. It is attributed to the positively charged group exhibiting an increased propensity for hydrogen bonding with water, thereby enhancing their affinity for water, consistent with our previous assumptions, indicating the important role of secondary introduction of hydrophobic long chains.

Interestingly, the second g(r) peak, representing the second solvation shell, is located at ~4.2 Å for both NiPc and NiPc-NH₃⁺, whereas no peak is observed in the case of NiPc-NH₃⁺/-SHP. This finding further suggests that long-chain hydrophobic groups not only alter the distribution of interfacial water but also disrupt the hydrogen bonding network of water farther from Ni (the second solvation shell), consistent with ATR-SEIRAS (Fig. 3e). The broken hydrogen bonding produces a more hydrophobic environment, which allows CO₂ to reside near the catalytic site and be reduced more readily²⁴. On the other hand, the broken hydrogen bonding inhibits proton jumping from H₃O⁺ to the catalyst via the Grotthuss mechanism²⁴, allowing more free water to enter (Fig. 3c), which further facilitates CO₂RR and suppresses HER (Fig. 2b)³⁸. In general, our MD simulations agree well with these RDE and ATR-SEIRAS results. Combining all our analyses, the acidic CO₂RR performance on NiPc-NH₂/CNT-SHP catalyst can be convincingly ascribed to the advantageous synergies between -NH₃⁺ groups and long-chain hydrophobic modifications in optimizing the interfacial microenvironment. This effects likely reduce H₃O⁺ and H₂O coverage at the Ni catalytic unit (Fig. 4f), enrich CO₂ near the interface, and thus improve selectivity under high current operation (Fig. 2c). Given the outstanding performance of NiPc-NH₂/CNT-SHP in acidic CO₂RR, we proceeded to optimize it for evaluation in a PEM reactor.

Electrochemical performance test in acid PSE reactor and TEA analysis

To enable continuous CO production under acidic conditions in a practical system, we constructed a PEM reactor incorporating NiPc-NH₂/CNT-SHP as the cathode. Given the operating conditions in a flow cell, 0.05 M K₂SO₄ + 0.05 M H₂SO₄ was initially used as the anolyte to mitigate (bi)carbonate precipitation. As potassium ions migrate across the membrane in hydrated form, they alter the catalyst’s local environment, increasing alkalinity at the cathode interface.¹¹, However, CO production remained minimal (Supplementary Fig. S48), likely due to high proton flux at the catalyst/membrane interface⁵¹. Increasing the K⁺ concentration (0.5 M K₂SO₄ + 0.05 M H₂SO₄) slightly improved FE_CO (Supplementary Fig. S49), but salt precipitations were observed in the flow channels after more than ten hours of testing (Supplementary Fig. S50), likely due to the local generation of high proton concentrations at the catalyst/membrane interface during the oxygen evolution reaction on the anode side. These protons will then be promptly transported across the PEM to the cathode side from the strong acid interface, which suppresses K⁺ transport from the anolyte to the cathode, hindering the regulation of the interfacial environment for CO₂RR, a process distinct from the interfacial environment of the catalyst in the acidic environment of the flow cell. The low FE_CO in the PEM reactor prompted us to establish a strong acid buffer interval to mitigate the impact of the strong acid interface on CO₂RR performance.

Instead of a conventional PEM reactor design, we developed a PSE reactor (Supplementary Figs. S51, 52) featuring two chambers separated by a PEM, where a conductive anionic PSE layer functions to (1) minimize iR-drop between the cathode and anode, (2) transfer protons while maintaining an acidic interfacial environment (Fig. 5a), (3) alleviate the strong-acid interface effect on CO₂RR performance. The system employs NiPc-NH₂/CNT-SHP on a gas diffusion layer (GDL) as the cathode and IrO₂-coated titanium mesh as the anode. The pure CO₂ was continuously supplied for the electrochemical reduction of CO₂ to CO. Meanwhile, the anode chamber was circulated with 0.5 M H₂SO₄ for OER, which can produce hydrogen protons. These protons will then be transported through the membrane and PSE layer to the cathode. They interact with diffused cations, regulating the local catalytic environment at the catalyst/PSE interface to facilitate CO generation. Additionally, bicarbonate at the catalyst/PSE interface reacts with protons to regenerate CO₂, mitigating salt deposition and carbon loss. Protons produced by water oxidation at the anode enter the middle chamber to balance the charge that has been spent.

Fig. 5: Electrochemical performance test in the acid PEM-PSE reactor.

a Schematic illustration of reducing CO₂ to CO in our acid PSE reactor. b I–V curve (without iR compensation), c, d FE_CO, and e j_CO (without iR compensation) for CO₂ RR using the reactor with (Resistance: 2.40 ± 0.44 Ω) or without (Resistance: 9.40 ± 1.37 Ω) PSE layer. f The chronopotentiometry stability test of the acid PSE reactor by directly flowing 0.1 M K₂SO₄ + 0.01 H₂SO₄ solution (pH = 2.27 ± 0.22 Ω) in the middle chamber at the current density of 100 mA cm⁻² and the corresponding FE_CO, black arrow: renewing electrolyte. g Comparison of our work with other relevant studies in terms of j_total, cell voltage, stability, and FE_CO for full-cell acidic CO₂RR. The error bars represent the standard deviation of three independent measurements. Source data are provided as a Source Data file.

The I–V curve of our PSE reactor flowing 0.1 M K₂SO₄ + 0.01 M H₂SO₄ solution in the PSE layer was plotted in Fig. 5b, where a high-concentration salt solution can be used. In contrast, we use the low salt concentration to mitigate salt precipitation during long-term testing for practical demonstrations. The PSE reactor’s cell voltage progressively rose as the total current density increased, which achieved a current density of 300 mA cm^–2 at the cell voltage of ~3.96 V. In contrast, the reactor without the PSE layer exhibited higher cell resistance (Supplementary Fig. S53b), leading to increased cell voltage at the same current (Fig. 5b). Owing to the bubbles generated from the anode disturbing the membrane, an apparent fluctuation was found in the constant current test of the PSE reactor without conductive solid electrolyte (Supplementary Fig. S54b). With the PSE reactor, NiPc-NH₂/CNT-SHP achieved a total current density of 150 mA cm⁻² at ~3.37 V, significantly lower than ~5.11 V in the absence of the PSE layer.

The strong acidity at the catalyst/PSE interface led to slightly lower FE_CO in the PSE reactor compared to the non-PSE system at currents below 120 mA cm^–2 (Fig. 5c). This discrepancy can also be mitigated by increasing the salt concentration (Supplementary Fig. S55). However, as the test current increases, the increased K⁺ facilitates the enhancement of interfacial alkalinity, resulting in comparable FE_CO ( ~ 98%) between both systems at 120 mA cm^-2 (Fig. 5d). Due to the lack of solid electrolyte support, the gas generated by the anode fluctuates on the electrical signal, thus the constant current test stops at 150 mA cm^-2 without PSE reactor (Fig. 5b). Notably, FE_CO remained above 90% across a wide current range (100–300 mA cm^–2), with NiPc-NH₂/CNT-SHP achieving j_CO of ~148.38 mA cm^–2 at ~3.37 V in the acidic PSE reactor, compared to ~4.60 V without PSE. At 3.96 V, j_CO reached 295.26 mA cm^-2 (Fig. 5e).

Electrolysis stability is one of the most crucial yet challenging components for evaluating practical applications in acidic conditions. Benefiting from the NiPc-NH₂/CNT-SHP, which features a stable Ni molecular site, a superhydrophobic interface, and a reliable design, our PSE reactor exhibits long-term stability in producing CO under acidic conditions. Even though a low concentration of K⁺ was used, a slight salt was also found on the back of the electrode. Consequently, the serpentine channel was rinsed with deionized water every 40 h, and the cathodic electrolyte was refreshed approximately every 60 h under a sealed state. This procedure prevents salt precipitation from obstructing the channel and ensures adequate K⁺ availability for CO₂ reduction reactions (CO₂RR). Notably, no organic products were detected in the liquid electrolyte after the 20-hour stability test, indicating the absence of liquid-phase products (Supplementary Fig. S56) and confirming the high stability of the PSE. The PSE reactor stability was evaluated by maintaining a 100 mA cm⁻² current density for 200 h with a high FE_CO above 80%. The whole-cell voltage was maintained at ~3.17 V (Fig. 5f). Following the 200 h test, the stability of the catalyst was confirmed by XPS, XAS, TEM, and contact angle measurements (Supplementary Figs. S57–60). Compared with the original PSE (Supplementary Fig. S52), no significant morphological changes were observed in the PSE after a 200 h test (Supplementary Figs. S61a, b), further confirming the structural stability of the PSE. Here we also conducted additional experiments to demonstrate that heat treatment significantly enhances energy efficiency, especially under high operational current densities (Supplementary Fig. S62). Additionally, NiPc-NH₂/CNT-SHP achieved a high j_CO of 210.45 mA cm^–2 at a cell voltage of 4.17 V in alkaline conditions and exhibited long-term stable operation at 50 and 100 mA cm^–2 (Supplementary Figs. S63–65), further underscoring its superior performance. Consequently, the NiPc-NH₂/CNT-SHP catalyst offers extended operational time, comparable cell voltage, and FE_CO relative to existing reports on CO₂RR under acidic conditions (100 mA cm^–2, Supplementary Table S4, Fig. 5g)^11,20,23,52, underscoring the advanced nature of this catalyst design.

Based on the electrochemical performance of our acidic PSE reactor system for CO₂ electrochemical reduction, we conducted a techno-economic analysis (TEA) to evaluate its economic viability and adaptability, adapting previously reported methodologies. Several process assumptions were made based on current feedstock prices and electrolyzer performance, with a sensitivity analysis accounting for potential future fluctuations in these parameters. The study of CO production costs indicates a total system cost range of approximately US$327.68 to US$351.91 per ton of CO at current densities between 100 and 300 mA cm^-2 (Supplementary Fig. S66a). Taking 100 mA cm^-2 as an example, the major cost contributors are as follows: electrolyzer and installation ($43.75 per ton of CO), balance-of-plant (BOP) components ($9.22 per ton of CO), electricity consumption ($198.47 per ton of CO), input chemicals ($63.41 per ton of CO), and other operational expenses ($31.51 per ton of CO). Furthermore, a detailed global sensitivity analysis shows that generation costs are highly sensitive to variations in electricity prices (Supplementary Fig. S66b, ranging from $0.02 to $0.06), cell voltages (ranging from 2.0 to 5.0 V), FE_CO (ranging from 100 to 50%), and current density (400 to 20 mA cm⁻²). This is because electricity prices are directly related to operating costs, and cell voltages are tied to the amount of electricity consumed. The Faradaic efficiency of products and current density affect the size and capital cost of electrolyzers for producing 1 ton of CO (Supplementary Note 1). Other parameters, such as stack cost, exert a moderate effect on costs. Overall, developing strategies to prevent declines in FE_CO and current density under specific voltage changes, as well as monitoring electricity policy and price fluctuations, will be crucial for ensuring the economic feasibility of acidic CO₂RR in the future.

Discussion

A nickel-based heterogeneous molecular electrocatalyst (NiPc-NH₂/CNT-SHP) featuring -NH₂ functional groups and grafted long-chain hydrophobic molecules has been successfully synthesized. Under acidic conditions, the -NH₂ group is in situ protonated to form -NH₃⁺, facilitating acidic CO₂RR at high current. The designed catalyst exhibited good stability, activity, and selectivity for the electrochemical reduction of CO₂ to CO under acidic conditions, outperforming its counterparts lacking hydrophobic and amino groups. In acidic electrolytes, the catalyst achieves a FE_CO of nearly 100%, a current density (j_CO) of 426 ± 14 mA cm^-2, and a CO₂ utilization rate of 76%. In comparison, the maximal FE_CO of above 99% and j_CO of 390 ± 7 and 414 ± 40 mA cm^–2 can be achieved in alkaline and neutral electrolytes, respectively. MD simulations, RDE test, in-situ ATR-SEIRAS and in-situ XPS proved that the positively charged -NH₃⁺ and superhydrophobic interface disrupt the protonated water (H₃O⁺)-rich network, influencing the migration of free water, H-bonded water, and H₃O⁺. This disruption inhibits H₃O⁺ invasion and optimizes the distribution of free water, thereby enhancing selective CO₂RR performance under acidic conditions. Notably, NiPc-NH₂/CNT-SHP demonstrated outstanding FE_CO ( > 80%) and operational stability (>200 h) at industrially relevant current densities in an acidic PEM-PSE reactor, which would be a promising demonstration of the use of renewable electricity for the continuous generation of CO through CO₂ reduction at a more feasible scale.

Methods

Chemicals

Ethanol (CH₃CH₂OH, ≥99.7%), nickel chloride (NiCl₂, ≥99%), 1-pentanol (≥99.0%), 1, 8-diazabicyclo[5.4.0]undec-7-ene (DBU, ≥98.0%), N, N-dimethylformamide (DMF, ≥99.9%), potassium hydroxide (KOH, ≥85.0%), hydrochloric acid (HCl, 36% ~ 38%), potassium bicarbonate (KHCO₃, ≥99.9%), sulfuric acid (H₂SO₄, ≥99.8%), thionyl chloride (SOCl₂, ≥99%) were all purchased from Sinopharm Chemical Reagent Co., Ltd. Nickel phthalocyanine (NiPc, ≥85%) was purchased from Aladdin chemical company. Carboxyl-functionalized carbon nanotube (CNT-COOH, ≥95%), graphitized carbon nanotube (CNT), and octadecylamine (≥90%) were purchased from Macklin Chemical Company. 4-Aminophthalonitrile (≥98.0%) was purchased from Bidepharm Chemical Company. High-purity carbon dioxide gas (CO₂, 99.999%) and high-purity argon (Ar, 99.999%) were purchased from Jiangsu SoPo Gas Co., Ltd. (China). Millipore water (18.2 MΩ·cm) was used throughout all experiments.

Synthesis of amino-substituted nickel phthalocyanine

NiPc–NH₂ was synthesized by dissolving 8 mmol 4-aminophthalonitrile, 1 mmol NiCl₂, and 2 mL 1,8-diazabicyclo[5.4.0]undec-7-ene in 60 mL 1-pentanol under ultrasonic treatment. After Ar purging, the solution was sealed in a 100 mL Teflon-lined autoclave and heated at 180 °C for 8 h. During the solvothermal reaction, base-promoted cyclization of phthalonitrile yielded crude amino-substituted nickel phthalocyanine. The product was sequentially treated in 1 M HCl and 1 M NaOH at 90 °C for 1 h each, filtered, washed with water, and purified via Soxhlet extraction (methanol) to afford pure NiPc–NH₂ as a dark green-blue solid.

Synthesis of NiPc-NH₂/CNT-SHP and NiPc-NH₂/CNT catalyst

Commercial carboxylated CNTs were vacuum-dried at 110 °C overnight. 200 mg of CNTs were dispersed in 40 mL SOCl₂ and 4 mL DMF, degassed, and refluxed at 70 °C for 24 h under Ar to yield acyl chloride-functionalized CNTs (CNT–COCl). The product was filtered, quenched with 1 M KOH to remove residual SOCl₂, washed with DMF, and then vacuum-dried. Subsequently, 200 mg of CNT–COCl was dispersed in 200 mL DMF and sonicated for 30 min, followed by the addition of 40 mg NiPc–NH₂ and further sonication. After degassing, the mixture was refluxed at 90 °C for 12 h under argon. Then, 3 g of octadecylamine was added, and the mixture was refluxed for another 12 h. The resulting NiPc–NH₂/CNT–SHP was collected by filtration, washed with DMF, ethanol, and water, and dried under vacuum. NiPc–NH₂/CNT was prepared identically but without octadecylamine.

Synthesis of NiPc@CNT catalyst

Commercial NiPc (40 mg) was dispersed in 100 mL of DMF and sonicated for 30 min. Separately, graphitized CNTs (200 mg) were dispersed in 100 mL of DMF and sonicated for 30 min. The two solutions were mixed under vigorous stirring, followed by continuous stirring for 24 h. The resulting NiPc@CNT was collected by vacuum filtration, washed with DMF and ethanol until the filtrate was colorless, and dried under vacuum.

Synthesis of NiPc@CNT-SHP catalyst

200 mg of CNT-COCl was dispersed in 200 mL of DMF and sonicated for 30 min. Subsequently, 40 mg of NiPc was added, followed by an additional 30 min of sonication. The mixture was degassed and refluxed at 90 °C under an argon atmosphere for 12 h. Then, 3 g of octadecylamine was added, and the reaction was continued under reflux at 90 °C for an additional 12 h. The resulting NiPc@CNT-SHP was collected by filtration, washed sequentially with DMF, ethanol, and water, and dried under vacuum.

Flow cell test

3 mg of samples, 1 mL of ethanol and 30 μL of Nafion (5 wt%) solution were mixed and ultrasonically treated for 1 h to obtain the ink solution. The obtained homogeneous ink was airbrushed onto 1×3 cm² hydrophobic carbon paper (Sigracet 28 BC gas-diffusion layer) to achieve a loading of 1 mg cm^-2, and then dried overnight at 60 °C. An Ag/AgCl electrode (3.5 M KCl, The Ag/AgCl reference electrode was calibrated in 0.5 M H₂SO₄ under H₂-saturated conditions using two Pt foil electrodes) and a Pt electrode were chosen as the reference (reaction area: 1 cm²) and counter electrodes, respectively. All potentials from the three-electrode experiments were converted to versus RHE with 80% iR correction by the following Eq. (1):

$${{{\rm{E}}}}\left({vs}.{{{\rm{RHE}}}}\right)={{{\rm{E}}}}\left({vs}.{{{\rm{Ag}}}}/{{{\rm{AgCl}}}}\right)+0.2046\, {{{\rm{V}}}}+0.059\times {{{\rm{pH}}}}-{iR}u$$

(1)

where i is the current and $Ru$ is the solution resistance. The electrochemical performance of the samples was measured using an electrochemical workstation (Shanghai, CHI1140C). The gas products were analyzed using online gas chromatography (GC-2030; Shimadzu) equipped with a thermal conductivity detector (TCD) and a flame ionization detector (FID). Liquid products were detected using ¹H NMR spectroscopy with DMSO as an internal standard. Each cathode electrode was activated by cyclic voltammetry for 20 min before the constant current test. In alkaline and neutral conditions, 1 M KOH (50 mL) and 1 M KHCO₃ (50 mL) were used as the cathode and anode electrolytes, respectively. In acidic conditions, 0.5 M K₂SO₄ (pH=1) was used as the cathode electrolyte and 0.5 M H₂SO₄ (50 mL) was used as the anode electrolyte. Furthermore, the pH of 3 M KCl (50 mL) was adjusted to 0.4 using concentrated sulfuric acid, which was used as the cathode electrolyte, and 0.5 M H₂SO₄ (50 mL) was used as the anode electrolyte. Before the electrochemical test, the electrolyte was saturated with argon gas for 30 min. The electrolyte was circulated through the cathodic and anodic chambers using a peristaltic pump at rates of 10 mL min⁻¹ and 40 mL min⁻¹, respectively. In all flow cell measurements, the test potential was corrected for 80% ohmic resistance. The cell resistances (R) were measured by electrochemical impedance spectroscopy (EIS) under open-circuit voltage. The flow rate of inlet CO₂ gas through the gas chamber was controlled to be 2 ~ 25 mL min⁻¹ using a 100-sccm-range MFC, and a soap film flowmeter measured the outlet flow rate. The electrochemically active surface area (ECSA) and electrochemical impedance spectroscopy (EIS) were tested in an H-type cell by an electrochemical workstation (Shanghai, CHI760e). The cathodic electrolyte is CO₂-saturated 0.5 M KHCO₃, and the anodic electrolyte is 0.5 M KHCO₃. The ECSA experiment was carried out on a glassy carbon electrode (3.0 mm in diameter) at the non-CO₂ Faradaic efficiency potential. The ECSA is proportional to the C_dl value. The loading of the electrocatalyst was 0.2 mg cm⁻². The EIS experiment was conducted at a potential of −0.60 V vs. RHE.

Porous solid electrolyte reactor test

Electrochemical measurements were performed on a BioLogic VSP-3e workstation. The PSE reactor featured a NiPc-NH₂/CNT-SHP catalyst coated on a 1 cm² gas diffusion layer (GDL) as the cathode. A 0.25 mm PTFE gasket with a 1.0 cm² window and an anion exchange membrane (AEM) was positioned between the cathode and the solid electrolyte. An IrO₂/Ti mesh served as the anode, separated from the solid electrolyte by a Nafion 117 proton exchange membrane. The anode compartment was supplied with 0.5 M H₂SO₄ (50 mL) at a flow rate of 1.0 mL min⁻¹ using a peristaltic pump, and the 0.01 M K₂SO₄ (50 mL) was used as the electrolyte in the middle chamber to reduce resistance. The central chamber contained a 2.5 × 2.5 cm² window filled with Dowex 50 W X8 (H⁺ form), serving as the porous solid electrolyte. During CO₂RR tests, CO₂ gas was delivered to the cathode at a rate of 20 mL min⁻¹ via a mass flow controller (Alicat). Prior to measurements, the cell was conditioned at a low current for ~20 min to stabilize the voltage and product selectivity.

Acid membrane electrode assembly electrolytic cell test

Electrochemical measurements were performed using a Bio-Logic VMP3e workstation. The cell assembly included a 0.25 mm polytetrafluoroethylene gasket with a 1.0 cm² window, a NiPc–NH₂/CNT–SHP catalyst coated on a 1 cm² GDL (1 mg cm^-2) as the cathode, and a commercial IrO₂ catalyst air-brushed onto Ti mesh (IrO₂/Ti) as the anode. The anode electrolyte consisted of either 0.05 M K₂SO₄ + 0.05 M H₂SO₄ (50 mL, stored at room temperature 25 °C) or 0.5 M K₂SO₄ + 0.05 M H₂SO₄ (50 mL, stored at room temperature 25 °C), delivered at 10 mL min^_1 via a syringe pump. The anode and cathode chambers were separated by a Nafion HP membrane (20 µm). CO₂ gas was supplied to the cathode at a flow rate of 20 mL min⁻¹ using a mass flow controller (100 sccm range).

Acid porous solid electrolyte reactor test

Electrochemical measurements were performed using a Bio-Logic VMP3e and a CHI 660e workstation, both equipped with a CHI 680c current booster. The cathodic compartment consisted of a stainless-steel chamber with a serpentine flow field and four ports: two for CO₂ gas circulation and two for electrolyte flow through a 1 mm thick middle chamber. A 0.5 mm silicone gasket with a 1.0 cm² window was used to ensure tight contact between the cathode and the solid electrolyte. The cathode received 20 mL min⁻¹ of CO₂ (controlled by a 100 sccm Alicat MFC). IrO₂/Ti mesh served as the anode, separated from the solid electrolyte layer by a Nafion HP membrane (20 µm). The anode was supplied with 0.5 M H₂SO₄ (50 mL) at 2.0 mL min⁻¹, while the catholyte consisted of 0.1 M K₂SO₄ + 0.01 M H₂SO₄ (50 mL).

In situ electrochemical ATR-SEIRAS measurements

In situ ATR-SEIRAS measurements were performed using a Pike Veemax III ATR cell equipped with a single-reflection Si crystal coated with a thin Au film, operating in internal reflection mode. Spectra were collected on a Thermo Fisher, Nicolet^TM iS50 spectrometer (U.S.). A three-electrode configuration was employed, using NiPc–NH₂/CNT–SHP as the working electrode, Ag/AgCl as the reference, and Pt felt as the counter electrode. The electrolyte (0.5 M K₂SO₄+H₂SO₄, pH=1) was saturated with CO₂ for 30 min prior to measurement. Spectra were recorded over a potential range from the open-circuit potential (OCP) to –1.6 V vs. RHE.

In situ electrochemical XPS measurements

In situ X-ray photoelectron spectroscopy (XPS) was employed to monitor real-time changes in the surface chemical states during electrochemical CO₂ reduction. Prior to measurements, clean and flat electrode samples (diameter ≤10 mm) were prepared and mounted in an in situ electrochemical cell connected to an ultrahigh vacuum (UHV) chamber (base pressure ≤1×10⁻⁹ mbar). XPS experiments were conducted using a Thermo Scientific K-Alpha system, and electrochemical control was achieved with a CorrTest workstation (CS350M). During the measurement, the electrochemical environment was precisely regulated, allowing for the stepwise acquisition of XPS spectral information (N 1 s, C 1 s) at different applied potentials. Particular attention was paid to shifts in binding energy, intensity variations, and the emergence of features in the XPS spectra to elucidate the interfacial processes occurring during operation.

Computational details

DFT and ab initio molecular-dynamics (AIMD, Supplementary Data 1) simulations were performed with the Vienna ab initio Simulation Package (VASP)⁵³ using projector augmented-wave (PAW) potentials⁵⁴ and a 400 eV plane-wave cutoff. Exchange–correlation was described by the PBE functional⁵⁵ within GGA and augmented by the DFT-D3 dispersion correction⁵⁶. All calculations were spin-polarized, and Brillouin-zone integration employed a Γ-only k-mesh⁵⁷, suitable for the isolated-molecule cell. A cubic box of 20 × 20 × 20 ų containing one NiPc unit was used. Geometry optimizations converged at 1 × 10^-5eV for energies and 0.01 eV Å⁻¹ for forces. AIMD simulations employed a Nosé–Hoover thermostat at 300 K with 1 fs time steps, setting H masses to those of deuterium⁵⁸. Reaction barriers were obtained via the slow-growth enhanced sampling method⁵⁹.

Since explicit AIMD of bulk water is computationally prohibitive, we construct a·finite-radius water-droplet model that mimics the solvation environment of NiPc molecule and thus makes AlMD· tractable. In this approach, the outer tier of water molecules feel a Lennard-Jones-like repulsive force·so·that the water moleclules do not drift too far from the central Ni atom.

All molecular-dynamics (MD) simulations were carried out in the canonical (NVT) ensemble at 298.15 K with initial random velocities, each extending beyond 20 ps. Equations of motion were integrated via the velocity–Verlet algorithm with a 1 fs time step, and temperature was maintained by a Nosé thermostat (Q = 0.01). Covalent interactions (bonds, angles, and dihedrals) were described by the Universal Force Field, with partial charges from the QEq (spell out) method. Long-range electrostatics and van der Waals forces were treated by Ewald summation (accuracy = 0.001 kcal mol⁻¹) using a 12 Å cutoff under periodic boundary conditions.

All three systems were contained in 25× 25 × 25 Å cubic boxes. To maintain the density of liquid water, 478, 469, and 451 water molecules were added to the system 1 (NiPc), system 2 (NiPc-NH₃⁺), and system 3 (NiPc-NH₃⁺/-SHP), respectively. Since the electrolyte employed in the experiments was the acidic K₂SO₄ solution, all MD simulations were performed with H₃O⁺, K⁺ and SO₄²⁻ to maintain charge neutrality. To ensure a uniform H₃O⁺ concentration, each system contained three H₃O⁺ ions. In system 1, since the NiPc molecule is neutral, two SO₄²⁻ anions and one additional K⁺ cation were added to maintain charge neutrality. In system 2, the protonated NiPc–NH₃⁺ molecule carries three NH₃⁺ groups, considering the three H₃O⁺ ions in solution, three SO₄²⁻ anions were required for charge neutrality. System 3 (NiPc–NH₃⁺/-SHP) was identical to system 2, except that an SHP molecule was positioned 4.2 Å from the Ni center.

Rotating disk electrode experiments

The diffusion limiting current of the reduction of hydronium ions was calculated according to Levich Eqs. (2–3)⁶⁰:

$${i}_{{platea}u}=0.62n{FA}{D}^{\frac{2}{3}}{{v}^{-}}^{\frac{1}{6}}{c}_{O,{H}^{+}}{\omega }^{\frac{1}{2}}$$

(2)

$$S=0.62n{FA}{D}^{\frac{2}{3}}{{v}^{-}}^{\frac{1}{6}}{c}_{O,{H}^{+}}$$

(3)

where n is the number of electrons transfer in the reaction, $F$ is the Faraday constant (9.65×10⁴C mol⁻¹), A is the electrode surface area, $D$ is the diffusion coefficient of hydronium ions, $v$ is the kinematic viscosity of electrolyte, ${c}_{O,{H}^{+}}$ is the bulk concentration of hydronium ions, and ω is the rotating speed of the RDE (unit: rad s⁻¹), ${i}_{{platea}u}$ is the current. $S$ is the linear fitting slope.

The kinetic current density can be obtained from the Koutecky-Levich Eq. (4):

$$\frac{1}{{J}_{{tot}}}=\frac{1}{{J}_{k}}+\frac{1}{{J}_{{plateau}}}$$

(4)

where$\,{j}_{{tot}}$ is the total current density, ${j}_{k}$ is the kinetic current density and ${j}_{{plateau}}$ is the diffusion limiting current density.

Evaluation of Faraday efficiency

The Faradic efficiency of the gas product was calculated by the following Eq. (5):

$${{{{\rm{FE}}}}}_{x}=\frac{{n}_{x}\times F(C/{mol})\times V({ml}/\min )\times {10}^{-6}({m}^{3}/{ml})\times v\left({vol}\%\right)\times 1.013\times {10}^{5}(N/{m}^{2})}{8.314(N{{{\rm{\cdot }}}}m/{mol}{{{\rm{\cdot }}}}K)\times T(K)\times {I}_{{total}}(C/s)\times 60(s/\min )}$$

(5)

${n}_{x}$: the number of electrons required to produce a product molecule (H₂, CO = 2);

F: the Faradaic constant (96485 C mol⁻¹);

V: the flow rate of CO₂ bubbling;

v: the volume ratio of gas product calculated from the peak area of GC current signal;

T: 298.15 K;

${I}_{{total}}$: the steady-state cell current (C/s).

Evaluation of partial current density

The partial current density of gas product was calculated by the following Eq. (6):

$${j}_{x}={{FE}}_{x}\times {j}_{{total}}$$

(6)

${j}_{x}$: the partial current density (mA cm⁻²) of gas product;

${{FE}}_{x}$: the Faradic efficiency of the gas product;

${j}_{{total}}$: the total current density (mA cm⁻²).

Evaluation of TOF

The turnover frequency (TOF, s⁻¹) value of the gas product was calculated by the following Eq. (7):

$${{TOF}}_{x}=\frac{{{FE}}_{x}\times {j}_{{total}}}{{n}_{x}\times F\times \omega \times {m}_{{cat}}/{M}_{{ni}}}$$

(7)

ω: the content of nickel (Ni) in the hybrid catalysts;

${m}_{{cat}}$: the amount of the tested catalyst;

${M}_{{ni}}$: the relative molecular weight of Ni.

Material characterizations

The morphologies of as-prepared samples were studied by using TEM (Japan, JEOL-2100F), HRTEM (Japan, JEOL-2100F), HAADF-STEM (FEI Titan) with energy dispersive X-ray spectroscopy (EDX), and SEM performed with a field-emission scanning electron microanalyzer (Hitachi S-4800 II, Japan). Furthermore, the atomic information of NiPc-NH₂/CNT-SHP was characterized using an FEI Themis Z transmission electron microscope operated at 200 kV and equipped with double spherical aberration (Cs) correctors. The physical structure information of the samples was analyzed using an X-ray diffractometer (XRD, Bruker AXS Company, Germany) equipped with monochromated Cu Kα radiation (λ = 1.54178 Å) and an XPS spectrometer (ESCA PHI500). The Ni amount loading of the catalysts was quantified by inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7800). ¹H NMR spectroscopy was performed with a 500 MHz spectrometer.

X-ray absorption spectroscopy (XAS)

The sample was analyzed using XAS at the Shanghai Synchrotron Radiation Facility in China. The spectra were processed and analyzed by the software codes (Athena)^61,62.