Quantitative study of oxygen evolution reaction using LiNi_0.5Mn_1.5O₄ thin-film electrodes Open Access

Water electrolysis converts water into hydrogen and oxygen.^1–4 Ir and Ru oxides have attracted considerable attention as catalysts for oxygen evolution reactions (OER) because of their high catalytic activity and durability.^5–9 Because these noble metals are rare and expensive, it is highly desirable to develop alternatives composed of abundant 3d transition metals, such as Ni, Fe, and Co.^10–13 However, 3d transition metal oxide catalysts exhibit lower catalytic activity, that is, higher overpotentials, than Ir and Ru oxide catalysts.^14,15 Consequently, elucidating the OER mechanism is necessary to lower the overpotentials of 3d transition metal oxide catalysts.

Oxides containing Ni³⁺ exhibit high OER catalytic activity, as demonstrated in NiOOH,¹⁶ Ni_xFe_1−xOOH,^17–19 Ni_3−xCo_xO₄,^20,21 and LaNiO₃.^22–24 In contrast, oxides containing Ni²⁺, such as Ni(OH)₂, NiO, and NiFe₂O₄, exhibit lower OER activity than Ni³⁺-containing oxides. In Ni²⁺-containing oxides, Ni³⁺ is suggested to contribute during the catalytic cycle.²⁵ Therefore, understanding the chemical state of Ni is critical for developing Ni-oxide catalysts with high OER activity.

LiNi_0.5Mn_1.5O₄ (LNMO) is an intriguing compound that exhibits OER catalytic activity despite the formal charge of 2+ for Ni²⁶ and 4+ for Mn.^27,28 The oxidation state of Ni can be tuned from 2+ to 4+ by the insertion/removal of Li, indicating the potential for enhanced OER activity. To date, the OER catalytic activity of LNMO has only been studied in powder form, leading to less accurate current density values. The difficulties in quantitative studies using powders are as follows: (1) determining the separate contributions of various additives to the reaction is difficult, (2) the potential might be underestimated by the voltage drop owing to the low electrical conductivity of LNMO (∼10⁻⁶ Scm⁻¹)²⁹ and the grain boundaries, and (3) measuring the surface area of the powder is difficult. Thin-film electrodes are suitable platforms for quantitative evaluations. LNMO thin films fabricated on a flat current collector have the following advantages: (1) only the catalyst surface is exposed, (2) the voltage drop is minimized by uniformly applying a potential from the current collector, and (3) the surface area can be defined if the surface roughness is small. However, there are no reports on the quantitative evaluation of OER activity using LNMO thin-film electrodes.

In this paper, we report the quantitative evaluation of the OER using LNMO thin-film electrodes. LNMO thin films deposited on an Au current collector exhibit a current density of 6.6 and ∼2.6 mA cm⁻² for geometric and roughness-modeled areas, respectively, at 1.78 V vs. a reversible hydrogen electrode (V_RHE). The current density using the estimated area is comparable to that of the previously reported value for the powder form.²⁷ X-ray photoelectron spectroscopy (XPS) revealed the presence of Ni³⁺ and Mn²⁺ in both the as-grown and post-OER LNMO thin films. These results suggest the importance of Ni³⁺ originating from the electron transfer between Ni²⁺ and Mn⁴⁺.

The LNMO thin-film electrode catalyst was fabricated using the following process. A direct-current magnetron sputtering method was employed to deposit a current collector (30 nm Ti layer and 100 nm Au layer) on an Al₂O₃(0001) substrate at room temperature. Using a radio-frequency (RF) magnetron sputtering method, a 120-nm thick LNMO thin film was deposited on a current collector using radio-frequency magnetron sputtering. A target with an excess Li composition of Li_1.2Ni_0.5Mn_1.5O₄ (diameter: 48 mm, Toshima Manufacturing Co., Ltd.) was used to compensate for the Li deficiency.^30,31 During sputtering, the substrate temperature, RF power, and deposition time were set at 500 °C, 100 W, and 60 min, respectively. The total pressure and total flow rate during RF sputtering were fixed at 1.0 Pa and 20.0 SCCM, respectively. Ar and O₂ gases were introduced at concentrations of 98% and 2.0%, respectively.

The structural and compositional characterization of the fabricated films was performed using x-ray diffraction [XRD, Bruker, D8 DISCOVER, Cu Kα rays (wavelength: 0.15406 nm)], Raman spectroscopy (HORIBA, XploRA PLUS, wavelength: 532 nm), scanning electron microscopy (SEM, Hitachi High-Tech, S-5500), and energy-dispersive x-ray analysis (EDX, Bruker AXS, XFlash® 5010). The electronic states of the thin-film surfaces were investigated using XPS (ULVAC-PHI, VersaProbeIII PHI5000). The water droplets on the LNMO thin-film electrode retrieved from the electrolyte were removed using an air duster to perform XPS. The LNMO thin film was then dried overnight (∼8 h) in a vacuum desiccator. The samples were subsequently transferred to the XPS apparatus in ambient air.

The electrochemical measurements were performed using a three-electrode electrochemical cell. The electrolyte used was a 0.1M KOH aqueous solution prepared by adding KOH (Guaranteed Reagent, FUJIFILM Wako Pure Chemical Corp.) to ultrapure water (>19.0 MΩ cm, Merck Millipore, Direct-Q UV3). Fe impurity removal was not performed. Fe was not detected by XPS on the surface after OER (Fig. S1 in the supplementary material). We believe that the effect of impurities below a detection sensitivity (∼1%) is negligible. The LNMO thin films on a current collector, Pt coil, and Ag/AgCl were used as the working, counter, and reference electrodes, respectively. Linear sweep voltammetry (LSV), cyclic voltammetry, chronopotentiometry, and electrochemical impedance spectroscopy (EIS) were performed. The geometric surface area of the working electrode was determined through image processing using ImageJ software.³² The quantification of oxygen evolution through chronopotentiometry was conducted using gas chromatography (Shimadzu Corp., GC-8A). Tafel analysis was performed based on the results of the LSV. EIS measurements were performed with a sinusoidal voltage variation of 20 mV over the frequency range of 100 mHz to 10 MHz at 1.70, 1.75, 1.80, and 2.00 V_RHE. The potential drop during measurement was compensated by an automatic function in the instrument (Biologic SP-50) during EIS measurements. A sinusoidal voltage of 100 kHz and 10 mV amplitude was used to compensate for a feedback ratio of 85%. The results were analyzed using pyZwx software.³³ Initially, the two-pole logarithmic Hilbert transform method³⁴ was used to exclude measurement points that did not meet the causality and stability criteria. Subsequently, fitting was performed using the equivalent circuit models, as discussed later. Additionally, the uncompensated resistance of the electrochemical cell was corrected by the EIS results.

First, we characterize the structures of the fabricated films [Fig. 1(a)]. The out-of-plane XRD patterns in Fig. 1(b) show the diffraction from the spinel-type LNMO and the diffraction from the Au current collector. The two-dimensional XRD pattern of LNMO is elongated in the ψ direction, indicating that the film is a weakly (111)-oriented LNMO polycrystalline thin film. The Raman spectrum also confirms the formation of LNMO [Fig. 1(c)], where the peaks at ∼494 and ∼610 cm⁻¹ correspond to the F_2g and A_1g modes of LNMO, respectively.³⁵ The significant background observed in the Raman spectrum was due to the reflection of the excited laser to the Au current collector underneath. These results confirm the formation of a weakly oriented polycrystalline LNMO thin film on the Au current collector.

SEM and EDX measurements confirm the uniformity of the crystal grain structure and composition of the thin film [Fig. 1(d)]. The SEM image shows crystal grains approximately 50–100 nm in size. Additionally, the EDX mapping of each element shows a homogeneous distribution of Ni, Mn, and O. Table S1 in the supplementary material presents the results of elemental composition measurements using EDX. The Mn/Ni ratio shows a value close to stoichiometry, which is approximately 3.5. The signal of Al originating from the substrate is also detected. This detection makes it challenging to quantify the O content within the LNMO films. We note that Li is not detectable by EDX.

The LSV curve of the device structure [Fig. 2(a)] shows the increase in the current density at ∼1.6 V_RHE, which reaches 6.60 mA cm⁻² at 1.78 V_RHE; here, the reaction area is a geometric area estimated from a photograph [Fig. 2(a)]. Considering the increase in real area due to surface roughness, the current density may decrease to 0.4 times. We establish two types of surface three-dimensional models to estimate the real area: a rectangular prism and a square-pyramid-shaped model [see Figs. S2(a) and S2(b) in the supplementary material]. The arithmetic mean roughness (R_a) was used as the evaluation scale for surface roughness. The R_a of the LNMO thin film was 35 nm through atomic force microscopy measurements. Hence, the roughness heights were set to match R_a. Compared to the directly measured simple geometric surface areas in this study, the areas in Figs. S1(a) and S1(b) in the supplementary material are 2.4 and 2.7 times larger, respectively. As a result, the estimated current density is ∼2.6 mA cm⁻². In such a scenario, the oxygen evolution current density of the LNMO thin film electrode is comparable to that of the octahedral LNMO powders reported previously, 2.92²⁷ and 0.14²⁸ mAcm⁻² at 1.78 V_RHE. The turnover frequency (TOF) was calculated to be 2.7 × 10 s⁻¹. From the electrical current in the vicinity of 1.3 V_RHE (corresponding to the change from Ni²⁺ to Ni³⁺), the number of electrons transferred is ∼1.6 × 10¹³, which corresponds to the number of active sites.³⁶ From the current density (6.6 mA cm^-2) at 1.78 V_RHE and electrode area (0.10 cm²), we obtain TOF = 2.7 × 10 s⁻¹. We note that the use of Ag/AgCl in 0.1M KOH may induce an electrode liquid junction potential of approximately 30 mV.^37,38

Subsequently, we investigated the rate-determining step of the OER from the Tafel slope³⁹ determined from the current densities ranging from 0.5 to 10 mA cm⁻². We obtained a Tafel slope of 69.8 mV dec⁻¹ [Fig. 2(b), inset], which is in good agreement with the previously reported value for octahedral LNMO powders comprising (111) facets (70 mV dec⁻¹).²⁷ Theoretically, the Tafel slope of ∼70 mV dec⁻¹ suggests that the conversion of adsorbed OH species to O²⁻ is the rate-limiting step of OER.³⁹ Accordingly, OH species adsorbed on the LNMO thin film may slowly convert to O²⁻.³⁹ It should be noted that when the current values are large >100 mAcm⁻², steady-state measurements are essential. However, since the current values in this study are small (0.5–5 mA cm⁻²), we believe that the error of Tafel analysis noted in the literature⁴⁰ is negligible. As for the potential used to calculate the Tafel gradient, it exceeds the equilibrium potential by more than 120 mV.⁴¹ In the present study, we conclude that the error is negligible because it does not fall into the critical region.

Chronopotentiometry was performed to evaluate the durability of the LNMO catalysts [Fig. 2(c)]. The observed potential increase over 90 h was 0.18 V, indicating sufficient durability of the catalyst. Furthermore, the stable evolution of oxygen during chronopotentiometry is confirmed by gas chromatography (Fig. S3 in the supplementary material) and Faraday efficiency is ∼90%.

The crystal structure of LNMO remains unchanged before and after the OER. The changes in the peak structure of the XRD and Raman spectroscopy [Figs. 3(a) and 3(b), respectively] are absent before and after the OER, showing that LNMO maintains its spinel-type structure during and after the OER with no degradation or dissolution. We speculate that the decrease in the 111 peak intensity is due to Li desorption from LNMO in the OER process. The Li 1s XPS shows the decrease in Li intensity after OER (Fig. S4 in the supplementary material), indicating that Li desorbed from the surface. We believe that the crystallinity has decreased and the diffraction intensity has also decreased because of the Li desorption. SEM images [Fig. 3(c)] also indicate no change in the LNMO electrode surface before and after the OER. Therefore, the structure of LNMO films remains intact during the OER, and the electrochemical measurements in this study reflect the intrinsic activity of LNMO.

To investigate the surface electronic states of the OER, we performed XPS (Fig. 4). The O 1s spectrum consists of two peaks at 529.4 and 531.0 eV [Fig. 4(a)], attributed to lattice oxygen and chemisorbed water on the surface, respectively.⁴² After the OER, an increase in the peak corresponding to chemisorbed water suggests an increase in the amount of adsorbed water molecules on the LNMO surface after the immersion of the samples in the KOH aqueous solution.

The Mn 2p spectra [Fig. 4(b)] indicate negligible changes before and after the OER measurements. The XPS spectra exhibit two peaks at 653.5 and 641.8 eV, corresponding to Mn 2p 1/2 and Mn 2p 3/2, respectively,⁴³ along with a broad tail centered at ∼636.4 eV. In particular, the broad tail at ∼636 eV is lower than the binding energy observed for Mn³⁺,⁴⁴ suggesting a lower valence state of Mn²⁺.⁴⁵ 

The Ni 2p 3/2 spectra [Fig. 4(c)] comprise three peaks at 861.2, 856.0, and 854.8 eV, assigned to the satellites from the shake-up process, Ni³⁺, and Ni²⁺, respectively.²⁶ Surprisingly, we confirm the presence of a formal charge of Ni³⁺ even in the as-grown LNMO. Similar to the XPS spectra of Mn, no changes were observed in the peak positions or relative peak areas of Ni before and after the OER measurements. Therefore, the presence of Ni³⁺ on the LNMO surface may have contributed to the observed OER activity. The origins of Ni³⁺ states remain unclear. It is essential to determine the valence states within thin films. Given that XPS is surface-sensitive, a future challenge lies in evaluating the interior using hard XPS or x-ray absorption spectroscopy, which have larger probing depths.

We estimated the proportions of Mn²⁺ and Ni³⁺ using the ratio of the peak areas (I) in the XPS spectra. Here, we denote the fraction of Mn²⁺ as x (the fraction of Mn⁴⁺ corresponds to 1 − x). The value of x/(1 − x) is approximately equal to I_Mn2+/I_Mn4+ ∼ 0.06, yielding x ∼ 0.057. Here, I_Mn2+ and I_Mn4+ are the areas of the fitted Gaussian functions at around 636.4 and 641.8 eV, respectively. This indicates that ∼5.7% of the original Mn⁴⁺ was converted to Mn²⁺ and that 0.16 electrons were accepted in Mn per chemical formula unit of LNMO. The fraction of Ni³⁺ (y) was similarly evaluated. The value of y/(1 − y) was approximately equal to I_Ni3+/I_Ni2+ ∼ 0.56, yielding y ∼ 0.36. Here, I_Ni3+ and I_Ni2+ are the areas of the fitted Gaussian functions at around 855.5 and 854.4 eV, respectively. These evaluations suggest that 0.17 electrons are donated from Ni per chemical formula unit of LNMO. The value of donated electrons from Mn (0.16 e⁻) is in good agreement with the accepted electrons in Ni (0.17 e⁻), suggesting an electron transfer from Mn and Ni in the LNMO thin films.

The binding energy shift observed in LNMO is about 0.3 eV. Adsorption of K is suspected as the reason for this shift. Indeed, when LNMO is immersed in an electrolyte solution, K in the electrolyte is detected on its surface (Fig. S5 in the supplementary material). Electron doping by K adsorption or Li/K exchange near the surface may induce a shift in the binding energies.

To examine the OER mechanism on LNMO, we conduct EIS at 1.70, 1.75, 1.80, and 2.00 V_RHE. Figures 5(a) and 5(b) show the Nyquist plots, indicating three distinct resistance components: the uncompensated resistance corresponding to the real-axis intercept on the high-frequency side and two resistance components from the capacitive semicircle. The uncompensated resistance primarily consists of the wiring and electrolyte resistances.

The Nyquist plots were fitted to three candidate equivalent circuit models (X, Y, Z; see Table S2 in the supplementary material).⁴⁶ Fitting was conducted using the Levenberg–Marquardt method within a confidence region (weight: absolute value of impedance; power factor: 2.0), resulting in calculated values within a certain range and within the confidence region. Models Y and Z exhibited relative errors exceeding 100% (Table S2 in the supplementary material). In contrast, model X exhibited errors of approximately 30% or less, except for 2.0 V_RHE, where the error remains within 60%. Considering that a single model should be applicable at any potential in electrode reactions, model X was determined to be the most appropriate in this study.

The curves of the parameters obtained by fitting model X (Table I) are depicted as solid lines in Figs. 5(a) and 5(b). The equivalent circuit Model X [Fig. 5(a), inset] consisted of three resistance components (R_s, R_a, R_c) and two pseudocapacitance components (CPE_d and CPE_a). Note that CPE ≡ A(jω)^−α is defined, where A is the CPE constant, α is the fractal parameter, ω is the frequency, and j is the imaginary unit. When α ∼ 1, the CPE value is equivalent to the capacitance, whereas when α ∼ 0, the CPE value is equivalent to the resistance. According to previous reports,⁴⁷ the resistance components (R_s, R_a, and R_c) and two pseudocapacitance components (CPE_d and CPE_a) correspond to the uncompensated resistance (R_s), resistance derived from the adsorbed OH intermediates (R_a), charge transfer resistance at the thin film and electrolyte interface (R_c), pseudocapacitance of the electric double layer (CPE_d), and pseudocapacitance derived from the adsorbed OH intermediates (CPE_a).

The phase vs. frequency plot (Fig. S6 in the supplementary material) exhibits a peak at 10 Hz when 1.70 V_RHE. The phase angle approaches 0° with increasing potential. These trends are similar to those observed in previous impedance measurements for alkaline electrolysis using Ni electrodes.⁴⁸ Therefore, we discuss the impedance results based on the interpretations in Ref. 47 as follows.

First, R_s remained nearly constant and independent of the applied potential. Considering that the setup of the water electrolysis cell was uninfluenced by the applied potential, this is reasonable. In contrast, the values of R_a decrease significantly with increasing potential and R_c decreases beyond 1.75 V_RHE [Figs. 5(c), and 5(d), respectively]. This trend is analogous to that reported in Ref. 46. According to this literature, not only the charge transfer resistance at the thin film and electrolyte interface (R_c) but also the contribution of the charge transfer resistance to adsorbed OH intermediates (R_a) is highlighted. Therefore, for the LNMO thin-film electrode, a similar suggestion was made regarding the contribution of the charge-transfer resistance to the adsorbed OH species.⁴⁶ This agrees with the Tafel plot results, indicating that the conversion of the adsorbed OH intermediates is the rate-determining step [Fig. 2(b)]. Consequently, reducing the charge-transfer resistance to the adsorbed OH intermediates at lower applied potentials may lead to enhanced OER activity.

Finally, we discuss the constant-phase elements. With an increase in potential, the capacitance of the electric double layer (CPE_d) significantly increases, while that of the constant-phase element associated with adsorption (CPE_a) decreases (Table I). In particular, the reduction in CPE_a was similarly observed in the alkaline water electrolysis of Fe oxides.⁴⁸ According to this literature, CPE_a appears owing to the adsorption of OH intermediates, and at high potentials, these intermediates almost entirely cover the surface (i.e., a coverage rate of 1). Therefore, for the LNMO thin-film electrode, we assume that at potentials above 1.75 V_RHE, the adsorbed OH intermediates almost completely cover the surface, ensuring a sufficient supply of reactants to the reaction sites. Consequently, we estimate that the adsorbed OH intermediates are adequately supplied to the reaction active sites at potentials above 1.75 V_RHE, leading to reduced R_a. In contrast, at 1.70 V_RHE, the significant value of CPE_a suggests that OH adsorption intermediates are not adequately supplied, resulting in a reduced OER current.

We quantitatively evaluated the OER using LNMO polycrystalline thin films as electrode catalysts. XPS measurements indicate the presence of Ni³⁺ and Mn²⁺ on both the as-grown and post-OER LNMO surfaces, suggesting the formation of Ni³⁺ due to the charge transfer between the Ni and Mn ions. In addition, we discussed the reaction mechanism through electrochemical impedance spectrometry measurements. In future studies, higher OER activity is anticipated in Li_1−xNi_0.5Mn_1.5O₄ (x = 0.00–0.50) with high-valence Ni electronic states by the Li desertion.

See the supplementary material for details on the equivalent circuit model and a conceptual diagram of the surface model. The supplementary material also includes the results of EDX, XPS, and oxygen evolution reactions.

This study was supported by JSPS Kakenhi (Grant Nos. 18H03876, 21K18892, 21H00141, 22H01953, 17H05216, 19H02596, 19H04689, 18H03876, and 18H05514), the JST-PRESTO (Grant No. JPMJPR17N6), JST-CREST (Grant No. JPMJCR1523) programs, and DX-GEM (Grant No. JPMXP1122712807). K.H. acknowledges funding from the Advanced Human Resource Development Fellowship for Doctoral Students (Tokyo Institute of Technology). XPS measurements of this work were supported by “Advanced Research Infrastructure for Materials and Nanotechnology in Japan (ARIM)” of the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Grant No. JPMXP1223UT0202. We are also grateful to the platform of Analytical Instruments for Chemistry and Materials Science (Tokyo Institute of Technology) for providing the equipment for the Raman and SEM measurements. We are thankful to Professor Takanabe and Assistant Professor Obata for the support of gas chromatography.

The authors have no conflicts to disclose.

Kentaro Hatagami: Conceptualization (lead); Investigation (lead); Writing – original draft (lead); Writing – review & editing (equal). Kazunori Nishio: Writing – review & editing (supporting). Ryota Shimizu: Conceptualization (supporting); Writing – review & editing (equal). Taro Hitosugi: Funding acquisition (lead); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.