Graphite-conjugated phthalocyanines through dioxin linkages: a molecularly programmable platform for heterogeneous CO₂ electrocatalysis

Abstract

Combining the structural programmability of molecular catalysts with the robustness and direct electrode coupling of heterogeneous systems has long been pursued in electrocatalysis. Herein, we report a 1,4-dioxin conjugation strategy that covalently anchors metal phthalocyanines onto graphitic electrodes while simultaneously enabling programmable peripheral functionalization, thereby achieving molecular-level tunability within a heterogeneous platform. Perfluorinated Ni-phthalocyanine was grafted onto graphitic electrodes via cyclization between edge catechols and fluoro-groups, forming graphite-conjugated Ni-phthalocyanine (GC-NiPc). Using the same chemistry, substituted catechols bearing electron-donating/withdrawing groups (methyl, hydrogen, aldehyde, carboxyl, nitro, tetrachloro, and pyridyl) were incorporated onto the phthalocyanine periphery of GC-NiPc to tune the electronic structure of the phthalocyanine ring, resulting in an approximately 100 mV shift in redox potential that correlates linearly with Hammett constants of the substituents. This electronic modulation translated directly into altered CO₂RR activity, with pyridine-modified GC-NiPc-Py achieving a high CO turnover frequency (1841 h⁻¹) and above 99% faradaic efficiency over a wide potential window. In situ ATR-IR spectroscopy and DFT calculations further identified *COOH formation as the rate-determining step, with electron-withdrawing groups lowering the associated activation barrier. Compared with physisorbed analogues, the covalent conjugation strategy improves catalyst utilization and long-term stability, while programmable substituent engineering further enables systematic optimization of catalytic performance.

Graphite–phthalocyanine conjugation through 1,4-dioxin linkages enables substituent-controlled electronic tuning of Ni–N₄ active sites, establishing a molecularly defined heterogeneous platform for optimizing CO₂ reduction activity.

Introduction

Combining the structural programmability of molecular catalysts with the robustness and direct electrode coupling of heterogeneous systems has long been pursued in electrocatalysis. Molecular M–N₄ catalysts, including metal porphyrins, phthalocyanines, and related macrocyclic complexes, occupy a central position in molecular catalyst design because they provide atomic-level structural precision and extensive tunability. Their electronic properties can be systematically tuned through modifying the metal center,^1,2 axial ligands,^3–5 and peripheral substituents,^6–9 leading to wide applications in the CO₂ reduction reaction (CO₂RR),^6,10,11 oxygen reduction reaction (ORR)^12,13 and other electrocatalytic processes. Despite their versatility, the practical implementation of molecular M–N₄ catalysts is hindered by their insufficient stability,^14,15 incomplete site exposure caused by π–π stacking-induced aggregation,¹⁶ and weak electronic coupling to conductive supports.¹⁷

Immobilization strategies have therefore been pursued to integrate molecular M–N₄ motifs into heterogeneous architectures.^18–21 Physisorption of aromatic macrocycles onto carbon supports through π–π stacking offers a straightforward immobilization approach that preserves molecular identity. However, such non-covalent adsorption results in slow interfacial electron transfer and susceptibility to desorption. Axial coordination bridging offers a mild strategy for immobilizing M–N₄ catalysts onto carbon supports while facilitating directional electron transfer,^19,22 however, the intrinsically dynamic nature of coordination bonds often compromises the long-term stability of the catalyst under operating conditions.²³ Although molecular M–N₄ catalysts can be converted into metal–nitrogen–carbon (M–N–C) materials through high-temperature (above 500 °C) pyrolysis to improve catalytic stability,²⁴ this process often obscures the molecular structure of the active sites and eliminates peripheral functional groups that are essential for precise electronic modulation.²⁵

To precisely control the structure and activity of heterogeneous M–N₄ active sites while preserving molecular-level tunability, several covalent integration strategies have been explored.²⁶ For example, incorporating functionalized monomers during the growth of Fe-phthalocyanine polymers on graphite surfaces enables modulation of the electronic structure of the Fe centers and their corresponding oxygen reduction reaction (ORR) activity.²⁷ However, the exact positional distribution and spatial uniformity of doped functional groups within these polymeric architectures remain insufficiently defined, complicating the establishment of rigorous structure–activity relationships. Furthermore, covalent integration strategies have recently emerged as an attractive approach.^28–31 Surendranath and coworkers demonstrated that conjugating Co-porphyrins to graphite edges via pyrazine linkages significantly enhanced ORR kinetics and reduced overpotential by strengthening electronic coupling.^28,29 Similarly, Rao and co-workers enhanced the CO₂RR performance of Ni phthalocyanine by conjugating the molecular catalyst to graphite edges, thereby facilitating interfacial electron transfer.³² Beyond stabilizing molecular catalysts on conductive supports, these strategies can also provide opportunities to regulate the local chemical environment surrounding the active sites.^30,33 For example, Ye and coworkers reported that covalent anchoring of molecular catalysts onto carbon nanotubes, followed by installation of tunable quaternary ammonium groups, could establish a stable charged and hydrophobic interfacial layer that regulates local water/proton transport and suppresses the competing hydrogen evolution reaction (HER), thereby modulating CO₂RR activity.³⁰ Similarly, covalently constructed layered architectures based on Co-phthalocyanine have been further engineered with alkyl-substituted quaternary ammonium groups to generate locally ionic, hydrophobic, and gas-philic microenvironments around the molecular catalytic centers, providing an additional means of controlling CO₂RR reactivity.³³ Covalent anchoring of molecular catalysts onto functionalized carbon nanotubes has also been demonstrated to improve their integration with conductive supports and enhance CO₂RR performance.³¹ These studies demonstrate that covalent immobilization can not only mitigate molecular association and improve catalyst retention, but also provide a platform for engineering the interfacial microenvironment of molecular M–N₄ catalysts. Despite its conceptual advance, the strategy relies on amine-functionalized macrocycles that require multistep synthesis and limit substituent diversification, thereby restricting its general applicability. To develop a more versatile conjugation strategy, we turned to the cyclization reaction between catechols and fluorinated aromatic groups, a well-established chemistry widely employed in the construction of covalent organic frameworks (COFs).^34–38 This reaction proceeds through nucleophilic aromatic substitution and oxidative annulation to form 1,4-dioxin linkages, generating extended π-conjugated structures under mild conditions with broad functional group tolerance. By adapting this chemistry to graphite edge functionalization, we envisioned a platform capable of simultaneously achieving covalent integration with graphitic electrodes and programmable peripheral functionalization of molecular catalysts. Unlike strategies that primarily address molecular immobilization or interfacial microenvironment engineering, this approach provides a common covalent anchoring motif while allowing systematic modulation of the molecular catalyst through peripheral functionalization, thereby enabling a more direct investigation of structure–activity relationships in heterogeneous M–N₄ electrocatalysis.

Herein, we synthesized graphite-conjugated Ni-phthalocyanine catalysts (GC-NiPc) by covalently anchoring perfluorinated Ni-phthalocyanine (NiPcF₁₆) onto graphitic edges through nucleophilic aromatic substitution, forming the 1,4-dioxin linkage. Using the same 1,4-dioxin chemistry, a series of functional groups, including methyl (–CH₃), hydrogen (–H), aldehyde (–CHO), carboxyl (–COOH), nitro (–NO₂), tetrachloro (–4Cl), and pyridyl (–Py), were introduced onto the phthalocyanine periphery to afford GC-NiPc-R catalysts. Electrochemical measurements revealed that electron-withdrawing/donating functionalization systematically shifted the [NiPc]⁻/[NiPc]²⁻ redox potential, correlating with the Hammett constant of the substituents. Notably, the logarithmic CO₂RR turnover frequency (TOF) exhibited a linear dependence on the [NiPc]⁻/[NiPc]²⁻ redox potential. The optimized GC-NiPc-Py catalyst achieved a TOF of 1841 h⁻¹ at −0.80 V_RHE with a faradaic efficiency for CO of 99.8%, while retaining 98% of the initial current density after continuous operation in an H-cell with 0.5 M CO₂-saturated KHCO₃. Control experiments confirm that graphite conjugation enhances catalytic performance and stability relative to physisorbed analogues. In situ infrared spectroscopy and density functional theory calculations further show that electron-withdrawing substituents reduce the activation barrier for the rate-determining *COOH formation. This work presents a versatile platform for rationally tuning molecule–support interactions and electronic structures, offering a generalizable approach for advancing molecular electrocatalysis.

Results and discussion

Synthesis and structural characterization of graphite-conjugated phthalocyanines

The carbon support (acid-treated acetylene black) was pretreated by reducing ortho-quinone edge sites to catechol moieties using Na₂S₂O₄ under a N₂ atmosphere (see the SI for details). The conversion of C O to C–OH functionalities was confirmed by the attenuation of the carbonyl stretching band at 1740 cm⁻¹ and the concomitant enhancement of the O–H stretching band at approximately 3600 cm⁻¹ in the IR spectrum (Fig. S1). NiPcF₁₆ was grafted onto the graphitic carbon support through a nucleophilic aromatic substitution reaction (S_NAr) catalyzed by potassium carbonate,³⁴ forming graphite-conjugated Ni-phthalocyanine, denoted as GC-NiPc. Notably, this catechol-fluoro cyclization chemistry has previously been employed in the synthesis of dioxin-linked covalent organic frameworks,^34–37,39 which proceeds efficiently under mild conditions. To functionalize GC-NiPc with other substituents, different functionalized catechols were used to react with the remaining fluorine sites (Fig. 1) to afford GC-NiPc-R (R = CH₃, H, CHO, COOH, NO₂, 4Cl, and Py). The products were thoroughly washed with DMSO, H₂O and acetone to remove residual catechols and unreacted NiPcF₁₆. The increase in characteristic C–O–C stretching bands in IR spectra (Fig. S1) confirms the formation of 1,4-dioxin linkages bridging the phthalocyanine macrocycle with both the graphitic support and the substituted catechols. Taking GC-NiPc-4Cl as an example, the uniform distribution of Ni, N, F, and Cl species within the material was further confirmed by elemental mapping using transmission electron microscopy (TEM) coupled with energy-dispersive X-ray spectroscopy (EDX), indicating that the functionalized catechol groups were homogeneously incorporated into the material without observable aggregation of Ni species (Fig. 1c and S2).

Fig. 1. (a) Synthesis of GC-NiPc-R. The right panel shows the proposed local structure of the NiPc with functional groups grafted on the edge of the graphite carbon. (b) Diagrams of functional groups in GC-NiPc-R with different electron-withdrawing/donating characteristics. (c) TEM images and EDX-based elemental mapping images of GC-NiPc-4Cl, showing the successful modification of the –4Cl functional group and uniform distributions of elements throughout the particles.

The compositions of GC-NiPc and GC-NiPc-R were characterized by inductively coupled plasma optical emission spectroscopy (ICP-OES), thermogravimetric analysis (TGA) and X-ray photoelectron spectroscopy (XPS). ICP-OES revealed similar Ni loadings of 1.1–1.2 wt% across all the samples. TGA revealed two major weight-loss steps at approximately 400 °C and 500 °C, attributed to the decomposition of phthalocyanine moieties and carbon supports, respectively. At 700 °C, a residual mass of 1.7–1.8 wt% was obtained, corresponding to NiO formation, in good agreement with the Ni content determined by ICP-OES (Fig. S3 and Table S1). Moreover, leaching experiments in DMF at 100 °C indicate that the NiPc species are predominantly covalently conjugated to the carbon edge site (Fig. S4), although we cannot completely exclude the presence of a small fraction of physisorbed phthalocyanine molecules. XPS analysis provided further insight into the substitution extent. Compared with pristine perfluorinated NiPc (F/Ni = 16), the F/Ni ratio in GC-NiPc decreases to ∼14, consistent with an average substitution of approximately two adjacent fluorides per macrocycle to form a single 1,4-dioxin linkage anchoring the phthalocyanine to the graphite carbon surface while leaving the remaining C–F groups intact (Fig. S5 and Table S2). Upon additional functionalization to generate GC-NiPc-R derivatives, the F/Ni ratio decreases substantially to 7.4(–CH₃), 8.0(–H), 9.0 (–CHO), 9.3(–COOH), 9.1(–NO₂), and 7.7(–Py), respectively. The O 1s spectra of GC-NiPc showed a decrease in C–OH species at 532.1 eV, accompanied by the emergence of C–O–C components, consistent with the conversion of surface catechol groups into dioxin linkages (Fig. 2a).⁴⁰ Because XPS probes only the near-surface region, variations in local NiPc coverage, carbon-surface oxidation, and the relative exposure of the carbon support can therefore alter the apparent C–O–C/C–OH ratio even when the average degree of secondary substitution is similar. In contrast, the F/Ni ratio compares two elements associated directly with the NiPc unit and is less affected by oxygen-containing groups on the carbon support. The relatively narrow range of F/Ni ratios (7.4–9.3) therefore provides more reliable evidence that the GC-NiPc-R samples have comparable average degrees of substitution. In addition, the N 1s spectra further confirmed the successful incorporation of post-synthetically introduced phthalocyanine and substituents. The pyrrolic nitrogen (398.8 eV) within the phthalocyanine macrocycles was clearly observed in the N 1s XPS spectra of all the GC-NiPc-R samples. Furthermore, the peak at 401.6 eV in the GC-NiPc-Py sample was assigned to pyridinic nitrogen, while the pronounced peak at 406.5 eV in the N 1s spectrum of GC-NiPc-NO₂ was attributed to the –NO₂ functional groups (Fig. 2b and S6). The relative peak intensities and corresponding nitrogen ratios are consistent with the expected molecular composition, further supporting the successful incorporation of the pyridyl and nitro substituents. The organic –Cl functional groups were also detected by Cl 2p_3/2 and 2p_1/2 peaks at 201.4 eV and 203.1 eV, respectively (Fig. S7), which are consistent with previous reports.²⁷

Fig. 2. X-ray photoelectron spectroscopy (XPS) measurement of GC-NiPc-R. (a) O 1s binding region showing the existence of C–O and C–O–C. (b) N 1s binding energy region representing the existence of the nitro group, pyridinic and pyrrolic nitrogen. (c) Ni 2p binding energy region confirming Ni²⁺ with a similar coordination environment. The right panel shows the local structure of the NiPc with functional groups.

Interestingly, only approximately half of the fluorine atoms from NiPcF₁₆ are substituted during the formation of GC-NiPc-R. This partial substitution is consistent with literature reports showing that fluorine atoms at the peripheral β-positions of perfluorinated phthalocyanines are more reactive toward nucleophilic substitution than those at the sterically hindered non-peripheral α-positions.^34,41 Based on these experimental observations, we propose a structural model for GC-NiPc-R in which preferential substitution occurs at the β-positions of the phthalocyanine. In this model, a pair of β-fluorides undergoes cyclization with catechol groups to form 1,4-dioxin linkages that covalently anchor the macrocycle to the graphite carbon surface, while the remaining accessible β-sites are functionalized with substituted catechols (–R groups). Nevertheless, since the spectroscopic measurements reflect averaged signals, the possibility of partial substitution at α-positions in a fraction of the phthalocyanine units cannot be completely excluded.

Electronic structure modulation of GC-NiPc-R

The coordination environment and electronic structure of the Ni center were further investigated by XPS and X-ray absorption spectroscopy (XAS). XPS measurements show Ni 2p_3/2 and 2p_1/2 peaks at 856.1 and 873.4 eV, respectively (Fig. 2c), corresponding to divalent Ni. The negligible shift in binding energy upon functionalization indicates that the oxidation state of Ni remains largely unchanged. The Ni K-edge X-ray absorption near-edge structure (XANES) spectra of physically supported NiPcF₁₆/C and covalently grafted GC-NiPc and GC-NiPc-Py all exhibit characteristic features of a square-planar D_4h coordination geometry (Fig. 3a).^42,43 The weak feature at approximately 8332 eV corresponds to the 1s → 3d transition, while the peak at around 8339 eV is attributed to the 1s → 4p transition. The slight changes in near-edge peak intensity observed for GC-NiPc and GC-NiPc-Py relative to NiPcF₁₆/C suggest subtle charge redistribution resulting from the covalent attachment of phthalocyanine to the graphite carbon substrate. The first derivative XANES spectra of NiPcF₁₆/C, GC-NiPc, and GC-NiPc-Py exhibit identical peak positions, indicating that the oxidation state of Ni remains unchanged upon covalent modification in all the samples (Fig. S8).

Fig. 3. (a) Ni K-edge XANES spectra and (b) EXAFS fitting results of NiPcF₁₆/C, GC-NiPc and GC-NiPc-Py. Wavelet transform of χ(k) of (c) NiPcF₁₆/C and (d) G C-NiPc-Py. (e) Square Wave Voltammetry (SWV) of GC-NiPc-R with different functional groups on GCE in 0.5 M KHCO₃/H₂O electrolyte, showing the reversible redox behavior of [NiPc]⁻/[NiPc]²⁻. (f) Redox potential of GC-NiPc-R determined from SWVs as a function of the Hammett constant of functional groups.

The extended X-ray absorption fine structure (EXAFS) spectra of the three NiPc-based catalysts exhibit a main peak at 1.4 Å (Fig. 3b and S9–S11), corresponding to the Ni–N coordination. The fitted coordination number of Ni–N is 4, indicating that the primary coordination environment of Ni remains unchanged after covalent modification. Additional features in the 2–3 Å range possibly arise from multiple scattering of Ni–N–C paths and Ni–Ni scattering associated with phthalocyanine stacking. Fitting analysis revealed Ni–Ni single scattering in NiPcF₁₆/C, whereas only Ni–N–C multiple scattering was present in GC-NiPc and GC-NiPc-Py (Table S3–S5). This distinction is further supported by wavelet transform (WT) analysis: NiPcF₁₆/C exhibits a pronounced intensity in the high-k region (∼8 Å⁻¹) at radial distances of 2–3 Å, characteristic of Ni–Ni scattering, whereas this feature is absent in GC-NiPc-Py (Fig. 3c, d, and S12). This observation further supports the presence of stacked NiPcF₁₆ in physically supported NiPcF₁₆/C. For comparison, this feature is absent in covalently grafted GC-NiPc-Py, suggesting a highly dispersed distribution of the Ni centers.

While XAS and XPS measurements reveal no significant changes in the oxidation state or coordination environment of the Ni centers, cyclic voltammetry (CV) provides a more sensitive probe of the electronic structure of the NiPc center. Both NiPcF₁₆/C and GC-NiPc display four distinct redox couples (Fig. S13). Previous studies on molecular NiPc generally assign these redox waves to ligand-centered processes associated with the phthalocyanine π-manifold.⁴⁴ pH-dependent measurements also revealed that the redox transition of NiPc is a proton-coupled process (Fig. S14). Accordingly, in the present GC-NiPc samples, the low-potential redox features centered at −0.01 and −0.08 V_RHE are attributed to the [NiPc]/[NiPc]⁻ and [NiPc]⁻/[NiPc]²⁻ couples, whereas the high-potential features centered at approximately 0.56 and 0.81 V_RHE are assigned to the [NiPc]/[NiPc]⁺ and [NiPc]⁺/[NiPc]²⁺ processes of the phthalocyanine ligand. The 1,4-dioxin linkage remains electrochemically inert within this potential window. Compared with physically supported NiPcF₁₆/C, covalently grafted GC-NiPc exhibits systematic shifts in the redox peak potentials. Furthermore, the peripheral functional groups in the GC-NiPc-R series (R = CH₃, H, CHO, COOH, NO₂, 4Cl, and Py) systematically tune the redox potentials. We focus on the reversible [NiPc]⁻/[NiPc]²⁻ redox couple located between −0.21 and −0.11 V_RHE (Fig. 3e and S15) as an electrochemical descriptor of the electronic structure of the NiPc moiety and its modulation by the peripheral substituents. Electron-withdrawing substituents shift this redox couple toward more positive potentials, whereas electron-donating substituents induce a negative shift (Fig. 3f, S15 and S16), consistent with previous studies of molecular macrocyclic complexes. However, the substituents are located farther from the Ni center, and their electronic effects are transmitted through an additional phenylene unit and the 1,4-dioxin linkage. Consequently, the magnitude of the redox-potential shifts is smaller than that typically observed for substituents directly attached to the macrocyclic framework.^6,45,46 The redox potential shows a linear dependence on the Hammett constants of the substituents, following the order: –Py > –4Cl > –NO₂ > –CHO > –H > –CH₃ (Fig. 3f and S15c), indicating that the electronic influence is attenuated but remains measurable through the extended conjugated framework.⁴⁶ This linear relationship suggests that functional groups modulate the electron density of the phthalocyanine π-system through substituent effects, without altering the primary Ni–N₄ coordination environment. Under the measurement conditions, the –COOH groups are expected to exist predominantly in their deprotonated –COO⁻ form. The redox potential of the [NiPc]⁻/[NiPc]²⁻ couple for the –COOH-functionalized sample deviates from the linear Hammett correlation observed for the other substituents. This deviation may arise from its negative charge and proton-binding ability, which may alter the local electrostatic environment and participate in proton transfer, shifting the redox potential beyond that predicted by its Hammett constant.

In previously reported graphite–pyrazine conjugates, strong electronic coupling hybridizes the molecular orbitals with the graphite band states, thereby suppressing discrete metal-centered outer-sphere redox peaks.^28,47,48 In addition, recent studies have shown that strong electrostatic coupling can also suppress discrete outer-sphere redox features when adsorbed macrocycle molecules reside within the electrochemical double layer.^49,50 In our graphite-conjugated phthalocyanines, we observe a reversible NiPc redox couple in the −0.21 to −0.11 V_RHE range, which is assigned to a NiPc redox process, indicating that the molecular redox states remain electrochemically addressable after surface attachment. Although the 1,4-dioxin junction is conjugated, its polarized C–O bonds likely provide weaker orbital coupling than a pyrazine linkage, limiting hybridization between the NiPc orbitals and graphite band states. Moreover, anchoring NiPc at the edges of the graphitic carbon may reduce the extent of electrostatic interaction between the NiPc and the carbon support compared with the macrocycle adsorbed directly onto the basal plane. Therefore, the edge-localized configuration, together with the limited electronic delocalization through the 1,4-dioxin linkage, may reduce both the electronic and electrostatic coupling between the NiPc macrocycle and the graphitic substrate and contribute to the preservation of the molecular redox states in the GC-NiPc system. The retention of the molecular redox couple further enables direct modulation of the molecular electronic structure through peripheral substituent engineering, providing a platform for systematically tuning the catalytic properties of the immobilized NiPc catalysts.

Optimizing the CO₂RR activity of GC-NiPc-R by substituent engineering

The CO₂RR activity was evaluated using linear sweep voltammetry (LSV) and chronoamperometry (CA) measurements, with a hydrophilic carbon cloth (1 mg cm_geo⁻²) serving as the working electrode in an H-cell containing CO₂-saturated 0.5 M KHCO₃. Control LSV measurements under an N₂ atmosphere showed substantially suppressed cathodic currents compared with those recorded under CO₂ (Fig. 4a and S17–S25), confirming that the enhanced current response mainly originates from the CO₂RR rather than background hydrogen evolution. LSV measurements revealed cathodic currents for GC-NiPc with an onset potential of −0.58 V_RHE (Fig. S26). In contrast, the NiPcF₁₆/C sample, prepared by physically adsorbing NiPcF₁₆ onto acetylene black to achieve comparable Ni loading, exhibited lower electrocatalytic activity. The enhanced activity of the GC-NiPc electrodes can be attributed to the well-dispersed molecular NiPc active sites, as the covalent attachment at the graphite edge promotes a higher density of exposed active sites and facilitates efficient electron transport between the electrode and catalytic centers. The introduction of electron-withdrawing and electron-donating groups (R = CH₃, H, CHO, COOH, NO₂, 4Cl, and Py) enabled continuous tuning of the CO₂RR onset potential from −0.63 to −0.48 V_RHE across the GC-NiPc-R series (Fig. S23). In particular, the –COOH, –NO₂, and –Py functional groups induced positive shifts of 20, 30, and 100 mV in the onset potential, leading to enhanced CO₂RR activity compared to GC-NiPc. Notably, the CO₂RR activity exhibits a clear correlation with the [NiPc]⁻/[NiPc]²⁻ redox couple, which reflects the substituent-dependent electronic structure of the catalyst (Fig. 4b).

Fig. 4. CO₂RR activity of GC-NiPc-R in CO₂-saturated 0.5 M KHCO₃ solution. (a) Linear sweep voltammetry of GC-NiPc-R at a scan rate of 10 mV s⁻¹. (b) Correlation between j_CO at −0.65 V_RHE and the redox peak potential. (c) Mass spectra of ¹³CO detected during the CO₂RR under a ¹³CO₂ atmosphere for GC-NiPc-Py. Comparison of TOF_CO, FE(CO) and FE(H₂) for (d) GC-NiPc and NiPcF₁₆/C and (e) GC-NiPc-R under different potentials. (f) Correlation between TOF_CO of GC-NiPc-R at −0.65 V_RHE and the redox peak potential. All potentials are reported versus the reversible hydrogen electrode without iR correction. Error bars represent the standard deviation from three independent measurements.

The similar Tafel slopes observed for all catalysts (Fig. S27) suggest a common reaction mechanism with a shared rate-determining step.⁵¹ However, the Tafel slopes obtained for GC-NiPc-R are relatively high compared with the theoretical value expected for a purely kinetically controlled rate-limiting electrochemical PCET step. Because the catalysts were evaluated on porous carbon-paper electrodes, CO₂ transport within the porous electrode and across the gas–liquid–solid interface may contribute to the measured current. Therefore, the measured slopes should be regarded as apparent Tafel slopes, and mixed kinetic–mass-transport control cannot be excluded. Electrochemical impedance spectroscopy (EIS) further supported this trend. Nyquist plots reveal decreasing charge-transfer resistance with increasing electron-withdrawing character of the substituents, indicating facilitated interfacial electron transfer (Fig. S28). To exclude surface area effects, electrochemically active surface area (ECSA), estimated from double-layer capacitance measurements (Fig. S29), was found to be comparable among GC-NiPc-Py, GC-NiPc, and NiPcF₁₆/C. Moreover, quantitative integration of the redox peaks for GC-NiPc and NiPcF₁₆/C (Fig. S30) reveals that, at identical loadings, GC-NiPc exhibits a slightly higher number of electrochemically accessible reduced sites than NiPcF₁₆/C, consistent with suppressed aggregation and enhanced site exposure. Meanwhile, the surface area-normalized kinetic current density at −0.65 V_RHE increases with increasing [NiPc]⁻/[NiPc]²⁻ redox peak potential (Fig. 4b and S31). This observation rules out geometric effects and confirms that the activity enhancement originates from electronic modulation induced by substituents via covalent graphite conjugation.

Chronoamperometry measurements were carried out at various potentials to quantify product distributions. Gaseous products were quantified by gas chromatography, and liquid products were analyzed by NMR spectroscopy. Carbon monoxide (CO) was identified as the primary product of the CO₂RR, whereas hydrogen (H₂) was detected as a by-product from the hydrogen evolution reaction (HER). To verify that the generated CO originated from the electrocatalytic CO₂ reduction process, isotopic ¹³CO₂ is employed as the carbon source in the reaction. The gaseous products were subsequently analyzed by mass spectrometry. A distinct signal at m/z = 29, corresponding to ¹³CO, demonstrates that the carbon source of CO is indeed from the CO₂ source (Fig. 4c). All the NiPc-based catalysts exhibit high CO selectivity, maintaining the FE(CO) above 90% within −0.55 to −0.8 V_RHE (Fig. 4d, e, S32 and S33). Notably, GC-NiPc exhibited an outstanding FE(CO) of more than 97% across the potential range of −0.65 to −0.8 V_RHE, surpassing the performance of physically adsorbed NiPcF₁₆/C. Upon incorporation of the pyridine functional group, GC-NiPc-Py further enhanced FE(CO) of above 99% over a wide potential range. To systematically evaluate substituent effects, product distributions at −0.65 V_RHE were analyzed. All the GC-NiPc-R products exhibited a high FE(CO) (above 97%), except for the electron-donating –CH₃ derivative, which showed a noticeable decrease (Fig. S34).

Comparison of the electrochemically accessible NiPc sites in NiPcF₁₆/C and GC-NiPc at different catalyst loadings revealed that increasing the areal loading promotes aggregation/stacking of catalyst particles (Fig. S30), which can partially hinder electrolyte penetration into the inner regions of the catalyst layer, thereby reducing the fraction of electrochemically accessible active sites. The turnover frequencies (TOFs) for CO formation were therefore calculated based on both the total NiPc population and the electrochemically accessible NiPc population to distinguish the effects of site accessibility from the catalytic activity of the accessible sites. Notably, GC-NiPc exhibits a higher TOF than physisorbed NiPcF₁₆/C even when normalized to the electrochemically accessible NiPc population (Fig. S35), supporting enhanced catalytic activity per electrochemically accessible NiPc site upon covalent conjugation. Furthermore, the TOF increases systematically with the electron-withdrawing strength of the substituents, with GC-NiPc-Py reaching 846 h⁻¹ at −0.65 V_RHE (Fig. 4f and S34). The logarithmic TOF values show an approximately linear increase with the Hammett constants of the substituents (Fig. 4f). This trend closely follows the variation in redox potential and LSV activity discussed above, establishing a clear correlation between substituent-induced electronic modulation and catalytic kinetics. In addition to electronic effects, pyridyl moieties have also been reported to promote local CO enrichment through secondary interactions and to assist proton delivery to reaction intermediates during the CO₂RR.⁵² While the proton-relay effect is not explicitly investigated in this work, such secondary coordination and microenvironmental effects may further contribute to the enhanced activity observed for GC-NiPc-Py. Covalent anchoring also significantly improves catalyst stability. At −0.65 V_RHE, GC-NiPc and GC-NiPc-Py retained stable CO partial current densities over 10 h, whereas physisorbed NiPcF₁₆/C rapidly decreased within 0.5 h (Fig. S36). Even at −0.8 V_RHE, GC-NiPc-Py retains a CO selectivity above 99% (Fig. 4e), demonstrating its excellent stability under highly reducing conditions.

Mechanistic insights into the CO₂RR on GC-NiPc-R

To elucidate the evolution of intermediates during the CO₂ reduction reaction (CO₂RR), in situ attenuated total reflection infrared (ATR-IR) spectroscopy was performed on the GC-NiPc-Py catalyst under applied potentials, enabling direct monitoring of potential-dependent intermediate formation along the reaction pathway. As shown in Fig. 5a, the C–O stretching and O–H vibrational bands characteristic of the *COOH intermediate separately emerged at 1372 cm⁻¹ and 1312 cm⁻¹, respectively, appearing at potentials more negative than −0.20 V_RHE. In addition, a progressively intensified band centered at 1930 cm⁻¹ was observed under more negative potentials (−0.45 V_RHE), which is attributed to the C O stretching vibration of the *CO intermediate adsorbed on the catalyst surface. The bands observed at approximately 1630 cm⁻¹ are attributed to the bending vibration mode of H₂O.¹⁰ Time-resolved monitoring at −0.65 V_RHE (Fig. 5b) reveals the rapid emergence of the *CO signal and a gradual increase in the *COOH-related bands, indicating continuous formation and accumulation of this key intermediate during CO₂ reduction. As the reaction proceeded, the intensity of the *COOH intermediate peak gradually increased, indicating the sustained formation and accumulation of this key reaction intermediate during CO₂ reduction. After returning to open-circuit conditions and rinsing of the electrode surface with 0.5 M KHCO₃, the ATR-IR spectral features (Fig. S37) returned to their initial state, demonstrating that the catalyst retained its structure after electrochemical operation, thereby indicating that the catalyst retained its structural integrity under the applied reaction conditions.

Fig. 5. In situ ATR-IR spectra of GC-NiPc-Py during the CO₂RR: (a) collected over the potential range from 0.6 to −1.4 V_RHE and (b) long-term monitoring under −0.65 V_RHE. Calculated electron density distribution of the HOCO and LUCO for (c and d) NiPcF₁₆/C and (f and g) GC-NiPc. Partial density of states of (e) GC-NiPc-H and (h) GC-NiPc-Py. (i) Calculated free energy diagrams of NiPcF₁₆/C, GC-NiPc, GC-NiPc-H and GC-NiPc-Py for the CO₂RR.

To further understand the electronic origin of the activity trend, DFT calculations were performed. For physically adsorbed NiPcF₁₆/C, the highest occupied crystal orbital (HOCO) and lowest unoccupied crystal orbital (LUCO) are primarily localized on the graphite layers (Fig. 5c, d, and S38), indicating inefficient electronic communication between the carbon support and the NiPc unit. In covalently grafted GC-NiPc, the LUCO becomes predominantly localized on the NiPc moiety (Fig. 5f, g, and S39), suggesting more efficient electron injection into the NiPc catalytic center, which provides a plausible explanation for the enhanced CO₂RR activity observed for GC-NiPc relative to physisorbed NiPcF₁₆/C. Following the introduction of phenyl (–H) and pyridyl (–Py) functional groups, no significant alteration in the HOCO and LUCO distributions was observed (Fig. S40 and S41). However, the –Py functionalized system exhibits a reduced band gap compared with the –H analogue (Fig. 5e and h), accompanied by a downward shift of the LUCO energy level (Fig. S42, referenced to the vacuum level). This is consistent with the experimentally observed positive shift in reduction potential from CV. Furthermore, Mulliken population analysis (Fig. S43) revealed the development of distinct electron-rich and electron-deficient environments around the Ni center for the –H and –Py modified systems, respectively.⁶ These results collectively demonstrate that substituent-induced electronic modulation tunes both the local charge distribution and frontier orbital energetics of the catalytic center.

DFT calculations were further performed to construct the Gibbs free energy profile of the CO₂RR on NiPc-based catalysts (Fig. 5i). The CO₂RR at the Ni center proceeds via a proton-coupled electron transfer (PCET) process to form the *COOH intermediate, which is identified as the rate-determining step of the reaction.⁵³ Subsequently, *COOH undergoes a second PCET step to form adsorbed *CO, followed by desorption to release CO and regenerate the active site, completing the catalytic cycle. Upon grafting NiPcF₁₆ onto the graphite edge, the free energy barrier for *COOH formation, ΔG(*CO₂ → *COOH), decreases slightly from 1.73 eV to 1.69 eV, indicating a modest thermodynamic stabilization of the key intermediate. To further elucidate substituent effects, Gibbs free energy profiles for the CO₂RR on GC-NiPc bearing –H and –Py substituents were also calculated. Substitution with –H increases ΔG(*CO₂ → *COOH) to 1.88 eV, likely because replacing electron-withdrawing F atoms with dioxin-linked phenyl groups increases the electron density of the NiPc center and destabilizes *COOH formation. By contrast, the more electron-withdrawing –Py substituent decreases ΔG(*CO₂ → *COOH) to 1.64 eV, facilitating the rate-determining step. These results demonstrate that electron-withdrawing substituents effectively modulate the rate-determining step by stabilizing the *COOH intermediate, thereby accelerating CO formation. Together, the electrochemical and computational results demonstrate that molecular-level electronic programming of graphite-conjugated NiPc catalysts governs their CO₂RR kinetics.

Conclusions

In summary, we report a series of graphite-conjugated NiPc-R catalysts in which electronically tunable functional groups enable systematic modulation of Ni–N₄ active sites. Covalent dioxin linkages enhance electrochemical stability and preserve the modular electronic tunability of the metal centers. The redox potential of the NiPc core was systematically tuned, which directly translates into predictable modulation of CO₂RR activity. Notably, pyridine-functionalized GC-NiPc achieves both high activity and above 99% CO faradaic efficiency over a wide potential window, together with enhanced long-term operational stability. In situ ATR-IR and DFT calculations confirmed the rate-determining formation of *COOH intermediates and revealed that electron-withdrawing groups lower the free energy barrier for the initial proton-coupled electron transfer step. These findings establish a versatile and modular platform for the rational design of molecularly well-defined, graphite-supported M–N₄ catalysts with tunable active-site energetics. More broadly, this strategy offers a general design principle for translating molecular catalysts into robust, electronically programmable heterogeneous platforms for diverse energy-conversion reactions.

Author contributions

Chenyu Zhu: conceptualization, investigation, formal analysis, visualization, writing – original draft, methodology, formal analysis, software; Lei Gao: formal analysis, software; Yulun Wu: investigation; Yu-Hao Gu: investigation; Xinyu Xu: investigation; Ya Yin: investigation; Hai-Xu Wang: review & editing; Shuai Yuan: conceptualization, writing – review & editing, project administration, supervision, funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Data availability

Supplementary information (SI): materials, instrumentation details, synthesis procedures, DFT calculation details, and additional data, including IR, TEM, TGA, XPS, XAS, ICP-OES, and electrochemical measurements. See DOI: https://doi.org/10.1039/d6sc06126e.