The energy crisis and environmental problems have increased the urgency for the development of renewable energy sources. Hydrogen is considered one of the most promising candidates because of its high energy density and non-polluting characteristics [1]. Among the many different hydrogen production technologies, water splitting is an appealing technique for energy conversion and storage, functioning by converting solar or wind-derived electricity to hydrogen fuel [2-4]. At present, challenges remain in commercialized water electrolysis systems. The practical applications of water-splitting are very limited because splitting reactions, including the anodic oxygen evolution reaction (OER) and cathodic hydrogen evolution reaction (HER), are all strongly uphill reactions with large overpotentials (actual cell voltage is 1.8–2.4 V, whereas the thermodynamic value is 1.23 V) [5]. Therefore, it is urgent and meaningful to explore efficient electrocatalysts for the water splitting system. Noble metal (Pt or Ir)-based catalysts have long been the benchmarking materials for HER and OER, but they are costly and scarce. Therefore, the design and construction of efficient bifunctional catalysts via the exposure of more HER and OER catalytically active sites simultaneously in the same electrolyte (alkaline, acid, or neutral) are urgently needed and remain a challenge. Currently, low-cost earth-abundant transition metal compounds have been developed as possible alternatives, especially for use in alkaline electrolytes [6, 7].

To date, some non-noble metal-based bifunctional catalysts have been thoroughly explored in alkaline electrolytes [8, 9], including transition-metal based sulfides (e.g., NiCo₂S₄ and Ni₃S₂)[10, 11], oxides (e.g., NiCo₂O₄)[12], layered double hydroxides (LDH) (e.g., NiFe LDHs) [13-16], carbides [17], nitrides [18], and phosphides [19-21]. It should also be mentioned that the catalytic performances of these non-noble-metal-based nanomaterials have not been maximized, and, overall, the water splitting reaction still suffers from a high overpotential and sluggish reaction rate. Among the earth-abundant transition metal compounds, the transition metal phosphides have been intensively studied and are promising candidates for overall water splitting reaction [8, 22, 23]. Previous studies have indicated that optimizing the nanostructure and composition of the catalysts is key to maximizing the catalytic efficiency [24]. In the past decades, materials based on a metal–organic framework (MOF) have been shown to provide a suitable platform for designing porous nanostructure catalysts for water splitting because of their high metal content, ultra-high porosity, high specific surface areas, tunable pore structures, and easily functionalized surfaces [25, 26]. Inspired by these diverse configurations, MOF materials have been widely used in many fields, including gas storage/separation and chemical sensors [27-29]. More importantly, they serve as appealing sacrificial precursors or templates to prepare functional materials [30-32]. However, the design of transition metal phosphides derived from multi-component MOF structures as water splitting catalysts is still rarely reported.

Herein, we report a stepwise strategy to design and fabricate novel hierarchically porous cobalt phosphide nanocage@ferric-zinc mixed-metal phosphide nanotubes (CoP@ZnFeP) hollow nanocomposites derived from MOF precursors as overall water splitting electrocatalysts. The hierarchical structures are composed of CoP nanocages and ZnFeP nanotubes. The as-prepared flower-like CoP@ZnFeP hybrids show remarkable electrocatalytic activity for the HER and OER in 1.0 mol/L KOH electrolyte, having onset overpotentials of 50 and 148 mV and Tafel slopes of 76 and 53.9 mV/decade, respectively. The CoP@ZnFeP flower-like hybrids display enhanced catalytic activity compared with the mechanical mixture of CoP and ZnFeP particles. Importantly, the catalytic performance is comparable to that of commercial noble-metal catalysts. The CoP@ZnFeP hierarchical nanocomposites also show excellent activity toward overall water splitting, yielding a 10 mA/cm² water-splitting current on the application of just 1.6 V (24-h continuous galvanostatic electrolysis) in the two-electrode system. The excellent performance is mainly due to the nano-building-blocks (nanocages, nanoparticles, and nanotubes), which have a large surface area and well-defined porous channels that are advantageous for mass and charge transport in the electrolyte. This design methodology for mixed-metal phosphides derived from MOFs may create the opportunities to search for highly efficient and robust non-precious metal catalysts for applications in high-performance energy conversion and storage devices.

The Co₃O₄ nanocages were prepared according to a previously reported method [33]. In a typical synthesis, two solutions were first prepared. Solution A contained 1 mmol Co(NO₃)₂·6H₂O dissolved in 10 mL distilled water with stirring, and solution B contained K₃[Co(CN)₆]₂ (0.04 mmol) and polyvinylpyrrolidone (PVP) surfactant (0.3 g) dissolved in 10 mL distilled water. The transparent red solution A was slowly added to solution B and aged for 24 h at room temperature. The resulting precipitate was washed several times with distilled water and ethanol, respectively, and then dried overnight at 60 ℃ under vacuum. The resulting powder was finally decomposed at 400 ℃ for 1 h in a N₂ atmosphere in an oven. All chemicals were of analytical purity and used without further purification.

The hierarchical Co₃O₄@Fe-MOF-5 nanocomposites were prepared according to our previously reported method [34]. In a typical procedure, Co₃O₄ nanocages (25 mg), Fe(acac)₃ (75 mg), Zn(NO₃)₂·6H₂O (58 mg), and benzene-1, 4-dicarboxylic acid (H₂BDC, 2 mg) were dissolved in dimethylformamide (DMF)/ethanol solution (2 mL, v/v = 5/3). This solution was treated using ultrasonic waves for 30 min. The mixed solution was transferred to a 50-mL Teflon-lined autoclave and heated at 100 ℃ for 6 h. After cooling to room temperature, the hierarchical Co₃O₄@Fe-MOF-5 nanocomposites were obtained after centrifugation and washing several times with distilled water and ethanol, successively, followed by drying overnight (12 h) at 50 ℃ in a vacuum. For comparison purposes, Fe-MOF-5 was also prepared in the absence of Co₃O₄ nanocages in the DMF/ethanol solution.

The hierarchical CoP@FeZnP nanocomposites were prepared by phosphorization. In a typical procedure, a Co₃O₄@Fe-MOF-5 sample was placed in the center of a tube furnace, and a combustion boat containing NaH₂PO₂ (0.2 g) was placed in the upstream side of the quartz tube. Before heating, the tube was flushed with Ar. The center of the furnace was slowly elevated to 400 ℃ (heating speed = 1 ℃/min) under the protection of an Ar atmosphere. The synthetic process was carried out for 30 min. Finally, the tube was allowed to cool to room temperature. The obtained samples are denoted CoP@FeZnP. Note that NaH₂PO₂ can decompose, readily generating poisonous PH₃ in the process.For comparison, the Co₃O₄ nanocages and Fe-MOF-5 were also treated with the same phosphorization procedure. That is, the cobalt phosphide (denoted CoP) and Zn-Fe metal phosphide (denoted ZnFeP) were also synthesized.

X-ray diffraction (XRD) measurements were carried out on a Shimadzu XRD-6000 X-ray diffractometer using Cu K_α radiation. X-ray photoelectron spectra (XPS) were obtained on an Escalab 250Xi (Thermo Fisher Scientific) X-ray photoelectron spectrometer. The calibration of the binding energy was carried out by setting the binding energy of the C 1s peak to 284.4 eV. The transmission electron microscopy (TEM) images were obtained using a field-emission TEM (2100F, Japan) coupled with an energy-dispersive X-ray spectrometer (EDX) at an accelerating voltage of 200 kV. To prepare the specimens for TEM measurements, the as-grown substrates were immersed in 2 mL of ethanol and sonicated for 10 s.

Catalyst suspensions in 1.0 mL of an ethanol/water (v/v =1:5) solution were prepared. In addition, 5 µL Nafion solution (0.5 wt%) was also added as a binder. For the electrode preparation, 10 µL of the catalyst suspension (10 mg/mL) was dropped onto the surface of a pre-polished glassy carbon electrode (GCE). The electrocatalytic activities of the samples for the OER and HER were studied in 1 mol/L KOH solution and recorded on an Autolab potentiostat (Metrohm Autolab, Netherlands) at ambient temperature. All tests were performed in a three-electrode electrochemical cell with a saturated Hg/HgO reference electrode and a graphite rod counter electrode. The prepared GCE electrodes were used as the working electrodes to investigate the electrocatalytic activities toward the HER and OER. Electrochemical impedance spectroscopy (EIS) was performed from 200 kHz to 50 mHz and analyzed by fitting the experimental results to equivalent circuits using the program NOVA 2.1. The double-layer capacitance (C_dl) of the hollow structures was obtained by applying cycle voltammograms over a range of scan rates (20–180 mV/s). The measurement system was calibrated against the reversible hydrogen electrode (RHE). The corresponding polarization curves were calibrated after iR correction.

The design strategy for the preparation of the CoP@ZnFeP hierarchical hybrid is illustrated in Fig. 1. In a nutshell, three steps involving crystal growth, self-aggregation, and phosphorization were carried out. The TEM images show that the Co₃O₄ crystals have a well-defined hollow cubic structure with a uniform size distribution (ca. 120 nm), as shown in Fig. 1b. The high-magnification TEM image (Fig. 1c) indicates that the nanocages have a porous surface and are composed of numerous nanoparticles. In the second step, the self-aggregation process was performed in a DMF/ethanol solution, and Co₃O₄@Fe-MOF-5 hierarchical structures were obtained. The TEM shows that the Fe-MOF-5 nanotubes grow out from Co₃O₄ nanocage to form flower-like shapes. The Co₃O₄ nanocage is located at the center, and several Fe-MOF-5 nanotubes protrude like petals. As shown in Fig. 1d and 1e, the Fe-MOF-5 nanotubes had a well-defined morphology with tube length of 400 nm, diameter of 100 nm, and shell thickness of 5–10 nm (Fig. S1, see the Supporting Information).

Furthermore, we noticed that the Fe-MOF-5 hollow nanotubes were only generated in the presence of the Co₃O₄ nanocages. In a control experiment performed under identical synthetic conditions but with no Co₃O₄ nanocages added, only irregular aggregates of Fe-MOF-5 particles formed (Fig. S2). Actually, the Co₃O₄ nanocage served as the surface for nucleation and could be a useful substrate for controlling the growth of the Fe-MOF-5 nanotubes. It was found that the morphology of the hybrids depends strongly on the amount of Co₃O₄. A detailed description of the morphological evolution of Co₃O₄@Fe-MOF-5 has been reported by us previously [34].

Since most MOFs are nonconductive, MOFs have been rarely applied in the field of electrocatalysis [35, 36]. In particular, to enhance the electrical conductivity of the Co₃O₄@Fe-MOF-5 hierarchical structure, the Co₃O₄@Fe-MOF-5 hybrids were generated by phosphorization via calcination. Typical TEM images show that the flower-like hollow structure was well maintained after calcination, and the nanotubes became highly porous, consisting of numerous nanocrystals (Fig. 2a and 2b). In particular, numerous nanocrystals filled the nanotubes, forming plentiful surface-active sites and increasing the surface area for the facile exchange of proton or oxygen-containing intermediates [37, 38]. The high-resolution (HR)-TEM images (Fig. 2c and 2d) reveal that the phosphide nanoparticles have diameters of ca. 10 nm. Additionally, elemental mapping was carried out to determine the distribution of Zn, Fe, and P in the nanotubes, and these measurements confirm the homogenous distribution (Fig. 2e). The corresponding EDX spectrum shows that Zn, Fe, and P are present with an atomic ratio close to 1:2.5:4 (Fig. S3). We also measured the whole elemental composition of the CoP@ZnFeP hybrids, as shown in Fig. S4. In general, Co, Zn, Fe, and P were detected, having an approximate molar ratio of 1:1.1:2.2:5.7. The chemical composition of the composites was further explored by XRD. A typical XRD pattern is shown in Fig. S5 and contains the main reflections from the CoP@ZnFeP hierarchical hybrid. The peaks at 26°, 26.7°, and 31.4° mainly originate from the ZnP₂ phase and are matched well to the (1 0 4), (1 1 2), and (1 1 4) planes, respectively (JCPDF 24-1464) [39]. The peaks at 40.2° and 44.1° mainly originate from Fe₂P and matched well with the (1 1 1) and (2 0 1) planes, respectively (JCPDF 51-0943) [40, 41]. All the other peaks are characteristic of Co₂P (JCPDF 32-0306) and also matched well with the (1 2 1) and (2 0 1) planes [42, 43]. The results are consistent with the above-mentioned TEM analysis. The hierarchical CoP@ZnFeP nanocomposites were composed of numerous mixed nanocrystals.

Further insights into the valence and the surface chemical composition of CoP@ZnFeP were obtained through XPS measurements (Fig. 3). As shown in Fig. 3a, the Co 2p core level spectrum showed two major peaks, assigned to Co 2p_3/2 and Co 2p_1/2, respectively[44-46]. The peak can be fitted with four peaks. The dominant peaks around 782.1 and 797.8 eV were attributed to Co in Co₂P. Fig. 3b shows the Zn 2p core level XPS spectra [47]. The binding energy of the Zn 2p_3/2 peak was found at 1022.5 eV, and the other sharp peak at 1045.1 eV corresponds to Zn 2p_1/2. The XPS spectrum of Fe 2p core level spectrum can be deconvoluted into three main peaks (Fig. 3c). The peaks with binding energies of 711.2 and 714.1 eV are assigned to Fe₂P and, possibly, an Fe-based oxide. Another peak located at 724.5 eV contributed to the related satellite peak [48, 49]. The P 2p spectrum (Fig. 3d) contains two peaks at 129.1 and 130.2 eV, which are assigned to P 2p_3/2 and P 2p_1/2, respectively. The two binding energies are attributed to the phosphorus anions of the metal phosphides. The broader peak at 135 eV can be attributed to P–O species and is close to the values reported previously. The presence of O may arise from the superficial oxidation of the CoP@ZnFeP particles as a result of air contact [50].

The electrocatalytic HER activity of the CoP@ZnFeP hybrids was investigated in an alkaline electrolyte (1.0 mol/L KOH). The catalytic performance tests were conducted in a standard three-electrode system with a scan rate of 1.0 mV/s at room temperature. Fig. 4a shows the linear sweep voltammetry (LSV) curves of the CoP@ZnFeP hybrids, a mixture of CoP and ZnFeP (denoted CoP/ZnFeP), and commercial Pt/C (20%). It is evident that the 20% Pt/C catalyst shows the best HER activity. The CoP@ZnFeP hybrids exhibited a high current density under a low applied overpotential, where the onset potential was close to –50 mV (vs. RHE). The CoP@ZnFeP hierarchical structure reached a current density of –10 mA/cm² at a potential of –148 mV (vs. RHE). It is evident that the mixture of Co₂P and ZnFeP (CoP/ZnFeP) showed inferior HER catalytic performance. The onset potential was close to –100 mV (vs. RHE) and the mixture required a larger potential of –230 mV (vs. RHE) to deliver a current density of –10 mA/cm². The significant intrinsic activities of different samples determined the difference in the HER activities, which were revealed by comparing the Tafel slopes. As anticipated, the Tafel value of the CoP@ZnFeP hierarchical hybrid was 76 mV/decade, which is lower than that of CoP/ZnFeP mixture (ca. 136 mV/decade, Fig. 4b).

The electrocatalytic activities of the CoP@ZnFeP hierarchical hybrids for OER were also evaluated with a three-electrode electrochemical cell in 1.0 mol/L KOH solution. The polarization cures are shown in Fig. 4c. It was found that the CoP@ZnFeP hierarchical structure produced a small onset potential of 1.45 V (vs. RHE) and a high anode current density of 10 mA/cm² at a potential of 1.50 V (vs. RHE) and was superior to that of IrO₂. In contrast, the mixture of CoP and ZnFeP (CoP/ZnFeP) exhibited a low current density under the same applied overpotential, where the onset potential was close to 160 mV (vs. RHE). Meanwhile, the Tafel slope of the CoP@ZnFeP hierarchical structure catalyst was only 53.9 mV/decade, which is lower than that of IrO₂ (60 mV/decade) and CoP/ZnFeP (65.8 mV/decade), as shown in Fig. 4d. The results further indicate that the hierarchical structure showed superior OER catalytic performance compared to their bulk counterparts. Indeed, the HER and OER catalytic activity of the CoP@ZnFeP hierarchical structure is superior to most of the bifunctional electrocatalysts that have been reported in alkaline electrolytes (Table S1). The CoP@ZnFeP hierarchical structure was composed of different nano-building-blocks (nanoparticles, nanocages, and nanotubes), which yielded a large surface area and well-defined porous channels that are advantageous for mass and charge transport in an electrolyte. Previous research has indicated that only surface atoms or a very thin layer of the active electrode material plays a dominate role in the electrochemical reaction. Therefore, the hierarchical hollow structure results in more accessible active surface area for electrolyte permeation, which is beneficial for enhancing the catalytic activity [51]. The superior performance may also be attributed to the synergism between the different crystalline phases (ZnP₂, Co₂P, and Fe₂P). The results are also consistent with the scientific consensus that metal phosphides with complex metal compositions may improve the electrocatalytic activity [50, 52].

We further investigated the electrode kinetics using double-layer capacitance (C_dl) and electrochemical impedance spectroscopy (EIS) measurements. As shown in Fig. S6A and S6B, the C_dl values of the CoP@ZnFeP hierarchical structure were all higher than those of the mixture of CoP and ZnFeP (CoP/ZnFeP). Meanwhile, the EIS measurements showed the consistent order in charge-transfer resistance. As shown in Fig. S7, the EIS spectra show that the CoP@ZnFeP hybrid has a lower charge-transfer resistance, which results in faster OER and HER kinetics compared to that of the CoP/ZnFeP catalyst. Specifically, the C_dl values and equivalent circuit diagrams are different when the same catalyst was used for HER and OER. The difference may be ascribed to two factors. First, the reaction mechanism of the OER is more complicated than that of the HER because of the greater number of reaction steps. Secondly, the different active sites for the OER and HER may also be responsible for these differences, as reported previously.

Next, we utilized the CoP@ZnFeP hierarchical structure as a bifunctional catalyst. For the measurement of the catalytic activity for the overall water splitting reaction, a two-electrode configuration was used, and the CoP@ZnFeP hierarchical material was used as the anode and the cathode. The overall water-splitting characteristics in a two-electrode configuration during 24-h continuous galvanostatic electrolysis are shown in Fig. S8. Remarkably, the CoP@ZnFeP hierarchical hybrids showed excellent activity, yielding a 10 mA/cm² water-splitting current on the application of a potential of just 1.6 V across the two electrodes. However, the potential changed by almost 70 mV over 24 h. In addition, the formation of bubbles is an issue that cannot be ignored. During continuous galvanostatic electrolysis, hydrogen and oxygen bubbles were vigorously generated on the electrode surface. Most of the bubbles did not leave the electrode surface immediately, and these adherent bubbles reduced the effective active area, thus leading to an increase in the applied potential. Worse still, the catalysts had an extremely low adhesive force to the surface, leaving the surface with ease[53]. Finally but importantly, the TEM images show that the flower-like hierarchical structure was basically maintained after long-term electrochemical testing (24 h), as shown in Fig. S9.

In summary, we report a stepwise strategy to design and fabricate novel hierarchically porous CoP@ZnFeP hollow nanocomposites. The as-prepared flower-like CoP@ZnFeP hybrids revealed remarkable electrocatalytic activity compared with the mechanical mixture of CoP and ZnFeP. The excellent performance is mainly due to the nano-building-blocks, which showed a large surface area and well-defined porous channels, which are advantageous for mass and charge transport in the electrolyte. Our study shows the potential of these novel hierarchical structures to function as highly active, bifunctional electrocatalysts for the overall water splitting reaction. This design methodology may create opportunities to develop highly efficient and robust non-precious metal catalysts for applications in high performance energy conversion and storage devices, such as batteries, supercapacitors, and electrocatalysis.