Combustion and Flame

6/20/13
Lennard-Jones parameters for combustion and chemical kinetics
modeling from full-dimensional intermolecular potentials

Ahren W. Jaspera and James A. Millerb

a
Combustion Research Facility, Sandia National Laboratories, PO Box 969, Livermore,
CA 94551-0969, USA
b
Chemical Sciences and Engineering Division, Argonne National Laboratory, Argonne,
IL 60439, USA

Abstract. Lennard-Jones parameters for use in combustion modeling, as transport

parameters and in pressure-dependent rate-coefficient calculations as collision rate
parameters, are calculated from accurate full-dimensional intermolecular potentials.
Several first-principles theoretical methods are considered. In the simplest approach, the
intermolecular potential is isotropically averaged and used to determine Lennard-Jones
parameters. This method works well for small species, but it is not suitable for larger
species due to unphysical averaging over the repulsive wall. Another method considered
is based on full-dimensional trajectory calculations of binary collisions. This method is
found to be very accurate, predicting Lennard-Jones collision rates within ~10% of those
obtained via tabulated (experimentally-based) Lennard-Jones parameters. Finally, a
computationally efficient method is presented based on one-dimensional minimizations
averaged over the colliding partners’ relative orientations. This method is shown to be
both accurate and efficient. The good accuracy of the latter two approaches is shown to
be a result of their explicit treatment of anisotropy. The effects of finite temperature
vibrations and multiple conformers are quantified and are shown to be small. The choice
of potential energy surface has a somewhat larger effect, and strategies based both on
efficient semiempirical methods and on first-principles direct dynamics are considered.
Overall, 75 systems are considered, including seven baths, targets as large as heptane,
both molecules and radicals, and both hydrocarbons and oxygenates.

Keywords: Lennard-Jones parameters, anisotropy, intermolecular potential, ab initio,

quantum chemistry

1
1. Introduction
Non-bonding intermolecular potentials play an important role in combustion
chemistry. First and foremost (from our point of view), they govern collisional energy
transfer from highly vibrationally excited molecules. In turn, this energy transfer largely
controls the rates of unimolecular reactions at low pressure, both thermal and chemically-
activated––most notably the low-pressure-limit rate coefficients of thermal dissociation
reactions. However, these same potentials also determine transport properties––the
coefficients of diffusion, viscosity, heat conduction, and thermal diffusion––that are used
in flame modeling. 1 , 2 Although some work has appeared in these areas relatively
recently,3 there is still a dearth of information that can be applied directly in chemical
kinetics and combustion modeling.
In two recent papers,4,5 we have shown that a relatively inexpensive electronic-
structure method, MP2/aug'-cc-pVDZ, gives results for intermolecular potentials that are
very close to those calculated from the high-level QCISD(T)/CBS method for CH4
interacting with several small-molecule bath gases (He, Ne, H2, and CH4). Furthermore,
we parameterized a very efficient semiempirical potential energy surface for
hydrocarbons interacting with typical atomic and diatomic baths based on the
QCISD(T)/CBS interaction energies. Using these potentials we calculated energy transfer
rates using trajectories that give accurate low-pressure rate coefficients for the CH4 (+ M)
⇄ CH3 + H (+ M) reaction, with M being any one of several typical bath gas molecules.
Subsequent work (as yet unpublished for CH4 + H2O, CH3OH + He, and CxHy + M) has
reinforced our conclusion that the MP2 and semiemprical methodologies can be
workhorses for obtaining collisional energy transfer rates in highly vibrationally-excited
molecules, and hence accurate unimolecular rate coefficients at low pressure.
The purpose of the present investigation is considerably less ambitious than that
described above, but it is still important. We want to use the intermolecular potential
energy methods discussed above to calculate accurate Lennard-Jones collision rates (or
frequencies) with minimal computational effort. In unimolecular reaction rate theory,6,7
one is always referring to !E or !Ed , the average energy transferred in a collision

or the average energy transferred in a deactivating collision. For these quantities to have
meaning one must first define what a collision is. There is enormous flexibility in this

2
definition, 8,9 but by convention it is normally defined in terms of a Lennard-Jones
collision rate, which must be determined accurately. The origin of the use of the Lennard-
Jones collision rate lies in the theory of transport processes. It is worthwhile to review
this point briefly.
First, if one has a binary mixture of hard-sphere molecules, the collision rate of a
molecule of species i with molecules of species j is1
Zij = !" ij2 n j gij , (1)
1/2
! 8k T $
where n j is the number density of j-type molecules; gij = ## B && is the average relative
" !µij %

mi m j
speed for i, j collisions; k B is Boltzmann's constant; µij = is the reduced mass
mi + m j

1
for i, j collisions; T is the temperature; and ! ij = (! i + ! j ) is the collision diameter
2
(with ! i and ! j the diameters of the two spheres). More generally, we will define ! ij as

the point where the intermolecular potential V (r) intersects the r axis––this is the
collision diameter for hard spheres. In the same hard-sphere model the binary diffusion
coefficient is
1
3 3
3 (2! k BT / µij ) 2

Dij = , (2)
16 p!" ij2

where p is the pressure. Defining zij = Zij n j to be a hard-sphere “collision rate

coefficient”, the diffusion coefficient becomes

1
3 3
3 gij (2! k BT / µij ) 2
Dij = . (3)
16 pzij

Thus all of the dependence of Dij on the potential is absorbed into zij .

If one considers the Lennard-Jones potential,

(! $12 ! $6 +
" "
V (r) = 4!ij *## ij && ' ## ij && - , (4)
*" r % " r % -
) ,

3
where ! ij was defined above and !ij is the well depth [or any potential of the form

V (r) = ! f (r / " ) ], the diffusion coefficient can be written as

1
3 3
3 (2! k BT / µij ) 2
Dij = (5)
16 p!" ij2!(1,1)*

where Ω (1,1)* is a reduced collision integral that depends on the intermolecular potential.
Equation (5) suggests that, by analogy with the hard-sphere case, we define the collision
rate coefficient as
zij = !" ij2!(1,1)* gij (6)

One can view this as a general result with ! ij and !(1,1)* evaluated for the particular

potential under consideration ( !(1,1)* = 1 for hard spheres). Results analogous to Eq. (6)
can also be derived using the viscosity and conductivity,1 but in these cases !(2,2)* ,
another reduced collision integral, replaces !(1,1)* . We shall not be concerned with the
precise definition of the collision integrals here except to note that their values are readily
obtainable in tabular form. We are most interested in diffusion coefficients and hence in
!(1,1)* , but the Lennard-Jones potential parameters that we determine can be used to
calculate any of the reduced collision integrals. !(1,1)* and !(2,2)* are usually very similar
in magnitude, commonly differing by less than 10%.
One is tempted to think that this is all well and good, but real molecules are not
point particles (or hard spheres). They have structure––their potential depends on more
than just the distance between the centers of mass of the two colliding molecules.
Interestingly, high level quantum scattering calculations on ab initio potential energy
surfaces show that transport properties can be computed quite accurately by using only
the “spherically-averaged,” isotropic part of the potential, i.e. by assuming point particles,
at least for small molecules.10,11,12 Also, the decades-old practice of fitting experimental
data to Lennard-Jones potentials suggests that this simplification may be accurate.
In the present article we discuss different methods of obtaining Lennard-Jones
parameters from detailed, full-dimensional intermolecular potentials. We compare the
results with the experimental results available, normally by comparing collision rate
coefficients for the different sets of Lennard-Jones parameters. We primarily consider

4
cases where helium is one of the collision partners. This simplifies the electronic-
structure calculations and eliminates any effects of dipole-dipole interactions. The pure-
gas parameters can be determined from whatever combining rules one may wish to use,
as discussed recently by Brown et al.3
We would be remiss if we did not mention evidence that the Lennard-Jones
potential may be too repulsive at short distances, leading to the underprediction of some
diffusion coefficients at high temperatures by ~20%.13 Paul and Warnatz14 suggest that
an exponential repulsive part may be more appropriate at high temperatures. The present
methodology could be generalized to obtain potential parameters for such a potential as
well. However, for now the Lennard-Jones potential is still in common use, largely
because it is the potential used in CHEMKIN.15 Brown et al.3 recommend the use of
Lennard-Jones parameters in combustion modeling over more complicated forms, as well.
We will limit ourselves to determining those parameters in the present article. We also
want to note that transport properties, particularly thermal conductivity and viscosity,
depend on some other molecular properties, most notably dipole moments and
polarizabilities. These properties can be computed accurately using relatively low-level
electronic structure methods. We hope to address this issue in the near future.

2. Theory
Here we describe several methods for calculating isotropically-averaged (i.e.,
spherically-averaged) intermolecular potentials and/or Lennard-Jones collision
parameters for a molecule or radical (A) interacting with a bath gas atom or molecule (B).
All of the methods involve averaging the full-dimensional anisotropic intermolecular
potential, V(R), in some way over the colliding species’ relative orientation and internal
structures. This is simpler than evaluating the collision integrals directly using the full
anisotropic potential. Instead, Lennard-Jones collision rate coefficients, z, are obtained
from the calculated Lennard-Jones parameters σ and ε using Eq. (6) with6

!(1,1) * " [0.7 + 0.52log10 (k BT / ! )]#1 , (7)

where the “i,j” subscripts have been suppressed.
The total geometry of the interacting binary A + B system can be written as
R ≡ (R A , R B , R Ω , r ) , where RA labels the internal coordinates of the target species, RB

5
labels the internal coordinates of the bath gas (if any), and the remaining coordinates
define the relative orientation of the target and bath, R Ω , and their center-of-mass
separation, r. The full-dimensional intermolecular potential is defined relative to the
separated unrelaxed species, i.e.,
V(R) = E(R) – EA(RA) – EB(RB), (8)
where E(R) is the total energy of the interacting target and bath gas evaluated at R,
EA(RA) is the energy of the isolated target molecule or radical evaluated at RA, and
EB(RB) is the energy of the isolated bath gas atom or molecule evaluated at RB (which for
atomic baths is zero).
We consider two methods for calculating effective isotropic potentials. In the
simplest scheme, the internal structures of the target and bath are set to their experimental
A B
or calculated equilibrium geometries, R eq and R eq , and V(R) is averaged over R Ω at

fixed r. The isotropically-averaged potential is then

A B
V (r) = ! V (R eq
,R eq ,R "(n) ,r) / N , (9)
n=1,N

where n labels the N uniformly sampled orientations for each r. At finite temperatures,
the target and bath are not confined to their equilibrium structures.16 Equation (9) can be
modified to include this effect by sampling the internal configurations of the A and B
from thermal distributions as we sample over relative orientations, i.e.,
A B
V (r) = ! V (R (n)
,R (n) ,R "(n) ,r) / N . (10)
n=1,N

Equation (9) may be considered the low temperature limit of Eq. (10). Generally, Eq. (10)
will predict longer-ranged interactions due to the effect of anharmonic vibrations. Even at
elevated temperatures, however, one may expect Eqs. (9) and (10) to predict very similar
effective isotropic potentials for small molecules and for larger molecules that do not
feature large amplitude motions. Only for systems with multiple low-lying conformers
with different shapes (long chain species with conformers related via torsions, for
example) would one expect the effective anisotropic potential to depend non-negligibly
on temperature; for such systems Eq. (10) may differ from Eq. (9).
The desired Lennard-Jones parameters can be determined straightforwardly from
the isotropically-averaged potentials, with

6
 ε = − min r V (11a)

V (! ) = 0 , (11b)
where the second equation sets σ as the inner turning point of V evaluated at its
asymptotic energy. Equation (11) requires locating two critical points on V : its minimum
and its inner turning point. It may be more convenient again to set
ε = − min r V (12a)
but then to use the associated value of r, rmin, to approximate
! = 2 !1/6 rmin . (12b)
If the calculated isotropic potentials defined by Eqs. (9) and (10) were in fact Lennard-
Jones potentials, then Eqs. (11) and (12) would be entirely equivalent. The difference in
the predicted values of σ using Eqs. (11) and (12) is therefore one measure of the
deviation of V from the Lennard-Jones form.
In the applications reported below, Eqs. (9) and (10) were evaluated for several
A
normal alkane + He systems with N = 2000–10000 for each r. For Eq. (10), R (n ) was

sampled from thermal distributions of geometries determined via separate molecular

dynamics calculations subject to an Andersen thermostat.17 In addition to reporting V ,
we also report the minimum and maximum value of V sampled for each r. These extrema
provide a measure of the anisotropy of V.
Next we consider using classical trajectories to calculate Lennard-Jones
parameters directly,18 i.e., not in reference to an isotropically-averaged intermolecular
potential. In this approach, a small ensemble of target–bath gas collisions is prepared
using initial conditions appropriate for collisional energy transfer calculations,4,5 subject
to the additional constraint of zero impact parameter. Along each trajectory n, the
instantaneous value of V is monitored, and the most negative value of V encountered, Vn,
and its associated center-of-mass distance, rn, are recorded. The trajectory-based
Lennard-Jones parameters are then defined
ε = − ∑ Vn / N . (13a)
n =1, N

! = 2!1/6 "r n
/N. (13b)
n=1,N

7
 This approach for calculating σ and ε differs in several ways from the use of Eqs.
(9) and (10). First, the target and bath are vibrating during the full-dimensional trajectory
collisions, and these internal motions may, in general, react to the presence of the other
species during the collision; i.e., the trajectory interactions are “softer” and include
dynamical effects due to the collision that are not available to the static sampling and
averaging approaches described by Eqs. (9) and (10). Second, the trajectory-based
Lennard-Jones parameters are derived directly from the full-dimensional (anisotropic)
potential, whereas in Eqs. (9) and (10) the anisotropy is averaged over before the
Lennard-Jones parameters are determined. If the intermolecular potential were really
isotropic and if the dynamical effect of internal vibrations were small, the two approaches
would give similar results. The trajectory-based Lennard-Jones parameters therefore
include anisotropic effects that are not included in Eqs. (9) and (10) and may be
considered “effective” Lennard-Jones parameters that include this anisotropy. Finally, the
trajectories are subject to total energy constraints, and so the trajectory method never
samples highly repulsive interactions, whereas the isotropically averaged potentials in
Eqs. (9) and (10) will generally include contributions from geometries high on the
repulsive wall, even for r close to the minimum of the effective isotropic potential.
Full-dimensional molecular dynamics may be too expensive for practical
evaluations of Lennard-Jones parameters, and a somewhat simpler approach is
considered next. However, when one requires Lennard-Jones parameters to be used
alongside molecular dynamics studies of energy transfer, the present ensembles represent
a small fraction of the total cost of such a calculation.
We also note that one could use classical trajectories to calculate transport
properties. 19 , 20 An advantage of such an approach would be that one could avoid
assuming the isotropic Lennard-Jones form (or some other simplified form) for the
intermolecular potential. The goal of the present work is to validate convenient schemes
for obtaining transport and collisional parameters for use in building detailed models of
combustion and for pressure-dependent reaction rate coefficients, where, as discussed
above, it is both conventional and expedient to assume Lennard-Jones interactions. We
therefore do not consider using full-dimensional trajectories to calculate transport
properties directly in this work.

8
 Finally, we present a method that includes the effect of local anisotropy, as in the
trajectory-based method, but that does not require full-dimensional trajectories. This
method incorporates desirable aspects of both of the approaches described above. In the
A
simplest implementation, R ! is sampled uniformly and V (R eq , R eqB , R Ω( n) , r ) is minimized
with respect to r, and
d A B
V (R eq ,R eq ,R !(n) ,r) = 0 , (14)
dr
defines the values of Vn and rn. Lennard-Jones parameters are then calculated via Eq.
(13). Alternatively, one can perform an additional optimization to locate the inner turning
point, σn, for each R Ω(n ) and calculate

ε = − ∑ Vn / N . (15a)
n =1, N

!= !! n /N. (15b)
n=1,N

Finally, we note that one could sample over the internal geometries of the target and bath
gas along with the orientations, where Vn and rn are then obtained by minimizing
V (R (An) , R (Bn) , R Ω( n) , r ) with respect to r,

d A B
V (R (n) ,R (n) ,R !(n) ,r) = 0 . (16)
dr
Such an approach includes temperature and multiconformational effects, as discussed
above for Eq. (10).
As with the trajectory approach, the one-dimensional-minimization approach
results in effective Lennard-Jones parameters that include anisotropy. The present
method differs from the trajectory method in that it is much less computationally
demanding, with the tradeoff that dynamical collision effects are not included.
The results of any of the methods described above will depend sensitively on the
potential energy surface used. Several potential energy surfaces have been considered
here, including both direct dynamics and more efficient approximate representations. For
hydrocarbons targets, we make use of our recently developed analytic intermolecular
potential energy surfaces,5 which have been validated against high-level QCISD(T)/CBS
energy calculations as well as in dynamical calculations of energy transfer. These

9
potential energy surfaces (labeled “exp/6”) employ a tight binding model21 to describe the
internal potential of the target along with parameterized “separable pairwise”
Buckingham (i.e., exp/6) interactions for the target–bath atomic interactions.
Buckingham parameters for several baths are taken from Refs. 5 and 22. For targets
containing oxygen atoms and to further test the exp/6 parameterization for hydrocarbons,
we also consider high-level direct MP2/CBS and QCISD(T)/CBS calculations with
counterpoise corrections and CBS extrapolations based on aug-cc-pVDZ and aug-cc-
pVTZ basis sets. (One expects the more widely used and slightly more computationally
expensive CCSD(T) method to perform similarly to the QCISD(T) method.) Finally, we
test the accuracy of less expensive direct dynamics methods, including (briefly) DFT and
MP2 with small and medium-sized basis sets and our previously employed4,5 MP2/aug'-
cc-pVDZ potential.
The calculated binary Lennard-Jones parameters and collision rates are compared
with tabulated values collected from a variety of sources. The usual combining rules are
used, with the arithmetic mean used for σ and the geometric mean used for ε. Pure gas
values for the bath gases are taken principally from the tables distributed with
CHEMKIN,15 with Ne and Kr from Ref. 23: σ = 2.576, 2.749, 3.330, 3.684, 2.920, 3.681,
3.458 Å and ε = 7.098, 24.74, 94.87, 121.4, 26.41, 67.89, 74.64 cm–1 for He, Ne, Ar, Kr,
H2, N2, and O2, respectively. Parameters for several normal alkanes are taken from Tee et
al.24 and are based on experimental viscosities and second virial coefficients. Hexane is
not included in this tabulation, so its values are linearly interpolated from pentane and
heptane. Parameters for the other target molecules and radicals are taken from
CHEMKIN’s distributed tables.15

3. Results and discussion

3.1. Isotropically averaged potentials
The exp/6 potential was used to evaluate the isotropically averaged potentials
described by Eqs. (9) and (10) with N = 10000 for four normal alkanes + He, with the
resulting Lennard-Jones parameters summarized in Table 1. The isotropically-averaged
potentials defined by Eq. (9) are plotted in Fig. 1 for methane, ethane, and butane + He,
where it can be seen that the potentials are well described by the Lennard-Jones form.

10
The calculated values of σ obtained via Eqs. (11) and (12) differ negligibly for CH4 and
only by 4% for hexane, which supports the use of this functional form to describe the
isotropically-averaged intermolecular potential, at least for the region of the potential
near the well and inner turning point. One can notice some minor deviations from the
Lennard-Jones form at longer and shorter distances, particularly for the larger species.
Using Eq. (9), the calculated values of σ2 (the square of σ is proportional to the
collision rate, as shown in Eq. (6)) are larger than the tabulated ones by only 17% for
methane but by more than 100% for hexane, while the calculated values of ε are smaller
by 30–80%. Because the deviations from the tabulated values for these two parameters
are in different directions, these effects cancel somewhat when calculating the collision
rates. For example, for CH4 + He σ2 and ε differ from the tabulated values by +17% and
–30%, while the collision rates differ by only 10%.
Agreement with the tabulated collision rates is worse for the larger systems,
principally due to significant differences between the calculated and tabulated values of
σ. Furthermore, while the tabulated well depths ε increase almost linearly with size, the
calculated ones are similar for methane, ethane, and butane, and the calculated value of ε
for hexane is smaller than for the smaller species. These differences can be explained by
considering the anisotropy in the intermolecular potential, as indicated by minimum and
maximum sampled values of V shown in Fig. 1. Even for the smallest system considered,
CH4 + He (Fig. 1(a)), there is significant anisotropy in the potential for r < 4 Å. At the
inner turning point on V (r = 3.38 Å) for example, the minimum and maximum energies
on the anisotropic potential differ by 100 cm–1, which is 4x as large as ε. This suggests
that, even for a small and roughly spherical system like CH4 + He, one may expect
significant anisotropy and that predicted Lennard-Jones parameters will be sensitive to
the treatment of this anisotropy. Similarly, we note that the locations of the inner turning
points along the most and least attractive approaches for CH4 + He differ by almost 1 Å.
As may be expected, there is even more anisotropy for the larger species. For
C2H6 + He, one can identify two groups of approaches in the minimum-energy curve in
Fig. 1(b): one with a minimum near r = 4.2 Å that includes approaches to the –CH3 ends
of ethane with interactions similar in magnitude to the CH4 + He interactions, and one
with a minimum near r = 3.5 Å that includes shorter-ranged and more attractive

11
interactions as He approaches perpendicularly to the C–C axis of ethane. These two types
of approaches lead to even greater anisotropy for ethane than for methane. For butane,
there are more than two qualitatively different types of approaches, but we do not attempt
to enumerate them. Instead, we note that the variation in the locations of the inner turning
points for the most and least attractive approaches for butane + He is 2.3 Å!
This significant anisotropy explains the inaccuracy of Eq. (9) for large systems.
For example, when sampling over the butane + He interactions at the tabulated value of σ
= 3.8 Å, the averaged potential includes geometries very high up on the repulsive wall––
as high as 35,000 cm–1. For large enough molecules, these energies can be made
arbitrarily large, because the bath gas atom may be put arbitrarily close to an atom on the
target molecule. The isotropic potentials obtained via Eq. (9) are therefore contaminated
by the resulting “infinities.” As a practical matter, the exp/6 potentials are not expected to
be accurate so high up on the repulsive wall. Even if the potentials were accurate, it is
formally undesirable that this region of the potential should contribute so significantly in
determining the transport properties. Most importantly, perhaps, the use of Eq. (9), while
accurate for small systems, significantly overpredicts the tabulated collision rates for
butane and hexane. One may conclude that these overpredictions are due to inaccurate
averaging over the repulsive wall, which leads to erroneously large values of σ.
We note another complication that arises for the normal alkanes larger than
propane and more generally for any system with multiple conformers. Butane + He
parameters were predicted via Eq. (9) for two equilibrium structures: one corresponding
to the anti conformer of butane and one corresponding to the gauche conformer of
butane. As seen in Fig. 1(c), the gauche conformer results in a shorter-ranged
isotropically-averaged potential relative to that of the anti conformer and leads to
collision rates that are ~10% smaller (as shown in Table 1). One may anticipate this
effect to be more significant for larger systems. This ambiguity can be resolved by using
the temperature-dependent method defined by Eq. (10). This method properly weights the
two conformers according to their relative thermal populations. The drawback of this
approach is that the predicted Lennard-Jones parameters are temperature dependent.
Nonetheless, Eq. (10) may be useful for some systems. Even for systems without
torsions, the effect of finite-temperature vibrations can be seen in Table 1. For methane

12
and ethane + He, the use of Eq. (10) leads to collision rates that are very similar at 300 K
and that are only a few percent larger at 3000 K. We may conclude that––aside from
multiple conformer effects––the effect of finite-temperature vibrations is small, even at
high temperatures.

3.2. Effective parameters with anisotropy from full-dimensional trajectories

Ensembles of N = 2000 classical trajectories and the exp/6 potential were used as
described in Sec. II to calculate Lennard-Jones parameters for a variety of normal
hydrocarbons and for several atomic and diatomic baths. The results are compared with
tabulated values in Table 2. Agreement is excellent when He is the bath gas, with the
calculated Lennard-Jones collision rates differing from the tabulated ones typically by
less than 3%. For the other baths, the predicted collision rates again agree well with the
tabulated values, but the differences can be as large as 10%. The good agreement over
this wide variety of baths and for targets as large as heptane is encouraging.
For He, the individual Lennard-Jones parameters are very similar to the tabulated
values, with σ typically predicted within ~0.1 Å and ε predicted within a few cm–1. For
the other baths, the differences are somewhat larger, and the predicted collision rates
benefit from the cancellation discussed above. Bastien et al.25 have stressed that the
experimental measurements cannot fix the two parameters individually and have instead
identified a two-dimensional “trough” of suitable experimentally-inferred pairs of values
for σ and ε. This trough can be defined through the collision rate, as discussed in the
Introduction. The present differences in the individual parameters therefore do not
necessarily indicate poor performance of the method. Again, the overall agreement in the
predicted collision rates is very good.
Transport parameters are most often tabulated for use in kinetic modeling as pure
gas (self-interaction) Lennard-Jones parameters, which are then converted to binary
Lennard-Jones parameters via simple combining rules.3,25 To make a practically useful
list from the present results, the binary Lennard-Jones parameters calculated here would
need first to be converted to a list of self-interaction parameters. To do this, one could
assume some set of pure gas parameters for the bath gas and then calculate the pure gas
parameters for the target species by inverting the combining rule, as noted in the

13
Introduction. The results in Table 2 can be used to test the sensitivity of the results of this
procedure on the choice of the bath gas. Using tabulated parameters for the seven bath
gases, the inferred pure gas values of σ for CH4 vary from 3.17−3.83 Å and the inferred
pure gas values of ε vary from 82−323 cm 1. Notably, the largest value for σ and the
−

smallest value for ε are obtained for He, and H2 gives the smallest value for σ and the
largest value for ε. Similar sensitivities are obtained for the larger alkanes. One cannot
assign a preferred bath gas by comparison with tabulated values, as the tabulated values
themselves require assuming a combining rule. Instead we note that N2 is the most
important collider in many applications, and its values are close to the mean of the ranges
identified above. Furthermore, while the maximum deviation in σ is large for the
different baths, the root-mean-squared deviation in σ is only 7% (15% in the collision
rate). The relative deviation is larger for ε, but the collision rate is less sensitive to this
parameter. In general, we recommend the use of N2 as a collision partner when
generating pure gas collision rates. Alternatively, one could directly calculate pure gas
collision parameters. Often, however, these interaction potentials are not well known, and
the appropriateness of the combining rules used in subsequent kinetic modeling would
nonetheless remain in question.
The systems in Table 2 were selected to keep things manageable, but the
trajectory method and the exp/6 potential may be readily applied to unsaturated
hydrocarbon species and hydrocarbon radicals as well. This is demonstrated for ten C3Hx
+ He systems in Table 3. There are clear trends with respect to the number of hydrogen
atoms. The collision rates for the molecular species increase systematically with
increasing saturation—by 7% for each pair of hydrogen atoms added. The two C3H4
isomers, allene and propyne, have similar collision rates. In practice and due to a general
lack of information about radicals, radicals species are often assigned the collision
parameters of related hydrogenated or dehydrogenated molecules. The present results
suggest that it is a good approximation to adopt parameters of “nearby” hydrogenated
molecules, although the radical species’ collision rates are sometimes smaller by a few
percent.

14
3.3. Effective parameters with anisotropy from one-dimensional minimizations
The exp/6 potential was used to perform one-dimensional minimizations,
averaged over orientation, as described by Eqs. (14) and (16) with N = 2000 for four
normal alkanes + He. The results are shown in Table 4. The predicted collision rates are
within ~15% of the tabulated ones, and, unlike the isotropically-averaged potential
methods, the accuracy of the one-dimensional minimization approach remains good for
the larger systems. The one-dimensional minimization methods tend to overpredict the
tabulated rates, and this overprediction does increase somewhat with size.
Figure 2 shows representative one-dimension cuts for three normal alkanes + He.
The local minimum for each cut is indicated, and these values are averaged (for larger
ensemble sizes than shown in Fig. 2) to obtain the results in Table 4. Figure 2 may be
compared with Fig. 1. In Fig. 1, the potential is spherically averaged first, and then the
Lennard-Jones parameters are extracted. In Fig. 2, Lennard-Jones parameters for each
approach are determined first, and these are then averaged. Clearly, the latter approach
incorporates anisotropic information not available to the former approach. The poor
performance of the approach in Fig. 1 further demonstrates that local anisotropic
information is required for quantitative predictions.
Also shown in Fig. 2 are representative results from the trajectory method
discussed above. Note that while the trajectory approach results in values for Vn and rn
that are similar to those obtained via the one-dimensional minimizations, they are
systematically shorter-ranged and more attractive. These differences may be attributed to
dynamical effects not included in the one-dimensional minimizations. Specifically, the
colliding partners experience “softer” interactions than the rigid fragment approaches due
to vibrational deformations from the impact of the collisions.
There is relatively little sensitivity of the one-dimensinal minimization results on
the temperature of the target, with the application of Eqs. (14) and (16) predicting rates
differing by only a few percent. This is in contrast to the results shown in Table 1 for the
isotropic potential methods. We can explain this result by noting that the one-dimensional
minimizations depend only on the local (i.e., anisotropoic) potential, whereas the
isotropic averages depend on the global potential for fixed r. The choice of conformer of

15
the target molecule therefore has more of an effect on the predictions of the isotropically-
averaged methods in Table 1 than the one-dimensional minimization methods in Table 4.
The results in Table 4 are similar in overall accuracy to those predicted by the
trajectory method in Tables 2 and 3. There are systematic differences, however. The
trajectory method tends to underpredict the collision rates, whereas the one-dimensional
minimization method overpredicts them. If we compare the finite temperature results in
Table 4 with the trajectory results in Table 2, we can attribute the differences in the
predictions to dynamical effects in the trajectory calculations. These differences vary
from ~5% for methane to ~15% for hexane.
The one-dimensional minimizations are orders of magnitude less demanding
computationally than the trajectory method and are simple to implement, particularly if
Eq. (14) is used. We therefore recommend this approach as a practical method for
obtaining accurate Lennard-Jones collision parameters.

3.4. Sensitivity to the potential energy surface

We also consider the sensitivity of the predicted results to the choice of potential
energy surface. For these tests, we use the one-dimensional minimization method
exclusively, with finite temperature effects neglected, i.e., we use Eq. (14). Results for
several small systems + He are summarized in Table 5. Aside from relatively poor results
for H + He and CH + He, all of the potentials considered predict collision rates in fairly
good agreement with the tabulated values, typically within 15%. The larger differences
for the lighter species and for H + He in particular may be due to the representation of the
repulsive wall by the Lennard-Jones form.13,14 The present calculated values for σ and ε
for H + He agree with those based on the high level calculated potential energy surface of
Middha et al.13
The QCISD(T)/CBS method is the highest-level method considered, and it is
encouraging that it generally agrees with the somewhat less expensive MP2/CBS method.
The much less expensive MP2/aug'-cc-pVDZ method often agrees with the higher-level
methods, with differences as large as 10%. The exp/6 potential was originally
parameterized5,22 against QCISD(T)/CBS calculations for CH4 + He but neglects
nonpairwise interactions. Furthermore the exp/6 potential uses the CH4 + He-based atom-

16
atom interaction parameters for all hydrocarbon interactions. We have not developed
similarly general exp/6 parameters for oxygenates.
We emphasize that all of the potentials considered here may be expected to be
fairly accurate, either as first-principles approaches (MP2/CBS and QCISD(T)/CBS) or
because they have been previously validated against such high-level calculations (exp/6
and MP2/aug'-cc-pVDZ). To demonstrate the errors associated with choosing a bad
potential, we obtained results using the B3LYP and M06-2X density functional theory
methods. Both of these functionals have been successfully employed in a wide variety of
contexts throughout chemistry, but neither is explicitly designed to treat dispersion
interactions. The resulting Lennard-Jones parameters can by unphysical when these
methods are used. For example, the M06-2X/cc-pVDZ method predicts ε = 161 cm–1 and
σ = 2.30 Å for C + He, which are very different from the tabulated values of 18.8 cm–1
and 2.94 Å and the high level (QCISD(T)/CBS) calculated values of 19.5 cm–1 and 2.91
Å. The B3LYP/cc-pVDZ method also predicts innacurate values of ε = 80.2 cm–1 and σ =
2.33 Å for C + He. Density functional theories designed specifically for dispersion
interaction26 were not tested here. Instead we note the general difficulty in predicting
weak dispersion reactions and the good accuracy of the counterpoise corrected MP2 and
QCISD(T)/CBS methods.
In general, the calculation of accurate intermolecular potentials requires the use of
large augmented (diffuse) basis sets, basis set extrapolations, counterpoise corrections,
and some post-Hartree–Fock approach, preferably with perturbative triples. 27 Less
expensive potentials may accidentally perform well in predicting collision rates, however,
due to the cancellation discussed above. When a less predictive method, say, overpredicts
the strength of the interactions, it is almost certain also to be shorter-ranged, such that the
two errors cancel.
Finally, results for several systems relevant to combustion are reported in Table 6
calculated using the MP2/CBS potential and Eq. (14). General agreement with tabulated
collision rates is good, although there are notable exceptions. As noted above, the present
approach overpredicts the H + He collision rate by 50%, which may be due to the
treatment of the repulsive wall. The tabulated15 pure gas value of σ for HCCO is 2.5 Å,
which is much smaller than tabulated values of σ for related species. The predicted value

17
of σ for HCCO + He is 3.6 Å and is similar to other HxC2O species. This choice in the
tabulated value of σ for HCCO leads to values of zcalc/ztab close to 2 in Table 6 for HCCO
+ He. Otherwise, the predicted and tabulated collision rates typically differ by less than
20%.

4. Conclusions
Several methods for calculating binary Lennard-Jones collision parameters from
full-dimensional intermolecular potentials have been considered. While orientationally
averaging over real potentials does result in isotropically-averaged potentials that are well
described by the Lennard-Jones form, collision rates based on these potentials are not
accurate for systems larger than a few atoms. We demonstrated that, even for small
molecules, real potentials feature significant anisotropy in the region of the potential
relevant to transport. When orientationally averaging over real potentials, this anisotropy
leads to unphysical averaging over repulsive interactions and significant errors in
predicted transport parameters.
Two methods for calculating effective Lennard-Jones parameters directly from
full-dimensional anisotropic intermoleculars potential were considered. The first involves
small ensembles of full-dimensional trajectories and was shown to predict Lennard-Jones
collision rates in excellent (10%) agreement with tabulated values for systems as large as
heptane and for several baths. The predicted values of σ are typically smaller than the
tabulated values, and the predicted values of ε are typically larger than the tabulated
values. These two errors cancel when evaluating collision rates and therefore may not
indicate poor performance of the method, as emphasized by Brown and co-workers.3,25
The second method for calculating effective Lennard-Jones parameters is much
simpler and more computationally efficient than the trajectory approach and involves
one-dimensional minimizations averaged over relative orientations of the two species.
Again, agreement between the predicted and tabulated collision rates is excellent (10%).
The trajectory method tends to underpredict the collision rates, while the one-dimensional
minimization method tends to overpredict the collision rates. These systematic
differences may be attributed to dynamical effects resulting in “softer” interactions in the
trajectory approach.

18
 In principle, one can include multiconformer effects and, more generally, the
effect of temperature-dependent vibrations for the target and bath gas species for any of
these methods tested here. The present tests confirm that such effects are small and may
typically be neglected.
Both radical and molecular species were considered. The present results confirm
that Lennard-Jones parameters for radical species may be accurately approximated as
those of associated hydrogenated (+H) molecules. One can, however, identify a
systematic increase in the predicted collision rates of molecular species with increasing
saturation (+2H).
The consistency of simple combining rules (the arithmetic mean for σ and the
geometric mean for ε) with the present results was tested for several normal alkanes and
several baths. Pure gas Lennard-Jones parameters for the normal alkanes obtained from
binary parameters depend fairly sensitively on the choice of bath gas. Notably, the use of
He as the bath gas, systematically results in the largest pure gas normal alkane values for
σ and the smallest values of ε. The use of N2, on the other hand, results in pure gas
normal alkane values of σ and ε close to their average values. Although He was primarily
considered here, it may be more useful to use N2 as the collision partner when compiling
lists of transport parameters for use in modeling studies.
Finally, the sensitivity of the predicted parameters on the choice of the level of
theory used to describe the potential energy surface was considered. The previously fitted
semiempirical exp/6 intermolecular potential is found to be accurate, but its applicability
is limited to hydrocarbons interacting with typical atomic and diatomic baths. For
oxygenated species, two high level ab initio methods were considered: the counterpoise
corrected MP2 and QCISD(T) methods with CBS basis set extrapolations. Typically the
two methods were found to result in similar predictions. More approximate ab initio
methods were also considered. While the previously employed MP2/aug'-cc-pVDZ
method performed fairly well, other inexpensive ab intio methods including density
functional theory were shown to lead to significant errors. Notably, these results again
demonstrated the significant cancellation in the collision rate arising from opposite-sign
errors in the predicted values of σ and ε.

19
 The cumulative effect of the present predicted (typically) 0–20% changes to
collision rates in Table 6 on various properties of combustion modeling is not known.
The present approach provides not only a convenient means for obtaining transport
properties for species not previously tabulated but also provides a means for obtaining a
systematic set of transport properties. Such a systematic set may aid in future sensitivity
analyses of transport properties in combustion modeling.

Acknowledgments
A. W. J. was supported by the Division of Chemical Sciences, Geosciences, and
Biosciences, Office of Basic Energy Sciences, U.S. Department of Energy. Sandia is a
multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company,
for the United States Department of Energy under Contract No. DE-AC04-94-AL85000.
J. A. M. was supported under contract number DE-AC02-2006-CH11357 as part of the
ASC-HPCC (ANL FWP #59044).

20
Table 1. Calculated Lennard-Jones parameters based on effective isotropic potentials
compared with tabulated values and relative collision rates for four normal alkanes + Hea
Calculatedb Tabulatedc zcalc/ztab
Target Method σ, Å ε, cm–1 σ, Å ε, cm–1 300 K 1500 K 3000 K
CH4 Eq. (9) 3.38 19.8 3.13 28.7 1.09 1.11 1.11
Eq. (10) 3.39–3.43 19.6–18.7 3.13 28.7 1.09 1.12 1.14
C2H6 Eq. (9) 3.77 24.2 3.40 36.8 1.13 1.16 1.16
Eq. (10) 3.79–3.85 23.8–22.3 3.40 36.8 1.14 1.17 1.20
C4H10 Eq. (9)d 4.73 19.0 3.79 44.3 1.32 1.37 1.39
e
Eq. (9) 4.46 24.4 3.79 44.3 1.23 1.26 1.28
Eq. (10) 4.50–4.67 23.3–20.1 3.79 44.3 1.24 1.31 1.36
C6H14 Eq. (9)d 5.93 11.3 4.04 51.9 1.61 1.73 1.76
Eq. (10) 5.88–5.74 11.7–12.2 4.04 51.9 1.59 1.66 1.67
a
For the exp/6 intermolecular potential
b
Via Eq. (11)
c
From Ref. 24
d
Anti conformer
e
Gauche conformer

21
Table 2. Calculated Lennard-Jones parameters based on full-dimensional trajectories
compared with tabulated values and relative collision rates for several normal
hydrocarbons in several bathsa
Calculated Tabulatedb zcalc/ztab
–1 –1
Target Bath σ, Å ε, cm σ, Å ε, cm 300 K 1500 K 3000 K
CH4 He 3.20 24.1 3.13 28.7 1.01 1.02 1.02
Ne 3.25 53.7 3.21 53.6 1.02 1.02 1.02
Ar 3.47 128 3.50 105 1.04 1.02 1.01
Kr 3.51 163 3.68 119 1.00 0.97 0.96
H2 3.05 92.4 3.30 55.3 0.96 0.93 0.92
N2 3.59 122 3.68 88.7 1.04 1.01 1.00
O2 3.35 126 3.57 93.0 0.95 0.93 0.93
C2H6 He 3.36 36.3 3.40 36.8 0.97 0.97 0.97
Ar 3.62 168 3.78 134 0.98 0.96 0.96
N2 3.78 148 3.95 114 0.98 0.96 0.95
C3H8 He 3.55 43.4 3.78 36.3 0.92 0.91 0.90
Ar 3.80 194 4.16 133 0.94 0.90 0.90
N2 3.96 164 4.33 112 0.93 0.90 0.89
C4H10 He 3.71 48.3 3.79 44.3 0.98 0.97 0.97
Ne 3.76 80.6 3.88 82.8 0.93 0.93 0.94
Ar 3.95 214 4.17 162 0.98 0.95 0.94
Kr 3.94 292 4.34 183 0.96 0.91 0.90
H2 3.67 147 3.96 85.6 0.99 0.95 0.94
N2 4.12 178 4.34 137 0.97 0.95 0.94
O2 3.86 240 4.23 144 0.97 0.92 0.91
C5H12 He 3.84 52.6 3.93 48.3 0.97 0.97 0.97
Ar 4.04 235 4.31 177 0.96 0.93 0.93
N2 4.23 192 4.48 150 0.96 0.94 0.93
C6H14 He 3.96 56.4 4.04 51.9 0.98 0.97 0.97
Ne 3.96 92.2 4.12 97.0 0.91 0.91 0.92
Ar 4.16 251 4.41 190 0.97 0.94 0.94
Kr 4.15 344 4.59 215 0.96 0.91 0.90
H2 3.90 166 4.21 100 0.99 0.95 0.93
N2 4.37 201 4.59 161 0.97 0.95 0.94
O2 4.10 288 4.48 169 1.00 0.94 0.92
C7H16 He 4.06 58.9 4.15 55.3 0.97 0.97 0.97
Ar 4.24 265 4.52 202 0.96 0.93 0.93
N2 4.42 213 4.70 171 0.95 0.92 0.92
a
For the exp/6 intermolecular potential
b
From Refs. 15, 23, and 24

22
Table 3. Calculated Lennard-Jones parameters based on full-dimensional trajectories for
C3Hx + He compared with tabulated values for propane + Hea
zcalc/ztab
Target σ, Å ε, cm–1 300 K 1500 K 3000 K
HCCC 3.20 42.1 0.74 0.73 0.73
HCCCHb 3.21 44.0 0.75 0.74 0.74
H2CCCH 3.24 45.1 0.77 0.76 0.76
H2CCCH2 3.28 45.7 0.79 0.78 0.78
H3CCCH 3.31 44.6 0.80 0.79 0.79
H2CCHCH2 3.35 44.7 0.82 0.81 0.81
H3CCHCH2 3.43 44.2 0.86 0.85 0.85
H3CCH2CH2 3.56 41.8 0.91 0.90 0.90
H3CCHCH3 3.55 41.5 0.91 0.90 0.90
H3CCH2CH3 3.55 43.5 0.92 0.91 0.90
a
For the exp/6 intermolecular potential
b
Hydrogenated “parent” molecules in bold may be compared with the associated
radical(s) listed above them.

23
Table 4. Calculated Lennard-Jones parameters based on one-dimensional minimizations
compared with tabulated values and relative collision rates for four normal alkanes + Hea
Calculatedb Tabulatedc zcalc/ztab
Target Method σ, Å ε, cm–1 σ, Å ε, cm–1 300 K 1500 K 3000 K
CH4 Eq. (14) 3.32 21.5 3.13 28.7 1.07 1.08 1.09
Eq. (16) 3.33–3.35 21.4–21.0 3.13 28.7 1.07 1.09 1.10
C2H6 Eq. (14) 3.64 28.9 3.40 36.8 1.08 1.10 1.10
Eq. (16) 3.63–3.66 28.8–28.3 3.40 36.8 1.08 1.11 1.12
C4H10 Eq. (14)d 4.08 37.3 3.79 44.3 1.12 1.13 1.13
e
Eq. (14) 4.07 37.1 3.79 44.3 1.11 1.12 1.13
Eq. (16) 4.06–4.09 37.9–36.7 3.79 44.3 1.11 1.12 1.13
C6H14 Eq. (14)d 4.32 45.1 4.04 51.9 1.11 1.12 1.12
Eq. (16) 4.35–4.43 43.9–42.5 4.04 51.9 1.12 1.14 1.17
a
For the exp/6 intermolecular potential
b
Via Eq. (13)
c
From Refs. 15 and 24
d
Anti conformer
e
Gauche conformer

24
Table 5. Calculated Lennard-Jones parameters compared with tabulated values and
relative collision rates for several intermolecular potentials for small targets + Hea
Calculated Tabulatedb zcalc/ztab
Target PES σ, Å ε, cm–1 σ, Å ε, cm–1 300 K 1500 K 3000 K
H exp/6 2.92 3.67 2.31 26.7 1.15 1.23 1.26
MP2/A'DZ 3.37 2.89 2.31 26.7 1.48 1.60 1.63
MP2/CBS 3.35 2.32 2.31 26.7 1.42 1.54 1.58
QCISD(T)/CBS 3.17 4.49 2.31 26.7 1.39 1.48 1.51
H2 exp/6 3.01 6.58 2.75 13.7 1.06 1.09 1.10
MP2/A'DZ 3.14 5.81 2.75 13.7 1.13 1.16 1.18
MP2/CBS 3.07 6.64 2.75 13.7 1.11 1.14 1.15
QCISD(T)/CBS 2.98 9.55 2.75 13.7 1.11 1.12 1.13
C exp/6 3.08 16.0 2.94 18.8 1.07 1.08 1.08
MP2/A'DZ 2.58 45.5 2.94 18.8 0.92 0.88 0.87
MP2/CBS 2.57 39.3 2.94 18.8 0.88 0.85 0.84
QCISD(T)/CBS 2.91 19.5 2.94 18.8 0.99 0.99 0.98
CH exp/6 3.15 17.7 2.67 19.9 1.37 1.38 1.38
MP2/A'DZ 3.10 26.9 2.67 19.9 1.44 1.42 1.41
MP2/CBS 3.09 21.1 2.67 19.9 1.36 1.36 1.36
QCISD(T)/CBS 3.02 26.9 2.67 19.9 1.36 1.34 1.34
3
CH2 exp/6 3.21 19.2 3.19 26.6 0.95 0.97 0.97
MP2/A'DZ 3.49 13.6 3.19 26.6 1.06 1.09 1.10
MP2/CBS 3.47 9.72 3.19 26.6 0.99 1.03 1.04
QCISD(T)/CBS 3.35 14.4 3.19 26.6 0.99 1.01 1.02
1
CH2 exp/6 3.21 19.2 3.19 26.6 0.95 0.97 0.97
MP2/A'DZ 3.40 20.2 3.19 26.6 1.08 1.09 1.10
MP2/CBS 3.36 16.1 3.19 26.6 1.01 1.03 1.04
QCISD(T)/CBS 3.29 18.9 3.19 26.6 1.00 1.02 1.02
O exp/6 N/A
MP2/A'DZ 2.87 21.5 2.66 19.9 1.14 1.14 1.14
MP2/CBS 2.79 12.7 2.66 19.9 1.01 1.03 1.04
QCISD(T)/CBS 2.73 18.8 2.66 19.9 1.04 1.04 1.04
OH exp/6 N/A
MP2/A'DZ 2.89 27.3 2.66 19.9 1.25 1.23 1.23
MP2/CBS 2.89 16.7 2.66 19.9 1.14 1.14 1.15
QCISD(T)/CBS 2.80 24.0 2.66 19.9 1.15 1.14 1.13
H2 O exp/6 N/A
MP2/A'DZ 3.12 29.6 2.59 53.1 1.28 1.32 1.34
MP2/CBS 3.14 14.1 2.59 53.1 1.13 1.21 1.23
QCISD(T)/CBS 3.02 21.7 2.59 53.1 1.13 1.18 1.20
a
Calculated via Eq. (14)
b
From Ref. 15

25
Table 6. Calculated Lennard-Jones parameters for several small targets + Hea
Calculated Tabulatedb zcalc/ztab
–1 –1
Target σ, Å ε, cm σ, Å ε, cm 300 K 1500 K 3000 K
H 3.35 2.32 2.31 26.7 1.42 1.54 1.58
H2 3.07 6.64 2.75 13.7 1.11 1.14 1.15
C 2.57 39.3 2.94 18.8 0.88 0.85 0.84
CH 3.09 21.1 2.66 19.9 1.36 1.36 1.36
3
CH2 3.47 9.72 3.19 26.6 0.99 1.03 1.04
1
CH2 3.36 16.1 3.19 26.6 1.01 1.03 1.04
CH3 3.48 11.3 3.19 26.6 1.02 1.06 1.07
CH4 3.41 14.9 3.13 28.7 1.05 1.08 1.09
C2 3.35 23.1 3.10 21.9 1.18 1.18 1.18
C2H 3.52 16.0 3.34 32.1 0.97 1.00 1.01
C2H2 3.55 16.8 3.34 32.1 1.00 1.03 1.04
C2H3 3.64 15.8 3.34 32.1 1.04 1.08 1.10
C2H4 3.65 17.5 3.27 37.2 1.07 1.11 1.12
C2H5 3.72 17.7 3.44 35.3 1.02 1.06 1.07
C2H6 3.73 19.8 3.40 36.8 1.07 1.10 1.11
O 2.79 12.7 2.66 19.9 1.01 1.03 1.04
OH 2.89 16.7 2.66 19.9 1.14 1.14 1.15
H2 O 3.14 14.1 2.59 53.1 1.13 1.21 1.23
O2 3.13 18.3 3.02 23.0 1.03 1.04 1.04
HO2 3.20 19.8 3.02 23.0 1.10 1.10 1.11
H2 O2 3.26 21.6 3.02 23.0 1.15 1.16 1.16
CO 3.33 14.7 3.14 22.5 1.04 1.06 1.06
HCO 3.37 16.5 3.08 49.5 0.96 1.02 1.03
H2CO 3.35 20.5 3.08 49.5 0.99 1.03 1.05
H3CO 2.95 27.9 3.27 46.3 0.73 0.75 0.76
CH2OH 3.54 17.6 3.13 45.3 1.06 1.11 1.12
CH3OH 3.48 21.1 3.10 48.7 1.07 1.11 1.12
CO2 3.30 25.0 3.17 34.7 1.01 1.03 1.03
C2O 3.32 30.7 3.20 33.8 1.05 1.06 1.06
HCCO 3.57 22.1 2.54 27.2 1.90 1.92 1.93
CH2CO 3.55 24.3 3.27 46.3 1.03 1.07 1.08
CH3CO 3.75 19.2 3.27 46.3 1.10 1.15 1.16
CH2CHO 3.68 21.0 3.27 46.3 1.08 1.12 1.14
CH3CHO 3.71 22.8 3.27 46.3 1.12 1.16 1.17
a
Calculated via Eq. (14) and using the MP2/CBS intermolecular potential
b
From Ref. 15

26
Figure Captions

Fig. 1. Isotropically averaged intermolecular potentials for three normal alkanes in He.
The circles are the results of Eq. (9) for the exp/6 potential. The solid lines show
Lennard-Jones curves based on the calculated parameters. The dotted lines show
the minimum and maximum sampled energies along r. Two sets of results are
shown for butane: one for the anti conformer equilibrium geometry (black) and
one for the gauche conformer equilibrium geometry (blue).

Fig. 2. Anisotropic exp/6 intermolecular potentials for three normal alkanes in He. The
lines show one-dimensional center-of-mass cuts for 20 different orientations. The
circles indicate the local minima obtained via Eq. (14). The triangles indicate
local minima from 20 full-dimensional trajectories. The initial conditions for the
one-dimensional minimizations and for the trajectory results shown here are not
related.

27
Figure 1

Fig. 1. Isotropically averaged intermolecular potentials for three normal alkanes in He.
The circles are the results of Eq. (9) for the exp/6 potential. The solid lines show
Lennard-Jones curves based on the calculated parameters. The dotted lines show
the minimum and maximum sampled energies along r. Two sets of results are
shown for butane: one for the anti conformer equilibrium geometry (black) and
one for the gauche conformer equilibrium geometry (blue).

28
Figure 2

Fig. 2. Anisotropic exp/6 intermolecular potentials for three normal alkanes in He. The
lines show one-dimensional center-of-mass cuts for 20 different orientations. The
circles indicate the local minima obtained via Eq. (14). The triangles indicate
local minima from 20 full-dimensional trajectories. The initial conditions for the
one-dimensional minimizations and for the trajectory results shown here are not
related.

29
References
1
G. C. Maitland, M. Rigby, E. B. Smith, and W. A. Wakeham, Intermolecular Forces:
Their Origin and Determination (Clarendon, Oxford, 1987).
2
J. O. Hirschfelder, C. F. Curtiss, R. B. Bird, R. B. Molecular Theory of Gases and
Liquids (Wiley, New York, 1954), pp. 1110, 1212.
3
N. J. Brown, L. A. J. Bastian, and P. N. Price, Prog. Energy Combust. Sci. 37 (2011)
565.
4
A. W. Jasper and J. A. Miller, J. Phys. Chem. A 113 (2009) 5612.
5
A. W. Jasper and J. A. Miller, J. Phys. Chem. A 115 (2011) 6438.
6
J. Troe, J. Chem. Phys. 66 (1977) 4745. J. Troe, J. Chem. Phys. 66 (1977) 4758.
7
T. Baer and W. L. Hase, Unimolecular Reaction Dynamics: Theory and Experiments
(Oxford University Press, New York, 1996).
8
N. J. Brown and J. A. Miller, J. Chem. Phys. 80 (1984) 5568.
9
A. Fernandez-Ramos, J. A. Miller, S. J. Klippenstein, and D. G. Truhlar, Chem. Rev.
106 (2006) 4578.
10
P. J. Dagdigian and M. H. Alexander, J. Chem. Phys. 137 (2012) 094306.
11
P. J. Dagdigian and M. H. Alexander, J. Chem. Phys. 138 (2013) 164305.
12
L. Monchick and S. Green, J. Chem. Phys. 63 (1975) 2000.
13
P. Middha, B. Yang, and H. Wang, Proc. Combust. Inst. 29 (2002) 1361.
14
P. H. Paul and J. Warnatz, Proc. Combust. Inst. 27 (1998) 495.
15
R. J. Kee, F. M. Rupley, J. A. Miller, M. E. Coltrin, J. F. Grcar, E. Meeks, H. K.
Moffat, A. E. Lutz, G. Dixon-Lewis, M. D. Smooke, J. Warnatz, G. H. Evans, R. S.
Larson, R. E. Mitchell, L. R. Petzold, W. C. Reynolds, M. Caracotsios, W. E. Stewart,
P. Glarborg, C. Wang, C. L. McLellan, O. Adigun, W. G. Houf, C. P. Chou, S. F.
Miller, P. Ho, P. D. Young, D. J. Young, D. W. Hodgson, M.V. Petrova, K. V.
Puduppakkam, CHEMKIN, Reaction Design, San Diego, CA, 2010.
16
Quantum mechanically there is some distribution of geometries associated with the
zero point wave function even at 0 K.
17
H. C. Andersen, J. Chem. Phys. 7 (1979) 1.
18
A. W. Jasper and N. Hansen, Proc. Combust. Inst. 34 (2013) 279.

30
19
N. Taxman, Phys. Rev. 110 (1958) 1235.
20
K. Chae, P. Elvati, and A. Violi, J. Phys. Chem. B 115 (2011) 500.
21
Y. Wang, H. C. Mak, Chem. Phys. Lett. 235 (1995) 37. Liu, T.; Truhlar, D. G. TB,
version 1.0.1, University of Minnesota: Minnesota, 2004, see
comp.chem.umn.edu/tbpac. Liu, T. Ph.D. Thesis, University of Minnesota, 2000.
22
W. A. Alexander and D. Troya, J. Phys. Chem. A, 110 (2006) 10834.
23
J. O. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and
Liquids (Wiley, New York, 1954), pp. 1110, 1212.
24
L. S. Tee, S. Gotoh, and W. E. Stewart, I&EC Fundamentals 5 (1966) 356.
25
L. A. J. Bastien, P. N. Price, and N. J. Brown, Int. J. Chem. Kin. 42 (2010) 713.
26
J. Klimeš and A. Michaelides, J. Chem. Phys. 137 (2012) 120901.
27
J. Řezác and P. Hobza, J. Chem. Theory Comput. 9 (2013) 2151.

31