Water splitting is a sustainable source of hydrogen, a clean energy carrier for fuel cells that can be produced and used without generating carbon dioxide emissions. Accordingly, green hydrogen production through water splitting is considered one of the most ideal approaches to achieving carbon neutrality. However, green hydrogen has a significantly higher cost of production compared with generating gray hydrogen by fossil fuel reforming, and this remains one of the biggest obstacles to achieving a hydrogen-based society.

Considering this, the Department of Energy has set a target of reducing the cost of hydrogen to $1 per kilogram by 2030 [1], and researchers are investigating various strategies to reduce the cost of green hydrogen production. One effective method to lower the cost of hydrogen production is to utilize non-precious metal catalysts [2-7]. Among them, cobalt (Co) has been widely used in fields such as water electrolysis [4,8-11], lithium-ion batteries [12,13], biogas production[14], and carbon dioxide conversion technologies [15,16], owing to its exceptional catalytic properties [17,18]. However, the supply of Co is limited, and its demand has surged across various industries in recent decades, driving up its price significantly. As a result, Co-based catalyst strategies have become a less viable approach for reducing the cost of green hydrogen production [19]. Instead, these challenges are encouraging the active pursuit of materials that could significantly reduce Co usage, or be Co-free alternatives, for example in the battery field [20-23]. Like the ongoing efforts to reduce platinum (Pt) in water electrolysis, in the near future, development of Co-free non-precious metal catalysts or minimizing Co usage will be increasingly important.

In this study, we propose a strategy to significantly reduce the Co content in a catalyst-coated substrate (CCS) used as anodes for anion exchange membrane water electrolysis (AEMWE) systems. In this efficient Co reduction method, we added a Co solution to the slurry required to make a CCS electrode, then used the print transfer method with a prototype NiFe oxide. The developed electrode achieved an overpotential of 380 mV at a current density of 100 mA cm^-2, representing a 12.9% improvement compared with the overpotential (430 mV) of NiFe oxide electrodes formed at the same current density. Interestingly, our strategy to significantly reduce Co usage resulted in 72 times less Co consumption compared to NiFeCo catalysts synthesized through conventional hydrothermal synthesis. Moreover, it exhibited nearly identical oxygen evolution reaction (OER) performance, validating the high efficiency of this strategy. The electrode developed to reduce Co content in the CCS exhibited a significant performance increase compared with NiFe oxide electrodes at the same voltage. The developed CCS electrode manufacturing strategy has broad applicability for reducing the use of various expensive elements

To explore the effect of Co reduction, we prepared three types of electrodes, as shown in Fig 1. The first involved the synthesis of the NiFeCo powder via conventional calcination synthesis (Fig 1a). Then, the NiFeCo was mixed with PTFE as a binder for printing transfer manufacturing, and an appropriate amount of DI water was added to produce a slurry using a thinky mixer. The as-prepared slurry was uniformly spread onto a polyimide film using the CCS method, and Ni foam was then placed on top. The NiFeCo oxide was transferred onto the Ni foam using a hot-press.

The NiFe electrode was fabricated using the same process as that used for the NiFeCo electrode without the Co precursor. For the Co-reduced NiFeCo electrode (denoted as NiFe@Co), we replaced the DI water typically used in slurry preparation with a Co solution. The addition of a Co solution during slurry preparation ensured uniform coating of the Co solution on the surface of the NiFe oxide. All electrodes obtained through the fabrication process were of the same size, and it was verified that they had uniform surfaces (Fig 1c). During the hot-pressing process, the slurry with uniformly coated Co was transferred onto the Ni foam, forming a catalyst surface with NiFeCo, resulting in a core-shell configuration, such as NiFe@Co (Fig 1b).

Because the OER is a surface reaction, it was anticipated that by incorporating NiFeCo only on the catalyst surface, rather than throughout the entire surface and bulk, Co usage could be significantly reduced while obtaining an OER performance comparable to that of a conventional NiFeCo electrode. The results of this analysis are discussed in detail below.

To confirm the morphology and structure of the three catalysts, X-ray diffraction, SEM, and TEM measurements were conducted. The synthesized NiFe catalyst (Fig 2a) consisted of an inverse spinel structure (NiFe₂O₄, JCPDS 00-044-1485) and iron oxide (Fe₂O₃, JCPDS 033-0664) [19]. NiFeCo synthesized with Co exhibited the spinel structure of Co₃O₄ (JCPDS 042-1467) and crystalline structure of NiFe [24]. In contrast, NiFe@Co, which contained only trace amounts of Co, displayed a structure nearly identical to that of NiFe, confirming the absence of other structural byproducts. In the SEM and low-magnification TEM images, the three catalysts had nanoparticle morphologies with sizes in the range 17–30 nm, indicating that the Co solution used in preparing NiFe@Co did not significantly affect the formation of the catalysts’ structure (Fig 2b) [25, 26]. High-magnification TEM was used to measure the lattice spacing of the catalyst, confirming the presence of a 0.25-nm spacing, corresponding to the (311) plane of NiFe₂O₄ (Fig 2c) [27]. For NiFe@Co, a 2–3-nm amorphous Co layer was observed on the surface, while in the core, a NiFe lattice spacing of 0.25 nm corresponding to the (311) plane was visible. However, for NiFeCo, Co₃O₄ formed locally [28]. The energy-dispersive X-ray mapping of Ni, Fe, and Co by STEM analysis clearly showed a uniform distribution of NiFe@Co (Fig 2d). Therefore, as previously mentioned, it was confirmed that the addition of the Co solution during the electrode fabrication process did not cause significant structural or compositional changes.

XPS analysis was conducted to investigate the electronic interactions upon Co addition, and understand the fundamental evolution of the surface chemical states of NiFe@Co. In the Ni 2p spectrum (Fig 3a), a major signal appeared for Ni 2p_3/2 (854.7 eV), which corresponded to Ni²⁺ from nickel ferrite [29]. Comparison of the Ni 2p_3/2 spectra of NiFe, NiFeCo, and NiFe@Co revealed no significant changes, and the addition of Co did not have a significant impact on Ni. In the Fe 2p spectrum, the primary peak in Fe 2p_3/2 was attributed to Fe³⁺ from nickel ferrite [30]. Unlike the chemical state of Ni, the Fe 2p_3/2 in NiFeCo exhibited a noticeably higher binding energy (0.5 eV) than that of NiFe, indicating an increase in the Fe³⁺ content owing to the introduction of Co (Fig 3b). Interestingly, for NiFe@Co, the peaks appeared at the same positions as those of NiFe, indicating that the electronic structure was not altered by the addition of Co. However, the high-resolution spectrum of Co 2p showed a signal corresponding to the previous finding that a thin layer of Co was distributed across the catalyst surface [25]. In addition, the low signal was attributed to the trace amount of Co used in the electrode fabrication process (Fig 3c).

In the O 1s high-resolution spectra, NiFe and NiFeCo showed similar patterns, but NiFe@Co exhibited a slightly lower intensity. This was attributed to the amorphous layer of Co, which likely obscured part of the oxide signal [31]. Therefore, it was confirmed that the Co used in the electrode fabrication process was distributed across the NiFe surface (Fig 3d). The atomic percentages of Ni, Fe, Co, and O in each sample were determined using XPS. This indicated that each electrode was composed of similar amounts of Ni and Fe. In addition, NiFe@Co had a significantly lower Co content compared with that of NiFeCo (Fig 3e and f).

To confirm the OER activity of the extremely reduced Co content in the 1 M KOH electrolyte, we measured LSV polarization curves of NiFe, NiFeCo, and NiFe@Co with 85% iR-compensation (Fig 4a). NiFe@Co and NiFeCo exhibited a current density of approximately 250 mA cm^-2 at 1.7 VRHE, whereas NiFe showed approximately 150 mA cm^-2, indicating an improvement of approximately 67%. The NiFe electrode required an overpotential of 430 mV to achieve a current density of 100 mA cm^-2, while the NiFeCo and NiFe@Co electrodes exhibited lower overpotentials of 372 and 380 mV, respectively, at the same current density (Fig 4b). This indicates that the OER activity improved in the ternary system containing Co, which had more oxygen evolution activity sites than NiFe.

In addition, the change in charge owing to Co-incorporation was measured using cyclic voltammetry. NiFe@Co exhibited a charge approximately three times higher than that of NiFe (Fig 4c), which is thought to result from the amorphous Co layer enhancing the electrical conductivity of the electrode surface, thereby increasing the OER occurring at the surface [32-35].

Nyquist plots were used to investigate the influence on the overall electrochemical resistance, including the electrolyte/catalyst and catalyst/substrate interfaces [36] (Fig 4d). The charge transfer resistance between the electrolyte and catalyst was represented as the diameter of the semicircle in the low-frequency region. The smaller value clearly confirmed that the improved conductivity of the Co-containing catalyst was the reason for its increased activity.

However, NiFeCo and NiFe@Co showed similar activities, despite their different Co contents. Fig 4e shows the amount of Co precursor used to fabricate a single electrode (electrode size: 4.9 cm²) with the synthesized catalysts. NiFe@Co exhibited activity similar to that of NiFeCo, but the amount of precursor used was approximately 72 times less, demonstrating a significant reduction in the amount of Co required. To confirm the superior OER activity of NiFe@Co, we calculated the mass activity of the catalysts. The mass activity relative to the total weight of the catalyst was similar for NiFeCo (20 mg cm^-2) and NiFe@Co (20 mg cm^-2). However, when considering the weight of Co used in the catalyst, the mass activity of NiFe@Co was 183.93 mA g^-1_Co, which was approximately 67 times higher than that of NiFeCo (2.73 mA g^-1_Co), as shown in Fig 4f.

For application in water electrolysis systems, durability is as important as activity. The durability of the fabricated electrodes was evaluated at a sufficient current density, where 100 mA cm^-2 of oxygen evolution occurred actively. The NiFe@Co electrode maintained stable performance without significant voltage fluctuations, confirming its high stability. The NiFe catalyst retained its crystalline structure after the durability test with no significant changes in shape or size compared with its initial state (Fig 4g).

To evaluate the application of the Co-reduced electrode in a water electrolysis system, we constructed an AEMWE single cell, as shown in Fig 5a, and conducted experiments. NiFe@Co and NiFe were used as anode materials. From the polarization curves in Fig 5b, it can be seen that the activity significantly increased when the NiFe@Co electrode was used compared with when the NiFe electrode was used. This enhancement was attributed to the increased number of active sites provided by the Co layer, and improved conductivity due to the thin anhydrous oxide layer [37], confirming that these factors were also effective in the full cell.

To further verify this, we dropped water onto the NiFe and NiFe@Co electrodes and measured the contact angles. The NiFe electrode had a contact angle of 121.65°, whereas the NiFe@Co electrode exhibited a considerably lower contact angle of 74.34°, indicating higher hydrophilicity. This suggested that the water supply and oxygen gas desorption were more efficient, leading to higher activity. Since durability is as important as activity to enhanced cell performance, it was tested at a current density of 500 mA cm^-2, at which hydrogen generation was sustained [38]. Both cells retained their durability for 500 h without significant voltage changes. The cell with the NiFe@Co electrode had a more stable durability owing to its reduced overpotential (Fig 5c).

In AEMWE, the dry cathode method has the advantage of producing pure hydrogen gas[39]. To confirm the purity of the hydrogen produced in the AEMWE single cell, we conducted differential electrochemical mass spectrometry (Fig 5d) [40]. A consistent amount of hydrogen was generated each time when a current density of 500 mA cm^-2 was applied three times. In addition, the relative abundances of the gaseous ions generated from the cathode were analyzed using atomic mass spectrometry (Fig 5e). These results demonstrated that the Co-reduced NiFe@Co electrode prepared using a simple synthesis process could be applied not only in a three-electrode system but also in a full cell. This method also enabled performance enhancement, thereby significantly simplifying the fabrication process.

In this study, we developed a method to significantly reduce Co usage in a CCS process for AEMWE electrodes. This approach addresses the growing concerns over the decreasing availability and rising costs of Co. By incorporating a Co precursor in the water used in the electrode fabrication process, we achieved a 72-fold reduction in Co usage compared with those of electrodes with similar activities. The electrode’s performance was validated not only at the catalyst/electrode level but also in a single-cell system designed to simulate real-world conditions. Over the course of 500 h, hydrogen was consistently produced without significant voltage fluctuations. The generated hydrogen was confirmed to be pure, with no byproducts detected. We thus present a fabrication process that enables the production of electrodes with both significantly reduced Co content and high performance, offering a promising approach for sustainable and cost-effective catalyst development.