Molecular cobalt corrole complex for the heterogeneous electrocatalytic reduction of carbon dioxide

Electrochemical conversion of CO2 to alcohols is one of the most challenging methods of conversion and storage of electrical energy in the form of high-energy fuels. The challenge lies in the catalyst design to enable its real-life implementation. Herein, we demonstrate the synthesis and characterization of a cobalt(III) triphenylphosphine corrole complex, which contains three polyethylene glycol residues attached at the meso-phenyl groups. Electron-donation and therefore reduction of the cobalt from cobalt(III) to cobalt(I) is accompanied by removal of the axial ligand, thus resulting in a square-planar cobalt(I) complex. The cobalt(I) as an electron-rich supernucleophilic d8-configurated metal centre, where two electrons occupy and fill up the antibonding dz2 orbital. This orbital possesses high affinity towards electrophiles, allowing for such electronically configurated metals reactions with carbon dioxide. Herein, we report the potential dependent heterogeneous electroreduction of CO2 to ethanol or methanol of an immobilized cobalt A3-corrole catalyst system. In moderately acidic aqueous medium (pH = 6.0), the cobalt corrole modified carbon paper electrode exhibits a Faradaic Efficiency (FE%) of 48 % towards ethanol production.

M inimizing of the CO 2 concentration in the atmosphere is one of the most important challenges in our time 1,2 . Therefore, the electrochemical reduction of CO 2 to value added chemicals is a sustainable strategy to solve the growing energy crisis, which at the same time has the potential to mitigate environmental pollution. In the past years, the electrochemical reduction of CO 2 has been studied by several research groups to produce valuable products, for example carbon monoxide, formic acid, methane, ethanol, or methanol [3][4][5] . Particularly the transformation of CO 2 in high-density alcohols, especially methanol and ethanol, is a cherished goal for chemists and environmental engineers alike 6,7 . Such transformation of CO 2 to alcohols coupled with the oxidation of water to oxygen 8 is a promising strategy 9 . However, the low reactivity of carbon dioxide in water with its large energy barrier (ΔE = −1366.8 kJ mol −1 ) 10 and the competing hydrogen evolution reaction, impedes such transformation, which makes the development of catalysts for electrocatalytic CO 2 reduction to ethanol in aqueous environment a big challenge 1,2,10 . The thermodynamic reduction potential for CO 2 to methanol and ethanol is 0.03 and 0.09 V (vs. RHE), respectively which is kinetically disfavored. Hence, often CO 2 reducing catalysts end up accruing lot of energy to be operational at a higher potential. In this regard the use of a molecular catalyst with earth abundant elements, (Fe, Mn, Co, Cu, and Ni), especially with a cobalt metal center 11,12 is a viable alternative as it offers a high degree of tunability with product selectivity at a low overpotential. As early as in 1980s chemists have been successful in reducing CO 2 to CO via electrochemical methods employing catalysts containing different metals like Co 13 , Ni 14 , Re [15][16][17] , etc. Recently electrochemical reduction of CO 2 to ethanol has been studied in various ways 18,19 , by tuning the applied potential 20 , pH 21 , and nature of the electrolyte 22 with an aim to control the product selectivity and increase the Faradaic efficiency (FE) as well as to understand the underlying mechanistic pathways. Metal surfaces 23,24 , oxides 25 , and alloys 26,27 are the most explored examples which show good FE for CO 2 reduction but lack selectivity 28,29 and work at higher overpotentials, involving complex synthetic procedures 30,31 . Emergent materials like B and N co-doped nanodiamonds exhibit excellent FE (93 %) and selectivity for conversion to ethanol, but work at higher overpotentials 10 . Moreover such heteroatom-doped materials 32 often require a sophisticated synthetic procedure like chemical vapor deposition, making it hard for large scale implementation 33 . We now compare and contrast state of the art catalysts for CO 2 electro-reduction to ethanol, all of which work at higher overpotentials with shorter activity time and have lower FE as compared to the catalyst reported here (Supplementary Table 1). For instance, Cu(100) works at −0.97 V vs. RHE yielding ethanol with a FE% of 14.7 29 . Further, use of Cu based nano-particles in an ensemble fashion (trans-CuEn) showed a FE% of ethanol formation to be 17 at −0.86 V vs. RHE 34 . While, with tailoring of cubic Cu nanocrystals to an edge length of 44 nm, FE% of 80 was achieved for CO 2 reduction but the FE% for Ethanol formation was as low as 3.7 35  and N co-doping shows a very good FE% of 93 and selectivity toward ethanol 10 .
With the aim to synthesize effective and stable electrocatalysts for CO 2 reduction for the selective ethanol formation, we focus on a molecular Co-corrole catalyst. Metal corroles are structural similar to metal porphyrins with both the metal centers and ligands participating in multielectron redox processes and are promising candidates for efficient proton-coupled electron transfer [40][41][42] . These metal complexes stabilize radical intermediates thus providing an effective pathway to facilitate C-C step-up 43,44 . Cobalt and iron corroles have been previously found to be catalytically active for CO 2 reduction to CO 12 .

Results
Synthesis and electrochemical characterization of the catalyst. The Co-corrole was synthesized via four steps, where the first two steps leading to 5,10,15-trispentafluorophenylcorrole, were performed according to Gryko's procedure 48 . The electronwithdrawing properties of the C 6 F 5 functionalities render the corrole ring electron deficient 49 . Chemical modification with the -S-PEG(7)-OMe moieties at the three para-positions of the meso-C 6 F 5 groups was performed (Fig. 1a) to optimize anchoring and equal distribution of the catalyst on the electrode surface. The chemical syntheses and characterizations are described in detail in the methods section and in the Supplementary Information file (Supplementary Figs. 1-6). The immobilization process of Cocorrole over carbon paper is implemented via drop casting by using anacetonitrile solution of Co-corrole. The modified carbon paper electrodes are stable in aqueous solution due to the insolubility of the Co-corrole moiety in water resulting in the formation of a sustainable heterogenized catalyst with extended lifetime for electrocatalysis.
Electrochemical properties of the Co-corrole were investigated by cyclic voltammetry, with glassy carbon as working electrode in CH 3 CN under Argon and 0.1 M TBAPF 6 as supporting electrolyte. As shown in Fig. 1b, red curve, two one electron redox peaks at −0.5 V (Co(III)/Co(II)) and −1.5 V (Co(II)/Co(I)) vs. NHE are occurring ( Supplementary Fig. 7). The irreversibility of the redox peak at −0.5 V is likely due to the partial loss of the PPh 3 ligand 50 . In analogy to previous report by Kadish et al. 50 , these two couples Co(III)/Co(II) and Co(II)/Co(I) are metal centered redox processes. The reversible two one electron redox peaks at 0.73 and 1.12 V vs. NHE are ligand centered oxidations and correspond to the formation of a cationic and dicationic cobalt(III) corrole complex 50 . DFT calculations suggest that the one-step and the two-step reduction at −0.5 and −1.5 V of the cobalt(III)-corrole leads to enhanced π-back bonding which strengthens the Co-N bonds and the cobalt corrole macrocycle becomes planar (Fig. 1c). Therefore, demetallation is energetically disfavored, which enables catalysis at a single Co(I)-site. The cyclic voltammetry of Co-corrole under CO 2 (Fig. 1b, (Fig. 1d).
The main peak for Co 2p 3/2 at 780.18 eV, is located at a typical cobalt(II) position (e.g., 780.2 eV for CoO) and the main peak for N1s is at 398.53 eV. Supplementary Fig. 15 displays also a C1s XP spectrum taken after electrocatalysis reaction. We observe two signals at 284.6 and 286.2 eV, because not all aromatic carbon atoms in the Co-corrole are the same, due to a lowering of symmetry to C 2v for corroles relative to D 4h for porphyrins. The shake-up satellite at 289.0 eV is typical for organic molecules with extended conjugated π systems. The XPS scans show that the catalyst is stable in course of electrocatalysis ( Supplementary Fig. 15).
Heterogeneous CO 2 electroreduction. The heterogeneous electrochemical CO 2 reduction experiments were carried out with Co-corrole deposited on carbon paper with effective loadings of 0.2 mg cm −2 . The modified electrode was found to reduce CO 2 to ethanol and methanol in 0.1 M NaClO 4 at a potential of −0.8 V vs. RHE (pH = 6.0, 0.1 M phosphate buffer, Table 1). Controlled potential electrolysis (CPE) under CO 2 of Co-corrole modified carbon paper exhibits a TON = 196 and a TOF = 0.011 s −1 for the catalytic conversion of CO 2 to EtOH over 5 h (Fig. 2a-d). The quantification of products was performed using the observed 1 Hnuclear magnetic resonance (NMR)-and gas chromatography mass spectrometry (GC-MS) measurement (e.g., in Fig. 3a and Table 1, and Supplementary Notes 1-3). XPS analysis of the electrode materials before and after catalysis reveals that the catalyst is stable in course of electroreduction and the catalyst pertaining to the reduction process are molecular Co-corrole units ( Supplementary Fig. 15). Moreover, in the course of 5 h electrolysis at −0.8 V vs. RHE the Faradaic efficiency (FE%) for the ethanol production was measured at different intervals of time (Fig. 2a). Figure 2a illustrates that the FE =~48% stays constant throughout the electrocatalytic reduction process.     the FE% for ethanol was found in the range of 9-48 with the selectivity for C 2 over C 1 increasing with increase in the applied cathodic potential from −0.73 to −0.96 V vs RHE with a significant decrease in methanol production. CPE long-term measurements were performed at −0.73 V (pH = 6.0) and −0.  3). After electrolysis, the mass spectra indicated the formation of reduced product and the following peaks appearedm/z of 45-52 (Fig. 2d). Among the peaks obtained, the one with m/z value 52 (CD 3 CD 2 OD) can be assigned for hexa-deuterated ethanol. In the spectrum apart from the molecular ion peak, there are also peaks corresponding to other fragments with m/z = 50 which can be from CD 3 CD 2 O and m/z = 34 for CD 2 OD + which are characteristic fragments of CD 3 CD 2 OD . This deuteration further proves that the source and incorporation of protons in the reduced product is from the solvent and the source of ethanol is from CO 2 reduction. To confirm the source of carbon in the reduction products, we have conducted the reduction experiments with 12 CO 2 and 13 CO 2 . The 1 H and 13 C-NMR spectra after CO 2 reduction at the Co-corrole modified carbon paper electrode exhibit resonances for ethanol, methanol, acetic acid, and formic acid (Fig. 3a, b). The 13 C-NMR spectrum after the reduction of 13  indicating the reduction of 13 CO 2 and formation of C-C bond (Fig. 3b). HCOOH and CH 3 OH appear as singlets at 170.0 and 49.8 ppm in the 13 C-NMR spectrum (Fig. 3b), respectively. These results were further substantiated by GC-MS data, which shows a shift of m/z = 2 for the molecular ion peak of CH 3 CH 2 OH + on 13 C enrichment (Fig. 3c, d). The results for CH 3 OH + , when compared to 12 CO 2 reduction shows a shift by m/z = 1 (Fig. 3e, f). On analysis of the fragmentation patterns, we observe that for ethanol obtained by the reduction of 12 CO 2 , the peak at m/z = 29 corresponds to the CH 3 CH 2 + ion. On the other hand, when 13 CO 2 is used, the peak occurs at m/z = 31, which is due to the substitution of both 12 C centers with 13 C isotopes. Likewise, the peak obtained at m/z = 31 in case of 12 C enriched ethanol resembles the CH 2 OH + fragment which on 13 C enrichment shifts to m/z = 32. For 12 C enriched methanol, the peak at m/z = 15 resembles CH 3 + ion which shifts to m/z = 16 on 13 C enrichment. The results obtained from both the 13 C-NMR and GC-MS experiments prove that the source of ethanol as well as methanol is CO 2 (Fig. 3b-f).

Discussion
An in-depth elucidation of the mechanistic pathway of the reduction process is beyond the scope of this present work. Detailed investigations are underway in our laboratories. To increase the CO 2 reduction efficiency and to avoid hydrogen evolution at low-pH values, all experiments were performed under weak acidic conditions (pH = 6.0, 0.1 M phosphate buffer). CO 2 reduction under heterogenized conditions with Co-corrole modified carbon electrodes exhibits a redox couple being [Cocorrole] 1− /[Co-corrole] 2− , which was found to be at −0.8 V vs. RHE. This markedly resembles the redox behavior of the Cocorrole molecule in the solution, so the redox properties are unperturbed upon heterogenization. We thus propose the molecular characteristics of the electrocatalyst to be persistent upon heterogenization. The EPR spectrum obtained after electrochemical reduction and subsequent dosage of CO 2 at room temperature exhibits a rhombic S = 1/2 signal at g = 2 with a weak 59 Co hyperfine splitting and indicates the formation of Co (III)-CO 2 •− species (Supplementary Fig. 24).
The role of protons is extremely crucial in this work and sets this process apart from related CO 2 reduction processes. For instance, at pH = 6.0 the necessary protons are provided for the subsequent proton coupled electron transfer (PCET), due to this, reduction to methanol/ethanol takes place (Fig. 3a). By performing the same experiment at pH = 7.2, we experimentally observe the reduction of CO 2 to a mixture of formaldehyde, ethanol, methanol, acetate, and formate (see assignment of NMR resonances in the supplementary Fig. 20), and at a pH = 8.0 we detect only CO as the main reduction product ( Supplementary  Fig. 25a-c and Supplementary Notes 4,5). This implies that the rate of PCET drastically decreases at weakly to moderately basic pH values. For the successful chemical transformation of CO 2 to methanol and ethanol, this result is crucial in the present context.
The existence of the Co(III)-CO 2 •− can only be justified through the presence of CO 2 •− formed at a very high potential of ca. −1.5 V vs. RHE 52 . But in our case, the reduction wave at ca. −0.8 V vs. RHE is responsible for the CO 2 reduction. This implies that the Co(I) in the center of the corrole complex enables the reduction of CO 2 to the CO 2 •− intermediate at a lower cost of energy.
We suggest a mechanism similar to that proposed by Koper et al. 53 . which is illustrated in Fig. 4 The proposed reaction pathways for the formation of the methanol and ethanol were further substantiated by conducting reduction studies of possible intermediates like formic acid, formaldehyde, methanol, and glyoxal under the same reaction conditions reported for Co-corrole-carbon paper electrodes. To test this conjecture, further reduction of the 2e − reduced product formic acid was induced, where it yielded a mixture of methanol and ethanol which indeed implies formic acid to be an important intermediate. Cyclic voltammetric measurements show an increase in current density ( Supplementary Fig. 34a) upon the addition of 10 mM of formic acid. CPE was performed at −0.73 V vs. RHE over Co-corrole-carbon paper for 5 h and the products obtained were analyzed in GC-MS spectrum where the presence of the characteristic peaks of methanol (m/z = 32) and ethanol (m/z = 46) were observed . The obtained results were further confirmed by 1 H-NMR spectroscopy ( Supplementary Fig. 34b). The proposed HCOOH intermediate under the low potential of reactivity gets reduced to HCHO and methanol, which are the final products and does not get further reduced. To verify the role of oxaldehyde in the current mechanistic cycle, 0.1 mM OHC-CHO was externally added under the reaction condition and CH 3 CH 2 OH was obtained proving oxaldehyde to be a key intermediate for CH 3 CH 2 OH formation ( Supplementary Fig. 35a, b). The high selectivity for alcohol is primarily due to the formation of the formic acid intermediate followed by formation of HCO • intermediate which is readily reducible to methanol. In the case of ethanol, we believe the formation of oxaldehyde is due to coupling of two HCO • intermediate which opens up a more favorable route to form ethanol. To complete the studies, we performed comparative measurements with two similar Co-corroles (cobalt triphenylphosphine 5,10,15-tris(pentafluoro-phenyl) corrole (PPh 3 -CoTpFPC) and cobalt triphenylphosphine 5,10,15-triphenyl corrole PPh 3 -CoTPC)), which consist of (a) three meso-C 6 F 5groups or of (b) three meso-C 6 H 5 -groups. The results are illustrated in the Supplmentary Informations file of the paper (Supplementary Fig. 43). The main reduction products under the same reaction conditions at pH = 6 were assigned to formic acid, methanol, and acetic acid. Only trace amounts of ethanol were found employing these two catalyst systems.
To conclude, we have demonstrated the electrochemical reduction of CO 2 to ethanol and methanol with a Cocorrole-carbon paper electrode at a low potential (−0.8 V vs. RHE) with a FE of 48% over a time period of 5 h. The Cocorrole-carbon paper electrode can withstand extremely high operational time of up to 140 h marking the highest efficiency for a molecular electrocatalyst reported so far in the literature. This is accomplished by the formation of a OHC-CHO type intermediate using a MeO-PEG(7)-S-modified cobalt corrole molecular catalyst. The Co-corrole molecule tends to stabilize different radical intermediates at the metal site. Thus, when reacted with a greater number of electrons highly reduced products are formed. On simulating the reactivity of Co-corrole, we found both experimentally and theoretically that in contrast to the CO pathway our catalyst proceeds via a formic acid pathway. By applying a redox potential of −0.73 V vs. RHE, a mixture of methanol and ethanol is detected. The obtained reaction products can only be explained if formic acid is developed temporarily, which is subsequently reduced to the formyl radical (HCO • ). The HCO • forms methanol as well as glyoxal and ultimately the glyoxal is then reduced to form ethanol serving our catalyst to operate at a low over potential.
Preparation of the working electrode. The working electrode consists of cobalt (III)-corrole immobilized over carbon fiber paper (A = 1 cm 2 ). For this, a 0.5 mmol solution of cobalt(III)-corrole in acetonitrile was prepared and drop-casted over carbon fiber to make a loading of 0.2 mg cm −2 . Further, we preserve the electrodes at room temperature to remove the excess acetonitrile. Then these electrodes were washed with 0.1 M pH = 6 phosphate buffer to get rid of the acetonitrile completely. Then the electrodes were dried in presence of CaCl 2 under high vacuum.
Cyclic voltammetry and CPE. Cyclic Voltammetry and CPE for CO 2 reduction were performed with a workstation (CH Instruments, Model CHI400A) in a twocompartment, three-electrode electrochemical H-cell, consisting of a gas tight cell with a total volume of 30 mL. Carbon paper was used as the working electrode. Ag/ AgCl (E Ag/AgCl = 0.222 V filled with 0.1 M KCl) as the reference, and Pt wire as the counter. The reference electrode potential was calibrated with respect to the reversible hydrogen potential using platinum working electrode and platinum wire as counter electrode in 0.5 M H 2 SO 4 electrolyte in H 2 atmosphere. This calibration result showed a shift of −0.222 V versus the NHE. All experiments were carried out at 25°C. The pH value of the solutions were obtained using a EUTECH pH 510 pre-calibrated with Thermo Scientific pH 4.01, 7.0, and 10.01 buffer solutions. For all the measurements CO 2 was continuously purged into the solution. All the potentials are represented in RHE scale with iR correction.
For electrochemistry in non-aqueous medium, acetonitrile was used as the solvent of choice with a similar electrochemical setup with glassy carbon as the working electrode, Ag/AgCl (0.1 M KCl) as the reference and Pt wire as the counter electrode with 0.1 M tetrabutylammonium perchlorate or 0.1 M tetrabutylammonium hexafluorophosphate as the electrolyte. Under this electrochemical conditions, the redox behavior of 0.01 M ferrocene in acetonitrile was studied, which was further used as the internal standard.
Detection and quantification of the CO 2 reduced products. 1 H-and 13 C-NMR spectroscopy of the carbon dioxide reduced liquid products were recorded a on a Bruker Ascend 700 MHz Avance III NMR spectrometer equipped with a cryoprobe and on a JEOL ECS-400 NMR spectrometer. As internal standard, 20 mL aqueous solution of 20 mM phenol and 10 mM of dimethyl sulfoxide were used. After CPE, to 350 µL electrolyte, 200 µL D 2 O, and 50 µL of the internal standard were added and transferred into a NMR-tube. During the measurements, the water peak was suppressed to increase the signal intensity of the analytes. The CO 2 reduced products were further analyzed using GC-MS. Trace 1300 GC and ISQ QD single quadruple GC-MS instrument with a TG-5MS capillary column (30 m × 0.32 mm × 0.25 µm) supplied by Thermo Fisher Scientific and DB-624 capillary column (30 m × 0.32 mm × 0.25 µm) supplied by Agilent were used for the same. For gaseous analysis CarboPLOT 007 capillary column (25 m × 0.53 mm × 0.25 µm) supplied by Agilent was used for separation and TCD for detection.
Electrochemically active surface area (ECSA) calculation. ECSA value is obtained by using the following equation where C D = electrochemical double layer capacitance which is obtained from the slope of the current vs. scan rate plot in the non-Faradaic region and C S = specific capacitance of the sample and in this case, C S = 0.040 mF cm −2 55 .
Analysis of the CO 2 reduced products in 1 H-NMR and GC-MS. Compounds formed at different potentials were detected directly by 1 H-NMR (integrated with respect to DMSO (δ = 2.71 ppm) as an internal standard) with a triplet signal at δ = 1.17 ppm and a quartet at δ = 3.67 ppm indicating ethanol ( Supplementary Fig.  17); a singlet at δ = 3.37 ppm showing the presence of methanol ( Supplementary  Fig. 17). For detection of formaldehyde, 2 mL of the reaction mixture was taken in a 20 ml headspace vial with 25 μL of ethanol and was acidified with 100 μL 1% ptoluenesulfonic acid. This mixture was then heated at 60°C for 1 h and then GC-MS measurements were done 56  XPS measurements. XPS was performed by using a Theta Probe, Thermofisher, UK, using monochromatic Al Kα X-rays (hν = 1486.6 eV), spot size 400 microns and with a photoelectron take-off angle of 90°with respect to the surface plane. The binding energies were corrected using the C1s peak at BE = 284.6 eV that arises from adventitious hydrocarbon.