Nd doping enables efficient and durable manganese-based water oxidation catalysts under acidic conditions

Abstract

Mn oxides are attractive non-noble-metal catalysts for the oxygen evolution reaction in proton exchange membrane water electrolyzers, but their performance is limited by unstable Jahn–Teller-active Mn³⁺ centers and a narrow potential window. Here, we show that Nd doping stabilizes Jahn–Teller-active Mn³⁺ centers within a mixed-phase Mn oxide catalyst. The catalyst requires an overpotential of 404 mV to reach 100 mA cm^–2 and remains stable for more than 1000 h at 200 mA cm^–2. A proton exchange membrane water electrolyzer employing this catalyst operates stably for more than 600 h at 100 mA cm^–2. In situ Raman and X-ray spectroscopy, together with isotope-labeling experiments, support an oxide path mechanism and suppressed Mn overoxidation and dissolution. Phase-enriched reference experiments and density functional theory calculations show that Nd-doped α-Mn₂O₃ favors the oxide path mechanism through structural Mn³⁺ motifs, whereas β/R-MnO₂-rich domains provide structural robustness. Nd-induced 4f–2p–3 d orbital coupling enhances Mn–O covalency and stabilizes Mn³⁺ centers, thereby improving the activity–stability balance of Mn-based acidic oxygen evolution catalysts.

Developing durable non-noble-metal catalysts for acidic water oxidation remains challenging. Here, the authors show that neodymium doping stabilizes active Mn³⁺ species, enabling manganese oxide to operate for over 1000 h at 200 mA cm⁻².

Introduction

Proton exchange membrane water electrolyzers (PEMWEs) represent a promising technology for producing green hydrogen^1–3. Compared with alkaline and anion exchange membrane water electrolyzers, PEMWEs offer several key advantages, including high current density (>2 A cm^–2), production of high-purity hydrogen (99.9999%)^4–7, and rapid response to variable renewable electricity (<10 s)⁸. For practical applications, however, PEMWEs face significant challenges, particularly at the anode exposed to highly acidic and oxidative conditions⁹. These harsh environments accelerate catalyst degradation and limit long-term durability. Despite extensive studies on efficient and stable acidic oxygen evolution reaction (OER) electrocatalysts, IrO_x—an extremely scarce material¹⁰—remains the benchmark OER catalyst in PEMWEs¹¹. Recent analyses suggest that the global Ir supply (~7–8 tons per year) can support only tens of GWs of annual PEMWE installation capacity unless Ir loading is drastically reduced or recycling efficiency is substantially improved¹². Although various transition metal-based acidic OER catalysts, such as Co- and Ni-based systems, have been investigated, their stability remains insufficient under PEMWE operating conditions^13,14.

In this context, Mn has garnered significant interest as a promising active element for acidic OER catalysts due to its low cost and abundance (Supplementary Fig. 1). Recent studies have shown that structurally optimized γ-MnO₂, including M94%-type γ-MnO₂ and Nd-modified γ-MnO₂, can exhibit high durability under acidic OER conditions^15,16. However, achieving both high activity and high-current durability remains challenging for Mn-based acidic OER catalysts. While MnO₂ can operate stably within a narrow potential window¹⁷, highly active Mn³⁺ species are difficult to stabilize under acidic conditions. Mn³⁺, with a t_2g³ e_g¹ electronic configuration, can induce Jahn–Teller distortion and provide flexible Mn–O coordination environments that facilitate OER intermediate formation and conversion^18–20. However, Jahn–Teller-active Mn³⁺ species can readily disproportionate into Mn²⁺ and Mn⁴⁺ in acid²¹. Therefore, stabilizing Jahn–Teller-active Mn³⁺ motifs while maintaining the structural robustness of Mn oxides is a promising strategy to bridge the activity–stability gap of Mn-based OER catalysts under high-current-density acidic conditions. To date, however, very few Mn-based acidic OER catalysts have simultaneously achieved high activity together with long-term durability at current densities exceeding 100 mA cm^–2.

In this study, we report a Nd-doped Mn oxide (NMO) as an efficient and durable acidic OER electrocatalyst for PEMWEs. Nd doping transforms γ-MnO₂—originally composed of β and R phases—into a mixed ternary oxide comprising α-Mn₂O₃, β-MnO₂, and R-MnO₂, accompanied by stabilization of Jahn–Teller distorted Mn³⁺ and shortened Mn–Mn distances. Compared with γ-MnO₂, the optimized NMO exhibits substantially enhanced OER activity and stability in 0.5 M H₂SO₄, along with stable operation for >1000 h at 200 mA cm^–2. NMO can also be readily integrated into a membrane electrode assembly (MEA) and operates stably for >300 h at 200 mA cm^–2 and >60 h at 500 mA cm^–2 in a PEMWE. In situ Raman spectroscopy and isotope-labeling differential electrochemical mass spectrometry (DEMS) reveal early formation of labile ^*O species and support an OPM-like pathway during acidic OER. In situ X-ray absorption spectroscopy (XAS) confirms that Nd doping suppresses overoxidation and changes in the local coordination environment during OER. Phase-enriched reference experiments reveal that α-Mn₂O₃-rich domains provide activity-promoting structural Mn³⁺ motifs, whereas β/R-MnO₂-rich domains contribute structural robustness under acidic OER conditions. Density functional theory (DFT) calculations further reveal phase-dependent OER pathways, with Nd-doped α-Mn₂O₃ favoring an OPM pathway and Nd-doped β/R-MnO₂ remaining adsorbate evolution mechanism (AEM)-dominant with reduced OER barriers. Nd doping also stabilizes Jahn–Teller-active Mn³⁺ centers while enhancing Mn–O covalency through 4f–2p–3d orbital coupling. This study provides mechanistic insights and design principles for developing Ir/Ru-free acidic OER electrocatalysts with high activity and durability.

Results

Catalyst synthesis and characterization

NMO and pristine γ-MnO₂ were synthesized by thermal decomposition. NMOs were prepared and tested at various Nd concentrations; in NMO-x, x denotes the Nd molarity in the precursor solution, which was varied from 0.01 to 0.4. Nd was selected as a dopant to improve the efficiency and stability of Mn oxide-based OER catalysts owing to the unique electronic structures of lanthanides and the relative abundance of Nd. Lanthanides possess highly degenerate yet spatially localized 4f orbitals that are shielded by 5d and 6s valence electrons, thereby enabling modulation of the electronic structure of transition-metal oxide frameworks. We hypothesized that, through 4f–2p−3d orbital coupling (Supplementary Fig. 2), Nd can act as an electron buffer to stabilize catalytically active Jahn–Teller distorted Mn³⁺ species, suppress overoxidation, and enhance Mn–O covalency—collectively improving both the activity and stability of Mn oxide catalysts^22–25.

As-synthesized Mn oxides with and without Nd doping were structurally characterized by powder X-ray diffraction (PXRD) and high-resolution transmission electron microscopy (HRTEM) (Fig. 1a, b). Thermal decomposition of Mn precursors without Nd doping led to the formation of γ-MnO₂ (nsutite), composed of two distinct tunnel structures: “1 × 1” tunnels of β-MnO₂ (pyrolusite) and “1 × 2” tunnels of R-MnO₂ (ramsdellite) (Supplementary Figs. 3 and 4). With increasing Nd doping, the characteristic PXRD peaks of γ-MnO₂ gradually diminished, while those corresponding to α-Mn₂O₃ (bixbyite) became more pronounced (Supplementary Fig. 3), irrespective of calcination temperature (Supplementary Figs. 5 and 6). No impurity phases such as Nd_xO_y were detected (Supplementary Fig. 7). These structural changes were accompanied by red shifts of the (101) and (111) peaks of γ-MnO₂ in PXRD patterns (Supplementary Fig. 8), along with increased d-spacings observed in HRTEM images (Fig. 1b). The measured d-spacings of the (222) plane of α-Mn₂O₃ (2.85 Å) and the (121) plane of R-MnO₂ (2.26 Å) were slightly larger than their theoretical values (2.71 Å and 2.13 Å, respectively). Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) further revealed that NMO-0.2 formed a mixed-phase domain boundary composed of distinct α-Mn₂O₃ (Fig. 1b, R1) and R-MnO₂ (Fig. 1b, R2) phases, together with clear evidence of Nd incorporation into the Mn oxide lattice (Fig. 1c; Supplementary Figs. 9–13). Rietveld refinement confirmed that NMO-0.2 exhibited an increased fraction of R-MnO₂ and a decreased fraction of β-MnO₂ compared with γ-MnO₂. When calcined at 350 °C, the phase composition (mol%) was determined as α-Mn₂O₃:β-MnO₂:R-MnO₂ = 11.6:32.1:56.3 for NMO-0.2 and 0.0:94.1:5.9 for γ-MnO₂ (Fig. 1a; Supplementary Table 1). These results indicate that NMO-0.2 is not an α-Mn₂O₃-dominant catalyst, but rather a MnO₂-rich mixed-phase oxide containing a minority α-Mn₂O₃ phase. In this structure, the α-Mn₂O₃ phase may provide structural Mn³⁺ motifs, whereas the MnO₂-rich β/R framework constitutes the dominant structural component. This mixed-phase assignment was further supported by confocal Raman spectroscopy (Supplementary Fig. 14)^16,26,27. These results collectively suggest that doping with the substantially larger Nd³⁺ ion (0.983 Å), compared with Mn⁴⁺ (0.53 Å), induces the formation of α-Mn₂O₃ and structural deformation of the symmetric “1 × 1” tunnel structure of β-MnO₂, while promoting the formation of asymmetric “1 × 2” tunnel structure of R-MnO₂.

Fig. 1. Structural characterization of the NMO catalyst.

a Rietveld-refined XRD patterns of NMO-0.2 and γ-MnO₂ catalysts. b HRTEM image of the NMO-0.2 sample (left) and magnified views showing different d-spacing in the R1 and R2 regions (right). c AC-HAADF-STEM and corresponding EDS elemental mapping of NMO-0.2. d Normalized Mn K-edge XANES spectra of NMO-0.2 and reference materials, with fitted oxidation states at the absorption edge (inset: oxidation state of Mn as a function of photon energy). e k³-weighted FT-EXAFS spectra at the Mn K-edge. R denotes the radial distance from the absorbing Mn atom. f Schematic illustration of the Nd-induced local distortion strategy to tune the MnO₆ octahedral environment.

The elemental composition and surface electronic structure of the samples were analyzed using multiple characterization techniques. X-ray photoelectron spectroscopy (XPS), inductively coupled plasma-optical emission spectroscopy (ICP-OES), and energy-dispersive X-ray spectroscopy (EDS) confirmed the uniform distribution of Mn, Nd, and O in the NMO-x samples, with a gradual increase in Nd content corresponding to higher Nd precursor concentrations (Supplementary Table 2). X-ray absorption spectroscopy (XAS), XPS (Supplementary Figs. 15–18), and electron paramagnetic resonance (Supplementary Fig. 19) analyses consistently confirmed the formation of Mn³⁺, accompanied by the oxidation of Nd and the generation of oxygen vacancies (O_V). For instance, X-ray absorption near-edge structure (XANES) spectra at the Mn K-edge showed a red shift in the absorption threshold for NMO compared with γ-MnO₂ (Fig. 1d), indicating partial reduction of Mn valence state (Supplementary Fig. 20). Based on linear interpolation between MnO and MnO₂ reference spectra, the average Mn valence states were estimated to be 3.68 for NMO-0.2 and 3.91 for γ-MnO₂ (insets of Fig. 1d). Although the overall XANES line shape of NMO-0.2 remains closer to MnO₂ than to Mn₂O₃, the red-shifted absorption edge indicates that NMO-0.2 contains partially reduced Mn species within a MnO₂-rich framework. This mixed-valence character can be understood as arising from two Mn³⁺ contributions: structural Mn³⁺ associated with α-Mn₂O₃ phase and electronically generated Mn³⁺ by partial reduction of Mn⁴⁺ within the MnO₂-rich framework.

To further examine the local coordination structure, the Mn K-edge EXAFS spectra were analyzed (Fig. 1e, Supplementary Figs. 21, 22, and Supplementary Table 3). The overall EXAFS profile of NMO-0.2 remained broadly MnO₂-like, confirming that NMO-0.2 retains a MnO₂-rich framework rather than forming a Mn₂O₃-dominant structure. Compared with γ-MnO₂, however, NMO-0.2 showed reduced Mn–O peak intensity, a lower Mn–O coordination number, and shortened Mn–Mn distances, indicating that Nd doping induces local coordination distortion, O_V formation, and modified Mn–Mn connectivity within the MnO₂-rich mixed-phase framework.

These findings suggest that Nd doping modifies the Mn oxide framework not by producing a α-Mn₂O₃-dominant catalyst, but by stabilizing Mn³⁺-containing motifs within a MnO₂-rich mixed-phase oxide containing minority α-Mn₂O₃. Mn³⁺, with its high-spin d⁴ electronic configuration, induces asymmetric elongation of MnO₆ octahedra along the axial direction due to Jahn–Teller distortion (Fig. 1f). The concomitant generation of Mn³⁺ and O_V induced by Nd doping can promote octahedral tilting and distortion, which in turn shortens the Mn–Mn distances. Furthermore, the increased number of electrophilic O_V sites is expected to strengthen the adsorption of nucleophilic species such as water molecules and hydroxyl intermediates, potentially facilitating subsequent OER processes.

Catalytic activity and stability under acidic OER conditions

Based on these findings, the electrocatalytic performance of NMO-x and γ-MnO₂ catalysts for acidic OER was evaluated in 0.5 M H₂SO₄ using a three-electrode configuration. The performances of NMO and γ-MnO₂ prepared at 250, 300, 350, 400, and 450 °C were compared, and NMO consistently exhibited substantially lower overpotentials than γ-MnO₂, across all Nd levels and calcination temperatures (Supplementary Fig. 23). Among the tested conditions, NMO synthesized at 350 and 400 °C showed the highest activity. The OER activity of NMOs increased with Nd concentration in the precursor solution up to 0.2 M, beyond which a plateau was observed (Fig. 2a). Among the series, NMO-0.2 prepared at 350 °C exhibited the best overall performance in terms of both activity and stability, with consistently high activity in both the iR-corrected and non-iR corrected analyses, and was therefore selected for further investigations (Fig. 2b; Supplementary Figs. 23–26). These results suggest that beyond a certain threshold, Nd atoms are no longer effectively incorporated into the Mn–O framework, resulting in saturation of catalytic activity and deterioration of structural stability. NMO-0.2 exhibited substantially enhanced performance compared with γ-MnO₂ in terms of apparent activity, electrochemically active surface area (ECSA)-normalized activity, Brunauer–Emmett–Teller (BET) surface area-normalized activity, and turnover frequency (TOF) (Supplementary Figs. 27–31), with overpotentials of 308 vs. 380, 406 vs. 489, and 447 vs. 532 mV at 10, 100, and 200 mA cm^–2, respectively. Moreover, NMO-0.2 showed the lowest degradation rate of 185 μV h⁻¹ at 100 mA cm^–2, outperforming both the lower and higher Nd-doped samples (465 μV h^–1 for NMO-0.1 and 3279 μV h^–1 for NMO-0.4) (Supplementary Fig. 26). It also demonstrated near-unity Faradaic efficiency for OER at 100 mA cm^–2, indicating negligible degradation or side reactions (Supplementary Fig. 32).

Fig. 2. Electrochemical performance of the catalysts.

All electrochemical measurements were performed in 0.5 M H₂SO₄ (pH 0.25) using a three-electrode configuration at 25 °C. a Polarization curves of Mn-based catalysts recorded at 1 mV s^–1 with 100% iR-compensation. b Overpotentials at 10, 100, and 200 mA cm^–2. Data are presented as means $\pm$ standard deviations from three independent measurements (n = 3). c Concentration profiles of Mn and Nd measured by ICP-OES during OER operation at 100 mA cm^–2. d Ex situ UV–vis spectra and photographs of electrolyte solutions collected after electrolysis at various applied potentials for γ-MnO₂ (left) and NMO-0.2 (right). e Chronopotentiometric stability tests of γ-MnO₂ and NMO-0.2 at 200 mA cm^–2. The right panels compare both catalysts on Pt/Ti fiber felt under identical conditions, whereas the left panels present the corresponding measurements on carbon fiber paper for comparison.

Next, the stability of NMO-0.2 and γ-MnO₂ supported on carbon paper was evaluated. NMO-0.2 maintained stable operation for >200 h at 100 mA cm^–2, whereas γ-MnO₂ was rapidly deactivated within 80 h (Supplementary Fig. 33), likely due to its dissolution into high-valent MnO₄^– species. Concentration profiles of dissolved metal species at 100 mA cm^–2, monitored over 60 h by ICP-OES, revealed that Nd doping effectively suppresses Mn dissolution and improves catalyst stability (Fig. 2c). It is well-known that γ-MnO₂ is susceptible to oxidative corrosion through the formation of MnO₄^– species beyond its narrow stable potential window (1.6 to 1.75 V_RHE) under acidic conditions. Indeed, the electrolyte collected after electrolysis at 1.8 V_RHE showed a distinct pink coloration when γ-MnO₂ was used, with absorption peaks at 525 and 545 nm (Fig. 2d). In contrast, negligible absorbance was observed for NMO-0.2, suggesting an extended potential window of electrochemical stability.

Encouraged by these results, we further evaluated the high-current stability of γ-MnO₂ and NMO-0.2 on different electrode supports at 200 mA cm^–2 in 0.5 M H₂SO₄ (Fig. 2e). On carbon fiber paper, γ-MnO₂ rapidly deactivated within <30 h, whereas NMO-0.2 maintained operation for ~100 h before rapid deactivation, likely due to carbon corrosion and catalyst detachment under strongly oxidative acidic conditions. To minimize support-related degradation, both catalysts were further tested on platinized titanium fiber felt (Pt/Ti), a corrosion-resistant porous transport layer commonly used in PEMWE systems (Supplementary Fig. 34). Under this configuration, γ-MnO₂ exhibited improved stability compared with that on carbon fiber paper, but its performance began to degrade rapidly after ~300 h, indicating that the Pt/Ti support mitigates support-related degradation but does not fully prevent catalyst deactivation of our thermally prepared γ-MnO₂ under acidic OER condition. More importantly, NMO-0.2 showed high durability, operating stably for >1000 h at 200 mA cm^–2 under the same Pt/Ti fiber felt configuration. These results demonstrate that NMO-0.2 retains high-current durability on a practical corrosion-resistant support, showing competitive performance compared with reported Mn-based acidic OER catalysts (Fig. 2f; Supplementary Table 4).

Reaction mechanism and in situ characterization

To elucidate the underlying mechanism responsible for the enhanced catalytic activity upon Nd doping, a series of electrochemical analyses were performed. The distinct peaks corresponding to ^*OH adsorption/desorption in cyclic voltammetry (Supplementary Fig. 35) and the enhanced methanol oxidation reaction (MOR) activity of NMO-0.2 (Supplementary Fig. 36) indicate a high population of electrophilic oxygen intermediates, such as ^*OH and ^*O species, on the surface of the NMO catalysts. Upon introducing tetramethylammonium cations (TMA⁺)—a probe known to suppress lattice oxygen-mediated (LOM)-type catalysts by electrostatically pairing with long-lived, negatively charged surface peroxo species (O₂^2–)—the representative LOM catalyst RuO₂ showed pronounced OER inhibition^28,29. Interestingly, NMO-0.2 exhibited an additional redox peak attributed to the ^*O/^*O₂^2– transition but remained unaffected (Supplementary Fig. 37), suggesting that ^*O–O^* species formed on its surface are transient, more covalent (less anionic) and/or less accessible, and thus not efficiently trapped by TMA⁺. Taken together, these electrochemical probe experiments suggest that NMO-0.2 is unlikely to follow a conventional LOM pathway^15,18.

In situ Raman measurement was conducted to probe reaction intermediates under applied potentials ranging from open circuit voltage (OCV) to 1.95 V_RHE in 0.5 M HClO₄ (Fig. 3a, b and Supplementary Fig. 38). The peaks at ~510/~650, ~580, and ~930 cm^–1 are attributed to out-of-plane symmetric Mn–O stretching, in-plane Mn–O stretching along the octahedral chains, and ClO₄^– species, respectively^24,30,31. With increasing applied potential, a new band corresponding to oxygen-containing Mn–oxo intermediates, such as surface terminal Mn⁴⁺=O^24,32, emerged at ~750 cm^–1. This assignment is supported by the isotope-induced red shift from ~780 cm^–1 in 0.5 M H₂SO₄ containing H₂¹⁶O electrolyte to ~730 cm^–1 in 0.5 M H₂SO₄ containing H₂¹⁸O electrolyte (Supplementary Fig. 39)³³. NMO-0.2 exhibited this Mn⁴⁺=O feature at a lower potential than γ-MnO₂ (1.65 and 1.75 V_RHE, respectively), indicating faster kinetics of the intermediate transition from M–^*OH to M–^*O species facilitated by Mn³⁺ and O_V in NMO-0.2. It is also noteworthy that γ-MnO₂ displayed a peak at ~835 cm⁻¹ corresponding to dissolved high-valent MnO₄^– species at >1.75 V_RHE, whereas no such feature was observed for NMO-0.2 even at 1.95 V_RHE.

Fig. 3. Investigation of the OER mechanism through in situ spectroscopic studies.

In situ Raman measurements were performed in 0.5 M HClO₄, and in situ XAS measurements in 0.5 M H₂SO₄, both at 25 °C. a, b In situ Raman spectra of γ-MnO₂ and NMO-0.2 collected under applied potentials, with magnified views of the 550–850 cm^–1 region. c Schematic illustration of the proposed OER pathways on γ-MnO₂ and NMO-0.2 surfaces. d, e In situ Mn K-edge XANES spectra of γ-MnO₂ and NMO-0.2 measured at OCV before OER (OCV_pre), 1.55 V_RHE, 1.65 V_RHE, and OCV after OER (OCV_post). The right panels show magnified absorption-edge regions. f, g k²-weighted FT-EXAFS spectra of γ-MnO₂ and NMO-0.2 collected at OCV_pre, 1.55 V_RHE, 1.65 V_RHE, and OCV_post. R denotes the radial distance from the absorbing Mn atom. h Schematic illustration of phase-dependent role and stabilization of Jahn–Teller-active Mn³⁺ motifs in NMO-0.2, including structural Mn³⁺ in the α-Mn₂O₃ phase and electronically induced Mn³⁺ in the MnO₂-rich framework.

To further probe the proposed oxygen-participating mechanism, we performed in situ differential electrochemical mass spectrometry (DEMS) using ¹⁸O-labeled γ-MnO₂ and NMO-0.2 catalysts in ¹⁶O-containing 0.5 M H₂SO₄. In an OPM-type pathway, catalyst-associated ¹⁸O species can participate in O–O bond formation together with oxygen species supplied from H₂¹⁶O. Thus, the detection of isotope-labeled oxygen products at m/z = 34 and m/z = 36, corresponding to ³⁴O₂ (¹⁸O¹⁶O) and ³⁶O₂ (¹⁸O¹⁸O), respectively, provides evidence for the involvement of catalyst-associated oxygen species during OER (Supplementary Figs. 40 and 41)³³.

Compared with γ-MnO₂, NMO-0.2 exhibited more pronounced isotope-labeled oxygen signals during OER, with detectable m/z = 32, 34, and 36 signals corresponding to ³²O₂, ³⁴O₂, and ³⁶O₂, respectively (Supplementary Fig. 42). The appearance of mixed-isotope oxygen products supports the participation of catalyst-associated oxygen species in the O–O coupling process on NMO-0.2. Together with the in situ Raman observation of earlier formation of labile ^*O species, the isotope-labeling DEMS results support a predominantly OPM-like pathway in which surface or near-surface oxygen intermediates participate in O–O bond formation while the Mn–O framework is largely preserved (Fig. 3c).

In situ XAS analysis was performed in 0.5 M H₂SO₄ electrolyte to examine the potential-dependent changes in the Mn oxidation state and local coordination environment during acidic OER (Supplementary Fig. 43). In the Mn K-edge XANES spectra, γ-MnO₂ exhibited a clear blue shift of the absorption edge with increasing applied potential, indicating progressive Mn oxidation under OER conditions (Fig. 3d). At OCV after OER, the absorption edge of γ-MnO₂ did not fully recover to its initial position, suggesting irreversible structural/electronic changes in the MnO₂ framework. In contrast, NMO-0.2 showed only minor absorption-edge shifts during OER, and the edge position was restored after returning to OCV (Fig. 3e). This reversible behavior indicates that NMO-0.2 better preserves the Mn oxidation state and local electronic structure under acidic OER conditions.

The corresponding in situ Mn K-edge FT-EXAFS spectra further support this interpretation (Supplementary Fig. 44). For γ-MnO₂, the Mn–O feature in the first coordination shell shifted to a shorter radial distance with increasing applied potential, accompanied by a decrease in Mn–O peak intensity (Fig. 3f and Supplementary Fig. 45). After returning to OCV after OER, the Mn–O feature did not fully recover to its initial state, indicating irreversible local coordination changes under acidic OER conditions. In contrast, NMO-0.2 maintained a nearly unchanged Mn–O radial distance and preserved Mn–O peak intensity throughout the potential sequence (Fig. 3g and Supplementary Fig. 45). The Mn–O feature also returned close to its initial state after OER, indicating that the local Mn–O coordination environment is largely preserved.

Taken together, the in situ XAS results show that γ-MnO₂ undergoes irreversible Mn oxidation and local structural degradation under acidic OER conditions, whereas NMO-0.2 exhibits largely reversible oxidation-state and coordination responses. This behavior indicates that Nd incorporation helps preserve the Mn oxidation state and local Mn–O framework during OER, supporting the stabilization of Mn³⁺-containing motifs in the mixed-phase NMO structure (Fig. 3h). Such structural and electronic robustness contributes to the enhanced acidic OER activity and durability of NMO-0.2.

Based on the electrochemical, in situ Raman, in situ DEMS, and in situ XAS results discussed above, we propose an OER mechanism for NMO-0.2. Nd doping likely induces the formation of Mn–O–Nd motifs and O_v, shortens the Mn–Mn distance, and enriches the Mn³⁺ active centers with electrons through 4f–2p–3d orbital coupling. The electron-enriched surface Mn sites facilitate the formation of labile ^*O species with weakened Mn–O bonds. These labile ^*O species can readily couple with adjacent ^*O species, thereby accelerating O–O radical coupling via the oxide path mechanism (OPM) pathway (Fig. 3c)^24,34.

To experimentally examine the role of each Mn oxide polymorph in the mixed-phase NMO catalyst, phase-enriched Mn oxide reference catalysts were prepared using the same thermal-decomposition platform (Supplementary Fig. 46). Although complete phase purity could not be achieved because of the inherent coexistence of Mn oxide polymorphs during thermal decomposition, the dominant phase was systematically tuned by controlling the synthesis conditions. Electrochemical measurements showed that the α-Mn₂O₃-rich reference exhibited the highest OER activity (Supplementary Fig. 47), whereas the β/R-MnO₂-rich references displayed enhanced stability under acidic conditions (Supplementary Fig. 48). Nd incorporation further improved the activity and/or stability of all phase-enriched references. Therefore, the enhanced OER performance of NMO-0.2 is attributed to the combined contribution of minority α-Mn₂O₃-derived structural Mn³⁺ motifs and electronically induced Mn³⁺ within the dominant MnO₂-rich framework, rather than to an α-Mn₂O₃-dominant phase. These results provide the experimental basis for the phase-specific DFT analysis presented below, where each polymorph is modeled separately to clarify how Nd incorporation modifies its local structural and electronic properties.

Theoretical validation of mechanisms

Guided by the experimentally identified phase-dependent contributions, we next performed DFT calculations to examine how Nd incorporation modifies the local coordination structure and electronic properties of each Mn oxide polymorph (Supplementary Data 1). Based on the PXRD and HRTEM results, structural models were constructed for three representative Mn oxide polymorphs in the Nd-doped NMO catalyst: α-Mn₂O₃, β-MnO₂, and R-MnO₂ (Supplementary Fig. 49 and Supplementary Table 1). Although α-Mn₂O₃ is not present in the pristine γ-MnO₂ sample, the pristine α-Mn₂O₃ model was included as a phase-specific reference to evaluate how Nd incorporation modifies the α-Mn₂O₃ component observed in NMO. Nd-doped models were generated by placing Nd atoms at interstitial sites within the open-channel frameworks of each polymorph, considering the large ionic-radius mismatch between Mn and Nd (Supplementary Fig. 50).

The optimized DFT structures show trends consistent with the experimentally observed local structural evolution. The calculated Mn–O bond lengths agree with the EXAFS/XRD-derived values and reproduce the Nd-induced bond-length variation (Supplementary Fig. 51 and Supplementary Table 3), supporting the validity of the models for describing the local coordination environment of Nd-doped Mn oxides. The optimized Nd-doped structures further reveal local distortion of the surrounding Mn–O polyhedra, indicating that Nd doping modifies the Mn–O coordination environment. In Mn⁴⁺-dominant β-MnO₂ and R-MnO₂, electron donation from Nd can partially convert Mn⁴⁺ toward a Mn³⁺-like electronic state, producing electronically driven Jahn–Teller distortion of the MnO₆ octahedra. In contrast, α-Mn₂O₃, which is already Mn³⁺-dominant and structurally distorted, shows a less pronounced additional distortion upon Nd incorporation.

Although the observed distortion may also be affected by local strain arising from the large ionic size of Nd, control calculations in which extra electrons were artificially introduced into pristine β-MnO₂ and R-MnO₂ produced even stronger octahedral distortions than those in the Nd-doped Mn oxide systems (Supplementary Figs. 52 and 53). This result suggests that the Jahn–Teller distortion may arise predominantly from Nd-derived electron donation rather than by ionic-size-induced lattice strain alone. Additional oxygen-vacancy (O_V) calculations further indicate that O_V formation does not dominate the observed structural evolution. Together, these control calculations support that Nd-derived electron donation may contribute substantially to the Mn-centered Jahn–Teller distortion (Supplementary Fig. 54).

The partial density of states (PDOS) of the pristine and Nd-doped Mn oxides was then analyzed to clarify the electronic origin of the Nd-induced modulation (Fig. 4a–c). Upon Nd incorporation, the Mn 3d and O 2p states exhibit peak broadening and stronger orbital overlap near the Fermi level, indicating enhanced Mn–O electronic interaction, increased electron delocalization, and strengthened Mn–O covalency. In addition, the d–p band center difference (Δε_d–p), a key descriptor for OER activity, shifts closer to the Fermi level in the Nd-doped systems (−0.38, −1.19, and −1.15 eV for Nd-α-Mn₂O₃, Nd-β-MnO₂, and Nd-R-MnO₂, respectively) compared with the pristine phases (−0.43, −1.37, and −1.22 eV)³⁵. Bader charge analysis further confirms that Nd dopants increase the Bader charge of neighboring Mn atoms, indicating partial reduction of Mn from 1.79 to 1.40 in α-Mn₂O₃, 2.00 to 1.67 in β-MnO₂, and 1.72 to 1.30 in R-MnO₂ (Supplementary Fig. 55). These results demonstrate that Nd acts as an electron donor, modulating the local electronic structure and promoting Mn-centered electronic Jahn–Teller distortion.

Fig. 4. Nd-induced electronic reconstruction and phase-dependent OER mechanisms in Mn oxides.

a–c PDOS of pristine and Nd-doped α-Mn₂O₃, β-MnO₂, and R-MnO₂, showing enhanced Mn 3d–O 2p hybridization with Nd 4f contribution after Nd doping. PDOS is presented in arbitrary units (arb. units). d–f Gibbs free energy diagrams of AEM and OPM pathways on pristine and Nd-doped Mn-oxide slabs, showing an OPM-favored pathway in Nd-α-Mn₂O₃ (111) and AEM-dominant pathways in Nd-β-MnO₂ (110) and Nd-R-MnO₂ (100). The free-energy profiles are shown at electrode potentials of U = 0 and 1.23 V, where U denotes the applied electrode potential. g Schematic illustration of phase-dependent OER pathways, in which structural Mn³⁺ in Nd-α-Mn₂O₃ favors OPM, whereas electronically induced Mn³⁺ enhances AEM. h Schematic illustration of Nd 4f–O 2p–Mn 3d orbital coupling.

To clarify the phase-dependent role of Nd doping in OER, we compared the Gibbs free energy diagrams of AEM and OPM for pristine and Nd-doped Mn-oxide slabs (Fig. 4d–g and Supplementary Figs. 56–59). Pristine α-Mn₂O₃ (111), β-MnO₂ (110), and R-MnO₂ (100) preferentially follow the conventional AEM pathway (Supplementary Fig. 55), whereas Nd doping lowers the OER barriers in all three phases (Supplementary Figs. 57–59). Notably, only Nd-α-Mn₂O₃ favors the OPM pathway (Fig. 4d and Supplementary Fig. 57), consistent with the activity-promoting role of α-Mn₂O₃-rich references observed experimentally. In contrast, Nd-β-MnO₂ and Nd-R-MnO₂ remain AEM-dominant, indicating that Nd doping improves their activity mainly by facilitating the conventional adsorbate-mediated OER steps rather than by inducing dual-site OPM activation. These results suggest that Nd doping accelerates OER through phase-dependent pathways (Fig. 4g): structural Mn³⁺ in Nd-α-Mn₂O₃ promotes OPM, whereas electronically induced Mn³⁺ in Nd-β-MnO₂ and Nd-R-MnO₂ enhances AEM. This highlights that the catalytic role of Mn³⁺ is governed not only by oxidation state but also by its structural environment.

The electronic origin of the Nd-doping effect was further examined by analyzing the orbital energy alignment between Mn 3d and O 2p states. Nd doping introduces Nd 4f states that hybridize with Mn 3d and O 2p orbitals, forming 4f–2p–3d orbital coupling. This interaction acts as an electron-donating channel from Nd to the Mn–O framework, thereby reconstructing the local electronic structure of Mn sites. As shown in Fig. 4h, Nd doping shifts the Mn 3d–O 2p energy separation upward from Δε_d–p = −1.00 eV to −0.93 eV, indicating enhanced Mn–O orbital hybridization and increased covalent character at the active sites. Work-function calculations further suggest possible interphase charge redistribution between α-Mn₂O₃ and MnO₂-rich domains, which may additionally modulate the local Mn–O electronic structure (Supplementary Fig. 60).

Overall, these results show that Nd dopants function as 4f-orbital electron donors that simultaneously modulate the electronic and structural degrees of freedom in Mn oxides. Through 4f–2p–3d orbital coupling, Nd doping strengthens Mn–O covalency, induces Mn-centered Jahn–Teller distortion, and optimizes OER reaction energetics. The resulting OER pathway is therefore determined not simply by the presence of Mn³⁺, but by how Mn³⁺ is manifested in the local structure: structurally embedded Mn³⁺ in Nd-α-Mn₂O₃ enables dual-site O–O coupling through the OPM pathway, whereas electronically induced Mn³⁺ in Nd-β/R-MnO₂ mainly tunes adsorbate evolution through the AEM pathway. This establishes a direct structure–property correlation between Nd-induced orbital reconstruction, phase-dependent Mn³⁺ stabilization, and enhanced OER performance. These calculations use idealized static models that do not fully capture catalytic operating conditions; therefore, the DFT results should be interpreted as qualitative trends.

PEMWE performance evaluation

The NMO-0.2 catalyst was integrated into a membrane electrode assembly (MEA) for PEMWE operation. The MEA was fabricated by the catalyst-coated substrate (CCS) method using NMO-0.2 as the anode catalyst, Pt/C as the cathode catalyst, and Nafion^® N117 as the PEM (Fig. 5a and Supplementary Fig. 61). The NMO-0.2-based PEMWE achieved a cell voltage of 1.88 V at 1 A cm^–2 (Fig. 5b) and maintained stable operation for >600 h at 100 mA cm^–2, outperforming γ-MnO₂ (Fig. 5c). It also operated for >300 h at 200 mA cm^–2 (Fig. 5d) and for >60 h at 500 mA cm^–2 (Fig. 5e), demonstrating the PEMWE applicability of NMO-0.2 under high-current acidic OER conditions.

Fig. 5. Evaluation of PEMWE performance.

PEMWE polarization measurements were conducted at 80 °C, while durability tests were performed at 60 °C, with 0.5 M H₂SO₄ (pH = 0.25) fed at a flow rate of 1 mL min^–1. a Schematic illustration of the PEMWE system with an MEA prepared by the CCS method. b Polarization curves of PEMWE cells employing γ-MnO₂ or NMO-0.2 as the anode catalysts. c–e Cell-voltage profiles of PEMWE cells employing γ-MnO₂ or NMO-0.2 as the anode catalyst at 100 mA cm^–2, together with NMO-0.2-based PEMWE operation at 200 and 500 mA cm^–2. f XRD patterns of NMO deposited on the Pt/Ti felt before and after electrolysis at 200 mA cm^–2. The # symbols indicate the Ti substrate features, while the yellow-shaded regions highlight diffraction peaks showing significant changes after electrolysis at 200 mA cm^–2.

Although the NMO-0.2-based PEMWE showed prolonged operation, a gradual increase in cell voltage was observed during extended electrolysis, indicating partial performance degradation. To clarify the origin of this deactivation, XRD patterns of the NMO anode on Pt/Ti felt were collected before and after electrolysis at 200 mA cm^–2 (Fig. 5f). After electrolysis, the diffraction features corresponding to pristine NMO were weakened, while additional features associated with the deactivated NMO phase appeared, suggesting partial structural reconstruction of the Mn oxide catalyst under acidic and oxidative PEMWE conditions. The Ti substrate peaks remained clearly visible, confirming that the observed changes mainly originated from the NMO catalyst layer rather than the underlying substrate. These results indicate that the voltage increase during long-term operation is primarily associated with gradual structural reconstruction/deactivation of the NMO anode, rather than immediate failure of the MEA. Overall, the NMO-based PEMWE demonstrated promising efficiency and durability among reported non-noble-metal acidic OER catalysts (Supplementary Fig. 62 and Supplementary Table 5), while the post-electrolysis structural analysis further clarifies that partial reconstruction of the NMO anode is responsible for the remaining performance decay under high-current PEMWE operation.

Discussion

In summary, this study demonstrates that Nd doping enables efficient and durable OER performance of Mn-based catalysts under acidic conditions. In particular, Nd-doped Mn oxide (NMO) achieved a low overpotential of 308 mV at 10 mA cm^–2 and maintained stable operation for >1000 h at 200 mA cm^–2 under acidic conditions. When integrated into a PEMWE, NMO delivered a cell voltage of 1.88 V at 1 A cm^–2 and operated stably for >300 h at 200 mA cm^–2 (>60 h at 500 mA cm^–2). In situ XAS, Raman, and DEMS analyses collectively show that Nd doping stabilizes Jahn–Teller–active Mn³⁺ centers, suppresses Mn overoxidation and dissolution, and promotes a predominantly OPM-like pathway. DFT calculations further reveal that Nd-derived 4f–2p–3d orbital coupling enhances Mn–O covalency and induces a phase-dependent OER mechanism: Nd-doped α-Mn₂O₃ preferentially follows the OPM pathway, whereas Nd-doped β-MnO₂ and R-MnO₂ remain AEM-dominant with reduced OER barriers. Therefore, NMO may serve as a promising non-noble-metal OER catalyst for PEMWEs. These findings establish a rational design strategy for low-cost Mn-based OER catalysts, demonstrating that 4f-element doping can effectively stabilize catalytically active Mn³⁺ motifs and bridge the activity–stability gap in acidic OER.

Methods

Reagents and materials

All chemical reagents were used as received without further purification. Mn(NO₃)₂·4H₂O (99.9%), Nd(NO₃)₃·6H₂O (99.9%), MnO (99%), Mn₂O₃ (99.9%), Mn₃O₄ (97%), MnO₂ (99%), Nd₂O₃ (99.9%), RuO₂ (99.9%), H₂SO₄ (95–98%), H₂O₂ (30 wt%), C₂H₅OH (99.9%), CH₃OH (99.9%), tetramethylammonium chloride (TMACl, 98%), and isotope-labeled H₂¹⁸O water (97 atomic % ¹⁸O) were purchased from Sigma-Aldrich. Nafion^® ionomer dispersion (D520) was purchased from DuPont Co. Carbon fiber paper (TGP-H-060) was purchased from Fuel Cell Earth Co. Platinized titanium (Pt/Ti) fiber felt, titanium fiber felt, and platinum on carbon black (Pt/C, 20 wt%) were obtained from Fuel Cell Store. Nafion^® 117 polymer membrane (N117) was purchased from Nara Cell Tech Co. Milli-Q ultrapure water (18.2 MΩ·cm) was used throughout all experiments.

Catalyst synthesis

Mn oxide-based OER catalysts were deposited onto self-supported anode substrates. A series of NMO-x catalysts (where x denotes the molarity of Nd) and γ-MnO₂ were synthesized via a thermal decomposition method. Mn(NO₃)₂·4H₂O (3 M) and Nd(NO₃)₃·6H₂O with specific molar concentrations were dissolved in deionized water to prepare precursor solutions. The precursor mixture was then drop-cast onto pretreated anode substrates (carbon fiber paper and Pt/Ti fiber felt) with geometric surface area of 1 × 1 cm^–2, followed by calcination at 250, 300, 350, 400, and 450 °C for 6 h with a ramp rate of 24 °C min^–1. After natural cooling to room temperature at 25 °C, the prepared electrodes were rinsed thoroughly with deionized water and dried in an oven at 60 °C. The catalyst loading was approximately 8 mg cm^–2.

Characterizations

Powder X-ray diffraction (PXRD) was performed using a D/MAX 2500-V diffractometer (Rigaku) equipped with a Cu K$\mathrm{\alpha }$ radiation source ($\mathrm{\lambda }$ = 1.5406 Å) operated at 40 kV and 200 mA. High-resolution transmission electron microscopy (HRTEM) was carried out using a JEM–2100F microscope (JEOL) operated at 200 kV, and the corresponding energy-dispersive X-ray spectroscopy (EDS) mapping was obtained using an X-Max T-80 detector (Oxford Instruments). Aberration-corrected high-angle annular dark-field transmission electron microscopy (AC-HAADF-STEM) was performed using a JEM-ARM 300F (JEOL) operated at 200 kV. X-ray photoelectron spectroscopy (XPS) was conducted on an ESCALAB 250 spectrometer (Thermo Fisher Scientific) equipped with a monochromatic Al K$\mathrm{\alpha }$ X-ray source. Electron paramagnetic resonance (EPR) spectra were obtained using an EMXplus spectrometer (Bruker) at 77 K. Ultraviolet–visible (UV–vis) absorption spectra were measured using a V-730 (JASCO). Inductively coupled plasma-optical emission spectrometry (ICP-OES) was measured on a 700-ES instrument (Varian). Synchrotron-based X-ray absorption spectroscopy (XAS) measurements, including X-ray absorption near-edge structure (XANES) spectra and extended X-ray absorption fine structure (EXAFS) at Mn K-edge, were conducted at the 6D beamline of the Pohang light source (PLS) at the Pohang Accelerator Laboratory (PAL), South Korea. The acquired XAS data were processed using the Athena program (version 0.9.26) for background subtraction and pre-edge/post-edge calibration. All EXAFS spectra were K²-weighted and Fourier transformed in the K-range 2–12 Å^–1 using a Hanning window function. Confocal Raman spectra were recorded using an Alpha300 R (WITec) equipped with a 532 nm laser. Gas chromatograph (GC) was carried out using a 2010 Plus (Shimadzu) equipped with a barrier discharge ionization detector (BID) and helium as the carrier gas. N₂ adsorption/desorption isotherms and Brunauer–Emmett–Teller (BET) surface areas were measured using an ASAP 2420 (Micromeritics).

Electrochemical characterizations

All electrochemical measurements were performed in 50 mL of freshly prepared 0.5 M H₂SO₄ electrolyte solution (pH ~0.25 ± 0.05) using a three-electrode configuration connected to a SP–150 potentiostat (Bio-Logic Science Instruments). The electrolyte storage condition in sealed glass bottle at 25 °C until use. The Ag/AgCl reference electrode (3 M KCl) was calibrated against a reversible hydrogen electrode in the same electrolyte at 25 °C before the electrochemical measurements. All potentials measured versus Ag/AgCl were converted to the reversible hydrogen electrode (RHE) and corrected for iR drop scale according to the following equation:

$${\mathrm{E}}_{\left(\mathrm{Potential},\mathrm{V}\mathrm{vs}.\mathrm{RHE}\right)}={\mathrm{E}}_{\left(\mathrm{Potential},\mathrm{V}\mathrm{vs}.\mathrm{Ag/AgCl}\right)}+\underset{20}{{\mathrm{E}}_{\mathrm{Ag}/\mathrm{AgCl}}^0}+0.0591\times \mathrm{pH}$$ 1

where ${\mathrm{E}}_{\left(\mathrm{Potential},\mathrm{V}\mathrm{vs}.\mathrm{Ag/AgCl}\right)}$ is the experimentally measured potential versus Ag/AgCl, ${\mathrm{E}}_{\mathrm{Ag}/\mathrm{AgCl}}^0$ is 0.197 V, and the pH is 0.25 in 0.5 M H₂SO₄. Linear sweep voltammetry (LSV) was conducted at a scan rate of 1 mV s^–1, All LSV curves were corrected for 100% iR drop according to the following equation:

$${\mathrm{E}}_{\mathrm{Final}}={\mathrm{E}}_{\left(\mathrm{Potential},\mathrm{V}\right)}\text{-}\mathrm{iR}$$ 2

where ${\mathrm{E}}_{\left(\mathrm{Potential},\mathrm{V}\right)}$ is the measured potential, i is the current at ${\mathrm{E}}_{\left(\mathrm{Potential},\mathrm{V}\right)}$, and R denotes the solution resistance (R_s) value between working and reference electrodes. The uncompensated R_s between the working and reference electrodes was measured by electrochemical impedance spectroscopy (EIS) under open circuit voltage (OCV) condition.

Redox features were examined by cyclic voltammetry (CV). Electrochemically active surface area (ECSA) was determined from the double-layer capacitance (C_dl) according to the following equation:

$$\mathrm{ECSA}=\frac{{\mathrm{C}}_{\mathrm{dl}}}{{\mathrm{C}}_{\mathrm{s}}}$$ 3

where C_dl was obtained from CV curves in the non-Faradaic region at a scan rate ranging from 10 to 100 mV s^–1. Specific capacitance (C_s) was assumed to be 0.35 mF cm^–2. Specific activities were normalized to both the ECSA and BET surface area. Long-term stability was evaluated by chronopotentiometry (CP).

Turnover frequency (TOF) calculation

The TOF was calculated according to the following equation:

$$\mathrm{TOF}=\frac{\mathrm{j}\times \mathrm{A}\times \mathrm{\eta }}{\mathrm{n}\times \mathrm{e}\times \mathrm{m}}$$ 4

where j is the measured current density (A cm^–2), A is the surface area of the electrode (cm), η is the Faradaic efficiency (assumed to be 100%), n is the number of electrons transferred per O₂ molecules (4 for OER), e is the elementary charge of an electron (1.602 $\times$ 10^–19 C), and m is the number of active sites. The TOF values were calculated from a single.

Faradaic efficiency calculation

GC was used to quantify the amount of O₂ generated during electrolysis in a two-chamber, three-electrode cell separated by a Nafion^® 117 membrane. The Faradaic efficiency was determined from a single measurement. The Faradaic efficiency for oxygen evolution was calculated using the following equation:

$$\mathrm{Faradaic}\mathrm{efficiency}=\frac{{\mathrm{n}}_{{\mathrm{O}}_2,\exp }}{{\mathrm{n}}_{{\mathrm{O}}_2,\mathrm{theo}}}$$ 5

where ${\mathrm{n}}_{{\mathrm{O}}_2,\exp }$ and ${\mathrm{n}}_{{\mathrm{O}}_2,\mathrm{theo}}$ are the experimentally measured and theoretically calculated moles of O₂, respectively. ${\mathrm{n}}_{{\mathrm{O}}_2,\mathrm{theo}}$ was obtained using Faraday’s law of electrolysis:

$$\mathrm{m}=\frac{\mathrm{I}\times \mathrm{t}}{\mathrm{n}\times \mathrm{F}}$$ 6

where I is the applied current (A), t is the electrolysis time (s), n is the number of electrons transferred per O₂ molecule (4 for OER), and F is the Faraday constant (96,485 C mol^–1).

${\mathrm{n}}_{{\mathrm{O}}_2,\exp }$ was determined using the ideal gas law, assuming ideal gas behavior:

$${\mathrm{n}}_{{\mathrm{O}}_2,\exp }=\frac{\mathrm{P}\times \mathrm{V}}{\mathrm{R}\times \mathrm{T}}$$ 7

where V is the collected O₂ volume (L). T is the temperature (K), R is the ideal gas constant (0.08206 L·atm·mol^–1·K^–1), and P is the partial pressure of O₂ (atm). The partial pressure of O₂ was calculated using:

$${\mathrm{P}}_{{\mathrm{O}}_2}=\frac{{\mathrm{P}}_{\mathrm{Total}}-{\mathrm{P}}_{\mathrm{Water}}}{{\mathrm{P}}_{\mathrm{Total}}}\times 1\mathrm{atm}$$ 8

where ${\mathrm{P}}_{{\mathrm{O}}_2}$ is 21.1 mm Hg and ${\mathrm{P}}_{\mathrm{Total}}$ is 762 mm Hg.

ICP-OES analysis of dissolved species in the electrolyte

Chronopotentiometry measurements of NMO-0.2 and γ-MnO₂ were conducted at a current density of 100 mA cm^–2 for 60 h. Electrolysis was performed in a two-chamber, three-electrode cell separated by a Nafion^® 117 membrane. During electrolysis, fixed volumes of electrolyte were periodically sampled from the anode chamber at predetermined time intervals for analysis using ICP-OES.

Molecular probing analysis via methanol oxidation reaction (MOR)

To examine the abundance of surface ^*OH coverage, CH₃OH was used as a nucleophile to probe electrophilic ^*OH species through nucleophilic attack. The MOR was measured in 0.5 M H₂SO₄ containing various CH₃OH concentrations. All electrochemical measurements were performed using the same procedures described above.

Molecular probing analysis using tetramethylammonium (TMA⁺) cation

To capture peroxo-species (O₂^2–) bounded to the catalyst surface, a TMA⁺-containing H₂SO₄ electrolyte was prepared by mixing 0.5 M H₂SO₄ with TMACl (0.1 or 0.5 M, if used). To ensure complete dissolution of dense TMACl, the solution was stirred vigorously for 6 h. The catalyst-loaded electrodes were soaked in the TMA⁺-containing electrolyte for 1 h prior to electrochemical testing, which was performed following the same procedures described previously.

In situ Raman spectroscopy

In situ Raman spectroscopy was performed using a custom-designed in situ electrochemical cell equipped with an immersion objective lens (50×, N.A. = 0.8). The working electrode was prepared on carbon fiber paper, with a Pt wire and an Ag/AgCl electrode used as the counter and reference electrodes, respectively. A 0.5 M HClO₄ solution served as the electrolyte. To prevent electrolyte leakage, an additional hydrophobic carbon fiber paper was placed beneath the working electrode and affixed to a conductive tantalum (Ta) ring. Electrochemical potentials were applied using an SP–150 potentiostat. All Raman spectra were acquired at a laser power of 1 mW and 30 accumulations per spectrum. For ¹⁸O labeling in situ Raman measurements, 0.5 M H₂SO₄ electrolyte was prepared using isotope-labeled H₂¹⁸O water (97 atom% ¹⁸O) as the solvent. The Raman spectra were collected under the same electrochemical and optical conditions and compared with those measured in H₂¹⁶O electrolyte.

In situ XAS measurement

In situ Mn K-edge XAS measurements, including XANES and EXAFS, were conducted in fluorescence mode using a custom-designed electrochemical cell equipped with polyimide film windows to allow efficient X-ray transmission while minimizing absorption and scattering. The catalyst-coated working electrode was mounted on the front side of the cell to reduce the effective electrolyte/X-ray path length and minimize bubble-induced scattering during fluorescence-mode measurements. The catalyst-coated working electrodes were prepared by drop-casting a precursor solution of 3 M Mn(NO₃)₂·4H₂O and 0.2 M Nd(NO₃)₃·6H₂O onto hydrophilic carbon fiber paper. The electrodes were then annealed in air at 350 °C for 6 h to directly form a binder-free catalyst layer on the carbon fiber paper through a self-supporting thermal-decomposition process.

Electrochemical measurements were carried out in a three-electrode configuration using the catalyst-coated electrode as the working electrode, a Pt wire as the counter electrode, and an Ag/AgCl electrode as the reference electrode. Prior to in situ XAS measurements, the catalyst-coated electrodes were pretreated by electrolysis in 0.5 M H₂SO₄ for 1 h. Ar-degassed 0.5 M H₂SO₄ was used as the electrolyte. XAS spectra were collected at OCV before OER, 1.55 V_RHE, 1.65 V_RHE, and OCV after OER, with chronoamperometry performed during each acquisition. At each applied potential, XAS spectra were collected with an acquisition time of approximately 25 min per scan. After each spectrum was collected, gas bubbles attached to the electrode surface were physically removed before proceeding to the next potential. The measurements were repeated using independently prepared electrodes to confirm reproducibility.

The acquired XAS spectra were processed by subtracting the pre-edge background and normalizing the post-edge region. EXAFS analysis was performed by Fourier transformation of the k²-weighted EXAFS oscillations, and k³-weighted EXAFS oscillations were also examined to assess the high-k data quality.

In situ differential electrochemical mass spectrometry (DEMS) measurement with isotope ¹⁸O-labeling

In situ DEMS measurements with isotope labeling were conducted to probe the involvement of catalyst-associated oxygen species during OER. The DEMS system was connected to an electrochemical cell operated at ambient pressure, allowing the gaseous OER products generated at the working electrode to be continuously transferred to the mass spectrometer for real-time analysis. For isotope labeling, the catalyst-coated electrodes were electrochemically treated at a constant current density of 50 mA cm^–2 in 0.5 M H₂SO₄ prepared with H₂¹⁸O, allowing ¹⁸O-labeled oxygen species to be incorporated onto or exchanged with the catalyst surface. After labeling, the electrodes were thoroughly rinsed with H₂¹⁶O to remove physically adsorbed H₂¹⁸O, while chemically incorporated or surface-bound ¹⁸O species remained on the catalyst. The in situ DEMS measurements were performed under a constant current density of 50 mA cm^–2 in a three-electrode configuration using H₂¹⁶O-containing H₂SO₄. During OER, the gaseous O₂ products were continuously monitored by mass spectrometry. The signals at m/z = 32, 34, and 36 were assigned to ³²O₂ (¹⁶O¹⁶O), ³⁴O₂ (¹⁸O¹⁶O), and ³⁶O₂ (¹⁸O¹⁸O), respectively. Before each measurement, the electrolyte was purged with high-purity Ar to remove dissolved oxygen. The isotope-labeled oxygen signals were used to evaluate the participation of catalyst-associated oxygen species in O–O bond formation during acidic OER.

PEMWE system operation

Membrane electrode assemblies (MEAs) were fabricated using a Nafion^® 117 (N117) membrane via the catalyst-coated substrate (CCS) method. Prior to fabrication, the membrane was pretreated to remove organic contaminants and ensure complete protonation by sequential immersion in 5 wt% H₂O₂ for 1 h, Milli-Q ultrapure water for 1 h, and 1 M H₂SO₄ for 1 h at 80 °C, followed by storage in distilled water until use. The N117 membrane was cut into 5 $\times$ 5 cm² with a thickness of 180 μm. The MEAs employed Pt/Ti fiber felt as the anode porous transport layer (PTL) and Ti fiber felt as the cathode gas diffusion layer (GDL). The OER catalyst was directly synthesized on the Pt/Ti felt anode via a thermal decomposition method, yielding a catalyst loading amount of ~8 mg cm^–2. To prepare the HER catalyst ink, Pt/C (20 wt%) was dispersed in a 1:1 (v/v) mixture of deionized water and ethanol. Nafion^® ionomer was added to achieve an ionomer content of 40 wt% relative to the cathode catalyst. After prolonged ultrasonication for several days, a homogenous Pt/C ink was obtained and sprayed onto the Ti felt cathode to achieve a catalyst loading amount of 1 mg cm^–2 (corresponding to 0.2 mg_Pt cm^–2). The MEAs were assembled by hot-pressing the cathode and anode onto Nafion^® 117 membrane at 120 °C for 3 min under a pressure of ~8 MPa. The geometric active area of the MEAs were 2 × 2 cm² (4 cm²). All electrolysis tests were conducted in a single-cell PEMWE configuration at 80 °C, with anolyte water fed into the anode at a flow rate of 1 mL min^–1 through tubing with an inner diameter of 0.1 mm.

Computational details

First-principles calculations were performed within spin-polarized density functional theory (DFT) using the vienna ab initio simulation package (VASP)²³. The generalized gradient approximation (GGA) in the form of the Perdew–BurKe–Ernzerhof (PBE) functional was employed to describe the exchange–correlation energy, together with projector augmented-wave (PAW) pseudopotentials²⁴. A plane-wave kinetic energy cutoff of 500 eV was used for all calculations. Structural relaxations were carried out until the Hellmann–Feynman forces on all atoms were below 0.01 eV Å⁻¹.

On-site Coulomb interactions were treated using the DFT + U approach, applying effective Hubbard parameters (U − J) of 4.0 eV for Mn 3d states^13,25. For Nd, a U value of 5.0 eV was employed, selected as a representative value within the range (4–6 eV) reported in previous studies, thereby ensuring a balanced and reliable description of the localized Nd 4 f states²⁶. Brillouin-zone integrations were performed using MonKhorst−Pack K-point meshes of 4×4×4 for α-Mn₂O₃, 4×4×6 for β-MnO₂, and 6×4×2 for R-MnO₂ for unit cell structure²⁷. All structural and electronic visualizations, including isosurfaces and charge-density plots, were generated using the VESTA software³⁶.

Reaction free energies were evaluated using the computational hydrogen electrode (CHE) model^37,38. Under standard conditions, the chemical potential of a proton–electron pair was referenced to 1/2G (H₂), thereby enabling electrochemical proton–electron transfer steps to be described with respect to gaseous H₂ and liquid H₂O. The free energy of each adsorbed intermediate was calculated as

G = E_DFT + E_ZPE – TS(1)

where E_DFT denotes the total energy obtained from VASP calculations, while E_ZPE and TS correspond to the zero-point energy and entropic correction terms, respectively, derived from vibrational frequency calculations.

Author contributions

J.R. and J.H.L. co-supervised the entire work. H.K. and J.R. conceptualized the fundamental idea and designed the experiments. H.K. conducted experiments and characterizations, and visualized data. H.K., S.P., and H.P. performed the in situ spectroscopic measurements, while J.H. and S.J.K. assisted with the in situ mass spectroscopic analysis. H.L. assisted with detailed characterizations, and Y.K. contributed to the setup of the electrolyzer device. J.J. and J.W. performed the computational DFT calculations, while S.G.K. provided guidance on the detailed computational analysis. H.K., J.J., J.H.L., and J.R. co-wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Hui Chen, Cong-Qiao Xu, Yonggui Zhao and Jun Zhong for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the Basic Science Research Program (2021R1A2C2013684), the Regional Leading Research Center (RLRC) (RS-2023-00217778), and the Nano & Material Technology Development Program (RS-2024-00404361) through the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT. This work was also supported by the InnoCORE program of the Ministry of Science and ICT (1.250022.01) and the Industrial Technology Innovation Program (RS-2025-06642983). This research was supported by Basic Science Research Program through the NRF funded by the Ministry of Education (RS-2024-00413257).

Data availability

The data supporting the findings of this study are available within this paper, its supplementary information, and the accompanying source data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77291-9.