Mechanism of the Aerobic Oxidation of α-Pinene



A combined experimental and theoretical approach is used to study the thermal autoxidation of α-pinene. Four different types of peroxyl radicals are generated; the verbenyl peroxyl radical being the most abundant one. The peroxyl radicals propagate a long radical chain, implying that chain termination does not play an important role in the production of the products. Two distinct types of propagation steps are active in parallel: the abstraction of allylic H atoms and the addition to the unsaturated C[DOUBLE BOND]C bond. The efficiency for both pathways appears to depend on the structure of the peroxyl radical. The latter step yields the corresponding epoxide product, as well as alkoxyl radicals. Under the investigated reaction conditions the alkoxyl radicals seem to produce both the alcohol and ketone products, the ketone presumably being formed upon the abstraction of the weakly bonded αH atom by O2. This mechanism explains the predominantly primary nature of all quantified products. At higher conversion, co-oxidation of the hydroperoxide products constitutes an additional, albeit small, source of alcohol and ketone products.


Selective oxidations play an important role in the chemical value chain as they convert relatively cheap molecules into value-added products.1 Although various oxidants can be used, O2 is desired both from an economical and an environmental point of view. Improving both the efficiency and selectivity toward the desired products remains a prime objective of academic and industrial research.2 An important class of aerobic oxidations are radical-propagated autoxidations.14 Important bulk scale processes using this technology are, for example, the oxidation of p-xylene to terephthalic acid (44×106 tonnes per year), the oxidation of cyclohexane to cyclohexanol and cyclohexanone (6×106 tonnes per year), and the synthesis of ethylbenzene hydroperoxide (6×106 tonnes per year). The autoxidation mechanism of an unfunctionalized alkane, RH, in the absence of a catalyst, as assumed up to 2005,3 is summarized by Equations (1)–(6), ((2)), ((3)), ((4)), ((5)), ((6)):

equation image((1))
equation image((2))
equation image((3))
equation image((4))
equation image((5))
equation image((6))

Equation (1) is the homolytic dissociation of a hydroperoxide molecule. The O[BOND]O bond is the weakest bond in the system, rationalizing why one often adds a small amount of ROOH to initiate the reaction. The O-centered radicals produced in the initiation step [Equation (1)] are rapidly converted into C-centered radicals via Equations (2) and (3). Alkyl radicals react in a diffusion-controlled manner with O2 [Equation (4)], producing the chain-carrying peroxyl radicals, ROO. Peroxyl radicals also react with the substrate [Equation (5)], albeit slower than alkoxyl and hydroxyl radicals, making them the most dominant radicals in the system. Equation (5) regenerates the alkyl radical and closes the propagation cycle, which is repeated many times before the peroxyl radicals are destroyed in the chain termination step [Equation (6)], producing an equimolar amount of the alcohol (ROH) and ketone (Q[DOUBLE BOND]O). Equation (6) compensates Equation (1) and precludes a radical runaway. In fact, a radical quasi-steady-state is established,5 implying that the rate of chain termination is equal to the rate of chain initiation. The ratio of the rates of propagation and termination is called the chain length, usually a large number (≫1), justifying the term chain mechanism. Hence, according to this traditional mechanism one would expect a large [ROOH]/[Q[DOUBLE BOND]O] product ratio. However, this is not observed; the concentrations of the major end products (ROOH, ROH, and Q[DOUBLE BOND]O) are of the same order of magnitude. This implies that crucial chain-propagating reactions leading to the end products are missing in the above mechanism.6

In addition to Equation (5), the reaction of peroxyl radicals with the oxygenated products also needs to be considered. In particular, the abstraction of the αH atom of ROOH was found to be a very fast reaction, 20 and 50 times faster than Equation (5) for ethylbenzene7 and cyclohexane,6, 8 respectively. Ab initio calculations demonstrated that the resulting R−αHOOH radical does not exist and spontaneously dissociates to Q[DOUBLE BOND]O and OH.9, 10 This implies that the abstraction of the αH atom of the hydroperoxide is a fast and straightforward source of the ketone. The OH radical will rapidly abstract an H atom of ubiquitously present RH molecules, producing an alkyl radical. The overall exothermicity of these two subsequent steps is approximately 50 kcal mol−1 (1 cal=4.184 J), creating a sudden and local temperature increase (hot spot), markedly affecting the fate of the reaction products. Indeed, either the nascent products (ROOH, R, H2O, Q[DOUBLE BOND]O) can diffuse away from each other, or the caged ROOH and R products can react together, as shown in Equation (7):

equation image((7))

Although facing a higher activation barrier than the diffusive separation, Equation (7) can compete because of the local hot spot. The efficiency of the reaction depends on the stability of the alkyl radical: the more stable the radical, the lower its reactivity and the lower the efficiency of the cage reaction. This efficiency has been experimentally measured for the three different hydrocarbon substrates: cyclohexane,6, 8 toluene,11 and ethylbenzene,7 as 70, 55, and 20 %, respectively, and is in line with expectations. For the case of the cyclohexane oxidation, the alkoxyl radicals co-produced in Equation (7) are partially converted to additional alcohol upon reaction with the substrate [Equation (2)], and also partially isomerized into ω-formyl radicals [Equation (8)].

equation image((8))

In cyclohexane autoxidation, this radical has been identified as the most important precursor (ca. 80 %) of the ring-opened byproducts, such as 6-hydroxyhexanoic acid and adipic acid.12, 13 Up until a few years ago, it was wrongly assumed that the waste products originated from the over-oxidation of cyclohexanone. Yet, in the recent study above,13 CyOOH was unambiguously identified as the crucial precursor of both the desired products and the byproducts. This case study on the autoxidation of cyclohexane illustrates the power of a detailed quantitative mechanistic investigation.

The initiation reaction [Equation (1)] could not explain the observed initiation rates as the 40 kcal mol−1 activation energy barrier makes the reaction extremely slow. Additionally, Equation (1) is also very inefficient in generating free radicals. Indeed, in the liquid phase, the nascent RO and OH radicals would rather recombine in their solvent cage than diffuse away from each other and start a radical chain reaction. Experimentally we were able to measure the rate of radical formation during the autoxidation of cyclohexane and found it to be proportional to the initial cyclohexanone concentration.14 Quantification of the experimental data, combined with quantum chemical and theoretical kinetic calculations, allowed us to identify the true initiation reaction as a bimolecular reaction between cyclohexyl hydroperoxide and cyclohexanone. In this reaction, the OH radical breaking away from the hydroperoxide abstracts a weakly bonded αH atom of cyclohexanone, producing a resonance-stabilized ketonyl radical. This reaction faces a lower energy barrier and is therefore much faster than Equation (1). In addition, it is also much more efficient as the in-cage recombination is slower. The ketone concentration increases in a nearly exponential manner during the reaction. The rate of initiation also increases very quickly. Until now, this overlooked initiation mechanism could be identified as the core of the autocatalytic mechanism of cyclohexane oxidation. Substrates that lack products with weak and poorly accessible C[BOND]H bonds (e.g., ethylbenzene7) do not show this autocatalytic upswing. For such systems, the initiation reaction is a bimolecular reaction between the ROOH product with RH, yielding RO, H2O, and R.14 Note that these ROOH-based initiation mechanisms are only valid after the induction period. Currently, it remains unclear how the first radicals are generated. A mechanism that has often been suggested in the literature is the abstraction of a H atom from RH by O2. However, the lifetime of this reaction has been estimated at 2 billion years for the case of cyclohexane, even at 300 °C.15 It seems more likely that trace amounts of impurities, for example, ROOH, are responsible for the initial initiation.

Radical chain oxidations are not only used in the bulk chemical industry but also in the synthesis of valuable fine chemicals. An interesting example is the oxidation of the renewable olefin α-pinene to a mixture of various compounds used in the synthesis of fragrances and flavors. One important oxidation product, α-pinene oxide, is isomerized to campholenic aldehyde. This molecule is the starting material in the synthesis of sandalwood-like fragrances, for example sandalore (Givaudan) and polysantol (Firmenich).16 Verbenol, another component of the oxidation mixture, is a well-known aggregation pheromone of the bark beetle and is thus utilized in forestal pest control.17 Despite its industrial and academic interest, the basic chemistry behind the oxidation process is not well understood.18, 19 The biotechnological oxidation of α-pinene has also been investigated, but continues to be a challenging task, owing to the long reaction time.20

A detailed understanding of the molecular mechanisms of autoxidation conditions would not only be useful for optimizing reaction parameters and designing appropriate catalysts,2126 but could also inspire a broadening of the scope of the autoxidation substrate. Herein, experimental investigations are combined with quantum chemical calculations to gain a quantitative insight into the reaction mechanism of the aerobic oxidation of α-pinene.

Results and Discussion

Some preliminary overall observations

The oxidation of α-pinene was studied at 363 K and 1 bar (1 bar =105 Pa) of pure O2 (see Experimental Section). Figure 1 shows the evolution of the α-pinene conversion as a function of time. It can be observed how after an induction period of ca. 3 h the conversion starts to increase. However, the conversion upswing is weaker than during the oxidation of cyclohexane,14 but similar to the behavior of ethylbenzene.7

Figure 1.

Time evolution of the conversion of α-pinene at 363 K.

Although not always appreciated in the literature, a multitude of products is formed (Scheme 1). For example, at a conversion as low as 2 %, all of the products shown in Figure 2 are already present: α-pinene oxide (29.0 %), verbenyl-hydroperoxide (29.0 %), verbenol (13.5 %), verbenone (7.0 %), pinenyl-hydroperoxide (4.0 %), pinenol (0.5 %), pinocarvyl-hydroperoxide (6.0 %), pinocarveol (1.5 %), pinocarvone (1.0 %), myrtenyl-hydroperoxide (5.0 %), myrtenol (2.0 %), and myrtenal (1.5 %).27 At higher conversions (>10 %), additional products are formed upon over-oxidation or rearrangement of the primary products (e.g., campholenal and sobrerol, coming from α-pinene oxide). Figure 2 a shows the evolution of the most important products as a function of the sum of products, up to approximately 20 % conversion. This plot suggests that these products are mainly primary in origin, although one can also anticipate a secondary contribution to the production of verbenol and verbenone (see below). Regardless, the product distribution is much less conversion-dependent than for the autoxidation of cyclohexane. A similar evolution was observed for the other α-pinene oxidation products (see Figure 2 b for the myrtenyl products).

Scheme 1.

Main products observed during the thermal oxidation of α-pinene.

Figure 2.

a,b) Evolution of the most important α-pinene oxidation products as a function of the conversion. The reported selectivities are valid at an α-pinene conversion of 10 % at Σ[P]=630 mM.

The large number of products arises from the fact that radicals can abstract H atoms from two different C atoms (denoted a and d, see Scheme 2). Abstraction of an H atom at the a-site results in the formation of a resonance-stabilized radical from which both verbenyl and pinenyl products can be formed (Scheme 1). Abstraction at the d-site also leads to a resonance-stabilized radical, giving rise to pinocarvyl and myrtenyl products (Scheme 1). The experimental ratio between the verbenyl+pinenyl and the pinocarvyl+myrtenyl products equals approximately 3. Abstraction of other H atoms results in radicals that are not resonance-stabilized or radicals containing a lot of ring strain. These abstractions face substantially higher energy barriers and are therefore of much less importance,28 especially under the mild reaction temperature of 363 K.

Scheme 2.

Four different oxidation sites (denoted a–d) in α-pinene.

H abstractions by peroxyl radicals: Input from quantum-chemical calculations

The most abundant radicals in the system are peroxyl type radicals (see below) and it is very useful to investigate how fast they react with the different C[BOND]H bonds in the substrate. For computational simplicity we used a smaller model radical, tert-butyl-peroxyl. Abstraction of the two H atoms at the a-site faces slightly different energy barriers; 12.5 and 13.6 kcal mol−1, depending on the H atom’s cis or trans orientation toward the dimethyl bridge (B3LYP/6-311++G(df,pd)//B3LYP/6-31G(d,p) level of theory). Based on the typical prefactor per H atom (3.25×108M−1 s−1[7]) one can estimate a rate constant kabs,a(363 K) of 11.5 M−1 s−1. The abstraction of the three H atoms at the d-site was calculated to face a barrier of 13.5 kcal mol−1, resulting in a kabs,d(363 K) of 7.0 M−1 s−1. Note that the computed kabs,a/kabs,d of approximately 2 is in good agreement with the experimental (verbenyl+pinenyl)/(pinocarvyl+myrtenyl) product ratio of approximately 3. The reactivity difference of ca. 1 kcal mol−1 between the a- and d-site H atoms is mainly caused by the difference in C[BOND]H bond strength (ca. 80.7 and 82.6 kcal mol−1, respectively), stemming from the secondary and primary nature of these H atoms.28

The resonance-stabilized radicals are shown in Scheme 3, together with the Mulliken atomic spin densities on the most relevant atoms. Addition of O2 to these resonance-stabilized radicals can result in four different types of peroxyl radicals: R(a)[BOND]OO, R(b)[BOND]OO, R(c)[BOND]OO, and R(d)[BOND]OO.27

Scheme 3.

The two resonance-stabilized radicals, formed upon abstraction of an H atom at the a- and d-sites (left and right, respectively), and the Mulliken atomic spin densities on the most relevant atoms (B3LYP/6-311++G(df,pd)//B3LYP/6-31G(d,p)-level).

These peroxyl radicals give rise to a wide range of observed allylic oxidation products (see Scheme 1). It is interesting to note that the observed pinocarvyl/myrtenyl products ratio of 0.95±0.05 is in good agreement with the computed spin density ratio of positions b and d, that is, 1.05, in the radical formed upon the abstraction of a d-site H atom (Scheme 3, right). On the other hand, the experimental verbenyl/pinenyl products ratio of 9.5±0.3 is much larger than the ratio of calculated spin densities of 1.03 for the a- and c-sites, in the radical formed upon the abstraction of an a-site H atom (Scheme 3, left). Most probably, this deviation arises from different O2 addition kinetics; it is indeed likely that the methyl group at the c-site would sterically hinder an approaching O2 molecule and hamper its addition to this c-site.

Addition of peroxyl radicals to the unsaturated C[DOUBLE BOND]C bond

In addition to the allylic oxidation products, one also observes α-pinene oxide. Epoxidation of olefins under autoxidation conditions has been known for a long time and has been ascribed to the addition of peroxyl radicals to the C[DOUBLE BOND]C bond, followed by a unimolecular ring-closure, eliminating RO.3 In the case of α-pinene, these two steps are shown in Scheme 4.

Scheme 4.

Epoxidation of α-pinene under autoxidation conditions.

Note that ROO radicals preferably add to the unsubstituted C-atom (the b-site) to form the more stable tertiary radical. For the addition of tert-butylperoxyl radicals, a barrier of 13.4 kcal mol−1 was computed (B3LYP/6-311++G(df,pd)//B3LYP/6-31G(d,p) level of theory), independent of from which side the olefin is approached (cis or trans to the dimethyl bridge). Combining this barrier with the reported prefactor for the addition of CH3OO to propylene,29 (A9=4.0×108M−1 s−1), results in k9(363 K) of approximately 7 M−1 s−1, taking into account a reaction path degeneracy of 2 (addition can occur cis or trans to the dimethyl bridge). The barrier to the second step in the epoxidation mechanism, is computed to be 5.5 kcal mol−1, leading to a transition state theory (TST) calculated rate constant of k10(363 K)=2×109M−1 s−1. This reaction stands in competition with the addition of O2 to the radical adduct (Scheme 5),30 a reaction that is probably diffusion-controlled, (k11(363 K)≈2×109M−1 s−1).

Scheme 5.

Addition of O2 to a radical adduct; a reaction that can potentially compete with the epoxidation shown in Scheme 4.

Nevertheless, under our conditions (1 bar O2), the estimated O2 concentration is only approximately 35 mM,31 meaning that the addition of O2 to the radical adduct cannot compete with the epoxidation mechanism. The second step in the epoxidation mechanism was also found to be much faster than the reverse step of the slightly endothermic addition: k10(363 K)/k−9(363 K)>1000, based on a conventional TST calculation.

Chain length

Under the assumption that most of the products are produced in a chain propagation reaction (chain length≫1), the rate of α-pinene oxidation is given by Equation (9):

equation image((9))

where kabs refers to the rate constant for allylic H abstraction (kabs,a+kabs,d) and k9 represents the rate constant for the addition of ROO radicals to the C[DOUBLE BOND]C double bond (viz. the rate-determining step in the epoxidation mechanism). The rate constants have already been estimated at k9(363 K)≈7 M−1 s−1 and kabs(363 K)≈18.5 M−1 s−1. This allows to make an estimation of the ROO concentration; for example, at 3 % conversion this is approximately 4×10−7M. This concentration is of the same order of magnitude as the values observed during the autoxidation of ethylbenzene.7

Assuming that over-oxidation of the primary reaction products is negligible (see below), the chain length is given by Equation (10):

equation image((10))

Unfortunately, the rate constant for termination is not known. Moreover, it is possible that kterm is in reality a function of the conversion as different types of peroxyl radicals are being formed. Nevertheless, the most dominant ROO radical can be assumed to be the verbenyl peroxyl R(a)[BOND]OO. As the self-reaction rate constant of alkylperoxyl radicals decreases with the alkyl size,32 it is likely that, based on an equal or even higher degree of steric hindrance, kterm≤3×106M−1 s−1, that is, kterm for cyclohexenyl-peroxyl radicals. Based on these values, the lower limit of the chain length is plotted in Figure 3 as a function of the conversion. Owing to the nearly linear increase of the ROOH concentration as a function of Σ[P] (see Figure 2), the concentration of ROO also increases because the rate of chain initiation is proportional to [ROOH] (dΣ[P]/dt is indeed verified to be proportional to [ROOH]0.5 as expected1). As a consequence, the chain length is found to decrease upon increasing conversion, although it remains quite high (ν>50).

Figure 3.

Lower limit of the chain length as a function of the α-pinene conversion.

Based on the estimated ROO concentrations, and assuming radical quasi steady-state,5 one can also evaluate the initiation over termination ratio, kinit/kterm, via Equation (11).((11))

equation image((11))

In contrast to the situation encountered for cyclohexane oxidation, kinit/kterm does not increase as a function of the conversion (Figure 4). This means that there is probably no reaction product assisting in the initiation, consistent with our observation that the initial addition of 1 mol % of R(a)[BOND]OH, R(a)[DOUBLE BOND]O, or PO does not significantly influence the autoxidation rate (viz. dΣ[P]/dt at a given conversion does not change upon the addition of these reaction products). Actually, one observes a slight decrease in kinit/kterm, which is probably due to the involvement of additional termination steps at higher conversion.

Figure 4.

Evolution of the kinit/kterm ratio as a function of the conversion.

Primary or secondary nature of the reaction products

A very convenient way to verify whether a given product is of primary or secondary origin is to plot the ratio [Pi]/Σ[P] versus Σ[P]; if the product is (partially) produced in a primary step, the intercept should be non-zero and equal to the fraction in which it is produced.11 This analysis was performed for the most important products; the results are shown in Figure 5. The first striking observation is that all products appear to be predominantly primary in origin. While this is readily explained for PO, R(a)[BOND]OOH, and R(a)[BOND]OH, the primary source of Q(a)[DOUBLE BOND]O has so far not been identified. Also interesting is the observation that the R(a)[BOND]OOH selectivity appears to decrease with Σ[P] whereas the R(a)[BOND]OH and Q(a)[DOUBLE BOND]O contribution slightly increases. So far the reaction mechanism does not account for these observations. Another minor effect is the small increase in epoxide (PO) selectivity. The latter observation could be a result of the fast over-oxidation of myrtenal (Q(d)[DOUBLE BOND]O) to the corresponding myrtenic acylperoxyl radical (Q(d)([DOUBLE BOND]O)OO) upon abstraction of its weak aldehyde H atom. Acylperoxyl radicals are known to be good epoxidizing agents.33 This hypothesis is supported by an 8 % increase in the PO selectivity (at 2 % α-pinene conversion) upon the initial addition of 1 mol % of myrtenal (Q(d)[DOUBLE BOND]O).

Figure 5.

Evolution of [Pi]/Σ[P] versus Σ[P] for the most important α-pinene oxidation products: α-pinene oxide (PO, filled circles); verbenyl-hydroperoxide (R(a)[BOND]OOH, triangles); verbenol (R(a)[BOND]O, open circles); and verbenone (Q(a)[DOUBLE BOND]O, stars).

The origin of alcohol and ketone

The primary nature of the R(a)[BOND]OH product can readily be understood, based on the proposed epoxidation mechanism (Scheme 4). Indeed, this epoxidation mechanism co-produces alkoxyl radicals (RO), which rapidly (much more rapidly than peroxyl radicals) abstract allylic H atoms, directly yielding the alcohol [Equation (2)]. However, not only the alcohol, but also the ketone seem to have predominantly primary character. The hypothesis that the ketone would exclusively originate from chain termination can be rejected, based on the observed long chain length (see above). On the other hand, one observes an initial linear correlation of both R(a)[BOND]OH and the Q(a)[DOUBLE BOND]O with the epoxide (Figure 6), suggesting that both the alcohol and the ketone originate from the same species, that is, alkoxyl radicals. A likely mechanism for Q(a)[DOUBLE BOND]O formation from R(a)[BOND]O is the abstraction of the weakly bonded αH atom by O2. The analogous reaction of O2 with cyclohexoxyl radicals has been kinetically characterized in the range 225–302 K;34 extrapolation of the rate constant to 363 K predicts a value of 3×107M−1 s−1. Given an O2 concentration of approximately 35 mM, the pseudo-first-order rate constant can be estimated at ca. 106 s−1. It appears reasonable that the analogous reaction with R(a)[BOND]O radicals is significantly faster, given the higher stability of the enone product. Therefore it is likely that the O2 reaction with R(a)[BOND]O can compete with the abstraction of allylic H atoms from α-pinene, the pseudo-first-order rate constant of which can be roughly estimated at 3×107 s−1.35 It is important to emphasize that one does indeed observe a decreased alcohol/ketone ratio at high O2 pressures. However, at higher O2 pressures several other reactions also become important and complicate the overall chemistry. Those issues will be addressed in a dedicated publication.

Figure 6.

Correlation of [R(a)[BOND]OH] and [R(a)[DOUBLE BOND]O] with [PO], up to 25 % conversion.

Although it seems reasonable that the R(a)[BOND]O radicals are converted to both R(a)[BOND]OH and Q(a)[DOUBLE BOND]O, in a ratio of approximately 2.1±0.1 (Scheme 6), it does not explain why this ratio would decrease at higher conversion (e.g., Figure 6); for example, at 25 % conversion, the R(a)[BOND]OH/Q(a)[DOUBLE BOND]O concentration ratio has dropped to 1.45. Over-oxidation of R(a)[BOND]OH can be excluded as an additional source of Q(a)[DOUBLE BOND]O becoming important at higher conversion, based on our experiment where we initially added 1 mol % of R(a)[BOND]OH and where no change in product distribution could be observed.

Scheme 6.

Proposed fate of the R(a)[BOND]O radicals; the experimentally observed r/s ratio is 2.1±0.1.

The ring-opening of the R(a)[BOND]O radical via C[BOND]C cleavage can be neglected, despite the formation of a conjugated enone product (barrier of 11.5 kcal mol−1 at the reliable36 B3LYP/6-31G(d,p) level of theory, first-order rate constant ca. 2×106 s−1). The reason for this is the enhanced ring strain in the product where the unpaired electron would be localized at a C atom of the four-membered ring.

We believe that the gradual shift in the R(a)[BOND]OH/Q(a)[DOUBLE BOND]O ratio is correlated with the observed secondary contribution to both the R(a)[BOND]OH and Q(a)[DOUBLE BOND]O production, observed in Figure 5. Therefore we ascribe this effect to the partial co-oxidation of the R(a)[BOND]OOH product, initiated upon the abstraction of the weakly bonded αH atom (Scheme 7). As already emphasized (see Introduction), all radicals of the type R−αHOOH dissociate promptly to Q[DOUBLE BOND]O+OH.10

Scheme 7.

Proposed co-oxidation pathway for R(a)[BOND]OOH.

Based on the initial d[R(a)[BOND]OH]/d[PO] and d[Q(a)[DOUBLE BOND]O]/d[PO] values (0.51 and 0.23, respectively) one can extrapolate the amount of R(a)[BOND]OH and Q(a)[DOUBLE BOND]O coming from the R(a)[BOND]O radicals produced in the epoxidation (Scheme 4) at higher conversion, based on the observed PO yield. The additional amount of the R(a)[BOND]OH and R(a)[DOUBLE BOND]O products (the amount observed minus the estimated amount coming from the subsequent chemistry of the R(a)[BOND]O radicals formed in the epoxidation) should then be ascribed to the mechanism in Scheme 7. The ratio between these additional amounts of R(a)[BOND]OH and Q(a)[DOUBLE BOND]O coming from R(a)[BOND]OOH co-propagation is given by Equation (12):

equation image((12))

where f represents the fraction of {Q(a)[DOUBLE BOND]O+ROOH+R+H2O} products of the R(a)[BOND]OOH co-propagation undergoing the cage-reaction as shown in Scheme 7; r and s represent the fractions of R(a)[BOND]O radicals reacting with the α-pinene and O2, respectively (Scheme 6). As can be observed in Figure 7, the plot of the additional R(a)[BOND]OH versus Q(a)[DOUBLE BOND]O is linear (R=0.97) with a slope of 0.19 from which f can be estimated at 0.11. This means that 11 % of the caged {Q[DOUBLE BOND]O+ROOH+R+H2O} products will undergo the activated cage-reaction shown in Scheme 7. As a comparison, this cage-fraction was measured to be 0.7, 0.55, and 0.2 for cyclohexane,68 toluene,11 and ethylbenzene,7 respectively. It should thus be emphasized that the obtained value for f of approximately 0.11 is fully in line with the cage-efficiencies determined before for other substrates.

Figure 7.

Plot of the additional R(a)[BOND]OH versus the additional Q(a)[DOUBLE BOND]O yields, stemming from the co-oxidation of R(a)[BOND]OOH product.

Note that this cage reaction (fraction f in Scheme 7) causes a net destruction of R(a)[BOND]OOH, explaining why its selectivity decreases as a function of the conversion (see Figures 2 a and 5). It is interesting to note that the R(d)[BOND]OOH selectivity decreases three times faster (viz. the more pronounced leveling-off of the R(d)[BOND]OOH contribution compared to R(a)[BOND]OOH in Figure 2 a and b). This can be attributed to a higher rate constant for αH abstraction from R(d)[BOND]OOH, owing to the presence of two αH atoms and a slightly looser TS (viz. less steric repulsion of the dimethyl bridge). The cage efficiencies are predominantly determined by the stability of the R radicals; the precise structure of the hydroperoxide is probably less important.

One important issue that has not been addressed so far is the fate of the HO2 radicals, produced in the reaction of the alkoxyl radicals with O2 (see Scheme 6, fraction s). Several reactions can be proposed:((13)), ((14)), ((15)), ((16)), ((17))

equation image((13))
equation image((14))
equation image((15))
equation image((16))
equation image((17))

Equations (13) and (14) represent the H abstraction and epoxidation mechanism, respectively, analogous to the ROO chemistry. Assuming a similar reactivity of the HO2 radical as the ROO radical, the combined pseudo-first-order rate constant for the consumption of RH by HO2 can be estimated at (k13+k14)[RH]≤200 s−1. Equations (15 and (16) are self- and cross-termination reactions. Assuming that both rate constants are equally fast (presumably diffusion-controlled, k15k16≈2×109M−1 s−1), the rate of Equation (16) will be much faster than (15), owing to [HO2]≪[ROO]. The pseudo-first-order rate constant of Equation (16) can be estimated at k16[ROO]≈400 s−1 (for a conversion of 0.5–1.5 %). Equation (17) represents the conversion of the HO2 radical into ROO. The barrier of the model reaction HO2+CH3OOH was computed to be as low as 4.6 kcal mol−1 owing to the formation of pre- and post-reactive complexes. Combining this barrier with a typical prefactor for H abstraction by peroxyl radicals (3.25×108M−1 s−1[7]) and a total [ROOH]≥10 mM (viz. [R(a)[BOND]OOH]+[R(b)[BOND]OOH]+[R(c)[BOND]OOH]+[R(d)[BOND]OOH]) for conversions ≥0.5 %, leads to a pseudo-first-order rate constant ≥5×103 s−1. It is clear that Equation (17) is by far the fastest of all competing HO2 channels. This implies that for every observed ketone molecule, one ROOH molecule has been destroyed, at least at low conversions where over-oxidation of hydroperoxide can be neglected as a source of ketone. Indeed, an equimolar amount of ketone and HO2 is produced upon the reaction of O2 and RO (Scheme 6).

Chemo-selectivity and interconversion of peroxyl radicals

Based on the mechanisms detailed above, the epoxidation efficiency (E.E.), that is, repox/(repox+rabstr), of a certain R(x)[BOND]OO peroxyl radical is given by Equation (18):

equation image((18))

Whereas either R(x)[BOND]OH or Q(x)[DOUBLE BOND]O are formed subsequent to the epoxidation step and R(x)-OOH is formed upon the allylic H abstraction, one should also take into account the amount of R(x)[BOND]OOH which has been destroyed by the HO2, co-generated with the initial ketone (Scheme 6). Assuming that the rate constant of HO2 with ROOH is independent of the precise structure of the hydroperoxide,37 one has to account for the relative abundance of the specific R(x)[BOND]OOH species. Note that Equation (18) is only valid for low conversions where the over-oxidation of R(x)[BOND]OOH can still be neglected as a source of Q(x)[DOUBLE BOND]O, meaning one has to extrapolate to zero conversion. Using this approach, we obtained the following epoxidation efficiencies: 40±10 % for R(a)[BOND]OO; 30±10 % for R(b)[BOND]OO; 10±5 % for R(c)[BOND]OO; and 40±10 % for R(d)[BOND]OO radicals. These values are in line with the computed epoxidation efficiency of tBuOO radicals (E.E.tbutylOO⋅≈7/(11.5+7)=38 %), except for the sterically hindered R(c)[BOND]OO.

So far it has been assumed that the peroxyl radicals only react with the olefin substrate and abstract allylic H atoms, or add to the C[DOUBLE BOND]C bond. However, one should also consider the interconversion of different peroxyl radicals via Equation (19):

equation image((19))

The barrier of such a thermoneutral reaction is computed to be slightly higher than for Equation (18), that is, 4.8 kcal mol−1 for the model reaction CH3OO+CH3OOH, leading to a k19(363 K)≈4×105M−1 s−1. Therefore this reaction can compete with the allylic H abstractions and C[DOUBLE BOND]C addition reactions, even at very low hydroperoxide concentrations. For normal autoxidations, this interconversion is degenerate as only one type of ROO radical is present. However, during the autoxidation of α-pinene, at least four types of peroxyl radicals are formed (verbenyl, pinenyl, pinocarvyl, and myrtenyl), all featuring slightly different reactivities. Owing to the high rate of interconversion, all peroxyl radicals will be in equilibrium with each other. So far it remains an open question whether one could affect the equilibrium distribution of the peroxyl radicals upon the addition of an appropriate catalyst. It cannot be excluded that the various hydroperoxides would have a different reactivity towards, for example, transition metal ion catalysts, and that this could lead to a modified peroxyl radical contribution and hence a different selectivity.


The thermal autoxidation chemistry of α-pinene is investigated. The addition of O2 to resonance-stabilized radicals leads to the formation of several peroxyl radicals. Of these, the verbenyl peroxyl radical is the most abundant. These peroxyl radicals can abstract allylic H atoms, yielding hydroperoxide, or they can add to the C[DOUBLE BOND]C double bond, yielding the corresponding epoxide and alkoxyl radicals. Owing to the special structure of these alkoxyl radicals they cannot only react with the α-pinene substrate to form the alcohol, but O2 is also able to abstract their weakly bonded αH atom, thereby yielding the ketone and HO2 radical. The HO2 radicals mainly react with the hydroperoxide products, converting them to peroxyl radicals. At higher conversions, the over-oxidation of the hydroperoxide product, initiated upon the abstraction of its weak αH atom, forms a small but quantifiable source of additional ketone and alcohol. Whereas the ketone product is immediately produced upon αH abstraction, the additional alcohol is only formed in an activated cage-reaction subsequent to the αH abstraction step. The efficiency of this cage-reaction was quantified and is in line with previous results on activated alkane substrates, such as ethylbenzene. Over-oxidation of the other major products (alcohol, ketone, and epoxide) does not seem to be important as can be concluded from co-oxidation experiments where small amounts of these products were initially added. The chain length was found to be larger than 50, implying that chain termination does not play an important role in the formation of products. Furthermore it was discovered that the epoxidizing efficiency of the involved peroxyl radicals markedly depends on their precise structure. Further work is in progress to investigate the role of transition metal ion catalysts on the mechanism.

Experimental Section

The experiments were performed in a 50 mL glass reactor, stirred with a Teflon-coated propeller; the vessel was connected to a large O2 reservoir kept at 1 bar. The temperature was controlled by a thermostat, equipped with an immersion heater and thermocouple (standard run at 363±2 K). Samples (±250 μL) were withdrawn from the reactor and analyzed by gas chromatography (GC) (HP6890; HP-5 column, 30 m/0.32 mm/0.25 μm; flame ionization detector). n-Nonane (1 mol %, Sigma Aldrich, >99 %) was added to the α-pinene substrate (Sigma Aldrich, 98 %, devoid of stabilizers) and used as an inert internal standard. The hydroperoxide yields were determined via a double injection, with and without reduction of the reaction mixture by trimethylphosphine (1 M in toluene, Sigma Aldrich). From the obtained augmentation in alcohol content, the corresponding hydroperoxide yield was determined. Product identification was done with GC–MS using both split injection (Tinject=250 °C) and cool-on-column injection (Tinject=50 °C) to verify the thermal stability of the products. No difference in product distribution could be observed.

Quantum chemical calculations were performed with Gaussian03 software38 at the UB3LYP/6-311++G(df,pd)//UB3LYP/6-31G(d,p) level of theory.39 Earlier, this method was validated against several benchmark levels of theory (G2 M, G3 and CBS-QB3) for H abstraction reactions by peroxyl radicals.6 The reported relative energies of the stationary points on the potential energy surfaces (PESs), (the energy barriers Eb and reaction energies ΔE) were corrected for zero point energy (ZPE) differences. Rate constants of elementary reactions were estimated by TST, in terms of the complete partition functions of the transition state (TS) and the reactant(s) and product(s) as well as their relative energy difference, Eb. For certain reactions, featuring loose TSs with hindered internal rotations, known prefactors were combined with the computed barriers. In those cases, this procedure results in a more accurate estimation of the rate constant than relying on conventional TST calculations where all internal motions were treated as harmonic oscillations to compute the pre-factor.6


The authors kindly acknowledge financial support from the Swiss National Science foundation and the ETH Zurich.