Light-enabled deracemization of cyclopropanes by Al-salen photocatalysis

Privileged chiral catalysts—those that share common structural features and are enantioselective across a range of reactions—continue to transform the chemical-research landscape1. In recent years, new reactivity modes have been achieved through excited-state catalysis, processes activated by light, but it is unclear if the selectivity of ground-state privileged catalysts can be matched. Although the interception of photogenerated intermediates by ground-state cycles has partially addressed this challenge2, single, chiral photocatalysts that simultaneously regulate reactivity and selectivity are conspicuously scarce3. So far, precision donor–acceptor recognition motifs remain crucial in enantioselective photocatalyst design4. Here we show that chiral Al-salen complexes, which have well-defined photophysical properties, can be used for the efficient photochemical deracemization5 of cyclopropyl ketones (up to 98:2 enantiomeric ratio (e.r.)). Irradiation at λ = 400 nm (violet light) augments the reactivity of the commercial catalyst to enable reactivity and enantioselectivity to be regulated simultaneously. This circumvents the need for tailored catalyst–substrate recognition motifs. It is predicted that this study will stimulate a re-evaluation of many venerable (ground-state) chiral catalysts in excited-state processes, ultimately leading to the identification of candidates that may be considered ‘privileged’ in both reactivity models.

Overall, the discovery that Salen catalysts can be used to control enantioselectivity in the excited state is of great value, and the transformation described very interesting.The mechanistic investigations, however, seem relatively preliminary and left several unanswered questions: -In the final sentence of the results section (p12), just prior to the conclusion, the authors propose a hypothesis for enantioenrichment: namely, that it is the conformational dynamics of the acyclic triplet diradical, and its ability to cross to the singlet state in different conformations, that most likely determine deracemization.This seems to be a key point of the study both in terms of how the desymmetrization works and mechanistic novelty, however, there is no experimental or computational evidence that directly addresses this point.
-A series of 3 substrates with different reduction potentials, -1.96 V, -2.12 V and > -2.2V were studied to support the mechanistic proposal, since only compound 1 reacts.However, out of all of these, compound 1 is also the only one that give an irreversible CV, suggesting a bond breaking event (in this case the cyclopropyl C-C).To me this suggests that you need a stabilizing gem-diPh to lower the C-C opening barrier -i.e., reduction may not be what is limiting reactivity of the other two substrates.Also, due to the fact that reduction of 1 is irreversible, the measured redox potential is approximate and very similar to 19.Square-wave voltammetry is necessary to obtain a reliable value.Overall, the experimental data provided to advance mechanistic arguments is limited, consisting of reduction potentials and an absorption spectrum.
-While the Al-Salen is proposed to act as a photoredox catalyst, a simpler explanation might be that it is a bathochromic shift agent.The redox potential of the Al-Salen is measured to result in uphill/endergonic reduction of the substrate, although the authors claim that e-transfer is "conceivable under LA activation of the substrate".However, in the computational studies, it is discussed that only weakly-bound complexes could be found and that there is no appreciable Alcarbonyl interaction.Doesn't this undermine the suggestion of LA activation?-The computational protocol looks appropriate for the study, although there is a possibility that only taking the best conformer from CREST and then optimizing with DFT will miss important structures.Are the key comparisons changed quantitatively by taking more xTB confomers forward for DFT optimization?-The idea that the different diastereomeric complexes give rise to different E-and Z-intermediates is potentially very interesting -although it was not easy to understand based on the shape of the "chiral pocket" why this is the case, even with the current SI Figure.This could be explained in more detail using 3D structures in the SI.
-In Figure 4 the carbon-centered radicals are in the wrong place for the two triplet INT1 species.
Referee #2 (Remarks to the Author): In this manuscript, Gilmour and coworkers report the deracemization of cyclopropyl ketone substrates using a chiral Al-salen catalyst under 400nm LED irradiation.The reaction scope is limited: cyclopropyl ketone substrates must possess stabilizing diaryl or diester groups.Selectivity is typically moderate-to-good (7:1 to 10:1), although some excellent (49:1) selectivites are observed.The authors nicely anchor scope limitations to mechanistic observations, which is instructive and scholarly.
Despite synthetic limitations, this work is remarkable in several ways.
First, deracemization is a highly significant research challenge with the potential to transform synthetic access to enantioenriched molecules.Very few catalytic deracemization reactions have been reported, and among those sparing examples, most require highly engineered specific substrate-catalyst interactions to achieve high levels of enantioinduction.The only exception -in my opinion -is Luo's approach (ref 51) which allows deracemization alpha to carbonyls.This work advances an Al-salen framework as a chiral photocatalyst, which allows a ketone functional group to be leveraged as a recognition element and thus an important step towards synthetically useful deracemizations.
Second, this study offers a new reactivity framework (C-C bond scission) for deracemization, by breaking/reforming C-C bonds rather than C-H bonds (Knowles, Bach, Hu, etc) or C=C bonds (Bach, Luo).This in principle opens the door for deracemization of fully substituted stereocentres.
Third -and perhaps most importantly -the identification of metal salen complexes as chiral photocatalysts is a potential breakthrough due to their synthetic accessibility and validated stereoinduction across reaction platform -hence the authors privileged catalysis pitch.
My reservations about the work are twofold.First, whether the authors' "privileged catalysis" argument can be appropriately validated, as at its core the argument makes claims beyond the present scope of work.Are Al-salen species uniquely viable as photocatalysts?What scope of salen ligands will be tolerated without disrupting desirable photophysical properties?Obviously this may depend on the transformation at hand.However additional screening data would be helpful herewhat other salen structures were explored?Is Al required, or would Cr-salen or Mn-salen be effective?Are suboptimal catalysts suffering from poor enantioinduction or perturbation of photophysical properties?
The second concern relates to the mechanism of enantioinduction.DFT study fails to answer most pressing question of why C-C scission is unselective but the *nearly* microscopically identical C-C bond reformation is highly enantioselective.I appreciate that these steps are NOT microscopically identical, but why are the transition state structures for the scission and reformation steps so different?Are the energetics of the two steps very different?Some additional information here is important.

Minor concerns:
The coordination of rac-1 and Al-1 is explored by DFT, but no discussion of an analogous experimental measurement.Is there a ground state interaction detected between substrate and ketone?
It would be useful to run the controls in Scheme 4a with scalemic substrate (instead of racemic) to assess background reactivity (racemization) that cannot be detected using rac-1.
Referee #3 (Remarks to the Author): The manuscript by Gilmour and Neugebauer et al. describes the visible-light mediated deracemization of cyclopropanes using a bifunctional Al-Salen catalyst.This work represents a conceptually novel approach to asymmetric photocatalysis by combining the LUMO-lowering activation of the ketone as provided by the Lewis acid nature of the Al with the reversible homolytic fission of the cyclopropyl ketone, and asymmetric induction provided therein by the ligand.This manifold is distinct from the chiral H-bonding donor-acceptor energy transfer and ground state chiral LA activation processes.
While this procedure has conceptual merit, the scope and application is somewhat limited by virtue of the reaction design.Further elaboration of the products could be explored, perhaps Baeyer-Villiger oxidation of products could yield the corresponding ester which ultimately could deliver BMS-505130 or Cipralisant via the aldehyde to alkyne etc.In addition, can the authors comment on the feasibility of performing the deracemization on benzoylcyclobutane, as the B-scission process is viable given appropriate substituents on the ring, although the scission is slower.Moreover, how are substituents at the other cyclopropyl carbon tolerated, such as gem-dimethyl.One could imagine the deracemization of cylopropanes based upon the Cypermethrin scaffold.If alkenyl substituents are used, can the delocalized radical perhaps close to afford a cyclopentene ring system with asymmetric induction.Additionally, can 3-membered heterocycles undergo similar transformations?
With regards to the mechanism, it would be appropriate to appropriate to see the UV/Vis and CV of the ligated catalyst and ketone to compare with the free species, this would indicate how much the LA is reducing the reduction potential of the ketone.Is a bathochromic shift in the triplet energy upon coordination of the Al visible at near-IR wavelengths to compare to the computed values?This may also enable some catalyst design.While it is not immediately clear from the images in the SI where the HOMO and LUMO are located in the excited state, if, as perhaps may be expected by analogy to other photocatalysts, the HOMO (and hence reducing ability) resides on the ligand framework, can substitution on the salen Ar rings with EDG's (such as alkoxys or amines) modulate the redox potential such that additional FGs can be tolerated (cmpds.19 and 30 for instance).Such rational photocatalytic design away from the normal salen ligand would undoubtedly add to this work.
Have the authors looked into the possibility (computationally or mechanistically) of Al-Cl bond fission, via an LMCT-type process, which would liberate a chlorine radical capable of undergoing HAT.Whether this could lead to enantioenrichment if the LA was bound seems unlikely but is this a competitive pathway energetically?Does the quantum yield of the reaction correlate to the absence of chain processes as purported by the mechanistic studies?
The manuscript is well written and schemes clearly presented, references are adequate.The SI is, again, well-constructed and the purity of compounds satisfactory.
Overall, while I believe this is a conceptual advance in asymmetric photocatalysis, the applicability of this reaction and utility of products remains more modest.If the authors can demonstrate some additional scope (4-membered ring, heterocycle) and/or derivatization to a compound(s) of interest, I believe this is of an appropriate standard to be published in Nature.
Author comments: This is an excellent, non-trivial aspect of the transformation that the referee has requested be further illuminated.To that end, we have considerably extended the conformer search for the possible triplet structures, including DFT optimisation of more than 20 possible candidate structures per enantiomer.This substantially expands the computational analysis and full details have been added to the supporting information.Figure 4 and the accompanying text in the manuscript have also been altered.To further analyse the origin of enantioenrichment, we have also interrogated the interconversion between the energetically most favourable structures and demonstrated that the isomerisation barrier is small.In addition, we have confirmed (by means of Spin-Flip-TDDFT calculations) that these triplet structures are energetically close to the (open-shell) singlet potentialenergy surface, which should lead to comparatively fast intersystem crossing.These results fully support the hypothesis that the conformational dynamics of the acyclic triplet diradical is most likely responsible for the deracemisation.
The manuscript text, supporting Information and Figure 4 have all been modified to accommodate these new findings.The text reads as follows "To test the assumption that the conformational process required for deracemization occurs in the long-lived triplet state, the charge separated triplet diradicals (R/S)-3 INT1 were considered as initial intermediates after intersystem crossing.In contrast to the initial complex, these species exhibit a short Al-O bond in the preferred conformations, as expected for a ketyl radical.The free energy barrier for ring opening to triplet diradicals 3 INT2 is rather low for both configurations (6 and 8 kcal/mol for (S)-and (R)-  S2, S3).However, conformations (E)-/(Z)-3 INT2", leading to the opposite (S) stereoisomers can be found within a free energy range of 1 kcal/mol for both diradicals, and epimerization could easily be achieved by rotation of the diphenyl methyl radical group to the opposite side of the enol plane.Furthermore, we also have identified a possible transition structure for the conversion of (E)-to (Z)-3 INT2, ( 3 TS2) with a free energy barrier of only 7 kcal/mol for this thermodynamically favorable process.This confirms that comparably fast processes can shift the conformational population of triplet diradical (Z)-3 INT2 and that the thermodynamic stability of these intermediates does not neccessarily determine the product configuration.Furthermore, we have confirmed by means of Spin-Flip-TDDFT calculations (Table S6) that (E)-and (Z)-3 INT2 are energetically close to the (open-shell) singlet potential-energy surface, which would lead to facile intersystem crossing.Our computational investigations indicate that there is no significant stereoselection before and during the excitation process, and in the formation of the 1,3-diradical.Fast conformational interconversions of diradicals (E/Z)-3 INT2 can pre-select conformations which lead to the observed accumulation of the (S)cyclopropane during bond formation in the singlet state."-A series of 3 substrates with different reduction potentials, -1.96 V, -2.12 V and > -2.2V were studied to support the mechanistic proposal, since only compound 1 reacts.However, out of all of these, compound 1 is also the only one that give an irreversible CV, suggesting a bond breaking event (in this case the cyclopropyl C-C).To me this suggests that you need a stabilizing gem-diPh to lower the C-C opening barrier -i.e., reduction may not be what is limiting reactivity of the other two substrates.Also, due to the fact that reduction of 1 is irreversible, the measured redox potential is approximate and very similar to 19.Square-wave voltammetry is necessary to obtain a reliable value.Overall, the experimental data provided to advance mechanistic arguments is limited, consisting of reduction potentials and an absorption spectrum.
Author comments: We appreciate the referee highlighting this aspect of the study and for the very helpful suggestions regarding the mechanistic investigations in a more general sense.The three substrates 1, 19, and 20 were designed to probe the influence of structural modifications on reaction efficiency.In the case of the methyl ester (20), our intention was to demonstrate that the phenyl ketone is required for photocatalysis.The inclusion of compound 19 was intended to demonstrate that replacement of the gem-diPh by gem-diEt only mildly effects the reduction potential but lowers the stability of the radical resulting from C-C bond scission, thereby lowering the rate of this step and preventing deracemization.As highlighted by the referee, this notion is corroborated by the irreversibility of reduction of 1 compared to 19.In contrast, 20 shows a significantly more negative reduction potential, suggesting that initial photochemical charge transfer from the catalyst is not feasible.We have modified the accompanying text in the manuscript in an attempt to make this more transparent and are grateful to the reviewer for highlighting this potential source of confusion.The referee has suggested that square-wave voltammetry be used to obtain reliable values, but we have been unable to access the electrodes required to perform this specific work.However, we believe that the measured reduction potential of 1 obtained by standard CV is reliable enough to draw the mechanistic conclusions described in the study.We hope that this response fully addresses the referee´s comments.
The text in the manuscript now reads as follows "Comparison of compound 19 (vs 1) was intended to demonstrate that the replacement of the gem-diPh by gem-diEt, although having a marginal impact on the reduction potential, lowers the stability of the radical resulting from C-C bond scission: this was expected to lower the rate of this step and thus prevent deracemization (e.r. 50:50).This notion is supported by the irreversible reduction of 1 in contrast to 19.Replacing the phenyl ketone by a methyl ester, as in the case of compound 20, results in a more negative reduction potential (<2.2 V), which suggests that the initial photochemical charge transfer from the catalyst is not feasible.Furthermore, control reactions with scalemic 19 and 20 were conducted and confirmed that no detectable background racemization was occurring.To complement these experimental data, a detailed computational study was conducted to further investigate the deracemization mechanism (Fig. 4c)." -While the Al-Salen is proposed to act as a photoredox catalyst, a simpler explanation might be that it is a bathochromic shift agent.The redox potential of the Al-Salen is measured to result in uphill/endergonic reduction of the substrate, although the authors claim that e-transfer is "conceivable under LA activation of the substrate".However, in the computational studies, it is discussed that only weakly-bound complexes could be found and that there is no appreciable Al-carbonyl interaction.Doesn't this undermine the suggestion of LA activation?
Author comments: The notion of aluminum-salen complexes acting as bathochromic shift reagents is intriguing and certainly something that we did consider in light of the elegant work by Profs Bach and Yoon.To further investigate this possibility, we performed additional UV/vis absorption spectroscopy experiments to identify any shifts in the substrate absorption upon catalyst addition.This was also conducted with strongly Lewis acidic Et2AlCl.These experiments did not reveal any shift in the absorption maxima, which supports the working hypothesis.The experimental details and absorption spectra (shown below) have been added to the Supporting Information.Furthermore, we have modified the manuscript to stress that whilst Lewis acid activation was a consideration in the initial reaction design (akin to ground state activation modes), this was evolved based on DFT insight: this indicates that a van-der-Waals complex (no Al-ketone interaction) in the ground state undergoes charge-transfer upon excitation.The lack of experimentally observed ground-state catalyst-substrate interactions in the UV/vis spectrum is also in line with this mechanistic picture.
The following text has been added to the manuscript: "To lend experimental support to this finding, spectroscopic and CV investigations were conducted and neither shifts in the absorption maxima nor in the cyclic voltammograms were observed (please see the Supporting Information).Therefore, Lewis acid activation of the substrate in the ground state was discounted, contrary to our initial hypothesis."To investigate the possibility of the aluminium-salen complex acting as a bathochromic shift reagent, NMR studies were performed (vide infra).The substrate and the catalyst were added in differing ratios to NMR tubes and dissolved in deuterated acetone.No shifts were observed by 13 C NMR spectroscopy, which further supports the working hypothesis.We fully agree that this experiment is an important control and thank the referee for suggesting it.We also respectfully direct the referee to the CV measurements in the response to Referee #3, which are consistent with these data.

Zoomed in region (measured in acetone-d6)
acetone-d6 1 + Al-1: 1:1 -The computational protocol looks appropriate for the study, although there is a possibility that only taking the best conformer from CREST and then optimizing with DFT will miss important structures.Are the key comparisons changed quantitatively by taking more xTB confomers forward for DFT optimization?
• Referee #2 (Remarks to the Author): In this manuscript, Gilmour and coworkers report the deracemization of cyclopropyl ketone substrates using a chiral Al-salen catalyst under 400nm LED irradiation.The reaction scope is limited: cyclopropyl ketone substrates must possess stabilizing diaryl or diester groups.Selectivity is typically moderate-togood (7:1 to 10:1), although some excellent (49:1) selectivites are observed.The authors nicely anchor scope limitations to mechanistic observations, which is instructive and scholarly.Despite synthetic limitations, this work is remarkable in several ways.
First, deracemization is a highly significant research challenge with the potential to transform synthetic access to enantioenriched molecules.Very few catalytic deracemization reactions have been reported, and among those sparing examples, most require highly engineered specific substrate-catalyst interactions to achieve high levels of enantioinduction.The only exception -in my opinion -is Luo's approach (ref 51) which allows deracemization alpha to carbonyls.This work advances an Al-salen framework as a chiral photocatalyst, which allows a ketone functional group to be leveraged as a recognition element and thus an important step towards synthetically useful deracemizations.
Second, this study offers a new reactivity framework (C-C bond scission) for deracemization, by breaking/reforming C-C bonds rather than C-H bonds (Knowles, Bach, Hu, etc) or C=C bonds (Bach, Luo).This in principle opens the door for deracemization of fully substituted stereocentres.
Third -and perhaps most importantly -the identification of metal salen complexes as chiral photocatalysts is a potential breakthrough due to their synthetic accessibility and validated stereoinduction across reaction platform -hence the authors privileged catalysis pitch.

Author comments:
We greatly appreciate the generous evaluation from referee 2 and, in particular, for the summary of how the work complements the other conceptual pillars of asymmetric photocatalysis.
As the referee notes in point 2, this work provides a new chiral photocatalyst platform for the deracemization via reversible C-C bond scission.In the manuscript, we have highlighted the value of socalled privileged catalysts in ground state processes and questioned if a comparable suite of privileged photocatalysts will emerge in the future.The referee touched on this in her/her/their report, and this motivated us to demonstrate proof of concept in a second, non-related transformation (please see below).We are most grateful to the referee for encouraging us to develop this further.
My reservations about the work are twofold.First, whether the authors' "privileged catalysis" argument can be appropriately validated, as at its core the argument makes claims beyond the present scope of work.Are Al-salen species uniquely viable as photocatalysts?What scope of salen ligands will be tolerated without disrupting desirable photophysical properties?Obviously this may depend on the transformation at hand.However additional screening data would be helpful here -what other salen structures were explored?Is Al required, or would Cr-salen or Mn-salen be effective?Are suboptimal catalysts suffering from poor enantioinduction or perturbation of photophysical properties?
Author comments: This is a great point and we appreciate the opportunity to comment and revise the manuscript.Salen complexes have found broad application in asymmetric synthesis in ground state regimes and they are often referred to as privileged chiral catalysts.When developing this project, we were very conscious of the value that a toolkit of privileged chiral photocatalysts would confer to complete current strategies.The manuscript was intended to phrase any possible generality with salenphotocatalysis as more of a question than a claim, and to extend the generality of Al-salen complexes from the group state to excited state reactivity paradigms.However, we fully appreciate that the term "privileged" should be used with extremely caution and we apologise for any lack of clarity.We have identified all incidences where the term appears in the manuscript and modified the text where necessary to ensure that our intentions are clear.We appreciate the referee highlighting this important point.
We are confident that Al-salen photocatalysis will find other applications and, motivated by the referee´s suggestion, we are happy to share a preliminary validation of an enantioselective 6πelectrocyclization that occurs at ambient temperature (up to 84:16 er.Please see below).Although this is beyond the scope of this study on cyclopropane deracemisation, we fully agree with the reviewer that an additional example would strengthen the manuscript and hint towards generality.We have therefore added these data to the supporting information and included a line in the main text of the manuscript.Since the focus of the paper is deracemization, we feel that the work is best placed in the Supporting Information but we would be happy to move it to Figure 5 if that is preferable.The following statement has been added to the manuscript as reference 62.
"To further demonstrate the synthetic potential of Al-salen photocatalysis, catalyst Al-1 has been utilized in an enantioselective 6π-electrocyclization.Please see the supporting information for full details." The referee has also raised an excellent point regarding the impact of structural alterations on catalyst performance.Early on in the project, and in addition to our original screening data, we explored a set of alternative salen complexes (please see below).The non-redox active metal aluminium is unique in its ability to generate an active chiral salen photocatalyst, whereas other metal-based salens were unreactive.This observation is in line with earlier studies on the photophysical properties of various metal-salen complexes (refs 34 & 35).It is interesting to note that substitution of the tBu group results in lower levels of enantioinduction.These data have been added to the supporting information and we apologise for not including them in the original submission.
We also respectfully direct Referee #2 to the next report where we detail the impact of structural modifications around the salen periphery.Ligands bearing only a methoxy group in the ortho or para positions were unfortunately insoluble in an array of common solvents (such as DCM, acetone or toluene).We therefore did not explore their photocatalysis behavior further due to this limitation.Substituting the p-tBu group with p-OMe did provide a catalyst that was soluble but no enantioenrichment was observed under the standard reaction conditions.The cyclopropane was recovered in racemic form.We consider that the success of the title reaction with a commercial catalyst will lower the barrier to utilising the methodology.The second concern relates to the mechanism of enantioinduction.DFT study fails to answer most pressing question of why C-C scission is unselective but the *nearly* microscopically identical C-C bond reformation is highly enantioselective.I appreciate that these steps are NOT microscopically identical, but why are the transition state structures for the scission and reformation steps so different?Are the energetics of the two steps very different?Some additional information here is important.
modification will play an important part in future studies to expand salen photocatalysis.In the specific cases mentioned by the referee, we have prepared a number of analogues and explored their suitability.Although some of this data was presented in the supporting information, we fully appreciate that more information would be beneficial to readers.
We have performed a range of UV/vis and NMR investigations to probe for possible (Lewis acid) catalyst -substrate interactions in the ground state but such phenomena were never detected (please also see the response to Referee #2 above).This is fully consistent with the DFT analysis.Furthermore, we performed CV measurements to further investigate catalyst and ligand interactions as proposed.However, the shift of the substrate's reduction potential upon catalyst addition in the CV is negligible.Regarding the helpful suggestion to investigate more electron-rich ligand frameworks, we have prepared the ligands bearing a methoxy group in the ortho or para positions.Unfortunately, these catalysts were insoluble in an array of common solvents (such as DCM, acetone or toluene).We therefore did not explore their photocatalysis behavior further due to this limitation.Substituting the the p-tBu group with p-OMe did provide a catalyst that was soluble but no enantio-enrichment was observed under the standard reaction conditions.The cyclopropane was recovered in racemic form.
We have included this information (summarised below) in the supporting information.
Catalyst screening added to the supporting information: Have the authors looked into the possibility (computationally or mechanistically) of Al-Cl bond fission, via an LMCT-type process, which would liberate a chlorine radical capable of undergoing HAT.Whether this could lead to enantioenrichment if the LA was bound seems unlikely but is this a competitive pathway energetically?Does the quantum yield of the reaction correlate to the absence of chain processes as purported by the mechanistic studies?

3INT1,
respectively): this step is associated with the back transfer of charge to the metal ligand.During these studies, we did not find significant spin density on the catalyst moiety of 3 INT2.The configuration of the double bond of the enol radical in 3 INT2 depends on the cyclopropyl configuration in the low energy conformation of 3 INT1: the (Z)-enol radical is formed from (S)-3 INT1, whereas the ring opening of the (R)ketyl radical furnishes the (E)-enol radical.Conformational analysis of (E)-/(Z)-3 INT2 reveals that both isomers attain conformations that would yield the (R) cyclopropane upon ring closure, as measured by the dihedral angle around the C • H-CH2 bond (Tables a 0.1 mmol scale according to General Procedure E.
Reactions performed on a 0.1 mmol scale according to General Procedure E.