Covalency-reinforced oxygen evolution reaction catalyst

The oxygen evolution reaction that occurs during water oxidation is of considerable importance as an essential energy conversion reaction for rechargeable metal–air batteries and direct solar water splitting. Cost-efficient ABO3 perovskites have been studied extensively because of their high activity for the oxygen evolution reaction; however, they lack stability, and an effective solution to this problem has not yet been demonstrated. Here we report that the Fe4+-based quadruple perovskite CaCu3Fe4O12 has high activity, which is comparable to or exceeding those of state-of-the-art catalysts such as Ba0.5Sr0.5Co0.8Fe0.2O3−δ and the gold standard RuO2. The covalent bonding network incorporating multiple Cu2+ and Fe4+ transition metal ions significantly enhances the structural stability of CaCu3Fe4O12, which is key to achieving highly active long-life catalysts.


Results
Catalytic activity of Fe 4 þ -perovskites. The OER catalytic performance of the Fe 4 þ -perovskites CFO, SFO and CCFO is compared with that of BSCF and RuO 2 , together with a nominally isoelectronic perovskite, LaMnO 3 (LMO), in Fig. 2. The tetravalency of the Fe ions for CCFO was confirmed via Fe K-edge X-ray absorption spectra ( Supplementary Fig. 2). To exclude geometrical effects, the current density per oxide surface area mA cm À 2 oxide À Á , in which the surface areas were determined using Brunauer-Emmett-Teller (BET) analysis, was adopted as the vertical axis in the voltammograms in this study (Supplementary Note 2, Supplementary Fig. 3, and Supplementary Table 1). As the OER activity of BSCF is strongly dependent on the synthesis conditions 9,12,19,20 , two different BSCF samples calcined at 950 and 1,100°C (BSCF 950 and BSCF 1100 , respectively) were tested. Figure 2a shows the obtained linear sweep voltammograms, and it can be seen that CCFO exhibits the highest OER activity of the catalysts tested here. The overpotential of CCFO for OER (Z ¼ 0.31 V), which was determined based on the onset potentials at 0.5 mA cm À 2 oxide , is the lowest of the examined substances, while its specific activity (current density at 1.6 V versus RHE) is the highest (Fig. 2a,b). The CCFO Tafel slope (51 mV dec À 1 ) is as low as those of the SFO and CFO (63 and 47 mV dec À 1 , respectively; Fig. 2c). The excellent properties of CCFO, which are attributed to the presence of the Fe 4 þ ions, exceed those of BSCF 1100 and RuO 2 . On the other hand, the BSCF performance reported by Suntivich et al. 9 is superior to that of the CCFO examined in this study. This is attributed to the difference in the synthesis conditions of these particular samples, because BSCF has exhibited different OER performance in a number of reports 9,12,19,20 (see also, the XRD profiles of the tested BSCF samples in Supplementary Fig. 4). Thus, we are unable to definitively conclude that CCFO exhibits superior OER performance to BSCF in this study, and further investigations are required in order to compare the intrinsic OER activities of these substances. However, the intrinsic superiority of the Fe 4 þ ions can be confirmed by comparison between AFe 4 þ O 3 (A ¼ Ca, Sr) and non-Fe 4 þ oxides. The overpotentials of SFO and CFO are 0.41 and 0.39 V, respectively, which are comparable to those of BSCF (Z ¼ 0.38 and 0.36 V for BSCF 950 and BSCF 1100 , respectively) and are lower than that of RuO 2 (Z ¼ 0.49 V). By comparing the specific activities (current densities at 1.6 V, Fig. 2b), it can be seen that SFO and CFO have high activities comparable to those of BSCF 1100 and RuO 2 . On the other hand, LMO exhibits poor OER catalytic activity; its overpotential cannot be defined because of the overly small current density, while the specific activity is only a centesimal fraction of that exhibited by AFeO 3 . This is because the secondary descriptor, electronegativity 2,13 , predominates the OER activity in these cases.
As the B-site ion (Fe 4 þ ) is identical among CFO, SFO and CCFO, the excellent OER activity of CCFO can be attributed to its particular structure, that is, it is a quadruple perovskite incorporating ordered A 0 -site Cu ions. It should be noted that a reference Cu 2 þ -Fe 3 þ complex oxide examined in this study, CuFe 2 O 4 spinel, did not exhibit high OER activity ( Supplementary Fig. 5), thus, the combination of a Cu 2 þ O 4 square and Fe 4 þ O 6 octahedron is a key factor enhancing the OER activity, as will be discussed below.
Stability of Fe 4 þ -perovskites. The Fe 4 þ -perovskite stability under OER conditions was tested. Fig. 2d-f show the cyclic voltammograms (CV) of SFO, CFO and CCFO for continuous 100 cycles. In the SFO case, the OER current density is suppressed even in the anodic sweep of the first cycle, because of the degradation of the SFO; this implies the dissolution of metal ions 15 . On the other hand, the increase in current density for 100th cycle is possibly attributed to the increase in electrochemical surface area in amorphization 12 . For the CFO, the OER current density increases slightly in the first B10 cycles, and then gradually decreases over the 100 cycles. As can be seen in the Tafel plots of the 3rd and 100th cycles for the SFO and CFO (Fig. 2g,h), both the Tafel slopes are increased after 100 cycles; this clearly suggests the degradation of the SFO and CFO under OER conditions (also see the increases in overpotentials for 100 cycles in Supplementary Note 3 and Supplementary Fig. 6). Considering the fact that the highest occupied molecular orbitals are dominated by the O 2p orbitals in SFO and CFO, as shown in Fig. 1a, the above results are explained by the trend that the highly elevated O 2p band centre (or the deep Fe 3d orbital) increases the activity but decreases the stability, as suggested for cobalt perovskites 6 . However, CCFO is remarkably stable up to 100 cycles, in spite of the fact that it has the same electronic configuration as SFO and CFO. For CCFO, the current density increases in the first B10 cycles and remains almost unchanged. The CCFO Tafel slope does not vary significantly, rather it improves slightly over the 100 cycles (Fig. 2i). This corresponds to a slight improvement in the catalytic activity. Thus, we conclude that CCFO is an excellent OER catalyst that satisfies both the activity and stability requirements.
To determine the difference in stability for these Fe 4 þperovskites, the surface structures of SFO, CFO and CCFO were investigated using high-resolution transmission electron microscopy (HRTEM) both before and after the 100-cycle OER measurements. Figure 3 shows surface HRTEM images of SFO, CFO and CCFO samples, as-synthesized, as-cast, and after the 100-cycle OER measurements. Well-crystalline surface structures can be observed for all the as-synthesized powders and the bulk crystallinity of all the samples was retained after the 100-cycle OER measurements (see also, the electron diffraction patterns in Supplementary Fig. 7). However, thin amorphous layers (B5 nm) formed on the surfaces of all the as-cast catalysts. Further amorphization gradually occurred during the OER cycles in the case of the SFO and CFO samples, resulting in thick amorphous layers about 20 nm after 100 cycles. These amorphous layers caused suppression of the OER reaction of these two perovskites with decreased current density as shown in Fig. 2d,e. In contrast, CCFO retained the thin amorphous layer (B5 nm) even after 100 cycles, and no erosion was observed. If the amorphous layers isolated only the catalyst surfaces from the electrolyte, the catalytic activities would converge on the lower levels equally. However, the experimental results suggest that the amorphous layers reflect the bulk properties to some extent. The significant improvement in the CCFO stability in comparison with that of AFe 4 þ O 3 can be attributed to the transformation of the covalent bonding networks. In a simple AFe 4 þ O 3 perovskite, the B-site Fe 4 þ ions are covalently bound to the oxide ions, whereas the A-site alkaline-earth metal ions have ionic characteristics 15 . Thus, the A-site ions are easily dissolved in the electrolyte during OER 12 . In contrast, the covalent Cu-O bonds in square-planar units of CCFO aid formation of the covalent bonding network and prevent progressive amorphization during the OER measurements.  In these types of perovskites, the A-sites are occupied by alkaline, alkaline-earth or rare-earth metal ions, the A 0 -sites by Jahn-Teller active ions such as Cu 2 þ and Mn 3 þ , and the B-sites by d-block transition metal ions. 3D electron density maps of SFO (equi-density level: 0.4 Å À 1 ) and CCFO (equi-density level: 0.5 Å À 1 ) were obtained from maximum entropy method analysis of synchrotron X-ray powder diffraction data. The shaded cross-sections indicate the (110) and 1 4 3 0 À Á planes of SFO and CCFO, respectively. The widespread covalent network incorporating the Cu, Fe and O ions is exemplified by CCFO. These illustrations were drawn using the VESTA3 program 26 . The synchrotron X-ray powder diffraction patterns and Rietveld refinement results are shown in Supplementary Fig. 1

Discussion
Here, we demonstrate the structural features of CCFO that are associated with OER catalytic activity. When we assume that the local crystal structures of CCFO are reflected on the surface at a certain level, several factors that increase the catalytic activity are considered. Fig. 4 proposes three possible OER routes for SFO and CCFO. The left route is the conventional Eley-Rideal (ER)-type mechanism for SFO and CCFO. In the ER-type mechanism, OH À adsorbates are bound to B-site Fe ions on the surface (Fig. 4a), in which the rate-determining step is considered to be the formation of the O-O bond (reaction 2) or the subsequent deprotonation (reaction 3) 9   reaction of the B-site ions. In both cases, the electron charges are transferred to the Fe ions efficiently through strong Fe-O covalent bonds. An almost identical mechanism is most likely valid on the Cu-terminated surface of CCFO (Fig. 4b). The Cu and Fe ions can tolerate the redox reactions in the Cu 2 þ /Cu 3 þ and Fe 4 þ /Fe 5 þ states, leading to the stable and high OER activity exhibited by CCFO. It should be noted that the Langmuir-Hinshelwood (LH)-type reaction can occur through the direct formation of the O-O bond between the neighbouring oxygen atoms connected to the nearest neighbouring Fe ions, because of the short distance (Fig. 4c). The oxygen-oxygen distance is shortened to B2.6 Å by heavily bent Fe-O-Fe bonds (B140°) for CCFO. This oxygen-oxygen distance is comparable to that of a-Mn 2 O 3 , in which the LH-type mechanism is thought to dominate 21 . In contrast, the oxygen-oxygen distances for simple cubic perovksite SFO is B3.9 Å because of the linear Fe-O-Fe bonds ( ¼ 180°). This oxygen-oxygen distance is too large to permit the oxygen atoms to interact with each other and form oxygen molecules. In the LH-type reaction, one of the two possible rate-determining steps in the ER-type reaction (that is, the deprotonation of the oxyhydroxide group to form peroxide ions) is skipped, resulting in the acceleration of the reaction. Thus, the realization of the LH-type reaction is another major specificity of CCFO. In summary, the Fe 4 þ -perovskite CCFO exhibits promising OER activity. Further, CCFO has a widespread covalent bonding network that enhances its stability. The cationic arrangements of this substance provide a further increase in the OER activity. These findings indicate that the covalent network consisting of multiple transition metal ions in CCFO plays a crucial role in the activity and stability of the OER catalysis. In addition, the various unexplored A 0 -B ion couplings in quadruple perovskites may provide further high-performance, high-stability and costeffective OER catalysts.

Methods
Sample preparation. SFO, CFO and CCFO were synthesized via a high-pressure synthesis method. LMO and BSCF were obtained using a polymerized method, while CuFe 2 O 4 was synthesized via the inverse coprecipitation method. RuO 2 (99.9%) was used as purchased from RARE METALLIC, Co, Ltd. The sample preparation details are given in the Supplementary Methods.
Characterization. X-ray diffraction patterns of reference oxides were obtained using a laboratory X-ray diffractometer (Rigaku Ultima IV) with Cu Ka radiation. Synchrotron X-ray powder diffraction patterns of the Fe 4 þ -perovskites were obtained at the SPring-8 BL02B2 beamline. Fe K-edge X-ray absorption spectra of the CCFO and Fe 3 þ reference oxides were collected at room temperature and in absorption mode at the SPring-8 BL01B1 beamline. Crystal structure refinements of CFO, SFO and CCFO were conducted based on the obtained synchrotron X-ray powder diffraction data using a Rietveld refinement program RIETAN-FP 22 . Electron density analysis of SFO and CCFO was performed using the Dysnomia maximum entropy method program 23 . HRTEM images were collected using a JEOL JEM-2100F.
Preparation of catalyst inks. The catalyst inks were prepared by reference to the methods reported by Suntivich et al. 9,24 and Jung et al. 7 K þ ion-exchanged Nafioni was used as a immobilizing binder, which did not prevent the transport of dissolved O 2 to the catalyst surface. A B3.33 wt.% K þ ion-exchanged Nafion suspension was prepared by mixing a 5 wt.% proton-type Nafion suspension (Sigma-Aldrich) and 0.1 M KOH aqueous solution at 2:1 by volume. The pH of the 5 wt.% proton-type Nafion suspension was initially B1 and 2 and was changed to B11 after mixing. The catalyst inks of the perovskites and reference oxides (RuO 2 and IrO 2 , Sigma-Aldrich) were prepared by mixing 50 mg of oxide, 10 mg of acetylene black (AB), and 0.3 mL of B3.33 wt.% K þ ion-exchanged Nafion suspension. The volumes of the inks were adjusted to 10 mL by the addition of tetrahydrofuran (Sigma-Aldrich). Thus, the final concentration of the catalyst inks was 5 mg oxide mL À 1 ink , 1 mg AB mL À 1 ink and B1 mg Nafion mL À 1 ink . A rotating ring-disk electrode (BAS Inc, Japan) consisting of a glassy carbon (GC) disk of 0.4 cm in diameter and a Pt ring part of 0.7 and 0.5 cm outer and inner diameter, respectively, was used as a working electrode after mirror polishing with 0.05 mg alumina slurry (BAS Inc). Then, 6.4 mL of catalyst ink was drop-cast onto the GC disk part (0.2 Â 0.2 Â p cm 2 ). The catalyst layer on the GC disk part was dried overnight in vacuum at room temperature, and was composed of 0. 25  For the catalysis evaluation of the perovskites for OER, the potential of the catalystmodified GC part was controlled from 0.3-0.9 V versus Hg/HgO (1.226-1.826 V versus RHE) at 10 mV s À 1 . For all measurements, the current density was iR-corrected (R ¼ B43 O) using the measured solution resistance, and capacitancecorrected by taking the average between the anodic and cathodic scans 9 . All the OER currents are shown relative to the surface area of the oxide catalysts estimated using BET analysis (BELSORP-max, BEL Japan, Inc, Japan).