Modifying redox properties and local bonding of Co3O4 by CeO2 enhances oxygen evolution catalysis in acid

Developing efficient and stable earth-abundant electrocatalysts for acidic oxygen evolution reaction is the bottleneck for water splitting using proton exchange membrane electrolyzers. Here, we show that nanocrystalline CeO2 in a Co3O4/CeO2 nanocomposite can modify the redox properties of Co3O4 and enhances its intrinsic oxygen evolution reaction activity, and combine electrochemical and structural characterizations including kinetic isotope effect, pH- and temperature-dependence, in situ Raman and ex situ X-ray absorption spectroscopy analyses to understand the origin. The local bonding environment of Co3O4 can be modified after the introduction of nanocrystalline CeO2, which allows the CoIII species to be easily oxidized into catalytically active CoIV species, bypassing the potential-determining surface reconstruction process. Co3O4/CeO2 displays a comparable stability to Co3O4 thus breaks the activity/stability tradeoff. This work not only establishes an efficient earth-abundant catalysts for acidic oxygen evolution reaction, but also provides strategies for designing more active catalysts for other reactions.

T he fast depletion of fossil fuels and increasing greenhouse effect demand sustainable strategies to produce carbonneutral fuels using renewable electricity 1 . Electrocatalytic water splitting has been considered a promising approach to generate hydrogen as a clean and renewable energy carrier 2 . Proton exchange membrane (PEM) electrolyzers operated in acidic media have shown great promises for large-scale applications [3][4][5] . Despite substantial recent advances in the discovery of robust and active earth-abundant electrocatalysts for acidic hydrogen evolution reaction (HER) 1,6-8 , the development of high-performance yet cost-effective electrocatalysts for the sluggish four-electron oxygen evolution reaction (OER) is challenging [9][10][11] especially in acidic media, which contributes to a major energy loss in the overall water splitting process and is a bottleneck for realizing practical PEM electrolyzers 3,12 . Most OER catalysts show inferior activities in acidic media compared to in alkaline media and require higher overpotentials to achieve comparable catalytic current densities. Moreover, the stability issues are more severe in acidic OER, and even noble metal-based catalysts (such as RuO 2 and IrO 2 ) experience dissolution and degradation 13,14 . Furthermore, the often observed tradeoff between activity and stability in acidic OER catalysts [13][14][15][16] complicates the catalyst design. As a result, there have been very limited choices of earth-abundant OER catalysts that are both active and stable in acidic media [17][18][19][20] . Cobalt (Co)-based catalysts such as Ba[Co-POM] 17 , hetero-N-coordinated Co single atom catalyst 21 , CoFePbO x 18 , Co 2 TiO 4 22 , and Co 3 O 4 23-25 are promising for acidic OER; however, the mechanistic details have rarely been studied for these emerging OER catalysts in acidic media.
The active site structures and catalytic mechanisms of cobalt oxide OER catalysts have been primarily investigated in alkaline and neutral media [26][27][28][29][30][31] , little is known about these catalysts in acidic media. The exact configuration of the active sites responsible for the O-O bond formation still remains debatable, but the generation of high-valence-state Co IV is accepted to be involved in the pre-OER redox processes of different types of cobalt oxide OER catalysts since they share the common active sites 26,31,32 . The further oxidation of the neighboring Co redox centers to form dimeric Co IV Co IV takes place at high potentials 33,34 , and thus causes a large energy loss to bypass this potentialdetermining process for the catalytic OER 31 . Besides, these prominent pre-OER redox features also suggest that the Co IV Co IV intermediates are stabilized and could suffer from a slow catalytic turnover process for OER 35,36 . Therefore, a better understanding of the relationships between redox properties and catalytic activity is the key to design more efficient (Co-based) OER catalysts and to enhance catalytic activity by regulating redox properties, which remains elusive and largely underexplored especially in acidic media.
In this work, we enhance the intrinsic catalytic activity of Co 3 O 4 by introducing nanocrystalline CeO 2 to form a heterogeneous Co 3 O 4 /CeO 2 nanocomposite and establish Co 3 O 4 /CeO 2 nanocomposite as an active acidic OER catalyst. CeO 2 has been well documented as (co-)catalyst in thermal catalysis due to its excellent redox properties and oxygen storage capacity 37 . Although CeO 2 has been introduced into a number of electrocatalyst systems to enhance the overall performance for various electrocatalytic reactions 38 including the alkaline OER [39][40][41] , how it impacts the catalytic activity remains controversial and its contribution to the redox properties of the electrocatalysts has not yet been discussed. Now we show that the introduction of CeO 2 (meaning phase-pure CeO 2 nanocrystallites are interdispersed among phase-pure Co 3 42 , which is further proved by the powder X-ray diffraction (PXRD) pattern of Co 3 O 4 /CeO 2 (Fig. 1e). Selected area electron diffraction patterns of both samples displayed similar diffraction rings due to the polycrystalline nature (insets of Fig. 1a, b). The inner to outer diffraction rings can be indexed to the (111), (220), (311), (400), (511), (440) planes of Co 3 O 4 (JCPDS 43-1003), consistent with the PXRD patterns (Fig. 1e) and the spinel oxide crystal structure of Co 3 O 4 (Fig. 1f) 43 . The introduction of CeO 2 decreased the crystallinity of Co 3 O 4 , as the average crystalline domain sizes of Co 3 O 4 and Co 3 O 4 /CeO 2 estimated from the widths of the (311) diffraction peaks using the Scherrer equation were 13.9 and 9.7 nm, respectively (Supplementary Fig. 5). From the HRTEM images (Fig. 1c, d), the lattice spacings of 0.243 and 0.467 nm were assigned to the (311) and (111) planes of Co 3 O 4 , respectively, and that of 0.312 nm was attributed to the (111) plane of CeO 2 . Nanoscale crystallites of CeO 2 exhibit an average domain size of~5 nm based on the Scherrer analysis of the PXRD peak ( Supplementary Fig. 6) and are evenly dispersed among phase-pure Co 3 O 4 crystallites with numerous interfacial contact regions. Elemental mappings further confirmed the successful introduction of Ce in Co 3 O 4 /CeO 2 (Fig. 1g). The bulk and surface Ce metal contents in Co 3 O 4 /CeO 2 [defined as Ce/(Ce + Co) × 100%] were determined as 9.1 and 6.6 atomic percent (at%) using energy-dispersive X-ray spectroscopy (EDS) and X-ray photoelectron spectroscopy (XPS), respectively (Supplementary Table 1).
Electrocatalytic properties of Co 3 O 4 /CeO 2 nanocomposites. The substantial differences in the redox properties and acidic OER catalytic performances between the Co 3 O 4 and Co 3 O 4 /CeO 2 catalysts on FTO electrodes are shown by cyclic voltammetry (CV) recorded in 0.5 M H 2 SO 4 solution (Fig. 2a). Three sets of pre-OER redox features are observed in Co 3 O 4 (the corresponding cathodic peaks are denoted as C1, C2, and C3 in the order of increasing potential, see Fig. 2b), which can be ascribed to the following equilibria involving dimeric Co redox centers 26,31,33 : Co II Co III ↔ Co III Co III ↔ Co IV Co III ↔ Co IV Co IV (see proposed detailed structural motifs in Supplementary Fig. 7). In contrast, Co 3 O 4 /CeO 2 displayed no obvious pre-OER redox features and a much lower onset potential for acidic OER ( Fig. 2a and Supplementary Fig. 8b), suggesting the redox properties of Co 3 O 4 can be effectively regulated by the introduction of CeO 2 .
Note that CeO 2 itself shows no redox feature and very poor activity toward OER in acid ( Supplementary Fig. 9). The Co 3 O 4 / CeO 2 catalyst prepared by introducing a nominal 10 at% Ce metal content during the electrodeposition process exhibited the highest acidic OER catalytic performance ( Supplementary Fig. 10) and was therefore studied in the rest of this work. The overpotentials required for Co 3 O 4 and Co 3 O 4 /CeO 2 (10 at% Ce) to reach a geometric catalytic current density of 10 mA cm -2 on FTO electrodes were 507 ± 5 and 423 ± 8 mV, respectively, showing a substantial improvement of~84 mV after the introduction of CeO 2 (Fig. 2a inset). The Tafel slopes of the acidic OER on Co 3 O 4 and Co 3 O 4 /CeO 2 were 110.8 and 88.1 mV dec -1 , respectively (Fig. 2c). Both are in the range of 60-120 mV dec -1 , indicating a mixed kinetic control mechanism 44 . A second linear Tafel region was observed in Co 3 O 4 (in the overpotential range of 350-425 mV shaded in pink), which originates from the chargeaccumulation process due to the oxidation of dimeric Co IV Co III to Co IV Co IV . In contrast, Co 3 O 4 /CeO 2 only exhibits a single linear Tafel region with a smaller slope of 88.1 mV dec -1 , which suggests that the OER catalytic onset takes place at a much lower overpotential of~300 mV without noticeable charge accumulation of dimeric Co redox centers.
The intrinsic acidic OER catalytic activities of Co 3 O 4 and Co 3 O 4 /CeO 2 catalysts on FTO electrodes were further extracted based on double-layer capacitance (C dl ) measurements and electrochemically active surface area (ECSA) normalization. The C dl values of Co 3 O 4 (7.31 mF cm -2 ) and Co 3 O 4 /CeO 2 (23.26 mF cm -2 ) ( Supplementary Fig. 11) showed that the introduction of CeO 2 substantially increased the ECSA. Nevertheless, after normalizing the geometric catalytic current density by the ECSA derived from C dl (see Methods for details) 45 , Co 3 O 4 /CeO 2 still displayed a much lower OER catalytic onset potential than Co 3 O 4 and a much higher ECSA-normalized catalytic current density of 23.7 μA cm -2 at the overpotential of 450 mV, which doubled that of Co 3 O 4 at the same overpotential (Fig. 2d). These results confirm that Co 3 O 4 /CeO 2 features enhanced intrinsic OER catalytic activity compared to Co 3 O 4 in acidic media.
We further examined the electron transfer kinetics of Co 3 O 4 and Co 3 O 4 /CeO 2 catalysts on FTO electrodes using electrochemical impedance spectroscopy (EIS) at different potentials and extracted the charge transfer resistance (R ct ) of the catalytic OER from EIS fitting using the Voigt circuit model (Supplementary Fig. 12 and Supplementary Table 2) 46 . At the potentials between 1.566 and 1.616 V vs. reversible hydrogen electrode (RHE), the charge accumulation process due to the oxidation of dimeric Co redox centers dominated on the Co 3 O 4 catalyst, whereas the catalytic OER already took place on the Co 3 O 4 /CeO 2 catalyst. As a result, the R ct values of Co 3 O 4 were one order of magnitude higher than those of Co 3 O 4 /CeO 2 (Supplementary Table 2). Once OER dominated on Co 3 O 4 after the oxidation of dimeric Co IV Co III to Co IV Co IV at the higher potential of 1.716 V vs. RHE, its R ct substantially decreased to be on the same order of magnitude as that of Co 3 O 4 /CeO 2 (Supplementary Table 2). These EIS results suggest that the catalytic OER on Co 3 O 4 takes place efficiently only after overcoming the sluggish kinetic step associated with the charge accumulation process to form dimeric Co IV Co IV , and the introduction of CeO 2 effectively regulates the redox properties of Co 3 O 4 and substantially enhances the electron transfer kinetics of the catalytic OER at a much lower overpotential.
We further verified that the enhanced catalytic activity of Co 3 O 4 /CeO 2 could not be attributed to the decreased crystallinity of Co 3 O 4 due to the introduction of CeO 2 (see earlier discussions of Fig. 1e and Supplementary Fig. 5 Fig. 14c, d), indicating the catalytic activity enhancement in Co 3 O 4 /CeO 2 originates from the regulated redox properties rather than sample crystallinity. To shed light on the pre-OER redox mechanisms of Co 3 O 4 and understand their relationships to the catalytic activity, we conducted pH-dependence analysis of the C3 peak on the Co 3 O 4 catalyst in H 2 SO 4 solution in the pH range of 0.48-1.24 ( Fig. 3a and Supplementary Fig. 15a). The peak potential vs. standard hydrogen electrode was plotted against the solution pH ( Fig. 3a inset). The slope of 95.9 ± 4.8 mV per pH unit suggests a 2 e -/3 H + coupled redox process 47 , which is different from the 59 or 120 mV per pH unit expected for a 1 e -/1 H + or 1 e -/2 H + process, respectively 48 . In addition, CV curves of Co 3 O 4 recorded at different scan rates in 0.5 M H 2 SO 4 solution ( Fig. 3b and Supplementary Fig. 16) reveal the first-order power law relationship between the three cathodic peak current densities and the scan rate ( Fig. 3b inset), suggesting that the C3 peak is associated with a surface capacitive process 49,50 . Thus, this crucial third redox feature of Co 3 O 4 corresponds to a 2 e -/3 H + surface capacitive process of Co IV Co III ↔ IV Co IV , consistent with the proposed structural motifs in Supplementary Fig. 7. Moreover, this prominent 2 e -/3 H + redox feature of Co 3 O 4 also indicates that the dimeric Co IV Co IV intermediate is partially stabilized and therefore cannot undergo a rapid catalytic turnover process to produce O 2 and return to the lower valence resting states 34,51 , thus resulting in an increased overpotential to drive the catalytic reaction 35,36 . In contrast, the absence of this pre-OER redox feature in Co 3 O 4 /CeO 2 suggests that the introduction of CeO 2 effectively destabilizes the dimeric Co IV Co IV intermediate and accelerates the catalytic turnover process, which leads to the enhanced acidic OER activity of the nanocomposite catalyst.
Since the oxygen source for acidic OER is H 2 O, the cleavage of HO-H bond and the proton transfer properties are important factors that could affect the catalytic activity, similar to the case of alkaline HER 52 . Therefore, we collected the CV curves of both  Fig. 3c) 34,53 . To separate the KIE from the reaction thermodynamics, linear sweep voltammetry curves were presented on the overpotential scale, and the KIE value was calculated based on the catalytic current density in the protonic vs. deuteric solution at the same overpotential (Fig. 3d, also see Methods for details). For both catalysts, the KIE values in OER potential regions fluctuated around the upper limit of secondary KIE (~1.5) with the absence of primary KIE, indicating that proton transfer is not rate-limiting for the acidic OER on both catalysts 34,53 . In addition, the pH-dependence analysis of the catalytic current densities at fixed overpotentials showed that the reaction order with respect to pH is close to zero on the RHE scale for acidic OER on both catalysts ( Supplementary Fig. 15), indicating the catalytic reaction is less dependent on the proton concentration in the electrolyte for both catalysts. These results suggest that the enhanced acidic OER activity of Co 3 O 4 /CeO 2 is unrelated to the proton transfer properties of the nanocomposite.
We further conducted temperature-dependent kinetic analysis of both Co 3 O 4 and Co 3 O 4 /CeO 2 catalysts to extract the apparent activation energy (E app ) and pre-exponential factor (A app ) of the acidic OER and to examine how the introduction of CeO 2 affects the catalytic mechanism. CV curves of both catalysts on FTO electrodes were recorded in 0.5 M H 2 SO 4 solution in the temperature range of 25-65°C ( Supplementary Fig. 18). As expected, the catalytic performances of both catalysts increased at elevated temperatures ( Fig. 3e and Supplementary Fig. 18). The E app values of both catalysts at fixed overpotentials were calculated from the Arrhenius equation (Fig. 3f and Supplementary Fig. 19, also see Methods for details) 54,55 . To completely capture the potential-dependent evolution of E app , the analysis was performed both below and above the catalytic onset potential. On both catalysts, the E app value reached its maximum around the respective catalytic OER onset potential (Fig. 3f) onsets (Fig. 3f), while more obvious differences are observed in the A app (Supplementary Fig. 20). The similar E app suggests that the introduction of CeO 2 does not alter the rate-determining step and the kinetic barrier for the formation of reaction intermediates, but rather enhances the intrinsic activity of the same type of catalytic active site in Co 3 O 4 by modifying the entropy of activation (i.e., the number of active intermediates that enter the rate-determining step) and the interfacial concentration of active sites, as higher A app is extracted for Co 3 O 4 /CeO 2 at the same overpotential [56][57][58] . Therefore, these KIE, pH-and temperaturedependence analyses exclude several other factors, so we attribute the enhanced acidic OER activity to the regulation of the redox properties in Co 3 O 4 /CeO 2 resulted from the modified local bonding environment, as explained below.  Fig. 22a, d). Ultraviolet photoelectron spectroscopy (UPS) (Supplementary Fig. 23) showed larger work function in Co 3 O 4 /CeO 2 than pure Co 3 O 4 , suggesting the electronic structure in Co 3 O 4 /CeO 2 was slightly modified due to possible electronic interactions between Co 3 O 4 and CeO 2 . XAS is more sensitive to subtle changes in the oxidation states and the local bonding environments throughout the nanocomposite samples. According to the relative absorption edge positions in the Co K-edge X-ray absorption near-edge spectra (Fig. 4a), the Co 3 O 4 /CeO 2 exhibited a slightly higher Co oxidation state than the as-synthesized Co 3 O 4 , and the Co oxidation states in both catalysts increased and became similar after OER testing (inset of Fig. 4a). The absorption edge energies were further determined by an integral method 59 Fig. 24) does not necessarily result in changes in the pre-OER redox features ( Supplementary  Fig. 14a). Besides the higher Co oxidation state, the changes in local bonding environment of Co 3 O 4 induced by CeO 2 were also observed, as revealed by extended X-ray absorption fine structure (EXAFS) (Fig. 4c and Supplementary Fig. 25). Fourier transforms of k 3 -weighted Co K-edge EXAFS spectra of both Co 3 O 4 and Co 3 O 4 /CeO 2 catalysts displayed three major signals associated with the Co-O, Co-Co oct (octahedral site), and Co-Co tet (tetrahedral site) scattering paths (Fig. 4c). Compared to the assynthesized Co 3 O 4 (Fig. 4c red trace), a shorter Co-O bond distance was observed in the Co 3 O 4 /CeO 2 (Fig. 4c blue trace) due to the higher positive charge density at the Co centers 61 after the electron redistribution from Co 3 O 4 to CeO 2 , as illustrated in the bottom scheme in Fig. 4d. More importantly, the bond distances in Co 3 O 4 /CeO 2 remained the same after OER testing (Fig. 4c light blue trace), and the crystal structure barely changed, as shown by the identical intensity ratio of Co-Co oct and Co-Co tet scattering paths (I oct /I tet ) before and after OER testing (lower panel of Fig. 4b). In contrast, there were distinct changes in the bonding distances in Co 3 O 4 after OER reaction (Fig. 4c light red curve), namely the shortening of both Co-O and Co-Co tet bonds and the elongation of Co-Co oct bond, as illustrated in the top scheme in Fig. 4d. Moreover, the I oct /I tet ratio in Co 3 O 4 displayed an obvious increase from 1.44 to 1.52 after OER testing (lower panel of Fig. 4b), suggesting the crystal structure of Co 3 O 4 underwent dynamic changes during OER reaction, as revealed by the prominent three sets of pre-OER redox features, which might be similar to the formation of active structure motifs during OER reactions in alkaline or neutral media 26,29 .
In situ Raman studies of the OER reaction mechanisms. The intensities of all Raman peaks at higher applied potentials decrease substantially (Fig. 5b, c lower panel), which was usually accompanied with the increase in average valence state of Co atoms 63 . When the applied potential was finally switched back from 1.87 to 1.22 V vs. RHE, the peak intensities partially recovered (lower panel in Fig. 5c) and the CoOOH species was clearly detected again.
To understand the evolution of the local bonding environments at the catalyst surface during the OER process, the peak position, intensity, and full width at half maximum (FWHM) of the Raman A 1g peak (~690 cm -1 ) were extracted by fitting with Lorentzian function (Fig. 5b, c). The shift in the peak position as a function of applied potential can be interpreted as either the change in crystallinity (e.g., red-shift with broadening in FWHM happens when the crystallinity decreases dramatically), or the generation of strain/stress (i.e., lattice contraction/extension) 64,65 . Since the marginal variations in the peak FWHM suggested the crystalline domain sizes of both samples remain relatively constant during the OER process ( Supplementary Fig. 27), the observed peak position shift should result from the lattice contraction/extension and surface reconstruction due to the changing local bonding environments. More importantly, the peak positions shift in opposite directions on these two catalysts as the potential goes over the OER catalytic onsets (Fig. 5c upper panel). Co 3 O 4 /CeO 2 showed a redshift in the A 1g peak position after the onset of OER at 1.52 V vs. RHE. Red-shifts in Raman signals are commonly observed in OER catalysts (CoO x 63,66 , NiOOH 67 , NiFe, and CoFe oxyhydroxides 68 ) at OER operating potentials, and they generally reflect the characteristic vibration for local bonding environment at the outer layer of catalysts with oxidized active site during OER. Thus, the generation of active Co IV species that can participate in a fast and efficient OER process should lead to the observed red-shift of the Raman signals. In contrast, blue-shifts in Raman signals usually suggest lattice contraction and charge redistribution 64,69 . Unlike the more active Co 3 O 4 /CeO 2 , the pure Co 3 O 4 catalyst would go through substantial charge-accumulation surface reconstruction (Co III Co IV ↔ Co IV Co IV ) at~1.62 V around the onset for OER. The Co IV species generated during this process are stabilized and cannot a b c onset of OER participate in fast OER turnover since the reduction peak could be still observed when the potential was scanned backwards, thus they lead to a blue-shift in the Raman signals (Fig. 5c). Another interesting difference is that the peak position of Co 3 O 4 /CeO 2 at 1.22 V vs. RHE remains almost unchanged before and after applying the higher potential sequence, suggesting the flexibility in the local bonding environment of Co 3 O 4 in the composite catalyst. However, the peak position of Co 3 O 4 cannot fully recover after the same potential cycle, with the final peak at~1 cm -1 higher in wavenumber accordingly (Fig. 5c upper panel and Supplementary  Fig. 28), which is consistent with the positive charge accumulated at the Co center with shorter Co-O bond in the Co 3 O 4 sample after OER (Fig. 4a-c). Together with the ex situ XAS results, the in situ Raman results clearly demonstrate that the bonding environment surrounding Co centers is modified in the Co 3 O 4 /CeO 2 catalyst, which allows the active Co sites to be more readily oxidized and avoid the substantial potential-determining surface reconstruction that would otherwise form stabilized dimeric Co IV Co IV with charge accumulation and lattice contraction. As Co IV is the key intermediate to start OER process, the more facile formation of Co IV species and destabilization of Co IV Co IV in Co 3 O 4 /CeO 2 would allow faster OER kinetics thus enhance the catalytic activity.
Electrode performance and stability of Co 3 O 4 /CeO 2 nanocomposites. We further optimized the overall electrode performance by replacing the FTO substrate with high-surface-area three-dimensional carbon paper substrate that facilitates electron and ion transport and gas bubble release. To reach a geometric catalytic current density of 10 mA cm -2 in 0.5 M H 2 SO 4 solution, Co 3 O 4 /CeO 2 on carbon paper electrode only required an overpotential as low as 347 mV, which is only 46 mV higher than that needed for the benchmark RuO 2 catalyst on carbon paper electrode ( Supplementary Fig. 29). A comprehensive comparison shows that Co 3 O 4 /CeO 2 is an efficient earth-abundant metal oxide-based electrocatalysts reported to date for the acidic OER (Supplementary Table 3). Lastly, we examined the acidic OER stability of the Co 3 O 4 / CeO 2 catalyst, since the tradeoff between activity and stability has usually been observed in acidic OER catalysts 15,16 Fig. 30a, c). The cobalt dissolution rate of Co 3 O 4 /CeO 2 also coincided with that of Co 3 O 4 in 0.5 M H 2 SO 4 solution ( Supplementary Fig. 30b). The metal dissolution rates of both catalysts were also investigated under open circuit condition without an applied bias ( Supplementary  Fig. 31). Both catalysts showed inferior stability under open circuit condition compared to their respective stability under anodically biased OER condition, suggesting that the applied bias is important for the long-term stability of earth-abundant Co oxides during acidic OER operation 70

Discussion
In conclusion, Co 3 O 4 /CeO 2 nanocomposite is established as an active earth-abundant OER electrocatalyst in acidic media. The overpotentials required for Co 3 O 4 /CeO 2 to achieve a geometric catalytic current density of 10 mA cm -2 on FTO and carbon paper electrodes are~423 and 347 mV, respectively, making it an efficient earth-abundant electrocatalysts for acidic OER. In-depth electrochemical characterizations using the KIE, pH-, and temperature-dependence analyses, together with in situ Raman and ex situ XAS structural characterizations of the Co 3 O 4 /CeO 2 catalyst before and after OER testing, consistently reveal the microstructural states of the catalysts and their changes through the OER processes. The introduction of nanocrystalline CeO 2 modifies the electronic structures and creates a more favorable local bonding environment in Co 3 O 4 that allows the Co III surface species to be easily oxidized into OER-active Co IV species and suppresses the charge accumulation of Co 3 O 4 under electrochemical conditions, which are the keys to bypassing the potential-determining redox step in Co 3 O 4 that result in substantial surface reconstruction and thus enhancing the acidic OER activity. Interestingly, Co 3 O 4 /CeO 2 also breaks the activity/ stability tradeoff by featuring enhanced activity but comparable acidic OER stability and better open circuit stability in comparison with Co 3 O 4 . We hope these findings could stimulate future studies to further elucidate the active site structures and the catalytic mechanisms of nanocomposite OER catalysts using other in situ and/or operando techniques. This work not only establishes an active earth-abundant nanocomposite catalyst (Co 3 O 4 /CeO 2 ) for OER in acidic media, but also stimulates mechanistic understandings and provides an effective strategy to design more efficient and stable nanocomposite electrocatalysts for acidic OER or other reactions in the future.

Methods
Chemicals. All chemicals were purchased from Sigma-Aldrich and used as received without further purification, unless noted otherwise. Deionized nanopure water (18.2 MΩ • cm) from a Thermo Scientific Barnstead water purification system was used for all experiments.
Synthesis of Co 3 O 4 and Co 3 O 4 /CeO 2 on FTO or carbon paper. The corresponding metal hydroxide precursors were first synthesized on the substrates by electrodeposition from a solution of the corresponding metal nitrate(s) with a total concentration of 0.1 molar (mol). For synthesizing the Ce-doped Co(OH) 2 precursor, 10 mol percent (mol%) of Co(NO 3 ) 2 in the solution was replaced with Ce (NO 3 ) 3 . Note that the as-received carbon paper substrate (Fuel Cell Earth, TGP-H-060) was Teflon-coated; therefore, it was first treated with oxygen plasma at 300 W power for 15 min for each side and then annealed in air at 700°C for 1 h to make the surface hydrophilic. Prior to the electrodeposition, the FTO and carbon paper substrates were successively washed with acetone, ethanol, and nanopure water. During the electrodeposition, an Ag/AgCl reference electrode and a Pt mesh counter electrode were used, and a constant potential of -1.0 V vs. Ag/AgCl was applied on the substrates for 3 and 10 min in the case of FTO and carbon paper, respectively. During the electrodeposition, the reduction of nitrate generated OHand a local alkaline environment near the substrate, and subsequently metal hydroxides were formed on the substrate 71 : Co 2þ þ 2 OH À ! CoðOHÞ 2 ð2Þ After the electrodeposition, the metal hydroxide precursors were dried at 80°C for 6 h, and then annealed in air at 400°C (or 300 or 500°C as specifically discussed) for 2 h in a muffle furnace to transform into oxides.
Structural characterizations. SEM and EDS were conducted on a Zeiss Supra 55VP field emission SEM equipped with a Thermo Fisher Scientific UltraDry EDS detector. The accelerating voltage for SEM and EDS were 3 and 15 kV, respectively. Transmission electron spectroscopy images and elemental mappings were collected using a JEM-2100F microscope equipped with an Oxford energy-dispersive X-ray analysis system, with the accelerating voltage of 200 kV. PXRD was performed on a Bruker D8 Advance powder X-ray diffractometer using Cu Kα radiation. XPS was performed on a Thermo Scientific K-Alpha XPS system with an Al Kα X-ray source. UPS was collected on a Thermo ESCALAB 250Xi XPS system with a He I source gun. The Raman spectra were collected on a Thermo Fisher Scientific DXRxi Raman imaging microscope with a 532 nm laser. The ICP-MS analysis was carried out on a Shimadzu ICPMS-2030 spectrometer. The XAS were collected in the transmission mode at the Advanced Photon Source Beamline 10-BM-B at the Argonne National laboratory. To collect the Co K-edge in the energy window from 7.450 to 8.650 keV, a 71/29 N 2 /He gas mixture was used in the I 0 ion chamber to achieve 10% absorption, while a 68/32 N 2 /Ar gas mixture was used in the I t ion chamber to achieve 70% absorption (calculated using Hephaestus at an energy of 7.709 keV). The Co foil standard was used for the energy calibration.
Electrochemical measurements. All electrochemical measurements were conducted in a conventional three-electrode setup using a Bio-Logic SP-200 potentiostat. The Co 3 O 4 or Co 3 O 4 /CeO 2 catalyst grown on FTO or carbon paper was directly used as the working electrode, along with an Ag/AgCl reference electrode and a Pt mesh counter electrode in 0.5 M H 2 SO 4 solution. CV was performed at the scan rate of 5 mV s -1 . EIS was collected in the frequency range from 100 kHz to 50 mHz. All CV curves were manually iR-corrected based on EIS results. To extract the double-layer capacitance (C dl ), CV was collected in pre-OER potential region at various scan rates from 10 to 60 mV s -1 . The relationship between ECSA (cm 2 ) and C dl (mF) is shown in Eq. (3): where C s is general specific capacitance, which is a constant of 0.035 mF cm -2 in the literature 45 .
All potentials were reported versus the RHE using Eq. (4): EðRHEÞ ¼ EðAg=AgClÞ þ 0:059 pH þ 0:197 ð4Þ The operational stability of the catalyst was tested by running chronopotentiometry tests at a constant geometric catalytic current density of 10 mA cm -2 in 0.5 (or 0.05) M H 2 SO 4 solution for 50 (or 100) h.
Reaction order with respect to pH. To extract the reaction order with respect to pH for the acidic OER, the electrochemical measurements of the catalysts were conducted in H 2 SO 4 solutions with different pH values. The reaction order with respect to pH was calculated using Eq. (5) 27,72 : where j is the catalytic current density at a fixed overpotential η. where the term of +0.013 originates from the difference in the standard equilibrium potentials of the deuterium couple (D 2 /D + ) and the proton couple (H 2 /H + ) 53 .
The overpotentials of the OER in the protonic and deuteric solution were determined by Eqs. (7) and (8), respectively 53 : The KIE was calculated using Eq. (9): where j H 2 O and j D 2 O are the catalytic current density in the protonic and deuteric solution, respectively, at the same overpotential (η) 72 .
Apparent activation energy. To extract the apparent activation energy (E app ) for the acidic OER, the electrochemical measurements of the catalysts were conducted in 0.5 M H 2 SO 4 solution at different temperatures. For heterogeneous electrocatalytic reaction, the current density can be expressed from apparent activation energy (E app ) in the Arrhenius Eq. (10) 56,57 : where A app is the apparent pre-exponential factor, R is the ideal gas constant (8.314 J K -1 mol -1 ), T is the temperature in Kelvin (K). Therefore, E app can be further calculated from fitting the slope of the Arrhenius plot using Eq. (11) 54,56 : ∂ðlog 10 jÞ ∂ð1=TÞ while the intercept of log 10 j vs. 1/T plot is the logarithm of A app 57 .
Average Co valence state. The absorption edge energies of the XAS spectra were first determined by an integral method shown in Eq. (12) 59 : where μ 1 = 0.15 and μ 2 = 1 are the lower and upper limit, respectively, of the normalized absorption intensity that are used for the integral. The average Co valence states were then calculated by fitting the absorption edge energies determined earlier into an experimental equation developed by Dau et al. 34

Data availability
The data that support the findings in the paper can be found in the Source Data. Additional data presented in the Supplementary Information are available from the corresponding author upon reasonable request. Source Data are provided with this paper.