In-situ local phase-transitioned MoSe2 in La0.5Sr0.5CoO3-δ heterostructure and stable overall water electrolysis over 1000 hours

Developing efficient bifunctional catalysts for overall water splitting that are earth-abundant, cost-effective, and durable is of considerable importance from the practical perspective to mitigate the issues associated with precious metal-based catalysts. Herein, we introduce a heterostructure comprising perovskite oxides (La0.5Sr0.5CoO3–δ) and molybdenum diselenide (MoSe2) as an electrochemical catalyst for overall water electrolysis. Interestingly, formation of the heterostructure of La0.5Sr0.5CoO3–δ and MoSe2 induces a local phase transition in MoSe2, 2 H to 1 T phase, and more electrophilic La0.5Sr0.5CoO3–δ with partial oxidation of the Co cation owing to electron transfer from Co to Mo. Together with these synergistic effects, the electrochemical activities are significantly improved for both hydrogen and oxygen evolution reactions. In the overall water splitting operation, the heterostructure showed excellent stability at the high current density of 100 mA cm−2 over 1,000 h, which is exceptionally better than the stability of the state-of-the-art platinum and iridium oxide couple.

H ydrogen has high energy density (120-140 MJ kg −1 ) and is an environmentally friendly clean energy source that can be produced through water splitting 1,2 . The state-ofthe-art water splitting catalyst for hydrogen production is composed of precious noble metals, and it has several drawbacks such as high-cost, limited reserves, and durability 3 . Currently, the Irand Ru-based compounds and Pt-group metals are considered the most efficient catalysts for the anodic oxygen evolution reaction (OER) and the cathodic hydrogen evolution reaction (HER), respectively 3,4 . Because each half-cell reaction requires a different pH environment for yielding the best performance, the adoption of such electrodes in integrated electrochemical water splitting has been limited in practical settings. To achieve efficient overall water splitting for hydrogen production, OER performance is also important because the OER process can become a kinetic bottleneck in HER owing to its complex and slow overall reaction involving concerted electron-proton shift steps [3][4][5] . Therefore, bifunctional catalysts that are earth-abundant and can exhibit concurrently excellent HER and OER performance are required urgently 6,7 . ABO 3 perovskite oxides (A: rare-earth or alkaline earth element, B: transition metal ion) have received significant attention as potential alternatives to precious metal-based catalysts (e.g., RuO 2 and IrO 2 ) owing to their strong catalytic activity, robust stability, and compositional flexibility 8,9 . Thus, considerable efforts have been devoted to understanding the mechanisms of OER and HER on perovskite oxides, and molecular orbital studies have suggested that cobalt-based oxides can be used as active catalysts in OER and HER 10 . Among the various ABO 3 perovskite oxide catalysts, La 1-x Sr x CoO 3-δ has been extensively studied for its strong catalytic activity. Liu et al. synthesized hierarchical mesoporous La 1-x Sr x CoO 3-δ by electrospinning and showed improved OER activity and redox stability in a lithium oxygen battery 11 . Mefford et al. showed the manner in which OER performance can be improved by exploiting the oxygen vacancy defect in La 1-x Sr x CoO 3-δ 12 . Grimaud et al. demonstrated that the O 2 generated from the lattice oxygen of La 1-x Sr x CoO 3-δ significantly influenced OER 13 .
Both theoretical and experimental investigations on transition metal dichalcogenides (TMDs) have revealed the great potential of TMDs as hydrogen generation catalysts owing to their high catalytic activity; robustness to CO, CO 2 , and O 2 ; affordability; and scalability 14,15 . Of the various TMDs, molybdenum diselenide (MoSe 2 ) is considered a promising HER catalyst because of its relatively superior electrochemical catalytic activity and chemical stability compared to other TMDs 16,17 . The Gibbs free energy of MoSe 2 for hydrogen adsorption is close to zero, and its hydrogen coverage is greater than those of other TMDs 17 . However, owing to low conductivity of the intrinsic 2H-phase MoSe 2 , MoSe 2 -based composite structures, such as MoSe 2 /carbon cloth 16 , MoSe 2 /n + p-Si 18 , and MoSe 2 /graphene 19 , have been typically used to improve the electrochemical activity of intrinsic MoSe 2 . In addition, inducement of the semiconducting (2H) to metallic (1T) phase transition in TMDs has been considered to improve the performance of TMDs-based composite electrochemical catalysts because the metallic phase can improve their intrinsic electrocatalytic nature 20 . However, the phase transition process of TMDs is rather complex and time-consuming, and it requires an inert environment owing to the highly reactive materials involved, such as alkali metals 20,21 .
Apart from performance, the durability of water splitting catalysts is an important criterion from the commercial perspective. Chemical instability results in catalyst decomposition in the electrode during continued operation, which hinders long-term catalyst stability. Therefore, developing earth-abundant bifunctional catalysts that are based on non-precious metal elements and concurrently offer excellent HER and OER performance along with robust chemical stability is indispensable to ensure the industrial viability of electrochemical water splitting.
In present work, we devise a composite perovskite oxide-TMD heterostructure composed of MoSe 2 (denoted MoSe 2 ), La 0.5 Sr 0.5 CoO 3-δ (denoted LSC only), and Ketjenblack carbon (denoted KB) as a bifunctional electrocatalyst for overall water electrolysis. The LSC, MoSe 2 , and KB heterostructure (denoted LSC&MoSe 2 ) offers considerably better HER and OER performances (onset potential, Tafel slope) than LSC and KB heterostructure (denoted LSC) or MoSe 2 alone. Interestingly, an in-situ local phase transition in MoSe 2 (from 2H-to 1T-MoSe 2 ) is observed during the formation of LSC&MoSe 2 , possibly because of spontaneous electron transfer from Co to Mo. This charge transfer is expected to enhance the intrinsic conductivity of MoSe 2 and increase the amount of Co-O and Co-OH in LSC, which can enhance the water splitting catalytic activity. When LSC&MoSe 2 || LSC&MoSe 2 electrode was applied to overall water splitting, the initiation potential was observed at 1.52 V, and the proposed electrode exhibited excellent overall water electrolysis stability over 1000 h at a high current density of 100 mA cm −2 , which is far superior performance compared to that of the Pt/C || IrO 2 electrode.

Results
Morphological and structural properties of LSC&MoSe 2 . LSC&MoSe 2 was prepared using the high-energy ball milling process with the optimum weight ratio of LSC:MoSe 2 :KB = 6:3:1 determined by the electrochemical analyses (Supplementary Figure 1-3; see Experimental Section for details). Morphological and structural analyses of the composite electrocatalyst were first performed by transmission electron microscopy (TEM) and scanning electron microscopy (SEM). Figure 1a, b show brightfield TEM image and high-angle annular dark-field (HAADF) image of LSC&MoSe 2 , along with energy-dispersive spectroscopy (EDS) elemental mapping, which clearly illustrates the presence of the associated elements (La, Sr, Co, O, Mo, Se, and C) in LSC&MoSe 2 . Further analysis on the morphology of LSC&MoSe 2 heterostructure was carried out by SEM ( Supplementary Figure 4). It can be seen that MoSe 2 nanoflakes are randomly distributed and adsorbed onto the LSC surface without causing any noticeable aggregation, which can contribute to the increase of overall surface area of the composite structure. The SEM-EDS and elemental quantitative analysis for LSC&MoSe 2 further revealed that each constituent atomic component in the composite structure was clearly observed with expected elemental ratio (Supplementary Figure 5). Figure 1c shows a high-resolution TEM (HR-TEM) image of LSC&MoSe 2 , which highlights the presence of MoSe 2 (red). The inset in Fig. 1c shows the fast Fourier transform (FFT) pattern of MoSe 2 viewed along the <001> zone axis, indicating highly crystalline structures corresponding to the (100) plane of 2H-MoSe 2 with a lattice spacing of 0.28 nm. A HR-TEM image and the corresponding FFT pattern (inset, white section) of LSC along the <110> zone axis are shown in Fig. 1d, which highlight the crystalline structure of LSC corresponding to the (001) plane with a lattice spacing of 0.40 nm. The crystal structures of LSC, MoSe 2 , and LSC&MoSe 2 were analyzed further by using X-ray diffraction (XRD), as shown in Fig. 1e. The peaks related to LSC and MoSe 2 can be observed clearly in the XRD pattern of LSC&MoSe 2 , indicating the wellmixed state of LSC and MoSe 2 without the presence of any additional phase, along with the amorphous state of KB.
We then performed Brunauer-Emmett-Teller (BET) analysis on LSC&MoSe 2 and LSC to investigate the effect of MoSe 2 on the surface area of the heterostructure. We also measured the BET surface area of MoSe 2 and LSC only as 32.55 and 10.47 m 2 g −1 , respectively (Supplementary Figure 6). However, as shown in Fig. 1f, the addition of MoSe 2 to LSC led to a notable increase in the total surface area of the composite structure, where the surface area of LSC&MoSe 2 (142.09 m 2 g −1 ) was more than thrice that of LSC (39.95 m 2 g −1 ). Statistical analysis of the aforementioned samples is provided in Supplementary Figure 7. Increase in the BET surface area for LSC&MoSe 2 can be attributed to the additional MoSe 2 nanoflakes present that are adsorbed onto the LSC surface. Such a remarkable increase in surface area improves water splitting by increasing the total number of active electrocatalysis sites for both HER and OER 22 . The pore size of LSC&MoSe 2 was investigated using the Barrett-Joyner-Halenda (BJH) method. The pore size distribution in Supplementary  Figure 8 shows that mesoporous pores (2-50 nm) are present on the surface of the heterostructure. We analyzed the BJH pore size distributions of MoSe 2 and LSC only to investigate the origin of mesoporous characteristics in LSC&MoSe 2 (Supplementary Figure 9). In MoSe 2 , a sharp peak primarily centered at around 4 nm is observed, indicating the presence of mesoporous structure, whereas the broad peak centered at 55 nm appears for the LSC only. This result is consistent to the BET surface area analysis (Supplementary Figure 6), where N 2 adsorption-desorption isotherms of MoSe 2 shows type-IV characteristics with a hysteresis loop at relative pressures (P/P 0 ) from 0.45 to 1.0 and LSC only shows much less conspicuous type-IV character than that of MoSe 2 within similar P/P 0 range, which suggests that the mesoporous characteristics observed from LSC&MoSe 2 heterostructure mainly originated from the MoSe 2 23 . This mesoporous-pore-sized catalyst with large surface area is  Electrochemical performance. At first, the hydrogen evolution reaction (HER) activities of the proposed catalyst were investigated for various catalyst configurations by examining rotating disk electrode (RDE) polarization curves obtained under N 2saturated 1 M KOH (Fig. 2a). The corresponding Tafel plots were derived from the obtained polarization curves (Fig. 2b). The potential was referenced to the reversible hydrogen electrode (RHE), as obtained from the calibration profile (Supplementary Figure 10). Pt/C exhibits excellent HER activity with an onset potential of 0.0 V vs. RHE and sharply increasing current density with a low Tafel slope value of 28 mV dec. -1 , which indicates the Volmer-Tafel reaction is its primary HER pathway 24 . In contrast, MoSe 2 , known as an active HER catalyst under acidic media 25 , shows rather inferior HER activity with an onset potential of −0.42 V and Tafel slope of 237 mV dec. -1 in alkaline media. LSC presents moderate HER activity, as reported previously 12 with an onset potential of −0.33 V and Tafel slope of 95 mV dec. -1 , suggesting the Volmer-Heyrovsky reaction is the dominant HER pathway 26 . Interestingly, the addition of MoSe 2 to LSC (LSC&MoSe 2 ) facilitated the HER activities of LSC by positively shifting the onset potential to −0.20 V, a 0.13 V increase from that of LSC. This improvement is noteworthy in that the onset potential usually depends on the intrinsic property of the catalyst, and the combination of two low-performance materials, with onset potentials of −0.33 V (LSC) and −0.42 V (MoSe 2 ), respectively, led to an enhanced overall onset potential of −0.20 V (LSC&MoSe 2 ), demonstrating the synergistic effect of heterostructure formation. The Tafel slope of the composite structure improved to 34 mV dec. -1 from 95 mV dec. -1 (LSC) and 237 mV dec. -1 (MoSe 2 ), respectively, approaching that of Pt/C (28 mV dec. -1 ), which suggests the Volmer-Tafel mechanism is the preferred pathway in the HER process of LSC&MoSe 2 . This result is comparable to that of the high-performance emerging HER catalysts under identical electrolyte conditions (1 M KOH), and it shows the lowest Tafel slope among the survey groups (Supplementary Table 1), which indicates faster increment in current with small overpotential and lower activation barrier toward rapid reaction rate.
In addition, we investigated the oxygen evolution reaction (OER) activities of the prepared catalysts under N 2 -saturated 1 M KOH. Figure 2c presents the OER polarization curves, and the corresponding Tafel slopes are derived, as shown in Fig. 2d. For the OER analysis, IrO 2 , known as the state-of-the-art catalyst for OER, was adopted as a reference catalyst 27 . IrO 2 exhibits excellent OER performance with the onset potential of 1.51 V vs. RHE and Tafel slope of 75 mV dec. -1 . In the case of LSC, active OER performance was observed with the onset potential of 1.54 V and Tafel slope of 90 mV dec. -1 , but MoSe 2 exhibited an inactive OER process along with an increase in overpotential. Notably, LSC&MoSe 2 presented considerably improved OER activities compared to that of LSC with an onset potential of 1.52 V and sharply increasing current density with a low Tafel slope of 77 mV dec. -1 . In the same electrolyte tested for HER (1 M KOH), the OER performance achieved in this work is comparable to that of recently reported high-performance catalysts (Supplementary Table 2). These results demonstrate that the formation of LSC&MoSe 2 leads to simultaneous enhancements in both the HER and OER activities, suggesting that LSC&MoSe 2 has potential as a highly efficient overall water splitting catalyst.
Analysis of LSC&MoSe 2 properties. To elucidate the origin of enhanced catalytic activity in LSC in the presence of additional MoSe 2 , various analyses, including spectroscopic, chemical, and electrical characterizations, were performed to investigate the potential interaction between the two substances. Figure 3a shows the fluorescence emission spectra (FL) of MoSe 2 , LSC only, and KB for various configurations. For the FL measurement, each sample (MoSe 2 , LSC&MoSe 2 , LSC only and MoSe 2 heterostructure (denoted LSC/MoSe 2 ), and MoSe 2 and KB heterostructure (denoted MoSe 2 /KB) was prepared using the same concentration of MoSe 2 (0.33 mg/mL). Detailed summary of abbreviations for each material system studied in this work is provided in Supplementary Table 3. Pristine 2H-phase MoSe 2 is semiconducting in nature 28 , and LSC only and KB are highly conductive materials with metallic features 29 . A clear FL peak was observed for MoSe 2 at 825 nm, whereas substantial FL quenching occurred in the MoSe 2 composite containing LSC only and KB, that is, LSC&MoSe 2 , LSC/MoSe 2 , and MoSe 2 /KB. In addition, the FL peak was not present in the case of KB, and only a weak FL spectrum was observed in the case of LSC only, as shown in Supplementary Figure 11. However, the FL peak was observed in the case of LSC. These results suggest that charge transfer occurs readily among MoSe 2 , LSC only, and KB while that between LSC only and KB is limited. Moreover, optical absorption measurements were performed to further examine the effect of local phase  Figure 3b shows the UV-vis-NIR spectrum of MoSe 2 , which highlights the semiconducting feature of the 2H-phase with two exciton peaks at 700 and 800 nm 30 . In LSC&MoSe 2 , such excitonic peaks disappeared, indicating that the phase transformed TMDs induced metal-like characteristics in LSC&MoSe 2 31,32 (Fig. 3c).
Because intrinsic catalyst activity is closely related to catalyst surface properties (e.g., electronic structure, vacancy, defect, etc.), the surface characteristics of LSC&MoSe 2 and LSC were investigated by X-ray photoelectron spectroscopy (XPS). We focused on the chemical state of Co 2p, as shown in Supplementary Figure 12, which has the dominant effect on the catalytic activity of LSC-based composite materials 22 . The Co 2p peak consists of two spin-orbit doublets, Co 2p 3/2 and Co 2p 1/2 , with two satellites. The Co 3+ and Co 2+ of LSC are positioned at 779.6/794.2 eV and 781.5/797.0 eV, whereas those of LSC&MoSe 2 are located at 780.6/795.2 eV and 782.5/798.0 eV, respectively, indicating a binding energy upshift of 1 eV in LSC&MoSe 2 relative to LSC. This shift in the XPS peak was attributed to changes in the oxidation state of the LSC owing to the presence of additional MoSe 2 . The Co 2p of LSC&MoSe 2 is to have higher oxidation states than the LSC owing to the electron transfer at the interface of LSC and MoSe 2 while maintaining the overall electroneutrality of LSC&MoSe 2 33,34 . The high electronegativity of Mo may have altered the electronic structure (σ*-orbital occupancy) of Co by decreasing the electron density in Co 35 . The σ*-orbital occupancy (e g ) of close to 1 of the perovskite oxides catalyst enhanced the binding of oxygen species at the B site, which contributed to improved OER performance 36 . The large ratio of high oxidation numbers in the Co species is related to the e g orbital filling of close to 1, which is a desirable feature for OER catalysts 22,37 . Therefore, the enhanced OER activity of LSC&MoSe 2 can be attributed to its higher Co 3+ /Co 2+ ratio than that of LSC (Supplementary Table 4). The chemical state of O 1s in the catalyst, too, can influence its OER performance 38 Figure 13). A high density of surface-active oxygen species on the catalyst surface, which is associated with oxygen vacancies and surface hydroxyl groups, is known to improve OER performance 6,38 , and the relative ratio between lattice and surface-active oxygen can be used as an indicator for evaluating OER performance 39 . As shown in Supplementary Table 5, the higher C O /A O ratio of LSC&MoSe 2 than that of LSC, i.e., metals with high oxidation states which enhances the adsorption ability for oxides in the LSC&MoSe 2 surface 38 , indicates larger surface coverage of hydroxide species in LSC&MoSe 2 , which can improve its intrinsic OER performance.
In addition, the surface adsorption capability of the catalyst was analyzed by conducting thermal gravimetric analysis (TGA) to elucidate the different electrocatalytic performances of LSC&MoSe 2 and LSC. LSC&MoSe 2 and LSC were exposed to wet-air conditions before the TGA measurement to facilitate the adsorption of water, H, and OH groups from the atmosphere. As shown in Fig. 3d, the relative weight loss (%) was larger in the case of LSC&MoSe 2 than that in the case of LSC, implying that LSC&MoSe 2 possesses a higher surface adsorption capability, which can improve its electrocatalytic performance 6 .
Ultraviolet photoelectron spectroscopy (UPS) analysis was conducted to investigate the charge transfer between LSC and LSC&MoSe 2 to elucidate the improved catalytic performance of LSC&MoSe 2 . The increase in the work function of the catalyst increases proton concentration in the electronic double layer of the catalyst, which facilitates easy progress of the Volmer step (H + +e -=H ad ), which initiates the HER process, with the least overpotential 40,41 . Moreover, it increases the preexponential factor and the rate constant. This reduces the metal-hydrogen (M-H) bond strength, which helps improve the exchange current density 42,43 . Figure 3e shows the secondary electron cutoff energies of LSC and LSC&MoSe 2 , from which the work function values are derived. The addition of MoSe 2 to LSC increased the work function value from 5.15 of LSC to 5.33 eV of LSC&MoSe 2 . The UPS result indicated that the incorporation of Mo with higher electronegativity into Co, the active site in perovskite oxide catalysts, led to an increase in the work function value of LSC&MoSe 2 , which improved its HER performance by increasing the exchange current density of LSC&MoSe 2 . This increase in the work function can induce a decrease in the adsorption bond strength between the catalyst and hydrogen, which brings the Gibbs free energy closer to the thermoneutral point of catalyst-H*, leading to improved catalytic activity of HER.
The improved catalytic performance of LSC&MoSe 2 was examined by means of electrochemical impedance spectroscopy (EIS) analysis. Figure 3f shows the Nyquist plots of both the HER and OER of MoSe 2 , LSC, and LSC&MoSe 2 , from which the charge transfer resistance (R ct ) of the catalyst between the electrode and electrolyte was obtained. The R ct of MoSe 2 is 1500 Ω cm 2 in both HER and OER, which is highest charge transfer resistance compared to the others. In contrast, the R ct of LSC&MoSe 2 is 5.71 and 4.36 Ω cm 2 for HER and OER, respectively, while that of LSC is 10.16 and 6.93 Ω cm 2 , respectively. This result suggests that rapid electron transport is feasible in LSC&MoSe 2 owing to the additional MoSe 2 , which can improve conductivity of the catalyst. In brief, the improved electrocatalytic performance of LSC&MoSe 2 in HER and OER can be attributed to various combinatorial effects of enhanced electrical and chemical properties in the catalyst that benefit the HER and OER aspects simultaneously.
Synergetic effect for improved electrochemical performance. The morphology of the crystal structure of LSC&MoSe 2 was investigated to further elucidate the dramatic improvement in the electrochemical performance of LSC&MoSe 2 compared to other catalyst configurations such as MoSe 2 and LSC. Figure 4a shows a HR-TEM image of LSC&MoSe 2 . Interestingly, after formation of the heterostructure of LSC with 2H phase MoSe 2 , the lattice structures of 1T-MoSe 2 (metallic, octahedral) and 2H-MoSe 2 (semiconducting, trigonal prismatic) were observed simultaneously. Each phase of the MoSe 2 lattice structure can be observed clearly in the zoomed-in image presented in Fig. 4b, c along with the corresponding schematic diagram of the lattice structures. The 2H-MoSe 2 region showed the typical hexagonal Mo-Se atomic arrangement with a 0.283 nm interlayer distance, corresponding to the (100) plane of MoSe 2 . In contrast, 1T-MoSe 2 exhibited a distinctively different crystal structure that was attributed to electronic structure rearrangement 44 Table 6. To verify whether the ball milling process has any effect on the phase transition of MoSe 2 , MoSe 2 /KB was synthesized under the same ball-mill process as LSC&MoSe 2 and investigated through XPS and HR-TEM analysis. As shown in Supplementary Figs. 15-16, no signature of phase transition is observed and only 2H-phase MoSe 2 is present in ball-milled MoSe 2 /KB. These results suggest that the local phase transition in MoSe 2 from the 2H-phase to the 1T-phase occurred in-situ during formation of the heterostructure between LSC and 2Hphase MoSe 2 . 1T-MoSe 2 is typically synthesized via alkali metal treatment of 2H-MoSe 2 in inert environment 20,21 . Surprisingly, in the present work, the presence of 1T-MoSe 2 was observed locally in LSC&MoSe 2 . We hypothesize that this in-situ phase transition in MoSe 2 is induced by electron transfer from Co to Mo 47 , as illustrated schematically in Fig. 4d. The electron transfer between Mo and Co is expected to improve the electrochemical catalytic activity of MoSe 2 and LSC, as described in Fig. 4e. In the case of MoSe 2 , additional electrons alter the Mo 4d-orbital configuration from the occupied 4d z 2 level to incompletely filled d xz , d yz , d yx orbital, which induces the phase transition 44,48 . As a result, the intrinsic conductivity of MoSe 2 improves, which improves the electrochemical performance of the catalyst. This electron transfer is beneficial for the catalytic activity of LSC to achieve efficient water splitting. The Co in LSC becomes more electrophilic after the electron transfer, which alters the electronic structure of Co by upshifting the d-band center 35 . Such an up-shift of d-band center strengthens the adsorption capability of OH -49 . Therefore, the enhanced OHaffinity and electrophilicity in Co, that is, increase in Co-O and Co-OH in LSC, can ultimately improve the overall water electrolysis performance. These charge transport phenomena bring synergistic effects to enhance both the HER and OER performance from the heterogenous composite structure of LSC&MoSe 2 in addition to the performance improvement factors mentioned in Analysis of LSC&MoSe 2 properties section.
Theoretical elucidation of charge transfer in LSC&MoSe 2 . To theoretically demonstrate the charge transfer phenomenon in LSC&MoSe 2 heterostructure, we performed the density functional theory (DFT) calculations (see calculation details in the Experimental section). For this purpose, we first examined the relative stability of two plausible terminations of LSC (001) surface, corresponding to CoO 2 -termination (denoted CoO 2 -t.) and (La, Sr)O-termination (denoted (La,Sr)O-t.), by surface energy (γ) calculations (see Supplementary Fig. 17 and Surface energy calculations for details in the Experimental section). As a result, (La, Sr)O-t. was predicted to be predominantly exposed on the LSC nanoparticle surface due to its lower surface energy (i.e., γ = 0.58  Accordingly, we built the LSC&2H-MoSe 2 heterostructure based on this termination (Fig. 5a). By the Bader charge analysis 50 , it was revealed that the charge transfer occurred from LSC into MoSe 2 with an amount of 0.772 e, whose direction was in accordance with our previous expectation. To better understand these electron movements, the projected density of states (PDOS) on each d-orbital of Co and Mo in LSC&2H-MoSe 2 was further investigated (Fig. 5b). The asymmetric configuration of spin-up and spin-down DOS for LSC indicated the ferromagnetic nature, whereas the symmetric configuration for MoSe 2 represented its non-magnetic character 51,52 . Driven by the electron transfer across the interface, the d-band center (ε d ) of Co in LSC shifted upward about 0.1 eV with respect to the Fermi level when coupled with 2H-MoSe 2 , indicating that LSC became more electrophilic. In a previous study, the enhanced electrophilicity of LSC was reported to strengthen the OH − affinity, which has a beneficial effect to improve the OER catalytic performance 35 . Meanwhile, the ε d of Mo in 2H-MoSe 2 was significantly downshifted by~0.9 eV with respect to the Fermi level when combined with LSC, implying that additional electrons transferred into 2H-MoSe 2 side. These extra d-electrons are expected to promote the 2H-to 1T-phase transition by destabilizing the 2H-MoSe 2 phase, as similarly observed in the MoS 2 system by Gao et al.'s work 53 . To further clarify this, we compared the relative stability for both phases of MoSe 2 in the presence or absence of LSC (Fig. 5c, d and  Supplementary Figure 18). The endothermicity of 1T-MoSe 2 (relative to its 2H-phase) clearly decreased from 0.65 to 0.21 eV per MoSe 2 unit when LSC was combined, indicating that the phase transition from 2H-to 1T-phase could occur more easily. Further, the local DOS of MoSe 2 in the LSC&MoSe 2 heterostructure showed that the intrinsic conductivity can be significantly enhanced by closing the energy gap due to the transition from semiconducting 2H-to metallic 1T-phase, which can improve the HER catalytic performance (Supplementary Figure 19) 54 . These results theoretically elucidate the beneficial effects of charge transfer phenomenon in LSC&MoSe 2 for both HER and OER performance.
Overall water splitting of LSC&MoSe 2 || LSC&MoSe 2 . An overall water splitting test was conducted to evaluate the performance and stability of the developed LSC&MoSe 2 electrocatalyst. By adopting the prepared LSC&MoSe 2 catalyst, we evaluated the overall water splitting performance in deaerated alkaline media containing N 2 -saturated 1 M KOH. The water splitting test was conducted with the three-electrode configuration by using an Ag/AgCl reference electrode to separate the overall cell reaction into the cathodic and anodic reactions of HER and OER, respectively. The electrochemical overall water electrolysis actively generated hydrogen (cathode) and oxygen (anode) gases, as shown in Fig. 6a. Figure 6b shows the HER and OER polarization curves, denoted E cathode and E anode , respectively, measured during the water electrolysis reaction, and the obtained cell potentials (E cell = E anode -E cathode ) of Pt/C || IrO 2 (Pt/C for cathode and IrO 2 for anode) and LSC&MoSe 2 || LSC&MoSe 2 (LSC&MoSe 2 for both cathode and anode). Also, the overall water electrolysis performance of bare Ni foam is investigated and compared to the published works (Supplementary Figure 20). The reference state-of-the-art catalyst pair, Pt/C || IrO 2 , performed well in both HER and OER, as observed previously in the halfcell-configured RDE polarization profiles. In the case of Pt/C || IrO 2 , overall water splitting was observed from near 1.38 V. LSC&MoSe 2 || LSC&MoSe 2 exhibited slightly lower HER performance than Pt/C, but the OER performance was higher than that of IrO 2 . Although overall water splitting was observed at a slightly higher value of 1.52 V, current density increased sharply as the potential increases. The determination of the onset potential values for each catalyst can be seen in Supplementary   Figure 21. Notably, LSC&MoSe 2 || LSC&MoSe 2 outperformed Pt/ C || IrO 2 after 2.16 V owing to its concurrently effective HER and OER kinetics. To investigate the electrochemical stability and durability of the LSC&MoSe 2 catalyst, we measured its chronopotentiometric profiles at a current density of 100 mA cm -2 , as shown in Fig. 6c. For Pt/C || IrO 2 , gradual degradation was initiated from 16 h and drastic degradation occurred after 60 h, leading to cell failure. In contrast, LSC&MoSe 2 || LSC&MoSe 2 exhibited remarkably stable operation over 1000 h with negligible fluctuation in performance, demonstrating its exceptionally high electrochemical durability in HER and OER, even at a high current density of 100 mA cm -2 . Furthermore, the energy efficiency of overall water electrolysis at the current density of 100 mA cm −2 was calculated to be 63.9% as shown in Supplementary Note 1. Considering the energy efficiency for conventional alkaline electrolysis is <70% with the use of noble metal-based electrode 55 , the proposed LSC&MoSe 2 catalyst can be considered as highly efficient for the water electrolysis. Supplementary Figure 22 shows SEM images of the LSC&MoSe 2 electrode after 1000 h of the overall water electrolysis test; clogging or electrochemical/physical damage was not observed on the catalyst after the test. The chemical state of LSC&MoSe 2 after the stability test was further examined via XPS analysis. As shown from Supplementary Figure 23 and Tables 7 and 8, the ratio of Co 3+ /Co 2+ and surface-active oxygen/lattice oxygen showed almost negligible changes even after the 1000 h of overall water splitting measurement, indicating the excellent stability of the proposed LSC&MoSe 2 as the electrocatalyst. Progress of the overall water splitting test with H 2 and O 2 bubble generation from the cathode and anode, respectively, is shown in Supplementary Movie 1. The overall water electrolysis stability achieved in this work was compared with that of the representative overall water splitting catalysts reported to date, as shown in Fig. 6d and summarized in Supplementary Table 9. Despite the harsh test environment (highest current density) in the present study, the developed catalyst exhibited overwhelmingly high overall water electrolysis stability compared to that achieved in other works.

Discussion
We demonstrated that a composite of perovskite oxides and MoSe 2 exhibits excellent electrocatalytic performance and stability in overall water electrolysis. Formation of LSC&MoSe 2 led to several synergetic effects, including increased specific surface area for both the HER and OER, enhanced surface adsorption capability, favorable kinetics for the Volmer-Tafel pathway, and decreased charge transfer resistance, which contributed to its improved electrochemical catalytic performance. Interestingly, electron transfer between Co and Mo induced a local phase transition in MoSe 2 and increased the amounts of Half-cell analysis. Half-cell measurements were conducted in typical threeelectrode configuration using a Pt wire and Ag/AgCl electrode (saturated KCl filled) as the counter electrode and reference electrode, respectively. The rotating disk electrode (RDE) tests were carried out by using catalysts of 20 wt% Pt/C (Alfa Aesar), IrO 2 (Alfa Aesar), LSC, MoSe 2 , and LSC&MoSe 2 on RRDE-3A (ALS). Each catalyst was prepared in the form of an ink by dispersing 10 mg of the catalyst in 1 mL of a binder solution (45:45:10 = ethanol:isopropyl alcohol:5 wt% Nafion solution (Sigma-Aldrich), volumetric ratio) followed by a bath sonication process. Then, HER and OER profiles were investigated in N 2 -saturated 1 M KOH aqueous solution at a scan rate of 10 mV s −1 by drop-coating 5 µL of each catalyst ink onto glassy carbon disk electrode, where the area is 0.13 cm 2 . A calibration in reversible hydrogen electrode (RHE) was experimentally determined at a scan rate of 1 mV s −1 in H 2 -saturated 1 M KOH, where platinum wire was used as the working and counter electrode and Ag/AgCl as the reference electrode. All half-cell profiles were iR compensated by measuring the resistance of solution (1 M KOH). All electrochemical tests were carried out using Biologic VMP3.
Overall water splitting test. The overall water electrolysis tests were proceeded in three-electrode configuration using Ag/AgCl reference electrode. The cathode and anode were prepared by electro-spraying the prepared catalyst ink onto a Ni mesh current collector with a catalyst loading density of 1 mg cm −2 . The measurements were proceeded in deaerated 1 M KOH aqueous solution. The current density was normalized with the geometric area of the catalyst. All electrochemical tests were conducted using Biologic VMP3.
Calculation details. Spin-polarized DFT calculations were carried out using the Vienna ab initio simulation package (VASP) 56 within the projector-augmented wave (PAW) method 57 . The electron exchange-correlation energy was treated by the generalized gradient approximation (GGA) with Perdew-Burke-Ernzerhof (PBE) functional 58 . The DFT +U method within Dudarev's approach 59 was also adopted with U = 4.3 eV and J = 1.0 eV for Co-3d and U = 4.0 eV for Mo-4d, respectively. The energy cutoff for the plane-wave basis set was set as 400 eV and PAW data sets were used with following valence electronic states: 5 s 2 , 5p 6 , 5d 1 , 6 s 2 for La; 4 s 2 , 4p 6 , 5 s 2 for Sr; 3d 8 , 4d 1 for Co; 2 s 2 , 2p 4 for O;4 s 2 , 4p 6 , 4d 6 for Mo; and 4 s 2 , 4p 4 for Se, respectively. Geometry optimizations were performed using conjugated gradient (CG) method until the net force on each atom reached <0.02 eV Å −1 and the total energy was changed within 10 −6 eV per atom. Dipole slab corrections were also applied to all slab model calculations. The Monkhorst-Pack scheme of k-point grid 60 was set to Γ points for geometry optimization, and 3 × 2 × 1 k-points in the Brillouin zone for DOS analysis, respectively. Bader analysis 50 was used to calculate the atomic charges.
Model systems for calculation. To construct the LSC&MoSe 2 heterostructure, each slab model for LSC surface and MoSe 2 layer was separately modeled in advance. First, the unit cell structure of LaCoO 3 (LCO) was fully relaxed by optimizing both atomic positions and lattice parameters, which were well matched with experimentally reported values 61 . Subsequently, the LCO bulk structure was cleaved along the (001) plane with two plausible terminations (i.e., CoO 2 termination and LaO termination). Note that we considered symmetric slab models of LCO (001), which consisted of 5 atomic layers, to remove the fictious dipole moment in the slab. The bottommost two layers were fixed to their bulk positions. Next, a 2 ffiffi ffi 2 p 3 ffiffi ffi 2 p 1 supercell structure was created, and subsequently half of La atoms were replaced into Sr atoms to attain the stoichiometry of La 0.5 Sr 0.5 CoO 3 system (i.e., a = 10.91 Å, b = 16.37 Å, c = 25.00 Å, 156 atoms). For MoSe 2 slab model, a 2 × 5 × 1 supercell structure of orthorhombic unit cell for 2H-MoSe 2 was created (i.e., a = 11.49 Å, b = 16.78 Å, c = 25.00 Å, 60 atoms). Finally, the LSC&MoSe 2 heterostructure was built by combining the LSC and MoSe 2 slabs with minimized lattice mismatch <3% (i.e., a = 11.20 Å, b = 16.57 Å, c = 35.00 Å, 216 atoms). The vacuum was sufficiently applied to avoid the self-interaction in z-direction.
Surface energy calculations. To evaluate the relative stability of complementary terminations in the LSC (001) surface, we calculated the surface energy (γ) by the sum of cleavage energy (γ u ) and relaxation energy (γ r ), which was previously reported by Heifets et al 62 ., The cleavage energy can be obtained as follows, where E u slab ðCoO 2 À t:Þ and E u slab ððLa; SrÞO À t:Þ are unrelaxed CoO 2 -and (La,Sr) O-terminated slab energies, E bulk is the total energy of bulk unit cell, N is the formula unit of slab models, and A is the surface area. The factor of four in the denominator arises from the four cleaved surfaces of two terminations. Next, we can calculate the relaxation energies for each CoO 2 -t. and (La,Sr)O-t. as follows, γ r ¼ ðE r slab ðXÞ À E u slab ðXÞÞ=2AðX ¼ CoO 2 À t:orðLa; SrÞO À t:Þ ð 3Þ where E r slab ðXÞ is a slab energy after relaxation.

Data availability
The data measured, simulated, and analyzed in this study are available from the corresponding author on reasonable request.