The solvent effect on two competing reaction mechanisms involving hypervalent iodine reagents ( λ 3...
Transcript of The solvent effect on two competing reaction mechanisms involving hypervalent iodine reagents ( λ 3...
The Solvent Effect on Two Competing ReactionMechanisms Involving Hypervalent Iodine Reagents(k3-Iodanes): Facing the Limit of the StationaryQuantum Chemical Approach
Oliver Sala,[a,b] Hans Peter L€uthi,*[a] and Antonio Togni[b]
Trifluoromethylation of acetonitrile with 3,3-dimethyl-1-
(trifluoromethyl)21k3,2- benziodoxol is assumed to occur via
reductive elimination (RE) of the electrophilic CF3-ligand and
MeCN bound to the hypervalent iodine. Computations in gas
phase showed that the reaction might also occur via an SN2
mechanism. There is a substantial solvent effect present for
both reaction mechanisms, and their energies of activation are
very sensitive toward the solvent model used (implicit, micro-
solvation, and cluster-continuum). With polarizable continuum
model-based methods, the SN2 mechanism becomes less
favorable. Applying the cluster-continuum model, using a shell
of solvent molecules derived from ab initio molecular dynam-
ics (AIMD) simulations, the gap between the two activation
barriers (DDG‡) is lowered to a few kcal mol21 and also shows
that the activation entropies (DS‡) and volumes (DV‡) for the
two mechanisms differ substantially. A quantitative assessment
of DDG‡ will therefore only be possible using AIMD. A natural
bond orbital-analysis gives further insight into the activation
of the CF3-reagent by protonation. VC 2014 Wiley Periodicals,
Inc.
DOI: 10.1002/jcc.23727
Introduction
Nowadays, the CF3 group is widely used to alter physico-
chemical properties of organic and inorganic compounds.
Among other, trifluoromethylation may lead to increased lipo-
philicity, enhancing the uptake of the compound into biologi-
cal systems.[1] In recent years, hypervalent iodine compounds
such as 3,3-dimethyl-1-(trifluoromethyl)-1k3,2-benziodoxol (rea-
gent 8, abbreviated DMTB) (Fig. 1) have become important
reactants in the field of organic and inorganic chemistry.[2] The
reagent transfers the electrophilic trifluoromethyl-group (CF3)
to a vast array of nucleophiles.[3] A novel unexpected Ritter-
type reaction, in which N-trifluoromethyl amidines are
obtained by N–CF3 bond-formation, has been discovered in
our laboratories.[4] The elucidation of the reaction mechanism
in view of the optimization of the reagent and/or the reaction
conditions is the main topic of this article.
DMTB and its model compound DMTI (1) are also referred
to as k3-iodanes. The superscript denotes the number of
ligands bound to the hypervalent iodine atom. The ligands of
k3-iodanes are typically arranged such that the most electro-
negative ligands are trans to each other, and the least electro-
negative ligand is positioned orthogonally, resulting in a
T-shaped structure.[5] In the case of DMTB and DMTI, the CF3
carbon atom, the iodine and the oxygen atom form a 3-
center-4-electron (3c4e) bond which is relatively weak and
thus accounts for the high reactivities of these reagents. In
practice, the reactivity can be further enhanced through proto-
nation of the reagent by a catalytic amount of Br�nsted acid.
In the Ritter-type reaction (Fig. 2a), acetonitrile (MeCN) is N-tri-
fluoromethylated by the protonated DMTB, H-DMTBe1 (9),
through a 1,2-shift giving rise to the intermediate 12. Subse-
quent coordination of benzotriazole (13) to the iodine atom fol-
lowed by a reductive elimination (RE) yield the uncommon N-
(trifluoromethyl)-amidine 17 and the iodoalcohol 16. The forma-
tion of the N-trifluoromethyl (aceto-) nitrilium species, whether
bound to the reagent (12) or unbound (6, Fig. 2b), marks the
rate determining step. Figure 2b shows that 6 can be formed
either by RE or bimolecular nucleophilic substitution (SN2). Both
reaction mechanisms lead to the same products, the nitrilium
cation 6 and the iodoallyl alcohols 7a and 7b.
The RE reaction pathway proceeds through a transition state
(TS) in which the two ligands (CF3 and MeCN, or, benzotriazole
and N-trifluoromethyl nitrilium ion) and the iodine atom are
arranged in a triangular geometry (compare 4, 11, and 14 in
Fig. 2). These TSs are formed after coordination of the sub-
strates to the reagent giving rise to the complexes 3, 10, and
14. Previous theoretical and experimental studies have focused
on the RE mechanism, which is assumed to be the prevailing
mechanism involving k3-iodanes.[6,7] Nevertheless, preliminary
computational studies have shown that, at least in gas phase,
the SN2 mechanism is preferred (left side in Fig. 2b).[8,9]
[a] H. P. L€uthi
ETH Zurich, Laboratorium f€ur Physikalische Chemie, Vladimir–Prelog–Weg
2, 8093 Zurich, Switzerland
E-mail: [email protected]
[b] A. Togni
ETH Zurich, Laboratorium f€ur Anorganische Chemie, Vladimir–Prelog–Weg
2, 8093 Zurich, Switzerland
Contract grant sponsor: Swiss National Science Foundation SNF (Project
137712); The authors acknowledge the support of Dr. Marcella Iannuzzi
(University of Zurich) for the assistance with the ab initio-MD
simulations)
VC 2014 Wiley Periodicals, Inc.
2122 Journal of Computational Chemistry 2014, 35, 2122–2131 WWW.CHEMISTRYVIEWS.COM
FULL PAPER WWW.C-CHEM.ORG
Even though the RE mechanism is considered to be the prevail-
ing mechanism, from an experimental point of view an SN2 mecha-
nism can not be excluded based on nucleofugality arguments.[10–12]
Provided the case, however, that the electrophilic CF3-carbon
would be stereogenic, the two reaction mechanisms would
give rise to two different products. While retention of the con-
figuration is expected for RE, inversion would result from an
SN2 reaction mechanism. A mechanism involving free CF3 radi-
cals would possibly lead to loss of stereochemical information.
Since this trifluoromethylation reaction was never realized in
gas phase, the question arises if the solvent—here MeCN or
CH2Cl2 dichloromethane (DCM)—has a major influence on the
reaction mechanism. In a situation with two competing reaction
mechanisms involving ionic reactants in a dipolar solvent, the
solvent effect is expected to be crucial. Not only should there
be a difference in reactivity and selectivity between gas phase
and condensed phase but also between solvents. Also note that
if performed in acetonitrile, the reaction actually is a solvolysis.
There are two general possibilities to represent a solvent in
quantum chemical calculations, namely, the explicit description
of the solvent molecules, or their representation in terms of a
reaction field by applying a polarizable continuum model
(PCM). The possibility of hydrogen-bond formation and interac-
tion with the 3c4e-bond of the reagent with the solvent limits
the applicability of a PCM-based solvent model. Alternatively,
other solvation models could be used, as described, for exam-
ple, by Truhlar and coworkers and Marzari and coworkers.[13,14]
A fully explicit treatment of the solvent, conversely, calls for an
ab initio Molecular Dynamics (AIMD) simulation. Within a sta-
tionary (time-independent) quantum chemical treatment, a par-
tially explicit approach (microsolvation) may still be sufficient to
account for the most important solute–solvent interactions.
Finally, microsolvation can be combined with a PCM-based
method to include the bulk of the solvent, which is referred to
as the cluster-continuum model. Despite the fact that iodanes
have become important reactants, there are only few theoretical
Figure 1. Left: Structures of DMTB and the model compound DMTI (3,3-dimethyl-1-(trifluoromethyl)-1k3,2-iodoxol) (reagent 1) applied in this study (See
Supporting Information section comparison between model compound and reagent for the eligibility of this simplification). Right: Idealized MO diagram of
a 3c4e-bond, in which the bonding and the nonbonding (nb) MO’s are doubly occupied.
Figure 2. a) During the novel Ritter-type reaction MeCN is N-trifluoromethylated and attacked subsequently by benzotriazole (13). Instead of an insertion
reaction (1,2-shift) a RE can occur, leading to 6, which is attacked subsequently by benzotriazole. b) The reaction sequence investigated in this study shows
the two competing mechanisms out of which the RE is commonly accepted as being the most likely one.
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 2122–2131 2123
investigations concerning the reactivity, selectivity, and reaction
mechanisms of the reagents mentioned above.[6,7,15]
In this study, the two competing reaction mechanisms (RE
and SN2) are investigated. The solvent effect is modeled by
applying a PCM method, microsolvation, as well as the cluster-
continuum model. The effect of the entropy is discussed, fol-
lowed by a consideration of the influence of other solvents.
Additionally, we will explore the response of structure, bonding
and reactivity to solvation and as well to reagent protonation
by means of a natural bond orbital (NBO) analysis. The following
principal questions are addressed: (a) is there a solvent effect
which discriminates between the two limiting reaction mecha-
nisms? and (b) is a stationary quantum chemical approach
adequate for qualitative or even quantitative predictions?
Computational Details
All equilibrium and TS structures were optimized using the
B3LYP,[16–18] BP86,[19,20] and M06-2X[21] density functional methods
using the Dunning augmented correlation consistent basis sets
(double- and triple-f),[22,23] along with the pseudopotential Stutt-
gart-K€oln-MCDHF-RSC-28-ECP[24] for iodine. We also performed sec-
ond order M�ller–Plesset perturbation theory energy calculations
(MP2/(aug-)cc-pVXZ-pp//BP86/aug-cc-pVTZ-pp; X: D,T). The solvent
effect was represented utilizing a polarizable continuum model (IEF-
PCM) with default parameter settings,[25–28] microsolvation and the
cluster-continuum model.[29] A discussion of error bars of solvation
free energies concerning these models is found in the litera-
ture.[12,29,30] All structures were optimized with the corresponding
solvent model. For comparison, computations with the SMD solvent
model[30,31] were carried out (SMD keyword in Gaussian). All calcula-
tions were performed with the Gaussian G09 suite of programs.[32]
The TS were searched by following the negative eigenvalue vector
of guess structures (TS keywords in G09), followed by an internal
reaction coordinate analysis. We report free energies DG obtained
for 298.15 K and 1 atmosphere with harmonic frequencies.
The counterpoise correction for basis set superposition
shows a systematic shift of all energies (Supporting Informa-
tion section basis set superposition error correction); therefore,
the energies presented here refer to the uncorrected values.
To explore the structure of the first solvation shell of H-
DMTIe1 (2) in MeCN, AIMD simulations were performed with
the Becke and Lee–Yang–Parr (BLYP)[16,17] exchange correlation
functionals as implemented in Quickstep,[33] which is part of
the CP2K program (version 2.5.13191).[34] Goedecker–Teter–Hut-
ter (GTH) pseudopotentials[35] were used with a split valence
Gaussian basis set of double-f quality, designed specifically for
these pseudopotentials, and with polarization functions for all
atoms.[36] The Born–Oppenheimer AIMD simulations were per-
formed with a time step of 0.5 fs in the isothermal-isobaric
ensemble using an isotropic, cubic cell (NPT-I) at 300 K with
periodic boundary conditions. The density cutoff for the Gaus-
sian and plane wave basis was set to 500 Ry.
For the analysis of the electronic structure of the various
stationary points an NBO analysis was made (version 5.9 imple-
mented in G09 release C1), applying the M06-2X/aug-pVDZ-pp
method.[37]
In absence of experimental data or high quality benchmark
values—CCSD(T)/TZ calculations were attempted unsuccess-
fully—we will monitor the results of the calculations using the
methods introduced above to ensure consistency of the pre-
dictions. Regarding the consistency of the results, we observe
that in gas and in condensed phase (acetonitrile, IEF-PCM)
B3LYP and BP86 with an aug-cc-pVTZ-pp basis set predict simi-
lar relative energies for all stationary points. For activation
energies DG‡ the maximal deviations between the two meth-
ods for the two TS amount to 2.0 and 1.7 kcal mol21. In gen-
eral, the M06-2X functional is in close accordance with B3LYP
(see Supporting Information Tables 4 and 5).
A reduction of the basis set to aug-DZ quality leads to a
maximal deviation for DG‡ of 2.6 kcal mol21 (BP86, SN2). In
this article, the terms activation energy, activation barrier and
barrier height are used interchangeably.
The effect of dispersion on the computed activation energies as
determined by the B3LYP and BP86 D2 functionals is a systematic
shift to higher values for both TSs (max. B3LYP 3.2 kcal mol21).
In view of the activation barriers discussed here—25 to 40 kcal
mol21, depending on the method and mechanism—the devia-
tions (error bars) discussed above are quite significant. Still, for
clarity, we will focus on the results obtained using the BP86/aug-
cc-pVDZ-pp method. Fortunately, the conclusions of this study
do not depend on the method used. Major deviations between
methods, whenever occurring, will be mentioned in the text.
Results and Discussion
The SN2 and RE reaction pathways in gas phase
As seen from Figure 3 and Table 1, the calculations confirm a
preliminary study[7] that in gas phase the SN2 reaction mecha-
nism is (kinetically) favored over the RE: the BP86 calculations
predict an SN2 activation barrier of 26.1 kcal mol21, and as
much as 35.7 kcal mol21 for the RE (see also Figure 6 and Sup-
porting Information Figure 5). The structures of the two TSs
are shown in Figure 4.
In gas phase, the reaction is slightly endergonic (DRG5
3:6 kcal mol21). There are two conformations of the iodo-
allylic alcohol 7: the s-cis 7a and the s-trans 7b conformer,
which in most cases is somewhat lower in energy than 7a
(Fig. 2b). The values in parentheses in Table 1 (and in all
other tables giving reaction energies) refer to the energy of
the reaction yielding 7b. The predictions of the other meth-
ods are at least qualitatively the same [see Supporting Infor-
mation section double-f calculations in the gas phase and in
acetonitrile (IEF-PCM)].
The coordination of MeCN to the iodine atom of the proto-
nated reagent leads to an equilibrium prior to the formation
of the two TSs. The complex with the acetonitrile coordinated
to the iodine atom is favored by 5.2 kcal mol21 (See Figs. 3
and 4). Given the fact that the coordinated and uncoordinated
reactants interconvert rapidly, the Curtin–Hammett principle
applies. This means that the product ratio will not only be
determined by the difference between the respective activa-
tion energies DDG‡ (DG‡RE2DG‡
SN2), but that it will also depend
on the free energy difference DG between the interconvertible
FULL PAPER WWW.C-CHEM.ORG
2124 Journal of Computational Chemistry 2014, 35, 2122–2131 WWW.CHEMISTRYVIEWS.COM
reactants. Applying the Curtin–Hammett principle to our case,
this means that the course of the reaction is determined by
the absolute difference DDG‡CH between the RE and the SN2
activation energies (Fig. 3). Therefore, the activation energies
are measured from the lowest stationary point on the poten-
tial energy surface (PES), that is, H-DMTIe1-MeCN in Figure 3.
Remember, that here the reaction products are identical.
Alternatively, the difference between the activation energies
can be also calculated without taking into account DG2;3 (Table
1, non-Curtin–Hammett). The energy released on coordination
of MeCN to iodine is assumed to fully dissipate to the bulk sol-
vent (surrounding the solute) and, therefore, is not available any
more for reaching the TS. In essence, the DDG‡ values become
smaller in the condensed phase and larger in gas phase, which
leads to the conclusion that the RE mechanism will become
more favored. Still, the SN2 barriers remaining lower. In contrast
to the Curtin–Hammett scheme, BP86 now also predicts a
smaller DDG‡ in solution. Note, that the main conclusions are
not affected by a non-Curtin–Hammett consideration.
The protonation of DMTI prior to the trifluoromethylation
reaction is crucial in view of the optimization of the reaction
conditions (Fig. 2).
Activation of the reagent by protonation. On protonation a
change in the energetic order of the frontier orbitals is observed:
in the protonated form, the LUMO changes from an IACvin to an
IACF3 antibonding orbital (Fig. 5). Inspection of the natural
atomic charges also shows that the protonation of the reagent
leads to a substantial increase of the charge of the CF3 carbon
atom from 0.82 to 0.99, indicating a higher electrophilicity of the
group. This also affects the nature of the 3c4e-bond and offers
an explanation for the enhanced reactivity of the protonated rea-
gent toward electron donating reactants.
The analysis of the 3c4e-bond in terms of second-order per-
turbation theory of NBO donor-acceptor interactions DEð2Þij
[38]
Table 1. Free energies (DG) in kcal mol21 at 298 K and 1 atm. with BP86/
aug-cc-pVDZ-pp.
Mechanism Solvent model DG‡CH DRG DG‡
non-CH
SN2 Gasphase 26.1 3.6 (2.2) 20.9
RE Gasphase 35.7 3.6 (2.2) 35.7
DDG‡ Gasphase 9.6 – 14.8
SN2 MeCN IEF-PCM 26.0 211.6 (212.3) 26.0
RE MeCN IEF-PCM 36.9 211.6 (212.3) 36.4
DDG‡ MeCN IEF-PCM 10.9 – 10.4
SN2 MeCN SMD 31.2 210.0 31.2
RE MeCN SMD 42.6 210.0 40.2
DDG‡ MeCN SMD 11.4 – 9.0
The difference between the activation energies DDG‡ is obtained by
assuming both a Curtin–Hammett (CH) and a non-Curtin–Hammett
situation.
Figure 4. The (pre)-equilibrium between 2 and 3 and the structures of the
two TSs in the gas phase as outlined in Figure 2b are shown. TS-RE intera-
tomic distances: NACCF3: 2.493 A, CCF3 AI: 3.447 A, NAI: 3.164 A, IAO:
3.139 A. TS-SN 2 interatomic distances: NACCF3: 2.130 A, CCF3AI: 3.171 A,
IAO: 3.040 A (BP86/aug-cc-pVDZ-pp).
Figure 3. The energy diagram (E 5DG) of the two competing reaction mechanisms as outlined in Figure 2b is shown with the preequilibrium between H-
DMTIe1 (2) and H-DMTIe1-MeCN (3) depicted in Figure 4. The energy of the reactants H-DMTIe1 and acetonitrile defines the reference point on the energy
scale. The assignment of the corresponding activation barriers (DG‡CH) and difference (DDG‡
CH) in a Curtin–Hammett (CH) scheme is outlined. The index CH
is used only in this figure for clarity (298 K, 1 atm.). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 2122–2131 2125
shows that in DMTI there is a strong orbital interaction between
the oxygen nonbonding electron pair and the antibonding
IACF3 orbital (nO ! r?I2CF3) in DMTI giving rise to a 3c4e-bond.
The interaction is reduced drastically on protonation of DMTI
from 89.2 to 10.6 kcal mol21. This goes along with a reduction
of the IAO natural bond order from 0.65 to 0.05. At the same
time the bond dissociation energy (BDE) drops from 58.6 to 15.6
kcal mol21. For the IACF3 bond, conversely, we observe a slight
increase of the bond order from 0.84 to 0.90 and of the BDE
from 37.1 to 44.1 kcal mol21. Hence, after protonation, the
F3CAIAO bond of H-DMTIe1 (2) is best described as consisting
of an I–CF3 bond and a weak IAO interaction.
More details (natural resonance theory analysis, BDE, frontier-
orbital interaction diagram) can be found in the Supporting Infor-
mation (section structure, bonding and reactivity of the reagent).
SN2 and RE reaction pathways in solution (PCM-based
methods)
Reaction in Acetonitrile. For the reaction in solution we find
that relative to the gas phase the SN2 activation energy DG‡
remains unchanged, but that the RE barrier is increased slightly:
by 1.2 to 36.9 kcal mol21 (Table 1). The B3LYP (and all other, see
Supporting Information section method validation and double-
zeta calculations in the gas phase and in acetonitrile (IEF-PCM))
calculations, however, do not support this trend and even pre-
dict a significantly lower activation barrier for the RE mechanism.
Still, the SN2 reaction pathway remains favorable also in solution,
even though—with one exception—the DDG‡ has changed in
favor of the RE mechanism. In a non-Curtin–Hammett scheme,
exception lifted, all DDG‡ values are decreased significantly in
condensed phase (Table 1; Supporting Information Table 5).
The solvation energies DG‡solv (PCM) of both TSs are very simi-
lar; the DDG‡solv (DGRE
solv2DGSN 2solv) amounts to 11.3 kcal mol21, in
close accordance to the SMD method (Table 2). At a first glance,
Figure 5. The three lowest unoccupied (in terms of NBO occupancies) natural
bond orbitals in solution. LUMO12 of DMTI corresponds to U? in the 3c4e-
bond representation shown in Figure 1; U and Unb are not listed (see Support-
ing Information section structure bonding, reactivity of the reagent). Note the
change in LUMO orbital character on protonation of the reagent in solution.
Figure 6. Free energy (DG) diagram of SN2 and RE in gasphase and in
MeCN (IEF-PCM and SMD) (298 K and 1 atm.). Since the products of both
mechanisms are the same, the reaction profile, compared to Figure 3, is
presented in a more compact form. The energies calculated with the PCM-
method SMD are depicted in yellow. The first step of the reaction is the
protonation of DMTI. The product H-DMTIe1 (together with x noninteract-
ing MeCN; x 5 1–3; here x 5 1) is set as reference point on the energy
scale. From this species either an SN2 occurs (following the black lines), or
a MeCN coordinates to the iodine atom leading to a RE pathway (red
lines). The relative energy of the product 7b is in parentheses. (B3LYP, M06-
2X, and MP2: Figure 4 in Supporting Information). [Color figure can be
viewed in the online issue, which is available at wileyonlinelibrary.com.]
Table 2. Free energy of solvation, differences between free energies of
solvation DDGsolv of different species and the gap between the two acti-
vation barriers of the competing reaction pathways in MeCN in kcal
mol21 at 298 K and 1 atm.
DGsolv
aug-cc-pVDZ-ppaug-TZ
PCM SMD PCM[a] PCM
DMTI (1) 25.4 – 24.5 –
H-DMTIe1 (2) 245.5 257.5 245.4 247.9
H-DMTIe1-MeCN (3) 239.6 249.9 241.1 239.8
TS-SN 2e1 (5) 240.2 247.2 238.6 239.8
TS-REe1 (4) 238.9 245.4 237.9 240.3
Products 6 1 7a 260.5 259.8 – 260.2
Products 6 1 7b 259.8 269.7 259.3 –
DDGsolv 1 DDG‡solv
2 versus 3 25.9 27.6 24.3 28.1
4 verus 5 1.3 1.8 0.7 20.5
[a] BP86-D2.
Table 3. Free energies of activation and solvation in kcal mol21 of vari-
ous solvents (298 K and 1 atm. IEF-PCM/B3LYP/aug-cc-pVDZ-pp).
Solvent Dielectric const. (e)
DG‡
DDG‡
DG‡solv
SN2 RE SN2 RE
None 1 (gas phase) 24.6 39.7 15.1
Toluene 2.37 25.6 38.1 12.5 221.0 223.6
Acetic acid 6.25 25.9 36.0 10.1 232.4 237.4
Dichloromethane 8.93 26.2 35.9 9.7 234.8 240.2
Nitroethane 28.29 25.4 34.0 8.6 238.8 245.3
Acetonitrile 35.69 25.1 34.1 9.0 239.6 245.7
DMF 37.22 25.6 33.3 7.7 239.2 246.6
Water 78.36 26.0 35.5 9.5 240.1 245.7
Formamide 108.94 26.5 35.8 9.3 240.2 246.0
NMFM 181.56 26.5 35.3 8.8 240.5 246.8
FULL PAPER WWW.C-CHEM.ORG
2126 Journal of Computational Chemistry 2014, 35, 2122–2131 WWW.CHEMISTRYVIEWS.COM
it appears that the BP86 functionals predict similar polarities for
both TSs applying these two solvent models. However, all other
functionals used predict the solvation energy of the RE transition
structure to be greater than the SN2 one (e.g., B3LYP: DDG‡solv5
26:1 kcal mol21, Supporting Information Table 4).
In condensed phase, the reaction becomes exergonic, lower-
ing the energy of reaction DRG to around 212 kcal mol21. The
interatomic distance N–CCF 3 increases by 0.31 and 0.16 A in
the TS-RE and TS-SN2 structure, respectively. Whereas the IACCF3
distance increases by 0.2 A in the TS-RE, the distance decreases
slightly (0.06 A) in the TS-SN2 structure. All other bond lengths
are not affected significantly on solvation in MeCN.
Solvent Screening. Given the polar reactants and TSs, the
response of the activation energy to solvent polarity may be con-
siderable. In DCM, which is frequently used for trifluoromethyla-
tion reactions with the reagent 8 and which is much less polar
than acetonitrile (�58:93 vs. 37.22), the computed activation
energy DDG‡ increases from 9.0 to 9.7 kcal mol21 (Table 3).
If, however, the dielectric constant � increases, the activation
energy of the SN2 mechanism is increased as well, while the
barrier of the RE is lowered. Interestingly, this behavior just
holds true until an � of around 36 is reached. A further increase
of � has no longer a significant impact on DDG‡ (Table 3).
Microsolvation and cluster-continuum modeling
In this section, the solvent is treated partially explicitly, to also
account for orbital interactions between the solute and the
solvent. The results of the B3LYP calculations in the regime of
microsolvation and cluster-continuum model are presented in
the Supporting Information.
To identify those solute–solvent interactions that request an
orbital-based description, we analyzed the trajectory of the
AIMD simulation of the protonated reagent H-DMTIe1 in
(explicit) acetonitrile solution. This reveals two preferred coordi-
nation sites of the solute: for an acetonitrile bound to the rea-
gent OH group, or an acetonitrile bound to the iodine atom, we
find a radial distribution function g(r) of 3.5 and 1.6, respectively
(Fig. 7). There is no other coordination site showing such high
probabilities of finding a solvent molecule being coordinated.
At first, one MeCN solvent molecule is coordinated to the OH-
group via hydrogen-bonding, or to the iodine atom, forming a k4-
iodane cluster (1:1 adduct). Next, two explicit solvent molecules
will be coordinated to the reagent giving rise to a k5-iodane clus-
ter (1:2 adduct) or a mixed k4-iodane cluster (1:2 adduct). Finally,
the bulk solvent effect was modeled using IEF-PCM for both clus-
ters, an approach referred to as cluster-continuum model.[29]
The free energy and entropy of solvation (DG�solv;DS�solv) apply-
ing the cluster-continuum model are calculated as outlined by
Pliego and Riveros (Table 4).[29] Note that the superposition of
Figure 7. Radial distribution function g(r) N to I (violet) and N to O (red),
obtained by AIMD simulation of 104 ps with 0.5 fs time steps at 300 K,
NPT ensemble, applying periodic boundary conditions (PBC), in a cubic box
of 19.2 A length (BLYP/DZVP-MOLOPT-SR-GTH). [Color figure can be viewed
in the online issue, which is available at wileyonlinelibrary.com.]
Table 4. Free energy of solvation DG�solv in kcal mol21 and entropy of solvation DS�solv[a] at 298 K and 1 atm. of H-DMTIe1 and the transition states TS-SN
2e1 and TS-REe1 (BP86/aug-cc-pVDZ-pp).
Species[b]
Free energies / kcal mol21 Entropies / cal mol21 K21
DG�
clust DG�solv nDGvap DG�solv (H-DMTIe1) DS�
clust DS�solv nDSvap DS�solv (H-DMTIe1)
H-DMTIe1 245.38 245.38 21.67 21.67
H-DMTIe1 (MeCN)1 25.22 239.55 1.76 243.01 226.54 21.48 14.21 213.81
H-DMTIe1 (MeCN)1 210.06 238.82 1.76 247.12 222.57 23.93 14.21 212.29
H-DMTIe1 (MeCN)2 24.37 239.58 3.52 240.43 250.18 3.06 28.41 218.71
H-DMTIe1 (MeCN)2 211.00 236.14 3.52 243.62 253.70 21.15 28.41 226.44
H-DMTIe1 (MeCN)3 27.76 239.59 5.28 242.07 282.71 10.41 42.62 229.68
Transition states DG�
clust DG�solv nDGvap DG�solv (TSe1) DS�
clust DS�solv nDSvap DS�solv (TSe1)
TS-SN 2e1 240.20 240.20 3.50 3.50
TS-SN 2e1 (MeCN)I1 20.15 237.99 1.76 236.38 220.17 20.39 14.21 26.35
TS-SN 2e1 (MeCN)OH1 20.98 239.06 1.76 238.28 227.24 5.33 14.21 27.70
TS-SN 2e1 (MeCN)OH;I2 0.49 229.53 3.52 225.52 244.53 242.67 28.41 258.79
TS-REe1 238.90 238.90 2.20 2.20
TS-REe1 (MeCN)I1 1.74 238.21 1.76 234.71 224.15 21.52 14.21 211.46
TS-REe1 (MeCN)OH1 21.73 236.60 1.76 236.57 226.87 22.62 14.21 215.28
TS-REe1 (MeCN)OH;I2 0.89 237.67 3.52 233.26 251.88 4.35 28.41 219.11
[a] Entropies (DS) in gas phase for H-DMTIe1 , TS-SN 2 and TS-RE: 119.13, 153.55 and 160.84 e.u., respectively. [b] Sequence of coordination: to I, OH,
twice to I, once to I and OH each and twice to I once to OH. To calculate DS‡ (vide infra, section microsolvation and cluster-continuum modeling), the
largest absolute DS�solv values were taken (bold numbers).
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 2122–2131 2127
microsolvation and PCM requires corrections for the solvent mole-
cules considered explicitly. The largest absolute value of the free
energy of solvation DG�solv represents the closest result to the
experimental value (bold number in Table 4). It remains unclear if
this observation is applicable to TSs. Nonetheless, for H-DMTIe1
this is an indication that the solvent coordination to the OH group
alone is sufficient (largest absolute DG�solv-value: 247.12 kcal
mol21). For the two TSs the largest DDG�solv-value of 27.74 kcal
mol21 is obtained by twofold solvent coordination, which is, there-
fore, necessary to discriminate effectively between the two TSs.
After coordination of the explicit solvent(s), following the
reaction pathways as before, one more MeCN molecule—act-
ing as reactant—is coordinated to the iodine atom for the RE.
For the SN2 mechanism no binding interaction between the
reactant and the reagent was found as this is the case for the
RE mechanism.
Using microsolvation and assuming a Curtin–Hammett
scheme, DDG‡ amounts to 9.0 kcal mol21 in favor of the SN2
mechanism for the reaction with one explicit solvent molecule
coordinated to the reagent OH-group, evaluated using the
BP86 functional. This compares to a DDG‡ of 10.9 kcal mol21
for the IEF-PCM/BP86 method. With explicit solvent coordina-
tion at the iodine atom, the SN2 mechanism is favored by 11.4
kcal mol21. Coordination of two explicit solvent molecules
does by no means show an additive effect: the SN2 mecha-
nism is still favored by 10.0 kcal mol21 (Table 5 and Fig. 8).
Adding the effect of the bulk solvent applying the cluster-
continuum model, both singly coordinated reagents (to the
OH group or to the iodine atom) do not change DDG‡ signifi-
cantly. The SN2 mechanism remains favored by as much as
11.3 kcal mol21, in both cases. However, a substantial decrease
of DDG‡ to 1.8 kcal mol21 is obtained by applying the cluster-
continuum model with two MeCN coordinated to both sites
(Table 5, Fig. 10).
Obviously, a thorough solvent description requires both,
explicit orbital interaction as well as the representation of the
bulk. The results obtained so far lead to the conclusion that
the solvent effect is substantial, leading to discriminative solva-
tion energies for both TSs. Hence, the two mechanisms are
affected differently by the influence of the solvent. When using
a PCM-based description, DDG‡ is reduced from roughly 15 to
9 kcal mol21 (B3LYP) (Table 6). The same trend is observed for
microsolvation. The combination of the two models to a
cluster-continuum approach leads to further significant reduc-
tion of DDG‡, meaning that the barriers are rather close in
energy. It is very important to note that all these considera-
tions are based on the free energy differences between the
two mechanisms. The reason for evolution may therefore also
be due to entropy effects. We, therefore, decided to take into
account entropy effects explicitly.
Table 5. Free energies (DG) in kcal mol21 of the microsolvation and the cluster-continuum solvent model. Notice the difference between the two free
energies of solvation of the transition states DDGzsolv applying the cluster-continuum solvent model, which amounts to 27.8 kcal mol21.
Curtin–Hammett mechanism Solvent model
DG‡ DG‡solv
1-OH 1-I 2 1-OH 1-I 2
SN 2 Microsolv. 30.9 25.9 32.4
RE Microsolv. 39.9 37.3 42.4
DDG‡ Microsolv. 9.0 11.4 10.0
SN 2 Cluster-continuum 29.7 29.8 42.5 238.3 236.4 225.5
RE Cluster-continuum 41.0 41.1 44.3 236.6 234.7 233.3
DDG‡ and DDG‡solv Cluster-continuum 11.3 11.3 1.8 1.7 1.7 27.8
Non-Curtin–Hammett
SN 2 Microsolv. 30.0 25.9 32.4
RE Microsolv. 39.9 36.7 39.2
DDG‡ Microsolv. 9.9 10.8 6.8
SN 2 Cluster-continuum 29.7 27.5 39.0 238.3 236.4 225.5
RE Cluster-continuum 37.5 36.2 39.3 236.6 234.7 233.3
DDG‡ and DDG‡solv Cluster-continuum 7.8 8.7 0.3 1.7 1.7 27.8
Hence, if DDG‡solv is significant, a considerable solvent effect is observable. (298 K, 1 atm., BP86/aug-cc-pVDZ-pp).
Figure 8. Free energy (DG) diagram of the SN2 and RE mechanisms apply-
ing microsolvation (upper part) and the cluster-continuum model (lower
part) with one explicit MeCN solvent molecule bound the OH-group (black)
or to the iodine atom (violet). (298 K, 1 atm., BP86/aug-cc-pVDZ-pp). In a
first step DMTI is protonated (not shown here), followed by coordination
of one of the two MeCN to the OH-group or iodine atom. From these spe-
cies an SN2 can occur with the second MeCN (lines to the SN2-TS), or the
second MeCN first coordinates to the iodine atom, followed by RE (lines to
the RE-TS). Two MeCN molecules are coordinated to the iodine atom in
the case of the RE pathway (Figure 9, right hand side). [Color figure can be
viewed in the online issue, which is available at wileyonlinelibrary.com.]
FULL PAPER WWW.C-CHEM.ORG
2128 Journal of Computational Chemistry 2014, 35, 2122–2131 WWW.CHEMISTRYVIEWS.COM
The influence of the entropy
According to the Eyring equation,[39,40] the reaction rate con-
stant k will depend on the activation entropy DS‡, that is, the
larger DS‡, the larger k.
k5kBT
hexp 2
DG‡
RT
� �5
kBT
hexp
DS‡
R
� �exp 2
DH‡
RT
� �
Applying the cluster-continuum model, the entropies of activa-
tion amount to 52.3 and 5.3 cal mol21 K21 for the RE and the SN2
reaction pathway, respectively. Apparently, the RE mechanism is
entropically favored by as much as DDS‡5 47 e.u. (entropy units;
cal mol21 K21) in acetonitrile solution. The DS‡ values are derived
from the entropies of solvation DS�solv presented in Table 4. Accord-
ing to the Eyring equation, as long as DDH‡ < 14.1 kcal mol21, the
RE rate constant will be larger than that for the SN2 process. With
the BP86 functional a DDH‡ of 11.8 kcal mol21 is obtained, leading
to kRE > kSN2 and a negative DDG‡ of 22.4 kcal mol21.
The large and positive activation entropy for the RE in solu-
tion indicates that the electrostriction in the RE will be less pro-
nounced, a consequence of a more shielded charge distribution
in the TS. Taking the difference of the polarity of the two TSs
into account, a discriminatory effect of electrostriction is
expected. In terms of dipole moments, the SN2 TS [11.0 (g), 11.8
(PCM) Debye] is much more polar than the RE TS [3.98 (g), 8.5
(PCM) Debye]. With regard to the dielectric saturation effect
(solvent dipole ordering), there are no conclusions possible
based on this model. Furthermore, the distinctly positive DS‡
for the RE also points at a unimolecular reaction mechanism.
In gas phase, DDS‡ amounts to 7.3 e.u. (DS‡S N 2; ðgÞ5226.0 e.u.
and DS‡RE;ðgÞ5218.8 e.u.). This is not sufficient to compensate for
the enthalpic contribution, hence, the SN2 mechanism is favored.
Figure 9. Structures of stationary points on the PES in gas phase (left: microsolvation at the OH group; right: at the iodine atom). The energy of the
hydrogen-bond between the nitrogen of the acetonitrile and OH-group of the reagent amounts to 10.1 kcal mol21 which is in the range of an ordinary
hydrogen-bond (4–14 kcal mol21). The coordination of the second MeCN to iodine is perpendicular to the NAIACvin - and the CF3AIAO “bond”. [Color fig-
ure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
Figure 10. Left: Free energy (DG) diagram of the SN 2 and RE pathways applying microsolvation (upper part) and the cluster-continuum model (lower part)
with two explicit MeCN solvent molecules coordinated to iodine and to the OH-group (298 K, 1 atm.). Right: Structures of stationary points on the PES in
gas phase (microsolvation). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 2122–2131 2129
Applying PCM, the activation entropies are now positive (DS‡RE:
45.5 and DS‡SN 2: 39.5 e.u.), but still not entropically discriminating
between the two TSs as observed for the cluster-continuum
model. The reduction of DDG‡ observed from gas phase to PCM
condensed phase is, therefore, governed by the enthalpies.
B3LYP confirms this observation, however, without quantita-
tive agreement. The importance of entropic effects, however,
is evident. Therefore, a comprehensive description of the
entropy effect is required to obtain quantitatively correct
results. However, increasing the number of solvent molecules
in the first solvation shell is excluded, due to the convergence
difficulties encountered and sampling the phase space is
extremely cumbersome. Furthermore, the use of the harmonic
approximation is debatable. Anharmonic corrections for the
lower vibrational modes may have an influence on the com-
puted entropy.[29]
This demonstrates the necessity for an integral description
of the entropy and the limits of the stationary approximation.
This means that AIMD has to be applied to improve the
description of the entropy effect.
The activation volumes DV‡ for the two mechanisms, in partic-
ular DDV‡, corroborate the conclusions derived from the effect
of entropy. A difference in activation volumes of 244 cm3 mol21
was found by AIMD simulations. Hence, the solvent effect is
stronger for the SN2 mechanism due to larger electrostriction.
The positive signs of both activation volumes (DV‡S N 25197 cm3
mol21 and DV‡RE5153 cm3 mol21) indicate that the reaction
mechanisms experience concertedness in the rate determining
step and that both rate determining steps are unimolecular. In
sum, the CF3AI and IAO bond elongations in the SN2 TS in com-
bination with the specific geometrical arrangement of the CF3-
group and MeCN are such that the structural contribution to the
activation volume is larger than in the RE-TS. Since in solution
both TSs are more polar than the reactants an additional envi-
ronmental volume change term must be taken into account,
resulting from solute–solvent interactions. This additional term
accounts for electrostriction and dielectric saturation and is also
larger for the SN2 TS than for the RE TS. The magnitude of DDV‡
supports the obvious change in mechanism.
Conclusions
This study shows that there is a substantial solvent effect on
both of the two limiting mechanisms for the reaction of rea-
gent 1 with acetonitrile, and that the predicted reaction barriers
strongly depend on the solvent model used (PCM-based, micro-
solvation, and cluster-continuum). Therefore, it remains difficult
to make a final statement as to which reaction pathway will be
favored. However, there is a trend toward the RE mechanism,
the more elaborate the solute–solvent description is.
While in gas phase the SN2 mechanism is clearly favored, a PCM-
based solvent description does not affect significantly the differ-
ence between activation barriers DDG‡ (DG‡RE2DG‡
S N2). Numeri-
cally, similar results are obtained by microsolvation, that is, by
explicit orbital interactions. The combination of the two
approaches—the cluster-continuum model—leads to a substantial
reduction in the gap between the two activation energies: on the
Ta
ble
6.
Sum
mar
yo
fDD
Gz
(RE
vs.
S N2
)in
kcal
mo
l21.
Solv
en
tm
od
el
aug
-cc-
pV
DZ
-pp
aug
-cc-
pV
TZ
-pp
B3
LYP
B3
LYP
[a]
BP
86
BP
86
[a]
B3
LYP
-D2
B3
LYP
-D2
[a]
BP
86
-D2
M0
6-2
XM
06
-2X
[a]
MP
2M
P2
[a]
B3
LYP
B3
LYP
[a]
BP
86
BP
86
[a]
MP
2M
P2
[a]
Gas
ph
ase
15
.12
0.4
9.6
14
.81
4.6
18
.91
0.5
13
.11
3.1
19
.32
4.6
14
.61
9.4
12
.21
6.5
20
.22
7.2
IEF-
PC
M9
.07
.61
0.9
10
.41
0.7
9.6
11
.17
.16
.96
.33
.88
.36
.41
1.7
7.8
7.3
6.1
SMD
10
.18
.61
1.4
9.0
MS
1-O
H1
3.4
15
.99
.09
.9
MS
1-I
14
.01
1.5
11
.41
0.8
MS
22
1.3
16
.21
0.0
6.8
CC
1-O
H9
.15
.61
1.3
7.8
CC
1-I
9.8
7.7
11
.38
.7
CC
23
.72
3.0
1.8
0.3
[a]
No
n-C
urt
in–
Ham
me
ttsc
he
me.
MS,
Mic
roso
lvat
ion
;C
C,
clu
ste
r-co
nti
nu
um
;(2
98
Kan
d1
atm
.).
FULL PAPER WWW.C-CHEM.ORG
2130 Journal of Computational Chemistry 2014, 35, 2122–2131 WWW.CHEMISTRYVIEWS.COM
basis of free energies (DG), the cluster-continuum model predicts
that the SN2 mechanism is favored by 1.8 kcal mol21 only.
The analysis shows that in the cluster-continuum model there
is a considerable difference in activation entropies between the
two pathways, favoring the RE mechanism. Further exploration
of the effect of entropy within the stationary models used here
turns out to be difficult, as an extension of the explicit solvent
shell poses massive computational problems. It is evident, how-
ever, that all contributions to entropy will have to be considered,
possibly also those not accounted for in this work (electrostric-
tion and dielectric saturation). To give a definite answer to the
question of the favored pathway, ab initio molecular dynamics
(AIMD) simulations have to be performed.
As part of an NBO analysis of the electronic structure of the
reagent, we show that the protonation nearly cleaves the
3c4e-bond present in the neutral form of the k3-iodane. More
importantly, the experimental observation that the k3-iodane
reagent is activated after protonation is related to the shift of
the LUMO from the IACvin to the IACF3 bond.
Keywords: hypervalent bonding � iodanes � competing reac-
tion mechanisms � solvent effect � cluster-continuum model
How to cite this article: O. Sala, H. P. L€uthi, A. Togni. J. Comput.
Chem. 2014, 35, 2122–2131. DOI: 10.1002/jcc.23727
] Additional Supporting Information may be found in the
online version of this article.
[1] K. M€uller, C. Faeh, F. Diederich, Science 2007, 317, 1881.
[2] V. V. Zhdankin, P. J. Stang, Chem. Rev. 2008, 108, 5299.
[3] T. Liang, C. N. Neumann, T. Ritter, Angew. Chem. Int. Ed. Engl. 2013, 52, 8214.
[4] K. Niedermann, N. Fr€uh, E. Vinogradova, M. S. Wiehn, A. Moreno, A.
Togni, Angew. Chem. Int. Ed. Engl. 2011, 50, 1059.
[5] H. Pinto de Magalh~aes, H. P. L€uthi, A. Togni, Org. Lett. 2012, 14, 3830.
[6] J. Malmgren, S. Santoro, N. Jalalian, F. Himo, B. Olofsson, Chem. Eur. J.
2013, 19, 10334.
[7] V. V. Zhdankin, Hypervalent Iodine Chemistry: Preparation, Structure and
Synthetic Applications of Polyvalent Iodine Compounds; Wiley: West Sus-
sex, United Kingdom, 2014.
[8] H. Pinto de Magalh~aes, Reactivity of electrophillic trifluoromethylating
hypervalent lambda-3-iodane reagents, Master’s Thesis, ETH Z€urich, 2011.
[9] O. Sala, Quantum chemical investigation of the solvent effect on two
competing trifluoromethylation reaction mechanisms involving
lambda-3-iodane reagents, Master’s Thesis, ETH Z€urich, 2012.
[10] M. Ochiai, Top. Curr. Chem. 2003, 224, 5.
[11] T. Okuyama, T. Takino, T. Sueda, M. Ochiai, J. Am. Chem. Soc. 1995,
117, 3360.
[12] K. Stanek, R. Koller, A. Togni, J. Org. Chem. 2008, 73, 7678.
[13] A. V. Marenich, C. J. Cramer, D. G. Truhlar, J. Chem. Theory Comput.
2013, 9, 3649.
[14] C. Dupont, O. Andreussi, N. Marzari, J. Chem. Phys. 2013, 139, 214110.
[15] T. Okuyama, H. Yamataka, Can. J. Chem. 1999, 77, 577.
[16] C. Lee, W. Yang, R. Parr, Phys. Rev. B 1988, 37, 785.
[17] A. D. Becke, Phys. Rev. A 1988, 38, 3098.
[18] A. Becke, J. Chem. Phys. 1993, 98, 5648.
[19] J. P. Perdew, Phys. Rev. B Condens. Matter 1986, 33, 8822.
[20] J. P. Perdew, Phys. Rev. B Condens. Matter 1986, 34, 7406.
[21] Y. Zhao, D. Truhlar, Theor. Chem. Acc. 2008, 120, 215.
[22] R. A. Kendall, T. H. Dunning, Jr., R. J. Harrison, J. Chem. Phys. 1992, 96, 6796.
[23] D. E. Woon, T. H. Dunning, Jr., J. Chem. Phys. 1993, 98, 1358.
[24] K. A. Peterson, B. C. Shepler, D. Figgen, H. Stoll, J. Phys. Chem. A 2006,
110, 13877.
[25] S. Miertus, E. Scrocco, J. Tomasi, Chem. Phys. 1981, 55, 117.
[26] M. Cossi, B. Mennucci, J. Tomasi, Chem. Phys. Lett. 1994, 228, 165.
[27] R. Cammi, J. Tomasi, J. Comput. Chem. 1995, 16, 1449.
[28] E. Cancs, B. Mennucci, J. Tomasi, J. Chem. Phys. 1997, 107, 3032.
[29] J. R. Pliego, J. M. Riveros, J. Phys. Chem. A 2001, 105, 7241.
[30] S. A. Martins, S. F. Sousa, J. Comput. Chem. 2013, 34, 1354.
[31] A. V. Marenich, C. J. Cramer, D. G. Truhlar, J. Phys. Chem. B 2009, 113, 6378.
[32] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R.
Cheeseman, G. Scalmani, V. Barone, B. Mennucci, G. A. Petersson, H.
Nakatsuji, M. Caricato, X. Li, H. P. Hratchian, A. F. Izmaylov, J. Bloino, G.
Zheng, J. L. Sonnenberg, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J.
Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven,
J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro, M. Bearpark, J. J. Heyd, E.
Brothers, K. N. Kudin, V. N. Staroverov, R. Kobayashi, J. Normand, K.
Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, N.
Rega, J. M. Millam, M. Klene, J. E. Knox, J. B. Cross, V. Bakken, C. Adamo,
J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R.
Cammi, C. Pomelli, J. W. Ochterski, R. L. Martin, K. Morokuma, V. G.
Zakrzewski, G. A. Voth, P. Salvador, J. J. Dannenberg, S. Dapprich, A. D.
Daniels, Farkas, J. B. Foresman, J. V. Ortiz, J. Cioslowski, D. J. Fox, Gaussian
09 Revision D.01; Gaussian Inc.: Wallingford CT, 2009.
[33] J. VandeVondele, M. Krack, F. Mohamed, M. Parrinello, T. Chassaing, J.
Hutter, Comput. Phys. Commun. 2005, 167, 103.
[34] CP2K Open Source Molecular Dynamics, 2013. Available at: http://
www.cp2k.org. Accessed on August 2013.
[35] S. Goedecker, M. Teter, and J. Hutter, Phys. Rev. B, 1996, 54, 1703.
[36] J. VandeVondele, J. Hutter, J. Chem. Phys. 2007, 127, 114105.
[37] B. Chan, L. Radom, J. Phys. Chem. A 2013, 117, 3666.
[38] F. Weinhold, C. R. Landis, Discovering Chemistry with Natural Bond
Orbitals; Wiley: Hoboken, New Jersey, 2012.
[39] H. Eyring, J. Chem. Phys. 1935, 3, 107.
[40] M. G. Evans, M. Polanyi, Trans. Faraday Soc. 1935, 31, 875.
Received: 23 May 2014Revised: 8 August 2014Accepted: 12 August 2014Published online on 15 September 2014
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 2122–2131 2131