On the role of anionic ligands in the site-selectivity of oxidative C-H functionalization reactions of arenes

Article abstract of DOI:10.1039/c3sc00017f

The replacement of an acetate ligand for carbonate leads to a reversal in site-selectivity in the Pd-mediated C-H oxidative coupling of benzo[h]quinoline with 1,3-dimethoxybenzene. This report describes Density Functional Theory studies designed to elucidate the origin of this selectivity change. These studies focused on two key mechanistic steps: C-H activation and C-C bond-forming reductive elimination. We considered monometallic and bimetallic reaction pathways for acetate and carbonate conditions. The favored C-H activation pathway proceeds via a concerted metalation deprotonation (CMD) mechanism, independent of the nature of anionic ligand (acetate versus carbonate). The predicted selectivity is ortho/para for the C-H activation for both the acetate and carbonate-ligated Pd complexes. Further, we determined that the reductive elimination step is greatly facilitated by the coordination of benzoquinone (by ΔΔG^? ～ 20 kcal mol ^-1) and is predicted to be meta-meta selective with both anionic ligands. Overall, the DFT studies indicate that the anionic ligand does not induce a mechanism change at the elementary steps, and the predicted selectivity at all steps is equivalent for carbonate and acetate, no matter whether a dinuclear or mononuclear pathway is considered. These studies lead us to propose that the role of the anionic ligand is to control which step of the mechanism is overall selectivity-determining. This proposal has been tested experimentally using appropriately designed experiments. Notably, the insoluble base MgO as an acid trap under acetate conditions (with the aim of making the C-H insertion step less reversible), gave rise to predominant ortho/para selectivity in the presence of acetate, in analogy to the results previously seen under carbonate conditions.

Full text of DOI:10.1039/c3sc00017f

Chemical Science
View Article Online
View Journal
EDGE ARTICLE
On the role of anionic ligands in the site-selectivity of
oxidative C–H functionalization reactions of arenes†
Cite this: DOI: 10.1039/c3sc00017f
a
b
b
*
Italo A. Sanhueza, Anna M. Wagner, Melanie S. Sanford
a
*
and Franziska Schoenebeck
The replacement of an acetate ligand for carbonate leads to a reversal in site-selectivity in the Pd-mediated
C–H oxidative coupling of benzo[h]quinoline with 1,3-dimethoxybenzene. This report describes Density
Functional Theory studies designed to elucidate the origin of this selectivity change. These studies
focused on two key mechanistic steps: C–H activation and C–C bond-forming reductive elimination. We
considered monometallic and bimetallic reaction pathways for acetate and carbonate conditions. The
favored C–H activation pathway proceeds via a concerted metalation deprotonation (CMD) mechanism,
independent of the nature of anionic ligand (acetate versus carbonate). The predicted selectivity is
ortho/para for the C–H activation for both the acetate and carbonate-ligated Pd complexes. Further, we
determined that the reductive elimination step is greatly facilitated by the coordination of
benzoquinone (by DDG‡ ꢀ 20 kcal molꢁ1) and is predicted to be meta–meta selective with both anionic
ligands. Overall, the DFT studies indicate that the anionic ligand does not induce a mechanism change
at the elementary steps, and the predicted selectivity at all steps is equivalent for carbonate and acetate,
no matter whether a dinuclear or mononuclear pathway is considered. These studies lead us to propose
that the role of the anionic ligand is to control which step of the mechanism is overall selectivity-
determining. This proposal has been tested experimentally using appropriately designed experiments.
Notably, the insoluble base MgO as an acid trap under acetate conditions (with the aim of making the
C–H insertion step less reversible), gave rise to predominant ortho/para selectivity in the presence of
acetate, in analogy to the results previously seen under carbonate conditions.
Received 3rd January 2013
Accepted 22nd April 2013
DOI: 10.1039/c3sc00017f
www.rsc.org/chemicalscience
is controlled by catalyst structure and/or reaction conditions
have been reported.2,3 Furthermore, the mechanistic origins of
Introduction
Direct functionalization of C–H bonds via palladium catalysis
has become an increasingly attractive strategy for the synthesis
and derivatization of natural products and pharmaceutically
relevant building blocks.1 Despite vast progress in this eld over
the past 10 years, several key challenges remain. In particular,
the control of site-selectivity in Pd-catalyzed C–H functionali-
zation requires signicant further development, in both reac-
tion discovery and mechanistic understanding.2,3 The most
frequently applied approach to control site-selectivity is the use
of substrates that contain ortho-directing groups, which pre-
organize the Pd-catalyst through coordination.1,4 In marked
contrast, only a few examples of reactions where site-selectivity
the observed selectivities generally remain poorly understood.
In this context, Sanford and co-workers have recently demon-
strated that site-selectivity in Pd-mediated oxidative coupling
reactions is highly dependent on the anionic ligand associated
with the Pd-center.5 When performing stoichiometric experi-
ments with cyclopalladated [BzqPd(II)X]2 (Bzq ¼ benzo[h]quin-
oline) and a large excess of 1,3-dimethoxybenzene (300 equiv.)
in the presence of 4 equiv. of DMSO, and 1 equiv. of benzo-
ꢂ
quinone (BQ) at 150 C (see Scheme 1), it was discovered that
the use of acetate as the anion (X) resulted in a 16 : 1 preference
for the meta–meta oxidative coupling product A. In contrast,
a
¨
Laboratory for Organic Chemistry, ETH Zurich, Wolfgang-Pauli-Strasse 10, 8093
¨
Zurich, Switzerland. E-mail: schoenebeck@org.chem.ethz.ch; Web: http://www.
schoenebeck.ethz.ch/
bDepartment of Chemistry, University of Michigan, 930 North University Avenue, Ann
Arbor, Michigan 48109, USA. E-mail: mssanfor@umich.edu; Web: http://www.umich.
edu/ꢀmssgroup/
† Electronic supplementary information (ESI) available: Computational details,
absolute energies, thermochemistry, and Cartesian coordinates of all calculated
structures; full ref. 6 and experimental details. See DOI: 10.1039/c3sc00017f
Scheme 1 Anion-dependent selectivity in oxidative coupling.
This journal is ª The Royal Society of Chemistry 2013
Chem. Sci.

View Article Online
Chemical Science
Edge Article
substitution of acetate for carbonate gave high selectivity for the as a function of ligand. It is important to note that there are
ortho/para coupling product B (11 : 1 ratio). The origin of this several unknown variables in this sequence. Most importantly,
selectivity reversal is currently not understood,5 and the rational the coordination environment at palladium in the various
design of new, more selective reactions as well as the applica- intermediates and transition states is unknown. Benzoquinone
tion of these transformations to wider classes of substrates will could coordinate to several possible intermediates;5,7 further-
require a detailed understanding of the mechanistic details and more, there is no information available on the precise nature of
factors controlling selectivity. We herein report Density Func- ligands (L).
tional Theory6 studies designed to elucidate the origin of
Given the large number of possible intermediates, we chose
selectivity in this Pd-mediated oxidative coupling between the following strategy to explore the anion effect on selectivity in
benzo[h]quinoline and 1,3-dimethoxybenzene as a function of the oxidative coupling of benzo[h]quinoline and 1,3-dimethoxy-
anionic ligand (Scheme 1). We also demonstrate the design and benzene. First, we calculated the C–H activation transition state
implementation of experiments to conrm our computationally preference as a function of anion (acetate versus carbonate) and
derived conclusions.
mechanism. Then we independently studied the BQ-promoted
reductive elimination (Steps 2/3) and calculated the transition
states and relevant intermediates prior to elimination for the
different anionic ligands and for a number of potential ligation
states. These calculations provide critical insights into the
origins of selectivity reversal as a function of anionic ligand. The
insights gleaned from the DFT studies were then subsequently
tested experimentally.
Results and discussion
Mechanistic studies suggest that this transformation proceeds
via the pathway illustrated in Scheme 2. The two key steps with
respect to site-selectivity are (i) the C–H activation of aryl-H
(Step 1) and (ii) C–C bond-forming reductive elimination (Step
3). Importantly, the reductive elimination step is greatly facili-
tated by the presence of benzoquinone (BQ).5,7 Detailed exper-
imental mechanistic studies suggest that this is due to
coordination of the BQ to a Pd-intermediate (Step 2), and this is
supported by recent computational studies of an oxidative Heck
reaction.8 In addition to facilitating the reductive elimination,
the concentration of benzoquinone can also affect site-selec-
tivity under some conditions. For example, when acetate is the
anionic ligand, a high concentration of BQ (ꢀ10 equiv.) gives
rise to much lower selectivity (A : B ꢀ 1 : 1). Interestingly, and in
marked contrast, with carbonate as the counterion, the A/B
selectivity is independent of BQ concentration.5
On the basis of this general mechanism, the changes in
selectivity as a function of ligand (acetate vs. carbonate) could
potentially be due to: (1) a change in the mechanism of the C–H
activation step upon changing the anionic ligand at Pd, (2) a
change in the site-selectivity of the C–H activation step (with no
change in the overall mechanism), (3) a change in the selectivity
of the quinone-promoted reductive elimination step, or (4) a
change in the selectivity-determining step of the overall reaction
Computational details and choice of system for calculations
All calculations were performed with Gaussian09,6 unless
otherwise stated. The structures presented in the main manu-
script were optimized in the gas-phase using uB97XD9/6-
31G(d,p) and LANL2DZ10 (for Pd and Na). Frequency calcula-
tions were performed on all gas-phase optimized geometries to
verify the nature of all stationary points as either minima or
transition states. Single point energies were then calculated
with M06L11 with SDD12 for Pd (and Na) and 6-31++G(d,p) or
def2-TZVPP. The solvation model COSMO-RS13 was employed to
account for solvation. The standard state was converted to 1 M
in solution (+1.89 kcal molꢁ1).
Dispersive effects are expected to be particularly relevant in
considerations of monomeric versus dimeric pathways and the
Pd/BQ interaction (but see later discussion for potential
overestimation of these effects). Potential inaccuracies in the
analyses of mono versus dimeric pathways due to the basis set
superposition error (BSSE)14 may also arise; we therefore also
considered alternative methods for energy calculations (e.g.
PBE0-D3, PBE0) or basis set (def2-TZVPP) in addition to the
above computational approach. The results are discussed in the
main manuscript.
We undertook the majority of preliminary conformational
searches with uB97XD/3-21G optimization, with later rene-
ment at uB97XD/6-31G(d,p) and M06L treatment. A method
and basis set comparison study for the ortho/para pathway (for
acetate) from Fig. 2 (presented in the ESI†) suggest that this was
adequate. The results of this study are summarized in the ESI
(page S6†) and compare the basis sets 6-31+G(d), 6-31G(d,p) and
3-21G with LANL2DZ (for Pd) for geometry optimizations with
the methods uB97XD or B3LYP15 as well as the single point
energy methods: PBE0, PBE0-D3 (ref. 16) and M06L with 6-
31++G(d,p), SDD (Pd) basis set or def2-TZVPP.11,24 This study
showed that there is barely any dependency of the activation
barriers of the individual steps on the basis set.
Scheme 2 Proposed mechanism of C–H functionalization.5
Chem. Sci.
This journal is ª The Royal Society of Chemistry 2013

View Article Online
Edge Article
Chemical Science
In order for the solvation model, COSMO-RS, to be appli- cyclopalladated dinuclear Pd complex in the C–H arylation of
cable, the experimental conditions need to be homogeneous, so pyridine N-oxide.23 We therefore calculated CMD C–H activation
that the effect of solvation on reactivity can be modeled. The by the dinuclear and mononuclear Pd(II) complex for both
original reaction conditions by Sanford et al. (see Scheme 1, i.e. acetate and carbonate and the two possible sites on the arene
300 equiv. of 1,3-dimethoxybenzene and 4 equiv. DMSO) were (ortho or meta in 1,3-dimethoxybenzene). We applied the
heterogeneous when carbonate was employed as the anionic following naming convention: TSCMD(m–m) or TSCMD(o–p) that
ligand. Thus, we initially performed an experimental solvent would lead to the meta–meta product A or the ortho/para product
screen with varying amounts of DMSO (4 to 300 equivalents) to B, respectively (see Fig. 1 and Scheme 1).
render the reaction homogeneous. The selectivity only changed
slightly under the homogeneous reaction conditions (1,3-
dimethoxybenzene/DMSO, 300 : 50) and the overall selectivity
trend was unaffected (see Table S1–S3, p. S2 in the ESI†). We
also determined experimentally that the counter-cation of the
carbonate salt (Cs+ and Na+ were tested) had no marked effect
on the selectivity (see p. S2 in the ESI†).
On the basis of this data, we dꢁecided to perform the
computational studies using NaCO3 or OAcꢁ as anionic
ligands and applying COSMO-RS calculations for the solvent
mixture.
Study of step 1: C–H activation as a function of anionic
ligand and mechanism. As shown in Scheme 3, there are several
potential mechanisms for the C–H activation step including: (a)
electrophilic aromatic substitution (SEAr),17 (b) concerted met-
alation deprotonation (CMD)18 or (c) a Heck-type arylation
pathway.1,4,19–21
Previous preliminary DFT investigations by the Sanford
group suggested against an electrophilic aromatic substitution
(SEAr) mechanism based on a poor correlation of the relative
energies of the intermediates with experimental data.5a Analo-
gous conclusions have also recently been made in related work
by Zhang et al.22 We therefore focused our investigations on
mechanisms (b) and (c), i.e. the CMD and the Heck type path-
ways (Scheme 3).
The CMD pathway features a six-membered transition state
in which the anionic ligand (acetate or carbonate) participates
in a concerted deprotonation of the arene, while Pd inserts into
the C–H bond (Scheme 3). Recent studies by Hartwig and co-
workers have suggested the potential direct involvement of a
Fig. 1 (a) CMD is the favored C–H activation mechanism. Selectivity (DDG‡) for
ꢁ
CMD-C–H activation for acetate (top) and NaCO3 (bottom), derived from dinu-
clear (Di, red) and mononuclear Pd-complexes (Mon, blue);25 calculated at
COSMO-RS (DMSO/DMB26 ¼ 50/300) M06L/6-31++G(d,p) (SDD for Pd)//uB97X/
6-31G(d,p) (with LANL2DZ for Pd & Na), including standard state conversion.29
Free energies in kcal molꢁ1. (b) Selectivity (DDG‡) for CMD-C–H activation for
ꢁ
acetate (top) and NaCO3 (bottom), derived from dinuclear (Di, red) and mono-
nuclear Pd-complexes (Mon, blue);25 calculated at COSMO-RS (DMSO/DMB26
¼
50/300) M06L/def2-TZVPP//uB97X/6-31G(d,p) (with LANL2DZ for Pd & Na),
including standard state conversion.29 Free energies in kcal molꢁ1
Scheme 3 Mechanistic possibilities for C–H activation.
This journal is ª The Royal Society of Chemistry 2013
.
Chem. Sci.

View Article Online
Chemical Science
Edge Article
Fig. 1a and b illustrate the relative activation barriers and
selectivities (DDG‡) for the CMD C–H activation transition states
to generate the ortho/para (right) or meta–meta isomer (le) with
ꢁ
AcOꢁ (top) and NaCO3 (bottom) as the anion, calculated at
M06L/6-31++G(d,p) (Fig. 1a) and M06L/def2-TZVPP (Fig. 1b)
levels of theory with COSMO-RS solvation (for DMSO/DMB26).24
In the case of the M06L/6-31++G(d,p) basis set, the mono-
metallic and bimetallic pathways are essentially iso-energetic
for acetate, but the dimeric pathway is favored for carbonate
(Fig. 1a).25 For def2-TZVPP basis set (see Fig. 1b), the C–H acti-
vation by a monomeric Pd(II) complex is mostly preferred for
carbonate and acetate ligands. Based on our calculations, C–H
activation at the ortho position [TSCMD(o–p)] is favored over meta
[TSCMD(m–m)] by DDG‡ z 4–7 kcal molꢁ1. This preference is
independent of the nature of the anionic ligand (AcOꢁ versus
NaCO3ꢁ) and nuclearity of Pd-complexes.
We next turned our studies to the Heck type mechanism
(c, Scheme 3). In previous studies by Yu and co-workers, a
change in mechanism from CMD to Heck upon ligand-alter-
ation has been claimed for the C–H activation.3j The Heck-type
pathway involves a four-membered transition structure as
illustrated in Scheme 3.1,22,27 In analogy to the CMD mechanism
above, for each anion, the two alternative isomeric transition
states were calculated. We found that the predicted activation
free energy barriers are considerably higher for the Heck type
pathway (DDG‡ > 20 kcal molꢁ1), suggesting that it is not in
competition with the CMD mechanism. This indicates that a
change in mechanism as the origin of selectivity for acetate
versus carbonate is highly unlikely.
Fig. 2 Free energy profile (in kcal molꢁ1) for acetate (top) or sodium carbonate
(bottom) as anionic ligands. The ortho/para-pathway (mononuclear) was calcu-
lated and various coordination possibilities considered. Calculated at COSMO-RS
(DMSO/DMB ¼ 50/300) M06L/6-31++G(d,p) (SDD for Pd & Na)//uB97X/6-
31G(d,p) (LANL2DZ for Pd & Na), including standard state conversion.29
Overall, these studies indicate that the favored C–H activa-
tion pathway proceeds via a CMD mechanism with ortho/para the reductive elimination with respect to the various possible
selectivity. This matches the selectivity obtained experimen- coordination states.
tally under carbonate conditions, but does not match that seen
Our calculations show that reductive elimination in the
with acetate. We next assessed the subsequent reductive absence of benzoquinone has a prohibitively high barrier of DG‡
elimination step for further insight into the origin of site- > 30 kcal molꢁ1 (see Fig. 2, TSRE1 and TSRE2). Coordination of BQ
selectivity.
to the bis-aryl Pd center signicantly decreases the barrier for
Study of steps 2/3: reductive elimination as a function of reductive elimination to roughly DG‡ z 13–15 kcal molꢁ1, i.e.
anionic ligand and coordination sphere. The precise structural TSRE3. The lowest energy pathway examined is therefore the one
features of the reductive elimination step are currently involving dissociation of the acid from Int1, NaHCO3 respec-
unknown. As indicated in Scheme 2, benzoquinone coordina- tively, followed by reductive elimination via the benzoquinone-
28
tion is believed to occur, but the geometry of the resultant coordinated TSRE3
complex is unknown. Moreover, it is not clear whether coor-
.
Although the reaction is difficult to achieve, requiring the
dination of the ligand (L ¼ conjugate acid of anionic ligand) reactant in large excess (as co-solvent) and 150 ꢂC reaction
occurs during the reductive elimination. Another unknown is temperature over prolonged time, the calculated activation free
whether the reductive elimination takes place via a mono- or energy barriers are still overall rather high. For the carbonate
bimetallic pathway. We initially considered a mononuclear pathway we calculate an overall activation free energy barrier of
pathway and studied the reductive elimination from Pd DG‡ ¼ 35.3 kcal molꢁ1 to reach TSRE3 (see Fig. 2); for acetate the
complexes containing a variety of ligands, involving (i) ben- barrier is even higher: DG‡ ¼ 41.4 kcal molꢁ1. This seems to be a
zoquinone coordination, (ii) benzoquinone-free reductive consequence primarily of dispersion that stabilizes the cyclo-
elimination with only L, or (iii) reductive elimination from a palladated Pd(II) acetate starting complex strongly.30,31 While
complex containing neither benzoquinone nor L as ligands. dispersion is arguably crucial in the system considered (p–p
Fig. 2 gives the free energy proles for the ortho/para-pathways interaction in Pd-dimer, Pd/BQ interaction in reductive elim-
(B-selectivity) for the two anionic ligands, AcOꢁ (top) or ination) the current methodology might overestimate these
NaCO3ꢁ (bottom), and takes into account various coordination effects.30,31
possibilities. The corresponding gure for the meta–meta
To also assess the likelihood of bimetallic, rather than
pathways (A-selectivity) is given in the ESI (Fig. S2, page S4†). monometallic, reductive elimination, we calculated TS geome-
The latter shows analogous activation free energy barriers in tries analogous to TSRE3 from a dinuclear Pd(II). Fig. 3 compares
Chem. Sci.
This journal is ª The Royal Society of Chemistry 2013

View Article Online
Edge Article
Chemical Science
the free energies of the bimetallic (TSRE3-Di) and monometallic
(TSRE3-Mon) reductive elimination transition states. The mono-
nuclear pathway was found to be favored for both anionic
ligands (see Fig. 3). Based on our calculations, the ligand (L)
dissociates from the mononuclear complex prior to reductive
elimination and the TS containing benzoquinone favors the
meta–meta product A under all conditions tested.
Summary of mechanistic ndings. The available data
suggest that with both anionic ligands the lowest energy
pathway for C–H activation involves a CMD mechanism. The
calculations also indicate that the inherent selectivity of C–H
activation does not change upon moving from acetate to
carbonate (Fig. 1). With both of these anionic ligands, C–H
activation to form the ortho/para isomer is favored. This pref-
erence holds, independent of whether C–H activation occurs at
a mononuclear or dinuclear Pd(II) species. Our calculations
further show that the BQ-promoted reductive elimination step
favors the meta–meta isomer, for both mononuclear and dinu-
clear Pd-complexes with both anionic ligands.
Origin of site-selectivity. Having established the favored
reaction pathways for C–H activation and reductive elimination,
we can now compare the predicted selectivities for the two
anionic ligands for di- and mononuclear pathways. Fig. 4a and
5a give an overview of the crucial selectivity-determining tran-
sition states considering bimetallic C–H activation, calculated
with M06L/6-31++G(d,p)/SDD and the COSMO-RS solvation
model for DMSO/DMB26 (50 : 300). Fig. 4b and 5b compare the
Fig. 4 (a) Overview of selectivity controlling steps for the acetate system and
bimetallic C–H activation at COSMO-RS (DMSO/DMB26 ¼ 50 : 300) M06L/6-
31++G(d,p) (SDD for Pd)//uB97X/6-31G(d,p) (with LANL2DZ for Pd & Na). Free
energies illustrated in kcal molꢁ1 29 (b) Overview of selectivity controlling steps for
.
the acetate system at COSMO-RS (DMSO/DMB26 ¼ 50 : 300) M06L/def2-TZVPP//
uB97X/6-31G(d,p) (with LANL2DZ for Pd & Na). Free energies illustrated in kcal
molꢁ1
.
29 At COSMO-RS (DMSO/DMB26 ¼ 50 : 300) M06L/6-31++G (d,p) (SDD for
Pd), the selectivity is 1.1 kcal molꢁ1. Energies in parenthesis are based on B3LYP
geometries [COSMO-RS (DMSO/DMB26 ¼ 50 : 300) M06L/6-31++G(d,p) (SDD for
Pd & Na)//B3LYP/6-31 + G(d) (with LANL2DZ for Pd & Na)].
selectivities for monometallic pathways at the M06L/def2-
TZVPP level of theory.
When acetate acts as the ligand, (Fig. 4a and b), the C–H
activation step favors the ortho/para isomer (dotted red path).
However, the subsequent reductive elimination via TSRE3(o–p) is
close in energy to the C–H activation TS of the competing meta–
meta pathway. Because the C–H insertion step should be
reversible (based on the relative barriers for the two steps, see
Fig. 4a and b), the mixture is therefore expected to partially
equilibrate. The overall predicted selectivity (DDG‡) is small. We
predicted the same small selectivity (DDG‡ z 1 kcal molꢁ1) also
for alternative methods (i.e. PBE0-D3, see ESI, Fig. S3†).
However, throughout the course of the reaction, acetic acid will
be generated in the reaction mixture. With increasing acid
concentration, the reverse reaction of C–H activation would be
expected to be favored, making the reductive elimination the
crucial selectivity-determining step (we present experimental
support of these proposals in the next section).
ꢁ
Fig. 3 Selectivity (DDG‡) for reductive elimination for acetate (top) and NaCO3
(bottom), derived from dinuclear (red) and mononuclear Pd-complexes (blue);
calculated at COSMO-RS (DMSO/DMB26 ¼ 50/300) M06L/6-31++G(d,p) (SDD for
Pd & Na)//uB97X/6-31G(d,p) (with LANL2DZ for Pd & Na), including standard
state conversion.29 Free energies in kcal molꢁ1. The activation free energy barriers
are given in Fig. 2.
This journal is ª The Royal Society of Chemistry 2013
Chem. Sci.

View Article Online
Chemical Science
Edge Article
In the course of the experiment, NaHCO3 will be generated
that could potentially participate in the reaction pathway. We
tested this possibility experimentally, employing NaHCO3 as the
anionic ligand source in place of Na2CO3 under otherwise
identical reaction conditions. We found that NaHCO3 also gives
rise to efficient conversion, and similar selectivity to that found
previously with Na2CO3, i.e. the A : B ratio was 1 : 8 in favor of
ortho/para product B (see page S5 in the ESI†). Thus, if NaHCO3
were to coordinate to Pd(II), this would not affect the overall
selectivity. These results were conrmed computationally also;
see Fig. S4 on page S5 in the ESI† for the computed selectivity
prediction. The calculated selectivity is in favor of ortho/para,
albeit to a smaller extent than with NaCO3ꢁ. Thus, the overall
selectivity that is experimentally observed under Na2CO3
conditions may likely arise from a mixture of Na2CO3 and
NaHCO3 derived reaction pathways.32
In summary, our computational data suggest that the ligand
(acetate versus carbonate) does not induce a change of mecha-
nism at the elementary steps, and the predicted selectivity at all
steps is equivalent for carbonate and acetate. Instead, the
anionic ligand controls which step of the mechanism is overall
selectivity-determining.33 For acetate, the reductive elimination
step is more strongly selectivity-controlling. For carbonate, the
C–H activation predominantly impacts selectivity.
Experimental investigations. If the above proposals are
correct, then it should be possible to obtain greater amounts of
the ortho/para product (i.e. the B-product) in the presence of
acetate, if the reaction conditions are altered so as to render the
C–H insertion step less reversible. This would disfavor equili-
bration and thus less meta–meta product via TSCMD(m–m) (Fig. 4a
and b) would be produced. In order to decrease the reversibility
of the reaction, the generated “free” acetic acid would need to
be removed from the reaction mixture. We hypothesized that
magnesium oxide (MgO), which is commonly used to neutralize
weak acids, could be used as an additional additive to the
original reaction conditions for this purpose (illustrated in
Scheme 1 (ref. 5)). MgO is an ideal base additive because it is
insoluble, which should preclude its direct participation in the
mechanism (via coordination and thereby serving as an anionic
Fig. 5 (a) Overview of selectivity controlling steps for the carbonate system,
considering bimetallic C–H activation; calculated at COSMO-RS (DMSO/DMB26
¼
50 : 300) M06L/6-31++G(d,p) (SDD for Pd
& Na)//uB97X/6-31G(d,p) (with
LANL2DZ for Pd & Na). Free energies illustrated in kcal molꢁ1
.
(b) Overview of
29
selectivity controlling steps for the carbonate system, considering monometallic
pathways; calculated at COSMO-RS (DMSO/DMB26 ¼ 50 : 300) M06L/def2-TZVPP.
Free energies illustrated in kcal molꢁ1
.
29 At COSMO-RS (DMSO/DMB26 ¼ 50 : 300)
M06L/6-31++G (d,p) (SDD for Pd & Na), the selectivity is 5.0 kcal molꢁ1. Energies
in parenthesis are based on B3LYP geometries [COSMO-RS (DMSO/DMB26
¼
50 : 300) M06L/6-31++G(d,p) (SDD for Pd & Na)//B3LYP/6-31 + G(d) (with
LANL2DZ for Pd & Na)].
For the carbonate system, the selectivity was experimentally ligand, for example).
found to be overall ortho/para-selective (see Scheme 1).
To our delight, upon addition of MgO, the oxidative
Computationally, the C–H activation for the meta–meta coupling of 1,3-dimethoxybenzene with cyclopalladated benzo-
pathway, i.e. TSCMD(m–m), was found to be the highest energy [h]quinoline in the presence of acetate as anionic ligand,
TS on the free energy surface. It is therefore disfavored (see showed a decrease in meta–meta selectivity (A : B ¼ 5.0 : 1
Fig. 5a and b). The ortho/para reductive elimination TS versus A : B ¼ 6.3 : 1 with no added MgO), see Fig. 6.
(TSRE3(o–p)) is lower in energy than the C–H activation TS for Furthermore, when the number of equivalents of MgO was
the competing meta–meta-pathway (TSCMD(m–m)) by DDG‡
¼
increased from 8 to 15 (thereby enhancing the efficiency of
1.3–4.3 kcal molꢁ1 (see Fig. 5a and b). The carbonate pathway acid removal), this trend continued (A : B ¼ 3.3 : 1, with 100%
therefore favors the ortho/para product B. Notably, the relative yield). These results are consistent with our computationally
barriers for the C–H activation (DG‡ z 33 kcal molꢁ1) and derived conclusions.
reductive elimination step (DG‡ z 35 kcal molꢁ1) in the
It was previously found that the addition of greater equiva-
carbonate system are similar, which is expected to render this lents of BQ under the acetate conditions renders the reaction
transformation less reversible than the analogous acetate more B-selective.5 When the equivalents of BQ were varied with
reaction. This is also expected to contribute to the observed increasing amounts of MgO, increasing quantities of the B-
selectivity for B. Applying the alternative method PBE0-D3, the product were obtained (see Fig. 7). It should be noted that
almost identical selectivity prole was obtained; (see ESI, though the addition of MgO to the acetate conditions may not
Fig. S3 on page S4†).
lead to a completely irreversible system, which should have
Chem. Sci.
This journal is ª The Royal Society of Chemistry 2013

View Article Online
Edge Article
Chemical Science
Our data suggest that the favored C–H functionalization process
involves
a mononuclear C–H activation for acetate and
carbonate conditions. The C–H activation involves a CMD
mechanism, independent of the nature of anionic ligand
(acetate versus carbonate) and nuclearity of the Pd-complex. The
predicted selectivity is ortho/para for the C–H insertion. The
reductive elimination is greatly facilitated by the coordination
of benzoquinone (by DDG‡ ꢀ 20 kcal molꢁ1) and is predicted to
be meta–meta selective. The anionic ligand (acetate versus
carbonate) does not induce a change in mechanism at the
elementary steps, and the predicted selectivity at all steps is
equivalent for carbonate and acetate, regardless of whether
dinuclear or mononuclear Pd complexes are considered.
However, the ligand does control which step of the mechanism
is overall selectivity-determining. We have tested these conclu-
sions experimentally using a number of appropriately designed
experiments. Notably, using the insoluble base MgO as an acid
trap under acetate conditions (with the aim of making the C–H
insertion step less reversible), gave rise to predominant ortho/
para (B) selectivity in the presence of acetate.
Fig. 6 Experimental tests of calculations. Enhanced B-selectivity observed in the
presence of MgO under acetate conditions.
Overall this combination of theory and experiment provides
a detailed and compelling picture of the origin of site-selectivity
in Pd-mediated oxidative cross-coupling. It is particularly
notable that the hypotheses generated from DFT studies were
conrmed through rationally designed experimental investiga-
tions. We anticipate that similar approaches will provide valu-
able insights into the origin of site-selectivity in a variety of
other Pd-catalyzed C–H functionalization reactions.
Fig. 7 Experimental tests of calculations. Increased B-selectivity observed in the
presence of MgO under acetate conditions.
selectivities similar to those seen under carbonate conditions,
addition of more MgO does lead to more of the B-product.
Finally, we hypothesized that the addition of external acid to
this system should reverse these effects, giving rise to greater
meta–meta selectivity (A-product), since the C–H insertion
would resume reversibility. We tested this scenario in Fig. 8. As
predicted, the proportion of A-product increased in a dose
dependent manner upon the addition of acid. This is consistent
with the hypothesis that the acid moves the system back
towards the original acetate system.
Acknowledgements
We thank the Swiss National Science Foundation and ETH
¨
Zurich for nancial support of the computational studies. The
experimental studies are based upon work supported by the US
Department of Energy [Office of Basic Energy Sciences] DE-
FG02-08 ER 15997. AMW thanks the US National Science
Foundation for a graduate fellowship. Calculations were per-
formed on the ETH high performance cluster “Brutus”.
Conclusions
Notes and references
We have conducted DFT investigations to elucidate the origin of
site-selectivity in the oxidative coupling of 1,3-dimethoxy-
benzene and benzo[h]quinoline as a function of anionic ligand.
1 For reviews, see: (a) L. Ackermann, R. Vicente and
A. R. Kapdi, Angew. Chem., Int. Ed., 2009, 48, 9792; (b)
D. Alberico, M. E. Scott and M. Lautens, Chem. Rev., 2007,
107, 174; (c) T. W. Lyons and M. S. Sanford, Chem. Rev.,
2010, 110, 1147; (d) R. Giri, B.-F. Shi, K. M. Engle,
N. Maugel and J.-Q. Yu, Chem. Soc. Rev., 2009, 38, 3242; (e)
L. McMurray, F. O'Hara and M. J. Gaunt, Chem. Soc. Rev.,
2011, 40, 1885.
2 S. R. Neufeldt and M. S. Sanford, Acc. Chem. Res., 2012, 45,
936.
3 (a) A. N. Campbell, E. B. Meyer and S. S. Stahl, Chem.
Commun., 2011, 47, 10257; (b) M. H. Emmert, A. K. Cook,
Y. J. Xie and M. S. Sanford, Angew. Chem., Int. Ed., 2011,
50, 9409; (c) A. J. Hickman and M. S. Sanford, ACS Catal.,
2011, 1, 170; (d) X. Wang, D. Leow and J.-Q. Yu, J. Am.
Chem. Soc., 2011, 133, 13864; (e) Y. Izawa and S. S. Stahl,
Fig. 8 Experimental tests of calculations. Upon addition of external AcOH to the
system with MgO, the selectivity favors product A once again.
This journal is ª The Royal Society of Chemistry 2013
Chem. Sci.

View Article Online
Chemical Science
Edge Article
Adv. Synth. Catal., 2010, 352, 3223; (f) Y.-H. Zhang, B.-F. Shi
and J.-Q. Yu, J. Am. Chem. Soc., 2009, 131, 5072; (g)
D. J. Schipper, L.-C. Campeau and K. Fagnou, Tetrahedron,
2009, 65, 3155; (h) S. Potavathri, A. S. Dumas, T. A. Dwight,
D. C. Powers, D. Benitez, E. Tkatchouk, W. A. Goddard, III
and T. Ritter, J. Am. Chem. Soc., 2010, 132, 14092; (h)
D. C. Powers, E. Lee, A. Ariafard, M. S. Sanford, B. F. Yates,
A. J. Canty and T. Ritter, J. Am. Chem. Soc., 2012, 134, 12002.
G. R. Naumiec, J. M. Hammann and B. DeBoef, 12 On the performance of ECPs for DFT, see: (a) X. Xu and
Tetrahedron Lett., 2008, 49, 4050; (i) L. Eberson and
D. G. Truhlar, J. Chem. Theory Comput., 2012, 8, 80; (b)
K. A. Peterson, D. Figgen, M. Dolg and H. Stoll, J. Chem.
Phys., 2007, 126, 124101.
¨
L. Jonsson, Acta Chem. Scand., Ser. B, 1974, 28, 771; (j)
K. M. Engle, D.-H. Wang and J.-Q. Yu, J. Am. Chem. Soc.,
2010, 132, 14137; (k) S. I. Gorelsky, D. Lapointe and 13 (a) The solvation contribution from COSMO-RS was
K. Fagnou, J. Org. Chem., 2012, 77, 658; (l) S. L. Gorelsky,
Organometallics, 2012, 31, 794; (m) A. Petit, J. Flygare,
A. T. Miller, G. Winkel and D. H. Ess, Org. Lett., 2012, 14,
3680; (n) E. Clot, O. Eisenstein, N. Jasim, S. A. Macgregor,
J. E. McGrady and R. N. Perutz, Acc. Chem. Res., 2011, 44, 333.
4 A. R. Dick and M. S. Sanford, Tetrahedron, 2006, 62, 2439.
5 (a) T. W. Lyons, K. L. Hull and M. S. Sanford, J. Am. Chem.
calculated with COSMOtherm (F. Eckert, A. Klamt,
COSMOtherm, Version C2.1, Release 01.11, COSMOlogic
GmbH & Co. KG, Leverkusen, Germany, 2010), based on
BP86/TZVP single-point energy calculation using Orca (F.
Neese, ORCA, Ver. 2.9 (Rev 0), MPI for Bioinorganic
Chemistry, Germany, 2011) at 298 K; (b) A. Klamt, F. Eckert
and W. Arlt, Annu. Rev. Chem. Biomol. Eng., 2010, 1, 101.
Soc., 2011, 133, 4455; (b) K. L. Hull and M. S. Sanford, J. 14 (a) H. B. Jansen and P. Ros, Chem. Phys. Lett., 1969, 3, 140; (b)
Am. Chem. Soc., 2007, 129, 11904; (c) K. L. Hull and
M. S. Sanford, J. Am. Chem. Soc., 2009, 131, 9651.
6 M. J. Frisch et al., Gaussian09, Revision A.01, for full
reference, see ESI.†
7 For other examples of BQ-promoted reductive elimination,
see: (a) J. S. Temple, M. Riediker and J. Schwartz, J. Am.
Chem. Soc., 1982, 104, 1310; (b) J. E. Backvall, S. E. Bystrom
and R. E. Nordberg, J. Org. Chem., 1984, 49, 4619; (c)
J. E. Backvall, R. E. Nordberg and D. Wilhelm, J. Am. Chem.
B. Liu and A. D. McLean, J. Chem. Phys., 1973, 59, 4557.
15 (a) A. D. Becke, J. Chem. Phys., 1993, 98, 5648; (b) C. Lee,
W. Yang and R. G. Parr, Phys. Rev. B, 1988, 37, 785.
16 (a) C. Adamo and V. Barone, J. Chem. Phys., 1999, 110, 6158;
(b) S. Grimme, J. Antony, S. Ehrlich and H. Krieg, J. Chem.
Phys., 2010, 132, 154104; (c) S. Grimme, J. Comput. Chem.,
2006, 27, 1787. For application of DFT-D in the study of
organometallic catalysis, see: (d) C. L. McMullin, J. Jover,
J. N. Harvey and N. Fey, Dalton Trans., 2010, 39, 10833.
Soc., 1985, 107, 6892; (d) A. C. Albeniz, P. Espinet and 17 For proposals of the SEAr mechanism, see: (a) M. Catellani
B. Martin-Ruiz, Chem.–Eur. J., 2001, 7, 2481; (e) M. S. Chen,
N. Prabagaran, N. A. Labenz and M. C. White, J. Am. Chem.
Soc., 2005, 127, 6970; (f) X. Chen, J. J. Li, X. S. Hao,
C. E. Goodhue and J. Q. Yu, J. Am. Chem. Soc., 2006, 128,
and G. P. Chiusoli, J. Organomet. Chem., 1992, 425, 151; (b)
C. C. Hughes and D. Trauner, Angew. Chem., Int. Ed., 2002,
41, 1569; (c) E. J. Hennessy and S. L. Buchwald, J. Am.
Chem. Soc., 2003, 125, 12084; (d) B. S. Lane, M. A. Brown
and D. Sames, J. Am. Chem. Soc., 2005, 127, 8050.
´
78; (g) M. Perez-Rodrıguez, A. A. C. Braga, M. Garcia-
Melchor, M. H. Perez-Temprano, J. A. Casares, G. Ujaque, 18 (a) V. I. Sokolov, L. L. Troitskaya andO. A. Reutov, J. Organomet.
A. R. deLera, R. Alvarez, F. Maseras and P. Espinet, J. Am.
Chem. Soc., 2009, 131, 3650; (h) M. P. Lanci, M. S. Remy,
W. Kaminsky, J. M. Mayer and M. S. Sanford, J. Am. Chem.
Soc., 2009, 131, 15618; (i) G. Yin, Y. Wu and G. Liu, J. Am.
Chem. Soc., 2010, 132, 11978; (j) A. Ishikawa, Y. Nakao,
H. Sato and S. Sakaki, Dalton Trans., 2010, 39, 3279.
Chem., 1979, 182, 537; (b) A. D. Ryabov, I. K. Sakodinskaya and
A. K. Yatsimirsky, J. Chem. Soc., Dalton Trans., 1985, 2629; (c)
´
M. Gomez, J. Granell and M. Martinez, Organometallics,
´
1997, 16, 2539; (d) M. Gomez, J. Granell and M. Martinez, J.
Chem. Soc., Dalton Trans., 1998, 37; (e) B. Biswas,
M. Sugimoto and S. Sakaki, Organometallics, 2000, 19, 3895;
(f) D. L. Davies, S. M. A. Donald and S. A. Macgregor, J. Am.
¨
8 C. Skold, J. Kleimark, A. Trejos, L. R. Odell, S. O. Nilsson Lill,
´
P.-O. Norrby and M. Larhed, Chem.–Eur. J., 2012, 18, 4714.
9 J.-D. Chai and M. Head-Gordon, Phys. Chem. Chem. Phys.,
2008, 10, 6615.
10 On the appropriateness of LANL2DZ as ECP for Pd, see: (a)
L. E. Roy, P. J. Hay and R. L. Martin, J. Chem. Theory
Comput., 2008, 4, 1029; (b) G. T. de Jong, M. Sola,
L. Visscher and F. M. Bickelhaupt, J. Chem. Phys., 2004,
121, 9982.
Chem. Soc., 2005, 127, 13754; (g) D. Garcıa-Cuadrado,
A. A. C. Braga, F. Maseras and A. M. Echavarren, J. Am. Chem.
Soc., 2006, 128, 1066; (h) M. Lafrance and K. Fagnou, J. Am.
Chem. Soc., 2006, 128, 16496; (i) M. Lafrance, C. N. Rowley,
T. K. Woo and K. Fagnou, J. Am. Chem. Soc., 2006, 128, 8754;
(j) D. Garcia-Cuadrado, P. de Mendoza, A. A. C. Braga,
F. Maseras and A. M. Echavarren, J. Am. Chem. Soc., 2007,
129, 6880; (k) S. I. Gorelsky, D. Lapointe and K. Fagnou, J.
Am. Chem. Soc., 2008, 130, 10848.
11 (a) Y. Zhao, N. E. Schultz and D. G. Truhlar, J. Chem. Theory
Comput., 2006, 2, 364; (b) Y. Zhao and D. G. Truhlar, J. Chem. 19 J. Le Bras and J. Muzart, Chem. Rev., 2011, 111, 1170.
Phys., 2006, 125, 194101; (c) Y. Zhao and D. G. Truhlar, J. 20 (a) D. Lapointe and K. Fagnou, Chem. Lett., 2010, 39, 1118; (b)
Phys. Chem. A, 2006, 110, 13126. For use of M06L in related
systems: (d) M. C. Nielsen, E. Lyngvi and F. Schoenebeck,
J. Am. Chem. Soc., 2013, 135, 1978; (e) A. J. Canty,
A. Ariafard, M. S. Sanford and B. F. Yates, Organometallics,
2013, 32, 544; (f) Y. Minenkov, A. Singstad, G. Occhipinti
and V. R. Jensen, Dalton Trans., 2012, 41, 5526; (g)
Y. Boutadla, D. L. Davies, S. A. Macgregor and A. I. Poblador-
Bahamonde, Dalton Trans., 2009, 5820; (c) I. V. Seregin and
V. Gevorgyan, Chem. Soc. Rev., 2007, 36, 1173; (d) C. J. Engelin
and P. Fristrup, Molecules, 2011, 16, 951; (e) D. Balcells, E. Clot
and O. Eisenstein, Chem. Rev., 2010, 110, 749; (f) J. Guihaume,
E. Clot, O. Eisenstein and R. N. Perutz, Dalton Trans., 2010, 39,
Chem. Sci.
This journal is ª The Royal Society of Chemistry 2013

View Article Online
Edge Article
Chemical Science
10510; (g) D. H. Ess, R. J. Nielsen, W. A. Goddard III and 28 Attempts to locate TSs involving acid and BQ-coordination
R. A. Periana, J. Am. Chem. Soc., 2009, 131, 11686. led to dissociation of the acid in all cases.
21 Another alternative involving oxidative addition of the C–H 29 The standard state was converted to 1 M in solution (+1.89
bond to PdII to form a PdIV aryl hydride intermediate has kcal molꢁ1).
previously been excluded as the mechanism. For examples, 30 In this context, it has recently been found that dispersion-
see: (a) Y. Boutadla, D. L. Davies, S. A. Macgregor and
A. J. Poblador-Bahamonde, Dalton Trans., 2009, 5820–5831; (b)
D. G. Musaev, A. Kaledin, B. F. Shi and J. Q. Yu, J. Am. Chem.
Soc., 2012, 134, 1690; (c) M. Albrecht, Chem. Rev., 2010, 110, 576.
22 S. Zhang, L. Shi and Y. Ding, J. Am. Chem. Soc., 2011, 133,
20218.
corrected functionals have the tendency to overestimate
the binding energy of large ligands to transition metal
complexes in solution, see: H. Jacobsen and L. Cavallo,
¨
ChemPhysChem, 2012, 13, 562. See also: P. Ronnholm,
¨
S. O. Nilsson Lill, J. Grafenstein, P.-O. Norrby,
M. Pettersson and G. Hilmersson, ChemPlusChem, 2012,
77, 799.
23 Y. Tan, F. Barrios-Landeros and J. F. Hartwig, J. Am. Chem.
Soc., 2012, 134, 3683.
31 In line with this, at COSMO-RS (DMSO/DMB ¼ 50/300) PBE0/
6–31++G(d,p) (SDD for Pd)//uB97XD/6–31G(d,p) (LANL2DZ
for Pd & Na) method of theory (i.e. a method that accounts
less well for dispersion) the barrier for C–H activation
involving acetate is calculated to be lower (DG‡ ¼ 29.6 kcal
molꢁ1).
24 For application of DFT methods for metals, see: (a)
G. Frenking, Theoretical Aspects of Transition Metal
Catalysis (Topics in Organometallic Chemistry), Springer-
Verlag, Berlin Heidelberg, 2005; (b) T. Strassner and
M. A. Taige, J. Chem. Theory Comput., 2005, 1, 848; (c)
C. J. Cramer and D. G. Truhlar, Phys. Chem. Chem. Phys., 32 In line with this, our experimental selectivity analysis of the
2009, 11, 10757; (d) G. T. de Jong and F. M. Bickelhaupt, J.
Chem. Theory Comput., 2006, 2, 322.
Na2CO3 reactions over time revealed that the selectivity was
highly ortho/para selective early on in the reaction (30 : 1,
aer 0.5 h), but less later on (8 : 1, aer 5 h). See ESI,†
page S7 for further information.
25 The bimetallic pathways are also favored for carbonate and
acetate if COSMO-RS (DMSO/DMB ¼ 50/300) PBE0–D3/6-
31++G(d,p)/SDD//wB97XD/6–31G(d,p) (LANL2DZ for Pd & 33 This scenario was presented as one mechanistic possibility
Na) is applied. (See Fig. S1 on page S3 in the ESI†).
26 DMB ¼ 1,3-dimethoxybenzene.
in our original report of this reaction (see ref. 5a, endnote
24). We originally disfavored it based on a few preliminary
experiments. However, the computations presented herein
in conjunction with the more extensive additional
experiments provide strong support in its favor.
27 (a) S.-Y. Tang, Q.-X. Guo and Y. Fu, Chem.–Eur. J., 2011, 17,
13866; (b) B. Glover, K. A. Harvey, B. Liu, M. J. Sharp and
M. F. Tymoschenko, Org. Lett., 2003, 5, 301.
This journal is ª The Royal Society of Chemistry 2013
Chem. Sci.