Porosity Engineering Within Ni─N─C Hollow Spheres for Ampere‐Level CO₂ Reduction Electrocatalysis

ABSTRACT

CO₂ electroreduction typically requires large overpotentials to sustain high reaction rates, which at current densities approaching 1 A cm⁻² inevitably intensify competition from the hydrogen evolution reaction (HER), making it difficult to simultaneously sustain high product selectivity (>95%). In this study, hollow Ni─NC (H‐Ni‐NC) was synthesized using nanoscale silica sphere templates and atomic Zn as a sacrificial pore‐former within the carbon shell. The optimized H‐Ni‐NC achieves a current density of −1.0 A cm⁻² with over 95% CO Faradaic efficiency in a flow cell. The high performance was attributed to the hollow sphere architecture enriched with gas‐permeable through‐pores. Notably, the Zn dosage selectively modulates the shell through‐porosity, without altering hollow sphere macrostructure or the active‐site structure, thereby enabling a systematic investigation of porosity effects. Finite‐element simulations provide a qualitative framework suggesting a trade‐off between reactant transport and active surface area with increasing porosity in hollow spherical catalysts. This work underscores that engineering of the nanoscale mass‐transport environment surrounding active sites plays a critical role in designing high‐efficiency electrocatalysts.

Atomic Zn regulates shell through‐porosity in hollow Ni─NC, enabling direct control over the local mass‐transfer environment during CO₂ electroreduction. Low porosity restricts reactant transport and limits access to catalytic sites, whereas increased porosity establishes continuous pathways that facilitate reactant diffusion and product removal.

1. Introduction

Anthropogenic CO₂ emissions currently outpace natural sequestration processes [1, 2]. To address this imbalance, the electrochemical conversion of CO₂ into value‐added chemicals and fuels has emerged as a promising route for closing the carbon cycle and storing renewable energy [3, 4]. Among various products, CO is a key industrial feedstock for the Fischer–Tropsch process [5]. The sluggish kinetics of CO₂ electroreduction arise not only from the intrinsic difficulty of CO₂ activation, but also from mass transport limitations at high reaction rates [6]. Even for intrinsically active catalysts with early onset potentials, increasing the current density to industrially relevant (e.g., >200 mA cm⁻²) inevitably requires large overpotentials, under which hydrogen evolution reaction (HER) is strongly promoted, making it challenging to simultaneously achieve high current density and high CO selectivity [7, 8]. Meeting this challenge requires electrocatalysts with not only high intrinsic activity and sufficient active‐site density, but also, crucially, superior mass transport properties at high reaction rates [5, 9, 10, 11, 12].

In general, mass transport limitations in CO₂ electroreduction have been addressed through two complementary strategies: polymer‐assisted interfacial modification [13, 14, 15] and geometric regulation [16, 17, 18, 19] of catalyst architectures. Polymer‐based modification, exemplified by hydrophobic additives such as polytetrafluoroethylene (PTFE), can enhance gas transport and suppress the HER. However, it may simultaneously hinder proton transport and obscure active sites, thereby limiting the extent of current density enhancement. On the other hand, geometric regulation of catalyst structures can enhance the accessibility of active sites and facilitate the transport properties of reactant species. Importantly, it also establishes the fundamental framework for efficient three‐phase transport, upon which interfacial tuning can provide additional performance gains.

Single‐atom catalysts (SACs), particularly those with metal‐N₄ coordination embedded in carbon matrices (M─N─C, M═Ni, Fe, and Co), have emerged as leading candidates for the CO₂‐to‐CO conversion due to their near‐unity selectivity and high atom utilization efficiency [20, 21, 22, 23, 24, 25]. To further boost their performance under mass transport‐limited conditions, geometric regulation of the carbon support, especially through hierarchically porous frameworks, has been widely employed [26, 27, 28, 29, 30, 31, 32]. However, the synthesis of such well‐defined hierarchies is challenging because the formation of different pore types is intrinsically coupled. More importantly, the effects of geometric regulation are often researched only through overall electrocatalytic performance. Such performance metrics are influenced by multiple factors, including electrochemically active surface area, active‐site density and structure, mass transport, and ionic conduction. As a result, it is difficult to isolate and normalize the contribution of pore architecture, and clear structure‐transport‐performance correlations. This gap in control has led to a fundamental, yet underexplored, question in catalyst design: how to selectively engineer the pore network around active sites to enhance gas diffusion and overall catalytic performance while largely preserving the overall structural framework of the catalyst?

In this study, we address this challenge by developing a cooperative‐templating strategy for engineering shell through‐porosity in hollow carbon spheres hosting atomically dispersed Ni–N₄ sites (denoted as H‐Ni‐NC). Our approach employs nanoscale silica spheres to define the hollow spherical morphology and, innovatively, uses atomic zinc as a sacrificial template to tailor the nano‐porosity of the carbon shell. By simply varying the Zn dosage, we achieve orthogonal control over the shell's pore architecture without affecting the sphere's diameter, shell thickness, or the local coordination of Ni sites. The catalyst (H‐Ni‐NC‐250) achieves an ampere‐level current density (−1.0 A cm⁻²) with >95% CO Faradaic efficiency and exhibits robust stability. Gas permeation measurements demonstrate that engineered shell porosity can modulate mass‐transport resistance. The finite‐element simulations further provide a qualitative framework suggesting a trade‐off between reactant transport and available active sites with increasing porosity. This work moves beyond reporting an active catalyst and highlights the importance of engineering the nanoscale mass‐transport environment for high‐current‐density electrocatalysis.

2. Results and Discussion

2.1. Cooperative‐Templating Strategy for Pore Structure Engineering

As shown in Figure 1a, a cooperative‐templating strategy using nanoscale silica spheres and atomic Zn was employed to synthesize hollow Ni–NC (H‐Ni‐NC) materials with tunable through‐pores in the carbon shell. In a preliminary screening, silica sphere diameters of 200, 500, and 1000 nm were evaluated (Figure S1), and the 200 nm spheres exhibited the best electrochemical performance. Because altering the diameter unavoidably convolutes various structural parameters, such as specific surface area, shell thickness, pore architecture, and interparticle packing geometry, we chose to fix the sphere size at 200 nm in this study to investigate the specific effects of shell porosity engineering. The synthesis involved three key steps. In the first step, Zn²⁺/Ni²⁺‐catalyzed polymerization of formamide (FA) was conducted using homemade silica nanospheres (∼200 nm, Figure S2) as a template to obtain a poly‐formamide‐coated SiO₂ precursor (NiZn‐pFA@SiO₂) [25]. Subsequently, pyrolysis at high temperature converted this precursor into Ni‐NC@SiO₂. During this process, the Zn species, serving as an atomic template, were removed and thereby generated through‐pores within the carbon shell. Benefiting from their ability to volatilize at elevated calcination temperatures, Zn species are widely utilized as polymerizing nodes and architectural “fences” to isolate target single atoms. As documented in previous reports, this volatilization process simultaneously drives the formation of porous structures. Finally, etching away the SiO₂ template yielded the hollow‐structured Ni–NC (H‐Ni‐NC) material. By varying the amount of Zn added, a series of H‐Ni‐NC materials with tunable porosity were synthesized (Figure S3), denoted as H‐Ni‐NC‐n (n = 200, 250, and 300), where ‘n’ represents the concentration (mM) of Zn²⁺ in the FA during the solvothermal step.

FIGURE 1.

Synthesis and structural characterizations of H‐Ni‐NC‐250 catalyst. (a) Schematic of the synthesis for hollow Ni–NC materials featuring tunable through‐pores within the carbon shell. (b) TEM image, (c) HAADF‐STEM image with its corresponding EDS elemental maps, and (d,e) aberration‐corrected HAADF‐STEM images of the H‐Ni‐NC‐250 material. (f) Normalized Ni K‐edge XANES spectra for the H‐Ni‐NC‐250 material and reference samples. (g) Fourier‐transformed R‐space EXAFS spectra and the corresponding EXAFS fitting curves (dotted lines) at the Ni K‐edge for H‐Ni‐NC‐250 material and reference samples (inset: schematic of the Ni–N₄ doped graphene structural model used for EXAFS fitting; grey, purple, and brown spheres represent Ni, N, and C atoms, respectively).

2.2. Hollow‐Structured Carbon Spheres With Isolated Ni–N₄ sites

As shown in Figure 1b, the H‐Ni‐NC‐250 material exhibits a uniform hollow spherical structure with an average diameter of 210 nm. High‐resolution transmission electron microscopy (HRTEM) images reveal a thin carbon shell of approximately 5 nm in thickness, containing both through and blind micropores (Figure 2f). Inductively coupled plasma optical emission spectrometry (ICP‐OES) and x‐ray photoelectron spectroscopy (XPS) results confirm the successful incorporation of Ni, with loadings of 4.79 and 4.50 wt.%, respectively, (Figure S10 and Table S2). The absence of diffraction peaks corresponding to metallic Ni or other crystalline phases in the XRD pattern (Figure 2a), together with the lack of visible nanoparticles in high‐resolution TEM images (Figure S4), indicates that no large aggregates are present. High‐angle annular dark‐field scanning transmission electron microscopy (HAADF‐STEM) imaging and corresponding EDS elemental mapping further demonstrate the homogeneous distribution of Ni throughout the carbon shell (Figure 1c). At the atomic scale, isolated bright spots in the aberration‐corrected HAADF‐STEM image (Figure 1e) provide direct evidence for the atomic dispersion of Ni species. Notably, a concentration gradient of Ni is observed from the interior to the exterior of the shell (Figures 1d, and S5), suggesting that active sites are available on both the inner and outer surfaces, with a higher density of active sites located in the inner layer. Within such an inner‐shell‐enriched Ni distribution, the introduction of shell through‐porosity can significantly increase the accessibility of the reactant gas to the interior of the shell, thereby boosting the utilization efficiency of these inner active sites.

FIGURE 2.

Effect of Zn dosage on material structure. (a) XRD patterns, (b) Ni 2p XPS spectra, (c) TEM images, (d) N₂ adsorption–desorption isotherms, and (e) pore volume statistics of H‐Ni‐NC‐200, H‐Ni‐NC‐250, and H‐Ni‐NC‐300 materials. (f) HR‐TEM image of H‐Ni‐NC‐250 material.

The oxidation state and local coordination environment of the Ni sites in the H‐Ni‐NC‐250 material was further investigated by x‐ray absorption spectroscopy (XAS). As shown in Figure 1f, the Ni K‐edge x‐ray absorption near‐edge structure (XANES) spectra reveal that the absorption edge of H‐Ni#x02010;NC‐250 is positioned close to that of Ni²⁺ in NiO and distinctly higher than that of Ni⁰ in the Ni foil, indicating an oxidation state near +2 for the Ni species [28, 29, 30]. The Fourier‐transform k²‐weighted extended x‐ray absorption fine structure (EXAFS) spectrum of H‐Ni‐NC‐250 exhibits a prominent peak at ∼1.35 Å, which corresponds to the first‐shell Ni–N coordination [33, 34, 35]. The absence of a Ni─Ni scattering peak near 2.2 Å further confirms the atomic dispersion of Ni without metallic aggregation [36, 37]. Quantitative EXAFS fitting clearly revealed the local coordination environment. The optimal fit for the H‐Ni‐NC‐250 spectrum was achieved with Ni─N scattering paths (Figure 1g). A first‐shell coordination number of ∼4, as derived from the fitting (Table S1), confirms that the Ni sites predominantly adopt a Ni–N₄ configuration.

2.3. Atomic Zn‐Mediated Through‐Pore Engineering

Comprehensive characterizations were performed to investigate the effect of Zn dosage on the H‐Ni‐NC‐n materials. As shown in Figure 2a, the XRD patterns of all three H‐Ni‐NC‐n samples exhibit similar diffraction peaks corresponding to graphitic carbon, with no shift in position or change in intensity. This indicates that the crystalline structure of the carbon matrix is not influenced by the Zn concentration. Furthermore, the absence of diffraction peaks for metallic Ni (PDF#45‐1027) rules out the formation of Ni nanoparticles, corroborating the results in Figure 1. Raman spectra (Figure S6) show characteristic D and G bands at approximately 1370 and 1544 cm⁻¹, respectively. The intensity ratio of the D to G band (I _D/I _G) serves as a robust indicator of structural defect levels. The nearly identical I _D/I _G ratios across different Zn loadings further confirm the structural consistency of the graphitic carbon framework.

The electronic structure was further investigated by XPS. As shown in Figure 2b, the high‐resolution Ni 2p spectra exhibits characteristic binding energies at 872.5 and 855.2 eV [38], indicating a valence state close to Ni²⁺, which matches well with the XANES findings (Figure 1f). Notably, the practically invariant peak positions across all H‐Ni‐NC samples confirm that they share highly similar local electronic structures. For nitrogen speciation, the N 1s fitting shows that the N species exist in 5 different coordination environments across the samples (Figure S7). Except for the M‐N signal intensity, which naturally rises as the Ni content increases, the rest of the nitrogen species show no obvious differences (Figure S8 and Table S3). Moreover, the positions of the N 1s and C 1s (Figure S9) peaks remain unchanged with varying Zn dosage, further confirming the stability of the electronic environment in the N‐doped carbon matrix.

The influence of Zn dosage on the hierarchical pore structure was also studied. TEM (Figures 2c and S14) and SEM (Figure S3) images reveal that the spherical diameter (∼200 nm) and shell thickness (∼4.5 nm) of the catalysts are not affected by Zn content. Although the overall hollow spherical morphology is preserved, N₂ adsorption–desorption isotherms (Figure 2d) show clear differences, indicating Zn‐dependent tuning of the shell porosity. The Type IV isotherms with H3 hysteresis loops reflect a hierarchical pore architecture [39], comprising narrow slit‐shaped micropores and macropores. The narrow slit‐shaped micropores enable gas transport into the carbon shell, and the macropores act as internal cavities. Higher Zn dosage reduces the total N₂ uptake, indicating a decrease in the surface area accessible to N₂. The pore size distribution further reveals a reduction in micropore volume with increasing Zn content (Figures 2e, S15, and Table S4). Combined with the nanoscale contrast variations resembling pore‐like features in the carbon shell via HRTEM image (Figure 2f), it can be inferred that low Zn concentrations generate isolated micropores in the shell, while further increases in Zn content cause these micropores to merge and form interconnected through‐pores. Besides, XPS survey spectra (Figure S11) and ICP‐OES data (Figures S12 and S13) indicate that the enhancement in through‐pore porosity is accompanied by a gradual loss of active sites.

2.4. Zn‐Dependent Tuning of CO₂RR Efficiency

Based on the results presented herein, this study demonstrates that tuning the Zn dosage enables modulation of the shell through‐porosity within the carbon shells while largely preserving the diameter and shell thickness of the hollow spheres. In conventional hierarchical porous materials, pore structure modification typically leads to concurrent changes in micro‐/mesopores and macropores, making it difficult to isolate the effects of specific pore types. In contrast, our synthesis strategy enables selective engineering of micropores while preserving the macroporous framework intact. This approach provides a model system to study the influence of shell porosity from other morphological variables, which we systematically investigated using the CO₂ reduction reaction (CO₂RR) as a probe.

The electrocatalytic CO₂RR performance of the H‐Ni‐NC‐n (n = 200, 250, and 300) catalysts was evaluated in both an H‐cell (Figure S16) and a flow cell (Figure S23). The polarization curves obtained in the flow cell (Figure 3a and S18a) show that H‐Ni‐NC‐200 reaches current density of −0.7 A cm⁻², while H‐Ni‐NC‐250 and H‐Ni‐NC‐300 achieve current densities of −1.0 and −0.9 A cm⁻², respectively. Notably, achieving ampere‐level current densities remains a considerable challenge in CO₂RR research (Figure S19 and Table S5), indicating that porosity engineering in hollow nanospheres plays an important role in performance improvement (Note S1). Across the entire tested current range, all three catalysts maintain CO Faradaic efficiencies above 90% (Figures 3b, S17, and S18b). It is worth noting that the local pH and carbonate species may play a critical role in CO₂RR at high current densities, which may also influence the Faradic efficiency (Note S2) [40, 41, 42].

FIGURE 3.

Electrochemical CO₂RR performance of H‐Ni‐NC‐200, H‐Ni‐NC‐250, and H‐Ni‐NC‐300 catalysts. (a) Polarization curves. (b) Faradaic efficiency (FE) for CO. (c) ICP‐normalized nominal TOF values for CO production. (d) Applied potential and the corresponding TOF measured at current densities of −0.1 and −0.7 A cm⁻² in flow cell. Flow‐cell CO₂RR measurements were conducted using 1 M KHCO₃ electrolyte with a flow rate of 20 mL min⁻¹ and a CO₂ flow rate of 20 sccm, and the reported potentials were obtained after iR correction. (e) Long‐term stability test of H‐Ni‐NC‐250 conducted in a membrane‐electrode‐assembly (MEA) cell with humidified CO₂ (20 sccm) at the cathode and 0.1 M KHCO₃ electrolyte at the anode with an electrolyte flow rate of 20 mL min⁻¹; inset shows a schematic of the MEA‐cell configuration.

The overpotential reflected in polarization curves is generally composed of three distinct contributions: kinetic overpotential, ohmic overpotential, and mass transport overpotential. Consequently, to further decouple the kinetic and mass transport resistance, we performed a Tafel slope analysis on the low‐current region of these H‐Ni‐NC polarization curves (Figure S27a). Based on the derived Tafel slopes and intercepts, the kinetically controlled polarization curves were extrapolated (represented by the light‐colored curves in Figure S27b–d). Since the experimental polarization curves (the dark‐colored curves in Figure S27b–d) had already been corrected for high‐frequency resistance to eliminate the ohmic overpotential contribution, the discrepancy between these two curves directly represents the mass transport overpotential. It can be clearly observed that the mass transport resistance of H‐Ni─NC decreases significantly with increasing Zn dosage.

To elucidate the relationship between Zn dosage and catalytic performance, we first evaluated the double‐layer capacitance (C _dl) of the H‐Ni‐NC‐n series (Figure S20) to compare their electrochemically active surface areas (ECSAs). The results show that C _dl decreases with increasing Zn dosage, reaching the lowest value of 12.1 mF cm⁻² for H‐Ni‐NC‐300. Nitrite stripping measurements was performed to estimate the relative density of electrochemically accessible sites [43, 44], revealing a similar decreasing trend in the electrochemically accessible of Ni active sites with increasing Zn dosage (Figure S24 and Table S6). Together with the BET results, these observations indicate that excessive Zn facilitates the interconnection of micropores to form through‐pores at the expense of total surface area and accessible active‐site density. It should be noted that, despite having a lower surface area and active‐site density than H‐Ni‐NC‐200 catalyst, the H‐Ni‐NC‐250 catalyst achieves a considerably higher current density (−1 A cm⁻²), showing that gas diffusion is an important contributor under high‐current conditions.

To directly investigate the impact of gas diffusion on CO₂RR performance, turnover frequency (TOF) values were calculated to account for differences in active‐site density [45]. In the Figure 3c, the nominal TOF values were estimated by normalizing the CO production rate to the total Ni content determined by inductively coupled plasma optical emission spectroscopy (ICP‐OES). Nevertheless, the TOF values continue to increase with increasing shell through‐porosity. In particular, the H‐Ni‐NC‐200 catalyst exhibits a tendency toward TOF saturation at high current density, whereas the more porous H‐Ni‐NC‐250 and H‐Ni‐NC‐300 catalysts continue to show increasing TOF values, suggesting that enhanced local gas accessibility and alleviated transport limitations facilitate more effective utilization of active sites under high‐current‐density conditions. In addition, the TOF values normalized by electrochemically accessible Ni sites estimated from nitrite stripping measurements (Figure S26) exhibit higher absolute values while maintaining a trend consistent with the ICP‐normalized nominal TOF values. Figure 3d illustrates that gas diffusion is not the dominant limitation at a low current density of −0.1 A cm⁻², as reflected by the similar potentials of the three catalysts, indicating their comparable intrinsic activity [27]. In contrast, at a high current density of −0.7 A cm⁻² where gas‐transport limitations become pronounced, the potential of H‐Ni‐NC‐200 is significantly higher than that of the other two samples, despite its higher active site density, highlighting the important role of enhanced gas access in boosting the reaction kinetics. Nevertheless, while H‐Ni‐NC‐300 achieves a higher TOF, it exhibits an overpotential comparable to that of H‐Ni‐NC‐250 and a lower limiting current density. This suggests that the reduction in active‐site density associated with higher porosity partially offsets the advantages of enhanced gas transport. The H‐Ni‐NC‐250 catalyst exhibits the highest current density among the catalysts investigated, reaching −1.0 A cm⁻² and demonstrating excellent CO₂RR performance.

To further probe the interfacial electrochemical behavior of the catalysts, electrochemical impedance spectroscopy (EIS) measurements were performed. As shown in Figure S24b, the DRT result reveals five distinct relaxation processes. In line with earlier literature [29, 46], the prominent peak at low frequencies (1–10 Hz) corresponds to macroscopic mass transport. Our data shows that the intensity of this mass transport feature drops continuously as the Zn dosage increases, pointing to reduced mass transport resistance within the more porous hollow frameworks, which is consistent with the results derived from the polarization curve decoupling analysis. Meanwhile, the mid‐frequency peaks (10–1000 Hz) are generally assigned to the apparent proton‐coupled electron transfer (PCET) steps in CO₂RR, which reflect the intrinsic kinetic effects. Notably, the integrated areas of these mid‐frequency γ peaks are very similar to each other, aligning with the coordination environment and electronic structures determined from our characterization data (Figures 2a, b and S6–S10).

A membrane electrode assembly (MEA) was further constructed to evaluate the catalyst stability. As shown in Figure S21, the MEA equipped with the H‐Ni‐NC‐250 cathode achieved a current density of 0.45 A cm⁻² at a cell voltage of 3.8 V. Furthermore, during a 100 h stability test conducted at 100 mA cm⁻², the system showed no noticeable performance decay while maintaining a CO Faradaic efficiency consistently above 90%, demonstrating the good stability of the H‐Ni‐NC‐250 catalyst.

2.5. Gas‐Diffusion Enhancement via Through‐Pore Engineering

As discussed above, increasing the Zn content promotes the connection of micropores into through‐pores within the shell, thereby enhancing gas diffusion, a key factor in achieving high current densities. To quantify the influence of the through‐pore architecture on mass transport, the CO₂ permeation performance was evaluated by fabricating the catalysts into GDEs with a loading of 0.5 mg cm⁻². The assessment was performed using a homemade permeation setup (Figure 4a). As shown in Figure 4b, the measured gas permeability decreased in the order: H‐Ni‐NC‐300 (9605 GPU) > H‐Ni‐NC‐250 (9327 GPU) > H‐Ni‐NC‐200 (8275 GPU). By comparing GDEs with catalyst loading to bare carbon paper (Figure S22), we further extracted the gas‐phase mass transport resistance attributable to the catalysts. The mass transfer resistance of H‐Ni‐NC‐250 was 35% lower than that of H‐Ni─NC‐200, and H‐Ni‐NC‐300 showed a further 15% reduction relative to H‐Ni‐NC‐250, demonstrating a clear correlation with increased through‐porosity. Accordingly, we propose a gas transport mechanism within porous hollow spheres (Figure 4c). In the assembled 3D hierarchical structure, more through‐pores in the shell provide expanded pathways for gas transport. This reduces resistance, accelerates mass transfer, and ultimately increases the local CO₂ concentration at the active sites.

FIGURE 4.

Through‐pore engineering enhanced gas‐diffusion. Electrochemical CO₂RR performance of H‐Ni‐NC‐200, H‐Ni‐NC‐250, and H‐Ni‐NC‐300 catalysts. (a) Measurement equipment for CO₂ gas permeations. MFC: mass flow controller; GC: gas chromatograph; f: volumetric flow rate; p: pressure. (b) CO₂ permeance of the H‐Ni‐NC‐200, H‐Ni‐NC‐250, and H‐Ni‐NC‐300 gas diffusion electrodes (GDEs). (c) Schematic illustration of the gas transport mechanism in porous hollow spheres.

Finite element simulations were conducted to clarify the role of through‐pore porosity in governing gas transport and its influence on the reaction rate. To capture the key structural characteristics of the H‐Ni‐NC‐n catalysts, a simplified spherical‐shell model with a diameter of 200 nm and a shell thickness of 5 nm was adopted, consistent with the electron microscopy observations (SEM, TEM and HAADF‐STEM). Through‐pores with a size of 1.2 nm were introduced into the shell according to the pore‐size distribution analysis (Figure S15). For simplification, the CO₂RR was mainly simulated on the inner shell surface, motivated by the experimentally observed higher local Ni density toward the inner shell region (Figure 1e), and were designed to evaluate the diffusion behavior of CO₂ gas within the 1.2 nm micropores. Four models with different porosities were constructed and labeled as Φ = 0.60, 0.30, 0.15, and 0.075. The reaction was approximated as a first‐order surface reaction with respect to CO₂ concentration, and the apparent first‐order surface reaction‐rate constant (k) was varied as 1, 2, 5, 10, and 20 × 10⁵ s⁻¹ (Figure 5a). With increasing surface reaction‐rate constant, the CO₂ concentration inside the sphere decreased. At a given surface reaction‐rate constant, higher shell porosity resulted in a higher internal CO₂ concentration. Such enhanced local CO₂ availability is expected to facilitate CO₂RR and suppress the competing HER [47], especially under high‐current‐density conditions. This qualitative trend is consistent with the experimentally observed CO₂RR behavior in both H‐cell and flow‐cell configurations. The CO Faradaic efficiency increased with shell porosity in the H‐cell measurements (Figure S16e), while the higher‐porosity H‐Ni‐NC‐250 and H‐Ni‐NC‐300 catalysts maintained >95% CO Faradaic efficiency at current densities above 500 mA cm⁻² in the flow cell. These observations, together with the qualitative simulation results, suggest that porosity‐dependent mass transport may contribute to maintaining high CO selectivity under high‐current‐density operation.

FIGURE 5.

Simulating the mechanism of through‐pores in enhancing gas diffusion and reaction within spherical shells. (a) CO₂ concentration distribution within a porous shell as functions of porosity and surface reaction‐rate constant. (b) Internal CO₂ concentration. (c) CO₂ reaction rate under varying porosity and surface reaction‐rate constant.

As shown in Figure 5b the internal CO₂ concentration (c/c ₀, normalized to the inlet concentration) within the spherical shell as a function of porosity and surface reaction‐rate constant. The internal CO₂ concentration increases notably with increasing through‐pore porosity, suggesting improved local CO₂ accessibility within the hollow structure. However, this enhancement becomes progressively less pronounced at higher porosity. For example, at a surface reaction‐rate constant of 5 × 10⁵ s⁻¹, increasing the porosity from = 0.075 to Φ = 0.60 (an eightfold increase) raises the c/c₀ value by only 4.9 times. This indicates that, within this simplified model, the enhancement in CO₂ concentration does not scale linearly with porosity, suggesting a saturating effect as through‐pores become more abundant.

In the model, the reduction in available active surface area with increasing shell porosity was incorporated through the remaining solid‐shell fraction (1‐Φ). Moreover, as illustrated in Figure 5c, increasing through‐pore porosity simultaneously reduces the available active surface area, which partly offsets the beneficial effect of enhanced CO₂ accessibility. These simulations provide a qualitative framework for understanding how enhanced gas transport and reduced available surface area jointly influence catalyst performance. Additionally, the simulations suggest that efficient gas transport becomes increasingly important at higher reaction‐rate conditions, where local CO₂ depletion becomes more significant (Figure 5c). These results suggest that gas transport and active‐site density could both be important factors governing the performance of porous hollow CO₂RR catalysts.

3. Conclusion

This study demonstrates a rational cooperative‐templating strategy for engineering through‐pores in hollow‐structured nitrogen‐doped carbon supports hosting atomically dispersed Ni–N₄ sites. By utilizing zinc species as a sacrificial atomic template, we achieve modulation of shell through‐porosity while largely preserving the macro‐scale hollow‐sphere morphology and Ni–N₄ coordination structure. This approach investigates critical design parameters, revealing a trade‐off: increasing porosity enhances gaseous reactant and product transport but inevitably reduces the volumetric density of active sites due to framework loss. Among the catalysts investigated, H‐Ni‐NC‐250 catalyst achieves an exceptional current density of −1.0 A cm⁻² with >95% CO Faradaic efficiency and demonstrates robust stability. Finite‐element simulations provide a qualitative framework for exploring the possible roles of gas transport and available active sites in the performance of porous hollow catalysts. Beyond presenting a high‐performance catalyst, this work establishes pore architecture engineering at the sub‐nano scale as a vital, generalizable design principle for next‐generation electrocatalysts and other heterogeneous systems where efficiency is governed by mass transport. Given the extensive use of porous electrodes, such as metallic foams, nanoporous metals, carbon nanotube‐based catalysts, and porous solid oxide electrolysis cell electrodes, the insights here offer broad implications for porous electrode design.

4. Experimental Section

4.1. Material

Tetraethyl orthosilicate (TEOS), ammonia, ethanol, and 2‐propanol were purchased from Aladdin Reagent Co., Ltd. Formamide was purchased from Damao Chemical Reagent Factory. NiCl₂, deuterium oxide (D₂O), and dimethyl sulfoxide (DMSO) were purchased from Alfa Aesar Co., Ltd. ZnCl₂ and Nafion 117 solution (∼5 wt.% in a mixture of lower aliphatic alcohols and water) were purchased from Sigma‐Aldrich Co., Ltd. KHCO₃ was purchased from Shanghai Macklin Biochemical Technology Co., Ltd. KOH was purchased from Fuchen Chemical Reagent Co., Ltd. Carbon dioxide, nitrogen, and argon were purchased from Bejing Huatong Jingke Chemical Co., Ltd. Deionized water with a resistivity ≥ 18 MΩ was used to prepare all aqueous solutions. All of the reagents were of analytical grade and were used directly without further purification.

4.2. Synthesis of SiO₂

Silica (SiO₂) nanoparticles were synthesized using a modified Stöber method [48]. First, 150 mL of ethanol, 50 mL of water, and 7 mL of ammonia were mixed in a beaker and stirred for 20 min. Then, 6 mL of TEOS was rapidly added, and the mixture was stirred for 10 h. Next, the mixture was centrifuged to collect a white precipitate, purified by deionized water three times and once with ethanol. Finally, the precipitate was dried in a freeze‐dryer for 12 h, yielding a hard template of SiO₂ nanoparticles with an average size of 200 nm.

4.3. Synthesis of H‐Ni‐NC With Different Through‐Pore Porosity

H‐Ni─NC was synthesized via FA condensation and carbonization, with modifications from our previous research [25]. First, 5 mM NiCl₂, 250 mM ZnCl₂, and 1 g SiO₂ hard template (200 nm average size) were added into 30 mL FA, sonicated for 30 min and then stirred for another 30 min. Then, the mixture was transferred into a Teflon‐lined stainless autoclave and heated at 180°C for 12 h. The obtained brown powder was collected by centrifugation, washed 3 times with deionized water, and dried in a freeze‐dryer for 24 h. Next, the brown powder underwent pyrolysis heated at a rate of 5°C min⁻¹ to 900°C, and maintained for 2 h under an argon flow of 100 mL min⁻¹. Finally, the product underwent alkaline (6 M KOH water solution) leaching in 60°C for 12 h to remove SiO₂ hard template to get the H‐Ni‐NC‐250 catalyst.

By maintaining all other experimental conditions constant and adjusting the concentration of ZnCl₂, H‐Ni‐NC catalysts with different through‐pore porosity were synthesized, denoted as H‐Ni‐NC‐n (where n = 200, 250, and 300, representing the concentration of ZnCl₂ in the FA solution in mM).

4.4. Material Characterization

SEM images were obtained using a Zeiss SUPRA 55 at an accelerating voltage of 10 kV. TEM was carried out on a FEI Tecnai G2 20 STwin (operated at 100 kV). HR‐TEM was performed on a JEOL 2100 operated at 200 kV. Aberration‐corrected HAADF‐STEM images and element mapping images were recorded on a JEOL JEM‐ARM200F TEM/STEM with a spherical aberration corrector, operated at 200 kV. The XAS of the Ni K‐edge was recorded at the 1W1B end station of the Beijing Synchrotron Radiation Facility (BSRF, operated at 2.5 GeV with a maximum current of 250 mA), and the energy was tuned by a Si (111) monochromator. The sample powder was uniformly coated on adhesive tape (Scotch Magic Tape, 1 × 0.5 cm²) for the XAS recording. XRD patterns were performed on a MeasSrv F9XDZ42 with Cu Kα (λ = 0.154 nm) radiation at a scan rate of 10°C/min. Raman spectra were recorded on a LabRAM Aramis (HORIBA Jobin Yvon S.A.S, Laser 532). XPS spectra were recorded on a Thermo Electron ESCALAB 250 (X‐ray Source: Al Kα line), where the binding energy was calibrated by C 1s (284.83 eV). ICP‐OES was carried out using a Thermo Fisher iCAP 7400. N₂ adsorption–desorption isotherms were measured on a Micromeritics ASAP 2460.

4.5. Electrochemical Measurements

The electrochemical measurements were conducted using a Corrtest CS2350H workstation under ambient temperature and pressure. For H‐cell tests, two compartments were separated by a Nafion 117 proton exchange membrane. A carbon rod served as the counter electrode, and an Ag/AgCl electrode (saturated KCl) was used as the reference electrode. The working electrode was prepared by dropping 100 µL of catalyst ink onto a 1 × 1 cm² YLS 060 (Toray) carbon fiber paper with a mass loading of 1 mg cm⁻². The catalyst ink was prepared by dispersing 5 mg of catalyst powder into 990 µL of ethanol and 10 µL of 5 wt% Nafion solution, followed by ultrasonication for 30 min. A CO₂‐saturated 0.5 M KHCO₃ solution was used as the electrolyte, and the cathode compartment was supplied with 20 sccm CO₂. The EIS tests were performed by applying at −0.7 V versus RHE with 5 mV amplitude in the frequency range of 100 kHz ∼ 100 mHz.

For the flow‐cell, the cathode and anode were separated by an anion exchange membrane. The GDE was prepared by brush‐painting the catalyst ink onto a 2 × 2 cm² YLS 060 (Toray) carbon fiber paper until a mass loading of 0.5 mg cm⁻² catalyst was achieved. The catalyst ink was prepared by dispersing 10 mg of catalyst powder into a mixture of 4 mL of 2‐propanol and 20 µL of 5 wt% Nafion solution, followed by ultrasonication for 30 min. GDE, nickel foam, and Ag/AgCl electrode (saturated KCl) served as working electrode, counter electrode and reference electrode, respectively. The geometric surface area of the catalyst was 1 cm². A 1.0 M KHCO₃ solution was used as the electrolyte. A peristaltic pump was used to control the flow rate at 20 sccm for both the cathode and anode sides. Additionally, 20 sccm of CO₂ was introduced into the gas flow channel of the cathode. All potentials during the H‐cell and flow‐cell tests were calculated with respect to the reversible hydrogen electrode (RHE) scale using the following equation:

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

For flow‐cell measurements, the reported potentials were further corrected for iR drop using E (iR‐corrected V vs. RHE) = E (V vs. RHE)—R _sol × I. The nominal TOF for CO production based on the total Ni content determined by ICP‐OES was calculated using the following equation:

$$TOF\left(h^{-1}\right)=\frac{j_{CO}/nF}{m_{cat}\times w/M_{Ni}}$$

where j _CO is the CO partial current density, n is the number of electrons transferred for CO formation, F is the Faradaic constant, m _cat is the catalyst loading on the electrode (0.5 mg cm⁻²), w is the Ni weight fraction determined by ICP‐OES (Table S2), M _Ni is the atomic mass of Ni (58.69 g mol⁻¹). The TOF values were calculated by normalizing the CO production rate to the total Ni content measured by ICP‐OES, assuming all Ni species are electrochemically accessible.

For the MEA‐cell, Sustainion X37 (dioxide materials) served as the anion exchange membrane. The GDE was prepared using the same method as the flow‐cell. In this setup, the GDE functioned as the cathode, while Ti‐melt coated with IrO₂ was utilized as the anode. The cathode was fed with 20 sccm CO₂, and the anode was circulated with a 0.1 M KHCO₃ solution at a flow rate of 20 sccm. Note that the testing of the MEA‐cell did not undergo iR correction.

During electrocatalysis tests, the gas products were analyzed online by gas chromatography (GC, SHIMADZU 2014C) and the liquid products were detected by a 600 MHz ¹H ¹D liquid NMR spectrometer (Bruker). The Faradaic efficiency of products was calculated by the following equation:

$$F{E}_i=\frac{n_i\times N\times \mathrm{F}}{C}\times 100\%$$

where n _i is the number of moles of target product, N is the number of electrons transferred, F is the Faradaic constant, and C is the total charge passed during electrolysis.

4.6. Nitrite Stripping Measurements

The relative electrochemically accessible Ni‐site densities of the H‐Ni─NC catalysts were estimated using the in‐situ nitrite stripping technique [49]. Catalyst inks were prepared by dispersing 5 mg of catalyst in 500 µL of isopropanol (IPA). The catalyst ink was then drop‐cast onto a glassy carbon electrode (geometric area: 0.196 cm²) with a catalyst loading of 0.5 mg cm⁻². A 0.5 M acetate buffer solution (pH = 5.2) was used as the electrolyte for the nitrite stripping measurements. The detailed test protocol is described below: (1) CV measurements were performed in N₂‐saturated electrolytes between 1.0 and 0.3 V versus RHE to remove residual O₂ from the electrolyte, then the CV curves (baseline) were tested between 0.4 and –0.3 V versus RHE at a scan rate of 10 mV/s in N₂‐saturated electrolyte. (2) For nitrite adsorption, the working electrode was immersed in a 125 mM NaNO₂ solution at 300 rpm for 5 min under open circuit potential (OCP). Subsequently, the working electrode was washed in DI water under OCP at 300 rpm for 1 min. Then, the working electrode was washed in acetate buffer solution under OCP at 300 rpm for 5 min. Finally, the working electrode was washed in DI water under OCP at 300 rpm for 1 min. (3) CV measurements were performed in N₂‐saturated electrolytes between 1.0 and 0.3 V versus RHE to remove residual O₂ from the electrolyte. Subsequently, stripping CV curves were recorded in N₂ saturated electrolyte at a scan rate of 10 mV/s between 0.4 and −0.3 V versus RHE.

The MSD was calculated as follows:

$$MSD\left[molsite{g}^{-1}\right]=\frac{Q_{strip}\left[C{g}^{-1}\right]}{n_{strip}F\left[Cmo{l}^{-1}\right]}$$

where Q _strip was obtained by integrating the differential stripping curves, n _strip is the number of electrons in the reduction of an adsorbed nitrite anion, we used a n _strip value of 5 based the reported by Kucernak et al. [50], and F is Faradaic constant.

The TOF normalized by electrochemically accessible Ni sites estimated from nitrite stripping measurements was calculated as follows:

$$TOF\left[s^{-1}\right]=\frac{I_{CO}\left[A{g}^{-1}\right]}{nF\left[Asmo{l}^{-1}\right]\times MSD\left[mol{g}^{-1}\right]}$$

where I _CO is CO partial current normalized by catalyst mass, n is the number of electrons in the reduction of CO₂ to CO, which is 2, and F is Faradaic constant.

4.7. Gas Permeance Measurements

The prepared GDE sample was fixed in a home‐made permeation apparatus (Figure S12) sealed with rubber O‐rings. The CO₂ feed flow rate was kept constant at 30 and 30 sccm N₂ was used as sweep gas. The pressures at both sides were still constant at 1 bar during the measurement process. A calibrated gas chromatograph was used to detect the component concentration on the permeate side after the measurement system ran stable.

The permeance of CO₂ (P_CO2) was calculated as follows:

$$P_{\mathrm{CO}2}=\frac{N_{\mathrm{CO}2}}{A\mathrm{\Delta }{p}_{\mathrm{CO}2}}=\frac{F_{\mathrm{CO}2}}{\mathrm{\Delta }{p}_{\mathrm{CO}2}}$$

where N_CO2 is the permeation rate of CO₂ (mol s⁻¹), Δp_CO2 is the partial pressure difference of CO₂ (Pa), and A is the membrane area (m²). F_CO2 is denotes the molar flux of CO₂ (mol m⁻² s⁻¹). The unit GPU is used for the gas permeance, where 1 GPU = 3.3928 × 10⁻¹⁰ mol m⁻² s⁻¹ Pa⁻¹. Every permeance was calculated by the average of five data points.

4.8. Finite‐Element Method Simulations

We constructed a finite‐element method simulations using COMSOL Multiphysics software. The model includes a hollow sphere with a diameter of 200 nm, a shell thickness of 5 nm, and micropores with a diameter of 1.5 nm distributed on the shell surface. To investigate the effect of porosity on flow characteristics, we set four different porosity values: 0.075, 0.14, 0.3, and 0.6. During the simulation process, the “Free and Porous Media Flow, Brinkman” interface in the computational fluid dynamics (CFD) module was employed to simulate the coupled free‐flow and porous‐medium transport behavior at the microscale. The velocity field in the fluid medium was solved using the laminar Navier–Stokes equations. Inside the nano‐sphere, the fluid flows in the form of a porous medium, and we employed the Brinkman equation to describe this phenomenon. The mass transfer process was simulated through the diffusion equation, strictly adhering to Fick's law of diffusion. Furthermore, we defined a reaction on the surface of the shell where CO₂ is reduced to CO, with the surface reaction‐rate constant (k) varying as 1, 2, 5, 10, and 20 × 10⁵ s⁻¹. In the calculation of concentrations, we used the “transport of diluted species” model, taking into account the effects of diffusion and reaction, thereby obtaining the concentration distribution of CO₂ and CO in the geometric space under different porosity and reaction rate conditions.

Author Contributions

Jiage Yu: Conceptualization, methodology, investigation, validation, visualization, writing – original draft, writing – review and editing, and data curation. Xiaoming Sun: Conceptualization, methodology, data curation, supervision, project administration, writing – review and editing, and visualization. Jiazhan Li: Writing – review and editing, funding acquisition, project administration, supervision, conceptualization, methodology, data curation, and visualization. Wei Liu: Conceptualization, methodology, writing – review, and editing. Yuan Wang: Data curation. Qianxi Yang: Data curation. Benqiang Tian: Data curation. Lu Bai: Data curation. Boyuan Li: Data curation. Haoyang Wu: Data curation. Yun Kuang: data curation. Qingyi Zhu: data curation. Linlin Zhou: data curation. Xiangrong Jin: Data curation.

Funding

This work was supported by the National Natural Science Foundation of China (NSFC 22309008) and the Fundamental Research Funds for the Central Universities (buctrc202401).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adma74095‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.