Epoxidation of unfunctionalized olefins by Mn(salen) catalyst using organic peracids as oxygen source: A theoretical study

Computational Details

Because quantum mechanical studies on the epoxidation in the real system containing numerous bulky substituents on the salen catalyst and aryl-conjugated olefin are prohibitive, we used a chemically relevant model. Our model comprised fully conjugated salen catalyst in (R,R) conformation without any bulky substituents and ethylene as olefin probe. We used peracetic acid as an oxygen source. To study the axial-ligand effect we used imidazole (because it is known to enhance enantioselectivity of epoxidation) (30–32) and acetate (the natural by-product from the peracetic acid). We believe that this model should allow us to identify the true reactive species and adequately describe the mechanism of the epoxidation. Although it may be argued that no enantioselectivity can be achieved with such a model, we find that this model provides qualitative understanding of the origin of enantioselectivity, which can be elaborated further with more realistic models in the future.

All calculations were performed with the gaussian 03 program suite (36) with the density functional theory (DFT), specifically with the Becke three-parameter hybrid functional combined with Lee–Yang–Parr correlation functional (B3LYP) (37–39). Preference for the B3LYP for the Mn(salen) systems has been justified in our recent studies (22, 40–42). Throughout this work we used a double-ζ LANL2DZ (BS1) (43) basis set with associated Hay–Wadt nonrelativistic effective core potential on Mn (44, 45) for geometry optimizations using the unrestricted B3LYP method.‡ Unless otherwise stated, for the optimized structures we performed and used in discussion single-point energy calculations with triple-ζ quality basis set BS2, comprised of Stuttgart–Dresden basis set (8s7p6d1f)/[6s5p3d1f], with associated effective core potential on Mn (46), and 6-311+G(d) basis set (47) on the other atoms. Where applicable, spin contamination has been projected out by using the Yamaguchi approach (48, 49). Vibrational frequency calculations at the B3LYP/BS1 level have been performed only for the transition states to confirm their nature. Therefore, the energetics presented below does not include zero-point energy or entropy corrections.

Results and Discussion

The present study focuses on the Mn(salen)-catalyzed epoxidation of the ethylene by peracetic acid. For the reference, the barrier of the bimolecular catalyst-free Prilezhaev epoxidation by peracetic acid is calculated to be 16.4 kcal/mol (≈17 kcal/mol in ref. 50). Thus, below we disregard the pathways of the Mn(salen)-catalyzed epoxidation, which depend on bimolecular elementary steps with barriers >17 kcal/mol.

The O—O Bond Cleavage in Mn(salen) Acylperoxo Complexes to Form oxo Complexes. Interaction of peracids with the chiral Mn(salen) catalyst produces a number of acylperoxo complexes of different geometric structure and oxidation state of the central Mn atom, as shown at the top of Fig. 1 (42). In Fig. 1 we use Mn^N–cisYZ and Mn^N–trans-L notation, where Y and Z stand for the atoms of the salen ligand situated trans to peroxo oxygen and carbonyl oxygen of the acylperoxo ligand, respectively, L represents an exogenous ligand, and N reflects the oxidation state of the Mn. The most important geometric isomers are trans, cisON, and cisNO and the oxidation states are II, III, and IV. These acylperoxo complexes can produce various oxo species via O—O bond cleavage, as discussed below and summarized in Fig. 1.

Mn^II complexes. Mn^II–acylperoxo complexes cleave the O—O bond to produce exclusively Mn^IV-oxo species. The lowest barriers on the ground sextet-state potential energy surface in the absence of an axial ligand are 14.6 and 17.3 kcal/mol for cisON and cisNO isomers, respectively. With the axial NH₃ ligand (model of imidazole), the barrier is 16.6 kcal/mol in Mn^II–trans-NH₃. The relative energies of the product Mn^IV–cisON, –cisNO, and –trans-NH₃ oxo species in the ground quartet state are –9.7, –12.8, and –25.3 kcal/mol, suggesting a spin crossover during the reaction.

Mn^III complexes. The O—O bond cleavage in Mn^III–acylperoxo complexes has been discussed elsewhere (41, 42). In summary, in the absence of axial ligand the O—O bond cleaves via a quintet-triplet spin-crossover pathway with overall 16.5 and 18.4 kcal/mol activation for the cisON and cisNO isomers to produce corresponding Mn^V–cis-oxo species in the triplet state, with relative energies of 2.6 and 2.0 kcal/mol, whereas formation of trans-oxo species is unfavorable.

In the presence of the axial ligand L = NH₃ and HCO^– ₂ (models of imidazole and acetate), the O—O bond of the Mn^III–trans-L-acylperoxo complexes cleaves with overall 13.7 and 12.4 kcal/mol activation energy to produce chameleon trans-L-oxo species with 8.4 and 5.6 kcal/mol relative energies. By “chameleon” we imply that product trans-oxo species can be regarded as either associated Mn^IV–trans-L-oxo and RCO₂ radical or Mn^V–trans-L-oxo and RCO₂ anion.

Mn^IV complexes. The Mn^IV–acylperoxo complexes in their ground quartet state are inert to the O—O bond cleavage and consequently have been characterized by ESR (14–16, 19), UV/visible, Raman, extended x-ray absorption fine-structure, and x-ray absorption near-edge spectroscopies (12).

Potential oxidants for olefin epoxidation. The acylperoxo complexes may participate in the olefin epoxidation directly or through various oxo species. Fig. 1 depicts all potential oxidants in the presence and absence of axial ligands.

Although our calculations show formation of the square pyramidal Mn^V–trans-oxo species to be impractical (42), this hypothetical “naked” oxo species helps to appreciate the axial-ligand effect and has been studied theoretically (5, 6, 22–25). The ground state of the Mn^V–trans-oxo species has been predicted to be singlet, triplet, or quintet. An axial ligand L such as imidazole or acetate destabilizes the singlet state of the octahedral Mn^V–trans-L-oxo by 12–13 kcal/mol, leaving either triplet or quintet as the ground state (22). Furthermore, the axial ligand changes the triplet state with two α electrons localized mainly on Mn to the triplet state with three α on Mn and one β on O³, the antiferromagnetic analog of the quintet state. Energy-wise, the triplet state for the acetate and imidazole is 1.3 kcal/mol lower and 0.3 kcal/mol higher than the quintet, respectively. Similar effects have been reported for acetonitrile and N-oxide axial ligands (23). Because the four unpaired electrons in the triplet and quintet states reside in similar orbitals and only differ by spin coupling, (22) we suppose that both states would have similar reactivity. Therefore, below we discuss the epoxidation mechanism by Mn^V–trans-L-oxo species in the quintet state.

The Mn^V–cis-oxo species have a triplet ground state. Similar to the triplet state of Mn^V–trans-L-oxo species the cis oxo species have three α electrons on Mn and one β electron on O³ (22, 42). However, the energy difference between the antiferromagnetic triplet and ferromagnetic quintet states in the cis oxo species is substantially larger than in the trans case. In particular, the quintet states of the cisON and cisNO oxo species are 4.6 and 7.3 kcal/mol higher than the corresponding triplet states. Despite high relative energy, the epoxidation by the quintet cis oxo species is of considerable interest, because the final complex of the Mn^III(salen)L with the epoxide has the quintet ground state (22).

The Mn^IV–trans-L-oxo species with and without an axial ligand as well as the Mn^IV–cis-oxo species have quartet ground states. The lowest doublet and sextet states are substantially higher (≈7–17 kcal/mol) even in the absence of an axial ligand (22). Therefore, we consider only quartet ground state in the epoxidation.

Below, we study the epoxidation by acylperoxo and oxo complexes, respectively.

Epoxidation by Acylperoxo Complexes. Here we discuss direct epoxidation of the ethylene by acylperoxo complexes. Fig. 2 introduces the notation for olefin approaches to the reactive MnO³ fragment. We label the approaches as K_M, where K is N¹, N², O¹, O², or O⁵ and is used to designate the orientation of the O³–C¹ vector along the salen molecule. The C¹ is the ethylene carbon atom forming the bond with O³. Two letters, such as N¹N², are used for K to represent the midpoint between the two atoms. The label M is A(–90°), B(90°), D(0°), or U(180°), and reflects the MnO³C¹C² dihedral angle (in parentheses) to specify orientation of the olefin along the approach vector O³–C¹.

Epoxidation by acylperoxo complexes in the absence of axial ligand. For each geometric isomer and each oxidation state of cis acylperoxo complexes in the corresponding ground spin state, we found structurally similar transition states with olefin approaching in a side-on fashion with B orientation, namely O²_B for cisON and O¹O²_B for cisNO, as summarized in Fig. 3.

Following the reaction coordinate from the transition states to the products indicates that the epoxidation by acylperoxo complexes in the ground spin states is asynchronous (O³C¹ distance is shorter than O³C², as seen in Fig. 3) concerted, regardless of the oxidation state. However, because these epoxidation pathways are bimolecular with barriers >17 kcal/mol, they are unlikely to compete with Prilezhaev epoxidation.

The transition states for the epoxidation by trans-acylperoxo complexes in the high spin states seem to have higher barriers than for the cis-acylperoxo complexes and have been discarded.

We also sought epoxidation transition states by the Mn^III–cis-acylperoxo complexes in the low-lying excited triplet state, which is the most reactive in the O—O cleavage. However, instead of direct epoxidation we found olefin assisted O—O bond cleavage transition states producing cis oxo species with 14.9 and 17.2 kcal/mol barriers for cisON and cisNO isomers, respectively. These barriers are lower than in the absence of the olefin (16.5 and 18.4 kcal/mol, respectively).

Epoxidation in the presence of an axial ligand. We do not discuss direct epoxidation by the trans-L-acylperoxo complexes, because in the presence of the axial ligand the O—O bond cleavage has overall 12.4–13.7 kcal/mol activation energy even without the substrate assistance (41). Thus, trans-L-acylperoxo complexes should effectively convert to the trans-L-oxo species.

In brief, direct epoxidation by acylperoxo complexes in any oxidation state cannot compete with catalyst-free Prilezhaev epoxidation. On the other hand monomolecular O—O bond cleavage in Mn^II– and Mn^III–cis and –trans-L-acylperoxo complexes to produce corresponding oxo species is facile. Therefore, below we investigate epoxidation by the oxo species.

Epoxidation by oxo Complexes. Here we elucidate the epoxidation mechanism by various oxo species. We first address epoxidation by Mn^V–cis-oxo species (potential oxidants in the absence of axial-ligand) and then discuss epoxidation by Mn^V- and Mn^IV–trans-L-oxo species (potential oxidants in the presence of axial ligand). To elaborate the axial-ligand effect on the enantioselectivity we compare the latter case with Mn^V–cis- and hypothetical trans-oxo species without axial ligands.

Epoxidation by cis-oxo complexes. In the preceding paragraph we assumed that the pattern of reactivity for quintet ferromagnetic and triplet antiferromagnetic states of the Mn^V-oxo species should be alike due to similarity of the electronic structure. We verify this assumption by studying both spin states rigorously, although quintet cis oxo species should not contribute to enantioselectivity (42). Table 1 summarizes all transition states optimized with a variety of olefin approaches and orientations for the Mn^V–cis-oxo complexes, in the quintet and triplet states, with their important geometrical parameters, imaginary vibrational frequencies, and relative energies. Recall that the relative energies of the triplet and quintet states of cisON-oxo species are 2.6 and 7.2 kcal/mol, whereas those of cisNO-oxo species are 2.0 and 9.3 kcal/mol, respectively.

The cisON-oxo species resembles trans-L-oxo species in orientation of the reactive MnO³ fragment and is discussed first. For the quintet-state cisON-oxo species we found two pairs of pseudosymmetric side-on approaches {O⁵_A; O²_B} and {O⁵_B; O²_A}, as shown in Table 1. The carboxylate ligand in the cisON oxo species is not involved in any hydrogen bonding with the chiral diimine bridge and readily interacts with the incoming olefin, which results in large differentiation within pairs of pseudosymmetric approaches. Specifically, the quintet energies for {O⁵_A; O²_B} and {O⁵_B; O²_A}, respectively, are {13.6; 16.2} and {13.8; 17.0} kcal/mol relative to the quintet cisON-acylperoxo complex.

For the triplet-state cisON-oxo, as expected, we found approaches {O²O⁵_A; O²_B} and {O⁵_B; O²_A} similar to those of the quintet state. However, the relative energies of these transition states, {7.6; 6.9} and {7.2; 7.5} kcal/mol, are substantially lower than those of the quintet states. Surprisingly, the energy differences within the triplet-state pairs are qualitatively different from the quintet case. This finding suggests that spin state may actually affect the enantioselectivity of the epoxidation.

The cisNO-oxo species is very different from the cisON and trans-L-oxo species. The most prominent feature of the cisNO-isomer is that the reactive MnO³ fragment is coplanar to both nitrogen atoms of the chiral diimine bridge and is consequently remote from the chiral centers (see Fig. 1). Furthermore, the carboxylate ligand forms a hydrogen bond with one of the chiral centers of the diimine bridge. Perhaps the most important feature is that the O¹N¹ side phenyl ring is bent toward the MnO³ fragment such that it completely prevents side-on olefin approach from the O¹N¹ side (see Fig. 1). Last, the alkyl group R of the carboxylate is very close to the MnO³ fragment and provides additional restrictions to the olefin approach, which effectively divides the approaches into two groups, one from the O¹O² side and the other from the N¹O⁵ side.

For the quintet-state cisNO-oxo species we found two pseudosymmetric pairs of olefin approaches, {O¹O²_A; O¹O²_B} and {N¹O⁵_A; O⁵_U}, shown in Table 1. The relative energies of these transition state are {15.9; 17.0} and {14.8; 16.6} kcal/mol, respectively. Similar to the cisON, several hydrogen bonds are involved in latching the olefin.

For the triplet-state cisNO species we found similar pairs of transition states {O¹O²_A; O¹O²_B} and {N¹O⁵_A; N¹O⁵_B}. The energies for these pairs are {6.4; 5.9} and {7.7; 9.9} kcal/mol, respectively.

The most favorable epoxidation transition states for both triplet cis oxo species have the olefin in the B orientation. Although triplet-state cisON- and cisNO-oxo species have comparable relative energies for epoxidation, the barrier for the formation of the cisNO oxo species is 1.9 kcal/mol higher than for cisON, suggesting that cisNO species might be less important in the epoxidation.

Epoxidation by trans-L-oxo complexes. In the presence of axial ligands the O—O bond cleavage results in the chameleon trans-L-oxo species, which acts as either Mn^V– or Mn^IV–trans-L-oxo species depending on the axial ligand and reaction conditions (41). Here we investigate the asymmetric induction step by isolated Mn^V– and Mn^IV–trans-L-oxo species with L = imidazole and acetate.

For the quintet-state Mn^V–trans-L-oxo species we find that with neutral axial ligand imidazole, the epoxidation is asynchronous (O³C¹ and O³C² distances differ by as much as 0.7 Å) concerted, whereas with the anionic acetate, the reaction switches to a stepwise (forming Mn^IV radical intermediate) mechanism. The latter agrees with previous studies on a smaller model with axial ligand Cl^– (5, 7–9). The epoxidation by Mn^V–trans-L-oxo species in the triplet spin state is stepwise regardless of the axial ligand.

We located two pairs of perfectly side-on pseudosymmetric olefin approaches, {O¹_B; N²O²_A}, {O¹_A; O²_B}, for the quintet Mn^V–trans-L oxo with both ligands, as seen in Table 2. The energy differences within the pairs of olefin approaches with axial ligands are 0.6–0.9 kcal/mol even for ethylene, which is an increase from 0.1 kcal/mol for the hypothetical naked Mn^V–trans-oxo species (refer to Table 3, which is published as supporting information on the PNAS web site). Because the ethylene in all the transition states samples only one phenyl ring at a time, the only explanation to the increase in the energy difference is the folding of the catalyst along the NO edges due to the axial ligand. With the axial ligand, the N²O² side phenyl ring is bent upward, disfavoring olefin approaches along that side. This finding supports the suggestion that asymmetric folding may enhance the enantioselectivity of the epoxidation (23, 29).

For the quartet ground state of Mn^IV–trans-L-oxo species epoxidation is asynchronous stepwise regardless of the axial ligand. For Mn^IV–trans-L-oxo species we also found two pseudosymmetric pairs of approaches, {O¹O²_A; O¹O²_B} and {N²_B; N¹_A}; however, these approaches are drastically different from those of Mn^V–trans-L-oxo species. Although approaches {O¹O²_A; O¹O²_B} sampling the twist of the phenyl rings have the lowest energy, the energy difference between the two is extra small (0.2 kcal/mol). On the other hand, the approaches {N¹_A; N²_B}, which sample the asymmetric diimine bridge, have an energy difference of 0.5–0.6 kcal/mol, which is comparable with the energy induced by the bend in the catalyst observed in Mn^V–trans-L-oxo species. Note that, because epoxidation by the Mn^IV–trans-L-oxo species is a bimolecular reaction with the activation energies >18 kcal/mol, it should not contribute to enantioselectivity. For the same reason, we do not discuss epoxidation by Mn^IV–cis-oxo species.

In contrast to Mn^V–cis-oxo species, the most favorable epoxidation transition states for the Mn^V–trans-L-oxo species have the olefin approach in the A orientation. Taking into account that substituted olefins add to MnO³ regioselectively, the change of orientation from B to A due to axial ligand in our model corresponds to inversion of absolute configuration of epoxide with prochiral olefins.

Final Ring Closure of the Radical Intermediate. Let us briefly discuss the evolution of the system beyond the first transition state TS1 in the stepwise mechanism, which is relevant to the cis-trans scrambling of the epoxides and regeneration of the catalyst. The ring closure in the radical intermediate has been studied by others for Mn^V–trans-oxo species on smaller models (5–9).

Ring closure in the cis-oxo species. In the case of Mn^V–cis-oxo species, both quintet and triplet spin states form radical intermediates, with one unpaired electron on the C² atom coupled to three unpaired electrons on the Mn atom in ferromagnetic and antiferromagnetic fashion, respectively. Because the coupled electrons are well separated, the energies of these intermediates in both spin states are very similar. For the cisON intermediates the energies are –11.4 and –11.3 kcal/mol, whereas for the cisNO isomer they are –11.0 and –11.2 kcal/mol, in the quintet and triplet states, respectively. The transition state for the ring closure (TS2) is energetically higher for the triplet state, which is an indication of the spin crossover. In particular, for cisON the relative energies of the TS2 are –8.7 and –3.7 kcal/mol, whereas for cisNO they are –9.0 and –6.9 for the quintet and triplet states, respectively. The final product is the epoxide weakly coordinated to the salen complex. Here the difference between the quintet and triplet energies is the largest. Namely, for cisON and cisNO the relative energies of the quintet products are –34.8 and –37.3 kcal/mol, whereas those for the triplet products are –21.1 and –30.8 kcal/mol.

Ring closure in the trans-L-oxo species. For L = imidazole the reaction in the quintet state is concerted and leads directly to the epoxide with retention of the original stereoconfiguration. For the Mn^V–trans-L-oxo species with L = acetate, the Mn^IV–trans-L radical intermediates in the quintet and triplet states are –16.9 and –16.8 kcal/mol, respectively, and the ring closure TS2s are –14.7 for the quintet state and slightly higher –12.3 kcal/mol for the triplet state. The final product has stronger preference for the quintet state, which is –40.0 kcal/mol, whereas the triplet product is –31.6 kcal/mol.

Because epoxidation by Mn^IV–trans-L-oxo species is unlikely, we do not discuss the ring closure in the corresponding Mn^III–trans-L-radical intermediate. Nevertheless details of this transformation are provided in Supporting Text, which is published as supporting information on the PNAS web site.

Concluding Remarks

To summarize, we have systematically studied the mechanism of the Mn^III(salen)-catalyzed epoxidation by peracids. We found that direct epoxidation by various acylperoxo complexes in three different oxidation states is unlikely because of higher activation energy than the noncatalyzed Prilezhaev epoxidation. Instead, the monomolecular O—O bond cleavage in the acylperoxo complexes occurs first, followed by epoxidation by the product oxo species. We find that only Mn^V–cis- and Mn^V–trans-L-oxo species can compete with the Prilezhaev reaction. The proposed reaction mechanism for epoxidation in the absence and presence of axial ligands is schematically drawn in Fig. 4 A and B, respectively. In addition, the top of Fig. 4A delineates a representative potential energy surface of the epoxidation reaction for the Mn^III–cisON-acylperoxo complex. Based on the activation energies we predict that the absolute rates for the O—O bond cleavage and subsequent epoxidation may be comparable depending on the concentrations of the catalyst, peracid, and olefin. To verify this prediction, experimental kinetics data on the epoxidation rates would be indispensable. Unfortunately, recent experimental mechanistic studies of the Mn^III(salen)-catalyzed epoxidation by peracids lack solid kinetics data due to very fast (within 1 min) epoxidation rates even at –78°C (14, 16–19, 35). Nevertheless, for closely related Mn^III(porphyrin) catalysts the rates of the Mn^V oxo species formation and/or subsequent epoxidation have been estimated in water and acetonitrile at room temperature and agree with our predictions (51, 52).

Importantly, we find that only for Mn^V–trans-L-oxo species with neutral imidazole axial ligand and only in the high-spin quintet state epoxidation proceeds in a concerted fashion with retention of original stereoconfiguration. In all other cases, epoxidation proceeds in a stepwise manner through an intermediate radical, potentially leading to cis-trans scrambling. Thus our results provide rationale for the experimentally observed sharp increase in the enantioselectivity of epoxidation of the terminal olefins by peracids at low temperature and in the presence of neutral axial ligand N-methylmorpholine-N-oxide (33–35). We explain that the dominant active oxygenating species in that epoxidation at low temperature is quintet Mn^V–trans-L-oxo species with L = N-methylmorpholine-N-oxide, which epoxidizes terminal olefins concertedly by analogy with the L = imidazole case.

Besides the mechanism of the epoxidation, we addressed the origin of enantioselectivity in the epoxidation by reactive Mn^V–cis and Mn^V–trans-L-oxo species within the limits of our model. For all oxo species in all considered spin states, we found multiple approaches of the olefin to the reactive MnO³ fragment. We call these approaches native, because our model is missing bulky substituents on the salen ligand and olefin substrate.

In the absence of an axial ligand the active oxygenating species are Mn^V–cis-oxo species, with the most favorable native olefin orientation being B, but in the presence of an axial ligand, the active oxygenating species change to Mn^V–trans-L with the most favorable olefin approach changed to A. Taking into account the fact that substituted olefins add to the MnO³ fragment regioselectively, this provides an explanation to the experimentally observed inversion of the absolute configuration of the epoxidation due to axial ligands (30–32).

Additionally, comparison of the native approaches for Mn^V– trans-L and hypothetical naked Mn^V–trans-oxo species suggests that asymmetric folding of the salen ligand in Mn^V–trans-L-oxo species induced by the axial-ligand strain increases the energy difference within the pairs of pseudosymmetric approaches and might provide a guidelines for the choice of axial ligands to enhance enantioselectivity of epoxidation (23, 29).

Our unsubstituted model demonstrates that enantioselectivity of epoxidation for small alkenes by all considered oxo species is low. Also, available experimental data show that enantioselectivity of epoxidation of the small alkenes remains low even for the best-substituted salen catalysts. Consequently, it is reasonable to suppose that larger olefins show better enantioselectivity due to multiple interactions with the critical chirality regions of the catalyst. Thus, investigation of the asymmetric induction requires accurate description of the weak noncovalent interactions between the olefin and the substituted catalyst. One of such interactions is dispersion-type π–π interaction of the aryl-substituted olefins with the catalyst. In such case, the results from molecular mechanics or quantum mechanics/molecular mechanics studies on the subject must be taken with extra caution, because the dispersion interaction may not be described accurately. Even full DFT calculations will not be adequate, because the dispersion is not accounted for by DFT methods. An accurate description of the dispersion-type interaction in transition metal systems is a challenge and requires at least an MP2 level of theory. Nevertheless, it may be achieved with an oniom (MP2:DFT) approach (53). Therefore, the present work is essential for near-future investigations of the origin of the enantioselectivity with rigorous description of the crucial weak interactions.