■ INTRODUCTION

Artificial photosynthesis for water splitting using renewable energy-based alternatives to fossil fuels offers potential to implement clean energy processes. [1][2][3][4][5] Improvement of the efficiency and stability of the electrocatalysts for the anodic reaction, the oxygen evolution reaction (OER), is imperative for the development of large-scale water splitting. 6,7 Homogeneous molecular electrocatalysts for water oxidation have been extensively investigated, and the tunable structures and welldefined active sites have led to detailed mechanistic understanding as well as the advancement of more active and longerlived catalysts. [8][9][10][11][12][13] There has been interest in first-row transition metal complexes due to the advantages of earth abundance and low cost compared to their noble metal counterparts. [14][15][16][17][18][19][20] In particular, cobalt-based molecular catalysts for electrochemical water oxidation have been studied due to their high catalytic activity. [21][22][23][24][25][26][27] Given the lability of Co II ligand bonds and the propensity for oxidative degradation, 28 the formation of CoO x using molecular Co complexes under oxidizing conditions is common. [29][30][31][32] Multidentate chelating ligands can provide stability to the metal center to help mitigate degradation under oxidative conditions. For example, Berlinguette and co-workers reported a pentadentate polypyridine Co II complex (a, Scheme 1) that catalyzes electrochemical water oxidation under basic conditions. 33,34 At pH > 10.2, the formation of CoO x was observed; however, the catalyst was proposed to remain homogeneous at lower pH values (6-9). Siewert and Gałezowska studied a pentacoordinate Co II complex as a molecular catalyst for water oxidation in which the ligand framework contained four NH imidazole units and a pyridine group. 35 Later, Anderson's group reported a related Co II complex stabilized by a tetraimidazolyl-substituted pyridine framework (b, Scheme 1) as a catalyst for electrochemical water oxidation from pH 7 to 9, 22 while complex b was found to degrade rapidly under oxidative conditions; the authors proposed that the initial molecular complex is the active catalytic species. Recently, Yang's group reported a Co II dipyridyldiamine complex (c, Scheme 1) with dimethylamine functional groups as pendant bases in the secondary coordination sphere as a catalyst for water oxidation in acetonitrile/water mixtures. 23 Interestingly, the analogous complex without the dimethylamine groups was found to be inactive for water oxidation, highlighting the importance of secondary sphere interactions. Another class of molecular catalysts for water oxidation is cobalt porphyrins (d, Scheme 1), which are active at both neutral and basic pH values. 24,36 An N,N′-bis(salicylidene)ethylenediamino Co II complex was found to be an active water oxidation electrocatalyst, but in this case, the activity was attributed to the degradation of the complex under oxidative conditions at pH 11 (e, Scheme 1). 29 A Co II complex containing two bidentate triazole carboxylate ligands was also found to be a catalyst for electrochemical water oxidation, forming an active film, in which the complex maintains its atomic structure, on the surface of the working electrode (f, Scheme 1). 37 Complexes based on Co III precursors also have been investigated as electrochemical water oxidation precatalysts. [38][39][40][41][42][43][44][45] Despite these advancements, the instability of molecular catalysts under oxidative conditions can pose significant challenges for their incorporation into electrolyzers. 41,46 In contrast, heterogeneous catalysts often exhibit higher electrocatalytic stability and can be directly incorporated into solidstate electrolyzers, but a major limitation of heterogeneous catalysts is the challenge associated with understanding and optimizing the structure of active sites. The integration of the molecular catalyst structure with heterogeneous materials is desirable since it offers the potential for device development presented by heterogeneous materials with the opportunity to tune catalytic active sites. 12,47,48 Recently, research efforts have focused on the preparation of supported electrocatalysts by anchoring molecular catalysts directly to an electrode or to a carbon support using covalent and noncovalent attachment methods. [47][48][49][50][51][52] Several examples have been reported for the synthesis of cobalt-heterogenized catalysts for water oxidation via van der Waals interactions between a cobalt complex with long alkyl chains and fluorinedoped tin oxide (FTO) or carbon electrodes (see, for example, g in Scheme 2). [53][54][55] Supported Co II phthalocyanine complexes are efficient molecular water oxidation electrocatalysts at pH 7 when physiosorbed onto FTO (h, Scheme 2) 56 or at basic pH (1 M KOH) when π-π-stacked onto carbon nanotubes. 57 Investigations of cobalt porphyrin complexes physiosorbed onto FTO (i, Scheme 2) demonstrated that the formation of a CoO x film on the electrode is responsible for the observed catalytic water oxidation activity. 30 A pyrene-modified Co II salophen complex immobilized on multiwalled carbon nanotubes (j, Scheme 2) was also found to degrade to cobalt hydroxides under oxidative conditions. 58 Covalent methodologies, such as direct amidation coupling, were used to immobilize Co III corroles onto carbon nanotubes as efficient catalysts for electrochemical water oxidation at pH 0, 7, and 14. 59 Immobilization of the molecular catalysts by π-π stacking using various aromatic groups of ligands can boost electronic π-delocalization between molecular catalysts and carbon supports and has emerged as a strategy to increase the stability of molecular catalysts. 48 Herein, we report supported catalysts for electrochemical water oxidation by immobilizing molecular cobalt complexes on ordered mesoporous carbon (OMC) supports via π-π stacking. The OMC support not only offers a substantial surface area and functional sites to immobilize cobalt complexes via π-interactions but provides an ordered tunnel matrix, which can facilitate electrochemical mass transfer.

We have recently reported studies demonstrating the ability of "capping arene" ligands to modulate the energetics of transition metal reactions that involve formal redox changes. [60][61][62][63][64][65] We have proposed that the structure of the capping arene ligand, which positions the pendant arene group and determines the extent of metal-arene bonding, can influence the relative stability of transition metal complexes in different oxidation states and geometries. 64 Given that transition metal catalysts cycle through multiple redox states during electrocatalytic water oxidation, we considered that the capping arene ligands might provide the ability to tune catalyst activity. Hence, cobalt complexes based on capping arene ligands were synthesized, characterized, and investigated as electrocatalysts for the OER. The resulting electrocatalysts were efficient for the OER, delivering a high turnover frequency (TOF) at low overpotential. Moreover, the electrocatalysts exhibited good structural stability under oxygen evolution reaction (OER) conditions, which was confirmed using X-ray absorption (XAS) spectroscopy and aberration-corrected transmission electron microscopy (TEM) characterization, although catalyst partial decomposition after hours of catalysis was observed. In addition, density function theory (DFT) calculations revealed the structural evolution of model molecular catalysts under OER conditions to delineate the catalytic process on the active sites of the electrocatalysts. Using DFT quantum mechanics calculations to probe plausible mechanisms for O-O bond formation, we determined that O-O bond formation most likely occurs through an intramolecular reductive elimination process in which oxo and hydroxo ligands on Co IV couple to form a Co II -OOH species. The structure of the capping arene ligand appears to influence the free energy of activation for this proposed ratedetermining redox reaction.

■ RESULTS AND DISCUSSION

Synthesis and Characterization of "Capping Arene" Co(II) Complexes. Capping arene ligands were selected as a platform for the preparation of cobalt complexes as we considered that properties of these ligands would render them useful for applications in electrochemical water oxidation. First, capping arene ligands are likely to be stable against oxidative degradation since they lack readily oxidized aliphatic C-H bonds and other chemical groups. Second, the capping arene moiety should enable strong π-π stacking, allowing the facile preparation of solid-state anodes on carbon materials. Third, the arene can potentially bind to the metal center with the dual role of providing stabilization under oxidative conditions (Scheme 3) and acting as a direct conduit of electrical conductivity from the metal center to the carbon material. Last, the capping arene ligands provide π-active ligands similar to bipyridyls that are known shuttle redox equivalents during electrochemical processes. 66 Two Co II complexes, (6-FP)Co(NO 3 ) 2 (1) (6-FP = 8,8′-(1,2-phenylene)diquinoline) and (5-FP)Co(NO 3 ) 2 (2) (5-FP = 1,2-bis(N-7-azaindolyl)benzene), were synthesized by mixing the corresponding proligand 6-FP or 5-FP and Co(NO 3 ) 2 •6H 2 O in acetonitrile at room temperature. Complexes 1 and 2 were isolated as purple crystals in yields of 75% (1) and 45% (2) (Scheme 4). They have been fully characterized by paramagnetic 1 H and 13 C NMR spectroscopy, elemental analysis, EPR spectroscopy, SQUID magnetometry, and single-crystal X-ray diffraction. Single crystals adequate for X-ray diffraction were obtained by vapor diffusion of diethyl ether into a CH 2 Cl 2 solution saturated with 1 or 2 (Figure 1). The geometry around the cobalt center in 1 can be described as distorted trigonal bipyramidal (TBP) with the value of the τ 5 distortion parameter of 0.69 significantly deviated from the ideal value of 1 for a TBP structure. 67 The equatorial plane results from the coordination of two nitrogen atoms of the 6-FP ligand and an oxygen atom from a bidentate nitrate ligand with the axial positions formed by the other oxygen from the bidentate nitrate and an oxygen atom from a monodentate nitrate ligand (Figure 1a). The Co-N bond distances, Co1-N1 2.0859 (11) Å and Co1-N2 2.1224(11) Å, are comparable Scheme 3. Schematic of "Capping Arene" Proligands 5-FP and 6-FP Discussed in This Work to those observed in other Co II complexes with chelating pyridine-based ligands. 23 The Co-O bonds exhibit two shorter distances and one longer distance (Co1-O2 2.0420(9) Å, Co1-O4 2.0551(10) Å, and Co1-O1 2.3244(10) Å). The Co II atom is located directly above C10 and C15 of the capping arene ring, but the Co-C distances of 2.921(4) Å and 2.698(5) Å are too elongated to indicate significant bonding interaction. For complex 2, four independent molecules are found in the unit cell of the solid-state structure, displaying two different geometries around the cobalt center. Two of the molecules present distorted square-pyramidal geometries and the other two present distorted octahedral geometries, depending on whether the second nitrate is bound k 1 or k 2 to the metal center (Figure 1 and the Supporting Information). The Co-N distances range from 2.056(6) Å to 2.124(6) Å, similar to those observed for complex 1. Analogous to complex 1, little to no arene-Co interaction is present in complex 2.

Complexes 1 and 2 exhibit well-behaved paramagnetic 1 H and 13 C NMR spectra. Hence, the wide-scan 1 H and 13 C NMR spectra of complexes 1 and 2 in CDCl 3 at room temperature display broad paramagnetically shifted peaks that are consistent with symmetric complexes. For complex 1, peaks range from 95 to -40 ppm in the 1 H NMR spectrum (Figure 2a) and from 640 to 80 ppm in the 13 C NMR spectrum (Figure 2b,c).

For complex 2, peaks range from 140 to 0 ppm in the 1 H NMR spectrum and from 425 to 120 ppm in the 13 C NMR spectrum (Figures S1 andS2, Supporting Information). The magnetic moments (μ eff ) of 1 and 2 were determined by the Evans NMR method to be 3.4(2) and 4.3 (2), respectively, indicating highspin character. Magnetic susceptibility traces of 1 and 2 measured as neat polycrystalline powders are shown in Figure 3. For 1 and 2, χT values of 2.55 and 2.52 cm 3 mol -1 K (μ eff = 4.6μ B and μ eff = 4.5μ B ) were obtained at room temperature, respectively. These χT values are in good agreement with a high-spin (S = 3/2) Co II and correspond to g ≈ 2.3 using

. The observed deviation of g from 2 is characteristic for Co II and originates from spin-orbit coupling. Down to 100 K, χT values remained constant. Upon decreasing temperature, these values were reduced to 1.25 cm 3 mol -1 K (1) and 1.4 cm 3 mol -1 K (2) at 2 K, indicating large magnetic anisotropy in the samples. The insets in Figure 3 show reduced magnetization curves for 1 and 2 at constant magnetic fields of 1, 4, and 7 T, respectively. These curves superimpose at low temperatures, suggesting zero field splitting (zfs) much larger than the Zeeman splitting induced by the external magnetic fields. Both paramagnetic NMR and SQUID magnetometry are well in accordance with a Co II S = 3/2 state with large zfs; this assignment is further supported by the lowtemperature EPR spectra depicted in Figure S3, Supporting Information. However, μ eff values obtained from solution paramagnetic NMR and polycrystalline SQUID measurements were found to be different. Moreover, frozen solution EPR and SQUID traces could not be simulated with one set of spin coupling parameters. These discrepancies indicate that complexes 1 and 2 in solution exhibit different magnetic properties and most likely also different geometric structures than in the crystalline phases.

The cyclic voltammetry of 1 (0.5 mM) was measured in MeCN under a dinitrogen atmosphere in the presence of tetrabutylammonium hexafluorophosphate as the supporting electrolyte using ferrocene as an internal standard. Complex 1 displays an irreversible oxidation peak at E p a = 1.09 V vs Fc + / Fc (Figure 4a), and complex 2 displays an irreversible oxidation peak at E p a = 1.11 V vs Fc + /Fc (Figure 4b), indicating that the products of oxidation are likely unstable under the electrochemical conditions. Both oxidation peaks for 1 and 2 are diffusion-controlled. Addition of 1 M KOH aqueous solution to both complexes 1 and 2 in MeCN produces a brown insoluble precipitate (most likely cobalt hydroxide), which prevented the study of molecular water oxidation.

Immobilization of "Capping Arene" Co(II) Complexes on OMC. Conductive OMC possesses a large surface area with an ordered framework structure, providing opportunities for application in heterogeneous electrocatalysis with potential advantages for mass transfer. 68 Our strategy for immobilization of Co complexes 1 and 2 involved attaching 1 or 2 on OMC via π-π stacking interactions between the arene ligand platform and the carbon support. The resulting material provides a heterogenized molecular electrocatalyst for the OER. The OMC support was obtained by the carbonization of oleic acid ligand in a self-assembled Fe 3 O 4 colloidal nanoparticle superlattice followed by the removal of the Fe 3 O 4 template. 68 Monodisperse Fe 3 O 4 nanoparticles were synthesized with oleic acid as a surface capping ligand. The TEM image in Figure S4, Supporting Information, shows that the resulting Fe 3 O 4 nanoparticles have a size of 9.5 ± 0.5 nm. The 2) in MeCN at variable scan rates. three-dimensional Fe 3 O 4 nanoparticle face-center cubic (fcc) superlattices were produced through slow evaporation of hexane solvent under ambient conditions, 68 which were then annealed at 500 °C in nitrogen gas to carbonize the oleic acid surfactant and subsequently washed in hydrochloric acid at 120 °C to remove the Fe 3 O 4 template (Figure S5, Supporting Information). The resulting OMC material, after a further treatment at 900 °C in forming gas (5% H 2 in N 2 ) to enhance the electrical conductivity and degree of graphitization as we reported before, 51 shows an ordered structure with a pore size of ∼7.5 nm and a wall thickness of ∼2.5 nm (Figure 5a).

The Co complex-loaded OMC was generated by sonicating the mixture of the molecular complex, OMC, and isopropanol (Figure 5b,c and Figure S6, Supporting Information). We anticipated that the Co complex attachment onto the OMC would be enhanced by π-π stacking interaction, similar to our previous work on Ir complex-loaded OMC. 51 Because of the porous structure and large surface area with facilitated diffusion from sonication, the OMC material should enable a high loading of the molecular Co complexes. The resulting Co complex-loaded OMC is labeled as 6-FP-Co-OMC-1 for 1 and 5-FP-Co-OMC-2 for 2. A control material was synthesized without FP ligand by loading Co(NO 3 ) 2 •6H 2 O onto OMC, which is marked as Co(NO 3 ) 2 -OMC-3. As shown in Figure 5b and Figure S6, Supporting Information, TEM images of 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 exhibit well-maintained OMC structures with homogeneous dispersion of cobalt complexes in the framework. The quantitative loading of Co in these samples was carried out with ICP-OES, which is 17.1, 19, and 4.3% for 6-FP-Co-OMC-1, 5-FP-Co-OMC-2, and Co(NO 3 ) 2 -OMC-3, respectively. Elemental mapping of cobalt on OMC shows homogeneous dispersion of Co on the OMC surfaces (Figures S7 andS8).

Electrocatalysis over Co Complex-Loaded OMC. The 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 catalysts were studied for the OER using a three-electrode system with a rotating disk electrode (RDE) as a working electrode. The electrochemical analyses were conducted in O 2 -saturated 1 M KOH aqueous electrolyte. The linear sweep voltammetry (LSV) plot from 0.6 to 1.67 V vs reversible hydrogen electrode (RHE) at a scan rate of 10 mV s -1 (Figure 6a) shows that 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 deliver a substantially higher current density than Co(NO 3 ) 2 -OMC-3 and pristine OMC at the same overpotential. As shown in Figure 6a, 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 reach current densities of 34.6 and 32.0 mA cm -2 at an overpotential of 400 mV, which are higher than those of Co(NO 3 ) 2 -OMC-3 (1.1 mA cm -2 ) and pristine OMC (3.5 mA cm -2 ). The overpotential of 6-FP-Co-OMC-1 at 10 mA cm -2 current density was 355 mV, lower than that of 5-FP-Co-OMC-2 (370 mV). Continuous LSV scan of 6-FP-Co-OMC-1 is provided in Figure S9, indicating that the performance of the immobilized Co molecular catalyst is stabilized after 10 scans. We found that 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 exhibited similar Tafel slopes of 67 and 55 mV dec -1 , suggesting similar OER kinetics (Figure 6b). The Faradic efficiencies of 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 are >90% at different current densities, indicating the high energy efficiency of both catalysts (Figure 6c).

The OER TOFs over 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 are summarized in Figure 6d. 6-FP-Co-OMC-1 delivers a TOF of 0.53 s -1 at an overpotential of 400 mV, which is higher than 5-FP-Co-OMC-2 with a TOF of 0.32 s -1 at the same overpotential (Figure 6d). Both are superior to Co(NO 3 ) 2 -OMC-3, which exhibits a TOF of 0.04 s -1 at 400 mV overpotential. The intrinsic activities of 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 appear to be better than many other supported Co catalysts in previous reports once overpotential is taken into account (Table 1). Also, our studies show a pronounced benefit of the capping arene ligands, as demonstrated by comparative electrocatalytic OER using 6-FP-Co-OMC-1, 5-FP-Co-OMC-2, and Co(NO 3 ) 2 -OMC-3, which presented activity that is ∼10 times that of Co(NO 3 ) 2 -OMC-3. While 6-FP-Co-OMC-1 and 5-FP-Co-OMC-2 give similar performance, the ratio of TOF using 6-FP-Co-OMC-1 versus 5-FP-Co-OMC-2 at 300, 350, and 400 mV are 3.75, 2.46, and 1.65, respectively. Assuming that the core molecular structure remains intact upon anchoring to the OMC material, these data indicate a more active site for 6-FP-Co versus 5-FP-Co, which is consistent with computational modeling of electrocatalytic water oxidation (see below).

The stability of 6-FP-Co-OMC-1 was studied with a chronoamperometry (CA) test at 300 mV overpotential. As shown in Figure 7a, the current density of the 6-FP-Co-OMC-1 catalyst retains 55% of the initial activity after 11 h of electrocatalysis. We found from TEM images in Figure 7b that the structure of OMC was maintained well with no visible nanoparticles of cobalt oxide. The high-angle annular dark field scanning TEM (HAADF-STEM) image acquired by an aberration-corrected STEM in Figure 7c shows that the catalyst after 11 h of electrocatalysis is present as a mixture of single-atom Co and cobalt oxide clusters on the OMC surface and the latter could arise from confined cobalt atomic species. Further investigation of the structure of the cobalt sites before and after the stability test was conducted by employing a spectroscopic probe of XAS. The Fourier-transformed Co Kedge extended X-ray absorption fine structure (EXAFS) of complex 1 and 6-FP-Co-OMC-1 before and after the CA stability test is presented in Figure 7d, Figures S10-S12, and Table S1, Supporting Information. The spectra of 1 and 6-FP-Co-OMC-1 exhibit very similar profiles with a predominant peak at an atomic distance of 1.59 Å corresponding to a Co-O/N bond, which is distinct to representative Co-O scattering pathways for Co oxide materials (2.4-2.5 Å). 7 The 6-FP-Co-OMC-1 material after the chronoamperometry test over 11 h exhibits similar characteristics of a Co-O/N bond with a peak at 1.62 Å while also showing new scattering DFT Calculations. We utilized DFT quantum mechanics calculations to probe plausible pathways for electrocatalytic water oxidation by complexes 1 and 2. This methodology has been previously validated to determine mechanisms and kinetics for numerous electrocatalytic processes including trinuclear Cu, 16 Co-doped TiO 2 , 7 IrO 2 , 71 and Fe-doped NiOOH. 72 Our computational studies aimed to determine the reaction pathway for electrocatalytic water oxidation that provides the most facile mechanism for O-O bond formation. Due to the size of the OMC-supported structure, we modeled the mechanism of electrocatalytic water oxidation using the molecular Co II complexes 1 and 2.

We modeled the mechanism of electrocatalytic water oxidation starting with the Co II In Figure 8, we also present the spin density surfaces. As expected, the unpaired spin resides on the oxo in both structures. For the 6-FP case, the Mulliken spin population on the oxo is 1.021, while for the 5-FP complex, the Mulliken spin population on the oxo ligand is 1.024. 73 The formation of ( 6 Path 2 involves two steps, the second step being ratedetermining. The O-O bond formation between the free hydroxide and the Co-oxo has 49.9 kcal mol -1 free energy above the Co II starting state (or 21.8 kcal mol -1 relative to the Co III oxo complex). The transition state for the second step has an imaginary mode at -656.9 cm -1 and an O-O distance of 1.98 Å. Path 2 is clearly less favorable than path 1.

The transition state for the concerted path 3 involves an eight-membered ring in which two water molecules and the OOH ligand are formed; this free energy barrier is calculated to be 51.5 kcal mol -1 above the starting state (or 23.4 kcal mol -1 above Co IV -O • ). The formation of (6-FP)Co(OH)-(OOH)(H 2 O) exhibits 5.9 kcal mol -1 free energy above the Co II starting state.

Path 4 involves a single step in which a free OH anion in solution attacks the oxo ligand to generate Co III -OOH. The O-O bond formation transition state possesses 41.7 kcal mol -1 free energy above the Co II starting state so that path 1 remains the most facile. Because our experiments are performed at pH 14, some finite concentration of OH anions is expected such that the direct nucleophilic attack of an OH anion on the Co-oxo is plausible. However, we find that the energetics are not favorable, likely because Co is reduced to Co(III) as opposed to Co(II) in path 1.

Of the four mechanisms for O-O bond formation, the intermolecular water nucleophilic attack of a hydroxy ligand outlined in path 1 provides the most facile pathway. Assuming that O-O bond formation is the rate-limiting step in catalysis, the overall free energy barrier for OER from (6-FP)Co-(OH) 2 (H 2 O) is 36.8 kcal mol -1 mol at an applied potential of 0.7 V versus RHE. Interestingly, the majority of this 36.8 kcal mol -1 barrier is due to Co oxidation (28.1 kcal mol -1 contribution) as opposed to O-O bond creation (8.7 kcal mol -1 contribution). In a related study, DFT calculations are employed to compare the intra-and intermolecular attacks of a hydroxide onto a Co IV -oxo species. 73 The intermolecular pathway was found to be lower in energy as the saturated metal center required decoordination of a pyridine before the intramolecular attack. Similar reaction pathways were also calculated for [(bpy)Cu II (OH) 2 ] (bpy = 2,2′-bipyridine) where the intermolecular hydroxy attack on a Cu III -oxyl was found to be favored over the intramolecular attack. 74 A more recent study on the mechanism of [(bpy)Cu II (OH) 2 ] water oxidation finds the intramolecular attack to be energetically favorable and attributes the formation of the O-O bond to the coupling of the oxyl moiety with a bound hydroxide ligand in an analogous manner to our proposed mechanism. 75 We perform the identical analysis for Co-catalyzed OER with the 5-FP ligand. Starting with (5-FP)Co(OH) 2 (H 2 O) at 0.0 kcal mol -1 , oxidation to (5-FP)Co(OH) 3 is endergonic at 11.1 kcal mol -1 ; further oxidation to (5-FP)Co(OH) 2 (O) exhibits uphill to 28.2 kcal mol -1 free energy (Scheme 7). We note that the free energy barrier for intramolecular reductive coupling is 0.7 kcal/mol lower for the 6-FP ligand compared to the 5-FP ligand. If we ignore Co II to Co IV oxidation, we see that O 2 formation through intramolecular reductive coupling with 6-FP requires a barrier of 8.7 kcal mol -1 ; with the 5-FP ligand, this barrier is 9.3 kcal mol -1 (4.2 kcal mol -1 higher).

Recall that oxidation from Co

■ CONCLUSIONS

The well-defined cobalt catalytic centers of (6-FP)Co(NO 3 ) 2 (1) and (5-FP)Co(NO 3 ) 2 (2) were effectively immobilized on conductive OMC, likely via π-π interactions, and were found to be active catalysts for electrocatalytic OER. The resultant catalysts were investigated with ex situ XAS, demonstrating reasonable stability and slow degradation of the supported complexes under the studied conditions (1 M KOH, pH 14) to cobalt oxide clusters after 11 h of electrolysis. Theoretical DFT calculations revealed the formation of an intermediate Co IV oxo species that can be converted to the hydroperoxo via intermolecular O-O reductive coupling between the oxo ligands and a hydroxide ligand. Our calculations suggest an explanation for the increased activity for 6-FP-Co-OMC-1 compared with 5-FP-Co-OMC-2. The structure of the 6-FP ligand provides more facile O-O bond formation through increased interaction of the capping arene π space and the Co-oxo antibonding orbital. Taken together, our experimental and theoretical results demonstrate an immobilization method for the heterogenization of molecular catalysts on conductive OMC and provide a strategy for further exploration of the use of capping arene ligands that are designed to optimize electrocatalytic water oxidation.

■ EXPERIMENTAL SECTION

Chemicals and Materials. FeCl 3 •6H 2 O (98%), 1octadecene (ODE, 90%), and oleic acid (OAc, 90%) were purchased from Sigma-Aldrich. Hexane (ACS-certified), 2propanol (IPA, ACS-certified), KOH (ACS-certified), ethanol (200 proof), and hydrochloric acid (HCl) were purchased from Fisher Scientific. Sodium oleate was obtained from Tokyo Chemical Industry.

General Methods. All reagents and solvents were purchased from commercial sources and used without further purification, unless stated otherwise. The proligands 5-FP and 6-FP were synthesized as previously described. 64,76 The UV-Vis absorption spectra were recorded using a Cary 60 UV-Vis spectrophotometer. The NMR spectra were referenced to tetramethylsilane (TMS) using residual proton signals ( 1 H NMR) or 13 C resonances ( 13 C NMR) of the deuterated solvents and were recorded using a Bruker AV800 spectrometer.

Synthesis of Fe 3 O 4 Nanoparticles. Monodisperse Fe 3 O 4 nanoparticles were synthesized according to a previously reported work. 50 Iron oleate was first prepared by reacting FeCl 3 •6H 2 O and sodium oleate with the mixture solvent of hexane, ethanol, and DI water at 50 °C for 4 h. The Fe 3 O 4 synthesis was carried out in a standard Schlenk line with moisture-free condition. The mixture of iron oleate (3.2 g), ODE (20 mL), and oleic acid (0.64 mL) was heated under vacuum to 100 °C and maintained at the temperature for 1 h to remove impurities. The system was then switched to a N 2 atmosphere, heated to 310 °C, and kept at the temperature for 1 h. Fe 3 O 4 was further washed twice with IPA and separated by centrifugation. The Fe 3 O 4 nanoparticles were dispersed in hexane and stored for further use.

OMC Preparation. The OMC was prepared by using a superlattice template of Fe 3 O 4 nanoparticles. Hexane dispersion of Fe 3 O 4 nanoparticles was placed in a ceramic boat with one glass slide covered on the top. The solvent was slowly evaporated in ambient conditions, which leads to the formation of a self-assembled superlattice. The carbonization of the oleic acid surfactant bound on nanoparticles was carried out in N 2 under 500 °C for 2 h. The obtained product was further washed with concentrated HCl at 120 °C for the removal of the Fe 3 O 4 template, leading to OMC. The OMC material was subsequently treated in forming gas at 900 °C for 2 h to enhance the graphitization degree and conductivity.

Synthesis of (6-FP)Co(NO 3 ) 2 (1). 6-FP (55.2 mg, 0.166 mmol) was added to a purple solution of Co(NO 3 ) 2 •6H 2 O (48.2 mg, 0.166 mmol) in acetonitrile (10 mL) and stirred overnight at room temperature, with no color change observed throughout the reaction. The solvent was removed in vacuo, and the resulting solid was extracted into dichloromethane, resulting in a violet solution (10 mL). After filtration through a fine porosity frit, crystals were obtained by the vapor diffusion method using diethyl ether. After 2 days at room temperature, bright purple crystals suitable for X-ray diffraction studies were obtained (64 mg, 75%). 1 H NMR (800 MHz, CDCl 3 ; all peaks appear as broad singlets) δ 90. Synthesis of (5-FP)Co(NO 3 ) 2 (2). 5-FP (93.6 mg, 0.302 mmol) was added to a purple solution of Co(NO 3 ) 2 •6H 2 O (91.7 mg, 0.315 mmol) in acetonitrile (10 mL) and stirred overnight at room temperature, with no color change observed throughout the reaction. The solvent was removed in vacuo, and the resulting solid was extracted into dichloromethane, resulting in a violet solution (10 mL). After filtration, crystals were obtained by the vapor diffusion method using diethyl ether. After 2 days at room temperature, bright purple crystals suitable for X-ray diffraction studies were obtained (66.7 mg, 45%). 1 H NMR (800 MHz, CDCl 3 ; all peaks appear as broad singlets) δ 135. SQUID Magnetometry. The temperature-dependent magnetization of 1 and 2 was measured as neat polycrystalline powder samples of solid material immobilized in eicosane by using a Superconducting Quantum Interference Device (SQUID, MPMS-7, Quantum Design, calibrated with a standard palladium reference sample, error < 2%) in the temperature range from 2 K to 300 K under a constant magnetic field of 1 T. The reduced magnetization data were obtained under constant magnetic fields of 1, 4, and 7 T.

EPR Spectroscopy. X-band cw EPR spectra were obtained using an X-band Bruker Elexsys E500 EPR spectrometer equipped with an ER4116DM dual mode resonator and an ESR 900 He cryostat. A total of 1.9 mg of 1 and 2 was dissolved in 1.5 mL of a 2:1 solvent mixture of toluene and CHCl 3 in a glove box. Two hundred microliters of each sample was transferred into standard EPR tubes for measurements. The EPR spectra were obtained at 10 K using a microwave frequency of 9.64 GHz, a microwave power of 20 mW, a lockin modulation amplitude of 0.7 mT, a time constant of 40.96 ms, 4096 points, and a modulation frequency of 100 KHz.

Co-Loaded OMC Preparation. The mixture of Co molecular complex and OMC (weight ratio, 1/1) was dispersed in IPA and sonicated for 0.5 h. The Co-loaded OMC was collected by centrifugation and washed one more time with IPA. The precipitant was dried in ambient conditions and stored for electrocatalyst ink preparation. The loading of Co was measured with ICP-OES analysis.

Electrocatalytic OER Measurement. All electrocatalytic performance was studied at room temperature in the O 2saturated 1.0 M KOH electrolyte. The testing cell is a threeelectrode setup, including a glassy carbon working electrode, a Pt foil counter electrode, and a Hg/HgO (1.0 M KOH) reference electrode. The electrochemical characterization was conducted with a BioLogic (Model VMP3) potential station. The electrocatalyst ink was prepared by sonicating Co-loaded OMC (concentration of 5 mg ml -1 ), IPA, and Nafion solution. The volume ratio of Nafion/IPA is 1/100. The glassy carbon electrode was polished with aluminum slurry before working electrode preparation. Electrocatalyst ink of 20 μL was spincast on glassy carbon. All the potentials were reported vs reversible hydrogen electrode (RHE) using the equation E(vs