Methane Activation by a Mononuclear Copper Active Site in the Zeolite Mordenite: Effect of Metal Nuclearity on Reactivity

Abstract

The direct conversion of methane to methanol would have wide reaching environmental and industrial impact. Copper containing zeolites can perform this reaction at low temperatures and pressures at a previously defined O₂ activated [Cu₂O]²⁺ site. However, after autoreduction of the copper containing zeolite mordenite and removal of the [Cu₂O]²⁺ active site, the zeolite is still methane reactive. In this study, we use diffuse reflectance UV-Vis, magnetic circular dichroism, resonance Raman, electron paramagnetic resonance, and X-ray absorption spectroscopies to unambiguously define a mononuclear [CuOH]⁺ as the CH₄ reactive active site of the autoreduced zeolite. The rigorous identification of a mononuclear active site allows a reactivity comparison to the previously defined [Cu₂O]²⁺ active site. We perform kinetic experiments to compare the reactivity of the [CuOH]⁺ and [Cu₂O]²⁺ sites and find that the binuclear site is significantly more reactive. From analysis of density functional theory calculations, we elucidate that this increased reactivity is a direct result of stabilization of the [Cu₂OH]²⁺ H-atom abstraction product by electron delocalization over the two Cu cations via the bridging ligand. This significant increase in reactivity from electron delocalization over a binuclear active site provides new insight for the design of highly reactive oxidative catalysts.

Graphical Abstract

1. Introduction

The direct conversion of methane to methanol is a reaction of paramount importance with both environmental and industrial interest. Given the strong C-H bond of methane (104 kcal/mol) and the tendency for overoxidation of the methanol product, the main route of selective methane oxidation utilizes a harsh and expensive process involving a syngas intermediate.¹ This precludes employing this syngas route at small scale oil extraction sites, leading to wasteful methane flaring and unnecessary greenhouse gas production. However, nature has solved this problem. The enzymes soluble and particulate methane monooxygenase (sMMO and pMMO) use iron and copper cofactors, respectively, to transform methane to methanol under ambient conditions. The iron active site in sMMO is a clearly defined binuclear site while the nature of the copper active site in pMMO is an open question with proposals for a mononuclear, binuclear, or trinuclear active site.^2–8 Similar to the enzymes, iron and copper exchanged zeolites can also selectively convert methane to methanol under relatively mild conditions, employing many of the characteristics that make their enzyme analogs so reactive but in a material that can be potentially scaled for industrial use.^9–11

Copper zeolites are prepared for methane reaction by first performing an aqueous exchange with a cupric salt and calcining the resulting material (heating in O₂ at 450°C), combusting the anionic portion of the Cu salt and dehydrating the zeolite. This calcined material can selectively convert methane to methanol. However, this Cu(II) active site can be reduced by heating in helium at high temperatures (≥500°C) concurrent with the release of O₂ in a process called autoreduction. The reduced active site can be reactivated for methane oxidation by addition of N₂O or O_2. We rigorously defined this active site as a [Cu₂O]²⁺ species.^12–14 Recently, there have been reports that the autoreduced zeolite of the topology mordenite (MOR), which should have its [Cu₂O]²⁺ active site reduced, is still able to oxidize methane to methanol. Whether or not this is due to a different active site or whether the autoreduced zeolite has been partially activated by O₂ leaks creating [Cu₂O]²⁺ sites has been an open issue.^15,16

In this study, we investigate the CH₄ reaction of the autoreduced MOR zeolite using electron paramagnetic resonance, X-ray absorption, diffuse reflectance UV-Vis, magnetic circular dichroism, and resonance Raman spectroscopies. We define the active site of the autoreduced zeolite as a mononuclear [CuOH]⁺ bound bidentate to an aluminate tetrahedral site (T-site). Although this active site has been proposed in past studies^17–19, it has not been rigorously experimentally demonstrated with multiple spectroscopies nor associated with the autoreduced zeolite.

Having a well-defined mononuclear active site allows comparison to the previously defined binuclear [Cu₂O]²⁺ active site to assess the effects of metal nuclearity on reactivity. This is important given that the reactivity of the [CuOH]⁺ site is an open question with claims of it being a non-reactive precursor site²⁰ to claims that it has a lower activation barrier for reaction with CH₄ than the binuclear [Cu₂O]²⁺ active site.¹⁹ We find experimentally that the [CuOH]⁺ site has a significantly higher barrier (ΔH^‡ = 15.7 kcal/mol) than the [Cu₂O]²⁺ active site (ΔH^‡ = 6.2 kcal/mol) for H-atom abstraction (HAA) from CH₄. Using density functional theory, we compare models for [CuOH]⁺ and [Cu₂O]²⁺ and elucidate that the weak ligand field and the low-coordination environment of the zeolite lattice activates both the mononuclear and binuclear Cu sites to perform methane activation with similar intrinsic barriers. However, the binuclear site is much more reactive in HAA due to its ability to stabilize the [Cu₂OH]²⁺ product of HAA by electron delocalization over the two Cu ions via a superexchange interaction through the bridging oxo ligand. The increased binuclear driving force due to delocalization indicates the importance of metal nuclearity for low temperature reactivity. Such information can greatly aid in designing better oxidation catalysts generally and in providing insight into the role that the higher nuclearity models of pMMO can play in enhancing C-H bond cleavage by delocalization of the electron gained from HAA.^4–8

2. Results and Analysis

2.1. Physical Properties

Autoreduced Cu-MOR samples were prepared in three steps: aqueous exchange with Cu(II) acetate on the parent Na-MOR zeolite, calcination with O₂ at 450°C to burn off acetate and dehydrate, and heating in helium at 500°C overnight to generate the autoreduced form (see Section S1). The Cu/Al ratio was determined to be 0.45 from inductively coupled plasma atomic emission spectroscopy (Table S1). These steps do not perturb the inherent structure of the zeolite lattice and no large copper oxide clusters were detected by powder X-ray diffraction (PXRD) (Figures S1). The microporosity was determined to be similar for Na-MOR and Cu-MOR (0.18 cm³/g) by N₂ physisorption, in agreement with literature values (Figure S2).²¹

2.2. Characterization and CH₄ Reaction of Autoreduced Cu-MOR

Autoreduced Cu-MOR has two distinct electron paramagnetic resonance (EPR) active S = 1/2 Cu(II) sites (Figure 1A black). These two axial species have the following EPR parameters^22,23: g_|| = 2.27, A_|| = 191 × 10⁻⁴ cm⁻¹, g_⊥ = 2.09, A_⊥ = 18–42 × 10⁻⁴ cm⁻¹, referred to as MOR-A, and g_|| = 2.32 A_|| = 169 × 10⁻⁴ cm⁻¹, g_⊥ = 2.07, A_⊥ = 9–22 × 10⁻⁴ cm⁻¹ referred to as MOR-B (with sharp g_⊥ hyperfine features observed in other studies).²⁴ These two sites account for 55±11% of total copper based on spin quantification. Reaction of the autoreduced zeolite with CH₄ at 200°C selectively reduces MOR-A, lowering the amount of EPR active copper in the zeolite from 55±11% to 25±5% (Figure 1A green). This occurs concurrent with the production of methanol obtained through aqueous or steam extraction of the CH₄ reacted sample. Thus MOR-A and MOR-B account for 30±12% and 25±5% respectively of total copper present in the autoreduced Cu-MOR. The remaining copper is not EPR detectable (down to 5 K and a field region of 0–4000 G).

Figure 1.

EPR (A) and XAS (B) of autoreduced (black), CH₄ reacted (green), and N₂O Activated (pink) Cu-MOR.

X-ray absorption spectroscopy (XAS) provides a complimentary technique to analyze the ratio of total Cu(I)/Cu(II) in the sample using the pre-edge and the rising edges (the extended fine structure region did not provide useful information due to heterogeneity as presented in Explanation S1). XAS of autoreduced MOR reveals overlapping features in the 8983–8985 eV region associated with the 1s → 4p transition of the Cu(I) in the sample along with an 8979 eV peak that is the 1s → 3d transition of the Cu(II) present (Figure 1B black).²⁵ Reaction of the autoreduced Cu-MOR sample with N₂O results in total oxidation as there is no remaining Cu(I) feature in the XAS spectrum from 8983–8985 eV (2.5±2.5% Cu(I)) (Figure 1B pink and Explanation S2). This N₂O reaction of autoreduced Cu-MOR also results in the appearance of a 22,000 cm⁻¹ charge transfer (CT) band in the diffuse reflectance UV-Vis spectrum (DR UV-Vis) (Figure S3) that corresponds to the previously assigned [Cu₂O]²⁺ active sites.¹³ The bicuprous precursors of these [Cu₂O]²⁺ active sites in autoreduced Cu-MOR quantify to 25±14% total copper (Explanation S2). This [Cu₂O]²⁺ species is absent in the autoreduced Cu-MOR as evidenced by the lack of the characteristic 22,000 cm⁻¹ CT band in DR UV-Vis (Figure 2A) and the lack of characteristic vibrations in resonance Raman (rR) with excitation at 458 nm (vide infra). From a combination of the above EPR and XAS results, we estimate that the autoreduced Cu-MOR has 30±12% MOR-A, 25±5% MOR-B, and 25±14% bicuprous sites that are precursors to [Cu₂O]²⁺ active sites. Given the measurement errors (in particular the Cu(I) estimate from XAS) there is the open possibility of up to 20±19% unaccounted for copper either from error in Cu(I) quantification or from Cu(II) that is EPR silent due to magnetic coupling but pXRD silent due to small cluster size.

Figure 2.

(A) DR UV-Vis of autoreduced (black) and CH₄ reacted (green) Cu-MOR. (B) Eyring plot of loss of DR UV-Vis signal from 20,000 to 22,000 cm⁻¹ for the reaction of autoreduced Cu-MOR with CH₄.

In parallel with the loss of the MOR-A EPR signal (with the production of CH₃OH), reaction of the autoreduced zeolite with CH₄ results in an increase in intensity in the 8983–8985 eV region of the XAS spectrum and loss of intensity in the pre-edge at 8979 eV, both consistent with Cu(II) reduction (Figure 1B green). Given the single turnover conditions, the additional Cu(I) upon CH₄ reduction directly relates to active site quantity. The CH₄ reacted sample shows a 45±20% increase in Cu(I) XAS intensity relative to the autoreduced sample (Explanation S2), within error of the 30±12% loss of the MOR-A site observed in EPR. Although the error in these quantifications leaves open the possibility of other methane reactive sites, loss of the MOR-A signal is sufficient to explain the reduction observed in XAS. Thus, we can directly correlate loss of EPR signal intensity to active site reactivity and assign MOR-A as the CH₄ reactive site in autoreduced Cu-MOR. This reaction produces 0.1 mol methanol/mol total Cu (Table S1), equivalent to 0.33±0.14 mol methanol/mol MOR-A (Note that two mononuclear Cu(II) sites are required for producing one mol of methanol, vide infra). After extraction of methanol, addition of oxygen and dehydration by heating in helium to 500°C allows regeneration of the autoreduced zeolite (Figures S4 and S5) that can subsequently react with CH₄ to form the same amount of methanol

Using this information, we can model the evolution of copper species in the zeolite under various conditions (Scheme 1). After aqueous exchange of copper acetate into the zeolite, calcination removes acetate ions and dehydrates the zeolite to leave copper ions bound to the lattice. This results in the [Cu₂O]²⁺ site assigned in our previous study¹³ and MOR-A and MOR-B analyzed here. The identity of the MOR-A site will be discussed in Sections 2.3 and 2.4. The MOR-B site is thought to be ligated by two aluminum sites²³, lowering its oxidizing potential and making it unreactive with CH₄ as revealed in this study (Figure 1A). The [Cu₂O]²⁺ sites can be selectively reduced by heating in helium to result in the autoreduced material which is the focus of this study. The [Cu₂O]²⁺ site can be regenerated through oxidation in O₂ or N₂O. Through EPR, XAS, DR UV-Vis, and rR (vide infra) we demonstrate that in the autoreduced sample only the MOR-A site reacts with CH₄ and is selectively reduced in this reaction. The MOR-A Cu(II) active site’s ability to break the strong C-H bond of CH₄ motivates our investigation into its geometric and electronic structure explored below.

Scheme 1.

Composition of species in Cu-MOR under various conditions.

2.3. Spectroscopic Definition of the MOR-A Active Site

To gain more insight into the nature of the MOR-A site, DR UV-Vis spectra of the autoreduced and CH₄ reacted samples were collected (the spectrum of dried Na-MOR was collected as a comparison in Figure S6). Autoreduced Cu-MOR has overlapping features which lose intensity across the spectral envelope upon reaction with CH₄ (Figure 2A). The broadness of the features makes them difficult to parse for chemical information. There is also the issue of contributions from multiple species simultaneously, e.g.. MOR-A, and MOR-B. Magnetic Circular Dichroism (MCD) is an optical technique that provides complementary information on electronic transitions to DR UV-Vis.²⁶ While DR UV-Vis has electronic transitions from all species in the zeolite, the MCD spectrum of autoreduced Cu-MOR reveals transitions that are all C-terms (temperature and field dependent transitions), indicating they are specifically the electronic transitions of the paramagnetic MOR-A and MOR-B sites rather than Cu(I) or potential EPR silent Cu(II) in the sample (Figures S7 & S8). The MCD spectrum of autoreduced Cu-MOR reveals low energy, low intensity excited states at 9,000 and 11,200 cm⁻¹, intense overlapping features above 10,000 cm⁻¹ with a maximum at 13,800 cm⁻¹, and two higher energy features with low MCD intensity at 20,700 cm⁻¹ and 29,500 cm⁻¹ (Figure 3). The CH₄ reacted Cu-MOR MCD spectrum (with no MOR-A EPR signal) has a positive feature around 9,000 cm⁻¹ and no feature at 21,000 cm⁻¹, indicating the MOR-A site has a negative feature at 9,000 cm⁻¹, overlapping negative features around 13,500 cm⁻¹, and a weak negative feature at 21,000 cm⁻¹ (Figure S9). MCD C-term intensity originates from spin-orbit coupling (SOC) (large for metal centered transitions) whereas DR UV-Vis intensity reflects the electric dipole moment of the transition. Thus metal centered d-d transitions (that are electric dipole forbidden but with high SOC) have large MCD/diffuse reflectance intensity ratios relative to ligand to metal charge transfer bands (that are electric dipole allowed but with low SOC). Consequently, we can use the combination of DR UV-Vis and MCD to assign MOR-A’s low energy features centered at 13,500 cm⁻¹ as multiple Cu d-d transitions (large MCD to diffuse reflectance intensity ratios) and its higher energy features at 20,700 and 29,500 cm⁻¹ as CT bands (small MCD to diffuse reflectance intensity ratios).

Figure 3.

DR UV-Vis (top) and MCD (bottom) spectra of autoreduced Cu-MOR. ● - Dots correspond to the equally colored Resonance Raman spectra of Figure S11, measured at different laser energies corresponding to the location of the dot on the x-axis. The relative intensities of the rR spectra (normalized to the lattice vibration at 480 cm-1) correspond to the height of the dot in the figure. This gives the resonance Raman excitation profile.

Assignment of MOR-A’s CT bands using MCD and DR UV-Vis provides an opportunity to obtain vibrational information on the active site using resonance Raman spectroscopy. Laser excitation of autoreduced Cu-MOR at 458 nm enhances three vibrational features at 552, 750, and 981 cm⁻¹ that are lost upon reaction with CH₄ (Figure 4A black vs green). There are also no clear resonance enhanced [Cu₂O]²⁺ vibrations (very intense in rR at 460 and 530 cm⁻¹) nor any vibrations at energies above 1000 cm⁻¹ (Figure S10). We thus see no evidence of [Cu₂O]²⁺ active sites in autoreduced Cu-MOR. The vibrations we do observe profile with the 20,700 cm⁻¹ MOR-A CT band, associating these vibrations with the MOR-A active site (Figures 3 top & Figure S11).

Figure 4.

(A) Resonance Raman spectrum of autoreduced (black), ¹⁸O labeled autoreduced (purple), and CH₄ reacted (green) Cu-MOR. λ_ex = 458nm. (B) DFT predicted rR vibrations and isotope shifts of T1 Site 2 model of [CuOH]⁺.

To gain further information on the vibrations, ¹⁸O and D isotope labels were incorporated into the active site by repeatedly steaming the autoreduced zeolite at 200°C with H₂¹⁸O or D₂O and subsequently drying the zeolite in He at 500°C. This procedure reproduces the DR UV-Vis and EPR of the autoreduced zeolite (Figures S12 & S13) and maintains the integrity of the active site while incorporating an isotope label. Water has been used in past studies to incorporate isotope labels into the zeolite through isotope scrambling after hydration.²⁷ Although there are no observable isotope shifts in the 552 and 981 cm⁻¹ MOR-A rR features of the H₂¹⁸O treated sample, there is a red shift of the 750 cm⁻¹ feature by 35 cm⁻¹, indicating that an exchangeable oxygen is associated with this vibration (Figure 4A purple). Treating the 750 cm⁻¹ vibration as a harmonic oscillator Cu-O stretch where the O belongs to an extra lattice ligand predicts a red shift to 716 cm⁻¹ as observed. This 750 cm⁻¹ vibration indicates a strong Cu-O bond corresponding to a short 1.71 Å bond length from Badger’s rule.²⁸ Steaming and heating with D₂O does not detectably perturb any features (Figure S14). This is expected as 5 cm⁻¹ is the smallest isotope shift that would be detectable on the 750 cm⁻¹ feature based on instrument resolution and the broadness of the feature, and a 4 cm⁻¹ red shift is predicted based on our [CuOH]⁺ active site model vide infra).

The S = ½ EPR signal and the strong Cu-O stretch vibration provide handles to define the MOR-A site. The binuclear [Cu₂O]²⁺ site first identified by Woertink et al. is an oxo bridged binuclear Cu(II) site. Thus, the overall spin must either be S = 1 or S = 0, and a [Cu₂O]²⁺ is excluded for MOR-A.¹² A binuclear site that could match the S = ½ MOR-A spin would be a localized mixed valent Cu(I)Cu(II) site. This possibility was eliminated by N₂O oxidation of essentially all Cu(I) (Figure 1B) with no change in the MOR-A EPR signal (Figure 1A). Trinuclear Cu(II) centers have been proposed to exist in Cu-MOR²⁹ and are S = ½; however, they exhibit a large derivative shaped pseudo-A term in MCD that is not observed for MOR-A (Figure 3).³⁰ This limits MOR-A to a mononuclear Cu(II) site. Amongst the mononuclear site possibilities, an end-on superoxo and a Cu(II) oxyl can be excluded because they would have triplet ground states.^31,32 An end-on hydroxoperoxo would match the spin, but its resonance Raman spectrum would exhibit a Cu-O stretch at lower energy (~624 cm⁻¹) and a higher intensity, isotope-sensitive O-O stretch at > 800 cm⁻¹ that is not present in the observed rR spectrum (Figure 4A).³³ The remaining possibility is a mononuclear cupric hydroxide. This matches the spin, and a DFT model for the site matches the DR UV-Vis, MCD, and rR experimental data well (vide infra). We thus assign MOR-A as a [CuOH]⁺ bound to the zeolite lattice. This site has clear spectroscopic handles that allow its assignment in other lattices, including in the autoreduced copper-containing MFI topology (Figures S15–17), indicating it is not unique to the MOR topology.

Thus, the CH₄ oxidation active site in autoreduced Cu-MOR is a [CuOH]⁺ species that is not reduced upon autoreduction, distinctly different than the [Cu₂O]²⁺ active site that is lost upon heating in helium. A [CuOH]⁺ methane reactive active site has been proposed in the literature,^17–19 but here we provide the first rigorous definition of this active site as well as its association with methanol production by autoreduced zeolites. The reactivity of this [CuOH]⁺ site also lacks consensus in the literature with reports that it is a non-reactive precursor site²⁰ to reports that it is has a lower barrier than the highly reactive [Cu₂O]²⁺ site.¹⁹ The reactivity of this spectroscopically well-defined [CuOH]⁺ active site with CH₄ can now be compared to the [Cu₂O]²⁺ active site defined earlier. We chose to compare to the [Cu₂O]²⁺ in the CHA topology. This model was experimentally validated by comparing the CuOCu angle of multiple generated [Cu₂O]²⁺ structures to the CuOCu angle determined by normal coordinate analysis of the resonance Raman vibrations.¹⁴ We chose this model to compare to the [CuOH]⁺ site in MOR given its existing detailed computational analysis of its reactivity¹⁴ and the fact that it contains only a single [Cu₂O]²⁺ site in contrast to the multiple [Cu₂O]²⁺ sites present in Cu-MOR.¹³

Woertink et al. observed a CH₄/CD₄ kinetic isotope effect in Cu-MFI, indicating that the rate limiting step in the reaction of the [Cu₂O]²⁺ site with CH₄ in Cu-MFI is the C-H bond cleavage by the copper active site.¹² Thus, CH₄ transport is faster than C-H bond cleavage by the [Cu₂O]²⁺ in MFI. This can be reasonably assumed for the [CuOH]⁺ in MOR as it has larger channels, thus more facile CH₄ transport, and yet a slower reaction is observed (vide infra). This indicates the rate must be related to the reactivity of the Cu active site. Autoreduced Cu-MOR was reacted with CH₄ and the decay of the MOR-A 21,000 cm⁻¹ CT band was monitored using DR UV-Vis to extract a first order rate constant (Figure S18). This was repeated for multiple temperatures to generate an Eyring plot (Figure 2B). We estimate a ΔH^‡ of 15.7 kcal/mol and a ΔS^‡ of −39.5 cal/mol·K for the barrier to CH₄ oxidation by [CuOH]⁺. These are compared to values for CH₄ activation by [Cu₂O]²⁺ in CHA in Table 1 where the MOR-A site is experimentally found to have a 9.5 kcal/mol higher enthalpic barrier.¹⁴ This is a significant reactivity differential given the range in ΔH^‡ between all identified [Cu₂O]²⁺ sites is less than 5 kcal/mol.^12–14 We now create an experimentally validated DFT model for the [CuOH]⁺ active site and model the reactivity of this site in parallel with the [Cu₂O]²⁺ site in CHA to better understand this large difference in activation barrier for C-H abstraction from methane.

Table 1.

Comparison of the kinetic parameters for the MOR-A [CuOH]⁺ site with the previously studied [Cu₂O]²⁺ site in Cu-CHA.¹⁴

	ΔH‡ (kcal/mol)	ΔS‡ (cal/mol-K)
[CuOH]+	15.7	−39.5
[Cu2O]2+	6.2	−51.7

2.4. MOR-A Active Site Models

Density Functional Theory (DFT) models for the [CuOH]⁺ site were constructed and compared to experimental data. MOR has four distinct T-sites (T1–4) with typically over 80% of the alumina concentrated at T1, T3, and T4^34,35 which were then used to generate active site models (for visualization of the MOR T-sites we direct readers to the following publications^34,36). Each T-site has 6 edges where a [CuOH]⁺ can bind bidentate to two oxygen atoms of the aluminate site. A [CuOH]⁺ was placed at each of these locations for each T-site and optimized, generating 18 initial models (example structure Figure 5). Due to steric constraints or lack of a local minimum, some edges could not reasonably bind the [CuOH]⁺, leaving the 11 models listed in Table S2. In all models, Cu binds bidentate to the anionic oxygen atoms of the aluminate T-site. The relative stability of each model was evaluated by calculating the binding energy for each [CuOH]⁺. This was calculated as the energy of exchanging a Na⁺ bound to the T-site for a bound [CuOH]⁺ (Equation 1) in the aqueous exchange step (modeled with a water polarizable continuum model).

Figure 5.

T1 Site 2 [CuOH]⁺ active site model. (Color scheme: red = O, pink = Al, gray = Si, mustard = Cu).

$$\text{Binding energy}=\text{E}\left({\text{Na}}^{+}\text{ bound}\right)-\text{E}\left({\left[\text{CuOH}\right]}^{+}\text{ bound}\right)\left(\text{calculated relative between sites}\right)$$ (Equation 1)

11 structures were obtained with stable geometries. The [CuOH]⁺ binds to the lattice with a 15 kcal/mol range in energies across the various sites (Table S2). To probe the effect of second-sphere interactions on site stability, each model was truncated to just the [CuOH]⁺, the bound aluminate, and the adjacent silica of the aluminate (example structure Figure S19). The six most stable large models maintain the highest binding energies after truncation, indicating the stability is predominantly related to the first coordination sphere. The most stable sites place the [CuOH]⁺ close to the SiOAl planes of both O atoms and have narrower ∠SiOAl angles (<140° increasing p-character) to allow better bonding with the aluminate site. Note that all stable structures have a CuOH angle of 120–125° with the CuOH plane perpendicular to the xy plane formed by the T-site oxygen atoms, the Cu, and the oxygen of the hydroxide (Figure 5).

Time dependent density functional theory (TD-DFT) and vibrational frequency calculations were performed on the 11 potential sites to compare with the experimental data (Table S2). The predicted most intense OH⁻ to Cu(II) CT transition energy ranges from 24,400–26,200 cm⁻¹ and the Cu-OH stretch vibration ranges from 682–706 cm⁻¹ across the 11 models, thus the variability is small. The T1 Site 2 model in Figure 5 has a Cu-O stretch of 700 cm⁻¹, reasonably close to the 750 cm⁻¹ vibration observed experimentally, while also having one of the highest binding energies. Thus, this was chosen as a reasonable active site model, but it is likely that the [CuOH]⁺ binds in a range of stable locations with very similar spectral features.

An additional possible model was evaluated by adding a water ligand to the T1 Site 2 model. This four-coordinate model optimizes to a distorted square planar geometry (Figure S20) with a decrease in the Cu-O stretch to 600 cm⁻¹ and a higher energy CT band at 27,000 cm⁻¹. Adding the water thus removes the low coordination environment necessary to form the strong Cu(II)-OH bond observed in rR, excluding this as a possible structure for MOR-A.

More in-depth spectroscopic analysis was performed on the T1 Site 2 [CuOH]⁺ (Figure 5) model. The TD-DFT ligand field energies (Figure S21) match the bands observed in MCD (Figure 3). In addition to reproducing the Cu-O vibration, this T1 Site 2 model also reasonably predicts the low and high frequency rR peaks and their lack of isotope perturbations as diagrammed in Figure 4B. The 552 cm⁻¹ vibration is the symmetric lattice mode (also observed in the rR spectra of all [Cu₂O]²⁺ sites)^12–14 and the 982 cm⁻¹ vibration is a symmetric lattice bending mode where the Al T-Site Cu^II-OH site is predicted to distort along this mode upon OH⁻ to Cu(II) CT excitation (Explanation S3). We thus assign the MOR-A structure to a 3-coordinate trigonal planar Cu(II) complex bound bidentate to the zeolite lattice with a strong extra lattice OH⁻ ligand (Figure 5). The short Cu-OH bond in this model is a direct result of the O 2p-orbital (perpendicular to the O-H bond) in the xy plane π donating into the Cu(II) dx²-y² ½ occupied orbital. This results in the β LUMO in Figure 6 left which is the frontier molecular orbital important for the reactivity as explored below.

Figure 6.

β LUMO of [CuOH]⁺ (left) and β LUMO and β LUMO+1 of [Cu₂O]²⁺ (right). Mulliken orbital contributions given for each orbital. (Color scheme: red = O, pink = Al, gray = Si, mustard = Cu).

2.5. Reaction Coordinate Calculations: Comparison of [CuOH]⁺ and [Cu₂O]²⁺ Zeolite Active Sites

Having a well-defined mononuclear Cu(II) site and its experimental reactivity with CH₄, we can now evaluate why its activation barrier for this reaction is almost 10 kcal/mol higher experimentally than that of the binuclear [Cu₂O]²⁺ site defined previously.^12–14 Second-sphere effects have been demonstrated to play a significant role in CH₄ activation in zeolites³⁶, so larger models (83 atoms for [CuOH]⁺ and 144 for [Cu₂O]²⁺) that include the complete second-sphere around the active site were used to evaluate CH₄ reactivity (see Section S1.11). Unlike the [Cu₂O]²⁺ site which has two oxidizing equivalents to take methane fully to methanol, the [CuOH]⁺ active site requires an additional site for methane oxidation. Snyder et al. demonstrated in the large pore *BEA iron zeolite that the methyl radical generated by HAA escapes the active site and reacts with a second active site.³⁷ In the case of [CuOH]⁺ in large pore MOR, cage escape and reaction of a methyl radical with a second [CuOH]⁺ site is necessary to fully generate methanol and would be consistent with the 0.33±0.14 mol methanol/mol MOR-A produced (with some overoxidation). The sequential reactions of two [CuOH]⁺ sites with methane and then the methyl radical generated in the first reaction were calculated using DFT (Figure 7A). The 1^st [CuOH]⁺ site performs HAA on CH₄, the resulting methyl radical migrates from hydrogen bonding to the [CuOH₂]⁺ 1^st product site into the large channel of MOR and subsequently reacts with a 2^nd [CuOH]⁺ site. This generates two Cu(I) sites, a [Cu(H₂O)]⁺ and a [Cu(CH₃OH)]⁺, in an overall exothermic reaction (ΔH = −61.2 kcal/mol of CH₄). In this reaction coordinate, the barrier to HAA is 27.7 kcal/mol while the barrier for migration of the methyl radical to exothermically react with the second [CuOH]⁺ site is 8.4 (33 minus 24.6) kcal/mol.

Figure 7.

DFT calculated H-atom abstraction reaction coordinate. (A) Two [CuOH]⁺ sites in MOR oxidizing methane to methanol with the methyl radical traveling through the zeolite channel between them. (B) Comparison between the [CuOH]⁺ in MOR (green) and the [Cu₂O]²⁺ in CHA (purple) with the final methyl radical in vacuum to compare adsorption effects. Reactant complex energies are set to zero.

The HAA reaction coordinate for the [CuOH]⁺ site in MOR is compared to the previously calculated reaction coordinate for the [Cu₂O]²⁺ site in CHA¹⁴ in Figure 7B. The two sites have similar methane adsorption energies to the active site (9±0.6 kcal/mol, Figure 7B far left to reactant complex) and similar methyl radical adsorption energies to the H-atom abstracted first product sites (12.6±0.1 kcal/mol, Figure 7B far right to product complexes), ruling out adsorption differences contributing to the barrier difference. However, in going from the reactant complex to the transition state, the [CuOH]⁺ site has a calculated 10.8 kcal/mol higher energy barrier, about the same as the 9.5 kcal/mol barrier difference observed experimentally. Marcus theory was used to extract the intrinsic barrier, i.e. the barrier with no thermodynamic driving force, using the method from Srnec et al..³⁸ The intrinsic barrier for the [CuOH]⁺ and [Cu₂O]²⁺ sites are about the same (11.8±0.6 kcal/mol) (Table 2). Thus, the difference in ΔH^‡ reflects the thermodynamic difference in the HAA reaction, a result of a difference in O-H bond strength in the HAA products (Table 2 last column; determined from the extremes of Figure 7B relative to the calculated C-H bond strength of CH₄ with the same functional and basis set). The O-H bond strength of [Cu₂OH]²⁺ is ~15 kcal/mol higher than the O-H bond strength of [CuOH₂]⁺. The origin of this difference is discussed below.

Table 2.

DFT calculated H-atom abstraction transition state energies, Marcus theory corrected intrinsic barriers, and product OH bond strength (calculated with respect to the 103.3 kcal/mol calculated C-H bond strength of CH₄) for [CuOH]⁺ in MOR and [Cu₂O]²⁺ in CHA.

ΔH (kcal/mol)	HAA TS	Intrinsic Barrier	Product OH Bond Strength
[CuOH]+	27.7	12.3	74.6
[Cu2O]2+	16.9	11.2	89.8

3. Discussion

In past studies, autoreduced Cu-MOR was observed to produce methanol upon reaction with methane.^15,16 In this study, we combine EPR, XAS, DR UV-Vis, MCD, and rR spectroscopies to define the active site for this reaction as a mononuclear [CuOH]⁺ center bound bidentate to the zeolite lattice (Figure 5 and Table 3). This is distinct from the [Cu₂O]²⁺ active site which is reduced upon heating in helium to bicuprous sites that are reactivated by N₂O or O₂.^12–14

Table 3.

Summary of spectroscopic features and CH₄ reactivity of the [CuOH]⁺ (MOR-A) active site.

	[CuOH]+
g values	g‖ = 2.27, g⊥ = 2.09
A values (cm −1 )	A‖ = 191 × 10−4 A⊥ = 18–42 × 10−4
Cu-O stretch (cm −1 )	750
Other rR vibrations (cm −1 )	552, 981
OH− to Cu(II) dx2-y2 CT band (cm−1)	20,700
LF transitions (cm −1 )	8,900–16,500
CH4 ΔH‡ (kcal/mol)	15.7
CH4 ΔS‡ (cal/mol*K)	−39.5
Percentage of [CuOH]+ in total Cu	30±12%

We also identified this site in Cu-MFI with similar EPR parameters and a similar Cu-O stretch despite the very different topological structure of the MFI lattice. In both Cu-MOR and Cu-MFI the [CuOH]⁺ binds bidentate to a single T-site. However, this is not the case in all topologies. In In LTA and FAU³⁹ for example no EPR parameters corresponding to the [CuOH]⁺ site assigned here are observed. We speculate that in lattices with an adjacent aluminate and silicate T-site positioned to bind the copper tridentate, different mononuclear Cu^II sites could form with different spectroscopic handles.

Additionally, because of the ability to selectively reduce the [Cu₂O]²⁺ site through heating in helium, we are able to isolate the single [CuOH]⁺ site as the sole source of methanol production in autoreduced Cu-MOR. This contrasts with the N₂O activated zeolite which still has the [CuOH]⁺ site but also has [Cu₂O]²⁺ active sites. This is also true in N₂O activated Cu-MFI (Figures S15–17).¹² This suggests that analyzing autoreduced Cu zeolites, in addition to the activated materials, is important in elucidating active site contributions to reactivity. This also provides further support for the use of site-selective techniques like resonance Raman and high-resolution techniques like EPR for analyzing active sites in Cu zeolites. In the case of N₂O activated Cu-MOR and Cu-MFI bulk techniques like EXAFS contain signals from multiple Cu sites simultaneously and limit the ability to parse for relevant information.

Although this [CuOH]⁺ site has been invoked multiple times in the literature, there have been competing proposals of reactivity.^19,20 We find that the [CuOH]⁺ site has a 10 kcal/mol higher barrier for HAA from CH₄ than the comparably well-defined [Cu₂O]²⁺ site. Through computational modeling, we find that these two sites have similar intrinsic barriers. This is a result of the similarly high covalency of their frontier molecular orbitals: [CuOH]⁺ (27% O character in the β LUMO) and [Cu₂O]²⁺ (20% O character in the β LUMO and 33% O character in the β LUMO+1) (Figure 6). This allows strong orbital overlap with the H-C bond of CH₄ for H-atom abstraction. The low coordination number enforced by the Al T-sites of the zeolite lattice results in strong hydroxo and oxo ligand to Cu(II) π-donor bonds for both motifs. This is amplified by the relatively weak donation by the zeolite Al T-site bidentate ligation. In the case of the mononuclear MOR-A site studied here, the strong Cu-OH π bond is revealed by its high energy Cu-O stretch and its low energy ligand to metal CT transition.

Thus, given the similar intrinsic barriers, the difference in ΔH^‡ reflects the thermodynamic difference in the HAA reaction, a result of the difference in O-H bond strength in the [CuOH₂]⁺ and [Cu₂OH]²⁺ first products. To understand the origin of the O-H bond strength difference, the H-atom transfer was separated into reduction and protonation steps (Figure S22). The [Cu₂O]²⁺ site is 16 kcal/mol easier to reduce than the [CuOH]⁺ site, about equal to the enthalpy difference calculated for addition of the H-atom to each site by HAA. The resulting reduced [Cu₂O]⁺ and [CuOH]⁺ sites have roughly the same proton affinities due in part to both having a Mulliken charge of −0.9±0.05 on the extralattice oxygen ligand. Thus, the thermodynamic difference in the reaction reflects the greater electron affinity of the oxo bridged binuclear as compared to the mononuclear [CuOH]⁺ site.

This higher affinity in the binuclear site appears to correlate with the delocalization of the added electron from HAA over the two Cu centers as the unpaired electron is equally shared over both copper ions (Figure S23C). To evaluate the impact of this delocalization on the electron affinity, a model [CuOZn]²⁺ zeolite site was generated by replacing one of the Cu atoms in the [Cu₂O]²⁺ site with a closed shell d¹⁰ Zn²⁺. This leaves the single d⁹ Cu²⁺ to accept the electron and localizes the added electron only on one metal in the reduced structure. The electron affinity of this model is compared to the mononuclear and binuclear Cu structures below:

$$\begin{gathered} {\left[{\text{Cu}}_2\text{O}\right]}^{2+}+{\text{e}}^{-}\to {\left[{\text{Cu}}_2\text{O}\right]}^{+} \Delta \Delta \text{H}=0\left(\text{defined}\right) \\ {\left[\text{CuOH}\right]}^{+}+{\text{e}}^{-}\to \left[\text{CuOH}\right] \Delta \Delta \text{H}=16\text{ kcal}/\text{mol} \\ {\left[\text{CuOZn}\right]}^{2+}+{\text{e}}^{-}\to {\left[\text{CuOZn}\right]}^{+} \Delta \Delta \text{H}=20\text{ kcal}/\text{mol} \end{gathered}$$ (Equation 2)

Localizing the electron in the binuclear structure lowers the electron affinity by 20 kcal/mol relative to [Cu₂O]²⁺. The [CuOH]⁺ site is actually easier to reduce than the [CuOZn]²⁺ structure due to the greater antibonding character of the oxo-Cu bridge (Figure S24) and electronic relaxation effects (geometric relaxation has a negligible effect). This indicates that delocalization of the electron over the two Cu^1.5 ions substantially increases the electron affinity of the [Cu₂O]²⁺ site.

To quantify this delocalization contribution to the electron affinity in [Cu₂O]²⁺, Piepho, Krausz, Schatz (PKS) theory for the potential energy surfaces of mixed valence sites was applied.^40,41 In PKS theory, mixed-valent structures have competing stabilizing contributions. The electronic coupling between the two metal ions (H_ab) creates stabilization through delocalization of the unpaired electron, favoring a totally symmetric structure with half integer oxidation states. This is opposed by vibronic trapping (λ) which lowers the energy as the structure becomes asymmetric to trap the electron on one metal. Here, one metal’s bonds contract, and the others’ expand, localizing the electron on the metal with longer bonds. These competing contributions were evaluated using DFT calculations to map the potential energy surface along a vibrational mode where the structure distorts from a symmetric delocalized structure to the asymmetric localized structure. This Q_- vibrational mode was created by optimizing the one electron reduced [CuOZn]⁺ structure where the additional electron is localized on the Cu and subsequently replacing the Zn(II) with Cu(II) without optimization. This creates the [Cu₂O]⁺ asymmetric structural limit in this vibration. To complete the vibrational mode, the change in atomic positions from the delocalized to the localized structure was used to create points of varying distortion (example structures in Figure S25). The DFT calculated energies of the points along this vibration are given by the solid squares in Figure 8. The DFT calculated potential energy surface was then fit with a potential energy equation that includes the electronic coupling associated with the electron delocalization (H_ab) and vibronic trapping that localizes the electron one center (λ), where k_- is the vibrational force constant in the Q_- mode and $\text{x}=\left(\frac{{\text{k}}_{-}}{\lambda }\right){\text{Q}}_{-}$ (Equation 3).

Figure 8.

DFT calculated change in energy of [Cu₂O]⁺ upon distortion along the Q_- mode (black squares). Fit of black squares to Equation 3 (purple). Plot of potential energy surface with parameters from fit with H_ab set to zero to simulate localized structure (green).

$$\text{E}=\frac{1}{2}\left(\frac{{\lambda }^2}{{\text{k}}_{-}}\right){\text{x}}^2-\sqrt{\frac{1}{2}{\left(\frac{{\lambda }^2}{{\text{k}}_{-}}\right)}^2{\text{x}}^2+ {\text{H}}_{\text{ab}}^2}$$ (Equation 3)

From the purple fit in Figure 8, the derived H_ab term is 29 kcal/mol which overcomes the vibronic trapping, resulting in the minimum at the symmetric delocalized structure. To estimate the effect of the delocalization on the total energy, the H_ab term was then set to zero, and the potential energy surfaces were re-plotted to simulate the localized structures (Figure 8 green). This surface contains two local minima resulting from vibronic trapping of the extra e⁻ on either Cu with energies at +17 kcal/mol. Thus, the electron delocalization stabilizes the HAA first product by 17 kcal/mol and without this delocalization the electron affinity of the [Cu₂O]²⁺ site would actually be lower than that of the [CuOH]⁺ site by 1 kcal/mol (Equation 2).

This electron delocalization is a direct result of the superexchange interaction between the two copper ions through the bridging oxo ligand. As seen in Figure 9 (extracted from DFT contours given in Figure S23) the 3dx²-y² orbitals on the two Cu ions form symmetric and antisymmetric combinations. The symmetric combination undergoes a π bonding/antibonding interaction with the bridging oxo 2p-orbital (bottom and top molecular orbitals) while the antisymmetric combination remains non-bonding (middle orbital). The energy of excitation from the non-bonding orbital (middle) to the antibonding symmetric β LUMO (top) is equal to 2 H_ab. TD-DFT calculations predict this transition at 18400 cm⁻¹, giving an H_ab of 27 kcal/mol, in reasonable agreement with the value obtained from mapping the potential energy surface in Figure 8 (29 kcal/mol). Thus, the strong π donor oxo bridge facilitates the electronic interaction between the copper ions, delocalizes the added electron gained from reaction with CH₄, lowers the reaction barrier, and enables low temperature methane oxidation. Four [Cu₂O]²⁺ sites have been rigorously defined so far.^10,12,13,36 For all of these sites, the DFT models of the HAA [Cu₂OH]²⁺ product have significant delocalization of the added electron across both copper centers (Table S3). Thus, product electron delocalization appears to be a general driving force in this active site motif.

Figure 9.

Molecular orbital diagram for the bonding interaction of 2 Cu 3dx²-y² orbitals with the bridging oxo 2p-orbital. The highest energy orbital is ½ occupied while the lower energy orbitals are fully occupied.

4. Conclusion

Electronic coupling between metals in mixed-valent systems has enabled rapid electron transfer in bioinorganic chemistry, the synthesis of new molecular electronics, and other important chemical phenomena.^42–44 Here, we illuminate a new application of mixed-valent metal chemistry: improving methane oxidation and catalysis. To make copper containing zeolites industrially viable for methane oxidation, a high density of low temperature reactive active sites is necessary. Although there have been conflicting reports in the literature on the reactivity of a mononuclear [CuOH]⁺ site, here we elucidate that the site is clearly less reactive than the binuclear [Cu₂O]²⁺ site. Thus, better zeolite design should concentrate on forming bridged multinuclear copper species rather than monomeric sites. We also suggest that a bridged multinuclear Cu site would potentially allow the low temperature reactivity of pMMO by delocalizing the electron gained from HAA and driving the reaction. This delocalization principle also applies to design of industrial catalysts generally. New active sites in redox catalysis could benefit from increased metal nuclearity with a strong bridging ligand, facilitating electronic coupling, allowing electron delocalization, and ultimately enabling low temperature reactivity.