Hydrogen Bonds Dictate O₂ Capture and Release within a Zinc Tripod

Abstract

Six directed hydrogen bonding (H-bonding) interactions allow for the reversible capture and reduction of dioxygen to a trans-1,2-peroxo within a tripodal zinc(II) framework. Spectroscopic studies of the dizinc peroxides, as well as on model zinc diazides, suggest H-bonding contributions serve a dominant role for the binding/activation of these small molecules.

The reversible capture of dioxygen as peroxide (O₂²⁻) is required for myriad biological and abiological reactions spanning from oxidases to metal-air batteries.¹ Biological systems leverage well-positioned secondary coordination sphere interactions, such as hydrogen bonding (H-bonding), within their active metal site(s) to achieve selective O₂ binding, activation, and transfer.² To emulate this principle, recent synthetic systems have demonstrated that H-bonds can facilitate O₂ capture in the superoxo or peroxo state, although binding is typically coupled with a metal-based redox event.³ In contrast, capture of the O₂ unit using H-bonds as the primary binding interaction is very rare,⁴ and was recently enabled by six directed H-bonds within a cryptand-type macrocycle.⁵ In this case, O₂ capture was facilitated by preorganization of a molecular capsule with a binding pocket size-matched for a diatom. Given that host/guest inclusion is highly sensitive to size complementarity,⁶ the use of a preassembled binding pocket complicates the role that the H-bonds serve in capture and/or stabilization. An open tripodal ligand provides one test of whether preassembly is required to capture O₂ using H-bonds without a redox-active metal.

Our group is working to evaluate how the precise structural, electronic, and cooperative modes in the secondary coordination sphere can be used to regulate reactivity.⁷ Recently, we reported a family of p-substituted tris(6-(p-R-phenylamino-2-pyridylmethyl)amine ligands (L^R, R = CF₃, H, OMe) that provide electronically tunable −NHAr H-bond donors in the secondary coordination sphere.^7f This ligand framework provided the first structural characterization of an H-bonded (trans-1,2-peroxo)dicopper complex in which a combination of six H-bonds and two Cu(II) centers encapsulate the O₂²⁻ unit. We hypothesized that if the six H-bonds to peroxide were key factors that allowed isolation, then the templating metal and the reductant could be separated. Zinc(II), the ubiquitous redox-inactive d-block metal, was selected to test this hypothesis (Fig. 1). Tripodal ligand scaffolds containing H-bond donors within the secondary coordination sphere have been recently popularized⁸ and have previously been templated on zinc.⁹ Zn₂O₂ fragments remain exceedingly rare and are limited to polymetallic aggregates with greater than two metals.¹⁰ Herein, we report dioxygen capture and reduction enabled by H-bonds via (trans-1,2-peroxo)dizinc complexes as well as the reverse—dioxygen release.

Figure 1.

Reversible O₂ capture and reduction enabled by H-bonding interactions.

An H-bonded dizinc peroxide was obtained from dioxygen and reductant in the presence of [ZnL^H]²⁺. Saturation of an equimolar MeCN solution of L^H and Zn(OTf)₂ with O₂ followed by rapid addition of bis(cyclopentadienyl)cobalt(II) at room-temperature afforded the (trans-1,2-peroxo)dizinc complex, [(L^H)₂ Zn₂ O₂][OTf]₂ (1^H), in 89% yield (Fig. 2).¹¹ An alternative synthesis of 1^H was also developed: sequential addition of H₂O₂ and NⁱPr₂Et to a MeCN solution of L^H and Zn(OTf)₂ afforded 1^H in 58% yield.^12–13 1^H is thermally stable in CD₃CN for >18 h at 50 °C in an inert atmosphere and is also moisture tolerant.¹⁴ However, the addition of a competitive H-bond acceptor (i.e. chloride) induces degradation.¹⁵

Figure 2.

Synthesis of 1^R and molecular structure of 1^H (50% probability ellipsoids) with bond distances of 1^R.

The structural metrics of 1^H were elucidated by single-crystal X-ray diffraction (XRD) and revealed a trans-1,2-peroxo binding mode with six directed H-bonds to the O₂²⁻ fragment, establishing the first example of a discrete (trans-1,2-peroxo)dizinc species. For each ‘(L^R)Zn’ fragment, one −NHPh group engages in H-bonding interactions with the proximal oxygen (N-O_proximal = 2.738 Å; N-H-O = 158.97°) while two engage the distal oxygen (N-O_distal = 2.837 Å (for both); N-H-O = 176.16 and 169.98°). The N-O distances and N-H-O angles are consistent with moderate-strength H-bonding interactions.¹⁶ The N-C_pyr distances range 1.3587(14)-1.3675(14) Å and are consistent with single-bonds where the N-H has not been deprotonated.¹⁷ The O₂ motif (O1-O1’ = 1.4954(13) Å) is more reduced than dioxygen¹⁸ or superoxide,¹⁹ and comparable to main-group trans-1,2-peroxides including [O₂ (B(C₆ F₅)₃)₂]²⁻ (1.488 Å)²⁰ and L₂B₂O₂ ²¹ (1.484 Å; L = subporphyrin).²² Notably, the O-O distance is the same as isostructural [(L^H)₂ Cu₂O₂]²⁺ (1.477(5) Å), which suggests that the ligand scaffold itself may serve a key role in regulating the structure, rather than the metal.

To interrogate the requirement of H-bonding for peroxide capture, we evaluated an isosteric ligand variant, tris(6-phenoxy-2-pyridylmethyl)amine (TPA^OPh), that does not contain H-bond donors. In contrast to L^H, when TPA^OPh and Zn(OTf)₂ were combined and treated with H₂O₂ and ⁱPr₂NEt, we observed demetalation, rather than the formation of a (trans-1,2-peroxo)dizinc (see SI). Attempts to synthesize the zinc analogue of Karlin’s [(TPA)₂ Cu₂ O₂]²⁺ (TPA = tris(2-methylpyridylamine)²³ resulted in products derived from [(TPA)Zn(OH)]⁺.^24–25 These results highlight the synergistic effect of H-bonds with the Zn center in 1^R to both capture and stabilize O₂²⁻.

Given the role of H-bonds for O₂²⁻ capture, we assessed the extent to which activation of the O–O unit could be tuned by H-bond donor strength. We selected ligand variants that feature identical steric properties surrounding the Zn₂O₂ core, yet vary in acidity of the NH, and thus H-bond donor strength. Given the highly coupled ligand structure, substituent modification will necessarily alter both H-bond donor strength and ligand electronics.²⁶ For example, an electron-withdrawing aniline (e.g. p-CF₃) will afford a better H-bond donor at the expense of a weaker ligand donor strength. Four para-substituted anilines with Hammett substituent (σ_p) constants ranging −0.83 (R = NMe₂) to 0.54 (R = CF₃)²⁷ were used to prepare the series of (trans-1,2-peroxo)dizinc complexes (1^R; R = NMe₂, OMe, CF₃; Fig. 2).

The electronic environment provided by each ligand variant in 1^R tracked with the methylene resonances (coupled doublets) of the C₃-symmetric ¹H NMR spectra. For example, electron-deficient 1^CF3 exhibits the most downfield resonances (4.22 and 4.08 ppm), while the more electron-rich 1^NMe2 features the most upfield resonances (4.00 and 3.86 ppm). These resonances show a good correlation when plotted against Hammett constants (see SI) and provide a descriptive measure of electronic environment provided by each ligand scaffold. In contrast to the direct relationship between ligand electronic environment and the methylene resonances, the −NH resonance involved in H-bonding interactions, which is a composite of −NH—O_proximal and −NH—O_distal, does not exhibit the same trend. 1^H displays the most downfield shift (10.21 ppm) while 1^OMe and 1^NMe2 exhibit the same shift (10.14 ppm). The absence of a clear trend contrasts with the previously reported series of (L^R)CuCl,^7f where the −NH resonances exhibit a linear correlation with Hammett constants, and suggests the −NH—O interactions in 1 are not adequately described by simple H-bond donor/acceptor contributions, but may also be influenced by the electronics at zinc.

To probe the origin of the ¹H NMR discrepancies, the structural metrics of 1^R were examined by single-crystal XRD. Complexes 1^CF3, 1^OMe, and 1^NMe2 are isostructural to 1^H with six directed H-bonds to the peroxide. The electronic substitutions have negligible consequence on the ability of zinc to dimerize about the O₂²⁻ unit—the Zn-Zn distances range from 4.719 (1^CF3) to 4.784 (1^OMe) Å. The Zn-O distance reports on the electronic environment of the TPA-ligand: the most electron-deficient variant, 1^CF3, displays the shortest Zn-O distance while the most electron-rich variant, 1^NMe2, contains the longest (1.9507(16) and 1.991(3) Å, respectively). The H-bonding interactions within the four species are comparable with N-O_proximal and N-O_distal distances ranging from 2.619–2.741 and 2.811–2.945 Å, respectively.²⁸ Within the series of complexes, the O–O bond length exhibits a variable extent of activation and ranges from 1.483(6) (1^NMe2) to 1.524(3) Å (1^CF3). Notably, the O–O bond in 1^R does not correlate with either the electronic character of the TPA-ligand—as assessed by Hammett constants of para-aniline substitution—or the H-bond donor strength. This directly contrasts the [(L^R)₂Cu₂O₂]²⁺ series whose LMCT-energy correlated to TPA-ligand electronics,^7f further suggesting multiple competing factors contribute to the overall description of the Zn₂O₂ unit.

Oxidation studies were pursued to assess the reversibility of H-bond mediated O₂ capture.²⁹ Addition of PhICl₂ to 1^H cleanly produces [(L^H)ZnCl][(OTf)] (76% yield) concomitant with gas evolution. A Clark electrode was used to confirm O₂ release during oxidation. Because this analytical technique requires aqueous conditions, the reaction was repeated with a water-soluble oxidant. Injection of 1^H into an aqueous solution containing [NH₄]₂[Ce(NO₃)₆] and [Bu₄N][Cl] caused a rapid increase in dissolved O₂, which reached a plateau after 3 minutes, corresponding to approximately 47% yield (Fig. 3, see SI). [(L^H)ZnCl][(OTf)] was formed as the (L^H)Zn-containing compound in a similar isolated yield as O₂ (40%).³⁰ These studies confirm that capture and release is dictated by the direction of electron flow in solutions containing [(L^H)Zn]²⁺.

Figure 3.

A) Oxidative release of dioxygen from 1^H. B) Dioxygen evolution trace detected by Clark electrode. C) Molecular structure of [(L^H)ZnCl]⁺ (50% probability ellipsoids).

Azide is a spectroscopic analogue for peroxide due to its similar frontier orbital manifold.³¹ In contrast to peroxo-units whose vibrational modes can be challenging to identify,³² metal-azides feature intense bands that are sensitive to electronic perturbations. We thus sought to interrogate the electronic and H-bonding contributions imparted on the O₂²⁻ or N₃⁻ unit within a set of isostructural (L^R)Zn complexes.³³ Octahedral (L^R)Zn(N₃)₂ complexes (2^R) were targeted because they feature both axial and equatorial azide environments, and only the axial azide can engage in H-bonding interactions. Complexes 2^R were synthesized by treating an acetone solution of L^R and Zn(ClO₄)₂·6H₂O with excess NaN₃ (Fig. 4). The ¹H NMR spectra exhibit downfield −NH resonances, consistent with H-bonding, which are dependent on TPA-ligand electronics. The furthest downfield −NH and methylene resonances (δ = 9.36 and 4.16, respectively) correspond to the most electron-deficient variant, 2^CF3, while the furthest upfield −NH and methylene resonances (δ = 8.90 and 4.05 ppm respectively), correspond to the most electron-rich complex, 2^NMe2.³⁴ The trend of the −NH resonance position contrasts with the series of 1^R (a composite of −NH—O_proximal and −NH—O_distal interactions) but is analogous to the chloride series, (L^R)CuCl.^7f

Figure 4.

Synthesis of 2^R and molecular structure of 2^H (30% probability ellipsoids).

Molecular structures for 2^R were determined by single-crystal XRD and establish an octahedral geometry in which the axial azide is engaged in H-bonding to the aniline-NH groups. All three H-bonding interactions are directed to the α-nitrogen of the azide; for 2^NMe2, the N_H-N_azide distances range 2.919–3.049 Å—consistent with moderate-strength H-bonding interactions.¹⁶ The Zn-N_{3(equatorial)} bond lengths reflect the electronic perturbations of the TPA-ligand (2^CF3 = 2.063(3); 2^NMe2 = 2.1357(15) Å); in contrast, minimal variation is displayed within the Zn-N_3(axial) bond (2^CF3 = 2.055(3); 2^NMe2 = 2.0741(15) Å). Furthermore, the bond distances for the previously reported (TPA)Zn(E₃)₂ (E₃ = -N₃³⁵ or -NCS³⁶) compounds are consistent with the equatorial, but not axial azide units of 2^R. This disparity suggests that the polarization of the axial azide ligand in 2^R may be governed by H-bonding interactions rather than the overall electronic environment of the TPA-ligand. Vibrational spectroscopy was employed as a complementary metric to decipher azide polarization (similar to gauging CO activation of M-CO species),³⁷ given the low precision in experimentally determined N—N distances.

The independent impact of H-bonding interactions and electronic character in 2^R was evident by solid-state IR spectroscopy.³⁸ Each complex displays two distinct ν_(N3)asymm modes that were identified by DFT analysis as the ν_(N3)-equatorial and ν_(N3) -axial.³⁹ In each complex, the H-bonded axial-azide is shifted to higher energy than that of the equatorial azide. Plots of the ν_(N3)-equatorial and ν_(N3)-axial shift for each complex verses their Hammett substituent constant (Figure 5) exhibit different slopes. The energies of ν_(N3)-axial (with H-bonding) are minimally perturbed across the series of complexes (2^CF3 = 2076; 2^NMe2 = 2069;Δ = 7 cm⁻¹) as compared to the energies of ν_(N3)-equatorial (without H-bonding) (2^CF3 = 2038; 2^NMe2 = 2057; Δ = 19 cm⁻¹) and is consistent with the trend in crystallographically determined Zn–N_{3(equatorial)} bond distances. We propose that the H-bonds in 2^R serve as the primary activating interaction for the axial azide, similar to previously reported (L^OH)CuF (L^OH = tris(6-hydroxy-2-methylpyridyl)amine), where halide (F-) binding was dictated by H-bonds, rather than a Cu(I)-F interaction.⁴⁰ By extension, the intramolecular H-bonding interactions in 2^R attenuate the electronic influence of ligand variation by reducing the covalency between Zn(II) and azide. We propose that this same phenomenon can be applied to rationalize the spectroscopy of 1^R. The non-linear correlation of the O-O bond distances as well as the ¹H NMR −NH resonances, with respect to Hammett constants, both imply the six H-bonding interactions to the O₂²⁻ unit play a greater role in substrate activation than the electronic influence of the supporting TPA-ligand acting on the zinc center.

Figure 5.

Linear free energy relationship of ν_(N3) (neat, ATR) and Hammett constants for 2^R.

We have demonstrated that an H-bond appended tripodal zinc complex assembles to capture peroxide derived from dioxygen and electrons. This is the first example of a discrete (trans-1,2-peroxo)dizinc complex and its isolation was facilitated by H-bonding interactions. The captured peroxide can be liberated as dioxygen via two-electron chemical oxidation. The TPA derivatives, tris(6-(p-R-phenylamino-2-pyridylmethyl)amine, provide tunable secondary sphere H-bond donors that augment the stabilization of otherwise unstable Zn₂O₂ units. Analysis of a series of related zinc-diazide complexes revealed that H-bonding interactions serve as the primary component responsible for substrate capture and activation in the absence of a redox-active metal.