J. Phys. Chem.

A 2006, 110, 10863-10871 10863

Quantum Chemical Investigation of Low-Temperature Intramolecular Hydrogen Transfer

Reactions of Hydrocarbons

Jim Pfaendtner, Xinrui Yu, and Linda J. Broadbelt*

Department of Chemical and Biological Engineering, Northwestern UniVersity, EVanston, Illinois 60208-3120
ReceiVed: March 16, 2006; In Final Form: July 26, 2006

The B3LYP functional was evaluated as a method to calculate reaction barriers and structure-reactivity
relationships for intramolecular hydrogen transfer reactions involving peroxy radicals. Nine different basis
sets as well as five other MO/DFT and hybrid methods were used in comparing three reactions to available
experimental data. It was shown that B3LYP/6-311+G(d,p) offers a good compromise between speed and
accuracy for studies in which thermodynamic and kinetic data of many reactions are required. Sixteen reactions
were studied to develop structure-reactivity relationships to correlate the activation energy with the heat of
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

reaction. As long as no structural heterogeneities were present in the transition state ring, a simple Evans-
Polanyı́ relationship was shown to capture the activation energy as a function of heat of reaction for reactions
Downloaded via NORTHWESTERN UNIV on October 27, 2023 at 19:58:13 (UTC).

in the 1,5-hydrogen shift family. For peroxy radicals undergoing self-abstraction of a hydrogen atom in the
1,5-position, the activation energy was calculated as Ea (kcal mol-1) ) 6.3 + ∆Hrxn (kcal mol-1). For reactions
with a carbonyl group embedded in the ring of the transition state, the activation energy of peroxy radicals
undergoing self-abstraction was correlated as Ea (kcal mol-1) ) 18.1 + 0.74*∆Hrxn (kcal mol-1). The impact
of the size of the transition state ring on the activation energy and pre-exponential factor was also probed,
and it was shown that these effects can be described using simple nonlinear and linear fits, respectively.

1. Introduction reactions are driven by the growth of peroxy radicals (RO2•),

which are formed from the very fast addition of molecular
The low-temperature oxidation of hydrocarbons is of interest oxygen to primary radical species. Subsequent isomerization
to researchers in a number of different fields. Examples include of RO2• species leads to hydroperoxyalkyl radicals (•QOOH)
the oxidative degradation of lubricating oils1,2 and combustion that are able to directly yield molecular products or introduce
of fuels in novel engines.3 Despite decades of detailed experi- bifunctional moieties along the hydrocarbon backbone. Wijaya
mental and theoretical investigations, a number of fundamental et al.3 have recently performed a thorough investigation of the
questions remain unanswered about elementary step reaction fate of •QOOH in the low-temperature regime. Herein we
networks that can describe the low-temperature condensed-phase investigate the isomerization reactions that lead to the formation
oxidation of large hydrocarbons. One main obstacle in answering of •QOOH.
these questions is the accurate resolution of the kinetic Intramolecular hydrogen transfer reactions of RO2• species
parameters of the important elementary reactions. are difficult to study experimentally, and kinetic parameters must
The fundamental reaction pathways in condensed-phase often be inferred from data rather than directly measured.
oxidation are basically well established and have been reviewed Denisova and Denisov12 tabulated a number of the experimen-
extensively including recent comprehensive reviews by Emanuel tally available rate constants for unimolecular hydrogen shifts
and Gál4 and Denisov and Denisova.5 However, successful of peroxy radicals. In their review of condensed-phase oxidation,
generation of a large-scale complex reaction network requires Mill and Hendry13 give accepted experimental values for the
kinetic data for hundreds or thousands of reactions, and in the pre-exponential factor of intramolecular abstraction by peroxy
case of condensed-phase oxidation, published data can be radicals of log A (1/s) equal to 12.5, 11.5, and 11.0 for ring
contradictory or lacking altogether. Therefore, the use of kinetic sizes (transition state) of 5, 6, and 7, respectively. Van Sickle
correlations, or structure-reactivity relationships, has become and co-workers studied the 1,5-hydrogen shift (six-membered
widespread in the modeling community as a means of estimating ring formed) of 2-peroxy pentane.14 They reported a unimo-
rate coefficients of reactions in large-scale mechanisms. In the lecular rate constant of 0.87 s-1 at 373 K. The 1,5-hydrogen
case of bimolecular atom-transfer reactions, the Blowers-Masel shift of 2-peroxy-2,4-dimethylpentane was reported by Mill and
correlation has been shown to be particularly effective in Montorsi15 to have a rate constant of 18 s-1 at 373 K. Mill and
estimating the activation energy.6 The Blowers-Masel correla- Hendry13 derived a rate constant of 8 s-1 at 373 K from the
tion is an improvement upon the classic Evans-Polanyı́ data of Rust16 for the 1,6-hydrogen shift of 2-peroxy-2,5-
correlation,7 which has been shown to work well for a number dimethylhexane. Given the experimentally accepted rate con-
of different elementary reaction types in free radical chemistry.8-10 stants and pre-exponential factors, it is possible to derive
One reaction class that is of particular interest in the early activation energies of 19.7, 17.5, and 17.2 kcal mol-1 for these
phase of oxidation is intramolecular hydrogen transfer.11 These three reactions, respectively.
There have been few attempts to theoretically characterize
* Corresponding author phone: (847) 491-5351; Email: the intramolecular hydrogen shift reactions of peroxy radicals.
[email protected]. Denisova and Denisov12 estimated activation energies by
10.1021/jp061649e CCC: $33.50 © 2006 American Chemical Society
Published on Web 08/25/2006
10864 J. Phys. Chem. A, Vol. 110, No. 37, 2006 Pfaendtner et al.

applying an analogue of the Marcus equation. Their method

requires a priori knowledge of both the dissociation energies
as well as the frequencies of bonds being broken and formed
in the reaction. To accurately apply this method to a wide range
of systems, a rigorous quantum mechanical treatment would
be necessary in order to obtain these unknown quantities. An
alternative approach is to use transition state theory (TST) to
obtain rate constants applicable in the condensed phase as well
as in the high-pressure limit.
Quantum mechanical studies of this reaction class are limited
to the works by Chan and co-workers11 and Merle and
co-workers.17 Chan et al. studied the self-abstraction by peroxy
radicals for a small series of aliphatic hydrocarbons using the
BHandHLYP functional with a 6-311G(d,p) basis set. It was
recently demonstrated3 that BHandHLYP has the potential to
be a predictive tool for studying kinetics of some peroxy radical
reactions, but not without including corrections for internal
rotation and errors associated with the peroxy ligand. Merle et
al. studied the unimolecular isomerization of the n-propylperoxy
radical using B3LYP/6-31+G(d,p) and CBS-QB3.
The purpose of the present work is to study the intramolecular
hydrogen transfer of several peroxy radicals using quantum
mechanical calculations. The effect of substituents and transition
state (TS) ring size will be used to elucidate kinetic correlations
suitable for calculating rate constants in the high-pressure limit.
In the past several years, the B3LYP functional has been used
extensively in similar kinetic studies. Balderas and co-workers
have recently provided an extensive evaluation of many quantum
mechanical methods for several addition reactions.18 However,
a broader consensus as to which basis set is appropriate for
various systems is still lacking. To help guide selection of an
appropriate method, we studied the three reactions for which
experimental data is available as noted above using the B3LYP
functional with seven different basis sets as well as BHandH-
LYP/6-311G(d,p), MP2/6-311+G(d,p), G3(MP2), G3//B3LYP,
and CBS-QB3 methods. The BHandHLYP/6-311G(d,p) func-
tional and basis set were selected for comparison with previously
obtained results using this level of theory.3,11 These reactions
are pictured as reactions 1-3 in Figure 1. A larger set of
reactions (reactions 4-11 in Figure 1) was studied at the
B3LYP/6-311+G(d,p) level of theory to develop kinetic cor- Figure 1. Intramolecular hydrogen abstraction reactions involving
relations for estimating rate constants or developing reaction peroxy radicals studied using quantum chemistry. For brevity, each
reaction is given a numerical identification.
mechanisms.
type basis sets as well correlation consistent basis sets of
2. Computational and Theoretical Details
Dunning. Within the Pople formalism split-valence double and
All molecular orbital (MO) and density functional theory triple-ζ basis sets optionally including d, f, and p-type polariza-
(DFT) calculations were completed using the Gaussian 0319 tion functions on first or second row atoms as well as an optional
software package. At each level of theory and basis set, the single set of diffuse functions on first-row atoms were used.
geometries of all reactants and products were optimized to The different basis sets examined were: 6-31G(d), 6-31G(d,p),
minimum energy structures using the Berny algorithm.20 Transi- 6-311G(d,p), 6-311+G(d,p), 6-311+G(df), and 6-311+G(3df,-
tion states were identified as first-order saddle points on the 2p). For further comparison, restricted open-shell wave functions
potential energy surface. After locating a TS of a particular ring were used in two of the B3LYP cases; unless noted all
size, other similar transition states were located efficiently and calculations are done with unrestricted, open-shell wave func-
successfully using the quadratic synchronous transit (QST3)21 tions. A single calculation using a correlation consistent basis
method. All transition states were first validated by verification set was carried out employing polarized valence triple-ζ
that only one large amplitude imaginary frequency was present functions augmented with diffuse functions (B3LYP/cc-
in the proposed TS. The true reactant and product(s) corre- pVTZ+d). Additionally, for comparison with results obtained
sponding to each TS were obtained by using intrinsic reaction previously,3,11 the Becke23 half and half functional, BHandH-
coordinate following.22 LYP, was used with a single basis set, 6-311G(d,p). Structures
Transition states were determined using a number of different optimized using MP2/6-311+G(d,p) allowed further compari-
MO and DFT prescriptions. In response to its immense sons to be made. The results were additionally benchmarked
popularity in the literature, we screened Becke’s23 three using high-level extrapolation methods. Two composite methods
parameter hybrid functional, B3LYP, most extensively. Seven that have been shown to provide accurate results in radical
different basis sets were employed including Pople Gaussian- chemistry,24 CBS-QB325 and G3//B3LYP,26 were used. Calcula-
Low-Temperature Oxidation of Hydrocarbons J. Phys. Chem. A, Vol. 110, No. 37, 2006 10865

tions were also performed using G3(MP2)27 in order to probe bonds, angles, and dihedrals were analyzed using the internal
the effect of a high-level correction using MP2 optimized mode analysis provided by Gaussian 03. In large molecules,
geometries. Accepted B3LYP scaling factors of 0.9806 for zero- there is often substantial mixing of vibrational modes resulting
point energies and 0.9614 for other calculations, e.g., partition in a number of low frequencies corresponding to multiple
functions, were used.28 Scott and Radom optimized these scaling different internal rotations. Potential problems arising from this
factors specifically for the 6-31G(d) basis set, but given the mixing of vibrational modes have recently been outlined.39 Our
absence of scale factors for other basis sets, it has become present approach is to use the internal mode analysis to identify
commonplace to scale results using these values for B3LYP, a single low-frequency vibration for each internal rotation in a
regardless of the basis set.17,29 Finally, solvation effects were given molecule. Thus, the HO partition function for a molecule
studied using a polarizable continuum model (PCM).30 The with n dihedral angles was corrected by (i) removing the
solvent cavity was established using radii obtained from the contribution of n low-frequency vibrational modes and (ii)
universal force field31 with hydrogen atoms explicitly defined. including the contribution of n internal rotations.
Full geometry optimization and energy calculations were Once the partition functions for each reactant and transition
performed with n-heptane as the solvent using B3LYP/6-31G- state were corrected for internal rotation, the rate constant as a
(d). function of temperature was calculated using the standard
Using the optimized geometry and frequencies obtained from expression,
a Hessian calculation, the total microcanonical partition function
can be calculated for each molecule within the rigid-rotor kbT Qq(T)
harmonic oscillator (HO) assumption. The formulas for elec- kTST ) rpd*κ(T) exp(-Eo/RT) (2)
tronic, translational, rotational and vibrational partition functions
h Qr(T)
are well established.32 However, the HO approximation incor-
rectly treats low-frequency rotations as harmonic oscillators. If where rpd is the reaction path degeneracy, determined by the
left uncorrected, this can induce substantial error in calculated number of identical hydrogen atoms a particular radical can self-
kinetic and thermodynamic data. Internal rotations whose barrier abstract, Qq and Qr are the total partition functions for the TS
to rotation is much less than kT are characterized as free rotors and reactant, respectively, and Eo is the reaction barrier, i.e.,
and corrected easily.33 In the temperature range of interest to the zero-point corrected difference between the ground-state
the current study (300-1000 K), the barrier to rotation for all electronic energies of the TS and the reactant. The quantity κ(T)
dihedrals is on the order of kT, therefore requiring that internal is the correction for quantum tunneling. To quantify the effect
rotations be treated as hindered rotors. Our approach in treating of quantum tunneling on the reactions of interest here, we
internal rotation was similar in principle to previously reported calculated various approximations to the transmission coefficient
methods.29,34 First, the hindrance potential energy surface (PES) including the Wigner40 correction, the Eckart approximation,41
for each rotation about polyvalent single bonds was obtained the zero-curvature approximation (ZCT),41 and the small
by scanning the dihedral angle in 30° increments. The dihedral curvature tunneling (SCT)41 approximation. Calculation of the
angle was fixed and the rest of the structure was allowed to Eckart, SCT, and ZCT corrections was performed using the
relax with the geometry optimized using the B3LYP/6-31G(d) software package “Virtual Kinetic Laboratory” of Zhang and
level of theory. Although rigid scans are more computationally Truong.42 Once the temperature-dependent partition functions
efficient, significant errors were encountered due to the branch- were known, rate constants as a function of temperature were
ing of the molecules under consideration. In particular, the calculated, and the Arrhenius parameters A and Ea were obtained
peroxy radical and hydroperoxide moieties gave nonphysical from a straight-line fit of ln k vs 1/T. Unless otherwise stated,
PES results when the superstructure was kept rigid. Examples we used a temperature range of 300-1000 K for regression of
of this are provided in the Supporting Information. To preserve A and Ea. To calculate the enthalpy of reaction, all thermody-
the structure of the reactive center during relaxed PES scans of namic parameters were obtained from ensemble energy averages
transition states, all coordinates that constituted the ring of each 〈E〉 using standard formulas.32
transition state were frozen, while all other coordinates except Thermochemistry of Peroxide Species. Although the B3LYP
the dihedral angle being scanned were optimized. Once the PES functional has been repeatedly shown to give accurate results
for each rotation was obtained, the one-dimensional Schrödinger for the heat of reaction and heat of formation, especially when
equation was solved in order to obtain the energy levels for compared to other methods of similar computational cost, it has
each rotation. This step was performed using the Fourier grid been documented that B3LYP systematically gives errors in the
Hamiltonian (FGH) method developed by Marston and co- bond dissociation energy of the O-O and O-H bonds in
workers.35,36 To obtain the correct energy levels, the numerical hydroperoxide moieties.3,43 The standard approach is to use
algorithm for the FGH method was implemented using published isodesmic reactions to obtain a corrected heat of formation.44
source code.37 The energy levels from this calculation were used Although this correction for reactants and products is straight-
to calculate the partition function for each internal rotation as forward, an analogous treatment for the TS is not obvious.
a function of temperature: Therefore, the heats of reaction reported are the difference in
the quantum chemical enthalpies between the reactant and

Qir )
1
σ i
r
( )
∑gi exp -kT
i
(1)
product species, which offered a method that was consistent
with the manner in which the activation energy was calculated
for the structure/reactivity relationships that were developed.
Treatment of Rotational Isomers. It was observed that in
where σr is the internal symmetry number of the rotating top some cases the IRC-derived reactants were not the minimum
and gi is the degeneracy of the ith energy level, i.29 energy structures. Subsequent rotation of the C-O bond in the
In making the correction to the HO partition function, the peroxy radical or rotation about the backbone of the molecule
vibrational modes corresponding to internal rotation must be could yield a reactant lower in energy typically by no more
identified and their contribution to the HO partition function than 1 kcal mol-1. While these were relatively subtle differences,
removed.34,38 For each vibration, the contributions from various it was still important to account for them in the evaluation of
10866 J. Phys. Chem. A, Vol. 110, No. 37, 2006 Pfaendtner et al.

methods and development of the structure/reactivity relation- using B3LYP/6-31G(d). As shown in Table 1, the PCM model
ships. For reactants in which a lower energy structure was found, does not result in any appreciable change in the calculated
the following modified rate constant was used: kinetic properties or enthalpy of reaction. This agreement
between gas-phase TST calculations and condensed-phase
kapparent ) Krot
eq (T)*kTST(T) (3) kinetic data has also been observed for other reaction classes.46
However, it is important to note the rate of hydrogen abstraction
where Krot
eq is the equilibrium constant between two rotational
in the condensed phase has also been shown to be a strong
isomers: function of the polarity of the solvent.47
These results add to the growing database of activation
-∆G rot(T)/RT
eq ) e
Krot (4) energies predicted by B3LYP, which collectively suggests that
B3LYP does not exclusively under- or overestimate reaction
Thus, the activation energy and pre-exponential factor were barriers. Furthermore, it is not generally possible a priori to state
regressed from a fit of ln(kapparent) versus 1/T. A full derivation in which direction B3LYP will err. For example, Wijaya and
of this relationship is provided in the Supporting Information. co-workers3 report an underestimation by B3LYP of the barrier
height for different cyclization reactions while Henry and
3. Results and Discussion Radom46 report that B3LYP always overpredicts the barrier for
the cyclization of but-3-enyl-radical. The heats of reaction are
Effect of Basis Set on Results Obtained from the B3LYP
not available from experiment, but two observations can be
Functional. Results for the comparison of 16 different methods
drawn based on the results provided. First, results from very
and the 3 different reactions studied in detail (reactions 1-3 in
small basis sets deviate strongly from the CBS and G3 methods,
Figure 1) are listed in Table 1. The first 11 entries in each section
with errors as large as 4 kcal mol-1. Moving to very large basis
of the table list B3LYP calculations performed with increasingly
sets improves the agreement with the CBS and G3 results, with
larger basis sets. The one non-Pople basis set used was listed
some B3LYP heats of reaction higher than the CBS-QB3 values
as the last of these 11 for clarity. The bottom five lines in each
by only 1 kcal mol-1. Finally, it is worthwhile to note that the
section of Table 1 contain results for these reactions with
barriers calculated using B3LYP/6-311+G(3df,2p) differ from
different MO, DFT, and hybrid methods. The results show that
those calculated using single point calculations from the same
the choice of the basis set has a strong impact on both the
method but with geometries optimized using B3LYP/6-31G-
reaction barrier and the heat of reaction. As expected, increasing
(d). Using the smallest basis set (6-31G(d)) for geometry
the size of the basis set generally improves the agreement
optimization and the largest basis set (6-311+G(3df,2p)) for
between the B3LYP results and experimental data or results
single point calculations gives activation energies that may
from the high-level compound methods. With one exception in
deviate significantly from the experimental, CBS, and G3 results.
which the calculated activation energy was equal to the
experimental value, all of the B3LYP methods overestimate the It is next interesting to compare the five higher level methods
experimental activation energies. This corresponds to an un- in the bottom of each section of Table 1 with each other. Barriers
derprediction of the reaction rate coefficient. One possible calculated at the MP2 and G3(MP2) levels of theory are in very
explanation for this could be the type of calculation used to poor agreement with the experimental data for all three reactions.
estimate the contribution of quantum tunneling, but our results To test if the cause of this discrepancy was from the optimized
show that the type of tunneling correction used did not change geometries or from the single point/hybrid energies, we
the calculated activation energies significantly. We calculated calculated the G3//B3LYP energy of the reactant and TS of the
the Eckart, ZCT, and SCT transmission coefficients for reactions MP2/6-311+G(d,p) optimized structures for reaction 2. The
1 and 2 using the B3LYP/6-31G(d) and B3LYP/6-311+G(d,p) calculated activation energy was nearly identical to the full G3//
basis sets. The Wigner correction was a good approximation B3LYP activation energy, thus indicating that the energies
of the SCT transmission coefficient (within a factor of 2 or less calculated using MP2 are poor. The BHandHLYP/6-311G(d,p)
at the temperature range of interest (373 K)). Accordingly, we method also substantially overpredicts the activation barrier
calculated all kinetic data in Table 1 using the Wigner correction. compared to the CBS-QB3 method. Despite their poor predic-
Furthermore, the transmission coefficient had a maximum value tions of the barrier height, BHandHLYP/6-311G(d,p), MP2/6-
of 3.7 at 300 K for the reactions we studied, which is much 311+G(d,p), and G3(MP2) still give pre-exponential values that
lower than that observed for intermolecular hydrogen transfer are in good agreement with the value obtained from CBS-QB3,
reactions of small molecules studied by Truong and co-workers45 and the heats of reaction are in reasonable agreement with the
and Gonzalez-Lafont and co-workers.41 The small contribution B3LYP values using large basis sets. In comparing the two
of tunneling is consistent with the adiabatic ground state hybrid methods, the results reveal that in the case of the first
potential clearly showing a Very late transition state. The and second reactions, CBS-QB3 gives results that are much
potential is nearly flat as the reaction goes from the transition closer to the experimentally reported values, whereas the
state to the product hydroperoxide radical (forward direction experimental values for the third reaction more closely match
as depicted in Figure 1). Because tunneling is not observed to the G3B3 results. However, each activation energy was obtained
have a strong impact in these reactions, it is not surprising that from a single source without any repeatability reported.
for each reaction in Table 1, a plot of ln k versus 1/T gave a Therefore, the comparison set is too limited to draw any
perfect straight-line fit, thereby validating our method of definitive conclusion about the superiority of either of these
obtaining the Arrhenius parameters A and Ea. methods. Additionally, in all cases, the calculated rate coef-
Because we are seeking kinetic correlations for use in ficients (kTST(373 K)) from the CBS-QB3 method give the best
condensed-phase hydrocarbon oxidation chemistry, we tested agreement to the available experimental data. For calculating
the sensitivity of the calculated data to the presence of a nonpolar rate coefficients, the G3//B3LYP method performs noticeably
solvent, i.e., a typical oxidation substrate of interest. We used better than pure B3LYP methods for reaction three, but at a
the PCM model to simulate a solvent cavity of n-heptane and comparable (or lower) level of accuracy for reactions 1 and 2.
reoptimized the reactants, TS, and products for reactions 1-3 For example, in the data set for reaction 1, B3LYP/cc-pVTZ+d
Low-Temperature Oxidation of Hydrocarbons J. Phys. Chem. A, Vol. 110, No. 37, 2006 10867

TABLE 1: Comparison of Basis Set and Method for Three Intramolecular Hydrogen Abstraction (Reactions 1-3 in Figure 1)
reactants transition state
methoda Eob ∆Hrxn c Ead log Ae kTST(373 K)f νirg 〈S2〉h νir g νii 〈S2〉h
reaction 1j C(CH3)2(O2 •)CH
2CH(CH3)2 f C(CH3)2(O2H)CH2C‚(CH3)2
6-31G(d) 19.8 17.8 21.1 10.8 2.73 × 10-2 47,75,110,181,207,229,242 0.753 193,201,205,225 -1626 0.756
6-31G(d)k 19.4 17.2 20.6 10.7 4.26 × 10-2
RO 6-31G(d)l 20.0 17.7 21.2 10.8 2.38 × 10-2 47,75,111,181,207,229,242 194,201,205,225 -1695
6-31G(d,p) 18.0 15.5 19.2 10.8 3.54 × 10-1 48,76,111,180,209,219,243 0.753 195,200,206,226 -1610 0.756
6-311G(d,p) 17.7 14.3 19.0 10.7 3.69 × 10-1 47,76,111,178,202,215,243 0.754 194,197,206,227 -1653 0.756
6-311+G(d,p) 17.7 13.7 18.7 10.6 4.39 × 10-1 48,75,103,178,208,228,242 0.754 193,200,205,227 -1654 0.756
RO 6-311+G(d,p)l 17.8 13.6 18.8 10.6 3.83 × 10-1 48,75,104,178,208,228,242 194,200,205,227 -1713
6-311+G(df) 20.2 16.7 21.3 10.7 1.65 × 10-2 55,77,99,178,213,220,245 0.754 194,203,206,228 -1694 0.757
6-311+G(3df,2p) 17.4 13.4 18.5 10.7 7.24 × 10-1 47,72,101,177,208,228,242 0.755 192,199,204,226 -1675 0.757
6-311+G(3df,2p)m 16.6 13.4 19.1 10.8 4.05 × 10-1 47,75,110,181,207,229,242 0.753 193,201,205,225 -1626 0.753
cc-pVTZ+d 17.4 13.6 18.5 10.7 7.24 × 10-1 44,74,106,178,204,227,240 0.754 194,196,204,226 -1666 0.756
MP2/6-311+G(d,p)e 34.1 17.7 35.7 10.6 4.79 × 10-11 60,66,93,180,221,232,259 0.763 208,215,228,252 -2033 0.794
BHandHLYP/6-311G(d,p) 23.4 14.7 24.7 10.7 1.68 × 10-4 51,76,119,187,217,234,256 0.757 206,210,219,242 -2133 0.765
G3(MP2) 43.7 13.7 44.8 10.9 4.45 × 10-16 48,71,124,193,241,252,267 0.761 262,302,315,336 -2777 0.797
G3\\B3LYP 17.9 12.9 19.2 10.8 3.54 × 10-1 47,75,110,181,207,229,242 0.753 193,201,205,225 -1626 0.756
CBS-QB3 16.3 12.3 17.5 10.6 2.22 44,79,111,178,200,219,244 0.754 185,202,215,238 -1653 0.756
experimental 17.5 11.5 18
reaction 2 CH3C(O2•)H(CH2)2CH3 f CH3C(O2H)HCH2C‚HCH3
6-31G(d) 24.0 19.6 23.1 10.9 2.31 × 10-3 71,79,108,169,206 0.753 193,210 -1628 0.757
6-31G(d)k 23.5 19.1 22.5 10.8 4.13 × 10-3
RO 6-31G(d)l 24.9 20.2 23.9 10.8 6.24 × 10-4 78,86,107,176,241 187,208 -1711
6-31G(d,p) 22.7 17.9 21.7 10.8 1.21 × 10-2 73,84,106,176,234 0.753 189,206 -1637 0.756
6-311G(d,p) 22.5 16.8 21.5 10.8 1.59 × 10-2 77,82,104,176,232 0.753 189,197 -1682 0.756
6-311+G(d,p) 22.3 16.3 21.3 10.8 2.08 × 10-2 71,79,108,176,230 0.754 190,203 -1690 0.757
RO 6-311+G(d,p)l 22.5 16.0 21.5 10.8 1.59 × 10-2 71,81,108,176,230 191,203 -1753
6-311+G(df) 25.1 19.2 24.1 10.8 4.76 × 10-4 68,78,104,176,231 0.754 191,206 -1726 0.757
6-311+G(3df,2p) 22.1 15.8 21.1 10.8 2.73 × 10-2 72,80,109,176,227 0.755 187,202 -1716 0.757
6-311+G(3df,2p)m 20.6 19.7 10.9 2.27 × 10-1 71,79,108,169,206 0.753 193,210 -1628 0.757
cc-pVTZ+d 21.3 15.3 20.4 10.8 7.02 × 10-2 39,76,106,167,200 0.754 188,201 -1703 0.757
MP2/6-311+G(d,p)n 40.6 19.3 39.4 10.7 4.10 × 10-13 67,73,109,175,242 0.763 213,215 -2098 0.796
BHandHLYP/6-311G(d,p) 27.6 16.0 26.3 10.8 2.45 × 10-5 54,80,112,171,214 0.756 202,211 -2190 0.765
G3(MP2) 40.9 15.3 38.9 9.9 1.28 × 10-13 73,87,117,179,237 0.761 124,163 -2979 0.800
G3\\B3LYP 21.7 14.7 20.8 10.9 5.15 × 10-2 71,79,108,169,206 0.753 193,210 -1628 0.757
CBS-QB3 20.2 14.4 19.3 10.9 3.90 × 10-1 47,77,107,167,202 0.754 189,197 -1682 0.756
experimental 19.7 11.5 8.7 × 10-1
reaction 3j C(CH3)2(O2•)(CH2)2CH(CH3)2 f C(CH3)2(O2H)(CH2)2C‚(CH3)2
6-31G(d) 20.0 16.6 19.9 9.4 5.48 × 10-3 49,79,99,139,198,207,230,230 0.753 197,204,219,230 -1628 0.757
6-31G(d)k 19.7 16.2 19.6 9.4 8.22 × 10-3
RO 6-31G(d)l 20.3 16.4 20.1 9.4 4.19 × 10-3 49,79,99,140,198,207,230,231 198,204,220,231 -1758
6-31G(d,p) 18.3 14.3 18.0 9.4 7.12 × 10-2 50,80,100,141,197,209,230,231 0.753 199,205,219,231 -1654 0.756
6-311G(d,p) 17.9 13.1 17.7 9.3 8.48 × 10-2 50,81,100,143,193,207,225,229 0.754 197,206,216,232 -1695 0.757
6-311+G(d,p) 17.8 13.1 18.0 9.3 5.66 × 10-2 49,76,95,138,198,209,226,228 0.754 196,207,218,232 -1702 0.757
RO 6-311+G(d,p)l 17.9 12.8 18.1 9.3 4.94 × 10-2 49,76,95,138,199,208,226,229 196,207,218,232 -1773
6-311+G(df) 20.2 16.0 20.4 9.3 2.22 × 10-3 51,72,96,140,201,211,227,232 0.754 196,207,220,232 -1740 0.757
6-311+G(3df,2p) 17.7 12.8 18.0 9.3 5.66 × 10-2 46,72,92,133,198,206,226,228 0.755 194,205,216,231 -1730 0.757
6-311+G(3df,2p)m 16.7 19.3 9.4 1.23 × 10-2 49,79,99,139,198,207,230,230 0.753 197,204,219,230 -1675 0.757
cc-pVTZ+d 17.8 12.9 18.0 9.3 5.66 × 10-2 46,75,95,137,196,206,226,227 0.754 195,204,215,231 -1721 0.757
MP2/6-311+G(d,p)n 34.2 18.4 34.9 9.3 7.07 × 10-12 60,86,108,143,203,223,228,242 0.763 213,223,227,238 -2164 0.794
BHandHLYP 23.4 13.7 23.0 9.2 5.28 × 10-5 50,85,106,149,206,218,233,241 0.757 210,220,223,239 -2214 0.765
G3(MP2) 47.0 12.0 44.7 7.5 2.03 × 10-19 43,79,103,153,217,239,246,257 0.761 83,95,198,254 -2875 0.797
G3\\B3LYP 17.9 10.6 16.7 9.4 4.11 × 10-1 49,79,99,139,198,207,230,230 0.753 197,204,219,230 -1675 0.757
CBS-QB3 16.5 10.6 15.9 9.3 9.62 × 10-1 50,81,100,144,194,208,226,229 0.754 197,206,216,232 -1695 0.757
experimental 17.2 11 8
a Unless alternate MO or hybrid method is given, the B3LYP functional is used and only the basis set is provided. b Reaction barrier (in kcal/

mol). c Heat of reaction (in kcal/mol) at 298 K. d Arrhenius activation energy (in kcal/mol). e Pre-exponential factor (in 1/s). f Rate coefficient (in
1/s) at 373 K. g Frequencies (in cm-1, unscaled) which were identified as hindered rotors. h Expectation value of the total spin for the reactants and
the transition states. i Imaginary vibrational mode (in cm-1) of the transition state. j Lower energy rotational isomer of IRC-derived reactant or
product was found, and eq 3 was used to calculate the pre-exponential factor and activation energy for the forward reaction. k Geometry optimization
and energy calculation performed using UB3LYP/6-31G(d) and a PCM model using the properties of n-heptane to describe the solvent cavity.
l Calculation performed with restricted-open shell wavefunctions. m Geometry optimization performed at UB3LYP/6-31G(d) with single point and

frequency calculation at given basis set. n MP2 scaling factors of 0.9670 (ZPE) and 0.9434 (fundamental frequencies) were used according to the
recommendation of Scott and Radom.28

gives a rate coefficient of 7.24 × 10-1 (1/s), whereas the G3//B3LYP for the reactants of reactions 2 and 3 in Figure 1.
prediction of G3//B3LYP is 3.54 × 10-1 (1/s). Both calculations The largest basis sets that most closely approximate the hybrid
are far from the single experimental value of 18 (1/s), but the methods cost slightly more than the CBS-QB3 calculation,
cc-pVTZ+d value is a factor of 2 closer. whereas G3//B3LYP is noticeably more expensive than the other
Given the variability of the results from all of the different methods. The high computational cost of G3//B3LYP is due
methods and basis sets, it is instructive to compare their almost entirely to the single point calculation using the MP4/
computational cost. Figure 2 gives the relative computational 6-31G(2df,p) level of theory. On the basis of the relatively good
cost for B3LYP using the 6-31G(d), 6-311+G(d,p), 6-311+G- performance of the B3LYP/6-311+G(d,p) calculations on all
(3df,2p), and cc-pVTZ+d basis sets, as well as CBS-QB3 and three reactions and its modest computational cost, we selected
10868 J. Phys. Chem. A, Vol. 110, No. 37, 2006 Pfaendtner et al.

pre-exponential factor to decrease with increasing ring size. The

proportional decrease in the pre-exponential factor with increas-
ing ring size is clearly seen in the plot of the pre-exponential
factor as a function of ring size in Figure 3. Although the
correlation is empirical, it can be explained based on changes
in internal rotations, and the straight line behavior of the pre-
exponential factor as a function of ring size may offer a simple
template for the development of structure-reactivity relation-
ships for systems in which intramolecular hydrogen transfer
plays an important role.
A Simple Structure-Reactivity Relationship to Describe
1,5- Intramolecular Hydrogen Transfer. For reactions in the
1,5-hydrogen shift family, structural changes were introduced
via substitution of various groups and atoms along the backbone
to create diversity in thermodynamic and kinetic parameters of
the reaction as delineated by reactions 1-2 and 4-9 in Figure
1. Starting with a very small substrate (reaction 6), we
systematically introduced additional substitutions both near the
hydrogen undergoing abstraction as well as the peroxy radical
Figure 2. Comparison of computational cost for the B3LYP functional
with several basis sets and two hybrid methods. Two reactants are moiety. This approach ensured that a broad range of heats of
compared: CH3C(O2•)HC3H7 (white bars) and C(CH3)2(O2•)(CH2)2CH- reaction was explored, thus creating a more general structure-
(CH3)2 (striped bars). All times are relative to the B3LYP/6-31G(d) reactivity relationship. It is interesting to note that even though
calculation of CH3C(O2•)HC3H7. almost all degrees of substitution were explored, there were no
forward reactions with a heat of reaction less than 8.5 kcal
this basis set to study other substituent and structural effects in mol-1. In addition, we investigated the effect of introducing
this reaction family. Our performance criterion was based on structural changes to the TS ring by adding a carbonyl linkage
the ability of a given method to accurately calculate the to reactions 6, 2, and 5 to create carbonyl-containing reactants
activation energy. This is based on an approach for estimating in reactions 7-9, respectively. Thermodynamic and kinetic data
reaction rate coefficients that is often applied to large reaction for 16 reactions (forward and reverse pairs) considered are
mechanisms that consist of thousands of reactions in which a summarized in Table 3, and the heats of reaction and activation
representative frequency factor is assumed for all reactions in energy data in Table 3 are plotted in Figure 4. Also included in
a given reaction family and the activation energy for individual Table 3 are data for reaction 8 calculated using CBS-QB3.
reactions is calculated using structure-reactivity relationships
For the forward reactions, the first five reactions listed in
based on thermodynamic properties.8,48,49 Although CBS-QB3
Table 3 fall on an Evans-Polanyı́7 correlation, i.e., Ea ) Eo +
would be the best choice based on agreement of the calculated
R∆Hrxn. The parameters regressed from these data are Eo )
rate coefficients with experimental data, a full exploration of
substrate size and substituent effects is not feasible using such 6.3 kcal mol-1 and R ) 1. The regressed transfer coefficient
an expensive hybrid method due to the large number of reactions indicates that this reaction family is more sensitive to the
proposed for study in Figure 1 as well as the poor scaling of influence of substituents than intermolecular hydrogen transfer,
hybrid methods. Therefore, we selected B3LYP/6-311+G(d,p) which is often approximated with a transfer coefficient of 0.8.6
as a good compromise between cost and accuracy for calculating In this case, the approximated activation energy is the heat of
activation energies. Combining the calculated activation energies reaction added to the intrinsic reaction barrier. The best fit for
with a representative frequency factor based on experimental the activation energy of the reverse reactions as a function of
data would yield values of rate coefficients in very good heat of reaction is a straight line with a slope equal to zero and
agreement with experiment. an intercept equal to the intrinsic reaction barrier for the forward
Effect of Ring Size. To probe the effect on the rate coefficient reactions of 6.3 kcal mol-1. This enthalpic consistency is
for this class of reactions of the relative position from which clarified by examining the relationship Ea ) ∆Hq + nRT from
the hydrogen was abstracted, we varied the ring size of the transition state theory, where n is the molecularity of the
transition state from four to seven. The four-membered ring reaction. When manipulated algebraically, this equation reveals
corresponds to a 1,3-hydrogen shift and the seven-membered that for forward and reverse reaction pairs of reactions in the
ring results from a 1,6-hydrogen shift. The degree of substitution intramolecular hydrogen transfer reaction family, the sum of
of the carbon atoms attached to the reacting hydrogen and to the transfer coefficients is equal to 1, and the intrinsic barriers
the OO• moiety was the same for all four reactions so that rate are equal, as observed here. In the practice of creating detailed
constants could be compared fairly. The four reactions examined kinetic models, overall thermodynamic consistency would be
are reactions 1, 3, 10, and 11 in Figure 1. The activation energies enforced by using only one of the two Evans-Polanyı́ relation-
and pre-exponential factors are summarized in Table 2. The ships to calculate the rate coefficient in one direction while the
relationship between ring size and activation energy is a other would be calculated based on a known equilibrium
decreasing, nonlinear function as seen in Figure 3. The decrease constant. However, the two distinct Evans-Polanyı́ relationships
in the activation energy is particularly marked beyond a ring are clearly enthalpically consistent within the approximation of
size of six as the ring strain becomes minimized. As the ring Ea ) ∆Hq + nRT, and thus, each relationship may be used
size of the TS is increased, an additional internal rotation is independently to estimate the activation energy depending on
lost or frozen within the TS ring structure. Thus, the partition the specific type of reaction of interest. A transfer coefficient
function for the reactant increases more than the partition of one is also known to correspond to a very late transition
function for the TS as the ring size is increased, causing the state that is consistent with the IRCs calculated.
Low-Temperature Oxidation of Hydrocarbons J. Phys. Chem. A, Vol. 110, No. 37, 2006 10869

TABLE 2: Comparison of Activation Energy and Pre-Exponential Factor as the Ring Size of the Transition State Is Varied
from 4 to 7a
reaction TS ring size Eab log A (1/s)
CH3CH(O2•)CH3 f CH3COCH3 + HO• c 4 41.6 12.6
(CH3)2C(O2•)CH(CH3)2 f (CH3)2C(O2H)C‚(CH3)2 5 28.0 11.5
C(CH3)2(O2•)CH2CH(CH3)2 f C(CH3)2(O2H)CH2C‚(CH3)2 6 18.7 10.6
C(CH3)2(O2•)(CH2)2CH(CH3)2 f C(CH3)2(O2H)(CH2)2C‚(CH3)2 7 18.0 9.3
a
Geometry optimization and energy calculations were performed using B3LYP/6-311+G(d,p). b Quantity in kcal mol-1. c The 1,3-hydrogen
shift creates the unstable •COOH containing radical, which directly yields the β-scission products shown in the table.

transfer reactions. We varied Eo to fit our data to the Blowers-

Masel relationship, but the fit was very poor due to the inability
of the Blowers-Masel relationship to asymptotically approach
a nonzero value for strongly exothermic reactions. A modified
Blowers-Masel relationship in which a constant was added to
the equation for each regime in eq 5 fit the data well with a
value of Eo ) 3.7 kcal mol-1 and a constant factor of 6.2 kcal
mol-1. This suggests that the intrinsic barrier for strongly
exothermic reactions approaches a value of 6.2 kcal mol-1
instead of zero for this class of reactions. However, because of
the lack of data in the endo- to exothermic transition region, it
is not possible to estimate whether the Blowers-Masel relation-
ship is superior to the simple Evans-Polanyı́ relationship.
Reactions 7-9 are also plotted in Figure 4 and clearly fall
on a different correlation than the corresponding reactants
without carbonyl groups. The calculation for reaction 8 was
repeated using CBS-QB3 to verify that higher-level methods
also capture this difference. The data in Table 3 show that
B3LYP/6-311+G(d,p) and CBS-QB3 are in very good agree-
ment. Linear regression shows that the data for reactants
containing a carbonyl linkage in the TS ring have a different
Evans-Polanyı́ relationship. The intrinsic barrier, Eo, is 18.1
kcal mol-1 and the endothermic/exothermic transfer coefficients,
R, are 0.74 and 0.26, respectively. Although these transfer
coefficients are much more similar to typical hydrogen transfer
values, the intrinsic barrier is much larger than typical hydrogen
transfer values of 7-10 kcal mol-1. One possible explanation
for this is increased ring strain due to the carbonyl linkage. Also,
in the product structure, the carbonyl group offers significant
stabilization to the radical center on the adjacent carbon atom,
resulting in a heat of reaction that is much lower than the alkyl
counterpart. We also explored the effect of tunneling on this
reaction class given the distinct structure-reactivity relationship
and the markedly different heats of reaction compared to the
analogous reactions without the intervening carbonyl group. For
all three reactions, we calculated the SCT, ZCT and Eckart
transmission coefficients as a function of temperature. The
Eckart model was observed to significantly overestimate the
SCT values in the low temperature region. Conversely, the ZCT
Figure 3. Variation of Arrhenius activation energy and pre-exponential and Wigner transmission values were noticeably lower than the
factor with transition state ring size. A regression line is given to show SCT values. Using eq 2 with the transmission coefficient based
the variation of each parameter as the ring size is increased. on the SCT method, we recalculated the activation energies
using rate constants over a small temperature range (300-600
Blowers and Masel6 have proposed a modified form of the
K). The SCT/TST activation energies were lower than the
Marcus equation to estimate the activation energy of hydrogen
corresponding Wigner/TST activation energies with the devia-
transfer reactions for endo- and exothermic reactions using a
single formulation as shown in eq 5. tion between the two methods (EWigner a - ESCTa ) ranging
-1

{
between 1.6 and 2.8 kcal mol ; however, the same form of
for ∆Hrxn/4Eo < -1 structure-reactivity relationship was observed, with Eo and R
0
values of 15.6 kcal mol -1 and 0.83/0.17.
Ea ) Eo(1 + ∆Hrxn/4Eo) for -1 e ∆Hrxn/4Eo e 1 (5)
2

∆Hrxn for ∆Hrxn/4Eo > 1 4. Conclusions

Their proposed relationship is second order in the heat of The B3LYP functional was evaluated as a method to calculate
reaction and nicely captures the curved transition region between reaction barriers and structure-reactivity relationships for
exothermic and endothermic reactions in bimolecular hydrogen intramolecular hydrogen transfer reactions involving peroxy
10870 J. Phys. Chem. A, Vol. 110, No. 37, 2006 Pfaendtner et al.

TABLE 3: Data Used in Development of Structure-Reactivity Relationship for 1,5-Hydrogen Shift Familya
forward reaction reverse reaction
number reaction ∆Hrxn b Eab log A (1/s) Eab log A (1/s)
1 C(CH3)2(O2•)CH2CH(CH3)2 f C(CH3)2(O2H)CH2C‚(CH3)2 c 13.7 18.7 10.6 5.1 9.7
2 CH3C(O2•)H(CH2)2CH3 f CH3C(O2H)HCH2C‚HCH3 16.3 21.3 10.8 5.8 10.7
4 (CH3)2CHCH2C(O2•)(CH3)CH2CH(CH3)2 f 8.6 16.8 11.3 6.1 11.1
(CH3)2C‚CH2C(O2H)(CH3)CH2CH(CH3)2
5 (CH3)CH(O2•)CH2CH(CH3)2 f (CH3)CH(O2H)CH2C‚(CH3)2 c 14.9 21.6 10.6 8.0 10.9
6 (CH3)CH(O2•)CH2CH3 f (CH3)CH(O2H)CH2C‚H2c 19.6 26.2 11.2 7.1 11.6
7 (CH3)CH(O2•)COCH3 f (CH3)CH(O2H)COC‚H2 11.6 26.3 11.1 15.3 12.5
8 (CH3CH(O2•)HCOCH2CH3 f (CH3CH(O2H)HCOCH‚CH3 4.0 21.3 11.6 17.6 12.0
8 (CH3CH(O2•)HCOCH2CH3 f (CH3CH(O2H)HCOCH‚CH3 d,e 2.7 20.8 11.7 17.8 11.0
9 (CH3)CH(O2•)COCH(CH3)2 f (CH3)CH(O2H)COC‚(CH3)2 1.5 19.1 10.4 17.4 10.1
a
When applicable, lower energy rotational isomers were used, according to eq 3. Unless noted, all calculations were performed using B3LYP/
6-311+G(d,p) b Quantity in kcal mol-1. c Lower energy rotational isomer of IRC-derived reactant was found and eq 3 was used to calculate the
pre-exponential factor and activation energy for the forward reaction. d Value not used in regression of structure/reactivity relationship. e Reaction
thermodynamics and kinetics calculated using CBS-QB3.

shift family. A pre-exponential factor of log A (s-1) ) 10.9

and Ea ) 6.3 + ∆Hrxn, where Ea and ∆Hrxn are in kcal mol-1,
are recommended if there are no carbonyl groups in the
transition state ring structure. Rate coefficients for reverse
reactions in this family can be approximated using a pre-
exponential factor of 10.8 (log A (s-1)) and an activation energy
of 6.3 kcal mol-1. If a carbonyl group within the ring structure
was present, an Evans-Polanyı́ relationship Ea ) 18.1 +
0.74∆Hrxn (R ) 0.26 for exothermic reactions) was shown to
capture the data well. The vast difference in the parameters of
the two distinct correlations demonstrates that different structure-
reactivity relationships are warranted. On the basis of the results
obtained, B3LYP/6-311+G(d,p) is an attractive choice for
obtaining insights into structure-reactivity relationships.

Acknowledgment. This work was partially supported by the

NSF IGERT program (Award Number DGE-0114429) and the
National Center for Supercomputing Applications under TG-
CTS050021 utilizing the Pittsburgh Supercomputing Center.
Support from the Inter-American Materials Collaboration Pro-
Figure 4. Plot of activation energy versus heat of reaction for all
reactions studied in the 1,5-hydrogen shift family. All of the B3LYP/
gram of the National Science Foundation (DMR-0303435) is
6-311+G(d,p) data from Table 3 are included in the figure, but two also acknowledged. We are grateful for helpful discussions with
different regressions were performed to reflect the different relationships Peter Vansteenkiste and Dr. Veronique Van Speybroeck and
observed. The open and closed circles indicate all data that are forward/ also acknowledge the use of their code in calculating reduced
reverse reaction pairs of the alkylperoxy radicals (reactions 1-2 and moments of inertia and fitting rotational hindrance potentials.
4-6 in Figure 1). The open and closed diamonds indicate the data that
contain carbonyl moieties within the TS ring structure (reactions 7-9 Supporting Information Available: Comparison of rigid
in Figure 1). The recommended structure-reactivity relationships are
given within the figure with units of kcal mol-1. and relaxed PES scans for the C-O bond of a peroxy radical
and a derivation of the apparent rate constant for combined
radicals. Nine different basis sets as well as five other MO/ rotational isomers and hydrogen transfer. This material is
DFT and hybrid methods were used in comparing three reactions available free of charge via the Internet at http://pubs.acs.org.
to available experimental data. It was shown that B3LYP/6-
311+G(d,p) offers a good compromise between speed and References and Notes
accuracy for studies in which thermodynamic and kinetic data (1) Blaine, S.; Savage, P. E. Ind. Eng. Chem. Res. 1992, 31, 69-75.
of many reactions are required. The BHandHLYP functional (2) Jensen, R.; Korcek, S.; Mahoney, L.; Zinbo, M. J. Am. Chem. Soc.
and methods based on MP2 geometries gave very poor 1981, 103, 1742-1749.
predictions of experimental activation energies, and their (3) Wijaya, C.; Sumathi, R.; Green, W. H., Jr. J. Phys. Chem. A 2003,
107 (24), 4908-4920.
reaction barriers disagreed substantially from those obtained with (4) Emanuel, N.; Gál, D. Modelling of Oxidation Processes. Proto-
CBS-QB3 and G3//B3LYP, whose applicability has been type: The Oxidation of Ethylbenzene; Akadémiai Kiadoó: Budapest, 1986.
extensively tested. (5) Denisov, E. T.; Denisova, T. G. Handbook of Antioxidants, Bond
We probed the effect of varying the transition state ring size Dissociation Energies, Rate Constants, ActiVation Energies and Enthalpies
of Reactions; CRC Press: Boca Raton, 2000.
as well as the effect of substituent groups for the 1,5-hydrogen (6) Blowers, P.; Masel, R. I. AIChE J. 2000, 46 (10), 2041-2052.
shift family. Clear trends were observed when the ring size was (7) Evans, M.; Polanyı́, M. Trans. Faraday Soc. 1938, 34, 11-29.
varied and all other substituents were held constant. As expected, (8) Boock, L. T.; Klein, M. T. Ind. Eng. Chem. Res. 1994, 33 (11),
increasing the ring size led to decreased activation barriers as 2554-2562.
(9) De Witt, M. J.; Dooling, D.; Broadbelt, L. J. Ind. Eng. Chem. Res.
well as decreasing pre-exponential factors. Evans-Polanyı́ 2000, 39 (7), 2228-2237.
relationships were shown to capture the activation energy as a (10) Kruse, T. M.; Wong, H.-W.; Broadbelt, L. J. Macromolecules 2003,
function of heat of reaction for reactions in the 1,5-hydrogen 36 (25), 9594-9607.
Low-Temperature Oxidation of Hydrocarbons J. Phys. Chem. A, Vol. 110, No. 37, 2006 10871

(11) Chan, W.; Hamilton, I.; Pritchard, H. J. Chem. Soc., Faraday Trans. (28) Scott, A.; Radom, L. J. Phys. Chem. 1996, 100 (41), 16502-16513.
1998, 94 (16), 2303-2306. (29) Van Speybroeck, V.; Van Neck, D.; Waroquier, M.; Wauters, S.;
(12) Denisova, T. G.; Denisov, E. T. Kinet. Catal. 2001, 42 (5), 620- Saeys, M.; Marin, G. B. J. Phys. Chem. A 2000, 104(46), 10939-10950.
630. (30) Cancés, E.; Mennucci, B.; Tomasi, J. J. Chem. Phys. 1997, 107
(13) Mill, T.; Hendry, D. In ComprehensiVe Chemical Kinetics; Else- (8), 3032-3041.
vier: New York, 1980; Vol. 16, pp 2-87. (31) Rappé, A.; Casewit, C.; Colwell, K.; Goddard, W.; Skiff, W. J.
(14) Sickle, D. V.; Mill, T.; Mayo, F.; Richardson, H.; Gould, C. J. Am. Chem. Soc. 1992, 114 (25), 10024-10035.
Org. Chem. 1973, 38 (26), 4435-4440. (32) McQuarrie, D. A.; Simon, J. D. Molecular Thermodynamics;
(15) Mill, T.; Montosori, G. Int. J. Chem. Kinet. 1973, 5, 119-136. University Science Books: Sausalito, CA, 1999.
(16) Rust, F. J. Am. Chem. Soc. 1957, 79 (15), 4000-4003. (33) Frenkel, M.; Marsh, K. N.; Wilhoi, R. C.; Kabo, G. J.; Roganov,
(17) Merle, J. K.; Hayes, C.; Zalyubovsky, S.; Glover, B.; Miller, T.; G. N. Thermodynamics of Organic Compounds in the Gas State; Thermo-
Hadad, C. J. Phys. Chem. A 2005, 109 (16), 3637-3646. dynamics Research Center: College Station, TX, 1994; Vol. 1.
(18) Gomez-Balderas, R.; Coote, M. L.; Hendry, D.; Radom, L. J. Phys. (34) Sumathi, R.; Carstensen, H.; Green, W. H., Jr. J. Phys. Chem. A
Chem. A 2004, 108 (15), 2874-2883. 2001, 105 (28), 6910-6925.
(19) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, (35) Marston, C. C.; Balint-Kurti, G. G. J. Chem. Phys. 1989, 91 (6),
M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. 3571-3576.
N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; (36) Balint-Kurti, G. G.; Dixon, R.; Marston, C. C. Int. ReV. Phys. Chem.
Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; 1992, 11, 317-344.
Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; (37) Balint-Kurti, G. G.; Ward, C. L.; Marston, C. C. Comput. Phys.
Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, Commun. 1991, 67, 285-292.
X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; (38) Sebbarand, N.; Bockhorn, H.; Bozzelli, J. W. J. Phys. Chem. A
Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; 2005, 109 (10), 2233-2253.
Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; (39) Cauter, K. V.; Speybroeck, V. V.; Vansteenkiste, P.; Reyniers, M.
Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, F.; Waroquier, M. ChemPhysChem 2006, 7, 131-140.
S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. (40) Hirschfelder, J. O.; Wigner, E. J. J. Chem. Phys. 1939, 7 (8), 616-
D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. 628.
G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; (41) Gonzalez-Lafont, A.; Truong, T. N.; Truhlar, D. G. J. Chem. Phys.
Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, 1991, 95 (12), 8875-8894.
M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; (42) Zhang, S.; Truong, T. N. VKLab version 1.0. University of Utah,
Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 2001.
03, revision C.02; Gaussian, Inc.: Wallingford, CT, 2004. (43) Brinck, T.; Lee, H.; Jonsson, M. J. Phys. Chem. A 1999, 103 (35),
(20) Peng, C.; Ayala, P. Y.; Schlegel, H. B. J. Comput. Chem. 1996, 7094-7104.
17 (1), 49-56. (44) Redfern, P. C.; Zapol, P.; Curtiss, L. A.; Raghavachari, K. J. Phys.
(21) Peng, C.; Schlegel, H. B. Israel. J. Chem. 1993, 33, 449-454. Chem. A 2000, 104 (24), 5850-5854.
(22) Gonzalez, C.; Schlegel, H. B. J. Chem. Phys. 1989, 90 (4), 2154- (45) Truong, T. N.; Duncan, W.; Tirtowidjojor, M. Phys. Chem. Chem.
2161. Phys. 1999, 1 (6), 1061-1065.
(23) Becke, A. J. Chem. Phys. 1993, 98 (7), 5648-5652. (46) Henry, D. J.; Radom, L. In Quantum-Mechanical Prediction of
(24) Coote, M. L.; Wood, G. P. F.; Radom, L. J. Phys. Chem. A 2002, Thermochemical Data; Understanding Chemical Reactivity; Cioslowski, J.,
106 (50), 12124-12138. Ed.; Kluwer Academic Publishers: Norwell, MA, 2001; Vol. 22, pp 161-
(25) Montgomery, J. A.; Frisch, M. J.; Ochterski, J. W.; Petersson, G. 197.
A. J. Chem. Phys. 1999, 110 (6), 2822-2827. (47) Avila, D. V.; Ingold, K. U.; Lusztyk, J.; Green, W. H., Jr.; Procopio,
(26) Baboul, A. G.; Curtiss, L. A.; Redfern, P. C. J. Chem. Phys. 1999, D. R. J. Am. Chem. Soc. 1995, 117 (10), 2929-2930.
110 (16), 7650-7657. (48) Broadbelt, L. J.; Pfaendtner, J. AIChE J. 2005, 51 (8), 2112-2121.
(27) Curtiss, L. A.; Redfern, P. C.; Raghavachari, K.; Rassolov, V.; (49) Van Geem, K. M.; Reyniers, M. F.; Marin, G. B.; Song, J.; Green,
Pople, J. A. J. Chem. Phys. 1999, 110 (10), 4703-4709. W. H., Jr.; Matheu, D. M. AIChE J. 2006, 52 (2), 718-730.