Skip to main content
NCBI home page
RSC Advances logoLink to RSC Advances
. 2026 Oct 5. Online ahead of print. doi: 10.1039/d6ra06796d

Trimetallic synergy in Ni2−x−yCoxFeyP precatalysts enables robust oxygen evolution

Faria Sultana a, Stephanie L Brock a,✉
PMCID: PMC13636294  PMID: 42835587

Abstract

Developing earth-abundant and stable oxygen evolution reaction (OER) catalysts remains important for practical utilization of alkaline water electrolysis. Previous efforts revealed that bimetallic, Mn-containing phosphides, including Co2−xMnxP, Fe2−xMnxP, and Ni2−xMnxP, are promising OER precatalysts forming relatively stable oxyhydroxide catalysts upon oxidation of phosphide to solubilized phosphate under alkaline conditions. Notably, the addition of Mn produces more active catalysts than monometallic Co2P, Fe2P or Ni2P, with optimal activity (minimized overpotential) found near 1 : 1 M : Mn ratio. In the present work, the role of Mn in catalysis is addressed indirectly by probing trimetallic phosphides comprising Co, Fe, and Ni without Mn. Ni0.67Co0.67Fe0.67P nanocrystals exhibit enhanced catalytic activity relative to their mono- and bimetallic counterparts, achieving an overpotential (η10) of 276.5 ± 3.7 mV at 10 mA cm−2. The composition was further varied to create Ni-rich (NiCo0.5Fe0.5P), Co-rich (Ni0.5CoFe0.5P) and Fe-rich (Ni0.5Co0.5FeP) phases, with the Ni-rich phase exhibiting a lower overpotential (η10 = 257.0 ± 5.0 mV) than the 1 : 1 : 1 trimetallic phase, while the Co- and Fe-rich phases proved less active. This non-monotonic trend indicates that activity is not governed simply by the concentration of an individual metal, but instead arises from an optimized Ni–Co–Fe environment in which Ni enrichment most effectively promotes the catalytically active state. The stability of Ni0.67Co0.67Fe0.67P was evaluated through chronoamperometry (15 h bulk electrolysis at 1.7 V vs. RHE) and X-ray photoelectron spectroscopy (XPS) before and after catalysis. The current density showed minimal decay (∼6%, from ∼32 to ∼30 mA cm−2), indicating robust performance. Notably, relevant to the bimetallic Mn-containing systems, the trimetallic (Co, Fe, Ni) phases are both more active and more stable under active OER conditions. This study demonstrates that trimetallic synergy between Co, Fe, and Ni results in enhanced alkaline OER activity and stability relative to bimetallic systems, including those containing Mn.

Developing earth-abundant and stable oxygen evolution reaction (OER) catalysts remains important for practical utilization of alkaline water electrolysis.graphic file with name d6ra06796d-ga.webp

Introduction

Solar (or wind)-driven water electrolysis presents an attractive pathway to clean hydrogen production, yet its efficiency is curtailed by kinetic barriers for both the water oxidation (oxygen evolution reaction, OER) and water reduction (hydrogen evolution reaction, HER) processes. However, it is the OER process, which is responsible for generating the protons and electrons employed in HER, that is the limiting step. Stable catalysts capable of reducing the overpotential (additional potential over thermodynamic) are needed, with the state-of-the-art comprising phases such as IrO2 and RuO2. The scarcity and cost associated with precious metal-containing catalysts has spurred interest in identifying earth-abundant inexpensive alternatives, among them precatalysts comprising phosphides of 3d metals. While a substantial body of work has been published on this topic, a broader understanding of how structural and compositional changes in the phosphide precatalyst influences the activity and stability of the in situ derived OER catalyst remains elusive. Variations in size (degree of polydispersity), shape, phase purity, and the types of support make direct comparison inherently difficult. For example, temperature programmed reduction of metal oxide with ammonium phosphates usually results in the formation of relatively large particles leading to low-surface areas,1–3 but the surface area can be increased by the inclusion of organic additives.4–6 Likewise, solvothermal reduction of metal halide with Na3P also leads to the formation of larger particle sizes due to the quick temperature rise during the exothermic metathesis process.7 On the other hand, a wide variety of supports with different chemistries and surface areas are employed such as, FTO, carbon cloth, graphene, graphite plates, carbon nanotubes, carbon paper, metal mesh, metal foam, and metal foils.8–12 Since the nature of the catalyst–support interface is expected to govern the catalytic activity, it is not possible to compare activity across different supports.

The Brock lab has established a strong foundation in synthesizing transition metal phosphide (TMP) materials through colloidal routes with an organophosphorus source to produce discrete, phase-pure and compositionally homogenous samples with low polydispersity. Among precatalyst candidates for OER, we have focused on bimetallic Mn-containing phosphides because of their low cost and redox flexibility, which can orchestrate the in situ generation and dynamic evolution of catalytically active sites during oxygen evolution.13,14 As a conductive support, we use Ketjan-300J carbon black in all cases, probing how variations in metal stoichiometry dictate OER activity and stability in M2−xMnxP, where M = Fe, Co and Ni.15–17 Among these three, Ni2−xMnxP (x = 1) exhibited the highest activity for alkaline OER, requiring the lowest overpotential (η10 = 280 mV), compared with Co2−xMnxP (x = 1, η10 = 330 mV) and Fe2−xMnxP (x = 0.9, η10 = 440 mV). Interestingly, all three systems showed similar trends: (1) the highest activity was shown for equal ratio of both metals (note, x = 0.9 was the most Mn-rich phase attainable for Fe2−xMnxP); (2) deactivation of the catalysts was observed under active conditions (10–20% over 15 h), accompanied by dissolution of Mn into the KOH electrolyte.

In the present work, we seek to probe the compositional anomaly (equal ratios of metal produce the most active catalysts in bimetallic phases) by investigating compositional effects in a trimetallic M2P system. At the same time, we seek to establish whether elimination of Mn can lead to more stable systems. Thus, we report OER activity of a previously developed trimetallic phosphide precatalyst incorporating Ni, Co, and Fe—three transition metals with significant value in OER electrocatalysis.18 Various kinds of phosphide systems based on each of these individual elements or their combinations synthesized through colloidal routes have recently been studied by several labs. For example, a synthesis of bimetallic phosphide nanorods has been reported using Co2P as the core and MPx (M = Cu, Ni, Co, Mn) as the shell, followed by annealing on a carbon nanotube (CNT) support to form homogeneous, bimetallic phosphide nanorods. Among the various compositions examined, CNT-Co1.33Fe0.67Px showed the highest OER activity (η10 = 293.4 mV), with this specific Fe content identified as optimal. In that study, a trimetallic phosphide (CNT-Co1.01Fe0.43Ni0.56Px) was also synthesized, but its OER performance was not evaluated.19 High entropy metal phosphide materials are another important subclass for catalyzing OER. NiCoFeMnCrP particles of ∼100 nm in diameter, nearly equimolar in the different metals, and adopting the Ni2P structure type, were synthesized via a sol–gel/thermal reduction process. The pentametallic phosphide was active for OER with η10 = 220 mV.20 For comparison, a “NiCoFeP” phase was reported with η10 = 335 mV, but the composition of this phase was not reported, nor was it subjected to post-catalysis characterization or stability testing. In a separate study, tetrametallic phosphide nanoparticles of CoNiFeCuP (as-synthesized composition based on atomic percentages: Co 19.3%, Ni 19.7%, Fe 15.2%, Cu 14.5%, P 31.3%) were prepared via a colloidal method, and evaluated for OER, delivering an overpotential of η10 = 244 mV.21 In this study, OER performance was also reported for a trimetallic phase “CoNiFeP”, which formed as ca. 120 nm spiky aggregates of 6 nm diameter nanorods, exhibiting an overpotential η10 = 275 mV; as was the case for “NiCoFeP”, post-catalysis characterization and stability tests were not reported.

In this work, we compare the activity of the homogenous phase, Ni2−x−yCoxFeyP (x = y = 0.67) to monometallic and Ni-based bimetallic phosphides, followed by evaluation of Ni-rich vs. Co-rich vs. Fe-rich compositions. We find that the trimetallic system outperforms the bimetallic systems with the Ni-rich trimetallic system the most active phase (η10Ni0.67Co0.67Fe0.67P = 276.5 ± 3.7 mV vs. ηNi-rich10 = 257.0 ± 5.0 mV). We also conducted a comprehensive stability assessment of the Ni0.67Co0.67Fe0.67P nanocrystals using multiple characterization techniques, including LSV cycling, chronoamperometry (15-hours bulk electrolysis), and XPS analyses before and after catalysis. Post-catalysis elemental mapping and ICP-MS analysis of the spent electrolyte was further employed to corroborate and complement the XPS findings. As described below, we find that Ni0.67Co0.67Fe0.67P nanoparticles outperform Mn-based phases in terms of stability and activity, despite the fact that iron is partially solubilized during the catalytic process. In the broader context, this work contributes to the establishment of a general understanding of how compositional variations in colloidally synthesized phosphide precatalysts influence their electrocatalytic performance.

Experimental section

Chemicals

All the chemicals used in this study along with their manufacturers and purity are listed in Table S1. Oleylamine (OAm) was stored over activated molecular sieves (3 Å) for at least 2 days before use and deoxygenated by purging argon on a Schlenk line. Nickel acetylacetonate, Ni(acac)2 and cobalt acetylacetonate, Co(acac)2 were handled in air. The remaining synthesis precursors were handled and stored with care inside an argon filled glove box. For the nanoparticle purification, nanoparticle–carbon black composite synthesis and ink preparation, the chemicals were used as received.

Synthesis of Ni2−x−yCoxFeyP nanoparticles

Into a 100 mL two neck Schlenk flask equipped with a 12-inch-long West condenser, an internal probe temperature controller (sensor type-K) and a medium size magnetic stir bar (5/8″ × 5/16″), (2 − x − y) mmol of nickel acetylacetonate, (Ni(acac)2), x mmol of cobalt acetylacetonate, (Co(acac)2), 2 mL of dried oleylamine (OAm), 3 mL of octadecene (ODE) and 5 mL of n-trioctylphosphine (TOP) were combined in an argon filled glove box at room temperature. The reaction flask was transported to a dual manifold Schlenk line and placed in a flexible glass fabric type 100 mL heating mantle regulated by the internal probe temperature controller. Glass wool was used to insulate the reaction system from the external atmosphere. The reaction system was first degassed at 110 °C for 20 minutes under constant stirring, followed by an argon purge for 20 minutes. This pump-purge sequence was repeated twice more. The formation of Ni–Co–P alloy was initiated by increasing the reaction temperature to 230 °C and subsequently aging for 1 hour. Next, y mmol of Fe(CO)5 in 3 mL of TOP was injected to the reaction flask (syringe loaded in the glovebox) to form a Ni–Co–Fe–P alloy via continued heating for an additional hour at 230 °C. Finally, to obtain nanocrystals of Ni2−x−yCoxFeyP, the reaction temperature was raised to 300 °C. Throughout all the reactions, (1) argon was continuously purged at a low flow rate to minimize the possibility of air exposure during the synthesis; and (2) during heating, the reaction time was recorded only after the system reached the designated set temperature. After 3 hours of annealing at 300 °C, the reaction was stopped, and the reaction flask was cooled down to room temperature overnight. Then the reaction flask was disconnected from the Schlenk line for nanoparticle purification.

For the trimetallic nanocrystals having equal metal ratios, 1.34 mmol of each precursor was used, whereas for the Nirich system, 1.0 mmol of Ni(acac)2 and 0.5 mmol of both Co(acac)2 and Fe(CO)5 were taken. Similarly, for the Corich and Ferich trimetallic systems, 1.0 mmol of Co and 1.0 mmol of Fe precursors were taken respectively, keeping the remaining two metal precursors equal to 0.5 mmol each.

Purification of Ni2−x−yCoxFeyP nanoparticles

The post-synthesis mother liquor was transferred to a 50 mL falcon tube and combined with 5–10 mL chloroform to disperse the organically capped nanocrystals. The mixture was sonicated for 5 min, after which ∼15 mL ethanol was added as an antisolvent to induce precipitation of the trimetallic phosphide nanoparticles. The suspension was centrifuged at 4000 r.p.m. for 15 min, and the product was redispersed in chloroform. This dispersion–precipitation procedure was repeated three times. The purified nanoparticles were collected and stored for subsequent characterization.

Nanoparticle–carbon black composite preparation

The purified nanoparticles were completely dried by continuously purging argon; then 50 mg of nanoparticles and 25 mg of Ketjen-300J carbon (C) black (2 : 1) were mixed in 5–10 mL of hexane and this mixture was sonicated for 30 minutes to make the nanoparticle–carbon black composite. The composite was collected by adding ethanol after sonication and centrifuging at 4000 r.p.m. for 10 minutes. Finally, the air-dried composite was transferred to an alumina boat placed in a fused silica flow tube in a furnace and annealed at 400 °C for 1 h with a ramp rate of 5 °C min−1 under reducing atmosphere (5% H2, balance N2). The effluent was purged through a hypochlorite solution acting as a getter in the unlikely event phosphine is produced. The annealed composite was collected and stored in the glove box.

Ink preparation

15 mg of annealed composite was mixed with 2 mL of ethanol, 1 mL of iso-propanol and 1 mL of nanopure water and sonicated for 20 minutes. Next, 1 mL of 5 wt% Nafion was added to the mixture and sonicated for 10 more minutes to obtain the precatalyst ink for electrochemical analysis.

Powder X-ray diffraction (PXRD)

Powder X-ray diffraction (PXRD) patterns were collected using a Bruker D2 PHASER diffractometer equipped with a Cu Kα radiation source (λ = 0.15406 nm) operating at 30 kV and 10 mA. Samples were prepared by drop-casting concentrated dispersions of the nanomaterials in chloroform onto a low background quartz sample holder and dried under ambient conditions. Diffraction data were acquired over a 2θ range of 30–80° with a step size of 0.02° at 0.5 s per step. The collected diffraction patterns were processed using Bruker EVA software, and peak positions were compared with reference Powder Diffraction Files (PDFs) from the International Centre for Diffraction Data (ICDD) to confirm phase purity and assign crystalline phases.

Transmission electron microscopy (TEM), energy-dispersive X-ray spectroscopy (EDS), and scanning transmission electron microscopy (STEM)

Transmission electron microscopy (TEM) for imaging and energy-dispersive X-ray spectroscopy (EDS) for average elemental mapping were performed using a Talos F200X S/TEM instrument. Samples were prepared by dispersing as-synthesized nanoparticles in chloroform and post-catalysis composite samples in methanol, followed by drop-casting onto carbon-coated 200-mesh Cu grids and drying under ambient conditions. Area specific elemental mapping was performed in high angle annular dark field (HAADF) mode. Particle size distributions for the nanoparticles were determined by analyzing TEM images with ImageJ software, with a minimum of 200 particles evaluated per sample.

X-ray photoelectron spectroscopy (XPS)

High-resolution Ni 2p, Co 2p, Fe 2p, and P 2p spectra were acquired on a Thermo Scientific Nexsa X-ray photoelectron spectrometer equipped with a monochromatic Al Kα (1486.7 eV) source and hemispherical analyzer. All spectra were acquired with a 50 eV pass energy, 0.1 eV energy step, and a dwell time of 100 ms per point. The C 1s signal at 284.8 eV was used as an internal reference for charge correction and binding-energy calibration.

Inductively coupled plasma mass spectrometry (ICP-MS)

Agilent 7700 inductively coupled plasma-mass spectrometer (ICP-MS) was used to quantify Ni, Co, Fe and P species dissolved in the post-electrolysis electrolyte. Following the 15 hours long bulk electrolysis, 1 mL of the KOH electrolyte was collected, digested with 2% HNO3 and 0.5% HCl, and diluted to 15 mL with the same HNO3–HCl mixture. Calibration curves were constructed using external standards (0, 1, 5, 10, 25, 50, 100, 150, and 200 ppb from mixed element standard, IV-ICPMS-71A, each prepared in 2% HNO3 and 0.5% HCl to a final volume of 15 mL.

Electrochemical measurements

Cyclic voltammograms were recorded using an EC Epsilon potentiostat equipped with a rotating disk electrode (RDE) in a conventional three-electrode configuration, employing an Ag/AgCl reference electrode and a Pt wire counter electrode. A glassy-carbon disk (3 mm diameter, geometric surface area 0.07 cm2) served as the working electrode, onto which 10 µL of catalyst ink was drop-cast (0.286 mg cmgeo−2 loading). Measurements were performed in 1.0 M KOH, wherein catalysts were activated by cycling with 20 segments prior to data collection. CVs were then acquired at a scan rate of 10 mV s−1 with electrode rotation at 1600 rpm. Potentials were referenced to Ag/AgCl and converted to the reversible hydrogen electrode (RHE) scale according to ERHE = EAg/AgCl + 0.197 + 0.059 × pH (pH = 14). 95% iR correction was applied for all polarization studies using the internal compensation function in the Epsilon software, which follows ERHE = EAg/AgCl + (0.197 + 0.059 × pH) − iR. The overpotential at 10 mA cm−2 was extracted directly from the measured LSV curve using the total recorded current density, with no background correction.

Double-layer capacitances (Cdl) were determined in the non-faradaic region using CV scans at 5, 10, 25, 50, 75, and 100, 200, 400 and 800 mV s−1 within ±50 mV of the open-circuit potential. Cdl values were obtained from the slope of the current versus scan-rate plots. Both anodic and cathodic currents were plotted against scan rates, and the obtained slopes were averaged to get the final Cdl value. Electrochemically-active surface areas (ECSA) were calculated as Cdl/Cs, where a Cs value of 0.04 mF cm−2 was employed, consistent with reported values in alkaline media.

Bulk electrolysis and faradaic efficiency

Bulk electrolysis experiments were conducted in a custom-built H-type cell using 1.0 M KOH as the electrolyte, with a Pt counter electrode and an Ag/AgCl reference electrode. Carbon cloth (1 cm2) was kept in the oven at 150 °C for at least 4 hours, after which 140 µL of catalyst ink was dropcasted and the electrode was dried under an IR lamp. Oxygen evolution was quantified using a GOW-MAC Series 400 gas chromatograph equipped with a thermal conductivity detector (TCD). Helium and nitrogen served as the carrier and internal standard gases, respectively. Faradaic efficiency (FE) was determined by comparing the theoretical O2 amount, calculated from total charge passed, with the experimentally measured O2 content in the headspace. FE calculations are explained in detail in the SI Section S1.

Results

Synthesis and characterization of Ni0.67Co0.67Fe0.67P

The synthesis of trimetallic Ni0.67Co0.67Fe0.67P nanoparticles was conducted in three steps (Scheme 1) modifying our previous published route by dissolving the Fe precursor in n-trioctylphosphine (instead of in octadecene) and increasing the annealing time for crystallization from 1 h to 3 h.18 First, the acetylacetonate precursors (M(acac)2, where M = Ni, Co) were combined with the reducing agent and weak ligand oleylamine (OAm), along with the phosphidation agent/ligand trioctylphosphine (TOP), and octadecene (ODE) to form the Ni–Co–P alloy. Second, the Fe precursor, zero-valent Fe(CO)5, was added to form the trimetallic phosphorus alloy at 230 °C. In the third step, the overall reaction temperature was increased to 300 °C to induce crystallization of the trimetallic phosphide nanoparticles.

Scheme 1. Colloidal synthetic protocol for Ni2−x−yCoxFeyP nanoparticles.

Scheme 1

Open in a new tab

The particles were isolated by adding solvent, chloroform, to disperse the hydrophobic ligand-capped nanoparticles and precipitated out by addition of anti-solvent, ethanol. The resultant Ni0.67Co0.67Fe0.67P nanoparticles were characterized by powder X-ray diffraction (PXRD), transmission electron microscopy (TEM), energy-dispersive X-ray spectroscopy (EDS), and scanning transmission electron microscopy (STEM) (Fig. 1). The obtained PXRD pattern (Fig. 1a) matches the hexagonal parent structure Ni2P (Fe2P-type), with no additional peaks. TEM (Fig. 1b) reveals the particles to be faceted, asymmetric cubes of an average length of 13.9 ± 0.2 nm and an average width of 11.4 ± 0.1 nm. The line scan elemental mapping in STEM mode (Fig. 1c and d) shows the distribution of the metals in the nanocrystals to be homogenous and co-localized with phosphorus. Atomic percentages of each metal and phosphorus were obtained from the elemental mapping analysis: Ni = 20.9 ± 0.5%, Co = 20.7 ± 0.9%, Fe = 21.6 ± 0.8% and P = 36.7 ± 1.6% (Ni : Co : Fe = 1.01 : 1 : 1.04) and the metal to P ratio is 1.72 : 1. This modest deviation from the 2 : 1 ratio is attributed to ligated TOP on the particle surface. These data are consistent with the formation of ca. 13 nm nanoparticles of a Fe2P-type phosphide solid solution comprising equimolar Ni, Co and Fe, with no evidence of crystalline impurities.

Fig. 1. Characterization of nanoparticles of targeted composition Ni0.67Co0.67Fe0.67P (a) PXRD pattern and reference pattern (Ni2P); (b) TEM image and size distribution histogram; (c) STEM-EDS mapping showing a line scan; (d) line scan result of (c) showing the co-localized, equimolar distribution of Ni, Co and Fe.

Fig. 1

Open in a new tab

Evaluation of geometric surface area-normalized OER activity

The colloidal nanoparticles are synthesized in the presence of ligands to controllably tune their size.22,23 The synthesized nanoparticles were made into a composite with carbon black (Ketjen-300J carbon), which also serves to interface the particles to a conductive support. To facilitate access to active sites during OER, the ligands are removed from the surface via high temperature (400 °C) annealing under a reducing environment. After annealing, the composite is incorporated into an ink comprising alcohol, nanopure water and Nafion. Finally, the OER activity was evaluated in 1.0 M KOH solution (pH = 14) via 10 cycles (20 segments) of cyclic voltammetry (CV) from 1.0 to 2.0 V for activation, followed by linear sweep voltammetry (LSV) with 95% iR correction (uncompensated raw LSV is also provided along with the 95% iR corrected LSV in Fig. S1). The activation process is performed to transform the precatalyst phosphide into a hydroxy/oxyhydroxide catalyst surface-bound to the phosphide. During cycling of the trimetallic Ni0.67Co0.67Fe0.67P in this region, the reversible redox couple between Ni2+ and Ni3+ is apparent as shown in Fig. S2. The current value in mA was normalized to the geometric surface area of the glassy carbon electrode (3 mm diameter, surface area = 0.07 cm2). For comparison, along with the trimetallic phosphide, the bimetallic, NiCoP and NiFeP and the monometallic Ni2P and Co2P nanoparticles were also synthesized and characterized as described in SI Section S2 and Fig. S3, S4 and made into ink following the same procedure as for the trimetallic phosphide.

During LSV measurements, all Ni-containing samples, with the exception of NiFeP, displayed an anodic peak attributed to Ni2+ → Ni3+ conversion (Fig. 2a). The position of this peak changes depending on the presence and identity of secondary (and tertiary) metals, reflecting their electron-donating/withdrawing abilities. As summarized in Table 1, with Ni2P as a precatalyst, the highest potential is attained (1.43 V), whereas substitution with Fe (50%) results in complete suppression of the peak within the potential window, and Co (50%) reduces the potential relative to the Ni2P precatalyst to 1.34 V. When both Fe and Co are incorporated into Ni2P to form Ni0.67Co0.67Fe0.67P, a potential that is intermediate between Ni2P and NiCoP is obtained (1.38 V).

Fig. 2. (a) OER polarization curve for the trimetallic Ni0.67Co0.67Fe0.67P (red), bimetallic NiCoP (sky blue), bimetallic NiFeP (green), and monometallic Ni2P (purple); (b) overpotential at 10 mA cm−2 current density (geometrically normalized) for all four systems plus monometallic Co2P (dark blue) and previously-reported Fe2P15 (orange).

Fig. 2

Open in a new tab

Table 1. Precatalysts with their corresponding Ni2+ → Ni3+ anodic peak potential (vs. RHE), the required overpotential at 10 mA cm−2 and 40 mA cm−2, and Tafel slopes. The error bars in the overpotential measurement correspond to two duplicate trials from two sets of samples for each system except for NiFeP, where four trials were performed. For Fe2P, the η10 value was taken from the literature15.

Precatalysts Ni2+ → Ni3+ Peak potential (V) Overpotential @10 mA cmgeo−2 (mV) Overpotential @40 mA cmgeo−2 (mV) Tafel slope (mV dec−1)
Ni2P 1.43 336.5 ± 1.2 438.0 147
NiFeP Not detected 290.5 ± 16.5 450.0 173
NiCoP 1.34 325.5 ± 7.0 401.0 111
Ni0.67Co0.67Fe0.67P 1.38 276.5 ± 3.7 351.0 103
Co2P N/A 400.0 ± 11.3 519.0 N/A
Fe2P N/A 590.0 N/A N/A
Open in a new tab

Fig. 2b compares the overpotentials at 10 mA cm−2 (η10) for the four Ni-based systems shown in Fig. 2a, along with two additional monometallic phosphides, Co2P and Fe2P (for Fe2P, the η10 value was taken from the literature15) and the overpotential data are collected in Table 1. The monometallic phosphide precatalysts exhibited the highest overpotentials, scaling as Ni2P < Co2P < Fe2P; incorporation of Fe or Co into Ni2P resulted in a significant drop vs. Ni2P (336.5 mV for Ni2P vs. 290.5 mV for NiFeP vs. 325.5 mV for NiCoP), and incorporation of both Fe and Co to form Ni0.67Co0.67Fe0.67P resulted in a further decrease in overpotential to 276.5 mV. In order to account for differences in electrochemically active surface area (ECSA) from sample-to-sample, CV analyses were conducted in the non-faradaic region for Ni0.67Co0.67Fe0.67P, NiCoP, NiFeP and Ni2P and these data were used to compute relative ECSA data (Fig. S5–S8). The same trend found when normalizing to geometric surface area was also found in the ECSA-normalized current density vs. potential curve, confirming the superiority of the trimetallic Ni0.67Co0.67Fe0.67P in terms of overpotential requirement (Fig. S9). Moreover, the four systems shown in Fig. 2a were further evaluated by plotting log J (where J denotes the current density) as a function of potential and fitting the line to obtain the Tafel slope. As illustrated in Fig. 3, the Tafel slopes of these systems, which reflect their reaction kinetics, indicate that Ni0.67Co0.67Fe0.67P exhibits the fastest kinetics (103 mV dec−1), followed by that of NiCoP (111 mV dec−1). On the other hand, both the monometallic Ni2P and bimetallic NiFeP had much higher slopes (147 and 173 mV dec−1, respectively) indicative of slower kinetics in these phases. These trends follow those observed for overpotential; however, NiFeP activity is an outlier when compared at higher current densities, as reflected in the large Tafel slope. This is evident when overpotential values are compared at higher current densities (40 mA cm−2 see Table 1) with NiFeP requiring the highest overpotential compared with Ni0.67Co0.67Fe0.67P, NiCoP and Ni2P. Notably, the latter three catalysts maintain a trend similar to that observed at 10 mA cm−2.

Fig. 3. Tafel plot for Ni0.67Co0.67Fe0.67P (red), bimetallic NiCoP (sky blue), bimetallic NiFeP (green), and monometallic Ni2P (purple). The plotted log J range corresponds to the interval between 9 and 20 mA cm−2.

Fig. 3

Open in a new tab

Stability of the Ni0.67Co0.67Fe0.67P-derived catalyst for OER

Aside from being an active OER catalyst, for practical viability, a catalyst should be stable under operating conditions. The Ni0.67Co0.67Fe0.67P-derived catalyst short-term stability was assessed by the following method: 140 µL of Ni0.67Co0.67Fe0.67P was dropcast on carbon cloth and utilized as the working electrode for a 15 h controlled potential electrolysis experiment (CPE) conducted at 1.7 V. As shown in Fig. 4, the initial current density value at 1.7 V was 31.50 mA cm−2, which decreased to 29.52 mA cm−2 over the 15 h study accounting for 6% reduction in current density. The changes in current density at fixed potential over time (Fig. 4) are also reflected in changes in the CV as a function of cycling (Fig. S2) with the 100th CV cycle showing a decrease in the Ni2+ oxidation potential relative to earlier cycles, while the reduction potential shifts first to lower potential (at 50 cycles), and then back to higher potential at 100 cycles. Faradaic efficiency (FE) measurement of our trimetallic system was evaluated for 1 h of bulk electrocatalysis in a closed system at 1.5 V vs. RHE, which gave 96.3% FE value (Section S1); potentials higher than 1.5 V could not be used as the pressure of rapid oxygen production pushed electrolyte from the working electrode side to the counter electrode side. The FE measurement was conducted once, for the purpose of confirming the main process is OER with minimal current loss associated with catalyst activation (i.e., FE > 95%).

Fig. 4. Controlled potential electrolysis of the trimetallic system for 15 h at 1.7 V vs. RHE.

Fig. 4

Open in a new tab

Composition and speciation changes for Ni0.67Co0.67Fe0.67P pre-vs. post-OER

Ni0.67Co0.67Fe0.67P nanoparticle–carbon composite, along with the pre-catalyst ink before and after catalysis, were characterized by X-ray photoelectron spectroscopy (XPS, Fig. 5) to elucidate speciation changes at each stage.

Fig. 5. XPS analysis of the trimetallic phosphide system at three stages for determining speciation changes: (a) nanoparticle–carbon composite, (b) precatalyst ink before catalysis, and (c) post catalysis.

Fig. 5

Open in a new tab

Beginning with the Ni 2p region, peak deconvolution was somewhat complicated, particularly for the Ni 2p1/2 signal of the pre-catalyst ink, due to the presence of a broad feature spanning 855.0–869.0 eV. This feature is commonly observed in Nafion-containing samples, as the FKL2 Auger peak of Nafion appears in this binding energy range. Nevertheless, analysis of the composite indicates that the initial Ni 2p3/2 peak position corresponds to metallic Ni, which was also observed in the pre-catalysis sample. After catalysis, the Ni 2p3/2 peak exhibited a slight shift to higher binding energy relative to the metallic state, suggesting oxidation of Ni is occurring. Metallic Co 2p3/2 peaks were observed with high intensity in both the composite and the pre-catalysis ink at binding energies of 778.4 eV and 778.6 eV, respectively. In contrast, following catalysis, the Co 2p3/2 peaks shifted toward higher binding energies, indicating oxidation of the metallic Co species. The Co 2p1/2 region also showed consistent trends. In the Fe 2p region, a relatively low-intensity Fe 2p3/2 metallic peak was observed, accompanied by Fe2+ features at higher binding energy, centered at approximately 711.8 eV. In the pre-catalysis ink, a very weak metallic Fe 2p3/2 peak appeared at a similar binding energy as in the composite; however, in the pre- and post-catalysis samples the higher oxidation state peak became significantly broadened, extending up to ∼730 eV. The P 2p3/2 spectra of the composite and pre-catalysis ink exhibited a peak at 129.8 eV, characteristic of metal phosphides. A second peak was observed at 133.7 eV for the composite and 133.9 eV for the pre-catalysis ink, corresponding to phosphate or phosphite (P–O or POx) species. Notably, the post-catalysis ink did not show detectable signals associated with either phosphide or phosphate species, indicative of loss of P during active catalysis.

The post catalysis ink on the carbon cloth from the 15 h controlled potential electrolysis experiment was sonicated in methanol to form a black slurry, from which a drop was used for STEM mapping analysis as shown in Fig. 6. Compared to as-prepared nanoparticles, significant Fe loss was evident, as the metal ratio after catalysis remained nearly equimolar in Ni and Co, but deficient in Fe—Ni : Co : Fe = 1.19 : 1.15 : 1 (Ni = 33.2 ± 1.2%, Co = 32.1 ± 1.6%, Fe = 27.7 ± 1.8, P = 6.9 ± 1.5). Moreover, the elemental composition analysis also demonstrated significant loss of P (M : P = 13.5 : 1), consistent with the loss of P signal in the XPS. Based on EDS comparison of the as-prepared Ni0.67Co0.67Fe0.67P to the spent ink, approximately 13 at% Fe and 81 at% P is lost in the 15 h CPE experiment. To corroborate the findings from EDS mapping, we performed ICP-MS on the electrolyte (1 M KOH) post-catalysis, comparing it to the blank KOH. As shown in Table 2. Ni and Co are not detected in the electrolyte post-catalysis, but the concentration of Fe increases relative to the baseline by ∼10 ppb (Fe is a known impurity in KOH) and significant concentrations of P (∼375 ppb) are detected. In contrast to the EDS mapping, calculations of Fe and P loss based on ICP-MS suggest much smaller losses during the 15 h CPE experiment, ∼0.7% Fe and 27% P (see Section S3 for calculations). The discrepancy is attributed to losses occurring during introduction of the nanoparticles to the carbon support, which is done in solution before making the ink (see Experimental section). This is consistent with prior work on NiMnP that showed a change in the Ni : Mn : P ratio between as-prepared nanoparticles (1.0 : 1.0 : 1.07) and the ink formulation used in testing (1.0 : 0.85 : 0.72).

Fig. 6. STEM/EDS mapping of the post-catalysis sample collected from the carbon cloth via sonication in methanol.

Fig. 6

Open in a new tab

Table 2. ICP-MS analysis of the trimetallic precatalyst sample: KOH (blank) vs. after catalysis (15 h of controlled potential electrolysis at 1.7 V vs. RHE). BDL = below detection limit.

Sample P (ppb) Fe (ppb) Co (ppb) Ni (ppb)
KOH (blank) BDL 43.5 BDL BDL
KOH (post catalysis) 375.4 54.1 BDL BDL
Open in a new tab

Comparing Ni0.67Co0.67Fe0.67P with Nirich (Ni1.0Co0.5Fe0.5P), Corich (Ni0.5Co1.0Fe0.5P) and Ferich (Ni0.5Co0.5Fe1.0P) trimetallic phosphides

To further investigate how modulating the metal content influences OER activity we introduced three additional phases having high fractions of one of the three metals being considered: Nirich (Ni1.0Co0.5Fe0.5P), Corich (Ni0.5Co1.0Fe0.5P) and Ferich (Ni0.5Co0.5Fe1.0P) trimetallic phosphides. PXRD data and TEM images with corresponding size distribution histograms are shown in Fig. S3 and S4, respectively. The actual compositions (from EDS analysis) are provided in Section S4. An interesting feature in the PXRD patterns (Fig. S3a) is observed near 2θ = 54.3°, where the (300), (002), and (211) reflections overlap into a single peak in the Ni-rich system; a similar behavior is also seen for Ni0.67Co0.67Fe0.67P. In contrast, the Corich and Ferich systems exhibit a clear splitting of this peak into two reflections. This splitting is usually observed in Fe2P, in which the (002) plane appears at a lower angle than the (300) plane. Notably, in the Corich and Ferich samples, the (002) reflection is shifted toward lower angles, suggesting an increase in d-spacing and corresponding changes in lattice parameters, likely due to the larger atomic radii of Co and Fe compared to Ni.

The metal rich systems were evaluated for their OER catalytic activity and compared to the equimolar, Ni0.67Co0.67Fe0.67P trimetallic system as shown in Fig. 7 from two duplicate trials. Here, the Nirich system showed superior performance in terms of OER activity compared to the Ni0.67Co0.67Fe0.67P nanocrystals, whereas both the Corich and Ferich nanocrystals showed relatively lower activity i.e., required higher overpotential to reach 10 mA cmgeo−2 current density.

Fig. 7. (a) OER polarization curve for the trimetallic Ni0.67Co0.67Fe0.67P in red, Nirich (Ni1.0Co0.5Fe0.5P) in lavender, Corich (Ni0.5Co1.0Fe0.5P) in turquoise, and Ferich (Ni0.5Co0.5Fe1.0P) in rust; (b) overpotential at 10 mA cm−2 current density for the four systems in Fig. 6a.

Fig. 7

Open in a new tab

To better understand the activation process, the first CV cycle was plotted and compared to the 10th cycle (without iR compensation), corresponding to 20 segments (Fig. 8 and S2). Two common trends were observed across all plots, regardless of the system. Compared with the first CV cycle, the 10th cycle showed: (1) a shift of the cathodic Ni3+/Ni2+ wave toward lower potential, and (2) an increase in the area under the cathodic wave corresponding to the reduction of Ni3+ to Ni2+.24 For the case of the Nirich and Ferich samples, the anodic Ni2+/Ni3+ wave also shifted to lower potential at the 10th cycle vs. the 1st.

Fig. 8. Cyclic voltammetry of Ni0.67Co0.67Fe0.67P in red, Nirich (Ni1.0Co0.5Fe0.5P) in lavender, Corich (Ni0.5Co1.0Fe0.5P) in turquoise, and Ferich (Ni0.5Co0.5Fe1.0P) in rust. CV 1 in solid line, CV 10 in dotted line.

Fig. 8

Open in a new tab

Discussion

Multimetallic synergy leads to lower overpotentials

In general, the addition of multiple metals to form M2P nanoparticles is established to decrease the overpotential for the resultant in situ derived OER catalysts. In our prior work, we prepared a series of nanoparticles of 2 : 1 M2P phases, Fe2−xMnxP (0 < x < 0.9),15 Co2−xMnxP (0 < x < 1.4),16 and Ni2−xMnxP (0 < x < 1.5).17 These we evaluated along with their monometallic counterparts Fe2P, Co2P and Ni2P as OER precatalysts (note that attempts to synthesize thermodynamically stable Mn2P25 phase by colloidal routes have not been successful to date). In all cases, while the monometallic phosphides were active in producing oxygen, the bimetallic Mn-containing phosphides were always more active in terms of the overpotential measured at a current density of 10 mA cm−2, i.e. η10bimetallic < η10monometallic. This trend is also reflected in two studies of high-entropy phosphides of first row transition metals vs. relevant multimetallic phases, where η10 followed the trend NiCoP > (NiCoFe)2P > (NiCoFeMn)2P > Ni0.4Co0.4Fe0.4Mn0.4Cr0.4P,20 in a 2021 study, and NiCoP > Ni0.67Co0.67Cu0.67P > Ni0.67Co0.67Fe0.67P > Ni0.56Co0.56Fe0.44Cu0.44P,21 in a 2025 study. Each of these high-entropy phosphide studies produced a trimetallic Ni–Co–Fe–P phase with Ni2P-type structure similar to the materials we report here. However, in one case the composition is not reported (we describe this phase as (NiCoFe)2P) and extensive testing, other than reporting overpotential, was not performed (i.e., no post-catalysis testing or compositional modifications within the trimetallic phase). However, the limited data provided in these references are comparable to what we report here. Thus, the OER overpotential of our NiCoP (η10 = 325.0 mV) and Ni0.67Co0.67Fe0.67P (η10 = 276.5 mV) phases closely matches those previously reported (NiCoP: η10 = 316.0,20 332.0 21 mV (NiCoFe)2P, Ni0.67Co0.67Fe0.67P: η10 = 283.0,20 275.0 21 mV, respectively); regardless of differences in particle size/morphology and testing characteristics. Collectively, these data suggest that the stoichiometry-driven trends are real and reproducible.

Activating the pre-catalysts

As with the bimetallic Mn-containing phosphide system, activation at positive potentials led to significant solubilization of phosphorus (presumably as phosphate), while XPS showed that the surface metal species were oxidized, suggesting the formation of oxyhydroxides, expected to constitute the active catalyst on the phosphide particle surface. The Ni-containing systems provide a convenient probe of this reconstruction through redox waves corresponding to Ni2+ oxidation to Ni3+ and subsequent reduction to Ni2+ within the potential window. By following changes in these nickel redox features upon cycling, insight can therefore be gained into how the phosphide precatalyst evolves into the catalytically active phase, at least with respect to the Ni component.

The behavior of these phosphide OER precatalysts under anodic polarization can be rationalized by comparison with the corresponding metal oxide/(oxy)hydroxide systems, because transition-metal phosphides undergo substantial surface reconstruction under alkaline OER conditions to generate oxidized, (oxy)hydroxide-like surface species. In such systems, Fe and Co have distinct effects on the Ni2+/Ni3+ redox transition. Fe incorporation generally shifts the oxidation of Ni2+ to Ni3+ to more positive potentials, consistent with stabilization of Ni2+ through changes in the local electronic environment and charge–transfer interactions. Importantly, this does not imply lower OER activity. Ni–Fe oxyhydroxides can exhibit substantially greater activity than pure Ni oxyhydroxides because catalytic performance is governed not only by the accessibility of Ni3+ but also by the binding energies of OER intermediates. While Ni and Fe oxyhydroxides alone bind relevant intermediates too weakly and too strongly, respectively, Fe sites embedded within a Ni-rich environment can provide more favorable adsorption energetics.26,27

Co incorporation, in contrast, generally shifts the Ni redox transition cathodically, facilitating oxidation of Ni2+ to Ni3+ at lower applied potentials. Consistent with this behavior, NiCoP in the present study exhibits a cathodic shift of the Ni2+/Ni3+ redox couple relative to Ni2P. When Fe and Co are incorporated simultaneously, the redox response reflects the competing electronic influences of the two metals, with Fe shifting the couple back toward more positive potentials relative to the Co-containing system. These effects are consistent with electronic hybridization and charge transfer among Ni, Co, and Fe species.28 In particular, Co may facilitate formation of the conductive Ni(iii)OOH phase at lower applied potentials, while Fe can modify the electronic structure and intermediate-binding properties of the reconstructed oxyhydroxide surface. Similar behavior has been reported for Ni–Fe–Co mixed-metal oxide systems.29

As shown in Fig. 8, cycling of the trimetallic systems also produces an increase in the cathodic current associated with reduction of high-valent Ni species. Together with the changes in redox potential, this suggests that as the catalyst evolves from a metallic phosphide to a metal oxyhydroxide supported on residual metal phosphide, the local environment of Ni becomes increasingly favorable for reversible Ni2+/Ni3+ redox chemistry and catalytic turnover. In the case of the most active catalyst (Ni-rich), the oxidation potential also shifts to lower values upon cycling, suggesting that formation of the catalytically relevant high-valent Ni species becomes increasingly facile. Thus, the observed evolution of the Ni2+/Ni3+ redox response can be attributed to the distinct but complementary electronic interactions introduced by Fe and Co incorporation. These interactions modify the local electronic environment of Ni while also tuning the properties of the reconstructed oxyhydroxide surface, providing a plausible basis for the enhanced OER activity relative to Ni2P.

Stoichiometry: finding the “sweet spot” and comparison with current literature

With respect to stoichiometry, among the three Mn-based systems, the lowest overpotentials were obtained for 1 : 1 M : Mn ratio (or, in the case of Fe, as near to 1 : 1 as synthetically possible) and the overpotential, η10, followed the trend: Fe1.1Mn0.9P > CoMnP > NiMnP. Thus, Ni-based phases were the most active, attributed in part to the larger electronegativity difference between Ni and Mn. However, the 1 : 1 M : Mn “sweet spot” appears to be unique to the Mn-containing bimetallic systems, as, for example, the most active phase in the (Co,Fe)2P precatalyst system is Co1.33Fe0.67P.19 Moreover, in the present system, it is not the equimolar Ni0.67Co0.67Fe0.67P that is the most active, but rather the Nirich phase, Ni1.0Co0.5Fe0.5P. Specifically, the η10 value trend revealed Nirich (257.0 ± 5.0 mV) < Ni0.67Co0.67Fe0.67P (276.5 ± 3.7 mV) < Corich (295.5 ± 12.5 mV) < Ferich (308.7 ± 5.7 mV). This is perhaps not unexpected, given the fact that the monometallic Ni2P is so much more active than Co2P or Fe2P, and we can consider the Ni–Co–Fe phosphides to be substituted versions of Ni2P, where Co and Fe work synergistically to lower the overpotential. Based on studies of Fe-containing layered metal oxyhydroxide catalysts such as NiFeOOH30,31 and CoFeOOH,32 the trimetallic system containing iron along with nickel and cobalt likely benefits from access to Fe(iv), which is expected to form during OER polarization. Fe(iv) is known to stabilize the O radical, which facilitates O–O coupling and eventual release of O2.32,33 However, given that FeOOH is intrinsically insulating (conductivity 2.41 × 10−10 S cm−1),34 the incorporation of Fe is anticipated to impair the electronic conductivity of the system. On the other hand, NiOOH and CoOOH have much higher intrinsic conductivities (10−2 to 101 S cm−1),35–38 and can therefore compensate for the low conductivity of FeOOH, leading to optimal activity when combined.

A broader comparison with recently reported multimetallic phosphide OER catalysts further places the activity of the present Ni-rich composition in context. Although Ni1.0Co0.5Fe0.5P reaches 10 mA cm−2 at an overpotential of 257.0 ± 5.0 mV, several trimetallic Ni–Co–Fe phosphides have been reported with comparable or lower η10 values. For example, Zhao et al.39 reported a NiFeCo-P@CAP/NF trimetallic phosphide (Ni : Fe : Co = 5 : 1 : 1; CAP = crystalline-amorphous phase) requiring only 202 mV to achieve 10 mA cm−2 in 1.0 M KOH. However, it was not entirely clear whether the heterophase NiFeCo-P@CAP/NF structure identified by XRD was the same as the phosphide with Ni : Fe : Co = 5 : 1 : 1 ratio, and other molar ratios within the trimetallic system were not evaluated for OER, limiting direct composition–activity comparisons. Porous NiCoFePx nanosheets, with relatively similar atomic percentages of Ni and Co (P = 31.6%, Ni = 23.9%, Co = 17.9%, O = 14.2%, C = 7.8%, and Fe = 4.5%) and synthesized through three major stages (NiCoFeLDH → NiCoFeOx → NiCoFePx), were reported to exhibit an η10 of approximately 259 mV,40 which is essentially comparable to the 257.0 ± 5.0 mV obtained in the present work. Thus, the present precatalysts should be considered competitive rather than record-performing on the basis of η10 alone. Importantly, direct comparison of literature overpotentials should be made cautiously because catalyst loading, substrate, morphology, electrolyte composition, etc. differ among studies. One important aspect that distinguishes our work from the broader pool of trimetallic phosphide-based studies is the thorough post-catalytic characterization of the catalyst system. This enables the observed structural and compositional changes to be directly associated with the evolution of the precatalyst during OER. Thus, the significance of the present system lies in the systematic control of Ni/Co/Fe stoichiometry within a common Ni2P-type structural framework, which allows the compositional dependence of OER activity to be directly assessed and identifies the Nirich composition as the optimum within this trimetallic series.

Short term operational stability and metal dissolution

For a catalyst to be relevant to practical applications, operational stability is a necessary criterion. In such cases, Mn-containing bimetallic systems usually suffer from instability issues (decrease in the current density at constant potential) and this correlates to Mn leaching into the electrolyte. The solubility of Mn is well known; for the model system LiMn2O4, Mn dissolution occurs even in the pre-OER region, at about 1.4 V vs. RHE, whereas additional dissolution also took place around 1.6 V vs. RHE.41 As for the NiMnP system, as the number of CV cycles increased to 100, and eventually to 200, the Ni2+/Ni3+ peak intensity decreases and shifts to higher potentials, which suggests that Ni2+ to Ni3+ conversion is becoming more hindered over time, a process that is critical for O2 evolution in Ni-based catalysts.17 The hindrance is attributed to the loss of Mn, and consequent reduction of charge transfer from Mn to Ni. In contrast, for Ni0.67Co0.67Fe0.67P, we see no loss of peak intensity (Fig. S2) and minimal shifting of the Ni2+/Ni3+ peak. Moreover, when comparing performance via constant potential electrolysis over 15 h, we find that Ni0.67Co0.67Fe0.67P can reach higher current densities at lower applied potentials relative to NiMnP, consistent with the lower Tafel slope reported for Ni0.67Co0.67Fe0.67P (103 mV dec−1) vs. NiMnP (186 mV dec−1); Ni0.67Co0.67Fe0.67P also demonstrates modest activity losses (6–10%) relative to NiMnP (20%), as shown in Table 3. An important point to note is that, although our trimetallic precatalyst demonstrated good short-term operational stability under the investigated OER conditions, extended durability testing, complemented by post-electrolysis structural and compositional analyses, is necessary to comprehensively assess its long-term practical stability.

Table 3. Comparison between bimetallic NiMnP and trimetallic Ni0.67Co0.67Fe0.67P for constant potential electrolysis i.e., bulk electrolysis of 15 hours.

Pre-catalyst Applied potential (V vs. RHE) Initial current density (mA cm−2) Final current density (mA cm−2) % Current density retained
NiMnP 2.10 20.0 16.0 80.0
Ni0.67Co0.67Fe0.67P 1.70 31.5 29.5 93.7
″ 1.60 14.3 12.9 90.2
″ 1.50 3.40 3.20 94.1
Open in a new tab

However, the trimetallic phosphide is also subject to metal leaching—in this case, Fe. Based on ICP-MS of the electrolyte, we find that trace amounts of Fe (∼0.7%) are released over a 15 h CPE test. A recent study on Co1−xFexOOH, where 0 ≤ x ≤ 1 by Burke et al.32 revealed that the stability of that system was disrupted for compositions 0.5 < x < 1, with 18–38 mass% of the catalysts lost during the 2 h long polarization process conducted at a constant potential of 1.58 V vs. RHE, whereas for x = 1 i.e., FeOOH, 44 mass% was lost, reflecting the solubility of FeOOH. However, when 0 < x < 0.5, no mass loss was detected.32 Similarly, in our M2P type system, Ni0.67Co0.67Fe0.67P, Fe constitutes approximately 33.3% of the total metal content and is more susceptible to dissolution under alkaline conditions compared to the relatively stable Ni and Co, which together account for 66.7% of metal content. The presence of Ni and Co, like in the CoFeOOH system,32 or in NiFe oxyhydroxide electrodeposited films42—likely helps stabilize the smaller Fe fraction, thereby imparting improved long-term stability to the trimetallic catalyst.

Conclusions

This study demonstrates that the OER performance of transition-metal phosphide precatalysts comprising Ni, Co, and Fe is strongly governed by multimetallic synergy and metal stoichiometry wherein incorporation of multiple metals consistently lowers the overpotential relative to monometallic and bimetallic phosphides. Moreover, by comparing select phases to previous literature reports of similar compositions, we find that the observed activity enhancement is reproducible across different particle sizes, morphologies, and testing conditions, in contrast to our original hypothesis. Within the Ni–Co–Fe phosphide series, unlike the Mn-based phosphide systems, maximum activity was not achieved at the equimolar composition. Instead, the Ni-rich phase, Ni1.0Co0.5Fe0.5P, exhibited the lowest overpotential, with η10 = 257.0 ± 5 mV. The superior performance of the Ni-rich trimetallic system can be attributed to the complementary functions of Ni, Co, and Fe. Indeed, electrochemical activation of the Ni–Co–Fe phosphides caused favorable shifts in the nickel redox potential and increased cathodic currents indicating the formation and regeneration of catalytically relevant high-valent metal species, at least for Ni. Excessive Fe incorporation may, nevertheless, reduce conductivity and increase susceptibility to dissolution, explaining why the equimolar and Fe-rich compositions are less active than the Ni-rich phase. Finally, the Ni–Co–Fe system also offers substantially improved durability compared with Mn-containing bimetallic phosphides, and much lower Tafel slopes, enabling higher current densities to be achieved for modest changes in potential. These findings demonstrate that increasing compositional complexity alone does not guarantee optimal performance in transition metal phosphide OER pre-catalysts; rather, activity depends on achieving a favorable balance among the constituent metals.

Author contributions

F. S. conducted the experimental investigation, analyzed and validated the data, wrote the original manuscript draft, and reviewed and edited the manuscript. S. L. B. conceived the project, acquired the funding, supervised the research, and reviewed and edited the manuscript. All authors have given approval to the final version of the manuscript.

Conflicts of interest

The authors declare no competing financial interest.

Supplementary Material

RA-OLF-D6RA06796D-s001

Acknowledgments

We thank the National Science Foundation (NSF) for support (DMR-1904775). This work used PXRD, TEM, UPS/XPS, and ICP–MS from the Lumigen Instrument Center, Department of Chemistry, Wayne State University, purchased with assistance from NSF grant numbers 1427926, 2018587, and 1849578. FS is grateful to Wayne State University for partial support via a Rumble Fellowship.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). These include a Table of chemicals used in the synthesis; Figures of electrochemical cycles, powder X-ray diffraction plots, electron microscopy images, electrochemically active surface area (ECSA) plots, OER polarization curves normalized to ECSA; descriptions of faradaic efficiency calculations, synthesis of monometallic and bimetallic phases for comparison, calculations of Fe and P loss from ICP-MS data, and compositional data for different metal-rich systems. Supplementary information is available. See DOI: https://doi.org/10.1039/d6ra06796d.

References

  1. Stinner C. Tang Z. Haouas M. Weber T. Prins R. Preparation and 31P NMR characterization of nickel phosphides on silica. J. Catal. 2002;208:456–466. doi: 10.1006/jcat.2002.3577. [DOI] [Google Scholar]
  2. Rodriguez J. A. Kim J.-Y. Hanson J. C. Sawhill S. J. Bussell M. E. Physical and chemical properties of MoP, Ni2P, and MoNiP hydrodesulfurization catalysts: time-resolved X-ray diffraction, density functional, and hydrodesulfurization activity studies. J. Phys. Chem. B. 2003;107:6276–6285. doi: 10.1021/jp022639q. [DOI] [Google Scholar]
  3. Deng C. et al., Templated-preparation of a three-dimensional molybdenum phosphide sponge as a high performance electrode for hydrogen evolution. J. Mater. Chem. A. 2016;4:59–66. doi: 10.1039/C5TA05453B. [DOI] [Google Scholar]
  4. Wang R. Smith K. J. Hydrodesulfurization of 4, 6-dimethyldibenzothiophene over high surface area metal phosphides. Appl. Catal., A. 2009;361:18–25. doi: 10.1016/j.apcata.2009.03.037. [DOI] [Google Scholar]
  5. Yang S. Liang C. Prins R. Preparation and hydrotreating activity of unsupported nickel phosphide with high surface area. J. Catal. 2006;241:465–469. doi: 10.1016/j.jcat.2006.05.014. [DOI] [Google Scholar]
  6. Cheng R. et al., Synthesis and characterization of high surface area molybdenum phosphide. Appl. Catal., A. 2007;316:160–168. doi: 10.1016/j.apcata.2006.08.036. [DOI] [Google Scholar]
  7. Jarvis R. F. Jacubinas R. M. Kaner R. B. Self-propagating metathesis routes to metastable group 4 phosphides. Inorg. Chem. 2000;39:3243–3246. doi: 10.1021/ic000057m. [DOI] [PubMed] [Google Scholar]
  8. Sun H. et al., Self-supported transition-metal-based electrocatalysts for hydrogen and oxygen evolution. Adv. Mater. 2020;32:1806326. doi: 10.1002/adma.201806326. [DOI] [PubMed] [Google Scholar]
  9. Tian J. Liu Q. Asiri A. M. Sun X. Self-supported nanoporous cobalt phosphide nanowire arrays: an efficient 3D hydrogen-evolving cathode over the wide range of pH 0–14. J. Am. Chem. Soc. 2014;136:7587–7590. doi: 10.1021/ja503372r. [DOI] [PubMed] [Google Scholar]
  10. Jiang P. et al., A cost-effective 3D hydrogen evolution cathode with high catalytic activity: FeP nanowire array as the active phase. Angew. Chem. 2014;126:13069–13073. doi: 10.1002/ange.201406848. [DOI] [PubMed] [Google Scholar]
  11. Li Y. et al., 3D self-supported Fe-doped Ni2P nanosheet arrays as bifunctional catalysts for overall water splitting. Adv. Funct. Mater. 2017;27:1702513. doi: 10.1002/adfm.201702513. [DOI] [Google Scholar]
  12. Tian J. Liu Q. Cheng N. Asiri A. M. Sun X. Self-supported Cu3P nanowire arrays as an integrated high-performance three-dimensional cathode for generating hydrogen from water. Angew. Chem. 2014;126:9731–9735. doi: 10.1002/ange.201403842. [DOI] [PubMed] [Google Scholar]
  13. Görlin M. et al., Oxygen evolution reaction dynamics, faradaic charge efficiency, and the active metal redox states of Ni–Fe oxide water splitting electrocatalysts. J. Am. Chem. Soc. 2016;138:5603–5614. doi: 10.1021/jacs.6b00332. [DOI] [PubMed] [Google Scholar]
  14. Zhong H. et al., Fundamental understanding of structural reconstruction behaviors in oxygen evolution reaction electrocatalysts. Adv. Energy Mater. 2023;13:2301391. doi: 10.1002/aenm.202301391. [DOI] [Google Scholar]
  15. Li D. Baydoun H. Kulikowski B. Brock S. L. Boosting the catalytic performance of iron phosphide nanorods for the oxygen evolution reaction by incorporation of manganese. Chem. Mater. 2017;29:3048–3054. doi: 10.1021/acs.chemmater.7b00055. [DOI] [Google Scholar]
  16. Li D. Baydoun H. Verani C. N. Brock S. L. Efficient water oxidation using CoMnP nanoparticles. J. Am. Chem. Soc. 2016;138:4006–4009. doi: 10.1021/jacs.6b01543. [DOI] [PubMed] [Google Scholar]
  17. Aghabozorgi F. Perera S. S. Brock S. L. Ni2–xMnxP Nanoparticles as Earth-Abundant Precatalysts for Electrochemical Water Oxidation. Chem. Mater. 2024;36:2461–2472. doi: 10.1021/acs.chemmater.3c03215. [DOI] [Google Scholar]
  18. Su’a T. Poli M. N. Brock S. L. Homogeneous nanoparticles of multimetallic phosphides via precursor tuning: Ternary and quaternary M2P phases (M= Fe, Co, Ni) ACS Nanosci. Au. 2022;2:503–519. doi: 10.1021/acsnanoscienceau.2c00025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Zhang Y. et al., Programmable synthesis of multimetallic phosphide nanorods mediated by core/shell structure formation and conversion. J. Am. Chem. Soc. 2020;142:8490–8497. doi: 10.1021/jacs.0c02584. [DOI] [PubMed] [Google Scholar]
  20. Lai D. Kang Q. Gao F. Lu Q. High-entropy effect of a metal phosphide on enhanced overall water splitting performance. J. Mater. Chem. A. 2021;9:17913–17922. doi: 10.1039/D1TA04755H. [DOI] [Google Scholar]
  21. Jeong H. et al., High-entropy CoNiFeCuP nanoparticles as efficient and durable catalysts for alkaline oxygen evolution reaction. J. Mater. Chem. A. 2025;13:16995–17005. doi: 10.1039/D5TA01149C. [DOI] [Google Scholar]
  22. Muthuswamy E. Savithra G. H. L. Brock S. L. Synthetic levers enabling independent control of phase, size, and morphology in nickel phosphide nanoparticles. ACS Nano. 2011;5:2402–2411. doi: 10.1021/nn1033357. [DOI] [PubMed] [Google Scholar]
  23. Layan Savithra G. H. et al., Rational design of nickel phosphide hydrodesulfurization catalysts: controlling particle size and preventing sintering. Chem. Mater. 2013;25:825–833. doi: 10.1021/cm302680j. [DOI] [Google Scholar]
  24. Mattinen M. et al., Dynamics of precatalyst conversion and iron incorporation in nickel-based alkaline oxygen evolution reaction catalysts. Cell Rep. Phys. Sci. 2024;5:102284. doi: 10.1016/j.xcrp.2024.102284. [DOI] [Google Scholar]
  25. Data retrieved from the Materials Project for Mn2P (mp-1849) from database version v2025.09.25, 2020, 10.17188/1193131 [DOI]
  26. Louie M. W. Bell A. T. An investigation of thin-film Ni–Fe oxide catalysts for the electrochemical evolution of oxygen. J. Am. Chem. Soc. 2013;135:12329–12337. doi: 10.1021/ja405351s. [DOI] [PubMed] [Google Scholar]
  27. Swierk J. R. Klaus S. Trotochaud L. Bell A. T. Tilley T. D. Electrochemical study of the energetics of the oxygen evolution reaction at nickel iron (oxy) hydroxide catalysts. J. Phys. Chem. C. 2015;119:19022–19029. doi: 10.1021/acs.jpcc.5b05861. [DOI] [Google Scholar]
  28. Stevens M. B. et al., Ternary Ni-Co-Fe oxyhydroxide oxygen evolution catalysts: Intrinsic activity trends, electrical conductivity, and electronic band structure. Nano Res. 2019;12:2288–2295. doi: 10.1007/s12274-019-2391-y. [DOI] [Google Scholar]
  29. Bates M. K. Jia Q. Doan H. Liang W. Mukerjee S. Charge-transfer effects in Ni–Fe and Ni–Fe–Co mixed-metal oxides for the alkaline oxygen evolution reaction. ACS Catal. 2016;6:155–161. doi: 10.1021/acscatal.5b01481. [DOI] [Google Scholar]
  30. Smith A. M. Trotochaud L. Burke M. S. Boettcher S. W. Contributions to activity enhancement via Fe incorporation in Ni-(oxy) hydroxide/borate catalysts for near-neutral pH oxygen evolution. Chem. Commun. 2015;51:5261–5263. doi: 10.1039/C4CC08670H. [DOI] [PubMed] [Google Scholar]
  31. Xiao H. Shin H. Goddard III W. A. Synergy between Fe and Ni in the optimal performance of (Ni, Fe)OOH catalysts for the oxygen evolution reaction. Proc. Natl. Acad. Sci. U. S. A. 2018;115:5872–5877. doi: 10.1073/pnas.1722034115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Burke M. S. Kast M. G. Trotochaud L. Smith A. M. Boettcher S. W. Cobalt–iron (oxy) hydroxide oxygen evolution electrocatalysts: the role of structure and composition on activity, stability, and mechanism. J. Am. Chem. Soc. 2015;137:3638–3648. doi: 10.1021/jacs.5b00281. [DOI] [PubMed] [Google Scholar]
  33. Leonov I. Greenberg E. Belov M. P. Rozenberg G. K. Abrikosov I. A. Mott transition and magnetic collapse in iron-bearing compounds under high pressure. High Pressure Res. 2017;37:96–118. doi: 10.1080/08957959.2017.1302445. [DOI] [Google Scholar]
  34. Mokhnatska L. V. et al., Ultrafine beta-FeOOH and Fe3O4 obtained by precipitation method: comparative study of electrical and electrochemical properties. Phys. Chem. Solid State. 2020;21:680–688. doi: 10.15330/pcss.21.4.680-688. [DOI] [Google Scholar]
  35. Zhou T. et al., Enhanced asymmetric supercapacitor performance of NiOOH electrodes from hexagonal Ni(OH)2 nanosheets via low-concentration OH−-mediated growth and galvanostatic oxidation. J. Energy Storage. 2025;137:118558. doi: 10.1016/j.est.2025.118558. [DOI] [Google Scholar]
  36. He B. Bai F. Jain P. Li T. A review of surface reconstruction and transformation of 3d transition-metal (oxy) hydroxides and spinel-type oxides during the oxygen evolution reaction. Small. 2025;21:2411479. doi: 10.1002/smll.202411479. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Poulain R. et al., Electronic and chemical properties of nickel oxide thin films and the intrinsic defects compensation mechanism. ACS Appl. Electron. Mater. 2022;4:2718–2728. doi: 10.1021/acsaelm.2c00230. [DOI] [Google Scholar]
  38. Hu J. et al., Improved interface charge transfer and redistribution in CuO-CoOOH p-n heterojunction nanoarray electrocatalyst for enhanced oxygen evolution reaction. Adv. Sci. 2021;8:2103314. doi: 10.1002/advs.202103314. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Zhao S. et al., Crystalline and amorphous phases: NiFeCo tri-metal phosphide as an efficient electrocatalyst to accelerate oxygen evolution reaction kinetics. Electrochim. Acta. 2022;426:140788. doi: 10.1016/j.electacta.2022.140788. [DOI] [Google Scholar]
  40. Helal G. et al., Electrochemical water splitting enhancement by introducing mesoporous NiCoFe-trimetallic phosphide nanosheets as catalysts for the oxygen evolution reaction. RSC Adv. 2024;14:17202–17212. doi: 10.1039/D4RA02344G. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Bisen O. Y. Baumung M. Tatzel M. Volkert C. A. Risch M. Manganese dissolution in alkaline medium with and without concurrent oxygen evolution in LiMn2O4. Energy Adv. 2024;3:504–514. doi: 10.1039/D3YA00434A. [DOI] [Google Scholar]
  42. da Silva G. C. Igel C. S. Cherevko S. Setting the Record Straight: A Systematic Study on Fe and Ni Dissolution in NiFe-Based Oxygen Evolution Reaction Catalysts. ACS Catal. 2026;16:8633–8641. doi: 10.1021/acscatal.6c01228. [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

RA-OLF-D6RA06796D-s001
RA-OLF-D6RA06796D-s001.pdf (1.8MB, pdf)

Data Availability Statement

The data supporting this article have been included as part of the supplementary information (SI). These include a Table of chemicals used in the synthesis; Figures of electrochemical cycles, powder X-ray diffraction plots, electron microscopy images, electrochemically active surface area (ECSA) plots, OER polarization curves normalized to ECSA; descriptions of faradaic efficiency calculations, synthesis of monometallic and bimetallic phases for comparison, calculations of Fe and P loss from ICP-MS data, and compositional data for different metal-rich systems. Supplementary information is available. See DOI: https://doi.org/10.1039/d6ra06796d.

Articles from RSC Advances are provided here courtesy of Royal Society of Chemistry

RESOURCES