Highly efficient water oxidation via a bimolecular reaction mechanism on rutile structured mixed-metal oxyfluorides

Mixed-metal oxides are generally considered to be the highest-performance catalysts for alkaline water oxidation. Despite significant efforts dedicated to understanding and accelerating their efficiency, most works have been limited investigations of Ni, Co, and Fe oxides, thus overlooking beneficial effects of hetero-anion incorporation. To this end, we report on the development of Co0.5Fe0.5O0.5F1.5 oxyfluoride materials featuring a rutile crystal structure and porous morphology via a scalable and green synthetic route. The catalyst surface, enhanced through electron withdrawing effects imparted by the fluoride ions, give rise to highly effective catalytic sites for electrochemical water oxidation. In particular, their performance across metrics of Tafel slope (27 mV/dec), mass activity (846 A/g at 1.53 V vs. RHE), turnover frequency (21/s at 1.53 V vs. RHE), overpotential (220 mV for 10 mA/cm), and stability (27 days of continuous operation) largely surpasses most known Cobased catalysts. Mechanistic studies suggest that this performance is driven by a bimolecular, oxygen coupling reaction mechanism through proximal active sites on the catalyst surface, thus enabling a new avenue for achieving accelerated oxygenic electrocatalysis.


Introduction
Impacts on global climate and environmental decay are increasingly evident as consequences of human activities 1 . Mitigating these effects will ultimately be underpinned by the transition to sustainable means of energy harvesting and consumption, within which electrochemical technologies stand to play a critical role [2][3][4] . Widespread adoption of electrochemical water and CO2 reduction to form H2 or C-based fuels is hampered by the kinetics of the oxygen evolution reaction (OER) and thus, much work in the field is dedicated towards OER catalyst discovery 5 . To this end, mixed-metal oxides have emerged as the top class of materials for this reaction in alkaline conditions 6 . The recent decade of work in the field has yielded important gains in understanding the roles of incorporating Fe impurities, lattice strain, substrate effects, interfaces and exfoliation. [7][8][9][10][11] In an orthogonal direction, the modification of the catalytically active sites through hetero-element incorporation has yielded substantial gains in performance over purely Ni/Co/Fe oxides, though most efforts focus on cationic incorporation 12 . In contrast, anionic incorporation is seldom used, largely due to synthetic challenges, especially in the direction of fluoride substitution. However, initial reports suggest that hetero-anionic compounds hold much promise as OER catalysts, though the factors driving these observations are not fully settled [13][14][15][16][17][18][19] . Often, metal oxides catalyze the OER through sequential proton-coupled electron transfer steps. This inevitably limits their performance via scaling relations as all intermediates share a M-O bond. 20 Breaking these scaling relations for enhanced OER performance can be accomplished by accessing new reaction mechanisms such as radical oxygen coupling and lattice oxygen involvement, and the discovery of new OER catalyst compositions ultimately underpins access to such reaction pathways.
Numerous synthetic routes have been used to prepare functional transition metal oxyfluorides to control the chemical composition with specific fluorination degree, particle size, morphology, and stabilization of desired crystal structures [21][22][23][24][25] . In this work, we report on the two-step preparation of a new cobalt iron oxyfluoride with a rutile type structure, Co0.5Fe0.5O0.5F1.5, obtained by calcination of the hydrated fluoride CoFeF5(H2O)7. Subsequent electrocatalytic evaluations unveiled that Co0.5Fe0.5O0.5F1.5 is one of the highest performing cobalt-based OER catalysts reported to date, considering metrics of overpotential, Tafel slope, mass activity, and stability. Complementary lines of computational and mechanistic studies were carried out to elucidate that the existence of a bimolecular reaction mechanism that is deemed to be the origin of the high efficiency of Co0.5Fe0.5O0.5F1.5.

Catalyst Synthesis and Characterizations
A soft chemistry method was chosen to generate Co0.5Fe0.5O0.5F1.5. A detailed description can be found in the supporting information. In brief, the catalyst was prepared in a controlled thermal decomposition of a hydrated fluoride precursor in mild conditions. The precursor CoFeF5(H2O)7 was precipitated by thermally assisted evaporation of a hydrofluoric acid (HF) solution containing metallic salts. The thermal decomposition of CoFeF5(H2O)7 was followed by combined thermogravimetric analysis (TGA) and thermodiffraction. Three thermal events appeared on the TGA curve and thermodiffractograms that correspond to the structural evolution of CoFeF5(H2O)7 across four domains (1) 100°C-220°C: Co 2+ Fe 3+ F5(H2O)7 → Co 2+ Fe 3+ F5 + 7H2O (37.5%, 38.0%) (2) 220°C-280°C: Co 2+  The hydrolysis reactions (2) and (3) are further confirmed by the absence of significant weight loss in the TGA profile performed under dry air at temperatures from 220°C (Fig. 1a).
A disordered model based on metal atoms located in 2a Wyckoff positions (WP) (Co: 50%, Fe: 50% confirmed by SEM-EDS, Fig. S7c) and the 4f WP statistically occupied by oxygen atoms (25%) and fluorine atoms (75%) was used for Rietveld refinement of Co0.5Fe0.5O0.5F1.5 PXRD pattern (space group P42/mnm). A good fit was obtained in spite of the low crystallinity with the cell parameters: a = b = 4.6916(5) Å, c = 3.1098(3) Å, V = 68.450(1) Å 3 (Fig. 1e). The broadening of the Bragg diffraction peaks, compared to the diffraction pattern of the precursor (Fig. S1), reveals small coherent diffraction domains. Indeed, an emerging porosity is observed which is likely related to the precursor's decomposition during which HF and H2O gas molecules act as a self-generated porogen ( Fig. 1c). Upon thermal treatment SBET increases from 3 m 2 .g -1 to 24 m 2 .g -1 confirming the porogen effect of the H2O and HF released. N2 adsorption/desorption isotherms shows type IV hysteresis corresponding to a mesoporous structure and the BJH pore-size distribution analysis (Fig. 1d) confirms an average pore diameter less than 10 nm, in good agreement with the TEM observation ( Fig. 1c). 57 Fe Mössbauer spectra (Fig. 1f)  and 712.7 eV (Fig. 1g). These values match that of Fe 3+ and fall between that of Fe in a pure oxide and a pure fluoride environment. The Co 2p3/2 peak has a primary feature at 781.7 eV, with a satellite at 786.4 eV which is in between the expected ranges of pure Co fluorides and oxides (Fig. 1h). These measurements illustrate the strong inductive effects imparted through the fluoride components within the lattice hypothesized earlier to influence the catalytic properties of the composite material.

Electrocatalytic Performance
The catalytic properties of Co0.5Fe0.5O0.5F1.5 were next evaluated through a host of electrochemical techniques. In a standard 3-electrodes configuration, a cyclic voltammogram (CV) of the catalyst deposited onto a carbon paper electrode in 1M KOH electrolyte showed a redox peak centered at 1.1 V vs. the reversible hydrogen electrode (RHE). This peak was attributed to the Co(II/III) redox couple ( Fig. 2a and inset). Electrocatalytic current initiated around 1.4 VRHE requires 220 mV overpotential to reach 10 mA/cm 2 and 265 mV to reach 100 mA/cm 2 . On the reverse scan, a reductive peak around 1.4 VRHE was observed and attributed to the reduction of Co(IV) back to Co(III) and was more evident with increasingly positive scan limits in the CV (Fig. 2a inset). These assignments are in agreement with previous observations of Co-oxides the literature 26,27 while Feoxides typically do not show such redox behavior under these conditions 28 . The oxidation peak of the Co(III/IV) redox couple was not as visible as it was overshadowed by the catalytic current which initiated alongside of the Co(III/IV) oxidation. This signifies that Co(IV) was likely involved in the OER catalytic cycle. We note that there is probably a surface restructuration that occurs as the redox behavior of the 1 st CV differs from the 2 nd , though no further changes occur afterward (Fig. S10a,b).
This indicates that there may be an initial surface restructuration occurring toward a stable surface induced by the application of positive potentials and the environment of Co becomes O/OHterminated at the surface as suggested by the changes observed on XPS spectra ( Fig. S10c-f).
Moreover, the peak shift of the Co +2 oxidation into Co +3 to lower value between the 1 st and the 2 nd cycle is in agreement with an exchange OH/F at the surface; the M-F is more ionic compared to M-O leading to higher potential 29 .
After 1.4 VRHE, the catalytic current rose exceptionally fast and this was reflected in the low Tafel slope of 27 mV/dec (Fig. 2b). A similar Tafel slope was measured if the catalyst was deposited on a glassy carbon rotating disk electrode (RDE) at 0.1 mg/cm 2 in a configuration with minimized masstransport limitations. In the same configuration, we quantified that mass-activity reached 846 A/g at 300 mV overpotential (Fig. 2c). Similarly, the turnover frequency (TOF) was measured to be 21 /s at this voltage, which was quantified by the redox-active Co atoms that were deduced through integration of the redox peaks in the CV assuming the Co is indeed the OER active site. If measured by the total mass loading, the TOF was approximately 0.46 /s, suggesting that approximately 1 out of every 45 Co atoms was electrochemically active. The fluorine component was deemed important for the catalytic performance of Co0.5Fe0.5O0.5F1.5 because when the catalyst was annealed at 500 o C to convert to CoO/CoFe2O4, the Tafel slope, onset potential and overall activity significantly diminished (Fig. S11).
The stability of Co0.5Fe0.5O0.5F1.5 was next evaluated through chronopotentiometry, switching between several current densities over the course of 27 days without pause (Fig. 2d). Rotating ring disk electrode (RRDE) measurements confirmed that the current was indeed originating from the OER  Chronopotentiometric evaluation of Co0.5Fe0.5O0.5F1.5: the activity did not noticeably decrease through 27 days of measurement (d).

Mechanism Study
In order to probe the reaction mechanism and understand the roles of Co and Fe atoms for Co0.5Fe0.5O0.5F1.5, we first performed density functional theory (DFT) computations. Here, we chose (110) facet because of its lowest surface energy and thus the most stable surface in rutile structures. 24,30 Two reaction pathways are considered here, adopting different O-O bond formation strategies, which are the water nucleophilic attack (WNA) and the interaction of two metal-oxo units (I2M) (Fig. S13).
These two mechanisms are reported to be the primary reaction pathways for the OER. 31  From the energy profiles ( Fig. 3c and f), it is apparent that the I2M mechanism has a lower energy barrier (4.02 and 3.51 eV, for structures 1 and 2, respectively) for its rate-limiting step (RLS, which is the desorption of O2) and thus is more favorable than the WNA mechanism (7.77 and 7.91 eV, for structures 1 and 2, respectively) for both model catalysts without any applied potentials (i.e., at U = 0 V). For both WNA and I2M mechanisms, their first two steps are the same electrochemical reactions with electron transfer. However, the last two steps in the I2M mechanism are chemical reactions without any electron transfer, whereas the last two steps in the WNA mechanism still contain electron transfer. Therefore, a large enough applied potential can theoretically suppress the limiting energy barrier in the WNA mechanism without any impact on the limiting energy barrier in the I2M mechanism to compel the two limiting energy barriers to equal. However, due to the large difference between the two limiting energy barriers at U = 0 V, only if an applied potential that is greater than 3.75 V for CoFeOF(110)-1 and 4.40 V for CoFeOF(110)-2, allows the reaction pathway shift from the I2M to the WNA mechanism. Since the applied potentials employed in this study are much smaller than 3.75 V and 4.40 V, we posit that the I2M mechanism containing an *O-O* coupling is the reaction pathway for Co0.5Fe0.5O0.5F1.5 catalyst.
Furthermore, a comparison of the energy profiles on both model catalysts, illustrates that CoFeOF(110)-2 has a lower limiting energy barrier at U = 0 V and requires a higher applied potential to shift from the I2M pathway to the WNA pathway. This implies that the structure of CoFeOF(110)-2 following the WNA pathway is more favorable for the OER, and the existence of  We next took to further elucidating the mechanism of the OER on Co0.5Fe0.5O0.5F1.5 catalyst surfaces through pH-dependent electrochemical experiments. CVs of the catalyst were recorded from pH 11 to nominally pH 14 ( Fig. 4a). At pH 11, the current attributed to Co redox peaks was considerably lower until approximately pH 12, whereas the reduction of Co(IV) increased continually until pH 14. Only at pH 14, the same amount of charge was integrated under the Co(IV) reduction as the Co(III) reduction (Fig. 4b).
In addition to the rise in Co(IV), the catalytic current also increased more rapidly at higher pH electrolytes, quantified here by a decrease in Tafel slope (Fig. 4c). Finally, the absolute potential (here plotted on vs. the Standard Hydrogen Electrode, SHE) of the Co redox potentials and the catalytic current exhibited differing shifts as a function of electrolyte pH (Fig. 4d).
First, the Co(II/III) redox couple shifted 91 mV/pH, indicative of a 2e -, 3H + process. This super-Nernstian behavior often signifies that the redox transition is coupled with not only the deprotonation of a group adsorbed on the Co, but also with a more extended deprotonation of sites within the lattice with a pKa of approximately 12, as at this pH value, the magnitude of the redox wave stops increasing 26,32,33 . On the other hand, the Co(III/IV) redox wave shifted 63 mV/pH, indicating that this process was a simple 1e -, 1H + (or 2e -, 2H + ) process. The current necessary to attain 10 mA/cm 2 , as measured in galvanostatic mode, showed a shift of 102 mV/pH. Such a dependence of pH on OER catalysis has previously been attributed to a decoupling of proton transfer, electron transfer, and catalytic steps in the OER cycle and has often implied the active role of lattice oxygen in reaction pathway 34,35 . The next measurement in the series was the use of chemical probes to deduce the presence of particular surface intermediates. We first introduced 1M methanol into the electrolyte as adsorbed nucleophilic *OH groups on metal oxide surfaces were previously found to react with alcohols 36 . Indeed, we observed a slow onset of methanol oxidation just past the Co(II/III) redox transition pointing to the initiation of reactive *OH coverage at this potential value (Fig. 4e). On the other hand, using 1M tetramethylammonium hydroxide (TMAOH) as the electrolyte, previously argued to specifically interact with active oxygen species and thus hinder the OER cycle 37  Nernstian and super-Nernstian pH shifts (d). *OH was evidenced above 1.1VRHE through an increase in current in the presence of methanol while active oxygen species were evidenced when TMA cations were added to the electrolyte (e). Isotope measurements (f, g) indicated that the catalyst performance is most hampered by K 18 OH rather than KOD. Raman spectroscopy indicated hydration of Co0.5Fe0.5O0.5F1.5 upon the application of positive potentials (h) shows three distinct species forming as a function of applied potential.
With above data in mind, we were able to construct a plausible mechanism for the OER on Co0.5Fe0.5O0.5F1.5 surfaces (Fig. 5). First, a 2e -, 3H + process occurs at 1. isotopic measurements. A plausible set of electron and water/hydroxide transfer steps was added finally to complete the OER cycle. Overall, the data implies a decoupling of key proton transfer, electron transfer and chemical steps and a lattice oxygen involving mechanism on Co0.5Fe0.5O0.5F1.5, a departure from the conventional route of purely proton-coupled electron transfer steps, though one that is highly efficient. Further, the active site here is deemed to be a Co-Fe bimolecular site. Returning back to conventional limitations in OER catalysis, this mechanism, which circumvents the standard PCET route, appears to be the origin of the Co0.5Fe0.5O0.5F1.5 exceptional performance. In fact, in recent years, key high performing alkaline OER catalysts also bypassed the standard PCET mechanism. 27,32,35,37,44,45 While the exact nature of every step and free-energy landscape of the OER on different crystal facets and active sites is not yet elucidated, nor are the precise effects of the fluorine components on the Co and Fe species deciphered, there are many routes forward to take in understanding this system and building upon the extracted insights to design next-generation materials. Both soft and hard X-ray absorption spectroscopies would be key in capturing the elementspecific electronic changes throughout the catalytic cycle. Vibrational spectroscopy, especially with the aid of isotope labelling and time-resolved measurements would also be instrumental in detecting every intermediate in the reaction pathway to render a closer match between the proposed model and reality. In the direction of OER catalyst innovation, the exploration of novel anionic components substituted into transition metal oxides has shown to be a fruitful avenue to pursue and one can point to recently innovated oxyhalide catalysts. 13,15,16,19 Concluding Remarks The work put forth in this study highlights the utility of exploring hetero-anionic composition in the discovery of efficient catalytic materials. A mixed anion compound, Co0.5Fe0.5O0.5F1.5, was fabricated through a controlled thermal conversion route and, the resultant material exhibited exceptionally high activity for the OER. A set of mechanistic investigations were carried out to elucidate key steps in the reaction mechanism and through this process, we found the OER proceeding

Supporting Information
Highly efficient water oxidation via a bimolecular reaction mechanism on rutile structured mixed-metal oxyfluorides  Operando Raman spectroscopy was conducted in a similar fashion as mentioned above, except a Ag/AgCl reference electrode was employed, and a custom-built Teflon spectroelectrochemical cell was used. An immersion objective was chosen to obtain the highest intensities, and measurements at each potential were recorded over a period of 5 minutes to attain a steady-state spectrum. A Renishaw Invia spectrometer was used with a 514 nm laser for these experiments. To build the models for Co0.5Fe0.5O0.5F1.5, we chose two possible structures to determine the roles of Co and Fe atoms, where the interchange between Co and Fe atoms takes place intermittently (CoFeOF(110)-1) or continuously (CoFeOF(110)-2). (110) facets were chosen since they were reported to be the most stable surface in rutile structures. 24,30 After the primitive unit cell was optimized through cell optimization, we employed the Atomic Simulation Environment (ASE) codes to create the slabs containing 10 Å of vacuum above and below the atoms (i.e. 20 Å of vacuum in total) for geometry optimization. 52 For both cell and geometry optimizations, the convergence criteria for the maximum force acting on each atom were set to be 0.023 eV/Å, and the BFGS method was used as the optimizers. An applied potential (U) to the system changes the energy of one electron by -eU, where e is the electron charge 53 .
The Bader charge analysis was performed using the Bader analysis program written by Henkelman et al. 52,[54][55][56] Meanwhile, the charge density plots were generated from the cube files containing the electronic density with the help of the cube cruncher utility within CP2K's tool-collection.
The adsorption energy of O2 (Eads) was calculated by using the following equation 53 , where E(slabs+O2), E(slabs), and E(O2) represent the total energy of the slabs with O2 adsorbed on it, slabs, and O2, respectively.

Synthetic approach
The new Co-Fe based oxyfluoride with rutile structure was prepared by thermal decomposition of the corresponding crystalline hydrated phase (CoFeF5(H2O)7). A first attempt to synthesize the precursor CoFeF5(H2O)7 was performed using solvothermal synthesis leading to a multiphase system with CoFe2F8(H2O)2 as an impurity ( Figure S1). While the XRD patterns of MFeF5(H2O)7 (M = Fe, Co, Ni) exist in the crystallographic database, their structures are unresolved. In our preparation, single crystals have been isolated and two new structures of hydrated fluorides phase, with the same formulation CoFeF5(H2O)7 were solved using single crystal X-ray diffraction. The first one was determined in the triclinic P-1 space group, and the second one was in the monoclinic C2/m space group. The details of the structure determination and the X-ray atomic coordinates are summarized in Table S1- (Table S8 and Table S9).
The crystal structure of pink crystalline CoFeF5(H2O)7 was determined by single crystal X-ray diffraction ( Fig. S1 and Table S1-S7), and Rietveld refinement ( Fig. S3 and Table S8-S10) of the powder X-ray diffraction pattern (PXRD) confirmed the phase purity. Scanning Electron Microscopy (SEM, Fig. S7b) revealed the formation of CoFeF5(H2O)7 micro-sized particles which is in good accordance with the sharpness of the peaks in the diffraction pattern and a specific surface area (SBET determined by N2 sorption) less than 3 m 2 g -1 . The homogenous distribution of Co, Fe, O and F and the                        Figure S4: Differences between 1 st and 2 nd CVs of Co0.5Fe0.5O0.5F1.5 in 1M KOH illustrate that a surface reconstruction process likely occurs (a, b). From the XPS data on the catalyst after 10 CV cycles (0.8 to 1.55 VRHE at 20 mV/s) and after extended electrolysis (10 hrs at 1.6VRHE), the restructuration stops after the first CV (c-f). The largest changes occur to the cobalt and oxygen spectra and indicate that the surface cobalt species becomes terminated with O/OH.