Synlett 2010(7): 999-1026  
DOI: 10.1055/s-0029-1219790
ACCOUNT
© Georg Thieme Verlag Stuttgart ˙ New York

Recent Mechanistic Insights in the Singlet Oxygen Ene Reaction

Mariza N. Alberti, Michael Orfanopoulos*
Department of Chemistry, University of Crete, 71003 Voutes Campus, Heraklion, Crete, Greece
Fax: +30(2810)545001; e-Mail: orfanop@chemistry.uoc.gr;

Further Information

Publication History

Received 11 November 2009
Publication Date:
25 March 2010 (online)

Biographical Sketches

Mariza N. Alberti was born in Athens, Greece in 1978. She received her BSc degree (1st Class Honors) in 2001 from the University of Crete. She obtained her MSc degree in 2004 and her PhD in 2007 from the University of Crete under the direction of Professor M. Orfanopoulos. She is currently working as a postdoctoral researcher at the same university. Her research activity has been mainly carried out in the field of the mechanistic elucidation of electrophilic addition reactions and the study of photocatalytic processes.
Michael Orfanopoulos received his BSc from the University of Patras, his MSc from the University of Toledo, and his PhD from Case Western Reserve University in 1979. He was a postdoctoral fellow at Stanford University with Professor H. S. Mosher and at UCLA with Professor C. S. Foote. In 1982 he became a research scientist at the NRC Demokritos, Athens, Greece, and in 1985 he joined the University of Crete, where he is currently Professor of Organic Chemistry. His research focuses on: (a) chemical and photochemical catalysis involving polyoxometalates and reactive oxygen species including singlet oxygen, and (b) fullerene-based carbon nanostructures, photosynthesis of fullerene and het­erofullerene derivatives.

Abstract

Singlet oxygen reacts with alkenes, which bear allylic hydrogens, in an ene fashion to afford allylic hydroperoxides. This reaction, apart from its synthetic usefulness, has received extensive mechanistic attention. Numerous experimental studies (e.g., trapping of intermediates, deuterium kinetic isotope effects, regio- or stereoselectivity studies etc.) and to a lesser extend computational work, support a stepwise mechanism with the formation of a three-membered ring (perepoxide-like) intermediate. This Account mainly highlights our group’s earlier and recent experimental efforts to ascertain facts relating to this concept.

1 Introduction

2 Theoretical Calculations

3 Kinetic Isotope Effects

4 Regioselectivity

4.1 ‘cis Effect’ Selectivity

4.2 Anticis Effect’ Selectivity

4.3 The Large Group Nonbonding Effect

4.4 Geminal Selectivity with Respect to Allylic or Vinyl Substituent

4.5 Electron-Withdrawing Group at the α- and β-Position

4.6 Regioselective Self-Sensitized Oxygenation of Fullerene Derivatives

4.7 Solvent and Electronic Effects

4.7.1 Site Selectivity in the ¹O2 Ene Reaction of α,β-Unsaturated Esters and Acids

4.7.2 Site Selectivity in the ¹O2 Ene Reaction of Allylic Alcohols

4.7.3 Syn Selectivity of β,β-Dimethylstyrene

4.7.4 Site Selectivity of Isobutenylarenes

4.7.5 ‘Push-Pull’ Electronic Effect

5 Studies on Diastereoselectivity

5.1 Diastereoselectivity in Self-Sensitized Oxygenation of a Fullerene Derivative

5.2 Diastereoselectivity in the ¹O2 Ene Reaction of Chiral Functionalized Alkenes

6 Stereochemistry

7 Hypersensitive Probes in the ¹O2 Ene Reaction

8 The ¹O2 Ene Reaction in Confined Media

9 Concluding Remarks

1 Introduction

Singlet molecular oxygen (¹O2, ¹ Δg) [¹] plays a crucial role in the degradation of materials, [²] as well as in biological [³] and therapeutic processes. [4] Over the last several years, the use of ¹O2 as a reagent in organic synthesis has been receiving continuous and remarkable attention. [5] It is instructive to note that although ¹O2 was discovered more than 70 years ago, [6] until the early 1960s was considered to be of rather limited interest as a research subject. Then, it was the pioneering work of Foote and Wexler [7] that provided evidence for the formation of ¹O2 as the reactive intermediate in solution via two independent routes: (1) a photochemical reaction and (2) a chemical reaction by using a hypochlorite/hydrogen peroxide system. This breakthrough brought ¹O2 into the mainstream of chemical research. Since then, several reaction modes of ¹O2 with unsaturated organic molecules have been reported. The more classical ¹O2 reactions are: (1) the [4+2] cycloaddition [8] [9] with conjugated dienes or anthracenes to yield endoperoxides, (2) the [2+2] cycloaddition [8] [¹0] with enol ethers, enamines or electron-rich alkenes to yield 1,2-dioxetanes, and (3) the so-called ene or Schenck [¹¹] reaction with alkenes to form allylic hydroperoxides.

Among the various types of ¹O2 reactions, the ene reaction has arguably drawn the most extensive experimental and theoretical attention. [¹²] The title reaction has been applied to the asymmetric synthesis of α-hydroxy ketones or α-hydroxy aldehydes as well as diols, from the corresponding ketones or aldehydes in the presence of catalytic amount of an amino acid. [¹³] Furthermore, the allylic hydroperoxides that result from the ¹O2 ene reaction have proven to be synthetically useful intermediates. Indeed, ¹O2-mediated allylic oxidation was successfully used as an essential step in the synthesis of several natural products, such as the 9,13-diepoxy labdane diterpene amoenolide K, [¹4] the enantiomerically pure lactones dihydromahubanolide B and isodihydromahubanolide B, [¹5] (±)-proto-quercitol, [¹6] (±)-vibo-quercitol, [¹7] (+)-asteriscanolide [¹8] and the second-generation litseaverticillols D-G, I and J. [¹9]

It is worth mentioning that the ¹O2 ene reaction was applied towards the synthesis of naturally occurring artemisine, which contains a 1,2,4-trioxane subunit (indicated with red in Scheme  [¹] ). [²0] Current research focuses on the synthesis as well as the antimalarial assessment of structurally simplified 1,2,4-trioxanes. The ¹O2 ene reaction followed by peroxyacetalization provides an attractive approach for this purpose. For instance, the photooxidation of allylic alcohol 1 afforded the diastereomeric β-hydroperoxyalcohols 2a,b (Scheme  [¹] ). [²¹] Interestingly, this and related photooxidations were performed in a more environmentally friendly manner by using tetraarylporphyrines embedded in polystyrene beads. [²¹] [²²] The final products 3a-d were obtained by acetalization with the appropriate ketones. In a similar manner, Singh and co-workers prepared a wide number of 1,2,4-trioxanes [²³] as well as 1,2,4-trioxepanes (seven-membered rings) and 1,2,4-trioxocanes (eight-membered rings). [²4] Moreover, several of these 1,2,4-trioxanes showed significant antimalarial activity. [²³a-d ] The aforementioned synthetic route was also used for the preparation of more complex molecules 4-6 which contain a 1,2,4-trioxane substructure (indicated with blue color in Scheme  [¹] ). [²5] Recently, this efficient synthetic approach was successfully used for the preparation of a series of chiral 1,2,4-trioxanes. [²6]

Scheme 1 Synthesis of 1,2,4-trioxane moiety via the ¹O2 ene reaction

The mechanistic details of the ¹O2 ene reaction, although studied for many years, are still a subject of controversy. The main question being whether the reaction is concerted or involves intermediates. A concerted mechanism in which the characteristic bond shifts take place through a six-membered ring transition state (7, Scheme  [²] ) has been favored by many researchers. The initially proposed synchronous pathway [²7] was challenged by a stepwise mechanism [²8] involving either an open biradical (8), [²9] an open zwitterion/dipolar (9) [³0] or a perepoxide (10) [³¹] intermediate. Some researchers consider, instead of the polar perepoxide, an exciplex intermediate (11). [³²] This intermediate constitutes an excited-state charge-transfer complex between ¹O2 and the substrate. Finally, dioxetanes (12) have also been reported as intermediates. [³³] However, the pathway through dioxetanes has been found to give only carbonyl cleavage products rather than rearrange to allylic hydroperoxides. [³4] Among the aforementioned intermediates, in the ¹O2 ene reaction of simple alkenes, the formation of the perepoxide (PE) is the most popular and found support from both experimental and computational work (see above sections). On the other hand, in the photooxidation of electron-rich alkenes, such as enamines, [³5] enol ethers [³6] or esters, [³7] silyl enol ethers [³8] and dienes, [³9] there was a correlation between PE and open zwitterionic intermediates. Additionally, most of these photooxidations afforded mixtures of ene and cycloaddition products, whose relative abundance depended on solvent polarity.

Intermolecular trapping of PE intermediate in the ¹O2 ene reaction with simple alkenes has fascinated researchers for many years. Interestingly, pinacolone, [] sulfoxides, [] phosphites, [] sulfenate and sulfinate esters [44] have been used to trap a PE intermediate in the ¹O2 addition to bi­adamantylidene. Trimethyl phosphite and triphenyl phosphite have also been used to trap intermediates in reactions of ¹O2 with diethyl sulfide [45] and trans-cyclooctane, [46] respectively. Despite these interesting findings, Tonachini and co-workers, based on theoretical studies, found that the trimethyl phosphite mediated abstraction of the terminal oxygen atom from a PE and an open-chain intermediate (such as intermediate 8 or 9) affords the same epoxide product. [47] Accordingly, they suggested that trapping experiments cannot discriminate between a PE and an open-chain intermediate pathway.

Scheme 2 Proposed mechanisms for the ¹O2 ene reaction

Importantly, kinetic isotope effects with tetrasubstituted, [³¹] [48] trisubstituted [49] and cis-disubstituted [50] alkenes ­revealed that the ¹O2 ene reaction proceeds via the irreversible formation of a PE intermediate. On the other hand, in similar studies performed on trans-butene-d 3 [] and 2,5-dimethyl-2,4-hexadiene-d 6, [40] a partial equilibration of the PE intermediate with the reactants was postulated. Furthermore, consistent with the intermediacy of a PE in the ¹O2 ene reaction was the lack of Markovnikov-type directing effects, [¹a] [7a] its stereochemical features [] and the observed regioselectivities. []

Despite these seemingly unambiguous experimental results on the ¹O2 ene reaction, in the early 2000s, the PE intermediate was challenged by modern computational studies. [54] Particularly, these calculations indicated a PE- or a biradical-based pathway, when CASSCF wave functions or unrestricted DFT methods were used, respectively. However, a ‘two-step no-intermediate’ mechanism operated when restricted DFT methods were used.

There is no doubt that the precise mechanism of the ¹O2 ene reaction remains obscure. Recently, to gain deeper insight into this mechanism, we synthesized and assayed more informative substrates bearing phenyl-substituted cyclopropyl groups as the mechanistic probes. [55] Additionally, we have investigated the stereochemistry of the ¹O2 ene reaction with simple non-functionalized alkenes. [56] These novel results, which are presented later (Section 6), have important implications regarding the precise mechanism of this reaction. Surprisingly, unlike the previous theoretical work, [54] a more recent computational study was in good agreement with our earlier and recent experimental findings. [57]

The scope of the current account is to summarize the key mechanistic features of this classical ene reaction, focusing on the advances made in our laboratory.

2 Theoretical Calculations

Theoretical and computational methods are a powerful tool in exploring reaction paths as well as in understanding the reaction mechanisms. Using these methods, transition state (TS) structures and intermediates can be located on the potential-energy surface (PES). Although the theoretical calculations have revealed several reaction mechanisms, in the case of ¹O2-mediated reactions they have encountered considerable complicating and challenging problems (due to the unique electronic structure of ¹O2). [58]

Early and recent theoretical calculations for the mechanism of the ¹O2 ene reaction are notably contradictory. In the mid-1970s, according to semiempirical MINDO/3 calculations, the lowest energy approach of ¹O2 to propene led to the formation of cis- and trans-methylperepoxide intermediates (which were calculated to be similar in energy); the formation of these intermediates was 16 kcal/mol exothermic with a barrier to addition of 11 kcal/mol. [59] Subsequently, both cis- and trans-methylperepoxides undergo rearrangement to the ene product, namely propene-3-hydroperoxide; this rearrangement was predicted to proceed with a barrier of 21 kcal/mol.

In the early 1980s, a combination of ab initio (GVB-CI) calculations and thermochemical methods of estimating substituent effects was in disagreement with the previously reported MINDO/3 results. [60] In particular, Harding and Goddard predicted that a PE would be 10-16 kcal/mol above the energy of simple starting alkenes-¹O2 and that a biradical intermediate would be preferred by 6-8 kcal/mol. Likewise, ab initio (STO-3G) calculations and semiempirical MINDO/3 calculations at the unrestricted Hartree-Fock level predicted that biradicals, formed from simple alkenes and ¹O2, were more stable than perepoxides. [] The findings in this study also indicated that the biradical intermediates appeared inconsistent with experimentally observed regioselectivity and stereospecificity. It is worth mentioning that CASSCF calculations, concerning the ¹O2 [2+2] cycloaddition, suggested that the initial formation of a biradical was favored and a PE was attainable only by passing through this biradical intermediate. []

It is interesting to point out here that the preference for hydrogen abstraction from cis-substituted groups in trisubstituted alkenes (the so-called ‘cis effect’) could be better rationalized by a concerted rather than a stepwise biradical mechanism. [] In agreement with this study, ab initio molecular orbital calculations supported a concerted mechanism for the reaction of 2-aminopropene [64] or propene [65] with ¹O2. In the early 1990s, two alternative reaction pathways, a concerted (which proceeded via a 6-membered ring TS) and a stepwise (involving strained PE- like intermediates) were located on the PM3 semiempirical PES for the reaction of propene with ¹O2. [66] In 1996, ab initio molecular orbital studies supported a concerted mechanism for the reaction of olefins (bearing allylic hydrogen atoms) with ¹O2. [67] In this case, a non-radical TS with a PE-like conformation was the most favorable. In the early 2000s, the reaction of ¹O2 with 1,3-cyclohexadiene was investigated at the DFT(B3LYP) and CASPT2 levels. In particular, Sevin and McKee reported a two-step mechanism for the formation of 1-hydroperoxy-2,4-cyclohexadiene and a less favored concerted TS for the formation of 1-hydroperoxy-2,5-cyclohexadiene. [68]

Among the more recent contributions toward understanding the ¹O2 ene reaction mechanism is the collaborative effort by the Singleton, Houk and Foote groups. [54] Based on high level ab initio calculations and experimental kinetic isotope effects, [54b] [69] they generally reported that the ¹O2 ene reaction with simple alkenes proceeded via two transition states without an intervening intermediate. This mechanism was styled as ‘two-step no-intermediate’. In particular, for the reaction of ¹O2 with cis-2-butene or tetramethylethylene, CCSD(T) single-point energies were computed on a grid of B3LYP structures. These calculations revealed that the PES bifurcated after the first TS. [70] This TS (with the symmetry of the PE) did not involve hydrogen abstraction by the trailing oxygen. The second TS (with the elusive PE-like structure) lay near a valley-ridge inflection (VRI) point. [] At the VRI point, the path fell off to one site or the other while abstracting a hydrogen from either terminal methyl group. It should also be mentioned that dynamic effects dictated the product ratio. In Figure  [¹] , there is depicted a model PES with sequential transition states (TS1 and TS2) and a valley-ridge inflection (VRI) point. The most representative reaction pathway is an intrinsic reaction coordinate (IRC) which is defined as the steepest descent pathway on a PES. This pathway is shown by a dotted black line, while the expected reaction trajectories are shown by transparent arrows (Figure  [¹] ).

In 2003, Tonachini and co-workers investigated the gas-phase mechanism of the ¹O2 ene reaction with propene. [] The main result of this theoretical study was that a two-step ene pathway passing through a polar biradical intermediate was sharply favored. On the other hand, in this system a PE pathway was ruled out; the PE was located higher in energy than the biradical by 12 kcal/mol at the DFT(MPW1K), DFT(B3LYP) and CASSCF levels of theory. In the mid-2000s, a stepwise pathway passing through a polar biradical intermediate was also suggested for the ¹O2 ene reaction with E-2-methyl-but-2-enal; this substrate was chosen as a simple example model system of an α,β-unsaturated carbonyl compound. [] Moreover, this mechanism was able to interpret the regioselectivity of the reactant E-2-methyl-but-2-enal. In 2008, studies using B3LYP/6-31G* and CASMP2 calculations were employed to study the ¹O2 ene reaction with tetramethylethylene and trans-cyclooctene. [74] In the case of alkenes such as tetramethylethylene, the ¹O2 ene reaction proceeded through a two-step no-intermediate’ mechanism. On the other hand, the title reaction with trans-cyclooctene was predicted to occur by a stepwise mechanism involving a PE intermediate; this intermediate was formed through a polarized biradical intermediate. The change in mechanism occurred because the trans-cyclooctene imposed a large strain in the TS for hydrogen abstraction.

Figure 1

Almost simultaneously with the preparation of this account, a detailed computational study was reported by Acevedo and Sheppard. [57] This theoretical work provided valuable and updated information for the mechanism of the ¹O2 ene reaction. Specifically, the addition of ¹O2 to tetramethylethylene was investigated, using novel three-dimensional potentials [of mean force (3-D PMF) calculations coupled to multidimensional mixed quantum and molecular mechanics (QM/MM) simulations], in three different explicit solvents: water, DMSO, and cyclo­hexane. These simulations predicted an alternative ­free-energy surface which supported a traditional stepwise mechanistic interpretation featuring an intermediate with the symmetry of a PE; this is in accordance with the mechanism derived from previous experimental results. Furthermore, this computational study provided insight into the effect of solvent on the ¹O2 ene reaction. It is noteworthy that the charge separation present in the PE intermediate is sensitive to solvent polarity as well as to hydrogen bonding. Consistent to this fact, a direct correlation of increasing PE stability with increasing solvent polarity was found. Last but not least, increasing the polarity of the solvents also increased the relative energy barrier for product formation, solidifying the PE’s role as an intermediate and not a TS in solution.

3 Kinetic Isotope Effects

Kinetic isotope effects (KIEs) [75] measurements are one of the most powerful tools used in the mechanistic studies on the ¹O2 ene reactions. More generally, the isotopic substitution is a useful technique which provides clues to the pathway of the reaction. An isotopic substitution will greatly modify the reaction rate when the isotopic replacement is in a chemical bond that is broken or formed in the rate-limiting step. In this case, the change is termed a primary isotope effect. In order for such a measurement to be meaningful, the competing isotopes (usually H and D) must be stereochemically as well as electronically equivalent. Taking into consideration these requirements, intra­molecular and intermolecular isotope effects have frequently been measured. Importantly, significant intermolecular primary deuterium isotope effects provide strong evidence for hydrogen abstraction in the rate-determining step of the reaction. However, high intramolecular (product) and simultaneous low intermolecular (kinetic) isotope effects are evidence for an intermediate, with an isotopic competition on the second (product-determining) but not on the first (rate-determining) step.

Over the last 40 years, several KIE measurements for the ¹O2 ene reaction have been reported. In the early 1970s, Kopecky and Van de Sande found low intermolecular KIEs in comparing cis-13-d 0 versus cis-13-d 6 and trans-13-d 0 versus trans-13-d 6 (Figure  [²] ). [76] In 1980, Stephenson and co-workers also reported a negligible intermolecular KIE for the ¹O2 ene reaction of 14-d 0 versus gem-14-d 6 (Figure  [²] ). [³¹b] Moreover, in an independent work, Gollnick and co-workers measured a low intermolecular KIE between compared 14-d 0 and 14-d 12 (Figure  [²] ). [77]

Figure 2 Intermolecular KIEs on the ¹O2 ene reaction of tetrasubstituted alkenes (vs = versus)

The first detailed mechanistic study by means of intramolecular primary KIEs on the ¹O2 ene reaction was reported in the early 1980s. [48] In this study, three types of isotope competition, namely cis, trans, and geminal, were examined. In particular, these types of isotope competition were arranged to occur in the ¹O2 ene reaction of cis-14-d 6, trans-14-d 6, and gem-14-d 6 (Figure  [³] ). Substantial intramolecular isotope effects (k H/k D = 1.38-1.45) were found with cis-related methyl and deuteriomethyl groups in tetramethylethylenes (trans-14-d 6 and gem-14-d 6). On the other hand, small or negligible intramolecular isotope effects (k H/k D = 1.04-1.09) were obtained with trans-related deuteriomethyl groups in cis-14-d 6. Based on these results, a concerted pathway where all methyl groups (because of symmetry) should be equally competitive was excluded and a stepwise mechanism with rate-determining the formation of the PE intermediate was established. In addition, these findings eliminate the formation of an open biradical or dipolar intermediate (OI). Substantial and essentially identical isotope effects were expected from the open biradical/dipolar intermediates OI1 and OI2 (Figure  [4] ), inconsistent with the experimental results. In a similar manner, if the intermediates OI3 and OI4 (Figure  [4] ) were equally formed, then no isotope effect would be measured; this was again in contradiction with the experiment.

Figure 3 Intramolecular KIEs on the ¹O2 ene reaction of TMEs

Figure 4 Open biradical/dipolar (OI) and perepoxide (PE) intermediates in the ¹O2 ene reaction of TMEs-d 6

Notably, a small intermolecular and a substantial intramolecular isotope effect in tetramethylethylenes (TMEs) provides strong evidence for a stepwise mechanism with the formation of PE as the most favorable intermediate in the rate-determining step of this classical ene reaction. In particular, addition of ¹O2 to cis-14-d 6 formed equally two distinct perepoxides PE1 and PE2 (Figure  [4] ). Assuming that these perepoxides do not interconvert to each other or to starting materials, no isotopic discrimination is expected in the following product-determining step. Thus, in this case, a negligible intramolecular isotope effect should be observed, which was in agreement with the experiment. On the other hand, addition of ¹O2 to trans-14-d 6 or gem-14-d 6 gave perepoxide PE3 or PE4 (Figure  [4] ), respectively. These intermediates were able to abstract either hydrogen or deuterium in the product-determining step, and therefore a substantial isotope effect was observed. Importantly, the specific isotope competition required that perepoxides PE1-PE4 were formed irreversibly.

Similar results were observed in the case of trisubstituted [49] and cis-disubstituted [50] alkenes. For instance, a very small KIE (average k H/k D = 1.07) was found in the intermolecular competition between cis-15-d 0 and cis-15-d 4 (Figure  [5] ). [50] However, a significant intramolecular (product) isotope effect (k H/k D = 1.50) in substrate cis-15-d 2 was determined (Figure  [5] ). More generally, the observed KIEs on the ¹O2 ene reaction of trisubstituted and cis-disubstituted alkenes were explained by the irreversible formation of PE intermediates.

Figure 5 Intermolecular and intramolecular KIEs on the ¹O2 ene reaction of disubstituted and trisubstituted alkenes

In 1990, the ¹O2 ene reaction of cis- and trans-2-butenes-d 3 was studied. [] Particularly, cis-butene-d 3 (cis-16-d 3, Figure  [5] ) gave an intramolecular KIE of k H/k D = 1.38, close to that observed with trans-14-d 6 and gem-14-d 6. On the other hand, the highly unreactive trans-butene-d 3 (trans-16-d 3, Figure  [5] ) showed an intramolecular KIE of k H/k D = 1.25, which was larger than that observed with cis-14-d 6. The latter isotope effect was rationalized in terms of partial reversion of the PE intermediate to the starting materials (trans-16-d 3 and ¹O2). In previous work done in our laboratory, we examined the ¹O2 ene reaction of 2,5-dimethyl-2,4-hexadiene-d 6 (17-d 6, Figure  [5] ). [40] In this study, when the reaction solvent was CHCl3, a substantial primary isotope effect of k H/k D = 1.58 was measured. In a similar manner with the case of trans-16-d 3, this isotope effect could be the result of partial reversion of the PE intermediate to the starting materials. Ultimately, this proposed inversion could be attributed to the fact that diene 17-d 6 is more nucleophilic than the simple trisubstituted alkenes and the energy barrier for the formation of the PE intermediate is lower.

4 Regioselectivity

The study on site selectivity has offered valuable information on the mechanism of several organic reactions. In the case of the ¹O2 ene reaction, it has proven to be a powerful mechanistic tool in order to determine the trajectory of the enophile attack. Two reviews have extensively covered the regioselectivity in the ¹O2-mediated allylic oxidation. [] However, we briefly review the most important regioselective features, which provide the foundation to help understand the reaction mechanism. We also mention several empirical rules, such as the ‘cis effect’, the anticis effect’, the large-group nonbonding effect and the ‘gem effect’. These effects were proposed to allow a priori determination of the regioselectivity in ¹O2 ene reaction with a variety of alkenes. Ultimately, we present some interesting solvent and electronic effects that were carried out in our laboratory.

4.1 ‘ cis Effect’ Selectivity

The reaction of ¹O2 with enol ethers [78] and acyclic [79] or cyclic [80] trisubstituted alkenes showed strong preference for hydrogen abstraction from the more substituted site of the double bond (Figure  [6] ). This startling selectivity went unrecognized more than 20 years of ¹O2-alkene ene reaction investigations. For example, ¹O2-mediated oxidation of trimethylethylene (28) gave equal amounts of photooxidized products 29 and 30 (Scheme  [³] ). This result led to the conclusion that the ene reaction proceeds without any selectivity. Considering that product 29 was obtained from H-abstraction from either methyl group (a) or (b) of 28, the relative reactivity of these groups was unknown. The lack of Markovnikov directing effects in this system has been interpreted as evidence against a biradical or ionic intermediate. However, the stereospecific deuterium labeling and subsequent photooxidation of the alkene 18, and similarly alkenes 19 and 20 (Figure  [6] ), revealed the hidden regioselectivity of the ¹O2 ene reaction. This unanticipated selectivity, where the more substituted site of the double bond is the more reactive one, is referred as to as ‘cis effect’ (Figure  [6] ).

Figure 6 ‘Cis effect’ selectivity in the ¹O2 ene reaction of enol ethers and trisubstituted alkenes (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Scheme 3

Several models have been proposed in order to explain this effect. In the late 1970s, Frimer et al. studied the primary and secondary isotope effects in the photooxidation of 2,3-dihydro-γ-pyran and suggested the importance of initial orbital interaction between the incoming ¹O2 and the enol ether. [] In 1980, Stephenson reported that an interaction between the LUMO orbital of the oxygen and the HOMO orbital of the reactant alkene stabilized the TS for the PE formation. [] In the same year, Schulte-Elte and Rautenstrauch proposed that in the case of a cycloalkene the allylic hydrogens at the axial position were more reactive; the orbital overlap between oxygen and allylic hydrogens was optimum in such a conformation. [80] In 1981, Houk and co-workers, based on STO-3G semiempirical calculations, postulated that the lower the rotational barrier of a specific alkyl group at the double bond, the higher the reactivity of this group toward hydrogen abstraction by ¹O2; [] this mechanism was later challenged by work accomplished in our laboratory (see Section 4.4). In addition, Schuster and co-workers demonstrated that trans alkenes showed distinctly lower normalized (negative) entropies of activation Δ S for the ¹O2 ene reaction than cis. [] Notably, they suggested that there was a type of hydrogen bonding interaction in the rate-limiting step of PE formation between the allylic hydrogens and the intermediate.

In conclusion, most of the proposed models are consistent with the existence of an interaction between the incoming ¹O2 and two allylic hydrogens that highly stabilizes the transition state TS1, versus TS2, of PE formation (Figure  [7] ).

Figure 7

4.2 Anti cis Effect’ Selectivity

Unlike the ‘cis effect’ selectivity, photooxidation of certain acyclic [84] and cyclic [85] trisubstituted alkenes 31-35 impressively illustrated a strong preference for hydrogen abstraction on the less substituted site of the double bond (Figure  [8] ). This selectivity was defined as the anticis effect’.

Figure 8Anticis effect’ selectivity in the ¹O2 ene reaction of certain trisubstituted alkenes (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Examination of the possible transition states leading to the major (anti) and minor (syn) PE intermediates provided reasonable mechanistic rational into the anticis effect’ selectivity. In a representative example (trisubstituted alkene E-33; Figure  [8] ), the nonbonding interactions of TS1, involving the large tert-butyl group and the incoming ¹O2, were expected to be stronger than those in TS2, where this steric interaction is less pronounced (Figure  [9] ).

Figure 9

Notably, the anticis effect’ selectivity in the ¹O2 ene reaction of trisubstituted alkenes is related: (a) to the degree of crowdedness on the more substituted site of the alkene; (b) to the nonbonding interactions in the newly formed double bond; and (c) to the lack of interaction of the incoming ¹O2 with two allylic hydrogens.

4.3 The Large Group Nonbonding Effect

Earlier work in our laboratory revealed an unexpected regioselectivity in the ¹O2 ene reaction with non-symmetrical cis- [86] and trans-alkylsubstituted alkenes. [5³a] [87] In particular, it was found that the allylic hydrogens next to the large alkyl substituent were more reactive than these next to the small alkyl substituent. This is clearly demonstrated in Figure  [¹0] . For example, when L (larger group) is isopropyl or tert-butyl and s (smaller group) is hydrogen, compounds 36 and 37 respectively, the preferential abstraction of allylic hydrogen adjacent to the L group is greater than 70%. In compound 39 (L = triphenyl group), the regioselectivity is almost exclusive (95%) on the L-substituted site. Notably, as the size of s becomes larger, the regioselectivity toward L decreases. This is demonstrated with compounds 40 and 41, where the preferential hydrogen abstraction is only slightly different on the two sites of the double bond. Notice also in compound 41, where L and s are phenyl and isopropyl groups respectively, competition for the two allylic sites leads to the nearly equal hydrogen abstraction from the two methylene sites. This result indicates further that nonbonding interactions play a more important role than conjugation with the π system of the phenyl ring in the TS of this reaction. Last but not least, a similar trend in regioselectivity for a series of geminal dimethyl and diethyl trisubstituted alkenes was observed (compounds 42-45, Figure  [¹0] ).

Figure 10 The large group nonbonding effect selectivity in the ¹O2 ene reaction of disubstituted and trisubstituted alkenes (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

This regioselectivity was mainly rationalized by examining the possible transition states leading to the allylic hydroperoxides. [5³a] [86] In the case of a cis-disubstituted alkene, a plausible mechanism of the ¹O2 ene reaction is depicted in Scheme  [4] . Specifically, in the TS2, which leads to the major product, the repulsive 1,3-nonbonding interactions between the oxygen atom and the large group (L) are smaller than those of the TS1. Notably, the TS2 is expected to have lower energy than the TS1. Taking into account that the formation of a PE intermediate in the photooxidation of trans-disubstituted alkenes is reversible, [] [40] the 1,3-nonbonding interactions in the product forming transition states between the oxygen and the alkyl substituents appear to control the site selectivity in a similar fashion with cis-disubstituted alkenes. The transition states in the hydrogen abstraction step also help to explain the observed change in site selectivity for trisubstituted alkenes 42-45. In a similar manner, the TS4 is expected to be lower in energy than the TS3 where the 1,3-nonbonding interactions are present (inset, Scheme  [4] ).

Scheme 4

In the early 2000s, a regioselectivity trend, opposite to what has been described in Figure  [¹0] , was reported for the ¹O2 ene reaction with prenylated dihydroxyacetophenones. [88] In particular for alkene 46, the allylic hydrogens of methyl groups were more reactive than the allylic hydrogens next to the phenyl ring due to a stabilizing interaction of the negatively charged oxygen of the PE1 intermediate with the phenolic hydrogen, ortho to the prenyl site chain (Scheme  [5] ). Generally, this site selectivity outlined another effect, named ‘phenolic assistance effect’, in competition with the large group nonbonding effect.

Scheme 5 Green and red numerical values indicate percentage of double-bond formation in the ene adducts

4.4 Geminal Selectivity with Respect to Allylic or Vinyl Substituent

Clennan and Chen assayed several substrates by replacement of an allylic hydrogen in TMEs with a series of functional groups. [89] These alkenes underwent ¹O2 ene reaction with surprising geminal selectivity (with respect to the allylic functionality). Some representative alkenes and their regio-limitations are presented in Scheme  [6] .

Three possible explanations were provided to rationalize the observed regioselectivity: (a) electronic repulsions between the lone pairs of the heteroatoms and the negatively charged oxygen of the PE; (b) rotational barrier differences within the methyl groups of the substrate (which will be mentioned later in this section); and (c) anchimeric assistance from the allylic substituent resulting to regioselective opening of the possible PE intermediate by an SN2 mechanism (Scheme  [6] ).

Scheme 6 Green and red numerical values indicate percentage of double-bond formation in the ene adducts

In order to examine comprehensively the factors affecting geminal selectivity, we examined the ¹O2 ene reaction with a series of alkenes free of heteroatoms in the allylic position [90] as well as 1-neopentyl substituted cycloalkenes. [] The photooxidation of these alkenes showed a strong preference for hydrogen abstraction from the group that is geminal to the large alkyl group. Some representative examples are summarized in Figure  [¹¹] .

Figure 11Geminal selectivity on the ¹O2 ene reaction of alkenes bearing a heteroatom or a large alkyl substituent at the allylic position (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Examination of the possible transition states leading to the major and the minor products in the reaction of ¹O2 with L-alkyl substituted alkenes (such as 49-51, Figure  [¹¹] ), provided insight into the geminal selectivity. In this case, a possible reaction mechanism is depicted in Scheme  [7] . Each of the two perepoxides PE1 and PE2 could lead though transition states TS1-TS4 to products. TS1 was expected to have lower energy than TS2 and TS4, due to the relief of 1,3-repulsions. Therefore, this transition state led to the major product. On the other hand, TS2 and TS4 were unfavorable because of the repulsive 1,3-nonbonding interactions. Both of these transition states led to the minor products. Last but not least, in TS3 although there was a relief of 1,3-nonbonding interactions between the oxygen atom and the L group, two geminal substituents (L and Me) would adopt a cis conformation in the newly forming double bond, which was highly unfavorable. Indeed, no product derived from this pathway was detected. It is important to point out that for 1-neopentyl substituted cycloalkenes (such as 52-54, Figure  [¹¹] ), the variation in the percentage of geminal regioselectivity was probably due to the combination of two effects: (a) the 1,3-nonbonding interactions between the tert-butyl group and the oxygen atom and (b) the different conformational arrangements of the allylic hydrogens in the ring systems. [80]

Scheme 7

In our laboratory, we further examined if the geminal selectivity was achieved in cases where the bulky substituent was at the vinyl position. [90] [] Indeed, a similar trend in regioselectivity was observed. Some key examples are given in Figure  [¹²] . It is worth mentioning that for styrene-type substrates (57, Figure  [¹²] ) the presence of a para-substituent did not alter the geminal selectivity. This observation indicated that nonbonding interactions were more important than electronic effects of the para-substituent phenyl ring in determining the stability of the TS of the product-determining step. The rationalization of the geminal selectivity observed for substrates 55-57 and 58-63 was exactly the same as previously mentioned in the case of substrates 49-51 and 52-54, respectively.

Figure 12Geminal selectivity on the ¹O2 ene reaction of alkenes bearing a large alkyl substituent at the vinyl position.a a Green and red numerical values indicate percentage of double-bond formation in the ene adducts

It is also worth mentioning that the ¹O2 ene reaction with vinyl silanes [] and stannanes [] exhibited a high degree of geminal selectivity with respect to the silyl of stannyl groups, respectively. Although steric factors might be responsible for the geminal selectivity, the selectivity was mainly attributed to electronic factors. For instance, in the case of the silanes, it was proposed that antibonding interactions of the C-Si s-bond and the lone pair of the non-terminal oxygen of the PE intermediate were responsible for the high degree of site selectivity.

In an earlier report, an alternative explanation of the geminal selectivity was based on a computational model proposed by the Houk research group. [] According to this model, the lower the calculated rotational barrier, the higher the reactivity of the alkyl group. In 1990, similarly, Clennan and co-workers performed molecular mechanics calculations (MM2) and found that the methyl group geminal to the neopentyl group in alkene 50 (red numerical values in Figure  [¹³] ) showed the lowest rotational barrier. [89b] Therefore, this methyl group was expected to be the most reactive which was in agreement with the observed experimental regioselection [90] (blue numerical values in Figure  [¹³] ). In addition, alkene 64 had a much higher rotational barrier (5.76 kcal/mol) than the methyl groups and was totally inactive to ¹O2. [89b]

Figure 13 Red and blue numerical values indicate the calculated rotational barriers and the percentage of double-bond formation in the ene adducts, respectively

However, work from our laboratory showed that the barrier to rotation did not always predict the regioselectivity in the ¹O2 ene reaction of alkenes. [94] In particular, we calculated rotational barrier values for the allylic methyl groups, with the ab initio (STO-3G) method, in a variety of di- and tri-substituted alkenes. Then, we compared these values with the observed experimental regioselectivity. Some representative results are shown in Table  [¹] . For alkenes E-33 and Z-33 the syn methyl groups had lower rotational barriers than the anti ones by ca. 0.5 kcal/mol. Although the proposed theoretical model [] required lower reactivity of the anti methyl than the syn methyl in these alkenes, the experimental results showed the opposite. A similar discrepancy between barriers to rotation and ene reactivity held for substrates 20 and 65.

Table 1 Relative Yields of Ene Products and Rotational Barriers of Methyl Groups
Substrate Ene Product (%)a Rotational Barriersb

E-33
š 76

š 24
1.64

1.11

Z-33
š 74

š 26
1.63

1.11

20
š 14

š 86
1.64

0.40

65
š 36

š 64
1.22

1.45

a Numerical values indicate percentage of double-bond formation in the ene adducts.
b In kcal/mol (STO-3G).

4.5 Electron-Withdrawing Group at the α - and β-Position

For alkenes bearing an electron-withdrawing group at the α-position, such as ketone, [95] aldehyde, [96] carboxylic acid, [97] ester, [98] cyano, [96a] amide, [96a] aldimine, [99] sulfoxide [¹00] and oxazoline, [¹0¹] a high degree of geminal selectivity with respect to the functional group was demonstrated. Representative examples are shown in Figure  [¹4] .

Figure 14Geminal selectivity on the ¹O2 ene reaction of alkenes bearing an electron-withdrawing group at the α-position (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Many mechanisms (involving the intermediacy of trioxanes, PE, exciplex, open dipolar or open biradical) were proposed in order to rationalize this selectivity. However, it was widely supported [96] [¹0¹] [¹0²] that the mechanism proceeded through an intermediate exciplex which had the structural requirements of a PE. Moreover, in this case the 1,3-nonbonding interactions did not contribute substantially to the regioselection. It was also reported that, in the hydrogen abstraction step, the PE intermediate opens preferentially at the C-O bond next to the unsaturated moiety, due to the forthcoming conjugation in the adduct. Therefore, TS1 is energetically the most favorable (Scheme  [8] ).

As part of our continuous interest in the field, we examined the site selectivity in the ¹O2 ene reaction of alkenes bearing an electron-withdrawing group at the β-position. [¹0³] Some results are summarized in Figure  [¹5] . For the carbonyl derivatives (75-78), the methylene hydrogen atoms were statistically almost five times more reactive than those on the methyl groups. On the other hand, for alkenes 79 and 80, the reactivity next to the allylic substituent decreased significantly. This behavior was attributed to the fact that the highly negatively polarized oxygen of the S-O and P-O bonds exhibited unfavorable repulsions with the negatively charged oxygen of the TS2 (Scheme  [8] ).

Scheme 8

The results observed in the case of the disubstituted unsaturated ester 81 and acid 82 were similar to that of the trisubstituted substrates 75 and 76, respectively. This revealed that any unfavorable interactions between oxygen and carbonyl in the more substituted site of the double bond seem to prevent the formation of the intermediate in the less substituted site of the double bond (anticis effect’ selectivity, Section 4.2).

Figure 15 Site selectivity on the ¹O2 ene reaction of alkenes bearing an electron-withdrawing group at the β-position (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

We suggested that there are three competing factors, which could affect the site selectivity in the ¹O2 ene reaction with alkenes bearing an electron-withdrawing group at the β-position: (1) the driving force to form the new double bond in conjugation with the functionality in the allylic hydroperoxide product; (2) the 1,3-nonbonding interactions between the positively charged oxygen of PE and the allylic functionality, which favored again the conjugated product; and (3) the electronic repulsions between PE and the allylic functionality favoring the unconjugated product.

4.6 Regioselective Self-Sensitized Oxygenation of Fullerene Derivatives

One of the most important and potentially useful properties of fullerenes is their efficient photoexcitation to a triplet excited state and subsequent energy transfer to form ¹O2 in high quantum yields. [¹04] In the case that a fullerene adduct bears an oxidizable group, in the presence of oxygen and light, self-sensitized oxidation was reported. [¹05] Only two reports so far in the literature deal with the ene regioselectivity in the autooxidation of fullerene derivatives.

The first case study concerned the ene reaction of several fullerene derivatives, prepared by the thermal [4+2] cycloaddition of conjugated dienes to C60. [¹06] The C60-cyclohexenes 83 and 84 showed strong preference for the formation of allylic alcohols with endocyclic double bonds (Figure  [¹6] ). Interestingly, cycloalkene 85 gave the allylic alcohol with an exocyclic double bond as a major product. Rubin and co-workers suggested that, due to favorable interactions between the incoming ¹O2 and the methylene allylic hydrogens, the endo-PE resulting from TS1 was preferably formed, rather than the exo isomer derived from TS2 (Figure  [¹7] ). [¹06] However, additional factors such as favorable electrostatic or electronic interactions between the negative oxygen of the developing endo-PE intermediate and the electron-deficient C60 may also contribute to the observed endo regioselectivity.

Figure 16 Regioselectivity in the self-sensitized oxygenation of fullerene derivatives (numerical values indicate percentage of double-bond formation in the ene adducts)

Figure 17

The second case study, which was conducted in our laboratory, concerned the ene reaction of a series of alkenyl-linked[C60] derivatives 86 (Figure  [¹6] ). [¹07] This reaction showed a consistent regioselectivity for double-bond formation next to the fullerene substituent. These observations indicated that nonbonding steric interactions played a more important role than the conjugation with the π system of the phenyl ring in the TS of this reaction. Furthermore, as the size of the R group became larger, the regioselectivity towards the fullerene group decreased (Figure  [¹6] ). Examination of the possible transition states TS3 and TS4 (leading to the major and minor products, respectively) provided an insight into this regioselectivity (Figure  [¹7] ). In TS3, the nonbonding interactions involving the large fullerene moiety were smaller than those at TS4. As the size of the substituent on both sites of the double bond became similar (e.g., R = tert-butyl) the nonbonding interactions in TS3 and TS4 became isoenergetic, leading to equal amounts of the two isomeric ene products.

4.7 Solvent and Electronic Effects

4.7.1 Site Selectivity in the ¹ O 2 Ene Reaction of α,β-Unsaturated Esters and Acids

Previously reported studies showed that the site selectivity as well as the rate of the ¹O2 ene reaction with alkenes was practically independent of solvent polarity. [¹08] On the other hand, a small variation in the distribution of the ene products of some non-functionalized olefins, by changing the solvent polarity was reported in the past. [87b] [¹09] However, no mechanistic explanation was offered to account for the observed solvent effects. It is rather difficult to rationalize these results based on any of the currently proposed mechanisms of the ¹O2 ene reaction. Nevertheless, it is generally accepted that product distribution depends substantially on solvent polarity and reaction temperature, only in substrates where both ene and dioxetane products may be produced. [³6] [¹¹0]

In the early 1990s, we investigated the reaction of ¹O2 with α,β-unsaturated esters. [¹0²b] It was found that the site selectivity in the ¹O2 ene reaction of the above mentioned substrates depended on solvent polarity. As seen from Table  [²] , the hydrogen abstraction from the methyl group, which is geminal to the ester functionality in compound 87, producing adduct 87a, decreased substantially as the solvent polarity increased. For instance, the ratio of ene products 87a/87b decreased by a factor of 5 on going from CCl4 to the more polar solvent DMSO. More generally, it was noted that there was a surprising correlation between the dielectric constant (ε, Table  [²] ) of the solvent and the distribution of the ene products; by increasing the dielectric constant of the solvent, the percentage of 87b increases substantially.

Table 2 Solvent Effect on the Site Selectivity in the ¹O2 Ene Reaction of 87

Solvent e (20 ˚C)a 87a/87b
CCl4  2.24 95:5
C6H6  2.28 94:6
Me2CO 20.70 88:12
MeCN 36.64 85:15
DMSO 47.24 80:20

a Taken from Lange’s Handbook of Chemistry, 15th edition.

This correlation was rationalized by examining the possible transition states (Scheme  [9] ). In TS2 the oxygen was oriented syn with respect to the ester group, and the net dipole moment is expected to be larger than in TS1, where the oxygen was placed anti to the ester group. Therefore, TS2 is more polar than TS1 and was expected to be preferentially stabilized by polar solvents. Accordingly, the ratio 87a/87b decreased with the increase of solvent polarity.

Scheme 9

To verify the above mechanistic assumption, the solvent dependence of the ene products derived from the photooxidation of the isomeric α,β-unsaturated esters E/Z-88 was examined (Table  [³] ). In the case of E-88, allylic hydroperoxides 88a and 88b were formed from two different PE intermediates. Specifically, when the oxygen atom of the PE was placed anti to the ester group, 88a was obtained, whereas 88b was formed from the syn position. For E-88, the expected solvent effect was found; in going from CCl4 to DMSO, the product ratio 88a/88b changed by a factor of 2.5 (Table  [³] ). On the other hand, for Z-88, both 88a and 88b adducts were formed from the same PE intermediate, and the ene product distribution was insensitive to solvent polarity (Table  [³] ). Notice that methylene hydrogens of both substrates are inactive to the ene mode.

Table 3 Solvent Effect on the Site Selectivity in the ¹O2 Ene Reaction of Z-88 or E-88

88a/88b
Solvent E-88 Z-88
CCl4 85:15 95:5
C6H6 83:17 95:5
MeCN 75:25 98:2
DMSO 70:30 93:7

In the early 2000s, Stensaas and co-workers investigated the photooxidations of tiglic acid, angelic acid, 2,3-dimethyl-2-butenoic acid and their corresponding methyl esters, using ¹O2 in CD3OD and CD3OD-D2O solvent mixtures and compared them with aprotic solvents with different dielectric constants. [¹¹¹] In particular, for tiglic acid and its methyl ester, the ene product distribution was independent of solvent polarity and hydrogen-bonding effect (between protic solvent and substrate). This result was in accordance with our findings in the ¹O2 ene reaction of α,β-unsaturated ester Z-88 and it was rationalized in terms of the ‘cis effect’. On the other hand, the photooxidation of 89 and 90 appeared to be sensitive to solvent polarity (Table  [4] ). Specifically, for compounds 89 or 90, a 23% or 9% increase of allylic hydroperoxides 89b or 90b, respectively, was achieved covering a range of dielectric constants from 2.3 to 42.4.

Table 4 Solvent Effect on the Site Selectivity in the ¹O2 Ene Reaction of 89 or 90 a

Solvent ε 89a/89b 90a/90b 91a/91b 92a/92b
C6D6  2.3 86:14 89:11 100:0 100:0
CD3OD 33.0 79:21 85:15  87:13  91:9
CD3CN 33.6 78:22 83:17  80:20  86:14
CD3OD-D2O (9:1) 37.7 70:30 82:18  81:19  90:10
CD3OD-D2O (4:1) 42.4 63:37 80:20 - -
CD3OD-D2O (3:2) 51.8 - -  79:21  87:13

a Taken from reference [111].

This increase was mainly attributed to the stabilization of the TS leading to the more polar PE intermediate, which was again in agreement with our results obtained in the photooxidations of α,β-unsaturated esters 87 and Z-88.

Interestingly, the findings concerning the photooxidations of 91 and 92 showed that increasing the solvent polarity did not necessarily increase the amount of minor products 91b and 92b, respectively (Table  [4] ). Accordingly, it was reported that hydrogen bonding between the solvent and the carboxyl group stabilized TS1 (Scheme  [¹0] ), which produced the major product 91a or 92a. Placing the negatively charged pendant oxygen away from the site of the PE with the hydrogen-bonded carboxyl group minimized electronic interactions and steric interference with the hydrogen-bonded solvent molecules. Moreover, in the case of α,β-unsaturated acid 92, both ‘cis effect’ and hydrogen-bonding interaction contributed to the formation of 92a.

Scheme 10

In a more recent work, Stensaas and co-workers concluded that principally four factors dictated the site selectivity in ¹O2 ene reactions of α,β-unsaturated esters and acids in protic solvents: the ‘cis effect’, the polarity of the solvent and substrate, and the most important hydrogen-bonding interactions between the solvent and substrate.

The above authors have also studied the aqueous photooxidations of some α-substituted alkene salts. In this case, the major factor dictating the product distribution of ene products was the hydrogen-bonding interactions between the water and the substrate. [¹¹²] In contrast, when some β-substituted alkene salts were photooxidized under similar conditions, a profound change in the observed regiochemistry was found. This change was attributed to the interplay of a number of interactions including developing conjugation, 1,3-nonbonding interactions and electronic effects. [¹¹²]

4.7.2 Site Selectivity in the ¹ O 2 Ene Reaction of ­Allylic Alcohols

Significant solvent effects on the regioselectivity of the title reaction were also observed with allylic alcohols E-93 and Z-93. [¹¹³] It is worth mentioning that for these allylic alcohols, ¹O2 could interact with only one allylic hydrogen or deuterium on each site of the alkene double bond. Therefore, no ‘cis effect’ would have been expected in this case. The results are summarized in Table  [5] . For E-93, the syn adduct was formed by hydrogen abstraction, while for Z-93 by deuterium abstraction; we defined as syn the adducts formed by allylic hydrogen abstraction which was on the same site of the double bond as the hydroxyl. As seen from Table  [5] , photooxidations of E-93 and Z-93 in CCl4, CHCl3 and MeCN showed similar syn selectivity. These findings indicated that the syn/anti selectivity is independent of the specific labeling of the methyl groups. For E-93, the ratio of syn/anti decreased by a factor of 6 on going from CCl4 to MeOH. This solvent dependence is the highest ever reported.

Table 5syn/anti Selectivity in the ¹O2 Ene Reaction of E-93 and Z-93

Substrate Solvent syn/anti
E-93 CCl4 75:25
Z-93 CCl4 72:28
E-93 C6H6 73:27
E-93 CHCl3 66:34
Z-93 CHCl3 65:35
E-93 Me2CO 42:58
E-93 MeCN 41:59
Z-93 MeCN 40:60
E-93 MeOH 33:67

Examination of the possible transition states in Scheme  [¹¹] helped to understand the observed stereoselectivity. In non-polar solvents a favorable hydroxyl-oxygen interaction during the formation of the PE is very important in stabilizing TS1 where oxygen was syn to the hydroxyl. The interaction between oxygen and one allylic hydrogen (or deuterium) existed in both TS1 and TS2. The only difference between these transition states was the stabilizing hydroxyl-oxygen interaction. Subsequently, PE1 and PE2 intermediates collapsed in a faster step through TS3 and TS4, respectively, to the ene products. Importantly, the ratio of syn/anti selectivity reflected the relative stabilities of TS1 and TS2. Furthermore, the polar solvents interacted with the hydroxyl group through hydrogen bonding and reduced its ability to interact with the incoming oxygen. Therefore, the activation energy of TS1 increased significantly and led to a reversal of selectivity. The observed syn/anti selectivity was mainly attributed to hydroxyl-­oxygen interactions and steric factors.

Scheme 11

4.7.3 Syn Selectivity of β,β-Dimethylstyrene

In previous work in our laboratory, we examined the photooxidation of β,β-dimethylstyrene. [¹¹4] In this case, the ene products were followed by significant amounts of 1,2-dioxetane as well as two diastereomeric diendoperoxides [¹¹5] and exhibited a fascinating solvent-dependent syn selectivity. To distinguish the syn/anti selectivity of the ene products produced from the two geminal methyls, the anti-methyl group was specifically deuterium labeled. Subsequently, the photooxidation of 94 in several solvents revealed that there was a preference for hydrogen abstraction from the methyl syn to the phenyl ring (Table  [6] ). In addition, the magnitude of this selectivity depended on solvent polarity. Specifically, on increasing the dielectric constant of the solvent, a substantial rise in the amount of hydrogen abstraction from the syn methyl group occurred. For instance, the ratio of syn/anti ene products increased by a factor of 3.4 on going from CCl4 to MeOH (Table  [6] ).

Table 6Syn Selectivity in the ¹O2 Ene Reaction of 94

Solvent ε (Temp ˚C)a syn -94/anti-94
CCl4 2.24 (20) 56:44
C6H6 2.28 (20) 57:43
CHCl3 4.87 (25) 63:37
MeCN 36.64 (20) 71:29
MeOH 33.0 (20) 82:18

a Taken from Lange’s Handbook of Chemistry, 15th edition.

In this study, we also measured the intermolecular KIE for the competition of 94-d 0 with its deuterated analogue 94-d 6 in CHCl3 (Figure  [¹8] ). [¹¹4] Likewise in other trisubstituted alkenes, [49] the intermolecular KIE was absent. This result again indicated an irreversible formation of the PE intermediate.

Figure 18 Intermolecular KIEs on the ¹O2 ene reaction of β,β-­di­methyl­styrenes (vs = versus)

Scheme 12

A plausible mechanism that accounts for the observed syn selectivity is depicted in Scheme  [¹²] . In TS1, the incoming oxygen is oriented toward the more substituted site of the double bond. While there is only one allylic hydrogen interaction with ¹O2, the TS1 leads to the major ene product. In TS2, which led to the minor anti product, ¹O2 again interacts with only one allylic hydrogen. Thus the extra stabilization of TS1 (major product) versus TS2 (minor product) might result from ‘favorable interactions’ between ¹O2 and the phenyl ring. In TS1, the electron-donating ability of the phenyl group stabilized the partial positive charge, which was developed on the benzylic carbon of the double bond. Simultaneously, the negative charge, which was developed on the oxygen in TS1, was stabilized (through space) by the partially positive phenyl ring. Therefore, the overall effect stabilized better the syn transition state TS1 than the anti TS2, where this effect was absent. On increasing the solvent polarity, the stabilization of the more polar TS1 became more significant.

4.7.4 Site Selectivity of Isobutenylarenes

In light of the unexpected syn selectivity in the ¹O2 ene reaction of β,β-dimethylstyrene in a variety of solvents, [¹¹4] the electronic effect of the reaction of ¹O2 with labeled isobutenylarenes was studied. [¹¹6] [¹¹7] It is worth mentioning that the photooxidation of para-substituted β,β-dimethylstyrenes (95a-d, Table  [7] ) in CHCl3 or MeCN, apart from the ene adducts, afforded a large amount of oxygenated products arising from [4+2] or [2+2] addition. In each case, the ene adducts were isolated and then the ratio of syn/anti selectivity was determined. These results are summarized in Table  [7] . For the ¹O2 ene reaction of al­kenes 94 and 95a-d in CHCl3, the site selectivity was found to depend on the electronic nature of the aryl substituents. In particular, electron-withdrawing substituents, such as 4-CF3 and 4-F, favored in about 40% hydrogen abstraction from the syn methyl group of the double bond. On the other hand, in the case of an electron-donating substituent, such as 4-OMe, the reactivity of the anti methyl group was slightly higher than that of the syn methyl. When the photooxidations of 95a-d were run in MeCN, a similar trend for syn selectivity was found (Table  [7] ).

Table 7 Site Selectivity in the ¹O2 Ene Reaction of 94 and 95a-e

syn/anti Selectivity
Substrate Z CHCl3 MeCN
94 H 63:37 71:29
95a OMe 46:54 54:46
95b CF3 74:26 76:24
95c F 68:32 70:30
95d Me 55:45 57:43

The mechanistic rationalization presented in Figure  [¹9] was used in order to explain the aforementioned results. In TS1, the syn approach of ¹O2 to the double bond was better stabilized by the partial positive charge on the phenyl ring, due to the electron-withdrawing character of 4-CF3, than in TS2, where the positive charge was reduced due to the electron-donating ability of the 4-OMe substituent. Accordingly, TS1 led to higher syn selectivity (74% or 76% in CHCl3 or MeCN, respectively), whereas TS2 led to the lowest syn selectivity (46% or 54% in CHCl3 or MeCN, respectively).

Figure 19

During our ongoing investigation of the site selectivity in the ¹O2 ene reaction we examined the photooxidation of alkene 96 (Scheme  [¹³] ). In this case, an abnormally high reactivity for the syn methyl group (86% relative yield) was found in either non-polar and polar solvent. Unlike isobutenylarenes 94 and 95a-d, where the syn selectivity was attributed to electronic effects of the phenyl ring, the selectivity of ¹O2 with 96 might be attributed to steric reasons. Thus, the favorable conformation of the phenyl ring would be the one which places the ring almost perpendicular to the plane of the double bond. First of all, this assumption may explain the formation of only ene adducts (no [4+2] or [2+2] derived products) in the reaction of ¹O2 with 96. Furthermore, the remarkable syn selectivity could be attributed to the fact that in TS2 the substantial nonbonding interactions involving the ortho-methyl substituents and the incoming ¹O2 are much larger than those in TS1, where these interactions are diminished (Scheme  [¹³] ).

Scheme 13 Green and red numerical values indicate percentage of double-bond formation in the ene adducts

4.7.5 ‘Push-Pull’ Electronic Effect

Recently, our focus was on investigating the influence of the ‘push-pull’ electronic effect on the site selectivity in ¹O2 ene reaction. [¹¹7] For this purpose, alkene 97 was synthesized and assayed (Table  [8] ). It is noteworthy that this alkene has three distinctive characteristics: (1) two geminal para-substituted phenyl rings with an electron-donating (-OMe) and an electron-withdrawing (-CF3) group, (2) sterically equivalent sites of the alkene double bond, assuming that variation of para-substitution is far enough to sterically influence the reaction center, and (3) one of the methyl groups is specifically labeled in order to reveal the site selectivity on the ene products. The photooxidations of 97 were run in various solvents nonpolar and polar (aprotic or protic). The results are summarized in Table  [8] . In particular in aprotic solvents, there was found a slight but persistent preference in the order of 2-10% for the formation of 97b. In protic solvents, this preference increased to the range between 15 and 20%.

Assuming that the steric effect was identical to both sites of the double bond, the small but noticeable site selectivity might be due to the variation of the electron density on the two sites of the double bond. A possible mechanistic rationalization is depicted in Scheme  [¹4] . It seemed highly probable that the geminal phenyl groups, due to steric reasons, could considerably be out of the plane of the alkene double bond. Therefore, their electronic effects (through resonance) are not fully developed to the benzylic carbon of the double bond during the ¹O2 addition. Moreover, in TS2, the relative favorable interaction of the 4-CF3-phenyl ring with the incoming oxygen was slightly higher than in TS2 (due to similar reasons mentioned in Section 4.7.4). It is important to point out that the difference in the electronic interactions between ¹O2 and the two para-substituted phenyl rings in 97 might be relatively smaller than these between ¹O2 and para-substituted alkenes 95a and 95b (Figure  [¹9] ); the phenyl rings in trisubstituted alkenes 95a and 95b were more conjugated with the alkene double bond by comparing to the tetrasubstituted alkene 97. Ultimately, it was not obvious why for alkene 97 the site selectivity increased from 2-5% in aprotic solvents and to 15-20% in protic solvents. A number of additional factors might contribute to this finding, such as (1) hydrogen bonding, (2) better stabilization of the more polar TS, and (3) conformational variations of the phenyl rings in protic solvents.

Table 8 Site selectivity in the ¹O2 ene reaction of 97

Solvent ε (Temp ˚C)a 97a/97b
CCl4  2.24 (20) 46:54
CHCl3  4.87 (25) 48:52
Me2CO 20.7 (20) 49:51
MeCN 36.64 (20) 49:51
CF3CH2OH 27.68 (20) 40:60
MeOH 33.0 (20) 42:58
CD3OD-D2O (4:1) 42.44 (20) 42:58

a Taken from Lange’s Handbook of Chemistry, 15th edition.

Scheme 14

5 Studies on Diastereoselectivity

It seems likely that ¹O2 could be utilized in a highly stereoselective manner in the ¹O2 ene reaction for the preparation of a variety of valuable oxyfunctionalized target molecules. Indeed, high diastereoselectivities [¹¹8] in the title reaction were observed mainly by Adam and co-workers. [] [¹¹9] On the basis of Adam’s studies, the factors governing π-facial selectivity in the ¹O2 ene reaction were categorized as steric, stereoelectronic, electronic and conformational. [] In the case of steric control, nonbonding repulsion between a substrate and reagent made one π face of the double bond more accessible to the attacking ¹O2 than the other. The term ‘stereoelectronic control’ attributed to these cases in which one face of the substrate displayed a higher π-electron density due to special features of the overall geometry of the substrate molecule. On the other hand, the term ‘electronic control’ included electrostatic attractions or repulsions as well as hydrogen bonding interactions between a substituent of the substrate and the incoming ¹O2. Ultimately, the conformational features dealt with the proper alignment of the allylic hydrogens for abstraction. From the aforementioned factors, only the electrostatic interactions were sensitive to solvent polarity. In this section, we present findings from three earlier studies conducted in our laboratory.

5.1 Diastereoselectivity in Self-Sensitized Oxygenation of a Fullerene Derivative

As already mentioned in Section 4.6, if a fullerene adduct bears an oxidizable group, in the presence of oxygen and light, self-sensitized oxidation may take place. In the late 1990s, we investigated the self-sensitized oxygenation of fullerene derivative 98 (Scheme  [¹5] ). [¹²0] This adduct was chemically synthesized by the [2+2] photochemical cycloaddition of 2,5-dimethyl-2,4-hexadiene to C60. Subsequently, derivative 98 underwent facile self-sensitized photooxidation and produced a mixture of the threo/erythro allylic hydroperoxides 98a and 98b as the only products, in a 55:45 ratio (Scheme  [¹5] ); the stereochemistry of these adducts was not assigned. Noteworthy, this finding was in contrast with that of chiral non-functionalized alkenes, which afforded higher diastereoselectivities. []

Scheme 15

5.2 Diastereoselectivity in the ¹ O 2 Ene Reaction of Chiral Functionalized Alkenes

In the 1990s, Adam and co-workers observed moderate to high diastereoselectivities in the photooxidations of chiral functionalized alkenes 99 (Table  [9] ). In particular, the ¹O2 ene reaction of chiral allylic alcohols, [¹²¹] amines and ammonium salts [¹²²] displayed a high threo selectivity in non-polar solvents. On the other hand, electron-accepting X substituents (e.g., SO2Ph, CO2Et, CO2H, Cl and Br) [¹²³] promoted highly regioselective ¹O2 ene reactions in moderate to good erythro stereocontrol in non-polar solvents.

Table 9 Diastereoselectivity in the ¹O2 ene reaction of chiral functionalized alkenesa

X R Solvent threo/erythro
OH Me CCl4 93:7
OH Me CHCl3 92:8
NH2 Me CCl4 95:5
NH2 Me CDCl3 92:8
NH3 +Cl- Me CDCl3 94:6
SO2Ph Me CCl4 ˜5:95
CO2Et H CCl4 22:78
CO2H H CCl4 21:79
Cl Me CDCl3 15:85
Br Me CCl4 11:89

a Taken from references [121-123].

The threo-selective photooxidation of allylic alcohols, amines, and ammonium salts, whose conformational preference was fixed by 1,3-allylic strain, [¹²4] was rationalized in terms of stabilization of the threo-configurated exciplex (threo-Ex, Scheme  [¹6] ) by hydrogen bonding between the substituent X and the incipient negatively charged ¹O2. Therefore, threo-PE intermediate and threo-99 adduct were formed preferentially (Scheme  [¹6] ). However, in the photooxidation of electron-poor allylic alcohols, [¹5b] the threo selectivity was attributed to stereoelectronic effects, rather than to intramolecular hydrogen bonding. As already pointed out, when the allylic substituent did not coordinate to oxygen (e.g., SO2Ph, CO2Et, CO2H, Cl and Br), the title reaction favored the formation of the erythro-99. This selectivity was attributed to steric and electrostatic repulsion of the X substituent and the negatively charged oxygen molecule in the threo-Ex; thus, threo-Ex was destabilized and the erythro-PE intermediate and erythro-99 product were formed favorably (Scheme  [¹6] ).

Scheme 16

To test the aforementioned steering effect between the hydroxyl group and oxygen, we examined the diastereoselectivity of chiral allylic alcohol 100 (Table  [¹0] ). [¹¹³b] Taking into account that the ¹O2 ene reaction proceeded through the formation of a PE intermediate, we defined as syn-PE the intermediate where the oxygen was placed syn to the hydroxyl group, and as anti-PE where the oxygen was placed on the less-substituted site of the double bond. The measured regioselectivity and diastereoselectivity in the photooxidation of 100 in CHCl3 or MeCN are shown in Table  [¹0] . In CHCl3 or MeCN, the syn/anti methyl reactivity was 84:16 or 82:12, respectively. This reactivity was solvent-independent and it was rationalized in terms of the ‘cis effect’ (Section 4.1).

Table 10 Regioselectivity and Diastereoselectivity in the ¹O2 Ene Reaction of 100

threo/erythro Selectivity
Solvent syn anti
CHCl3 77:7 13:3
MeCN 63:19 13:5

As shown in Table  [¹0] , in CHCl3, the syn-PE resulted in a 11:1 threo to erythro diastereoselectivity. On the other hand, the anti-PE afforded threo diastereoselectivity (threo/erythro = 4.3:1), but to a minor extent compared to the syn-PE. For the syn-PE, it seemed likely that oxygen interacted with the hydroxyl group in a TS with the minimum 1,3-allylic strain (TS syn - threo , Figure  [²0] ) and this directed the reaction diastereoselectivity. For the anti-PE, we suggested that the threo diastereoselectivity could be attributed to the transition state TS anti - threo (Figure  [²0] ). In TS anti - threo , oxygen approached the double bond from the face where the least steric repulsions were developing. Ultimately, the conformations which led to erythro adducts seemed to be less favorable.

Figure 20

In the more polar MeCN, a decrease of the threo/erythro ratio was found in both syn and anti intermediates. This decrease was more significant in the case of the syn intermediate. For the syn-PE, solvation of hydroxyl group through hydrogen bonding reduced the favorable oxygen-hydroxyl steering effect. Therefore, the threo/erythro ratio decreased from 11:1 in CHCl3 to 3.3:1 in MeCN. Similarly, for the anti-PE, the threo/erythro selectivity decreased from 4.3:1 in CHCl3 to 2.6:1 in MeCN. This could be attributed to the solvation of the hydroxyl group, which increased its bulkiness and reduced the energy difference between TS anti - erythro and TS anti - threo (Figure  [²0] ).

In a collaborative study, Adam’s and Orfanopoulos’ groups investigated the regioselectivity and diastereoselectivity of the ¹O2 ene reaction with tetrasubstituted and 1,3-allylically strained chiral allylic alcohol 101 (Table  [¹¹] ). [¹²5] Although this tetrasubstituted substrate is structurally closely related to the trisubstituted allylic alcohol 100, it was envisaged that with the additional methyl group in the trix position, a more precise view of the enophilic reaction coordinate could be obtained. In particular, the steric demand of the two sites of the double bond is more comparable in 101 than in 100 and the allylic hydrogen atoms of the trix-methyl group provide additional coordination for the incoming ¹O2 (‘cis effect’, Section 4.1). Moreover, the trix-methyl group can induce 1,2-allylic strain (¹,² A) with the stereogenic lone site, in addition to the already existing 1,3-allylic strain (¹,³ A) between the twix and lone substituents. The measured regioselectivity and diastereoselectivity in the ¹O2 ene reaction of 101 in CDCl3 or CD3OD are shown in Table  [¹¹] . Clearly, the regioselectivity was found to be solvent-independent. Furthermore, in twix-101 and twin-101, the formation of threo stereoisomers was favored over the erythro.

The above mentioned findings were rationalized by the intermediacy of threo- and erythro-PE intermediates. The formation of these intermediates preceded by more loosely aggregated enophile/substrate encounter complexes (ECs); the threo-EC and erythro-EC structures for 101 are illustrated in Scheme  [¹7] . For the threo-EC case, 1,3-allylic strain was minimized, such that the allylic hydroxyl group pointed above the substrate plane engaging in hydrogen bonding with the ¹O2 attack from the top π face. This EC preferably afforded the threo-configured PE and then the corresponding ene product. On the other hand, 1,2-allylic strain was minimized for the erythro-EC structure, the hydroxy group points below the substrate plane, and ¹O2 attacked from the bottom π face forming the erythro-configured PE and ene product. Since in all cases the threo-configured ene product prevailed, it seems reasonable to assume that 1,3-allylic strain dominated rather than 1,2-allylic strain.

Table 11 Regioselectivity and Diastereoselectivity in the ¹O2 Ene Reaction of 101

Selectivity
twix-101 twin-101 trix-101
Solvent (threo/erythro) (threo/erythro)
CDCl3  24  40 36
(95:5) (86:14)
CD3OD  22  38 40
(70:30) (60:40)

Scheme 17

Once ¹O2 had selected the favored π face of the double bond by hydrogen bonding (threo attack conditioned by ¹,³ A strain), it had still the choice for hydrogen abstraction between the two sites of the double bond affording either the twix regioisomer (syn site) or the twin/trix regioisomers (anti site). In particular, in the favourable threo-EC structure, the ¹O2 enophile was still sufficiently mobile to specify the regioselectivity between the two sites of the double bond, the twix regioisomer through TS syn or the twin/trix regioisomers through TS anti (Scheme  [¹7] ). As seen from Table  [¹¹] , the twin/trix regioisomers were preferred over the twix one. This indicated that the coordination of ¹O2 with the allylic hydrogen atoms (‘cis effect’, Section 4.1) on the anti site was more effective than hydrogen bonding with the hydroxy group on the syn site.

It should also be mentioned that the threo/erythro diastereoselectivity was lowered in MeOH due to competitive hydrogen bonding with this protic solvent. Specifically, the overall threo/erythro diastereoselectivity of twix-101 regioisomer dropped from 95:5 in CDCl3 to 70:30 in CD3OD (Table  [¹¹] ). Similarly, for the twin-101 regioisomer the threo/erythro diastereoselectivity decreased from 86:14 to 60:40 in CDCl3 versus CD3OD. At this point, we suggested that steric effects were responsible for this observation. The allylic hydroxy group was solvated by MeOH, and therefore its relative size was larger than the methyl group of the lone substituent. Consequently, the favored conformation of the substrate changed to a rotamer, in which the solvated hydroxy group was at the periphery of the alkene, as shown in MeOH-EC (Figure  [²¹] ). For this rotamer, the enophilic attack was still favored from the threo site, because the erythro attack was sterically hindered by the methyl group of the lone substituent.

Figure 21

6 Stereochemistry

The first detailed stereochemical investigation of the ¹O2 ene reaction was reported in 1980. [¹²6] In particular, it was examined the correlation between the stereochemistry of C-O bond-making to C-H bond-breaking in the title reaction with the optically and isomerically pure alkene 102 (Scheme  [¹8] ). Photooxidation of 102 resulted in 82% abstraction from the benzylic group and exclusive formation of the trans allylic hydroperoxide. [50] Concerning the involvement of a PE intermediate, approach of ¹O2 from the top face of the double bond would form the R tertiary hydroperoxide and in order to produce trans alkene would abstract D. Similarly, ¹O2 approach from the bottom face of the double bond would form the S allylic hydroperoxide with H abstraction. These two possible products were labeled as R D-102 and S H-102 (Scheme  [¹8] ). Indeed, in this case, chiral alkene 102 reacted in a suprafacial manner with ¹O2 from either the top, or the bottom face of the double bond forming R D-102 and S H-102, respectively. No ‘crossover products’ (products which are derived from the antarafacial attack) were observed. It is important to mention that proton integration in the vinyl region of the ¹H NMR spectrum of ene products revealed no isotopic discrimination (k H/k D = 1.01). Moreover, the diastereomeric ratio, which was obtained by ¹H NMR examination of the methyl protons in the presence of a chiral shift reagent, was found to be S H/R D = 1.0. Overall, it was concluded that, due to the observed stereoisomeric-isotopic relationship, the ¹O2 ene reaction was a highly stereospecific suprafacial process.

Scheme 18

Taking into account the aforementioned findings, we have recently investigated the stereochemistry of the ¹O2 ene reaction with an optically active alkene. [56] This alkene, namely (S,S)-cis-1,4-diphenyl-2-butene-1,4-d 2 (103, Figure  [²²] ), has three distinctive characteristics: (1) chirality at the two reactive allylic carbons C1 and C4, by virtue of stereospecific deuteration, (2) different groups at both ends of the double bond such that the ene adducts would contain a new stereogenic center, and (3) a C 2 symmetry axis such that the two faces of the double bond are equivalent.

Figure 22

The photooxidations of 103 were run in CHCl3, Me2CO, MeCN and MeOH. In all cases the ene products, which are depicted in Scheme  [¹9] , were obtained in high yield. As expected, only the trans allylic hydroperoxides were found in the reaction mixture. [50] Since the two faces of the double bond were equivalent, we shall present here the mechanistic possibilities considering only one of them. Approach of ¹O2 from the top face would abstract D atom and form a new R stereogenic center, whereas abstraction of H would form a new S stereogenic center (Scheme  [¹9] ). We defined the new stereogenic centers as R D or S H and the two diastereomeric ene adducts were labeled (R D,R)-103 and (S H,R)-103.

The ratio of products (S H,R)-103/(R D,R)-103, which is the result of isotopic H/D competition for abstraction at the two allylic centers of alkene 103, was proportional to the primary intramolecular isotope effect (Scheme  [¹9] ). In the photooxidation of 103 in CHCl3, integration of the vinylic signals of (S H,R)-103 and (R D,R)-103 stereoisomers was used in order to determine the primary isotope effect which was found to be equal to k H/k D = 1.20 ± 0.05. The same isotope effect was measured when Me2CO, MeCN or MeOH were used as the reaction solvents. On the other hand, the diastereomeric ratio was determined by ¹H NMR examination of the diastereotopic benzylic protons of (S H,R)-103 and (R D,R)-103. It is worth mentioning that these protons were measurably separated in the absence of any chiral shift reagent. In CHCl3, the ratio of the newly formed stereogenic centers was measured to be S H/R D = 1.23 ± 0.05. The unequal formation of these stereogenic centers (10% ee of the S enantiomer) was due to the isotopic competition for abstraction at the chiral allylic centers of the olefin 103. Identical diastereomeric ratios were found when Me2CO, MeCN, or MeOH was used as the solvent in the photooxidation of 103, under similar reaction conditions. Herein, it is important to emphasize the correspondence of the diastereomeric ratio (S H,R)/(R D,R) of 1.23 with the isotopic k H/k D ratio of 1.20.

Scheme 19

These findings were best rationalized by involving the formation of a PE intermediate (Scheme  [¹9] ). From this intermediate, abstraction of deuterium and subsequent C-O bond formation led to the R D stereogenic center with only hydrogen remaining in the product double bond. Similarly, abstraction of hydrogen led to the formation of the S H stereogenic center. Notably, these results confirmed that the ¹O2 allylic oxidation of simple olefins is a highly stereospecific suprafacial process, independent of solvent polarity. Had the crossover products [(R H,R)-103 and (S D,R)-103] been formed, the ¹H NMR would have been more complicated. Moreover, the observation of an isotope effect, which matched exactly the stereogenic ratio, made it difficult to argue that the title reaction proceeded through an open biradical or dipolar intermediate. In that case, open intermediates OI1 and OI2 would be formed in equal amounts (Figure  [²³] ). Thus, neither (R H,R)-103 nor (S D,R)-103 products would be preferentially formed, and because of free rotation around the previous C-C double bond, a non-stereospecific reaction leading to racemic products could be expected.

Figure 23

7 Hypersensitive Probes in the ¹ O 2 Ene Reaction

The aim of our recent study [55] was to investigate from another point of view the possible involvement of an open biradical/dipolar intermediate in the ¹O2 ene reaction. For this purpose, we designed and assayed informative alkenes bearing the 2,2-diphenylcyclopropyl group as a mechanistic probe. [55] It is noteworthy that substituted cyclopropyl groups have been used as traps for other radical intermediates, [¹²7] since they involve the rapid rearrangement of the cyclopropylcarbinyl radical (104) to the homoallyl radical (104a, Scheme  [²0] ). [¹²8] Newcomb and co-workers reported that 2,2-diphenylcyclopropylcarbinyl radical (105) rearranged exceedingly fast with experimental rate constant of 5 × 10¹¹ s at room temperature. [¹²9] This probe has been often used as a radical clock [¹³0] to quantify a radical lifetime.

Scheme 20

The photooxidations of either E-106 or Z-106 (Table  [¹²] ) were run in CHCl3, Me2CO and MeCN, or MeOH. In all cases, isomeric ene allylic hydroperoxides bearing an intact cyclopropyl group were formed and isolated exclusively. Due to the fact that these ene adducts contained two stereogenic centers, the reaction mixture consisted of two pairs of diastereomeric hydroperoxides (total of eight isomers); thus, the ¹H NMR analysis was considerably complicated. Gratifyingly, after the reduction of the reaction mixture with Ph3P, each diastereomer of the two pairs 106a,b and 106c,d (Table  [¹²] ) was separated and characterized by ¹H NMR spectroscopy.

Table 12 The ¹O2 Ene Reaction of Alkenes E-106 and Z-106 a

Alkene 106a,b 106c,d
E-106 23 77
Z-106 70 30

a Identical product ratios were found when Me2CO, MeCN, or MeOH was used as the solvent in the photooxidation of 106, under similar reaction conditions.

Interestingly, for the photooxidation of E-106 in CHCl3, there was found a substantial preference for hydrogen abstraction from the methyl group geminal to the 2,2-diphenylcyclopropyl substituent of the double bond (Table  [¹²] ). This regioselection was attributed to the previously mentioned large-group nonbonding effect (Section 4.3). For the photooxidation of Z-106 in CHCl3, the more substituted site of the double bond was the more reactive (Table  [¹²] ). This site selectivity was in accordance with the ‘cis effect’ (Section 4.1). Furthermore, these trends in regioselectivity were independent of solvent polarity.

The proposed mechanism that could account for the ¹O2-addition to alkene E-106 is presented in Scheme  [²¹] . Importantly, the ¹O2 ene reaction of this substrate should lead to distinctly different products depending on the adopted mechanistic path. For instance, when a PE intermediate was invoked, ene products with cyclopropyl groups intact could be formed. On the other hand, if an open intermediate OI (biradical or dipolar) with a lifetime greater than 10-¹¹-10-¹² s [¹²9] [¹³¹] had been formed, the ring-opened products might have been detected. In our case, the exclusive formation of the unrearranged products 106a-d and the observed site selectivity were better rationalized by the intermediacy of a PE. Subsequently, in TS1 leading to the minor ene products 106a,b, the nonbonding interactions involving the large diphenylcyclopropyl group, were expected to be stronger than those in TS2 where these interactions are much less pronounced. A similar mechanism was also suggested for the photooxidation of alkene Z-106.

Scheme 21

As additional evidence for the nature of the intermediate in ¹O2 ene reaction, we used the second-generation hypersensitive probe (107, Scheme  [²²] ). This probe was capable of distinguishing between biradical and dipolar intermediates. [¹³²] Consequently, in ring opening of this cyclopropyl carbinyl system, the phenyl group stabilizes an incipient radical more effectively than the methoxy group, whereas the methoxy group favors an incipient carbocation.

Scheme 22

Specifically, we carried out the photooxidations of E/Z-108 in CHCl3, Me2CO, MeCN, and MeOH. In all cases, after the reduction of the reaction mixture with Ph3P, we managed to isolate by column chromatography the mixture of allylic alcohols 108a-d and 108e-h in which the cyclopropyl ring remained intact (Scheme  [²²] ). In agreement with our previous results in this section, the proposed mechanism that could account for the formation of the allylic alcohols 108a-h seemed to be similar to that already presented in Scheme  [²¹] .

In conclusion, two hypersensitive radical probes were used for the investigation of ¹O2-mediated ene reaction in various solvents. In all the reaction conditions used herein, no rearranged products were detected. Our findings, in spite of their limitations, seemed unreasonable to argue for a biradical or dipolar mechanism.

8 The ¹ O 2 Ene Reaction in Confined Media

In recent years, researchers have paid a great deal of attention to performing photooxidations in microreactors. [5³b] [¹³³] Considering the ¹O2 ene reaction, the dye-supported Y-type zeolites [¹³4] have been widely used as confined media. [¹³5] The measured intramolecular and intermolecular isotope effects in the intrazeolite allylic oxidation of deuteriumlabeled TMEs were very similar [¹³6] to those reported in solution (see Section 3). Hence, the intrazeolite reaction mechanism is stepwise involving the irreversibly formation of PE-like intermediate in the rate-determining step. Notably, the regioselectivity [¹³7] as well as the diastereoselectivity [¹³8] of the title reaction is remarkably affected within the zeolite cavities. Generally, the substantial changes in the ene selectivity are explained considering: (1) the electrostatic interaction between the pendent negatively charged oxygen atom in the PE-like intermediate and the alkali metal cation (mainly Na+) of zeolite, (2) conformational effects within the zeolite supercages and (3) the interaction between the π system of the reactant and the alkali metal cation of zeolite.

Interestingly, Tung and co-workers investigated the photooxidation of α-pinene sensitized by 9,10-dicyanoanthracene (DCA) [¹³9] in Nafion-Na+ membranes [¹40] or surfactant vesicles. [¹4¹] The oxidation within these microreactors selectively gave either the ¹O2-mediated, or the superoxide radical anion (O2 -˙) mediated products (depending on the status and location of the substrate and sensitizer molecules in the reaction media). In contrast, in homogeneous solution both the energy transfer and the electron transfer pathways were observed. [¹4²] Last but not least, in recent years the study of the ¹O2 ene reaction using alternative microreactors, such as cyclodextrin, [¹4³] polystyrene beads, [²²a] [¹44] pentasil zeolites, [¹45] mesoporous silica SBA-15, [¹46] octa acid capsules, [¹47] and IRA-resin, [¹48] has received considerable attention.

9 Concluding Remarks

Recent years have seen a burgeoning in research interest directed toward unraveling the chemistry of ¹O2. It is generally accepted that, among the ¹O2 additions to unsaturated substrates, the ene reaction has attracted the most extensive experimental and theoretical attention. This attention has been motivated by its powerful synthetic potential, since ¹O2 ene reaction exhibits fascinating regio- as well as stereoselectivity. In this account, we presented an overview of the work accomplished in our laboratory regarding the mechanism of this reaction without, however, ignoring the equally important contributions of other research groups in the field. In the light of the majority of experimental and theoretical findings, we strongly suggest that the ¹O2 ene reaction proceeds via the irreversible formation of a PE-like intermediate. We finally believe that the diversity of ¹O2 will continue to fascinate researchers in the fields of mechanistic and synthetic organic chemistry as well as in biological systems.

Acknowledgment

We gratefully acknowledge our collaborators whose names appeared in the references and their valuable contributions made the described work possible. The Foundation for Education and European Culture is warmly acknowledged for providing a one-year fellowship to M.N.A.

Scheme 1 Synthesis of 1,2,4-trioxane moiety via the ¹O2 ene reaction

Scheme 2 Proposed mechanisms for the ¹O2 ene reaction

Figure 1

Figure 2 Intermolecular KIEs on the ¹O2 ene reaction of tetrasubstituted alkenes (vs = versus)

Figure 3 Intramolecular KIEs on the ¹O2 ene reaction of TMEs

Figure 4 Open biradical/dipolar (OI) and perepoxide (PE) intermediates in the ¹O2 ene reaction of TMEs-d 6

Figure 5 Intermolecular and intramolecular KIEs on the ¹O2 ene reaction of disubstituted and trisubstituted alkenes

Figure 6 ‘Cis effect’ selectivity in the ¹O2 ene reaction of enol ethers and trisubstituted alkenes (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Scheme 3

Figure 7

Figure 8Anticis effect’ selectivity in the ¹O2 ene reaction of certain trisubstituted alkenes (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Figure 9

Figure 10 The large group nonbonding effect selectivity in the ¹O2 ene reaction of disubstituted and trisubstituted alkenes (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Scheme 4

Scheme 5 Green and red numerical values indicate percentage of double-bond formation in the ene adducts

Scheme 6 Green and red numerical values indicate percentage of double-bond formation in the ene adducts

Figure 11Geminal selectivity on the ¹O2 ene reaction of alkenes bearing a heteroatom or a large alkyl substituent at the allylic position (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Scheme 7

Figure 12Geminal selectivity on the ¹O2 ene reaction of alkenes bearing a large alkyl substituent at the vinyl position.a a Green and red numerical values indicate percentage of double-bond formation in the ene adducts

Figure 13 Red and blue numerical values indicate the calculated rotational barriers and the percentage of double-bond formation in the ene adducts, respectively

Figure 14Geminal selectivity on the ¹O2 ene reaction of alkenes bearing an electron-withdrawing group at the α-position (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Scheme 8

Figure 15 Site selectivity on the ¹O2 ene reaction of alkenes bearing an electron-withdrawing group at the β-position (green and red numerical values indicate percentage of double-bond formation in the ene adducts)

Figure 16 Regioselectivity in the self-sensitized oxygenation of fullerene derivatives (numerical values indicate percentage of double-bond formation in the ene adducts)

Figure 17

Scheme 9

Scheme 10

Scheme 11

Figure 18 Intermolecular KIEs on the ¹O2 ene reaction of β,β-­di­methyl­styrenes (vs = versus)

Scheme 12

Figure 19

Scheme 13 Green and red numerical values indicate percentage of double-bond formation in the ene adducts

Scheme 14

Scheme 15

Scheme 16

Figure 20

Scheme 17

Figure 21

Scheme 18

Figure 22

Scheme 19

Figure 23

Scheme 20

Scheme 21

Scheme 22