Introduction

Proton-coupled electron transfer (PCET) reactions of small molecules are central to prospective energy conversion and storage schemes that promise to replace carbon-based fuel sources1,2. For example, water splitting to make oxygen and hydrogen requires orchestrated removal of 2H+ and 2e from the H2O molecule despite the large bond dissociation free energy (BDFE) of its O-H bonds (113.0 kcal/mol)3. In the context of fuel cell technologies, the optimal catalysts for this anodic water oxidation are IrOx nanomaterials4 that rely on a precious metal, iridium, with insufficient earth abundance to support the global-scale energy economy5. Similarly, prospective use of ammonia as a clean energy source requires three PCET events per NH3 molecule despite its large N-H BDFE (100.3 kcal/mol)6,7. PCET reactions also have relevance to frontier areas of organic synthesis8,9,10. Ideally, catalysts could facilitate PCET by weakening X-H bonds (X = OH or NH2) through chemical interactions with the small molecules. However, coordination of H2O or NH3 to most metal ions induces acidification but not bond weakening, i.e., H+ transfer but not accompanying e- transfer needed for energy transduction. For example, the classic Werner complex, [Co(NH3)6]3+, has a pKa of 13 that is significantly lower than that of ammonia11, yet it maintains a high BDFEN-H of 105 kcal/mol12. Instances of coordination-induced bond weakening13 (CIBW) often involve acidification of the X-H bond by coordination to a highly reduced metal center, e.g. TiIII, MoI, BiII12,14,15,16, thus enabling H+ transfer from the acidified X-H ligand coupled to e- transfer from the reducing metal ion during net PCET (Fig. 1a)3. In an extreme case, the BDFEO-H of [Sm(OH2)n]2+ has been estimated to be in the 26-39 kcal/mol range17,18, and similar behavior is observed upon coordination of NH3 to SmII19.

Fig. 1: Coordination-induced bond weakening (CIBW) motifs.
figure 1

a Traditional motif involving a reducing and Lewis acidic metal ion, (b) the multisite PCET motif evident in the OEC of photosystem-II, (c) lack of weakening with typical AlIII ions, (d) the multisite PCET scheme reported here relying on redox non-innocence of the ligand. BD(F)EO-H values for Al-OH2 complexes were estimated using DFT calculations.

Given the importance of PCET to renewable energy, it is critical to consider how CIBW can be implemented using the most earth-abundant metals, which in some cases do not readily access low-valent states. In this regard, it is noteworthy that water oxidation during photosynthesis does not follow the CIBW paradigm described above but rather follows a multisite PCET paradigm. Here, H2O coordinates to a redox-innocent Ca2+ ion that is itself incorporated into a redox-active manganese-oxo cluster that collectively forms the oxygen-evolving complex (OEC) of photosystem-II20,21. As such, PCET at the OEC involves acidification of H2O by the Ca2+ Lewis acid combined with e- transfer from the [Mn4] electron reservoir (Fig. 1b), rather than co-localizing H+ and e transfer functions at a single metal center.

The most abundant metal on earth is aluminum. Although the common Al3+ ion acidifies water dramatically (pKa = 5.0), typical AlIII compounds maintain high BDFEO-H values that are typically >90 kcal/mol. (Fig. 1c) Our group recently discovered that the heterobinuclear complex, LAl(Me)Fp (1, L- = [HC(CMeNdipp)2]-, dipp = 2,6-di-iso-propylphenyl, Fp- = [FeCp(CO)2]-), serves as a masked source of the [LAlMe] radical (A) that is formally an AlII complex but is better formulated as AlIII coordinated by the radical dianion, [L]2-22,23. Since conjugation of the AlIII Lewis acid to the [L]2- electron reservoir bears similarities to the OEC motif (Fig. 1d), we hypothesized that A would promote CIBW by an analogous, multisite PCET mechanism. Here, we demonstrate CIBW during O-H activation of water/alcohols and N-H activation of an amine at ambient conditions. Not only does this represent a rare example of CIBW by the most abundant metal on earth24,25,26, but also computational modeling suggests an unusual degree of CIBW. Most importantly, this discovery points to redox non-innocence27,28,29 as a useful design element for enabling PCET with earth’s most abundant metals like aluminum and calcium.

Results

Rapid reactions between 1 and O-H substrates (i.e., H2O and alcohols) or N-H substrates (i.e., NH3 and amines) were observed at ambient conditions (vide infra). Thus, the question of reaction pathway immediately arose. Our previous works provided experimental and computational evidence that 1 dissociates reversibly at ambient conditions by Al-Fe homolysis, producing small equilibrium concentrations of the A/Fp frustrated radical pair (FRP)30,31,32 that can cooperatively activate O-coordinating substrates22,23. The majority of available mechanistic data probed CIBW of the C=O π-bonds in CO2 and O = CPh2, the latter of which allowed for spectroscopic characterization of the [LAl(Me)(OCPh2)] radical due to stabilizing delocalization of the unpaired spin into the benzophenone π-system22. We can now report that this behavior extends to nitrogen-containing π-systems, as well, providing evidence that the FRP is capable of engaging substrates with N-coordinating groups. Addition of pyridine to 1 produces Fp2 along with C-C coupled dialuminum complex 2 (Fig. 2a), which presumably forms via pyridine adduct B that places significant unpaired spin density in the pyridine π-system and triggers diradical coupling at the 4-position. The X-ray crystal structure of 2 indicates localized π-bonding consistent with disruption of pyridine aromaticity: the C2-C3 and C5-C6 distances show double-bond character [1.336(3) Å] while the C3-C4 and C4-C5 distances show single-bond character [1.509(3)−1.516(4) Å], and the N-C2 and N-C6 distances are elongated [1.395(3)−1.400(3) Å] compared to pyridine (1.340 Å)33. The C4-C4’ distance is also indicative of single bonding [1.572(4) Å], consistent with the pyramidalized, C(sp3)-like geometries at these centers.

Fig. 2: Experimental evidence for Al/Fe frustrated radical pair (FRP) behavior.
figure 2

a Pyridine diradical coupling induced by coordination to [LAlMe]•, (b, c) divergent regioselectivity during ring opening of (±)-propylene oxide consistent with 3 acting as a masked frustrated Lewis pair (FLP) and 1 acting as a masked frustrated radical pair (FRP). Crystal structures are shown as thermal ellipsoids (50% probability) for key atoms and wireframes for C atoms in L. Pyridine H atoms are shown in calculated positions, and all other H atoms are omitted for clarify.

Having established that the putative FRP is reactive towards both O- and N-coordinating substrates, the next question was whether observed CIBW of π-systems extends to σ-frameworks. Since we previously showed that 1 is capable of cooperative C-O σ-bond cleavage of cyclohexene oxide22 and tetrahydrofuran23, it was implied that intermediate A induces CIBW of σ-bonds as well as π-bonds. However, far less definitive mechanistic data had been gathered in support of the FRP mechanism for these C-O σ-bond activation reactions involving substrates that lack π-systems to stabilize unpaired spin density. Thus, since σ-bond activation would also be relevant to targeted X-H cleavage by PCET, we sought to establish the FRP mechanism for C-O σ-bond substrate activation by 1. In particular, we sought to rule out an alternative, polar pathway involving dissociation of 1 by Al-Fe heterolysis to generate the [LAlMe]+/Fp- frustrated Lewis pair (FLP)32. Particularly useful in this regard is the sister compound of 1, LAl(Me)Wp (3, Wp- = [CpW(CO)3]-), which is reported here. Unlike 1, which features a direct Al-Fe bond, complex 3 lacks any direct Al-W interaction in favor of an isocarbonyl bridge (i.e. Al···O ≡ C-W, see Fig. 2b). We preliminarily view the structure of 3 as featuring the [LAlMe]+ (A+) cation having formed a dative adduct with the Wp- anion through one of the CO oxygen atoms. In other words, whereas we expected 1 to act as a masked source of the A/Fp FRP, we hypothesized that 3 would serve as a masked source of the A+/Wp- FLP in solution.

Accordingly, 1 and 3 were found to exhibit opposite regioselectivity during ring-opening C-O σ-bond activation of (±)-propylene oxide. The reaction of 3 and (±)-propylene oxide produced 4, which putatively forms via transition state C resembling SN2 attack of Wp- on epoxide coordinated to A+ (Fig. 2b). Here, the regioselectivity is consistent with well-known polar epoxide ring-opening pathways that are usually controlled by sterics and, thus, occur at the less substituted carbon center34. On the other hand, the reaction of 1 and (±)-propylene oxide produced 5, which putatively forms via D that is the result of coordination-induced C-O σ-bond cleavage by intermediate A (Fig. 2c). Here, the regioselectivity is dictated by the preference for placing radical character on a secondary rather than primary carbon center in D35. Collectively, these results are consistent with 1 following an FRP mechanism and 3 following an FLP mechanism during σ-bond activation reactions. Structures of 4 and 5 were assigned definitively by 1H NMR spectroscopy, and the assignment of 5 was verified by X-ray crystallography.

Having established that 1 can activate σ-bonds through FRP pathways, next we closely examined the reactions between 1 and substrates containing X-H bonds (Fig. 3a). At temperatures ranging from −30 °C to room temperature, rapid reactions were observed between 1 and the O-H substrates H2O, MeOH, iPrOH, and tBuOH as well as the N-H substrate iBuNH2. Monitoring these reactions by in situ NMR spectroscopy showed the formation of LAl(Me)X products 6 and stoichiometric FpH. Formation of FpH was evident in these experiments from the appearance of resonances at 4.1 ppm (Cp) and −11.7 ppm (Fe-H) in the 1H NMR spectra (see Supplementary Fig. 22)36.

Fig. 3: Examination of X-H activation pathways.
figure 3

a Experimental X-H bond activation reactions, (b) pseudo-first order rate constants (excess X-H, 283 K) determined by 1H NMR spectroscopy, (c) reaction thermodynamics computed by DFT. Gibbs free energy values are given in units of kcal/mol, dCN values are average C-N distances for each compound, and error bars (95% confidence intervals) on kobs are roughly ± 0.002 min−1.

To gain further insight into the X-H activation mechanism, we measured pseudo-first order rate constants (excess X-H, 283 K) as a function of X. Effectively no variation in reaction rate was observed across the series, nor was a kinetic isotope effect (KIE) evident when comparing MeOH and MeOD (Fig. 3b). The fact that X-H activation rate is independent of its pKa (e.g. iPrOH vs. iBuNH2) is inconsistent with the X-H cleavage involving simple H+ transfer and, instead, implies that e- transfer is involved in limiting the rate. The fact that X-H cleavage occurs independently of steric hindrance (e.g. MeOH vs. tBuOH) implies that the rate-limiting step does not involve substrate coordination. The absence of a measurable KIE indicates that X-H activation occurs after the rate-limiting step. Collecting these observations, we propose that the rate-limiting step in each of the reactions is Al-Fe cleavage from 1 to reveal the A/Fp FRP. Both CIBW of the X-H substrate and subsequent X-H cleavage by PCET would, then, occur after rate-determining formation of intermediate A. Consistent with the fact that masked FLP 3 proceeds by a different mechanism than masked FRP 1, we found that X-H cleavage by 3 does have a pKa dependence. For example, complex 3 was found to activate more acidic iPrOH but not less acidic iPrNH2.

To examine the viability of our mechanistic proposal, we used unrestricted DFT calculations (M06/def2-TZVPD//PBE1PBE/def2-SVP in toluene CPCM solvent, see Supplementary Information for details) to model key reaction steps and intermediates for H2O activation by 1 (Fig. 3c). As we calculated previously, homolytic Al-Fe cleavage of 1 to produce the A/Fp FRP is endergonic (ΔG = +28.9 kcal/mol)22. Coordination of H2O to A to produce [A·H2O] was calculated to be slightly exergonic (ΔG = −9.3 kcal/mol). Finally, net H-atom transfer from [A·H2O] to Fp to produce 6a and FpH was calculated to be highly exergonic (ΔG = −47.7 kcal/mol). Thus, the overall H2O activation process is thermodynamically favorable by approximately ΔG = −28.1 kcal/mol. The calculated endergonic Al-Fe cleavage step and subsequent exergonic steps for water coordination and H-atom transfer is consistent with experimental observations indicating that both H2O coordination and O-H bond cleavage occurring after the rate-determining step. It is noteworthy that the energy profile of H2O activation is distinct from that previously calculated for CO2 activation in that CO2 coordination to A was calculated to be slightly endergonic22. Thus, pseudo-first order rate constants previously measured22 for CO2 activation were likely composites of the two elementary rate constants for Al-Fe cleavage and CO2 coordination. On the other hand, the pseudo-first order rate constants for X-H activation in this study are more likely pure measurements of Al-Fe homolysis to unmask the key FRP intermediate.

Towards understanding the origin of CIBW in this system, next we examined the electronic configurations of key intermediates. In starting complex 1, the β-diketiminate ligand is in its closed-shell [L]- form and has correspondingly short C-N distances indicative of significant C=N double-bond character (Fig. 3c). Al-Fe homolysis produces intermediate A in which the β-diketiminate ligand is reduced to its [L]2- form, with elongated C-N distances due to population of the ligand π* manifold with additional electron density. This reduced ligand character is maintained upon H2O coordination to AlIII to form [A·H2O] according to the calculated C-N distances. Furthermore, in [A·H2O], the unpaired electron is localized on the β-diketiminate ligand’s π* system rather than at Al according to examination of its SOMO (Fig. 4) and spin density distribution. The electronic configuration of the [A·H2O]/Fp radical pair was found to depend on interaction distance. At long distances (>4 Å), the lowest-energy configuration of the [A·H2O]/Fp pair is a weakly coupled, open-shell singlet diradical (S2 > 0) where one electron remains in the β-diketiminate π* system and the other electron is localized on the Fe center. The triplet state is several kcal/mol higher in energy. At shorter interaction distances (<4 Å), it is thermodynamically favorable for e- transfer to occur such that the β-diketiminate π* electron is transferred to the Fe center to generate Fp- and closed-shell [LAl(Me)(OH2)]+. Populating this polar electronic configuration induces barrierless proton transfer (Fig. 4). Upon net hydrogen atom transfer to Fp, the contracted C-N distances calculated for 6a imply re-oxidation to the [L]- state. Thus, perhaps the best description of this PCET step is that it asynchronously couples H+ transfer from acidified H2O to e- transfer from the β-diketiminate electron reservoir. However, though these results imply that e- transfer precedes H+ transfer, the available data cannot elucidate the degree of coupled timing for e- transfer and H+ transfer along the asynchronous PCET pathway. It is worth contrasting this behavior with a recently reported computational model26 for X-H cleavage by a heterobinuclear Al-M complexes24 with diradical character25. While facile X-H cleavage processes were observed in that system, they are proposed to involve a concerted, 2e- pathways and thus do not strictly qualify as PCET reactions enabled by CIBW.

Fig. 4: Asynchronous proton-coupled electron transfer (PCET).
figure 4

Singly-occupied molecular orbital for [A·H2O] (0.03 isosurface) calculated by DFT (left), and a summary of lowest-energy electronic configurations as a function of [Al-OH2]···Fe distance (right). The compiled data indicate asynchronous PCET in which ET leads PT.

Finally, we calculated the degree of CIBW by evaluating the BDFEO-H for [A·H2O]. DFT calculations with the M06 functional gave a value of only 1.0 kcal/mol, and other DFT functionals (e.g. PBE1PBE) gave similarly small BDFE value. CCSD(T)/def2-TZVP energies on a model complex confirmed this very low bond energy for the O-H bond in [A·H2O]. These results imply that the BDFEO-H for [A·H2O] is significantly lower than that of [Sm(OH2)n]2+17,18, meaning that the degree of CIBW induced by A in the current system is without precedent. Moreover, the near-zero BDFEO-H for [A·H2O] raises another mechanistic possibility: upon H2O coordination to A, [A·H2O] can spontaneously liberate H to form 6a directly without any interaction with Fp. This hypothetical step can also be considered PCET, as it would involve loss of H+ from coordinated H2O coupled to loss of e- from the reduced β-diketiminate ligand. This alternative pathway, which is somewhat analogous to C-O cleavage induced by epoxide coordination to A (Fig. 2c), cannot be ruled out. In fact, if Al-Fe homolysis from 1 produced a solvent-caged A/Fp FRP, it is possible that both pathways (asynchronous PCET from [A·H2O] to Fp and direct H liberation from [A·H2O] followed by H/Fp recombination) are operative. Whereas DFT calculations indicate that the coordinated water molecule in [A·H2O] is not acidified significantly compared to free H2O, the degree of CIBW in the current system is driven mainly by the strong reducing potential of the 1e--reduced β-diketiminate ligand in [A·H2O] (see SI for detailed calculations).

Discussion

Homolysis at ambient conditions of Al-Fe complex 1 produces small equilibrium concentrations of a frustrated radical pair (FRP) consisting of Fp and the AlIII-containing radical species A. Previously established coordination-induced bond weakening (CIBW) of C=O π-bonds by A is now shown definitively to extend to C-O σ-bonds, representing a metal/ligand cooperative analogue of established TiIII chemistry37. This FRP behavior has been applied to CIBW of O-H and N-H σ-bonds. Experimental and computational analysis of these X-H activation reactions indicate that A induces a significant degree of CIBW. Not only is this a rare example of CIBW by aluminum, the most abundant metal on earth, but it establishes a metal/ligand cooperative paradigm for multisite PCET wherein the metal center acidifies a coordinated X-H bond while the redox non-innocent ligand conducts e- transfer in a concerted but asynchronous manner. The use of earth-abundant metals in PCET reactions will be critical to various renewable energy schemes. The results reported here will inform the molecular design features of such systems, opening CIBW to abundant but redox inert metals through metal/ligand cooperativity.

Methods

Experimental details (synthesis & characterization, reactivity studies, kinetics measurements) are provided as Supplementary Information. Computations were performed as follows. The PBE1PBE38 functional (ultrafine integration grid), def2-SVP basis set39, and conductor-like polarizable continuum model (CPCM)40 for toluene were used for geometry optimizations and vibrational frequency characterization in Gaussian 1641. Single-point energies were calculated with M0642/def2-TZVPD using ORCA43. CCSD(T) calculations were executed in ORCA. 3D structures were created using CYLview44. Further computational considerations are provided as Supplementary Information.