Herein, we report a bimetallic cobalt‐based phosphide zeolitic imidazolate framework (BCP‐ZIF) with optimized composition as a bifunctional catalyst for overall water splitting. Importantly, the low H* adsorption energy and high conductivity of our synthesized Cu x Co 3– x P/nitrogen‐doped carbon (NC) catalyst, especially Cu 0.3 Co 2.7 P/NC, exhibit excellent catalytic performance in alkaline media (1.0 m KOH, pH = 13.5), featuring an ultralow overpotential of 0.19 V for OER and –0.22 V for HER at a current density of j = 10 mA cm −2 . To the best of our knowledge, the OER overpotential (at j = 10 mA cm −2 ) of the Cu 0.3 Co 2.7 P/NC catalyst is the lowest reported to date measured by the rotating disk electrode (RDE) system. Resulting from dopant‐induced CoP x phase behavior of the BCP‐ZIFs, we demonstrate that upon mild Cu doping of Cu x Co 3– x P/NC ( x = 0.3), an optimal electrocatalytic performance can be obtained with lowest charge‐transfer resistance and highest reaction kinetics. More impressively, our homemade electrolyzer with Cu 0.3 Co 2.7 P/NC as both cathode and anode catalysts has outperformed the Pt/C and RuO 2 paired counterparts by showing 60% higher current density at a voltage of 1.74 V. In addition, it demonstrates a long‐term durability with little deactivation after 50 h continuous operation in alkaline electrolyte.

Sustainable hydrogen fuel production driven by water electrolysis is one of the most practical solutions for energy storage and conversion. 5 To re‐distribute the intermittent power, excess electrical energy is converted and stored in the chemical form (H 2 ), which later can be used for power generation when needed. A typical water‐splitting system involves two half‐reactions: the oxygen evolution reaction (OER) at anode and hydrogen evolution reaction (HER) at cathode. 6 These two reactions combined determine the overall water‐splitting efficiency. Currently, there are two major water‐splitting media, acid and basis. Acidic electrolyte (low pH) is relatively well received due to the compatibility of mass producible membrane, while basic media (high pH) has the advantage of low cost and less precious metal demanding. Conventionally, commercially available Pt and RuO 2 are considered the state‐of‐the‐art HER and OER catalysts, which can efficiently increase the reaction rate and lower the overpotential. 7 The low natural abundance and prohibitive market price, however, are big hurdles for adopting Pt and RuO 2 on the industrial scale electrolysis system. 8 Exploring alternative nonprecious metal and stable water‐splitting catalysts with high efficiency becomes urgent and important. The promising alternatives for HER include transition metal‐based sulfides, carbides, and phosphides, such as MoS 2 , 9 WS 2 , 10 NiS, 11 MoC , 12 CoP, 13 and NiP. 14 As for OER catalysts, transition metal and their oxide/hydroxide species, including Co 3 O 4 /C, 15 CoO, 16 NiCo LDH (layered double hydroxide), 17 and MnO 2 , 18 etc., are usually preferred due to their high activities. Particularly, the formation of bimetallic HER and OER catalysts by introducing secondary metal ion is believed to be more active and stable due to tailored electronic and surface properties of the host structure. For example, Boettcher and co‐workers have shown a dramatic increase in OER activity upon addition of Fe cations in first‐row transition metals. 19 Zou and co‐workers revealed that by incorporating Zn ion, the Co‐based MOF (metal organic framework) catalysts demonstrate excellent HER performance comparable to commercial Pt/C. 20 However, transition metal phosphide based bimetallic catalysts are rarely studied and tested, despite their possible superiority in catalyzing OER and HER for overall water splitting.

During the past few decades, tremendous amount of renewable energy‐based infrastructures and facilities has been designed and constructed to work in concert with the calling of “green evolution.” Facing the fact that the nonrenewable energy has been increasingly unsustainable, the energy harvested in the “green” ways greatly diversifies the source of current power supply systems and makes us less vulnerable to the possible environmental crisis. Among different renewable energy sources, wind, solar, and hydropower are considered the most accessible and relatively well‐established ones. However, the unpredictable and intermittent nature of these green energy sources has introduced other concerns, which might cost even more energy waste than the traditional fossil fuel cycles. 1 For example, the electricity generated by hydropower is almost constant during day and night, but the power need varies during peak and off‐peak times. 2 These concerns become renown in certain regions when the renewable energy makes up of the great portion of their power supply. 3 Incorporating a buffer section between the intermittent energy source and the power grid is necessary to compensate the peak time by supplying stored energy, and collecting the excess power during off‐peak period. 4 The advancement in energy storage and conversion technologies is therefore of great importance to make the “green evolution” truly evolutional.

2 Results and Discussions

2.1 Structure of BCP‐ZIFs The synthesis route of BCP‐ZIFs is illustrated in Scheme 1; two metal precursors and MeIM (2‐methylimidazole) linkers are used to assemble ZIF self‐templates, which subsequently undergo carbonization and phosphidation to obtain the final catalysts.21, 22 Figure 1a shows a typical scanning electron microscopy (SEM) image of Cu 0.3 Co 2.7 /NC‐ZIFs with uniform polyhedrons (ca. 500 nm) and smooth surface. The morphology of as‐obtained ZIFs is consistent with other reported literatures.[13, 23] After carbonization and phosphidation at elevated temperature, the size and shape polyhedral structures are maintained with relatively rougher surface, where bimetallic Cu/Co nanoparticles are encapsulated inside the nitrogen‐doped carbon matrix (Figure 1b,c). The distribution of elements in BCP‐ZIFs is further investigated by elemental mapping. Figure 2d–h confirms the successful phosphidation and presence of bimetallic Cu/Co in Cu 0.3 Co 2.7 P/NC. The highly overlapped P and Co in SEM mapping also indicate the formation of CoP x species. Scheme 1 Open in figure viewerPowerPoint Schematic drawing of the synthesis route of bimetallic cobalt‐based phosphide Cu 3 Co 2.7 P/NC. Figure 1 Open in figure viewerPowerPoint a) SEM images of typical morphology of Cu 0.3 Co 2.7 /NC‐ZIFs with well‐defined polyhedral structure. b) After carbonization at 900 °C for 3 h for Cu 0.3 Co 2.7 P/NC. c) TEM image of Cu 0.3 Co 2.7 P/NC with bimetallic cobalt encapsulated in the nitrogen/carbon matrix. d–h) Elemental mapping of Cu 0.3 Co 2.7 P/NC confirming the presence of Co, Cu, P, and C. Figure 1 Open in figure viewerPowerPoint XRD patterns showing the progress of metallic cobalt and CoP x phases upon Cu doping. The crystallinity and phase information of the bimetallic cobalt‐based phosphide are confirmed by X‐ray diffraction (XRD) in Figure 1 and Figure S2 of the Supporting Information. All investigated samples display highly crystallized structure with distinct characteristic peaks. The metallic cobalt and cobalt phosphides can be indexed as Co (PDF#15‐0806), Co 2 Si‐type Co 2 P (PDF#32‐0306), and MnP‐type CoP(PDF#65‐2593).24 It is worth mentioning that the incorporation of Cu2+ neither introduces any observable Cu peaks nor associated peak shifts in the XRD patterns, suggesting the homogeneous substitution of Cu ion into CoP x matrix without forming CuCo alloys.25 The homogeneity of substituting Cu2+ ion might attribute to its similar ionic radius and electronegativity as Co2+ ion.26 Alternating CoP , Co 2 P, and Co phase fraction, however, is clearly observed with increasing Cu2+ doping level. By measuring the ratio of peak intensities, it is clear that the metallic Co phase fraction is larger with higher Cu2+ doping level, accompanied by a decreasing Co 2 P phase present. For our Cu x Co 3– x P/NC catalysts, the composition of Co, CoP, and Co 2 P mixed phases is believed to have a key effect in promoting electrochemical performance due to their different electroactivity and conductivity.24 Building on the XRD survey of different phase behaviors, we further investigate the electronic states associated with Cu doping using X‐ray photoelectron spectroscopy (XPS). Similar to other carbonized ZIF structures using MeIM as an organic linker, the nitrogen dopants in three major forms are clearly observed in N1s high‐resolution spectra (Figure 3a).27 In Figure 2b, the presence of deconvoluted peaks at 129.4 and 130.3 eV suggests the successful phosphidation of cobalt by forming CoP and Co 2 P, while a single peak at 133.9 eV represents some partially oxidized phosphate species.[13, 28] Besides confirming the presence of nitrogen and phosphide, the catalytically active Co element was further deconvoluted to gain more insight into the relationship between its complex oxidation states and transition metal (Cu) doping effect. It is noted that with increasing Cu content, from Cu 0.3 to Cu 1 , stronger Co02p 1/2 and Co02p 3/2 peaks of metallic Co are present (Figure 2c). This is identical to what was observed in XRD patterns, indicating that the presence of more concentrated Cu ions in the Co host structure might favor the formation of the cobalt metal. Oxidized Co derivatives (Co2+ and Co3+) show more complicated but reasonable behaviors compared to the XRD results. Upon mild Cu doping (Cu 0.3 ), the Co2p spectra show lower Co3+/Co2+ ratio for both Co2p 1/2 and Co2p 3/2 regions, corresponding to a more dominant Co 2 P than CoP phase. This is consistent with weaker CoP characteristic peaks in XRD of Cu 0.3 Co 2.7 P/NC. Interestingly, when introducing more Cu dopants, an increase of Co3+ is first observed for Cu 0.6 Co 2.4 P/NC, followed by a decreasing Co3+ content in Cu 1 Co 1 P/NC. Combining the XRD and XPS results, it is highly possible that the presence of Cu in Cu x Co 3– x P/NC can inhibit phosphidation of metallic cobalt, which leads to the phase transition of CoP x species, featured with higher Co2+ content upon mild Cu doping and the formation of Co metal with increasing Cu concentration. Another phenomenon associated with Cu doping is the peak shift for the oxidized Co species. All cobalt oxidation states, Co2+2p 1/2 , Co2+2p 3/2 , Co3+2p 3/2 , and Co3+2p 1/2 , exhibit a left shift of the bonding energy with increasing Cu content. The trend of oxidized Co binding energy indicates a complex CoP bonding behavior with weaken CoP bonds in the presence of Cu. A comparison of molar ratio of Cu/Co in Cu x Co 3– x P/NC between starting materials and surface composition obtained by XPS can be seen in Table S1 (Supporting Information). The mixed Co oxidation states and associated phase behaviors are essential in determining their catalytic activity, which are shown in the electrochemical performance below. Figure 2 Open in figure viewerPowerPoint a) High‐resolution XPS spectra of N2p, including pyridinic, pyrrolic, and quaternary N. b) XPS survey of P2p with the presence of phosphide and phosphate peaks. c) Deconvoluted spectra for Cu x Co 3– x P/NC with different Cu doping concentration.

2.2 Electrochemical Performance of BCP‐ZIFs in Alkaline Media The incorporation of second transition metal into cobalt‐based phosphide and thus induced mixed CoP x phase are of great importance to understand the relationship between elemental composition and their corresponding activities. The electocatalytic OER of Cu x Co 3– x P/NC was first evaluated by linear sweep voltammetry (LSV) using RDE in N 2 ‐saturated alkaline electrolyte (1 m KOH). As revealed in Figure 4a, Cu 0.3 Co 2.7 P/NC shows obviously superior performance compared to other Cu composition with the unprecedentedly low overpotential of η = 0.19 V at 10 mA cm−2 with a catalyst mass loading of 0.4 mg cm−2. Among different transition metal‐doped BCP‐ZIFs, Cu dopant presents the best OER catalytic activity, which is not only better than that of Zn 0.3 Co 2.7 P/NC (η = 0.33 V), Ni 0.3 Co 2.7 P/NC (η = 0.28 V) at j = 10 mA cm−2, but outperforms commercial RuO 2 (η = 0.27 V), as shown in Figure 3b. The optimal OER performance of Cu 0.3 Co 2.7 P/NC becomes clearer at high potential with rapid current increase and high current density j = 325 mA cm−2 at η = 0.77 V versus RHE (reversible hydrogen electrode). Besides such high catalytic activity, Cu 0.3 Co 2.7 P/NC also displays excellent stability in electrochemical OER (Figure 3c). In fact, there is merely 0.05 V positive shift for η at 10 mA cm−2 and negligible change of current density for bias larger than V = 1.68 V after 1000 cycles. In Figure 3d, the kinetic analysis of OER is carried out by measuring the slope of linear part of Tafel plots. For CoP/NC, Cu 0.3 Co 2.7 P/NC, Cu 0.6 Co 2.4 P/NC, Cu 1 Co 2 P/NC, and RuO 2 , the Tafel slopes are 57, 44, 50, 65, and 63 mV dec−1, respectively. The exchange current density is derived from Tafel plots, using the equation η/b = log (j/j 0 ), where b is the Tafel slope, j is the current density, and j 0 is the exchange current density. The obtained j 0 of Cu 0.3 Co 2.7 P/NC is 2.63 × 10−9 A cm−2, which is higher than the state‐of‐the‐art commercial RuO 2 , j 0 = 2.51 × 10−9 A cm−2. The low Tafel slope and high exchange current of Cu 0.3 Co 2.7 P/NC indicate that the Cu/Co molar ratio in Cu x Co 3– x P/NC plays an important role in determining OER catalytic activity. With mild Cu doping in Cu 0.3 Co 2.7 P/NC (x = 0.3), the electrochemical activity reaches its maximum, while continuously increasing the doping content results in decreasing catalytic activity. The composition‐dependent electrochemical active surface area of Cu x Co 3– x P/NC is also taken into account for its contribution to OER activity. The trend similar to the Tafel slopes is observed by measuring the electrochemical double layer capacitance (C DL ), where the C DL versus composition forms a volcano‐like plot with Cu 0.3 Co 2.7 P/NC (30.9 F g−1) sitting at the tip of volcano, CoP (20.5 F g−1), Cu 0.6 Co 2.4 P/NC (19.2 F g−1), and Cu 1 Co 2 P/NC (9.1 F g−1) residing at the wings (Figure 3e). The measured electrochemical impedance spectroscopy (EIS) also confirms the effect of Cu dopant on the charge‐transfer resistance increasing from 5.6 Ω for Cu 0.3 Co 2.7 P/NC to 5.8 Ω for Cu 0.6 Co 2.4 P/NC, 6.1 Ω for CoP/NC, and 6.6 Ω for Cu 1 Co 2 P/NC (Figure 3f). The low charge‐transfer resistance demonstrates a better conductivity with low Cu‐doping content.29 Moreover, the Cu 0.3 Co 2.7 P/NC displays a straighter line along the imaginary axis than the other composition, indicating a low diffusion resistance.30 As expected, both EIS and kinetic analysis are high in accordance with Cu x Co 3 –x P/NC OER performance displayed in LSV plots. In addition, the intrinsic catalytic activities with different Cu content are studied on the basis of turn over frequency (TOF), in which Cu 0.3 Co 2.7 P/NC shows the highest value of 5.82 × 10−3 s−1, larger than that of Cu 0.6 Co 2.4 P/NC (2.72 × 10−3 s−1), Cu 1 Co 2 P/NC (1.96 × 10−3 s−1), CoP/NC (2.87 × 10−3 s−1), and RuO 2 (0.61 × 10−3 s−1) (Table S2, Supporting Information). It is also worth noting that the electrochemical performance of Cu 0.3 Co 2.7 P/NC, including ultralow overpotential, high exchange current density, and low Tafel slope is among the best nonprecious metal OER catalysts (Table S3, Supporting Information). Figure 3 Open in figure viewerPowerPoint a) LSV plots of Cu x Co 3– x P/NC and commercial RuO 2 measured in 1 m KOH electrolyte using RDE. b) Comparison of OER performance of Zn 0.3 Co 2.7 P/NC, Ni 0.3 Co 2.7 P/NC, Cu 0.3 Co 2.7 P/NC, and RuO 2 . c) LSV plots showing strong stability of Cu 0.3 Co 2.7 P/NC after 1000 potential cycles. d) Normalized Tafel plots of Cu x Co 3– x P/NC and commercial RuO 2 show superior reaction kinetic with Cu 0.3 Co 2.7 P/NC catalyst. e) Volcano‐like plot of electrical double‐layer capacitance C DL of Cu x Co 3– x P/NC with different Cu concentration. f) Nyquist plots derived from EIS measurements, showing lower charge‐transfer resistance of Cu 0.3 Co 2.7 P/NC. We next assess the HER performance of BCP‐ZIFs in order to screen the optimal bifunctional catalyst for our final water‐splitting tests. The HER experiments are first carried out by LSV in the same N 2 ‐saturated 1 m KOH solution resembled to OER testing. As expected in Figure 5a, Cu 0.3 Co 2.7 P/NC exhibits a low overpotential at 10 mA cm−2 of η = 0.22 V, which is smaller than that of Cu 0.6 Co 2.4 P/NC and Cu 1 Co 2 P/NC with higher Cu content. The TOF values calculated for different Cu x Co 3– x P/NCs and Pt are shown in Table S2 (Supporting Information), where Cu 0.3 Co 2.7 P/NC exhibits a value of 24.29 × 10−3 s−1, much higher than other Cu composition. To compare the influence of different transition metal dopants, Zn 0.3 Co 2.7 P/NC and Ni 0.3 Co 2.7 P/NC catalysts are also investigated and shown in Figure 4b. The overpotential of Cu 0.3 Co 2.7 P/NC is found to be 0.06 V and 0.04 lower than its Zn‐doped and Ni‐doped counterparts, respectively. As previously reported, the transition metal can be more positively charged in the presence of P atoms, in which the negatively charged P can attract the hydrogen proton at low coverage and desorb H 2 at high coverage in electrochemical HER.31 Thus, one can expect that an increasing atomic percentage of P phase in CoP x (CoP vs Co 2 P) renders better HER performance. The electronic conductivity, however, is usually compromised for metal phosphides with low transition metal to P ratio, featured by weaker metallic character.18, 32 Thus a perfect balance between activity and conductivity is desired for maximizing electrochemical performance. In our bimetallic cobalt‐based phosphides, such a balance is achieved by Cu 0.3 Co 2.7 P/NC with the smallest amount of Cu ions doped in the cobalt lattice. Moreover, as revealed by both XRD and XPS surveys, Cu 0.3 Co 2.7 P/NC shows the lowest amount of Co0, while Cu 1 Co 2 P/NC consists of a majority of metallic Co and Co2+ species. Therefore, we tentatively propose that the CoP and Co 2 P species are the decisive components for the optimized activity in Cu x Co 3– x P/NC catalyst. Other than high HER catalytic activity, Cu 0.3 Co 2.7 P/NC also exhibits strong stability, as shown by a small negative shift of 0.023 V at j = 10 mA cm−2, and indistinguishable difference of current density beyond η = 0.39 V after 1000 continuous cycling (Figure 4c). Although there is a lack of consensus on the HER reaction process in alkaline media, regardless of any possible routes the HER proceeds, hydrogen atom adsorbed on the catalyst surface (H ads ) is always involved during the reaction. In order to examine the reaction pathway in our alkaline media, coverage of adsorbed hydrogen occurred in either Volmer–Tafel or Volmer–Heyrovsky pathway is adopted as an analogous model as in acidic media. The theoretical Tafel slopes of each of the following reactions are 120, 30, and 40 mV dec−1, respectively33 Figure 4 Open in figure viewerPowerPoint a) LSV plots of Cu x Co 3– x P/NC and commercial Pt/C measured with RDE at 2000 rpm. b) Effect of different transition metal doping (Zn, Ni, and Cu) in determining HER performance of BCP/NC‐ZIFs. c) Stability test of Cu 0.3 Co 2.7 P/NC before and after 1000 continuous potential cycling. d) Tafel plots of Cu x Co 3– x P/NC and commercial Pt/C. In Figure 4d, the calculated Tafel slope of Cu 0.3 Co 2.7 P/NC is 122 mV dec−1, which is slightly larger than that of Pt/C (105 mV dec−1) and smaller than its doped and undoped counterparts (123 mV dec−1 for CoP/NC, 131 mV dec−1 for Cu 0.6 Co 2.4 P/NC, and 129 mV dec−1 for Cu 1 Co 2 P/NC). In spite of the hydrogen evolution mechanism in alkaline media that remains elusive, the reaction is more likely governed by Volmer–Heyrovsky, close to pure Ni with a similar Tafel slope of 120 mV dec−1.34

2.3 Optimized Electronic Structure and H* Adsorption Energy Density functional theory (DFT) analysis was also carried out to explore the origin of superior catalytic process upon foreign metal doping. The Nyquist plot discussed above shows a decreased charge‐transfer resistance upon mild Cu doping, which is closely associated with improved electrical conductivity of the catalysts. It is obvious that all four systems, CoP, Co 2 P, Cu‐CoP, and Cu–Co 2 P, demonstrate the metallic feature since the Fermi level lies within the conduction band (Figures S9–S12, Supporting Information). Around the Fermi level, most of the densities of states in pristine and doped CoP come from the hybridization of metal d‐states and phosphorous p‐states, while in pristine and doped Co 2 P mainly from metal d‐states. Figure 6a,b,d,e shows the change of bonding charge distribution in the crystal structure of Cu‐doped CoP and Co 2 P from pristine CoP and Co 2 P. In pristine CoP and Co 2 P, electron accumulation (yellow cloud) can be seen in the space between a Co atom (red) and a P atom (red), which indicates the formation of covalent bonds between these atoms. By replacing one Co atom by a Cu atom (purple), it is easy to observe a certain amount of charge accumulation around Co atoms while less accumulation around Cu atoms in both Cu‐doped CoP and Co 2 P systems. Moreover, no charge accumulation is found between Cu and P atoms. This may explain the enhancement in conductivity in Cu‐doped systems since the introduction of Cu decreases the number of covalent bonds (CoP) in the systems by creating more free electrons. Notably, since Co 2 P possess more electrons per unit cell than CoP, more free electrons will be released with Cu doping in Co 2 P‐riched phase, which are beneficial for a better electrical conductivity. Figure 5 Open in figure viewerPowerPoint Red circles represent Co atoms; blue circles represent P atoms; purple circle represents Cu atoms; and yellow clouds represent charge accumulation. a) Bonding charge distribution for pristine CoP. Charge accumulation can be seen in between Co and P atoms. b) Bonding charge distribution for Cu‐doped CoP and electron depletion around the Cu atom. c) Dramatic lower H* adsorption energy of both CoP and Co 2 P host structures upon Cu doping. d) Bonding charge distribution for pristine Co 2 P. e) Bonding charge distribution for Cu‐doped Co 2 P. Hydrogen adsorption energy is another critical factor in determining the ultimate catalytic activity. One of the agreements in searching for a good electrocatalyst is to possess a moderate H adsorption free energy, which greatly affects the efficiency of adsorption and desorption steps when splitting water. Commercial Pt has the most desirable H adsorption free energy around 0.09 eV,20 approaching the thermoneutrality. Our DFT analysis reveals that the Cu doping significantly changes the value of hydrogen adsorption free energy in both CoP and Co 2 P systems. For instance, in Figure 5c, the hydrogen adsorption free energy is 0.48 eV for pristine CoP (001) surface, and 0.21 eV for Cu‐doped CoP (001) surface, while the doping of Cu also changes the hydrogen adsorption free energy from 0.54 eV in pristine Co 2 P (001) surfaces to 0.28 eV in Cu‐doped Co 2 P (001) surfaces. Contrary to its pristine counterpart, Cu‐doped Co phosphide shows excellent catalytic activity via a synergistic effect of superior electrical conductivity of Co 2 P phase and low hydrogen adsorption of CoP phase.