Coordination-induced O-H/N-H bond weakening by a redox non-innocent, aluminum-containing radical

Several renewable energy schemes aim to use the chemical bonds in abundant molecules like water and ammonia as energy reservoirs. Because the O-H and N-H bonds are quite strong (>100 kcal/mol), it is necessary to identify substances that dramatically weaken these bonds to facilitate proton-coupled electron transfer processes required for energy conversion. Usually this is accomplished through coordination-induced bond weakening by redox-active metals. However, coordination-induced bond weakening is difficult with earth’s most abundant metal, aluminum, because of its redox inertness under mild conditions. Here, we report a system that uses aluminum with a redox non-innocent ligand to achieve significant levels of coordination-induced bond weakening of O-H and N-H bonds. The multisite proton-coupled electron transfer manifold described here points to redox non-innocent ligands as a design element to open coordination-induced bond weakening chemistry to more elements in the periodic table.

The computational part needs to be improved considerably.
1) DFT calculations have been performed only for modeling the reaction intermediates in the proposed mechanism for H2O activation by 1, thus reporting only thermodynamic free energies and NOT kinetic free energy barriers, which are fundamental for identifying the rate-determining step.Indeed, the assumed RDS (homolytic Al-Fe cleavage of 1) is rather highly endothermic (ΔG = +24.5 kcal/mol), which means that the activation barrier should be expected to be even much higher (that could be not completely consistent with a reaction occurring at a temperature from -30°C to room temperature).A transition state for the PCET (3TS) has been located at a lower level of theory (which is then not consistently comparable to intermediates) on the triplet energy surface.I would expect here a spinforbidden reaction, where a spin-crossing from a singlet to a triplet PES should occur.Therefore, calculations on both the PESs would be needed to properly characterize the mechanism (at least to verify that the TS energy is lower on the triplet than on the singlet PES) and confirm the asynchronicity of the PCET process.
2) In the BDFE calculation for eq 2, an error of about 14 kcal/mol has been assessed with respect to the experimental value (63 kcal/mol), which suggests that the used computational set up is not sufficiently reliable (for instance, relativistic effects were not included, the basis set quality and functionals other than PBE0 were not tested).
3) It would be helpful to present the electronic structures or, at least, the calculated spin density of the radical species to show where the unpaired electron is (de)localized in the intermediates.The authors claim that the β-diketiminate ligand is redox non-innocent and that upon reduction the extra electron populate the ligand π* manifold.Did the authors calculate the electronic structure?Changes in intraligand bond distances cannot be used as indicative of the ligand oxidation state.At DFT level, oxidation state is a very critical issue.
4)It may be useful a comparative study with recent computational results showing that the M-Al bond is able to activate, via a concerted, diradical-like mechanism, the O-H and N-H bonds.Moreover, analogy of complex 1 with CaMn4 oxygen-evolving complex in photosystem-II should be more deeply justified.
Although the reported experimental evidence is very important in the field of small molecule activation processes, based on the above critical issues, I am afraid that I cannot recommend the acceptance of this manuscript in Nature Communications as it stands.
Mankad and coworkers present an interesting example of multisite proton coupled electron transfer that occurs through the homolylsis of an Fe-Al bond that results in an Al(III) complexed to a radical anion ligand that induces significant bond-weakening in small molecules coordinated to the Al.Overall, the work is quite interesting and the mechanistic work and analysis supports the mechanistic hypotheses in the manuscript.I am strongly in favor of acceptance, but have a number of comments for the authors to consider in advance of publication.Some of the description and citations in the introduction neglect seminal work.I recommend that either seminal work be cited in place of the work cited in the manuscript, or that it be added.
In the section describing the frontier work in PCET in synthesis, the citation of the recent Chem.Rev. by Knowles and colleagues is highly relevant, but I do not believe the work of Studer is appropriate.While certainly photocatalytic, the reaction requires the synthesis of a high MW sacrificial phosphine that limits the utility of the system.Earlier work by Knowles (J.Am.Chem. Soc., 2015, 137, 6440-6443) is really more appropriate to cite.I am not sure that I agree with the description of the basis for coordination induced bond weakening, especially the use of invariably.While in most instances coordination of a small molecule to a low-valent metal metal reduces the pKa of the bound of the bound ligand, in some cases it is quite modest.In other cases, significant X-H bond-weakening occurs at less proximate sites although the impact of increasing acidity is almost non-existent (ie coordination of amides to Ti(III), Sm(II), etc.) Sm-water bond-weakening was established by Flowers several years earlier (2015) than the cited Mayer report (J.Am.Chem. Soc. 2015, 137, 11526-11531).The estimate of bond-weakening by Flowers described in a follow-up small review (Dalton Trans. 2019, 48, 16142-16147) is smaller but consistent with the value determined by Mayer and Kolmar employing thermochemical cycles but is more relevant to synthetic systems in organic media.The value by Mayer is an overestimate since aqueous potentials and pKa's were used not relevant to the solvent employed in the reduction of an enamine in THF.I believe this was also recently pointed out by Peters and coworkers in a recent JACS publication as well.This should be corrected in the introduction and in the results and discussion section below Figure 3.It isn't necessary to get into this level of detail, but certainly the range of bond weakening based on experimental evidence can be described briefly.
In Figure 1 a, the incorrect structure is shown.Upon addition of water to low-valent titanocene, chloride ions are solvated by water and displaced to the outer sphere.A better representation is shown above and supported by EPR, voltammetry, and computational studies carried out be Gansauer and coworkers (Angew. Chem. Int. Ed., 2012, 51, 3266-3270).I recommend this structure be shown.
The discussion around Figure 1b should be referred to as multisite PCET.This terminology is wellestablished for the description of biological systems such as the OEC of photosystem II and synthetic systems.1b is also a classic example of acidification of a bound small molecule coupled to ET from a redox center in the language used to describe coordination induced bond weakening.
In the second paragraph of the results and discussion, it may be useful to cite recent work of Knowles on the coordination induced bond weakening of cyclopropanes (J.Am.Chem. Soc. 2022, 144, 34, 15488-15496) since in some ways it is analogy to the current system.
In the conclusion, the authors discuss the extension to the weakening of C-O bonds.I think this should be reworded.In the reduction of epoxides by low-valent titanocene, coordination induced bond weakening is an important feature of the first step in reduction.Although it doesn't involve X-H bond weakening, coordination of an epoxide to Ti(III) clearly weakens the C-O bond and is a classic example of coordination induced bond weakening.There needs to be more context in the description of the present Al-Fe system to differentiate from more classic systems.
Mankad and coworkers report a remarkable discovery based on coordination-induced bond weakening of protic substrates on coordination to a aluminium radical.These radical is derived from the homolysis of a Fe-Al heterometallic.The authors have provided convincing evidence not only for radical generation (which expands and builds upon prior work from the group, e.g J. Am.Chem.Soc. 2022, 3210) but also that a PCET event is at play, rather than simply an acidification of O-H or N-H bond on coordination.
The results are important and will find broad interest as they suggest a general strategy for use of H2O, NH3, simple alcohols and amines through activation at main-group radical intermediates.
Ultimately the information could inform design of new catalysts.The authors are also correct in stressing then key point of novelty that this is the first time such reactivity has been observed at aluminium, the most abundant metal in the earth's crust.
I am very supportive of publication and have the following suggestions.
(1) The narrative of the current draft is a little awkward.The work is framed in terms of PCET but the bulk of the initial results and discussion, captured in Figure 2, focuses on reactivity of heterometallics.
At times this feels likely two stories.Can the authors edit this section to reduce the amount of content, or better draw the connection to the PCET results -which are the key novelty.
(2) Additional analysis of the electronic structure of A by DFT, including a spin-density plot would help argue the ligand-based character of the radical.
(3) The estimation of the BDE of [A-H2O] radical is a little unsatisfying.I think it is appropriate to put an upper-bound on this, but the lack of correlation between the computational values and experimental data in benchmarking of model is a bit worrisome.The use of diffuse functions in the basis-set is known to be important to properly model metal hydride complexes, I would strongly suggest that the authors investigate Ahlrich's basis set that includes both polarisation and diffuse e.g.def2-QZVPPD or def2-TZVPD and compare data to those in Table S1.The functional group dependence should also be explored including both hybrid and mGGA functionals.
(4) It is not uncommon to not be able to observe 13C resonance of sites directly bonded to Al by 1D NMR, but usually these are resolved by 2D methods.Can the authors find these for 6a-e by 1H-13C HSQC experiments.
(5) Can 183W satelittles not be observed in the NMR data of 4? If not a note should be added to the SI to clarify.(6) There is a serious issue with the characterisation data reported for 6d.This compound does not contain W, something is wrong here.Please check the data and assignments.
Reviewer #4: Remarks to the Author: The structural data presented by the authors does support their conclusion, but RS15 and SS1 need refining with sensible cutoff on the data resolution, see comments below.For compound 2 the description in the manuscript of the numbering and bond distances are not consistent with the CIF, I think the atom numbers were changed at somepoint, see below.Other standard information is missing from the CIFs that the authors need to fill in.

RS15
While the data was collected to 0.55 angstroms, data is considered observed is I/sigma

In addition to changes indicated below in response to reviewer comments, the length of the abstract was cut down to fall within Nat. Commun. regulations, and several typos were fixed. All new text in the manuscript is highlighted in yellow.
Reviewer #1 (Remarks to the Author): Mankad et al. report an experimental investigation on the coordination-induced O-H and N-H bond weakening using a previously studied heterobinuclear [LAl(Me)Fp] complex 1, L -= [HC(CMeNdipp)2]-, dipp = 2,6-di-iso-propylphenyl, Fp-= [FeCp(CO)2]-.The work follows on other studies of the same authors on the same system, where they provide experimental and computational evidence that complex 1 dissociates reversibly at ambient conditions by Al-Fe homolysis, producing small equilibrium concentrations of the The present work provides a clear experimental evidence of coordination-induced O-H and N-H bond weakening, which is an outstanding result.However, the nature and reactivity of this class of heterobinuclear complexes is not well-contextualized in the Introduction, since references to previously studied metal-aluminyl systems which have experimentally shown reactivity with small molecules are lacking (https://doi.org/10.1038/s41557-018-0198-1).Metal-aluminyl systems have been also thoroughly characterized computationally (https://doi.org/10.1021/jacs.1c06728) to the extent that, very recently, computational studies have demonstrated that O-H and N-H bond activation could be both kinetically and thermodynamically feasible, through a similar diradical-like mechanism (https://doi.org/10.1002/chem.202203584).Within this broader context, the conclusion of this work:..."this is the first example of coordination-induced bond weakening by aluminum"..., is questionable.

Response: Thank you to Reviewer 1 for pointing out these omissions. We have made edits to the wording in the Introduction paragraph about CIBW by aluminum and added these three references.
The computational part needs to be improved considerably.
1) DFT calculations have been performed only for modeling the reaction intermediates in the proposed mechanism for H2O activation by 1, thus reporting only thermodynamic free energies and NOT kinetic free energy barriers, which are fundamental for identifying the rate-determining step.Indeed, the assumed RDS (homolytic Al-Fe cleavage of 1) is rather highly endothermic (ΔG = +24.5 kcal/mol), which means that the activation barrier should be expected to be even much higher (that could be not completely consistent with a reaction occurring at a temperature from -30°C to room temperature).A transition state for the PCET (3TS) has been located at a lower level of theory (which is then not consistently comparable to intermediates) on the triplet energy surface.I would expect here a spin-forbidden reaction, where a spin-crossing from a singlet to a triplet PES should occur.Therefore, calculations on both the PESs would be needed to properly characterize the mechanism (at least to verify that the TS energy is lower on the triplet than on the singlet PES) and confirm the asynchronicity of the PCET process.

We confirmed that charge separated species (bond heterolysis) are ~60 kcal/mol higher in energy (consistent with nonpolar, non-hydrogen bonding solvent) than Al and Fe radicals (bond homolysis). Also, as expected, there is no potential energy barrier for bond homolysis. There is the normal, smooth dissociation. While there could be a variational transition state in principle, finding this would be an enormous effort and would likely show that the barrier for bond cleavage is within the error of DFT
compared to the calculated bond energy.Therefore, the bond energy is likely a reasonable, but not perfect, estimate of the kinetics for this bond cleavage.Unfortunately, DFT likely overestimates this bond energy.While it would be nice to calculate it with CCSD(T), the full system is needed since the bulky aryl groups greatly modulate the bond energy, making this system is too large even for modern compute systems (we tried; even with ORCA's fast solver it was too large).

As suspected by the reviewer, long-range interaction of the Al and Fe radicals has a lower energy open-shell singlet configuration, not a triplet configuration. The triplet state is not involved in the reaction. Orbital swapping was required to obtain the correct open-shell singlet configuration, and we confirmed the configuration with a CASSCF calculation. Importantly, when water is coordinated to the Al metal center to generate Al(H2O) and the Fe center can approach the proton of water, at a distance of ~4 Angstroms electron transfer becomes thermodynamically favorable. The electron transfer happens from the Al ligand to the Fe metal center, and this results in a switch from an openshell singlet configuration to a closed-shell singlet configuration; this induces barrierless proton transfer. A detailed discussion of these new calculations is now contained in the manuscript.
2) In the BDFE calculation for eq 2, an error of about 14 kcal/mol has been assessed with respect to the experimental value (63 kcal/mol), which suggests that the used computational set up is not sufficiently reliable (for instance, relativistic effects were not included, the basis set quality and functionals other than PBE0 were not tested).

Response: With the new understanding about the reaction thermodynamics and electronic states it was no longer necessary to invoke eq 1-3. We have now removed the section discussing these equations and instead added an entirely new paragraph describing the direct calculation of the Al-water O-H BDFE. We used both DFT (M06, PBE1PBE) and CCSD(T) calculations to verify this very small bond energy.
3) It would be helpful to present the electronic structures or, at least, the calculated spin density of the radical species to show where the unpaired electron is (de)localized in the intermediates.The authors claim that the β-diketiminate ligand is redox non-innocent and that upon reduction the extra electron populate the ligand π* manifold.Did the authors calculate the electronic structure?Changes in intraligand bond distances cannot be used as indicative of the ligand oxidation state.At DFT level, oxidation state is a very critical issue.

Response: Thank you, this is now shown in revised Figure 4. The unpaired electron is, indeed, completely localized on the ligand backbone.
4)It may be useful a comparative study with recent computational results showing that the M-Al bond is able to activate, via a concerted, diradical-like mechanism, the O-H and N-H bonds.Moreover, analogy of complex 1 with CaMn4 oxygen-evolving complex in photosystem-II should be more deeply justified.

Response: We added the following text to contrast our report with the previous work on the Al-Au complexes: "It is worth contrasting this behavior with a recently reported computational model for X-H cleavage by a heterobinuclear Al-M complexes with diradical character. 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." As suggested by
Reviewer 2, we have included the established terminology of "multisite PCET" to describe both the CaMn4 OEC and our reported system.We hope this further justifies the analogy.We have also removed mention of the OEC in the Results & Discussion section to avoid overemphasizing this analogy.
Although the reported experimental evidence is very important in the field of small molecule activation processes, based on the above critical issues, I am afraid that I cannot recommend the acceptance of this manuscript in Nature Communications as it stands.
Response: We thank Reviewer 1 for the thorough and rigorous review.After incorporation of revisions in response to these comments and requests from other reviewers, we are confident the manuscript will be acceptable for Nat.Commun.
Reviewer #2 (Remarks to the Author): Mankad and coworkers present an interesting example of multisite proton coupled electron transfer that occurs through the homolylsis of an Fe-Al bond that results in an Al(III) complexed to a radical anion ligand that induces significant bond-weakening in small molecules coordinated to the Al.Overall, the work is quite interesting and the mechanistic work and analysis supports the mechanistic hypotheses in the manuscript.I am strongly in favor of acceptance, but have a number of comments for the authors to consider in advance of publication.

Response: Thank you to Reviewer 2 for the supportive evaluation.
Some of the description and citations in the introduction neglect seminal work.I recommend that either seminal work be cited in place of the work cited in the manuscript, or that it be added.
In the section describing the frontier work in PCET in synthesis, the citation of the recent Chem.Rev. by Knowles and colleagues is highly relevant, but I do not believe the work of Studer is appropriate.While certainly photocatalytic, the reaction requires the synthesis of a high MW sacrificial phosphine that limits the utility of the system.Earlier work by Knowles (J.Am.Chem. Soc., 2015, 137, 6440-6443) is really more appropriate to cite.

Response: We have added the Knowles JACS 2015 references as requested. We have chosen to keep the Studer reference due to the connection to p-block radical chemistry.
I am not sure that I agree with the description of the basis for coordination induced bond weakening, especially the use of invariably.While in most instances coordination of a small molecule to a low-valent metal metal reduces the pKa of the bound of the bound ligand, in some cases it is quite modest.In other cases, significant X-H bond-weakening occurs at less proximate sites although the impact of increasing acidity is almost non-existent (ie coordination of amides to Ti(III), Sm(II), etc.) Response: We have replaced "invariably" with "often".After that change, with all due respect to Reviewer 2, we feel that our description is accurate as written.
Sm-water bond-weakening was established by Flowers several years earlier (2015) than the cited Mayer report (J.Am.Chem. Soc. 2015, 137, 11526-11531).The estimate of bond-weakening by Flowers described in a follow-up small review (Dalton Trans. 2019, 48, 16142-16147) is smaller but consistent with the value determined by Mayer and Kolmar employing thermochemical cycles but is more relevant to synthetic systems in organic media.The value by Mayer is an overestimate since aqueous potentials and pKa's were used not relevant to the solvent employed in the reduction of an enamine in THF.I believe this was also recently pointed out by Peters and coworkers in a recent JACS publication as well.This should be corrected in the introduction and in the results and discussion section below Figure 3.It isn't necessary to get into this level of detail, but certainly the range of bond weakening based on experimental evidence can be described briefly.

Response: Thank you to Reviewer 2 for the helpful suggestions. We have updated this part to include a range of BDFE values (26-39 kcal/mol) for aqueous Sm(II) and added the requested citation of Flowers.
In Figure 1 a, the incorrect structure is shown.Upon addition of water to low-valent titanocene, chloride ions are solvated by water and displaced to the outer sphere.A better representation is shown above and supported by EPR, voltammetry, and computational studies carried out be Gansauer and coworkers (Angew.Chem.Int. Ed., 2012, 51, 3266-3270).I recommend this structure be shown.

Response: Thank you. The figure has been updated, and the ACIE 2012 by Gansauer et al. is now cited.
The discussion around Figure 1b should be referred to as multisite PCET.This terminology is wellestablished for the description of biological systems such as the OEC of photosystem II and synthetic systems.1b is also a classic example of acidification of a bound small molecule coupled to ET from a redox center in the language used to describe coordination induced bond weakening.
Response: Thank you to Reviewer 2 for correcting our terminology.The "multisite PCET" term has been added where appropriate, which also helps address a concern of Reviewer 1.
In the second paragraph of the results and discussion, it may be useful to cite recent work of Knowles on the coordination induced bond weakening of cyclopropanes (J.Am.Chem. Soc. 2022, 144, 34, 15488-15496) since in some ways it is analogy to the current system.

Response: Thank you. This work is now cited in the third paragraph of Results & Discussion.
In the weakening of N-H bonds, recent work has demonstrated significant weakening in Sm(II)-NH3 N-H bonds (J.Org.Chem. 2022Chem. , 87, 1689Chem. -1697)).

Response: Thank you, this work is now cited in the Introduction.
In the conclusion, the authors discuss the extension to the weakening of C-O bonds.I think this should be reworded.In the reduction of epoxides by low-valent titanocene, coordination induced bond weakening is an important feature of the first step in reduction.Although it doesn't involve X-H bond weakening, coordination of an epoxide to Ti(III) clearly weakens the C-O bond and is a classic example of coordination induced bond weakening.There needs to be more context in the description of the present Al-Fe system to differentiate from more classic systems.
The discussion around Figure 1b should be referred to as multisite PCET.This terminology is wellestablished for the description of biological systems such as the OEC of photosystem II and synthetic systems.1b is also a classic example of acidification of a bound small molecule coupled to ET from a redox center in the language used to describe coordination induced bond weakening.
In the second paragraph of the results and discussion, it may be useful to cite recent work of Knowles on the coordination induced bond weakening of cyclopropanes (J.Am.Chem. Soc. 2022, 144, 34, 15488-15496) since in some ways it is analogy to the current system.
In the conclusion, the authors discuss the extension to the weakening of C-O bonds.I think this should be reworded.In the reduction of epoxides by low-valent titanocene, coordination induced bond weakening is an important feature of the first step in reduction.Although it doesn't involve X-H bond weakening, coordination of an epoxide to Ti(III) clearly weakens the C-O bond and is a classic example of coordination induced bond weakening.There needs to be more context in the description of the present Al-Fe system to differentiate from more classic systems.
Ti (OH 2 ) n n = 1,2 This mean a massive amount of noise was added to this data set.If the data are cut at 0.72 Angstroms then the high electron density peaks around the W drop from ~4 electrons to ~1.5 electrons.In cutting the data all the AlertA and B's are sorted out and do not occur and leading to a better refined structure.Why are there no hydrogen atoms on the Cyclopentadienyl ligands?Why are its hydrogen not included in the chemical formula?This mean a massive amount of noise was added to this data set.Also R(merge) even at 0.71 angstroms is 63.96% which again is adding more noise to the data.From my experience R(merge)s of greater than 30% stop adding more noise than useful data.The authors may want to go back at look at the processing of the original data and determine if the crystal was dying in the beam and if possible cut the data at that point.However, if the data were cut at 0.82 Angstrom all the AlertA and B's are no longer triggered.The bonds listed above are not consistent with the number in the CIF provided, as I believe these should be in the thirties, C2-C3 are in the backbone of the main ligand.
(I) > 2, see INTENSITY STATISTICS FOR DATASET below.Therefore from ~0.72 to 0.55 angstroms the data is on average unobserved.
[LAlMe]. and Fp.frustrated radical pair (FRP) that can cooperatively activate oxygenated substrates such as CO2.Here, the authors show stoichiometric activation of O-H and N-H containing substrates including H2O, MeOH, iPrOH, tBuOH as well as iBuNH2.Based on NMR and kinetic measures, they propose that the rate-determining step (RDS) of the reaction mechanism is Al-Fe cleavage from 1 to produce [LAlMe]./Fp.FRP followed by substrate (H2O) coordination to[LAlMe].togive[LAlMeH2O]., from which H-atom transfer to Fp. occurs via H+/e-transfer from the redox non innocent ligand (PCET).