Additivity of substituent effects in aromatic stacking interactions

Article abstract of DOI:10.1021/ja504378p

The goal of this study was to experimentally test the additivity of the electrostatic substituent effects (SEs) for the aromatic stacking interaction. The additivity of the SEs was assessed using a small molecule model system that could adopt an offset face-to-face aromatic stacking geometry. The intramolecular interactions of these molecular torsional balances were quantitatively measured via the changes in a folded/unfolded conformational equilibrium. Five different types of substituents were examined (CH₃, OCH₃, Cl, CN, and NO₂) that ranged from electron-donating to electron-withdrawing. The strength of the intramolecular stacking interactions was measured for 21 substituted aromatic stacking balances and 21 control balances in chloroform solution. The observed stability trends were consistent with additive SEs. Specifically, additive SE models could predict SEs with an accuracy from ±0.01 to ±0.02 kcal/mol. The additive SEs were consistent with Wheeler and Houk's direct SE model. However, the indirect or polarization SE model cannot be ruled out as it shows similar levels of additivity for two to three substituent systems, which were the number of substituents in our model system. SE additivity also has practical utility as the SEs can be accurately predicted. This should aid in the rational design and optimization of systems that utilize aromatic stacking interactions.

Full text of DOI:10.1021/ja504378p

Article
pubs.acs.org/JACS
Additivity of Substituent Effects in Aromatic Stacking Interactions
Jungwun Hwang, Ping Li, William R. Carroll,† Mark D. Smith, Perry J. Pellechia, and Ken D. Shimizu*
Department of Chemistry and Biochemistry, University of South Carolina, Columbia, South Carolina 29208, United States
S
* Supporting Information
ABSTRACT: The goal of this study was to experimentally test the additivity
of the electrostatic substituent effects (SEs) for the aromatic stacking
interaction. The additivity of the SEs was assessed using a small molecule
model system that could adopt an offset face-to-face aromatic stacking
geometry. The intramolecular interactions of these molecular torsional
balances were quantitatively measured via the changes in a folded/unfolded conformational equilibrium. Five different types of
substituents were examined (CH3, OCH3, Cl, CN, and NO2) that ranged from electron-donating to electron-withdrawing. The
strength of the intramolecular stacking interactions was measured for 21 substituted aromatic stacking balances and 21 control
balances in chloroform solution. The observed stability trends were consistent with additive SEs. Specifically, additive SE models
could predict SEs with an accuracy from 0.01 to 0.02 kcal/mol. The additive SEs were consistent with Wheeler and Houk’s
direct SE model. However, the indirect or polarization SE model cannot be ruled out as it shows similar levels of additivity for
two to three substituent systems, which were the number of substituents in our model system. SE additivity also has practical
utility as the SEs can be accurately predicted. This should aid in the rational design and optimization of systems that utilize
aromatic stacking interactions.
stacking and aid in rational design of host−guest, drug−
receptor, and substrate−catalysts systems that incorporate
aromatic stacking interactions. In addition, the verification of
additive SEs could help differentiate the different theoretical
models of the aromatic stacking SEs.24−28
Our strategy was to prepare and study a series of small
molecule model systems (Scheme 1), which could form and
INTRODUCTION
■
Aromatic stacking interactions play an important role in
determining the structure, property, and function of many
synthetic1−4 and biological systems.4−7 For example, attractive
interactions of aromatic surfaces have been cited as a major
stabilizing interaction in nucleic acid and protein structures,8
host−guest complexes,9,10 solid-state structures,11 and tran-
sition states of asymmetric catalysts.12,13 A common strategy for
modulating the strengths of aromatic stacking interactions is via
the introduction of substituents on the aromatic rings.
Theoretical14−18 and experimental19−22 studies have found
that electron-withdrawing groups stabilize and electron-
donating groups generally destabilize aromatic stacking
interactions. In addition, theoretical studies have predicted
that the electronic substituent effects (SEs) in multisubstituted
aromatic rings will be additive.23 For example, Sherrill et al.
demonstrated an excellent linear correlation between the
interaction energy and the number of substituents. However,
these theoretical studies were carried out in vacuo and examined
the aligned face-to-face stacking geometry, where one aromatic
ring is directly over the opposing ring. In this study, we
experimentally test whether the electrostatic SEs are additive
for the more commonly observed offset aromatic stacking
geometry (Figure 1). SE additivity would provide a simple
means of rationally designing systems that utilize aromatic
Scheme 1. Representation of (a) the Molecular Torsional
Balance Designed To Measure the SEs of an Intramolecular
Aromatic Stacking Interaction via Changes in the Folded/
Unfolded Equilibrium Ratio and (b) (b) the Control Balance
Designed to Measure the Solvent and Repulsive Lone Pair to
π Interactions of the Oxygen Linker
Figure 1. General depiction of additive SEs for aromatic stacking
interaction in the offset face-to-face geometry.
Received: May 1, 2014
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
measure the strength of an intramolecular offset face-to-face (or
parallel) aromatic stacking interaction. Five different sub-
stituents (OCH3, CH3, Cl, CN, NO2) were introduced on the
aromatic ring yielding a total of 21 different combinations. This
range of substituents and substitution patterns allowed us to
systematically test whether the SEs of the monosubstituted
systems could be added together to accurately predict the SEs
for the multisubstituted systems (Figure 1).
There have been a number of experimental studies that have
examined the additivity of SEs in face-to-face aromatic stacking
interactions. However, these previous studies examined a
limited number of substituents, measured the interaction
energies via an indirect method, and yielded to opposing
conclusions. Two separate studies examined the additivity of
SEs in aromatic stacking interactions indirectly via measuring
changes in a rotational barrier. Cozzi and Siegel showed that
benzene rings with varying numbers of fluorines appeared to
show SE additivity for the stacking interaction within a rigid
1,8-diarylnaphthalene model system.29 Conversely, Waters and
Rashkin found that the SEs for CH3, CF3, and F groups fell
short of additivity by an average of 19% in their
benzylpyridinium model system.22 A possible explanation for
the opposing conclusions is the indirect method of measuring
aromatic stacking energies used in both studies. The face-to-
face aromatic stacking interaction energies in the ground states
were measured via the rotational barrier of one of the aromatic
surfaces. However, both systems could form additional edge-to-
face or edge-to-substituent interactions in their transition states.
Therefore, the observed SE trends could be a combination of
the trends for the face-to-face and these other interactions.
Another experimental study of note is Schneider’s compre-
hensive study of the SEs in the host−guest interactions of
porphyrins in water.30 The observed additive SEs were
attributed to dispersion and solvophobic effects as opposed
to aromatic stacking interactions.
Figure 3. Top view of the folded conformer of the X-ray crystal
structure of a two-armed version of unsubstituted balance 1u
highlighting the intramolecular aromatic stacking interaction between
the phenyl ether of the arm (top) and the phenanthrene shelf
(bottom).31 For clarity, the bicyclic framework is hidden.
measured via a conformational equilibrium (Scheme 1). Due to
restricted rotation around the Caryl−Nimide single bond, 1 and 2
adopt distinct folded and unfolded conformers that form and
break the intramolecular stacking interaction. Thus, the folded/
unfolded conformer ratio provides a sensitive ( 0.008 to 0.03
kcal/mol) measure of the interaction.33,34 This ratio was easily
measured from the peak areas of the two conformers in the 1H
NMR spectra, which were in slow exchange at room
temperature. Third, the balance exclusively forms the face-to-
face geometry and cannot form alternative stacking geometries
due to the rigidity of the bicyclic framework. Molecular
modeling simulations confirmed that there was insufficient
distance between the aromatic surfaces in the folded conformer
to form the edge-to-face geometry. Fourth, we could introduce
up to three substituents on the arm of the balance that would
only electrostatically attenuate the stacking interaction. The X-
ray structure (Figure 3) showed that substituents at these
positions could not form steric or dispersion interactions with
the aromatic shelf. Due to the offset geometry, the benzene arm
juts out beyond the edge of the phenanthrene shelf, and
substituents at the para- and meta-positions are not over the
aromatic shelf. The elimination of these secondary interactions
was important, as experimental24,22 and theoretical studies35
have shown that they can disrupt the electrostatic SE trends of
interest. While the crystal structure shows the two meta-
positions are in different geometries, they are equivalent on the
NMR time scale. This is because, as shown by calculations by
Datta et al.,36 the arm is moving back and forth rapidly between
the two outer benzenes of the phenanthrene shelf.
Our recently reported molecular torsional balance model
system (Scheme 1 and Figure 2) has a number of attractive
Finally, control balances 2a−u could be used to measure and
remove other possible influences on the folded/unfolded ratios
in balance 1. These included repulsive the lone pair to π
interactions of the oxygen linker31,36,37 or solvent effects of the
substituted aryl ether arms.38,39 The control balances 2a−u
contained the same substituted aromatic arms but had a smaller
benzene shelf that could not form the aromatic stacking
interactions. This can be visualized using the crystal structure in
Figure 3. The control balance 2 contains only the central
benzene ring of the phenanthrene shelf in 1, which can form
lone pair to π interactions with the linker ether oxygen. Thus,
the aromatic interaction energies were measured using the
difference in the folding energies of the aromatic stacking
balances 1 and the control balances 2 (ΔG1 − ΔG2).
Figure 2. Folded conformers of aromatic stacking balance 1 and
nonstacking control balance 2, which contain 21 different substituted
benzenes arms (a−u).
attributes for the study of SE additivity.31,32 First, we previously
established that this rigid bicyclic model system forms a well-
defined intramolecular aromatic stacking interaction in the
folded conformer. The X-ray crystal structure (Figure 3)
showed that the aromatic surfaces of the phenyl ether arm
and the phenanthrene shelf were held in a parallel geometry
and at a proper distance (3.76 Å, centroid-to-plane). Second,
the aromatic stacking energy can be easily and accurately
RESULTS AND DISCUSSION
■
First, the 21 substituted balances 1a−u and the matching 21
substituted control balances 2a−u (Figure 2) were prepared.
These were all rapidly and efficiently assembled using a
common modular synthesis route (Scheme 2). The substituted
B
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
have a shorter benzene shelf that cannot form intramolecular
aromatic stacking.
Scheme 2. Modular Synthetic Route Used to Prepare
Balances 1 and 2 with Substituted Arms a−k, n, o, r−t
The ability to measure the intramolecular stacking
1
interactions was confirmed by examination of the H NMR
spectra. The presence of distinct folded and unfolded conformers
1
was observed via the separate set of peaks in the H NMR
spectra (CDCl3, 25 °C) for most of the protons in 1a−u and
2a−u. Thus, the two conformers were in slow exchange, and
folded/unfolded ratio could be easily measured from the
corresponding peak areas. The succinimide methine protons
were used for this analysis because they were singlets in a
relatively clear region of the spectra (4.2−4.7 ppm). To ensure
the accuracy and consistency of the measurement, the line-
fitting method was applied to high-concentration NMR samples
(30 mM).42,43 In this manner, the folded/unfolded ratios for 1
and 2 were measured (Table 1), and the corresponding folding
energies (ΔG1 and ΔG2) were calculated.
Table 1. Measured Folding Energies (kcal/mol) of Balances
1 (ΔG1) and Control Balances 2 (ΔG2) Systems and SEs
(SEx = (ΔG1x − ΔG2x) − (ΔG1u − ΔG2u))
a
a
b
arm
substituents
ΔG1
ΔG2
SEmeasd
a
b
c
d
e
f
m-CH3
0.40
0.44
1.50
1.31
1.56
1.35
1.52
1.22
1.45
1.38
1.28
1.25
1.30
1.51
1.58
1.47
1.31
1.61
1.26
1.41
1.29
1.56
1.40
−0.19
0.05
p-CH3
m,m-(CH3)2
m,p-(CH3)2
m,m,p-(CH3)3
p-Cl
0.29
−0.35
−0.08
−0.27
−0.15
−0.54
−0.92
−0.73
−0.30
−0.39
−0.36
−0.94
−0.31
−0.63
−1.00
0.01
0.35
0.34
0.15
g
h
i
m-Cl
−0.02
−0.46
−0.37
0.03
aromatic ring of the arms and the aromatic shelves were
synthesized independently and then condensed together in the
last step. Five representative substituents were chosen spanning
the range from electron-donating (CH3, OCH3) to electron-
withdrawing (Cl, CN, and NO2).40 A variety of substitution
patterns were examined including 1 unsubstituted (1u), 9
monosubstituted (1a, b, f, g, n, o, r, s), 10 disubstituted (1c, d,
h−m, p, q, t), and 1 trisubstituted (1e) arms.
m,m-(Cl)2
m,p-(Cl)2
p-Cl, m-CH3
m-Cl, p-CH3
p-NO2
j
k
l
−0.01
0.22
m
n
o
p
q
r
p-NO2, m-CH3
p-CN
−0.28
0.24
m-CN
−0.24
−0.31
0.34
All of the aromatic arms 4 were assembled via an SNAr
reaction. The majority of the arms (4a−k, n, o, r−t) were made
by the reaction of the appropriately substituted phenolate with
2-fluoronitrobenzene to form diphenyl ethers 3. Then 2-nitro
groups were reduced to the primary amines. For arms with
strong electron-withdrawing substituents (l, m, p, q), their
phenolates were not sufficiently nucleophilic. Therefore, 4l, m,
p, q were prepared in one step via the SNAr reaction of 2-
aminophenol with the appropriately substituted fluoro- or
nitrobenzene (not shown).
m,p-(CN)2
p-NO2, m-CN
p-OCH3
0.33
−0.16
−0.15
−0.61
0.00
s
m-OCH3
m,p-OCH3
none
0.21
t
0.03
u
0.48
a
ΔG1 and ΔG2 were measured at 25 °C with an error of < 0.008 and
b
0.03 kcal/mol, respectively. SE with an error of < 0.03 kcal/mol.
The aromatic shelves were synthesized via the Diels−Alder
reaction of maleic anhydride with either phencyclone or
diphenylisobenzofuran to yield the endo-bicyclic anhydrides 5
or 6 containing phenanthrene or benzene surfaces.41 Finally,
thermal condensation reaction of a substituted aniline arm
(4a−u) and a bicyclic anhydride shelf (5 or 6) yielded the endo-
bicyclic 1 and 2.
The ability of balances 1 and 2 to form the desired
intramolecular aromatic stacking interactions was established by
1H NMR. The expected upfield shifts were observed for the
protons on the substituted benzene arms of 1a−u, which were
indicative of the benzene arm being positioned over the
phenanthrene shelf. In particular, the ortho-protons (adjacent to
the ether oxygen) of the benzene arms shifted from 1.0 to 1.2
ppm upfield. By comparison, upfield shifts were not observed
for the benzene arm protons of control balances 2a−u, which
The folding energies of 1 and 2 (ΔG1 and ΔG2) were
analyzed to verify that folded/unfolded equilibrium provided a
measure of the intramolecular aromatic stacking energies in 1.
The ΔG1 values (−0.46 to 0.48 kcal/mol) were consistently
lower in energy than the ΔG2 values (1.22 to 1.61 kcal/mol).
This was consistent with the expected stabilization of the folded
conformer in 1 by aromatic stacking interactions. The average
difference of −1.3 kcal/mol was consistent with previous
measures of benzene−benzene stacking interactions in organic
solution.20,24
Hammett plot analyses were conducted to establish that the
electrostatic SEs in our aromatic stacking model system were
similar to those in previous theoretical and experimental
studies.19−22 To isolate the intramolecular stacking energies,
the difference in folding energies of 1 and 2 (ΔG1 − ΔG2) of
the monosubstituted balances were plotted against their σmeta
C
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
parameters (Figure 4). The Hammett σmeta parameter was used
because it more closely correlates to the purely electrostatic SEs
Table 2. Comparison of the Calculated (SEcalcd) and
Measured (SEmeasd) multisubstituent effects. SEcalcd were
calculated from the sum of the SEmeasd for the constituent
mono-SE
Figure 4. Hammett σm plot of ΔG1 − ΔG2 for the monosubstituted
balances with substituents in the meta- (blue) or para-substituents
(red) (the substituents from left to right CH3, H, OCH3, Cl, CN,
NO2).
than σpara, which also includes resonance effects.15,24 Interest-
ingly, the meta- and para-systems displayed distinct linear
trends with slopes of (−0.95 and −0.55 kcal/mol). The origins
of this positional dependence will be discussed in more detail
below. However, the overall sign and magnitudes of these
slopes were consistent with those measured by Hunter (−0.62
kcal/mol)21 and Gung (−1.06 kcal/mol)20 for face-to-face
aromatic stacking interactions in organic solvents.
The importance of the control balances 2 in improving the
accuracy of the analysis was tested by comparing the Hammett
plots with (ΔG1 − ΔG2) and without (ΔG1) the balance 2
corrections. Although the Hammett plots without balance 2
had similar magnitude negative slopes, there was considerably
more scatter in the linear correlation. For example, the R2 value
for the para-substituted values for ΔG1 was 0.69 as compared
to 0.97 for ΔG1 − ΔG2. The improved correlation with ΔG2
suggests that secondary factors other than the electron-
withdrawing and -donating effects of the substituents influence
the folding ratios in 1. These secondary factors include
solvation effects, changes in the dipoles of the folded and
unfolded conformers, and resonance effects on the oxygen linker
that could modulate the repulsive lone pair to π interactions.
Once the viability of our model system was established, the
additivity of the SEs was assessed. The SEs in 1 were calculated
using eq 1. The SE for an arm with substituents x was defined
as the difference in the folding energies of 1 and 2 containing
the substituents x. In addition, the SE(x) was normalized by
subtracting out the stacking interactions in the unsubstituted
balances 1u and 2u. Thus, the SE for the unsubstituted arm
(SEu) was 0.0 kcal/mol.
a
SEmeasd (kcal/mol) in CDCl3 at 25 °C with an error of < 0.03 kcal/
b
mol. SEcalcd of disubstituents with an error of < 0.04 kcal/mol.
c
SEcalcd of trisubstituents with an error of < 0.05 kcal/mol.
groups (arms c, d, e, s), electron-withdrawing Cl, NO2, and CN
groups (arms h, i, p, q), and mixtures of electron-donating and
withdrawing groups (arms j, k, m).
However, there was not perfect agreement between the
measured and calculated SE values in Table 2. Therefore,
additional analyses were conducted to determine whether these
deviations were systematic or random error. First, a correlation
plot of the calculated and measured SE values (Figure 5)
showed an excellent linear correlation with an R2 value of 0.96
SE(x) = (ΔG1x − ΔG2x) − (ΔG1u − ΔG2u)
(1)
and a slope near unity (0.99). The residuals (SEcalcd − SEmeasd
)
An initial analysis of the SEmeasd values for the 11
multisubstituted arms in Table 2 were consistent with SE
additivity. There was a good correlation between to the SEmeasd
and SEcalcd values calculated from the sum of the SEmeasd for the
individual substituents. For example, the SEmeasd (−0.35 kcal/
mol) for the m,m-(CH3)2 arm c is very similar to the SEcalcd
value (−0.37 kcal/mol) calculated from the sum of two SEmeasd
from the m-CH3 arms. The agreement between the measured
and calculated SEs spanned a diverse set of substituent types.
These included arms with electron-donating CH3 and OCH3
had a standard error 0.02 kcal/mol, which was less than the
error of the analysis of 0.03 kcal/mol. Most importantly, the
SEmeasd values were not consistently lower than the SEcalcd
values. In addition, a plot of the residuals did not show any
systematic variations. Specifically, the residuals for arms with
electron-donating or -withdrawing groups did not show
systematic variance.
Building on the success of the above additivity model, a more
complex additivity model was developed to provide a more
accurate estimate of the SEs. A multivariate approach was
D
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(1.00). An additional advantage of this analysis is that it could
be used to predict SEcalcd2 values for monosubstituted arms.
The existence or absence of SE additivity can also be used to
test fundamental models of the aromatic stacking interaction.
Two general models have been proposed for the origins of the
SEs (Figure 7). In the direct interaction model, the substituents
Figure 5. Correlation plot of the calculated (SEcalcd) measured
(SEmeasd) SEs for the aromatic stacking interactions in the 11
multisubstituted arms in 1 and 2. The SEcalcd values are based on
simple additivity model based on the sum of the individual SEmeasd
values for each multisubstituted arm.
employed similar to that used by Schneider et al. for the
analysis of SE additivity in porphyrin host−guest complexes.30
This multivariate model incorporated data from the multi-
substituted and monosubstituted arms. Thus, each of the 9
individual SEs (m-CH3, p-CH3, m-Cl, p-Cl, m-CN, p-CN, m-
OCH3, p-OCH3, p-NO2) were estimated based on multiple SE
measurements instead of just the monosubstituted SE measure-
ments used in the previous analysis.
To estimate the 9 individual SEs, a matrix of 21 algebraic
equations was created using equation (eq 2) to calculate the SE
for each substituted arm in Table 1. Equation 2 calculates the
SE for a substituted benzene via the sum of the number of the
individual substituents of type z (nz) multiplied by the SE of
substituent z (SEz). Using the solver function in Excel, a set of
SEz values was found for the 9 different types of substituents
that best fit the SEcalcd2 and SEmeasd values for all 21 arms.
Figure 7. Representations of (a) the direct and indirect SE models for
the aromatic stacking interaction and (b) the presence and absence of
additive SEs for the direct and indirect SE models.
interact directly with the edges of the opposing aromatic
ring.27,35,44 Alternatively in the indirect interaction model, the
substituents modulate the electrostatic potential of the attached
π-system. The polarized aromatic ring, in turn, will have a
stronger or weaker electrostatic attraction for the opposing
aromatic ring.14,45
SEcalcd2 = ∑ nz·SEz
(2)
The two SE models predict different degrees of SE additivity.
The direct model predicts additive SEs.44 Each substituent
forms a separate interaction with the edge of the opposing
aromatic ring. In contrast, the indirect model predicts SEs that
are smaller than estimated by an additivity model. In the
indirect model, the first substituent will impart the largest
electrostatic polarization of the attached π-system. However,
each additional substituent has a successively smaller SE as the
buildup of charge on the aromatic ring makes the π-surface
more difficult to polarize.
The improved accuracy was evident from the better
correlation of the SEcalcd2 and SEmeasd values. The standard
error of the difference was only 0.01 kcal/mol. A plot of
SEmeasd versus SEcalcd2 (Figure 6) also showed the improved
correlation with an R2 = 0.99 and a slope was closer to unity
Therefore, the additivity of the SEs in our experimental
model systems appeared to support the direct SE model. The
near unity slopes of the correlation plots (Figures 5 and 6) and
trendlines intersecting the origin provide support for the 1:1
correlation of the SEmeasd to the additivity-based SEcalcd values.
However, a concern was that the deviation for the indirect
model from the additivity model could be too small to be
accurately measured. To address this concern, we estimated the
deviations from the additivity model for the indirect model by
calculating the electrostatic potential (ESP) for a series of
benzene rings with varying numbers of substituents (0−6). ESP
provides a measure of the electrostatic polarization of the π-
system by substituents and was shown by Hunter et al. to
correlate to the measured SEs in the aromatic stacking
interactions.21,24 The ESP calculations were made using the
Figure 6. Correlation plot of the measured (SEmeasd) and calculated
SEs (SEcalcd2) for all 21 arms using a simple additive model based on
the sum of the constituent monosubstituted arms.
E
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
same method and level of theory (B3LPY, 6-31G*) as used by
Hunter et al. Three representative substituents were examined
that matched the substituents used in our model system: two
electron-withdrawing groups (CN and Cl) and one electron-
donating (CH3) group. ESP for symmetrical substitution
patterns is shown in Figure 8 for benzenes bearing 2, 3, and
4 substituents. Similar trends were observed when the ESPs for
asymmetric substitution patterns.
Quantitatively, the ratios of the SE for the meta- and para-
positions averaged 1.7:1 as judged by the ratio of the slopes for
their Hammett plots (Figure 4).
The indirect model predicts a weaker positional SE
dependence than the direct model. To the first approximation,
the indirect model should not have a positional dependence, as
the meta- and para-substituents polarize the π-system of the
attached aromatic ring to a similar extent. However, the
orientation of the polarization of the π-surface will differ
depending upon the position of the substituent relative to the
dipole or quadrupole of the opposing aromatic surface.
Therefore, we cannot entirely rule out the possibility that the
indirect model, as the magnitude of its positional dependence is
not easily calculated.
CONCLUSIONS
■
In this study, we have utilized a small molecule model system to
experimentally test the additivity of the SEs for the offset
stacking aromatic interaction in organic solution. A series of
molecular torsional balances 1a−u and control balances 2a−u
with 21 different substituted benzene arms were prepared
containing 5 different substituents (CH3, OCH3, Cl, CN, NO2)
that ranged for strongly electron-withdrawing to electron-
donating groups. The formation of intramolecular aromatic
stacking interactions and the ability to electrostatically influence
the interaction energies by the introduction of substituents in
our model system were confirmed by NMR, X-ray crystallog-
raphy, and the observation of linear Hammett plots for the
monosubstituted systems. These SEs ranged from slightly
destabilizing (+0.05 kcal/mol for p-CH3) substituted arm b to
strongly stabilizing (−1.00 kcal/mol for m,p-(CN)2).
An analysis of the SEs for the 11 multisubstituted balances
demonstrated that simple additivity models could provide a
good estimate ( 0.01 to 0.02 kcal/mol) of the measured SEs.
Specifically, the SEs for multisubstituted systems could be
accurately estimated from the sum of the individual SEs (eq 2).
These individual SEs could be estimated from SEmeasd of the
monosubstituted balances or using via multivariate analysis
using the SEmeasd of the mono- and multisubstituted balances.
The observation of SE additivity for the aromatic stacking
interaction in this solution-phase study was consistent with
previous theoretical studies conducted in vacuo.23
The additivity of SEs for the aromatic stacking interactions
has both theoretical and practical applications. First, the SE
additivity and the positional dependence of the SEs in our
model system provide support for the direct SE model, in
which the SEs are due to the direct interaction of the
substituents with the edge of the opposing aromatic ring.
However, due to the limitation in the number of substituents in
our system and the measurement error, we cannot definitively
exclude the indirect SE model. Second, the additivity of the SEs
for the aromatic stacking interaction has practical applications.
This should aid in the rational design and optimization of
systems that utilize aromatic stacking, as the magnitude of the
stabilization or destabilization by substituents can be accurately
predicted using an additive SE model. As demonstrated in this
study, the aromatic stacking energies of multisubstituted
benzenes can be accurately predicted from the measured
stacking energies of the analogous monosubstituted systems.
Alternatively, the stacking energies of monosubstituted systems
can be predicted from stacking energies of multisubstituted
systems. For example from Table 2, the stacking energy of m-
Figure 8. Calculated ESPs (kcal/mol) at the center of benzene rings
bearing 0−6 symmetrically positioned substituents (CN, Cl, and
CH3). The broken lines show the expected ESPs for a perfectly
additive system.
The ESP calculations showed the expected deviations from
additivity (Figure 8). The first substituent had the largest effect,
and each successive substituent had a smaller effect. This
yielded asymptotic ESP curves that increasingly deviated from
the linear additivity model with an increasing number of
substituents. However, the deviations from the additivity model
were small for benzenes with 2 and 3 substituents, which was
the number of substituents in our model systems. The average
difference from the additivity model was only −8% with 2
substituents and −11% with 3 substituents. Furthermore, the
majority of these differences were due to the electron-donating
CH3 groups that had small SEs, and thus small variations
became large percentage differences. Excluding the CH3 group
data, the deviations from the additivity model were only slightly
larger than the error in our model system. Therefore, we cannot
definitively rule out the indirect SE model.
We examined other ways of differentiating the direct and
indirect SE models. Larger deviations (>20%) from the
additivity models were predicted for benzenes with 4 or more
substituents (Figure 8). However, our current model system is
limited to 3 substituents because one position is occupied by
the oxygen linker, and the two adjacent ortho-positions can
form additional steric and dispersion interactions. Alternatively,
the distinct SE trends for the meta- and para-substituents in our
model system provide support for the direct SE model. The
direct SE model was originally developed for the aligned
stacking geometry. However, the direct SE model has also been
shown to be applicable to the offset stacking geometry.44 A key
difference in the offset geometry is that the substituents are no
longer symmetrically arranged. Thus, the direct SE model
predicts that substituents will have different magnitude SEs
depending upon the distance of the substituents from the edge
of the opposing aromatic surface. This positional dependence
was observed in our system, as the closer meta-substituents
have a much stronger influence than the para-substituents.
F
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
CH3 can be estimated from half of the measured stacking
energy of m,m-(CH3)2.
REFERENCES
■
(1) Resconi, L.; Cavallo, L.; Fait, A.; Piemontesi, F. Chem. Rev. 2000,
100, 1253−1345.
(2) Roesky, H. W.; Andruh, M. Coord. Chem. Rev. 2003, 236, 91−
119.
(3) Zhao, D. H.; Moore, J. S. Chem. Commun. 2003, 807−818.
(4) Meyer, E. A.; Castellano, R. K.; Diederich, F. Angew. Chem., Int.
Ed. 2003, 42, 1210−1250.
(5) Salonen, L. M.; Ellermann, M.; Diederich, F. Angew. Chem. Int.
Ed. 2011, 50, 4808−4842.
(6) Draper, D. E. J. Mol. Biol. 1999, 293, 255−270.
(7) Whitten, D. G.; Chen, L. H.; Geiger, H. C.; Perlstein, J.; Song, X.
D. J. Phys. Chem. B 1998, 102, 10098−10111.
(8) Riley, K. E.; Hobza, P. Acc. Chem. Res. 2013, 46, 927−936.
(9) Klosterman, J. K.; Yamauchi, Y.; Fujita, M. Chem. Soc. Rev. 2009,
38, 1714−1725.
(10) Schneider, H. J. Angew. Chem., Int. Ed. 2009, 48, 3924−3977.
(11) Moulton, B.; Zaworotko, M. J. Chem. Rev. 2001, 101, 1629−
1658.
EXPERIMENTAL SECTION
■
Measurement of the Folded/Unfolded Ratios. The folding
ratios for balances 1a−u and control balances 2a−u were measured
from the relative peak areas of the succinimide methine protons in the
1H NMR spectra (CDCl3, 25 °C, 30 mM) as previously described.31,32
Briefly, the peak areas of the succinimide methine singlets in the 4.2 to
4.7 ppm region for the folded and unfolded conformers were measured
using the line fitting method as implemented in the Mnova NMR
software. In cases where these peaks overlapped with other protons,
the folded/unfolded ratios for the succinimide methine protons were
verified by comparison with the folded/unfolded ratios measured from
the most downfield aromatic protons (8.0−8.7 ppm). The folded and
unfolded peaks were assigned based on comparison of their peak areas
to the upfield shifted doublet for the ortho-proton of the N-aryl ring
(4.6−4.7 ppm for balance 1, 5.7−5.8 ppm for balance 2), which was
assigned to the unfolded conformer. This peak was assigned to the
unfolded conformer because this proton is directly over the aromatic
shelf in the unfolded conformer. The corresponding folding energies
(ΔG at 298 K) were calculated from the equation: ΔG = −RT
ln[folded]/[unfolded].
(12) Wei, Y.; Held, I.; Zipse, H. Org. Biomol. Chem. 2006, 4, 4223−
4230.
(13) Ruano, J. L. G.; Aleman, J.; Alonso, I.; Parra, A.; Marcos, V.;
Aguirre, J. Chem.Eur. J. 2007, 13, 6179−6195.
(14) Hunter, C. A.; Sanders, J. K. M. J. Am. Chem. Soc. 1990, 112,
5525−5534.
(15) Watt, M.; Hardebeck, L. K. E.; Kirkpatrick, C. C.; Lewis, M. J.
Am. Chem. Soc. 2011, 133, 3854−3862.
(16) Wheeler, S. E.; McNeil, A. J.; Muller, P.; Swager, T. M.; Houk,
K. N. J. Am. Chem. Soc. 2010, 132, 3304−3311.
(17) Lee, E. C.; Kim, D.; Jurecka, P.; Tarakeshwar, P.; Hobza, P.;
Kim, K. S. J. Phys. Chem. A 2007, 111, 3446−3457.
(18) Sinnokrot, M. O.; Sherrill, C. D. J. Am. Chem. Soc. 2004, 126,
7690−7697.
Error Analyses. The error in measuring the folded/unfolded ratio
from the peak areas in the 1H NMR spectra was conservatively
estimated at 1% for balances 1a−u. This estimate is based on the
literature studies that have found that concentrations can be accurately
measured by 1H NMR with an accuracy of <1%.46−48 In particular, the
use of line shape analysis software can improve the accuracy to <0.3%
even for overlapped peaks.49 To ensure this level of accuracy, the
NMR spectra were collected at a high concentration (30 mM). At this
concentration, the minor conformer concentration (folded or unfolded)
remained ≥10 mM for the entire series of balances 1a−u.
Despite the measurement of the folded/unfolded ratios under similar
conditions, a larger measurement error of 5% was estimated for
control balances 2a−u. Control balances 2a−u had folding ratios
ranging from 1:8 to 1:15. Thus, at 30 mM, the concentration of the
minor isomer could be as low as 2 mM. At this lower concentration,
we conservatively estimated the peak area measurement error at 5%
based on literature precedent.46
(19) Cozzi, F.; Cinquini, M.; Annunziata, R.; Dwyer, T.; Siegel, J. S. J.
Am. Chem. Soc. 1992, 114, 5729−5733.
(20) Gung, B. W.; Xue, X. W.; Reich, H. J. J. Org. Chem. 2005, 70,
3641−3644.
(21) Cockroft, S. L.; Hunter, C. A.; Lawson, K. R.; Perkins, J.; Urch,
C. J. J. Am. Chem. Soc. 2005, 127, 8594−8595.
(22) Rashkin, M. J.; Waters, M. L. J. Am. Chem. Soc. 2002, 124,
1860−1861.
Propagation of these measurement errors in the folded/unfolded
ratios through the Gibbs free energy equation gave errors of 0.008
and 0.03 kcal/mol for the measured folding energies for 1 and 2
(ΔG1 and ΔG2), respectively. Propagation of these errors in ΔG1 and
ΔG2 through the SE definition equation (eq 1) yielded an error of
0.03 kcal/mol in SEmeasd values.
(23) Ringer, A. L.; Sinnokrot, M. O.; Lively, R. P.; Sherrill, C. D.
Chem.Eur. J. 2006, 12, 3821−3828.
(24) Cockroft, S. L.; Perkins, J.; Zonta, C.; Adams, H.; Spey, S. E.;
Low, C. M. R.; Vinter, J. G.; Lawson, K. R.; Urch, C. J.; Hunter, C. A.
Org. Biomol. Chem. 2007, 5, 1062−1080.
(25) Ringer, A. L.; Sherrill, C. D. J. Am. Chem. Soc. 2009, 131, 4574−
4575.
ASSOCIATED CONTENT
■
S
* Supporting Information
(26) Cozzi, F.; Annunziata, R.; Benaglia, M.; Baldridge, K. K.;
Aguirre, G.; Estrada, J.; Sritana-Anant, Y.; Siegel, J. S. Phys. Chem.
Chem. Phys. 2008, 10, 2686−2694.
Experimental details, synthetic procedures, H and 13C NMR
1
spectra, X-ray data. This material is available free of charge via
the Internet at http://pubs.acs.org.
(27) Wheeler, S. E.; Houk, K. N. J. Am. Chem. Soc. 2008, 130,
10854−10855.
(28) Seo, J.-I.; Kim, I.; Lee, Y. S. Chem. Phys. Lett. 2009, 474, 101−
106.
AUTHOR INFORMATION
■
Corresponding Author
shimizu@mail.chem.sc.edu
(29) Cozzi, F.; Ponzini, F.; Annunziata, R.; Cinquini, M.; Siegel, J. S.
Angew. Chem., Int. Ed. Engl. 1995, 34, 1019−1020.
(30) Liu, T. J.; Schneider, H. J. Angew. Chem. Int. Ed. 2002, 41,
1368−1370.
Present Address
†Tennessee Tech University, Chemistry, Foster Hall (FOST)
107/Box 5055, Cookeville, TN 38505
(31) Carroll, W. R.; Pellechia, P.; Shimizu, K. D. Org. Lett. 2008, 10,
3547−3550.
(32) Li, P.; Zhao, C.; Smith, M. D.; Shimizu, K. D. J. Org. Chem.
2013, 78, 5303−5313.
Notes
The authors declare no competing financial interest.
(33) Paliwal, S.; Geib, S.; Wilcox, C. S. J. Am. Chem. Soc. 1994, 116,
4497−4498.
ACKNOWLEDGMENTS
■
(34) Mati, I. K.; Cockroft, S. L. Chem. Soc. Rev. 2010, 39, 4195−4205.
(35) Arnstein, S. A.; Sherrill, C. D. Phys. Chem. Chem. Phys. 2008, 10,
2646−2655.
This work was supported by the National Science Foundation
(CHE 1310139, CHE 1048629, and CHE 0911616).
G
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(36) Nijamudheen, A.; Jose, D.; Shine, A.; Datta, A. J. Phys. Chem.
Lett. 2012, 3, 1493−1496.
(37) Mooibroek, T. J.; Gamez, P.; Reedijk, J. CrystEngComm 2008,
10, 1501−1515.
(38) Hunter, C. A. Angew. Chem., Int. Ed. 2004, 43, 5310−5324.
(39) Muchowska, K. B.; Adam, C.; Mati, I. K.; Cockroft, S. L. J. Am.
Chem. Soc. 2013, 135, 9976−9979.
(40) Hansch, C.; Leo, A.; Taft, R. W. Chem. Rev. 1991, 91, 165−195.
(41) Yoshitake, Y.; Misaka, J.; Setoguchi, K.; Abe, M.; Kawaji, T.; Eto,
M.; Harano, K. J. Chem. Soc., Perkin Trans. 2 2002, 1611−1619.
(42) Soininen, P.; Haarala, J.; Vepsalainen, J.; Niemitz, M.;
Laatikainen, R. Anal. Chim. Acta 2005, 542, 178−185.
(43) Pauli, G. F.; Goedecke, T.; Jaki, B. U.; Lankin, D. C. J. Nat. Prod.
2012, 75, 834−851.
(44) Wheeler, S. E. J. Am. Chem. Soc. 2011, 133, 10262−10274.
(45) Hunter, C. A.; Lawson, K. R.; Perkins, J.; Urch, C. J. J. Chem.
Soc. Perk. Trans. 2 2001, 651−669.
(46) Rizzo, V.; Pinciroli, V. J. Pharm. Biomed. Anal. 2005, 38, 851−
857.
(47) Bauer, M.; Bertario, A.; Boccardi, G.; Fontaine, X.; Rao, R.;
Verrier, D. J. Pharm. Biomed. Anal. 1998, 17, 419−425.
(48) Barding, G. A. J.; Salditos, R.; Larive, C. K. Anal. Bioanal. Chem.
2012, 404, 1165−1179.
(49) Soininen, P.; Haarala, J.; Vepsalainen, J.; Niemitz, M.;
Laatikainen, R. Anal. Chim. Acta 2005, 542, 178−185.
H
dx.doi.org/10.1021/ja504378p | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX