Electrocatalytic Water Oxidation by a Homogeneous Copper

1 is EPR silent at 77 K in acetonitrile and taken together with the 1H NMR suggests ..... To gain deeper insight into the landscape of intermediates a...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/JACS

Electrocatalytic Water Oxidation by a Homogeneous Copper Catalyst Disfavors Single-Site Mechanisms Sara J. Koepke, Kenneth M. Light, Peter E. VanNatta, Keaton M. Wiley, and Matthew T. Kieber-Emmons* Department of Chemistry, University of Utah, Salt Lake City, Utah 84112-0850, United States S Supporting Information *

ABSTRACT: Deployment of solar fuels derived from water requires robust oxygen-evolving catalysts made from earth abundant materials. Copper has recently received much attention in this regard. Mechanistic parallels between Cu and single-site Ru/Ir/Mn water oxidation catalysts, including intermediacy of terminal Cu oxo/oxyl species, are prevalent in the literature; however, intermediacy of late transition metal oxo species would be remarkable given the high d-electron count would fill antibonding orbitals, making these species high in energy. This may suggest alternate pathways are at work in copper-based water oxidation. This report characterizes a dinuclear copper water oxidation catalyst, {[(L)Cu(II)] 2 -(μ-OH) 2 }(OTf) 2 (L = Me2TMPA = bis((6-methyl-2-pyridyl)methyl)(2-pyridylmethyl)amine) in which water oxidation proceeds with high Faradaic efficiency (>90%) and moderate rates (33 s−1 at ∼1 V overpotential, pH 12.5). A large kinetic isotope effect (kH/kD = 20) suggests proton coupled electron transfer in the initial oxidation as the rate-determining step. This species partially dissociates in aqueous solution at pH 12.5 to generate a mononuclear {[(L)Cu(II)(OH)]}+ adduct (Keq = 0.0041). Calculations that reproduce the experimental findings reveal that oxidation of either the mononuclear or dinuclear species results in a common dinuclear intermediate, {[LCu(III)]2-(μ-O)2}2+, which avoids formation of terminal Cu(IV)O/Cu(III)−O• intermediates. Calculations further reveal that both intermolecular water nucleophilic attack and redox isomerization of {[LCu(III)]2-(μ-O)2}2+ are energetically accessible pathways for O−O bond formation. The consequences of these findings are discussed in relation to differences in water oxidation pathways between Cu catalysts and catalysts based on Ru, Ir, and Mn. {[(bpy)2(OOH)Ru(IV)]2-(μ-O)-[(bpy)2(OH)Ru(IV)]}4+.11 This type of O−O bond forming mechanism is designated water nucleophilic attack (WNA, Figure 1).12 A similar complex

1. INTRODUCTION The thermodynamically challenging step in natural and artificial photosynthesis is the oxidation of water to dioxygen, 2H2O → O2 + 4H+ + 4e− (ΔE° = 1.23 V, ΔG = 113 kcal mol−1).1 Deployment of solar fuels derived from water at scale requires robust catalysts that evolve oxygen with high efficiency and are likely derived from earth abundant materials.2 In Nature, water oxidation occurs at the Mn4CaO5 active site of the oxygenevolving complex (OEC) of photosystem II. The mechanism of enzymatic O−O bond formation in the OEC remains yet unresolved and is believed to occur by either attack of water/ hydroxide on an electrophilic Mn(V)O adduct or by a μoxo−Mn(IV)-O• radical coupling mechanism.3,4 The latter mechanistic possibility is based exclusively on computational studies,5 while many synthetic homo- and heterogeneous systems have been developed that mechanistically parallel the former.6−9 For example, in the well-studied blue dimer of Meyer and co-workers,10 {[(bpy)2(H2O)Ru(III)]2-(μ-O)}4+ (bpy = 2,2′-bipyridine), the key step of catalysis is generation of the O−O bond, which occurs via attack of the Lewis base water on the high-valent intermediate {[(bpy)2(O)Ru(V)]2-(μO)} 4+ to generate the hydroperoxide intermediate © 2017 American Chemical Society

Figure 1. Three general mechanisms for creation of O−O bonds.

reported by Brudvig, Crabtree, and co-workers,13 {[(terpy)Mn(IV)(O)]-(μ-O)2-[(terpy)Mn(V)(O)]}+ (terpy = 2,6-bis(2pyridyl)pyridine), also oxidizes water via WNA.14 Given that in these examples the water attacks the terminal Ru(IV)O/ Mn(V)O site, rather than the bridging (μ-O) site, it was suggested that single-site catalysts would be sufficient for Received: April 1, 2017 Published: May 30, 2017 8586

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society

Figure 2. Selected homogeneous water oxidation catalysts based on copper. Complex 1: ref 40, and this work; complex 2: ref 39; complex 3: refs 28 and 34; complex 4: refs 35 and 36; complex 5: ref 42; complex 6: refs 43 and 44; complex 7: ref 45; complex 8: ref 46; complex 9: refs 47 and 46; complex 10: ref 48; complex 11: ref 49; complex 12: ref 50; complex 13: ref 51; complex 14: ref 52; complex 15: ref 53; complex 16: ref 54; complex 17: ref 55; complex 18: ref 56.

catalysis to proceed.15 Indeed, many single-site Ru,16,17 Mn,18,19 and Ir20,21 catalysts have since been developed which proceed by analogous WNA mechanisms. An alternative mechanism based on radical coupling of two high-valent metal-oxyl moieties has also been proposed (I2M, Figure 1), but examples of this mechanism are rare. For example, an extremely fast single-site Ru catalyst has been proposed to undergo I2M largely due to the fast rate of the reaction, which depletes OH− at the electrode, opening up this alternate pathway to generating an O−O bond.22 Recently, late first-row transition metal water oxidation catalysts have emerged as potential earth abundant species for practical applications. Electrocatalysts based on Fe, Co, Ni, and Cu have been identified,23 as have catalysts that function under chemical oxidation.24−27 Of particular interest in this report are homogeneous copper catalysts (Figure 2), which have seen remarkable growth since the initial report by Mayer and coworkers.28 Mayer’s initial report draws motivation29 from an observation made earlier by Tolman and co-workers,30 wherein reversible O−O bond formation with a dicopper peroxide complex, {[(tacn)Cu(II)]2-(μ-η2:η2-O2)}2+, was demonstrated to be in equilibrium with a complex wherein the O−O bond is broken, {[(tacn)Cu(III)]2-(μ-O)2}2+. Thus, with copper, an alternative O−O bond forming mechanism exists that is designated redox isomerization (RI, Figure 1).31 The Mayer catalyst, {[(bpy)Cu(II)]2-(μ-OH)2}2+ (Figure 2, complex 3), was found to oxidize water with a fast turnover frequency (TOF, 100 s−1) at a 750 mV overpotential at pH 12.5. The authors noted in keeping with historical observations32,33 that they observed electron paramagnetic resonance (EPR) signals consistent with dissociation of the catalyst into species with noninteger spin, which was interpreted as formation of the mononuclear species [(bpy)Cu(OH)2] at high pH. The linearity of the current as a function of copper concentration was interpreted as a rate-determining step that involves a single

copper and thus catalysis proceeding via [(bpy)Cu(OH)2]; however, over the narrow concentration window it might be expected that catalysis proceeding via the dinuclear or higher nuclearity complexes would be indistinguishable. The situation is potentially more complex; at neutral pH, Zheng and coworkers reported a tetranuclear variant, [(bpy)4Cu4(μ2-OH)2(μ3-OH)2(H2O)2]2+ based on crystallography and mass spectrometry, which had high catalytic activity and stability.34 In an attempt to generate more efficient catalysts, Lin and coworkers generated ligands with substituents at the 6,6′positions of bpy.35 The 6,6′-dihydroxy-2,2′-bipyridine complex (Figure 2, complex 4) demonstrated water oxidation at lower overpotential than the bpy parent, due to hydrogen bonding of the hydroxyl substituent to the bound hydroxo/aquo ligands. The influence of hydrogen bonding and nuclearity of the complexes on water oxidation was further expanded by Papish and co-workers.36 Electronic structure calculations of the proposed mechanism based on the assumption of a mononuclear catalyst revealed net hydrogen atom abstraction from the ligand, rather than the copper-bound hydroxide. This is due to involvement of a redox-active molecular orbital on the ligand, which generates an electrophilic Cu(III)−OH species that can undergo WNA to generate a bound hydroperoxide.35 Higher level calculations performed by de Ruiter and Buda on the parent bpy system invoke a terminal Cu(III)O moiety upon oxidation of the mononuclear [(bpy)Cu(OH)2], which is consistent with the lower level calculations on 6,6′-dihydroxy2,2′-bipyridine given that the −OH substituents are not available on the ligand.37 It is noteworthy that the crystal structure of 6,6′-dihydroxy-2,2′-bipyridine copper hydroxide by Lin and co-workers reveals a 1D polymer,35 and an earlier report by Papish and co-workers revealed that this ligand can support higher nuclearity copper structures,38 which suggests along with the Zheng tetranuclear variant34 the possibility of 8587

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society

reported reduced dioxygen species would allow for anchoring of possible reaction trajectories with independently accessible intermediates and thereby allow for validation of the computational methods. Thus, we screened a variety of copper complexes with scaffolds derived from tmpa (tmpa = tris(2pyridylmethyl)amine). This strategy was fruitful, yielding {[(Me2TMPA)Cu(II)]2(μ-OH)2}2+ (Me2TMPA = bis((6methyl-2-pyridyl)methyl)(2-pyridylmethyl)amine, Figure 3) as

higher nuclearity species in bpy analogues. This concern was noted by Mayer and co-workers.29 Nonetheless, mechanistic proposals for copper that involve intermediacy of high-valent terminal Cu-oxo/oxyl species are appealing because (i) the mononuclear species dominate in solution and (ii) mechanistic parallels to the well-defined water oxidation mechanisms of single-site Ru and Ir complexes described above. For example, Meyer and co-workers recently reported a tripeptide-ligated copper catalyst (Figure 2, complex 5) that on the basis of observation of two electrochemical oxidation waves in cyclic voltammetry, both of which are pH dependent, proposed a mechanism for rate-limiting O−O bond formation by reaction of Cu(IV)O (Cu(III)−O•) with water.41,42 Similar arguments for a heavily modified version of the polypeptide scaffold (Figure 2, complex 12)50 were made based on analogous electrochemistry and observation of a small kinetic isotope effect (KIE) of 2, which was interpreted as a potential concerted O atom−proton transfer (APT) pathway.57 The electrochemistry of several other copper-based complexes has been interpreted analogously, with the common themes of intermediacy of high-valent terminal Cu−O moieties and ratelimiting O−O bond formation via APT pathways.43,47,49,52 While the mechanistic parallels to Ru and Ir are enticing, it is worth noting that while complexes of Ru and Ir with terminal oxo groups are readily available, no terminal copper oxo/oxyl has ever been observed.58,59 Moreover, late transition metal oxo species have historically been disfavored due to filled antibonding orbitals in metals with high d electron counts,60 which may suggest alternate pathways could be at work in copper-based water oxidation. This possibility has received some recent attention. In particular, a few examples based on intramolecular water nucleophilic attack pathways have recently been predicted. These mechanisms are reminiscent of a proposal from Milstein and co-workers on photoinduced oxygen evolution from a single-site Ru dihydroxo species.61 For example, Maseras, Llobet, and co-workers reported a single-site copper catalyst (Figure 2, complex 13) with a redox-active ligand, wherein calculations predict initial oxidation to [(L•)Cu(III)(OH)], followed by weak coordination of an additional hydroxide to the copper, leading to an unusual single-electron transfer WNA, which extrudes dioxygen.51 The redox-active 6,6′-dihydroxy2,2′-bipyridine complex (Figure 2, complex 4) discussed above could also be envisioned to undergo a similar extrusion of dioxygen.35 Both of these complexes with redox-active ligands potentially avoid the problematic Cu(IV)O/Cu(III)−O• intermediates by enabling highly electrophilic −OH moieties without the need to access Cu(IV). As an alternative to avoiding the problematic terminal oxo functionality, yet another strategy has been reported by Liao, Zhang, and co-workers (Figure 2, complex 18),56 in which calculations predict a terminal copper-bound hydroxo, which performs WNA on a Cu2(μ-O) species. It is important to note that despite reactivity patterns that suggest Cu2(μ-O) would behave as an electrophile,62 to our knowledge, the reaction of Cu2(μ-O) with water to generate an O−O bond has not been demonstrated. These issues prompted us to consider if (i) putative singlesite copper catalysts could catalyze water oxidation more generally by accessing dinuclear intermediates and thus avoid Cu(IV)O/Cu(III)−O• intermediates, and if so (ii) would redox isomerization or intermolecular WNA O−O bondforming mechanisms be favored. We reasoned that identification of a water oxidation catalyst that had previously

Figure 3. Copper catalyst under investigation in this report, {[(Me2TMPA)Cu(II)]2(μ-OH)2}2+.

a water oxidation catalyst, which is the subject of this report.63 Characterization of this species by spectroscopy and electrochemistry revealed that it too dissociates in aqueous solution, but that the dinuclear species must be the active catalyst based on the involvement of a proton in the rate-determining step. Computational evaluations of accessible trajectories from both mononuclear and dinuclear species were instructive. The calculations revealed that oxidations of both mononuclear and dinuclear complexes are competitive, but result in common dinuclear intermediates rather than high-valent mononuclear intermediates. The calculations also revealed that Cu(IV)O/ Cu(III)−O• intermediates are highly disfavored due to their basicity and availability of dimerization routes and that intermolecular WNA is competitive with redox isomerization for O−O bond formation. These experimentally calibrated calculations have major implications for the design of copperbased water oxidation catalysts and suggest that the mechanism of copper-based single-site catalysts should be re-evaluated in light of the potential for forming dinuclear intermediates.

2. EXPERIMENTAL SECTION 2.1. General Methods. All reagents and solvents were purchased from commercial sources and used without further purification unless stated otherwise. Water (18 MΩ·cm) was obtained using a Milli-Q Ultrapure water purification system. Electronic absorption spectra were collected on an Agilent 8453a spectrometer equipped with a Unisoko Cool-Spec cryostat. Continuous-wave electron paramagnetic resonance spectra were collected on a Bruker EMX-Plus spectrometer with a TM4102 resonator equipped with a syringe pump. Spectra were digitally filtered in Origin (OriginLab, Northampton, MA, USA) postcollection using a low-pass Fourier filter with a cutoff frequency of 0.03 Hz. Mass spectra were collected on a Waters TQD Acquity UPLC with an Acquity H Class UPLC. 1H NMR spectra were collected on a Varian Inova 400 MHz spectrometer. Tetramethylsilane was used as an internal standard. 2.2. Synthesis of {[(Me2TMPA)Cu(II)]2(μ-OH)2}(OTf)2 (1). Me2TMPA was prepared analogously to that reported previously64 by condensation of 2-picolylamine (1.6 mL, 15 mmol) with 6-methyl2-pyridinecarboxaldehyde (3.91 g, 32 mmol) in THF followed by reduction with sodium triacetoxyborohydride (9.9 g, 47 mmol). Complex 1 was prepared by a method analogous to that reported by Suzuki and co-workers for the perchlorate salt.65 Specifically, triethylamine (0.46 mL, 3.3 mmol) in THF (5 mL) was added dropwise to a solution of copper(II) triflate (0.2 g, 0.55 mmol) and 8588

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society Me2TMPA (0.19 g, 0.61 mmol) in 15 mL of THF under nitrogen. Upon addition, the solution turned from green to light blue and a light blue precipitate was formed that was collected by filtration (0.24 g, 80%). Complex 1 was recrystallized by slow diffusion of diethyl ether into acetonitrile to yield blue plates suitable for X-ray crystallography. UV−vis (acetonitrile): λmax [nm] (ε [M−1 cm−1]) = 260 (18 500), 306 (3500), 655 nm (90);. IR (KBr): ν(OH) [cm−1] = 3610 (m, sharp). 2.3. Electrochemistry. Unless otherwise noted, electrochemistry was performed on a CH Instruments 720E potentiostat equipped with a Pine Research Instrumentation MSR rotator. All electrochemical experiments were performed at room temperature. Electrochemistry of acetonitrile solutions contained 0.1 M tetrabutylammonium hexafluorophosphate (TBAPF6) recrystallized from ethanol as supporting electrolyte and employed a nonaqueous Ag/Ag+ reference electrode (0.01 M AgNO3) referenced externally to ferrocene (0.624 V vs NHE).66 Aqueous solutions contained 0.1 M sodium trifluoromethanesulfonate (NaOTf) as supporting electrolyte and NaOH were used to adjust pH. Deuterium isotope experiments were performed in deuterium oxide (D2O, 99.9%) with pD adjusted with NaOD.67 Aqueous electrochemistry employed a commercial (CH Instruments) Hg/HgO (1 N NaOH) electrode (0.140 V vs NHE) externally referenced to a saturated calomel electrode (0.241 V vs NHE). All experiments employed a platinum mesh counter electrode. Cyclic voltammetry (CV) experiments were performed with a threeelectrode system in a one- or two-compartment electrochemical cell. The reference and counter were as described above. In normal CV experiments, a 3.00 mm (0.0707 cm2) geometric-diameter glassy carbon (GC) working electrode (BASi) was used. Rotating disc electrode CVs (RDE-CVs) used a 5.00 mm (0.196 cm2) geometricdiameter GC electrode (Pine Research Instruments). GC working electrodes were polished with a 0.05 mm Al2O3 slurry, sonicated for 3 min in ethanol, then again in ultrapure water, and then rinsed prior to each sample. Solutions were purged with nitrogen for 10 min, and a stream of nitrogen was maintained over solutions for the duration of the experiments. Rotating ring-disk electrode (RRDE) voltammetry employed a fourelectrode setup with a Digi Ivy 2023 bipotentiostat and a Pine E6 series ring-disk working electrode containing a 5.00 mm (0.196 cm2) geometric-diameter GC and 6.50 mm i.d., 7.50 mm o.d. (0.110 cm2) platinum ring (Pine Research Instruments). The electrode collection efficiency (N) was determined by amperometry in 5.0 mM K3Fe(CN)6 and 0.1 M KCl. A potential of 0.566 V vs NHE was applied to the disk and 1.02 V vs NHE on the ring, and the rotation rate was 1600 rpm. The collection efficiency was calculated using

N=

energies, analytical frequency calculations were performed, and no imaginary frequencies were found. Solvated single-point energies were determined using the TPSSh functional,72 the same basis sets as for geometry optimizations, and the PCM solvation model with the solvent parameters of water, unless otherwise noted. The Gibbs free energy of each species was determined by adding the solvated singlepoint SCF energy to the thermal correction from the respective frequency calculation. The free energy of the water-solvated proton was corrected from the value at pH = 0 to the working experimental pH (typically 12.5 unless otherwise noted) using the expression given by Winikoff and Cramer:73 ° ′ ΔGproton = ΔGproton − C(pH) where ΔG′proton is the corrected value, ΔG°proton is the consensus value of −265.9 kcal mol−1,74 and C is 1.23 kcal mol−1. Since determination of the free energy of a hydroxide anion is a challenge to describe computationally, in calculations where a hydroxide anion is a reactant, water was instead substituted as the reactant and a solvated proton added as a product to provide a net OH− reactant. This serves to cancel out the error in the hydroxide ΔG. Calculated potentials are reported relative to the normal hydrogen electrode (NHE) at pH 12.5 using the following expression:

ΔGrel = ΔGcalc − nF(ΔEcalc) where ΔGrel is the free energy relative to NHE, ΔGcalc is the DFTderived ΔGproducts − ΔGreactants, n is the number of electrons transferred, F is Faraday’s constant (23.06 kcal mol−1 V−1), and ΔEcalc is 4.124 V, which was determined by correction of the absolute potential of NHE (4.54 V).75 This correction was accomplished by reducing the experimental NHE value by the difference between the experimental (1.145 V) and DFT-calculated (0.729 V) reduction potential of LCuOH2 (L = N,N′-bis(2,6-diisopropylphenyl)-2,6pyridinedicarboxamide) experimentally measured by Dhar and Tolman.76

3. RESULTS AND ANALYSIS 3.1. Characterization of 1. The dicopper dihydroxide complex 1 was prepared by metalation of Me2TMPA and was characterized by spectroscopy and crystallography. The cationic fragment of 1, {[(Me2TMPA)Cu(II)]2(μ-OH)2}2+ (Figure S1 and Table S1), was reported previously by Suzuki as the perchlorate salt65 and is analogous in geometric structure with minor differences in metrical parameters (Figure S2 and Table S2). The copper ions and hydroxide anions are coplanar, with average Cu−O(H) distances of 1.921 and 1.945 Å and Cu···Cu 2.936 Å, with the triflate anions hydrogen bonding to the hydroxide protons (Figure S2). The coordination environment about each copper is square planar with the apical nitrogen and the unsubstituted pyridine arm in the plane of the copper and hydroxides (1.998 and 2.068 Å, respectively). The methylpyridyl arms are weakly interacting with the copper in the axial positions (2.259 and 2.303 Å). The optical spectrum of 1 in acetonitrile shows an intense band at 260 nm (18 500 M−1 cm−1), a shoulder at 306 nm (3500 M−1 cm−1), and a weak band at 655 nm (90 M−1 cm−1), which are assigned on the basis of energy and intensity as ligand π → π*, OH → Cu LMCT, and Cu d → d, respectively (Figure S3). Electrospray mass spectrometry revealed a mass envelop at a mass-to-charge ratio (m/z−1) of 945.7, which is consistent with {[(Me2TMPA)Cu]2(μ-OH)2}(OTf)+ (945.1856 g mol−1), which also matches the expected isotope distribution (Figure S4). The 1H NMR spectrum of 1 in acetonitrile displayed paramagnetically shifted resonances (Figure S5). 1 is EPR silent at 77 K in acetonitrile and taken together with the 1H NMR suggests an overall integer spin system. Since the copper ions are in the +2 oxidation state, and thus d9, an integer spin state

Ir Id

where Ir is the ring current, Id is the disk current, and N is the collection efficiency of the electrode, which was determined to be 0.22 ± 0.03 (22 ± 3%). Controlled potential electrolysis (CPE) experiments were performed in a two-compartment, glass frit separated cell with a 3.00 mm (0.0707 cm2) geometric-diameter GC working electrode (CH Instruments). The other compartment contained the counter and reference electrodes. 2.4. Calculations. Density functional theory (DFT) calculations were performed with the electronic structure package Gaussian09 revision D.01.68 Mulliken populations and Mayer bond orders were computed from the Gaussian09 output with QMForge,69 and molecular orbitals were visualized using Lumo.70 The model of 1 was built based on the molecular structure of the isolated dication from crystallographic data; other species were built from this model as a starting point and were optimized using the hybrid functional B3LYP71 within the unrestricted formalism. The basis sets on Cu, O, and N atoms were of triple-ξ quality (6-311G*) and a double-ξ quality (6-31G) basis set on C and H atoms. Models were optimized using tight convergence criteria and ultrafine integration grids as implemented in Gaussian09. To ensure stationary points on the potential energy surfaces were obtained and to provide zero-point 8589

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society

Figure 4. Lowest unoccupied molecular orbitals of 1 that display opposite signed unpaired spin density in orbitals of dx2−y2 character, consistent with an antiferromagnetically coupled singlet model (right).

suggests that 1 is a dinuclear adduct in acetonitrile. To further explore the electronic structure of 1 in solution, variabletemperature (VT)-NMR was collected in acetone for a wider temperature window (Figure S7). The temperature dependence of the four most paramagnetically shifted peaks were evaluated within the Heisenberg−Dirac−Van Vleck effective spin Hamiltonian / = −2JS1·S2 where S1 and S2 are the singlet (antiferromagnetically coupled) and triplet (ferromagnetically coupled) states, respectively, by applying a Boltzmann population to the Curie law as follows:77 δobs = δdia +

Cδ ⎡ e−2J / kT ⎤ ⎢ ⎥ T ⎣ 1 + 3e−2J / kT ⎦ Figure 5. RDE-CVs were performed at 100 mVs−1 using a 5 mm GC electrode at 2000 rpm in pH 12.5 NaOTf solution (black) and 1 mM 1 in NaOTf solution (blue). (Inset) RDE CVs at 2000 rpm were performed at 100 mV s−1 using a 5 mm GC electrode in pH 7 phosphate buffer (black) and 1 mM ferrocene carboxylic acid in phosphate buffer (red).

where δobs is the observed chemical shift, δdia is the diamagnetic contribution to the chemical shift, Cδ is a collection of terms that are empirically fit, T is the temperature, k is the Boltzmann constant, and −2J is the energy separation between the two energy states S1 and S2 (Figure S8). Within this formalism, −2J was found to be 66 cm−1, and thus 1 has an antiferromagnetically coupled ground state with a low-lying thermally accessible paramagnetic excited state. The magnitude of the coupling is consistent with other Cu2(μ-OH)2 adducts in the literature that are well-known to be dependent on the Cu−O−Cu angle (Figure S9).78 CV of 1 in acetonitrile revealed an irreversible oxidative wave at 2.02 V (Figure S10). Density functional theory calculations on 1 were performed and reproduced the crystallographically derived geometric structure and spectroscopically deduced electronic structure. Specifically, DFT predicts a square-planar coordination environment about copper, with weakly associated methylpyridyl ligand arms at 2.578 Å. Complex 1 is well described as a broken-symmetry (BS) singlet (ST = 0), which, after spin purification to the true singlet, is calculated to be 0.26 kcal mol−1 (91 cm−1) lower in energy than the triplet, consistent with the paramagnetism observed in the 1H NMR spectrum. Opposite signed unpaired spin density is found on the Cu(II) ions in molecular orbitals of dx2−y2 character, which leads to the overall antiferromagnetically coupled ground-state description (Figure 4). The electronic pathway that mediates the superexchange interaction is through the in-plane O p orbitals, consistent with the dependence of the magnitude of the magnetic exchange on the Cu−O−Cu angle. 3.2. Electrocatalytic Water Oxidation by 1. A large, irreversible current is observed in the RDE-CV of 1.0 mM solutions of 1 with an onset potential of ∼1.2 V at a scan rate of 100 mV s−1 in 0.1 M NaOTf at pH 12.5 (Figure 5). The current is in excess of what would be expected for a oneelectron process, determined by comparison to RDE-CV of a 1.0 mM ferrocene monocarboxylic acid solution (Figure 5,

inset). The current flow is linearly dependent on the concentration of 1 up to ∼0.6 mM (Figure 6), after which

Figure 6. Concentration dependence by RDE-CVs at 100 mV s−1 on 0 to 1 mM 1 using a 5 mm GC electrode at 2000 rpm in pH 12.5 NaOTf solution. Inset: Concentration dependence of current density at 1.73 V.

depletion of hydroxide anion at the electrode has been implicated to limit the rate of catalysis.22 These observations suggest 1 is competent for catalytic water oxidation. Under identical conditions for the above RDE-CV, the catalytic activity of a previously reported catalyst, {[(bpy)Cu]2-(μOH)2}2+, was slightly better, with a lower onset potential (∼1.0 V) and larger current densities at equivalent potentials compared to 1. Comparable catalytic responses by CV were 8590

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society

12.5 the thermodynamic potential for water oxidation is 0.493 V; thus this rate is at an overpotential of 1.007 V). For comparison, Mayer and co-workers reported the kcat for {[(bpy)Cu]2-(μ-OH)2}2+ was 100 s−1 at 0.750 V overpotential.28 The rate of water oxidation by solutions of 50 μM 1 as determined by the current flow using RDE-CV at 1.5 V is dependent on the pH of the solution (Figure 7). The pH

observed with indium tin oxide and boron-doped diamond electrodes. 3.2.1. Dioxygen Production. Upon passing current through aqueous solutions containing 1 above 1.2 V vs NHE, bubbles are formed at the working electrode (Figure S12 inset). The composition of these bubbles was determined to be dioxygen by several methods as follows. First, after sweeping through the catalytic potential window of the catalyst at oxidizing potentials, the potential was then swept cathodically where an irreversible peak was observed at −0.11 V, which is due to accessing the O2/O2− couple (Figure S12). This peak disappears upon sparging the solution with dinitrogen and is only observed when sweeps at high potentials precede cathodic scans. This peak is also observed after bubbling dioxygen into the background solution. Second, controlled potential electrolysis at 1.62 V vs NHE was performed in a sealed vessel connected to a flask of alkaline pyrogallol solution (Figure S13). The dioxygen produced by electrolysis was trapped by the pyrogallol and then quantified spectrophotometrically (Table S3). The oxygen produced in the presence of catalyst was 2.6 times higher than that produced by electrolysis in the absence of catalyst, consistent with amperometry experiments (see Section 3.2.3). Finally, to further investigate the efficiency of oxygen production, RRDE voltammetry was performed. Initial CV experiments were performed to determine potentials for ring and disk. The disk was held at a potential of 1.54 V vs NHE, which results in water being oxidized by 1 to dioxygen (2H2O → O2 + 4H+ + 4e−), and the ring was held at −0.309 V vs NHE, which reduces the dioxygen produced to hydrogen peroxide (O2 + 2H+ + 2e− → H2O2).79 Faradaic efficiency is calculated according to

ε=

Figure 7. RDE CVs of 50 μM 1 at 100 mV s−1 using a 5 mm GC electrode at 2000 rpm in pH 7.38 to 13.65 NaOTf solution. Inset: pH dependence of current density at 1.5 V.

dependence is nonlinear, with essentially no current flow from pH 7.5 to pH 11, with a marked increase in current above pH 11. Only minor intensity changes were observed in the UV−vis of 1 between 7.5 and 13 as monitored in the d → d bands at 685 and 838 nm with decreases observed above pH 13 (Figure S20). To assess the possibility of proton involvement in the rate-determining step of catalysis, the isotope dependence of catalysis by 1 was explored by RDE-CV in D2O (Figure 8). Solutions of 1 in pD 12.5 D2O show significantly attenuated current; at 1.5 V vs NHE 0.0537 mA (0.273 mA cm−2) was observed in the deuterated solution, whereas 0.316 mA (1.61 mA cm−2) was observed in the proteo solution. The kinetic isotope effect was calculated at 1.5 V according to81

2Ir NId

where ε is the Faradaic efficiency, Ir is the ring current, Id is the disk current, and N is 22% at 1600 rpm, as determined in Section 2.3. The factor of 2 in this equation comes from the balance of the two relevant half-reactions. Under these conditions, the current density observed was 2.7 ± 0.2 mA cm−2 at the disk (Id) and 0.58 ± 0.1 mA cm−2 at the ring (Ir), which equates to a Faradaic efficiency of 112 ± 23%.80 The Faradaic efficiency of {[(bpy)Cu]2-(μ-OH)2}2+ was reported as 90% and is the same within error; however, it should be noted that oxygen evolution was measured by an alternate method.28 Attempts to re-evaluate {[(bpy)Cu]2-(μ-OH)2}2+ using the RRDE method described for 1 resulted in decomposition by rapid copper deposition on the platinum ring. 3.2.2. Rate of Water Oxidation. Catalytic water oxidation by 1 appears to proceed at more modest rates than {[(bpy)Cu]2(μ-OH)2}2+ given the more anodic onset of catalytic current. Nonetheless, the near unity Faradaic efficiency allows for an estimate of the apparent rate constant (kcat) in the kinetic regime assuming an EC mechanism at low catalyst concentrations according to the following:

KIE =

kcat,H2O kcat,D2O

⎛ icat,H2O ⎞2 ⎟⎟ = ⎜⎜ ⎝ icat,D2O ⎠

where the ratios of the apparent rate constants (kcat) are proportional to the catalytic current (icat) ratio squared, which yielded a KIE of 20, which is significantly larger than the KIE reported for water oxidation by [(TDImP)Co] (TDImP = 5,10,15,20-tetrakis(1,3-dimethylimidazolium-2-yl)porphyrin), which was 2.8.36 EPR experiments of the D2O solution of 1 revealed partial dissociation of 1 to the extent observed in H2O solutions (see Section 3.3), and thus, the sizable primary KIE for 1 suggests proton involvement in the rate-determining step of catalysis. 3.2.3. Stability and Phase under Electrocatalytic Conditions. Motivated by recent reports that demonstrate that bare copper salts82,83 and electrodeposited CuO84,85 are effective water oxidation catalysts, experiments were performed to determine the lifetime and nature of the catalytically active species to assess these possibilities. Of particular concern was a report from Du and co-workers that CuO electrodeposited from [(tmpa)Cu(H2O)](ClO4)2 is an active catalyst.86 Initial rinse tests were performed by RDE-CV with a GC electrode at

⎛ ⎞ ic nc 1 = ⎜⎜0.359 3/2 kcat ⎟⎟ ip np ⎝ ⎠ ν

where ic is the catalytic current, ip the noncatalytic current, nc the number of electrons transferred in the catalytic process, np for a noncatalytic process, and ν the scan rate. The plot of ic/ip versus ν−1/2 (Figure S14) yields a kcat of 33 s−1 at 1.5 V (at pH 8591

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society

example, 50 μM solutions of 1 yield a peak at 1.61 V (3 mV s−1, 1.42 mA cm−2). The position of this peak is scan rate dependent, with increasing scan rate increasing peak potential; at a scan rate of 4000 mV s−1, the peak of the 50 μM 1 solution shifted to 2.03 V (Figure 10). The current at the peak was

Figure 10. Scan rate dependence of 50 μM 1 by CV at 3 (light gray), 5 (black), 10 (gray), 20 (red), 50 (orange), 100 (light green), 200 (green), 500 (light blue), 1000 (blue), 2000 (dark blue), and 4000 (purple) mV s−1 using a 3 mm GC electrode in pH 12.5 NaOTf solution. Inset: Scan rate dependence of the catalytic current for each peak potential versus the square root of scan rate for 50 μM 1 by CV.

linearly dependent on the square root of the scan rate, which is indicative that the electrochemical process is diffusion limited and thus homogeneous.87 A heterogeneous process, such as catalyst adsorbed to the electrode surface, would result in a linear relationship between current and scan rate, which is not what is observed (Figure S19). We also attempted to measure solutions of copper triflate in the absence of ligand; however, the copper was not soluble under the conditions employed, and no precipitates formed during electrochemical experiments of 1, which excludes this possibility. Further evidence comes from amperometry of 1 under identical conditions at a potential of 1.35 V (Figure 11), which displayed immediate onset of a stable current (1.60 mA cm−2, 0.94 mA cm−2 above background), indicative that no chemical process must occur to achieve catalysis. Controlled potential electrolysis was performed under significantly more forcing conditions at 1.79 V, which yielded an average current density

Figure 8. Isotope effect by RDE CVs at 100 mV s−1 with (blue) and without (black) 0.5 mM 1 using a 5 mm GC electrode at 2000 rpm in (top) protonated pH 12.5 and (bottom) deuterated pD 12.5 NaOTf solution. The KIE was calculated at 1.5 V to be 20.

1600 rpm (Figure 9), cycling first in the pH 12.5 0.1 M NaOTf background solution, then 25 cycles at 20 mV s−1 in the

Figure 9. RDE CVs at 20 mV s−1 using a 5 mm GC electrode at 1600 rpm in 0.1 mM NaOTf pH 12.5 background solution (dark blue), then 25 scans with 1 mM 1 (light blue), followed by rinsing the electrode and replacement into fresh background solution (red).

presence of 1.0 mM 1. The electrode was removed from the solution and visually inspected, which revealed no deposition of material. The electrode was briefly rinsed with buffer and recycled in background solution, the voltammogram of which revealed no increase in current over the initial background voltammogram, which would be expected if catalytically active material deposited on the electrode surface. At low concentrations, an oxidative peak can be resolved by CV, which is not present in solutions in the absence of 1; for

Figure 11. Amperometry using a 3 mm GC electrode at 1.35 V over ∼5 h in a 0.1 mM NaOTf pH 12.5 solution (black) and in the presence of 1 mM 1 (red). 8592

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society

Figure 12. Left: Concentration-dependent EPR spectrum of 1 with [Cu] of 0.4 mM (red), 0.8 mM (orange), 1.2 mM (green), 1.6 mM (blue), 2.0 mM (purple) in electrolyte buffer at pH 12.5. Right: EPR intensity as a function of total copper concentration and fit to Keq = [CuOH]2/ [Cu2(OH)2] = 0.0041 (---).

Figure 13. Left: Speciation of 1 upon dissolution into aqueous solutions of pH 12.5 as determined by DFT. The numbers over the arrows represent the ΔG of the transformation and L = Me2TMPA. Right: Relative concentrations of 1 (dark blue), [(Me2TMPA)CuOH]+ (light blue), and [(Me2TMPA)CuOH2]2+ (red) as a function of pH with ΔG = 3.2 kcal mol−1 and as a function of ΔG of dissociation at pH 12.5 (inset).

of 3.23 mA cm−2 compared to a background of 1.95 mA cm−2, which gives a background-subtracted current density of 1.28 mA cm−2. Under these forcing conditions, the current was stable for 31/2 h, after which the current began to slowly decrease for the next 41/2 h. Over this 8 h period, which corresponds to a turnover number (TON) of 1.43, the solution pH dropped from 12.41 to 9.53 and the solution changed color from a light blue to green, quantified as a loss of the 838 nm electronic absorption band in the optical absorption spectrum (Figure S18). No coating on the electrode was observed over the 8 h experiment, and the CV wave shape changed to match that of the background after 8 h of CPE. Together, these observations suggest that 1 functions as a homogeneous electrocatalyst with poor stability at ∼1.3 V overpotential. This turnover number is lower than that reported for [(F3TMPA)Cu](ClO4)2 (F3TMPA = tris(2-fluoro-6-pyridylmethyl)amine) (TON ≈ 5) at lower overpotential,39 and thus it is anticipated that the TON of 1.43 is actually a lower limit due to the forcing conditions required as a result of moderate turnover rate of 1, which degrades the catalyst. 3.3. Aqueous Speciation of 1. Mayer and co-workers noted that in the water oxidation precatalyst {[(bpy)Cu]2(OH)2}(OAc)2 (OAc = acetate anion), dissolution into aqueous electrolyte solutions led to pH-dependent dissociation of the dinuclear species into mononuclear fragments that were assigned as [(bpy)Cu(OH2)2]2+ at low pH or [(bpy)Cu(OH)2] at high pH.28 Although 1 was found to be a dimer in organic solvents, we sought to assess if dissociation could occur with 1 in aqueous solutions. Indeed, EPR experiments of 1 at 2 mM total copper concentration at room temperature under flow conditions revealed an EPR signal (Figure 12). This signal was quantified by double integration to yield a concentration of

paramagnetic copper of 1.2 mM. As the dimer is diamagnetic and thus EPR silent, the signal must originate from a paramagnetic product, which is interpreted as formation of a mononuclear species. Assuming a two-species equilibrium between mononuclear and dinuclear species, the spin quantitation yields a Keq of 0.0041, which indicates the dinuclear species being favored over the mononuclear species by ΔG = 3.2 kcal mol−1. Electronic structure calculations were performed to further explore this equilibrium and the nature of the mononuclear species. Of the possible species present in pH 12.5 solution, only 1 and the LCuOH species (L = Me2TMPA) are predicted by DFT to be present to any significant degree. Addition of another hydroxo or aquo ligand to 1 is unfavorable by 32.2 and 7.5 kcal mol−1, respectively (Figure 13). These structures differ in that the hydroxo groups all bridge in the former, while in the latter the water coordinates one copper ion, resulting in one methylpyridyl arm of each ligand dissociating from copper and stabilizing the water via hydrogen bonding (Figure 13). With respect to the mononuclear species, LCuOH2, LCu(OH2)2, LCu(OH2)(OH), and LCu(OH)2 were evaluated and found to be 5.1, 13.5, 8.4, and 24.7 kcal mol−1 uphill from LCuOH. This suggests that the EPR signal observed experimentally is resultant from LCuOH, as [LCu]2(μ-OH)2 is EPR silent. DFT predicts 1 to be favored over LCuOH by 1.2 kcal mol−1, in good agreement with the experimentally derived value of 3.2 kcal mol−1. Van Eldik and co-workers reported that [(tren)Cu(H2O)]2+,88 which bears some structural resemblance to [(Me2TMPA)Cu(H2O)]2+,89 has a pKa of 9.09. The experimental pKa of [(Me2TMPA)Cu(H2O)]2+ is not known, but is predicted by DFT to be 8.76, which is close to the value 8593

DOI: 10.1021/jacs.7b03278 J. Am. Chem. Soc. 2017, 139, 8586−8600

Article

Journal of the American Chemical Society reported for the tren complex above. Therefore, we computationally considered the relative concentrations of 1, LCuOH, and LCuOH2 (L = Me2TMPA) as a function of pH (Figure 13). At high pH, DFT predicts vanishingly small concentrations of LCuOH2, while at low pH LCuOH2 is favored over 1 and LCuOH by 10.7 and 11.9 kcal mol−1, respectively, which results in LCuOH2 being the dominant species in solution at low pH. It should be noted that the accuracy of these calculations depends on the energy of the solvated proton and the first water solvation shell in solution that is not explicitly treated, and thus some variance in the magnitude of these values is to be expected. Titrations monitored optically revealed only minor changes in the electronic absorption spectrum (Figure S20) as a function of pH up to pH 13, above which changes in the spectrum occur, implying stability of the catalyst above pH 13 may be an issue. EPR experiments over a pH range of 11 to 13 revealed only a slight change (