Theoretical calculations of carbon-oxygen bond dissociation enthalpies of peroxyl radicals formed in the autoxidation of lipids

Article abstract of DOI:10.1021/ja034182j

Theoretical calculations were carried out to provide a framework for understanding the free radical oxidation of unsaturated lipids. The carbon-hydrogen bond dissociation enthalpies (BDEs) of organic model compounds and oxidizable lipids (R-H) and the carbon-oxygen bond dissociation enthalpies of peroxyl radical intermediates (R-OO^.) have been calculated. The carbon-hydrogen BDEs correlate with the rate constant for propagation of free radical autoxidation, and the carbon-oxygen BDEs of peroxyl radicals correlate with rate constants for β-fragmentation of these intermediates. Oxygen addition to intermediate carbon radicals apparently occurs preferentially at centers having the highest spin density. The calculated spin distribution therefore provides guidance about the partitioning of oxygen to delocalized carbon radicals. Where the C-H BDEs are a function of the extent of conjugation in the parent lipid and the stability of the carbon radical derived therefrom, C-OO^. BDEs are also affected by hyperconjugation. This gives way to different rates of β-fragmentation of peroxyl radicals formed from oxygen addition at different sites along the same delocalized radical. We have also studied by both theory and experiment the propensity for benzylic radicals to undergo oxygen addition at their ortho and para carbons which, combined, possess an equivalent unpaired electron spin density as the benzylic position itself. We find that the intermediate peroxyl radicals in these cases have negative C-OO^. BDEs and, thus, have rate constants β-fragmentation that exceed the diffusion-controlled limit for the reaction of a carbon-centered radical with oxygen.

Full text of DOI:10.1021/ja034182j

Published on Web 04/19/2003
Theoretical Calculations of Carbon-Oxygen Bond
Dissociation Enthalpies of Peroxyl Radicals Formed in the
Autoxidation of Lipids
Derek A. Pratt, Jeremy H. Mills, and Ned A. Porter*
Contribution from the Department of Chemistry and the Center in Molecular Toxicology,
Vanderbilt UniVersity, NashVille, Tennessee 37235
Received January 15, 2003; E-mail: n.porter@vanderbilt.edu
Abstract: Theoretical calculations were carried out to provide a framework for understanding the free radical
oxidation of unsaturated lipids. The carbon-hydrogen bond dissociation enthalpies (BDEs) of organic model
compounds and oxidizable lipids (R-H) and the carbon-oxygen bond dissociation enthalpies of peroxyl
radical intermediates (R-OO•) have been calculated. The carbon-hydrogen BDEs correlate with the rate
constant for propagation of free radical autoxidation, and the carbon-oxygen BDEs of peroxyl radicals
correlate with rate constants for â-fragmentation of these intermediates. Oxygen addition to intermediate
carbon radicals apparently occurs preferentially at centers having the highest spin density. The calculated
spin distribution therefore provides guidance about the partitioning of oxygen to delocalized carbon radicals.
Where the C-H BDEs are a function of the extent of conjugation in the parent lipid and the stability of the
carbon radical derived therefrom, C-OO• BDEs are also affected by hyperconjugation. This gives way to
different rates of â-fragmentation of peroxyl radicals formed from oxygen addition at different sites along
the same delocalized radical. We have also studied by both theory and experiment the propensity for benzylic
radicals to undergo oxygen addition at their ortho and para carbons which, combined, possess an equivalent
unpaired electron spin density as the benzylic position itself. We find that the intermediate peroxyl radicals
in these cases have negative C-OO• BDEs and, thus, have rate constants for â-fragmentation that exceed
the diffusion-controlled limit for the reaction of a carbon-centered radical with oxygen.
Scheme 1. Chain Propagation Steps and Products of Linoleate
Free Radical Oxidation
Introduction
The autoxidation of polyunsaturated fatty acids and esters
has been the focus of intense investigation because of its
potential importance in biology and medicine.1,2 This process,
shown in Scheme 1, has a slow propagation step, kp, that
involves hydrogen atom transfer to a peroxyl radical and a fast
step, kO2, where oxygen adds to the intermediate carbon radical.
Oxygen addition is reversible for many radicals, and the rate
constant for the reverse reaction, kâ, is also shown in the scheme.
Product mixtures obtained from the free radical chain oxidation
of diene fatty acids or esters consist principally of diene
hydroperoxides, as shown in Scheme 1. Thus, free radical
oxidation of linoleic acid or its esters, 1, gives five major product
diene hydroperoxides, 2-6.3,4 Two of these product hydroper-
oxides, 2 and 5, are cis, trans dienes having hydroperoxide
substitution at the 13 and 9 position of the C-18 chain. These
compounds are the major products formed in linoleate oxidations
in which millimolar concentrations of phenolic antioxidants
(ArOH), such as R-tocopherol, are present during the reaction.
Two other hydroperoxides, 3 and 6, having trans, trans diene
geometry are formed in oxidations carried out in the absence
of antioxidants.5 Recently, a nonconjugated hydroperoxide, 4,
was identified as a product formed from oxidations of methyl
linoleate to which ∼0.1 M R-tocopherol was added.6
Mechanistic studies suggest that one key to understanding
product distribution in diene fatty acid oxidation is the revers-
ibility of oxygen addition to intermediate pentadienyl radicals.
Thus, as illustrated in Scheme 2, loss of oxygen from diene
peroxyl radicals competes with hydrogen atom trapping of these
peroxyls by phenolic antioxidants. The distribution of products
derived from linoleate depends therefore on the rates of
(1) (a) Marnett, L. J. Carcinogenesis 2000, 21, 361-370. (b) Sevanian, A.;
Ursini, F.. Free Radical Biol. Med. 2000, 29, 306-311. (c) Spiteller, G.;
Spiteller, D.; Jira, W.; Kiessling, U.; Dudda, A.; Weisser, M.; Hecht, S.;
Schwarz, C. Peroxide Chem. 2000, 179-208.
(2) (a) Steinburg, D.; Parhsarathy, S.; Carew, T. E.; Khoo, J. C.; Witztum, J.
L. N. Engl. J. Med. 1989, 320, 915-924. (b) Chisholm, G. M.; Steinburg,
D. Free Radical Biol. Med. 2000, 28, 1815-1826.
(3) Chan, H. W.-S.; Levett, G. Lipids 1977, 12, 99-104.
(5) Porter, N. A.; Wujek, D. G. J. Am. Chem. Soc. 1984, 106, 2626-2629.
(6) (a) Brash, A. R. Lipids 2000, 35, 947-952. (b) Tallman, K. A.; Pratt, D.
A.; Porter, N. A. J. Am. Chem. Soc. 2001, 123, 11827-11828.
(4) (a) Porter, N. A. Acc. Chem. Res. 1986, 19, 262-268. (b) Porter, N. A.;
Caldwell, S. E.; Mills, K. A. Lipids 1995, 30, 277-290.
9
10.1021/ja034182j CCC: $25.00 © 2003 American Chemical Society
J. AM. CHEM. SOC. 2003, 125, 5801-5810
5801

A R T I C L E S
Pratt et al.
Scheme 2. Reversible Oxygen Addition to Pentadienyl Radicals
RSE(Y) ) BDE(Y-CH2-H) - BDE(CH3-H) (1)
As has been noted previously,8 all methods agree well on this
simplest case.
Although the DFT models perform well for the most part,
there are some noteworthy exceptions. It has been reported
elsewhere that DFT underestimates X-H BDEs (X ) O, N)
for compounds with increasingly weak bonds (i.e., the deviation
in calculated BDEs from experimental data is larger the more
stable the incipient radical).14-17 It would appear that DFT also
underestimates C-H BDEs for compounds with increasingly
weak bonds as shown in Table 1 (cf. methane through to 1,4-
pentadiene). Improving the quality of the model improves the
agreement with experiment, especially for 1,4-pentadiene, but
the HLM result of 74.2 kcal/mol is still 2.2 kcal/mol lower than
the experimental value of 76.4 kcal/mol. Increasing the basis
set size to 6-311++G(3df,3pd) does not improve the agreement
(74.1 kcal/mol), suggesting that the basis set limit has already
been reached at 6-311+G(2d,2p). Moreover, using a large
correlation consistent basis set does not improve the result (74.4
by ROB3LYP/AUG-cc-pVTZ). The G3 compound method
predicts 75.6 kcal/mol, also low compared to experiment but
only by 0.8 kcal/mol. With only five heavy atoms in 1,4-
pentadiene, we were able to push the limits of theory and
perform a CBS-APNO18 calculation of the C-H BDE. CBS-
APNO finds 75.3 kcal/mol, only 1.1 kcal/mol higher than the
HLM result of 74.2 kcal/mol and in good agreement with the
G3 result.19
hydrogen atom transfer from ArOH to, and â-fragmentation of,
intermediate peroxyls as well as on the partitioning of oxygen
to the reactive sites on the delocalized radical intermediates.
Given the fundamental importance of the reactions of
delocalized carbon radicals such as 7 and 8 and the peroxyl
radicals formed from these species by oxygen addition to the
understanding of peroxidation mechanisms, we set out to
examine the important reactions of these radicals by theory.
Specifically, we seek answers to the following questions: (1)
What is the relative ease with which a hydrogen atom can be
removed from an unsaturated lipid? One anticipates that the
rate constants for hydrogen atom transfer should be related to
the bond dissociation enthalpy (BDE) of the relevant C-H bond.
A peroxyl radical is the partner in this reaction and peroxyl
radicals have O-H BDEs that are relatively independent of
structure,7 leaving the C-H BDE uniquely relevant to the rate
of the H atom transfer. (2) Where does the unpaired electron
spin density reside in the delocalized intermediate carbon-
centered radical, and does oxygen add preferentially to carbon
centers having more spin density? (3) Finally, following the
addition of oxygen to these radicals, how stable are the resulting
lipid peroxyl radicals toward â-fragmentation? Given the fact
that these reactions are endothermic and the reverse reaction,
oxygen addition, occurs without a substantial barrier, we believe
the rate constants for â-fragmentation should be uniquely
dependent on the C-OO• BDE. To provides answers to these
questions of interest, we report here a computational study to
determine the C-H BDEs in lipids of interest, the spin density
distributions in the incipient radicals formed from those lipids,
and the C-OO• BDEs in the lipid peroxyl radicals derived
therefrom, respectively.
While all methods give values within the error of the
experimentally determined C-H BDE of propene, G3 and
G3MP2 give values for the C-H BDE in toluene that are 2.6
and 4.6 kcal/mol, respectively, larger than experiment (88.5 (
1.5 kcal/mol). Even the DFT models give a C-H BDE 0.8-
2.1 kcal/mol higher than experiment. Indeed, while experiment
suggests that RSE(benzyl) < RSE(allyl) by only a couple tenths
(9) Curtiss, L. A.; Raghavachari, K.; Redfern, P. C.; Rassolov, V.; Pople, J.
A. J. Chem. Phys. 1998, 109, 7764.
(10) Curtiss, L. A.; Redfern, P. C.; Raghavachari, K.; Rassolov, V.; Pople, J.
A. J. Chem. Phys. 1999, 110, 4703-4709.
(11) Berkowitz, J.; Ellison, G. B.; Gutman, D. J. Phys. Chem. 1994, 98, 2744-
2765.
(12) Trenwith, A. B. J. Chem. Soc., Faraday Trans. 1 1982, 78, 3131-3136.
This C-H BDE, also measured by Clark et al. reference (Clark, K. B.;
Culshaw, P. N.; Griller, D.; Lossing, F. P.; Martinho Simoes, J. A.; Walton,
J. J. Org. Chem. 1991, 56, 5535-5539.) by photoacoustic calorimetry, was
found to be 77 kcal/mol. However, upon revision of this value to 82 kcal/
mol to correct for solvation (Laarhoven, L. J. J.; Mulder, P.; Wayner, D.
D. M. Acc. Chem. Res. 1999, 32, 342-349), the authors admit that addition
of t-BuO• to the double bond takes place under the experimental conditions,
and the two processes cannot be deconvoluted, leading to an artificially
high C-H BDE. Hence, we consider the older gas phase thermolysis value
obtained by Trenwith to be most reliable.
Results and Discussion
Choice of Methodology. To calculate the C-H BDEs in the
lipids of interest, we required a methodology that would
successfully reproduce available experimental data but also be
computationally tractable on the larger models of polyunsatu-
rated systems and of their corresponding peroxyl radicals.
Because of our past successes using model density functional
theory (DFT) calculations employing the B3LYP density
functional to predict X-H BDEs,8 we applied these models to
a short series of unsaturated hydrocarbons. The results are shown
in Table 1. Results obtained by the high accuracy compound
G39 and G3MP210 methods as well as some experimental data
are presented alongside for comparison.
(13) Pedley, J. B.; Naylor, R. D.; Kirby, S. P. Thermochemical Data of Organic
Compounds, 2nd ed.; Chapman and Hall: London, 1986.
(14) Pratt, D. A.; de Heer, M. I.; Mulder, P.; Ingold, K. U. J. Am. Chem. Soc.
2001, 123, 5518-5526.
(15) Pratt, D. A.; DiLabio, G. A.; Brigati, G.; Pedulli, G. F.; Valgimigli, L. J.
Am. Chem. Soc. 2001, 123, 4625-4626.
(16) Pratt, D. A.; DiLabio, G. A.; Valgimigli, L.; Pedulli, G. F.; Ingold, K. U.
J. Am. Chem. Soc. 2002, 124, 11085-11092.
(17) Wright, J. S.; Johnson, E. R.; DiLabio, G. A. J. Am. Chem. Soc. 2001,
123, 1173-1183.
(18) Montgomery, J. A., Jr.; Ochterski, J. W.; Petersson, G. A. J. Chem. Phys.
1994, 101, 5900. Ochterski, J. W.; Petersson, G. A.; Montgomery, J. A.,
Jr. J. Chem. Phys. 1996, 104, 2598.
We have included methane in the series to determine the
radical stabilization enthalpies (RSEs) of the substituents:
(19) It should be pointed out that the B3LYP-HLM C-H BDE for 1,3-pentadiene
is in good agreement with the experimental, G3, and G3MP2 values despite
the fact that the pentadienyl radical formed upon C-H homolysis is the
same as that for 1,4-pentadiene. This suggests that the observation that
DFT underestimates X-H BDEs (X ) O, N) for compounds with
increasingly weak bonds is not because of a problem with how the open
shell radical is handled but instead the closed-shell precursor.
(7) Luo, Y.-R. Handbook of Bond Dissociation Energies in Organic Com-
pounds; CRC Press: Boca Raton, FL, 2003; pp 166-170. Virtually all
ROO-H BDEs are roughly 88 kcal/mol.
(8) DiLabio, G. A.; Pratt, D. A.; LoFaro, A. D.; Wright, J. S. J. Phys. Chem.
A 1999, 103, 1653-1661.
9
5802 J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003

Carbon−Oxygen BDEs of Peroxyl Radicals
A R T I C L E S
Table 1. C-H BDEs in Methane, Toluene, Propene, and Pentadiene by Various Levels of Theoryi
g
method
CH sH
3
C6H CH sH
CH dCHCH sH
(CH dCH)2CHsH
5
2
2
2
2
B3LYP-LLMa
B3LYP-MLM1b
B3LYP-MLM2c
B3LYP-HLMd
G3
105.5 (0.0)
105.5 (0.0)
105.5 (0.0)
105.8 (0.0)
104.9 (0.0)
105.3 (0.0)
104.9 ( 0.1e (0.0)
89.3 (-16.2)
89.4 (-16.1)
90.1 (-15.4)
90.6 (-15.2)
91.1 (-13.8)
93.1 (-12.2)
88.5 ( 1.5e(-16.4)
86.9 (-18.6)
87.1 (-18.4)
87.9 (-17.6)
88.1 (-17.7)
87.5 (-17.4)
88.6 (-16.7)
71.9 (-33.6), 80.7 (-24.8)
72.6 (-32.9), 80.8 (-24.7)
74.1 (-31.4), 82.0 (-23.5)
74.2 (-31.6), 82.3 (-23.5)
75.6 (-29.3), 82.1 (-22.8)
77.3 (-28.0), 83.7 (-21.6)
76.4f (-28.5), 83.5 (-21.4)h
G3MP2
experiment
88.2 ( 2.1e (-16.7)
a (RO)B3LYP/6-311+G(2d,2p)//AM1/AM1. b (RO)B3LYP/6-311+G(2d,2p)//(U)MP2/6-31G(d)/(U)HF/6-31G(d). c (RO)B3LYP/6-311+G(2d,2p)//(U)B3LYP/
6-31G(d). d (RO)B3LYP/6-311+G(2d,2p)//(U)B3LYP/6-311+G(2d,2p). e Reference 11. f Reference 12. g Values in italics for 1,3-pentadiene. h Obtained from
the value for 1,4-pentadiene correcting for the difference in heats of formation between 1,4- and 1,3-pentadiene (7.1 kcal/mol); see ref 13. i RSEs are
included in parentheses. All values in kcal/mol.
Table 2. Spin Distribution in Allyl, Benzyl and Pentadienyl by
Various Levels of Theory
ment (and DFT) values suggest that the sum of the spin density
at the three ring positions is equal to the spin at the benzylic
position, but HF and MP2 find that there is substantially more
spin in the ring than at the benzylic position.
Where the B3LYP functional has been shown to underesti-
mate X-X and X-Y (X,Y ) non-hydrogen) BDEs, the B3P86
density functional has been found to be particularly useful for
these purposes,25 and thus, we have applied this to the series of
peroxyl radicals formed upon addition of oxygen to the radicals
shown in Table 2. We also include the unsubstituted methylp-
eroxyl radical for comparison. Again, we also provide G3 and
G3MP2 data for comparison. The results are shown in Table 3.
Structures of the peroxyl radicals are shown in Figure 1. Details
of alkylperoxyl radical structure provided by small basis HF/
MP2 calculations26 and DFT calculations27 have been presented
elsewhere.
a Methods: (A) UHF/6-31G(d); (B) UMP2/6-31G(d); (C) UB3LYP/6-
311+G(2d,2p); (D) UB3P86/6-311G(d,p). b Taken from ref 23. c Taken
from ref 24.
For methylperoxyl, the G3 methods give the best agreement
with experiment. B3P86-HLM overestimates the BDE by 1.4
kcal/mol, whereas B3P86-LLM and B3P86-MLM are in error
by some 4.8 and 2.5 kcal/mol, respectively. For benzylperoxyl,
SCF convergence difficulties with AM1, HF, and MP2 pre-
vented the calculation of the C-OO• BDE by both the LLM
and MLM B3P86 models as well as by G3 and G3MP2. B3P86-
HLM has no convergence problems and provides a BDE in
excellent agreement with experiment. For allylperoxyl and
pentadienylperoxyl, all methods give results that are in reason-
ably good agreement with experiment. It is interesting to note
that the difference between the HLM-calculated C-OO• BDEs
for allylperoxyl and benzylperoxyl is 2.4 kcal/mol and the
experimental value is 3.4 kcal/mol. This is consistent with the
notion (Vide supra) that the C-H BDE in toluene is indeed at
least 2 kcal/mol higher than that in propene.
of a kilocalorie, theory predicts RSE(benzyl) < RSE(allyl) by
2.2-2.5 (DFT) and 3.6-4.5 (G3) kcal/mol, respectively. This
is consistent with earlier theoretical calculations by Borden and
Hrovat.20 Some of our own kinetic measurements suggest that
this difference is ∼2 kcal/mol.21
Upon C-H bond breaking, propene, toluene, and 1,4-
pentadiene give rise to delocalized allyl, benzyl, and pentadienyl
radicals, respectively, as shown in Table 2, and we have
calculated the spin distribution in these radicals. The calculated
values are compared to the relative spin density distribution as
derived from the R-hyperfine coupling constants (R-hfcc’s)
obtained by EPR in Table 2.22
In the allyl radical, all four methods agree that the spin is
distributed evenly between C1 and C3 as symmetry requires.
In the pentadienyl radical, which also possesses C2V symmetry,
all of the theoretical methods suggest a slight bias for spin at
the central carbon atom (C3), but by varying amounts, and in
all cases, it is slightly underestimated relative to the EPR
experiment. For the benzyl radical, the calculated spin densities
by the DFT methods are in good agreement with the experi-
mental values, but the HF and MP2 calculated spin densities
differ substantially from the EPR experimental values. Experi-
When the substituent effects on the C-OO• bond in meth-
ylperoxyl are calculated, shown in parentheses in Table 3, an
initially unexpected trend results. We denote these substituent
effects as apparent radical stabilization enthalpies, RSE*:
RSE*(Y) ) BDE(Y-CH2-OO•) - BDE(CH3-OO•) (2)
(20) Hrovat, D. A.; Borden, W. T. J. Phys. Chem. 1994, 98, 10460-10464.
(21) The rate constant of â-fragmentation of the nonconjugated allylbenzylp-
eroxyl radical to the allylbenzyl radical is 1 order of magnitude slower
than the rate constant of â-fragmentation of the nonconjugated pentadi-
enylperoxyl of methyl linoleate. Given the plots in Figures 5 and 6, we
translate this to a difference in RSE of roughly 2 kcal/mol.
(22) Spin densities derived from unrestricted wave functions are reported.
Although spin densities from restricted wave functions yielded spin
distributions that follow the same trend as those from the unrestricted
calculations, they do not reproduce experimental relative spin distributions
as well. For example, the spin distribution in the benzyl radical from a
ROB3LYP/6-311+G(2d, 2p) calculation places spin at the benzylic, ortho,
and para positions in a ratio of 5.31:1:1.27.
Given that we are forming the same radicals (Table 2),
whether we break the C-H bonds shown in Table 1 or the
(23) Dust, J. M.; Arnold, D. R. J. Am. Chem. Soc. 1983, 105, 1221-1227.
(24) Davies, A. G.; Griller, D.; Ingold, K. U.; Lindsay, D. A.; Walton, J. C. J.
Chem. Soc., Perkin Trans. 2 1981, 633.
(25) DiLabio, G. A.; Pratt, D. A. J. Phys. Chem. A 2000, 104, 1938-1943.
(26) Boyd, S. L.; Boyd, R. J.; Shi, Z.; Barclay, L. R. C.; Porter, N. A. J. Am.
Chem. Soc. 1993, 115, 687-693.
(27) Kranenburg, M.; Ciriano, M. V.; Cherkasov, A.; Mulder, P. J. Phys. Chem.
A 2000, 104, 915-921.
9
J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003 5803

A R T I C L E S
Pratt et al.
Table 3. C-OO• BDEs in Methylperoxyl, Benzylperoxyl, Allylperoxyl, and 1-Pentadienylperoxyl by Various Levels of Theoryi
•
•
•
•
method
CH sOO
C6H CH sOO
CH dCHCH sOO
CH dCHCHdCHsCH sOO
3
5
2
2
2
2
2
B3P86-LLMa
B3P86-MLMb
B3P86-HLMc
G3MP2
G3
experiment
37.2 (0.0)
35.9 (0.0)
34.1 (0.0)
32.4 (0.0)
h
h
20.6 (-16.6)
21.2 (-14.7)
19.8 (-14.3)
19.7 (-12.7)
19.0 (-13.2)
15.3 (-21.9)
15.0 (-20.9)
14.0 (-20.1)
15.1 (-17.3)
13.9 (-18.3)
22.2 (-11.9)
h
h
32.2 (0.0)
32.7 ( 0.9 (0.0)d
21.8 ( 0.9 (-10.9)e
18.4 ( 0.6 (-14.3)f
13.4 ( 1.2 (-19.3)g
a
b
c
(RO)B3P86/6-311G(d,p)//AM1/AM1. (RO)B3P86/6-311G(d,p)//(U)MP2/6-31G(d)/(U)HF/6-31G(d). (RO)B3P86/6-311G(d,p)//(U)B3P86/-
6311G(d,p). d Reference 28. e Reference 29. f Reference 30. g Reference 31. h SCF failed to converge (QC as well). i RSE*s are included in parentheses (see
text). All values in kcal/mol.
Figure 2. Lipids of interest: (a) palmitoleate, (b) oleate, (c) linoleate, (d)
conjugated linoleate, (e) linoleidate, (f) linolenate, (g) arachidonate.
Table 4. Spin Distributions in Isomers of the 1,3-Dimethylallyl and
1,5-Dimethylpentadienyl Radicals by Various Levels of Theory
Figure 1. Minimum energy conformations of (a) methylperoxyl, (b)
benzylperoxyl, (c) allylperoxyl, and (d) pentadienylperoxyl. View shown
looking down the peroxyl C-O bond.
C-OO• bonds in Table 3, we would expect that the effect of
substituents on the C-H or C-OO• BDE would be the same.
We find that this is not the case. In fact, RSE* < RSE in all
cases. For example, with the HLM methodologies, (RSE -
RSE*) ) 3.3 (phenyl), 3.4 (vinyl), and 3.4 (dienyls) kcal/mol,
respectively. The corresponding experimental values are 5.5,
2.4, and 2.1, kcal/mol, respectively. This unexpected result is
due to negative hyperconjugation of the filled π-HOMO of the
benzylic/vinylic/dienylic substitutent and the empty σC-O* of
the alkylperoxyl radical:32
a Methods: (A) UHF/6-31G(d); (B) UMP2/6-31G(d); (C) UB3LYP/6-
311+G(2d,2p); (D) UB3P86/6-311G(d,p). b Derived from hyperfine cou-
pling constants taken from ref 34. c Oleate/palmitoleate model. d Linoleate/
linolenate/arachidonate model. e Conjugated linoleates model. f Linoleidate
model.
C-H BDEs in Lipids, Spin Distributions in Lipid Radi-
cals, and C-OO• BDEs in Lipid Peroxyls. It seems reasonable
to suggest that oxygen addition to delocalized radicals is related
to the spin distribution in the radical.34 We therefore considered
the spin density distribution in the delocalized radicals formed
from hydrogen atom abstraction at the activated (allylic or bis-
allylic) positions of precursor lipids. The results of these
calculations are shown in Table 4.
Entry 2 of Table 4 shows the spin density distribution in the
cis-trans allyl radical generated from removal of an oleate
allylic hydrogen. The calculations all suggest that there is greater
spin density at the cisoid end of this radical compared to the
Since both of the DFT-HLM models perform well in
predicting the relative spin density distribution in delocalized
radicals, the BDEs of the C-H bonds that are broken to form
these radicals and the BDEs of the C-OO• bonds formed upon
addition of oxygen to these radicals, we have confidence that
calculated results with these methods for lipids of biological
relevance reasonably reflect reality. The unsaturated lipids we
have studied are shown in Figure 2.
(28) Knyazev, V. D.; Slagle, I. R. J. Phys. Chem. A 1998, 102, 1770-1778.
(29) Fenter, F. F.; Noziere, B.; Caralp, P.; Lesclaux, R. Int. J. Chem. Kinet.
1994, 26, 171.
(30) Knyazev, V. D.; Slagle, I. R. J. Phys. Chem. A 1998, 102, 8932-8940.
(31) Zils, R.; Inomata, S.; Imamura, T.; Miyoshi, A.; Washida, N. J. Phys. Chem.
A 2001, 105, 1277-1282.
(33) Porter, N. A.; Mills, K. A.; Carter, R. L. J. Am. Chem. Soc. 1994, 116,
6690-6696.
(34) Bascetta, E.; Gunstone, F. D.; Walton, J. C. J. Chem. Soc., Perkin Trans.
2 1983, 603-613.
(32) Pratt, D. A.; Porter, N. A. Org. Lett. 2003, 5, 387-390.
9
5804 J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003

Carbon−Oxygen BDEs of Peroxyl Radicals
A R T I C L E S
Table 5. C-OO• BDEs in Model Peroxyl Radicals Formed in the Autoxidation of Oleates, Linoleates, and Their Isomers (C-H BDEs Are
Presented for Comparison)
a Taken from ref 33. b Taken from ref 6b. c Taken from ref 5. d Unpublished results. e Oleate/palmitoleate model. f Linoleate/linolanate/arachidonate model.
g Conjugated linoleates model. h Linoleidate model.
transoid end and the EPR hfcc’s34 are consistent with this
conclusion. The calculations and EPR both suggest that there
should be a preference for oxygen addition at the cisoid end of
the radical. In the case of oleate oxidation,33 45% of oxygen
adds at the transoid end of the intermediate cis-trans allyl
radical, whereas 55% of oxygen addition occurs at the cisoid
end of the radical. This result suggests that oxygen adds
preferentially to carbon centers in delocalized radicals that bear
the highest spin density.
The radical shown in entry 3 of Table 2 is the first-formed
carbon radical in the oxidation of linoleic acid or linoleate esters.
All of the calculations reported in the table for this radical
indicate that C4 bears the highest spin density of any of the
carbons, and EPR results support this conclusion. Oxidation of
linoleate esters in the presence of high concentrations of
R-tocopherol gives products in which 43% of oxygen addition
occurs at the C4 carbon while 27% of oxygen addition occurs
at each of the cisoid carbons at C2 and C6. Again, the results
support the notion that oxygen addition occurs preferentially at
sites of the highest spin density in a delocalized radical.
The unsymmetrical radical shown in entry 4 of Table 2 would
be formed in the oxidation of a Z,E isomer of linoleate. The
calculations all predict an even higher spin density at C4 of
this radical than is the case for the radical formed from linoleate,
which has a Z,Z alkene configuration. EPR experiments are
consistent with this result. Oxidation reactions of the Z,E isomer
have been reported but these studies were not carried out in the
presence of high concentrations of R-tocopherol, and the results
therefore do not reflect the kinetically controlled product
distribution. Indeed, no products resulting from oxygen addition
at C4 were observed, and none would be expected under the
low H atom donor conditions of oxidation used. We are currently
examining the oxidation of the Z,E and E,E linoleate isomers
under conditions where high concentrations of antioxidants are
present, and the results of these studies will be reported
elsewhere. Suffice it to say that we do indeed observe
preferential addition of oxygen at C4 of these radicals under
kinetically controlled oxidation conditions.
Theory provides guidance in predicting the sites of addition
of oxygen to delocalized radicals (Vide supra), and we anticipate
that calculations of the relevant C-H and C-OO• BDEs would
have predictive value for the rates of propagation for the slow
chain step, kp, and â-fragmentation of intermediate peroxyls,
kâ. The results of these calculations are presented in Table 5,
9
J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003 5805

A R T I C L E S
Pratt et al.
Figure 3. Minimum energy structures of the (a) oleate peroxyl radical
model, Table 5, entry 1; (b) conjugated linoleate peroxyl radical model,
Table 5, entry 8; and (c) nonconjugated linoleate peroxyl radical model,
Table 5, entry 5.
Figure 4. Logarithm of experimental rate constants for chain propagation
in the autoxidation of various hydrocarbons ((1) 1,4-cyclohexadiene; (2)
9,10-dihydroanthracene; (3) linoleate; (4) 1,4-pentadiene; (5) allylbenzene;
(6) tetralin; (7) oleate; (8) ethylbenzene; (9) cumene; (10) toluene; (11)
cyclopentane; (12) cyclohexane; (13) linolenate; (14) 1,3-cyclohexadiene;
(15) 1,4-dihydronaphthalene; (16) indene; (17) cyclohexene; (18) cyclo-
pentene; (19) indan; (20) p-xylene) plotted as a function of calculated C-H
BDEs. Chain propagation rate constants are taken from ref 35.
and minimum energy structures of representative peroxyl
radicals are shown in Figure 3.
As might be expected, there is a correlation of calculated
C-OO• BDEs and C-H BDEs derived from analogous
structures. Lipid molecules that have weak C-H bonds generally
correspond to peroxyl radicals that have weak C-OO• bonds,
since the same carbon radicals are formed upon bond breaking.
Of more practical interest is the correlation of the C-H and
C-OO• BDEs with the important reactions that they undergo
in free radical chain oxidation. Molecules that have low C-H
BDEs have high rates for hydrogen atom abstraction by a
peroxyl radical. Since this step is usually the slow step in
autoxidation, Scheme 1, it is of some interest to consider in
more detail the relationship between the calculated C-H BDEs
and experimentally determined kp’s for autoxidation. The
experimental rates of hydrogen atom transfer to peroxyl radicals
are plotted versus the DFT-calculated C-H BDEs of the
hydrocarbon precursor in Figure 4. The correlation is only
satisfactory, R2 ) 0.816, which is not surprising given the fact
that the experimental values of kp are obtained with difficulty
and have significant associated error. Most of the rate constants
used to generate Figure 4 come from the rotating sector method,
as implemented by Howard and Ingold,35 requiring that rates
of termination and initiation be determined as well as the rate
of substrate or oxygen consumption. Based upon this analysis,
the best fit of the data is described by eq 3. This equation
provides predictions of propagation rate constants for substrates
for which this value is not known. Given the difficulty associated
with the experimental determination of such rate constants, we
believe eq 3 may prove to be of some value.36
Figure 5. Logarithm of experimental rate constants for â-fragmentation
of lipid peroxyls ((1) oleate peroxyl radicals, ref 33; (2) conjugated linoleate
peroxyl radicals, ref 5; (3) nonconjugated linoleate peroxyl radical, ref 6b;
(4) cumylperoxyl radical, ref 37; (5) diphenylethylperoxyl radical, ref 38;
(6) R-vinyl benzylperoxyl radical, unpublished results; (7) nonconjugated
linoleiadate peroxyl radical, unpublished results; (8) nonconjugated cis/
trans linoleate peroxyl radical, unpublished results; (9) allylperoxyl radical,
ref 30; (10) conjugated pentadienylperoxyl, ref 31) plotted as a function of
calculated C-OO• BDEs.
cumylperoxyl, diphenylperoxyl, and pentadienylperoxyl. In
addition, we have recently measured kâ for R-vinyl benzylper-
oxyl, 6, in our laboratory. The experimental rate constants of
â-fragmentation are plotted versus the DFT-calculated C-OO•
BDEs in Figure 5. A linear correlation is obtained with an R2
) 0.923. Given that the rate constants for â-fragmentation span
7 orders of magnitude and were obtained over a period of more
than 20 years and by very different methods, we find that it is
perhaps not surprising that the correlation is not better. Based
upon this analysis, eq 4 best describes the relationship between
kâ and the calculated C-OO• BDE.39
log kp) -0.219(C-H BDE) + 18.9
(3)
Peroxyl radicals having low C-OO• BDEs also have high
rate constants for â-fragmentation, although the set of rate
constants available for comparison is rather small. In fact,
â-fragmentation rates have only been reported for peroxyl
radicals derived from oleates and linoleates, allylperoxyl,
(35) Howard, J. A. AdV. Free Radical Chem. 1972, 4, 49-173.
9
5806 J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003

Carbon−Oxygen BDEs of Peroxyl Radicals
A R T I C L E S
log kâ ) -0.46(C-OO• BDE) + 9.8
(4)
BDEs and Secondary Oxidation of Arachidonate and
Linolenate. All six of the primary oxidation products of
arachidonate and two of the four primary oxidation products of
linolenate are lipid hydroperoxides that have activated methylene
groups with very weak C-H bonds (Vide infra). Scheme 3 shows
the primary and secondary oxidation pathways of arachidonate.
To provide predictions about the propensity of primary
oxidation products shown in Scheme 3 to undergo further
oxidation and to provide predictions of the rate of fragmentation
for individual peroxyl radicals, we have calculated the C-H
and C-OO• BDEs of the structures shown in Table 6. Also
presented in Table 6 are the calculated kâ for the loss of oxygen
from the peroxyl as well as the kp calculated for H atom
abstraction from a C-H precursor by a peroxyl radical.
Of particular note is that four of the six primary hydroper-
oxide products derived from arachidonate can undergo second-
ary oxidation to give a heptatrienyl radical, while two of the
six give pentadienyl radicals upon secondary oxidation. Thus,
four of the six primary oxidation products of arachidonate and
two of four of linolenate are predicted to be more easily oxidized
than arachidonate and linolenate themselves, since removal of
a hydrogen can yield a heptatrienyl radical. When one compares
the C-H BDE 69.8 kcal/mol for entry 1, Table 6 with 72.7
kcal/mol for entry 5, Table 5, it is predicted that kp should
increase by roughly a factor of 5. Clearly, secondary oxidation
products will become a significant component of the oxidation
mixture formed from a polyunsaturated lipid if oxidation of more
than a few percent of the lipid takes place. Also of note is the
fact that â-fragmentation of peroxyl radicals shown in Table 6
generally occurs at high rates. For example, the peroxyl radical
shown in entry 1 of the table is an intermediate in one of the
secondary oxidation pathways for arachidonate, and the rate
constant for fragmentation of this peroxyl is such that it is
unlikely that appreciable quantities of this peroxyl will be
trapped, even with concentrations of 100 mM R-tocopherol
(since k ) 3.8 × 106 M-1 s-1 for R-TOH + ROO•,15 the rate
of trapping by R-TOH will be 0.1 M × [ROO•] × (3.8 × 106
M-1 s-1) ) 3.8 × 105 s-1[ROO•] compared to the rate of
â-fragmentation, which will be 7.4 × 107 s-1[ROO•]).
Figure 6. Calculated C-OO• BDEs of lipid peroxyls plotted as a function
of C-H BDEs of corresponding lipids.
dominantly at these interior positions, and that the lipid
hydroperoxides resulting from oxygen addition at these positions
will be the kinetic products of lipid autoxidation.
Hyperconjugative Stabilization of Peroxyl Radicals. In
Figure 6 is presented the calculated C-OO• BDE plotted versus
the calculated C-H BDE for the substrates shown in Tables 5
and 6 (and allylbenzylperoxyl/allylbenzene). While one antici-
pates that these BDEs should be closely related, the plot shows
that two different correlations exist for the data. The upper line
corresponds to the data for the peroxyls having only one vinylic,
conjugated dienylic or trienylic π system attached to the peroxyl
R carbon, while the lower line corresponds to peroxyls that have
two π systems (diene, triene, or arene) attached to the R center.
Thus, it appears that there is an effect related to the local
substructure of the peroxyl bearing carbon in addition to the
more general correlation with radical stabilization enthalpy and
conjugation in the parent lipid/peroxyl. If radical stabilization
and parent conjugation were the only factors operating, one
would not expect that two separate sets of data would be
observed in Figure 6.
The importance of hyperconjugative stabilization in peroxyl
radicals has been pointed out elsewhere,32,27 and we suggest
that this effect gives rise to the two correlations shown in Figure
6. The RSE for an allyl radical relative to a methyl radical is
16.7 (HLM: -17.7) kcal/mol, and one would therefore expect
a C-OO• BDE in allylperoxyl of 32.7-16.7 ) 16 (HLM:
34.1-17.7 ) 16.4) kcal/mol. Instead, the C-OO• BDE is 18.4
(HLM: 19.8) kcal/mol. Thus, the hyperconjugative interaction
between the σC-O* and the HOMO of the vinyl group is worth
approximately 2.4 (HLM: 3.4) kcal/mol. A similar analysis can
be done for comparison for the ethylperoxyl radical. The RSE
for an ethyl radical relative to a methyl radical is 3.8 (HLM:
4.8) kcal/mol, and one would expect a C-OO• BDE in
ethylperoxyl of 32.7-3.8 ) 28.9 (HLM: 34.1-4.8 ) 29.3) kcal/
mol. Instead, this BDE is known to be 35.5 (HLM: 35.1) kcal/
mol, which makes for the interaction between the combination
of σC-H and the σC-O* of 6.6 (HLM: 5.8) kcal/mol. From this
analysis, we conclude that a peroxyl bearing two π substitutents
(no alkyl substitutents) would have a lower C-OO• BDE than
that of such a radical bearing one π and one alkyl substituent,
since alkyl substituents further stabilize the peroxyl radical. This
In the heptatrienyl radical shown in Scheme 3, positive spin
density is found at four possible positions, with the inner
positions (C4 and C6 in Table 7) possessing the greatest spin
densities. This suggests that oxygen addition will occur pre-
(36) It is worth noting that the propagating radical in autoxidations of
cyclohexadienes, 9,10-dihydroanthracene, and 1,4-dihydronaphthalene was
likely hydroperoxyl and that rate constants for cyclopentane and cyclo-
hexane are for reactions with t-BuOO•. All rate constants are reported as
measured and not on a per hydrogen basis.
(37) Howard, J. A.; Bennett, J. E.; Brunton, G. Can. J. Chem. 1981, 59, 2253-
2260.
(38) Howard, J. A.; Chenier, J. H. B.; Yamada, T. Can. J. Chem. 1982, 60,
2566-2572.
(39) The kinetic approach to determine the linoleate values of kâ used in Figure
5 (entries for 2) was based upon a competition of fragmentation of the
linoleate peroxyl radicals and H-atom abstraction from 1,4-cyclohexadiene
by these radicals. The value of the rate constant for H-atom abstraction
used for comparison was kp (1,4-CHD) 270 M-1 s-1 in co-oxidations with
cyclohexadiene, whereas the values measured by Howard and Ingold for
this rate constant are closer to 1400. Taking this difference into consider-
ation, roughly a factor of 5 in the rate constant, we observe the corrected
linoleate points now give rise to an overall R2 ) 0.946 in the plot in Figure
5. We nevertheless use the “uncorrected” eq 5 as a means to make
predictions of kâ of peroxyl radicals for which the rate constant is unknown,
since the overall line of best fit does not really change: log kâ ) -0.46-
(C-OO• BDE) + 9.9.
9
J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003 5807

A R T I C L E S
Pratt et al.
Scheme 3. Arachidonate Primary and Secondary Oxidation Pathways
Table 6. C-OO• BDEs for Peroxyl Radicals Formed from Secondary Arachidonate Oxidation and C-H BDEs in Their Corresponding Lipid
Precursorsa
a Predicted based upon eq 4. b Predicted based upon eq 3.
Table 7. Spin Distribution in 1,7-Dimethylheptatrienyl by Various
Levels of Theory
both the conjugated and nonconjugated peroxyl radicals), as
shown for the comparison of the monoene, diene, and triene
systems described in Tables 5 and 6. Figure 7 shows the
calculated C-OO• BDEs in structures having extended conjuga-
tion. This leads to peroxyl radicals with nonconjugated π
systems having C-OO• BDEs of 2.4 and 1.3 kcal/mol, for the
tetraene (n)3) and pentaene (n)4) peroxyls, respectively.
Application of eq 4 suggests that the rate constants for
â-fragmentation of these peroxyls would be 5.0 × 108 and 1.6
× 109 s-1. Products formed from oxygen addition at these
centers would likely never be observed because of these
a Methods: (A) UHF/6-31G(d); (B) UMP2/6-31G(d); (C) UB3LYP/6-
311+G(2d,2p); (D) UB3P86/6-311G(d,p).
extremely rapid oxygen off-rates. For peroxyl radicals having
only conjugated π systems, the C-OO• BDE, and the rate of
â-fragmentation, levels off with extended conjugation, ap-
proaching a value of 105 s-1. We conclude that products
resulting from oxygen addition to these radicals may be observed
but only at the terminal (conjugated) positions. Such radicals
can be generated from, for example, the oxidation of caro-
tenoids.40
difference in hyperconjugation in the peroxyl C-OO• BDEs is
expected to be approximately equal to the difference between
the vinyl and methyl stabilization, 6.6-2.4 ) 4.2 (HLM: 5.8-
3.4 ) 2.4) kcal/mol, in reasonable agreement with the difference
of ∼2-2.5 kcal/mol between the data on the two lines in Figure
6.
Peroxyls Derived from Benzylic Radicals. Table 2 shows
that the spin density distribution in benzyl radicals places
Lipids with Extended Conjugation. Extending conjugation
makes the rates of propagation and â-fragmentation faster (for
9
5808 J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003

Carbon−Oxygen BDEs of Peroxyl Radicals
A R T I C L E S
Only cumyl alcohol could be detected. These reactions are
summarized in Scheme 4.
Calculation of the C-OO• BDEs for the quinoid peroxyl
radicals provides an estimate of the rate constant for â-frag-
mentation of these peroxyls. The calculated C-OO• BDEs for
the ortho and para addition products of -8.1 and -6.4 kcal/
mol, respectively, indicate that the rate constant for â-fragmen-
tation will exceed even the diffusion-controlled limit for a
bimolecular reaction, that is, kâ,ortho ) 3.4 × 1013 s-1 and kâ,para
) 5.5 × 1012 s-1 from eq 4, thus approaching the kinetic limit
for this kind of unimolecular reaction. Experiments to detect
products derived from quinoid peroxyls were bound to fail.
Summary
1. DFT can be used to calculate C-H and C-OO• BDEs to
a reasonable level of accuracy and to predict the relative spin
distribution in delocalized radicals.
Figure 7. C-OO• BDEs in lipid peroxyls as a function of double bond
2. DFT-calculated C-H BDEs correlate with experimental
rate constants for chain propagation for autoxidation. C-OO•
BDEs correlate with experimental rate constants for the
â-fragmentation of lipid peroxyl radicals.
number.
Scheme 4. Cumylperoxyl Trapping Experiments
3. C-OO• BDEs and kâ values of lipid peroxyl radicals are
determined largely by the conjugative stability of the parent
peroxyl as well as the radical stabilization of the carbon-centered
radical. Hyperconjugation may also play a role.
4. The primary oxidation products of arachidonate and
linolenate are more oxidizable than the parent lipids themselves.
The values of kâ for the peroxyl radicals formed in secondary
oxidation, which would be very difficult to elucidate experi-
mentally, can be predicted by eq 4. These predicted rate
constants may help in the mechanistic deconvolution of the
complex product mixtures that result from the autoxidation of
highly unsaturated lipids.
5. Extending conjugation increases kâ’s. These rates appear
to level off at about 105 for peroxyl radicals having only one π
system attached to the carbon bearing the peroxyl, and they
increase to 109 for peroxyl radicals having two π systems
attached to this carbon.
6. Quinoid peroxyl radicals formed from ring addition of
oxygen to benzylic radicals have negative BDEs and thus cannot
be trapped.
significant spin in the aromatic ring, and this suggests that
nonaromatic quinoid type peroxide products might form from
these radicals. A critical issue in this chemistry revolves around
the lifetime of these quinoid peroxyls. If they live long enough
to be trapped by a good H-atom donor such as R-tocopherol,
they should be found as a significant part of the product mixture.
Experimental Section
Theoretical Calculations. Geometries were optimized and vibra-
tional frequencies calculated using either B3LYP/6-311+G(2d,2p)
(C-H BDEs)8 or B3P86/6-311G(d,p) (C-OO BDEs)25 as implemented
in the Gaussian 9841 suite of programs compiled to run on an SGI
Origin. Vibrational frequencies were scaled by 0.9806, as suggested
by Scott and Radom.42 After enthalpic corrections of the energy to 298
K, the enthalpies of products and reactants of the bond homolysis (either
Several experiments were carried out with the intent of
trapping the kinetic product mixture of oxygen addition to cumyl
radicals. For these trapping experiments, we used cumylperoxyl
radicals generated photolytically from azocumene in the pres-
ence of oxygen and R-TOH. All efforts to find evidence of ring
oxygen addition to cumyl radicals failed, the only peroxide
observed in these experiments being cumyl hydroperoxide.
Quenching these reactions with triphenylphosphine, expecting
cumyl alcohol, and the two phenols 2-isopropylphenol and
4-isopropylphenol, which would form by aromatization of the
quinoid alcohols, also provided no evidence for their formation.
(41) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M.
A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgonery, J. A., Jr.; Stratmann,
R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin,
K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi,
R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.;
Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.;
Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz,
J. V.; Baboul, A. G.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.;
Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham,
M. A.; Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.;
Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.;
Gonzalez, C.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian
98, revision A.7; Gaussian, Inc., Pittsburgh, PA, 1998.
(40) Burton, G. W.; Ingold, K. U. Science 1984, 224, 569. It has been suggested
that â-carotene reacts with peroxyl radicals by addition to form a delocalized
carbon-centered radical, which does not add oxygen and carry the chain.
Although it is well-documented that addition occurs to give an intermediate
carotenoid carbon radical, the fate of this radical remains unclear.
(42) Scott, A. P.; Radom, L. J. Phys. Chem. 1996, 100, 16502.
9
J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003 5809

A R T I C L E S
Pratt et al.
R-H f R• + H• or ROO• f R• + O2) were differenced to give the
bond dissociation enthalpy (BDE) at 298 K. Where possible, the high
accuracy G3 and/or G3MP2 compound methods were employed for
comparison.
In some experiments, hydroperoxide products were immediately
reduced by triphenylphosphine before analysis by GC on an SPB-1
capillary column. The reduction was performed by adding a crystal of
triphenylphosphine to the reaction mixture after oxidation. A crystal
of BHT was also added as a stabilizer. The only oxygen-containing
products observed under any conditions of azocumene decomposition
were cumyl hydroperoxide and cumyl alcohol.
Cumylperoxyl Trapping Experiments. Azocumene was synthe-
sized using the method of Nelsen and Bartlett.43 Reactions were
conducted in 1 dram Teflon capped vials. Total reaction volumes were
100 µL, and the solvent was benzene. The concentration of azocumene
was 0.05 or 0.1 M, and R-tocopherol concentrations ranged from 0.1
to 0.75 M. Thermolysis of the azo compound was carried out at 55
and 65 °C from 8 to 30 h, and photolysis was with a 400 W medium
pressure mercury lamp in a quartz cuvette overnight. All products
(hydroperoxide, alcohol, and phenols; see Scheme 4) were commercially
available and were purchased from Aldrich Chemical Co. to be used
as authentic standards. Analysis of hydroperoxide products was by
normal phase HPLC with UV detection at 254 nm using a 0.3%
isopropyl alcohol/hexane solvent system.
Acknowledgment. We thank Keri A. Tallman for helpful
discussions and NIH HL17921, GM15431, P30 ES00267, and
CHE 9996188 for financial support. We also thank Dr. Jarrod
Smith and the Structural Biology Center at Vanderbilt for
generous access to their computational resources. D.A.P. thanks
NSERC Canada for their support.
Supporting Information Available: Cartesian coordinates and
thermochemical data for all computed structures. This material
is available free of charge via the Internet at http://pubs.acs.org.
(43) Nelsen S. F.; Bartlett, P. D. J. Am. Chem. Soc. 1966, 88, 137-143.
JA034182J
9
5810 J. AM. CHEM. SOC. VOL. 125, NO. 19, 2003