DFT Mechanistic Insights into Aldehyde Deformylations with Biomimetic Metal-Dioxygen Complexes: Distinct Mechanisms and Reaction Rules.

Aldehyde deformylations occurring in organisms are catalyzed by metalloenzymes through metal-dioxygen active cores, attracting great interest to study small-molecule metal-dioxygen complexes for understanding relevant biological processes and developing biomimetic catalysts for aerobic transformations. As the known deformylation mechanisms, including nucleophilic attack, aldehyde α-H-atom abstraction, and aldehyde hydrogen atom abstraction, undergo outer-sphere pathways, we herein report a distinct inner-sphere mechanism based on density functional theory (DFT) mechanistic studies of aldehyde deformylations with a copper (II)-superoxo complex. The inner-sphere mechanism proceeds via a sequence mainly including aldehyde end-on coordination, homolytic aldehyde C-C bond cleavage, and dioxygen O-O bond cleavage, among which the C-C bond cleavage is the rate-determining step with a barrier substantially lower than those of outer-sphere pathways. The aldehyde C-C bond cleavage, enabled through the activation of the dioxygen ligand radical in a second-order nucleophilic substitution (SN2)-like fashion, leads to an alkyl radical and facilitates the subsequent dioxygen O-O bond cleavage. Furthermore, we deduced the rules for the reactions of metal-dioxygen complexes with aldehydes and nitriles via the inner-sphere mechanism. Expectedly, our proposed inner-sphere mechanisms and the reaction rules offer another perspective to understand relevant biological processes involving metal-dioxygen cores and to discover metal-dioxygen catalysts for aerobic transformations.

Introduction

Aldehyde deformylations are chemical reactions taking place in organisms during metabolism and biosynthesis.[1,2] These biochemical reactions are catalyzed by metalloenzymes such as cytochrome P450 and aldehyde decarbonylase (ADO).[3−6] By binding to molecular oxygen, the metal cofactors of the metalloenzymes form metal–dioxygen active cores that perform these reactions.[7−10] Inspired by these biological processes, much effort has been devoted to the study of small-molecule transition-metal–dioxygen complexes (denoted TM–O2 complexes hereafter).[11−16] On the one hand, TM–O2 complexes serve as biomimetic models to deeply understand the relevant biological processes like deformylations; on the other hand, studying the chemistry of TM–O2 complexes could lead to the discovery of aerobic catalytic transformations.[17−40] The use of abundant molecular O2 as an oxidant offers a green and sustainable approach to produce valuable chemical products.

TM–O2 complexes often feature amphoteric reactivity, including the electrophilic reactivity as exhibited in the hydrogen atom abstraction reactions[41−43] and the nucleophilic reactivity as found in the aldehyde deformylations.[18,31,44−48] In the past two decades, using aldehydes such as 2-phenylpropionaldehyde (2-PPA), trimethylacetaldehyde (TMA), cyclohexanecarboxaldehyde (CCA), and other electrophiles as probes, the nucleophilic reactivity of a wide range of TM–O2 complexes has been investigated, where TM = V,[49,50] Mn,[26,31,40,44,45] Fe,[23,47] Co,[46,48] Ni,[28,32] and Cu.[18] Notably, using O2 as an oxidant, catalytic deformylation of 2-PPA has also been reported.[19,51,52] As de Visser and co-workers in 2021 have presented an excellent and comprehensive review on this topic,[14] we herein highlight the works relevant to the present study.

The copper–dioxygen core present in oxygenase (e.g., lytic polysaccharide monooxygenase[53]) and oxidase (e.g., heme-copper oxidases[54]) enzymes[55,56] encouraged the synthesis and the reactivity study of various biomimetic copper–dioxygen model complexes.[12,15] In this context, Tolman and co-workers in 2010 synthesized a copper (II)–superoxo complex [CuO2]K(18-crown-6) supported by a dianionic pyridine-2,6-dicarboxamide ligand.[20] In 2014, McDonald and co-workers studied the nucleophilic reactivity of the complex and found that the complex reacted with acyl chlorides and a range of aldehydes (eqs 1–4, Scheme ).[18] To improve the stability and solubility of [CuO2]K(18-crown-6), Tolman and co-workers in 2019 used phase-transfer agent Krpytofix222 (Krypt) to replace 18-crown-6, furnishing [CuO2][K(krypt)].[22] It was reported that the new complex ([CuO2][K(krypt)]) reacted with 2-PPA to give copper(II) enolate complex [LCu(OC=C(Me)Ph)]− in the presence of water in a different solvent (19:1 tetrahydrofuran (THF)/CH3CN). Because the two experimental studies were conducted with [CuO2]− complexes having different countercations under different experimental conditions, the question of whether the copper–dioxygen core possesses nucleophilic reactivity remains unclear. It is highly desirable to have an “ideal” approach that could avoid the effects of these undefinable experimental factors. While it is not easy for experimental studies, theoretical calculations offer a convenient approach to reach the goal. Theoretical calculations produce detailed geometric, energetic, and electronic information, which is important for deeply understanding reaction mechanisms and reactivity.

Scheme 1

Deformylation Reactions Studied in This Work

Mechanistically, a widely accepted mechanism to account for aldehyde deformylations with TM–O2 complexes is the outer-sphere nucleophilic attack mechanism,[18,31,44−48] termed as ONA mechanism hereafter. Under the mechanism, the O2 ligand undergoes nucleophilic attack at the aldehyde carbonyl carbon to start a deformylation reaction.[18,31,44−48] Using a quantum mechanics/molecular mechanics (QM/MM) modeling method, Hackett and co-workers demonstrated the ONA mechanism to be feasible for the deformylation catalyzed by sterol-14α-demethylase (CYP51).[57] In the experimental studies, the ONA mechanism was often postulated.[18,31,44−48] As exemplified in Scheme A,[18] under the ONA mechanism, the reaction begins with nucleophilic attack to form an intermediate, followed by Baeyer–Villiger oxidation or hydrogen atom abstraction pathways, leading to the deformylation products. Because the mechanism involves abstraction of α-hydrogen of aldehydes (2-PPA and CCA) and TMA does not have an α-hydrogen, the deformylation of TMA (eq 2) terminates after the decomposition of the nucleophilic attack intermediate. As the mechanism could account for reactions in eqs 1–3 reasonably, it is puzzling that TMA could undergo deformylation (eq 2), but the more electrophilic benzaldehyde could not (eq 4), although it is possible that the nucleophilic attack is not the rate-determining step.

Scheme 2

Previously Proposed Outer-Sphere Deformylation Mechanisms (A–C) and Our Proposed Inner-Sphere Mechanisms (D)

For clarity, the −1 charge states of Cu-containing structures and +1 charge state of Mn-containing structures were omitted.

As the ONA mechanism is commonly accepted, Sastri and de Visser et al.[33,41] in 2016 and 2017 first discovered that the ONA mechanism could not be applied for their 2-PPA deformylation with the cationic manganese (III)–peroxo complex supported by dimethyl-2,4-di(2-pyridyl)-3-benzyl-7-(pyridin-2-ylmethyl)-3,7-diazabicyclo[3.3.1]nonan-9-one-1,5-dicarboxylate ligand, denoted [MnO] (quintet) hereafter. They proposed an electrophilic deformylation mechanism, where the peroxo ligand of [MnO] first abstracts the α-H atom of 2-PPA, denoted α-HAA mechanism (Scheme B).[41] Their density functional theory (DFT) calculations indicated that the nucleophilic attack transition state (TS) is 8.5 kcal/mol higher than the α-HAA TS and the TS for the nucleophilic attack at the in situ generated enol species is slightly higher than the α-HAA TS by 1.2 kcal/mol. More interestingly, the same group in 2019 found that the replacement of aldehyde α-hydrogen with deuterium could switch the deformylation product of α-[D1]-CCA from conventional cyclohexanone to cyclohexanecarboxylic acid.[58] This finding led them to propose a mechanism in which the peroxo ligand first extracts the aldehyde hydrogen (denoted ald-HAA mechanism, Scheme C).[58]

In 2020, Baik–Cho joint group[59] and our group[60] independently carried out DFT calculations to gain insight into the dioxygenase-like reactivity of [CoIII(TBDAP)(O2)]+ (denoted ], TBDAP = N,N-di-tert-butyl-2,11-diaza[3.3](2,6)-pyridinophane) toward nitriles (R–C≡N; R = Me, Et, and Ph).[61] These studies showed that the reaction starts with nitrile coordination to form a five-ring metallacycle as a key intermediate, followed by O–O bond cleavage, finally leading to the activation product (vide infra). Our further analyses indicated that the O–O bond cleavage could be essentially considered as a cobalt-aided [1,3]-sigmatropic rearrangement.[60] Because aldehyde may also coordinate to TM–O2 complexes, we envisioned that the deformylations with TM–O2 complexes could start with aldehyde coordination, thus initiating an inner-sphere pathway that is different from the known outer-sphere pathways (i.e., ONA,[18,57] α-HAA,[41] and ald-HAA[58] pathways). Using the experimental reactions (eqs 1–4) as representatives,[18] our DFT study demonstrated that these deformylations indeed prefer the inner-sphere mechanism (Scheme D) over the outer-sphere mechanisms (Scheme A–C).[18,41,57,58] Compared to the outer-sphere mechanisms, a marked difference lies in that the inner-sphere mechanism involves homolytic aldehyde C–C bond cleavage prior to O–O bond homolytic cleavage to generate an alkyl radical. The C–C bond cleavage is enabled via O2 ligand radical activation in a way similar to a second-order nucleophilic substitution (SN2) process; thus, we term the mechanism as the radical SN2-like mechanism. Expectedly, the inner-sphere mechanism will complement the known outer-sphere mechanisms for deeply understanding the reactivity of TM–O2 complexes and offer a new way of thinking for the rational development of catalysts to perform aerobic catalytic transformations.

Computational Methods

Using experimental compounds, all structures were optimized at the B3LYP/6-31G** level in the experimental solvent (a solvent mixture with THF/dimethylformamide (DMF) = 3:1 for the ] system[18] and ] system,[57] benzene solvent for the ] system[61]). The solvent effects of the solvent mixture were described by the generic polarizable continuum model (PCM) solvent model with a dielectric constant of 14.88 (=ε(THF) × 0.75 + ε(DMF) × 0.25 = 14.88, where ε(THF) = 7.43 and ε(DMF) = 37.22) and those of benzene solvent were described by the PCM solvent model.[62,63] Depending on the nature of a species, the B3LYP calculations could be restricted B3LYP (RB3LYP) for closed-shell singlet species, unrestricted B3LYP (UB3LYP) for open-shell species, or broken-symmetry B3LYP (BSB3LYP, implemented with the “guess = mix” keyword in the Gaussian program) for open-shell singlet species. Particular attention was paid to the closed-shell singlet species. When the wave function of a closed-shell singlet species was found to be unstable, the open-shell singlet was recalculated with BSB3LYP and reported. In the nomenclature hereafter, we specify the spin multiplicity of a structure with a left superscript. A left superscript OS represents an open-shell singlet, and no superscript means a closed-shell singlet. Harmonic frequency analysis calculations at the same level were performed to verify the optimized geometries to be minima with no imaginary frequency or transition states having unique one imaginary frequency. With the optimized structures, the energies were further improved by M06L[64](PCM)/6-311++G(d,p)//B3LYP(PCM)/6-31G(d,p) single-point calculations with solvent effects simulated by the PCM solvent model using the experimental solvent (supra infra). The M06L single-point calculations maintained the wave function nature in the geometric optimizations using the keyword “guess = read” in the Gaussian program. Harmonic vibration frequencies at the B3LYP(PCM)/6-31G(d,p) level were used to correct the single-point energies to free energies at the experimental temperature (193.15 K for the ] system, 298.15 K for ] and ] systems). Considering that the ideal gas model used in the Gaussian program inevitably overestimates the entropy contribution to the free energy for reactions in a solvent, a −2.6 kcal/mol correction was applied for each component decrement (i.e., correcting the overestimation of the entropy penalty) and vice versa.[65−69]

The crystal structure of ] has not been obtained, but that of the precursor [Cu]-(4-Bu-pyridine) was reported.[20] In addition, ] was characterized to have a triplet ground state, an O–O bond vibrational frequency of 1104 cm–1, and a characteristic UV–vis absorption at 627 nm.[20] Using these relevant experimental results, we evaluated the reliability of our protocol by comparing to other protocols (M06L(PCM)//B3LYP-D3(PCM), M06[70−73](PCM)//B3LYP-D3[74](PCM), B3LYP-D3(PCM)//B3LYP-D3(PCM), MN15[75](PCM)//MN15(PCM)) with different basis set combinations (6-311++G**//6-31G** or Def2TZVP[76−78]//Def2SVP[76−79]). Detailed results in the Supporting Information (Section SI1) indicate that the protocol (M06L(PCM)/6-311++G**//B3LYP(PCM)/6-31G**) has the best overall performance in reproducing these available experimental results. In addition, we also confirmed that the reaction system could not involve [Cu]–O–O–[Cu] species (see Section SI2 for details).

The natural bond orbital (NBO) charges,[80−83] the Mulliken spin populations, and biorthogonal molecular orbitals (BMOs) were obtained at the B3LYP(PCM)/6-31G(d,p) level. All standard DFT calculations were performed using the Gaussian 09 program.[84]

Results and Discussion

Deformylation Mechanism of 2-PPA with

The ONA mechanism was often applied to account for the deformylations with TM–O2 complexes as deformylating reagents.[18,31,44−48] We first investigated whether [CuO] could act similarly. At a glance, because of the anionic nature of [CuO], the ONA mechanism appears reasonable. Nevertheless, the nucleophilic attack to give IM2a is highly endergonic by 24.5 kcal/mol (Figure ). The geometric optimization to locate a singlet counterpart of IM2a led to separate [CuO] and 2-PPA. These results suggest that the ONA mechanism may not operate in this reaction. We next considered the α-HAA and ald-HAA mechanisms. Similar to the previous finding,[41] the α-HAA TSs, TS1a (ΔG≠ = 18.0/15.9 kcal/mol), are substantially lower than (ΔG = 24.5 kcal/mol). The ald-HAA TSs, TS1b (ΔG≠ = 23.4/22.3 kcal/mol), are higher than the α-HAA TSs but still lower than IM2a.

Figure 1

Reaction pathway for the deformylation of 2-PPA (eq 1), along with the relative free energies in kcal/mol. The values in black and orange in IM2 and TS2 are the bond lengths in angstroms and NBO charges in e, respectively. The −1 charge states of the Cu-containing structures were omitted for clarity.

Essentially, the three processes (ONA, α-HAA, and ald-HAA) take place in an outer-sphere fashion without the participation of the copper center. Although the α-HAA and ald-HAA TSs are not inaccessible, our previous study[60] inspired us to explore whether the reaction could undergo an inner-sphere mechanism, starting with carbonyl coordination to the copper center. Compared to the outer-sphere mechanism, we foresaw that the inner-sphere mechanism could be benefited from the coordination interaction and Cuδ+···Oδ− Coulomb attraction. Figure illustrates our computed pathways. First, the carbonyl group of 2-PPA undergoes end-on coordination with [CuO] by crossing a low barrier of 2.8 kcal/mol (TS1), forming a slightly stable intermediate IM2. Proceeding forward, two pathways were examined. The pathway to break the O–O bond, giving IM3a, is highly endergonic by 37.8 kcal/mol, unambiguously excluding the possibility. The alternative breaks the aldehyde C4–C5 bond of 2-PPA. Remarkably, the TS (TS2) to break the bond is 24.0 kcal/mol lower than IM3a. In addition, we considered the two-state reactivity mechanism[85] to break the C4–C5 bond, but the open-shell singlets IM2 and TS2 are higher than their corresponding triplet counterparts. The energetic difference of 1.7 kcal/mol between TS2 and TS2 indicates that the two-state reactivity can only play a minor role if any.

The homolytic C4–C5 bond cleavage affords IM3 and a benzyl radical R. The bond cleavage substantially elongates the O1–O2 bond from 1.31 Å in IM2 to 1.45 Å in IM3; thus, the O–O bond can break with a barrier of 11.2 kcal/mol (TS3 relative to IM3). In comparison, intermediate IM3a led by directly breaking the O1–O2 bond of IM2 has relative energy of 37.8 kcal/mol. Therefore, the C4–C5 bond cleavage greatly facilitates the O–O bond cleavage. Furthermore, we considered whether the O1–O2 bond cleavage could be facilitated by associating with the R radical, as described by TS3a; however, both TSs are higher than TS3, suggesting that the association of R cannot help the O1–O2 bond cleavage. It is worth noting that, referring to TS3a, the bulky ligand of the [CuO] complex prevents R associating with O1.

The O–O bond cleavage leads to IM4, in which O1 bears a single electron. Thus, the R radical couples with O1, giving IM5. Finally, IM5 undergoes intramolecular hydrogen atom transfer via TS4, affording the experimental products acetophenone and formic acid. Examining the inner-sphere pathway, the ketone oxygen atom originates from [CuO], which agrees with the isotope labeling experiment with [CuO] as the deformylating reagent. To our delight, a complex similar to IM2 has been characterized spectroscopically for the deformylation of 2-PPA with an analogue of [CoO] during our preparation of the paper, which could be an experimental evidence to support our proposed mechanism (see the box in Figure ).[86]

Overall, the reaction is exergonic by 71.3 kcal/mol with a rate-determining barrier of 14.7 kcal/mol for breaking the aldehyde C–C bond, indicating that the reaction could take place facilely. The rate-determining TS2 (ΔG≠ = 13.8 kcal/mol) is lower than the intermediate IM2a (ΔG = 24.5 kcal/mol) led by nucleophilic attack; TS1a (ΔG≠ = 18.0/15.9 kcal/mol), TS2a (ΔG≠ = 20.4 kcal/mol), and TS3b (ΔG≠ = 20.5 kcal/mol) involved in the α-HAA pathway; and TS1b (ΔG≠ = 23.4/22.3 kcal/mol) for the ald-HAA process. The favorability of the inner-sphere mechanism could be attributed to the favorable interactions (coordination and Coulomb attraction) between the copper center and the carbonyl oxygen of 2-PPA (see the NBO charges and Cu···O3 distances given in IM2 and TS2) and the high stability of the benzyl radical (vide infra). Compared to the ONA and α-HAA mechanisms in which the O–O bond breaks prior to the C–C bond cleavage, the inner-sphere mechanism first breaks the C–C bond, which facilitates subsequent O–O bond cleavage.

To corroborate the favorability of the inner-sphere mechanism, we further compared the energetics of the four mechanisms at the other five levels of DFT calculations (see Section SI3 for detailed results). In spite of some numerical differences, the energetic results from the six levels of calculations consistently indicate that the inner-sphere mechanism is most favorable.

Comparing the Deformylation Mechanism of 3[CuO2]− and Acetonitrile Activation Mechanism of

Acetonitrile activation with [CoO] involves O–O bond cleavage but not C–C bond breaking.[59,60] It is insightful to compare the key steps of the two systems. In the following, we gain insight into the differences and similarities of the two systems (Figure ).

Figure 2

Comparing the key steps involved in 2-PPA deformylation with [CuO] and acetonitrile activation with [CoO]. For clarity, the −1 charge states for Cu-containing structures and +1 charge states for Co-containing structures were omitted.

First, the O2 ligand in [CuO] adopts end-on coordination with unequal Cu–O bond lengths (1.91 and 2.55 Å, respectively), while the O2 ligand in [CoO] features side-on coordination with equal Co–O bond lengths (ca. 1.88Å). The coordination difference indicates that a Cu(II) oxidation state is preferred in [CuO], while a Co(III) oxidation state is preferred in [CoO], although Cu(III)–peroxo complexes were synthesized with strongly electron-donating bidentate-diketiminate or anilido-imine ligands.[87,88] Due to the coordination difference, the oxygen ligand in [CuO] bears a single electron (Mulliken spin population: O2 1.17), while the ligand in [CoO] possesses no radical character with the two single electrons residing at the cobalt center (Mulliken spin populations: Co 1.88, O2 0.02). However, the O2 ligand in [CoO] can gain a single electron upon acetonitrile coordination, as indicated by the Mulliken spin population of O2 (0.94) in . The Cu(II) oxidation state preference explains the general feature that the Cu center in the intermediate of the [CuO] system only interacts with one of the two neighbor oxygen atoms tightly, while the Co center in the intermediate of the [CoO] system can interact with neighboring two atoms tightly.

Second, IM1 converts to a loose five-ring structure IM2, while IM1 isomerizes to a tight five-ring metallacycle IM2. Indicated by the Mulliken spin populations and the O–O bond length in IM2, the O2 ligand in IM2 maintains its radical and partial double bond character in IM1. In contrast, the O2 ligand in IM2 bears no single electron with the two single electrons located at the Co center and the O–O bond is a regular single bond. In addition, the O2–C4 bond in IM2 is only partially formed with a bond length of 1.65 Å, while the bond in IM2 is a well-formed C–O single bond with a bond length of 1.36 Å. To gain deeper insight into the origin of these differences, we first go over the isomerization mechanism of IM1 to IM2, proposed in our previous study (Figure ).[60] To begin with, the single α-electron located on the O2 ligand in IM1 activates an acetonitrile N≡C π-bond, forming the C–O σ-bond in IM2 and leaving a single α-electron on the N=C double bond. Then, the single α-electron on the N=C double bond interacts with a pair of nonbonding electrons on Co to form the Co–N bond in IM2. With this mechanism established, we understand why IM1 is converted to IM2, rather than an intermediate similar to IM2. Figure A compares the key relevant biorthogonal molecular orbitals (BMOs) of IM1 and IM1. The two intermediates have similar singly occupied BMOs (SOBMOs) dominated by O2 ligands (see SOBMO189Cu/SOBMO128Co), which are approximately perpendicular to the Cu–O–O/Co–O–O plane. In IM1, the C=O π-bond orbital of 2-PPA (DOBMO137Cu, doubly occupied BMO) lying in the Cu–O–O plane is nearly perpendicular to SOBMO189Cu. Therefore, the O2 ligand radical cannot activate the C=O bond effectively, preventing the single-electron activation mode (see box C in Figure B). Nevertheless, the carbonyl group can be activated through coordination activation, as illustrated by the electron flow chart in box A (Figure B). In principle, the full activation of the C=O π-bond would result in IM2-RA; however, the activation may not be so effective. Overall, IM2 can be represented by the two resonance structures IM2-RA and IM2-RB, without involving IM2-RC. The resonance structure description agrees with the relatively long lengths of Cu–O1 (R = 2.34Å), C4–O2(1.65 Å), and O3=C4 (1.30 Å) and shortened Cu–O3 (1.92 Å) in IM2. In contrast to the 2-PPA unit in IM1, the acetonitrile unit in IM1 has two π-bonds (see DOBMO100Co and DOBMO112Co in Figure A). The perpendicular DOBMO100Co is nearly parallel to the single-electron orbital (SOBMO128Co), which is favorable for single-electron activation. Thus, single-electron activation can take place effectively to convert IM1 to IM2 with the mechanism described by TS2 in Figure .

Figure 3

Mechanism for the isomerization of IM1 to IM2. For clarity, the +1 charge states for Co-containing structures were omitted.

Figure 4

(A) Comparing the singly occupied biorthogonal molecular orbitals of IM1 and IM1. (B) Mechanism for converting IM1 to IM2. For clarity, the −1 charge states for Cu-containing structures and +1 charge states for Co-containing structures were omitted.

Third, IM2 undergoes aldehyde C4–C5 bond cleavage, giving an alkyl radical, while IM2 breaks the O1–O2 bond directly (Figure ). A straightforward explanation for the difference would be that the O–O bond in the former is a partial double bond (R = 1.31Å), while the bond in the latter is a single bond (R = 1.46Å). Previously, we have shown that the O–O bond cleavage is essentially a cobalt-aided one-electron [1,3]-sigmatropic rearrangement (for electron flow chart description, see Figure ). In the following, we analyze the mechanism for breaking the C–C bond in IM2. As shown in Figure , IM2 cannot break the O–O bond because the process to afford IM3a is highly endergonic by 37.8 kcal/mol. The high energy of IM3a can be attributed to its unfavorable characteristics such as the disfavored Cu(III) oxidation state and hypervalent carbon. Therefore, an approach to circumvent the hypervalent carbon problem is to break the aldehyde C4–C5 bond before O–O bond cleavage, which can be achieved through the activation of the O2 ligand radical. Figure A illustrates the principle for IM2 to break the C–C bond via TS2. The C4–C5 σ-bond orbital in TS2 points outward the five-ring plane, which is symmetrically suitable for the interaction of the C4–C5 σ-bond orbital with the single-electron orbital of the O2 ligand, as depicted by SOBMO189Cu of the bond cleavage TS TS2. As illustrated by the electron flow chart of TS2, the bond cleavage takes place in a way similar to an SN2 process but is enabled by single-electron activation of the O2 ligand radical, rather than an electron pair of a nucleophile in the conventional SN2 reaction. Because of the bond cleavage feature and the step is the rate-determining step of the deformylation mechanism, we propose to term the inner-sphere deformylation mechanism as the radical SN2-like mechanism.

Figure 5

(A) Mechanism to break the C–C bond in TS2. (B) Mechanism to break the C–C bond in TS1Ba (see Figure for TS1Ba). The −1 charge state of TS2 and +1 charge states of TS1Ba were omitted for clarity.

Figure 6

Key steps for the reactions of [CoO] with 2-PPA and 2-PPN, along with the relative energies in kcal/mol. The +1 charge states of the Co-containing structures were omitted for clarity.

As compared in Figure , intermediate led by C4–C5 bond cleavage is similar to IM2, except for the Cu–O3 coordination bond in the former, while the Co–N covalent bond in the latter (due to the different preference of oxidation states, Cu(II) versus Co(III)). In particular, the ligand O1–O2 bond in IM2 becomes a single bond (R = 1.45 Å) in IM3; thus, the O–O bond can break via a mechanism depicted by the curly arrows in TS3. Comparing the Mulliken spin populations of IM4 and IM3, the O–O bond cleavage of IM3 leads to an α-electron located on O1, while the process of IM2 leads to an α-electron and a β-electron located at O1 and N, respectively. Thus, IM3 undergoes intramolecular radical–radical coupling to form the N–O bond, completing a [1,3]-sigmatropic rearrangement to give IM4. In comparison, IM4 undergoes intermolecular radical–radical coupling with R, giving IM5 for subsequent processes to complete the deformylation. The occurrence of the formal Cu(III) oxidation state in IM4 can be ascribed to the strong π-donor nature of the oxo ligand. On the other hand, the Cu–O3 distance (2.18 Å) is substantially longer than that (1.93 Å) in the X-ray structure of a Cu(II) complex [CuOC(O)H]− supported by a similar ligand in [CuO],[89] indicating that the Cu–O3 bond is only a weak covalent bond and IM4 is not a well-formed Cu(III) species.

After understanding the differences and similarities between 2-PPA deformylation with [CuO] and acetonitrile activation with [CoO], we further questioned whether [CoO] could also promote deformylations via a similar mechanism, whether [CuO] could activate acetonitrile, and whether the benzyl substituent of 2-PPA plays a role in C–C bond cleavage. To answer these questions, we computed the key steps for the reactions of [CoO] with 2-PPA and 2-phenylpropanenitrile (N≡CCH(Me)Ph, 2-PPN), respectively (Figure ). For the reaction of [CoO] with 2-PPA, the quintet pathway shown in Figure is slightly more favorable than the triplet pathway (see Section SI4 for detailed results). The quintet results are discussed below and the discussion holds true for the triplet case. The C–C bond cleavage is similar to that of [CuO] (see Figure ). First, the coordination complex IM1A is similar to IM2. Second, IM1A also favors C–C bond cleavage over O–O bond cleavage, albeit to less extent, compared to the case of [CuO]. As IM3a is 24.0 kcal/mol higher than C–C bond cleavage TS TS2, the energy difference between IM2A and TS1A is much smaller (5.3 kcal/mol). The large difference can be attributed to the different preferred oxidation states of Cu(II) and Co(III) because the preferred Co(III) oxidation state in IM2A avoids the unfavorable effect in IM3a due to the disfavored Cu(III) oxidation state.

For the reaction of [CoO] with 2-PPN, the five-ring metallacycle is similar to IM2, as shown in Figure . Interestingly, IM1B could break its O–O bond via two reaction modes. The mode depicted by TS1Bb is similar to that of TS3 in acetonitrile activation, as shown in Figure , but the mode described by TS1Ba has no counterpart, which breaks O–O and C–C bonds concertedly. Energetically, the latter mode is slightly more favorable than the former. Considering that the aldehyde C–C bond cleavage in IM2 is activated by the single electron on the O2 ligand (Figure A), it is interesting to note that the O2 ligand in (Figure ) possesses no single electron and the N=C bond lies in the ring plane, but can also undergo C–C bond cleavage, which we understand as follows. Referring to Figure B, the homolytic O1–O2 bond cleavage results in O2 with β radical character (denoted (O2)• hereafter). Radical (O2)• could activate either of its neighboring C4–C5 bond or C4=N3 bond, depending on which activation is more feasible. If the substituent is a good radical-leaving group such as a benzyl group (vide infra), the (O2)• radical would prefer to activate the C–C bond (see SOBMO151Co of TS1Ba), forming the C4=O2 double bond and releasing an alkyl radical. Referring to Figure , if the substituent is a poor radical-leaving group such as a methyl group in , the (O2)• radical would prefer to act on the C=N π-bond, forming a C4=O2 double bond and leaving a single electron on the nitrogen center in IM3. This activation mechanism explains why the C–C and O–O bond cleavages of takes place concertedly because there is a need to generate the reactive (O2)• radical for the subsequent C–C or C=N bond activation. As demonstrated by the much more stable IM2B than IM2B, the C–C bond cleavage to release a benzyl radical has the advantage to avoid the destabilizing effect due to the location of a single electron on nitrogen to maintain the preferred Co(III) oxidation state. These studies indicate that the benzyl substituent of 2-PPA and 2-PPN plays a role in allowing the C–C bond cleavage, which will be analyzed in detail later.

The activation of acetonitrile and 2-PPN with [CuO] was also examined. However, no intermediate similar to either IM2 or IM2 could be located, probably due to the following reason. Because the anionic [CuO] is less electrophilic than the cationic [CoO], the former could not activate nitrile as effectively as the latter through nitrile coordination, which makes the N≡C π bond activation difficult for the O2 ligand radical of [CuO].

On the basis of the discussion above, we envision that [CoO] could react with 2-PPA but [CuO] could not activate acetonitrile. Indeed, it was reported that an analogue of [CoO] could deformylate 2-PPA during our preparation of the paper.[86] A five-membered intermediate characterized spectroscopically (see the box in Figure ) is similar to our IM1A. In addition, we predict that [CoO] may react with 2-PPN in a way different from acetonitrile activation and call experimental verification.

Deformylation Mechanisms of TMA/CCA with

2-PPA has an α-H atom with ketone as the deformylation product. We further computed the deformylation pathways of aldehyde TMA without an α-H atom. In addition, CCA (with an α-H atom) deformylation also afforded a ketone product, providing another case to corroborate our mechanism. Figure shows our computed pathways for the deformylations of TMA and CCA. Proceeding along the blue pathway (PathATMA/CCA), TMA/CCA first undergoes a process similar to that of 2-PPA, breaking the aldehyde C–C bond via TS2 to generate IM2 and an alkyl radical R/R. However, the subsequent processes are different. First, the O–O bond cleavage in the case of 2-PPA takes place directly without the participation of the benzyl radical generated from C–C bond cleavage. Differently, in the case of TMA/CCA, the association of the R/R radical with the O2 atom lowers the O–O bond cleavage barriers by 3.6/3.2 kcal/mol (the energetic difference between TS3a and TS3). Note that, due to the steric hindrance of the ligand, the alkyl radical cannot associate with the O1 atom. We will rationalize the difference in the next section. Second, because of the above difference, the subsequent processes are different. For CCA, due to the association of the alkyl radical R in O–O bond cleavage, the process results in cyclohexyl formate (IM3) and copper-oxo species [CuO]. The two species then react via a sequence including H-atom abstraction via and nucleophilic attack via TS5, giving IM5. Finally, the cleavage of the C–O bond via TS6 leads to the experimental products cyclohexanone and formic acid (eq 3). Note that, an alternative to the nucleophilic attack, IM4 and radical (IM3)• could undergo rebound via TS5a, but TS5a is significantly higher than TS5 and TS6, excluding the possibility. Because TMA does not have an α-H atom, the reaction terminates after O–O bond cleavage, giving the experimental product, tert-butyl formate (eq 2).

Figure 7

Deformylation pathways of TMA and CCA, along with the relative free energies in kcal/mol. The −1 charge states of the Cu-containing structures were omitted for clarity.

We also examined the outer-sphere ONA, α-HAA, and ald-HAA pathways. The nucleophilic attack TS (TS1b, ΔG≠ = 24.1/26.3 kcal/mol) is much higher than the rate-determining C–C bond cleavage TS (TS2, ΔG≠ = 16.5/21.2 kcal/mol). PathBCCA describes the α-HAA pathway. Although the α-HAA TS (TS1a, ΔG≠ = 20.7, 21.4 kcal/mol) is slightly lower than the rate-determining barrier TS2 (ΔG≠ = 21.2 kcal/mol) for C–C bond cleavage, the mechanism encounters a high barrier TS3a (ΔG≠ = 30.2 kcal/mol) for the nucleophilic attack of [CuO] at the enol intermediate. PathCCCA describes the ald-HAA pathway of CCA. Although the ald-HAA TS (TS1c) is only slightly higher than the rate-determining barrier TS TS2, the recombination TS (TS2c, ΔG≠ = 31.8, 32.9 kcal/mol) is much higher than TS2 (ΔG≠ = 21.2 kcal/mol). In the case of TMA, the only alternative is ald-HAA, but TS1a for the process is 5.6 kcal/mol higher than TS2. These results exclude the outer-sphere ONA, α-HAA, and ald-HAA mechanisms for the two deformylation reactions, further affirming the preference of the inner-sphere mechanism.

Key Factors to Enable the Radical SN2-like Deformylation

In addition to 2-PPA (A1), TMA (A2), and CCA (A3), McDonald et al.[18] also carried out deformylations of other aldehydes including propionaldehyde (A4), acetaldehyde (A5), and benzaldehyde (A6), but no reactions were observed for A4–A6 (eq 4). These experimental results encouraged us to find a controlling parameter to enable deformylation at the aldehyde end. On the basis of our proposed inner-sphere mechanism involving aldehyde C–C bond cleavage as the rate-determining step to generate an alkyl radical, we reasoned that the stability of the radical could play an important role in promoting the aldehyde C–C bond cleavage. As shown in Figure , we calculated the bond cleavage barriers (ΔG≠) and the stabilization free energies (ΔG) of the formed radicals relative to benzyl radical R (see Section SI5 for details on how to estimate ΔG). Remarkably, the barriers ΔG≠ are linearly correlated to the stabilization energies ΔG of the radicals with a regression coefficient of 0.95. The excellent linear correlation of ΔG≠ to ΔG demonstrates that the stability of the radical originating from the aldehyde substituent is a key factor for the deformylation reactions. Because radicals R–R are much less stable than R–R, the rate-determining barriers of A4–A6 are significantly higher than those of A1–A3, explaining the experimental observations of no reaction for A4–A6 (eq 4). In addition, the stability of the radical explains why benzaldehyde (A6) could not undergo deformylation, while TMA (A2) could. Because the corresponding phenyl radical (R) of benzaldehyde is 17.6 kcal/mol less stable than the tert-butyl radical (R) of TMA, the former has much higher C–C bond cleavage barrier than the latter (33.5 versus 16.5 kcal/mol). Experimentally, various derivatives of A6 (p-R-PhC(O)H, R = NMe2, OMe, Me, Cl, NO2) were also attempted for their deformylations, but no reaction was observed.[18] Our calculations showed that the radicals corresponding to these derivatives have relative stabilization energies higher than that of A6 (>25.4 kcal/mol, see Section SI5). Therefore, these derivatives would have rate-determining barriers higher than that (33.5 kcal/mol) of A6, explaining the observations of no reaction for these A6 derivatives. The rationalizations of these experimental observations support our proposed radical SN2-like mechanism. Moreover, the stability of the radical also elucidates the aforementioned different O–O bond cleavage mechanisms of TMA and CCA from that of 2-PPA. The corresponding radical (R) of 2-PPA is much more stable than R• and R of TMA and CCA by 7.8 and 12.5 kcal/mol, respectively. Thus, in the case of 2-PPA, the entropic penalty due to associating a benzyl radical to IM3 could override the energetic benefit of radical–radical association (see the process from IM3 to TS3a in Figure ).

Figure 8

Relationship between the rate-determining C–C bond cleavage barrier and the stability of the radicals relative to benzyl radical.

Next, we considered the geometric factor of TM–O2 complexes in enabling deformylations via the inner-sphere mechanism. Examining the structures of [CuO] and [CoO], the inner-sphere mechanism requires TM–O2 complexes to have available room for aldehyde coordination. In this context, we computed 2-PPA deformylation with the porphyrin-ligated [FeO] model complex of CYP51 (Figure A). Previously, Hackett et al. demonstrated the feasibility of the ONA mechanism.[57] In agreement with their study, our calculations predicted that [FeO] favors a low-spin doublet state and the quartet is 6.4 kcal/mol higher than the doublet. Not surprisingly, the geometric optimization to locate a coordination complex similar to IM2 or IM1 or IM2 between [FeO] and 2-PPA drove the two parts apart due to the steric hindrance of the rigid planar porphyrin ligand. Thus, the outer-sphere mechanisms (ONA, α-HAA, and ald-HAA) are possible alternatives. For the ONA mechanism, the nucleophilic attack intermediate IM2a has relative free energy of 7.2 kcal/mol, which is significantly lower than that (24.5 kcal/mol) of the copper counterpart IM2a. The barrier TS2a (ΔG≠ = 18.8 kcal/mol) to break the aldehyde C–C bond of IM2a is also lower than that of the copper counterpart TS2b (ΔG≠ = 22.6 kcal/mol). Furthermore, we examined whether the reaction could undergo α-HAA and ald-HAA mechanisms (Figure B). For the α-HAA mechanism, the TS2b (ΔG≠ = 9.6 kcal/mol) is not high, but the pathway needs to cross a high barrier of 23.8 kcal/mol (TS3b) for nucleophilic attack at the enol intermediate (referring to Scheme B). The TS2c for ald-HAA reaches 28.4 kcal/mol. These barriers are higher than TS2a (ΔG≠ = 18.8 kcal/mol), indicating that the ONA mechanism is most favorable.

Figure 9

(A) Comparing the outer-sphere nucleophilic attack of [CuO] and [FeO] at 2-PPA. (B) Key TSs involved in 2-PPA deformylation with [FeO] via the α-HAA mechanism (TS2b and ) and ald-HAA mechanism (). (C) Outer-sphere nucleophilic attack of [CuO] with a strong nucleophile benzoyl chloride. The values in the parentheses are free energies (in kcal/mol) relative to reactants.

Figure A also compares the outer-sphere nucleophilic attacks of [FeO] and [CuO] at 2-PPA. The NBO charge (q) and Mulliken spin population results of IM2a and IM2a indicate that the processes are essentially different. As the carbonyl oxygen in IM2a is an O– anion with a Mulliken spin population of 0.00 and charge of −0.92 e, that in IM2a is an O• radical with a Mulliken spin population of 0.67 and charge of −0.46 e. In addition, the nucleophilic attack of [FeO] significantly increases the Mulliken spin population at the Fe center from 0.31 in [FeO] to 0.92 in IM2a, while the process of [CuO] only slightly increases the Mulliken spin population at the Cu center from 0.57 to 0.62. According to the Mulliken spin population and NBO charge results of IM2a, we reasoned that the nucleophilic attack of [FeO] takes place via a two-electron mechanism, while the process of [CuO] proceeds via a one-electron mechanism, as compared in the black boxes. Due to the difference, as shown by the Mulliken spin population results of TS2b and TS2a, the C–C bond cleavage of IM2a is homolytic, giving benzyl radicals, while the process of IM2a is heterolytic, giving benzyl anions. In addition to the geometric factor, the stronger nucleophilicity of the dianionic [FeO] than the monoanionic [CuO] should also contribute to these differences.

In the radical SN2-like process, the leaving group is an alkyl radical. Intrigued by the question of what if the substrate is more electrophilic than 2-PPA and bears a good anionic leaving group, in this context, we further studied the nucleophilic attack of [CuO] at benzoyl chloride. Experimentally, it has been found that [CuO] could react with acyl chlorides.[18]Figure C shows the intermediates formed via inner-sphere (IM2, CS denoting closed-shell singlet) and outer-sphere (IM2a) mechanisms, respectively. Interestingly, regardless of the inner- or outer-sphere mechanisms, the attacks are conventional SN2 processes, breaking the C–Cl bond heterolytically; no intermediate similar to IM2 (Figure ) or IM2a (Figure A) could be formed. In addition, the intermediates prefer singlet as the ground state (IM2/IM2a is 15.3/13.4 kcal/mol lower than their triplet counterparts), which is different from the attack of [CuO] at 2-PPA. Although the nucleophilic attack with benzoyl chloride is different from that with 2-PPA, the inner-sphere pathway is still favored, IM2/IM2 being lower than IM2a/IM2a. The preference could be attributed to the Coulomb attraction and the weak coordination between the positive copper center and the carbonyl oxygen in IM2. The results demonstrate that the inner-sphere pathway, which could take the advantage of the favorable interaction between the metal center and the functional group of substrates, should not be overlooked even if a deformylation reaction proceeds via the conventional nucleophilic attack.

Summarizing the discussion in this section, two factors could be crucial for TM–O2 complexes to undergo the inner-sphere mechanism with aldehydes. First, TM–O2 complexes feature the coordination environment that allows aldehyde end-on coordination to the metal center. Second, aldehydes feature a substituent corresponding to a relatively stable radical as a good radical-leaving group.

Rules for the Reactions of TM–O2 Complexes with Aldehydes or Nitriles via the Inner-Sphere Mechanism

On the basis of the present and previous studies,[59,60] we reasoned that the reactions of TM–O2 complexes with aldehydes or nitriles may follow the rules sketched in Scheme if the reactions undergo the inner-sphere mechanism. We clarify that the inner-sphere mechanism does not rule out the possibility that some TM–O2 complexes may favor outer-sphere mechanisms, as exemplified by the [FeO] complex in Figure .

Scheme 3

Rules for the Reactions of TM–O2 Complexes with Aldehydes and Nitriles via the Inner-Sphere Radical SN2-like Mechanism

The inner-sphere mechanism begins with the end-on coordination of the substrate aldehydes/nitriles to the metal center of TM–O2, forming coordination complexes IMI/IMIV. If TM–O2 is a metal–peroxo complex, the coordination renders the O2 ligand to gain a single electron.

For the reactions of TM–O2 with aldehydes, subsequent to the formation of IMI, if the radical R corresponding to aldehyde substituent is a poor radical-leaving group (R-type), the aldehyde C–C bond cannot be broken and the inner-sphere mechanism cannot proceed. If R is a good radical-leaving group (R- and R-type), the O2 ligand radical could activate the aldehyde C–C bond, generating an alkyl radical R and an intermediate IMII. Subsequently, the O–O bond cleavage takes place via two possible modes. If the R radical like a benzyl radical (R-type) is very stable, the O–O bond breaks directly via TSI-type TS (e.g., TS3 in Figure ), giving IMIII. Then, radical R and IMIII couples, followed by intramolecular hydrogen abstraction via TSa, finally affording the products, ketone and formic acid. If R is medially stable (R-type), the association of radical R with O could aid O–O bond cleavage, taking place via TSII-type TS (e.g., TS3 in Figure ). The O–O bond cleavage leads to formate ester. If the aldehyde (e.g., TMA) does not have an α-H, the reaction stops at the step and gives formate ester as the final product. If the aldehyde possesses an α-H, the M=O species abstracts the α-H in the formate ester via TSb (e.g., TS4 in Figure ), giving M–OH species, then the reaction affords the product.

For the reaction of TM–O2 with nitriles, the coordination complex IMIV first isomerizes to a five-ring metallacycle IMV via O2 ligand radical activation of the nitrile group, forming the metal–N and C–O bonds, followed by O–O bond cleavage. If the R group of a nitrile is a poor radical-leaving group like R-type, the O–O bond breaks via TSIII-type TS (e.g., TS3 in Figure ), followed by intramolecular radical–radical coupling, finally leading to the nitrile activation product. If the R group is a good radical-leaving group like R-type, IMV breaks O–O and C–C bonds concertedly via TSIV-type TS (e.g., TS1Ba in Figure ). For the nitriles with R-type substituent, TSIII- and TSIV-type modes would compete. To our knowledge, the activation of nitrile with R-type substituents has not been studied experimentally. It would be interesting to characterize the final products, and we invite experimental study to verify our rules.

Conclusions

In summary, we have performed DFT study of the deformylation mechanism of aldehydes with copper(II)–superoxo as a deformylating reagent. In contrast to the known outer-sphere mechanisms, including nucleophilic attack, α-hydrogen atom abstraction, and aldehyde hydrogen atom abstraction, the study found that the present deformylations prefer an inner-sphere mechanism that could be termed as the radical SN2-like mechanism. The mechanism proceeds via a sequence mainly consisting of end-on aldehyde coordination, aldehyde C–C bond cleavage, O–O bond cleavage, and α-hydrogen abstraction, finally giving ketone or formate ester as the product. The aldehyde C–C bond cleavage is the rate-determining step, which is achieved by the activation of the O2 ligand radical via an SN2-like fashion. The C–C bond cleavage greatly facilitates the subsequent O–O bond cleavage. Because the rate-determining step results in an alkyl radical corresponding to an aldehyde substituent, the stability of the alky radical is a key factor to enable the inner-sphere mechanism. In addition, the metal–dioxygen complex should possess a coordination environment suitable for aldehyde end-on coordination. Taking the deformylation mechanism and the mechanism for acetonitrile activation with cobalt(III)–peroxo into considerations together, we have derived reaction rules for deformylations and nitrile activations with metal–dioxygen complexes and identified four modes for breaking the O–O bond in metal–dioxygen complexes, depending on the properties of the substrates and the metal–dioxygen complexes. This study demonstrates that the inner-sphere mechanism could be favored over the known outer-sphere mechanism and should not be overlooked when studying the reactivity of TM–O2 complexes. Expectedly, the inner-sphere mechanism, as well as the reaction rules, will complement the known outer-sphere mechanisms, thus offering a new way of thinking for understanding the reactivity of metal–dioxygen core-related biological processes such as deformylations and the discovery of metal–dioxygen catalysts for catalytic aerobic transformations.