Article

pubs.acs.org/accounts

Computation and Experiment: A Powerful Combination to

Understand and Predict Reactivities
Published as part of the Accounts of Chemical Research special issue “Computational Catalysis for Organic
Synthesis”.
Theresa Sperger,† Italo A. Sanhueza,†,‡ and Franziska Schoenebeck*,†
†
Institute of Organic Chemistry, RWTH Aachen University, Landoltweg 1, 52074 Aachen, Germany
‡
Laboratory for Organic Chemistry, ETH Zürich, Vladimir-Prelog-Weg 3, 8093 Zürich, Switzerland
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
Downloaded via UNIV FED DO CEARA on July 31, 2018 at 17:44:16 (UTC).

CONSPECTUS: Computational chemistry has become an established tool for the study of the origins of chemical phenomena
and examination of molecular properties. Because of major advances in theory, hardware and software, calculations of molecular
processes can nowadays be done with reasonable accuracy on a time-scale that is competitive or even faster than experiments.
This overview will highlight broad applications of computational chemistry in the study of organic and organometallic reactivities,
including catalytic (NHC-, Cu-, Pd-, Ni-catalyzed) and noncatalytic examples of relevance to organic synthesis. The selected
examples showcase the ability of computational chemistry to rationalize and also predict reactivities of broad significance. A
particular emphasis is placed on the synergistic interplay of computations and experiments. It is discussed how this approach
allows one to (i) gain greater insight than the isolated techniques, (ii) inspire novel chemistry avenues, and (iii) assist in reaction
development. Examples of successful rationalizations of reactivities are discussed, including the elucidation of mechanistic
features (radical versus polar) and origins of stereoselectivity in NHC-catalyzed reactions as well as the rationalization of ligand
effects on ligation states and selectivity in Pd- and Ni-catalyzed transformations. Beyond explaining, the synergistic interplay of
computation and experiments is then discussed, showcasing the identification of the likely catalytically active species as a function
of ligand, additive, and solvent in Pd-catalyzed cross-coupling reactions. These may vary between mono- or bisphosphine-bound
or even anionic Pd complexes in polar media in the presence of coordinating additives. These fundamental studies also inspired
avenues in catalysis via dinuclear Pd(I) cycles. Detailed mechanistic studies supporting the direct reactivity of Pd(I)−Pd(I) with
aryl halides as well as applications of air-stable dinuclear Pd(I) catalysts are discussed. Additional combined experimental and
computational studies are described for alternative metals, these include the discussion of the factors that control C−H versus
C−C activation in the aerobic Cu-catalyzed oxidation of ketones, and ligand and additive effects on the nature and favored
oxidation state of the active catalyst in Ni-catalyzed trifluoromethylthiolations of aryl chlorides. Examples of successful
computational reactivity predictions along with experimental verifications are then presented. This includes the design of a
fluorinated ligand [(CF3)2P(CH2)2P(CF3)2] for the challenging reductive elimination of ArCF3 from Pd(II) as well as the
guidance of substrate scope (functional group tolerance and suitable leaving group) in the Ni-catalyzed trifluoromethylthiolation
of C(sp2)−O bonds. In summary, this account aims to convey the benefits of integrating computational studies in experimental
research to increase understanding of observed phenomena and guide future experiments.

1. INTRODUCTION as “physical organic chemistry”, has specialized in utilizing

The continuous search for the origins of chemical phenomena analytical tools as well as developing approaches (e.g., Hammett
has stimulated generations of organic chemists to make use of
physical tools to gain insight on structure, properties and Received: February 8, 2016
reactivity relationships. This subdiscipline, commonly referred to Published: May 12, 2016

© 2016 American Chemical Society 1311 DOI: 10.1021/acs.accounts.6b00068

Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research
plots, kinetic analyses, isotope effects...) to identify mechanistic
details and/or gain indirect information on the nature of
transition states. A more recent addition to the plethora of tools
is computational chemistry. Owing to the tremendous progress
in the development of hardware, software, and theoretical
methods, computational chemistry has evolved to a powerful tool
that allows one to gain mechanistic information and study the
nature of transition states directly, allowing therefore the analysis
of the origins of selectivity or the study of more complex
reactivity scenarios, such as those related to catalysis as well as
property prediction.
There are numerous computational tools available, each of
which has its own strengths and weaknesses, as well as
applications for which it has been developed.1 In reactivity
studies of small organic and organometallic molecules (∼60
atoms, excluding hydrogen), quantum mechanical approaches of
good accuracy are nowadays fast enough to be competitive with
experiments. To give an illustration of the tremendous progress
in possibilities: if we wanted to study the bond making and
breaking of a Diels−Alder reaction, one would typically optimize
the transition state (TS) in a static fashion employing quantum
mechanics. While calculation times of about 6 months were
necessary in the 1980s,2 using Hartree−Fock (HF) theory and a
small basis set on a supercomputer of that time, nowadays, the
exact same calculation can be done in roughly 15 s.3 For
comparison, a standard DFT approach takes 258 s for the same
TS optimization.4
Since calculations are nowadays this fast, sometimes even
faster than experiments, it is reasonable to question why
calculations are not universally applied, instead of doing
experiments. This is on the one hand ascribable to the accuracy
(or lack thereof) of the theoretical method employed, but also
the fact that the interesting reactivity problems tend to be more
complex, larger in size and with many more mechanistic
possibilities. Consequently, the time it takes to gain unambig-
uous mechanistic information is also increased. However, when
teamed with experiments, computations can be very valuable,
and the synergistic approach of computation and experiment will
arguably gain more insight than any of the isolated techniques. In
this account we describe our activities in using computational
chemistry to (i) understand organic and organometallic
transformations as well as (ii) guide our experiments in the
development of novel chemistry. In this context, we will
showcase the predictive capability of computational chemistry
and will emphasize the synergistic power of combining
calculations with experiments.
2. STUDYING ORGANIC (NONCATALYTIC)
TRANSFORMATIONS
Several decades of excellent physical organic chemistry research
provided chemists with an intuitive reactivity understanding of
potential and reasonable mechanistic scenarios for many organic
transformations.5 However, to be able to explain certain
selectivities or slight variations in electronic properties (e.g.,
polar versus radical mechanisms), experimental mechanistic
approaches have been limited to provide thorough under-
standing, since the nature of the transition state is the key
controlling factor. It is in these cases where computations have
provided valuable insight. Over the past years, our research group
has studied organic reactivity aspects in relation to functional
organic molecules and synthesis. These include the mechanistic
details of rearrangements of push−pull chromophores6 (includ-
ing those connected to fullerenes7), enantiomerization processes
1312
Article
of chiral organic cyclo-adducts,8 reactivity and selectivity aspects
of novel amide bond formation,9,10 the origins of selectivity in
natural product synthesis,11 and property calculations in a
medicinal chemistry context.12 Many of these investigations were
undertaken in collaboration with researchers from academia and
industry.
An example is our study of the enantiomerization mechanism
of strained cyclic buta-1,3-dienes, in collaboration with Diederich
and co-workers.8 This project aimed to prepare optically active
compounds that may be utilized as chiral memory units or chiral
sensors. However, key to application is that the molecule remains
chiral. Thus, an understanding (as well as quantification) of the
enantiomerization mechanism could aid the development of
appropriate molecules. While chemical intuition might suggest
that chair-conformers are adopted along the pathway, it was
computationally found that the energetically most feasible route
involves an initial flip from the chairlike geometry to the
energetically higher boat-conformer, followed by isomerization.
Notably, the predicted activation barrier for the process was in
excellent agreement with experimental measurements (Scheme
1a).
Scheme 1. (a) Enantiomerization of Cyclic Buta-1,3-dienes
via boat-TS and (b) Ring Contraction of Spirocyclopropane
Isoxazolidines
a
Energies (in kcal/mol).
Challenging synthetic problems have always stimulated
mechanistic investigations and reaction development. The
Carreira group developed an elegant total synthesis of
gelsemoxonine, involving a ring contraction of a spirocyclopro-
pane isoxazolidine as the key step.11 Experimental mechanistic
probes revealed that the process is highly stereoretentive,
indicating that the mechanism is either concerted or proceeds
via intermediates that cannot undergo rotation. Computations
revealed that initial N-protonation of the isoxazolidine is crucial
to trigger the ring contraction (lowering the barrier by 20 kcal/
mol relative to the nonprotonated pathway). The ethylene
extrusion was found to be concerted and no intermediate was
formed along the reaction coordinate, converting directly into
the products. While the N−O cleavage transition state was found
to have some biradical character (revealed through (U)CCSD-
(T), CASSCF, and RO−DFT methodology), there is a transition
to a polar mechanism along the reaction coordinate (Scheme
1b).
3. STUDYING ORGANOMETALLIC
TRANSFORMATIONS
Metals (whether heterogeneous, homogeneous, or within an
enzyme pocket) may induce reactivities that cannot be achieved
DOI: 10.1021/acs.accounts.6b00068
Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research
Table 1. Commonly-Employed DFT Methods in the Schoenebeck Group to Study Organic and Organometallic Reactivities
organic
geometry optimization B3LYP
energy calculation (single point energy correction) B3LYP, M06-2X
Scheme 2. Dispersion Allows Location of Bisligated Oxidative Addition TS of Aryl Triflate
through purely organic reactivity modes. Organometallic
chemistry has therefore been coined as a field that likely provides
future innovative synthetic transformations.13 Given the unique
reactivities, organometallic transformations are currently less
intuitive, making it often challenging to predict the reaction
outcome a priori. Consequently, the choice of ideal additive,
solvent, and metal for the desired reactivity and selectivity is
frequently a result of experimental screening and/or serendipity.
This is particularly the case for catalysis-related research. To
guide research toward innovative solutions, greater under-
standing of the underlying processes will be valuable. Using a
synergistic approach of computation and experiment, our group
succeeded in gaining greater understanding but also in guiding
novel experiments. In terms of applicable computational
methodology, our group recently undertook a statistical analysis
of commonly employed DFT methods in the study of
organometallic transformations (involving Pd, Ni, Rh, and Ir).
This analysis showed that the B3LYP method is rather popular in
calculations involving transition metals, in particular for the
optimization of structures.14 However, for energy calculations,
we and others generally employ methods that better account for
dispersion since many organometallic transformations involve
significant intramolecular dispersive interactions. Table 1
summarizes DFT-methods most frequently employed in our
group for the study of organic and organometallic investigations.
While many similarities exist between our organic and organo-
metallic computational approaches, we preferred the MO6L-
method in the computational evaluation of organometallic
systems.15,16
Although B3LYP can give excellent results and be sufficient in
geometry optimizations, if dispersion is not crucial, questions
addressing nuclearity or ligation state will require the
consideration of dispersion. An example from our lab is the
search for bisligated transition states for the oxidative addition of
aryl triflates to Pd(PtBu3)2.17 While methods that do not account
well for dispersion, such as B3LYP or PBE0, resulted in ligand
dissociation, the inclusion of dispersion (ωB97XD or B3LYP-D3
methodology) allowed for the first location of a bisligated
oxidative addition transition state with Pd(PtBu3)2 (Scheme 2).
Importantly, for bulky phosphine ligands, such as PtBu3, it is
commonly assumed that a monoligated “Pd(0)L” species would
be reactive, on the basis of the previous computational inability to
locate bisphosphine transition states as well as experimental
kinetic data. The inclusion of dispersion in calculations suggested
an associative mechanism with a substrate to be preferred over
the previously proposed dissociation of one phosphine ligand to
1313
Article
organometallic
B3LYP, B3LYP-D3, ωB97XD
M06L, M06, B3LYP-D3, PBE0-D3, ωB97XD
form a monoligated Pd(0) species. This implies that no “Pd(0)L”
is ever “free” in solution but rather a species that is bound to a
substrate or solvent.18,19
4. MECHANISTIC STUDIES OF TRANSFORMATIONS
RELATED TO CATALYSIS
4.1. General Remarks
Catalysis is highly important and arguably a key discipline where
chemists might make a contribution to finding solutions to global
challenges and creating a more sustainable life. It also
revolutionized the way chemists assemble molecules and has
consequently been recognized with several Nobel prizes.
The computational study of catalytic transformations differs
from its corresponding noncatalytic counterpart in several
aspects:20 (i) there are frequently various intermediates and
steps to consider within a complex catalytic cycle, which may in
turn depend on additives and conditions employed; (ii) several
competing pathways need to be evaluated, discriminated only by
very small energetic differences; and (iii) there is a need to
address several conformers that arise from flexible catalysts and/
or ligands.
An exemplification of the aforementioned challenges is our
study on the origins of stereoselectivity of an NHC-catalyzed
annulation reaction, developed by Bode and co-workers.21 While
a Claisen rearrangement was identified as the stereodetermining
step, the favored protonation states (i.e., enol or enolate) of the
transition state and intermediates were unknown. We inves-
tigated the factors that control stereoselectivity and found that
the protonated transition states for the two competing
stereochemical outcomes gave no stereoselectivity. Moreover,
the consideration of various potentially reactive conformers in
the context of a Boltzman-weighted average of the transition
state (TS) ensembles showed that presence of a proton would
diminish selectivity. Only a deprotonated pathway would result
in discrimination of stereoisomers, consistent with electrostatic
repulsion of the formed enolate with the π-system of the phenyl-
substituted acyl azolium-catalyst in the disfavored transition state
(Scheme 3). Notably, while different mechanisms were assumed
(Claisen rearrangement versus Michael addition), the computed
transition states are in this case structurally identical for both
Claisen rearrangement and conjugate addition.
A similar mechanistic analysis was conducted in collaboration
with the Enders group. In this study, also multiple mechanistic
scenarios were considered, such as anionic, neutral, and radical
pathways in addition to exo and endo selectivity modes of ring
DOI: 10.1021/acs.accounts.6b00068
Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research Article

Scheme 3. Enantioselective NHC-Catalyzed Annulation via ing, since the precise mechanisms and structural features of key
Deprotonated Pathway catalytic species are unknown. In this regard, we set out to
investigate the influences of catalyst, substrate, solvent, and base
on the selectivity of Cu-catalyzed aerobic oxidations of ketones
(Scheme 4).25 Key experiments were conducted that revealed

Scheme 4. Mechanistic Divergence in Cu-Catalyzed Aerobic

Oxidation Resulting from an α-Peroxo Ketone As a Common
Intermediate

closures in the sequential organo- and Ag-catalyzed synthesis of

spiropyrazolones.22 Our calculations at the CPCM (DCM)
M06/def2-TZVP//B3LYP/6-31G(d) level of theory suggested
that a neutral pathway possesses prohibitively high activation
barriers, and that instead an anionic pathway is adopted favoring
the 5-exo-dig cyclization. The role of Ag is to overcome the
endergonicity of this step, rendering the cyclization thermody-
namically favored. Notably, depending on the substrate, there
may be a divergence in mechanism. An alternative Ag-free that the selectivity of oxidation likely arises from a common
process was experimentally found to only occur in the presence intermediate, i.e., an α-peroxo ketone. While both C−C bond
of oxygen and could computationally be attributed to an cleavage and C−H functionalization were computationally found
oxidation-induced radical pathway. This alternative pathway was to be consistent with a radical mechanism, the precise formation
favored only for a phenyl-substituted substrate and significantly of alcohol from the peroxide intermediate remained unknown.
less for other substrates explaining their lack of reactivity under Thus, we synthesized the putative peroxide intermediate and
Ag-free conditions. experimentally uncovered a novel base-triggered reduction
These examples illustrate that the number of possibilities to mechanism.
explore in the context of catalysis is tremendously increased. This Similarly, combining calculations with experiments aided our
inherent complexity of the chemical system is a reason as to why mechanistic study on the exclusively trans-selective chlorocarba-
transformations are not uniformly calculated, instead of doing moylation of alkynes.26 In collaboration with the Lautens group,
experiments. However, the study of specific mechanistic we investigated two plausible mechanistic pathways that can
questions of single steps, particularly in a qualitative manner, account for the exclusive formation of trans-product (Scheme 5).
i.e., when comparing two systems relative to each other and when An ionic trans-chloropalladation pathway could be ruled out
teamed up with experiments is likely much more powerful and experimentally, since no halide exchange was observed upon
even predictive, as we will discuss below.
4.2. Combined Experimental and Computational Approach Scheme 5. Proposed Mechanistic Pathways of the trans-
to Gain Mechanistic Information on Reactivity- and Selective Chlorocarbamoylation of Alkynes
Selectivity-Controlling Factors in Catalysis
We have studied the effects of ligand, additive, and solvent to gain
insights on active catalytic species, potential mechanistic
pathways, and key controlling factors.14,20,23 In this context,
the use of a combined computational and experimental approach
has proven highly beneficial: on the one hand, computations
allow the study of key intermediates and also transition states,
which are challenging to monitor via conventional experimental
approaches. On the other hand, experiments give valuable insight
on computationally challenging scenarios (such as ionic
pathways, small energetic differences, method dependences)24
and can therefore be used to validate computational results or
assist computations. While this synergistic approach is, in
principle, applicable to a variety of metal-containing and metal-
free reactivities, we will showcase selected examples of our own
work in relation to Cu-, Pd-, and Ni-catalysis.
There has been significant interest in aerobic homogeneous
catalysis with nonprecious metals, but one of the key challenges is
its selectivity and control thereof. Additionally, the under-
standing of underlying selectivity-controlling factors is challeng-
1314 DOI: 10.1021/acs.accounts.6b00068
Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research
addition of an exogeneous halide. An alternative pathway
involving cis-carbopalladation and fast subsequent cis-trans
isomerization was further studied by means of calculations at
the CPCM (toluene) M06L/def2-TZVP//B3LYP/6-31G(d)
(LANL2DZ) level of theory. Calculations indicated that both cis-
carbopalladation as well as cis-trans isomerization proceed with
feasible energetic barriers of 14.7 and 2.4 kcal/mol, respectively.
The low barrier for isomerization is significantly favored over
direct reductive elimination of cis-product, thus leading to the
sole formation of trans-product as the kinetically and
thermodynamically favored product.
We have also applied computational investigations in the
development of a Ni-catalyzed protocol for the trifluorome-
thylthiolation of aryl and heteroaryl chlorides (see Scheme 6).27
Scheme 6. Ni(0)/Ni(II)-Catalyzed Trifluoromethylthiolation
of Aryl Chlorides
In this study, computations rationalized and aided the develop-
ment of the first efficient C-SCF3 functionalization of aryl
chlorides using Ni-catalysis. Our results suggest that the
reactivity is in line with a Ni(0)/Ni(II) pathway, since the
specifically prepared Ni(II)(Ar)(Cl) was shown to be a
competent precatalyst and intermediate. An alternative mecha-
nism involving Ni(I) as a reactive species could be ruled out, as
this species proved to be catalytically incompetent. Notably, in
this context we isolated and characterized a rare Ni(I)-monomer
Scheme 7. Using 4-Chlorophenyl Triflate As a Selectivity Probe and the Experimental Parameters That We Identified to Switch
Selectivity
1315
Article
with dppf (1,1′-bis(diphenylphosphino)ferrocene) and Cl. In
addition, computations and experiments helped to understand
the effect of ligand. Experiments revealed the favored catalyst
resting state as (L)Ni(cod) for L=dppf and Ni(L)2 for L=dppe.
The labile cod ligand in the resting state was computationally
revealed to be a key for efficient catalysis as it readily dissociates
prior to oxidative addition. By contrast, Ni(dppe)2 suffers from a
large ligand dissociation energy, which renders the system
unreactive under the mild reaction conditions employed.
Moreover, MeCN was identified as an efficient traceless additive,
resulting in a widely applicable protocol for the trifluorome-
thylthiolation of a variety of substrates and biologically relevant
molecules.
The question of the truly catalytically active species is a classic
problem in catalysis, especially when several ligation states need
to be considered. Another dimension of complexity arises when
charged species may also be involved. Here, preferences of
adduct formation cannot readily be determined, as different
charge states are involved. In this context, an enduring question
was the potential involvement of anionic Pd as catalytically active
species.28,29 While charged species were previously experimen-
tally detected through electrochemical studies, it was uncertain
whether they could also be reactive.30 We addressed this question
with a selectivity probe and deduced mechanistic information
indirectly through a combined experimental/computational
approach. We studied the chemoselectivity of Suzuki-Miyaura
cross-coupling of 4-chlorophenyl triflate and experimentally
discovered that a switch in solvent from THF to a more polar
solvent, such as DMF, results in reversal of selectivity from C−Cl
to C−OTf functionalization.31,32 A qualitative comparison of
experiments with computations suggested that the observed
selectivity switch is inconsistent with the reactivity of a
monoligated Pd(PtBu3) catalyst, which has generally been
proposed as the catalytically active species for this catalyst
system.33 Instead, calculations are in line with the reactivity of an
anionic Pd catalyst, which forms in the presence of coordinating
species. Control experiments of the computationally derived
conclusions in the absence of coordinating additives (e.g., KF)
and by employing noncoordinating stannane transmetallating
agents instead of boronic acids led to C−Cl addition even in
polar solvent (Scheme 7, left). Addition of KF, on the other hand,
reversed the selectivity again to C−OTf addition.
DOI: 10.1021/acs.accounts.6b00068
Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research
Scheme 8. Reactivity of Halide-Bridged Pd(I) Dimers: Pd(0)- (Red) versus Pd(I)-Catalysis (Green)
Utilizing the same mechanistic probe and its inherent
reactivity preference for addition of monoligated PdL at C−Cl
and bisligated PdL2 at C−OTf, we also examined favored
phosphine ligation states as a function of ligand and computa-
tional method. Implementation of dispersion in the calculation
versus no-dispersion showed that the predicted site selectivity of
mono- (C−Cl) versus bisphosphine (C−OTf) pathways differed
by roughly 20 kcal/mol.17 While for PtBu3, C−Cl addition was
calculated to be preferred (regardless whether dispersion was
considered), our computations suggested that the essentially
unexplored ligand, PiPrtBu2, could react via both mono- and
bisphosphine transition states, thus possessing the ability to
switch selectivity between C−Cl and C−OTf bond addition
depending on the concentration of ligand utilized (Scheme 7,
right).34 This suggests that dispersion is the selectivity-
controlling factor, indicating that it could become an additional
guiding principle in the design of novel ligands.
The synergy of computation and experiment was not only
advantageous for the explanation of selectivity and elucidation of
mechanistic pathways but also led us to the exploration of
intriguing reactivities. In this context, we discovered that halide-
bridged Pd(I) dimers can form under reaction conditions usually
encountered in Stille or Sonogashira cross-couplings. Our
investigations showed that Cu- and Ag-additives typically
employed in cross-coupling reactions lead to an oxidation of
the Pd(0)(PtBu3)2 species. In this process, the corresponding
halide-bridged Pd(I) dimer is formed under concomitant
reduction of Cu- or Ag-salt. This observation is consistent with
some of the contrasting observations of additive effects in cross
coupling reactions. In the case of Suzuki-Miyaura cross-
couplings, the effect of CuXn (X = Br, I; n = 1−2) salts depends
on the employed anion. For X = Br, the bromide-bridged Pd(I)
dimer is formed and readily liberates the catalytically active
monophosphine Pd(0) species resulting in increased reaction
rates. However, for X = I, an iodide-bridged Pd(I) dimer is
formed, which is more robust than the corresponding bromo
dimer and is less efficient in releasing Pd(0). Formation of this
dimer will become severe for challenging cross-coupling, e.g., the
Sonogashira reactions of aryl chlorides, since the formed Pd(I)
dimer readily reacts with alkynes to initiate polymerization.
We subsequently studied the reactivity of the dinuclear
complexes in greater detail. Specifically, we wanted to understand
whether the dimer will solely act as precatalyst for Pd(0) or
whether it may also react directly. In this context, we studied
catalysis derived from the bromide-bridged Pd(I) dimer
[(PtBu3)PdBr]2 in order to obtain insights into the likely active
catalytic species and applications in selective cross-coupling
reactions (Scheme 8).35 By means of combining calculations
with experiments we could show that the Pd(I) dimer reacts as a
precatalyst for Pd(0)-based Suzuki cross-couplings. The
calculations at the CPCM (THF) M06L/6-31+G(d) (SDD)//
1316
Article
B3LYP/6-31+G(d) (SDD) level of theory indicated that a direct
disproportionation is unlikely. Instead the Pd(I) dimer activation
is induced by a nucleophile.36 Thus, under additive-free
conditions, the Pd(I) dimer should remain stable and we
therefore next investigated the fundamental question whether a
Pd(I) dimer could react directly as a catalyst. Subjecting the
bromide-bridged Pd(I) dimer to an aryl iodide, we observed
formation of the corresponding aryl bromide. This indicated that
a halide exchange had taken place and was the first positive
indication that Pd(I) dimers might directly react with aryl
halides.37 We employed a combination of experimental and
computational tools to investigate whether the Pd(I) dimer was
indeed directly involved in this halide-exchange process and to
rule out alternative mechanisms. Our investigations showed that
a Pd(0)/Pd(II) mechanism was unlikely. Pd(II) intermediates
could not be observed during the stoichiometric reaction and
when independently synthesized were not yielding the product
under the analogous reaction conditions. Also, an alternative
Pd(I) radical mechanism was shown to be rather unlikely, since
our experimental studies generating Pd(I) radicals via known
methods only resulted in trace amounts of aryl bromide product.
In addition, the presence of radical trapping agents did not
significantly affect the reaction outcome. This was further
supported by calculations at the CPCM (THF) M06L/6-311+
+G(d,p) (SDD)//B3LYP/6-31G(d) (SDD) level of theory
which indicate that the radical mechanism is energetically
disfavored over the dinuclear pathway. Instead, the data suggest
that the halide-exchange process is likely to be thermodynami-
cally driven and can occur via a direct oxidative addition to the
Pd(I)−Pd(I) framework. Furthermore, kinetic measurements
support this mechanism by showing that the reaction follows a
first-order dependence in both aryl iodide and Pd(I) dimer.38
The dinuclear reactivity of Pd(I) not only represents an
alternative to the textbook Pd(0)/Pd(II) mechanism but also a
concept in catalysis that may have significant potential in
methodology development. First, we identified that the more
robust iodide-bridged Pd(I) dimer is completely bench-stable,
while Pd(0) catalysts are typically not. Thus, we subsequently
investigated the potential of dinculear Pd(I) catalysis for new
types of cross-coupling reactions such as the formation of C−
SCF339 and C−SeCF340 bonds. The prerequisites for the
successful application of the concept of dinuclear Pd(I) reactivity
of are 2-fold: (i) the employed nucleophile needs to be able to
replace the bridging moieties of the dinuclear Pd(I) catalyst, and
(ii) the nucleophile needs to be able to act as a bridging moiety
itself in order to preserve the dinuclear Pd(I) framework.
Applying these guidelines, we were able to develop catalytic
protocols for the formation of C−SCF3 and C−SeCF3 bonds
from aryl iodides and bromides.39,40 DFT-calculations and NMR
monitoring studies were in line with the direct reactivity of
DOI: 10.1021/acs.accounts.6b00068
Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research
Scheme 9. Ligand Design to Facilitate Ar−CF3 Reductive Elimination from Pd(II)
dinuclear Pd(I). In addition, we could show that the stable
catalytic entity can be recovered after reaction completion.
4.3. Computational Prediction and Guide to Experimental
Innovation
The design of chemical reactivity using computations as a
predictive tool constitutes a long pursued goal of computational
chemistry. However, examples in which computations have been
implemented in the design of chemical reactivity are rather scarce
or have not been tested experimentally.41−45
Therefore, we set out to investigate the challenging reductive
elimination step in Ar−CF3 bond formation from Pd(II)
complexes.46 This transformation has been proven difficult as
only very few ligands have been reported to trigger the activation
of the inert Pd−CF3 bond.47,48 By computationally comparing
the experimentally efficient Xantphos ligand with smaller bite
angle ligands, such as dppe, we uncovered that the latter can be
rendered efficient for reductive elimination of Ar−CF3 if small
and electrostatically repulsive ligand substituents are employed.
By using this criterion as a guiding principle, we computationally
designed a small bite angle ligand with CF3-subtituents and
tested our computational design experimentally. After synthesiz-
ing the designed ligand and its corresponding ArPd(II)CF3
complex, efficient reductive elimination to yield PhCF3 was
observed (Scheme 9). In contrast to steric effects, this study uses
electrostatic repulsion as a design principle presenting potential
for novel avenues in catalysis and new ideas for entirely novel
ligand motifs and reactivities.
While Ni has been shown to catalyze the functionalization of
the least reactive aromatic C−OMe bonds49, unreactive C−C or
C−H bonds were installed in those cases, excluding potential
back-reactions of the formed products. Our group set the goal of
developing the first Ni-catalyzed trifluoromethylthiolation of
C(sp2)−O bonds. Our experiments showed, that the SCF3-
moiety of the product reacts with the Ni(0)-catalyst, causing its
deactivation. Therefore, we computationally assessed the
reaction scope and examined possible leaving groups, relative
to the barrier of oxidative addition to Ar-SCF3 (Scheme 10). This
Scheme 10. Assessment of Leaving Groups Compatible with
Ni-Catalyzed Trifluoromethylthiolation
1317
Article
suggested that although a range of C−OR moieties can in
principle react with Ni, only the most reactive ones, triflates and
nonaflates, are compatible with the reactivity window set by the
C−SCF3 functionality. We verified these predictions exper-
imentally and subsequently developed the first efficient method
to functionalize C(sp2)−O bonds. Computations along with
selected key mechanistic experiments served as a valuable tool to
rapidly assess substrate scope while avoiding elaborate screening
efforts. Importantly, the calculations were done on a competitive
time scale (the complete TS-calculations took on average 11.5
h)50 to experiments, showing the potential of computationally
assisted explorations of scope.
5. CONCLUSION
We have showcased herein the power of the application of
computational chemistry in the development and understanding
of modern organic and organometallic chemistry. A particular
emphasis was given to a combined computational and
experimental approach, which allows for a greater and more
unambiguous mechanistic insight than any of the isolated
techniques may provide. Beyond rationalizing, we also showed
examples of successful reactivity predictions, computational
ligand designs and concepts to avoid a wasteful experimental
screening through rapid computational assessment. We hope
that this overview will further stimulate the synergistic use of
calculations and experiment.
■ AUTHOR INFORMATION
Corresponding Author
*E-mail: franziska.schoenebeck@rwth-aachen.de.
Notes
The authors declare no competing financial interest.
Biographies
Theresa Sperger studied chemistry at ETH Zürich and received her
M.Sc. in 2014. She subsequently joined the Schoenebeck group for her
Ph.D. as an Evonik Foundation scholar. Her research interests involve
the mechanistic study of catalytic reactions employing a combined
experimental and computational approach.
Italo A. Sanhueza received his M.Sc. degree in chemical engineering
from Uppsala University in 2011. He is currently pursuing his Ph.D. in
the Schoenebeck group focusing on the combined computational and
experimental study of organic and organometallic reactivity.
Franziska Schoenebeck studied chemistry at the TU Berlin and the
University of Strathclyde in Glasgow. She received her Ph.D. from the
WestCHEM Research School in Glasgow under the supervision of Prof.
Murphy before joining Prof. Houk as a postdoctoral researcher in 2008
at UCLA. She started her independant research at the ETH Zürich in
2010. Since 2013, she has been professor at the RWTH Aachen
University.
DOI: 10.1021/acs.accounts.6b00068
Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research
■ ACKNOWLEDGMENTS
We are grateful for the financial support from the RWTH Aachen
University, ETH Zü rich, MIWF NRW and the Evonik
Foundation (doctoral scholarship to T.S.).
■ REFERENCES
(1) Cramer, C. J. Essentials of Computational Chemistry Theories and
Models; John Wiley & Sons Ltd.: New York, 2002.
(2) Houk, K. N., private communication.
(3) Calculated at the HF/STO-3G level of theory, employing four
processors.
(4) Calculated at the B3LYP/6-31G(d) level of theory, employing four
processors.
(5) Anslyn, E. V.; Dougherty, D. A. Modern Physical Organic Chemistry;
University Science Books: Herndon, VA, 2005.
(6) Jayamurugan, G.; Gisselbrecht, J.-P.; Boudon, C.; Schoenebeck, F.;
Schweizer, W. B.; Bernet, B.; Diederich, F. Expanding the chemical space
for push-pull chromophores by non-concerted [2 + 2] and [4 + 2]
cycloadditions: access to a highly functionalised 6,6-dicyanopentaful-
vene with an intense, low-energy charge-transfer band. Chem. Commun.
2011, 47, 4520−4522.
(7) Yamada, M.; Schweizer, W. B.; Schoenebeck, F.; Diederich, F.
Unprecedented thermal rearrangement of push-pull-chromophore-
[60]fullerene conjugates: formation of chiral 1,2,9,12-tetrakis-adducts.
Chem. Commun. 2010, 46, 5334−5336.
(8) Chiu, M.; Tchitchanov, B. H.; Zimmerli, D.; Sanhueza, I. A.;
Schoenebeck, F.; Trapp, N.; Schweizer, W. B.; Diederich, F. Strain-
Accelerated Formation of Chiral, Optically Active Buta-1,3-dienes.
Angew. Chem., Int. Ed. 2015, 54, 349−354.
(9) Zhou, L.; Li, Z.; Zou, Y.; Wang, Q.; Sanhueza, I. A.; Schoenebeck,
F.; Goeke, A. Tandem Nucleophilic Addition/Oxy-2-azonia-Cope
Rearrangement for the Formation of Homoallylic Amides and Lactams:
Total Synthesis and Structural Verification of Motuporamine G. J. Am.
Chem. Soc. 2012, 134, 20009−20012.
(10) Zou, Y.; Ding, C.; Zhou, L.; Li, Z.; Wang, Q.; Schoenebeck, F.;
Goeke, A. Tandem Cross-Dimerisation/Oxonia-Cope Reaction of
Carbonyl Compounds to Homoallylic Esters and Lactones. Angew.
Chem., Int. Ed. 2012, 51, 5647−5651.
(11) Diethelm, S.; Schoenebeck, F.; Carreira, E. M. Mechanistic Insight
into the Spirocyclopropane Isoxazolidine Ring Contraction. Org. Lett.
2014, 16, 960−963.
(12) Hendriks, C. M. M.; Penning, T. M.; Zang, T.; Wiemuth, D.;
Gründer, S.; Sanhueza, I. A.; Schoenebeck, F.; Bolm, C. Pentafluor-
osulfanyl-containing flufenamic acid analogs: Syntheses, properties and
biological activities. Bioorg. Med. Chem. Lett. 2015, 25, 4437−4440.
(13) Seebach, D. Organic SynthesisWhere now? Angew. Chem., Int.
Ed. Engl. 1990, 29, 1320−1367.
(14) Sperger, T.; Sanhueza, I. A.; Kalvet, I.; Schoenebeck, F.
Computational Studies of Synthetically Relevant Homogeneous
Organometallic Catalysis Involving Ni, Pd, Ir, and Rh: An Overview
of Commonly Employed DFT Methods and Mechanistic Insights.
Chem. Rev. 2015, 115, 9532−9586.
(15) Averkiev, B. B.; Truhlar, D. G. Free energy of reaction by density
functional theory: oxidative addition of ammonia by an iridium complex
with PCP pincer ligands. Catal. Sci. Technol. 2011, 1, 1526−1529.
(16) Gusev, D. G. Assessing the Accuracy of M06-L Organometallic
Thermochemistry. Organometallics 2013, 32, 4239−4243.
(17) Lyngvi, E.; Sanhueza, I. A.; Schoenebeck, F. Dispersion Makes the
Difference: Bisligated Transition States Found for the Oxidative
Addition of Pd(PtBu3)2 to Ar-OSO2R and Dispersion-Controlled
Chemoselectivity in Reactions with Pd[P(iPr)(tBu2)]2. Organometallics
2015, 34, 805−812.
(18) McMullin, C. L.; Jover, J.; Harvey, J. N.; Fey, N. Accurate
modelling of Pd(0) + PhX oxidative addition kinetics. Dalton Trans.
2010, 39, 10833−10836.
(19) Vidossich, P.; Ujaque, G.; Lledos, A. Palladium monophosphine
Pd(PPh3): is it really accessible in solution? Chem. Commun. 2014, 50,
661−663.
1318
Article
(20) Tsang, A. S. K.; Sanhueza, I. A.; Schoenebeck, F. Combining
Experimental and Computational Studies to Understand and Predict
Reactivities of Relevance to Homogeneous Catalysis. Chem. - Eur. J.
2014, 20, 16432−16441.
(21) Lyngvi, E.; Bode, J. W.; Schoenebeck, F. A computational study of
the origin of stereoinduction in NHC-catalyzed annulation reactions of
α,β-unsaturated acyl azoliums. Chem. Sci. 2012, 3, 2346−2350.
(22) Hack, D.; Dürr, A. B.; Deckers, K.; Chauhan, P.; Seling, N.;
Rübenach, L.; Mertens, L.; Raabe, G.; Schoenebeck, F.; Enders, D.
Asymmetric Synthesis of Spiropyrazolones by Sequential Organo- and
Silver Catalysis. Angew. Chem., Int. Ed. 2016, 55, 1797−1800.
(23) Bonney, K. J.; Schoenebeck, F. Experiment and Computation: A
Combined Approach to Study the Reactivity of Palladium Complexes in
Oxidation States 0 to IV. Chem. Soc. Rev. 2014, 43, 6609−6638.
(24) Sperger, T.; Fisher, H. C.; Schoenebeck, F. Computationally
deciphering palladium-catalyzed reaction mechanisms. WIREs Comput.
Mol. Sci. 2016, 6, 226.
(25) Tsang, A. S. K.; Kapat, A.; Schoenebeck, F. Factors That Control
C−C Cleavage versus C−H Bond Hydroxylation in Copper-Catalyzed
Oxidations of Ketones with O2. J. Am. Chem. Soc. 2016, 138, 518−526.
(26) Le, C. M.; Hou, X.; Sperger, T.; Schoenebeck, F.; Lautens, M. An
Exclusively trans-Selective Chlorocarbamoylation of Alkynes Enabled
by a Palladium/Phosphaadamantane Catalyst. Angew. Chem., Int. Ed.
2015, 54, 15897−15900.
(27) Yin, G.; Kalvet, I.; Englert, U.; Schoenebeck, F. Fundamental
Studies and Development of Nickel-Catalyzed Trifluoromethylthiola-
tion of Aryl Chlorides: Active Catalytic Species and Key Roles of Ligand
and Traceless MeCN Additive Revealed. J. Am. Chem. Soc. 2015, 137,
4164−4172.
(28) Amatore, C.; Azzabi, M.; Jutand, A. Rates and mechanism of the
reversible oxidative addition of (Z)- and (E)-1,2-dichloroethylene to
low-ligated zerovalent palladium. J. Am. Chem. Soc. 1991, 113, 1670−
1677.
(29) Amatore, C.; Jutand, A. Anionic Pd(0) and Pd(II) Intermediates
in Palladium-Catalyzed Heck and Cross-Coupling Reactions. Acc. Chem.
Res. 2000, 33, 314−321.
(30) For selected previous computational studies that attempted to
address this question, see: (a) Goossen, L. J.; Koley, D.; Hermann, H. L.;
Thiel, W. Palladium Monophosphine Intermediates in Catalytic Cross-
Coupling Reactions: A DFT Study. Organometallics 2006, 25, 54−67.
(b) Goossen, L. J.; Koley, D.; Hermann, H. L.; Thiel, W. The Palladium-
Catalyzed Cross-Coupling Reaction of Carboxylic Anhydrides with
Arylboronic Acids: A DFT Study. J. Am. Chem. Soc. 2005, 127, 11102−
11114. (c) Ahlquist, M.; Fristrup, P.; Tanner, D.; Norrby, P.-O.
Theoretical Evidence for Low-Ligated Palladium(0): [Pd−L] as the
Active Species in Oxidative Addition Reactions. Organometallics 2006,
25, 2066−2073.
(31) Proutiere, F.; Schoenebeck, F. Solvent Effect on Palladium-
Catalyzed Cross-Coupling Reactions and Implications on the Active
Catalytic Species. Angew. Chem., Int. Ed. 2011, 50, 8192−8195.
(32) Proutiere, F.; Schoenebeck, F. Orthogonal Selectivities under
Pd(0) Catalysis with Solvent Polarity: An Interplay of Computational
and Experimental Studies. Synlett 2012, 23, 645−648.
(33) Schoenebeck, F.; Houk, K. N. Ligand-Controlled Regioselectivity
in Palladium-Catalyzed Cross Coupling Reactions. J. Am. Chem. Soc.
2010, 132, 2496−2497.
(34) Proutiere, F.; Lyngvi, E.; Aufiero, M.; Sanhueza, I. A.;
Schoenebeck, F. Combining the Reactivity Properties of PCy3 and
PtBu3 into a Single Ligand, P(iPr)(tBu)2. Reaction via Mono- or
Bisphosphine Palladium(0) Centers and Palladium(I) Dimer For-
mation. Organometallics 2014, 33, 6879−6884.
(35) Proutiere, F.; Aufiero, M.; Schoenebeck, F. Reactivity and stability
of dinuclear Pd(I) complexes: studies on the active catalytic species,
insights into precatalyst activation and deactivation, and application in
highly selective cross-coupling reactions. J. Am. Chem. Soc. 2012, 134,
606−612.
(36) Aufiero, M.; Scattolin, T.; Proutière, F.; Schoenebeck, F. Air-
Stable Dinuclear Iodine-Bridged Pd(I) Complex - Catalyst, Precursor,
or Parasite? The Additive Decides. Systematic Nucleophile-Activity
DOI: 10.1021/acs.accounts.6b00068
Acc. Chem. Res. 2016, 49, 1311−1319
Accounts of Chemical Research Article

Study and Application as Precatalyst in Cross-Coupling. Organometallics

2015, 34, 5191−5195.
(37) Bonney, K. J.; Proutiere, F.; Schoenebeck, F. Dinuclear Pd(i)
complexes-solely precatalysts? Demonstration of direct reactivity of a
Pd(i) dimer with an aryl iodide. Chem. Sci. 2013, 4, 4434−4439.
(38) Kalvet, I.; Bonney, K. J.; Schoenebeck, F. Kinetic and
Computational Studies on Pd(I) Dimer-Mediated Halogen Exchange
of Aryl Iodides. J. Org. Chem. 2014, 79, 12041−12046.
(39) Yin, G.; Kalvet, I.; Schoenebeck, F. Trifluoromethylthiolation of
Aryl Iodides and Bromides Enabled by a Bench-Stable and Easy-To-
Recover Dinuclear Palladium(I) Catalyst. Angew. Chem., Int. Ed. 2015,
54, 6809−6813.
(40) Aufiero, M.; Sperger, T.; Tsang, A. S.; Schoenebeck, F. Highly
Efficient C-SeCF3 Coupling of Aryl Iodides Enabled by an Air-Stable
Dinuclear Pd(I) Catalyst. Angew. Chem., Int. Ed. 2015, 54, 10322−
10326.
(41) Hopmann, K. H. Quantum chemical studies of asymmetric
reactions: Historical aspects and recent examples. Int. J. Quantum Chem.
2015, 115, 1232−1249.
(42) Sun, Z.; Wang, Q.; Xu, Y.; Wang, Z. A computationally designed
titanium-mediated amination of allylic alcohols for the synthesis of
secondary allylamines. RSC Adv. 2015, 5, 84284−84289.
(43) Gosiewska, S.; Raskatov, J. A.; Shintani, R.; Hayashi, T.; Brown, J.
M. The Origins of Enantioselectivity in Rh−Diene Complex Catalysed
Arylation of Cyclohex-2-enones. Chem. - Eur. J. 2012, 18, 80−84.
(44) Uhe, A.; Hölscher, M.; Leitner, W. Carboxylation of Arene C-H
Bonds with CO2: A DFT-Based Approach to Catalyst Design. Chem. -
Eur. J. 2012, 18, 170−177.
(45) Wang, Y.; Wang, J.; Su, J.; Huang, F.; Jiao, L.; Liang, Y.; Yang, D.;
Zhang, S.; Wender, P. A.; Yu, Z.-X. A Computationally Designed Rh(I)-
Catalyzed Two-Component [5 + 2+1] Cycloaddition of Ene-vinyl-
cyclopropanes and CO for the Synthesis of Cyclooctenones. J. Am.
Chem. Soc. 2007, 129, 10060−10061.
(46) Anstaett, P.; Schoenebeck, F. Reductive elimination of ArCF3
from bidentate Pd(II) complexes: a computational study. Chem. - Eur. J.
2011, 17, 12340−12346.
(47) Grushin, V. V.; Marshall, W. J. Unexpected H2O-induced Ar-X
activation with trifluoromethylpalladium(II) aryls. J. Am. Chem. Soc.
2006, 128, 4632−4641.
(48) Cho, E. J.; Senecal, T. D.; Kinzel, T.; Zhang, Y.; Watson, D. A.;
Buchwald, S. L. The palladium-catalyzed trifluoromethylation of aryl
chlorides. Science 2010, 328, 1679−1681.
(49) Cornella, J.; Zarate, C.; Martin, R. Metal-catalyzed activation of
ethers via C-O bond cleavage: a new strategy for molecular diversity.
Chem. Soc. Rev. 2014, 43, 8081−8097.
(50) Calculated on the RWTH Bull cluster, using 6 cores/10 GB
memory.

1319 DOI: 10.1021/acs.accounts.6b00068

Acc. Chem. Res. 2016, 49, 1311−1319