Increasing Electrocatalytic Oxygen Evolution Efficiency through Cobalt-Induced Intrastructural Enhancement and Electronic Structure Modulation.

Electrolytic water splitting using surplus electricity represents one of the most cost-effective and promising strategies for hydrogen production. The high overpotential of the oxygen-evolution reaction (OER) caused by the multi-electron transfer process with a high chemical energy barrier, however, limits its competitiveness. Here, a highly active and stable OER electrocatalyst was designed through a cobalt-induced intrastructural enhancement strategy combined with suitable electronic structure modulation. A carved carbon nanobox was embedded with tri-metal phosphide from a uniform Ni-Co-Fe Prussian blue analogue (PBA) nanocube by sequential NH3  ⋅ H2 O etching and thermal phosphorization. The sample exhibited an OER activity in an alkaline medium, reaching a current density of 10 mA cm-2 at an overpotential of 182 mV and displayed a small Tafel slope of 47 mV dec-1 , superior to the most recently reported OER electrocatalysts. Moreover, it showed impressive electrocatalytic durability, increasing by approximately 2.7 % of operating voltage after 24 h of continuous testing. The excellent OER activity and stability are ascribed to a favorable transfer of mass and charge provided by the porous carbon shell, synergistic catalysis between the three-component metal phosphides originating from appropriate electronic structure modulation, more exposed catalytic sites on the hollow structure, and chainmail catalysis resulting from the carbon protective layer. It is foreseen that this successfully demonstrated design concept can be easily extended to other heterogeneous catalyst designs.

Introduction

Developing sustainable n class="Chemical">hydrogen enpan>ergy to replapan> class="Chemical">ce nonrenewable fossil fuels has been attracting intense worldwide research interest. Hydrogen has many unrivaled advantages such as high calorific value, recyclability, and environmental friendliness. It is thus viewed as a viable candidate to meet future global terawatt‐level energy needs.[ , , , ] Among the reported hydrogen production technologies, electrocatalytic water splitting (EWS) driven by surplus electricity is one of the most cost‐effective and promising strategies. The raw materials needed for EWS are not only abundantly available, but energy conversion from electrical to chemical energy can also effectively fill the large gaps in energy supply requirements.[ , , , ] However, as an indispensable half‐reaction in EWS systems, the oxygen evolution reaction (OER) at the anode involves a multi‐electron transfer process with a high electrochemical energy barrier; it thus requires an advanced electrocatalyst to accelerate the sluggish kinetics of the overall reaction.[ , , , ] Although noble‐metal Ir/Ru‐based materials are considered as benchmark OER catalysts, they are extremely scarce, and exorbitant market prices severely impede their commercial application. Therefore, substantial efforts have been made to explore highly active and inexpensive catalysts that can replace traditional noble metals.[ , , , , ]

Transition‐n class="Chemical">metal‐based catalysts (e. g., oxides, phosphides, chalcogenides, and carbonitrides) have attracted tremendous interest in recent years owing to their impressive OER activities, natural abundance, and the consequent low costs.[ , , ] Among these, transition‐metal phosphides (TMPs) are considered to be one of the most studied electrocatalytic materials at present.[ , , ] However, monometal TMPs usually have poor electroconductivities and inferior stabilities, thus significantly limiting their expected catalytic activities.[ , , ] For this reason, multicomponent TMPs have been developed and have proven to be feasible alternatives for enhancing catalytic performance, both in theory and practice.[ , , ] In Ni−Fe−P, for example, density functional theory (DFT) calculations have revealed that the electronic structure of the Ni−P catalyst can be readily tuned with the help of extra Fe 3d states at the Fermi level, greatly reducing the charge transfer resistance and thus promoting its OER activity. In addition, EWS is usually performed in a strong alkaline/acidic solution, in which the employed electrolytes inevitably corrode or poison the catalytic sites, causing irreversible changes in the composition and/or structure of the catalyst surface.[ , , ] As a result, catalytic activity may drastically decrease as the electrocatalytic reaction proceeds. Therefore, effective protection of the catalyst is a major issue for its long‐term, repeated use.

In the last den class="Chemical">cade, pan> class="Chemical">Prussian blue analogue (PBA)‐derived thermal phosphorization strategies have been repeatedly demonstrated as feasible and excellent “best of both worlds” solutions for the preparation of mixed‐metal phosphide chainmail catalysts.[ , , ] On one hand, it is easy to introduce multiple metal species into the crystal structure of PBA materials by controllable displacement or adsorption; these metal species serve as the reaction precursors for the desired multicomponent TMPs. On the other hand, cyanide‐derived carbon layers formed in situ on the surfaces of the phosphides can inherit most of the rich porous structures of the PBA host materials and can exhibit excellent conductivities resulting from N‐doping. These characteristics allow for satisfactory chainmail protection of the internal active sites, favor the charge transfer, and permit precise adsorption of target substances involved in the catalytic process. However, if the PBA precursors are not mechanically strong, they can break into small particles during the high‐temperature phosphorization process, causing the subsequent secondary growth of small particles as well as agglomeration. Therefore, the loss of the impressive ordered and porous nanostructures during this thermal treatment process is clearly disadvantageous for efficient electrochemical reactions.

To address this, we previously designed a n class="Chemical">SiO2‐protected thermal phosphorization approach for the synthesis of an N‐doped carbon nanobox embedded with phosphide nanoparticles. Owing to the external structural enhancement provided by the SiO2 protective layer, the morphology of the catalyst was well maintained; however, subsequent etching was necessary to remove the SiO2, complicating the synthetic procedure. Alternatively, a polymer layer can be used to replace the original SiO2 coating. Although removal of the protective layer is no longer needed, the resultant active sites are buried deep in the carbon matrix, making it difficult for immediate or adequate contact with the electrolyte, resulting in a loss of their intrinsic catalytic activities. It is clear that the above external structural enhancement strategy is not only detrimental to the maximum exposure of the catalytic active sites, but also adds additional inconvenience to material preparation. To resolve these two issues, ingenious catalyst designs are necessary.

Previous researn class="Chemical">ch has demonpan>strated that Ni−Co PBA possesses shorter M−N and C−N bond lengths in comparison to Ni−Fe PBA, exhibiting higher bond energies and mechanical strength, which are particularly important for morphology retention during the thermal phosphorization process.[ , ] As such, it is reasonable to believe that the introduction of cobalt species into the crystal structure of Ni−Fe PBA can serve as an intrastructural enhancer, significantly improving the mechanical strength and allowing the preservation of the original morphology. In addition, the cobalt‐induced intrastructural enhancement can effectively compensate for the damaging effect of subsequent ammonia etching of the PBA structure, thus endowing the product with richer pore structures and more active sites. Moreover, the electronic interactions between three metal phosphide components can help to obtain the desired synergistic catalysis, thereby improving the electrocatalytic performances.

Owing to these n class="Chemical">considerationpan>s, we propose a pan> class="Chemical">cobalt‐induced intrastructural enhancement strategy, combined with proper electronic structure modulation, for enhancing the OER efficiency in the resulting carved carbon nanoboxes (CCN) embedded with tri‐metal TMP nanoparticles. Typically, uniform Ni−Co−Fe PBA nanocubes were first synthesized through a facile chemical precipitation reaction between a nickel source, the ligand potassium ferricyanide, and the intrastructural enhancer potassium hexacyanocobaltate. Second, NH3 ⋅ H2O was employed to etch the solid Ni−Co−Fe PBA nanocubes, which produced hollow nanocubes along with the isochronous generation of apical nanopores. Finally, a high‐temperature phosphorization treatment converted the hollow Ni−Co−Fe PBA nanocube to the desired catalyst −CCN embedded with Ni−Co−Fe−P in the cyanide‐derived carbon layer (termed as Ni−Co−Fe−P@CC−E‐XX). Here, CC is cyanide‐derived carbon layer, E denotes NH3 ⋅ H2O etching, and XX denotes the volume of NH3 ⋅ H2O used. Scheme 1 displays the design concept and synthesis route of Ni−Co−Fe−P@CC‐E‐15. The sample delivered outstanding OER activity in an alkaline medium, affording a current density of 10 mA cm−2 at an ultralow overpotential of 182 mV and a small Tafel slope of 47 mV dec−1, which is superior to the most recently reported state‐of‐the‐art OER electrocatalysts (Tables S1 and S2). This cobalt‐induced intrastructural enhancement with simultaneous electronic structure modulation proved to be a successful technique and can therefore be readily extended to other heterogeneous catalyst designs.

Scheme 1

The design concept and synthesis route of Ni−Co−Fe−P@CC‐E‐15.

The design n class="Chemical">conpan> class="Chemical">cept and synthesis route of Ni−Co−Fe−P@CC‐E‐15.

Results and Discussion

DFT n class="Chemical">calpan> class="Chemical">culations were carried out to prove the rationality of the proposed cobalt‐induced intrastructural enhancement. As shown in Figure 1a,c the optimized structures are ordered and crystalline, indicating that they are stable after relaxation. It is well known that the chemical properties of a material are particularly dependent on its stability. Hence, we calculated the cohesive energy (E coh) between the bulk PBA structure and the defect structure (Fe and Co) of the metal dimers. The defect model is formed by removing one metal atom from the bulk structure, and the defect formation energy is defined as the amount of energy required to initiate a reaction. The defect formation energy of metal atoms in their bulk crystals can be calculated using the following formula: ΔE b=E Defect+E Metal–E Bulk. Here, E Defect is the total energy of the PBA structure containing a defect, E Bulk is the total energy of the bulk PBA structure, and E Metal is the energy of one metal atom from pure bulk metal. A low defect formation energy indicates increased stability of metal dimers embedded in bulk PBA. Interestingly, the defect formation energy is −3.00 eV for Ni−Co PBA; this is favorable compared to Ni−Fe PBA (1.18 eV). As shown in Figure 1b,d for both the side and top view in the VESTA (Visualization for Electronic and STructural Analysis) code, the charge density difference map can be calculated as Δρ=ρ bulk−ρ defect−ρ metal, which exhibits the presences of bulk structure atoms, metal vacancies, and pure metal atoms. The oscillation between charge accumulation and depletion, represented by the cyan‐ and blue‐colored areas, denote positive and negative charge density differences, respectively. It is clear that an electron transfers from the nickel atoms to the neighboring iron or cobalt atoms. This suggests that the nickel atoms are more positively charged (due to charge loss) in the Ni−Co PBA system, indicating the formation of stronger, polarized Ni−Co covalent bonds in Ni−Co PBA compared to the Ni−Fe PBA system.

Figure 1

(a,b) Geometric structure in a 4×4×1 supercell and difference of charge densities of Ni−Fe PBA structure, respectively; (c,d) Geometric structure in a 4×4×1 supercell and difference of charge densities of Ni−Co PBA structure, respectively (an isosurface level is set to 0.001 e Å−3).

(a,b) Geometrin class="Chemical">c stpan> class="Chemical">ructure in a 4×4×1 supercell and difference of charge densities of Ni−Fe PBA structure, respectively; (c,d) Geometric structure in a 4×4×1 supercell and difference of charge densities of Ni−Co PBA structure, respectively (an isosurface level is set to 0.001 e Å−3).

To understand elen class="Chemical">ctronpan>ipan> class="Chemical">c properties, such as the total density of states (TDOS), of the Ni−Fe PBA and Ni−Co PBA surfaces, orbital hybridization was performed by decomposing the electron local function (ELF) distribution. As shown in Figure 2, the Ni−Fe PBA surfaces exhibit joint contributions from Fe and Ni in the valence band hybridized between −0.7 and 0 eV, that is, below the Fermi level, implying a strong d‐band transition metal contribution (Fe 3d and Ni 3d) to orbital hybridization. In contrast, there is a different feature in the Ni−Co PBA system, as shown in Figure 2b. The main contributions to the DOS are the Ni 3d bands below the Fermi level, located between −0.5 and 0 eV, and the Co 3d bands located between −2 and −1 eV. This indicates relatively weaker hybridization in the Ni−Co PBA system and stronger hybridization between the Fe 3d and Ni 3d orbitals near the edge of the valence band. This greater Ni−Co overlap population at a low energy level, in comparison to Ni−Fe, demonstrates a relatively stronger stabilization of the Ni−Co PBA system compared to the Ni−Fe PBA system. Full details of each projected density of states (PDOS) for the 3d bands of transition metals are shown in Figure 3. Figure 3a displays the PDOS for the transition metal centers of the Ni−Fe PBA and Ni−Co PBA systems. In these cases, the three t2g orbitals include d, d, and d 2− 2 character, and the eg orbitals present d 2 and d character. The d 2− 2 orbitals are the most stable and fully occupied. The d orbital is doubly occupied, and the d orbital couples with this orbital, destabilizing itself. Hence, the singly occupied d orbital on one Fe 3d is an unoccupied orbital, and on the other Ni 3d is an occupied orbital, which is expected to lead to weak orbital coupling. The orbital splitting shows a <0.5 eV shift of the energy states from −0.5 to 0 eV, corresponding to the energy of the t2g orbital (d 2−d 2). The orbital coupling has little effect on the other parts of the spectrum. Furthermore, the Fe 3d and Ni 3d orbitals of the Ni−Fe PBA surface did not exhibit a splitting of the bands at the Fermi level. The computed PDOS shows that the peak located around −0.5 to 0 eV in the Ni−Fe PBA is noticeably shifted to a higher energy level and the Fermi level moves across the Ni 3d orbitals; these orbitals are all occupied below the Fermi level. After introducing cobalt ions into the Ni−Fe PBA system, differential scanning calorimetry (DSC) analyses (shown in Figure S1) demonstrate that the resulting Ni−Co−Fe PBA exhibits better thermal stability with an exothermic peak at higher temperatures, implying that cobalt enhances the structure.

Figure 2

PDOS of (a) Ni−Fe PBA and (b) Ni−Co PBA system (the Fermi level is set at 0 eV).

Figure 3

Projected 3d orbital PDOS of (a,c) Ni−Fe PBA and (b,d) Ni−Co PBA system (the Fermi level is set at 0 eV).

n class="Chemical">PDOS of (a) Ni−Fe PBA and (b) Ni−Co PBA system (the Fermi level is set at 0 eV).

Projen class="Chemical">cted 3d orbital pan> class="Chemical">PDOS of (a,c) Ni−Fe PBA and (b,d) Ni−Co PBA system (the Fermi level is set at 0 eV).

The n class="Chemical">compositionpan> anpan>d pan> class="Chemical">crystal structure of the as‐prepared precursor were investigated by X‐ray diffraction (XRD). As revealed in Figure S2a, the well‐defined diffraction signals of the Ni−Co−Fe PBA nanocrystals can be attributed to the joint contributions of the cubic Ni−Fe PBA (ICDD card no. 01‐082‐2283) and the Ni−Co PBA (ICDD card no. 01–089‐3738), with exceptionally high crystallinity and purity. The corresponding energy‐dispersive X‐ray spectroscopy (EDS) shown in Figure S2b reveals the presence of all the component elements in Ni−Co−Fe PBA products, that is, Ni, Co, Fe, C, and N. The calculated atomic ratio of Ni/Co/Fe was 5.92 : 1 : 2.89, close to the theoretical value of 6 : 1 : 3, indicating that cobalt species have been successfully introduced into the main structure of Ni−Fe PBA as an intrastructural enhancer, rather than by simple physical adsorption. The Au and Pd peaks are from the sputtering process used during sample preparation for scanning electron microscopy (SEM), and the K signal originates from the potassium ions adsorbed by the PBA structure. SEM and transmission electron microscopy (TEM) images indicate that the resultant Ni−Co−Fe PBA precursor has a cubic structure with a smooth surface, along with a mean length of 120 nm (Figure S2c–f). Figures S3 and S4 depict the TEM/EDS line scans and elemental mapping of the component elements. These reveal that the cobalt is evenly distributed throughout the nanocubes, thereby contributing to its intrastructural enhancement effect and maintains the original cubic morphology of the reaction precursor. The Ni−Co−Fe PBA was then converted into a carved hollow nanocube using a controllable and preferential alkaline etching treatment, creating eight nanopores at the vertices of each nanocube, as seen in Figure 4a–d. These nanopores can survive the thermal phosphorization process and provide a favorable mass and charge transfer channel for efficient electrocatalytic reactions. The thickness of the shell is approximately 13 nm, as shown in Figure 4e. An ultrathin shell is highly desirable as it forms a porous structure through the loss of various volatile components during the thermal fabrication procedure. This is also advantageous for improving its OER activity. The resultant Ni−Co−Fe PBA carved nanoboxes barely change in composition or structure after NH3 ⋅ H2O etching, as revealed in the corresponding XRD pattern and EDS images (Figure 4f–h).

Figure 4

(a–c) SEM and (d,e) TEM images, (f) XRD pattern, (g) TEM/EDS line scans as well as (h) elemental mapping images of Ni−Co−Fe PBA carved nanoboxes.

(a–n class="Chemical">c) SEM anpan>d (d,e) TEM images, (f) XRD patternpan>, (g) TEM/EDS line span> class="Chemical">cans as well as (h) elemental mapping images of Ni−Co−Fe PBA carved nanoboxes.

The subsequent phosphorization pron class="Chemical">cess pan> class="Chemical">converts Ni−Co−Fe PBA carved nanobox into the expected product, Ni−Co−Fe−P@CC‐E‐15. The XRD pattern depicted in Figure 5a reveals that the obtained catalyst consists of three kinds of metal phosphides: Ni5P4 (ICDD card no. 00–018‐0883), CoP2 (ICDD card no. 01‐077‐0263), and FeP (ICDD card no. 00‐003‐1066). These are the result of the phosphorization of the metal centers in the tri‐metal PBA structure. It is expected that uniform mixing of the three metal elements in the precursor will cause close contact between the metal phosphide components, inducing strong synergistic catalytic effects and leading to enhanced OER performance. Raman spectroscopy was performed to determine the degree of graphitization of the carbon matrix by calculating the relative I D/I G intensity ratio. As evident in Figure S5, the D band signal at 1344 cm−1 may result from the defect state of the carbon atom lattice, while the G band at 1558 cm−1 shows the degree of graphitization. The I D/I G ratio is 0.98, indicating a high degree of graphitization and small structural defects in the cyanide‐derived carbon, which can improve the electronic transport and reaction kinetics involved in the OER process. The morphology of the obtained Ni−Co−Fe−P@CC‐E‐15 nanobox was further characterized by SEM, TEM, and EDS mapping. Unsurprisingly, the morphology of the nanobox did not change significantly, even after undergoing the harsh high‐temperature phosphorization process. In particular, the nanopore structures still remain on the surfaces, as revealed in Figure 5b–d. These existing nanopores can maximize the readily accessible catalytic sites for electrochemical water splitting and provide a rapid transfer channel to access the internal active sites. Figure 5e–g shows a high‐resolution (HR)TEM image of the sample, which contains three sets of interlayer distances: 0.238, 0.259, and 0.357 nm, assigned to the FeP, Ni5P4, and CoP2 components, respectively. Moreover, the EDS mapping images of the Ni−Co−Fe−P@CC‐E‐15 nanobox, depicted in Figure 5h, demonstrates that six elements, including Ni, Co, Fe, P, C, and N, outline the geometric profile of the four nanoboxes, further confirming the successful preparation of the Ni−Co−Fe−P@CC‐E‐15 nanobox.

Figure 5

(a) XRD pattern, (b,c) SEM, (d) TEM, (e–g) HRTEM, (h) high‐angle annular dark‐field STEM images and elemental mapping images of Ni−Co−Fe−P@CC‐E‐15.

(a) XRD pattern, (b,n class="Chemical">c) SEM, (d) TEM, (e–g) HRTEM, (h) high‐anpan>gle anpan>nular dark‐field STEM images anpan>d elemenpan>tal mapping images of n class="Chemical">Ni−Co−Fe−P@CC‐E‐15.

In addition, high‐resolution X‐ray photoelen class="Chemical">ctronpan> spectroscopy (HRXPS) was carried out to investigate the surface elemental composition and valence state of the resultant Ni−Co−Fe−P@CC‐E‐15. The HRXPS spectrum of the Ni 2p region (Figure 6a) can be deconvoluted into three sub‐bands with binding energies of 853.8, 855.9, and 861.3 eV; these can be assigned to nickel phosphide, NiO species, and a satellite signal, respectively.[ , ] The Co 2p spectrum shown in Figure 6b exhibits three distinct signals at 778.5, 781.6, and 785.1 eV, ascribed to Coδ+ species in the Co−P component, Co−O, and satellite peaks, respectively.[ , ] As seen in Figure 6c, the Fe 2p spectrum can be divided into three component signals at 707.2, 710.6, and 713.7 eV, corresponding to the Fe−P, Fe−O, and a satellite peak, respectively.[ , ] The P 2p spectrum (Figure 6d) exhibits two strong peaks at 129.6 and 134.3 eV, the former corresponding to metal phosphide and the latter assigned to the P−O bond.[ , ] The observed M/P−O signals may be as a result of surface oxidation of the product after prolonged exposure to air, consistent with previous reports. In Figure 6e, the peaks of the C1s spectrum at 284.6, 285.3, and 286.5 eV are consistent with the expected C−C, C−N, and C=O, respectively. Three component signals of pyridinic‐N (398.5 eV), pyrrolic‐N (400.1 eV), and quaternary‐N (401.3 eV) as seen in Figure 6f, confirm N‐doping in the obtained carbon matrix; this is conducive to enhancing the conductibility of the sample, which in turn promotes the OER activity.

Figure 6

HRXPS spectra of component elements of Ni−Co−Fe−P@CC‐E‐15: (a) nickel, (b) cobalt, (c) iron, (d) phosphorus, (e) carbon, and (f) nitrogen.

HRXPS spen class="Chemical">ctra of pan> class="Chemical">component elements of Ni−Co−Fe−P@CC‐E‐15: (a) nickel, (b) cobalt, (c) iron, (d) phosphorus, (e) carbon, and (f) nitrogen.

The OER an class="Chemical">ctivities of the catalyst products were studied in a 1.0 m KOH solution with a typical three‐electrode system. To determine the cause of the enhanced OER efficiency in Ni−Co−Fe−P@CC‐E‐15, five reference catalysts were designed and developed. In order to assess the influence of the volume of NH3 ⋅ H2O on the vertex pores and OER efficiency, Ni−Co−Fe PBA was etched with 18 and 12 mL of NH3 ⋅ H2O and then phosphorized to create Ni−Co−Fe−P@CC‐E‐18 and Ni−Co−Fe−P@CC‐E‐12 (Figures S6 and S7). To determine the function of the nanopore at the vertex, Ni−Co−Fe PBA solid nanocubes were phosphorized to fabricate Ni−Co−Fe−P@CC nanocubes, without the NH3 ⋅ H2O etching step used to create the desired vertex pores (Figure S8). This sample is identical in composition to Ni−Co−Fe−P@CC‐E‐15, the only difference being the absence of nanopores at its vertices. Lastly, as for the cobalt‐induced intrastructural enhancement and desired electronic structure modulation effects, Ni−Co and Ni−Fe bimetallic PBAs were synthesized and then etched with the same volume of NH3 ⋅ H2O to produce Ni−Co and Ni−Fe PBAs carved nanoboxes. These serve as the precursors for the formations of the Ni‐Co−P@CC‐E‐15 (Figures S9 and S10) and Ni−Fe−P@CC‐E‐15 (Figures S11–S13). The characterization results for these five reference samples are depicted in detail in the Supporting Information.

The OER an class="Chemical">ctivities of the as‐prepared Ni−Co−Fe−P@CC‐E‐15 and the five reference products were tested via linear scanning voltammetry (LSV). The graphite electrode used is relatively inert to the OER process and is therefore particularly suitable for supporting catalysts without interfering with an accurate assessment of their performances (not shown here). As evident in Figure 7a, the corresponding LSV plot reveals that the resultant Ni−Co−Fe−P@CC‐E‐15 delivers the best OER activity with ultralow overpotentials of 182, 238, and 250 mV to give current densities of 10, 100, and 250 mA cm−2, respectively, which are superior to those of commercial RuO2 catalysts (Figure 7a, 182 vs. 302 mV) and previously reported TMP catalysts (Tables S1 and S2). Movie 1, shown in the Supporting Information, confirms the intensive generation of bubbles with a continuous increase in the working voltage for a typical LSV run using Ni−Co−Fe−P@CC‐E‐15. The next best OER catalyst is Ni−Co−Fe−P@CC‐E‐18, yielding η 10=222 mV (η10 denotes the overpotential at current density of 10 mA/cm2), which is inferior to that of Ni−Co−Fe−P@CC‐E‐15 (182 mV) and greater than that of the third best Ni−Co−Fe−P@CC‐E‐12 (245 mV) catalyst. From the perspective of chemical reactions, the ammonia etching step can be understood as the controlled and preferential dissolution of the PBA component located at specific positions in a particular nanocube. This allows the creation of a carved nanobox with eight nanopores at each vertex. Moreover, an increase in ammonia consumption not only helps to expand the size of the pores, but also causes more metal components to be dissolved. More importantly, excessive alkaline etching will seriously lower the mechanical strength of the PBA; this is disadvantageous for maintaining its morphology in the subsequent processing steps. Therefore, it can be concluded that the lower OER activity of the Ni−Co−Fe−P@CC‐E‐18, compared to the Ni−Co−Fe−P@CC‐E‐15, may be attributable to agglomeration and insufficient active sites resulting from the excessive loss of metal species (Figure S6). The inferior catalytic performance of the Ni−Co−Fe−P@CC‐E‐12 sample can be ascribed to the blocked mass and charge transfer process caused by the unfavorable pore size obtained by insufficient ammonia etching (Figure S7). Therefore, it is no surprise that the Ni−Co−Fe−P@CC exhibits the fourth‐rank catalytic activity with η 10=302 mV, which may be ascribed to a lack of nanopores necessary for accelerating access to the interior active sites as well as the escape of product bubbles (Figure S8).

Figure 7

(a) Polarization curves, (b) NH3‐TPD profiles, (c) valence band spectra, (d) Tafel slopes, (e) Nyquist plots of catalysts toward OER, and (f) normalized chronopotentiometric v–t curves toward OER with Ni−Co−Fe−P@CC‐E‐15 as electrocatalyst. Colour codes: Ni−Co−Fe−P@CC‐E‐15 (red), Ni−Co−Fe−P@CC‐E‐18 (green), Ni−Co−Fe−P@CC‐E‐12 (blue), Ni−Co−Fe−P@CC (olive), Ni‐Co−P@CC‐E‐15 (wine), Ni−Fe−P@CC‐E‐15 (black), and RuO2 (orange).

(a) Polarization curves, (b) pan> class="Chemical">NH3‐TPD profiles, (c) valence band spectra, (d) Tafel slopes, (e) Nyquist plots of catalysts toward OER, and (f) normalized chronopotentiometric v–t curves toward OER with Ni−Co−Fe−P@CC‐E‐15 as electrocatalyst. Colour codes: Ni−Co−Fe−P@CC‐E‐15 (red), Ni−Co−Fe−P@CC‐E‐18 (green), Ni−Co−Fe−P@CC‐E‐12 (blue), Ni−Co−Fe−P@CC (olive), Ni‐Co−P@CC‐E‐15 (wine), Ni−Fe−P@CC‐E‐15 (black), and RuO2 (orange).

n class="Chemical">Next, we clarify the influence of cobalt‐induced intrastructural enhancement and electronic structure modulation on the OER activity. The Ni−Fe−P@CC‐E‐15 sample exhibited the lowest OER performance with an overpotential of 344 mV at 10 mA cm−2, which can be attributed to the following: DFT calculations and DSC analyses demonstrated that the introduction of cobalt ions can lead to the formation of Ni−Co−Fe PBA, thereby enhancing stability and mechanical strength, which are particularly important for maintaining morphology during thermal phosphorization. It can therefore be concluded that the Ni−Fe PBA‐carved nanobox fails to maintain its original morphology due to the lack of cobalt‐induced intrastructural enhancement, resulting in cracking into small nanoparticles and causing severe nanoparticle agglomeration. This is confirmed in Figure S12c,d. Furthermore, in the absence of Co ions, the Ni−Fe PBA can produce two kinds of metal phosphides after phosphorization: Ni5P4 (ICDD card no. 00–018‐0883), and FeP (ICDD card no. 00–003‐1066), with no CoP2 component in the final product, as shown in Figure S13a. However, on closer examination, it can be seen that the XPS signals from Ni or Fe in Ni−Co−Fe−P are slightly shifted to a higher binding energy compared to Ni−Fe−P. Moreover, the P peak (in the M−P bond) shows an obvious red shift in the binding energy, suggesting that partial electron transfers from the metal cations to the P anions are promoted when cobalt ions are introduced to generate Ni−Co−Fe−P tri‐metal TMPs (M−P signals in Ni−Co−Fe−P: 853.8, 707.2, and 129.6 eV for Ni, Fe, and P, respectively; M−P signals in Ni−Fe−P: 853.6, 707.0, and 129.8 eV for Ni, Fe, and P, respectively; see Figure 6 and Figure S13b–d). Therefore, the electronic structure of the obtained Ni−Co−Fe−P catalyst can be well regulated through synergistic effects between the three components, giving rise to the formation of metal species with higher valence states and stronger oxidizing abilities, thus facilitating the generation of M−OH as well as the precise adsorption of OH−. OH−, a strong Lewis base, can be chemically adsorbed on Lewis acidic surfaces by donating the lone‐pair electrons of oxygen. Interestingly, NH3, with a lone pair of electrons from nitrogen, is also a strong Lewis base. Thus, NH3 temperature‐programmed desorption (NH3‐TPD) characterization was used to indirectly evaluate the affinity of OH− toward the catalysts. As depicted in Figure 7b, it can be seen that the resultant Ni−Co−Fe−P@CC‐E‐15 delivered higher total peak areas than Ni−Fe−P@CC‐E‐15, which results in an increase in the amount of OH− adsorbed, a favorable electrophilic surface, and better OER performance, which can be further confirmed by the O1s XPS spectra of these two samples (Figure S14). Moreover, ultraviolet photoelectron spectroscopy (UPS) was performed to provide a further in‐depth analysis on the modification of the electronic environment of the nanohybrids. As revealed in Figure 7c, Ni−Co−Fe−P@CC‐E‐15 displays a valence band maximum value of 1.02 eV, much closer to the Fermi level (E F, set to 0 eV). This is smaller than that of Ni−Fe−P@CC‐E‐15 (1.21 eV), implying that the surfaces of Ni−Co−Fe−P@CC‐E‐15 nanohybrids display a better metallic character with higher density of states around E F, which can contribute to faster charge transport and faster kinetics for the OER process.[ , , ] Lastly, to further prove the possible electronic structure regulation, the OER performance of Ni‐Co−P@CC‐E‐15 was also measured, which exhibited the penultimate catalytic activity (η 10=330 mV, Figure 7a). Therefore, it can be concluded that the desired synergistic effects resulting from three M−P components can effectively accelerate the charge‐transfer and tune the adsorption energies of reaction molecules via optimization of the electronic structure, thereby lowering the energy barrier of the electrocatalytic reaction and enhancing the electrocatalytic performance.

The intrinsin class="Chemical">c catalytic efficiencies of the active sites in the six samples can be further assessed by turnover frequency (TOF). The calculation procedures are presented in Figure S15 and Table S1. As expected, the as‐prepared Ni−Co−Fe−P@CC‐E‐15 displayed the highest TOF value (12.85 s−1), superior to those achieved by Ni−Co−Fe−P@CC‐E‐18 (5.57 s−1), Ni−Co−Fe−P@CC‐E‐12 (1.26 s−1), Ni−Co−Fe−P@CC (0.19 s−1), Ni‐Co−P@CC‐E‐15 (0.08 s−1), and Ni−Fe−P@CC‐E‐15 (0.05 s−1). This is in agreement with the trend observed in the variation of the overpotential. In addition, the electrochemical active surface area (ECSA) is also a vital index for evaluating the OER efficiency. Therefore, the double‐layer capacitance (C dl) of the catalyst, which increases linearly with ECSA, was measured by cyclic voltammetry in the non‐Faradaic region. Figure S16 shows that the Ni−Co−Fe−P@CC‐E‐15 catalyst exhibits the highest C dl (1.08 mF cm−2), better than that of the other five catalyst products: 0.84, 0.71, 0.50, 0.42, and 0.33 mF cm−2 for Ni−Co−Fe−P@CC‐E‐18, Ni−Co−Fe−P@CC‐E‐12, Ni−Co−Fe−P@CC, Ni‐Co−P@CC‐E‐15, and Ni−Fe−P@CC‐E‐15, respectively. To determine the intrinsic activity of the electrocatalysts, polarization curves for OER normalized by ECSA was presented in Figure S17, which revealed that Ni−Co−Fe−P@CC‐E‐15 exhibited the highest normalized current density. Therefore, it can be concluded that the best OER activity of Ni−Co−Fe−P@CC‐E‐15 not only results from the relatively highest ECSA and largest number of exposed sites (Table S1), but also from the best intrinsic activity. These lead to optimal OER activity. With regard to the kinetics of the OER process, Tafel plots and electrochemical impedance spectroscopy (EIS) can clarify the catalytic performance. Generally, the smaller the Tafel slope, the lower the required OER overpotential of the catalytic process under the same kinetic current density or apparent current density. In Figure 7d, it can be seen that the resultant Ni−Co−Fe−P@CC‐E‐15 presents an ultrasmall Tafel slope of 47 mV dec−1, lower than the other catalysts: 66 mV dec−1 for Ni−Co−Fe−P@CC‐E‐18, 75 mV dec−1 for Ni−Co−Fe−P@CC‐E‐12, 90 mV dec−1 for Ni−Co−Fe−P@CC, 100 mV dec−1 for Ni‐Co−P@CC‐E‐15, 137 mV dec−1 for Ni−Fe−P@CC‐E‐15, and 101 mV dec−1 for RuO2. This confirms that Ni−Co−Fe−P@CC‐E‐15 possesses more advantageous charge transfer kinetics for a better OER reaction. This is further proved by EIS. As shown in Figure 7e, it is clear that the resultant Ni−Co−Fe−P@CC‐E‐15 has the smallest semiarc among the six catalyst products and exhibits the lowest charge transfer resistance, which contribute to its superior OER activity.

The robustness and durability of the n class="Chemical">catalyst is vital for prospective commercialized applications. Thus, chronopotentiometry was performed to evaluate the OER stability of Ni−Co−Fe−P@CC‐E‐15 in an alkaline electrolyte at 10 mA cm−2. As evident in Figure 7f, the operating potentials necessary to maintain this current density increases by approximately 2.7 % after 24 h of continuous testing. The morphology was unchanged, confirming the inherent structural stability (Figure S18a–f). However, as the OER process is performed in a strong alkaline environment, the surface composition of the catalyst can significantly change. Compared with the spectra shown in Figure 6, the metal signals in the M−P bond are substantially weakened and the P peak in the M−P bond almost disappears after OER testing, as confirmed in Figure S18g–j. This implies the possible formation of a superficial oxide layer on the catalyst surfaces during the OER process. A previous report has demonstrated that the in situ formation of O‐rich phosphide and phosphate‐FePi/NiFeP heterostructure on the surface of bimetal phosphide may significantly promote the interfacial charge‐transfer, thus improving the electrocatalytic activity. Recently, Guo et al. found that the surface of the heterometallic phosphide could be oxidized to oxide/(oxy)hydroxide after the OER process; thus, the in situ phase transformation may enhance the OER performance. Therefore, based on the XPS analysis after OER stability testing mentioned above and previous reports, it can be concluded that the in situ production of an M−P/oxyhydroxide interface can effectively accelerate the charge transfer, thus promoting the OER efficiency. Furthermore, by comparing the ratio of actual oxygen production with the theoretical value, the faradaic efficiency is determined to be 98 %, thereby confirming the absence of side‐reactions during the electrochemical process, as revealed in Figure S19.

In summary, the prominent OER an class="Chemical">ctivity of Ni−Co−Fe−P@CC‐E‐15 can be ascribed to the following: (1) The holes created at the apex by alkaline etching and the nanopores in the shell layer caused by the thermal phosphorization can provide favorable mass and charge transfer channels for efficient electrocatalytic reactions.[ , ] (2) Advantageous electronic structure regulation through the synergistic effect between three components facilitate to the generation of M−OH and the precise adsorption of OH−.[ , ] (3) The cobalt‐induced intrastructural enhancement can allow the hollow nanostructure of the catalyst to be maintained, endowing it with large ECSA as well as more active sites. (4) The carbon protective layer not only prevents/delays the trimetal TMPs from being inactivated by the electrolyte, but also favors mass and charge transport. (5) The in situ creation of an M−P/oxyhydroxide interface can effectively accelerate charge transfer, promoting the chemical reaction efficiency.[ , , , ]

Conclusions

A n class="Chemical">combinpan>ed pan> class="Chemical">cobalt‐induced intrastructural enhancement and proper electronic structure modulation strategy was developed to fabricate Ni−Co−Fe−P@CC‐E‐15 nanoboxes as a highly active and stable oxygen evolution reaction (OER) electrocatalyst. The as‐prepared Ni−Co−Fe−P@CC‐E‐15 catalyst exhibited excellent OER activity with ultralow overpotentials of 182, 238, and 250 mV, affording current densities of 10, 100, and 250 mA cm−2, respectively, with a small Tafel slope of 47 mV dec−1. Moreover, the obtained catalyst also displays remarkable long‐term durability with only a 2.7 % loss in OER performance after 24 h of continuous testing. This work provides new insights into the design and synthesis of inexpensive, efficient, and stable multicomponent transition‐metal phosphide (TMP)‐based chainmail catalysts for future electrocatalysis applications.

Experimental Section

Materials

n class="Chemical">Nickel(II) nitrate hexahydrate, pan> class="Chemical">nickel(II) sulfate hexahydrate, potassium hexacyanocobaltate(III), potassium hexacyanoferrate(III), trisodium citrate dihydrate, sodium hypophosphite, ethanol, and ammonium hydroxide were purchased from Sigma‐Aldrich.

Materials preparation

Preparation of n class="Chemical">Ni−pan> class="Chemical">Co−Fe PBA nanocubes: Typically, 0.006 mol of Ni(NO3)2 ⋅ 6 H2O and 0.009 mol of Na3C6H5O7 ⋅ 2 H2O were added into 200 mL of deionized (DI) water to produce solution 1. Then, 0.001 mol of K3[Co(CN)6] and 0.003 mol of K3[Fe(CN)6] were introduced into 200 mL of DI water to produce solution 2. Solutions 1 and 2 were thoroughly mixed for precipitation, followed by 7 days aging. The precipitates were collected with a centrifuge, washed with DI water and absolute ethanol five times, and then dried in vacuum at 50 °C for 24 h to afford the product.

Preparation of Ni−pan> class="Chemical">Fe PBA nanocubes: Typically, 0.006 mol of NiSO4 ⋅ 6 H2O and 0.0075 mol of Na3C6H5O7 ⋅ 2 H2O were added into 590 mL of DI water to produce solution 3. Then, 0.004 mol of K3[Fe(CN)6] was introduced into 10 mL of DI water to produce solution 4. Solutions 3 and 4 were thoroughly mixed for precipitation, followed by 24 h aging. The precipitate was collected with a centrifuge, washed with DI water and absolute ethanol five times, and then dried in vacuum at 50 °C for 24 h to afford the Ni−Fe PBA nanocubes.

Preparation of n class="Chemical">Ni−pan> class="Chemical">Co PBA nanocubes: Typically, 0.006 mol of Ni(NO3)2 ⋅ 6 H2O and 0.009 mol of Na3C6H5O7 ⋅ 2 H2O were added into 200 mL of DI water to produce solution 5. Then, 0.004 mol of K3[Co(CN)6] were introduced into 200 mL of DI water to produce solution 6. Solutions 5 and 6 were thoroughly mixed for precipitation, followed by 7 days aging. The precipitates were collected with a centrifuge, washed with DI water and absolute ethanol five times, and then dried in vacuum at 50 °C for 24 h to afford the product.

Preparation of Ni−pan> class="Chemical">Co−Fe PBA carved nanoboxes: In a typical run, 100 mL of DI water and 15 mL of NH3 ⋅ H2O were added into 50 mL of ethanol, followed by addition of 100 mg of Ni−Co−Fe PBA nanocubes. The resulting suspension was stirred at room temperature for 20 min to afford Ni−Co−Fe PBA carved nanoboxs. The product was collected with a centrifuge, washed with deionized water and absolute ethanol five times, and then dried in vacuum at 70 °C overnight. For the preparation of other carved samples, the fabrication process was the same as above except the amount of to‐be‐carved samples was replaced, and different ammonium hydroxide dosages were used.

Preparation of Ni−pan> class="Chemical">Co−Fe−P@CC‐E‐15 nanoboxes: 15 mg of Ni−Co−Fe PBA carved nanoboxes and 0.15 g of NaH2PO2 were loaded separately into two porcelain boats, and the two boats were situated side by side at the center of a tube furnace (Anhui Kemi Machinery Technology Co., Ltd) with the NaH2PO2‐containing boat being placed upstream. The calcination was conducted at 400 °C for 2 h with a heating rate of 3 °C min−1 under N2 atmosphere, followed by cooling to ambient temperature under N2 gas flow. For preparation of other phosphorized samples, the process was the same as above except the to‐be‐phosphorized material was different.

Materials characterizations

The n class="Chemical">compositionpan> anpan>d pan> class="Chemical">crystalline phase of the sample were investigated with powder XRD (Shimadzu XRD‐6000, Japan), EDS (Hitachi S‐4800 and JEM‐2100, Japan), XPS and UPS (Thermo ESCALAB 250XI, America). The morphology and microstructure of the product were observed with SEM (Hitachi S‐4800, Japan) and HRTEM (JEM‐2100, Japan). Raman spectra were recorded at room temperature in the spectral range of 1000–2200 cm−1 using a Raman spectromicroscope (LabRAM HR800, Horiba Jobin Yvon, France). DSC was conducted on Pyris 1 TGA thermal analyzer (PerkinElmer) from 50 to 500 °C with a heating speed of 5 °C min−1 in N2 atmosphere. The NH3−TPD experiment was carried out on flow apparatus (Micrometrics TP‐5080).

Electrochemical characterizations

The OER an class="Chemical">ctivity of the product was characterized on a CHI760D electrochemical workstation with a typical three‐electrode setup. The working electrode was prepared using the as‐prepared powders (85 wt%) as the active material and polyvinylidene fluoride (15 wt%) as the binder. They were mixed in N‐methylpyrrolidone (NMP) to form a sample suspension. The working electrode was fabricated by drop‐casting the sample suspension, sonicated for 30 min before use, onto a graphite electrode (1 cm×1 cm) and dried at 80 °C in an oven. The mass loading of the active material on the working electrode was controlled to be around 0.5 mg cm−2. A platinum foil counter electrode and an Hg/HgO reference electrode were employed to complete the three‐electrode setup. For the measurements, 1 m KOH (pH=13.9) aqueous solution was used as the electrolyte. The potential values for the OER in this study were converted and referred to the reversible hydrogen electrode (RHE) using the Nernst equation: E RHE=E Hg/HgO+0.098+0.059 pH, where E Hg/HgO is the experimentally measured potential against the Hg/HgO reference electrode. All LSV polarization curves were iR‐corrected with respect to the involved solution resistances. The over‐potential (η) was calculated using the equation: η=E RHE−1.23. Prior to electrochemical measurements, the working electrode was conditioned by cycling through the potential window of 0 to 0.8 V vs. Hg/HgO thirty times at a scan rate of 100 mV s−1. The polarization curves were recorded with a linear potential sweep at a scan rate of 2 mV s−1. EIS was conducted in the frequency range of 105 to 0.01 Hz with an AC amplitude of 5 mV and the applied potential set at E Hg/HgO=0.6 V. The ECSA of the catalysts was characterized from the double‐layer charging curves obtained from cyclic voltammetry at increasing scan rates within a non‐Faradaic potential window (0.297–0.397 V vs. RHE), in which no Faradaic redox reactions occur. The ECSA can be determined based on the following equation: ECSA=C dl/C s, where C s is the specific capacitance of a flat smooth surface of the electrode material, about 40 μF cm−2 according to the literature. Long‐term stability test was carried out with chronopotentiometric measurements. For calculation of the TOF, reductive negative scan peak areas were firstly determined from cyclic voltammograms recorded at a specific scan rate, for example 300 mV s−1. Charge (Q) can be obtained with the formula: Q=peak area/300 mV s−1. Assuming a single‐electron transfer reaction in the reduction process, the number of surface active sites (n) can be calculated with the equation: n=Q/(1×1.602×10−19). Finally, TOF values were obtained from TOF=j×N A/(4×n×F) (j=current density, N A=Avogadro number, F=Faraday constant).

DFT calculations

All our DFT n class="Chemical">calpan> class="Chemical">culations of fully optimized geometries and electronic properties were performed within Vienna ab initio simulation program (VASP) code, and related DFT calculation details are documented in the Supporting Information.

Conflict of interest

The n class="Chemical">authors declare no conflict of interest.

As a servin class="Chemical">ce to our pan> class="Chemical">authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

n class="Chemical">Clipan> class="Chemical">ck here for additional data file.

Supplementary

n class="Chemical">Clipan> class="Chemical">ck here for additional data file.