Triazenide-supported [Cu₄S] structural mimics of Cu_Z that mediate N₂O disproportionation rather than reduction

Abstract

As part of the nitrogen cycle, environmental nitrous oxide (N₂O) undergoes the N₂O reduction reaction (N₂ORR) catalyzed by nitrous oxide reductase, a metalloenzyme whose catalytic active site is a tetranuclear copper–sulfide cluster (Cu_Z). On the other hand, heterogeneous Cu catalysts on oxide supports are known to mediate decomposition of N₂O (deN₂O) by disproportionation. In this study, a Cu_Z model system supported by triazenide ligands is characterized by X-ray crystallography, NMR and EPR spectroscopies, and electronic structure calculations. Although the triazenide-ligated Cu₄(μ₄-S) clusters are closely related to previous formamidinate derivatives, which differ only in replacement of a remote N atom for a CH group, divergent reactivity with N₂O is observed. Whereas the formamidinate-ligated clusters were previously shown to mediate single-turnover N₂ORR, the triazenide-ligated clusters are found to mediate deN₂O, behavior that was previously unknown to natural or synthetic copper–sulfide clusters. The reaction pathway for deN₂O by this model system, including previously unidentified transition state models for N₂O activation in N–O cleavage and O–O coupling steps, are included. The divergent reactivity of these two related but subtly different systems point to key factors influencing behavior of Cu-based catalysts for N₂ORR (i.e., Cu_Z) and deN₂O (e.g., CuO/CeO₂).

---

Introduction

Anthropogenic nitrous oxide (N₂O) emissions have approximately 500 times the global warming potential of CO₂ per molecule¹ and represent the leading cause of ozone layer depletion.² Therefore, it is critical to understand the mechanisms by which both natural and synthetic catalysts convert N₂O to benign compounds.^3–5 In nature, N₂O is converted to N₂ according to eqn (1) as part of the nitrogen cycle.^6,7 This N₂O reduction reaction (N₂ORR) is catalyzed by nitrous oxide reductase (N₂OR),^8–10 a copper-dependent enzyme crucial to bacterial denitrification. The catalytic site of N₂OR is a copper–sulfide cluster known as Cu_Z,¹¹ which has consistently been found to contain a [Cu₄(μ₄-S)]ⁿ⁺ core in all its catalytically active forms.^12–15 Alternatively, heterogeneous catalysts are known to mediate the decomposition of N₂O (deN₂O) according to eqn (2). Although this disproportionation reaction is catalyzed most efficiently by Rh,¹⁶ explorations of earth-abundant alternatives have identified Cu-doped zeolites and CuO/CeO₂ as viable catalyst materials.^17–19

N₂O + 2H⁺ + 2e⁻ → N₂ + H₂O (1)

N₂O → N₂ + ½O₂ (2)

Our group has reported previous synthetic studies on structural^20–22 and functional^23–25 mimics of Cu_Z involving [Cu₄(μ₄-S)] cores supported by bridging diphosphine or formamidinate ligands.¹¹ One of these complexes, [Cu₄(μ₄-S)(NCN)₄]⁻ ([1^C]⁻), not only possesses an electronic structure similar to Cu_Z according to XAS analysis²⁴ but also was found to mediate a single turnover of N₂ORR, producing N₂ and O²⁻ along with [Cu₄(μ₄-S)(NCN)₄] (1^C) quantitatively (NCN = [MesN = CH-NMes]⁻, see Scheme 1).²³ In this reaction, it was proposed that one equivalent of [1^C]⁻ activates the N₂O substrate (mimicking Cu_Z in N₂OR) while another [1^C]⁻ equivalent acts as a sacrificial electron donor (mimicking the Cu_A electron transfer site in N₂OR). Based on computational modeling, μ-1,3-N₂O binding across a Cu–S edge of the cluster enabled by sulfide redox non-innocence was proposed,²⁴ although details of the N–O bond breaking pathway (e.g., the transition state structure) were not elucidated. This manuscript details the synthesis, characterization, and N₂O reactivity of an analogous [Cu₄(μ₄-S)(NNN)]ⁿ⁻ system (NNN = [MesN=N-NMes]⁻; n = 0, 1; see Scheme 1). Unlike its formamidinate analogue [1^C]⁻, the triazenide derivative [1^N]⁻ reacts with N₂O in a 1 : 1 stoichiometry upon reductive activation, mediating a single turnover of deN₂O rather than N₂ORR. Computational analysis of the deN₂O process includes transition state models for N–O cleavage and O–O coupling steps that were not identified in previous studies. The divergent selectivity of N₂ORR for [1^C]⁻vs. deN₂O for [1^N]⁻ points to key factors influencing behavior of Cu-based catalysts for N₂ORR (i.e., Cu_Z) and deN₂O (e.g., CuO/CeO₂).

Scheme 1 Divergent N₂O reactivity of Cu₄S model systems (Mes = 2,4,6-Me₃C₆H₂).

Results and discussion

Synthesis and characterization of the 1^N system somewhat parallel those of the 1^C system^21–24 but will be detailed here to facilitate important comparisons between complexes that are summarized in Table 1.

Table 1 Data comparison of triazenide- and formamidinate-ligated Cu₄S clusters

Property	Cu4S(NNN)4 (1^N)	[Cu4S(NNN)4]^− ([1^N]^−)	Cu4S(NCN)4 (1^C)^a	[Cu4S(NCN)4]^− ([1^C]^−)^b
a From Johnson et al.;^22 b From Johnson et al.^23 and Rathnayaka et al.;^24 c Generated by crystallography symmetry. d Mulliken populations for the LUMO and SOMO, respectively.
λ max (nm)	602, 768 (sh)	620, 934	470 (sh), 561	566
ε (M^−1 cm^−1)	1.2 × 10^4	4.1 × 10^3, 3.3 × 10^3	1.4 × 10^4	8.6 × 10^3
d (Cu⋯Cu) (Å)	2.405(3)	2.5853(16)	2.4226(6)	2.502(1)
	2.965(3)	2.6031(16)	3.0353(6)	2.809(1)
	2.405(3)^c	2.5853(16)^c	2.4226(6)^c	2.532(1)
	2.965(3)^c	2.6031(16)^c	3.0353(6)^c	2.831(2)
E° (V vs. [FeCp2]^+/0)	−0.85	irreversible	−1.28	irreversible
g tensors	n/a	2.143, 2.066, 2.005 (rhombic)	n/a	2.043, 2.090 (axial)
A(^63,65Cu) tensors (MHz)	n/a	≤122, 90, 100 (from H-strain)	n/a	15, 100
Redox-active MO^d	S 3p, 25%	S 3p, 26%	S 3p, 21%	S 3p, 20%
	Cu 3d, 11% each	Cu 3d, 12% each	Cu 3d, 13% each	Cu 3d, 14% each

---

Addition of a toluene solution of elemental sulfur to a bright yellow THF solution of dicopper(I) precursor, Cu₂(NNN)₂, immediately produced an inky blue solution from which [Cu₄(μ₄-S)(NNN)₄] (1^N) began to spontaneously crystallize. Complex 1^N is highly crystalline, showing only sparing solubility in THF, CH₂Cl₂, and CHCl₃ and no measurable solubility in CH₃CN, Et₂O, and toluene. Notably, 1^N is indefinitely stable on a benchtop in open air. The ¹H NMR of 1^N in CDCl₃ (Fig. S1†) is consistent with two inequivalent NNN environments (syn and anti to the μ₄-S ligand, respectively), each with restricted N–C bond rotation. A similar interpretation of the ¹H NMR spectrum of 1^C was attributed to intramolecular π-stacking interactions between neighboring mesityl groups evident by X-ray crystallography.²³ The UV-Vis-NIR spectrum of 1^N in CH₂Cl₂ (Fig. 1a) is dominated by an intense charge transfer transition at λ_max = 602 nm (ε = 1.2 × 10⁴ M⁻¹ cm⁻¹) with a shoulder at λ_max ≈ 768 nm. Another feature at higher energy (λ_max = 369 nm) is likely ligand-based, as it appears in all complexes examined, including previously reported Cu₂(NNN)₂.²⁶ For comparison, the analogous charge transfer transitions for 1^C, which is purple-colored, were observed at 561 (ε = 1.4 × 10⁴ M⁻¹ cm⁻¹) and ∼470 nm (shoulder).²²

Fig. 1 (a) Experimental and (b) computational data: UV-Vis-NIR of neutral (blue, CH₂Cl₂) and anionic (red, THF) forms of 1^N (left) and molecular structures of the neutral redox state of 1^N (right). The X-ray crystal structure is depicted as 50%-probability ellipsoids for non-C, H atoms and as wireframe for C atoms; H atoms are omitted; and only one component of the disordered Cu₄S core is shown. Computations were done at the B3LYP/def2TZVPP level of DFT using the CPCM implicit solvation model (CH₂Cl₂); computed UV-Vis-NIR spectra were generated using excitations calculated by TD-DFT and assuming 0.333 eV half-width at half height.

Complex 1^N crystallized with cubic symmetry in the R4̄3n space group, with the molecule residing on a crystallographic special position. The asymmetric unit (Fig. S3†) contains a single NNN ligand, a single Cu atom disordered over two positions, and a single S atom. Applying crystallographic symmetry elements reveals a cluster with four well-ordered NNN ligands related by a molecular S₄ axis, along with a disordered inorganic core with eight distinct Cu positions and two S positions (Fig. S3†). To assist with interpretation, a computational model of 1^N was built and optimized by DFT. (Here and throughout the study, computational models use methyl groups in place of mesityls.) The resulting structure (Fig. 1b) features a rectangle-based pyramidal Cu₄S core with Cu–Cu distances of 2.45 and 2.85 Å and Cu–S distances of 2.24 Å. The crystallographic model was then analyzed by grouping together four Cu atoms with one S position and the other four Cu atoms with the other S position to match the computational model as closely as possible. This analysis indicated a rectangular pyramidal Cu₄S core disordered over two positions, with experimental Cu–Cu distances of 2.405(3) and 2.965(3) Å and Cu–S distances of 2.176(2) Å (Fig. 1a). Grouping together the Cu and S atoms in alternative ways provided geometries dissimilar to the computational model, e.g., with unreasonably long Cu–S bonds. Additionally, this interpretation of the crystallographic data provides a model that is isostructural to 1^C, which was reported to have Cu–Cu distances of 2.4226(6) and 3.0353(6) Å within its rectangle-based pyramidal Cu₄S core.²²

The cyclic voltammogram of 1^N in ⁿBu₄NPF₆/THF (Fig. S6†) showed a reversible reduction at −0.85 V vs. FeCp₂^+/0 to the [1^N]⁻ state, and an irreversible reduction at lower potentials to the [1^N]²⁻ state that is unstable on this timescale. The redox behavior mirrors that of 1^C, which possesses a reversible reduction at −1.28 V and an irreversible reduction at lower potentials.²² Thus, the substitution of four formamidinate ligands for four triazenides shifts the redox potentials positively by approximately 0.4 V.

Chemical reduction of 1^N was carried out with either CoCp₂ (E°′ = −1.33 V vs. FeCp₂^+/0) or K[FeCp(CO)₂] (E°′ = −1.8 V vs. FeCp₂^+/0) in THF or CH₂Cl₂.²⁷ In each case, a subtle color change from inky blue to a duller blue was observed. While the [1^N][CoCp₂] salt was too insoluble for solution analysis, use of Kryptofix-222 to form [1^N][K(Krypt₂₂₂)] provided material suitable for characterization. As expected, only resonances for the cationic portion were observed by ¹H NMR (Fig. S7†) due to the S = 1/2 ground state of the [1^N]⁻ anion. The [1^N][K(Krypt₂₂₂)] salt is air-sensitive in solution, gradually converting to neutral 1^N when left on the benchtop. The UV-Vis-NIR spectrum of [1^N][K(Krypt₂₂₂)] in THF (Fig. 1a) has three main features, all of which have lower molar extinction coefficients than for corresponding peaks for 1^N. Charge transfer transitions at λ_max = 382 nm and λ_max = 620 nm (ε = 4.1 × 10³ M⁻¹ cm⁻¹) are minimally shifted in wavelength compared to 1^N. However, the feature for 1^N at λ_max = 768 nm shifts significantly to λ_max = 934 nm (ε = 3.3 × 10³ M⁻¹ cm⁻¹) for [1^N]⁻. For comparison, purple-colored [1^C]⁻ was reported to have a charge transfer band at λ_max = 566 nm (ε = 8.6 × 10³ M⁻¹ cm⁻¹), and no NIR data was reported in the previous study.²³ Generally, both 1^N and [1^N]⁻ have their optical transitions shifted to longer wavelengths by 40–50 nm compared to 1^C and [1^C]⁻.

Crystallization of [1^N][K(Krypt₂₂₂)] invariably gave thin plates with very weak X-ray diffraction, and X-ray quality crystals of other salts of [1^N]⁻ (e.g., [1^N][CoCp₂], [1^N][K(18-crown-6)]) were not obtained after repeated attempts. Compound [1^N][K(Krypt₂₂₂)] crystallizes with orthorhombic symmetry in the C222₁ space group, with both cation and anion portions residing on crystallographic special positions (Fig. S9†). The [K(Krypt₂₂₂)]⁺ unit exhibited severe disorder. Furthermore, within the [1^N]⁻ unit, the S atom was disordered over two positions and the NNN ligands exhibited significant disorder, too. Fortunately, the four Cu positions were well ordered. The crystallographic data indicates that, upon reduction from 1^N to [1^N]⁻, the Cu₄S core converts from a rectangular towards a square-based pyramid shape, with experimental Cu–Cu distances of 2.5853(16) and 2.6031(16) Å. A DFT model of [1^N]⁻ (Fig. S10†) was found to have computed Cu–Cu distances of 2.54 and 2.74 Å, which are also shifted from rectangular towards square shaped compared to the DFT model of 1^N (vide supra). The change from rectangular to square shape in the Cu₄ base was also observed previously upon reduction of 1^C to [1^C]⁻, the latter of which has experimental Cu–Cu distances of 2.502(1) and 2.809(1) Å.²³

Having optimized computational models for 1^N and [1^N]⁻, next we conducted TD-DFT calculations to better understand the electronic transitions evident by UV-Vis-NIR. As shown in Fig. 1b, the salient features of the experimental UV-Vis-NIR spectra are well captured in the computed spectra, especially when considering that the computational models used a truncated [MeN=N–NMe]⁻ ligand in place of the [MesN=N-NMes]⁻ ligand used experimentally. For [1^N]⁻, excitations calculated at 639 nm (experimental: 620 nm) and 961 nm (experimental: 934 nm) are both charge transfers into the LUMO (MO145β), which is a highly delocalized MO with 26% S 3p character, 48% Cu 3d character (12% per Cu), and significant contribution from the two NNN ligands syn to the μ₄-S atom (Fig. 2a). The transition calculated at 639 nm involves excitation of an electron from MO144β, which has 74% Cu 3d character. Thus, this transition can be viewed as a combination of Cu→S and Cu→NNN MLCT. The transition calculated at 961 nm involves excitation of an electron from MO141β, which is exclusively based on the four NNN ligands (<5% Cu 3d, <5% S 3p). Thus, this transition can be viewed as NNN→[Cu₄S] CT. The corresponding transitions for neutral 1^N are qualitatively similar but less readily interpreted because of some admixture of states via configuration interactions at each excitation wavelength (Fig. S4†).

Fig. 2 (a) Charge transfer transitions (0.03 isosurfaces) for [1^N]⁻ calculated by TD-DFT at the B3LYP/def2TZVPP level using the CPCM implicit solvation model (CH₂Cl₂). Experimental and simulated X-band EPR spectroscopy for [1^N][K(Krypt₂₂₂)] in frozen CH₃CN/CH₂Cl₂ (1 : 1) glass at 77 K: (b) 1st derivative and (c) 2nd derivative spectra. (d) Spin density (0.001 isosurface) for [1^N]⁻ calculated by DFT. EPR simulation parameters: g = 2.143, 2.066, 2.005; A = 193, 68, 98 MHz for 2 equivalent ¹⁴N nuclei; H-strain = 122, 90, 100 MHz; Gaussian line broadening = 0.43.

The X-band EPR spectrum of [1^N][K(Krypt₂₂₂)] at 77 K in a frozen CH₃CN/CH₂Cl₂ glass is shown in Fig. 2b. Fine structure was not well resolved in the first-derivative spectrum but was emphasized by plotting the second derivative (Fig. 2c). Surprisingly, hyperfine splitting due to the four ^63,65Cu nuclei (I_Cu = 3/2) was not evident. Instead, the spectra were best simulated with hyperfine splitting from two equivalent ¹⁴N nuclei. The simulation also required inclusion of significant “H-strain”, i.e., unresolved hyperfine coupling that can be assigned to a combination of the four ^63,65Cu centers and remaining ¹⁴N nuclei. This pattern can be rationalized by analyzing the spin density from the DFT model of [1^N]⁻, which is plotted in Fig. 2d. Of the total 1e⁻ spin, only 0.113e⁻ resides on each Cu center (0.452e⁻ total), meaning that most of the unpaired spin is ligand-centered. The largest contributor is the μ₄-S atom, which carries 0.300e⁻. Of the remaining 0.248e⁻ of NNN-centered spin, nearly all resides on the two NNN ligands syn to the μ₄-S ligand, with effectively none on the two anti-NNN ligands. Thus, the A(¹⁴N) coupling observed by EPR spectroscopy can be assigned to the central N-atoms of the two syn-NNN ligands. The small and apparently unresolved A(^63,65Cu) coupling is consistent with the high degree of covalency (i.e., μ₄-S and NNN redox non-innocence) apparent from calculations. Although the EPR spectroscopy of [1^C]⁻ differs in that it is an axial signal dominated by A(^63,65Cu) coupling,²³ nonetheless the redox-active MOs for both the 1^N and 1^C systems are similar to each other (Table 1).²⁴

No reaction was observed when [1^N][K(Krypt₂₂₂)] was exposed to N₂O (1 atm) in CH₂Cl₂. This behavior contrasts that of [1^C][K(18-crown-6)], which converts quantitatively to 1^C under the same conditions.²³ In the latter case, it was proposed that reversible N₂O binding to [1^C][K(18-crown-6)] to transiently form [1^C·N₂O][K(18-crown-6)] shifts the reduction potential such that a second equivalent of [1^C][K(18-crown-6)] can act as an electron donor, producing [1^C·N₂O][K(18-crown-6)]₂ and 1^C (i.e., N₂O-induced disproportionation of the copper cluster). The shift of the 1^N/[1^N]⁻ potential to more positive values apparently prevents such N₂O-induced disproportionation from occurring spontaneously for the triazenide derivative. Therefore, the reaction was repeated with a stronger sacrificial electron donor, CoCp₂, to reduce a transiently-formed [1^N·N₂O][K(Krypt₂₂₂)]. A 1 : 1 mixture of [1^N][K(Krypt₂₂₂)] and CoCp₂ in CH₂Cl₂ solution was observed to darken slightly upon exposure to N₂O (1 atm). Over 24 h, the major species in solution remained [1^N]⁻ according to UV-Vis-NIR spectroscopy. However, analysis of the reaction mixture by ¹H NMR spectroscopy indicated small conversion (∼10%) of CoCp₂ to [CoCp₂]⁺ along with formation of a new, diamagnetic product (2) containing the NNN ligand. The same product 2 was generated when exposing a 1 : 2 mixture of 1^N and CoCp₂ to N₂O (1 atm) or when replacing N₂O with excess Me₃NO under N₂. Compound 2 could be extracted into toluene or Et₂O, which served to separate it from [CoCp₂]⁺, [1^N]⁻, and/or 1^N. Washing 2 with pentane served to separate it from CoCp₂ (and excess Me₃NO where relevant, see Fig. S12†). At this point, compound 2 was found to have a brown color, with a UV-Vis-NIR spectrum in CH₂Cl₂ (Fig. 3a) showing a single distinct feature at λ_max = 420 nm but lacking any well-defined charge transfer transitions in the 600–1000 nm region as observed for 1^N and [1^N]⁻. From these collected observations, it can be concluded that (1) the reaction involves O-atom transfer from N₂O since Me₃NO can replace it, (2) the resulting product 2 is diamagnetic and neutral in charge based on NMR characterization and solubility properties, and (3) compound 2 likely lacks the μ₄-S atom that would produce low-energy CT electronic transitions.

Fig. 3 Characterization of complex 2: (a) experimental UV-Vis-NIR (left, CH₂Cl₂) and high-resolution mass spectrometry (right, positive EI mode) data; (b) computed UV-Vis-NIR spectrum (left) and molecular structure (right). Computations were done at the B3LYP/def2TZVPP level of DFT using the CPCM implicit solvation model (CH₂Cl₂); computed UV-Vis-NIR spectra were generated using excitations calculated by TD-DFT and assuming 0.45 eV half-width at half height.

Unfortunately, all reaction conditions explored consistently provided only low conversion (<15%) to 2, with the remaining mass balance being predominantly [1^N]⁻. Furthermore, repeated attempts at obtaining X-ray quality crystals of 2 failed. The identity of 2 was made clear by analyzing it using EI-HRMS. The parent ion mass and isotope distribution (Fig. 3a) indicate the empirical formula, Cu₄(O₂)(NNN)₄. To investigate this possibility further, the structure of 2 assuming that formula was modeled computationally. Two energy minima were located: one with μ₂:η² binding of an O₂ unit to one Cu from the periphery of the Cu₄ core, and another with μ₃:η²,η¹,η¹ binding of the O₂ unit to three Cu atoms in the capping position typically occupied by sulfide (Fig. S13†). The former isomer (Fig. 3b) was calculated to be lower in Gibbs free energy by −13.6 kcal mol⁻¹. The predicted UV-Vis-NIR spectrum (Fig. 3b) of this structure reproduces key features of the experimental spectrum, with a prominent absorbance at λ_max = 448 nm (experimental: 420 nm) that tails into the lower energy region of the spectrum without well-defined absorbance features in the 600–1000 nm region. The local structure within 2 of the Cu(O₂) unit in a nitrogen-rich environment resembles 1 : 1 Cu : O₂ species that have been long studied since seminal works of Kitajima, Tolman, and others.^28,29 Notably, the presence of an absorbance at ∼400 nm is characteristic of such compounds and is typically assigned to the π → π* transition of the bound O₂.³⁰ The calculated O–O distance of 1.37 Å in 2 and the short calculated Cu–O distances of 1.85 and 1.86 Å are all indicative of a highly reduced O₂ unit.^31,32 Curiously, the ¹H NMR spectrum for 2 shown only one set of NNN peaks, rather than the three sets expected based on the calculated structure. This is likely indicative of fluxionality due to the O₂ ligand rapidly hopping from Cu site to Cu site.

Based on these observations, a plausible mechanism for a hypothetical deN₂O reaction for is presented in Scheme 2. Reversible binding of N₂O to [1^N]⁻ would generate a small equilibrium concentration of [1^N·N₂O]⁻. Despite CoCp₂ not being a strong enough reductant to spontaneously reduce [1^N]⁻, reduction of [1^N·N₂O]⁻ could be spontaneous due to the π-accepting nature of N₂O raising the reduction potential of the Cu₄S core. The resulting [1^N·N₂O]²⁻ intermediate would, then, be the active catalyst in the “fully reduced” 4Cu^I redox state.³³ Based on previous computational modeling with [1^C·N₂O]⁻,²⁴ a μ-1,3-N₂O binding mode spanning a Cu–S cluster edge is proposed. Low conversion to 2 observed experimentally is likely due to inefficiency of this reductive activation process, i.e., most of the dissolved complex remains unactivated as [1^N]⁻.

Scheme 2 Plausible mechanisms for N₂ORR and deN₂O.

On the active cycle, loss of N₂ from [1^N·N₂O]²⁻ would generate [1^N·O]²⁻, which is proposed to feature μ₂-O²⁻ binding based on previous modeling for [1^C·O]²⁻.²³ For N₂ORR chemistry associated with the 1^C system, dissociation of O²⁻ from [1^C·O]²⁻ generates 1^C, which can hypothetically undergo 2e⁻ reduction under N₂O to regenerate [1^C·N₂O]²⁻. In the 1^N system, N₂ORR chemistry is not observed. Instead, a second N₂O molecule presumably intercepts [1^N·O]²⁻ to generate [1^N·O₂]²⁻ with loss of N₂. To complete the hypothetical deN₂O cycle, O₂ loss would form [1^N]²⁻, which would regenerate [1^N·N₂O]²⁻ upon N₂O coordination. On the other hand, if S²⁻ dissociation outcompetes O₂ dissociation, then off-cycle species 2 is formed. (Other pathways to 2 cannot be ruled out, vide infra.) Experimentally, examination of the headspace gas above the N₂O reaction revealed only trace quantities of N₂ and O₂, consistent with the off-cycle decomposition to 2 outcompeting on-cycle N₂O-for-O₂ substitution.

To gain further insight, the deN₂O cycle was modeled computationally. The reaction energy profile is shown in Fig. 4, which features models for the four key intermediates as well as the two transition states associated with N–O cleavage of N₂O. The computed intermediate [1^N·N₂O]²⁻ (A) is isostructural to [1^C·N₂O]²⁻ calculated previously,²⁴ with μ-1,3-N₂O binding across a Cu–S edge of the cluster and significant N₂O activation according to the bent O–N–N angle of 126°. Loss of N₂ from A to form [1^N·O]²⁻ (B) is exergonic by ΔG = −28.2 kcal mol⁻¹ and proceeds via transition state TS1 with a barrier height of ΔG^‡ = 12.9 kcal mol⁻¹. The structure of TS1 differs from that of reactant A in that the N₂O ligand occupies a μ-1,1-O binding mode spanning a Cu–Cu cluster edge in TS1. This binding mode has been previously proposed for a synthetic model complex³⁴ and for N₂O activation by Cu-ZSM-5 ³⁵ but differs from that proposed for Cu_Z itself.¹² At TS1, the N₂O molecule has undergone further activation as evidenced by the elongated N–O bond distance (1.57 Å) and contracted N–N bond distance (1.14 Å) relative to those in A (1.297 and 1.233 Å, respectively). The Cu–Cu edge of the cluster with the N₂O bound is short (2.54 Å) to accommodate the bridging ligand and is compensated by a long Cu–Cu distance of 3.35 Å along the edge opposite the N₂O ligand. The structure of B following N₂ loss features a μ-oxo ligand with a Cu–O–Cu angle of 87° and a Cu–Cu distance (2.56 Å) similar to that in TS1.

Fig. 4 Reaction intermediates and transition states for deN₂O calculated by DFT, with Gibbs free energy values given in units of kcal mol⁻¹. Computations were done at the B3LYP/def2TZVPP level of DFT using the CPCM implicit solvation model (CH₂Cl₂) except for TS2 (see ESI†).

Oxygenation of B by N₂O to form [1^N·O₂]²⁻ (C) was calculated to be exergonic by ΔG = −9.6 kcal mol⁻¹ and proceeds via transition state TS2. The structure of TS2 features the N₂O substrate approaching the Cu–O–Cu unit with an O⋯O distance of 1.84 Å. The N₂O unit is significantly activated in TS2, with a bent O–N–N angle of 143°, a long O–N distance of 1.50 Å, and a short N–N distance of 1.13 Å. Unlike compound 2 formed from S²⁻ loss, the structure of C features a μ₂:η¹-O₂ ligand. The calculated O–O distance of 1.35 Å in C is indicative of significant O₂ reduction,³¹ though the Cu–O distances in C (2.02 and 2.06 Å) are longer than those calculated for 2.

Loss of O₂ from C to form [1^N]²⁻ (D) was calculated to be exergonic by ΔG = −29.8 kcal mol⁻¹. Unlike intermediates B and C and transition states TS1 and TS2 that all involve bridging O-ligands along one cluster edge, the structure of D features a nearly square-shaped tetracopper core, with all calculated Cu–Cu distances between 2.66 and 2.78 Å. To complete the hypothetical catalytic deN₂O cycle, coordination of N₂O to D to regenerate A was calculated to be endergonic by ΔG = +25.2 kcal mol⁻¹. Thus, although the net deN₂O reaction is calculated to be favorable (ΔG_rxn = −42.4 kcal mol⁻¹), the rate-determining N₂O binding step prevents efficient catalysis under the experimental conditions examined, allowing for off-cycle conversion of C to inactive 2. At this time, the path from C to 2 is ambiguous: it could involve simple S²⁻ dissociation from C to form 2 directly but could also involve exergonic O₂ dissociation from C to form D followed by O₂-promoted conversion of D to 2.

Oxygenation of B to form C presumably involves nucleophilic addition of the N₂O oxygen to the electrophilic μ-oxo ligand of B (see TS2). In accord with that hypothesis, analysis of the frontier molecular orbitals of B indicated that the LUMO has significant (24%) oxygen 2p character (Fig. 5). One reason that the 1^N system is more likely than 1^C to do deN₂O chemistry might be the stabilization of this LUMO due to the electronegative nature of the NNN⁻ ligands relative to less electronegative NCN⁻, thus favoring O–O coupling by N₂O nucleophilic addition in the NNN⁻ case. A second contributing reason could be that the relatively electron-withdrawing NNN⁻ ligands make the tetracopper core more Lewis acidic than for the NCN⁻ analogues, thus causing O²⁻ loss required for N₂ORR to be less facile for B and instead shunting the 1^N system into the deN₂O pathway.

Fig. 5 Lowest unoccupied molecular orbital (0.03 isosurface) calculated for [1^N·O]²⁻ (B) by DFT at the B3LYP/def2TZVPP level using the CPCM implicit solvation model (CH₂Cl₂).

Conclusions

A redox pair of Cu₄(μ₄-S) clusters supported by triazenide ligands was synthesized and thoroughly characterized. While much of the characterization data mirrored those of the related formamidinate-supported system studied previously,^21–24 divergent reactivity behavior with N₂O was observed. While the formamidinate-ligated system was found to promote single-turnover N₂ORR akin to the biological Cu_Z cluster it mimics, the triazenide system was found to promote (inefficient) deN₂O reactivity that is a typical hallmark of heterogeneous catalysts like CuO/CeO₂. Computational modeling of the deN₂O pathway allowed for comparisons to be made with previously studied N₂ORR chemistry, enabling deeper understanding of factors influencing selectivity. A simple working model is as follows. The formamidinate system is relatively electron-rich, providing a “soft” tetracopper core that behaves like Cu_Z (that features a soft sulfide ligand) by readily dissociating O²⁻ upon N₂O deoxygenation to favor N₂ORR. On the other hand, the triazenide system is relatively electron-poor, providing a “hard” tetracopper core that behaves like CuO/CeO₂ (that features hard oxide ligation) by forming a tightly-bound and electrophilic μ-oxo intermediate upon N₂O deoxygenation to favor deN₂O. These insights stand to inform future N₂O-fixing catalyst designs.

Data availability

Spectral data is reproduced in ESI,† and raw data are available from the author upon request. Computational output coordinates have been uploaded as ESI.† X-ray crystallography data is available upon request from the CCDC under deposition numbers 2299540 and 2299541.

Author contributions

N. P. M. carried out experiments, conducted computational modeling, and wrote the manuscript.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

Funding was provided by NIH/NIGMS under grant R35 GM140850. Computational resources and services were provided by the Advanced Cyberinfrastructure for Education and Research (ACER) group at UIC. Dr Daniel McElheny (UIC) and Dr Samantha MacMillan (Cornell) assisted with collection of EPR and X-ray diffraction data, respectively.

Notes and references

1. J. Hansen and M. Sato, Proc. Natl. Acad. Sci. U. S. A., 2004, 101, 16109–16114.
2. A. R. Ravishankara, J. S. Daniel and R. W. Portmann, Science, 2009, 326, 123–125.
3. H. Tian, R. Xu, J. G. Canadell, R. L. Thompson, W. Winiwarter, P. Suntharalingam, E. A. Davidson, P. Ciais, R. B. Jackson, G. Janssens-Maenhout, M. J. Prather, P. Regnier, N. Pan, S. Pan, G. P. Peters, H. Shi, F. N. Tubiello, S. Zaehle, F. Zhou, A. Arneth, G. Battaglia, S. Berthet, L. Bopp, A. F. Bouwman, E. T. Buitenhuis, J. Chang, M. P. Chipperfield, S. R. S. Dangal, E. Dlugokencky, J. W. Elkins, B. D. Eyre, B. Fu, B. Hall, A. Ito, F. Joos, P. B. Krummel, A. Landolfi, G. G. Laruelle, R. Lauerwald, W. Li, S. Lienert, T. Maavara, M. MacLeod, D. B. Millet, S. Olin, P. K. Patra, R. G. Prinn, P. A. Raymond, D. J. Ruiz, G. R. van der Werf, N. Vuichard, J. Wang, R. F. Weiss, K. C. Wells, C. Wilson, J. Yang and Y. Yao, Nature, 2020, 586, 248–256.
4. S. R. Pauleta, M. S. P. Carepo and I. Moura, Coord. Chem. Rev., 2019, 387, 436–449.
5. W. C. Trogler, Coord. Chem. Rev., 1999, 187, 303–327.
6. N. Lehnert, B. W. Musselman and L. C. Seefeldt, Chem. Soc. Rev., 2021, 50, 3640–3646.
7. C. Ferousi, S. H. Majer, I. M. Dimucci and K. M. Lancaster, Chem. Rev., 2020, 120, 5252–5307.
8. S. R. Pauleta, S. Dell'Acqua and I. Moura, Coord. Chem. Rev., 2013, 257, 332–349.
9. S. R. Pauleta, C. Carreira and I. Moura, in Metalloenzymes in Denitrification: Applications and Environmental Impacts, ed. I. Moura, L. B. Maia, S. R. Pauleta and J. J. G. Moura, Royal Society of Chemistry, 2017, pp. 141–169.
10. C. Carreira, S. R. Pauleta and I. Moura, J. Inorg. Biochem., 2017, 177, 423–434.
11. S. C. Rathnayaka and N. P. Mankad, Coord. Chem. Rev., 2021, 429, 213718.
12. E. M. Johnston, C. Carreira, S. Dell'Acqua, S. G. Dey, S. R. Pauleta, I. Moura and E. I. Solomon, J. Am. Chem. Soc., 2017, 139, 4462–4476.
13. E. M. Johnston, S. Dell'Acqua, S. Ramos, S. R. Pauleta, I. Moura and E. I. Solomon, J. Am. Chem. Soc., 2014, 136, 614–617.
14. A. Pomowski, W. G. Zumft, P. M. H. Kroneck and O. Einsle, Nature, 2011, 477, 234–237.
15. A. Wüst, L. Schneider, A. Pomowski, W. G. Zumft, P. M. H. Kroneck and O. Einsle, Biol. Chem., 2012, 393, 1067–1077.
16. M. Konsolakis, ACS Catal., 2015, 5, 6397–6421.
17. M. H. Groothaert, K. Lievens, H. Leeman, B. M. Weckhuysen and R. A. Schoonheydt, J. Catal., 2003, 220, 500–512.
18. P. J. Smeets, M. H. Groothaert, R. M. van Teeffelen, H. Leeman, E. J. M. Hensen and R. A. Schoonheydt, J. Catal., 2007, 245, 358–368.
19. M. Zabilskiy, P. Djinović, B. Erjavec, G. Dražić and A. Pintar, Appl. Catal., B, 2015, 163, 113–122.
20. B. J. Johnson, S. V. Lindeman and N. P. Mankad, Inorg. Chem., 2014, 53, 10611–10619.
21. S. C. Rathnayaka, C. W. Hsu, B. J. Johnson, S. J. Iniguez and N. P. Mankad, Inorg. Chem., 2020, 59, 6496–6507.
22. B. J. Johnson, W. E. Antholine, S. V. Lindeman and N. P. Mankad, Chem. Commun., 2015, 51, 11860–11863.
23. B. J. Johnson, W. E. Antholine, S. V. Lindeman, M. J. Graham and N. P. Mankad, J. Am. Chem. Soc., 2016, 138, 13107–13110.
24. S. C. Rathnayaka, S. M. Islam, I. M. Dimucci, S. N. MacMillan, K. M. Lancaster and N. P. Mankad, Chem. Sci., 2020, 11, 3441–3447.
25. C.-W. W. Hsu, S. C. Rathnayaka, S. M. Islam, S. N. MacMillan and N. P. Mankad, Angew. Chem., Int. Ed., 2020, 59, 627–631.
26. P. Alayoglu, T. Chang, M. V. Lorenzo Ocampo, L. J. Murray, Y.-S. Chen and N. P. Mankad, Inorg. Chem., 2023, 62, 15267–15276.
27. N. G. Connelly and W. E. Geiger, Chem. Rev., 1996, 96, 877–910.
28. K. Fujisawa, M. Tanaka, Y. Moro-oka and N. Kitajima, J. Am. Chem. Soc., 1994, 116, 12079–12080.
29. D. J. E. Spencer, N. W. Aboelella, A. M. Reynolds, P. L. Holland and W. B. Tolman, J. Am. Chem. Soc., 2002, 124, 2108–2109.
30. N. W. Aboelella, S. V. Kryatov, B. F. Gherman, W. W. Brennessel, V. G. Young, R. Sarangi, E. V. Rybak-Akimova, K. O. Hodgson, B. Hedman, E. I. Solomon, C. J. Cramer and W. B. Tolman, J. Am. Chem. Soc., 2004, 126, 16896–16911.
31. D. A. Iovan, A. T. Wrobel, A. A. McClelland, A. B. Scharf, G. A. Edouard and T. A. Betley, Chem. Commun., 2017, 53, 10306–10309.
32. N. C. Tomson, K. D. Williams, X. Dai, S. Sproules, S. DeBeer, T. H. Warren and K. Wieghardt, Chem. Sci., 2015, 6, 2474–2487.
33. S. Ghosh, S. I. Gorelsky, P. Chen, I. Cabrito, J. J. G. Moura, I. Moura and E. I. Solomon, J. Am. Chem. Soc., 2003, 125, 15708–15709.
34. I. Bar-Nahum, A. K. Gupta, S. M. Huber, M. Z. Ertem, C. J. Cramer and W. B. Tolman, J. Am. Chem. Soc., 2009, 131, 2812–2814.
35. M.-L. L. Tsai, R. G. Hadt, P. Vanelderen, B. F. Sels, R. A. Schoonheydt and E. I. Solomon, J. Am. Chem. Soc., 2014, 136, 3522–3529.

---

Footnote

† Electronic supplementary information (ESI) available: Experimental section and spectra data, crystallographic data, computational output. CCDC 22995402299541. For ESI and crystallographic data in CIF or other electronic format see DOI: https://doi.org/10.1039/d3sc05451a

---