Electrochemical Reduction of N <sub>2</sub> O with a Molecular Copper Catalyst

triflate and MeCN is possible, if not likely, in solution. Additionally, a very broad resonance can be observed in the 31 P{ 1 H} NMR spectrum of 1-Cu (δ = -48 ppm) in concentrated samples, corresponding to the bridging P atoms of the ligand with a chemical shift similar to that of 1 (δ = -45.2 ppm).

Electrochemical Characterization. Cyclic voltammetry (CV) of 1-Cu in MeCN revealed three reduction processes. The first quasi-reversible wave at -0.8 V vs Fc + /Fc corresponds to a metal-centered Cu(II)/Cu(I) redox couple with a large peak-to-peak separation (ΔE p = 200 mV at 100 mV s -1 ) . A similarly large ΔE p was observed in the previously reported PY5Me 2 analog. 56 Scan rate studies show an increase in ΔE p of 1-Cu with increasing scan rates, which indicates a changing coordination environment upon reduction from Cu(II) to Cu(I). 57 The two remaining irreversible cathodic processes appear at E p,c = -2.15 V and E p,c = -2.75 V vs Fc + / Fc and are tentatively assigned as Cu(I)L/Cu(I)L •-and Cu(I)L • /Cu(I)L 2-ligand-based reductions, respectively (L = 1; Figure S15). These assignments are supported by the CV of 1 in MeCN, which shows that the substituted pyridine ligand can be reduced by 1 e -in the absence of a metal (Figure S23).

To confirm the presence of ligand-based reductions in the CV of 1-Cu, the control complex [(MeIm 4 P 2 Py)ZnOTf]-[OTf] (1-Zn) was synthesized following a similar procedure as its Cu(II) analog. An octahedral ligand field environment similar to that of 1-Cu is observed in the SXRD structure of 1-Zn (Figure S43). The average Zn-N Im (2.118(1) Å) bond is slightly shorter than the average Zn-N Py distance (2.153(4) Å) in its PY5Me 2 analog, consistent with the stronger donor properties of the imidazole donors. 58 Most notably, 1-Zn also contains an unusually long axial Zn-N Py bond (2.380(2) Å), which is again likely enforced by the longer bridging P-C bonds in ligand 1. 59 The CV of 1-Zn in MeCN shows three cathodic processes (Figure S21), with the first irreversible wave assigned as the Zn(II)L 2+ /Zn(II)L •+ redox couple (E p,c = -1.83 V). The remaining two, presumably ligand-based, irreversible reductions are at very negative potentials (E p,c = -2.54 V and E p,c = -2.73 V). Notably, the reduction of 1-Zn is milder than that of the analogous [Zn(PY5Me 2 )(MeCN)] 2+ complex, suggesting that the phosphine groups in 1 are more electron-withdrawing toward the apical pyridine ring. 60 The ligand-based reduction potential in 1-Zn is significantly shifted from those of 1-Cu (E p,c = -2.15 V) and the free ligand 1 (E 1/2 = -2.68 V). These di erences in potential can be attributed to the di erences in charge between the three species. Regardless of these di erences, all of these CV studies support the viability of a ligand-based reduction in this system.

Electrocatalysis. In dry solvent and electrolyte, the first two reduction potentials in the CV of 1-Cu under an atmosphere of N 2 O do not shift, suggesting that there is no binding of N 2 O prior to the reduction of the complex (Figure 3A). A catalytic current is observed upon the addition of water, and a slight anodic shift is observed with increasing equivalents (Figure 3B). The stoichiometry of this reaction is described in eq 1.

Reduction of N 2 O occurs after the ligand is reduced, as can be seen by the inflection in the catalytic wave after the Cu(I)L/ Cu(I)L •-redox couple (Figure 3A, blue trace). The slight anodic shift in the inflection point with increasing water concentration could suggest that water facilitates binding to Cu(I) and promotes N-O bond cleavage. Evidence of significant water coordination to Cu was not found, as the reduction potentials of 1-Cu remained unchanged with high concentrations of water in the absence of N 2 O (Figure S19). Similar electrochemical behavior is observed when the N n Bu 4 OTf supporting electrolyte is used instead of N n Bu 4 PF 6 , which excludes competitive OTf -binding in solution (Figure S18). The CV of 1-Zn in the presence of N 2 O shows that it is not a competent N 2 O reduction electrocatalyst, and we note that the reduction potential of ligand 1 is beyond the onset for the direct reduction of N 2 O with glassy carbon (see the SI for details). Both observations highlight the importance of Cu for electrocatalysis with this framework. Controlled potential electrolysis (CPE) in the presence of 1 atm N 2 O and 100 mM H 2 O using a reticulated vitreous carbon (RVC) working electrode was performed at -2.3 V for 1 h to investigate the product selectivity of 1-Cu. A linear increase of charge passed over time is seen up until ∼9 C is passed, after which charge consumption plateaus, indicating loss of activity (Figure 3C). This hypothesis is also supported by the slow drop in current over the course of electrolysis, which suggests catalyst degradation, as has been observed in related systems. 47 We note that only a low amount of background activity by the RVC electrode was observed in the absence of 1-Cu under identical conditions. Headspace analysis using TCD GC found N 2 to be the only gaseous product, with no detectable H 2 . Additionally, only a small amount of H 2 is produced in the absence of N 2 O under identical conditions (Figures S34 andS36), indicating that 1-Cu is not competent for proton reduction under these conditions. An average turnover number (TON) of 54 (2) was determined over the course of 1 h with a Faradaic e ciency of 83(8)% for N 2 . This nonquantitative Faradaic yield likely arises from decreasing activity over time due to catalyst degradation, presumably from competing reactions with water or the OH -product 61 as well as the limited stability of the reduced species. Further evidence of this is supported by the CV of 1-Cu, which shows the instability of the reduced species in the presence of water at slower scan rates (Figure S19). Finally, CV of the solution after CPE using a glassy carbon plate (Figure S27) shows no electrochemical features, suggesting that the decomposition product(s) is electrochemically inactive. Indeed, a significant amount of free ligand, which is electrochemically inactive at the applied potential, was identified by 1 H and 31 P{ 1 H} NMR spectroscopy after CPE (Figures S28 andS29).

Mechanistic Investigations. The electrochemical activity of 1-Cu prompted us to perform chemical and theoretical investigations of possible mechanistic steps. We hypothesized that any N 2 O reduction would necessarily proceed from an initially reduced congener of 1-Cu. We therefore investigated the reactivity of 1-Cu with the reducing agents. We initially tested catalytic chemical reduction of N 2 O using 0.1% Na/Hg as a reducing agent under conditions similar to those used in CPE experiments (see the SI). Although there is significant activity from Na/Hg and H 2 O in the absence of catalyst, the presence of 1 mol % 1-Cu more than doubles the amount of N 2 generated, with nearly all reducing agent consumed (Figure S33), supporting Cu-mediated catalysis. This observation also supports the use of Na/Hg as a surrogate for electrochemical reduction in mechanistic probes of 1-Cu.

We then attempted to obtain a more detailed characterization of the reduction products by chemically reducing 1-Cu with Na/Hg in MeCN (Scheme 2). In situ monitoring of this reduction by UV-visible spectroscopy shows the disappearance of the initial d-d transitions in the spectrum, followed by a slow growth of two bands at approximately 410 and 680 nm over the course of 1 h (Figure S38). We propose that these absorption bands are related to the catalytically relevant species, since the Cu(I) intermediate is expected to have no signal in the visible region, consistent with our observations.

To test this hypothesis, characterization of this putative oddelectron-reduced species was attempted using EPR spectroscopy. An EPR signal consistent with an S = 1/2 complex distinct from the EPR signal of the Cu(II) starting complex was observed in a MeCN frozen solution (33 K) (Figure 4A). The frozen-solution EPR spectrum of the reduction of 1-Zn (Figure S40) shows a similar but distinct signal that supports the viability of ligand-based radicals in this system. The signal from the reduction of 1-Cu can be simulated with g = 2.010, 2.013, and 2.012, A( 63 Cu) = 12, 11, and 7 MHz, and A( 14 N) = 42, 3, and 2 MHz, consistent with a primarily pyridine centered radical. We note this signal is qualitatively similar to a Scheme 2. Reduction of 1-Cu by 2e -using Excess Na/Hg previously reported 2,6-disubstituted pyridine radical (broad single line, peak-to-peak 120 G). 62 We also performed DFT calculations on possible reduced complexes to gain insight into the catalytically relevant species (Figure S44). Geometry optimizations of a singly reduced Cu(I) intermediate predict a four-coordinate geometry with one ligand arm dissociated. This prediction is consistent with the structural changes inferred from the quasi-reversible redox couple in the CV of 1-Cu at -0.8 V vs Fc + /Fc. Further reduction by an additional electron is ligand-based, as illustrated by the spin density, which is primarily on the pyridine donor with some delocalization onto the imidazole arms through the σ* of the P-C bond (Figure 4B). We have performed calculations of the EPR parameters of this complex, which match the values obtained from simulation of the experimental spectrum well. Namely, both DFT and simulation support moderate hyperfine coupling to N and Cu (Table S1).

It is important to note that there are multiple thermodynamically accessible isomers for these reduced species that may contribute to catalyst degradation, as has been reported in a similar N5 species for water reduction (see SI). 63 Thus, the depicted geometry of this doubly reduced intermediate is only a model, and other coordination geometries and ligation environments (di erent solvates, for example) are possible if not likely. We hypothesize that this reduced species then rapidly reacts with N 2 O and water to generate N 2 and OH -, consistent with the catalytic onset that is observed just beyond the second reduction event in the CV of 1-Cu. Indeed, CV with variable concentrations of 1-Cu and H 2 O suggests firstorder behavior in both of these reagents, consistent with this hypothesis (Figure S20). Regardless of the exact reduced speciation of 1-Cu, the spectroscopy and calculations support a vital role for ligand redox noninnocence as an electron storage/ shuttling mechanism for catalysis. This echoes possible roles for the multi-Cu cluster in biological N 2 O reduction and may point to a more fundamental requirement for additional redox cofactors in N 2 O reduction catalysis by Cu.

■ CONCLUSION We report the first example of a molecular Cu catalyst for N 2 O reduction. This complex, 1-Cu, enables electrocatalytic reduction of N 2 O with water to a ord N 2 with a high Faradaic e ciency. Catalytic reduction with a chemical reducing agent was also demonstrated with dilute Na/Hg. Electrochemical studies support the onset of catalysis after reduction of 1-Cu twice, and a combination of spectroscopy and theory supports the importance of ligand-based reductions in forming reduced intermediates. While 1-Cu is a highly unusual example of a Cubased catalyst for N 2 O reduction, we do note that previous molecular electrocatalysts with other transition metals show higher Faradaic e ciencies (>90%) and less decomposition. [43][44][45][46][47][48] For instance, [Re(2,2′-bipyridine)(CO) 3 Cl] boasts ∼200 turnovers over the course of 2 h with no significant catalyst degradation, albeit with a comparatively precious metal. 48 The findings reported here provide an initial proof-ofconcept validation for further e orts toward the design of new Cu-based molecular electrocatalysts for N 2 O reduction. In addition to improved performance metrics and increasing the stability of the active catalyst, there remain interesting mechanistic questions surrounding both electron transfer principles and details about N 2 O binding, reduction, and protonation events.