Electrochemical CO2 Reduction with a Heterogenized Iridium−Pincer Catalyst in Water

Immobilization of well‐defined homogenous (electro)catalysts onto conductive supports offers an attractive strategy for designing advanced functional materials for energy conversion. In this context, this study reports (i) the introduction of a pyrene anchoring group on a PNP−pincer IrI complex previously described as a selective catalyst for the electrodriven CO2 reduction (CO2RR) into CO in DMF/water mixtures, (ii) the comparison of its CO2RR activity in DMF/water mixtures with the ones of two pyrene‐free reference complexes, and (iii) its activity in pure water after immobilization onto carbon nanotubes (CNTs). Surprisingly, in homogeneous conditions we find HCOO−, instead of CO, as the main CO2 reduction product for the three catalysts. After immobilization on CNTs, even if non‐negligible competitive proton reduction reaction is observed in fully aqueous media, the complex is still able to drive CO2RR and produce HCOO− with a significantly lower overpotential with respect to solution studies.


Introduction
Direct electrochemical conversion of carbon dioxide into fuels and chemicals offers attractive prospects for the transition towards a sustainable "low-carbon" economy, and thus, is currently a domain of intense research. [1] In this field, the immobilization of well-defined transition metal complexes able to drive the CO 2 Reduction Reaction (CO 2 RR) onto conductive supports is an appealing strategy. [2] Indeed, heterogenized molecular catalysts often combine the benefits of heterogeneous catalysis [3] (catalyst recyclability, efficient electron transfer from the electrode support to the catalytic centers) with those of homogenous catalysis [4] (uniform catalytic sites, high control on the active site properties and tunability, simpler mechanistic investigations and relative ease to establish structure-activity correlations). Furthermore, it can drastically reduce drawbacks typical of homogeneous catalysts, like their diffusion-dependent kinetics and poor stability. The immobilization of a molecular catalyst can indeed help preventing its fast deactivation under turnover conditions by limiting its free diffusion in the bulk and the occurrence of detrimental intermolecular reactions involving highly reactive intermediates (dimerization, ligand degradation) or its poisoning with reaction products. [2b,5] Besides, once heterogenized, the catalyst of interest can be used regardless of its intrinsic solubility into the media selected, sometimes allowing to promote a remarkable new reactivity. [6] Heterogenization can, thus, offer an interesting approach for transposing commonly used catalysts from organic to aqueous conditions that are generally more desirable for the development of functional devices, while potentially broadening their scope of activity.
When the heterogenization of a molecular CO 2 RR catalyst is targeted, the nature of the catalyst and that of the supporting material, as well as the immobilization strategy have to be attentively planned.
The choice of multi-walled carbon nanotubes (MWCNTs) as supporting material is attractive since it combines an excellent electrical conductivity and porosity of the substrate with the possibility to easily functionalize it through non-covalent π-π stacking interactions. This anchoring mode does not affect the conductivity of MWCNTs and can result in excellent loading densities and surface stability of the complexes. [7] Significant precedents of CO 2 RR catalysts immobilized on carbon nanotubes by π-π stacking interactions include manganeseÀ bipyridine, [2b] nickelÀ cyclam, [8] iron-, copper-, cobalt-and nickel-complexes of porphyrins [2c,d] and phthalocyanines, [2c,d,9] and rheniumÀ diimine [10] complexes. In all these systems, working in aqueous or mixed organic/water medium, heterogenization allowed for a significant enhancement of the catalytic activity and/or stability of the catalysts compared to the homogeneous counterparts.
Concerning the nature of the molecular catalyst to be immobilized, we focus here on iridiumÀ pincer complexes. MetalÀ pincer complexes are potentially attractive for electrocatalytic CO 2 RR since they incorporate tridentate redox noninnocent ligands that can allow for the stabilization of critical catalytic intermediates. Iridium complexes are known as CO 2 reduction electrocatalysts since late 90s, [11] but the introduction of pincer ligands to support this activity has only been proposed by Brookhart et al. in 2012. [12] In this seminal work an iridium(III) dihydride complex featuring an anionic PCPÀ pincer ligand was demonstrated to be a selective electrocatalyst for the production of HCOO À in acetonitrile/water mixtures. A later extension of this work demonstrated the versatility of pincer ligand platforms for designing fully water-soluble catalysts. [13] Finally, one of the Brookhart complexes was successfully integrated onto carbon nanotube-coated gas diffusion electrodes (GDEs) by non-covalent binding, leading to a remarkable increase of turnover number from 40 to 2 · 10 5 while maintaining the selectivity of the complex for formate production. [14] Given this promising precedent, we decided to extend this strategy to a PNP pincer-supported Ir I complex (Ir H , Scheme 1) recently described by Brudvig et al. as a selective electrocatalyst for CO 2 reduction to CO in DMF/water mixtures. [15] In the following, we thus report (i) the functionalization of the ligand scaffold with a pyrene anchoring group and its complexation to the Ir I center; (ii) the comparison of the CO 2 RR activity of the pyrene-tagged complex (Ir OPyr , Scheme 1) in DMF/water mixtures with the activity of the two reference complexes Ir H and Ir OMe , the latter being substituted with a methoxy group in the para-position of the pyridine ring of the pincer (Scheme 1); and (iii) the immobilization of Ir OPyr on MWCNTs to perform CO 2 RR catalysis in fully aqueous medium.
Interestingly, using a mercury (Hg) pool working electrode we observe the formation of HCOO À as the main reduction product (instead of CO, as previously reported for Ir H using carbon paper working electrode) [15] for the three catalysts tested in homogeneous conditions. Gratifyingly, the selective generation of HCOO À is retained after immobilization of the pyrenetagged complex even in pure water, together with a significant decrease of the CO 2 RR overpotential.

Synthesis
In order to prepare iridiumÀ PNP@MWCNTs hybrids, the first step was the introduction of a pyrene anchoring group on the periphery of the PNPÀ pincer. More specifically, the PNP H ligand framework shown in Scheme 1 was substituted in the paraposition of the pyridine ring with a butylpyrene tag to afford Scheme 1. Multi-step synthesis of the ligand PNP OPyr and complexation: (i) K 2 CO 3 , MeCN, reflux overnight, 64 %; (ii) THF:MeOH 5 : 2, from 0°C to 60°C, 1 h, 95 %; (iii) KOH, THF, reflux 20 h, 77 %; (iv) THF (CH 2 Cl 2 ), from À 78°C to RT overnight, 83 %; (v) toluene, 80°C, 48 h, 60 %; (vi) acetone, 30 min, CO bubbling, 68 %. The reference complexes Ir H [15] and Ir OMe are depicted in the inset. the PNP OPyr ligand via the strategy displayed in Scheme 1 and briefly described thereafter. In the first step, 1-(4-bromobutyl)pyrene was reacted with diethyl 4-hydroxypyridine-2,6-dicarboxylate to afford derivative 1, wherein the pyrene and pyridine moieties are linked by an ethereal bond, via a classical Williamson-type reaction. Next, selective reduction of the ethyl ester groups by LiBH 4 followed by tosylation of the resulting diol yielded the electrophilic species 3 that can easily undergo nucleophilic substitution by bis-tert-butyl phosphineÀ borane to afford 4. The final ligand PNP OPyr was isolated after deprotection of the phosphines in the presence of 1,4diazabicyclo[2.2.2]octane (DABCO) as BH 3 -scavenger. The overall yield of this five-steps synthesis is of 23 % from 1-(4-bromobutyl)-pyrene.
The preparation of the novel Ir I À carbonyl pincer complexes, Ir OPyr and Ir OMe , was inspired by the synthesis of the unsubstituted Ir H complex. [15] The complexes were obtained in good yields (68-70 %) by mixing the respective ligands with the Ir I precursor [Ir(coe) 2 (acetone) 2 ]PF 6 in acetone before treating the mixture with CO. The formation of the metalÀ carbonyl adducts was confirmed by the observation of distinctive bands in the infrared spectrum of the compounds, centered at 1952 cm À 1 , 1956 cm À 1 , and 1962 cm À 1 [16] for Ir OPyr , Ir OMe , and Ir H , respectively, and corresponding to the stretching mode of metal-bound CO. The trend observed for the CO frequencies (i. e. a 4-10 cm À 1 decrease from Ir H to Ir OR ) is in good agreement with an increase of the donor character of the ligands from PNP H to PNP OR (R = Me or "Pyr"). Similarly to the Ir H parent compound, both Ir OR complexes display a plane of symmetry orthogonal to the pyridine plane, as attested by the presence of only one singlet in their 31 P NMR spectra.

Redox properties of the Ir complexes
The electrochemical properties of the complexes were investigated by Cyclic Voltammetry (CV) in DMF solutions at 0.5 mM with 0.1 M nBu 4 NPF 6 supporting electrolyte under Ar atmosphere, using glassy carbon (GC) as the working electrode ( Figure 1 and Table 1). In the following, all potentials are referred to the Fc + /0 couple. First, we focused on the Ir H and Ir OMe reference complexes lacking the pyrene moiety ( Figure 1). In good agreement with the previous report, [15] the CV of the unsubstituted Ir H exhibits an irreversible reduction peak centered at Ep c = À 2.09 V that can be attributed to the first monoelectronic reduction of the {(PNP)Ir} unit. Unsurprisingly, the first reduction occurs at a more cathodic potential for the Ir OMe complex, with Ep c = À 2.38 V, due to the stronger donor character of the methoxy-substituted ligand. Next, we investigated the pyrene-tagged Ir OPyr complex. In the cathodic direction, the CV displays two fully irreversible processes at Ep c1 = À 2.25 V and Ep c2 = À 2.40 V followed by a reversible system at E 1/2 = À 2.52 V (ΔEp = 92 mV). The linear dependence of the latter peak's intensity to the square root of the scan rate ( Figure S1), confirms that the process is diffusion-controlled and excludes the observation of adsorbed species under the present conditions. A comparison with the CV of pure pyrene under the same conditions (E 1/2 = À 2.50 V, ΔEp = 98 mV, see Supporting Information, Figure S2), permits to assign the reversible cathodic peak of Ir OPyr to the one-electron reduction of the pyrene moiety. The first two irreversible cathodic peaks (Ep c1 = À 2.25 V, Ep c2 = À 2.40 V) are thus attributed to the one-electron reduction of the {(PNP)Ir} unit (see DFT calculations below). The splitting of the electrochemical signature of Ir OPyr in two cathodic peaks is not observed in the case of Ir H and Ir OMe , therefore it must be related to the presence of the pyrene tag. More specifically, we tentatively attribute this splitting to the existence an equilibrium in the system related either to the presence of two conformations of the Ir OPyr [17] or labile aggregates in the electrolyte.
DFT calculations performed on the three complexes support the assignments of the cathodic peaks observed in the CVs. The structures of Ir OPyr , Ir OMe , and Ir H , were optimized using ORCA starting from Avogadro 3D-built structures and the frontiers orbitals were then calculated ( Figure 2 and Table 2). We found that Ir OMe and Ir H have quite similar LUMOs with an electronic distribution largely delocalized over both the Ir center and the   PNP ligand framework, indicating that the one-electron reduction is both metal-and ligand-based in good agreement with the analysis reported by Brudvig et al. [15] The second reduction is predicted to be much more energetic for these two complexes, since the LUMO + 1 is 0.61-0.88 eV upward the LUMO, and accordingly it is not observed in the CVs. Conversely, in Ir OPyr the LUMO and LUMO + 1, localized over the pyrene moiety and over the {(PNP)Ir} unit respectively, are almost degenerated (Δ LUMO-LUMO + 1 = 106 meV), which is in agreement with the observation of two successive one-electron reductions in the CV (even if the sequence of the two processes is not properly predicted at the level of DFT calculations used here).

CO 2 RR electrocatalysis in DMF solution
As a preliminary step prior to heterogenize the Ir OPyr complex, its electrocatalytic properties were investigated in homogeneous conditions and compared to those of the parent Ir H [15] and Ir OMe species to evaluate the effect of adding the pyrene moiety. Adapting the previously reported conditions, [15] the potential activity of Ir OPyr to drive CO 2 RR was first probed by CV using 0.5 mM solutions of the complex in DMF, under Ar or CO 2 saturated atmosphere, both in the absence and in the presence of water (4.6 M) as a source of protons. As shown in Figure 3, after saturating the mixture with CO 2 , an increase of the current intensity is observed close to the cathodic system corresponding to the {(PNP)Ir} reduction. This observation is coherent with the occurrence of a catalytic process, pointing towards CO 2 RR. This is further supported by the drastic increase in the intensity of the catalytic peak at E cat/2 = À 2.40 V upon addition of 4.6 M H 2 O. Importantly, after purging the latter solution with Ar, the catalytic wave is no longer observed, thus confirming that it can be mostly attributed to CO 2 RR and not to the hydrogen evolution reaction (HER) at the CV timescale. A rinse test was performed to exclude that surface-confined species were responsible for the catalytic activity observed (see Supporting Information, Figure S3): after recording a CV in the presence of 0.5 mM Ir OPyr in DMF 0.1 M nBu 4 NPF 6 + 4.6 M H 2 O under CO 2 , the GC electrode was gently washed with DMF and used again as the working electrode in a fresh electrolyte solution in the absence of the complex (DMF 0.1 M nBu 4 NPF 6 + 4.6 M H 2 O under CO 2 ). The resulting CV did not display any catalytic wave, confirming that the active species for CO 2 RR is homogeneous as previously observed for Ir H . [15] When comparing the CV recorded for Ir OPyr    and Figures S4 and S5), the higher apparent catalytic activity of the pyrene-derivate is striking: i cat /i p of 13.0 at E cat/2 = À 2.40 V for Ir OPyr vs 6.1 at E cat/2 = À 2.38 V for Ir OMe , and 2.0 at E cat/2 = À 2.07 V for Ir H . It should be noted that at the CV timescale the behavior of the Ir H complex recorded under CO 2 in our experiments is similar to the previously reported data. [15] The catalytic overpotentials for HCOO À production, which is the main product under the present conditions (see Controlled-Potential Electrolysis data below), have been calculated to be ( Table 3): [18] η(Ir OPyr ) HCOO -= 950 mV, η(Ir OMe ) HCOO À = 930 mV, and η(Ir H ) HCOO À = 620 mV. As expected, the values of η(Ir OPyr ) HCOOÀ and η(Ir OMe ) HCOOÀ are higher than η(Ir H ) HCOO -, due to the higher donor character of the corresponding ligands.
In an attempt to rationalize the apparent superior catalytic rates of Ir OPyr observed at the CV timescale, the combination of the following elements can be considered: 1) The presence of the electron rich ROÀ substituent in para position of the pyridine moiety that favorably tunes the electronic properties of the complex, as suggested by the higher value of i cat /i p for Ir OMe compared to Ir H (6.1 vs 2.0, respectively, see Table 3). 2) The covalent linkage between the RO-substituted {(PNP)Ir} and the pyrene moieties in Ir OPyr , which probably act in synergy to trigger catalysis (i cat /i p (Ir OPyr ) = 13.0). We suggest that the close reduction potentials of the two units allow pyrene, which in its free form also exhibits an apparent modest CO 2 RR activity (i cat /i p = 3.1), to play the role of electron mediator in the system.
Next, Controlled-Potential Electrolysis (CPE) experiments were carried out in DMF in the presence of 4.6 M H 2 O containing 0.1 M nBu 4 NPF 6 supporting electrolyte, for 0.5 mM solutions of Ir H , Ir OMe , Ir OPyr , and pyrene. A mercury working electrode was employed to limit both the background H 2 production and the potential formation of catalytically active metallic nanoparticles over time. For all the systems we run CPE at À 2.43 V vs Fc + /0 , which corresponds roughly to the middle of the catalytic wave observed in the CVs of the Ir OR complexes. When running 2 h CPE, fairly stable currents were observed in most of the cases (see Figures S7-11 for the chronoamperometry profiles). The products resulting from CO 2 RR and the competitive proton reduction reaction were quantified by gas (CO, CH 4. and H 2 ) and ionic chromatography (HCOO À , C 2 O 4 2À ). Among all products tested, CH 4 could not be detected in any significant amounts and will not be further discussed in the following.
The evolution of the products' distribution during CPE is displayed as a function of time in Figure 5 in the case of Ir OPyr . Tables 4 and 5 summarize the 2 h long CPE results obtained in homogeneous conditions for all the systems investigated.
In our case, a different selectivity was observed for Ir H complex than originally reported by Brudvig et al. under very    [15] we found HCOO À as the main product of CO 2 reduction with a FE of 71 %, whilst the complex was reported to be selective for CO production under all potentials tested in the original work. In order to determine if this variation of selectivity could be associated to the different nature of the electrode material employed (mercury vs carbon), we repeated the experiment using carbon paper as working electrode (see Supporting Information, Table S1, S2, Figure S12) in otherwise identical conditions. Interestingly, in this case both CO and HCOO À were obtained with similar FE (25 % and 34 %, respectively), and the proportion of H 2 produced increased significantly (41 % FE, Table S1). The discrepancy with the previous report can tentatively be rationalized by considering the coexistence of competitive, almost isoenergetic CO 2 RR pathways, where minor modifications of the reaction conditions (electrode material, substrates/catalyst concentrations, temperature, etc.) favor one over the other(s). Nevertheless, the formation of HCOO À as the main product of CO 2 RR in both cases agrees well with the known ability of iridiumÀ pincer complexes to form metalÀ hydride adducts prone to CO 2 insertion [12][13][14]19] via an ET H mechanism. [4] Isomerization of the metal-formate species involved in the ET H route, to a hydroxycarbonyl intermediate of a competitive ET M pathway, [4] can be envisaged under the Brudvig conditions, leading to CO release together with water as a side product. It is worth noting that a similar behavior was previously reported in the case of ruthenium catalysts. [20] Looking at Table 4, we observe that for all three Ir complexes investigated, the main product of CO 2 RR is invariably HCOO À (71 % for Ir H , 79 % for Ir OMe , and 80 % for Ir OPyr ), while CO (20 % for Ir H , 16 % for Ir OMe , and 13 % for Ir OPyr ) is obtained as a side product. Many other electrocatalysts have been previously reported for the selective production of HCOO À from CO 2 in homogeneous solution, based on various transition metals (Ir, Ni, Pt, Rh, Mn, Fe, Ru, Co) and supported by a variety of common ligands (porphyrins, polypyridyl ligands, aza-macrocycles, pincers, etc.). [4] We can note that all the iridium catalysts studied here lead to relatively close product distribution. The calculated FE of HCOO À , CO, and H 2 production remain in the range of 70-80 %, 10-20 %, and 5-10 %, respectively, indicating only a minor effect of the substitution of the ligand scaffold on the selectivity of the complexes for CO 2 RR in this case. Importantly, only a minor evolution of product distribution is observed over time (see Figure 5), which suggests that the catalytic systems remain fairly stable over time and that the contribution of potentially degraded species in the process is low to negligible in most of the cases. [22]

Heterogenization and electrocatalysis in water
After ensuring that the catalytic activity for CO 2 RR of Ir OPyr was preserved with respect to those of the parent catalysts Ir H and Ir OMe , we next explored its activity in fully aqueous solutions. With this purpose, we immobilized Ir OPyr on multi-walled carbon nanotubes (MWCNTs) electrodes through non-covalent π-π stacking interactions with the pyrene anchoring group. More specifically, an ink was prepared by sonication of MWCNTs in acetonitrile followed by the addition of Ir OPyr and Nafion as binder. The ink obtained was then deposited on glassy carbon (GC) electrode, which was then washed with acetonitrile to remove any nonspecifically bound complex, then dried. The amount of iridium incorporated in the Ir OPyr @MWCNTs/GC hybrid material was determined by ICP-OES to be 1.70 % (w/w) corresponding to a catalyst loading of 0.76 μmol · cm À 2 and 47 % grafting yield of the complex on MWCNTs. The chemicallymodified electrodes were then evaluated for CO 2 RR electrocatalysis in water (CO 2 -saturated aqueous 0.5 M KHCO 3 solution, pH = 7.4), and compared to bare MWCNTs deposits probed under identical conditions.
The CVs of electrodes coated with Ir OPyr @MWCNTs exhibited slightly higher cathodic currents under CO 2 as compared to bare MWCNTs deposits, in the range between À 0.8 to À 1.2 V vs Ag/AgCl (Figure 6), which suggests moderate CO 2 RR activity of the Ir OPyr @MWCNTs/GC hybrids in water.
Next, we run 2 h CPE experiments to confirm that the CO 2 RR activity was indeed maintained, and to compare the selectivity of Ir OPyr after heterogenization in these conditions. Based on the CV, the applied potential was fixed at À 1.10 V vs Ag/AgCl, corresponding to À 1.63 V vs Fc + /0 , for both Ir OPyr @MWCNTs and bare MWCNTs deposits. This corresponds to a~860 mV anodic shift compared to the same catalytic system studied in organic  medium (E cat/2 = À 2.40 V vs Fc + /0 ). Such an anodic shift of the catalytic reduction potential for CO 2 RR (+ 450 mV) was previously reported in the case of a NiÀ cyclam complex after its immobilization on carbon electrodes, via π-π interactions, [8b] and was attributed to the strong interactions between the complex and the electrode surface. In the present system, either a similar explanation and/or the stabilization of critical catalytic intermediate(s) in aqueous media can be proposed. Further investigations would, however, be required to conclude on the origin of the significant shift observed in the case of Ir OPyr . The charge passed during CPE of Ir OPyr @MWCNTs/GC hybrids is significantly higher than complex-free MWCNTs/GC (see Table 6), suggesting that the CO 2 RR activity of the complex is maintained within the hybrid electrode. Furthermore, fairly stable currents were observed during the CPE (see Figure S19). The FE and TON obtained for Ir OPyr @MWCNTs/GC and bare MWCNTs/GC are reported in Tables 6-7, the evolution of the products distribution as a function of time is displayed in Figure 7. It should be noted that TON were determined based on the total amount of Ir catalyst incorporated in the working electrode, as determined by ICP-OES, and not on its accessible electroactive fraction. This choice implies that TON reported in Table 7 are likely underestimated.
Notably, Ir OPyr retains an activity for CO 2 RR in water, leading primarily to the release of HCOO À , as observed in homogeneous organic solutions, as well as for the Brookhart Ir pincer complexes. [12][13][14] The faradaic efficiency for HCOO À production reaches 60 % after 2 hours, which makes the present system a rare example of significant production of HCOO À from CO 2 in aqueous conditions (FE > 50 %) catalyzed by an immobilized transition-metal complex. To the best of our knowledge, this was only previously reported for electropolymerized [Ru-(bpy)(CO) 2 ] n films, [23] or after immobilization of the abovementioned Brookhart IrÀ pincers [14] or Mn(bipyridine) complexes [2b] onto carbon nanotubes.
The only other secondary CO 2 RR products found in our heterogeneous system are C 2 O 4 2À (FE 17 %) and CO (FE 2 %). The presence of a significant amount of oxalate among the products of catalysis is compatible with the concurrent activation of a ET M [4] or an outer-sphere [21] pathway. Interestingly, when considering the evolution of the FE over time (Figure 7) we can observe a limited variation of the products distribution suggesting a relatively stable system. We can mostly note a slight decrease in the competitive processes to the benefit of HCOO À production, most likely due to slight modification of the microenvironment of the catalytic centers (local proton gradient buildup, dehydration of the Nafion film….) rather than to the evolution of the Ir complexes themselves under turnover conditions. This hypothesis is supported by the X-ray photoelectron spectroscopy (XPS) analysis of Ir OPyr @MWCNTs modified electrodes before and after 2 h CPE experiments ran at À 1.10 V vs Ag/ AgCl, which shows almost identical signatures for both the Ir and P contributions of the complexes (see Figures 8 and S33-34). The initial spectra recorded before CPE show the expected doublet from Ir I species in the iridium 4 f region at 61.9 eV (4f 7/2 ), together with a shoulder attributed to an Ir III species at 62.6 eV (4f 7/2 ). Similar signatures are recorded on pristine Ir OPyr powder, for which the 1 H and 31 P NMR spectra exclude the presence of Ir III co-products. The unexpected Ir III species are, thus, most likely generated in situ by exposure to X-ray radiation as recently reported for other hybrid materials featuring Ir I centers. [24] The phosphorus region features a P 2p 3/2   peak at 131.7 eV ( Figure S34) related to the coordinated phosphine ligands as well as another one at 136.4 eV corresponding to the PF 6 À counter-ions. By comparing the relative intensities of XPS peaks, a P/Ir ratio of 1.98 is obtained. This observation agrees perfectly with the expected stoichiometric phosphine:Ir ratio in Ir OPyr and, thus, indicates that the integrity of the complex is preserved upon grafting. Most importantly, the XPS spectrum recorded after 2 h turnover reveals the same pattern in both the Ir 4 f region (4f 7/2 at 62.0 eV) and the P 2p one (P 2p 3/2 at 131.9 eV), while maintaining a P/Ir ratio of 2.02. The latter is a definitive evidence of the retention of the initial molecular structure of the catalyst under catalytic conditions.

Concluding Remarks
We have synthesized a new PNPÀ pincer ligand tagged with a pyrene anchoring group (PNP OPyr ), which can find general application for the heterogenization of metalÀ pincer complexes onto carbon-based materials and their application for small molecules activation electrocatalysis. As a study-case we have used this ligand to heterogenize a previously reported pincerÀ iridium complex (Ir H ), known to catalyze the electrochemical CO 2 reduction reaction. [15] In homogeneous water/DMF mixtures the CO 2 RR selectivity of the pyrene-tagged Ir catalyst (Ir OPyr ), as well as that of the pyrene-free counterparts including the original Ir H complex, was found to differ from the previous studies: the main CO 2 reduction product being invariably HCOO À instead of CO. This observation may be related to iridium-formate and hydroxycarbonyl intermediates close in energy, which could allow switching selectivity by subtle variations of the experimental conditions.
After immobilization of Ir OPyr on MWCNTs, the system still demonstrates a bias for CO 2 RR over HER in pure water. Under these conditions, the CO 2 to HCOO À reduction activity is retained, which represents a rare case of significant formate production from CO 2 in aqueous conditions catalyzed by immobilized molecular complexes. We can also observe (i) the high stability of the catalyst under turnover conditions, as demonstrated by XPS measurements performed on the immobilized catalyst before and after CPE, and (ii) a surprisingly large anodic shift in the catalytic reduction potential for CO 2 RR. The latter can be tentatively attributed to a strong interaction of the complex with the carbon surface or the stabilization of critical catalytic intermediates in fully aqueous media, even if a severe structural rearrangement of the catalyst cannot be excluded at this point.
Our current efforts are focused on the exploration of the electrochemical CO 2 RR abilities of other metalÀ pincer complexes with pyridine-based PNP ligands, which could be successfully immobilized on electrode surfaces via π-π stacking interactions by employing the pyrene-tagged PNP OPyr ligand.
All the syntheses were carried out under an atmosphere of argon by using standard Schlenk techniques (most of the organic compounds) or in a nitrogen-filled glovebox (synthesis of 5, iridium complexes), while the purification procedures were performed under air atmosphere, with the exception of the isolation of compound 5.

1
H NMR (400 MHz), 31 P{ 1 H} (162 MHz) and 11 B{ 1 H} NMR (128 MHz) spectra were recorded on a Bruker Avance III 400 MHz spectrometer in suitable solvent, and spectra were referenced to residual solvent ( 1 H) or external standard ( 31 P{ 1 H}: H 3 PO 4 ). The infrared spectra were recorded on an Agilent Technologies Cary 630 FTIR spectrometer equipped with a standard transmission module, a diamond ATR one, and a DialPath sample interface. The low-resolution mass spectra were recorded either on an Amazon speed ion trap spectrometer or a LXQ type Thermo Scientific spectrometer, both equipped with an electrospray ionization source (ESI). The samples were analyzed in positive ionization mode by direct perfusion in the ESI-MS interface (ESI capillary voltage = 2 kV, sampling cone voltage = 40 V). The high-resolution mass spectra were recorded on a LTQ Orbitrap XL Thermo Scientific spectrometer equipped with an ESI source. À counter-ion initially associated to the pristine Ir OPyr complex. See Figure S34 for the peaks deconvolution.
Emission spectra were recorded in deoxygenated solvents at room temperature using a Fluoromax 4 (Horiba). The iridium content was determined on an ICP-OES ACTIVA Jobin Yvon apparatus from solutions obtained by treatment of the material with sulfuric acid and aqua regia in a Teflon reactor at 400-450°C. Milli-Q water was purified through a Millipore system.
The amounts of evolved hydrogen, carbon monoxide and methane were determined by sampling 50 μl of the headspace in Perkin Elmer Clarus 580 gas chromatograph equipped with a molecular sieves 5 Å column (30 m-0.53 mm), a methanizer and thermal conductivity (TD) and flame ionization (FI) detectors. Formate and oxalate anions were determined by ionic exchange chromatography (883 Basic IC, Metrohm).
X-ray photoelectron spectroscopy (XPS) analyses were carried out with a Versa Probe II spectrometer (ULVAC-PHI) equipped with a monochromatic Al Ka source (hν = 1486.6 eV) at CEA Grenoble. The core level peaks were recorded with constant pass energy of 23.3 eV. The XPS spectra were fitted with CasaXPS 2.3 software using Shirley background. Binding energies are referenced with respect to the adventitious carbon (C 1 s BE = 284.8 eV).
Electrochemistry and catalytic experiments. Electrochemical experiments were performed using a SP-300 Bio-Logic potentiostat in an air-tight three-electrode cell with a Ø 1.6 mm glassy carbon disk working electrode, a Pt wire as counter electrode and an Ag/AgCl, KCl (3 M) reference electrode separated from the bulk solution by a Vycor frit. Experiments were performed under argon or CO 2 using 0.5 mM concentrations of the complexes in 3 mL anhydrous electrolyte solution. Cyclic voltammograms (CVs) were collected at a scan rate of 100 mV s À 1 at room temperature. The working electrode was polished before each measurement on a MD-Nap polishing pad with a 1 μm monocrystalline diamond paste, rinsed with ethanol and dried under air. Controlled potential electrolysis (CPE) experiments were carried out with mercury or carbon paper working electrodes in a two compartment H-cell, where the counter electrode compartment was separated from the compartment containing the working and reference electrodes by a fritted glass. Electrolyte solutions were 0.1 M [n-Bu 4 N]PF 6 in anhydrous dimethylformamide and purged with argon for 10 minutes prior to use. For experiments under CO 2 atmosphere, the solution was purged with CO 2 for 45 minutes. Ferrocene was added at the end of electrochemical experiments as an internal standard. The catalytic experiments (i. e. the CPE experiments followed by the quantification of the CO 2 RR products by GC/HPLC) were repeated three times and the average values of faradaic efficiency (FE) and turnover numbers (TON) are reported in Tables 4-7, S1-S2. Standard deviations are in a range of~5 % from the mean value.
Computational methods. DFT calculations were performed using ORCA program package version 4.2.1 [31] using the long range and dispersion-corrected B3LYP/G hybrid functional. [32] The def2-SVP basis set was used for all atoms. [33] Calculations were performed with the presence of a solvent reaction field of dimethylformamide produced by the conductor-like polarizable continuum model (CPCM). [34]