Oxygen Evolution Activity of Amorphous Cobalt Oxyhydroxides: Interconnecting Precatalyst Reconstruction, Long‐Range Order, Buffer‐Binding, Morphology, Mass Transport, and Operation Temperature

Nanocrystalline or amorphous cobalt oxyhydroxides (CoCat) are promising electrocatalysts for the oxygen evolution reaction (OER). While having the same short‐range order, CoCat phases possess different electrocatalytic properties. This phenomenon is not conclusively understood, as multiple interdependent parameters affect the OER activity simultaneously. Herein, a layered cobalt borophosphate precatalyst, Co(H2O)2[B2P2O8(OH)2]·H2O, is fully reconstructed into two different CoCat phases. In contrast to previous reports, this reconstruction is not initiated at the surface but at the electrode substrate to catalyst interface. Ex situ and in situ investigations of the two borophosphate derived CoCats, as well as the prominent CoPi and CoBi identify differences in the Tafel slope/range, buffer binding and content, long‐range order, number of accessible edge sites, redox activity, and morphology. Considering and interconnecting these aspects together with proton mass‐transport limitations, a comprehensive picture is provided explaining the different OER activities. The most decisive factors are the buffers used for reconstruction, the number of edge sites that are not inhibited by irreversibly bonded buffers, and the morphology. With this acquired knowledge, an optimized OER system is realized operating in near‐neutral potassium borate medium at 1.62 ± 0.03 VRHE yielding 250 mA cm−2 at 65 °C for 1 month without degrading performance.


Introduction
The oxygen evolution reaction (OER) is key to supply electrons and protons for the electrocatalytic formation of sustainable fuels. [1] The performance of it in near-neutral media has 1. Do the precatalyst nature and the buffer type used for the reconstruction influence the properties of the finally formed CoCat catalyst? 2. How can the activity difference of various CoCat phases be explained on a molecular level and how can the OER properties of these phases be optimized? 3. What causes the different magnitudes of mass-transport limitations of various CoCat phases and how can they be minimized? 4. Can noble-metal-free OER electrodes be realized that, in near-neutral medium, can achieve high current densities at overpotentials in the range of those in strongly acidic or alkaline media?
To answer these research questions, we synthesized a set of four different OER catalysts that all contain a CoCat like structure but have different electrocatalytic properties. Two of these catalysts are the previously reported CoP i and CoB i . The other two catalysts were obtained by the reconstruction of a layered cobalt borophosphate model precatalyst, Co(H 2 O) 2 [B 2 P 2 O 8 (OH) 2 ]·H 2 O (called CoBP, for a structure description see Figure S1, Supporting Information), in borate or phosphate buffer. CoBP is ideal for this purpose, as it contains equal quantities of borate and phosphate in its structure and previously borophosphate derived catalysts showed excellent OER properties. [26][27][28] The aim of this manuscript is especially to interconnect different phenomena affecting the OER performance to develop a model consistent with previous reports and able to explain the activity differences and mass-transport limitations of the four CoCat phases and two cobalt oxides (CoO and Co 3 O 4 ). In this regard, the precatalyst reconstruction conditions, long-range order, buffer-binding, and morphology are considered.
Herein, we find that, during the OER, CoBP can be fully reconstructed into two different CoCat phases with distinct catalytic properties depending on the buffer (KP i and KB i ). Comparing these two catalysts with CoP i and CoB i shows that all catalysts have the same short-range order. Nevertheless, significant differences in the catalytic activity, Tafel slope, linear Tafel range, the buffer-binding and content, the degree of order/domain size, the number of available edge sites, and the morphology were observed. With this data set, we succeed in connecting the different electrocatalytic properties to variations in chemical structure and morphology of the four catalysts. Following these new insights, we designed a system that can perform the OER in near-neutral media at 1.63 V RHE achieving a current density of 250 mA cm −2 for 1 month without activity degradation. Supporting Information), inductively coupled plasma optical emission spectroscopy (ICP-OES, Figure S2 inset, Supporting Information), scanning electron microscopy (SEM) with energy-dispersive X-ray (EDX) mapping ( Figures S3 and S4, Supporting Information), transmission electron microscopy (TEM) with selected-area electron diffraction (SAED, Figure S5, Supporting Information), X-ray photoelectron spectroscopy (XPS; Figure S6, Supporting Information), and X-ray absorption spectroscopy including extended X-ray absorption fine structure (EXAFS; Figure 2c), and X-ray absorption near edge structure (XANES; Figure 2b) analyses. All methods confirm the formation of a pristine CoBP phase.

The Reconstruction of CoBP in KP i and KB i
To investigate the electrochemical properties of CoBP, we electrophoretically deposited it on FTO without any binder (for SEM/EDX see Figures S7 and S8, Supporting Information, and for XAS Figure 2b,c). With the obtained electrode, we conducted chronopotentiometry (CP) experiments at 1 mA cm −2 for 24 h in 0.1 m KP i (pH = 7.20) and 0.1 m KB i (pH = 9.24) buffer ( Figure 2a). The samples of CoBP after 24 h CP in KP i or KB i are called CoBP-P i and CoBP-B i , respectively. In the first 1.5 h of the CP experiment, the potential is decreasing in a similar way for both buffers. After that, it remains constant. The overpotentials at 1 mA cm −2 (η 1 ) of CoBP-P i and CoBP-B i are 379 ± 7 and 352 ± 7 mV.
After 24 h of CP, the initially highly crystalline CoBP phase is X-ray amorphous ( Figure S9, Supporting Information). Furthermore, ICP-OES shows that CoBP-P i contains no boron anymore and CoBP-B i has no phosphorous (inset Figure S9, Supporting Information). Thus, the initially crystalline CoBP phase reconstructed/transformed completely.

Quasi In Situ XAS
To understand this reconstruction, we performed XAS on samples freeze-quenched at η = 400 mV after 15 min and 24 h CP at 1 mA cm −2 (denoted as quasi in situ, see Characterization Details in Supporting Information for more information). [30] The XANES spectra reveal that in both buffers, the cobalt bulk oxidation state increases over time and is 3.2 after 24 h (Figure 2b, for oxidation state determination see Figure S10, Supporting Information). This oxidation state is consistent with quasi in situ measurements on CoP i and indicates that the CoBP phase fully reconstructed into a CoCat phase during operation. [23,30] Figure 6a). b) Co K-edge XANES data of various CoBP samples before measurement and quasi in situ (all samples were freeze-quenched at η = 400 mV) after 15 min or 24 h CP at 1 mA cm −2 . Figure S10, Supporting Information, shows the linear regression used to determine the oxidation states. c) Co K-edge EXAFS plots of the same samples like in (b). The two most important distances of the as-prepared CoBP (Co II -O and Co-P, dashed lines) and the quasi in situ formed CoCat phase (Co III -O and Co III -(O)-Co III , dotted lines) are depicted. The black lines show the simulation. EXAFS simulation parameters and experimental data in k-space are provided in Tables S2 and S5 and Figure S11, Supporting Information. and a CoCat model (Figure 2c and Figure S11 and Tables S2, S5, and S6, Supporting Information). The initial CoBP phase has two characteristic Co II -O and Co II -(O)-P EXAFS peaks (dashed lines in Figure 2c) that are distinguishable from the most pronounced Co III -O and Co III -(O)-Co III peaks of CoCat (dotted lines in Figure 2c). After 15 min, the Co II -O and Co II -(O)-P signal intensity decreased significantly already. For both buffers, EXAFS simulations of these spectra reveal a ratio of Co II -O to Co III -O of three to one indicating that around 25% of the CoBP material has converted into a CoCat related phase. After 24 h, the EXAFS spectra of both materials are consistent with a complete reconstruction into CoCat. A detailed discussion of this data can be found in Section 2.4.3.

XPS and Electron Microscopy
Additionally, ex situ XPS measurements were conducted after 15 min and 24 h of CP (Figures S12-S15, Supporting Information). After 24 h for both CoBP-P i and -B i , the Co 2p XPS reveals a mixed surface oxidation state of Co II and Co III . Such a reduced surface oxidation state compared to the bulk (bulk oxidation state is around three, see Section 2.4.3) is consistent with previous reports on CoCat. [31,32] The P 2p, B 1s, and O 1s spectra confirm the formation of a CoCat phase with either almost only a phosphate (CoBP-P i ) or only borate (CoBP-B i ) surface species. Surprisingly, the XPS measurements after 15 min show that the outer electrode surface did not reconstruct at all. This observation is remarkable, as such reconstructions are widely reported to start at the surface of the particles leading to the formation of core-shell structures, in which the core is the precatalyst and the shell an quasi in situ formed oxidic phase in contact with the electrolyte.
To localize the CoCat phase formed after 15 min, which was identified by XAS, SEM-EDX measurements from the electrode cross section were performed together with visible-light images from the front and reverse sides of the transparent electrode (Figure 3 and Figure S16, Supporting Information). This data reveals that the reconstruction starts near the conducting FTO electrode substrate and from there it proceeds to the top layers of the film. Thus, after 15 min, the newly formed CoCat phase is still covered with crystalline CoBP explaining the XPS results. A reason for this unprecedented kind of reconstruction is that CoBP is an insulator and that an anodic potential is a prerequisite for the reconstruction. Therefore, initially, only the material in direct contact with the conducting FTO is electronically wired to the anode. Thus, the reconstruction begins there and can then proceed at the interface of the crystalline CoBP with the newly formed, electron conducting CoCat to the outer CoBP particles of the film.
It is crucial to consider this manner of reconstruction when post catalytic characterizations on OER (pre)catalysts are performed, as the XPS, pXRD, and electron microscopic data after and before catalysis might be indistinguishable, because the reconstructed phase is hidden under the intact precatalysts. Thus, potentially leading to the false conclusion that a precatalyst does not reconstruct.
Additional to the SEM cross section investigation, further electron microscopic investigations have been performed. The SEM/EDX mappings show a homogeneous distribution of the  CoCat contains more cobalt and potassium and less oxygen than CoBP. The potassium and tin EDX peaks overlap which explains why the FTO is marked in green in the mapping. A deconvoluted EDX spectrum of the FTO shows that no substantial amount of potassium is present. This spectrum also contains an oxygen peak. Nevertheless, it is less intense than the one of the CoBP and CoCat phases, leading to the weak red color in the oxygen mapping for FTO. Overall, the mappings reveal, supported by the data of Figures S12 and S16, Supporting Information, that the reconstruction of CoBP to CoCat did not start at the outer electrode surface but at the FTO CoBP interface.
constituting elements for both CoBP-P i and -B i after 24 h CP (Figures S17-S20, Supporting Information).
TEM images of CoBP-P i and -B i show that the morphology after 24 h CP is a function of the buffer and that CoBP-B i comprises smaller particles (10-250 nm) than CoBP-P i (50 nm to several µm, Figure 4a,b and Figures S21 and S22, Supporting Information). Another difference between the phases can be seen in the SAED ( Figure S23, Supporting Information): for CoBP-P i , no clear diffraction rings are present, but for CoBP-B i rings can be seen. Thus, CoBPB i has a higher degree of long-range order and might be best described as nanocrystalline. This trend is confirmed by high-resolution (HR)-TEM. For CoBP-P i , spherical aberration corrected HRTEM with a field emission gun cathode and a point resolution of 75 pm could not resolve any lattice fringes or atom columns ( Figure  S21, Supporting Information, for a beam damage discussion see Figure S24, Supporting Information). However, for CoBP-B i , the same machine was capable to resolve atomic columns of probably orderly stacking crystalline domains of 1.5-3 nm

Comparison of the Reconstruction of CoBP with CoO and Co 3 O 4 under (Near-)Neutral OER Conditions
As we found that CoBP reconstructs completely under the herein applied OER conditions, we decided to investigate whether the cobalt oxide phases, CoO (rock salt structure) and Co 3 O 4 (spinel structure), behave similarly. In the first hours of the CP measurements, for CoBP, the potential to reach 1 mA cm −2 decreases indicating the continuous formation of an OER active phase. In contrast, for both  oxides, an increase in the required potential is observed at the beginning (Figure 5 inset). Furthermore, after 24 h, the η 1 is almost 100 mV lower for CoBP-P i compared to the two oxides with the same cobalt loading. These differences imply that the behavior of the oxides is substantially different. In this regard, quasi in situ XANES reveals a slight increase in the average oxidation state for CoO and an unchanged oxidation state for Co 3 O 4 ( Figure 5). Further, the smaller edge maximum for both oxides after the OER treatment indicates a minor amorphization. The quasi in situ EXAFS spectra of CoO and Co 3 O 4 after 24 h of OER could be fitted well with the crystal structures of the pristine materials ( Figure S26 and Tables S3 and S4, Supporting Information). Thus, cobalt oxides are more reluctant toward amorphization during OER than CoBP.

OER Activity
To investigate the influence of the catalyst type on the OER, the four amorphous CoCat phases (Figure 6a for an overview) together with the two crystalline cobalt oxides were tested in 0.1 m KP i buffer for 24 h at a low current density, 1 mA cm −2 ( Figure S27, Supporting Information, all electrodes had a cobalt loading of 1.6 ± 0.2 µmol determined by ICP-OES). Their activity declines in the order CoBPB i > CoB To check the influence of the buffer and its concentration during catalysis, Tafel slopes were recorded in 0.1 m and 1 m KP i and KB i using the most active catalyst, CoBP-B i ( Figure S28, Supporting Information). At 0.1 m, no significant difference between KP i and KB i were observed; nevertheless, 1 m KB i buffer yields higher current densities and a wider Tafel range than 1 m KP i (for more information on the different buffers, iR correction, and stirring see Figures S28 and S29, Supporting Information). Thus, we decided to continue our studies in stirred 1 m KB i . Steadystate current-voltage graphs of CoBP-P i , CoBP-B i , CoP i , and CoB i were obtained by iR-corrected 3 min CA measurements at various potentials ( Figure 6b). This data was also used to determine the Tafel slopes (Figure 6c) of all four compounds and normalized by the number of electrons transferred in the reduction peak (Figure 6d, see Section 2.5.1 and the therein discussed Figures S36-S38, Supporting Information, for more information). Under these conditions, again, CoBP-B i is most OER active and the Tafel slopes of the catalysts formed in KB i buffer show similar Tafel slopes (around 56 mV dec −1 ), which are lower than those of the catalysts formed in KP i buffer (around 68 mV dec −1 ). It is noticeable that previous reports described the Tafel slope of CoP i and CoB i to be around 59 mV dec −1 (correlating to 2.3 × RT/F); [12,33] however, when taking a closer look at the Tafel data of ref. [12[ differences between the slopes of CoP i and CoB i are clearly visible. Furthermore, other groups have reported different slopes than 59 mV dec −1 for various CoCat phases already. [34,35] Three trends concerning the activity can be deduced:   Figure S10, Supporting Information, shows the linear regression used to determine the oxidation states.

Quasi In Situ XAS
To gain further insights into the catalyst structure during OER, we performed quasi in situ XAS measurements on all catalysts (CA at η = 400 mV, Figures 7 and 8a). For all systems, XANES reveals a quasi in situ bulk cobalt oxidation state of 3.2 ( Figure 7) and an ex situ oxidation state of 2.9 ( Figure 7). These oxidation states are in line with the ones reported ex-and quasi in-situ for CoCat materials. [30] For all catalysts, the quasi in situ cobalt K-edge Fourier transformed EXAFS spectra indicate a layered structure formed by edge sharing [CoO 6 ] octahedra (Figure 8a-c and Figure S30, Supporting Information), as reported for CoP i and CoB i (CoCat materials in general) previously. [6][7][8] To gain deeper insights and to identify structural differences between the compounds, all spectra were simulated. The model used is based on five coordination shells (Figure 8b). All spectra could be simulated well based on this model (a simulation considering multiple scattering effects and one without is provided in Tables S5 and S6, Supporting Information). The radii of the five shells showed only minor variations between the different compounds confirming the structural model of Figure 8b, which is also consistent with the observed lattice distances of the TEM data ( Figure 4). In contrast to the radii, the populations of the Co-Co shells differ between the catalysts formed in KB i (larger populations) and KP i (Figure 8d,e). The size of a domain, a layer of edge sharing [CoO 6 ] octahedra (Figures 1 and 8b), correlates with the population of the Co-Co shells (for domains of different sizes and their respective populations see Figure S31 and  also previously identified in several comparisons of CoP i and CoB i . [6][7][8] A detailed discussion of the domain size and the EXAFS models can be found in the caption of Table S1, Supporting Information. A domain can be viewed as a nanosized crystallite and larger crystallites enable diffraction. Thus, the observation of electron diffraction for CoBP-B i and its absence for CoBP-P i (Figure 4 and Figure S21, Supporting Information) is consistent with the EXAFS data. Further, in contrast to CoBP-P i , for CoBP-B i , HR-TEM could directly visualize atomic columns of domains that probably stack in an ordered way (Figure 4).
After the initial formation, OER tests in different buffers for 24 h were performed. These tests did not change the XANES and EXAFS spectra of these catalysts ( Figures S32-S34, Supporting Information) showing that the structures remain after initial formation under the investigated conditions.

SEM
To identify differences in the morphology and surface area of the four catalysts, SEM images were taken of the catalyst films with the same cobalt loading (Figure 9). These images reveal large differences in the morphology of the films. CoP i has a comparably flat surface, and its film shows no porosity on the nanometer scale. CoBP-P i 's surface is not flat and it consists of spherical particles with a diameter of several hundred nanometers. CoB i also consists of spherical particles, but with a smaller size than those of CoBP-P i . CoBP-B i still contains spherical particles but is of a much smaller size than CoB i . CoBP-B i has by far the highest surface area. Furthermore, its film with the same loading is at least two times thicker than the other ones emphasizing its larger porosity. In this regard, it is noticeable that the as-deposited CoBP film ( Figure S7, Supporting Information) is thicker than the CoBP-P i and CoBP-B i films formed from it. Additionally, the TEM investigation of CoBP-P i and CoBP-B i (Figure 4 and Figures S21 and S22, Supporting Information) shows a significantly smaller particle size for CoBP-B i consistent with the SEM investigations.

Differences between CoBP-P i , CoBP-B i , CoP i , and CoB i
The OER activity follows the trend CoBP-B i > CoBP-P i ≈ CoB i > CoP i . The atomic structure of the systems is determined by the buffer used during the formation, whereas catalysts formed in KP i have smaller domains sizes and contain more buffer species while those formed in KB i have larger domain sizes and less buffer species. The morphology of the catalysts is affected by the precursor type and the buffer used during the formation, whereas Co 2+ precursor and KP i buffer leads to smaller particles while CoBP precursor and KB i buffer leads to larger ones and flatter films.

Factors for the Different OER Activities of CoBP-P i , CoBP-B i , CoP i , and CoB i
In this section, we first describe the different parameters used to measure the OER performance and interconnect them with the observables and models explaining the performance variations. In the end, we discuss which aspects are most relevant for the direct comparison of the four catalysts.    Figure 1). [36] Furthermore, Bergmann et al. showed that the number of edge sites correlates to the ratio of the current density at the Co II/III and Co III/IV reduction peak minima (only accessible edge sites contribute to the Adv. Mater. 2022, 34, 2207494   Figure 8. a) OER quasi in situ Co K-edge Fourier transformed EXAFS plots of the four catalysts shown in Figure 6a and one ex situ measurement as comparison. The quasi in situ samples were freeze-quenched at η = 400 mV after 24 h CP at 1 mA cm −2 in KB i buffer. Data of the same catalysts freeze-quenched in KP i buffer shows no significant differences ( Figures S33 and S34, Supporting Information). The most important distances with their structural motifs are depicted by the dashed vertical lines. The black curves show the simulation. EXAFS simulation parameters and experimental data in k-space are provided in Figure S30 and Table S5, Supporting Information. b) A domain of a CoCat-like structure. The white rings represent the five distances marked with dashed lines in (a). The gray lines mark the lattice planes that fit to the distances observed in the HRTEM (Figure 4). c) Side view of a domain with the missing distance from the HRTEM (Figure 4) with gray lines representing oxygen lattice planes. d,e) The population numbers of the second coordination shell and the sum of the fourth and fifth coordination shell all shown in (b). The larger these populations are, the bigger is the average domain size ( Figure S31 and Table S1, Supporting Information) and the higher the degree of order. [24,30] For (d) no multiple scattering was considered while for (e) a multiple scattering analysis was performed for the second coordination shell. Table S1, Supporting Information, includes a more detailed discussion of the two approaches with respect to the domain size. Table S6, Supporting Information, shows the simulation parameters including multiple scattering. Co II/III redox activity) and that a higher ratio of edge sites results in lower Tafel slopes. [36] Therefore, we recorded CVs of all four compounds ( Figures S35-S37, Supporting Information). While the samples reconstructed in KP i showed the same current density at the Co II/III and Co III/IV reduction peak minima, the samples formed in KB i buffer showed a larger current density at the Co II/III reduction-peak minima than the Co III/IV one. This observation is consistent with the smaller Tafel slope that was obtained for samples formed in CoB i .
As elaborated, the electrochemical data suggests that CoB i and CoBP-B i have more Co II/III redox active edge sites than CoP i and CoBP-P i . Our EXAFS analysis and previously reported pair distribution function analyses imply a smaller domain size for catalysts formed in KP i buffer compared to those formed in KB i , [6][7][8] probably because P i binds strongly at the edge sites inhibiting the further domain growth. [37] A smaller domain size correlates to more edge sites and thus should result in a higher Co II/III redox activity, better OER performance, and a lower Tafel slope. [38] To understand this apparent contradiction, the different binding modes of phosphate and borate on the catalysts' edge sites must be considered (for a detailed discussion and reaction mechanism see Figure S37, Supporting Information). [35,37,39] As our ICP-OES results of Table 1 and previous reports have shown, P i binds strongly and under our OER conditions irreversibly to the [CoO 6 ] domain edge sites, while B i can easily be dissociated ( Figure S39, Supporting Information). [37] The oxygen atoms of the phosphate group replace hydroxides in µ 2 -OH-bridged Co III moieties. Thus, P i binding hinders redox activity of these sites by disabling deprotonation in the proton-coupled electron transfer [40] and consequently inhibits their OER activity. We note that the catalysts reconstructed in KP i also contain twice as many buffer species than those reconstructed in KBi (Table 1) and that the phosphate groups of KP i remain in the catalyst during catalysis, in contrast to the borate ions of KB i that exchange, when operated in KP i buffer (Table 1). Thus, for the herein discussed cases, the inhibition by the buffer binding dominates the effect of the domains size on OER available edge sites.

Total Number of Redox Active Sites
Herein, in regimes where mass-transport limitations are negligible, catalysts with the same Tafel slope (same ratio of Co II/III to Co III/IV redox active sites, e.g., CoP i and CoBP-P i ) show different OER activities at the same loading (Figure 6b,c). These variations must originate from differences in the number of OER active sites per total cobalt loaded on the substrate. As OER active sites must be able to change their redox state during catalysis, the number of electrons transferred in the reversible Co II/III and Co III/IV redox peaks (e − redox ) has often been found to be proportional to the number of OER active sites (in the previous paragraphs, the ratio of the Co II/III and Co III/IV redox active sites was considered not their total number). e − redox qualitatively follows the trend of the four catalysts' activities (for two kinds of normalizations see Figures S36-S38, Supporting Information). Thus, it is one aspect to explain the activity differences between the samples formed in the same electrolyte (same Tafel slope). However, a normalization of the activity data by this quantity shows still significant differences revealing that other effects are also important, as described in the previous and following paragraphs.

Morphology: Surface Site versus Bulk Site Activity
A previous report on the difference between the bulk activity of CoP i compared to its surface activity (around 15 nm) showed that the surface activity is higher per loaded cobalt atom; nevertheless, at high catalysts loadings, the bulk activity dominates, as dramatically more cobalt sites are in the bulk compared to the surface. [10] We suggest that the difference in the activity between bulk and surface cobalt site does not mainly originate from different natures of active sites, as 1) no structural differences between the near-surface and the bulk phase has been observed, [10,11] as 2) the Tafel slopes and CV shapes of thin nearsurface activity dominated CoP i films are the same as those of films more than a magnitude thicker, [12,41] and as 3) we could only observe a small difference between the Tafel slope of a CoP i film with a loading of 0.04 µmol compared to 1.6 µmol ( Figure  S40, Supporting Information). The SEM images ( Figure 9) show remarkable differences in the morphologies and surface areas (catalyst-electrolyte interface) of the four catalysts. The catalysts with a larger surface area will benefit from the higher activity of the surface cobalt sites. In agreement with this, the surface area trend found herein agrees with the OER activity trend, even in regimes where mass transport is negligible. We note here that meaningful double layer (C dl ) or electrocatalytic active surface area (ECSA) measurements are challenging to obtain of these catalysts and are not reported herein or in the literature (for a detailed discussion see Figure S41, Supporting Information). Furthermore, compared to (near-)surface active catalysts, surface area normalizations are not suitable to obtain an intrinsic activity and the surface area is poorly defined and not straightforwardly connected to the catalytic performance in general, as within the whole electrolyte-permeable catalyst an interface between electrolyte/substrate and active sites exists and not only on the outer surface.

Mass Transport: Linear Tafel Range and Surface Area
The discussion of intrinsic activity and the number of active sites in the last two subchapters focused on low current densities, where mass transport is negligible, as linear Tafel slopes are still preserved. However, at higher current densities, such limitations can inhibit the OER. [11,12,[41][42][43][44] Mass-transport limitations can arise from insufficient electron transport or proton transport (as well as oxygen bubble detachment which is not discussed herein). Recently, it has been proven by quasi in situ conductivity measurements that electron transport is not a limiting factor. [12] However, proton transport has been found to be a limiting factor in various previous reports. [11,12,41,42] In a stirred setup ( Figure S29, Supporting Information), it can arise 1. In the electrolyte or electrode-electrolyte interface in form of a low-pH zone (pH gradient); [42,43,45,46] 2. Within the catalyst film. [12,41] Concerning (1), at higher potentials, we could observe an activity increase when using 1 m instead of 0.1 m KB i ( Figure S29, Supporting Information) indicating buffer dependent protontransport limitations. ICP-OES (Table 1) shows that the buffer concentration inside the film does not significantly increase when the electrolyte concentration is changed from 0.1 to 1 m KB i . Thus, (2) should not depend on the buffer concentration, and (1) must be responsible for the buffer concentration dependency of the activity. Regarding (1), in the simplest diffusion model of a stirred electrocatalytic solution, the stirred bulk solution has a constant concentration of all constituents and on the surface of the electrode exists a diffusion layer. [45,46] A higher buffer concentration will increase the concentration of proton carriers in this diffusion layer explaining the observed dependency. Alternatively, a valid explanation is also that the proton exchange between the catalyst film and the electrolyte buffers is the rate limiting step of the proton transport and that it depends on the buffer concentration.
Concerning (2), as such limitations will be the only ones that depend on the catalyst film thickness, they can be proven by comparing the linear Tafel range of two electrodes with different loadings ( Figure S40, Supporting Information). The linear Tafel range for CoP i catalysts with a lower loading extends to current densities that are over a magnitude larger than those of CoP i catalysts with 40-times higher loading. This comparison suggests that proton transport within the catalysts is also an important factor.
The proton mass transport of (1) and (2) can both be improved by increasing the surface area/outer catalystelectrolyte interface. This conclusion explains why CoBP-B i is considerably more active at higher current densities and has a wider Tafel range than the other catalysts (Figure 6c), as CoBP-B i has the highest surface area (Figure 9). In general, the Tafel range ( Figure 6c) and the surface area (Figure 9) of the catalysts show the same trend. Figure 10 summarizes the parameters causing the activity differences found between the four herein investigated catalysts. Catalysts formed from CoBP precursor are more active than those formed from Co 2+ due to their different morphology that creates a larger surface area (interface between the bulk electrolyte and outer catalyst) for CoBP derived systems. This difference leads to more surface sites, which have been shown to be more active than bulk sites, [10] and is beneficial for proton transport. Catalysts formed in KB i buffer are more active than those formed in KP i buffer, 1) due to differences in the morphology resulting in a larger surface area for KB i derived phases and 2) due to more accessible edge sites, as P i irreversible binds to edge sites leading to their inhibitions. Concerning point (2), it is worth mentioning that KB i -derived phases have a larger domain size than KP i derived ones, which results in less edge sites; however, the effect on available edge sites is dominated by the P i binding. Note that KP i derived phases have a molar Co to P i ratio of around 1-0.4. Thus, even if every second cobalt site is on the edge, 80% of these sites could be inhibited by KP i .

An Optimized System Operating Efficiently and Long-Term Stable at High Current Densities at Elevated Temperature
Due to the mass-transport limitations and inferior kinetics of the OER in near-neutral media compared to a strongly alkaline and acidic one, it is challenging to achieve high current densities at reasonable overpotentials. [14,47] Taking our best catalysts (CoBP-B i ) and considering our deduced concepts and previous reports, [44,[48][49][50] we performed the following optimizations of our system to meet this challenge: 1. Loading the catalyst on porous nickel foam. 2. Increasing the cobalt loading to 22 µmol cm −2 . 3. Raising the buffer concentration to 4 mol kg −1 . 4. Elevating the temperature to 65 °C.
Comparing this system to CoP i and CoBP-B i at room temperature on FTO (Figure 11a), we achieved an activity increment of around 475-times (CoP i ) and 40-times (CoBP-P i ) at 1.625 V RHE . The optimized system reaches 100 mA cm −2 at 1.585 V RHE which is still inferior but in range of the most active OER catalysts in strongly alkaline and acidic media under steady-state conditions. [4,47,51,52] In general, these results show that the gap between near-neutral and strongly alkaline and acidic OER is likely to decrease at elevated temperatures, as those temperatures partly compensate the proton-transport disadvantage of buffers compared to hydroxide and protons. [44,53] Furthermore, the solution resistance is decreased ( Figure S28c, Supporting Information) and the autoprotolysis of water is enhanced. [54] Long-term stability at higher potentials is a critical issue, as, in near-neutral media, cobalt-based catalysts rely on a selfhealing mechanism, which prevents complete dissolution of the catalyst by continuous redeposition (but see also ref. [55]). [56] We tested the long-term stability of our optimized system at a current density of 250 mA cm −2 at 65 °C for 1 month (Figure 11b). During the CP measurement, this current density was achieved at 1.62 ± 0.03 V RHE without any observable degradation of the activity over 1 month proving that long-term stable, near-neutral water splitting at elevated temperatures, high efficiencies, and industrially relevant current densities is realizable.

Conclusion
Herein, we have synthesized the cobalt borophosphate, CoBP, and, for the first time, investigated toward its OER activity. Our ex-and quasi in-situ analysis uncover an unreported kind of reconstruction of this material which does not start at the outer surface of the electrode but at the FTO-CoBP interface. After a prolonged time, CoBP could be fully reconstructed into two different CoCat phases depending on the buffer (CoP i or CoB i ) that are significantly more OER active than CoP i and CoB i under the same conditions (Figure 6a for an overview). With these four different CoCat phases, we performed a detailed electrochemical, structural, and morphological analysis enabling us to answer the research question (1)-(4) of the introduction: 1. Both the precatalyst type and the nature of the buffer used during the reconstruction are relevant for the OER activity Figure 10. Summary of the factors causing the activity difference of the four herein investigated fully reconstructed catalysts, which were obtained either in KP i or KB i electrolyte and either from Co 2+ (aq) or CoBP as precatalyst. The blue spheres are the cobalt atoms, the red ones oxygen, the purple-grey ones phosphorus, and the light-pink ones hydrogen.